Binding of amyloid peptides to domain-swapped dimers of other amyloid-forming proteins may prevent their neurotoxicity

Abstract

We propose that domain-swapped dimers, and possibly other small oligomers of amyloid-forming proteins, could exert chaperone-like action by binding amyloidogenic peptides and thus preventing their toxic effect. This action would be different to those of their monomeric form and facilitated by the new surface exposed after domain-swapping. Our idea is based mainly on studies of human cystatins, which are known to domain-swap in vitro and in vivo. It has been reported that human cystatin C and human stefin B (cystatin B) bind amyloid-beta (Aβ) and inhibit its fibril formation 1, 2. For stefin B, this binding depends on its oligomeric state, with stefin B tetramers binding the strongest and inhibiting Aβ fibril growth 2. To obtain a more general picture, we have studied other cystatins. The domain-swapped dimers of cystatin C, but not the monomer, were found to partially inhibit Aβ fibril formation. However, this did not hold for all cystatins, since stefin A domain-swapped dimers did not have a similar effect. Wilhelmus et al. 3 proposed that certain amyloid-forming proteins, such as gelsolin, apolipoproteins and heparan sulphate proteoglycans, could act as ‘amateur chaperones’. In this way, they would resemble crystallins (class of small heat shock proteins), whose oligomers composed from dimers exert chaperone activity 4. Here, we describe experimental data of the binding of Aβ peptide to domain-swapped oligomers of the three human cystatins: cystatin C, stefin B and stefin A – as opposed to monomers – and derive some speculative ideas on possible general roles of domain-swapped dimers and other small oligomers of amyloidogenic proteins as compared to monomers. Aberrant protein folding and aggregation are in general believed to play a pathological role. Amyloid fibril formation is a hallmark of neurodegenerative diseases from Alzheimer's (AD), Parkinson's and Huntington's disease, and other dementias, to various ataxias and motor neuron disease. A consensus has been reached that soluble oligomers of amyloid-forming proteins cause most damage to neural cells. It is known that the soluble, low Mr oligomers of Aβ are toxic 5, 6 and exert electrophysiological changes in neurons 7, finally acting on cognition and memory 8, 9. The toxicity of the soluble oligomers of amyloid-forming proteins is likely connected to membrane interaction. A vicious circle could occur via membrane permeation, leading to Ca2+ imbalance with other detrimental consequences for the cell 10, 11. Growing evidence indicates, however, that amyloid fibrils may also be functional, contributing to normal physiology 12, 13. Functional amyloids have been found in a wide range of organisms, from bacteria to mammals, with functions as diverse as scaffolding, regulation of melanin synthesis and epigenetic control of polyamines 12. The canonical amyloid β-sheet fold is a conformation that is highly resistant to proteolysis, self-replicating and able to function as a ‘molecular memory’ 12. Another example is c-Jun N-terminal kinase (JNK)-interacting protein-1 (JIP-1), a scaffolding protein, which mediates the intracellular trafficking of Aβ precursor protein (AβPP) by molecular motor kinesin-1 14. The biosynthesis of the pigment melanin occurs on the stack of human Pmel17 amyloid fibrils, acting as a template 13. At present there is no evidence for a possible physiological role of the prefibrillar oligomers. However, we propose to look for conditions under which it would be beneficial for the cell either to attack the membranes of pathogens (acting as pore-forming toxins) or temporarily enhance cellular metabolism by increasing Ca2+ entry, which could result from binding of the oligomers to the plasma membrane. Aβ peptide was found to be a physiological regulator of ion channel expression and hence neuronal excitability 15. Recently, anti-microbial properties of Aβ have been demonstrated, comparable to a known anti-microbial peptide LL-37 16. Of note, stefin B possesses pore-forming properties in vitro 11, 17. Domain-swapping has been proposed as one of the ways for ordered protein aggregation, ending in amyloid fibrils 18, 19. Domain-swapping is the process by which parts of two monomers unfold and then refold back into a dimer, exchanging parts of the chain and forming similar domains as in the monomer. The resulting structure can be a closed ended intertwined dimer with a ‘hinge region’, or higher order oligomer – when the process is propagated 20. On the other side, it has been demonstrated in a number of cases that domain-swapped dimerisation can lead to new function. For example, inducible nitric oxide synthase uses domain-swapping to properly align reductase and oxygenase domains to allow flavin-to-haem electron transfer 21. Domain-swapping has been shown to be important in interactions of scaffolding proteins, such as PDZ2 22 and viral (HIV) capsid-related polyproteins 23. Cystatins are known to domain-swap into dimers in vitro and in vivo 24-27. The stefin B tetramer was shown to consist of two such units 28. The physiological role of domain-swapped dimers and small oligomers of cystatins has not been demonstrated. Smaller oligomers, from dimers up, have also been reported in the case of Aβ peptide 29, β2-microglobulin 30 and α-synuclein 31 among others. It is reasonable to ask whether the presumably domain-swapped dimers of the above-mentioned amyloid-forming proteins have different binding affinities than the monomers. For example, it would be worth considering if they act differently in some signalling cascade or if they enter the nucleus, where the bi-lobed structure of the dimer could bind to either DNA or histones. Indeed, it has been recently reported that stefin B is a histone binding protein 32. Human cystatins 33 are cysteine proteinase inhibitors classified as C1 clan in the MEROPS sequence database 34. Type 1 cystatins (stefins A and B, family I25A) are predominantly intracellular inhibitors, while type 2 cystatins (family I25B, comprising cystatins C, D, E/M, F, G, S, SN and SA) are mostly extracellular, secretory proteins. The two families have had quite a different evolution 35. Cystatins (monomers) are claimed to act as cysteine protease inhibitors, both in vitro and in vivo. They contribute to apoptosis in EPM1 disease and reduce oxidative stress, at least partially by inhibition of cathepsins 36, 37. However, cathepsin-independent functions have also been reported. Di Giaimo et al. 38, using a yeast two-hybrid system, identified five proteins (β-spectrin, NF-L, RACK-1, TCrp and Mtrp) that interact with stefin B in the rat cerebellum, none of which is a protease. This seems to suggest that stefin B (cystatin B) could interact with RACK-1 (receptor for PKC) and thus interfere with PKC signalling. However, the interaction of RACK-1 with stefin B described in the cerebellum was not observed in thymocytes, indicating that the interaction of stefin B with RACK-1 is probably tissue specific 39. It remains to be seen if some of the interactions 32, 39 occur with oligomeric, rather than monomeric, stefin B. Cystatin C was shown to protect neurons from Aβ-induced toxicity 40, either by direct interaction with Aβ 1, 41, 42 or via promoting autophagy 43, the latter by an unknown mechanism. Both cystatins C and B are important for neural repair accompanying epileptic seizures 44, 45. In in vitro studies it was shown that cystatin C and Aβ interact and that amyloid fibril formation by Aβ is inhibited by cystatin C 1. We have shown that stefin B (cystatin B) also interacts with Aβ. We have shown further that binding of Aβ to stefin B is oligomer and conformation specific, which is a new feature, not previously reported for cystatin C. In the case of stefin B, domain-swapped dimers and tetramers preferentially bound Aβ 2. The same entities also inhibited Aβ fibril growth 2 (Fig. 1). Aβ fibrillation is shown in Fig. 1A, as measured by ThT fluorescence when probed alone and in presence of the wild-type stefin B and a more labile variant of Y31 stefin B; this latter forms domain-swapped dimers. It can be seen that the Y31 variant completely inhibited Aβ fibril formation. Figure 1B shows the effect of isolated monomers, dimers, tetramers and other small oligomers of the wild-type stefin B. The most complete inhibition of Aβ fibril growth is achieved by the stefin B wild-type tetramers, known to be composed of two domain-swapped dimers 28. Inhibition of Aβ(1-40) fibril formation by stefin B measured by ThT fluorescence. Aβ peptide concentration was 17 µM throughout, in pH 7.3, 40 °C. A: Aβ alone, 1:1 molar ratio of Aβ to Y31 stefin B (complete inhibition) and 1:1 molar ratio of Aβ to E31 stefin B. B: Aβ alone, 1:1 molar ratio to E31 stefin B monomers, dimers, tetramers and higher oligomers. Protein concentration of stefin B samples was 17 µM. This Figure is adapted from Fig. 2 2, using the same experimental data. Using ESI MS 2, it has been shown that Aβ and stefin B (Y31 variant) bind in a ratio of 1:2, which means that one molecule of Aβ could bind to a stefin B dimer and two molecules of Aβ to a stefin B tetramer 28. With the aim of generalising our findings, we examined amyloid fibril formation by Aβ in the presence of two other human cystatins, stefin A and cystatin C, separately for isolated monomers and domain-swapped dimers (Fig. 2). Cystatin C dimer inhibited Aβ fibril growth by 30% (Fig. 2) but not stefin A dimers. Aβ fibril formation was not inhibited by monomers of either cystatin. Fibril growth of Aβ(1-40) in the presence of monomers and dimers of two other cystatins, stefin A and cystatin C. Growth was measured by ThT fluorescence at 482 nm. A: Aβ fibril growth alone (squares), Aβ with stefin A monomers (circles), Aβ with stefin A dimers (triangles). Representative curves of two independent experiments, each performed in two parallels. B: Aβ alone (squares), Aβ with cystatin C monomers (circles), Aβ with cystatin C dimers (triangles). Each curve represents average of either two (Cys C monomer) or three (Aβ alone, Cys C dimer) independent measurements, each performed in two parallels. Stefin A dimer was prepared by heating a sample of monomeric protein to 85 °C for 20 min 25. Similarly, cystatin C dimer was prepared by heating a sample to 70 °C for 120 min. The dimers were then separated from monomers using Superdex 75. To explain the absence of the inhibitory effect of stefin A on Aβ fibril formation, sequence comparison of the three human cystatins was performed using Clustal W sequence alignment program (Fig. 3). Solvent accessibility was calculated using areaimol as part of the ccp4* program and is presented in the alignment (Fig. 3) by the colour code from blue (accessible) to yellow (not accessible). On the basis of the sequence alignment it is not possible to conclude that stefin A would differ much from stefin B, apart from more His and Phe residues in stefin B. Looking specifically at residues 46–78 (stefin numbering), which span the β-strands 2 and 3 with the corresponding turns, accessibility trend seems to be in accordance: parts of stefin A sequence are the least accessible to solvent. Whole sequences of stefins A and B domain-swapped dimers have been aligned with cystatin C domain-swapped dimer using CLUSTAL W software program. 1N9J, solution structure of the 3D domain swapped dimer of stefin A; 2OCT, stefin B (cystatin B) tetramer (only dimer was taken) and 1TIJ, 3D domain-swapped human cystatin C with amyloid-like intermolecular beta-sheets. A and B are the two chains intertwining in the dimer. PDB structures were used to calculate solvent accessible surface using areaimol, part of the ccp4* program. To mark solvent accessibility Aline software was used where blue represents solvent accessible residues and yellow not accessible. From the alignment it can be seen that there is a trend to more accessibility in stefin B and cystatin C, as opposed to stefin A. In vivo, overexpression of human cystatin C in the brain of AβPP-transgenic mice inhibited Aβ fibril formation and reduced cerebral Aβ deposition 42, 46. Most importantly, co-localisation of cystatin C with Aβ in AD brains 29 and greater amounts of the insoluble cystatin C:Aβ complex in controls than in AD patients 41 have been demonstrated. In cells expressing the Swedish mutant of AβPP 47, stefin B was found to co-precipitate with the C-terminal fragment of AβPP and to co-localise with Aβ inclusions 2, strongly suggesting that their interaction can also take place in cells. It has recently been confirmed that stefin B itself forms aggresome-like structures in the cell 48, 49. Using cystatins as an example, we have shown that Aβ preferentially binds the domain-swapped dimer units of cystatin C and stefin B, but not the monomer. However, our results also show that this is not general to all cystatins, as stefin A domain-swapped dimer does not show such binding. The next step would be to show whether the domain-swapped oligomers of these two cystatins bind some other amyloidogenic peptides, such as amylin, prion and α-synuclein fragment. To generalise even further, one could isolate monomers and dimers of some other globular proteins prone to form amyloid, such as αB-crystallin 50 (suspecting they might be domain-swapped) and try binding of the above-mentioned peptides. The overall conclusion based on human cystatins studies is that domain-swapped dimers and other small oligomers of amyloid-forming proteins can lead to binding of other amyloid peptides, thus exerting a chaperone-like action 3. It may also be that the selective binding properties of domain-swapped dimers of the two human cystatins could serve more specific functions, as has been shown with some other proteins forming domain-swapped dimers 21-23. This work was supported by programs P1-0140 (proteolysis and its regulation, led by B. Turk and to 2009 by V. Turk) and P1-0048 (Structural Biology, led by D. Turk) and a young researchers grant (to A. T. V.), financed by the Ministry for Science and Technology of the Republic of Slovenia via the Slovenian Research Agency (ARRS). We are thankful to Dr. Gregor Gunčar for his help with Fig. 3.