Molecular determinants for amyloid fibril formation : lessons from lung surfactant protein C

The term amyloid emanates from early histological examinations where tissue deposits associated with certain diseases were thought to be composed of carbohydrates, but amyloid is now known to be made up of proteins which have lost their native, soluble, structure and turned into insoluble bsheet polymers [1]. Amyloid fibrils are defined by their ability to stain with the dye Congo red and then show birefringence under polarised light, by the appearance of 5–10 nm unbranched fibrillar structures detectable by electron microscopy, and by typical X-ray diffraction pattern. Amyloid fibrils are likely to be built from a polypeptide chain that has adopted a cross-b-sheet structure, ie, the individual b-strands are oriented perpendicular to the fibril direction [2, 3]. The approximately 20 known amyloid diseases represent a perplexing medical problem, including such diverse entities as Alzheimer’s disease (fibrils from the amyloid bpeptide, Ab), the spongiform encephalopathies or prion diseases, including Creutzfeldt-Jakob disease in humans, scrapie in sheep and bovine spongiform encephalopathy (mad-cow disease) (fibrils from the prion protein, PrP), familial polyneuropathy (fibrils from transthyretin), type 2 diabetes mellitus (fibrils from islet amyloid polypeptide or amylin), systemic amyloidosis (fibrils from lysozyme or transthyretin), Parkinson’s disease (fibrils from a-synuclein) and Huntington’s disease (fibrils from huntingtin) [4–6]. Amyloid diseases occur in sporadic and hereditary forms, whilst the prion diseases are also infectious. Notably, the formation of amyloid fibrils is not necessarily the result of abnormal processes, but can be used for specific functions. They form a protective shell in the silk moth [7], and spiders’ web fibres have been proposed to have amyloid-like properties [8]. The proteins implicated in disease are invariably present in the corresponding amyloid deposits, but also other proteins and additional constituents can be Amyloid fibrils are polymers composed of proteins in b-sheet conformation, which are found in at least 20 diseases for which no cure is available. These diseases include Alzheimer’s disease and spongiform encephalopathies, in which the amyloid b-peptide (Ab) and the prion protein (PrP), respectively, form amyloid. All fibrils are morphologically very similar although the native structures of the corresponding proteins are widely different. Proteins that are not known to form fibrils in vivo can do so under conditions where unfolding intermediates are well populated. This indicates that fibril formation can arise from most, if not all, polypeptide chains under certain conditions, and that nature has found ways to avoid fibrillogenic protein conformations. In support of this, it was recently found that unpaired b-strands present in native proteins are prevented from forming intermolecular b-sheets, by strategic placement of prolines and charged residues for example. Structural studies of the lung surfactant-associated protein C (SP-C) have revealed determinants for amyloid fibril formation. The poly-valine a-helix of SP-C spontaneously converts to b-sheet aggregates in vitro and SP-C amyloid fibrils are found in pulmonary alveolar proteinosis. Ab, PrP, and SP-C harbour an a-helix which is strongly predicted to form a b-strand, and in all cases investigated so far such a-helix/b-sheet discordance correlates with the ability to form b-sheet aggregates and fibrils.

found.For example, the amyloid plaques found in Alzheimer's disease have been found to contain, in addition to Ab, proteoglycans, a 1 -antichymotrypsin and apolipoprotein E [9].Moreover, the protein serum amyloid P component (SAP) binds to amyloid deposits irrespective of what protein is the primary constituent.SAP-amyloid binding can be reduced by low molecular weight SAP ligands, a property which has been explored as a means to facilitate amyloid clearance [10].For some fibrilforming proteins, eg, Ab and PrP, both wild-type and mutant proteins form amyloid, while for others, e.g., lysozyme, destabilising point mutations are required for fibril formation [11].Point mutations related to inherited forms of human prion diseases do not cause formation of the amyloid form of PrP in vitro and are not as a rule destabilising [12].
All fibrils appear to be made from very similar cross-b-sheet structure [2], but the native states of fibril-forming proteins vary considerably.For example, PrP contains a structurally disordered part, three a-helices and a short b-sheet [13].Ab is mainly flexible, but under membrane-mimicking conditions its central (residues [15][16][17][18][19][20][21][22][23][24][25] and C-terminal parts (residues 30-40) form helices, see [2].The conversion of the soluble form of PrP to its fibrillar counterpart is accompanied by reduction in a-helix content and increase in b-sheet structure, and Ab fibril formation also involves a-helix to b-strand conversion [14,15].This indicates that proteins with a-helical structure in their native states must undergo a-helix3b-sheet conversion before or during fibrillation.Fibril formation, however, does not necessarily involve a-helix3bsheet conversion; transthyretin is a tetramer where each subunit is composed of all-b-sheet structure, and a-synuclein is structurally disordered in its native state.Amyloid fibril formation is promoted under conditions where conformations present during unfolding are well populated [16][17][18].A current understanding of the events leading to transformation of a native protein into a fibril with b-sheet structure is presented in figure 1. Salient features of this process are (i) local unfolding of the native protein, or dissociation of a multimeric protein, leading to exposure of regions which are prone to self-associate, (ii) intermolecular association, (iii) further structural rearrangement leading to increased b-sheet content, and (iv) acquisition of a cross-b-sheet structure and polymerisation into fibrils.
The ability to form amyloid fibrils is not limited to the proteins that form amyloid in human and mammalian disease.Fibrils similar to those found in disease can be formed in vitro from virtually any protein under partly denaturing conditions [19,20].Although any peptide sequence may be able to form a cross-b-sheet structure under some conditions, the ability to do so under physiological conditions is apparently limited to a few proteins.This raises the question: what distinguishes the small number of proteins that form amyloid under physiological conditions?In vitro fibril formation of a globular protein and of short peptides can be modulated by changes in the amino acid sequence [21][22][23].For natural amyloidogenic proteins, there is data indicating that specific regions of their amino acid sequences determine fibrillation.Ab is highly fibrillogenic, while peptides lacking residues 14-23 are not [24].In b 2-mi- croglobulin, which forms fibrils in haemodialysisrelated amyloidosis, five of seven b-strands are preserved in a partly unfolded fibrillogenic interme-diate, and largely the same region forms a rigid bsheet in b 2-microglobulin fibrils [25,26].Interestingly, when represented as synthetic peptides in isolation, only one of these strands forms fibrils in vitro [27].Unexpectedly, studies of lung surfactant protein C (SP-C) have lead to the identification of structural features, which appear to play a role for the fibrillogenic behaviour of fibril-forming proteins.

Lung surfactant protein C
SP-C is one of four surfactant-associated proteins which are involved in preserving alveolar patency and in lung host-defence [28,29].SP-C is evolutionarily highly conserved and the peptide is found only in lung surfactant [30].SP-C is mainly implicated in lung surface properties [31], and the characterisation of SP-C has been fuelled by the fact that it is found in preparations used for therapy of respiratory failure [32].However, SP-C binds lipopolysacharides, suggesting that it may also be involved in lung host-defence [33][34][35].SP-C is extremely hydrophobic owing to the presence of two fatty acyl chains and almost 30 residues (out of 35) with aliphatic side-chains (figure 2).The three-dimensional structure of SP-C encompasses an a-helix covering positions 9-34, which is perfectly adopted to span a lipid bilayer [36,37] (figure 2).
The helical structure of SP-C exhibits longterm stability when inserted into lipid micelles or membranes, but is metastable in aqueous organic solvents, i.e., once it opens up into a non-helical structure, it does not refold again but forms bsheet aggregates [38].The short-term stability of the SP-C poly-valine a-helix stems from tight interactions between the valine side-chains, giving high activation energy for SP-C helix unfolding [39,40].The structural conversion of SP-C from a-helix to b-sheet aggregates results in formation of amyloid fibrils, and occurs in association with pulmonary alveolar proteinosis [41].Whether these fibrils are formed as a result of altered intraalveolar metabolism in pulmonary alveolar proteinosis, or are involved in the pathogenesis of the disease remains to be investigated.The metastability of helical SP-C and the problems in folding SP-C into a helix may be involved in other human diseases.Familial forms of human interstitial lung disease are associated with point mutations in the proSP-C gene [42,43], and formation of aggresomes occurs in cells transfected with mutated proSP-C [44].The near-complete lack of mature SP-C observed in the disease cases is likely a dominant negative effect caused by interactions between mutant and wild-type proSP-C [45,46].Transgenic expression of the mature SP-C peptide leads to marked lung underdevelopment [47], while genetic ablation of SP-C gives variable effects on respiratory function [29,48].It is thus possible that interstitial lung disease is caused to a large extent by cellular injury due to SP-C aggregation.It will be interesting to find out whether bsheet formation and fibril formation of SP-C are associated with the familial forms of interstitial lung disease.Structure of SP-C.The cylinder represents the a-helix of SP-C, covering residues 9-34 in the structure determined by NMR, and which corresponds to the thickness of a fluid dipalmitoyl-phosphatidylcholinebilayer.The plus sign in the N-terminal part of the helix corresponds to the evolutionarily conserved Lys11-Arg12 dibasic pair.The two palmitoyl groups linked to Cys5 and Cys6 are depicted by wavy lines.The amino acid sequence of human SP-C is given using one-letter abbreviations for the amino acid residues.The discordant region is underlined.
The experimentally determined SP-C a-helix, composed essentially of a long stretch of valine residues, is unusual since valines are well-known to be over-represented in b-strands and under-represented in helices [36,49].Intriguingly, synthetic peptides with the SP-C amino acid sequence are inefficient in helix formation, and form insoluble aggregates [49].In contrast, peptides in which the poly-valine sequence of SP-C is replaced with that of a helical part of bacteriorhodopsin or a polyleucine stretch, readily form helices and do not aggregate [49,50].These sequences have high helical propensities.When three-dimensional structures in the protein data bank were compared with their predicted secondary structures, about three percent of the proteins turned out to contain a ≥ 7-residue a-helix strongly predicted to form a bstrand.These a/b discordances include PrP (helix 2, positions 179-191), Ab (positions 16-23), and SP-C (positions 12-27, figure 2).Coagulation factor XIII, a triacylglycerol lipase, and a transpeptidase likewise contain a long discordant helix.None of these proteins were known to be fibrillogenic, but were found to form fibrils in physiological-like solutions, suggesting a correlation between ahelix/b-strand discordant stretches and amyloid fibril formation [51].
Regarding Ab, two lines of evidence indicate that the discordant region is crucial for fibril formation; positions 16-20 are essential for formation of Ab intermolecular contacts and fibril formation [52], and destabilisation of the helix covering residues 17-24 is critical for a-helix3b-sheet conversion and fibril formation [53].Experimental support for the possibility that discordant helices represent underlying structural ambivalence can be found also in a study of isolated synthetic peptides corresponding to transmembrane helices of bacteriorhodopsin [54].This showed that of the seven transmembrane regions, only the discordant one (the most C-terminal of the helices) forms a hyperstable b-sheet aggregate, while five of the other regions form stable transmembrane helices and one region does not form any stable secondary structure.The authors pointed out that C-terminally located helices may have a lower intrinsic stability, and their efficient folding requires rapid folding of the segments which are synthesised first, so that they can serve as folding templates for the distal segments [54].It is possible that discordant helices require the presence of scaffolding, e.g., a folded protein nucleus or a lipid membrane, for proper folding in order to compensate for their low intrinsic potential for helix formation.
In PrP helix 2 is discordant, while helices 1 and 3 are correctly predicted to be helical [51].In line with this, an early prediction of PrP secondary structure gave contradictory results for the helix 2 region, suggesting that alternative conformations of PrP can coexist [55].It was recently found that PrP helix 2, in particular its C-terminal part, is "frustrated" in its helical state and this region may play a role in a3b transition [56].Moreover, in a crystal structure of dimeric PrP, the last turn of the discordant helix 2 is transformed to a b-strand, while helices 1 and 3 remain intact.It was suggested that formation of the dimeric form could represent a first step towards fibril formation of PrP [57].This shows experimentally that parts of the PrP discordant helix can change the structure into b-strand, as suggested from the secondary structure predictions, and lends support to the concept that discordant helices may represent starting sites for a-helix3b-sheet conversions.The Cys residue in PrP helix 2 makes a disulphide bond with helix 3.In the domain-swapped dimer, this disulphide bond has spontaneously rearranged from an intramolecular to an intermolecular interaction [57].This suggests that the disulphide connecting helices 2 and 3 can be opened in vivo.Interestingly, a reduced form of a recombinant PrP fragment (positions 91-231), but not the oxidised form, can form a soluble monomer composed almost entirely of b-sheet [58].This b-sheet-rich monomer was prone to aggregate into fibrils.
Modulation of the sequences of a/b discordant segments, so that their helical propensities are increased and b-strand propensities are decreased reduces fibril-formation.The mutations K16A/ L17A/F20A and V18A abrogate Ab fibril formation in vitro [52,59].These changes reverse the discordant nature of the 16-23 region of Ab.Likewise, by replacing valines in SP-C with leucines (SP-C(Leu)), an a-helix is found experimentally and by predictions [50,51].SP-C(Leu) does not form b-sheet aggregates or fibrils under conditions where SP-C forms abundant fibrils.Studies of SP-C and SP-C(Leu) using hydrogen/deuterium exchange and mass spectrometry indicate that unfolded SP-C rapidly converts into aggregates, while for unfolded SP-C(Leu) the lowest energy barrier is the one going back to helical conformation [60] (figure 3).Hydrogen/deuterium exchange in combination with NMR indicates that the activation energy for unfolding a-helical SP-C is ~100 kJ/mol [38].The a-helical conformation of PrP is likewise kinetically stabilised relative to a b-rich isoform, with an energetic barrier of ~80 kJ/mol [61].SP-C and PrP thus retains an a-helical fold not because of thermodynamic stability but because of a high energetic barrier to unfolding.
Studies of the two-helix protein acylphosphatase show that regions which determine folding and aggregation can be localised to different regions of the polypeptide chain [62,63].Mutations in mainly one of the a-helices influence the aggregation properties, while mutations in the other helix affect the folding properties [63].Analysis of SP-C and SP-C(Leu) gives some clues a-Helix/b-strand discordant helices and fibril formation as to what may identify an aggregation-determining region and a folding-determining region, respectively.There is no major difference between SP-C and SP-C(Leu) in size, polarity or structure of the native state (Figs. 2 and 3), but their secondary structure preferences differ significantly.The a-helix/b-strand propensity ratio for Val is 0.46, and for Leu 1.2 [60].Thus there is an excellent agreement between the secondary structure propensities and the fates of the unfolded peptides; SP-C forms b-sheet aggregates and SP-C(Leu) reforms an a-helix (figure 3).This supports the observations that partitioning between aggregation and folding can be attributed to conformational preferences of the polypeptide chain [62,64].However, studies of peptides derived from bstrands of b 2 -microglobulin gave no correlation between secondary structure propensities and fibril formation [27].One caveat is that the secondary structure propensities used in the predictions described herein are mainly derived from watersoluble proteins.Using secondary structure propensities determined for peptides in non-polar environments [65] a helical structure of SP-C is predicted, suggesting that its discordant nature could reflect that the proper non-polar membrane environment is not taken into account.However, also in non-polar environments SP-C forms bsheet aggregates [49,60,66], while SP-C(Leu) forms a helix in all solvents used [50].This shows that folding of the SP-C poly-valine segment into a helical conformation is not accomplished by placing the mature peptide in a hydrophobic environment; at least proSP-C is probably necessary for correct folding (see above).
Fibrillogenic proteins may under physiological conditions form partially unfolded intermedi-ates that are prone to oligomerise (figure 1).It appears reasonable to assume that if a discordant helix unfolds, the high b-strand propensity will make it less able to refold into a helical conformation compared to a region for which a helix is predicted, analogous to the situation observed for SP-C vs SP-C(Leu) [60] (figure 3).Unfolded discordant helices have high intrinsic preferences to adopt b-strand conformation and could then form b-sheets via protein oligomerisation.In the pathway of figure 1, unfolding of a discordant helix, relative to unfolding of a "normal" helix, would favour reaction 5 over reaction 2, thereby diverting a partly unfolded intermediate to fibril formation.This suggestion is in agreement with recent data.Analysis of b-sheet edges in a large sample of b-sheet proteins showed that all free edge strands were protected from making intermolecular bsheet interactions [67].The mechanisms involved include strategically placed prolines and charged residues, very short edge strands and loop coverage.In contrast, edge strands that natively form bsheet dimers or rings lack such protections.Naturally occurring monomeric b-sheet proteins have thus evolved to avoid edge-to-edge aggregation [67].Introduction of charged residues in do novo designed polypeptides was found to indeed prevent formation of amyloid-like fibrils and instead promote formation of monomeric b-sheets [68].Prevention of amyloid fibril formation thus includes avoidance of that unpaired b-strands that can form intermolecular contacts.Protection against intermolecular b-sheet formation has likely not evolved for regions that are natively helical, and transformation of a discordant helix into an unpaired b-strand may thus promote fibril formation.Folding and aggregation of SP-C and SP-C(Leu).Schematic representation of the different behaviour of SP-C compared to SP-C(Leu) regarding folding and aggregation, as derived from hydrogen/deuterium exchange in combination with NMR and mass spectrometry.Both peptides are helical in their native states (N).SP-C(Leu) (green) refolds into a helix after unfolding, while SP-C (red) forms b-sheet aggregates and amyloid fibrils after unfolding, with no detectable refolding of helical structure.The amino acid sequence of SP-C(Leu) is shown, see figure 2 for the amino acid sequence of SP-C.
The pathophysiological mechanisms in amyloid diseases may, in principle, include loss of functional protein due to deposition as insoluble amyloid, but most evidence indicates that the major underlying mechanism is rather destructive effects elicited by the deposits [69,70].Regarding Ab, the levels of both Ab  and Ab (1-42/43) in the cerebral cortex are significantly higher in Alzheimer's disease than in normal ageing [71].Notably, in Alzheimer's disease brains a marked shift in Ab from the predominantly soluble form in normal ageing to very large insoluble pools was seen.Näslund and co-workers found that the cortex levels of Ab ending at positions 40 and 42 were elevated early in dementia, and levels of both peptides strongly correlated with cognitive decline [72].These studies indicate an important role for Ab in mediating initial pathogenic events in Alzheimer's disease.However, no correlation is found between the amounts of Ab amyloid deposits and the degree of dementia, see eg, [73].The reason for this is not known but may be related to the question whether intermediates in the pathway leading to fibril formation are the dominant toxic species [74]?Evidence is accumulating that prefibrillar soluble aggregates, including species referred to as protofibrils, may be more toxic than the mature fibrils [75][76][77][78][79][80].Oligomeric, prefibrillar forms may form membrane-penetrating pores, similar in structure to pores made from bacterial toxins, and thereby cause cell lysis [81].It has been proposed that incorporation of the prefibrillar species into mature fibrils is a way of protecting against their toxic effects, although it remains possible that both oligomers and mature fibrils are destructive.
Obstruction of amyloid and/or oligomer/ protofibril formation may prevent the occurrence or progression of amyloidoses, and several ways have been explored to accomplish this [82].Compounds which can abrogate fibril formation by interfering with peptide-peptide contacts in fibrils, which are expected to interfere with reactions 5 through 8 in figure 1, have been identified [52,83,84].A potential drawback with such compounds is that they not only reduce fibril formation, but may also increase the amounts of oligomers/protofibrils, see figure 1.Another way to prevent fibril formation is to stabilise the native structure.This ap-proach is effective in preventing fibril formation of acylphosphatase under partly denaturing conditions [85].The possibility that a/b discordant helices are involved in the fibrillation process for some proteins suggest that stabilisation of these segments in helical conformation could prevent amyloid formation.Such an approach is analogous to the stabilisation of the native, non-amyloidogenic transthyretin tetramer, which prevents the dissociation into subunits which can form fibrils [86].Stability of the SP-C helix is inversely correlated with its potency for forming fibrils; removal of one or both of the palmitoyl groups of SP-C in vitro results in faster unfolding of the a-helical conformation, as well as faster and more prominent formation of insoluble aggregates [41,66].Moreover, fibrillar SP-C associated with pulmonary alveolar proteinosis is partly non-palmitoylated [41], while in healthy individuals SP-C is almost exclusively di-palmitoylated [87].
For Ab and SP-C, removal of the discordant nature by residue replacements abrogates fibril formation in vitro, see above and figure 3, further supporting that increasing the helical occupancy can reduce the fibril formation.This hypothesis is supported by the finding that optimal stabilisation of helical Ab by addition of trifluoroethanol completely prevents fibril formation [88].However, partial stabilisation of helical structure of Ab instead accelerates formation of b-sheet aggregates and fibrils [88].Furthermore, transient formation of a-helix structure prior to formation of b-sheet structure and fibrils has been observed, suggesting that partially helical forms of Ab may be on-pathway to fibril formation [89].Further studies are needed in order to understand the steps leading to fibril formation, so that this most devastating result of protein misfolding can be controlled.

Figure 1
Figure 1 Pathway for formation of amyloid fibrils.Schematic representation of the current understanding of molecular mechanisms involved in fibril formation.N represents the native structure, U a completely unfolded state, and I represents an intermediate in which parts of the polypeptide chain are in non-native conformation.Formation of intermolecular b-sheet mediated by unfolded regions in I leads to protein oligomerisation (O), which in turn leads to further formation of b-strand structure (B), formation of prefibrillar oligomers/protofibrils (P) and eventually to fibril formation (F).The scale bar in the electron micrograph represents 50 nm.

Figure 3
Figure 3 : jan.johansson@vmk.slu.seSwiss Medical Weekly: Call for papers Swiss Medical Weekly The many reasons why you should choose SMW to publish your research Official journal of the Swiss Society of Infectious disease the Swiss Society of Internal Medicine the Swiss Respiratory Society Med Wochenschr (1871-2000) Swiss Med Wkly (continues Schweiz Med Wochenschr from 2001) Editores Medicorum Helveticorum