Crystallizing ideas about Parkinson's disease

Genetic forms of Parkinson's disease (PD) are useful in helping us to understand the underlying pathophysiology of this disorder in both its rare familial forms and the more common “sporadic” type. Three genes are unambiguously causal in different PD families: α-synuclein, parkin, and DJ-1 (reviewed in ref. 1). Two recessive mutations have been found in DJ-1, a large deletion and an L166P point mutation (2). DJ-1 mutations are likely to result in loss of protein function, and, hence, deciphering the normal cellular role of DJ-1 will be key to understanding how mutations cause disease.

DJ-1 has a number of functions, several of which may be relevant to the pathways that underlie PD (3). It was cloned independently in different laboratories examining processes as varied as cellular transformation, RNA binding, and male fertility (4–6). For this reason, and because DJ1 is a member of a large superfamily of proteins, we have to be cautious in assigning a PD-specific function. In this issue of PNAS, the report by Wilson et al. (7) of the detailed crystal structure of DJ-1 adds considerably to our understanding of what DJ-1 is and what it is not. Two additional studies showing DJ-1 at different resolutions are now in press (8, 9). All three structures show a similar overall folding pattern and demonstrate that DJ-1 is a dimer, which is confirmed in two of these papers by additional techniques. There are similarities to previously solved structures within the DJ-1 superfamily, namely the bacterial proteins PH1704 (10), a protease, and heat shock protein 31 (Hsp31) (7), a chaperone from Escherichia coli. Although the different proteins in the superfamily have a general pattern of folding that is similar (Fig. 1), there are also substantial differences. For example, PH1704 is a hexamer compared with the dimeric nature of DJ-1 and Hsp31.

Fig. 1.

Similarities and differences in structures of different members of the DJ-1 superfamily. (A) Monomers of human DJ-1 and two structural homologues, the P. horikoshii protease PH1704 and the E. coli chaperone Hsp31. Each monomer was redrawn from database coordinates at the Research Collaboratory for Structural Bioinformatics (www.rcsb.org) to give broadly similar orientations for all three molecules. In this view, β-strands (green) are found in the center of the structure, surrounded by α-helixes (magenta). Note the presence of an extra α-helix at the C terminus of DJ-1 compared with PH1704, which strongly affects packing of the protein into a dimer. The larger Hsp31 protein is also organized in a similar fashion with a centralβ-sheet rich region surrounded by helices. Note also that all three proteins have a β-hairpin (oriented at the bottom of these figures), which plays a role in the dimerization of DJ-1. (B) A dimer of DJ-1 compared with a trimer of PH1704. DJ-1 has been crystallized as a dimer, the two monomers represented in blue and red. PH1704 is represented as a trimer to illustrate the relative orientation of the three monomers; the protein is actually hexameric and consists of two units similar to the structure shown forming a closed ring. The red monomer is in approximately the same orientation as the red monomer in the DJ-1 dimer.

In each of the recently identified structures of DJ-1, there are key residues that are highlighted in relation to the known or purported function of the protein. Given the relevance to human disease, locating the site of mutations is an important contribution. Leucine 166 is in the penultimate C-terminal α-helix, and the L166P mutation is strongly predicted to disrupt this structural motif. Our recent results suggest that L166P destabilizes the protein (D. W. Miller, R. Ahmad, S. Hague, M. J. Baptista, R. Canet-Aviles, C. McLendon, D. M. Carter, P.-P. Zhu, J. Stadler, J. Chandran, G. R. Klinefelter, C. Blackstone, and M.R.C., unpublished data), in line with the prediction that helixes G and H together contribute to dimer formation. This is also consistent with predictions previously made by Bonifati et al. (2), who based their structural model on the structural homologue PH1704.

Solving the crystal structure of DJ-1 allows us to test predictions about the function of the protein, or more accurately to suggest that some homology-based ideas are unlikely to be correct. In other members of the DJ-1 superfamily, a catalytic Glu/Asp–His–Cys triad is responsible for enzyme activity. Although this triad is present in DJ-1, it appears not to be catalytically competent as the arrangement of the conserved residues is not suitable for proton transfer, as occurs in the nucleophilic attack mechanism used in proteases, amidotransferases, or kinase members of the DJ-1 superfamily. In fact, two of the structural studies report negative results in protease and kinase assays (7, 9), although one paper does label a potential active site (8). Interestingly, a His–Cys pair is oriented in a similar manner to DJ-1 in some groups of proteases such as the caspases (12) that have been shown to act as cellular executioners. Amidotransferase activity is unlikely, because there is no suitable glutamine-binding site. Overall, these results suggest that DJ-1 is not likely to be an enzyme in any of the known classes represented by other members of the DJ-1 superfamily, but do not yet exclude the possibility that there is some form of catalysis.

The equivalent residue to the catalytically active cysteine in human DJ-1 is Cys-106, and structural data indicate that this is a reactive amino acid. Given the differences with known enzymes and the negative results reported above, the role of DJ-1 is still debatable. However, structural studies may give an important clue in explaining the previously reported observation of a response of DJ-1 to oxidative stress (13, 14). There is a shift in the isoelectric point of the protein when exposed to oxidizing agents, and it has been suggested that this might be indicative of a direct cysteine oxidation, perhaps to a cysteine sulfinic acid (13, 14). Wilson et al. (7) point out that Cys-106 is susceptible to radiation-induced oxidation. Honbou et al. (9) refer to unpublished observations that mutating Cys-53 to alanine abolishes the capacity of DJ-1 to respond to oxidative stress. Therefore, one or two specific residues act as a radical sensor and control the function of DJ1 in different intracellular conditions.

Specific residues sense radicals and control the function of DJ1 in different intracellular conditions.

Finally, Tao and Tong (8) have specifically analyzed a series of residues that are thought to be involved with interactions with SUMO. SUMO is one of a number of small ubiquitin-like modifiers whose cellular role is to control, for example, movement of proteins between the cytoplasm and nucleus. It was previously reported (unpublished observations in ref. 15) that DJ-1 is sumoylated at Lys-130 by the E3-like enzyme PIASxα. Tao and Tong (8) solved the crystal structure of an artificial K130R mutation that is hypothesized to be sumoylation deficient. K130 is on a surface-exposed helix and projects into the solvent, thus making it available for covalent modification by SUMO or other modifiers. Mutating this residue to an arginine had no substantial effects on the overall folding of the protein.

What, in summation, do these structural studies tell us about PD? At one level they represent an enormous advance, especially in comparison to the knowledge we have about other PD proteins. The structural biology of α-synuclein has been resolved for several years: in solution it has little structure, being natively unfolded, but adopts a helical conformation on binding to lipid membranes. The structure of parkin is not yet solved apart from a small region at the N terminus, which has been recently described by using NMR (16), and hence there is still much work to be done on this protein.

A key question that remains is how these three gene products relate to each other at the cellular level. Dominant mutations in α-synuclein have been shown to trigger toxicity in susceptible cell types (17). Like other proteins associated with neurological diseases, α-synuclein can aggregate, and it has been suggested that relatively soluble oligomeric forms mediate toxicity (18, 19). Parkin, on the other hand, is able to suppress toxicity mediated by mutant α-synuclein or other triggers (reviewed in ref. 20). Whether DJ-1 has a similar effect in suppressing toxicity is not yet known, but it is possible to use the structural analyses to make some conjectures.

First, the previously mooted idea (2, 13) that DJ-1 is a free radical sensor has greater credence with the observations that there is at least one cysteine with a propensity to react with oxidizing species (7). There are longstanding suggestions that the pathogenic pathway of PD (and other neurodegenerative disorders) involves oxygen and nitrogen radicals or other oxidizing species (21, 22). However, the various genes that unambiguously cause neurodegeneration are not generally those that control oxidative responses. A notable exception is the amyotrophic lateral sclerosis (Lou Gehrig's disease) mutations in the prototypic antioxidant enzyme Cu/Zn superoxide dismutase (SOD1). However, mutations generally do not affect enzyme activity and probably only affect free radical metabolism indirectly (23). Instead, the mutations, which are widely dispersed along the protein, promote protein aggregation events, as has been elegantly demonstrated recently by using structural biology techniques (24). Such data do not dismiss the role of free radicals in neurodegeneration, e.g., by promoting protein aggregation (reviewed in ref. 22), but call into question the primacy of oxidative events. If the genes that trigger neurodegeneration are unrelated to oxidative metabolism, then it is difficult to argue that free radicals trigger pathology, although perhaps there are differences between familial and sporadic disease. However, if DJ-1 is authentically an oxidative stress response protein, then this resurrects the idea that oxidation is a crucial and early event (some of these ideas have been discussed elsewhere; ref. 3).

If DJ-1 is a sensor for the oxidative state of a neuron in PD, what functions might it have? In their paper describing the identification of mutations in DJ-1, Bonifati et al. (2) suggested that on oxidation DJ-1 might mediate differential transcriptional and posttranscriptional responses important in maintaining neuronal viability (2). This hypothesis is based in part on previous observations that DJ-1 is part of an mRNA binding complex (6) and interacts with PIASxα (a modulator of the androgen receptor) (15), which might affect sumoylation and indirectly regulate transcription. There are several bacterial and yeast transcription factors that are capable of sensing oxidative or nitrative stress and cysteine to cysteine-sulfinic acid modifications occur (11). An alternate idea is that oxidation could affect protease activity of DJ-1 (9), although, as discussed above, it is far from clear that DJ-1 is an enzyme. This raises the idea that testing ideas of function of DJ-1 should be performed under both basal and oxidative conditions. Furthermore, design of mutants that are stable but lack the capacity to respond to oxidative stress would be extremely useful. If the key role of DJ-1 is to protect against oxidative events, then one would predict that these would be loss of function variants and would cause disease similar to the L166P mutant.

The different crystal structures of DJ-1, in addition to structural and sequence homologues in public databases, go some way toward understanding how mutations in this gene cause disease. Added to our growing knowledge about the functions and effects of mutations in α-synuclein and parkin, this helps us understand the familial forms of PD and perhaps the common sporadic form. In general, it seems likely that recessive mutations are found in genes that normally protect neurons from different forms of damage, and DJ-1 seems to fit into this category. The beguiling idea that DJ-1 plays a role in cellular responses to oxidative stress will require careful dissection in future studies.