Targeted Control of Kinetics of β-Amyloid Self-association by Surface Tension-modifying Peptides*

Brain tissue from Alzheimer's patients contains extracellular senile plaques composed primarily of deposits of fibrillar aggregates of β-amyloid peptide. β-Amyloid aggregation is postulated to be a major factor in the onset of this neurodegenerative disease. Recently proposed is the hypothesis that oligomeric intermediates, rather than fully formed insoluble fibrils, are cytotoxic. Previously, we reported the discovery of peptides that accelerate β-amyloid aggregation yet inhibit toxicity in vitro, in support of this hypothesis. These peptides contain two domains: a recognition element designed to bind to β-amyloid and a disrupting element that alters β-amyloid aggregation kinetics. Here we show that the aggregation rate-enhancing activity of the disrupting element correlates strongly with its ability to increase surface tension of aqueous solutions. Using the Hofmeister series as a guide, we designed a novel peptide with terminal side-chain trimethylammonium groups in the disrupting domain. The derivatized peptide greatly increased solvent surface tension and accelerated β-amyloid aggregation kinetics by severalfold. Equivalent increases in surface tension in the absence of a recognition domain had no effect on β-amyloid aggregation. These results suggest a novel strategy for targeting localized changes in interfacial energy to specific proteins, as a way to selectively alter protein folding, stability, and aggregation.

␤-Amyloid (A␤) 1 is a 40 -42-amino-acid fragment cleaved from membrane-bound amyloid precursor protein, containing sequences from both extracellular and transmembrane regions of the parent protein. Postmortem analysis of Alzheimer's disease brains reveals the presence of extracellular senile plaques composed primarily of deposits of A␤ fibrillar aggregates. The "amyloid hypothesis," that A␤ amyloid deposition is a causative factor in the onset of Alzheimer's disease, is supported by biochemical, genetic, and animal studies (1).
A␤ self-association proceeds from the random coil monomer, through ␤-sheet structure formation and oligomerization, fila-ment (or protofibril) initiation and growth, and then fibril assembly, growth, and deposition (2). The hypothesis, that A␤ is toxic only when aggregated into fibrils, is supported by a substantial body of data (3)(4)(5). Recently, an alternative hypothesis has been put forth: specifically, that a soluble intermediate in the fibrillogenesis pathway, rather than the fully formed fibrillar end product, is the most cytotoxic form of A␤ (2, 6 -10). The issue of the conformation and aggregation status of the toxic A␤ specie(s) remains controversial.
Several groups have reported the synthesis of compounds that interfere with A␤ aggregation and inhibit toxicity (11)(12)(13)(14)(15)(16)(17)(18)(19)(20). Our group chose a strategy employing hybrid peptides as inhibitors; these peptides contain a recognition domain, designed to bind specifically to A␤, and a disrupting domain, designed to interfere with normal A␤ aggregation (21). As the recognition domain, we chose residues 16 -20 (KLVFF) of full-length A␤; this region was identified as critical for A␤ self-association (22,23). Hybrid peptides with the strongest affinity for binding to A␤ were the most effective at protecting against A␤ toxicity (24). We identified hybrid peptides that, when mixed with A␤, inhibit A␤ toxicity while promoting more rapid formation of larger A␤ aggregates (21,25,26). If it proves to be true that intermediate oligomeric species in the A␤ aggregation pathway are the toxic species, the cytoprotection afforded by these compounds might result from their ability to reduce the concentration of toxic intermediate species.
The objective of the work reported here is to identify a plausible physicochemical basis for the action of hybrid peptides in accelerating A␤ aggregation. We demonstrate a strong positive correlation between the surface tension of aqueous solutions of active compounds and the ability of these compounds to increase the rate of A␤ aggregation. This concept is used to rationally design a modified peptide with markedly enhanced activity.

EXPERIMENTAL PROCEDURES
Peptides-A␤-(1-40) was purchased from Anaspec, Inc. (San Jose, CA). Protected amino acids, resin, and HBTU were purchased from Novabiochem. Betaine, ethyl acetate, and piperidine were purchased from Sigma. Diisopropylethylamine was purchased from Advanced Chemtech (Louisville, KY). KKKK was purchased from Bachem, Inc. (King of Prussia, PA). All other materials were purchased from Fisher Scientific or Sigma. KLVFFKKKKKK was synthesized by solid-phase peptide synthesis using Fmoc-protected amino acids and purified by HPLC as described previously (26). The dicyclohexamine salt of double Boc-protected lysine was converted to the free acid form using sulfuric acid and ethyl acetate extraction. To make the betaine-modified peptide, LVFFKKKKKK was synthesized on a Wang resin but with Mtt protection on the 6 C-terminal lysines. While still on the resin, the free acid form of double Boc-protected lysine was coupled at the N terminus in the presence of HBTU and diisopropylethylamine activators. Mttprotecting groups were cleaved using 1% trifluoroacetic acid followed by addition of piperidine. Betaine was coupled to the free amines to form an amide linkage in the presence of HBTU and diisopropylethylamine. 95% trifluoroacetic acid was added to cleave the resulting peptide from the resin and remove the Boc-protecting groups. The cleaved peptides were purified by reverse-phase HPLC (C4 column) using an acetonitrile/water gradient. Fractions were collected and analyzed by MALDI mass spectrometry. Purified peptides were stored as lyophilized powders at Ϫ70°C.
Surface Tension-The equilibrium surface tension of peptide solutions was measured using an FTÅ200 pendant drop tensiometer (First Ten Angstroms, Portsmouth, VA). A droplet of inhibitor solution was formed at the end of a blunt, 22-gauge stainless steel needle, and the shape of the droplet was imaged. The surface tension was measured by fitting the Young-Laplace equation to the contour of drop shape once equilibrium was reached.
Light Scattering-Peptide or betaine solutions were prepared by dissolving the compounds in double-filtered (0.22 m) PBSA. A␤ was dissolved in double-filtered (0.22 m) 8 M urea at a concentration of 2.8 mM for 10 min and then diluted into double-filtered PBSA or PBSA containing test compound, to 140 M A␤. All samples were at pH 7.4 and contained 0.4 M urea. Samples were quickly filtered through 0.45-m filters directly into clean light scattering cuvettes. Dynamic light scattering data as well as average scattered intensity at 90°s cattering angle were collected using a Coherent argon ion laser at 488 nm and a Malvern 4700 system, as described in more detail elsewhere (26).

RESULTS
Previously, we proposed a strategy for generating hybrid peptide compounds that modulate A␤ aggregation and inhibit A␤ toxicity (21). Effective hybrid peptides contain an N-terminal recognition domain, KLVFF, homologous to residues 16 -20 of A␤, and a C-terminal disrupting domain, a repeat sequence of non-homologous amino acids. We observed that KLVFFKKKKKK and KLVFFEEEE, but not KLVFF or KLVFFSSSS, protected cells from A␤ toxicity (25,26). Interestingly, protection from toxicity was accompanied invariably with an increase in the rate of aggregation of A␤.
Given that both cationic and anionic, but not polar uncharged, disrupting domains were capable of accelerating A␤ aggregation, we hypothesized that the disrupting domain acted by altering physical properties of the solvent. To test this, the surface tension of aqueous solutions of hybrid peptides (without A␤) was measured using a pendant drop method. Compounds with charged disrupting domains increased surface tension in a concentration-dependent manner, but compounds with polar uncharged disrupting domains, or the recognition element alone with no disrupting domain, had little effect on surface tension (Table I). Increased surface tension of aqueous solutions of the hybrid peptide correlated strongly with an increased rate of aggregation of mixtures of A␤ ϩ hybrid peptide (Fig. 1).
We next tested whether the disrupting domain alone was sufficient to accelerate A␤ aggregation. KKKK at 140 and 280 M increased solvent surface tension to 53.9 Ϯ 0.8 and 57 Ϯ 1 dyne/cm, respectively, similar to that of KLVFFKKKKKK at  Table I. the same concentrations (Table I), but even 420 M KKKK had no effect on A␤ aggregation, whereas 140 M KLVFFKKKKKK greatly accelerated A␤ aggregation (Fig. 2). These results indicate that modest increases in surface tension alone are insufficient to cause acceleration of aggregation and provide further support for the notion that the A␤ binding ability of the KLVFF recognition domain is required for activity. We reasoned that even greater acceleration of A␤ aggregation could be obtained with a KLVFF recognition element coupled to a disrupting domain that produced a greater solvent surface tension effect. To identify an appropriate candidate, we turned to the Hofmeister series, which lists ions in the order of their ability to stabilize protein folded structure (27). Protein structure stabilization by co-solutes correlates strongly with the ability of the co-solute to increase the surface tension of water (28). The Hofmeister cation series is N(CH 3 ) 4 ϩ Ͼ NH 2 (CH 3 ) 2 ϩ Ͼ NH 4 ϩ Ͼ K ϩ Ͼ Na ϩ Ͼ Cs ϩ Ͼ Li ϩ Ͼ Mg 2ϩ Ͼ Ca 2ϩ Ͼ Ba 2ϩ , with the cations on the left classified as kosmotropes (protein structure-stabilizing, or "salting-out") and those on the right as chaotropes (protein structure-destabilizing, or "salting-in") (27).
The lysine hexamer disrupting domain of our most effective hybrid peptide reported to date, KLVFFKKKKKK, contains terminal amines that are protonated at neutral pH (-(CH 2 ) 4 -NH 3 ϩ ). We reasoned by analogy to the Hofmeister series that a side chain with a methyl-substituted terminal amine group might have enhanced activity as compared with lysine. Be-taine ((CH 3 ) 3 N ϩ CH 2 COO Ϫ ) is a naturally occurring compound that contains the requisite methyl-substituted amino group and also contains a free carboxyl group allowing facile coupling to a lysine side chain. We modified the lysine side chains in the disrupting domain using the following strategy. KLVFF-KKKKKK was synthesized using standard Fmoc solid-phase synthesis techniques but with Mtt protection of the 6 C-terminal lysines and Boc protection of the N-terminal lysine. While maintaining the peptide on the resin, the Mtt groups were cleaved, and betaine was coupled to the lysine side chains via an amide linkage (Fig. 3). Cleavage from the resin produced a mixture of KLVFFKKKKKK derivatized with four, five, or six betaines, as confirmed by mass spectroscopy analysis (Fig. 3). A fraction highly enriched in the fully derivatized peptide, compound 1, was isolated by reverse-phase HPLC for further study (Fig. 3).
Compound 1 was extremely water-soluble. Size-exclusion chromatographic analysis confirmed that it was monomeric in PBSA (data not shown). Pendant drop measurements of aqueous solutions of the purified compound demonstrated greatly increased surface tension as compared with KLVFFKKKKKK (Table II).
The effect of compound 1 on A␤ aggregation was remarkable (Fig. 4). After 4 h of aggregation, the hydrodynamic diameter of A␤ in the presence of compound 1 was more than 100-fold greater than A␤ alone(ϳ2300 versus ϳ20 nm) and nearly 50 times greater than A␤ with KLVFFKKKKKK, which was our most active compound prior to the discovery of compound 1. Betaine at 1700 M increased the surface tension of an aqueous solution to 61.4 Ϯ 0.9 dyne/cm, equivalent to 140 M compound 1, but betaine at 3400 M had no measurable effect on A␤ aggregation kinetics (Fig. 4). These results show that surface tension can be used as a design strategy for improving activity of the disrupting domain of hybrid peptides and confirm that a specific recognition domain is required for activity. DISCUSSION Self-association of protein into large aggregates of ␤-sheet structure and fibrillar morphology is a feature of a number of diseases, including the primary amyloidoses and neurodegenerative diseases such as Alzheimer's. Many, perhaps even most, proteins and peptides can be coaxed into forming ␤-sheet fibrils under appropriate conditions of solvent, pH, and tem-perature (29), Alarmingly, fibril-forming peptides and proteins appear to be toxic to a wide number of cell types (29). The end product of aggregation, the fully formed amyloid fibril, may be the primary toxic species, but recent evidence suggests that instead, it is a structured kinetic intermediate that is killing cells (2)(3)(4)(5)(6)(7)(8)(9)(10). Indeed, our earliest studies provided the first published data, to our knowledge, linking accelerated aggregation with inhibition of toxicity (21).
Our observations that hybrid peptides with either anionic or cationic disrupting domains accelerate A␤ aggregation (26) led to the hypothesis that disrupting domains act by affecting solvent properties. In particular, an increase in surface tension of aqueous solutions by addition of co-solutes is strongly linked to changes in protein stability and protein aggregation (28). Indeed, we observed that those hybrid peptides that accelerated A␤ aggregation also measurably increased the surface  tension of aqueous solutions (Fig. 1). The increase in surface tension was mediated solely through the disrupting domain (Table I). We tested whether the disrupting domain alone was capable of affecting solvent properties and/or A␤ aggregation. Although a lysine tetramer at 420 M increased the surface tension, it failed to alter A␤ aggregation kinetics. This is not surprising, given that molar concentrations (Ͼ0.1-1 M or higher) of co-solute are generally required for a sufficient change in solvent properties to produce measurable changes in protein folding and aggregation (27,30,31).
These encouraging results served as a basis for rational design of novel compounds with greater efficacy. We used the Hofmeister series as a guide toward selecting functional groups with strong surface tension activity. The Hofmeister series has proven to be a reasonably reliable predictor of co-solute effects on protein structure, aggregation, and activity (e.g. Ref. 32). The strongest salting-out (kosmotropic) cations in the Hofmeister series are methylammonium ions (27). Several related compounds, such as betaine ((CH 3 ) 3 N ϩ CH 2 COO Ϫ ), sarcosine (CH 3 )H 2 N ϩ CH 2 COO Ϫ ), and trimethylamine N-oxide ((CH 3 ) 3 NO), are naturally occurring intracellular solutes that regulate osmotic pressure and modulate protein folding and enzyme function (e.g. Ref. [33][34][35]. These compounds also affect protein aggregation. For example, sarcosine stabilized the native conformation of the serpin ␣ 1 -antitrypsin and protected against thermal inactivation and aggregation (30). Similarly, betaine partially inhibited light chain amyloid fibril formation from immunoglobulin light chain (29). In contrast, trimethylamine N-oxide accelerated fibril assembly from A␤ (36). This apparent contradiction can be resolved by noting that kosmotropes, through a preferential exclusion mechanism, increase the surface tension (interfacial energy per unit area) of aqueous solutions, thereby producing a thermodynamic driving force to reduce surface area to minimize interfacial energy. Thus, kosmotropes drive the system toward more compact protein structures. For ␣ 1 -antitrypsin and immunoglobulin light chain, the most compact structure is the natively folded monomer, but since A␤ monomer is random coil, its most compact and folded structure is the ␤-sheet fibril.
Given these clues, we developed a method for synthesizing hybrid peptides with terminal trimethylammonium groups in the disrupting domain. The betaine-derivatized KLVFFK-KKKKK was remarkably active at increasing surface tension and was dramatically more effective at accelerating A␤ aggregation than KLVFFKKKKKK. Although betaine alone increased surface tension of aqueous solutions, the osmolyte did not measurably affect A␤ aggregation. These results indicate the specificity of the action of compound 1.
It is well established that co-solutes that increase solvent surface tension favor compact folded and/or aggregated protein structures. What is unique about the compounds described here is that the increase in surface tension is apparently localized via the recognition element to the solvent near the aggregating peptides. We hypothesize that this leads to a much greater localized change in solvent surface tension than that experienced in the bulk solvent or in the absence of the recognition element. This localized change in solvent properties then drives further growth of highly aggregated species. It is interesting to speculate that similar strategies could be used as novel mechanism of targeting interfacial changes to individual proteins to influence protein folding, stability, and aggregation in a highly specific manner.