Parkin solubility declines in the ageing human brain including in the Substantia nigra
Parkin’s biochemistry in the human brainstem vs. other regions of the neuroaxis has remained largely unexplored [73]. We serially fractionated 20 midbrain specimens (ages, 26–82 years) and > 40 cortices (ages, 5–85 years) from human subjects, which had been collected post mortem (Fig. 1, Supplementary Fig. 1; Supplementary Table 1, online resource). In control brain, we found that before the age of 20 years, nearly 50% of cortical parkin was found in soluble fractions generated by salt [Tris-NaCl; TS]- and non-ionic detergent [Triton X-100; TX]-containing buffers (Fig. 1a, b; Supplementary Fig. 1a, online resource). In contrast, after age 50 years, parkin was predominantly (> 90%) found in the 2% SDS-soluble (SDS) fraction and the 30% SDS extract of the final fractionation pellet (P). The same distribution was seen in adult midbrain (e.g., S. nigra; red nucleus), the pons (e.g., L. coeruleus), and the striatum (Fig. 1a, b; Supplementary Fig. 1a–c, online resource).
Intriguingly, in older individuals (ages, ≥ 50 years) approximately half of the detectable parkin remained soluble in the human spinal cord and skeletal muscle specimens, which had also been collected post mortem (Fig. 1c, d). We used univariate linear regression analysis to explore a correlation between soluble parkin (of TS- and TX-fractions relative to the total signal for parkin, plotted as %) and age in human control cortices (Fig. 1e). The regression coefficient of age was − 0.54 (at a 95% confidence interval (CI) of − 0.79 to − 0.29, P = 7.7e−05), where the multiple R-squared value was 0.302. When defining parkin solubility as a binary variable, i.e., the presence or absence of soluble parkin in TS- and TX-fractions (absent defined as less than 2% of total signal), and using logistic regression analysis, we found that the transition to insoluble parkin occurred between the ages of 28 years (with low sensitivity but high specificity values) and 42 years (with high sensitivity and low specificity values) (data not shown).
This age-dependent partitioning of parkin was not seen for any other protein examined, including two other PD-linked proteins, i.e., DJ-1 and α-synuclein (Fig. 1a, f), or for organelle-associated markers, e.g., cytosolic glyoxalase-1, peroxiredoxin-1 and -3; and endoplasmic reticulum-associated calnexin. Notably, mitochondrial markers, e.g., voltage-dependent anion channel (VDAC) and Mn2+-superoxide dismutase (MnSOD), also did not partition with parkin (Fig. 1g; Supplementary Fig. 1b, c, online resource; and data not shown). In contrast, parkin did co-distribute with LC3B, a marker of protein aggregation, foremost in brain specimens from older individuals (Fig. 1a, h; Supplementary Fig. 1c, online resource).
The age-associated loss in parkin solubility appeared unique to the human brain in that it remained predominantly soluble in the adult nervous system of other species, e.g., mice and rats as well as cynomolgus monkey, which were processed in the same way (Fig. 1i). Specifically, in brain lysates of two different wild-type strains of mice (C57Bl/6J, and a mixed 129S//FVB/N//C57Bl/6J background) aged to 18 and 22 months respectively, parkin remained present in the soluble fraction throughout their lifespan (Supplementary Fig. 1d, online resource; and data not shown).
In soluble fractions from older humans, we did not detect any truncated species of parkin using several, specific antibodies (data not shown). Despite the loss of parkin solubility with progression in age, PRKN mRNA was detectable in individual neurons isolated from the S. nigra and cortex throughout all age groups; there, transcript levels in neurons did not correlate with the subject’s age (Supplementary Fig. 1e, f, online resource).
Our analysis comprised samples with post mortem intervals (PMI) that spanned from 2 to 74 h (Supplementary Table 1, online resource). Using univariate linear regression analysis, we detected no correlation between parkin solubility in human control cortices (n = 41) and PMI length, where the regression coefficient for PMI measured − 0.15 (95% CI: − 0.76 to 0.46, P = 0.62), and the multiple R-squared value was 0.0064 (Fig. 1j). As expected, PMI did not correlate with the age of the deceased person (not shown). Likewise, wild-type parkin was found to be largely insoluble in striatal, midbrain and pontine samples isolated from aged subjects with PMIs as short as 2 to 6 h (Fig. 1k; Supplementary Fig. 2a, b, online resource). We further explored a possible contribution of PMI to parkin solubility by mimicking conditions of some of the human autopsy cases, using adult mice. This included a PMI length of up to 40 h, where animals were kept at room temperature for the first 14 h, followed by storage over 26 h at 4 °C before removal of their brain; in these cases, parkin remained in the soluble compartments (Fig. 1l and data not shown). While we cannot exclude that PMI length could affect parkin’s solubility in some cases, the age-dependent loss of parkin solubility observed in human brain samples of our cohort was not due to the PMI.
Further, we determined that the decline in detectable parkin solubility in the aged human brain did not differ based on the sex of the deceased person, such as when examined by univariate linear regression analysis or by multivariate analysis (data not shown); it was also not caused by either tissue freezing prior to protein extraction or the pH value of the buffer (Supplementary Fig. 2c–f, online resource). Moreover, employing the commonly used ‘RIPA buffer’ instead of our serial extraction buffers resulted in the release of parkin into the supernatant with some reactivity left in the pellet, as expected (Supplementary Fig. 2g, online resource).
Decline in parkin solubility correlates with rising hydrogen peroxide levels in the mammalian brain
We next explored a possible association between parkin distribution, age and oxidative changes. Using sister aliquots from the brain specimens examined above, we found that hydrogen peroxide (H2O2) concentrations positively correlated with age (Fig. 2a, b; see also Supplementary Table 1, online resource), as expected from the literature [54]. Using univariate linear regression analysis, we determined that the coefficient of age was 0.067 (95% CI: 0.035 to 0.098, P = 3e−04; multiple R-squared value, 0.4877).
In three brains from subjects with non-PRKN-linked parkinsonism, the levels of H2O2 were similar to those measured in age-matched controls (Fig. 2b). When analyzing parkin distribution vs. H2O2 concentrations in human cortices, we found that parkin solubility in human brain negatively correlated with H2O2, where the coefficient of the latter was − 4.2 (95% CI: − 7.92 to − 0.48, P = 0.029; multiple R-squared value, 0.2174) (Fig. 2c).
We next sought to dynamically model the observed correlation between higher ROS levels in the nervous system and reduced parkin solubility. We first used an ex vivo approach, whereby wild-type mouse brains were exposed to either saline or H2O2 during tissue homogenization. There, we saw a significant reduction in soluble parkin and an increase in insoluble parkin in the H2O2-exposed lysates (Fig. 2d, e). We next examined two in vivo models. In the first, wild-type mice were intraperitoneally injected with 40 mg/kg of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) toxin one hour before sacrificing them to induce acute oxidative stress, but no cell death [3]. Brains were serially fractionated, and parkin distribution was quantified across soluble and insoluble compartments. There, we measured a decrease of murine parkin in the soluble fraction and a corresponding rise in the insoluble fractions of MPTP- vs. saline-injected littermates (Fig. 2f, g).
In the second in vivo model, we observed a similar, significant shift in parkin distribution toward insolubility in adult mice that were haploinsufficient for the Sod2 gene, which encodes mitochondrial MnSOD, and which occurred in the absence of an exogenous toxin (Fig. 2h, i). Of note, in both models we confirmed the expected rise in H2O2 levels (see below and El-Kodsi et al. [17]). Moreover, in contrast to murine parkin, the solubility of endogenous Dj-1, which is encoded by a second, ARPD-linked gene, was not visibly altered under these elevated oxidative stress conditions, as monitored by SDS/PAGE/Western blotting (Fig. 2h).
Parkin is reversibly oxidized in the adult human brain
The correlation between parkin insolubility and H2O2 levels in the human brain suggested to us that the relation could be due to posttranslational, oxidative modifications. Indeed, in contrast to SDS-containing brain fractions analyzed under reducing conditions (+ DTT), when gel electrophoresis was performed under non-reducing (-DTT) conditions, we detected parkin proteins ranging in Mr from > 52 to 270 kDa, invariably in the form of redox-sensitive, high molecular weight (HMW) smears (right vs. left panel; Fig. 3a). We saw the same pattern in fractions prepared from control midbrains; no such reactivity was seen in SDS-extracts of parkin-deficient ARPD brains, thus demonstrating the specificity of protein detection.
We confirmed that reversible oxidation of brain parkin was also present in soluble (TS-, TX-) fractions, albeit at lesser intensities (Fig. 3b; data not shown). Of note, the formation of high Mr parkin was not due to secondary oxidation in vitro, because specimens had been processed and fractionated in the presence of iodoacetamide (IAA) prior to SDS/PAGE in order to protect unmodified thiols. These HMW parkin smears also did not arise from covalent ubiquitin-conjugation, such as due to auto-ubiquitylation of parkin, owing to the fact that such adducts cannot be reversed by reducing agents (e.g., DTT) (data not shown).
Because we predicted that the loss of parkin solubility was due to thiol-based, posttranslational oxidation events [50], we first sought to test this in vitro using purified, tag-less, full-length, recombinant (r-) parkin. There, we observed the H2O2 dose-dependent formation of HMW smears and loss of parkin solubility; however, r-parkin protein solubility was greatly recovered by adding DTT (Fig. 3c; Supplementary Fig. 3a, online resource) or β-mercaptoethanol (not shown). Demonstrating its sensitivity to bi-directional redox forces, the exposure of native r-parkin to excess DTT also rendered it increasingly insoluble (Supplementary Fig. 3b, online resource), likely due to loss of zinc-sulfur chelation in its four RING domains [31, 47]. Unlike r-parkin, the addition of up to 1 M DTT in the extraction buffer did not induce parkin’s extraction into a soluble phase (i.e., TS- or TX-fractions) in aged human brain tissue (Supplementary Fig. 3c, online resource).
We confirmed by mass spectrometry (MS) of the holoprotein carried out without any trypsin digestion that all 35 cysteine-based thiol groups of human r-parkin are principally accessible to alkylation by IAA (right vs. left panel; Supplementary Fig. 3d, online resource). These results unequivocally demonstrated that each parkin cysteine theoretically possesses the capacity to have its thiol be modified. Nevertheless, in these in vitro experiments we consistently observed a concentration-dependent change in r-parkin solubility, thereby suggesting that some thiols were more amenable than others to modification by reactive species (see below and summary in Supplementary Table 2, online resource).
Oxidative conditions alter parkin structure
The progressive insolubility of brain parkin and r-parkin due to redox stress suggested that the protein had undergone structural changes. Indeed, when we analyzed the effects of spontaneous oxidation using native r-parkin by far-UV-circular dichroism (Fig. 3d), soluble fractions initially contained both α-helically ordered as well as unstructured r-parkin proteins. Five days later, r-parkin preparations were separated by centrifugation and fractions re-analyzed. There, we found a marked shift to increased β-pleated sheet-positive r-parkin in insoluble fractions (Fig. 3d). Similarly, when we monitored r-parkin during spontaneous oxidization using dynamic-light scattering (Supplementary Fig. 3e, online resource), we observed a gradual shift in the hydrodynamic diameter from 5.1 nm, representing a folded monomer, to multiple peaks with larger diameters 5 h later. The latter indicated spontaneous multimer formation, which was partially reversed by the addition of DTT (right panel; Supplementary Fig. 3e, online resource). Thus, these structural and solubility changes of r-parkin were congruent with our immunoblot results for human brain parkin (Fig. 3a).
Hydrogen peroxide modifies parkin at multiple cysteines
To determine whether the oxidation of cysteines and/or methionine residues caused parkin insolubility, we analysed r-parkin that was treated with and without H2O2 and/or thiol-alkylating agents using liquid chromatography-based MS (LC–MS/MS). To differentiate reduced from oxidized cysteines we used a serial thiol-fingerprinting approach, which labelled reduced thiols with IAA, and tagged reversibly oxidized thiols with N-ethylmaleimide (NEM) after their prior reduction with DTT (Fig. 3e). The first test was to determine how progressive oxidation affected thiol accessibility. As expected, using the strong alkylating agent IAA on the nascent protein (and trypsin digestion to map individually modified peptides), we confirmed that the majority of parkin cysteines were reactive (Supplementary Fig. 3d; Supplementary Table 2, online resource). Intriguingly, when treating native r-parkin with lower H2O2 concentrations, we identified an average of 19 cysteines (54.3%) to be modified; in contrast, higher H2O2 concentrations increased this number to 32 cysteines (91.4%). These results suggested progressive unfolding of r-parkin with increasing oxidation (Supplementary Table 2, online resource).
Next, we sought to precisely identify the location of oxidized cysteine residues. Using Scaffold PTM-software, we found a rise in the number of oxidized residues (NEM-Cys, range of 3–26), which was proportional to the increase in H2O2 concentrations and appeared to begin in parkin’s RING1 domain at three residues, i.e., Cys238, Cys241 and Cys253 (Supplementary Table 2, online resource; Fig. 3i), but also involved Cys95 in its linker domain (Fig. 3h). Furthermore, when quantifying thiol modifications by MaxQuant software [10], we found a significant drop for the number of cysteines in the reduced state (IAA-cysteines) within the H2O2-treated samples (P = 0.0016; Fig. 3f), as expected.
In accordance, when comparing cysteine oxidation events in soluble and insoluble fractions of untreated vs. oxidized r-parkin preparations, the number of IAA-Cys was significantly decreased in the pellets (P < 0.0001; Fig. 3g). Of note, modifications at methionine residues did not correlate with r-parkin solubility. These collective results unequivocally demonstrated that H2O2-induced oxidation events at cysteine-based thiols are linked to both progressive, structural change and lesser solubility of human r-parkin.
Parkin is also irreversibly oxidized in adult human and mouse brains
We next sought to identify oxidation events at parkin cysteines in vivo by LC–MS/MS. To this end, we examined both cortex-derived, human parkin and brain parkin isolated from intraperitoneally, MPTP toxin- (vs. saline-) treated mice (Fig. 4). Specimens were processed with IAA during homogenization and fractionation to prevent any oxidation artefacts in vitro. Following immunoprecipitation and gel excision of endogenous parkin at the 50–53 kDa range (an example is shown in Supplementary Fig. 4a, b, online resource), we focused on cysteine mapping and the identification of thiol redox states (Fig. 4a). A graphic representation of theoretically possible, thiol-based redox modifications is provided in Supplementary Fig. 4c, online resource).
In human control cortices (n = 12 runs; summarized in Fig. 4a), we mapped a mean of 46.8 and 19.4% of parkin’s wild-type sequence in the soluble and insoluble fractions, respectively. There, we found cysteines in either a redox reduced state (IAA-alkylated Cys + 57; examples shown in Fig. 4b, d) or in oxidized states (e.g., to sulfonic acid Cys + 48). Irreversible oxidation events in human cortices occurred, for example, at Cys95 (Fig. 4c) and Cys253 (Fig. 4e). The relative frequencies of detection for parkin thiols that were found in a reduced state in vivo (and thus, were alkylated by IAA in vitro) in the soluble vs. insoluble fractions of the human brain were 67.3 vs. 38.1%, respectively (Fig. 4a).
Likewise, in saline- and MPTP-treated mouse brains (n = 6 runs), we mapped 25.0 and 51.5% of wild-type parkin, respectively (summarized in Fig. 4a). Interestingly, akin to the findings in the human brain, we identified the murine sequence-corresponding residue Cys252 in either a reduced or in irreversibly oxidized states (Fig. 4f, g). As mentioned, mice do not carry a cysteine at residue 95 (for sequence comparison, see below). The relative frequencies of detection for thiols that were in a reduced state in vivo (and thus, alkylated by IAA in vitro) in parkin from saline- vs. MPTP toxin-treated mouse brains were 92.9 vs. 68.2%, respectively (Fig. 4a). These collective results demonstrate that parkin cysteines are variably oxidized in adult mammalian brain.
Parkin thiols reduce hydrogen peroxide in vitro
A typical redox reaction involves the reduction of an oxidized molecule in exchange for the oxidation of the reducing agent (examples are shown in Supplementary Fig. 4c, online resource). We, therefore, asked whether parkin oxidation resulted in a reciprocal reduction of its environment (Fig. 5; Supplementary Fig. 5, online resource). Using r-parkin, we established that parkin could reduce H2O2 levels in a concentration-dependent manner in vitro (Fig. 5a; Supplementary Fig. 5h, online resource). This reducing activity was not enzymatic, in that it did not mirror the dynamics of catalase, and r-parkin did not possess peroxidase activity (Fig. 5a; Supplementary Fig. 5a, online resource). Rather, the reaction was dependent on parkin’s thiol integrity, because pre-treatment with NEM (or IAA) and pre-oxidation of the protein with H2O2 abrogated the ROS-reducing activity of r-parkin (Fig. 5b; Supplementary Fig. 5b, g, online resource). It thus appeared similar to the effect of glutathione (Fig. 5a; Supplementary Fig. 5a, e, f, online resource).
The anti-oxidant effect by r-parkin was also dependent on its intact Zn2+ coordination, because increasing concentrations of the divalent ion chelator, EDTA, abrogated the activity; the latter could be ameliorated by supplementing the reaction buffer with zinc (Supplementary Fig. 5c, online resource). As expected, the exposure of r-parkin to excess H2O2 (or excess DTT) led to the release of zinc ions from the nascent recombinant protein, as measured in vitro (Supplementary Fig. 5d, online resource).
Interestingly, RNF43 (a distinct E3 ligase that contains a zinc-finger domain), HOIP (an E3 ligase containing a RING domain) and bovine serum albumin (BSA, which akin to parkin has 35 cysteines), did not show any H2O2-lowering capacity (Fig. 5c, d; Supplementary Fig. 5e, online resource). Further, Parkinson’s-linked α-synuclein, which has no cysteines, also had no reducing effect (Fig. 5c, d). These results suggested that the cysteine-rich, primary sequence and the tertiary structure of r-parkin conferred anti-oxidant activity.
We next examined an additional, cysteine-containing, ARPD-linked protein, e.g., r-DJ-1 and two disease-linked variants of full-length r-parkin, p.G328E and p.C431F, as well as a C-terminal RING2-peptide of parkin (r-parkin321C). We also used a second ROS quantification assay for further validation and to expand our dose-dependency studies (Fig. 5e, Supplementary Fig. 5f–m, online resource). There, r-DJ-1 and r-parkin321C showed negligible H2O2-lowering capacity, and the two point-mutants conferred less activity than did wild-type, human r-parkin (Fig. 5e). As expected from typical redox reactions (Supplementary Fig. 4c, online resource), the lowering of ROS in vitro correlated with reciprocal r-parkin oxidation, as revealed by SDS/PAGE, which was performed under non-reducing conditions immediately after the reaction with H2O2 (Supplementary Fig. 5n, online resource).
These results suggested that the anti-oxidant activity by r-parkin was dependent on its reactive thiol content, which we examined next using the Ellman’s reagent. There, full-length r-parkin, r-parkin321C and r-DJ-1 showed the predicted number of reactive thiols, whereas the single point-mutant variants of r-parkin revealed fewer accessible thiols (Fig. 5f). From these results, we observed a linear correlation between thiol equivalencies and the degree of ROS reduction in vitro, demonstrating that a greater number of readily reactive and/or a greater number of accessible thiols in human parkin proteins corresponded with a more effective lowering of H2O2 (Fig. 5g).
Hydrogen peroxide levels are increased in parkin-deficient brain
To explore whether parkin oxidation conferred ROS reduction in vivo, we first quantified H2O2 concentrations in the brains of wild-type and prkn−/− mice. A trend, but no significant difference, was measured under normal redox equilibrium conditions. However, when analyzing brain homogenates from mice treated with MPTP-toxin vs. saline, carried out as above (Fig. 2), we found significantly higher H2O2 levels in the brains of adult prkn−/− mice compared to wild-type littermates (P < 0.001; Fig. 5h). Similarly, in adult humans H2O2 levels were significantly increased in the cortex of PRKN-linked ARPD patients vs. age-, PMI-, ethnicity- and brain region-matched controls [42] (P < 0.05; Fig. 5i). Specimens of three non-PRKN-linked patients with parkinsonism showed H2O2 levels comparable to those from age-matched normal cortices (Fig. 2b, red circles). We concluded that the expression of wild-type PRKN alleles contributes to the lowering of ROS concentrations in adult, mammalian brain.
Parkin prevents dopamine toxicity in cells in part by lowering hydrogen peroxide
To address the question of selective neuroprotection, we revisited the role of parkin in cellular dopamine toxicity studies [51, 104]. We first tested parkin’s effect on ROS concentrations in dopamine-synthesizing, human M17 neuroblastoma cells. There, dopamine exposure of up to 24 h caused a significant rise in endogenous H2O2 (P < 0.05; Fig. 5j), as expected. Wild-type PRKN cDNA expression effectively protected M17 cells against the dopamine stress-related rise in H2O2 levels (P < 0.0001; Fig. 5j). By comparing sister cultures that expressed similar amounts of exogenous parkin proteins, the E3 ligase-inactive p.C431F mutant had a partial rescue effect, whereas p.G328E, which we confirmed to retain its E3 ligase activity in vitro, showed no H2O2-lowering capacity in these cells (Fig. 5j; and data not shown).
Moreover, only wild-type parkin, but none of the mutant variants tested, increased the viability of M17 cells under rising dopamine stress conditions (P < 0.01; Fig. 5k; and data not shown). This protective effect also correlated with parkin insolubility and its HMW smear formation, as expected from previous studies [51]. These posttranslational changes in M17-expressed parkin were not reversible by DTT or SDS (Supplementary Fig. 6a, b, online resource), thereby suggesting irreversible dopamine-adduct formation. Notably, the protection from dopamine toxicity positively correlated with the level of PRKN cDNA transcribed, as confirmed in sister lines of M17 cells that stably express human, wild-type parkin. There, we estimated that ~ 4 ng of parkin protein expressed in healthy, neural cultures neutralized each μM of dopamine added during up to 24 h (Supplementary Fig. 6c, d, online resource).
Parkin binds dopamine radicals predominantly at primate-specific cysteine 95
We next explored which thiols of parkin were involved in the neutralization of dopamine radicals. Covalent conjugation of RES metabolites at parkin residues had been previously suggested [51, 104], but not yet mapped by LC–MS/MS examining the whole protein. Aliquots of r-parkin were exposed to increasing levels of the relatively stable dopamine metabolite aminochrome. As expected, this led to the loss of protein solubility and HMW species formation at the highest dose tested (Fig. 6a, b). These reaction products were then used to map modified residues by LC–MS/MS. Specifically, proteins corresponding to r-parkin monomer (51–53 kDa) and two HMW bands, one at ~ 100 kDa, the other near the loading well, were gel-excised (Fig. 6a), trypsin digested and analyzed.
There, we made the following four related observations: (i) Increasing aminochrome concentrations led to a significant decline in the total number of spectra readily identified by LC–MS/MS as parkin-derived peptides, both in the monomeric and HMW bands (P < 0.001 and P < 0.0001), respectively (Fig. 6c). This indicated to us either a marked loss in solubility (and thus, lesser accessibility by trypsin) or a rise in heterogenous, complex modifications, which rendered the analyte undetectable by LC–MS/MS, or both; (ii) Despite fewer spectra recorded, we identified a significant increase in the number of oxidized cysteines (such as irreversibly modified to sulfonic acid) following aminochrome exposure, in particular within the HMW bands of r-parkin (P < 0.0001; Fig. 6d); (iii) Under these conditions, four distinct forms of dopamine metabolites were found conjugated to parkin cysteines. Mass shifts of + 145, + 147, + 149 and + 151 were identified, which represented covalent attachment by indole-5,6-quinone, two variants of aminochrome (O = ; HO–), and dopamine quinone itself, respectively (Fig. 6e; Supplementary Fig. 7a, online resource); and (iv) Unexpectedly, we identified in Cys95 the most frequently dopamine-conjugated parkin residue (P < 0.0001; n = 98 spectra; Fig. 6e–g; Supplementary Fig. 7b–g, online resource). Other residues of r-parkin, which we identified to carry any one of the dopamine metabolites we tracked, included Cys166, Cys169, Cys182, Cys212, Cys238, Cys293, Cys360 and Cys365, but at a much lesser frequency (Fig. 6e, f; Supplementary Fig. 7h–o, online resource). No dopamine metabolite-related mass shifts were detected in the control samples that had not been exposed to aminochrome, as expected. We noted with interest that residue Cys95 of wild-type parkin, as the most frequently catalogued one to be modified by dopamine metabolites, is also primate sequence-specific (Fig. 6g, h).
Parkin augments melanin formation in vitro, which involves residue cysteine 95
The oxidation of dopamine in the presence of cysteine-containing proteins, which generates covalent adduct-carrying proteins, underlies structural characteristics during the formation of neuromelanin pigment in the human midbrain (and pons), of which biochemical aspects have been modeled ex vivo [18, 19]. Given the observed relations between r-parkin, dopamine radical conjugation, aggregate formation and protein insolubility, we next examined whether melanin formation was altered by the presence of parkin. Indeed, wild-type r-parkin augmented total melanin formation in a protein concentration- and time-dependent manner in vitro (Fig. 7a). Like the wild-type protein, two ARPD-linked, full-length r-parkin variants, p.C431F and p.G328E, also augmented melanin formation in vitro, when monitored over 60 min, whereas r-DJ-1 and BSA showed no effect under these conditions (Fig. 7b).
Interestingly, mutagenesis of residue Cys95 to alanine (p.C95A; Fig. 7c), which was confirmed by nucleotide- and protein sequencing (by LC–MS/MS), completely abrogated the enhancing effect by r-parkin on the polymerization rate of dopamine to melanin (Fig. 7d, e). Of note, in our study all the recombinant proteins heretofore analyzed were used after their N-terminal His-SUMO-tag had been removed; however, the p.C95A-mutant was resistant to enzymatic digestion of the tag from the parkin holoprotein. Therefore, both His-SUMO-r-parkin and His-SUMO-p.C95A were utilized (Fig. 7c–e). Importantly, in parallel experiments we saw no difference in the kinetics of melanin formation between wild-type r-parkin proteins that either carried a His-SUMO-tag or were tag-less (not shown). We concluded that under these in vitro conditions, residue Cys95 was highly relevant to enhanced melanin polymerization by human parkin.
Furthermore, when the p.C95A-variant of parkin was expressed in M17 cells and examined in our dopamine toxicity assay, the mutant protein showed only a partial effect in H2O2 lowering capacity when compared to wild-type parkin, even when p.C95A was expressed at much higher levels (Fig. 7f, g). These results were consistent with our collective LC–MS/MS results of oxidative modifications of parkin at Cys95 (shown in: Figs. 3h, 4c; Supplementary Table 2, online resource). We reasoned from these complementary ex vivo results that wild-type parkin could be associated with the synthesis of neuromelanin in vivo. Therefore, we sought to explore this further in dopamine neurons of human midbrain.
Anti-parkin reactivity localizes to neuromelanin in the Substantia nigra of adult control brain
Subcellular localization studies of parkin in human brains had previously been hindered by the lack of renewable antibodies (Abs) that reliably detect the protein in situ [73, 77, 81, 85]. We, therefore, developed and extensively characterized several, monoclonal Abs of the IgG2b-subtype using preparations of untagged, full-length, human r-parkin as immunogen. To this end, we generated four stable, epitope-mapped clones, i.e., A15165B, A15165D, A15165G, and A15165E. The performance and specificity of these clones had been confirmed by ELISA, dot blot analyses, SDS/PAGE/Western blotting under reducing conditions, which included the usage of ARPD brain extracts, immunoprecipitation from the human brain and indirect immunofluorescence in cellular studies (Supplementary Fig. 8a–c, online resource; Tokarew et al., manuscript in preparation). Importantly, clones A15165D, A15165G, and A15165E were able to specifically detect human parkin in human brain sections by immunohistological methods (see below).
Serial sections of adult, human midbrain from control subjects were developed by traditional immunohistochemistry (IHC) using metal-enhanced 3–3′-diaminobenzadine (eDAB), which generates a black signal for positive immunoreactivity. There, anti-parkin clones A15165D, A15165G and A15165E revealed dark, granular staining throughout the cytoplasm of pigmented cells (ages, ≥ 55 years) (Fig. 8a, b, d). Using sections of anterior midbrains from nine adult control subjects, ≥ 83% of the anti-tyrosine hydroxylase (TH)-positive neurons were also positive for parkin, as quantified by double labelling (Fig. 8c). Under these conditions and Ab concentrations, no anti-parkin signal was generated by clone A15165B, which had been successfully used in IP experiments above (Fig. 4a). Further, in brainstem nuclei outside the S. nigra, for example in neurons of cranial nerve III (CNIII) and the periaquaductal grey, as well as in sections of control cortices anti-parkin clones A15165D, -G and -E also stained vesicular structures adjacent to the nucleus, albeit at a much lesser intensity than pigmented neurons (Tokarew et al., manuscript in preparation).
Intriguingly, sections from younger control subjects (ages, ≤ 33 years) that were processed in parallel revealed less intense, anti-parkin reactivity in S. nigra neurons, which matched the paucity of their intracellular pigment (Fig. 8e); of note, mature neuromelanin consistently generates a brown color in sections developed without any primary Ab. The different immunoreactivities seen between younger vs. older midbrains suggested that the three anti-parkin clones (A15165D, -G and -E) likely reacted with an age-related, modified form of parkin in situ, because the PRKN gene is already expressed in dopamine cells at a young age (Fig. 1b; Supplementary Fig. 1a–d, online resource).
To confirm the specificity of the new anti-parkin clones, we serially stained midbrain sections from a 71 year-old, male ARPD patient, who was entirely deficient in parkin protein due to compound heterozygous deletions of PRKN exons 2 and 3 (Fig. 8f; Supplementary Fig. 9a–c, online resource) [38]. Development of serial sections with anti-parkin clones A15165E, -D and -G revealed no immunoreactivity in surviving midbrain neurons of the S. nigra from this ARPD subject. In the absence of parkin, there was no signal overlap between eDAB reactivity (black color) and either intracellular neuromelanin granules in surviving dopamine cells or with extracellular pigment (brown; Fig. 8f; Supplementary Fig. 9c, online resource). In parallel, development of midbrain sections from individuals with the diagnoses of dementia with Lewy bodies, of non-PRKN-linked, sporadic PD as well as of cases with incidental Lewy bodies readily demonstrated eDAB reactivity overlapping with neuromelanin for all three anti-parkin clones (Supplementary Fig. 9d–g, online resource; and data not shown). These results demonstrated that the staining by the three anti-parkin clones in our microscopy studies of post mortem human brain appeared specific.
Parkin frequently localizes to LAMP-3+-lysosomes within Substantia nigra neurons
Neuromelanin granules have been shown to occur in specialized autolysosomes [111]. When screening for co-localization of parkin reactivity with a variety of markers for subcellular organelles in sections of adult control brain, we detected that immunofluorescent signals by anti-parkin (green) and anti-CD63/LAMP-3 (red) antibodies strongly overlapped with pigmented granules of nigral neurons (Fig. 8g–i; see also Supplementary Fig. 9h, online resource).
Using confocal microscopy, we demonstrated that in adult midbrain anti-parkin signals, as generated by clone A15165E, and neuromelanin granules were frequently surrounded by circular, ~ 2 μM (in diameter)-sized rings of anti-LAMP-3 reactivity (Fig. 8i, j). A z-stack video for the parkin and LAMP-3 co-labelling studies is appended (Supplemental Information_video, online resource). We concluded that in the adult, human midbrain from neurologically healthy controls and in surviving neurons of subjects, who suffer from parkinsonism that is not linked to bi-allelic PRKN deletion, a pool of parkin appears physically associated with neuromelanin pigment in close association with juxtanuclear, lysosomal structures.