We recently reported that the Parkinson disease-linked E3 ubiquitin ligase, Parkin, is selectively recruited to dysfunctional mitochondria with low membrane potential to promote their autophagic degradation, suggesting that a deficiency of mitochondrial quality control may underlie the observed mitochondrial dysfunction in Parkin knockout Drosophila
and mice 
. How Parkin is able to distinguish damaged, depolarized mitochondria from healthy, polarized mitochondria, however, was unknown.
Here, we show that PINK1 selectively accumulates on depolarized mitochondria that have sustained damage. This selective accumulation is achieved by a novel mechanism, in which PINK1 is constitutively synthesized and imported into all mitochondria, but cleaved from healthy mitochondria by voltage-sensitive proteolysis (Figure S9
). On damaged mitochondria that have lost their membrane potential, however, PINK1 cleavage is inhibited, leading to high PINK1 expression on the dysfunctional mitochondria. Expression of mitochondrial PINK1 is required for the recruitment of Parkin to the dysfunctional mitochondria and for their selective elimination by Parkin. In addition, increased expression of PINK1 on the outer mitochondrial membrane is sufficient for Parkin recruitment and Parkin-induced mitophagy, suggesting that loss of membrane potential activates Parkin recruitment primarily through the up-regulation of mitochondrial PINK1.
This model offers a parsimonious explanation for several observations that have been made previously. Full-length mitochondrial PINK1 (~63 kDa) is cleaved into a short ~52-kDa form, but the short, primarily cytosolic form is unstable, raising the questions: why is PINK1 found both on the mitochondria and in the cytosol, and which form of PINK1 is active in the PINK1/Parkin pathway 
. Our results suggest that full-length mitochondrial PINK1 is the active form in the PINK1/Parkin pathway, and that cleavage of PINK1 into an unstable cytosolic form maintains low levels of PINK1 on healthy mitochondria in order to suppress the PINK1/Parkin pathway in the absence of mitochondrial damage. Additionally, this model provides an explanation for the observation that the uncoupler valinomycin (which can inhibit the TIM22/23 mitochondrial import pathway) blocks PINK1 processing but fails to block PINK1 import 
. Our model suggests that membrane potential is not required for PINK1 import but is required to selectively maintain low PINK1 expression on healthy mitochondria. This mechanism couples the collapse of mitochondrial voltage potential following mitochondrial damage to selective PINK1 accumulation on damaged mitochondria.
At present, it is unclear which protease(s) mediate the cleavage of PINK1 in mammalian cells. Although the intramembrane serine protease Rhomboid-7 appears to be required for PINK1 cleavage in Drosophila 
, our results suggest that PARL, its mammalian ortholog, is not required for PINK1 cleavage in mammalian cells. This situation is similar to that of OPA1, which also requires Rhomboid-7 for cleavage in Drosophila
, but does not require PARL for cleavage in mammalian cells 
In addition, determining how PINK1 cleavage is modulated by membrane potential will require further study. The protease itself may be sensitive to membrane potential and/or the PINK1 cleavage site may be available to the protease only in the presence of a membrane potential. Alternatively, the regulation of PINK1 cleavage by membrane potential may be indirect. That inhibition of PINK1 cleavage by mitochondrial depolarization up-regulates the PINK1/Parkin mitophagy pathway also raises the possibility that inhibitors of PINK1's protease might up-regulate the pathway and have some therapeutic benefit.
Our results suggest that PINK1 induces Parkin recruitment to a particular subset of mitochondria, following its accumulation, and there are several models for how PINK1 might induce Parkin recruitment. In the simplest, as PINK1 accumulates, Parkin may be recruited to mitochondria through a direct interaction with the accumulated PINK1. In support of this model, PINK1 appears to directly bind Parkin at least in some contexts 
. Alternatively, PINK1 may need to phosphorylate Parkin, a substrate of Parkin, or an adaptor between PINK1 and Parkin, and thereby increase Parkin's affinity for a substrate or receptor on mitochondria. Consistent with a role for phosphorylation in the activation of Parkin, we found a kinase-deficient version of PINK1 fails to rescue Parkin recruitment to mitochondria in PINK1 null MEFs (even though PINK1 KD appears to be processed identically to wild-type PINK1). We were unable to replicate findings suggesting that phosphorylation of two threonines in a conserved region of Parkin is sufficient to induce Parkin recruitment to mitochondria 
, but it is possible that Parkin may be phosphorylated by PINK1 elsewhere. If direct phosphorylation is sufficient to induce Parkin recruitment to mitochondria, however, it seems difficult to explain how Parkin can be targeted to a particular subset of mitochondria, as appears to occur in cells with bioenergetically diverse populations of mitochondria 
Mutations in Parkin and PINK1 are inherited primarily in a recessive manner, and loss of their function is thought to cause early-onset Parkinson disease. We find that patient mutations in PINK1 and Parkin disrupt the PINK1/Parkin mitochondrial turnover pathway at distinct steps, consistent with the potential relevance of this pathway for the development of Parkinson disease.
Mutations in Parkin's UBL or its deletion caused a moderate deficit in Parkin recruitment to depolarized mitochondria and induction of mitophagy. That deletion of the UBL only partially inhibited the recruitment of Parkin to mitochondria suggests that whereas this domain promotes the recruitment of Parkin to mitochondria, it is not absolutely necessary for recruitment or subsequent mitophagy. The UBL likely promotes recruitment of Parkin through interaction with a protein containing a ubiquitin-binding domain, as mutating residue isoleucine 44, which is critical for the interaction between UBLs and UBDs 
, to alanine resulted in a recruitment deficit similar to that caused by deletion of the UBL domain. The disease-causing mutations R42P, which causes global unfolding by NMR 
, and A46P lie on either ends of the beta-pleated sheet containing I44A, suggesting that these mutations may inhibit Parkin recruitment by disrupting the interaction between Parkin and UBD-containing proteins (Figure S7A and S7B
Mutations in key cysteine residues in the RBR domain or deletion of RING2, which is responsible for Parkin's ubiquitin ligase activity, severely disrupt both the recruitment of Parkin to mitochondria and its induction of mitophagy. Interestingly, the R275W mutation in RING1 of Parkin causes only a minor disturbance of Parkin recruitment to depolarized mitochondria but severely disrupts mitophagy, suggesting that recruitment and mitophagy can be experimentally disassociated.
The R275W polymorphism in Parkin and the G411S polymorphism in PINK1 have only been identified as heterozygous polymorphisms in cases of Parkinson disease 
. For this reason, the pathogenicity of these polymorphisms has been a matter of controversy. Our results show that the R275W Parkin mutation, which affects a highly conserved arginine residue, causes a significant loss of Parkin function in our mitophagy assay. This is consistent with in vivo data in Drosophila melanogaster
, demonstrating that Parkin R275W, unlike wild-type Parkin, fails to compensate for loss of endogenous Parkin. By contrast, we found that PINK1 containing the G411S polymorphism, which is conserved in vertebrates, but not invertebrates, could compensate for loss of endogenous PINK1, consistent with the view that PINK1 G411S may be a natural variant and not a disease-causing mutation.
The stringent dependence of Parkin recruitment on PINK1 under depolarizing conditions is a little surprising given that, when overexpressed, Parkin can partially compensate for PINK1 loss in Drosophila
and in mammalian cells 
. How Parkin overexpression compensates for PINK1 loss is not known, but there are several possible explanations. First, there may be mechanisms independent of PINK1 and depolarization that can recruit Parkin to dysfunctional mitochondria. Alternatively, Parkin may serve other functions in the cell that are independent of PINK1 and protect against mitochondrial dysfunction indirectly; or Parkin may function to some degree upon overexpression independently of mitochondrial docking, perhaps effecting mitophagy or other mitochondrial changes from the cytosolic compartment.
Stable loss or knockdown of PINK1 in mammalian cellular models and mice leads to a number of mitochondria-related abnormalities. Mitochondria in these cells or tissues exhibit electron transport chain (ETC) dysfunction, diminished membrane potential, increased reactive oxygen species production, mitochondrial fragmentation, and calcium dysregulation, among other abnormalities 
. Although some of these abnormalities may be a reversible consequence of others—for instance, mitochondrial fragmentation may be due to low membrane potential 
, and ETC dysfunction and decreased membrane potential may be, in part, a functional consequence of calcium dysregulation 
—other abnormalities may be due to irreversible dysfunction of specific mitochondrial proteins or protein complexes. For instance, Complex I and the putative Na+
transporter seem to be dysfunctional in cultured cells following PINK1 knockdown 
, whereas Complex I and II appear to be dysfunctional in the striatum of mice lacking PINK1 
Although the proximate cause of these abnormalities in PINK1 null cells remains obscure, one explanation may be the failure of PINK1/Parkin pathway to eliminate oxidatively damaged mitochondria, which accumulate over time as a natural consequence of metabolism and other cellular stresses. That Parkin null cells and tissues appear to share some of the same mitochondrial defects as PINK1 null cells and tissues supports the view that these abnormalities may be due to loss of a common PINK1/Parkin pathway 
. We cannot rule out that PINK1 may actively prevent mitochondrial damage and dysfunction, in addition to its signaling role in the PINK1/Parkin pathway. PINK1's interaction with HtrA2/OMI, for instance, appears to be independent of Parkin function in Drosophila 
Loss of PINK1 and Parkin affects some cell populations, like substantia nigral neurons, more than others, even though PINK1 and Parkin appear to be more widely expressed. Why some tissues are more vulnerable to loss of PINK1/Parkin than others is unclear, but it may relate to the degree of damage mitochondria sustain within that tissue (e.g., mitochondria in the SN are subject to greater oxidative stress than those in other neural tissues 
); the existence of redundant mitophagy pathways (e.g., mammalian tissues may contain pathways orthologous to those recently identified in yeast 
); the ability of the tissue to mitigate the damage by other means (a tissue composed of mitotic cells may be able to manage mitochondrial damage through cellular turnover rather than mitochondrial turnover); and mitochondrial demand within a particular tissue (neurons have high, local metabolic demands, and dopaminergic neurons are subject to especially high calcium fluxes that need to be buffered by mitochondria 
). Some or all of these factors may contribute to the special reliance of SN neurons on PINK1 and Parkin.
PINK1 and Parkin are a significant cause of autosomal recessive parkinsonism and have been genetically linked to a pathway that protects against progressive mitochondrial damage and dysfunction. We have found that PINK1 levels and subsequently Parkin recruitment to mitochondria are dramatically regulated by the bioenergetic state of individual mitochondria, and that this unique regulation may allow PINK1 and Parkin to promote the selective and efficient turnover of mitochondria that have become damaged. Loss of PINK1 or Parkin function due to pathogenic mutations can disrupt this mitochondrial turnover pathway which may lead to the accumulation of dysfunctional mitochondria in vulnerable tissues—with a resultant increase in oxidative stress, depression of metabolism, and, eventually, accelerated cell death, which has been observed in Drosophila
and, to a lesser extent, in mouse models of the disease 
. Together, these findings provide a biochemical explanation for the genetic epistasis between PINK1 and Parkin observed in Drosophila
, and support a novel, testable model of how loss of PINK1 and Parkin function may lead to autosomal recessive parkinsonism.