Intra-membrane Oligomerization and Extra-membrane Oligomerization of Amyloid-β Peptide Are Competing Processes as a Result of Distinct Patterns of Motif Interplay*

Background: It is currently unclear whether soluble oligomers of the amyloid-β peptide (Aβ) can cause neurotoxicity by direct membrane incorporation. Results: Aβ monomers but not the soluble oligomers readily insert into the membrane followed by self-assembly into membrane-embedded oligomers. Conclusion: Solution-phase oligomerization and membrane insertion of Aβ are mutually exclusive processes that proceed through distinct pathways. Significance: The competing intra- and extra-membrane oligomerization of Aβ may determine distinct neurotoxic mechanisms. Soluble oligomers of amyloid-β peptide (Aβ) are emerging as the primary neurotoxic species in Alzheimer disease, however, whether the membrane is among their direct targets that mediate the downstream adverse effects remains elusive. Herein, we show that multiple soluble oligomeric Aβ preparations, including Aβ-derived diffusible ligand, protofibril, and zinc-induced Aβ oligomer, exhibit much weaker capability to insert into the membrane than Aβ monomer. Aβ monomers prefer incorporating into membrane rather than oligomerizing in solution, and such preference can be reversed by the aggregation-boosting factor, zinc ion. Further analyses indicate that the membrane-embedded oligomers of Aβ are derived from rapid assembly of inserted monomers but not due to the insertion of soluble Aβ oligomers. By comparing the behavior of a panel of Aβ truncation variants, we demonstrate that the intra- and extra-membrane oligomerization are mutually exclusive processes that proceed through distinct motif interplay, both of which require the action of amino acids 37–40/42 to overcome the auto-inhibitory interaction between amino acids 29–36 and the N-terminal portion albeit via different mechanisms. These results indicate that intra- and extra-membrane oligomerization of Aβ are competing processes and emphasize a critical regulation of membrane on the behavior of Aβ monomer and soluble oligomers, which may determine distinct neurotoxic mechanisms.

Alzheimer disease (AD) 4 is the most common cause of dementia characterized by extracellular amyloid plaques and intracellular neurofibrillary tangles (1,2). The major component of amyloid plaque is aggregated amyloid-␤ peptide (A␤) with a length of 39 -43 residues. A␤ is a secreted peptide generated by sequential processing of amyloid precursor protein by ␤and ␥-secretases. Accumulating in vitro, in vivo, and genetic evidence indicates that A␤ is causally involved in the pathogenesis of AD, forming the basis of the prevalent amyloid cascade hypothesis (3,4). The A␤ aggregates, in particular the low molecular weight soluble oligomers, are emerging as the primary toxic species (1,2,5,6). Indeed, pathogenic A␤ soluble oligomers have been isolated from AD-affected human (7,8) and mouse model (9) brains, and the cerebrospinal fluid level of A␤ oligomers is considered to be a highly promising biomarker for AD diagnosis (10).
Interaction with the cell membrane has been proposed as one of the key mechanisms for A␤ to exert its neurotoxicity (11)(12)(13). For example, the membrane can promote aggregation of A␤ monomers, while after incorporating into the membrane, A␤ is able to form channel-like oligomers (14,15) or even disrupt membrane integrity (16,17) resulting in cell dysfunction. Although membrane insertion, oligomerization, and neurotoxicity of A␤ appear to be tightly coupled, the correlation between soluble and membrane-embedded oligomers of A␤ remains elusive. In addition, it is currently unclear whether soluble A␤ oligomers can act by direct membrane incorporation.
In the present study, we have characterized the interaction of membrane with A␤ in multiple assembly states and conclude that solution-phase oligomerization and membrane insertion of A␤ are mutually exclusive processes. Our results further indicate that the intra-and extra-membrane oligomerization of A␤ proceed via distinct pathways and may determine separate neurotoxic mechanisms. These results not only reveal an important regulation of membrane on the behavior of A␤ but may provide clues for designing stage-specific and A␤-targeted therapy.
Monomeric A␤ Preparation-A␤ monomer free of aggregation was prepared as described with minor modification (18). Lyophilized A␤ was stored in sealed glass vial at Ϫ80°C. After equilibrated for 30 min at room temperature, A␤ was dissolved to 1 mM in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Sigma) by extensive vortex. The peptide was further incubated at 4°C for 2 h with continuous shaking (ϳ70 rpm) and stored in small aliquots at Ϫ80°C after evaporating off HFIP by N 2 stream. Immediately prior to experiments, the HFIP-treated peptide aliquot was re-suspended to 5 mM in anhydrous dimethyl sulfoxide (DMSO, Sigma) by brief vortex followed by water batch sonication (1 min each).
Oligomeric and Fibrillar A␤ Preparation-For preparation of the low molecular weight soluble oligomer A␤-derived diffusible ligand (ADDL), 100 M monomeric A␤ was incubated in F-12 medium (Ham's F-12, BioSource, Australia) for 24 h at 4°C (18,19). A␤ protofibril (PF) was prepared by incubating 100 M A␤ monomer in TBS (50 mM Tris, 100 mM NaCl, pH 7.4) for 24 h at room temperature followed by 14,000 ϫ g centrifugation to remove large aggregates (11,20,21). A␤ fiber was prepared by incubating 500 M A␤ monomer in TBS for 2 weeks at room temperature and pelleted by centrifugation at 14,000 ϫ g.
Langmuir Film Balance-Monolayer experiments were conducted with a Trough-S microbalance (Kibron, Finland) as described previously (22). Lipids were spread onto the TBS buffer (10 mM Tris, 140 mM NaCl, pH 7.4) filled in the trough to achieve a stable initial surface pressure ( i ). A␤ was injected into the subphase to a final concentration of 600 nM through the side hole. The membrane pressure () was monitored at intervals of 1 s until a stable value was reached, usually within 5000 s. The measurements were performed at a temperature of 23.5 Ϯ 0.5°C with continuous stirring. In a typical assay with a constant surface area, the increase in surface pressure of monolayer (⌬) reflects the membrane insertion of A␤. Linear fitting of ⌬ versus i yields a straight line with negative slope, which intersects the x axis at the critical membrane insertion pressure ( c ). c represents the highest surface pressure of a monolayer below which a protein can insert, thereby quantitatively defining the membrane insertion capacity. The surface pressure of physiological lipid bilayer is ϳ30 -32 millinewtons (mN)/m (23)(24), indicating that a protein can insert into cell membrane only when its c is Ͼ30 mN/m. To detect the aggregation state of monolayer inserted A␤, experiments were conducted with a constant surface pressure, and A␤ insertion would result in surface area expansion. After 5000 s, monolayers were collected into tubes via negative pressure produced by vacuum for immunoblotting.
Liposome Experiments-Large unilamellar liposomes were prepared using a mini-extruder (Avanti) as described previously (22). After incubation of A␤ with liposomes for the indicated times, a 10-min centrifugation at 14,000 ϫ g was conducted to pellet large A␤ aggregates. The resulting supernatant was subjected to additional centrifugation and SDS-PAGE analysis as indicated in Fig. 3A. In some experiments, a 30-min treatment with acidic buffer (1 M NaCl, 20 mM NaH 2 PO 4 , pH 2.0 adjusted by acetic acid) was used to dissociate low affinity interaction of A␤ with liposomes. In cross-linking experiments, PBS was used instead of TBS. Alternatively, A␤ was mixed with DPPC in chloroform at a molar ratio of 1:200. The mixture was dried under an N 2 stream and resuspended in 10 mM TBS (pH 7.4) to 1 mg/ml of lipid concentration. Then a 20-min bath sonication was used to prepare reconstituted proteoliposome with integrated A␤ (25).
FRET Assay-Fluorescein (0.5 M)-labeled A␤ (donor) and 0.5 M tetramethylrhodamine-labeled A␤ (acceptor) were coincubated with liposomes at the indicated peptide/lipid ratio with continuous stirring under room temperature. An LS-55 fluorometer (PerkinElmer Instruments) was used to detect the fluorescence emission at 588 and 540 nm (5 nm slit width) with an excitation wavelength of 470 nm (2.5 nm slit width). The FRET ratio was calculated as F A(588 nm) /F D(540 nm) , where F A(588 nm) and F D(540 nm) represent acceptor and donor emission intensities, respectively.
EM-A 5-l droplet of A␤ sample (ϳ45 g/ml) was added to a freshly glow-discharged carbon-coated EM grid for 1 min followed by staining with 1% sodium phosphotungstate for 30 s. The grids were examined with a Tecnai G20 (FEI) transmission EM.

RESULTS
The more amyloidgenic isoform A␤1-42 was used and is referred to as A␤ hereafter in the subsequent experiments unless otherwise stated. To investigate the interaction of soluble A␤ oligomers with membrane, we prepared and carefully characterized four A␤ samples in different assembly states, i.e. monomer, low molecular weight soluble oligomer (ADDL, A␤-derived diffusible ligand), PF, and mature fiber following the established protocols (11, 18 -20). These A␤ species exhibited expected features in SDS-PAGE, EM observation, ThT fluorescence, and cytotoxicity assays ( Fig. 1 and its legend) (11, 18 -20, 26).

Soluble A␤ Oligomers Exhibit Impaired Membrane Insertion
Capability-We examined A␤-membrane interactions by using Langmuir film balance, which measures changes in surface pressure of lipid monolayer as an index of protein insertion. The term of "insertion" here means part of the tested molecule is incorporated into the hydrophobic core of monolayer resulting in the increase in monolayer surface pressure. The injection of A␤ monomer evoked an abrupt rise of the surface pressure of DPPC monolayer ( Fig. 2A), reflecting a strong insertion of A␤ monomer. By contrast, the addition of fiber resulted in only marginal changes, whereas ADDL and PF induced moderate surface pressure increases and required longer time to reach the equilibrium. Increasing the amount of injected ADDL did not rescue the insertion defects. Similar findings were also obtained with DPPS and lipid raft (a lipid raft-mimic component)-mimic monolayers (Fig. 2, B and C). Further quantitative analysis revealed that the A␤ monomer possessed a ϳ1.5to 4-fold higher affinity for lipid monolayer and ϳ1.5-fold stronger maximal monolayer insertion than ADDL (Fig. 2D). These findings indicated that prior oligomerization in solution impaired the intrinsic capability of A␤ in membrane insertion regardless of lipid composition. Because the lateral pressure of physiological lipid bilayer is ϳ30 mN/m (23,24), the lower values of the critical membrane insertion pressure ( c ) (Fig. 2E) suggest that ADDL is unable to directly insert membrane bilayer. Consistent with this speculation, after a 2-h co-incubation with liposomes ADDL still resided in solution phase, whereas A␤ monomer was primarily found to be associated with liposomes (Fig. 2F). The above results demonstrate that soluble A␤ oligomers exhibit impaired insertion capability and low affinity to membrane and emphasize that incomplete monomerization of A␤ will lead to an underestimation of the membrane insertion capacity.
Membrane-inserted A␤ Monomers undergo Rapid Oligomerization-Remarkably, incubating 1 M A␤ monomers with liposomes resulted in rapid occurrence of membrane-associated oligomers that co-sedimented with liposomes in ultracentrifugation, whereas no self-assembly could be observed for A␤ monomers incubated alone (Fig. 3A). FRET assay using A␤ monomers tagged with fluorescein and tetramethylrhodamine further revealed a fast kinetics of membrane-induced oligomerization showing almost no lag time (Fig. 3B). The A␤ oligomers were not dissociated from liposomes by harsh acidic treatment (Fig. 3A), suggesting they were membrane-embedded. Indeed, their assemble pattern was essentially identical as that of the channel-like oligomers within membrane prepared directly from A␤-lipid fusion mixtures ( Fig. 3C) (25). Because the soluble A␤ oligomer, i.e. ADDL, was unable to associate with liposomes (Fig. 2, E and F), the intra-membrane oligomers should derive from the assembly of inserted A␤ monomers instead of the insertion of A␤ oligomers formed in solution or on membrane surface. Accordingly, the interaction of A␤ monomers with monolayer did not follow the pattern of simple insertion; instead it fitted well with a two-phase process (Fig. 3D). The early and late phases most likely represent the initial insertion and the subsequent intra-membrane conformational change or oligomerization, respectively. Consistent with this speculation, the monolayer-inserted A␤ was largely oligomeric, whereas in . A␤ monomer primarily migrated as a single band at ϳ4.5 kDa in SDS-PAGE and was invisible under EM, whereas ADDL was characterized by bands corresponding to trimer/tetramer in SDS-PAGE and particle-like appearances with even size distribution (ϳ5 nm) under EM. There were additional high molecular weight bands in PF and fiber, and these two samples showed expected morphology. The absence of fiber in A␤ monomer, ADDL, and PF preparations were further verified by the comparable low ThT fluorescence as compared with that of mature fiber. D, cell viability of N2a cells treated with different sterile A␤ species for 48 h was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Results (n Ͼ 3) are given as mean Ϯ S.E.; *, p Ͻ 0.05; **, p Ͻ 0.005. At low concentrations PF and ADDL are significantly more toxic than monomer sample in cell viability assay.
the subphase A␤ remained monomeric (Fig. 3E). A scrambled A␤ control, on the other hand, exhibited only one-phase membrane insertion.
Intra-and Extra-membrane Oligomerization of A␤ Are Competing Processes-An interesting feature of the interaction between A␤ monomer and membrane was that, at non-saturating low concentration (1:200 molar ratio; 1 M), essentially all peptide translocated to liposomes with no detectable signal of A␤ monomer or oligomer left in the solution fraction (Fig. 3A). This suggests that A␤ monomers prefer membrane association to oligomerization in solution at least at low micromolar concentration. However, the presence of aggregation-boosting factor, i.e. zinc ion (27) (Fig. 4A), nearly abrogated the membrane insertion of A␤ monomer, although it did not suppress the surface pressure increase evoked by A␤17-42, an N-terminal truncated isoform without the zinc binding site (28) (Fig. 4B). As comparison, another bivalent ion, calcium, was much less potent than zinc in promoting A␤ self-assembly, and only marginally inhibited membrane insertion of A␤ monomer as expected. Nonetheless, the addition of zinc after A␤ incorporation into monolayer failed to reverse the insertion. These findings indicate that solution-phase oligomerization and membrane insertion of A␤ are competitive in nature, which is also in line with the observed inverse correlation between the aggregation extent and insertion capacity (Fig. 2).
Because A␤ monomers and soluble A␤ oligomers, such as ADDL, are more or less in continuous interconversion (29,30), we reasoned that the preferential incorporation of A␤ monomers into membrane might gradually shift the equilibrium of monomers 7 ADDL transition, thereby leading to the accumulation and oligomerization of A␤ in liposomes. Indeed, although ADDL showed no detectable membrane association after a 2-h incubation (Figs. 2F and 4C), prolonged incubation (6 -48 h) of ADDL with liposomes resulted in the occurrence of membrane-associated A␤ in liposome fraction (Fig. 4C) that was resistant to acidic striping (Fig. 4D) despite the fact that most of ADDL still remained in solution phase as observed by EM (Fig. 4E). Moreover, the slow accumulation of A␤ oligo- Because the c of ADDL is lower than 30 mN/m, the physiological lateral pressure of cell membrane, this suggests that, unlike A␤ monomer, ADDL is unable to directly insert membrane bilayer. F, 1 M A␤ monomer or ADDL was incubated with 200 M DPPC liposomes for 2 h at room temperature. A␤ in the liposome-bound (P1) and supernatant fractions (S1; please refer to the scheme shown in Fig. 3A for the detailed preparation protocol) were obtained by ultracentrifugation at 200,000 ϫ g for 30 min and probed by immunoblotting with 6E10. A␤ monomer was detected only in the liposome fraction while ADDL was exclusively detected in the solution fraction. mers in liposomes (Fig. 4C) suggests that they were formed by gradual intra-membrane assembly by inserted A␤ monomers but not direct incorporation of ADDL. Therefore, we conclude that the extra-and intra-membrane oligomerization (or membrane insertion) of A␤ are competing processes (Fig. 4F).
Distinct Pathways of Intra-and Extra-membrane Assembly of A␤-To dissect the mechanism of the intra-membrane oligomerization, we further analyzed the interactions of a panel of A␤ variants with membrane (Fig. 5, A and B). A␤1-28, the N-terminal extracellular portion of A␤, was unable to insert both monolayer and bilayer, highlighting the importance of the C-terminal trans-membrane portion of A␤1-42 in mediating membrane insertion. Indeed, even a short truncation in the C-terminal of A␤1-42 impaired its interaction with membrane as evidenced by the abrogated and moderately reduced liposome association of A␤1-36 and A␤1-40, respectively. Given that the two peptides possess sufficient high c to insert the physiological bilayer, these results indicate that aa 29 -36 are likely to be the basic unit for metastable/reversible membrane insertion of A␤, whereas aa 37-40/42 are additionally required for firm attachment. Moreover, aa 37-40/42, or a stable membrane integration, appear to be essential for the intra-membrane self-assembly, because A␤1-42 and A␤1-40 efficiently oligomerized within membrane, while no liposome-associated A␤1-36 oligomer could be observed. N-terminal truncations, however, did not impair membrane interactions of A␤ as expected. However, it resulted in a disparate self-assembly pattern: the intra-membrane oligomerization was somewhat with or without acidic buffer treatment were detected by immunoblotting with 6E10. The Ctrl. lane is A␤ monomer, which underwent identical treatment in the absence of liposomes. Almost all the A␤ monomers were sedimented with liposomes, and acid treatment was unable to release liposome-associated A␤. B, 0.5 M fluorescein-labeled monomeric A␤ (donor) and 0.5 M tetramethylrhodamine (TAMRA)-labeled monomeric A␤ (acceptor) were co-incubated with liposomes at the indicated peptide/lipid ratios, and FRET signal was collected and calculated as described under "Material and Methods." The rapid increase in FRET signal indicates membrane-induced efficient oligomerization of A␤. By contrast, no FRET signal could be detected when liposomes were absent or unlabeled peptide was used as acceptor. C, membrane-associated A␤ oligomers were prepared by reconstituting A␤-lipid fusion mixture (1:200 molar ratio) or incubating A␤ monomer with the preformed liposomes (1:200 molar ratio). Samples were cross-linked with bis(sulfosuccinimidyl) suberate before silver-staining SDS-PAGE. Lanes 1 and 2 are A␤ monomer controls without lipid or liposomes. Lanes 3 and 4 are pellet and supernatant fractions of reconstituted A␤-lipid fusion sample, respectively. Lanes 5 and 6 are pellet and supernatant fractions of liposome-A␤ incubation sample, respectively. It was evident that the two samples showed similar self-assembly patterns. D, exponential fitting of -t plots of A␤ monomer insertion into DPPC monolayer. The experimental data of wild-type A␤ insertion deviates from one-phase process fitting but fits well to a two-phase process, suggesting that after insertion A␤ undergoes further conformational changes. As a comparison, the kinetics of scrambled A␤ fit better to a one-phase process. E, 600 nM A␤ monomer or ADDL was injected into the subphase beneath DPPC with a constant surface pressure of 28 mN/m. After 5000 s, the monolayer and subphase fractions were separated for subsequent immunoblotting with 6E10. The control lane is the A␤ monomer or ADDL at the same concentration incubating for 5000 s. The monolayer-inserted A␤ was largely oligomeric while A␤ in the subphase remained monomeric. enhanced with A␤11-42 but completely diminished with A␤17-42. Therefore, aa 11-16 are responsible for the formation of stable intra-membrane A␤ oligomers. The failure of A␤17-42 to oligomerize within membrane indicates that the intact C-terminal alone is insufficient in mediating the formation of the stable intra-membrane oligomer. These results thus establish a C-terminal insertion initiated intra-membrane selfassembly pathway that ends up with the N-terminal stabilized oligomers.
We continued to investigate the oligomerization of the above A␤ variants in the absence of liposomes (Fig. 5C) to gain further insight into how membrane regulates A␤ self-assembly. 100 M instead of 1 M A␤ was used, because without liposomes the aggregation of A␤ is rather inefficient at a low level (16). Remarkably, A␤1-28 was most prone to aggregation as exemplified by the complete dimerization instantly after the conventional HFIP/DMSO treatment that works well to monomerize A␤1-42. Extending the C-terminal of A␤1-28 to 36 resulted in the least aggregative peptide, i.e. A␤1-36, indicating a motif within the sequence of aa 29 -36 antagonizes the N-terminal directed A␤ self-assemble. Indeed, such interaction could also explain the inability of aa 29 -36 to mediate firm membrane attachment of A␤ (please see under "Discussion"). Further extension of the C-terminal to 40 or 42 led to mild and signifi-cant reversion of the aggregation defect of A␤1-36, respectively. Hence, aa 37-40 and particularly aa 40 -42 are able to partially release the inhibition imposed by aa 29 -36. Additional experiments identified aa 1-10 as a repressor and aa 11-16 as an important motif in mediating A␤ oligomerization, because the aggregation of A␤11-42 and A␤17-42 was enhanced and markedly reduced, respectively. In contrast to the inability of A␤17-42 to form intra-membrane oligomer, this peptide retained significant self-assembly capability in the absence of liposomes, indicating that one or more motifs other than aa 11-16, most likely the C-terminal, also contribute to the solution-phase oligomerization of A␤.
Taken together, these results underscore that the self-assembly of A␤ is progressed as a result of complex interplay among multiple motifs in its own sequence. Interestingly, a common initiating event appears to be shared by the extra-and intramembrane oligomerization, i.e. overcoming the auto-inhibition conferred by the interaction between aa 29 -36 and the N-terminal portion through the action of aa 37-40/42 albeit via different mechanisms. This would release A␤ into a state permissive for solution self-assembly or stable membrane attachment. However, distinct pathways then ensue for extra-and intra-membrane oligomerization as exemplified by their differential requirement for aa 11-16. Indeed, this motif is exclu- monolayer, injection of A␤17-42 preincubated with Ca 2ϩ into the metal ion-free subphase; Zn 3 inserted A␤1-42, injection of Zn 2ϩ alone into the metal ion-free subphase beneath the monolayer with prior inserted A␤1-42. C, 1 M ADDL or A␤ monomer was incubated with DPPC liposomes for the indicated time and the liposome-bound (P1) and supernatant fractions (S1; please refer to the scheme shown in Fig. 3A for the detailed preparation protocol) were probed for A␤ signal by immunoblotting with 6E10. ADDL was incubated with DPPC liposomes for 24 h followed by 6E10 immunoblotting of liposome-bound (P2) and supernatant fractions (S2) after acidic buffer treatment (D) or EM observation (E). Prolonged incubation of ADDL with liposomes resulted in occurrence of liposome-associated A␤ oligomers, suggesting liposomes gradually shifts the equilibrium of monomers 7 ADDL interconversion, thereby leading to the accumulation and oligomerization of A␤ in liposomes. The scale bar represents 100 nm. F, the proposed model depicting the distinct interactions of A␤ monomer and soluble oligomer with membrane. In the absence of aggregation-boosting factors (e.g. zinc ion) in solution, A␤ monomers preferentially insert membrane and undergoes rapid intra-membrane oligomerization. Appreciable solution-phase oligomerization may ensue only at elevated A␤ concentrations or with the co-existence of aggregation-boosting factors (e.g. zinc) to counteract the avid membrane interaction. Soluble oligomers are unable to insert membrane; rather they may function primarily by binding to high affinity cell surface receptors.
sively required for the formation of stable intra-membrane oligomers (Fig. 5B) but seems to be somewhat dispensable in the extra-membrane self-assembly (Fig. 5C). Hence, except for facilitating the spatial proximity of A␤ molecules non-specifically, membrane insertion also enhances the self-assembly of A␤1-40/42 through regulating the pattern of motif interplay as these peptides oligomerized much more efficient within membrane than in solution.

DISCUSSION
Soluble A␤ oligomers play an essential role in the pathogenesis of AD (1, 2, 5), however, whether the membrane is among their direct targets that mediate the downstream neurotoxic effects remains elusive. Herein, we show that multiple soluble oligomeric A␤ preparations, including ADDL, PF, and zinc-induced A␤ oligomer, exhibit much weaker capability to insert membrane than that of A␤ monomer (Figs. 2-4), suggesting that solution-phase oligomerization masks the motif responsible for membrane insertion. Indeed, as part of the transmembrane segment of amyloid precursor protein, the C-terminal of A␤ not only mediates membrane insertion (31,32) but is intimately involved in extra-membrane self-assembly (33)(34)(35)(36)(37)(38). As such, it is plausible that these two processes are mutually exclusive, which is further corroborated by our findings that zincaccelerated solution-phase oligomerization markedly suppresses membrane insertion (Fig. 4). Although some reports propose that soluble A␤ oligomers may alter membrane conductivity by direct insertion (39,40), the reversibility of such effect by simple washing would, instead, argue for a weak interaction of non-insertion nature. Overall, it may be safe to conclude that stable soluble A␤ oligomers are incapable to insert membrane, although certain (transient) metastable oligomerization intermediates in solution-phase self-assembly pathway might still be competent in membrane insertion, likely due to prominent hydrophobic surface exposure as suggested by Streltsov et al. (33) and Nag et al. (41).
A␤ monomers appear to prefer membrane incorporation to solution-phase oligomerization at least at a low peptide level (Fig. 3, A and B). Moreover, following insertion A␤ monomers rapid assemble into oligomers within membrane, which are essentially identical in assembly pattern as channel-like oligomers (Fig. 3C). Hence, appreciable solution-phase oligomerization may ensue only with elevated A␤ concentrations or with the co-existence of aggregation-boosting factors (e.g. zinc) to counteract the avid membrane interaction. Although the inability of the soluble A␤ oligomer to insert or stably attach to the membrane suggests that this toxic species is unlikely to induce cell dysfunction via direct interacting with the membrane, high affinity cell surface receptors (42,43) have been identified to bind and mediate the adverse effects of soluble A␤ oligomers. Therefore, the competing intra-and extra-membrane oligomerization of A␤ may determine two distinct toxic mechanisms. The intra-membrane oligomerization of A␤ and the associated compromise of membrane integrity likely dominate in the earliest phase of AD initiation for the less pathogenic microenvironment (e.g. low A␤ production and insignificant accumulation of metal ion), whereas with the disease progression the extra-membrane self-assembly will gradually become favorable, and soluble A␤ oligomers may thus play a more prominent role via signaling through cell receptors. Moreover, despite being a vulnerable target membrane may also help to alleviate the acute attack of soluble A␤ oligomers by converting them to the less toxic intra-membrane species (Figs.  1D, 3A, 4C, and 4F).
By comparing the extra-and intra-membrane oligomerization behavior of A␤ variants, we are able to gain further insight into the complex interplay among multiple sequence motifs within A␤ underlying the conformational changes associated with its self-assembly (Fig. 5). Remarkably, among the examined peptides, A␤1-36 emerges as the unique one showing interesting properties. First, A␤1-36 is very inefficient in selfassembly, despite the fact that A␤1-28 (44) and aa 25-35 (45) are both highly aggregative in solution. Second, although it  Fig. 3A for the detailed preparation protocol) were probed by immunoblotting with 4G8. Almost no liposome association of A␤1-28 and A␤1-36 could be detected, whereas A␤1-42, A␤1-40, and A␤11-42 showed a high level of membrane association and intra-membrane oligomerization. Despite strong membrane association, no intra-membrane oligomer of A␤17-42 was observed. C, 100 M A␤ variants were incubated in TBS at room temperature for the indicated times and analyzed by silver-staining SDS-PAGE (the left panel) with or without cross-linking, or by immunoblotting with 4G8 (the right panel) without prior cross-linking by bis-(sulfosuccinimidyl) suberate (BS 3 ). The aggregation patterns in silver-staining SDS-PAGE with or without cross-linking are very similar. At such a high concentration, all the A␤ variants aggregated albeit with different kinetics and capacity. A␤1-28 is the most aggregative peptide showing no monomeric band in SDS-PAGE, whereas A␤1-36 is the least aggregative peptide, the majority of which remained monomeric after a 3-day incubation. evokes strong pressure increase in monolayer assays, A␤1-36 fails to stably associate with membrane, despite that the amino acid 30 -36 sequence is the most hydrophobic stretch in the sequence of A␤ and aa [25][26][27][28][29][30][31][32][33][34][35] have been reported to insert and form ion channels within the membrane (46,47). These results thus suggest an antagonistic interaction between aa 29 -36 and aa 1-28, the extracellular domain of A␤. Consistent with this notion, aa 30 -36 comprise the most represented segment that constitutes the major part of the C-terminal strand and participates in the intra-molecular ␤-hairpin formation by pairing the central hydrophobic cluster (aa [17][18][19][20][21] within the N-terminal strand, as revealed by extensive NMR characterization and x-ray crystallography (33)(34)(35)(36)(37)(38). Accordingly, such interaction may disable the respective functionalities dictated by the two respective sequences of aa 29 -36 and 1-28 and, hence, lock A␤1-36 in an auto-inhibited configuration.
Extension of the C terminus of A␤1-36 to 40/42 moderately/ markedly rescued the defect in solution-phase oligomerization and stable membrane insertion. Because the strand pairing between aa 30 -36 and the central hydrophobic cluster is largely retained as the basic folding unit in the structure models of A␤ resolved under different extra-membrane assembly states (33)(34)(35)(36)(37)(38), we propose that, instead of breaking the antagonistic interaction between these two segments, aa 37-40/42 may overcome the self-assembly defect via one of the following mechanisms: (i) it acts in concert with aa 30 -36 to form a longer ␤-hairpin that is more compatible for efficient aggregation as illustrated in the fibril structures (34,35); or (ii) it provides additional docking site or surface for stable inter-molecular interaction as revealed in the structures of soluble oligomers (33,36,38). By contrast, the pairing between aa 30 -36 and the central hydrophobic cluster in A␤1-40/42 is likely to be broken during membrane insertion to allow stable incorporation via the transmembrane portion. Indeed, the intriguing metastable or reversible membrane insertion of A␤1-36 suggests that the membrane can facilitate the reconfiguration of this peptide (or the hairpin) to expose the membrane-spanning segment (i.e. aa 29 -36). However, due to the shortened C terminus, there is probably only a small free energy difference between the autoinhibited solution state and the membrane-inserted state of A␤1-36, allowing an interconversion between these two states.
In monolayer experiments, the excessive amount of peptide and the interconversion equilibrium together allow the existence of a sufficient amount of membrane-inserted peptide in the steady state of insertion. By contrast, in bilayer experiments, during the separation of liposomes from the bulk solution by ultracentrifugation, the decrease in peptide-liposome contact favors the conversion from inserted to auto-inhibited state and, hence, results in depletion of A␤1-36 from liposomes. As such, the increased hydrophobicity of the longer C terminus of A␤1-40/42 will further stabilize the inserted conformation leading to more stable membrane incorporation.
Our findings also reveal an important role of the N-terminal hydrophilic portion, i.e. aa 1-17, in regulating the self-assembly of A␤. Deletion of aa 1-10 enhanced both intra-and extramembrane oligomerization albeit to different extents, indicating A␤11-42, the amyloidgenic product of the alternative ␤-secretase pathway, is more prone to oligomerize than A␤1-42 consistent with its proposed role in the early phase of AD (48). The ␣-secretase pathway product A␤17-42, on the contrary, is considered as a non-amyloidgenic isoform. Accordingly, we found this peptide was unable to form stable intramembrane oligomers and oligomerized in solution with a markedly reduced efficiency. Although the underlying mechanism awaits further exploration, these results pinpoint aa 11-16 as the critical motif in promoting A␤ self-assembly. Paradoxically, A␤17-42 has recently been reported to form ion channels in lipid bilayer (49). This finding, however, does not necessarily contrast to our results because: (i) the discrepancy may be due to different experimental conditions, such as peptide/lipid ratio (ϳ1/15 versus 1/200) and peptide level (10 -40 M versus 1 M); (ii) the dynamic assembly and disassembly of the channel observed by Jang et al. (49) are indicative of a weak inter-molecular interaction, which is in agreement with the inability of A␤17-42 to form the stable intra-membrane oligomer identified herein. Together, our current work emphasizes a critical regulation of membrane on the behavior of A␤ monomer and soluble oligomers, which may have implications in understanding the role of A␤ in AD pathology.