Musical chairs during mitophagy

Abstract

Mitophagy, or the autophagic degradation of mitochondria, is thought to be important in mitochondrial quality control, and hence in cellular physiology. Defects in mitophagy correlate with late onset pathologies and aging. Here, we discuss recent results that shed light on the interrelationship between mitophagy and mitochondrial dynamics, based on proteomic analyses of protein dynamics in wild-type and mutant cells. These studies show that different mitochondrial matrix proteins undergo mitophagy at different rates, and that the rate differences are affected by mitochondrial dynamics. These results are consistent with models in which phase separation within the mitochondrial matrix leads to unequal segregation of proteins during mitochondrial fission. Repeated fusion and fission cycles may thus lead to “distillation” of components that are destined for degradation.

Mitochondrial quality control is an important aspect of cellular physiology, with links to mitochondrial pathologies, both heritable and sporadic, as well as to “normal” cellular aging phenomena. In mammalian cells as well as in yeast, selective mitochondrial autophagy (mitophagy) seems to occur, independent of starvation-induced macroautophagy and apparently in response to mitochondrial stresses.

A central theme in mitochondrial physiology is the protean nature of the organelle. Mitochondrial material intermixes among the different mitochondria in the cell and this mixing occurs through repeated fusion and fission. The function of this mixing is somewhat obscure, as in both yeast and animal cells concomitant abrogation of both fission and fusion generates cells with morphologically normal mitochondria that can function in oxidative metabolism, but do not carry out intermixing of mitochondrial components. The role of mitochondrial fusion and fission in mitophagy has usually been approached from the perspective of generating “bite-sized” mitochondrial fragments, with the idea that this may be rate limiting. However, experiments with artificially fragmenting mitochondria in yeast, such as in mgm1^ts mutant yeast cells, refute this. To discern whether mitochondrial size or mitochondrial dynamics is the factor that determines the rate of mitophagy, we turned to stationary phase mitophagy in yeast. Stationary phase mitophagy is induced in yeast cells upon very long (~3 d) incubation in a respiratory medium, and several observations suggest that this is a bona fide quality control event. We found that stationary phase mitophagy is significantly slower in dnm1Δ cells, which lack a dynamin homolog that is essential for mitochondrial fission, relative to wild-type (WT) cells.

Yeast mitochondrial fission mutants such as dnm1Δ, have aberrant mitochondrial morphology, exhibiting very large clustered mitochondria. Several labs have shown that mutants that are defective in both fission and fusion, such as dnm1Δ mgm1Δ double-knockout cells (Mgm1 is a dynamin homolog that is required for mitochondrial fusion), have normal mitochondrial morphology but lack dynamics, i.e., intramitochondrial material does not intermix in these cells. The mgm1Δ dnm1Δ double mutants have additional properties; mgm1Δ cells have multiple tiny mitochondrial compartments, since they can perform fission but not fusion and they do not grow on a nonfermentable medium. When combined with the dnm1Δ mutant, however, the double mutants are able to respire and grow in nonfermentable medium. We took advantage of these properties: If the delayed mitophagy observed in dnm1Δ mutants is due to the increased size of the organelle, then the double mutant would be expected to show an enhanced rate, closer to that observed in WT cells. If the requirement for Dnm1 stems from other reasons, however, we would not expect this result. The outcome of the experiment did not support mitochondrial size as the determinant of the rate of mitophagy: the double mutant showed a rate of mitophagy indistinguishable from that of the dnm1Δ mutant, suggesting that the dynamic nature of the organelle is rate determining. To determine whether segregation of individual mitochondrial components occurs during mitophagy, and whether mitochondrial dynamics play a role in it, we turned to proteomics. If proteins segregate according to functionality, with defective molecules being sent for degradation, then different proteins might be expected to have different rates of attrition, and thus of degradation. While we cannot at this point distinguish between “defective” and “functional” molecules of the same protein species in vivo, we are able to compare the degradation rates of specific proteins. We conducted a SILAC screen geared at identifying proteins whose levels drop off with time, during stationary phase, at a rate commensurate with the onset of mitophagy, but whose levels do not follow this pattern in specific mutants. Thus, we compared protein dynamics in WT cells relative to atg32Δ, ptc6Δ/aup1Δ, and pep4Δ, and searched for mitochondrial proteins that did not drop off as rapidly in the mutants as in WT cells. While we were definitely able to identify such proteins, the SILAC data as such did not indicate rates of degradation, but merely reflected the steady-state tendencies of protein abundance levels. We further analyzed a small group of mitochondrial proteins implicated by the proteomic data as rapidly turning over through mitophagy (their levels were much higher in the mutants) and compared them with proteins whose levels were not affected by the mutations, suggesting that they were inefficiently mitophagocytosed. We could definitely discern, using the appropriate controls, between proteins which, over a 7-d stationary phase incubation, were very efficiently (> 50%) turned over via mitophagy and proteins that seemed to be partly or completely (0–20% turnover) excluded from mitophagy. Strikingly, when analyzed in a dnm1Δ mutant, those proteins that are efficiently mitophagocytosed in WT cells were now completely stabilized and showed few or no signs of mitophagy (0–5%) while the proteins that are inefficiently mitophagocytosed in the WT cells seemed unaffected. This result strongly corroborates the prediction of the alternative model, as explained above, in which mitochondrial dynamics functions to generate a selectivity filter for mitophagy. Indeed, protein segregation was observed, albeit of a surprising nature. Proteins that mitophagocytosed efficiently were distributed evenly in the matrix, while proteins that were inefficiently mitophagocytosed were sequestered in specific mitochondrial subdomains. This is somewhat counterintuitive as one might expect that proteins prone to degradation might be sequestered, and not vice versa. Yet another intriguing result was the fact that when looking at a protein with intermediate behavior, Hsp78 (~15–20% turnover by mitophagy), we observed a shift from a random distribution throughout the mitochondrial matrix on the first day of the incubation, to a sequestered pattern by the third day of the incubation, corroborating the correlation between sequestration and mitophagic efficiency.

The sequestration of the inefficiently mitophagocytosed protein species is intriguing. It seems to suggest that sequestration and phase separation is a mechanism of segregating out protein species that do not undergo mitophagy, from a milieu of proteins that actively and perhaps stochastically become entrapped in autophagosome-engulfed mitochondrial compartments. This finding suggests that further study may reveal additional segregation events within the same protein species, which we are currently unable to discern due to technical limitations.

A plausible explanation is a phase separation model in which the functioning protein components of the matrix act as a closely interacting gel phase, as opposed to freely diffusing in solution. The gel would be held together by numerous weak, transient, yet specific and cooperative, interactions between its molecular components. Defective proteins would be shed from the gel and would lose the ability to interact with the gel phase. Such molecules would generate a separate “solution” phase within the matrix, and with sufficient surface tension between the 2 phases, the detached proteins would coalesce into a “puddle” of defective molecules in an aqueous phase, much like oil droplets coalesce at the surface of a pot of chicken soup. Upon organelle fission, the puddle would end up in one or the other daughter due to these simple considerations.

These results, suggesting a role for mitochondrial dynamics in segregating matrix proteins during mitophagy, should also be viewed alongside a recent publication which showed that Dnm1, the GTPase required for mitochondrial fission, directly interacts with the autophagic machinery. It is unclear at this point whether this implies a direct role of Dnm1 in the segregation process, or whether this interaction signifies other roles for Dnm1 in mitophagy.

There is at least one additional layer of information in our results. Traditionally, protein trafficking assays single out a reporter molecule whose fate is taken to generalize the fate of the phase of matter in which the reporter is embedded. However, our results imply that in order to fully grasp the complexity of mitophagy, one needs to follow a battery of reporters, and that by following a single reporter we may be biasing our picture of the process.

Abeliovich H, Zarei M, Rigbolt KT, Youle RJ, Dengjel J. Involvement of mitochondrial dynamics in the segregation of mitochondrial matrix proteins during stationary phase mitophagy. Nat Commun. 2013;4:2789. doi: 10.1038/ncomms3789.