Since the discovery that mutations in the gene for DJ-1 cause early-onset Parkinson disease (PD), researchers have been puzzling over what the protein does. Does this DJ rock, march, or swing? Is it a protease, a kinase, or a redox switch (see ARF related news story)? A paper in the November PLoS Biology suggests, in fact, that it waltzes like a chaperone.

To narrow down the potential roles for DJ-1, Asa Abeliovich and colleagues at Columbia University, New York, tested the protein for protease, antioxidant, and catalase activity, finding that it had none of these properties. First author Shoshana Shendelman and colleagues then turned their attention to chaperone activity. Chaperones are produced by cells in response to a variety of stresses, including heat shock and the presence of reactive oxygen species (ROS), and they play a major role in keeping other proteins from denaturing. In bacteria, some chaperones contain an amino acid motif called the ThiJ domain, and the presence of this motif in DJ-1 had led to early speculation that it, too, is a chaperone.

Shendelman and colleagues tested this possibility by measuring DJ-1’s ability to prevent aggregation of citrate synthase and glutathione S-transferase in response to heat shock. The authors found that DJ was able to reduce aggregation by about 50 percent. Significantly, DJ-1 carrying the PD mutation (a proline for a leucine at position 166) failed to prevent aggregation, suggesting that the chaperone activity is what prevents the disease. That, plus having comparable activity to the human chaperone Hsp27, supports the contention that DJ-1 is a bona fide chaperone.

What about prior evidence that DJ-1 acts as a redox switch (see ARF related news story)? The redox state certainly seems to affect DJ-1, because in the presence of the reducing agent dithiothreitol, the protein completely failed to keep citrate synthase out of aggregates. When Shendelman then added the strong oxidizer hydrogen peroxide to the mix, DJ-1 was reactivated—though under these conditions it could only protect about 20 percent of the citrate synthase. Curiously, in contrast to previous studies showing that cysteine 106 is essential for a redox switch that translocates DJ-1 to mitochondria (see ARF related news story), the authors found that it is cysteine 53 that is crucial for the chaperone activity. In fact, a cysteine 106 mutant still protected cells from oxidative stress.

So what does DJ-1 do in vivo? Parkinson disease is characterized by the accumulation in dopaminergic neurons of dense protein aggregates called Lewy bodies (LB). α-synuclein is one of the major components of LBs. Could DJ-1 prevent aggregation of α-synuclein? Apparently it may. The authors found that wild-type DJ-1 prevented the generation of high molecular weight polymers that form when α-synuclein is warmed, whereas the PD mutant form of DJ-1 had no effect. This chaperone activity wasn’t just limited to the test tube, either. In neuroblastoma cells subjected to oxidative stress, overexpression of wild-type DJ-1, but not the mutant form, prevented accumulation of insoluble α-synuclein. And the authors found that in DJ-1-negative cells, detergent-insoluble synuclein also accumulated.

“Taken together, our data strongly support the notion that DJ-1 functions as a redox-dependent protein chaperone to mitigate molecular insults downstream of an ROS burst,” write the authors. And though they state that it is possible that DJ-1 plays additional roles in the mitochondria or nucleus, they also note that they did not observe any re-localization of the protein in cultured cells. But as Mark Cookson writes in an accompanying PLoS Primer, “it is not yet firmly established which activity of DJ-1 is most relevant to recessive parkinsonism.”

In an accompanying paper, Abeliovich and colleagues propose a cellular model for PD based on DJ-1-negative cells derived from embryonic stem (ES) cells. Joint first authors Shendelman, Cecile Martinat, and Alan Jonason, used DJ-1-negative ES cells to generate dopaminergic neurons. The authors found that DJ-1-negative ES cells had increased sensitivity to oxidative stress and differentiated into dopaminergic neurons much less efficiently than normal ES cells. In addition, those dopaminergic neurons that survived the process were much more sensitive to 6-hydroxydopamine, a toxin that leads to oxidative stress and apoptosis. The authors also found that cells lacking DJ-1 were more sensitive to inhibition of the proteasome, thus tying together two pathways that have been linked to PD—oxidative stress and protein degradation by the ubiquitin-proteasome system.

The authors write that “this study presents a novel, ES cell-based genetic approach to the study of neurodegenerative disorders.” Advantages of this would include the ability to study human rather than mouse dopaminergic neurons.—Tom Fagan.

Shendelman S, Jonason A, Martinat C, Leete T, Abeliovich A. DJ-1 is a redox-dependent molecular chaperone that inhibits α-synuclein aggregate formation. PLoS Biology. 2004 November;2:e362. Abstract

Martinat C, Shendelman S, Jonason A, Leete T, Beal MF, Yang L, Floss T, Abeliovich A. Sensitivity to oxidative stress in DJ-1-deficient dopamine neurons: and ES-derived cell model of primary Parkinsonism. PLoS Biology. 2004 November;2:e327. Abstract

Cookson MR. Molecules that cause or prevent Parkinson’s disease. PLoS Biol. 2004 Nov;2(11):e401. Epub 2004 Nov 16. No abstract available. Abstract


Make a Comment

To make a comment you must login or register.

Comments on this content

  1. In this study, Shendelman and coworkers report that DJ-1 inhibits α-synuclein aggregation, and suggest that DJ-1 would prevent α-synuclein toxicity. This is a step forward for the field because it shows a direct molecular link between recessive parkinsonism and PD. As others have commented (Hardy and Langston, 2004), it is not clear if different recessive and dominant genes delineate a single “pathway” in PD. The current paper suggests that there really are overlapping mechanisms.

    There are two controversial aspects of the report that bear further examination. The first is how this chaperone activity relates to the other reported functions of DJ-1, and which residues within the protein are responsible. For example, in our laboratory, substituting cysteine 106 with alanine precludes oxidation of DJ-1 in cells or in the crystal (Canet-Aviles et al., 2004). This is consistent with data from Kinumi et al. (2004) using mass-spectrometry where Cys106 was oxidized more readily than other cysteine residues. Again, in our hands, Cys106Ala fails to protect neurons against MPP+ toxicity (Canet-Aviles et al., 2004). Therefore, in some assays, Cys106 controls the function of DJ-1, which contrasts to the chaperone activity described by Shendelman et al., which is dependent on Cys53. It is possible that there are multiple functions of the protein and these two residues make differential contributions. So, depending on the assay, one may find different effects of mutating either Cys53 or Cys106.

    However, we need a little more replication before concluding definitively that DJ-1 has chaperone activity. Olzmann et al. (2004) have previously tested the chaperone function of DJ-1 using heat-induced aggregation of citrate synthase, similar to Fig. 1D in the Shendelman paper. In the Olzmann assay, none of the different DJ-1 variants tested had chaperone function, and they reported that this is also true under oxidizing conditions. Whether differences in assay conditions between the two reports account for these differing conclusions is not yet clear. Another interesting observation is that DJ-1 does not colocalize with α-synuclein inclusions in intact cells; a possible corollary to this is that DJ-1 does not label Lewy bodies but does decorate some tau inclusions (Rizzu et al., 2004).

    Correlation of this postulated chaperone function to the redox-sensitivity for DJ-1 therefore needs more experimental support. An extension to the work by the Abeliovich group would be to measure neuronal cell death as well as formation of intracellular inclusions. There are several models where α-synuclein kills neurons, and it would be interesting to see how overexpression of wt DJ-1 would compare with the ability of Cys106Ala and Cys53Ala mutants in preventing toxicity. Assuming that α-synuclein toxicity is a consequence of its tendency to aggregate, the simplest prediction of the Shendelman paper is that Cys53A would not protect neurons. However, if the redox-inactive Cys106Ala failed to protect cells but Cys53Ala DJ-1 did, this would imply that the chaperone activity is not required for neuroprotection.


    . How many pathways are there to nigral death?. Ann Neurol. 2004 Sep;56(3):316-8. PubMed.

    . The Parkinson's disease protein DJ-1 is neuroprotective due to cysteine-sulfinic acid-driven mitochondrial localization. Proc Natl Acad Sci U S A. 2004 Jun 15;101(24):9103-8. PubMed.

    . Cysteine-106 of DJ-1 is the most sensitive cysteine residue to hydrogen peroxide-mediated oxidation in vivo in human umbilical vein endothelial cells. Biochem Biophys Res Commun. 2004 May 7;317(3):722-8. PubMed.

    . Familial Parkinson's disease-associated L166P mutation disrupts DJ-1 protein folding and function. J Biol Chem. 2004 Feb 27;279(9):8506-15. PubMed.

    . DJ-1 colocalizes with tau inclusions: a link between parkinsonism and dementia. Ann Neurol. 2004 Jan;55(1):113-8. PubMed.

  2. We thank Drs. Canet-Aviles and Cookson for their thoughtful comments regarding our recently published manuscripts in PLoS. As they note, the main finding in our first paper is that DJ-1 chaperone activity is redox-regulated. Although other mammalian proteins, such as protein disulfide isomerase (PDI), have been shown to harbor redox-regulated chaperone activity, DJ-1 is unique in that its chaperone activity is induced in oxidative conditions and is abrogated in reducing conditions (this is the opposite of PDI). The redox-dependence of DJ-1 chaperone activity completely explains the prior mixed reports of DJ-1 chaperone activity in the literature. As mentioned by Dr. Canet-Aviles, Olzmann et al. (Olzmann et al., 2004) fail to detect chaperone activity for DJ-1, and importantly, they perform chaperone activity assays in the presence of the reducing agent β-mercaptoethenol (1mM). A second group, Lee et al. (Lee et al., 2003), do observe chaperone activity for DJ-1 in in vitro assays. However, Lee et al. did not investigate the effects of reducing agents on DJ-1 chaperone activity directly and therefore did not observe the redox regulation that we describe. Thus, to our knowledge, this is the first evidence of redox regulation of DJ-1 function.

    We have correlated the in-vitro chaperone function of DJ-1 with the ability of DJ-1 to rescue the oxidative stress sensitivity of DJ-1-deficient cells. In an accompanying paper, we describe the analysis of DJ-1 deficient cells and show that such cells display increased sensitivity to oxidative stress, consistent with a prior report (Taira et al., 2004). Furthermore, we went on to perform rescue experiments and show that whereas wild-type DJ-1 efficiently rescues the sensitivity phenotype, the L166P mutant fails to do so.

    It is important to point out that DJ-1 chaperone activity extends to at least three in-vitro substrates (citrate synthase, GST, and α-synuclein) and two in-vivo substrates (α-synuclein and NFL). Thus, DJ-1 activity is relatively broad and it is likely that there are multiple substrates in vivo. α-synuclein may be an especially interesting candidate for the relevant substrate in vivo in the context of PD, but a more likely scenario, given these data, is that DJ-1 has multiple PD-relevant substrates (similar to other small chaperones).

    We performed a structure-function analysis of DJ-1 chaperone activity and focused on cysteines because a prior analysis of an unrelated redox-dependent bacterial chaperone, Hsp33 (Jakob et al., 1999), revealed a role for cysteine modification in redox regulation. (It is important to emphasize that these are not naturally occurring mutations.) We find that C53 is required for chaperone activity of DJ-1, and that C106 is neither required for chaperone activity nor for redox regulation of the activity. Consistent with the in-vitro chaperone assays, we show that the C53A mutant is also defective in rescuing the sensitivity phenotype of DJ-1 deficient cells in vivo, whereas the C106A behaves similar to wild-type. Thus, we correlate the in-vitro chaperone function activity with the ability to rescue DJ-1 function in vivo.

    To our knowledge, two prior studies have touched upon the function of DJ-1 cysteine mutants in vivo, both by overexpression in tumor cells. Our in-vivo functional rescue analysis of cysteine mutants of DJ-1 is somewhat consistent with a prior study from Taira et al. This group overexpresses wild-type or mutant forms of DJ-1 in NIH3T3 cells and exposes these cells to oxidative stress in the form of hydrogen peroxide. They find that both the C53A mutant and a C106 mutant display reduced activity, although the C106 mutant appears to display partial activity. In contrast to these studies, Drs. Canet-Aviles and Cookson report that both C53A and wild-type DJ-1 protect tumor cells from oxidative stress in the form of MPTP. It is quite possible that these different results with respect to cysteine mutants may relate to the disparate assays employed for in-vivo DJ-1 function; for instance, our rescue experiments in knockout cells may not correlate directly with the activity of DJ-1 overexpression in tumor cell lines.

    It is interesting to relate the functional redox regulation of DJ-1 chaperone activity in vitro to prior studies that have reported structural redox modification of DJ-1 in the context of oxidative stress, as Drs. Canet-Aviles and Cookson point out. Mitsumoto et al. (Mitsumoto et al., 2001) originally described the redox structural modification of DJ-1, an isoelectric point shift, and this finding has been nicely confirmed by others (Canet-Aviles et al., 2004; Taira et al., 2004). Both of these groups went on to try and identify cysteine residues that are modified, but unfortunately the results have been conflicting. Taira et al. report that both C53 and C106 are required for the isoelectric point shift, whereas Canet-Aviles report that only C106 is required. Interestingly, the crystal structure data from Canet-Aviles’s paper does reveal evidence of modification of C53, albeit to a lesser extent than C106. Further structure-function studies are necessary to dissect to mechanism of action of DJ-1 chaperone activity and relate this to the structural modification of DJ-1; it is possible that C106 is structurally modified in the context of oxidative stress and yet is not required for DJ-1 function. It would be especially important to identify mutants that alter the redox dependence but do not abrogate the chaperone activity of DJ-1.


    . The Parkinson's disease protein DJ-1 is neuroprotective due to cysteine-sulfinic acid-driven mitochondrial localization. Proc Natl Acad Sci U S A. 2004 Jun 15;101(24):9103-8. PubMed.

    . Chaperone activity with a redox switch. Cell. 1999 Feb 5;96(3):341-52. PubMed.

    . Crystal structures of human DJ-1 and Escherichia coli Hsp31, which share an evolutionarily conserved domain. J Biol Chem. 2003 Nov 7;278(45):44552-9. PubMed.

    . Oxidized forms of peroxiredoxins and DJ-1 on two-dimensional gels increased in response to sublethal levels of paraquat. Free Radic Res. 2001 Sep;35(3):301-10. PubMed.

    . Familial Parkinson's disease-associated L166P mutation disrupts DJ-1 protein folding and function. J Biol Chem. 2004 Feb 27;279(9):8506-15. PubMed.

    . DJ-1 has a role in antioxidative stress to prevent cell death. EMBO Rep. 2004 Feb;5(2):213-8. PubMed.

  3. Although mutations in DJ-1 have been identified as the cause for an early-onset, autosomal recessive form of familial PD, little is presently known about the biochemical function of DJ-1, and it is unclear whether DJ-1 acts in a similar or distinct pathway compared to other familial PD gene products (α-synuclein, parkin and UCH-L1). The paper by Shendelman et al is very interesting because it suggests that DJ-1 is a redox-regulated chaperone that inhibits aggregation of α-synuclein.

    We agree with Drs. Canet-Aviles and Cookson that more experiments are necessary before concluding definitively that DJ-1 has a chaperone function. As described in our published paper (Olzmann et al., 2004), we failed to detect any chaperone activity for DJ-1 and its mutant forms (C106A and L166P) in suppressing the heat-induced aggregation of citrate synthase. Although we included 1 mM β-mercaptoethanol in the elution buffer during DJ-1 protein purification, the actual concentration of β-mercaptoethanol in the chaperone assays was less than 0.01 mM because purified DJ-1 protein was used at a dilution of 1:100 to 1:200. Furthermore, preincubation of DJ-1 with 300 μM H2O2 did not lead to any measurable chaperone activity in our assays. Thus, the discrepancy between our results and other reports (Lee et al., 2003; Shendelman et al., 2004) could not be explained by the redox-dependence of DJ-1 chaperone activity. One possible reason for this discrepancy might be due to differences in DJ-1 proteins and buffer compositions used in the chaperone assays. The DJ-1 protein used in our assays was untagged, whereas both Lee and Shendelman used His- and T7- double tagged DJ-1 (expressed from the vector pET-21a and pET-28a, respectively). Although Shendelman mentioned that removal of the His tag did not alter the DJ-1 chaperone function, it is unclear whether the extra peptide sequence introduced by the T7 tag and the multiple cloning site has any effect on the DJ-1 chaperone activity. Moreover, the differences in the DJ-1 purification procedures and/or in the chaperone assay conditions might also contribute to the different conclusions since the aggregation of citrate synthase is known to be affected by various buffer components (Buchner et al., 1998), such as ammonium sulfate and imidazole used in the Shendelman assay.

    Another controversial issue of the Shendelman paper that requires further studies is whether DJ-1 has protease function. DJ-1 shares significant sequence and structural homology with the PfpI family of intracellular proteases (e.g., PH1704) and with Hsp31, an E. coli chaperone protein that was recently shown to also have protease activity (Honbou et al., 2003; Huai et al., 2003; Lee et al., 2003; Tao and Tong, 2003; Wilson, M. A. et al., 2003). However, unlike PH1704 and Hsp31, DJ-1 does not have a Cys-His-Glu/Asp catalytic triad. Instead, DJ-1 contains the putative nucleophile Cys-106 which has the potential to form a Cys-His diad with His-126 (Tao and Tong, 2003). Although the lack of a complete catalytic triad has been used to argue against a protease function of DJ-1 (Lee et al., 2003; Wilson, M. A. et al., 2003), it has been proposed that DJ-1 may use the Cys-His diad to carry out its protease function (Tao and Tong, 2003). A fully functional Cys-His catalytic diad has been demonstrated in several cysteine proteases, such as the caspases (Stennicke and Salvesen, 1999; Walker et al., 1994; Wilson, K. P. et al., 1994). In support of the DJ-1 protease hypothesis, we have shown that purified recombinant DJ-1 protein possesses intrinsic protease activity in a fluorescence-based protease assay using BODIPY FL-labeled casein as a substrate (Olzmann et al., 2004). Furthermore, the protease activity was completely abolished by the mutation of Cys-106 to an Ala, supporting a function of DJ-1 as a cysteine protease. In contrast, Shendelman et al failed to detect protease activity using the succinylated casein assay. These differences might be accounted for by the sensitivity of the protease assays since the BODIPY casein assay is at least 100 times more sensitive than the succinylated casein assay (Jones et al., 1997). Furthermore, differences in DJ-1 proteins (untagged vs. tagged) and buffer compositions as mentioned earlier, might also contribute to the different conclusions about DJ-1 protease function.

    As described in our paper (Olzmann et al., 2004), the protease activity of purified DJ-1 is weak, implying that there might be activation mechanism(s) for increasing the catalytic efficiency of DJ-1 protease in vivo. Because of the potentially hazardous action of activated proteases in cells, proteases are generally synthesized as an inactive precursor called zymogen or are kept in a latent form before activation (Donepudi and Grutter, 2002; Neurath, 1986). The zymogen or the latent form of proteases often possesses low but significant intrinsic proteolytic activity (Stennicke and Salvesen, 1999), which is consistent with the weak protease activity detected in our assays. The most common known mechanism for activation of proteases is via specific proteolytic cleavage to remove an N-terminal peptide, often referred to as the propeptide (Neurath, 1986). For some cysteine proteases, such as caspases, the activation involves cleavage of an internal linker peptide (Donepudi and Grutter, 2002; Stennicke and Salvesen, 1999). Interestingly, based on the crystal structure of DJ-1, it has been proposed that the catalytic site of DJ-1 may be blocked by its C-terminal peptide (amino acids 175–189) and that DJ-1 protease may be activated by conformational change induced by specific signals or proteolytic cleavage (Honbou et al., 2003). Another known mechanism of activation is the use of a protein co-factor for substantially increasing protease activity, e.g., Apaf-1 as a cofactor for caspase-9 and fibrin for tissue plasminogen activator (tPA) (Donepudi and Grutter, 2002; Stennicke and Salvesen, 1999). Identification of such co-factor(s) and activation mechanism(s) as well as endogenous substrates of DJ-1 will facilitate the characterization of DJ-1 catalytic function because they should increase the catalytic efficiency of DJ-1 in the protease and/or chaperone assays.

    1 Buchner, J., Grallert, H., and Jakob, U. (1998). Analysis of chaperone function using citrate synthase as nonnative substrate protein. Methods Enzymol 290, 323-338. Abstract

    2 Donepudi, M., and Grutter, M. G. (2002). Structure and zymogen activation of caspases. Biophys Chem 101-102, 145-153. Abstract

    3 Honbou, K., Suzuki, N. N., Horiuchi, M., Niki, T., Taira, T., Ariga, H., and Inagaki, F. (2003). The Crystal Structure of DJ-1, a Protein Related to Male Fertility and Parkinson's Disease. J Biol Chem 278, 31380-31384. Abstract

    4 Huai, Q., Sun, Y., Wang, H., Chin, L. S., Li, L., Robinson, H., and Ke, H. (2003). Crystal structure of DJ-1/RS and implication on familial Parkinson's disease. FEBS Lett 549, 171-175. Abstract

    5 Jones, L. J., Upson, R. H., Haugland, R. P., Panchuk-Voloshina, N., Zhou, M., and Haugland, R. P. (1997). Quenched BODIPY dye-labeled casein substrates for the assay of protease activity by direct fluorescence measurement. Anal Biochem 251, 144-152. Abstract

    6 Lee, S. J., Kim, S. J., Kim, I. K., Ko, J., Jeong, C. S., Kim, G. H., Park, C., Kang, S. O., Suh, P. G., Lee, H. S., and Cha, S. S. (2003). Crystal structures of human DJ-1 and Escherichia coli Hsp31 that share an evolutionarily conserved domain. J Biol Chem 278, 44552-44559. Abstract

    7 Neurath, H. (1986). The versatility of proteolytic enzymes. J Cell Biochem 32, 35-49. Abstract

    8 Olzmann, J. A., Brown, K., Wilkinson, K. D., Rees, H. D., Huai, Q., Ke, H., Levey, A. I., Li, L., and Chin, L. S. (2004). Familial Parkinson's disease-associated L166P mutation disrupts DJ-1 protein folding and function. J Biol Chem 279, 8506-8515. Abstract

    9 Shendelman, S., Jonason, A., Martinat, C., Leete, T., and Abeliovich, A. (2004). DJ-1 is a redox-dependent molecular chaperone that inhibits a-synuclein aggregate formation. PLoS Biology November;2, e362.

    10 Stennicke, H. R., and Salvesen, G. S. (1999). Catalytic properties of the caspases. Cell Death Differ 6, 1054-1059. Abstract

    11 Tao, X., and Tong, L. (2003). Crystal structure of human DJ-1, a protein associated with early-onset Parkinson's disease. J Biol Chem 278, 31372-31379. Abstract

    12 Walker, N. P., Talanian, R. V., Brady, K. D., Dang, L. C., Bump, N. J., Ferenz, C. R., Franklin, S., Ghayur, T., Hackett, M. C., and Hammill, L. D. (1994). Crystal structure of the cysteine protease interleukin-1 beta-converting enzyme: a (p20/p10)2 homodimer. Cell 78, 343-352. Abstract40676

    13 Wilson, K. P., Black, J. A., Thomson, J. A., Kim, E. E., Griffith, J. P., Navia, M. A., Murcko, M. A., Chambers, S. P., Aldape, R. A., Raybuck, S. A., and Livingston, D. J. (1994). Structure and mechanism of interleukin-1 beta converting enzyme. Nature 370, 270-275. Abstract

    14 Wilson, M. A., Collins, J. L., Hod, Y., Ringe, D., and Petsko, G. A. (2003). The 1.1-A resolution crystal structure of DJ-1, the protein mutated in autosomal recessive early onset Parkinson's disease. Proc Natl Acad Sci U S A 100, 9256-9261. Abstract


News Citations

  1. Protection Against Parkinson’s—How the DJ Changes Station

Paper Citations

  1. . DJ-1 is a redox-dependent molecular chaperone that inhibits alpha-synuclein aggregate formation. PLoS Biol. 2004 Nov;2(11):e362. PubMed.
  2. . Sensitivity to oxidative stress in DJ-1-deficient dopamine neurons: an ES- derived cell model of primary Parkinsonism. PLoS Biol. 2004 Nov;2(11):e327. PubMed.
  3. . Molecules that cause or prevent Parkinson's disease. PLoS Biol. 2004 Nov;2(11):e401. PubMed.

Further Reading


  1. . DJ-1 is a redox-dependent molecular chaperone that inhibits alpha-synuclein aggregate formation. PLoS Biol. 2004 Nov;2(11):e362. PubMed.
  2. . Molecules that cause or prevent Parkinson's disease. PLoS Biol. 2004 Nov;2(11):e401. PubMed.
  3. . Sensitivity to oxidative stress in DJ-1-deficient dopamine neurons: an ES- derived cell model of primary Parkinsonism. PLoS Biol. 2004 Nov;2(11):e327. PubMed.

Primary Papers

  1. . DJ-1 is a redox-dependent molecular chaperone that inhibits alpha-synuclein aggregate formation. PLoS Biol. 2004 Nov;2(11):e362. PubMed.
  2. . Molecules that cause or prevent Parkinson's disease. PLoS Biol. 2004 Nov;2(11):e401. PubMed.
  3. . Sensitivity to oxidative stress in DJ-1-deficient dopamine neurons: an ES- derived cell model of primary Parkinsonism. PLoS Biol. 2004 Nov;2(11):e327. PubMed.