Mitochondrial LonP1 protease is implicated in the degradation of unstable Parkinson's disease-associated DJ-1/PARK 7 missense mutants

DJ-1/PARK7 mutations are linked with familial forms of early-onset Parkinson's disease (PD). We have studied the degradation of untagged DJ-1 wild type (WT) and missense mutants in mouse embryonic fibroblasts obtained from DJ-1-null mice, an approach closer to the situation in patients carrying homozygous mutations. The results showed that the mutants L10P, M26I, A107P, P158Δ, L166P, E163K, and L172Q are unstable proteins, while A39S, E64D, R98Q, A104T, D149A, A171S, K175E, and A179T are as stable as DJ-1 WT. Inhibition of proteasomal and autophagic-lysosomal pathways had little effect on their degradation. Immunofluorescence and biochemical fractionation studies indicated that M26I, A107P, P158Δ, L166P, E163K, and L172Q mutants associate with mitochondria. Silencing of mitochondrial matrix protease LonP1 produced a strong reduction of the degradation of the mitochondrial-associated DJ-1 mutants A107P, P158Δ, L166P, E163K, and L172Q but not of mutant L10P. These results demonstrated a mitochondrial pathway of degradation of those DJ-1 missense mutants implicated in PD pathogenesis.


Results
Degradation of human wild type DJ-1 and missense mutants. As previously stated in the introduction, degradation of ectopically expressed DJ-1 missense mutants could be affected by interactions with the cell endogenous DJ-1 and/or the use of tagged constructs. Accordingly, untagged wild type DJ-1 (WT) and the missense mutants were transfected into DJ-1-null mouse embryonic fibroblasts (MEFs). The mutants L10P, M26I, A107P, P158Δ, E163K, L166P and L172Q were unstable (Fig. 1A,B) after treatment of cells with cycloheximide (CHX) and showed different degradation rates. In contrast, DJ-1 WT and the missense mutants A39S, E64D, R98Q, A104T, D149A, A171S, K175E and A179T were not significantly degraded after treatment of the cells for 24 h with CHX (Fig. 2). The apparent degradation rate of DJ-1 missense mutants, evaluated by the corresponding half-life and ordered from shortest to longest half-life, was: L10P, P158Δ and L166P < A107P < L172Q < E1 63K < M26I (Supplementary Table I). We have previously shown that transfected human DJ-1 WT and point mutants M26I, R98Q, A104T, D149A are stable proteins in N2a mouse cells (expressing the endogenous mouse DJ-1) and L166P is unstable 43 . We extended those previous studies with similar assays to other missense DJ-1 mutants in the same cell line. In these experiments, the protein levels of the missense mutants A39S, A171S, K175E and A179T did not significantly change after treatment of transfected cells with CHX for 24 h (Supplementary Fig. 1A). In contrast, the point mutants L10P, A107P, P158Δ, E163K, L166P (shown again for comparison) and L172Q were degraded at different rates ( Supplementary Fig. 1B).
Although the proteasomal pathway is implicated in the degradation of DJ-L166P, its degradation is only partially prevented by proteasome inhibitors 43 . Therefore, we decided to re-evaluate those results and to test if inhibition of other proteolytic pathways can be more effective. To that end, transfected cells were treated with CHX for 24 h for those missense mutants that have shown longer half-lives: M26I, A107P, E163K and L172Q (Fig. 3A) and for 12 h for those missense mutants with shorter half-lives: L10P, P158Δ and L166P (Fig. 3B), in the absence or in the presence of several protease inhibitors. As shown in Fig. 3, addition of MG132 (proteasome inhibitor) or inhibitors of the autophagic-lyssosomal pathway (NH 4 Cl, NH 4 Cl in combination with leupeptin or 3-methyl adenine together with E64) had no significant effect on the degradation of the missense mutants or DJ-1 WT. Those results suggested that neither the proteasomal nor the autophagic-lysosomal pathways of protein degradation played a major role in the degradation of the unstable DJ-1 missense mutants.

Subcellular localization of human wild type DJ-1 and missense mutants.
A possible clue to understand the pathway of degradation of DJ-1 missense mutants could be their subcellular localization. DJ-1 protein is described to be localized both in the cytoplasm and in the nucleus 2,10,61 and to translocate to the mitochondria upon cell oxidative stress 45,62 . L166P missense mutant is reported to be localized in the cytoplasm and mitochondria 2,63 , E163K mutant was shown to be localized in the mitochondria 63,64 , as well as M26I 63 and L172Q 49 missense mutants. Therefore, the subcellular localization of the human DJ-1 WT and missense mutants was determined after transfection of DJ-1-null MEFs by indirect immunofluorescence. Figure 4 shows the results obtained and Supplementary Fig. 2 shows the analysis of the co-localization of DJ-1 and mitotracker fluorescence. Without doubt, it can be concluded that the immunofluorescence signal of M26I, A107P, P158Δ, E163K, L166P and L172Q co-localize with mitotracker fluorescence (Pearson coefficient ≥ 0.6). In contrast, it was observed a wide-spread pattern of fluorescence through the cell with some mitochondrial co-localization (naked-eye observation, Pearson coefficient > 0.2 and < 0.4) with mitotracker for E64D, R98Q, A104T, D149A and A171S missense mutants. Finally, a wide-spread non-mitochondrial distribution of fluorescence without significant co-localization with mitotracker fluorescence (Pearson coefficient ≤ 0.2) was found for DJ-1 WT, L10P, A39S, A171S and K175E. These results suggested that a good approach to study both, protein degradation and subcellular localization, would be to obtain fusion constructs of DJ-1 mutants with fluorescent proteins. To that end, constructs of M26I and L166P fused to the N-terminus of EGFP were produced. The results obtained after transfection in MEFs from DJ-1-null mice (see Supplementary Fig. 3) clearly indicated that the C-terminal tagging of the unstable L166P missense mutant greatly slows down its degradation rate compared to the untagged version (see Fig. 1  www.nature.com/scientificreports/ after 24 h of incubation in the presence of CHX. Furthermore, the fusion proteins did not reproduce the preferential mitochondrial localization observed by indirect immunofluorescence of the untagged M26I and L166P DJ-1 missense mutants (compare Supplementary Figs. 3, 4). In conclusion, the approach with fluorescent fusion constructs to study protein degradation and subcellular localization was discarded.
To get independent evidence of the subcellular localization, biochemical fractionation studies of DJ-1 mutants transfected in DJ-1-null MEFs were performed. Supplementary Fig. 4 shows the results obtained. Cell fractionation studies showed that DJ-1 point mutants M26I, A107P, P158Δ, L166P, E163K and L172Q showed significant association with the mitochondrial fraction, while they were also clearly present in the cytoplasmic fraction. This interpretation is further strengthened by comparison to similar fractionation studies with transfected DJ-1 WT and missense point mutants L10P, A39S, E64D, R98Q, A104T, D149A, A171S, K175E and A179T that showed a predominant cytoplasmic distribution and were absent in the mitochondrial fraction ( Supplementary Fig. 4).
To check for off-target effects of mLonP1 shRNA, human LonP1 (hLonP1) cDNA was used to rescue the phenotype of mLonP1 silenced cells respect to degradation of two of the unstable DJ-1 mutants. As shown in Fig. 6, over expression of hLonP1 completely rescued the inhibition of degradation of the unstable DJ-1 P158Δ and L166P mutants produced by mLonP1 shRNAs. Taken together, those results clearly allowed the conclusion that LonP1 is the mitochondrial protease is clearly involved in the degradation of unstable DJ-1 mutants A107P, P158Δ, E163K, L166P and L172Q mutants but not in the degradation of DJ-1 L10P mutant.

Discussion
DJ-1 WT is a stable protein when expressed by transfection in DJ-1-null MEFs, similar to what is found for endogenously expressed DJ-1 in other cells 36,38,43,[66][67][68] . The PD missense mutations A39S, E64D, A104T, D149A, K175E and the polymorphic missense variants R98Q and A171S are as stable as DJ-1WT, both in MEFs from DJ-1-null mice ( Fig. 1) and in N2a cells ( Supplementary Fig. 1), expressing the endogenous DJ-1 WT. The DJ-1 point mutants L10P, A107P, P158Δ, E163K, L166P and L172Q were degraded with similar half-lives in MEFs from DJ-1-null mice and in N2a cells. The M26I DJ-1 mutant is slowly degraded in MEFs from DJ-1-null mice (see Fig. 1 and Supplementary Table 1), but degradation is not apparent in N2a cells after treatment of cells with CHX for 24 h, as reported previously 43 . These results indicated that the unstable (and untagged) DJ-1 mutants, qualitatively, are unstable even in the presence of the expression of endogenous DJ-1 WT.
Several laboratories (including ourselves) have reported that the missense DJ-1 L166P mutant is degraded by the ubiquitin-proteasome pathway 2,36-43 . Similar conclusions were reported for the degradation of L10P, P158Δ 46,47 and L172Q 49 point mutants. Nevertheless, several groups reported that proteasome inhibitors are not   (Fig. 3). Similar results with those inhibitors have been reported previously for the degradation of DJ-1 L166P missense mutant 37,39,41 .
In the search for alternative pathways of degradation, experiments of subcellular localization of DJ-1 by immunofluorescence and biochemical cell fractionation studies were performed ( Fig. 4 and Supplementary Fig. 4, respectively). Those data showed that the DJ-1 point mutants M26I, A107P, P158Δ, E163K, L166P and L172Q, but not L10P, are significantly associated with mitochondria. Those results are in agreement with previously published work of the localization of M26I, P158Δ, E163K, L166P and L172Q 2,45,49,63,64,69 ; no data are available for A107P point mutant. The widely used experimental approach of using fusion constructs of the missense mutants with fluorescent proteins as reporters for both degradation and subcellular localization was discarded, as the DJ-1 L166P and M26I proteins fused to EGFP (Supplementary Fig. 3) did not meet the simple criteria  www.nature.com/scientificreports/ To study the possible involvement of mitochondria in the degradation of missense DJ-1 mutants, we hypothesized that mitochondrial matrix LonP1 could be implicated. The experimental evidence presented ( Fig. 5 and Supplementary Fig. 5) clearly indicated that mitochondrial LonP1 is implicated in the degradation of A107P, P158Δ, E163K, L166P and L172Q. We showed (Fig. 6) that the strong inhibition of their degradation by interrupting the expression of mouse LonP1 can be rescued by overexpression of human LonP1 (whose expression is not suppressed by the action of the specific mouse LonP1-targeting shRNAs), indicating the specificity of the observed inhibitory effects. Formally, it can be argued that rescue by a variant mLonp1 mutant construct insensitive to the shRNA may provide additional, but not essential, confirmation. Furthermore, while certainly hLonp1 is overexpressed in the rescue experiments, the results obtained suggest that those DJ-1 point mutants are also substrates for the hLonP1 protease, as expected. In contrast, there is no effect on L10P degradation. The elucidation of the degradation pathway of this missense mutant requires further investigation.
The localization of DJ-1 in the mitochondria and translocation to mitochondrial matrix have been studied 63,69 . The L10P mutation located in strand 1 prevents homodimer formation while the mutant interacts with DJ-1 WT 46,48 . The L10P mutation may hinder the mitochondrial localization, as that region of strand 1 of the 3D-protein structure 4,5,7 has been implicated in mitochondrial localization. The DJ-1 E18A point mutant (a-helix 1) localizes to the mitochondria, but the substitution by alanine of the wild -type sequence at leucine (a.a. 7), valine (a.a. 8), isoleucine (a.a. 9) and leucine (a.a. 10, changed to Pro in the missense mutant) within strand 1 prevents mitochondrial localization 69 . The fact that the mutants ΔC15 and M26I (shown also here, Fig. 4) also localized to the mitochondria 63 , further supports the role of the aminoacid sequence of strand 1 and a-helix 1 (aminoacids 5-28) in the N-terminus as the sequence that either prevents the translocation of DJ-1 to mitochondria, or www.nature.com/scientificreports/ promotes its cytoplasmic retention. But those N-terminal structural elements are clearly insufficient to explain the location to mitochondria, as DJ-1 with point mutations at the C-terminus having an intact N-terminal sequence, also localize to the mitochondria. Clearly further experiments will allow elucidating the structural requirements for the mitochondrial localization and matrix translocation of DJ-1 missense mutants. Mutations in LONP1 gene produce the CODAS syndrome with Cerebral, Ocular, Dental, Auricular and Skeletal abnormalities 70,71 , more recently a bi-allelic mutation (c.2282 C > T, (p.Pro761Leu) in LONP1 results in neurodegeneration with deep hypotonia and muscle weakness, severe intellectual disability and progressive cerebellar atrophy 72 . These observations clearly indicate a relationship of LonP1 with central nervous system abnormalities and neurodegeneration. LonP1 is implicated in the degradation of matrix mitochondrial proteins like aconitase 73 , 5-aminolevulinic acid synthase 74 , TFAM 75 , StAR 76 , complex 1 of the OXPHOS 77 , SDH5 78 and SDHB 79 of complex II of OXPHOS and unfolded proteins, like the matrix located OTC Delta used to promote an unfolding protein response (mitUPR) in mitochondria 80,81 . LonP1 also participates in the default degradation in mitochondrial matrix of the PD linked protein kinase encoded by PINK1/PARK6 gene. Under mitUPR conditions, PINK1 accumulates (increased accumulation was produced with silencing of LonP1) and recruits the PD-linked E3 ligase PARKIN/PARK2 promoting mitochondrial clearance by mitoautophagy 80,82 . Other groups have observed the accumulation of PINK1 upon silencing LonP1 without the need of concomitant induction of mitUPR response 83,84 . PINK1 is initially processed by the mitochondrial processing protease (MPP), other mitochondrial proteases like PARL, ClpPX and AFG3L2 also participate 82 . The processed PINK1, in contrast to the above observations of degradation by LonP1, would be rapidly degraded by the ubiquitin-proteasome pathway 85,86 after polyubiquitylation 87 . In Drosophila, overexpression of DJ-1 rescues the altered phenotype caused by the loss of PINK1, but not of Parkin 88 . Similarly, DJ-1 overexpression also rescues the increased sensitivity of Substantia Nigra to MPTP in PINK1 null mice 89 . Furthermore, the mitochondrial fragmentation phenotype of DJ-1-null cells can be rescued by overexpression of PINK1 or PARKIN 90,91 . Taken together all these results indicate that DJ-1 and PINK1 (PARKIN) probably act in parallel pathways and LonP1 is involved in DJ-1 point mutants and PINK1 degradation, with variable degrees of involvement of the ubiquitin-proteasome pathway and other degradation pathways that may also be subjected to cell-specific determinants.
In conclusion, the PD phenotype presented by patients harbouring homozygous mutations M26I, A107P, P158Δ, E163K, L166P and L172Q can be explained by a "loss of function", similar to the effects produced by DJ-1 mutation that resulted in strong down-regulation or absence of DJ-1 mRNA (deletions, CNV and splicing mutations) because of its instability due to its mitochondrial targeting and degradation mainly by matrix mitochondrial LonP1 protease. A similar situation is the case of patients with homozygous DJ-1 L10P mutation, while its main pathway of degradation remains to be determined. In contrast, the other point mutants (A39S, E64D, A104T, D149A, K175E and A179T) and the rare polymorphisms (R98Q, A171S) might not be truly pathogenic (likely sure for the polymorphic variants R98Q, A171S). Another possibility is that their half-life could still be lower than DJ-1 WT, but escape to our detection limits (24 h), requiring the use of quantitative proteomic studies using SILAC pulse-chase experiments and MS for its determination (far away from the scope and budget of our work). By those methods the DJ-1 WT protein shows a very long half-life (t1/2), from t1/2 = 187 h in mouse NIH3T3 cells 66 to t1/2 = 199 to 299 h as determine in vivo in mouse heart 92 . Accordingly, the pathogenetic mechanism for A39S, E64D, R98Q, A104T, D149A, A171S, K175E and A179T remains to be determined.

Materials and methods
Plasmid constructs. The vectors for expression of untagged human wild type DJ-1 (hDJ-1 WT) and missense mutants M26I, R98Q, A104T, D149A and L166P have been previously described 43  The EGFP fusion protein constructs of hDJ-1 M26I and L166P were obtained by gene synthesis of hDJ-1 cDNA missense mutants including restriction sites 5′ for XhoI and 3′ for AgeI and subcloning into the XhoI-AgeI sites of the pEGFP-N1 mammalian expression vector (GenScript). or N2a cells (3 × 10 5 per well) were seeded in 6-well culture plates and incubated with 700 µL of filtered lentivirus-containing medium (scramble control shRNA or mouse LonP1 shRNA) in the presence of 8 µg/mL polybrene to a final volume of 1.4 mL per well. Culture media was replaced 24 h after infection, 48 h after infection transduced cells were selected by addition of 8 µg/mL puromycin to the culture media for four days. Puromycinresistant cells were transiently transfected with the indicated human DJ-1 constructs and treated with CHX for studying protein degradation, as described above.
For rescue experiments in mouse LonP1 silenced MEFs, puromycin-resistant MEFs were transiently transfected with human LonP1, selected with culture medium containing 250 µg/mL zeocin for four days, transiently transfected with the indicated human DJ-1 constructs and treated with CHX for studying protein degradation, as described above.
Cell fractionation studies. For biochemical subcellular fractionation experiments, DJ-1-null MEFs were transfected with the indicated untagged human DJ-1 constructs and 48 h after transfection cells were washed with cold PBS for three times, pelleted by centrifugation at 110 × g for 5 min at 4 °C, suspended in lysis buffer (20 mM HEPES pH 7.4, 250 mM sucrose, 5 mM MgCl 2 , 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 25 mM NaF, 1 mM orthovanadate, 10 µM leupeptin, 1 µg/mL pepstatin and 1 mM PMSF) and incubated on ice for 10 min, with thoroughly pipetting up and down the suspension each 2 min. Next, the cell suspension was incubated for 5 min at room temperature with lysis buffer containing digitonin (to a final concentration of 50 µg/mL), cell lysis was verified by Trypan blue staining. Cell suspensions were then transfered to ice and centrifuged at 1000 × g for 5 min at 4 °C to remove nuclei, debris and non lysed cells. The supernatant was used as the total fraction (input) and was further centrifuged at 15,000 × g for 15 min at 4 °C. The supernatant was used as the cytosolic fraction whereas the pellet, containing mitochondria, was washed twice with 200 µL of lysis buffer without digitonin and centrifuged at 15,000 × g for 15 min at 4 °C. After the last wash, the pellet was resuspended in a final volume identical to the volume of the cytosolic fraction with a buffer containing 10 mM HEPES pH 7.4, 10 mM KCl, 1 mM DTT, 0.6% NP-40, 1 mM EDTA, 1 mM EGTA, 25 mM NaF, 1 mM orthovanadate, 10 µM leupeptin, 1 µg/ mL pepstatin and 1 mM PMSF, vortexed for 10 s and centrifuged for 30 s at 15,000 × g at 4 °C. The supernatant of this centrifugation was used as the mitochondrial fraction. Samples from the total input, and mitochondrial and cytoplasmic fractions were loaded onto SDS-PAGE and analyzed by Western and immunoblot, as described below. www.nature.com/scientificreports/ Western immunoblotting. After the treatments, cells were directly lysed in SDS-buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 20% glycerol, 10 µM leupeptin, 1 µg/mL pepstatin and 1 mM PMSF). Cell extracts were sonicated for 10 min on ice, centrifuged at 15,000 × g for 10 min and supernatants used to measure total protein concentration with BCA protein assay kit (Thermo Scientific-Pierce, Waltham, Massachusetts, USA). Equal amounts of total protein were loaded onto 10% or 14% SDS-PAGE for Western transfer to PVDF membranes and immunoblotting. Immunoblots were probed with mouse anti-DJ-1 monoclonal antibody (1:1000, MBL, Woburn, Massachusetts, USA, Clone 3E8); rabbit anti-DJ-1 polyclonal antibody (1:1000, Abcam, Cambridge, UK, ab18257); rabbit anti-LonP1 polyclonal antibody (1:1000, Abcam ab103809) and rabbit anti-Tim23 polyclonal antibody (1:1000, Abcam ab230253). Mouse α-Tubulin (1:10,000, DM1A, Sigma, Darmstadt, Germany) monoclonal antibody was used as loading control. The blots were developed with a peroxidase-labeled goat anti-mouse or anti-rabbit secondary antibody (1:5000, Biorad, Hercules, California, USA) and chemiluminiscence detection MF-ChemiBIS 3.2 (DNR Bio-Imaging Systems, Neve Yamin, Israel). Blots were analyzed by quantitative densitometry using Totallab TL100 software (version 1.0, TotalLab Ltd., Newcastle upon Tyne, UK) and protein levels were normalized respect to tubulin.
Immunofluorescence, confocal microscopy and image analysis. Cells grown on coverslips in 24-well plates were stained for 45 min with 250 nM MitoTracker Red (Invitrogen). The coverslips were then washed with cold PBS three times, fixed with 4% paraformaldehyde in PBS for 20 min at room temperature, permeabilized in PBS with 0.1% Triton X-100 and blocked with PBS and 3% BSA for 1 h at room temperature. The coverslips were processed for indirect immunofluorescence by incubation with rabbit anti-DJ-1 polyclonal antibody (1:200, Abcam) or rabbit anti-LonP1 polyclonal antibody (1:100, Abcam) for 3 h at room temperature, washed with PBS three times for 10 min followed by incubation with Alexa-488 or Alexa-647 fluorescentlabelled secondary antibodies (1:1000) in blocking buffer for 1 h. Next, coverslips were washed with PBS three times for 10 min and 1 μg/mL DAPI (4′,6-diamidino-2-phenylindole) was included for nuclear counterstaining in the second of the washing steps. Coverslips with transfected cells with EGFP fusion protein constructs of DJ-1 were processed for direct fluorescence visualization by incubation with 1 μg/mL DAPI for 5 min. Coverslips were mounted with Prolong Gold antifade reagent (Invitrogen) for confocal microscopy observation in a laser scanning microscope (Leica TCS SP5, Wetzlar, Germany).
To quantify the co-localization between MitoTracker fluorescence (red channel) and the immunofluorescence of transfected human DJ-1 constructs (green channel), single planes along the z-axis of the confocal fluorescence images obtained were analysed with the Image Correlation Analysis (ICA) plugin of ImageJ software 93 . After background subtraction, a single cell was defined as a region of interest (ROI) and quantitative co-localization analysis of the pixels for both channels (red and green) was measured using Pearson's correlation coefficient the hDJ-1/MitoTracker. Data are presented as mean ± s.e.m from at least 20 individual cells from two different experiments.
Statistical analysis. Quantitative data are reported as means ± s.e.m from three different experiments.
When indicated, values are expressed as mean ± upper and lower values from two different experiments. Statistical significance between groups was calculated using a two-tailored Student's t-test.