Amyloid fibril formation, structure and domain swapping of acyl‐coenzyme A thioesterase‐7

Acyl‐coenzyme A thioesterase (Acot) enzymes are involved in a broad range of essential intracellular roles including cell signalling, lipid metabolism, inflammation and the opening of ion channels. Dysregulation in lipid metabolism has been linked to neuroinflammatory and neurological disorders such as Alzheimer's and Parkinson's diseases. Structurally, Acot enzymes adopt a circularised trimeric arrangement with each monomer containing an N‐ and a C‐terminal hotdog domain. Acot7 spontaneously forms amyloid fibrils in vitro under physiological conditions. The resultant amyloid fibrillar structures were characterised by dye‐binding fluorescence assays, far‐UV circular dichroism spectroscopy, transmission electron microscopy and X‐ray fibre diffraction. Acot7 has an unusual mechanism of aggregation with no lag phase. The initial phase (~ 18 h) of aggregation involves conformational rearrangement within the oligomers to form species of enhanced β‐sheet character. The subsequent loss of α‐helical structure is accompanied by large‐scale amyloid fibril formation. The crystal structure of Acot7 revealed an unexpected arrangement of the two domains within the circularised trimeric structure, which is the basis for a proposed mechanism of amyloid fibril formation involving domain swapping during the initial phase of aggregation. Acot7 formed fibrils in the presence of its substrate arachidonoyl‐CoA and its inhibitors and maintained its enzyme activity during fibril assembly. It is proposed that the Acot7 fibrillar form acts as functional amyloid.


Introduction
Acyl-CoA thioesterase (Acot) enzymes are involved in a variety of important cellular roles including cell signalling, regulation of metabolic and enzymatic signals and inflammation. They hydrolyse long-chain fatty acids and regulate cellular concentrations of the activated fatty acids, acyl-coenzyme A and coenzyme A (CoA) [1][2][3]. Acot enzymes are ubiquitously expressed in both prokaryotes and eukaryotes. Mammalian Acot family members are broadly categorised into type I and type II classes based on their monomer masses [1,4]. Among the mammalian type II class, Acot7 and Acot12 have been extensively studied and are present in the brain and liver, respectively. Acot7 plays a role in neuronal fatty acid metabolism [5] and has been ascribed a novel inflammatory role through the production of arachidonic acid from its highly specific substrate arachidonoyl-CoA (AA-CoA), suggesting a link with eicosanoid metabolism [6].
Amyloid fibrils are long, highly ordered, b-sheetcontaining protein aggregates associated with a range of disorders such as Alzheimer's and Parkinson's diseases, and type-2 diabetes [7][8][9]. The conversion of the native protein to its amyloid fibrillar form occurs via a variety of intermediate oligomeric species. It is likely that these oligomers, not the fibrils, are cytotoxic [10][11][12]. Extensive fibril deposition, however, causes tissue and organ disruption in systemic amyloidosis [13,14]. In contrast, nonpathological, functional amyloid is also found in nature where it performs a variety of roles and the assembly of the fibrillar structure is associated with the gain of novel function or properties [8,15,16].
In this study, Acot7 was observed to undergo spontaneous amyloid fibril formation in vitro under physiological pH and temperature conditions. The closely related protein, Acot12, did not form amyloid fibrils under these conditions. Mechanistically, Acot7 amyloid fibril formation does not occur via a standard nucleation-dependent mechanism. Aggregation involves the trimeric oligomers acting as a nucleus to facilitate fibril formation. Furthermore, Acot7 formed amyloid fibrils in the presence of the substrate AA-CoA and enzyme inhibitors and was enzymatically active throughout the process of fibril formation. The X-ray crystal structure of Acot7 revealed that the protein adopts a trimeric, circular quaternary structure containing two hotdog domains per monomer, similar to the overall trimeric arrangement in human Acot12. However, the individual hotdog domains are arranged differently in the two proteins due to a domain swap in one of the Acot7 monomers, which provides a basis for the fibril formation of Acot7. It is concluded that Acot7 amyloid fibrils display characteristics of functional amyloid.

Acot7 forms amyloid fibrils under in vitro physiological conditions
Based on our observation that Acot7, but not Acot12, aggregates in solution at neutral pH to form discrete, SDS-resistant oligomers (Fig. 1), the ability of both proteins to form amyloid fibrils in vitro was investigated at pH 7.4 and 37°C. An in situ Thioflavin T (ThT) dye-binding assay revealed that Acot7 exhibited a slow and almost linear increase in fluorescence over 75 h, implying that it formed amyloid fibrils under these conditions ( Fig. 2A). Unusually, there was no lag phase before an increase in ThT fluorescence of Acot7 occurred. Due to the slow rate of Acot7 aggregation, a plateau of ThT fluorescence was not reached. By contrast, no significant increase in ThT fluorescence was observed for Acot12, implying that the protein did not form ordered aggregates ( Fig. 2A). Formation of long fibrillar structures by Acot7 was confirmed by transmission electron microscopy (TEM; Fig. 2B). Acot7 formed fibrils of more than a micrometre in length and width 11-13 nm (Fig. 2B). TEM confirmed that Acot12 did not form aggregates, including of a fibrillar nature, under these conditions (Fig. 2C).
Characterisation of the amyloid fibrillar structure of Acot7 The structural characterisation of Acot7 fibrils was performed using X-ray fibre diffraction and far-UV circular dichroism (CD) spectroscopy. X-ray fibre diffraction of purified Acot7 fibrils yielded a diffraction pattern with a meridional reflection at 4.7 A and an equatorial reflection at 8. 8-11.6 A (Fig. 3A), revealing that the fibril core is organised in a cross bsheet arrangement, confirming an amyloid fibrillar structure [17][18][19]. From previous study [2], Acot7 contains a mixture of a-helix and b-sheet secondary structures. The far-UV CD spectrum of Acot7 upon dissolution (i.e. at zero hours) was consistent with these structural elements ( Fig. 3B) with an ellipticity minimum at~208 nm corresponding to a-helical secondary structure and a broad minimum at~218 nm arising from b-sheet. The CD spectrum of purified Acot7 fibrils exhibited a very broad ellipticity minimum centred around~224 nm (Fig. 3B), implying the presence of significant b-sheet and some a-helix (which has an additional ellipticity minimum at~222 nm).

Amyloid fibril formation of Acot7 does not proceed by a standard nucleation-dependent mechanism
To investigate the molecular mechanism of Acot7 polymerisation, the dependence of the initial rate of aggregation on the concentration of Acot7 was assessed by in situ ThT fluorescence (Fig. 4A). For ThT fluorescence, a plot of ln(initial rate) versus ln ([Acot7]) was linear with a slope of 1.32 AE 0.08 implying apparent first-order kinetics and that the rate of Acot7 fibril assembly was linearly dependent on protein concentration (Fig. 4B) [20]. Aggregation profiles in the presence of the oligomer and fibril-sensitive probe Bis(Triphenylphosphonium) tetraphenylethene (TPE-TPP) [21] revealed a bi-phasic profile ( Fig. 4A) with the first phase occurring over 18 h postincubation of Acot7. This initial phase was not observed in the ThT fluorescence profile and is probably due to the generation of prefibrillar oligomers, which ThT does not recognise and bind to, unlike TPE-TPP [21,22]. The plot of ln(initial rate) versus ln([Acot7]) for the initial phase of the TPE-TPP profile was linear in Fig. 4B with a slope of 2.37 AE 0.18, nearly twice that of the ThT fluorescence reflecting that the dyes monitor different aggregation processes. For all concentrations of TPE-TPP, the second phase of TPE-TPP aggregation up to 100 h had a rate slightly slower than that observed in the ThT aggregation profile (Fig. 4A). Thus, at 250 lM Acot7, the rate of aggregation from 40 to 80 h was 21.0 AE 1.1Áh À1 for TPE-TPP fluorescence and 32.8 AE 1.6Áh À1 for ThT fluorescence. The aggregation profiles for both dyes showed that there was no significant lag phase during Acot7 amyloid fibril formation, i.e. that Acot7 undergoes spontaneous dye-sensitive aggregation. Thus, the use of both dyes enables the observation of the formation of prefibrillar oligomers (via TPE-TPP fluorescence), which is followed by the formation of the fibrillar species (as monitored by both dyes).
Previously, we and others have observed that small heat-shock molecular chaperone proteins (sHsps) such as aB-crystallin (aB-c) interact, in a concentrationdependent manner, with a variety of intermediately folded monomeric, prefibrillar (oligomeric) and amyloid fibrillar species to suppress their aggregation [23][24][25][26][27][28]. When suppressing amorphous aggregation and amyloid fibril formation, sHsps interact with target proteins at an early stage along their aggregation pathway [25,26]. To gain further insight into the mechanism of Acot7 fibril formation, Acot7 was incubated with aB-c at 1 : 1 and 1 : 2 aB-c:Acot7 molar subunit ratios. As revealed by ThT fluorescence, aB-c did not suppress or modulate the assembly process of Acot7 (Fig. 4C). Nucleationdependent amyloid fibril formation involves the formation of stable oligomers or nuclei during the lag phase that then sequester monomers leading to large-scale aggregation. The aggregation of Acot7 occurs over a relatively long time frame (> 100 h, Fig. 4C), conditions, which are favourable for efficient aB-c chaperone action as it prefers to interact with slowly aggregating target proteins [26,29]. Thus, the inability of aB-c to alter Acot7 fibril formation does not arise from aB-c being unable to rearrange itself for optimal chaperone interaction, for example via subunit exchange and dissociation from its oligomeric state. More likely, the ineffective chaperone action of aB-c is due to Acot7 aggregating via a nonstandard mechanism, as is evidenced by the absence of a lag phase and a plateau in its ThT fluorescence aggregation profile (Fig. 4A,C). Consistent with this, our previous studies showed that a-crystallin (a combination of the closely related sHsps, aA-crystallin and aB-c) did not prevent the aggregation of the serpin, a-trypsin, which aggregates via a nucleation-independent polymerisation mechanism involving initial dimer formation due to the reactive centre loop from one monomer inserting into the b-sheet of a nearby monomer [30].
The rate-limiting step in nucleation-dependent protein aggregation is the formation of the nucleus during the lag phase. The introduction of preformed fibrils or seeds (i.e. nuclei) circumvents this requirement for a nucleus and promotes fibril formation leading to an increase in the rate of aggregation and a concomitant reduction in the lag phase [31][32][33][34]. The assembly of Acot7 was measured by ThT fluorescence upon the addition of 1%, 5% and 10% v/v preformed Acot7 fibrils as seeds to a fresh solution of Acot7 (Fig. 4D). Little effect was observed on the rate of Acot7 aggregation in the presence of preformed Acot7 fibril seeds. By contrast, the addition of similar levels of seeds to a solution of A53T a-synuclein (the protein involved in Parkinson's disease, which aggregates via standard nucleation-dependent means) led to a decrease in the lag phase and an exponential increase in the rate of aggregation [20]. In addition, A53T a-synuclein aggregation is readily inhibited by the chaperone action of shaking, incubated in the absence (black solid circles) or presence of 1% v/v (red solid circles), 5% v/v (cyan solid circles) and 10% v/v (blue solid circles) preformed Acot7 fibrils as seeds. The ThT fluorescence due to binding to 1% v/v (red), 5% v/v (cyan) and 10% v/v (blue) preformed Acot7 fibrils was subtracted from the data. aB-c [27], unlike the effect of aB-c on Acot7 aggregation (Fig. 4C). Thus, from the above experiments, it is concluded that Acot7 amyloid formation does not occur via a standard nucleation-dependent mechanism.
Conformational changes during Acot7 amyloid fibril formation The secondary structural transitions in Acot7 during its fibril formation were monitored in real-time by far-UV CD spectroscopy over 56 h (Fig. 5A). Specifically, the change in ellipticity at 208 and 218 nm was measured, i.e. at minima corresponding to a-helical and bsheet conformations, respectively. The increase in Acot7 ellipticity at 208 nm (due to the loss of a-helix) was observed only after~18 h of incubation (Fig. 5B), a time that corresponds to the end of the first phase of the TPE-TPP aggregation profile ( Fig. 4A), i.e. the maximum formation of prefibrillar oligomers and immediately prior to large-scale amyloid fibril formation. The ellipticity at 218 nm steadily and linearly decreased over the first 40-odd hours of incubation, indicative of a gain of b-sheet over this period (Fig. 5B). The absolute value of the rate of gain of Acot7 b-sheet (À127.8 AE 2.2 [h]Áh À1 ) was comparable to the loss of a-helicity (+147.9 AE 2.5 [h]Áh À1 ), i.e. a ratio of 0.9, implying that the two processes are correlated. The CD data imply that Acot7 increases its bsheet content in a linear manner over the first~40 h of the experiment whereas the loss of a-helicity only occurs after~18 h of incubation. Thus, the a-helical regions of Acot7 retain their structure during the initial phase probably associated with oligomer formation, i.e. the Acot7 prefibrillar oligomers contain significant a-helical structure. Consistent with these structural changes, there was little light scattering (turbidity) at 350 nm observed over the first 18 h. After this time, however, light scattering increased in a sigmoidal manner consistent with large-scale aggregation due to amyloid fibril formation (Fig. 5C). Furthermore, the 18-h time-point correlated well with the end of the first phase of TPE-TPP fluorescence (Fig. 4A), which monitored the formation of prefibrillar oligomers of Acot7. The light scattering profile reached a plateau after~40 h (Fig. 5C) implying completion of aggregation, which is consistent with the end of bsheet formation as monitored by CD spectroscopy (Fig. 5B).
The X-ray crystal structure of Acot7 reveals a domain swap compared with the structure of Acot12 To gain detailed insights into the mechanism of fibril formation, we determined the X-ray crystal structure of Acot7. Crystals diffracting to 2. 6 A resolution were formed in 0.1 M ammonium citrate, pH 7.0 and 10% PEG 3350, and diffraction data from a single crystal were indexed and integrated into the space group P12 1 1. The Matthews coefficient predicted that three Acot7 monomers were present in the asymmetric unit, and the structure was solved by molecular replacement in PHASER [35] using the individual hotdog domains of Acot7. Three N-and C-terminal hotdog domains each could be placed reliably due to the quality of the electron density data, the low sequence conservation between the domains (30% amino acid sequence identity) and the presence of bound CoA in each of the three N-terminal domains. The latter observation is consistent with our previous structural study of the individual N-and C-terminal domains of Acot7, which showed that the N-terminal domains all contained the bound cofactor CoA [2,36]. Following rebuilding and refinement in COOT and REFMAC, respectively, the final model had Rwork and Rfree values of 0.231 and 0.268, respectively, no Ramachandran outliers, and good stereochemistry (Table S1). The three monomers in the asymmetric unit exhibited a similar structure, with the greatest root mean square deviation (RMSD) of 0.29 A. The structure of Acot7 revealed the presence of double-hotdog domains within each monomer (Fig. 6A), with each hotdog domain featuring a central sausage-like a-helix (the hotdog) surrounded by a fivestranded antiparallel b-sheet (resembling a bun) (Fig. 6A,B) [37]. The double-hotdog arrangement and quaternary structure (Fig. 6C) are similar to other Acot homologues, including Acot12, for which a crystal structure has been determined [38]. The N-and Cterminal domains within Acot7 are structurally equivalent and can be superimposed (Fig. 6D). The main interactions of CoA in the three N-terminal domains include the sidechains of Leu-108, Ser-136, His-138, Tyr-198, Arg-205 Asn-283, Phe-284, and His-285 (Fig. 6E,F). The positioning of CoA within the Nterminal domains is consistent with a previous report that, of two putative active sites within Acot7, only the Asn24/Asp213 active site is functional [2].
The Acot7 monomers associate into a circular trimeric quaternary arrangement (Fig. 6C) that is consistent with the arrangement of monomers in other Acot homologues, including Acot12 [36]. A predicted structure of Acot7 was produced by homology modelling using PYMOL against the structure of the acyl-CoA hydrolase protein from B. halodurans (PDB: 1VPM), which also exhibits a trimeric structure consisting of six hotdog domains. In comparison to this structure and that of the other Acot proteins, the crystal structure of Acot7 revealed an unexpected domain arrangement. Rather than the conventional Acot assembly as predicted, with hotdog domains positioned in an alternate N-C, N-C, N-C configuration (where N and C refer to the N-and C-terminal domains, respectively; Fig. 7A), Acot7 had a reversed orientation within one of its monomers, i.e. an N-C, N-C, C-N arrangement within the trimer (Fig. 7B). This is distinct from the quaternary arrangement of Acot12 that has been determined experimentally (Fig. 7C) [36]. Thus, one of the monomers in Acot7 underwent a domain swap relative to the two other monomers (Fig. 7B,D). This unexpected arrangement in the Acot7 assembly created three unique interfaces (Fig. 7E). Analysis in Proteins, Interfaces, Structures and Assemblies (PISA) [39] identified a total surface area for the trimer of~38 800 A 2 and a total buried area of 9400 A 2 . The N-N domain interface (interface I) consists of 1100 A 2 of buried surface area, 11 hydrogen bonds and six salt bridges. The N-C domain interface (interface II) is comprised of 980 A 2 of buried surface area, seven hydrogen bonds and six salt bridges while the C-C domain interface (interface III) is comprised of 880 A 2 of buried surface, five hydrogen bonds and one salt bridge. A summary of the interactions between the monomers at each of the interfaces is presented in Table S2. To understand why Acot7, but not other trimeric thioesterases such as Acot12, could be capable of a domain swap, the structural features of the individual domains in the Acot7 structure were examined. Through the superposition of the N-and C-terminal domains of Acot7 (Fig. 6D), it was apparent that the N-and C-terminal domains are capable of substituting for each other. This is consistent with our previous Acot7 structural work [2] where both individual domains of Acot7 formed the same double-hotdog domain monomer structure and higher-order trimeric structures. The ability of the Acot7 domains to substitute for each other, even in the context of the fulllength protein, despite harbouring a low sequence identity, is unique within the Acot family. The extent to which the N-and C-terminal hotdog domains of Acot12 were superimposable was investigated. The Acot12 domains were less superimposable than those in Acot7, and when a domain swap was performed with Acot12 as is present in the Acot7 trimer, a number of steric clashes occurred. Moreover, using PYMOL, the Acot7 N-terminal domain aligned with its Cterminal domain with an RMSD of 0.762 A whereas the Acot12 N-terminal domain aligned with its Cterminal domain with an RMSD of 0.920 A. The implication is that Acot12 would be less likely to undergo the same domain swap as Acot7. A detailed investigation of the ability of the domains in both proteins to be superimposed would require an analysis of the specific structural features of both domains in the Acot7 and Acot12 crystal structures.
The propensity of Acot7 to form amyloid fibrils is dependent on the arrangement of its hotdog domains To gain further insight into the structural aspects of the tendency of Acot7 to form amyloid fibrils, the fibril-forming propensity of the individual N-terminal and C-terminal domains of Acot7 was compared with that of the full-length protein containing both domains. From the time-dependent ThT fluorescence profiles in Fig. 8A, the C-terminal domain of Acot7 underwent fibril formation to a much greater extent (as is apparent from an approximately 10-fold greater magnitude of ThT fluorescence after 72 h of incubation) than full-length Acot7. To a lesser extent, the individual N-terminal domain of Acot7 was also more prone to form amyloid fibrils than the full-length protein (Fig. 8A) since the magnitude of its ThT fluorescence after 72 h of incubation was~2.2 times greater than that of full-length Acot7. TEM images confirmed the formation of fibrils for both individual domains which, in the case of the C-terminal domain, were densely clumped together (Fig. 8B,C). Thus, the association of an individual hotdog domain, particularly the C-terminal domain, with another individual domain of the same type leads to much greater fibril formation than for the full-length Acot7 protein containing both domains. It is concluded that linking the N-and C-terminal domains in full-length Acot7 modulates the amyloid fibril-forming propensity of the individual domains significantly. At a quaternary structural level, placing the domains in an alternate N-C, N-C, N-C domain arrangement in the circularised trimer, as in Acot12, stabilises the protein and negates the possibility of aggregation to form amyloid fibrils under physiological conditions ( Fig. 2A,C). Consistent with this, Acot12 does not undergo the same well-defined oligomerisation with time as Acot7, as monitored by its lack of a laddering profile in SDS/ PAGE (Fig. 1B).

Acot7 enzyme activity is not affected by amyloid fibril formation and the binding of its substrate and enzyme inhibitors does not prevent Acot7 fibril formation
No significant change in the thioesterase activity of Acot7 was observed during the protein's amyloid fibril formation over a time frame of 130 h (Fig. 9A). A 2.4-, 1.2-and 1.4-fold increase in the rate of 150 lM Acot7 fibril formation, as monitored by ThT fluorescence, occurred in the presence of 100, 150 and 300 lM concentrations of the substrate, AA-CoA (Fig. 9B). Furthermore, in the presence of an excess of inhibitors of Acot7 enzyme activity such as glutarate, malonate and tricarballylate, the rate of Acot7 amyloid fibril formation increased 4.9, 6.0 and 4.5-fold, respectively (Fig. 9C). Thus, Acot7 retains its enzyme activity upon amyloid fibril formation, and the presence of the substrate AA-CoA and inhibitors of Acot7 activity increases the propensity of the protein to form amyloid fibrils.
The active site of Acot7 is located at the interface between the two domains, with CoA housed mainly in the N-terminal domain (Fig. 6). The retention of enzymatic activity of Acot7 upon fibril formation implies that the active site, particularly within the N-terminal domain, retains its secondary structure, i.e. it is not part of the cross b-sheet fibril core of the protein. The much greater propensity of the C-terminal domain to form fibrils (Fig. 8) is consistent with this proposal.

Discussion
Crystallisation of the domain-swapped Acot7 trimer in an overall N-C, N-C and C-N arrangement of its hotdog domains (as schematically presented in Fig. 7B) is consistent with this species being the most stable trimeric arrangement of the protein. The ability of Acot7 to form amyloid fibrils in vitro (Fig. 2), in contrast to Acot12, which has a N-C, N-C, N-C arrangement of its domains (Fig. 7C), implies that the formation of the domain-swapped Acot7 trimer is a factor in facilitating the fibril formation of Acot7. The close structural similarity of the N-and C-terminal domains in Acot7, which is not the case in other Acot proteins such as Acot12, may facilitate the structural rearrangement required for domain swapping. The implication, therefore, is that the Acot7 domain-swapped trimer is the oligomeric species that is the basis or building block for amyloid fibril formation of Acot7, i.e. it is analogous to the nucleus in the nucleation-dependent mechanism of protein aggregation [14,[40][41][42]. Another factor in facilitating the fibril formation of Acot7 is the adjacency of two C-terminal domains, and to a lesser extent, the adjacency of two N-terminal domains in the domain-swapped trimer (Fig. 7B,E). As evident in the ThT fluorescence aggregation profiles of the individual domains of Acot7 (Fig. 8), the C-terminal domain of Acot7 is highly amyloidogenic, with the Nterminal domain being slightly more amyloidogenic than the intact Acot7 protein. On this basis, the adjacency of two N-and two C-terminal domains within the domain-swapped Acot7 trimer would promote fibril formation. By contrast, the alternate arrangement of N-and C-terminal domains (as in Acot12) stabilises the overall protein and suppresses the tendency for the individual domains (particularly the C-terminal one) to aggregate. The ability of regions in proteins adjacent to amyloidogenic polypeptide sequences to modulate and minimise the latter's propensity to aggregate has been described in a variety of amyloidogenic peptides and proteins, including aB-c and other mammalian sHsps [43][44][45][46][47]. Figure 10 provides a schematic, putative mechanism of Acot7 amyloid fibril formation. The initial phase of Acot7 aggregation (up to around 18 h postincubation) involves one or more of the six hotdog domains in the trimer undergoing structural rearrangement involving a domain swap. As a result, the amount of b-sheet secondary structure increases without large-scale fibril formation. The process is slow as it probably has high activation energy due to the conformational rearrangement that is required. After around 18 h, enough of the domain-swapped, trimeric oligomer is present to induce large-scale aggregation to form an amyloid fibril. Concomitantly, there is a steady loss of ahelicity and further gain of b-sheet associated with large-scale amyloid fibril formation, as monitored by an exponential increase in light scattering and a plateau after~40 h (Fig. 5).
The absence of a lag phase in the ThT and TPE-TPP aggregation profiles of Acot7 (Fig. 4A) is consistent with the domain-swapped trimer being the prefibrillar oligomeric nucleus, whereby some of this trimer is already present upon dissolution of Acot7. The domain-swapped trimer is aggregation-prone and amyloidogenic. The association of Acot7 monomers is not required for the nucleus to form during a lag phase prior to the exponential growth phase arising from the sequestration of partially folded monomeric proteins, as occurs during standard nucleation-dependent amyloid fibril aggregation. Additionally, no exponential increase in ThT fluorescence upon the addition of Acot7 fibril seeds is consistent with this proposal, since a quantity of the nucleus (i.e. the trimeric oligomer) is already present, and hence, the rate of increase in ThT fluorescence of Acot7 is essentially unaffected by the presence or absence of seeds.
Domain swapping that results in amyloid fibril formation occurs in other proteins. For example, a similar fibril-forming mechanism to that of Acot7, involving the production of amyloidogenic precursors via conformational conversion, has been proposed for the cysteine protease inhibitors, cystatins and stefins Fig. 10. Schematic mechanism of Acot7 amyloid fibril formation. The Acot7 monomer is composed of similarly structured N-and C-terminal hotdog domains. At the quaternary level, Acot7 is arranged as a circular trimer. The most stable form of Acot7 crystallises with one of the monomers having its domains swapped relative to the arrangement in other thioesterases, e.g. Acot12. As a result, two N-terminal and two C-terminal domains are adjacent to each other. In solution, the conversion is slow. The resultant domain-swapped trimer is the amyloidogenic oligomer, which then self-assembles rapidly via the association of adjacent C-terminal domains and adjacent N-terminal domains to form prefibrillar oligomers, which then undergo amyloid fibril formation. The C-terminal domain, in particular, is highly amyloidogenic, so their adjacency in the Acot7 domain-swapped trimer facilitates aggregation. There is no specific structural information about the nature of the Acot7 prefibrillar oligomers so they have not been included in the schematic. The dotted line in the fibril indicates the six domains within the Acot7 trimer.  [42]. For these proteins, amyloid fibril polymerisation incorporating domain swapping is described by a 'gain-of-interaction model' [48] whereby the monomers within the fibril are linked together by domains from an adjacent monomer in an analogous manner to that presented in Fig. 10 for Acot7 amyloid fibrils.
The mechanism of Acot7 amyloid fibril formation has similarities to that of j-casein fibril formation [20,49,50]. Both proteins form oligomers under physiological conditions. For j-casein, dissociation of the highly amyloidogenic monomer from the oligomer is the rate-determining step in amyloid fibril formation. For Acot7, domain swapping within the trimer, i.e. the oligomer, is the trigger for fibril formation. Thus, in a broad sense, the two proteins have a similar, nonstandard mechanism of aggregation; the presence of an oligomer governs the structural rearrangements for amyloid fibril growth. Hence, the ThT fluorescence aggregation profiles for both proteins are similar in that neither exhibits a lag phase nor is their aggregation affected to a significant extent by the presence of fibrillar seeds. The Acot7 mechanism of aggregation has similarities to the Nucleated Conformational Conversion mechanism of fibril formation [51,52] in which a monomeric intermediate (in this case, the domainswapped Acot7 trimer) is a transitional species. It aggregates to form a nucleated oligomer that then associates into amyloid fibrils. So, the addition of Acot7 seeds has little effect on Acot7 aggregation because the predominant alternately domain-arranged trimer is stable and has little proclivity to convert to the domain-swapped, amyloidogenic trimer.
Furthermore, aB-c is not able to inhibit Acot7 fibril formation because Acot7 aggregates via an unusual mechanism. For aB-c to function effectively as a molecular chaperone, it interacts with a dynamic, monomeric, intermediately folded target protein early along its aggregation pathway, for example, one that is undergoing conformational reorganisation as it participates in self-association [23,24,26]. The Acot7 trimer does not satisfy these criteria as it is already present as an oligomeric nucleus. The long time frame over which Acot7 aggregates (days) is favourable to aB-c chaperone activity [26,29], so this is not a factor in the chaperone's ineffectiveness.
Finally, Acot7 is enzymatically active for about a week as it aggregates and forms amyloid fibrils. Also, it forms fibrils in the presence of the substrate AA-CoA and enzyme inhibitors (Fig. 9). As a result, in vivo, Acot7 may be a type of functional amyloid [8,15,16]. The structural and functional significance, intracellularly, of Acot7 adopting an amyloid fibrillar form will be the focus of future investigations.

Expression and purification of Acot7 and Acot12 proteins
The amino acid sequences of the expressed Acot proteins used in this study are provided in Table S3. E. coli BL21 CodonPlus (DE3)-RIL Top 10 competent cells (Stratagene, La Jolla, CA, USA) and BL21 (DE3) pLysS competent cells (Novagen, Darmstadt, Germany) were transfected using the heat-shock method with 1 lL of plasmid DNA in pMCSG21 vector (Genscript, Piscataway, NJ, USA) for 45 s at 42°C. Cells were recovered in 200 lL of Luria broth media (1% w/v tryptone, 0.5% w/v yeast extract, 10 gÁL À1 NaCl) at 37°C for 45 min at 0.51 g. Cells were spread onto agar plates containing spectinomycin (100 lgÁmL À1 ) and incubated at 37°C. Following inoculation using a single isolated colony, cultures were produced using the transformed BL21(DE3) pLysS cells by adding 500 lL of cells to 500 mL of autoinduction expression base media (1% w/v tryptone, 0.5% w/v yeast extract, 1 mM MgSO 4 , 5% 209 NPS buffer (50 mM Na 2 HPO 4 , 50 mM KH 2 PO 4 , 25 mM (NH 4 ) 2 SO 4 ), 2% 509 5052 solution (0.5% glycerol, 0.05% glucose, 0.2% lactose) containing spectinomycin (100 lgÁmL À1 )). Cells were incubated at 37°C at 0.067 g for~28 h. Cells were harvested by centrifugation at 5403 g at 18°C for 30 min. Cell lysates were then centrifuged at 17 640 g at 18°C for 30 min, and the supernatant was filtered through a 0.45 lM syringe filter. The soluble cell extracts were injected using a superloop at 2 mLÁmin À1 into a nickel NTA column equilibrated with HIS buffer (50 mM sodium phosphate buffer, pH 8.0, 300 mM NaCl, 20 mM imidazole). Purified protein was eluted from the column using HIS buffer containing 500 mM imidazole. The protein was incubated with TEV protease to cleave the HIS tag at the N-terminal TEV cleavage site. Size exclusion chromatography was performed on an AKTA FPLC Superdex 200 20/60 filtration column. The purity of the protein was assessed by SDS/ PAGE using a 4-12% Bis-Tris gel.
Expression and purification of aB-crystallin aB-c, in the vector pET24d(+), was expressed in the BL21 (DE3) strain of Escherichia coli. The cell pellet was lysed, and supernatant was loaded on to an anion exchange column (HiTrap 16/10 DEAE Sepharose, GE Healthcare). Acot7 crystallisation and X-ray crystal structure determination Concentrated protein samples (40 mgÁmL À1 ) were screened for crystallisation conditions using the hanging drop method in 48-well plates (Hampton Research, Aliso Viejo, CA, USA) and commercially available screens. Briefly, 1.5 lL of protein was added to 1.5 lL of crystallisation solution on a glass slide above a reservoir containing 300 lL of crystallisation solution. Acot7 was incubated in the presence of CoA in a 37°C incubator. X-ray diffraction data were collected on the MX1 (ADSC Quantum 210r and ADSC Quantum 315r detector) and MX2 (Eiger 16M detector) beamlines at the Australian Synchrotron Facility, Melbourne [55,56]. The datasets were indexed and refined using IMOSFLM [57]. The data were scaled and refined using AIMLESS [58,59], PHASER MR [35], REFMAC [60] and PHENIX [61] software. COOT [62] software was used for structural modelling. The coordinates and associated structural data were deposited and validated to the Protein Data Bank (PDB) and issued the identification code 6VFY.
Monitoring amyloid fibril formation of Acot7

Solution turbidity measurements
The change in light scattering at 350 nm with time was used to monitor the aggregation of Acot7 at various concentrations in 20 mM phosphate buffer at pH 7.4, 37°C.
The assay was conducted in Greiner bio-one, half-area, clear-bottom, medium-binding 96-well plates, sealed with transparent film. Light scattering was monitored using a Biotek Synergy 2 microplate reader.

Far-UV circular dichroism spectroscopy
The far-UV CD spectra of 10 lM Acot7 were collected over a wavelength range of 180-260 nm, with a 0.5 nm stepincrement, a bandwidth of 1.5 nm and a scan rate of 4 s per increment at 37°C. Each spectrum was obtained with an Applied Photophysics Chirascan spectropolarimeter using a QS high-precision cell (Hellma Analytics, Muellheim, Germany) with a 0.01 cm path length. All samples were prepared in 10 mM phosphate buffer, pH 7.4. Each spectrum was an average of five accumulations, with the blank subtracted. All high-voltage tension records above 600 V were discarded.

Real-time far-UV circular dichroism spectroscopy
The far-UV CD spectrum of 10 lM Acot7 in 10 mM phosphate buffer, pH 7.4, 37°C was collected over a period of 65 h in a 0.01 cm path length cuvette sealed with sealing tape. The approximate scan time of each spectrum was 11 min, which was collected over a wavelength range of 180-260 nm with a 0.5 nm step-increment, a bandwidth of 1.5 nm and a scan rate of 4 s per increment. Amyloid seeding experiments 150 lM of fresh Acot7 in 20 mM phosphate buffer, pH 7.4, was incubated at 37°C with agitation either without (unseeded) or with (seeded, 1% v/v, 5% v/v and 10% v/v) purified Acot7 amyloid fibrils. Acot7 amyloid fibrils (seeds) were formed by incubating 500 lL of 150 lM protein in the above-mentioned amyloid fibril-forming conditions for 160 h. The fibrils were purified by centrifugation at 20 251 g for 1 h. The pellet was washed three times with 20 mM phosphate buffer, pH 7.4 and finally resuspended in 250 lL of the buffer. The in situ aggregation was monitored via ThT fluorescence as described above.
X-ray fibre diffraction Fibrillar Acot7 was formed by incubating 200 lM protein under the above-mentioned amyloid fibril-forming conditions for 160 h. The fibrils were purified by centrifugation at 14 100 9 g for 1 h and then resuspended in distilled water immediately before the formation of the fibril stalk. An Acot7 fibril stalk was prepared via the stretch frame method according to Serpell et al. [63]. Diffraction images were collected from the fibril samples using a Cu Ka Rigaku rotating anode source (wavelength 1.5418 A) and mar345 image plate detector (MarResearch GmbH, Norderstedt, Germany). Images were examined and reflections measured using MARVIEW (MarResearch).

Thioesterase activity assay
The reaction mixture to measure thioesterase activity of Acot7 contained 6.6 lL AA-CoA (from 0.33 mM stock), 190.4 lL 100 mM phosphate buffer, pH 7.4 and 1 lL of 200 lM Acot7 in a final volume of 200 lL. The enzyme activity of fresh and incubated Acot7 under amyloid fibrilforming conditions was monitored by measuring the absorbance at 412 nm over 30 min, immediately after the addition of 0.1 mM 5,5 0 -Dithiobis (2-nitrobenzoic acid; DTNB). The molar absorption coefficient, e 412 (13 600 M À1 Ácm À1 ), was used to calculate the cleavage of the thioester bond. Units of Acot7 enzyme activity (U) are expressed as moles of AA-CoA hydrolysedÁmin À1 at 37°C [3].

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Data collection and refinement statistics for Acot7. Table S2. A summary of the interactions between the monomers in each of the three interfaces in the Acot7 crystal structure. Table S3. Amino acid sequences of mouse Acot7, the N-and C-terminal domains of Acot7, and Acot12 proteins.