Key aromatic/hydrophobic amino acids controlling a cross-amyloid peptide interaction versus amyloid self-assembly

The interaction of the intrinsically disordered polypeptide islet amyloid polypeptide (IAPP), which is associated with type 2 diabetes (T2D), with the Alzheimer's disease amyloid-β (Aβ) peptide modulates their self-assembly into amyloid fibrils and may link the pathogeneses of these two cell-degenerative diseases. However, the molecular determinants of this interaction remain elusive. Using a systematic alanine scan approach, fluorescence spectroscopy, and other biophysical methods, including heterocomplex pulldown assays, far-UV CD spectroscopy, the thioflavin T binding assay, transmission EM, and molecular dynamics simulations, here we identified single aromatic/hydrophobic residues within the amyloid core IAPP region as hot spots or key residues of its cross-interaction with Aβ40(42) peptide. Importantly, we also find that none of these residues in isolation plays a key role in IAPP self-assembly, whereas simultaneous substitution of four aromatic/hydrophobic residues with Ala dramatically impairs both IAPP self-assembly and hetero-assembly with Aβ40(42). Furthermore, our experiments yielded several novel IAPP analogs, whose sequences are highly similar to that of IAPP but have distinct amyloid self- or cross-interaction potentials. The identified similarities and major differences controlling IAPP cross-peptide interaction with Aβ40(42) versus its amyloid self-assembly offer a molecular basis for understanding the underlying mechanisms. We propose that these insights will aid in designing intervention strategies and novel IAPP analogs for the management of type 2 diabetes, Alzheimer's disease, or other diseases related to IAPP dysfunction or cross-amyloid interactions.

and T2D (1). Increasing evidence suggests that cross-interactions between different amyloidogenic proteins or polypeptides, for which we here use the term "cross-amyloid interactions," modulate their self-assembly into amyloid fibrils and thus may link different diseases to each other (1)(2)(3)(4). However, underlying mechanisms have not been yet understood.
In the case of AD and T2D, epidemiological and pathophysiological evidence suggests that the two diseases are linked to each other (3,5,6). A possible molecular link could be the interaction between their key amyloidogenic polypeptides A␤ (AD) and IAPP (T2D) (3,7,8). In fact, a nanomolar affinity interaction between early prefibrillar A␤40 (42) and IAPP species has been shown in vitro to suppress amyloidogenesis, whereas seed amounts of A␤40 (42) fibrils are able to cross-seed IAPP amyloidogenesis in vitro and in animal models in vivo (3,(7)(8)(9). Importantly, the 37-residue-long IAPP, which in soluble form functions as a glucose regulatory neuropeptide and, similar to A␤40 (42), is present both in blood and CSF, has recently been found to co-localize with A␤ plaques in the brains of AD patients, suggesting a possible (patho-)physiological role for the cross-interaction between the two polypeptides (3,5,10,11).
Due to the highly dynamic nature and strong self-and co-aggregation propensities of IAPP and A␤40 (42), characterization of their hetero-assemblies at an atomic level via NMR or X-ray crystallography has not been yet possible. Uncovering the molecular determinants of their cross-interaction is, however, of high importance both for elucidating its potential role in disease pathogenesis and for designing possible intervention strategies and novel therapeutic IAPP analogs. In fact, the IAPP-A␤40 (42) interaction has already been successfully applied to design bioactive IAPP analogs or IAPP-derived peptides as inhibitors of amyloidogenesis of IAPP and/or A␤40(42) (5,8,16,17). In addition, pramlintide, a soluble and bioactive IAPP analog, which has been used in T2D treatment for several years, has recently been suggested to be a promising candidate for the treatment of AD and certain cancers as well (5,18,19).
By using a systematic alanine scan approach, fluorescence and far-UV CD spectroscopy, heterocomplex pulldown assays combined with NuPAGE gel electrophoresis and Western blotting, sedimentation assays, the thioflavin T (ThT) binding assay, transmission electron microscopy (TEM), and molecular dynamics (MD) simulations, we here uncover key molecular determinants of the IAPP cross-interaction with A␤40 (42) and characterize their role in IAPP solubility, conformation, and amyloid self-assembly. Our results provide a molecular basis for the interaction of IAPP with A␤40(42) versus its amyloidogenic self-assembly and for the design of intervention strategies and IAPP analogs with tailored self-/cross-interactions.

Interactions of IAPP "hot segments" with IAPP and A␤40
First, we asked which are the key residues of the interaction of the partial IAPP "hot segments" IAPP (8 -18) and IAPP (22)(23)(24)(25)(26)(27)(28) with IAPP or A␤40. We addressed this question by using alanine scanning and a fluorescence spectroscopic titration assay developed by us earlier (8,12,20,21). This assay is based on the principle that changes in fluorescence of fluorescently labeled proteins or polypeptides are very sensitive indicators of their interactions and can be used to determine their binding affinities (22).

Interactions of 8-Ala and 4-Ala-containing IAPP mutants with IAPP and A␤40(42)
To investigate this hypothesis, we synthesized and studied 17 mutants of full-length IAPP comprising one mutant with substitutions of all eight of the above identified residues by Ala, two quadruple-point mutants, four triple-point mutants, six double-point mutants, and four single-point Ala mutants (Fig. 1). The affinities of their interactions with prefibrillar IAPP, A␤40, and A␤42 species were then quantified by fluorescence spec-troscopic titrations (Fig. 2, supplemental Figs. S1 and S2, and Tables 2 and 3). Synthetic N ␣ -terminal fluorescein-labeled IAPP (Fluos-IAPP) or its mutants (5 nM) were titrated with IAPP, A␤40, and A␤42 (pH 7.4), and fluorescence spectra were recorded (8,12,20). Of note, nanomolar affinities (K d(app) values) have been determined previously for interactions of Fluos-IAPP with IAPP and A␤40(42) by using this methodology and were confirmed in the context of the current studies (Table 2, Fig. 2, and supplemental Figs. S1 and S2) (8,12,20).
Dramatically decreased binding affinities to both A␤40(42) and IAPP were found for 8A and 4A as compared with IAPP ( Fig. 2, supplemental Figs. S1 and S2, and Tables 2 and 3). For   (7). Previously identified "hot segments" of IAPP self-assembly and its hetero-assembly with A␤ are in boldface type (12), and introduced Ala substituents are shown in red.

Interactions of single-point mutants of IAPP with IAPP and A␤40(42)
To directly address the above hypothesis, interactions of single mutants A15, A16, A23, and A26 with A␤40(42) and IAPP were studied next (Fig. 2, supplemental Figs. S1 and S2, and Tables 2 and 3). All four mutants bound IAPP with only slightly weaker affinities than the affinity of IAPP self-assembly, suggesting that none of the four residues Phe 15 , Leu 16 , Phe 23 , and Ile 26 in isolation plays a key role in IAPP self-assembly. By contrast, strongly reduced affinities were found for the interactions of A23 and A26 with A␤40; the affinity of A23 was Ͼ200-fold (⌬⌬G Ͼ 3 kcal/mol) and that of A26 44-fold decreased (⌬⌬G ϭ 2.2 kcal/mol) as compared with the affinity of IAPP (Fig. 3).
However, substitution of Phe 15 or Leu 16 by Ala did not strongly affect IAPP affinity to A␤40 (⌬⌬G Ͻ 1 kcal/mol) (Fig. 3). These findings suggested that IAPP residues Phe 23 and Ile 26 (⌬⌬G Ͼ 2.0 kcal/mol) are hot spots of the IAPP interaction surface with A␤40 (Fig. 3). Importantly, the binding affinities of the above four single Ala mutants for A␤42 were very similar to the affinity of IAPP (supplemental Fig. S1 and Tables 2 and 3). Thus, none of the above mutations in isolation was able to affect the interaction of IAPP with A␤42.

Heterocomplex pulldown assays
To obtain more support for the above findings, mixtures of IAPP, 4A, and A23 with A␤40 or A␤42 were subjected to heterocomplex pulldown assays combined with NuPAGE gel electrophoresis and Western blotting (  (8,12). By contrast, no A␤40 was found in lanes corresponding to A␤40 bound to biotin-4A or biotin-A23 (Fig. 4), whereas the amounts of A␤42 bound to biotin-4A were strongly reduced as compared with A␤42 bound to biotin-IAPP (supplemental Fig. S3). These findings were thus in very good agreement with the results of the fluorescence titration studies.

Interactions of double/triple mutants of IAPP with IAPP and A␤40(42)
To identify possible direct or indirect interactions/coupling between the identified key residues and their effects on the stabilities of IAPP homo-assemblies or their hetero-assemblies with A␤40(42), interactions of double and triple mutants were studied next (Figs. 2 and 3, supplemental Figs. S1 and S2, and Tables 2 and 3). All hetero-assemblies between IAPP and double/triple mutants exhibited only weakly reduced stabilities (Ͻ10-fold) as compared with IAPP-IAPP assemblies (Tables 2  and 3). Also, most of the double/triple mutants were found to bind A␤40 with only weakly reduced (Ͻ10-fold) affinity as Figure 3. Identification of hot spots and key residues of IAPP self-assembly and its hetero-assembly with A␤40(42) via fluorescence spectroscopic titrations (A-C). Differences in binding free energy of Ala mutants (mut) and wild-type (wt) IAPP (⌬⌬G mut Ϫ wt or ⌬⌬G) toward wild-type IAPP (A), A␤40 (B), or A␤42 (C) (means Ϯ S.D. (error bars); n ϭ 3 assays). Mutants in boldface type are those with strongly diminished binding affinity to IAPP, A␤40, or A␤42 as compared with IAPP.

Key determinants of a cross-amyloid peptide interaction
compared with IAPP with two exceptions: A15,23, which exhibited a Ͼ200-fold weaker affinity, and A23,26, which exhibited a 19-fold weaker affinity for A␤40 as compared with IAPP (Tables 2 and 3).
The dramatically impaired binding of A15,23 to A␤40 was consistent with our finding that Phe 23 is a hot spot of the IAPP interaction surface with A␤40. On the other hand, the finding that binding affinities of several triple/double point mutants to A␤40 were only Ͻ10or 20-fold lower than the affinity of IAPP suggested that the presence of two or three Ala-substituted hot spots/key residues compensates for the detrimental effects of mutating a hot spot to Ala. Cooperative, direct/indirect interactions between key residues and/or structural perturbations caused by the mutations probably account for these findings (23). Importantly, in the case of the hetero-assemblies of A␤42 with double/triple Ala mutants, a strong destabilization (⌬⌬G Ͼ 2) was found only for the A15,23-A␤42 hetero-assemblies ( Fig. 3, supplemental Fig. S1, and Tables 2 and 3). These latter results suggested that the presence of either Phe 15 or Phe 23 is required for the interaction of IAPP with A␤42.

Effects of IAPP-GI mutants on amyloidogenesis of A␤40(42) and IAPP
We have reported previously that IAPP-GI, a double N-methylated analog of full-length IAPP, is a potent inhibitor of amyloidogenesis of A␤40, A␤42, and IAPP (8,20). Three mutants of IAPP-GI were synthesized, and their effects on the kinetics of amyloidogenesis of A␤40, A␤42, and IAPP were studied by using the ThT binding assay (supplemental Fig. S4). Mutants comprised 4A-GI, an IAPP-GI analog with all four key residues substituted by Ala, the IAPP-GI analog A23-GI with Phe 23 substituted by Ala, and the IAPP analog A15,23-GI, containing substitutions of both Phe 15 and Phe 23 by Ala. In strong contrast to IAPP-GI, the Ala-substituted IAPP-GI analogs were unable to block A␤40(42) and/or IAPP amyloidogenesis (supplemental Fig. S4). These results supported the suggestion that the four residues Phe 15 , Leu 16 , Phe 23 , and Ile 26 play an important role in IAPP-A␤40(42) and IAPP-IAPP interactions.

Self-assembly propensities of IAPP mutants
The self-assembly propensities of the mutants were studied next (Table 4 and data not shown). Fluorescence spectroscopic titrations showed that 8A and 4A exhibited a Ͼ1000-fold weaker self-assembly propensity than IAPP. By contrast, the K d(app) values of self-assembly of most other mutants were very similar to the K d(app) of IAPP self-assembly and to the K d (app) values of their interactions with IAPP (Tables 3 and 4) (data not shown). Only in the case of the self-assembly of A15,16 and of A23,26 were markedly reduced (10 -15-fold weaker) affinities found (Table 4). This latter result was consistent with an important role for homotypic interactions in early steps of IAPP selfassembly, as suggested for IAPP fibrils as well (Table 4) (14,15). Taken together, the studies on the self-assembly potency of the mutants provided additional evidence for a crucial, concerted role of residues Phe 15 , Leu 16 , Phe 23 , and Ile 26 in IAPP self-assembly.

Solubility of IAPP mutants
We next addressed the question whether mutations affected IAPP solubility by using a sedimentation assay (Fig. 5A). The focus was on mutants that exhibited impaired bindings to A␤40(42) and/or IAPP because such mutants could become leads for soluble IAPP analogs with tailored self-/cross-interaction properties. Peptides were incubated at 100 M in aqueous buffer (pH 7.4) for 7 days, and, following centrifugation, peptide amounts in supernatants and pellets were quantified (Fig. 5A). Of note, IAPP has previously been found to precipitate within a few days already at 1 M by using the same methodology (8,20). A dramatic increase of solubility (Ͼ100-fold) as compared with IAPP was found for four of the 17 mutants (i.e. 8A, 4A, A15,23,26, and A23,26) ( Fig. 5A and Table 4). Additional sedimentation assays at 50 or 20 M revealed that A15,23 was Ͼ50fold more soluble than IAPP and that the solubilities of the four single mutants were similar to the solubility of IAPP (Table 4 and supplemental Fig. S5). Thus, four of the six mutants identified here exhibiting impaired binding to IAPP or A␤40(42) (i.e. 8A, 4A, A23,26, and A15,23) had strongly improved solubilities compared with IAPP.

Secondary structure of IAPP mutants
Effects of Ala mutations on the conformation of IAPP were then investigated by far-UV CD spectroscopy ( Fig. 5 (B-E) and supplemental Fig. S6). The spectra of freshly dissolved IAPP

Key determinants of a cross-amyloid peptide interaction
and the mutants (10 M; pH 7.4) were of similar shapes and indicative of mixtures of random coil and ␤-sheet/␤-turn structural elements (Fig. 5, B-E). Also, no strong differences were observed between the magnitudes of the CD spectra of IAPP and most of the mutants (Fig. 5, B-E) (24). Only in the case of Ala(9 -12) and A16 were the magnitudes of the CD spectra strongly reduced compared with IAPP (Fig. 5, B and E). As expected for peptides with high oligomerization propensity, a clear concentration dependence was observed in the CD spectra even at low micromolar concentrations (data not shown) (24). However, oligomerization did not result in precipitation or strong changes of secondary structure contents of IAPP and most mutants over at least a 10-fold concentration range (5-50 M) (supplemental Fig. S6). Determination of secondary structure contents via deconvolutions of CD spectra obtained between 5 and 50 M suggested 30 -50% ␤-turn/␤-sheet, 40 -60% random coil and 5-15% ␣-helical contents for most mutants (supplemental Fig. S6) (25)(26)(27). For IAPP, 30 -40% ␤-turn/␤-sheet, 50 -60% random coil, and 10% ␣-helix were determined, consistent with earlier reported values (supplemental Fig. S6) (24). Together, the CD studies provided evidence that (a) mutations did not cause large structural changes in IAPP and (b) ␤-strand-loop-␤-strand conformers are major populations of folded species in IAPP and the mutants. The strongly impaired binding of A23, A26, A23,26, and A15,23 to A␤40(42) could thus be due to the substitution of IAPP residues making direct contacts with A␤40(42) by Ala, whereas the restored high-affinity binding of many double and all triple Ala mutants could be due to local structural changes related to the introduced substitutions. Such changes could be mediated for instance via local fold (de)stabilization and/or registry shifts in ␤-strands, yielding alternate A␤40(42) binding surfaces. The high conformational flexibility of both A␤40 (42) and IAPP strongly supports this suggestion.

Amyloidogenicity of IAPP mutants
Next, amyloidogenicities of IAPP and the 17 Ala mutants were studied by using the ThT binding assay in combination with TEM (Fig. 6). First, kinetics of amyloid formation were studied by incubating the peptides at 12 M (pH 7.4) for 7 days (Fig. 6, A-D). Of note, IAPP has previously been found to be amyloidogenic at a concentration 19-fold lower than the above one (625 nM) (8). Seven of the 17 mutants comprising the two triple mutants A15,16,23 and A15,16,26; the three double mutants A15,26, A16,23, and A16,26; and the two single mutants A15 and A26 were found to aggregate into ThT binding assemblies (Fig. 6, A-D). The presence of significant amounts of amyloid fibrils in their incubations was confirmed by TEM (Fig. 6, A-D). The above seven mutants were thus Ͻ20-fold less amyloidogenic than IAPP, which is consistent with the corresponding mutations having only weakly affected IAPP amyloidogenicity.
To identify mutations that strongly affected IAPP amyloidogenicity, the 10 mutants that did not form fibrils at 12 M were studied with regard to their amyloidogenicity at 62.5 M, corresponding to a 100-fold higher concentration than the concentration under which IAPP already forms fibrils (8) (Fig. 6,  E-H). No ThT binding was found in 14-day-aged incubations of the five mutants 8A, 4A, A(9 -12), A15,23,26, and A16,23,26, and the absence of fibrils was confirmed by TEM (Fig. 6, E and  F). By contrast, the three double mutants A15,16, A23,26, and A15,23 were amyloidogenic according to both the ThT binding assay and TEM (Fig. 6G). Notably, in the case of A16 and A23, weak ThT binding was observed, and TEM indicated the presence of both fibrils and significant amounts of amorphous aggregates (Fig. 6H).
Taken together, the above studies showed that mutations caused strong (Ͼ100-fold reduced amyloidogenicity; 5 mutants), medium (20 -99-fold reduced amyloidogenicity; 5 mutants), and weak/no (Ͻ20-fold reduced amyloidogenicity; 7 mutants) effects on IAPP amyloidogenicity (results summarized in Table 4). Their effects depended both on the nature of the substituted residues and on their position within the IAPP sequence. In this context, amyloidogenicities of A23, A26, A23,26, and A15,23, which exhibited strongly impaired binding to A␤40(42), were affected but not strongly reduced as compared with IAPP. Thus, the presence of Phe 23 and Ile 26 or Phe 15 and Phe 23 in isolation or combination with each other does not strongly affect IAPP amyloidogenesis. Importantly, the strongly reduced amyloidogenicity of the 4A mutant and the pronounced differences between the effects of triple/double mutations on IAPP amyloidogenicity suggested that cooperative interactions between Phe 15 , Leu 16 , Phe 23 , and Ile 26 underlie their effects on IAPP amyloidogenesis.

Effects of mutations on IAPP protofilament stability studied by MD simulations
To obtain more information about the mechanistic basis of the above-observed effects of mutations on IAPP amyloidogenicity, we next performed all-atom MD simulations in explicit water. The stability of a native IAPP protofilament model was compared with the stability of protofilament models of selected mutants. We focused on the four mutants A23, A26, A15,23, and 4A, which exhibited dramatically impaired binding to A␤40(42) but similar (A26), significantly reduced (A23 and A15,23), or strongly reduced amyloidogenicities (4A) as compared with IAPP (Table 4). We constructed hexameric protofilament models of IAPP and the mutants based on the IAPP fibril structure suggested by Eisenberg et al. (14) and recent results by some of us (28) (Fig. 7). Notably, the Eisenberg IAPP fibril structure has already been used in MD simulations by various groups to study both IAPP self-assembly and its heteroassembly with A␤40(42) (29 -31). Our simulations addressed the stability of the native IAPP protofilament as compared with the constructed protofilament models containing substitutions of Phe 23 by Ala (A23), of Ile 26 by Ala (A26), of Phe 15 and Phe 23 by Ala (A15,23), and of Phe 15 , Leu 16 , Phe 23 , and Ile 26 by Ala (4A) (Fig. 7).
Overall, the results of the MD simulations (Fig. 7) were in good agreement with the results of our ThT binding and TEM studies ( Fig. 6 and Table 4) and with the results of previous studies (12-14, 30, 32-39). During the 3 s of simulation, the native IAPP protofilament model remained stable and close to the starting structure as quantified by C␣ backbone root mean square deviation (RMSD) of the amyloid core region Phe 15 -Ser 29 (Fig. 7A). The MD simulations suggested that a major

Key determinants of a cross-amyloid peptide interaction
factor contributing to the stability of the U-shaped, ␤-strandloop-␤-strand motif of the IAPP protofilament is the hydrophobic core formed by residues located close to the loop region IAPP (18 -22) (14,30,31) (Fig. 7B). Thereby, interactions between hydrophobic and/or aromatic residues probably play an important role (14,30,31).
In the case of the A26 protofilament model, substitution of Ile 26 by Ala resulted in only a weak destabilization of the hydrophobic core, as evident by the fact that only few water molecules were found in the inner side of the strands after ϳ3 s of simulation time (Fig. 7, A and C). In addition, no significant changes in the Phe 23 -Phe 23 stacking interactions and the

Key determinants of a cross-amyloid peptide interaction
U-shaped structure of the A26 fibrils were observed. Thus, based on the MD simulations, the Ile 26 3 Ala mutation was not expected to cause a significant effect on IAPP amyloidogenicity, which was in good agreement with the findings of the amyloidogenicity assays.
In the case of the A23 mutation, both the tight hydrophobic core of the IAPP protofilament and the starting protofilament structure were retained despite the lack of Phe 23 -Phe 23 stacking interactions (Fig. 7, A and D). Thus, a not strongly reduced amyloidogenicity would be expected for A23 as compared with IAPP. Phe 23 -Phe 23 interactions could be more important on the quaternary structure level in fibrillar double/multilayer IAPP assemblies than for the single protofilament layer used here in the MD simulations. This hypothesis was in good agreement with the observed weak ThT binding and the significant amounts of amorphous aggregates found in A23 incubations in addition to the fibrillar assemblies.
For the A15,23 protofilament model, both the tight hydrophobic core and the protofilament structure were mostly retained; few water molecules were observed only in the peripheral region (Fig. 7, A and E). Thus, mutating both F15 and F23 by Ala was not expected to strongly affect IAPP amyloidogenicity as found above by the ThT binding and TEM studies.
Importantly, the MD simulations also suggested that the simultaneous substitution of Phe 15 , Leu 16 , Phe 23 , and Ile 26 by Ala (in 4A) causes dramatic effects on IAPP protofilament structure and stability (Fig. 7, A and F) as found above by the amyloidogenicity studies as well. This is reflected in the large RMSD with respect to the native IAPP protofilament structure (Fig. 7A). The MD simulations indicated formation of a water channel within the loop region, which was most likely due to gaps in the hydrophobic core. Furthermore, a markedly increased disorder in loop stacking was observed, possibly due to both the lack of Phe 23 -Phe 23 stacking interactions and the simultaneous destabilization of the hydrophobic core.

Key determinants of a cross-amyloid peptide interaction
In this context, our studies suggest that the above four residues and their interactions play a key role in the stability of the hydrophobic core of the IAPP amyloid core sequence IAPP (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29), which probably strongly contributes to both IAPP amyloid self-assembly and its cross-interactions with A␤40 (42). We also identified combinations of two or three of the above key residues whose substitution by Ala restores binding to A␤40 (42) even in the presence of hot spots that are substituted by Ala. Furthermore, our findings provided the first evidence for differences between the IAPP-A␤40 versus IAPP-A␤42 interaction interfaces. Structural differences between A␤40 and A␤42 and the highly dynamic and polymorphic nature of IAPP-A␤40(42) hetero-assemblies probably account for this latter finding (9, 12, 29, 41, 42).
The finding that each of the two aromatic residues Phe 15 and Phe 23 in isolation affects but is not of crucial importance for IAPP amyloidogenesis is in very good agreement with previous results (35,38,43). In fact, substitution of Phe 23 or Phe 15 by leucine has been reported to only weakly suppress (F23L mutant) or accelerate (F15L mutant) IAPP amyloidogenesis (35,38,43). Furthermore, non-additive effects on IAPP amyloidogenesis have been reported for various different multiple substitutions within the IAPP sequence, including F15L and F23L, which is also consistent with our findings (38,43). Regarding the two hydrophobic residues Leu 16 and Ile 26 , strongly reduced amyloidogenicities have been previously reported for mutants L16Q, I26P, and I26D (43)(44)(45). However, our studies showed that the conservative mutations L16A and I26A did not strongly affect IAPP self-assembly, suggesting that Leu 16 and Ile 26 in isolation are not key determinants of this process. Effects of Pro, Gln, or Asp substituents on peptide structure and amyloidogenicity most likely account for the apparent discrepancy between our findings and the previous ones.
The finding that specific aromatic/hydrophobic residues within the amyloid core of IAPP are able to control its crossinteraction with A␤40(42) but do not play a crucial role in its amyloid self-assembly is of high importance because it offers a molecular basis for deciphering mechanisms underlying the above and possibly cross-interactions between other amyloidogenic polypeptides as well (1,2,23,46). In fact, interactions of aromatic/hydrophobic residues play a variety of roles in protein-protein interactions, including amyloid self-assembly, and common molecular recognition rules are often involved (23,34,47). However, their roles in cross-interactions of intrinsically disordered proteins/polypeptides, including amyloidogenic ones, are largely unknown (47).
Finally, our studies generated a number of novel IAPP analogs that have very high sequence similarity to IAPP but strongly differ from it with regard to their amyloid self-assembly potency and/or their ability to cross-interact with A␤40 (42). For instance, we identified (a) the four IAPP mutants 8A, 4A, A15,23, and A23,26, which exhibit both strongly impaired binding to IAPP and/or A␤40(42) and improved solubility and/or amyloidogenicity, as compared with IAPP, and (b) the three IAPP mutants A23, A26, and A23,26, which are unable to interact with A␤40 but are still able to bind IAPP with high affinity. Thus, these peptides are suitable templates for designing novel IAPP analogs with tailored self-/cross-amyloid interactions and related functions (6,11).
In summary, here we identified single aromatic/hydrophobic residues within the IAPP amyloid core region that are able to control its interaction with A␤40(42) but not IAPP self-assembly. We also identified four aromatic/hydrophobic residues, which, in combination, are able to control both IAPP amyloid self-assembly and its cross-interaction with A␤40 (42). Furthermore, we devised a number of different full-length IAPP analogs that have a high sequence similarity to IAPP but distinct profiles regarding their amyloid self-assembly and/or their cross-interaction with A␤40(42). Our results provide a molecular basis for deciphering IAPP-A␤40(42) interactions and designing intervention strategies and novel IAPP analogs with optimized self-/cross-interactions as leads for therapeutics in T2D, AD, and possibly other devastating diseases linked to IAPP dysfunction and/or its cross-amyloid interactions (1,3,5,6,10,11,18,19).

. Schematic presentation of proposed IAPP residues and folds mediating cross-interactions between early prefibrillar IAPP and A␤40(42) species (top) versus IAPP amyloid self-assembly (bottom) based on our current results
Here identified IAPP hot spots/key residues are indicated by red/orange symbols in early disordered conformers only. The IAPP self-assembly-mediating sequence FGAIL is shown in yellow. The disulfide bridge between Cys 2 and Cys 7 in IAPP is indicated by a continuous line. Only a few heterodimers under the various possible hetero-assemblies are shown; previously identified IAPP binding sequences within A␤40 (42) are shown in purple (12,42).

Determination of binding affinities and ⌬⌬G values by fluorescence spectroscopic titrations
Fluorescence measurements were performed with a JASCO FP-6500 fluorescence spectrophotometer using an assay system that we have previously developed and applied for the determination of the affinity (K d(app) ) of the interaction of IAPP or its segments with IAPP, A␤40, and A␤42 (8,12,20). Briefly, excitation was at 492 nm, and fluorescence emission spectra were recorded between 500 and 600 nm within 2-5 min following solution preparation at 25°C.
Apparent affinities (K d(app) values) of the interactions of the partial IAPP segments (hot segments) with IAPP or A␤40 were determined by titration of freshly made solutions of synthetic N-terminal fluorescein-labeled IAPP segments and their Ala mutants (5 nM) with IAPP or A␤40 (up to 2 M) as described (12). K d(app) values of interactions of full-length IAPP or its Ala mutants with IAPP, A␤40, A␤42, or Ala mutants were determined by titration of freshly made solutions of synthetic N ␣ -terminal fluorescein-labeled IAPP (Fluos-IAPP) or mutant (Fluos-mutant) (5 nM) with IAPP, A␤40, A␤42, or the Ala mutant (up to 5 M; exception: Fluos-8A (or Fluos-4A) ϩ IAPP, up to 15 (or 10) M) as described (8,12,20).
Briefly, freshly made stock solutions of unlabeled peptides and their fluorescently labeled analogs in HFIP were used (8,12,20). All assays were performed in 10 mM sodium phosphate buffer, pH 7.4 (1% HFIP). Under these experimental conditions, freshly made solutions have been shown previously to contain fluorescently labeled IAPP (5 nM) mainly in a monomeric state, whereas titrations of fluorescently labeled mutants with nonlabeled ones suggest that at 5 nM, the mutants are also mainly present as monomers (see K d(app) values of self-association of mutants in Table 4) (12,20). The above experimental set-up allows for addressing interactions between early prefibrillar, mostly monomeric species of IAPP or its mutants with early prefibrillar species of A␤40(42), IAPP, or the mutants (8,12,20). Reversibility of binding of Fluos-IAPP/mutants to IAPP or A␤40(42) was verified. Of note, the presence of 1% HFIP has been shown previously to not affect binding affinities (12). Binding affinities were estimated by using 1:1 binding models; however, more complicated binding models may also apply due to the strong self-assembly potentials of these peptides (8,12,20). Determined K d(app) values are means Ϯ S.D. from three binding curves. The changes in binding free energy were calculated via the equation, ⌬⌬G ϭ RTln(K d(mut) /K d(wt) ), where R is the gas constant and T is absolute temperature). Determined K d(app) values of Fluos-IAPP interaction with IAPP, A␤40, and A␤42 were 9.7, 48.5, and 219 nM, corresponding to binding free energies, ⌬G values, of Ϫ10.9 kcal/mol (IAPP-IAPP), Ϫ9.9 kcal/mol (IAPP-A␤40), and Ϫ9.7 kcal/mol (IAPP-A␤42) (8,20). Of note, the K d(app) values of these latter interactions were determined in the context of the current fluorescence titration studies as well as internal controls and were very similar to the ones determined previously (8,12,20).

Biotin pulldown assays, NuPAGE, and Western blot analysis
A previously developed assay system for the pulldown of IAPP-A␤ heterocomplexes was applied (8,20). Briefly, synthetic N ␣ -terminal biotinylated IAPP (biotin-IAPP) or Ala mutants (2.5 M) were incubated (1 h) alone or in a mixture with freshly dissolved A␤40 or A␤42 (5 M) in 200 l of 10 mM sodium phosphate buffer (pH 7.4) at room temperature. Streptavidin-coated magnetic beads were washed, blocked, and incubated with assay buffer, and the preincubated peptide solutions (see above) were added to the beads as described (8,20). Nonspecific binding of A␤40(42) was determined by incubating A␤40(42) alone (5 M) with beads. After incubation (4 h), beadbound complexes were isolated by magnetic affinity, beads were washed, and reducing NuPAGE sample loading buffer was added. The beads were boiled (5 min), and supernatants containing complexes or peptides alone were subjected to NuPAGE electrophoresis in 4 -12% BisTris gels with MES running buffer according to the manufacturer's (Invitrogen) recommendations. Equal amounts of peptides were loaded in each lane; loaded amounts for A␤40 (42) controls (freshly dissolved peptides without incubation with beads) were 0.2 g (A␤40) and 0.3 g (A␤42). Thereafter, peptides were blotted onto nitrocellulose using an XCell S Blot Module blotting system (Invitrogen). A␤40(42) bound to biotin-IAPP or biotin-mu-tants was revealed by Western blotting using a polyclonal rabbit anti-A␤ antibody (Sigma) in combination with a POD-coupled secondary antibody and the SuperSignal West Dura Extended Duration Substrate staining solution (Thermo Scientific). Negligible nonspecific binding to the beads was found for A␤40 (42) under the applied conditions. Control blots were also performed to detect biotinylated peptides and were developed with streptavidin-POD (anti-Biotin in the figures). A molecular weight marker ranging from 3.5 to 260 kDa (Novex) was electrophoresed in the same gels.

Sedimentation assays
A previously developed sedimentation assay for the determination of solubility of IAPP and its analogs was applied (16,20). Briefly, IAPP and all mutants were incubated at 100 M in 10 mM sodium phosphate buffer, pH 7.4, at room temperature. Of note, under these conditions, IAPP has been previously shown to immediately precipitate (16,20). IAPP solubility was also tested at 1 M, where it has previously been found to precipitate within a few days (20). In addition, solubility of A15,23 and single Ala mutants was studied also at 50 and 20 M, respectively. At the time points 0 h and 7 days, solutions were centrifuged (20,000 ϫ g for 20 min), and peptide amounts in supernatants and pellets were quantified by the BCA protein assay (Pierce) (16,20). The amount of peptide present in a non-centrifuged sample at 0 h was also quantified in each experimental set-up and corresponded to the total peptide amount. A mutant was judged as soluble when Ͼ80% of mutant was present in the supernatant at the 7-day time point. Results in Fig. 5A and supplemental Fig. S5 show amounts (percentage of total) of soluble peptides present in supernatants at the end of the incubation process (7 days) except for 1 M IAPP (end of incubation at 5 days) (20).

Far-UV CD spectroscopy and determination of secondary structure contents
Far-UV CD measurements were carried out using a Jasco 715 spectropolarimeter. Spectra were collected between 195 and 250 nm at 0.1-nm intervals with a response time of 2 s and at room temperature; each spectrum is the average of three spectra. CD measurements were performed using freshly made solutions containing IAPP or Ala mutants at 5-50 M (as indicated in the figure legends) in aqueous 10 mM sodium phosphate buffer, pH 7.4, containing 1% HFIP, as described previously for IAPP (20,24). Briefly, CD solutions were obtained by diluting peptides from their freshly made HFIP stocks, and IAPP was in the lag time of its fibrillization process under these conditions (data not shown and results of previous studies) (8,16,20,24,53). The spectrum of the buffer was subtracted from the CD spectra of the peptide solutions. Secondary structure contents (supplemental Fig. S6) were calculated by deconvolutions of CD spectra of IAPP and the mutants performed with ContinLL at DichroWeb using the reference spectra set 7 (25)(26)(27)54).

ThT binding assays
Kinetics of amyloidogenesis of IAPP and the Ala mutants were determined at room temperature using previously estab-lished ThT binding assay systems (8,12,17,20). The effects of Ala mutations on IAPP amyloidogenicity were determined by two rounds of screening. First, kinetics of amyloidogenesis of IAPP and all Ala mutants were studied at 12 M (in 50 mM aqueous sodium phosphate buffer, pH 7.4, containing 100 mM NaCl and 0.5% HFIP), which corresponds to a 19-fold higher concentration than the concentration (625 nM) under which IAPP has been previously found to form fibrils within 24 h (20). ThT binding was measured at various time points up to 7 days. Mutants that did not bind ThT under the above conditions were subjected to a second round of screening at 62.5 M (in aqueous 10 mM Tris-HCl buffer, pH 7.4, containing 2% HFIP), corresponding to a 100-fold higher concentration than the concentration under which IAPP aggregates into fibrils, and amyloidogenesis was followed up to 14 days by the ThT binding assay (20). IAPP was included in each experimental set as control, and results of ThT assays were confirmed by TEM (Fig. 7).

TEM
TEM was applied to confirm the results of the ThT assays as described previously (8,20). Briefly, 10-l aliquots of solutions used for ThT assays were applied on carbon-coated grids at the end of the incubation process. Following washing with distilled water, grids were stained with aqueous 2% (w/v) uranyl acetate and examined with a JEOL JEM 100CX electron microscope at 100 kV (8,20).

Molecular dynamics simulation
Simulation models-The IAPP hexamer structure was extracted from the full-length, all-atom IAPP fibril model suggested by Eisenberg et al. (14), which is based on crystal fibril structures of IAPP segments NNFGAIL and SSTNVG. Within the amyloid fibril hierarchy, the IAPP hexamer corresponds to a so-called protofilament consisting of two antiparallel ␤-sheets connected by a loop region. This tertiary structure results from six IAPP monomers stacked on top of each other in ␤-hairpin conformation with an interpeptide spacing of 4.8 Å. In each monomer, residues Cys 2 and Cys 7 were connected by a disulfide bridge, and the relevant alanine mutations were inserted into an energy-minimized native IAPP hexamer structure.
Simulation protocol-The simulations were performed using the Amber14 simulation package with the all-atom force field ff14SB and explicit TIP3P water (55,56). 12 neutralizing chloride ions were added, and standard ionization states at pH 7 were used for ionizable side chains. The IAPP hexamers were placed into an octahedral simulation box with a minimum dis-

Key determinants of a cross-amyloid peptide interaction
tance of 12 Å to the box boundaries, resulting in system sizes of ϳ3,200 solute and ϳ11,000 solvent atoms. Before starting the integration in time, the energy of each model system was minimized using 10,000 steps of steepest-descent algorithm. The systems were then heated up to 300 K in steps of 50 K and 100 ps each. A subsequent equilibration phase was divided into 2 ns in the NVT ensemble and a further 4 ns in the NPT ensemble using the Berendsen weak-coupling scheme with a reference temperature of T ϭ 300 K, a reference pressure of p ϭ 1 bar, and coupling time constants of T ϭ 0.1 ps and p ϭ 1.0 ps (57). During the equilibration, the positions of heavy backbone atoms were restrained using a harmonic potential with a force constant of k restr ϭ 2.4 kcal/(mol⅐Å 2 ). In a subsequent relaxation phase, the position restraints were gradually removed by reducing k restr in steps of 0.5 kcal/(mol⅐Å 2 ) and 10 ps each.
The unrestrained production simulations were performed for ϳ3 s in the NPT ensemble at T ϭ 300 K and pressure of 1 bar using the standard leapfrog algorithm. The combination of the SETTLE algorithm for bond constraining and hydrogen mass repartitioning allowed the use of an integration time step of 4 fs (58).
Long-range Coulomb interactions were treated using periodic boundary conditions and the particle mesh Ewald method with a 12-Å real-space cut-off (59). Simulation frames were saved every 80 ps.
Simulation analysis-The resulting simulation trajectories were analyzed using the program Cpptraj version 14.05 (60).
To compare the stability of mutants with respect to the native IAPP hexamer, a C␣ backbone RMSD was calculated within the amyloid core segment Phe 15 -Ser 29 , which includes the loop region and the mutation-relevant parts of the ␤-sheets. Both peripheral monomers at the hexamer tips were excluded from RMSD calculation due to considerable conformational fluctuations inherent to small amyloid protofilament models (28).
The SASA within the hexamer's internal amyloid core was quantified using the LCPO algorithm of Weiser et al. (61). Only amino acids pointing toward the internal amyloid core within the mutation-relevant region were included in the calculation: Leu 16