Our data indicate that PINK1 is activated following mitochondrial membrane potential (Δψm) depolarization and that PINK1 directly phosphorylates Parkin at Serine 65 (Ser65
) inducing its activation. Previous observations indicating that Drosophila
dPINK1 and dParkin null flies have similar degenerative phenotypes [4
] and that over-expression of Parkin rescues the phenotype of dPINK1 null Drosophila
(but not the converse) are consistent with PINK1 acting upstream of Parkin [4
]. The ability of PINK1 to regulate mitochondrial dynamics in mammalian cells as well as Drosophila
has also been suggested to be dependent upon Parkin [6
]. The finding that humans with loss-of-function mutations in either PINK1 or Parkin display indistinguishable clinical presentation of PD also argues in favour of a major connection between PINK1 and Parkin in humans [18
Parkin belongs to the RING-in-between-RING (RBR) family of ubiquitin ligases and recent work suggests that Parkin as well as other RBR family members may function as RING/HECT hybrids that use the HECT-preferred E2, UbcH7, and catalyse transfer of ubiquitin to substrates via an intermediate thioester-linked ubiquitin adduct on a conserved cysteine in the RING2 domain (Cys431
]. It was recently proposed that the Ubl domain of Parkin acts as an auto-inhibitory domain by binding to the C-terminal region and preventing catalytic activity by sterically preventing ubiquitin binding to the C-terminus [24
]. However, the mechanism of activation of Parkin remained unknown [24
Inspection of all the available structures, both nuclear magnetic resonance (NMR) and crystal, reveals that Ser65
may sometimes lie within the fifth β-strand that makes up the ubiquitin-like fold and is partially exposed to the surface (PDB code 2KNB (solution, mouse) [31
]; PDB code 1IYF (solution, human) [32
]; see electronic supplementary material, figure S4) or sometimes may lie within a loop adjacent to the strand (PDB code 1MG8 (solution, mouse) [33
]; PDB code 2ZEQ (crystal, mouse) [34
]) suggesting some conformational flexibility around this element. It is therefore possible that upon interaction with PINK1, the Ubl domain undergoes a local conformational change enabling PINK1 to gain access to Ser65
for phosphorylation. One possibility is that phosphorylation at Ser65
may perturb the β-strand formation in the Ubl domain leading to loss of the autoinhibition and subsequent activation of Parkin. Indeed previous work has shown that this region of the Ubl domain in which Ser65
lies is critical for maintaining the autoinhibition [24
]. In future work it would be critical to study the biophysical interaction between PINK1 and Parkin, define the interacting domains in more detail and establish how phosphorylation of Ser65
impacts on this interaction leading to the activation of Parkin. The finding that phosphorylation of full-length Parkin is much less efficient than phosphorylation of the isolated Ubl-domain suggests that other factors might regulate the ability of PINK1 to phosphorylate Parkin at Ser65
, which would be interesting to investigate in future studies.
In this study, we have assessed Parkin E3 ligase activity employing an auto-ubiquitylation assay that establishes that following Ser65
phosphorylation, Parkin in the presence of an E1, and the E2, UbcH7, is capable of inducing the formation of short polyubiquitylated chains. In future work, it would be vital to repeat these assays employing physiological Parkin substrate(s) to confirm that Ser65
phosphorylation of Parkin does indeed enhance the ubiquitylation of a genuine Parkin substrate. In our assay, it is possible that these short polyubiquitylated chains are covalently attached to the E2, UbcH7, since in a recent study of the RBR family member, HHARI, similar ubiquitin species were produced by HHARI catalytic activity, which were attributed to HHARI E3-dependent ubiquitylation of UbcH7 [30
]. Further work will be required to determine the nature of these ubiquitin species and whether or not they are linked to UbcH7. It will also be exciting to test the effect of disease mutants located within and out with the Ubl domain as well as a mutant of cysteine (Cys431
) on the catalytic activity of Parkin.
Our data provide fundamental insights into the regulation of Parkin and elaborate a signalling pathway that may be central to neuronal loss in PD (). Our data would suggest that loss-of-function mutations in PINK1 would lead to suppression of Parkin E3 ligase activity and result in reduced ubiquitylation of Parkin's targets. This may explain why over-expression of Parkin in dPINK1 null Drosophila
restores ubiquitylation of targets and rescues the null phenotype [4
]. It is possible that the key Parkin targets are located at the mitochondria and indeed several candidate mitochondrial substrates for Parkin have been proposed, including Mitofusin1 [35
], VDAC1 [12
] and more recently PARIS [36
] and Miro [37
]. In future work, it would be vital to test whether phosphorylation of Parkin at Ser65
influences its ability to ubiquitylate these or other targets and define how this links to PD. Our data also suggest that small molecules that bind to and disrupt the Ubl domain-C-terminus auto-inhibitory interface may activate Parkin in a similar manner to Ser65
phosphorylation. If Parkin inactivation occurs in sporadic PD patients in addition to those harbouring PINK1 mutations, then such small molecule activators could hold significant therapeutic promise in combating disease progression in PD.
Figure 9. Model of Parkin activation by PINK1. Under basal conditions Parkin is kept in a closed inactive conformation by Ubl-mediated autoinhibition. Following mitochondrial depolarization, PINK1 phosphorylates Parkin at Ser65 thereby relieving Parkin autoinhibition (more ...)
In previous work, employing a positional scanning peptide library approach, we elaborated an artificial peptide substrate termed PINKtide that had the sequence WIpYRRS
PRRR. This was phosphorylated by an insect orthologue, TcPINK1, albeit weakly with a Vmax
of 8 U mg−1
and a Km
of 493 μM [3
]. This contrasts with optimal peptides for other active protein kinases that can usually be phosphorylated with a Vmax
of 100–1000 U mg−1
of less than 10 μM. Mutation of the +1 Pro in PINKtide to other residues tested inhibited phosphorylation, suggesting that this residue might represent a key determinant for PINK1 phosphorylation [3
]. However, the sequence encompassing Ser65
of Parkin, DLDQQS
IVHI, is quite dissimilar from PINKtide and does not possess a +1 Pro residue. Our Parkin data suggest that a +1 Pro residue is not an essential determinant for PINK1 phosphorylation. Previous studies, based on co-immunoprecipitation and co-localization experiments [38
], have reported that PINK1 and Parkin bind to each other and therefore it is possible that additional docking interactions between PINK1 and Parkin enable Ser65
to be efficiently phosphorylated. This may also explain why a short peptide, encompassing Ser65
, synthesized by our laboratory was not significantly phosphorylated by TcPINK1 (data not shown).
There has been one previous report that human PINK1 isolated from non-CCCP unstimulated cells can directly phosphorylate Parkin at a single threonine residue, Thr175
]. In that study a deletion fragment of PINK1 spanning residues 200–581 was used that would be predicted to be missing approximately the first 50 amino acids of the N-lobe of the PINK1 kinase domain including the conserved glycine-rich loop motif (residues 163–169), which in other kinases is essential for coordinating ATP. This construct of PINK1 would therefore not be expected to be active in our opinion. Moreover, in that report, the kinase-inactive mutant of the PINK1 [200–581] fragment still exhibited substantial kinase activity towards Parkin [25
]. Taken together these findings indicate that the phosphorylation of Thr175
observed in this study was likely to be mediated by a contaminating kinase. Our experiments have also identified Ser131
as a phosphorylation site in Parkin that is constitutively phosphorylated and not influenced by PINK1 or CCCP (a
lies within the linker region of Parkin between the Ubl and RING0 domain and unlike Ser65
is not fully conserved in lower organisms (e.g. leucine in Drosophila
). A previous in vitro
study has suggested that Ser131
may be phosphorylated by Cdk5 [40
], and further work would be required to define the importance of this phosphorylation site.
Our findings suggest that the full-length form of PINK1 becomes rapidly stabilized within 5 min of CCCP treatment and this also coincides with the disappearance of the cleaved form of PINK1 (a
). PINK1 also becomes activated at 5 min and reaches maximal activity at 40 min as assessed by monitoring phosphorylation of Parkin Ser65
within cells (c
). However, the timecourse of PINK1 autophosphorylation at Thr257
takes longer, requiring around 40 min with activation then being sustained for at least up to 3 h (a
). It is not uncommon for kinases to exhibit differential kinetics of catalytic activity for autophosphorylation as compared with substrate phosphorylation. The kinetics for kinase activity against a substrate generally occurs faster than autophosphorylation and is regarded as a more reliable read-out of kinase activation. Interestingly, previous work has suggested that Parkin can translocate to the mitochondria following CCCP treatment and that this is enhanced when PINK1 is co-expressed in cells [12
]. The kinetics of Parkin translocation when PINK1 is co-expressed is 5 min compared with 30 min when PINK1 is absent [14
]. Given this is similar to the kinetics of Parkin phosphorylation, it would be interesting to test whether phosphorylation of Ser65
influences the kinetics of Parkin translocation to mitochondria in cells.
What drives the stabilization of full length PINK1 at the mitochondria and destabilization of cleaved PINK1 is unknown at present. One proposal is that the PARL protease-mediated cleavage is rapidly inhibited following CCCP-induced depolarization in mitochondria thereby leading to stabilization of the full-length form of PINK1 [9
], while another report suggests that in the presence of CCCP, PINK1 is relocated to the outer mitochondrial membrane rendering it inaccessible by PARL [8
]. The striking disappearance of the cleaved form of PINK1 in the cytoplasm within 5 min (b
) also suggests that CCCP could trigger rapid degradation of this form of PINK1 or else inhibit maturation and processing of PINK1 in the cytosol by PARL-independent proteolysis. In future work, it would be important not only to investigate how full-length PINK1 is stabilized and the cleaved form of PINK1 destabilized, but also to determine whether simple recruitment of PINK1 to mitochondria is sufficient to induce its activation or whether additional depolarization and/or stabilization of the full-length form of PINK1 is a pre-requisite for subsequent activation of PINK1.
It would also be essential to discover the mechanism by which PINK1 is activated at the mitochondria following CCCP treatment. Our data suggest that activation of PINK1 can be observed after immunoprecipitation and extensive washing of the immunoprecipitate, which may be consistent with a covalent modification. Lambda phosphatase treatment substantially reduced PINK1 activation suggesting that phosphorylation is required for full activation (e). In future work, it would be essential to establish what are the key phosphorylation and/or other covalent modifications induced by CCCP and whether these are responsible for triggering activation of PINK1. Given the specificity of activation of PINK1 by agonists that depolarize the Δψm (a,b), it is also possible that PINK1 may become associated with a non-covalent activator at the mitochondrial membrane, such as another protein or a small molecule second messenger that is recruited following Δψm depolarization. It would also be interesting to investigate whether CCCP-mediated generation of an intermediate molecular species, such as reactive oxygen species, could modify PINK1 leading to its activation.
Our data suggest that PINK1, like many kinases, autophosphorylate at Thr257
and probably other residues we have not identified after it is activated. Based on multiple sequence alignments, Thr257
appears to lie within the centre of a region, which is predicted to be an insertion loop and so would likely be accessible for phosphorylation [3
]. This is also associated with an electrophoretic mobility shift on a polyacrylamide gel after CCCP treatment of wild-type but not kinase-inactive PINK1, which would be incapable of autophosphorylation (a
). It should be noted that this band-shift was best observed by resolving proteins on an 8 per cent isocratic polyacrylamide gel and was much less pronounced on gradient gels (see electronic supplementary material, figure S3). This may explain why the band-shift of wild-type PINK1 following CCCP treatment has not been reported before. Although our data suggest that autophosphorylation of Thr257
is not critical for triggering the activation of PINK1 we still feel that phospho-antibodies that recognize Thr257
are likely to be a useful reporter of human PINK1 activation, at least until a biomarker can be generated against the genuine trigger of PINK1 activation, which is currently unknown.
Given the large body of evidence implicating mitochondrial dysfunction in PD [26
], it would be important to explore in subsequent work whether Ser65
and perhaps Thr257
phosphorylation could have utility as specific biomarkers for PD progression. It would also be interesting to look at Ser65
phosphorylation in transgenic mouse models of PD (e.g. α-synuclein) to determine whether this pathway is implicated in other genetic forms of PD. It would also be important to examine Ser65
phosphorylation in brains and cell lines of patients with Parkin and PINK1 mutations and, more importantly, sporadic PD. In future work, it would be important to raise state-of-the-art biomarker-grade monoclonal antibodies against these residues.
In conclusion, we have added important new information that supports the notion that PINK1 and Parkin function in a common signalling pathway. Our data suggest that Δψm depolarization induces stabilization and activation of PINK1 at the mitochondria enabling it to directly phosphorylate Parkin at Ser65 within the N-terminal Ubl domain leading to activation of the Parkin E3 ligase activity. We also provide evidence that, once activated, PINK1 autophosphorylates at several residues and this is associated with an electrophoretic band-shift on a polyacrylamide gel. We have identified one of these autophosphorylated residues as Thr257 and provided evidence that this could serve as a reporter for PINK1 activation. Our findings provide reagents and a framework for exciting follow-up studies. Firstly, it will be crucial to understand how CCCP and other agents that induce Δψm depolarization activate PINK1. It will also be essential to identify the key substrates that Parkin ubiquitylates in response to Ser65 phosphorylation. Hopefully such information could provide valuable clues as to how disruption of the PINK1-Parkin signalling pathway leads to PD and whether this pathway is also disrupted in patients with the sporadic form of the disease. These studies are imperative as they could lead to new ideas for therapies to treat PD in the future, for example by developing a chemical probe that activates Parkin by disrupting the Ubl-domain-mediated autoinhibition of Parkin.