Parkin is part of a stable, non-covalent, ∼110-kDa complex in vivo
In contrast to SDS–PAGE, BN-PAGE is performed in non-denaturing conditions and thus permits separation of intact non-covalent protein complexes (Schägger et al., 1994; Vandenberghe et al., 2005a). We used this technique to determine the native molecular mass of parkin extracted from the brain stem and diencephalon of 2- to 4-month-old mice. Brain homogenates were separated by centrifugation into pellet (P) and supernatant (S) fractions (Fig. 1A). Fraction P was enriched in membranes, as demonstrated by Western blotting for the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit glutamate receptor subunit 2 (GluR2), a neuronal membrane protein; fraction S was mostly cytosolic, as indicated by Western blotting for the cytoplasmic protein α-tubulin (data not shown). After incubation with 1% Triton X-100, solubilized proteins from fractions P and S were further fractionated by centrifugation on 10–50% glycerol gradients (Fig. 1A). The resulting fractions (P1–12 and S1–12) were subjected to BN-PAGE followed by immunoblotting with CS2132, a commonly used, commercially available polyclonal anti-parkin antibody. This antibody detected a band in fraction S6 with a very high molecular mass (450–550 kDa; Fig. 2A). To check the specificity of the CS2132 antibody, we performed parallel experiments on brain extracts from parkin-null mice. However, the 450–550-kDa complex was equally detected in parkin-null extracts, indicating that it was not an authentic parkin complex (Fig. 2A). The observed lack of specificity of the CS2132 antibody was in agreement with the findings of Pawlyk et al. (2003).
Figure 2. BN-PAGE reveals a stable, non-covalent, ∼110-kDa parkin complex in brain. (A and B) Extracts of brain stem and diencephalon from wild-type and parkin knockout (KO) mice were separated into pellet (P) and supernatant (S) fractions, as described in Materials and methods and Fig. 1A. P and S fractions were further fractionated by glycerol gradient centrifugation. Twelve P and 12 S fractions (numbered from the top of the glycerol gradient to the bottom) were analysed by BN-PAGE, followed by Western blotting with either the polyclonal anti-parkin antibody CS2132 (A) or the monoclonal anti-parkin antibody PRK109 (B). The CS2132 antibody detected a 450–550-kDa band in fraction S6, which was also found in parkin-null extract (A). By contrast, the PRK109 antibody (B) revealed a band in fraction S5 at a lower molecular weight (indicated by the arrow), which was absent in parkin knockout brain. The PRK109 antibody also showed some minor, non-specific immunoreactivity at higher molecular weights (indicated by the asterisk) in S5 from both wild-type and parkin knockout. (C) The graph shows how BN gels were calibrated based on the relative mobilities of native protein size markers to determine the molecular weight (MW) of the parkin complex. Markers, denoted by black circles, were thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), lactate dehydrogenase (140 kDa) and bovine serum albumin (67 kDa). The dotted line indicates the estimated MW of the parkin complex. (D) In the left and middle panels, heating brain fraction S5 at 100 °C in 1.5% sodium dodecyl sulphate (SDS) for 10 min immediately prior to BN-PAGE disrupted the ∼110-kDa parkin complex and led to the appearance of monomeric parkin (indicated by the arrow). In the rightmost panel, approximately 50 ng of purified recombinant parkin was analysed by BN-PAGE and Western blot with PRK109 without heat treatment or addition of SDS, revealing a band (indicated by the arrow) with apparent molecular weight (∼50 kDa) consistent with that of monomeric parkin. (E) Omission of Triton X-100 from the extraction protocol did not change the native molecular mass of the parkin complex from brain or the amount of parkin extracted. Numbers to the left of (A), (B), (D) and (E) indicate MWs of the native protein size markers.
Download figure to PowerPoint
We therefore switched to using PRK8 and PRK109, two monoclonal antibodies that were previously demonstrated to be parkin specific (Pawlyk et al., 2003). These antibodies detected a single band concentrated in fraction S5, which represented authentic parkin because it was absent in parkin-null extracts (Fig. 2B). This native parkin band had an estimated molecular mass of ∼110 kDa, based on its mobility relative to native markers of known molecular mass (Fig. 2C; Lee et al., 2002). A small amount of parkin of similar molecular weight was detectable in fraction P5, but only after much more prolonged film exposure (data not shown). There was also some minor immunoreactivity at higher molecular weights (asterisk in Fig. 2B), but this was also found in the parkin knockout and thus probably resulted from non-specific binding. Remarkably, monomeric parkin was undetectable in native brain extracts (Fig. 2B). Importantly, purified recombinant parkin had an apparent molecular mass of ∼50 kDa on native gels (Fig. 2D), indicating that the ∼110-kDa parkin band represented a protein complex rather than an unusual tertiary conformation of the parkin monomer. Boiling brain extracts in 1.5% SDS for 10 min before BN-PAGE completely disassembled the ∼110-kDa parkin complex and led to the appearance of monomeric parkin (Fig. 2D), indicating that the parkin complex was held together by non-covalent interactions.
In the experiments shown in Fig. 2A–C brain extracts were solubilized in 1% Triton X-100. Although this is a mild detergent, we wondered whether any parkin interactions might still be disrupted. However, this was not the case, as shown by the fact that omission of Triton X-100 from the extraction protocol did not affect the apparent molecular mass of the parkin complex on native gels (Fig. 2E). Omission of Triton X-100 also did not diminish the amount of parkin extracted (Fig. 2E), consistent with the previous finding that mouse parkin is highly soluble and easily extracted from brain even in the absence of detergents (Pawlyk et al., 2003).
As an independent approach, we determined the molecular mass of native parkin by gel filtration chromatography of fraction S in the absence of detergents. Parkin eluted from the column as a single peak with an estimated molecular weight of ∼105 kDa (Fig. 3), consistent with the BN-PAGE results. Monomeric parkin was again undetectable.
Figure 3. Demonstration of the parkin complex in brain by gel filtration chromatography. (A) Fraction S prepared in the absence of detergents was applied to a gel filtration column. Fractions of 0.5 mL were eluted from the column and fractions 11–70 were analysed with dot blot for parkin using the PRK109 antibody. Fraction numbers are indicated to the left of the blot. Molecular weights (MWs) indicate the elution patterns of the protein size markers thyroglobulin (669 kDa), ferritin (440 kDa), lactate dehydrogenase (140 kDa), bovine serum albumin (67 kDa), vitamin D-binding protein (55 kDa) and protein phosphatase 2A phosphatase activator (36 kDa). (B) Calibration graph of the gel filtration column. The black circles denote the protein size markers. Kav = (Ve − Vo)/(Vt − Vo), where Ve = elution volume, Vo = void volume and Vt = total column volume. The dotted line indicates the estimated MW of the parkin complex.
Download figure to PowerPoint
The parkin gene is not only expressed in brain but also in other tissues, especially heart and skeletal muscle (Kitada et al., 1998). To find out whether the components of the ∼110-kDa parkin complex were also expressed outside the brain, we applied the protocol shown in Fig. 1A to mouse heart and skeletal muscle. The apparent molecular mass of parkin on native gels was similar in extracts from heart, skeletal muscle and brain (Fig. 4), suggesting that none of the components of the parkin complex was brain specific.
Figure 4. Formation of the ∼110-kDa parkin complex is conserved across tissues. S fractions were prepared from mouse heart, skeletal muscle and brain (Br.). The samples were analysed with glycerol gradient centrifugation, BN-PAGE and Western blotting with the monoclonal anti-parkin antibody PRK8. The arrow indicates the position of the ∼110-kDa parkin complex.
Download figure to PowerPoint
Next we performed immunoprecipitation experiments on brain extracts using the parkin-specific monoclonal antibodies and separated the immunoprecipitates by SDS–PAGE, in order to identify the components of the parkin complex by silver staining of the gels and mass spectrometry. However, comparison of silver-stained gels of immunoprecipitates from wild-type and parkin-null brains failed to reveal any specific bands (data not shown). This was possibly due to low abundance of the parkin complex in brain, low affinity of the monoclonal parkin antibodies, or masking of specific bands on the gel by the light or heavy chain bands of the immunoprecipitating antibody.
As an alternative approach for identification of the components of the native parkin complex, we compared the migration of parkin in glycerol gradients and native gels with that of previously published parkin-binding partners. However, the reported parkin-interacting proteins BAG5 (Kalia et al., 2004), CHIP (Imai et al., 2002), 14-3-3η (Sato et al., 2006), hSel-10 (Staropoli et al., 2003) and Hsp70 (Imai et al., 2002) migrated in characteristic, reproducible band patterns that clearly differed from that of parkin (Fig. 5). By contrast, immunoreactivity for α-tubulin and β-tubulin (Yang et al., 2005), two proteins with monomeric molecular masses of ∼50 kDa, partially overlapped with the 110-kDa parkin band (Fig. 6A). One possible interpretation was that the parkin band represented a stable complex of parkin with monomeric tubulin. Alternatively, the overlap of immunoreactivity could simply be due to the close proximity of the molecular weight of the α/β-tubulin heterodimer to that of the parkin complex. To discriminate between these two possibilities, we performed antibody supershift experiments (Fig. 6B). If parkin and tubulin coexist in a stable complex, incubation of extracts with anti-tubulin antibody before BN-PAGE should change the migration of parkin due to antibody binding to parkin–tubulin complexes (Yang et al., 2002; Vandenberghe et al., 2005a). However, this was not the case, arguing against the existence of a stable parkin–tubulin complex in these extracts (Fig. 6B). This conclusion was further supported by the similar migration pattern of tubulin from wild-type and parkin-null brain extracts (Fig. 6C). Not unexpectedly, immunoreactivity for the 104-kDa protein CASK (Fallon et al., 2002) also partially overlapped with the ∼110-kDa parkin band, but antibody supershift experiments and comparison with parkin-null extracts did not show evidence for the presence of a stable parkin–CASK complex (data not shown). Other published parkin-binding proteins, such as Rpn10 (Sakata et al., 2003), XAPC7 (Dachsel et al., 2005), UbcH7 (Imai et al., 2000; Shimura et al., 2001), cullin-1 (Staropoli et al., 2003), DJ-1 (Baulac et al., 2004), synphilin-1 (Chung et al., 2001), CDCrel-1 (Zhang et al., 2000), α-synuclein (Shimura et al., 2001), p38 (Corti et al., 2003), TRAF2 and IKKγ (Henn et al., 2007), equally failed to show detectable association with the parkin complex on native gels (data not shown). We did not check the migration patterns of the recently reported parkin interactors LRRK2 (Smith et al., 2005), RANBP2 (Um et al., 2006), Eps15 (Fallon et al., 2006) and ataxin-2 (Huynh et al., 2007), because these proteins, with predicted molecular weights of ∼285 kDa, ∼385 kDa, ∼142 kDa and ∼136 kDa, respectively, were too large to be subunits of the ∼110-kDa parkin complex.
Figure 5. Known parkin-interacting proteins do not show detectable association with parkin on native gels. (A–F) S and P fractions from brain were separated by glycerol gradient centrifugation and BN-PAGE, followed by Western blotting for the protein indicated. The dotted ellipses indicate the position of the parkin band, as determined by reprobing each blot with anti-parkin.
Download figure to PowerPoint
Figure 6. Lack of a detectable stable complex of parkin with tubulin. (A) Fraction S from brain was separated by glycerol gradient centrifugation, and fractions S4–S6 were analysed by BN-PAGE and Western blotting for parkin, α-tubulin or β-tubulin (β-Tub.). The dotted ellipses indicate the position of the parkin band, as determined by reprobing each blot with anti-parkin. In the rightmost panel, fraction S was boiled in 1.5% sodium dodecyl sulphate (SDS) for 10 min before BN-PAGE to identify the monomeric tubulin band. (B) Fractions S4–S6 were incubated for 90 min with 5 µg of control rabbit IgG, rabbit anti-α-tubulin (Tub. Ab), or buffer alone (Tris) before BN-PAGE. After Western blotting (WB) with mouse anti-α-tubulin (left panel), the blot was stripped and reprobed with mouse anti-parkin (right panel). Pre-incubation with polyclonal anti-α-tubulin fully shifted the α-tubulin bands to higher molecular weights due to antibody binding and cross-linking of tubulin complexes (left panel), but did not shift the parkin band (right panel). (C) The migration pattern of native β-tubulin from parkin knockout (KO) brain is similar to that in the wild-type extracts shown in (A).
Download figure to PowerPoint
The ∼110-kDa parkin complex is not a parkin homodimer
We considered the possibility that the ∼110-kDa parkin complex might be a parkin homodimer. However, this was highly unlikely for four reasons. First, the apparent molecular mass (∼50 kDa) of purified recombinant parkin on native gels (Fig. 2D) already suggested that parkin does not easily form homodimers. Second, parkin overexpression experiments in cell lines argued against a homodimeric composition of the parkin complex (Fig. 7). Endogenous parkin was not detectable with our BN-PAGE assay in untransfected COS1, SH-SY5Y, CHO or HEK-293 cells (data not shown). However, when we transfected parkin in COS1, SH-SY5Y or CHO cells, the protein accumulated predominantly as a monomer, with only minor formation of the ∼110-kDa complex (Fig. 7A–C). This strongly suggested that the ∼110-kDa complex was not a parkin dimer, but rather contained components that were present in limiting amounts in transfected COS1, SH-SY5Y and CHO cells. Interestingly, the abundance of the ∼110-kDa complex (relative to that of the parkin monomer) was dramatically higher in transfected HEK-293 cells (Fig. 7D) than in COS1, SH-SY5Y or CHO cells, suggesting that HEK-293 cells expressed higher endogenous levels of the binding partners required for complex formation.
Figure 7. Overexpressed parkin accumulates predominantly as a monomer in several cell lines. (A) COS1 cells were transiently transfected with parkin and extracted in 1% Triton X-100, followed by glycerol gradient centrifugation of the extract. Glycerol gradient fractions 2–11 were analysed by BN-PAGE and immunoblotting for parkin. Upon overexpression parkin migrated as a monomer (indicated by the arrow on the left of the first panel). After more prolonged film exposure, a small amount of the ∼110-kDa complex was detected, as illustrated by comparison with the endogenous parkin band (indicated by the arrowhead) from brain (Br.). (B–D) SH-SY5Y (B), CHO (C) and HEK-293 (D) cells were transiently transfected with parkin and extracted in 1% Triton X-100. Glycerol gradient fractions 2–6 were analysed by BN-PAGE and immunoblotting for parkin. (B) The results of both brief and more prolonged film exposure are shown. In the rightmost panel of (B), total Triton-soluble extract from transfected SH-SY5Y cells was boiled in 1.5% SDS for 10 min immediately before BN-PAGE to identify the monomeric parkin band. The arrows and arrowheads indicate monomeric parkin and the ∼110-kDa parkin complex, respectively. (E) Either 30 or 250 ng of parkin cDNA was transfected per cm2 of COS1 cell culture. For each transfection condition, 20 µg of total protein extract was examined by sodium dodecyl sulphate (SDS)–PAGE and Western blotting, showing different total parkin expression levels. The mobility of a 50-kDa molecular weight marker is shown on the left of the blot. (F) Extracts of COS1 cells transfected with either 30 or 250 ng of parkin cDNA per cm2 were analysed by glycerol gradient centrifugation, and fractions 3–5 were subjected to BN-PAGE and parkin immunoblotting to visualize monomer (arrow) and ∼110-kDa complex (arrowhead). In the right panel, total Triton-soluble extract from COS1 cells transfected with 250 ng/cm2 of parkin cDNA was boiled in 1.5% SDS for 10 min immediately before BN-PAGE to identify the monomeric parkin band. When only 10 ng/cm2 of parkin cDNA was transfected, the protein was not detectable with the glycerol gradient centrifugation and BN-PAGE assay (not shown).
Download figure to PowerPoint
Third, we also varied the transfection level of parkin in COS1 cells (Fig. 7E and F). With higher total parkin expression levels, there was relatively little change in the amount of complex, but there was a clear increase in the amount of monomer. Thus, formation of the ∼110-kDa complex again appeared to be limited by the endogenous levels of the required parkin-binding partners.
Finally, partial denaturation experiments did not support a (homo)dimeric structure of the parkin complex. Heating extracts from brain (Fig. 8A) or transfected COS1 cells (Fig. 8B) to 65 °C (instead of 100 °C as in Fig. 2D) in 1.5% SDS led to partial dissociation of the parkin complex and the appearance of three parkin bands with estimated molecular masses of approximately 80, 60 and 50 kDa. The most parsimonious explanation for this observation would be that the ∼60-kDa band represented a complex of parkin with a protein other than parkin itself, the ∼80-kDa band reflected the association of a third component and the ∼110-kDa band arose from the addition of yet another subunit. According to this simple model, the ∼110-kDa complex would be a tetramer. Alternatively, it was also possible that, for example, the ∼60-kDa band represented a complex of parkin with more than one other protein, in which case the ∼110-kDa complex would contain more than four components. Hence, the findings suggested that the ∼110-kDa complex was at least a tetramer. In any case, the results shown in Fig. 8 were not consistent with the ∼110-kDa complex being a parkin homodimer.
Figure 8. Partial dissociation of the parkin complex. (A) Heating fraction S from brain to 65 °C for 10 min in 1.5% sodium dodecyl sulphate (SDS) before BN-PAGE led to the appearance of three parkin bands (indicated by the arrows), suggesting that the ∼110-kDa complex was at least a tetramer. In the rightmost panel, fraction S was heated to 100 °C in 1.5% SDS for 10 min before BN-PAGE to dissociate the parkin complex completely and identify the position of the monomeric parkin band. (B) COS1 cells were transiently transfected with parkin and extracted in 1% Triton X-100. The cell extracts were treated in the same way as the brain extracts in (A).
Download figure to PowerPoint
PD-linked parkin mutations compromise parkin complex formation
Many pathogenic parkin mutations have been shown to cause a loss of parkin function by impairing its solubility, stability or catalytic activity (Henn et al., 2005; Sriram et al., 2005). However, a subset of PD-linked parkin mutations (including the A82E, K161N, K211N and R256C missense mutations) have little impact on parkin solubility, stability or interaction with known binding partners or on its intrinsic catalytic activity in a cell-free system (Hampe et al., 2006). The pathogenic mechanism of these mutations thus remains elusive. We therefore explored whether the A82E, K161N, K211N and R256C mutations disrupted the ∼110-kDa complex. We transfected wild-type, A82E, K161N, K211N and R256C parkin cDNA into COS1 cells to determine the relative amounts of parkin complex and monomer. Importantly, Triton-soluble parkin levels were similar for the wild-type, A82E, K161N, K211N and R256C variants, as shown by parallel control experiments (Fig. 9C and D), thus precluding confounding effects of parkin expression levels on the observed complex/monomer ratios. Interestingly, each of the four mutations significantly impaired formation of the ∼110-kDa parkin complex (Fig. 9A and B). Quantitatively, the A82E, K161N, K211N and R256C mutations caused a ∼40%, ∼60%, ∼50% and ∼30% reduction, respectively, of the amount of parkin complex relative to total parkin (Fig. 9B). Given the similar soluble parkin expression levels (Fig. 9C and D), the differences in parkin complex/monomer ratio between wild-type and mutant proteins indicated a reduced intrinsic ability of the mutants to form the ∼110-kDa complex. We also examined the R275W parkin variant, which was previously shown to have reduced, but non-zero, solubility in Triton X-100 (Hampe et al., 2006). The Triton-soluble level of R275W parkin was indeed lower than that of the other variants (Fig. 9C and D). A small amount of R275W parkin monomer was detected, but no complex was observed (Fig. 9A).
Figure 9. Comparison of ∼110-kDa complex formation between wild-type (WT) parkin and PD-linked parkin variants. COS1 cells were transiently transfected with 70 ng/cm2 of WT or R256C mutant parkin cDNA, 50 ng/cm2 of A82E or K161N parkin cDNA, 30 ng/cm2 of K211N cDNA and 120 ng/cm2 of R275W cDNA. (A) Transfected COS1 cells were extracted in 1% Triton X-100, followed by glycerol gradient centrifugation of the extracts. Fractions 2–6 of the gradient were analysed by BN-PAGE and parkin immunoblotting to visualize monomer (arrowhead) and ∼110-kDa complex (arrow). (B) In the experiments shown in (A), the amounts of parkin complex and parkin monomer were quantified. The graph represents the amount of parkin complex, expressed as a percentage of the sum of parkin complex and monomer. Asterisks denote significant difference (P < 0.05) from WT. (C and D) In parallel with each of the experiments shown in (A and B), 15 µg of Triton-soluble protein extract was analysed with SDS–PAGE and parkin immunoblotting to compare soluble parkin protein levels between WT and mutant variants. At the low signal intensity of the blots shown in (C), parkin appeared as a doublet due to the presence of an N-terminally truncated parkin species generated through an internal translation initiation site (Henn et al., 2005). No endogenous parkin signal could be observed by SDS–PAGE in untransfected (Untransf.) COS1 cells (C), except after very prolonged film exposures (not shown). The graph in (D) shows SDS–PAGE and Western blot quantification of the Triton-soluble parkin levels. Within each experiment, the level of the parkin mutants was normalized to that of WT parkin. There were no significant differences in soluble parkin levels between WT, A82E, K161N, K211N or R256C (P = 0.96 by one-way anova).
Download figure to PowerPoint