Apolipoprotein C-II Adopts Distinct Structures in Complex with Micellar and Submicellar Forms of the Amyloid-Inhibiting Lipid-Mimetic Dodecylphosphocholine

Abstract

The formation of amyloid deposits is a common feature of a broad range of diseases, including atherosclerosis, Alzheimer’s disease, and Parkinson’s disease. The basis and role of amyloid deposition in the pathogenesis of these diseases is still being defined, however an interesting feature of amyloidogenic proteins is that the majority of the pathologically associated proteins are involved in lipid homeostasis, be it in lipid transport, incorporation into membranes, or the regulation of lipid pathways. Thus, amyloid-forming proteins commonly bind lipids, and lipids are generally involved in the proper folding of these proteins. However, understanding of the basis for these lipid-related aspects of amyloidogenesis is lacking. Thus, we have used the apolipoprotein C-II amyloid model system in conjunction with x-ray and neutron scattering analyses to address this problem. Apolipoprotein C-II is a well-studied model system of systemic amyloid fibril formation, with a clear and well-defined pathway for fibril formation, where the effects of lipid interaction are characterized, particularly for the lipid mimetic dodecylphosphocholine. We show that the micellar state of an inhibitory lipid can have a very significant effect on protein conformation, with micelles stabilizing a particular α-helical structure, whereas submicellar lipids stabilize a very different dimeric, α-helical structure. These results indicate that lipids may have an important role in the development and progression of amyloid-related diseases.

Introduction

Amyloid formation and deposition are characteristic of several diseases including Alzheimer’s disease, Parkinson’s disease, atherosclerosis, and several forms of systemic amyloidosis, where amyloid deposits are formed in the circulatory system and liver (for a review see (1, 2)). The proteins that form these deposits in their respective diseases typically do not have much similarity in terms of sequence, secondary, tertiary, or quaternary structure. However, a large number of amyloidogenic proteins have roles in lipid transport, membrane deformation, and lipid metabolism suggesting a common functional basis that may be related to a propensity for these proteins to form amyloid (for a review see (3)). One example of a systemic amyloidosis-related disease is atherosclerosis, where a number of apolipoproteins have been found to form amyloid deposits in atherosclerotic plaques (2, 4, 5, 6). These amyloid deposits may play a role in propagating the inflammatory response in the disease and the development of atherosclerotic lesions (7, 8).

Apolipoprotein C-II (apoC-II) is a 79-amino-acid protein cofactor of lipoprotein lipase that circulates in the bloodstream bound to very low-density lipoproteins and chylomicrons. In a lipid-bound state it adopts a primarily α-helical structure (∼80% α-helix by circular dichroism spectroscopy) (9, 10, 11), however in the absence of lipids apoC-II is unstable and aggregates to form homogeneous β-sheet amyloid fibrils with all of the hallmarks of pathologic amyloid fibrils (12). The homogeneity, relative solubility, and the well-characterized kinetics of formation of these fibrils provide an ideal system for the study of factors that are involved in the formation of amyloid fibrils, particularly those formed by apolipoproteins. In addition to the advantages of being an excellent system for biophysical characterization of ordered aggregation, apoC-II is one of several apolipoproteins that accumulate in atherosclerotic plaques. In these plaques it is localized with serum amyloid P, a marker of in vivo amyloid deposits, where it may be involved in initiating early events in heart disease, including the induction of the macrophage inflammatory response (13).

Phospholipids significantly affect apoC-II fibril formation. High concentrations of micellar lipids cause apoC-II to adopt a native-like conformation, which does not aggregate (9, 10, 14, 15, 16). This constitutes the most general form of amyloid inhibition by these biological molecules, where on average one apoC-II molecule associates with one lipid micelle. NMR analysis of apoC-II in complex with micelles shows a fairly flexible pair of α-helices in a horseshoe-like arrangement (9, 10). In contrast, at high protein/lipid ratios, where multiple apoC-II molecules are expected to be associated with individual micelles, the presence of lipid surfaces is not sufficient to prevent aggregation, and alternate amyloid morphologies for apoC-II fibrils can be produced (15).

In comparison, at concentrations below their critical micelle concentration (CMC), phospholipids accelerate fibril formation (14, 15, 16). Our previous work shows that activation of apoC-II fibril formation by submicellar dihexanoylphosphatidylcholine (DHPC) occurs via rapid induction of tetrameric apoC-II species that structurally isomerize to become capable of nucleating fibril formation (16, 17). Addition of DHPC during the elongation phase of apoC-II fibril formation indicated that DHPC had no discernible effect on fibril elongation (17). These studies demonstrate that submicellar phospholipids and self-assembled phospholipid complexes can selectively target individual steps in the apoC-II amyloidogenic pathway.

Further investigation of the effect of submicellar lipid mimetics on apoC-II aggregation was conducted using a library of 96 lipid-like molecules (18). Interestingly, this library produced ∼30 compounds that inhibited the aggregation of apoC-II at a submicellar concentration. In general, these molecules stabilized an α-helical structure, with a molecular mass consistent with a dimer of apoC-II (18). These results indicate an additional mode of inhibition by lipid-like molecules.

Although we have information on the global secondary structure and stoichiometry of this oligomer, we lack any information about the physical structure of this apoC-II species. Thus, we have conducted a small-angle scattering study, using both x-rays and neutrons (SAXS, SANS), of the complex of apoC-II with dodecylphosphocholine (DPC), a representative inhibitor of fibril aggregation, to further elucidate the low resolution structure of both micelle stabilized and submicellar lipid-like molecule stabilized apoC-II complexes.

Materials and Methods

ApoC-II was expressed and purified as previously described (16). High purity (>99%) DPC was purchased from Sigma-Aldrich. All other reagents were of analytical grade.

Preparation of apoC-II for small-angle scattering measurements

ApoC-II was stored as ∼40 mg/mL stock in 5 M guanidinium hydrochloride (GuHCl) at −20°C. ApoC-II is completely unfolded in 5 M GuHCl, making this an ideal stock solution for initiating time courses and biophysical measurements of apoC-II’s folding and self-association behavior. To initiate refolding in the presence of DPC, the stock solution of apoC-II was diluted to a concentration of 4 mg/mL (∼448 μM) in phosphate buffer (100 mM sodium phosphate, pH 7.4) containing DPC (0.5 mM for submicellar DPC, 10 mM for micellar DPC). The samples were desalted using Nap5 desalting columns (GE Healthcare) equilibrated in phosphate buffer containing DPC (0.25 mM DPC for submicellar samples, 10 mM DPC for micellar samples), to remove the remaining GuHCl. The protein concentration of a 1 in 100 dilution of the eluate was determined using 280 nm absorbance and an extinction coefficient of 11,000 M⁻¹ cm⁻¹. The final apoC-II concentration, used for SAXS and SANS measurements, was 2 mg/mL (± 10%), consistent with the dilution from the desalting step. This concentration, although low for SANS measurements, is the highest concentration we can reasonably obtain with DPC. This is due to the CMC for DPC being ∼1 mM under these conditions and when approaching the CMC, there is the development of an equilibrium distribution between micellar and nonmicellar compound (16, 19). This equilibrium, along with a distribution of apoC-II associated with both the nonmicellar and the micellar compound, results in a polydisperse system that is difficult to measure by small-angle scattering techniques. Concentrations that exceed the DPC concentration result in aggregation (18). Samples were dialyzed against 100 mL of the appropriate phosphate buffer/DPC solution overnight before small-angle scattering (SAS) measurement. The dialyzed sample was centrifuged at 16,500 × g to remove any precipitate immediately before the SAS measurement. No significant change in protein concentration was observed after the dialysis step.

Thioflavin T measurements of amyloid formation

The prepared apoC-II/DPC complexes, along with apoC-II diluted from the GuHCl stock to 2 mg/mL, were incubated at room temperature for 96 h. Over this timeframe, aliquots (30 μL) were taken in triplicate at various time points (0, 5, 10, 30 min, and 1, 12, 24, 48, 96 h), and placed in a 96-well microplate with a clear bottom (Greiner). To each replicate, 170 μL of 100 mM phosphate buffer containing 35 μM thioflavin T (ThT) was added (final ThT concentration 30 μM). The ThT fluorescence intensity was immediately measured by exciting the sample with 444 nm light and monitoring the emission at 485 nm. A cutoff filter of 450 nm was used to reduce the contribution from scattered excitation light. Replicates were averaged, and a blank fluorescence measurement of a sample prepared exactly as described previously with the omission of apoC-II, was subtracted. The results of the assay were confirmed by centrifuging the final time point sample (96 h) at 100,000 × g and measuring the protein concentration of a 1 in 100 dilution of the supernatant and pellet fractions with a BCA assay kit (Pierce).

Small angle x-ray measurements

Measurements were made using the high-intensity undulator source on the SAXS/WAXS beamline of the Australian Synchrotron (Clayton, Victoria, Australia) (20). The beam size (full width at half-maximum at the sample) was 250 μm wide × 100 μm high with a total photon current of ∼2 × 10¹² photons s⁻¹. The q range measured was 0.005 to 0.50 Å⁻¹, where q = 4π sin θ/λ, 2θ is the x-ray scattering angle and λ, the x-ray wavelength, was set to 1.03 Å using a cryo-cooled Si (III) double-crystal monochromator. Absolute intensities were calibrated using water in the identical 1.5 mm ID quartz capillary used for analyzing the sample. The samples were slowly pumped through the capillary and 25 exposures were made at 1 s intervals. The exchange buffer from dialysis was used as the scattering blank.

SANS measurements

SANS experiments were conducted on the QUOKKA SANS beamline of the OPAL reactor, located at the Australian Nuclear Science and Technology Organization (ANSTO). Data were collected using a wavelength of 5 Å and two sample-to-detector lengths, a mid q-range length of 6 m, and a high q-range length of 1.3 m, resulting in a total q range of 0.002–0.56 Å⁻¹. Samples were introduced to the beam using 1 mm pathlength quartz cuvettes. Samples were prepared as described previously, however to conduct contrast matching a parallel set of samples was prepared in 40% D₂O, which matches the neutron scattering length density of the protein (as calculated by MULCh (21)) and thereby reduces its scattering signal. The exchange buffer of each sample was used as a scattering blank.

Data analysis

SAXS images were averaged and buffer subtracted using ScatterBrain V2.6 (http://www.synchrotron.org.au/aussyncbeamlines/saxswaxs/saxs-data-a-processing). SANS data were averaged, merged, and subtracted using IGOR Pro and the National Institute of Standards and Technology (NIST, Gaithersburg, MD) macros (22). Data were primarily analyzed using the ATSAS suite of SAS tools (23), applied in the following ways. Data were plotted using PRIMUS (24), radius of gyration, R_g, was calculated using AutoRg (25), the pair-distance distribution, P(r), and Porod volume were calculated using GNOM (26) and DATPOROD (25), respectively. De novo reconstruction from SAXS data was achieved using DAMMIN (27). This algorithm was run 10 times independently, and the resulting DAM models were superimposed, averaged, and filtered using DAMAVER in automatic batch mode (28). All de novo SAXS structures presented are the averaged and filtered structures. The SANS data were compared to existing structures for micelles (molecular dynamics (MD) model of a micelle containing 54 DPC molecules acquired from (29)) and for apoC-II associated with DPC micelles (from Research Collaboratory for Structural Bioinformatics (RCSP) Protein Data Bank (PDB): 1SOH (10)) with CRYSON (30). The micelle data was modeled as a sphere using BODIES (24), and de novo reconstruction of the SANS data was achieved through the use of MONSA (27), again with 10 independent analyses followed by superimposition and averaging. All SANS de novo structures are presented as the averaged and filtered structure. SUPCOMB (31) was used to overlay the de novo structure of apoC-II in association with micelles with the top 10 structures provided by the previous NMR analysis (PDB: 1SOH (10)). Errors are typically propagated from the data, and are evaluated through numerical solutions implemented by the various programs used to analyze the data.

Results

Inhibition of apoC-II aggregation by DPC

Previous ThT and analytical ultracentrifugation measurements indicate that the presence of excess submicellar and micellar DPC inhibits apoC-II aggregation over the protein concentration range of 0.3–1.0 mg/mL (10, 18). However, the effect of DPC at the concentrations of apoC-II above the detection limit of the SANS instrument (minimum concentration ∼2 mg/mL), where reliable data can be acquired, has not been investigated. To confirm that submicellar and micellar concentrations of DPC maintained their inhibitory effects at the higher concentrations of apoC-II, a ThT assay over 96 h was conducted on the apoC-II/DPC complexes (Fig. 1 A). ApoC-II alone rapidly self-associates into amyloid fibrils, forming ThT-positive aggregates before the first measurement, and reaching a plateau in signal by 5 min, which did not change significantly over the 96 h period. However, the complexes of apoC-II with submicellar and micellar DPC showed no change in ThT fluorescence over the time course, indicating no amyloid fibril formation. This result was confirmed by a brief pelleting assay after 96 h of incubation (Fig. 1 B). ApoC-II in the absence of detergent has <5% protein remaining in the supernatant, whereas both of the apoC-II/DPC complex samples contain 95% of the starting protein.

Figure 1.

Aggregation of apoC-II. (A) Measurement of the aggregation of apoC-II (2mg/mL) in the absence (circles) and presence of submicellar (0.25 mM, squares) or micellar DPC (10 mM, triangles) was conducted using a point-based ThT fluorescence assay. (B) The ThT measurement of the final point of the time course (solid bars) was confirmed using a pelleting assay, where the protein concentration of the supernate (gray bars) and pellet (open bars) was measured after centrifugation at 100,000 × g, via BCA assay.

SAXS analysis of the structure of apoC-II in the presence of submicellar DPC

We have previously investigated apoC-II in the presence of submicellar DPC using circular dichroism and analytical ultracentrifugation, indicating the formation of a dimer with a high content of α-helical secondary structure (18). However, we have little understanding of the molecular arrangement of apoC-II in complex with submicellar DPC. It is likely this complex is relatively flexible, and unlikely to crystallize well, thus SAXS is ideally suited to define the low-resolution structure of the apoC-II/DPC dimer complex. ApoC-II in the presence of submicellar DPC produced a high quality scattering pattern (Fig. 2 A), with a well-defined Guinier region (Fig. 2 C), indicating a monodisperse sample with an R_g of 32.1 ± 0.3 Å (Table 1) (an unfolded apoC-II monomer has an R_g of ∼18 Å, calculated from (32)). A Kratky plot of the data is consistent with the formation of a globular structure, with a well-defined peak at low q (Fig. 2 B). Analysis of the data with GNOM (26) provided a P(r) distribution with a maximal dimension, D_max, of 100 ± 0.3 Å, and an R_g of 31.8 ± 0.4 Å (Fig. 2 D). Setting R(min) to float from 0, and ignoring the first 35 and last 50 points of the scattering data did not significantly alter the P(r). De novo reconstruction of the scattering envelope using DAMMIN (27), produced an approximately cylindrical structure (Fig. 2, E and F, average structure of 10 independent reconstructions, nominal spatial deviation (NSD) of 0.883), which displays twofold symmetry despite no symmetry constraints being used in the analysis. Porod analysis of the data gave a molecular mass of 17 ± 2 kDa, consistent with the formation of a dimer of apoC-II (Table 1), and our previous analytical ultracentrifugation analysis of the dimer (18).

Figure 2.

SAXS analysis of the apoC-II/DPC complex. (A) SAXS data acquired at the Australian Synchrotron for apoC-II in complex with submicellar DPC (circles). The solid lines represent the DAMMIN fit of the data. (B) the Kratky plot of the data in (A) is consistent with a globular structure. (C) Guinier plot of the data in (A), indicating an R_g of 32.1 Å. (D) P(r) generated from the data in (A), indicating a D_max of ∼100 Å and an R_g of ∼31.8 Å. (E and F) The de novo structure of apoC-II in complex with DPC generated by DAMMIN (long axis, E; short axis, F). To see this figure in color, go online.

Table 1.

Values for D_max, R_g, and Molecular Mass Obtained from the Various Analyses of the Data in Figs. 2, 3, 4, and 5.

Sample	% D2O	Guinier Rg (Å)a	P(r) Rg (Å)b	P(r) Dmax (Å)b	Molecular Mass (kDa)c
SAXS: apoC-II + submicellar DPC	0	32.1 ± 0.32	31.8 ± 0.41	100.2 ± 1.2	17.1 ± 2
SANS: apoC-II + submicellar DPC	0	32.4 ± 0.42	31.2 ± 0.21	98.1 ± 1.1	18.1 ± 2.1
SANS: DPC micelle	40	16.9 ± 0.35	17.1 ± 0.24	40.1 ± 0.5	18.5 ± 1.2
SANS: DPC micelle + apoC-II	40	17.2 ± 0.20	17.0 ± 0.36	40.2 ± 0.4	18.0 ± 1.3
SANS: DPC micelle + apoC-II	0	20.6 ± 0.32	21.1 ± 0.35	61.1 ± 0.8	8.5 ± 1.1

^aCalculated using AutoRg; error provided by AutoRg from numerical assessment of the fit region to the data.

^bCalculated using GNOM; error in the Rg and D_max values is propagated from the error in the measurement.

^cCalculated using DATPOROD, error calculated by propagation from error in the data, and fit quality.

SANS analysis of the structure of apoC-II in the presence of submicellar DPC

SANS experiments of apoC-II in the presence of submicellar DPC in nondeuterated buffer produced similar profiles to the SAXS data (Fig. 3 A). Under these conditions (0% D₂O, submicellar DPC) the DPC does not contribute significantly to the scattering curve (MULCh calculated match point is 8.8% D₂O, Fig. S1 in the Supporting Material). SANS experiments in deuterated (40% D₂O) buffers, where the protein’s contribution to the scattering is reduced, resulted in no scattering above the background of the buffer (Fig. S2). Kratky (Fig. 3 B) and Guinier analysis (Fig. 3 C) of the 0% D₂O data indicated a globular, monodisperse sample, with an R_g of 32. 4 ± 0.4 Å (Table 1). P(r) analysis of the data with GNOM (Fig. 3 D) gave a D_max, of 98 ± 1 Å and an R_g of 31.2 ± 0.2 Å (Table 1), consistent with the values provided by the SAXS data. De novo reconstruction indicated a similar scattering envelope for the complex as the SAXS data (Fig. 3, E and F, average of 10 independent MONSA calculations, NSD 0.951). Porod analyses gave a molecular mass of 18 ± 2 kDa.

Figure 3.

SANS analysis of the apoC-II/DPC complex. (A) SANS data acquired at ANSTO for apoC-II in complex with submicellar DPC (circles). The solid lines represent the GNOM fit (red) and the MONSA fit (blue) of the data. (B) The Kratky plot of the data in (A) is consistent with a globular structure. (C) Guinier plot of the data in (A), indicating an R_g of 32.4 Å. (D) P(r) generated from the data in (A), indicating a D_max of ∼98 Å and an R_g of ∼31.2 Å. (E and F) The de novo structure of apoC-II in complex with DPC generated by MONSA (long axis, E; short axis, F). To see this figure in color, go online.

SANS analysis of apoC-II/DPC micelle complexes

ApoC-II in the presence of micellar DPC is difficult to study by SAXS, as the scattering is dominated by the larger micelles, which is highly challenging to deconvolute from the protein scattering profile. Thus, we have used contrast-variation SANS analysis, where the signal arising from the DPC micelle (or the apoC-II) can be reduced through the addition of D₂O, to characterize this complex. SANS indicated a well-defined structure both in deuterated and nondeuterated buffer systems (Figs. 4 and 5). In 40% D₂O, where protein scattering is suppressed, the presence of apoC-II made little difference to the shape of the DPC micelles (Fig. 4). The P(r) in the presence or absence of apoC-II are not significantly different, and gave values for R_g and D_max of ∼17 Å and ∼40 Å, respectively (Fig. 4 B; Table 1). Guinier analysis showed that micelles in the DPC alone sample and the sample in the presence of apoC-II are monodisperse, with R_g values of 16.9 Å and 17.2 Å, respectively (Fig. 4 C; Table 1). Further analysis of the DPC micelle only data with BODIES gave an excellent fit (Fig. 4 D, χ²= 1.089) to an ellipsoid with major and minor axes of 20.1 Å and 19.8 Å, respectively, indicating the formation of a sphere with a diameter of ∼40 Å. Prolate ellipsoid and cylinder models gave poor fits (χ² > 1.9). The data provided a molecular mass of 18 ± 1.2 kDa, giving an approximate stoichiometry of 50–56 DPC molecules/micelle. These parameters are similar to those of a series of MD simulations of a micelle of 54 DPC molecules (29). Fitting of a MD generated DPC micelle, containing 54 DPC molecules, (Fig. 4 F (29)) to the scattering data using CRYSON provides a good representation of the data (Fig. 4 E, χ² = 1.16), and are also consistent with a previous NMR/SAXS study of DPC micelle formation (33). However, other small-angle scattering studies of DPC micelles found slightly greater aggregation numbers, and that the micelles were more of a prolate ellipsoidal shape, and fit well to a core shell model with a minor axis of ∼16.1 Å and a major axis of ∼24.3 Å (34, 35). These differences may be due to differences in buffer composition, as DPC micelle formation and structure is quite dependent on the solution environment.

Figure 4.

SANS analysis of DPC micelles in 40% D₂O. (A) SANS data for DPC micelles alone (black circles) and DPC micelles in the presence of apoC-II (blue circles). Solid lines indicate the GNOM fits generated during calculation of P(r) (Red: DPC alone, orange: DPC + apoC-II). (B) P(r) for DPC micelles alone (red line), and for DPC micelles + apoC-II (black line). (C) Guinier analysis for DPC micelles alone (black circles, red line) and in the presence of apoC-II (blue circles, orange line). The two data sets are offset for clarity, but otherwise overlay. (D) Fit of DPC micelles to a BODIES model indicating that the data is consistent with a sphere with a diameter of 40 Å. (E) CRYSON fit to a MD simulation of a DPC micelle containing 54 individual DPC molecules. (F) DPC micelle structure used for CRYSON fit in (E). To see this figure in color, go online.

Figure 5.

SANS analysis of apoC-II structure in association with DPC micelles. (A) Data acquired for apoC-II in nondeuterated buffers. Solid line represents the MONSA fit (red line) and the CRYSON fit (blue line). (B) Kratky plot of the data in (A). (C) Guinier analysis, indicating an R_g of 20.6 Å. (D) P(r) calculated using GNOM, representing the ensemble of apoC-II structures likely to be present in the sample. (E–G) indicate various views of the top 10 NMR structures (various colors in cartoon format) represented in PDB: 1SOH superimposed with the de novo scattering envelope for apoC-II calculated using MONSA (gray transparent spheres). To see this figure in color, go online.

In nondeuterated buffer (Fig. 5), where there was insignificant contribution from the DPC micelles to the scattering data (Fig. S1), the Kratky plot indicated the formation of a globular structure (Fig. 5 B). Guinier analysis of the scattering data provided an R_g of 20.6 ± 0.3 Å (Fig. 5 C; Table 1), whereas P(r) analysis gave an R_g of 21.1 ± 0.4 Å and a D_max of 61.1 ± 0.8 Å (Fig. 5 D; Table 1). Reconstruction of these data indicated a horseshoe type structure (Fig. 5, E–G, average structure of 10 independent calculations, NSD 1.1), that agrees closely with the ensemble of structures obtained from NMR analysis of apoC-II in the presence of DPC micelles. A CRYSON fit to the first 10 consensus structures in 1SOH.pdb further supports that the data is representative of this structure (fit to the first model is shown in Fig. 5 A, the 10 structures gave χ² of between 1.12 and 1.15). The top 10 structures were chosen to give an accurate representation of the NMR ensemble structure. Porod analysis gave a molecular mass of 8.5 ± 1.1 kDa (Table 1).

Discussion

The results, showing that apoC-II adopts a horseshoe shaped α-helical conformation in the presence of DPC micelles and a cylindrical dimeric structure in submicellar DPC, indicate that there are two distinct mechanisms of inhibition of apoC-II amyloid fibril formation by lipid-like molecules. These mechanisms are well supported by our previous work. The first and most studied mode of inhibition is the induction of a native-like α-helical fold of apoC-II by micellar lipids; whereas the second, more recent observation, is inhibition of fibril formation through the induction of a symmetrical dimeric structure by specific binding of submicellar lipid molecules. The micelle-induced horseshoe structure is highly consistent with the ensemble of structures solved by MacRaild et al. (9, 10), and further characterized by Zdunek et al. (11), which contain ∼80% amphipathic α-helical structure. The conformations of the extended α-helical structure of these NMR structure ensembles showed some diversity, as long-range nuclear Overhauser effects were weak signals. Our results indicate that the horseshoe structure of apoC-II is stable on lipid-like surfaces. The data indicating a mass of 8.5 kDa for apoC-II (Table 1), and dimensions consistent with the monomeric NMR structures (Fig. 5), further confirm that under these conditions, apoC-II associates as 1 protein molecule per DPC micelle (or 1 apoC-II per 50–56 DPC molecules). This can account for some of the inhibitory activity of lipid surfaces, where the 1 protein per micelle interaction prevents close association of apoC-II molecules and thereby reduces misfolding, self-association, and aggregation. Indeed, increasing the protein content of lipid micelles, such that there is a ratio of more than one apoC-II molecule per micelle, results in the formation of rod-like fibrils (15). Interestingly, the binding of apoC-II does not seem to significantly affect the structure of the DPC micelle. This is consistent with the role of apoC-II as an exchangeable and dynamic cofactor for lipoprotein lipase, where it is expected to have little role in the formation and stability of lipoprotein particles.

The second mechanism, whereby individual, submicellar DPC molecules interact directly with apoC-II and induce a stable α-helical structure (∼60% α-helix by circular dichroism spectroscopy), is a relatively new observation (18). SAXS and SANS reconstructions are remarkably similar, and indicate a symmetrical cylindrical structure with an overall dimension of ∼100 Å, and a radius of ∼18 Å. These dimensions easily accommodate two apoC-II molecules; however the exact arrangement of the molecules in this scattering envelope is still unclear. Thus, we can only speculate on the basis of our previous work with other lipids and lipid mimetic molecules that show both inhibitory and activating activities at submicellar levels, as to the localization of DPC and the molecular basis of the DPC activity on apoC-II. DPC associates 2:2 with apoC-II in the dimer, suggesting that each individual apoC-II component of the dimer has a single DPC molecule associated with it (18). Analysis of the interaction of lipids with apoC-II using Förster resonance energy transfer measurements between nucleotide-binding domain-labeled lysophospholipids and Alexa-594-labeled S61C- mutants of apoC-II suggests that the lipid binding site is located within 12 Å of this residue (36). Thus, on the basis of linear distances between residues in an unfolded apoC-II, the DPC binding site may be located within the region S₅₄KSTAAMSTYTGIFT₆₈, This region includes an amyloidogenic sequence, and is heavily involved in the interactions required to form amyloid fibrils (37). MD investigations of the binding of lipids to peptides comprising the 60–70 amino acid region of apoC-II indicate that the main interactions occur between the rings of Y63 and F67 and the hydrophobic acyl chains of the lipid, which induce an extended conformation in the backbone of the peptide (38). In other analyses of the interaction of this particular peptide with lipids the formation of stable dimeric structures was observed (39). Further MD analyses of how this amyloidogenic peptide can self-associate in the presence of lipids suggest that the stabilization is mediated by promotion of intermolecule hydrophobic interactions (39). These results indicate that this region has a parallel configuration in its lipid-induced oligomeric form, suggesting that the DPC-induced full-length apoC-II dimer may also adopt this configuration. ApoC-II amyloid fibrils have an in-register, parallel conformation of β-sheet strands (37), so it is likely that this stabilized dimeric species may have some relation to the orientation of apoC-II molecules in the amyloid fibril nucleus. Indeed, in full-length apoC-II, the stabilization of tetrameric forms of the protein by lipids promotes the formation of amyloid fibril nuclei (16), with concomitant dissociation of the lipid molecule (36). Thus, submicellar phospholipids increase the rate of apoC-II fibril formation in a catalytic fashion. The stabilization of a dimeric species by DPC suggests that the conformational change in apoC-II required to become competent to form amyloid-forming nuclei is prevented. This may be due to a range of factors, but one mechanism may be a stronger interaction where the dissociation of the lipid-mimetic does not occur. Given the small differences between activating phosphatidylcholines and the inhibiting acylphosphocholine detergents, determining the affinity of lipids and lipid mimetics for transient oligomeric species is not trivial and it is likely that elucidation of the molecular basis of the submicellar DPC-apoC-II interaction and the resulting apoC-II dimerization will not be conclusively solved until high-resolution structures are available. These results may provide some insight into the influence of lipids on the formation of amyloid by all apolipoproteins. As reviewed by Hatters et al. (40), and more recently by Das et al. (41), the highly conserved structural features of the apolipoprotein superfamily also appear to confer a propensity to form amyloid structures. The entire superfamily is thought to be derived from duplication of a single 22 amino acid tandem repeat, which forms the basis of the prevalent class A amphipathic helix structure of the lipid interacting regions of the apolipoprotein (40). This structure results in a parallel binding orientation on the lipid surface, which is critical for the exchangeable properties of the lipoproteins, providing a less avid interaction to the lipid surface than a transmembrane domain. However, these regions also display significant structural instability in the absence of lipids, resulting in self-association, aggregation, and amyloid fibril formation. ApoC-II is a typical apolipoprotein in that in the presence of lipids it has a predominantly α-helical structure, with a very high proportion of amphipathic helix structure. Our previous work indicates that factors that influence apoC-II aggregation can also influence other apolipoprotein self-association events. For example, apoC-I responds to submicellar lipids to form dimers (19) and apoE has been demonstrated to alter its oligomerization properties in the presence of micellar and submicellar lipids (42). However, apoC-II is one of the smallest apolipoproteins, and has less additional sequence and structure to defend against destabilizing influences, such as the absence of lipid surfaces. Despite this observation, a stabilizer of apoC-II oligomers may give insight into the stabilization of other apolipoproteins, and may lead to therapeutic avenues for apolipoprotein amyloid-related diseases. Thus, the activity of these amyloid inhibiting compounds warrants further research to determine the structure activity relationships of the lipid mimetics, and the structural basis for the stability of the dimeric species.

We have also observed stabilization by lipid-mimetics and hydrophobic small molecules in Aβ_1–42 amyloidogenesis, where a range of lipid-mimetics (43), and the hydroxyquinolines clioquinol and PBT2 (44), interact with the peptide and stabilize it in a nontoxic dimeric conformation. Although the defining feature of these molecules is their hydrophobicity, and amphipathic nature, it should be noted that there was a differing pattern of inhibiting compounds for Aβ and apoC-II, and that relatively subtle differences in acyl chain length, headgroup size, and headgroup chemistry resulted in significantly different activities in regards to amyloid fibril formation (18, 43). These results suggest that a nonspecific hydrophobic interaction is not the mechanism of inhibition by these compounds, and that more defined, specific interactions involving both the acyl chain and headgroup mediate the interactions between these compounds and apoC-II. These results also suggest that the interaction with different specific hydrophobic molecules can modulate amyloid fibril formation by a wide range of amyloid-forming proteins, and may be a viable therapeutic approach that does not require absolute knowledge of the toxic oligomeric structures and conformations that are present in the various amyloid-related diseases (44). Indeed, a related mechanism has been demonstrated as the only effective treatment for transthyretin (TTR) systemic amyloidosis (45, 46, 47). In this case, amyloidogenesis involves release of monomeric protein from a native tetrameric structure, which then misfolds and aggregates. Therapeutics, based on the native TTR substrate thyroxine, have been developed that bind with high affinity (K_d 4 nM) to TTR and stabilize the tetrameric form of TTR, thereby halting amyloidogenesis (45, 46, 47). These observations suggest that a general therapeutic approach to target amyloid formation, where the aim is to stabilize an inert, nontoxic oligomer may be viable in a majority of amyloid-related diseases.

These results indicate that lipid molecules can play an important role in protein folding and aggregation. However, the role is poorly understood due to the complexity of biophysical properties of lipids. Here, we show that the micellar state of an inhibitory lipid can have a very significant effect on protein conformation, with micelles stabilizing a particular α-helical structure, whereas submicellar lipids stabilize what appears to be a very different dimeric, α-helical structure. These differences indicate that lipid homeostasis and modification may be a significant factor in the prevention, development, and progression of amyloid-related diseases.

Author Contributions

T.M.R., M.D.W., N.K., K.W., C.C.C., and R.B.K. conceived, designed, and conducted the experiments; T.M.R., K.W., R.B.K., C.L.M., and D.J.M. helped design experiments, analyze results, and interpret the findings. The article was written through contributions of all authors.

Editor: David Eliezer.