Two in-press papers in the Journal of Biological Chemistry reveal the crystal structure of DJ-1, a protein of unknown function that was recently implicated in the etiology of familial Parkinson's disease (see ARF related news story).

Working independently, Xiao Tao and Liang Tong at Columbia University, New York, and Fuyuhiko Inagaki and colleagues at Hokkaido University, Sapporo, Japan, report the three-dimensional structure of the DJ-1 homodimer to resolutions of 1.8 and 1.95 Angstroms, respectively.

Both labs describe a similar overall structure, though Tao and Tong tally up a few more b-sheets (11 vs. 9) and one less a-helix (8 rather than 9) than reported by the Japanese group. First author Kazuya Honbou, from the latter, compared the 3D structure of DJ-1 with other GAT family proteins (Class I glutamine amidotransferases), revealing that DJ-1 has an extra a-helix at the C-terminal end of the molecule. This finding is supported by Tao and Tong. The two reports are also in agreement that DJ-1 is structurally most similar to an intracellular cysteine protease from Pyrococcus horikoshii, suggesting a proteolytic role for the protein, though as Ted Dawson from Johns Hopkins Medical Institute points out, the structural similarity to E. coli heat shock protein 31 suggests that DJ-1 may act as a chaperone.

Both groups also agree that the potential active site on DJ-1 is unlike those found in other GAT members, which have a catalytic triad of amino acids comprising a cysteine, a histidine, and an acidic residue, either glutamic or aspartic acid. Both groups identify cysteine 106 and histidine 126 as likely to be involved in catalysis, but both also fail to detect any acidic residue close enough to form the third leg of the triad. Honbou and colleagues speculate that the additional helix may be endowed with a regulatory role, and prevent the acidic residue from accessing the triad site, whereas in GAT family dimers, the acidic residue is supplied by the other half of the dimer. Alternatively, Tao and Tong suggest that the catalytic site may be a cysteine/histidine diad, as found in the caspases.

Of prime interest to the neurodegeneration field is the role played by leucine 166, which when mutated to proline causes Parkinson's disease. Tao and Tong suggest that the mutation, which lies in the middle of the seventh a-helix, not only would cause this part of the molecule to unfold, but would also destabilize the dimer by disrupting hydrophobic interactions between the two polypeptide chains. This is also the outcome predicted by Honobu and colleagues. Thus, it would seem that a dimer of DJ-1 is essential to prevent Parkinson's, though exactly what role the protein plays is still a mystery.—Tom Fagan


  1. From the standpoint of someone who is interested in oxidative stress signaling, the crystal structure raises several interesting possibilities. First, the conversion of cys-53 to a cys-sulfinic acid is reminiscent of the change that occurs in the prokaryotic transcription factor OxyR in response to hydroperoxide in bacteria. In that scheme, hydrogen peroxide modifies cys-199. This leads to a conformational change in the OxyR protein, which allows transcription of genes that compensate for oxidative stress. OxyR is thus the prototypic redox sensor. It is possible that DJ-1 also functions as a redox sensor.

    If peroxide levels drift above a homeostatic threshold, then cys53 gets converted to a sulfinic acid, leading to a conformational change and an activation of protease activity. One consequence of this activation would be to activate the androgen receptor or other transcription factors as a compensatory response. Another possibility is that DJ-1 controls not survival, but executes death, and that DJ-1, like caspases or calpains, is involved in selective proteolytic cleavage of substrates that ensure cell death.

    I favor the scenario in which DJ-1 is part of a protective pathway because of the genetic association with PD (and the response to toxicants in sperm); however, it is possible that mutations associated with PD decrease the threshold for DJ-1 activation and thereby increase vulnerability of nigral neurons to trigger events irreversibly associated with death.

  2. These papers are important steps forward to clarify the exact function of DJ-1, and, in turn, to promote understanding of the pathogenesis of DJ-1-related and of classical Parkinson's disease.

    These studies report the crystal structure of human DJ-1, identifying a putative active site, and comparing the structure with other members of the DJ-1-ThiJ-PfpI superfamily.

    An important finding in both studies is that DJ-1 is able to form dimers. Moreover, both studies confirm that the residue mutated in the PARK7 patients (L166) is located in an a-helix region, and that the mutation (L166P) appears to induce severe structural consequences. If other disease-linked missense mutations in DJ-1 are identified, they will pinpoint other functionally important domains. This would complement the biochemical and structural biology studies in the common effort to elucidate the role of this fascinating protein.

  3. The solution of the crystal structure of DJ-1 is indeed a great step forward, and the localisation of the L166P mutation is very helpful. Vincenzo Bonifati and colleagues suggested that this might be in a helical region in their Science paper, and it is reassuring to see the crystal structure agree with their predictions.

    A word of caution is warranted, however, about the interpretation of structural data in terms of what it tells us about function. As Tom Fagan points out, a catalytic triad has not been identified. Crucially, the idea that DJ1 might have protease (or kinase) activity still needs to be tested.
    This leads to a bigger issue. Numerous cases are known of structural homologues with low sequence homology, and of many sequence homologues, all of which have very diverse functions. Therefore, arguing for function based on sequence or structural homology is weak. A key experiment now is to purify DJ1 and assay for a variety of possible activities, a list that I suggest should include: protease (caspase-like and serine protease-like), kinase, amidotransferase, and chaperone activity. Not an inconsiderable task, but potentially useful.

    Finally, the cys53ala mutation is a very useful observation, but I wish the paper could have shown the data for that. There are three conserved cysteines in DJ1. Do all of them sense redox states? A systematic analysis is warranted.

Make a Comment

To make a comment you must login or register.


News Citations

  1. New Parkinson’s Gene: DJ Mutations Make Neurons Change Their Tune

Further Reading

No Available Further Reading

Primary Papers

  1. . Crystal structure of human DJ-1, a protein associated with early onset Parkinson's disease. J Biol Chem. 2003 Aug 15;278(33):31372-9. PubMed.
  2. . The crystal structure of DJ-1, a protein related to male fertility and Parkinson's disease. J Biol Chem. 2003 Aug 15;278(33):31380-4. PubMed.