Dnm1p puncta assemble in fis1-L80P cells, but fission is impaired
Our previous studies suggested a model where Fis1p functions at two distinct steps in the mitochondrial fission pathway (Tieu and Nunnari, 2000
). Specifically, we proposed that Fis1p functions at an early step, to regulate Dnm1p assembly, and at a later step, with Mdv1p, to facilitate Dnm1p-dependent mitochondrial membrane constriction and division. To test this model of Fis1p function, we analyzed strains harboring novel FIS1
alleles for a mutation that would separate these two proposed functions. Interestingly, analysis of one FIS1
mutant indicated that, although mitochondrial fission was blocked, Dnm1p assembled into punctate structures characteristic of wild-type cells. Sequence analysis of the FIS1
locus in these cells revealed a point mutation that results in an amino acid change from leucine 80 to proline present in the conserved cytoplasmic region of Fis1p, in a position to mediate interactions with both Mdv1p and Dnm1p.
Examination of mitochondrial morphology in fis1-L80P cells with mitochondrial-targeted GFP revealed net-like mitochondrial structures, in contrast to the branched tubules seen in wild-type cells ( A; ). These structures were indistinguishable from those observed in mdv1-Δ, fis1-Δ, and dnm1-Δ cells and indicate that mitochondrial fission is impaired in fis1-L80P cells ( A). Furthermore, quantification of this mitochondrial morphology defect indicates that the frequency of net-like structures was almost as high as observed in mdv1-Δ and fis1-Δ cells, indicating that mitochondrial fission is severely defective in fis1-L80P cells ().
Figure 1. The fis1-L80P mutation disrupts mitochondrial fission, but Dnm1p-containing puncta have wild-type characteristics. Mito-GFP was used to visualize mitochondrial morphology. Dnm1p was visualized using a Dnm1p–GFP fusion protein. (A) Mitochondrial (more ...)
Quantification of mitochondrial morphology and Dnm1p-associated puncta
To determine whether Dnm1p assembly is affected in fis1-L80P
cells, we examined the steady-state localization pattern of Dnm1–GFP. Interestingly, the characteristics of Dnm1–GFP localization were not significantly altered as compared with wild-type cells ( B; ). The ability of Fis1-L80P protein to support the assembly of Dnm1p into punctate structures is in contrast to what is observed in fis1
cells where Dnm1p assembly is aberrant, as indicated by the presence of fewer, brighter Dnm1p–GFP-labeled punctate structures ( B; ; Mozdy et al., 2000
; Tieu and Nunnari, 2000
These observations suggest that the fission defect observed in fis1-L80P
cells is not the result of a defect in the assembly of Dnm1p-containing puncta, but rather occurs as a result of a defect at a later step in the division pathway. Significantly, the phenotypic characteristics of fis1-L80P
cells are similar to those observed in mdv1
cells (; Tieu and Nunnari, 2000
). Together with our previous observation that in dnm1
cells, Mdv1p remains associated with mitochondria in a Fis1p-dependent manner, these results support our hypothesis that Fis1p functions with Mdv1p to regulate fission at a post-Dnm1p assembly step.
Overexpression of Mdv1p suppresses the fission defect in fis1-L80P cells
We asked whether overexpression of Mdv1p could suppress the observed fission defect in fis1-L80P cells to further test our hypothesis that a Mdv1p-dependent, post-Dnm1p assembly step is specifically blocked in fis1-L80P cells. To overexpress Mdv1p in cells, we used the GAL1 promoter. When cells containing the GAL1-MDV1 plasmid were grown under inducing conditions, using the carbon source galactose, Mdv1p was overexpressed ~20-fold as assessed by SDS-PAGE and Western blotting with anti-Mdv1p antibodies ( A).
Figure 2. Overexpression of MDV1 suppresses the mitochondrial fission defect in fis1-L80P cells. All strains harboring either the empty pGAL1 vector or pGAL1-MDV1 vectors were grown in SRaf media to early log, subcultured into SGal media, and grown for an additional (more ...)
Examination of mitochondrial morphology in wild-type cells with mito-GFP indicated that overexpression of Mdv1p had no effect on mitochondrial structure (). This observation is in contrast to what has been observed in studies of Dnm1p, where overexpression results in mitochondrial fragmentation and indicates that, unlike Dnm1p, the concentration of Mdv1p in wild-type cells is not rate limiting for fission (Sesaki and Jensen, 1999
; Fukushima et al., 2001
). As expected, net-like mitochondrial structures observed in mdv1
cells were transformed into tubular branched structures upon Mdv1p overexpression, indicating that fission is restored and that Mdv1p expressed from the GAL1 promoter is functional (Tieu and Nunnari, 2000
; ). Significantly, mainly branched reticular mitochondrial structures, characteristic of wild-type cells, were observed when Mdv1p was overexpressed in fis1-L80P
, indicating that overexpression of Mdv1p suppresses the fission defect in fis1-L80P
cells (). Mitochondrial nets, however, persisted in fis1
cells when Mdv1p was overexpressed, demonstrating that suppression of the fission defect in fis1-L80P
cells is allele specific (). Taken together, these observations are consistent with our hypothesis that a Mdv1p-dependent post-Dnm1p assembly step is specifically blocked in fis1-L80P
and suggest that Mdv1p interacts with Fis1p during fission.
Mdv1p and Fis1p are in a complex
We previously used the two-hybrid assay to demonstrate that Dnm1p and Mdv1p interact (Tieu and Nunnari, 2000
). In this study, we also used the two-hybrid assay to determine whether Mdv1p and Fis1p interact. Interactions between activating domain (AD) fusion proteins and binding domain (BD) fusion proteins were assessed by monitoring the expression of the stringent GAL2
reporter gene. We tested both the full-length AD–FIS1
protein fusion and a construct lacking the transmembrane domain, AD–FIS1-
, with BD–MDV1
to assess protein–protein interactions. Cells expressing a combination of AD–FIS1
constructs displayed growth on media lacking adenine, indicating that Fis1p and Mdv1p specifically interact ( A; unpublished data). Cells harboring both the AD–FIS1-
plasmids displayed more robust growth on media lacking adenine. Thus, not surprisingly, the Fis1p transmembrane domain interfered with the Fis1p–Mdv1p interaction in the assay.
Figure 3. Fis1p interacts with Mdv1p by two-hybrid and coimmunoprecipitation analyses. (A) Two-hybrid analysis of MDV1 and FIS1. Interaction between AD fis1Δ128–155 and BD MDV1 two-hybrid vectors was assessed by growth of indicated transformants (more ...)
To test whether a Mdv1p–Fis1p interaction occurs within the context of the mitochondrial membrane, we performed immunoprecipitations with anti-Mdv1p from isolated detergent-solubilized mitochondria and whole cell extracts. We were unable to detect a Mdv1p–Fis1p complex in cell extracts, suggesting that a Mdv1p–Fis1p interaction might be labile in vitro. To overcome the possible instability associated with the Mdv1p–Fis1p interaction, proteins were cross-linked in vivo with the bifunctional, reversible cross-linker dithiobis(succinimidylpropionate) (DSP), and extracted under denaturing conditions before immunoprecipitation with antibodies. After immunoprecipitation, cross-links were reversed with a reducing agent and precipitates were analyzed by SDS-PAGE and Western blotting with anti-Mdv1p and Fis1p antibodies.
Western blot analysis of fractions from the anti-Mdv1p immunoprecipitation revealed that a significant fraction of Mdv1p from wild-type cells was present in the precipitate ( B, lanes 1–3). Significantly, we observed that a fraction of Fis1p from cross-linked extracts was reproducibly coimmunoprecipitated by anti-Mdv1p antibodies ( B, lanes 1–3). As a control for specificity, anti-Mdv1p antibodies were used to perform immunoprecipitations from DSP–cross-linked mdv1-Δ cell extracts ( B, lanes 4–6). Under these conditions, Fis1p was not observed in precipitates, indicating that the coimmunoprecipitation of Fis1p with anti-Mdv1p antibodies specifically is dependent on Mdv1p (, lanes 4–6). The Mdv1p–Fis1p interaction could also be detected by coimmunoprecipitation in extracts prepared from isolated mitochondria after DSP cross-linking (unpublished data). Thus, both two-hybrid and coimmunoprecipitation data indicate that Mdv1p and Fis1p interact within a complex in cells.
The Mdv1p–Fis1p interaction is abolished in fis1-L80P cells
Taken together, our genetic, biochemical, and two-hybrid analyses suggest that Fis1p interacts with Mdv1p to regulate a rate-limiting, post-Dnm1p assembly step in the fission pathway. In this context, these observations further suggest that an alteration in the Fis1p–Mdv1p interaction in fis1-L80P cells is specifically responsible for the observed defect in mitochondrial fission. To test this idea, we biochemically analyzed the Fis1p–Mdv1p interaction in fis1-L80P cells.
We first examined the stability and intracellular localization of Fis1p and Mdv1p in fis1-L80P cells. Wild-type and fis1-L80P cell extracts were fractionated by differential centrifugation, and analyzed by SDS-PAGE and Western blotting. Consistent with a mitochondrial localization, in extracts from both wild-type and fis1-L80P cells, the majority of Mdv1p and Fis1p cofractionated with porin, the mitochondrial marker, in the mitochondrial-enriched pellet fraction ( A, compare lanes 1–3 with 4–6). In addition, levels of Mdv1p and Fis1p were similar in fis1-L80P and wild-type cell extracts ( A, compare lanes 1–3 with 4–6). These results indicate that both Fis1p and Mdv1p are expressed stably and localized correctly to mitochondria in fis1-L80P cells. Thus, the fission defect observed in fis1-L80P cells is not simply the result of Fis1p and/or Mdv1p instability.
Figure 4. The Fis1p–Mdv1p interaction is disrupted in fis1-L80P cells and restored upon overexpression of Mdv1p. (A) Fractionation of cell extracts from wild-type and fis1-L80P cells by differential centrifugation. Fractions were analyzed by SDS-PAGE and (more ...)
To test whether the Mdv1p–Fis1p interaction is altered in fis1-L80P cells, we determined whether Fis1p coimmunoprecipitated with Mdv1p under conditions where a complex was detected in wild-type cells (). Interestingly, when Mdv1p was immunoprecipitated from DSP–cross-linked extract from fis1-L80P cells with anti-Mdv1p antibodies, we failed to detect Fis1p in the precipitates, in contrast to wild-type cells ( B, pGAL1). These results suggest that a complex containing Mdv1p and Fis1-L80Pp fails to form, or that interactions within the complex are weakened and thus harder to detect. Consistent with the latter possibility is our observation that overexpression of Mdv1p in fis1-L80P cells suppresses the mitochondrial fission defect. Thus, we tested whether increasing the amount of Mdv1p in the fis1-L80P cells could, by mass action, restore the Mdv1p–Fis1p interaction observed by coimmunoprecipitation with anti-Mdv1p antibodies.
Wild-type, mdv1-Δ, and fis1-L80P cells harboring pGAL1-MDV1 were grown in galactose to induce overexpression of Mdv1p, DSP cross-linked, and immunoprecipitated with anti-Mdv1p antibodies. Western blot analysis of fractions from wild-type cells indicated that Fis1p coimmunoprecipitated with Mdv1p either with or without overexpression of Mdv1p ( B, pGAL1 and pGAL-MDV1, lanes 1 and 4). Interestingly, overexpression of Mdv1p did not increase the fraction of Fis1p that coimmunoprecipitated with Mdv1p, suggesting that Mdv1p is not limiting for this interaction. As expected, when Mdv1p is overexpressed in mdv1-Δ cells, Fis1p can be observed in the immunoprecipitates, in contrast to immunoprecipitates from mdv1-Δ cells harboring the pGAL1 vector without the MDV1 gene ( B). Significantly, overexpression of Mdv1p in fis1-L80P cells restored the Mdv1p–Fis1p interaction as detected by coimmunoprecipitation ( B, pGAL1 and pGAL-MDV1, lanes 1 and 4). These observations indicate that the Mdv1p–Fis1p interaction is defective in fis1-L80P cells and that overexpression of Mdv1p can restore this interaction, as detected by coimmunoprecipitation. These data suggest that the mitochondrial fission defect observed in fis1-L80P cells is the result of a defective Fis1p–Mdv1p interaction, suggesting a role for this interaction at a late post-Dnm1p assembly step in mitochondrial fission.
Mdv1p functions as a molecular adaptor in fission
Data presented in this manuscript demonstrate that Mdv1p interacts with Fis1p during fission to catalyze a rate-limiting step. We had previously shown that Mdv1p also interacts with Dnm1p in punctate structures within cells in a Fis1p-independent manner (Tieu and Nunnari, 2000
). Thus, to gain further insight into the molecular mechanism of mitochondrial fission, we examined the regions of Mdv1p responsible for its interactions with Fis1p and Dnm1p.
Mdv1p contains at least three distinct regions: a novel NH2
-terminal extension region (NTE), a middle region predicted to form a C-C structure, and a COOH-terminal region that contains seven WD repeats predicted to form a seven-bladed propeller structure (WD) ( A; Tieu and Nunnari, 2000
). To analyze the structural features of Mdv1p required for its interactions with Dnm1p and Fis1p, we constructed GAL1-regulated GFP fusions to each of these putative domains and examined their localization patterns after expression in wild-type, mdv1-
, and fis1
cells. We also examined mitochondrial morphology, particularly in wild-type and mdv1
cells, to determine whether expression of Mdv1p domains produced dominant negative phenotypes or could complement the loss of MDV1
Figure 5. Mdv1p functions as a molecular adaptor during mitochondrial fission. (A) Predicted structural domains of Mdv1p (NTE, C-C, and WD [seven-WD repeat domain]). (B) Dual localization of mitochondria labeled with MitoTracker (in red, left) and GFP-tagged Mdv1p (more ...)
The localization of GFP-tagged Mdv1p domains and mitochondrial morphology in representative cells are presented and summarized in schematic form in B. As expected, GFP-tagged Mdv1p was localized to punctate structures primarily associated with mitochondria in wild-type, mdv1
, and fis1
cells, but not in dnm1
, consistent with our previous observations showing that GFP–Mdv1p interacts and colocalizes with Dnm1p in these structures in a Fis1p-independent manner ( B, panels 1–4). In addition, in mdv1
cells, mitochondrial tubular structures characteristic of wild-type cells were observed, indicating that, as previously published, GFP–Mdv1p is functional ( B, panel 2; Tieu and Nunnari, 2000
). Also as previously shown, GFP–Mdv1p was observed uniformly localized to mitochondria in dnm1
cells ( B, panel 4; Tieu and Nunnari, 2000
). The functional and biochemical data presented in this study indicate that Mdv1p's mitochondrial localization pattern in dnm1
cells reflects a Dnm1p-independent interaction between Mdv1p and Fis1p. Consistent with this interpretation, in cells lacking both Dnm1p and Fis1p, Mdv1p was observed in a diffuse pattern, associated with the cytosolic fraction (unpublished data; Tieu and Nunnari, 2000
Mitochondrial net-like structures were observed in mdv1-Δ strains expressing each Mdv1p region alone, indicating that no single region was sufficient for wild-type levels of mitochondrial fission ( B, panels 2, 6, 10, and 14; 100%, n = 50 in all strains). However, in wild-type, mdv1-Δ, and fis1-Δ cells expressing a GFP-tagged version of WD region of Mdv1p, GFP fluorescence was observed in punctate structures, primarily associated with mitochondria ( B, panels 9–12). The localization of GFP–WD to punctate structures was not observed in dnm1-Δ cells, indicating that Dnm1p is required for their formation ( B, panel 12). In addition, punctate structures labeled by GFP–WD also contained Dnm1p, as assessed by colocalization of Dnm1–dsRed in wild-type cells ( C). Taken together, these observations suggest that the WD region is sufficient to interact with Dnm1p.
GFP–NTE labeled both mitochondria and punctate structures in wild-type cells ( B, panels 5–8). Unlike GFP–WD, however, the punctate structures labeled by GFP–NTE probably do not reflect an interaction of NTE with Dnm1p and may be the result GFP–NTE self-aggregation because they also were observed in dnm1-Δ and dnm1-Δ fis1-Δ cells ( B, panel 12; unpublished data). In addition, GFP–NTE-labeled punctate structures were observed localized at the cell cortex, not associated with mitochondria. The dispersive mitochondrial labeling pattern observed for GFP–NTE in wild-type cells was not observed in fis1-Δ, suggesting that the NTE interacts specifically with Fis1p in cells ( B, panel 7). Consistent with this interpretation, GFP–NTE also was observed to be associated with mitochondria in dnm1-Δ cells ( B, panel 8). In contrast, GFP–WD was observed in a diffuse, cytosolic pattern in the majority of dnm1-Δ cells ( B, panel 12). Thus, the cytological analysis of Mdv1p domains suggests that the WD and NTE regions are each sufficient to interact with Dnm1p and Fis1p, respectively.
Interestingly, in all cell types examined, the GFP-tagged C-C region of Mdv1p was observed in a diffuse pattern, consistent with a cytosolic localization, suggesting that this region is not involved in mediating interactions with either Dnm1p or Fis1p ( B, panels 13–16). Interestingly, in contrast to the NTE region, expression of GFP-tagged C-C and WD in wild-type cells caused mitochondrial net-like structures to form ( B, panels 9 and 13; 80% net-like structures, n = 58, and 19% net-like structures, n = 59, respectively), indicating that these regions interfere with mitochondrial fission in a dominant negative manner.
To independently test our cytological observations, we also examined the structural features of Mdv1p required for its interactions with Dnm1p and Fis1p using the two-hybrid assay. We tested both the full-length Mdv1p and Dnm1p protein fusions and Fis1Δ128–155p with regions of Mdv1p to determine protein–protein interactions. An interaction was indicated by cells growing on media lacking adenine (indicated by + in ).
Two-hybrid interactions between Mdv1p and Mdv1p domains, Fis1Δ128–155p and Dnm1p
As previously shown, full-length Mdv1p was observed to interact with both Fis1-Δ128–155p and Dnm1p in this assay ( A; Tieu and Nunnari, 2000
). In addition, we observed that full-length Mdv1p interacted with full-length Mdv1p in this assay, suggesting that Mdv1p is able to oligomerize. The WD region was observed to interact with Dnm1p exclusively, consistent with our cytological observations using GFP-tagged WD. The NTE region was expressed as a BD or AD fusion only when the C-C region was included (NTE/C-C; unpublished data) and specifically displayed an interaction with Fis1Δ128–155p, consistent with our cytological analysis of the NTE region alone.
We also observed an interaction of NTE/C-C with full-length Mdv1p by two-hybrid analysis (). Indeed, two-hybrid analysis indicates that the C-C region alone was sufficient to specifically interact with Mdv1p, suggesting that Mdv1p forms a higher order homo-oligomeric structure via the predicted C-C region (). Analysis of the primary structure of Mdv1p by the MultiCoil algorithm indicates that Mdv1p has a high probability of forming a parallel dimeric C-C structure (Wolf et al., 1997
). It is possible, therefore, that the C-C region exerts its dominant negative effect on mitochondrial fission in wild-type cells by preventing the formation of an Mdv1p dimer ( B, panels 13–16). The localization of the GFP-tagged C-C region to the cytosol in cells further suggests that full-length Mdv1p interacts with both Dnm1p and Fis1p as a homodimeric or higher order complex.
A molecular model for mitochondrial fission
Data presented in this paper support a model, similar to the one recently proposed where Fis1p plays two distinct and separable roles during mitochondrial fission (Shaw and Nunnari, 2002
). Early in the fission pathway, Fis1p targets Dnm1p to mitochondrial membranes and regulates its assembly probably via a direct interaction. Dimeric Mdv1p coassembles with Dnm1p in Dnm1p-containing punctate structures where it functions, specifically at a rate-limiting step where fission is triggered. Here we have provided genetic and biochemical evidence that Fis1p also functions at this rate-limiting step, by interacting with Mdv1p. Our observations also indicate that Mdv1p plays the role of a molecular adaptor, whose two functional domains are separated by a C-C region. Our analysis supports a model where the WD domain of Mdv1p mediates an interaction with assembled Dnm1p. Given that expression of the WD region interferes with mitochondrial fission in a dominant negative manner, we infer that the localization of Mdv1p to assembled Dnm1p structures is important for its ability to stimulate fission. Our data also support a role for the NTE region of Mdv1p as a molecular switch that interacts in a regulated manner with Fis1p, triggering conformational changes within an assembled Dnm1p structure that bring about the division of mitochondrial membranes. Elucidation of the stoichiometry of Dnm1p, Fis1p, Mdv1p, and any other factors within the complex(es) we have identified will help answer the question of how these components function in generating the force required to coordinately divide the outer and inner mitochondrial membranes.