Steric Crowding of the Turn Region Alters the Tertiary Fold of Amyloid-β18–35 and Makes It Soluble*

Aβ self-assembles into parallel cross-β fibrillar aggregates, which is associated with Alzheimer's disease pathology. A central hairpin turn around residues 23–29 is a defining characteristic of Aβ in its aggregated state. Major biophysical properties of Aβ, including this turn, remain unaltered in the central fragment Aβ18–35. Here, we synthesize a single deletion mutant, ΔG25, with the aim of sterically hindering the hairpin turn in Aβ18–35. We find that the solubility of the peptide goes up by more than 20-fold. Although some oligomeric structures do form, solution state NMR spectroscopy shows that they have mostly random coil conformations. Fibrils ultimately form at a much higher concentration but have widths approximately twice that of Aβ18–35, suggesting an opening of the hairpin bend. Surprisingly, two-dimensional solid state NMR shows that the contact between Phe19 and Leu34 residues, observed in full-length Aβ and Aβ18–35, is still intact in these fibrils. This is possible if the monomers in the fibril are arranged in an antiparallel β-sheet conformation. Indeed, IR measurements, supported by tyrosine cross-linking experiments, provide a characteristic signature of the antiparallel β-sheet. We conclude that the self-assembly of Aβ is critically dependent on the hairpin turn and on the contact between the Phe19 and Leu34 regions, making them potentially sensitive targets for Alzheimer's therapeutics. Our results show the importance of specific conformations in an aggregation process thought to be primarily driven by nonspecific hydrophobic interactions.

Alzheimer's disease pathology has been associated with the aggregation of amyloid ␤ (A␤), 6 a 39 -43-amino acid-long pep-tide (1,2). In this process, the unstructured monomers of A␤ get converted into amyloid fibrils composed of hairpin-shaped monomeric units assembled in a parallel ␤ sheet arrangement (3,4). The central part of this hairpin structure consists of a hydrophilic region flanked by hydrophobic ␤-sheet-forming segments at both ends (5)(6)(7)(8)(9)(10). This particular shape is believed to be dictated by the specific pattern of hydrophobic and charged residues in the A␤ sequence and has been identified even in soluble A␤ aggregates (7,(11)(12)(13)(14)(15)(16)(17), including the very early stage oligomers of A␤ (18). Chemically constraining the side chains of Asp 23 and Lys 28 accelerates the kinetics of A␤ aggregation by stabilizing the hairpin structure (5). On the other hand, introducing a negative charge at Ser 26 disrupts the Asp 23 -Lys 28 salt bridge, reduces the plasticity of the turn region (Gly 25 -Gly 29 ), and stabilizes the soluble monomeric and oligomeric assemblies of A␤ (19). This hairpin turn thus seems to be a critical factor dictating the self-assembly of A␤ under physiological conditions. Studying the influence of the turn region on the properties of A␤, separately from that of the distal terminal regions and without changing the electrostatics, can yield valuable information on the logic of A␤ assembly.
A␤ aggregation appears to be primarily hydrophobic in nature. In fact a contact between hydrophobic regions containing Phe 19 and Leu 34 is one of the earliest contacts formed during the aggregation of amyloid ␤ (18). In a generic hydrophobic aggregate, the burial of the hydrophobic surface can in principle proceed in an intermolecular fashion, which would appear to render the turn unnecessary. However, such "straight chain" hydrophobic interactions would end up sandwiching the charged midsection between the two hydrophobic flanking regions (type II; Fig. 1), which would be energetically costly. A hairpin-shaped aggregate on the other hand relieves the hydrophilic region from such constraints by exposing the turn region (type I; Fig. 1). It is likely that the glycine residue in the turn region, Gly 25 , facilitates the hairpin bend, because it allows the peptide to have enough space to avoid steric hindrance. If Gly 25 is removed altogether, the peptide will not have significantly different hydrophobicity, but a tight turn would likely be sterically unfavorable. The neighboring residues Ser 26 and Val 24 have bulky side chains and may find it difficult to fit within a tight turn without Gly 25 . Consequently, those properties that depend on the turn will change.
We have earlier studied the 18 -35 fragment of A␤ (A␤ 18 -35 , hereafter called S), which has shown that the turn can form without any contribution from the terminal regions, and the solubility and other biophysical properties of the peptide remain similar to that of A␤ (20). Here we study a mutant of the A␤ 18 -35 fragment, lacking the residue Gly 25 ( 18 VFFA-EDV 24 _ 26 SNKGAIIGLM 35 , hereafter called S ⌬G25 ; residue numbers are according to the full-length A␤). If aggregation is critically dependent on the formation of the turn, then this mutation is expected to result in a more soluble peptide, having very different physical and biological properties. If it forms the fibrils at all, it would be interesting to investigate whether it forms parallel (type IIA; Fig. 1) or antiparallel (type IIB; Fig. 1) fibrils without a turn or whether it is similar to S, which makes fibrils containing a hairpin turn just like A␤ (type I; Fig. 1).
To verify our hypothesis, we have determined the biophysical properties of S ⌬G25 , and these characteristics were compared with those of S, determined earlier (20). In addition, many of the properties were also compared with those of ⌬G29 and G25P mutants of S (hereafter called S ⌬G29 and S G25P , respectively) and the ⌬G25 mutant of full-length A␤ 40 (hereafter called A␤ ⌬G25 ). S ⌬G29 examines whether the position of the glycine deletion is critical. The behavior of S G25P on the other hand would tend to force a turn (in the monomers containing the cis form of the proline) or open up the turn (for the trans form). The aggregation kinetics and solubility of the peptides were studied by measuring optical absorbance of the centrifuged supernatant as a function of time. The sizes of the monomers and the formation of oligomeric aggregates in the solution were measured using fluorescence correlation spectroscopy (FCS), using fluorescently labeled peptides. The secondary conforma-tion of the peptides in the solution state is probed using thioflavin-T (thio-T) binding, CD, and solution state NMR. The effect of S and S ⌬G25 on the aggregation and fibril formation traits of A␤ 40 were probed by performing cross-seeding experiments with peptide mixtures. The morphology of the aggregates formed by S ⌬G25 and A␤ ⌬G25 were probed by transmission electron microscopy. Conformations of the aggregates were assessed with the help of IR and solid state NMR (ssNMR) spectroscopy. The intermolecular arrangements were specifically probed by cross-linking neighboring tyrosine residues, which assays in-register parallel ␤-sheet arrangements. The results yielded by these studies help us determine the role played by the hairpin turn in general, and the Phe 19 -Leu 34 interaction in particular (21,22), in the self-assembly of A␤ under physiological conditions.
Aggregation Kinetics and Solubility-The peptides were dissolved and adjusted to pH 12 with NaOH at different concentrations ranging from 1 to 2 mM and incubated for 30 min. The peptide is mostly in the monomeric or small oligomeric form in basic conditions (24). Then it was mixed with 20 mM phosphate buffer (containing 146 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 0.8 mM MgSO 4 , 0.4 mM KH 2 PO 4 , and 20 mM Na 2 HPO 4 ) maintained at pH 7.4. Sodium azide (anti-bacterial) concentration in buffer is kept at 0.5 mM. The dilution factor was usually 10, and the final pH of the solution was verified to be within 0.1 pH unit of 7.4. The time of mixing was considered as the zero time of aggregation, and concentration was measured using the absorbance at 230 nm using the peptide backbone as a probe of concentration and also by tyrosine fluorescence in case of A␤ and A␤ ⌬G25 . The solutions were constantly rotated at 20 rpm at room temperature (298 K). The initial reading was recorded without any centrifugation immediately after mixing. Later time points are measured from the supernatant of sample cen- trifuged at 2000 ϫ g for 20 min. The supernatant was mixed back with the precipitate after each measurement and rotated in a centrifuge tube with rpm of 20 for efficient mixing. The aggregation was studied for 7 days. Jasco double beam absorption spectrophotometer (model no. V-530; Jasco) was used for all measurements with necessary day to day blank corrections. To assess the effect of possible cross-seeding, aggregation kinetics of 100 M A␤ 40 in presence of 100 M S or S ⌬G25 were monitored using the fluorescence of tyrosine remaining in the supernatant with time.
Fluorescence Correlation Spectroscopy-FCS measurements were performed on a home-built instrument and following the methods established earlier (25). Tetramethylrhodamine-labeled peptides were used at a concentration ratio of 1:1000 (labeled:unlabeled) as fluorescent markers for the peptide aggregates. MEMFCS, a fitting routine developed specifically for such measurements (26), was used to obtain a size distribution from the FCS data in a model-free manner. The instrument was calibrated using the dye rhodamine B.
Circular Dichroism-CD measurements were performed on 100 M peptide solution incubated in 2 mM phosphate buffer at pH 7.4 for 6 and 16 h. The solutions were centrifuged at 2000 ϫ g for 20 min. The measurements were performed with the solutions both before centrifugation and after centrifugation (with the supernatant). Far UV CD spectra for the peptides were measured over a range of 195-260 nm in a cuvette of 1-mm path length in a J-810 spectropolarimeter (Jasco).
Thioflavin-T Binding Studies-Peptide samples (100 M each) were prepared in phosphate buffer (pH ϳ7.4) containing 5 M thio-T. The peptide samples were diluted from pH 11 aqueous solution. thio-T fluorescence from the samples were recorded in a FluoroMax 3 spectrofluorimeter (Horiba Scientific) using 444-nm excitation as a function of time after preparation of the sample in pH 7.4 buffer.
Electron Microscopy-Solutions containing 1 mM S ⌬G25 or 100 M A␤ ⌬G25 at pH 7.4 phosphate buffer were incubated at room temperature for 72 h. 10 l of these solutions were placed on carbon-coated 100 mesh copper grids (Electron Microscopy Sciences, Hatfield, PA) and allowed to be adsorbed for 2-3 min. The grids were blotted and mildly washed with MilliQ water three or four times. Then the samples were stained by 0.1% of uranyl acetate followed by drying under an infrared lamp. The samples were examined with a transmission electron microscope (LIBRA 120, EFTEM; Carl Zeiss). The fibril widths were analyzed with ImageJ (open source software).
Nuclear Magnetic Resonance Spectroscopy-A 400 M solution of S ⌬G25 peptide incubated overnight was subjected to solution state NMR measurements. All measurements were performed on 500-MHz Buker AVIII spectrometer using a 5-mm double channel direct probe equipped with Z-gradients.
For ssNMR measurements, selective isotopically enriched peptide was allowed to aggregate from a starting concentration of 1-2 mM in the similar buffer system used for solubility measurements. The aggregates were isolated from the soluble peptide by centrifugation for 1 h. The pellet thus collected was washed with deionized water twice by resuspending it in water and centrifuging for 1 h each time. The final pellet so obtained was rapidly frozen using liquid nitrogen, lyophilized, and packed in a 2.5-mm magic angle spinning (MAS) rotor. All ssNMR measurements were performed on 700-MHz Bruker AVIII NMR spectrometer at MAS frequency ( r ) of 12 and 17 kHz using a 2.5-mm triple resonance MAS probe. 1 H dipolar decoupling was accomplished using swept frequency, twopulse phase modulation ( ϭ 15°) decoupling scheme (27). Two-dimensional 13 C-13 C through-space NMR spectra were recorded using second order dipolar recoupling schemes of PARIS-xy (m ϭ 1, n ϭ 0.5) (28) and mixing periods of 20 (at r ϭ 17 kHz) and 320 (at r ϭ 12 and 17 kHz) ms, respectively.
All one-dimensional data were processed and analyzed using TopSpin 3.2. All two-dimensional spectra were processed with TopSpin 3.2 and analyzed using Analysis 2.3.1 (29). Secondary chemical shifts were calculated by subtracting the temperature and pH adjusted sequence-corrected random coil chemical shifts from the observed chemical shifts for each amino acid (30,31).
Infrared Studies-200 M of A␤ 40 , 200 M A␤ ⌬G25 , and 1 mM of S ⌬G25 were incubated separately for 72 h in phosphate buffer (pH 7.4). All the solutions were subsequently centrifuged at 16,000 ϫ g for 15 min, and the supernatants were discarded. Precipitates so obtained were lyophilized and subjected to IR studies. The FTIR spectra of all the samples were recorded in a Nicolet 6700 FTIR instrument (Thermo Electron Corporation).
Tyrosine Cross-linking Experiments-We prepared separate peptides with one extra tyrosine residue at the N terminus of S and S ⌬G25 for these experiments. 200 M of A␤ 1-40 , 200 M of tyr-S, and 1 mM of tyr-S ⌬G25 were incubated for 14 days in pH 7.4 phosphate buffer (20 mM Na 2 HPO 4 , 150 mM NaCl, 5 mM KCl, and 2 mM NaN 3 ; pH adjusted by adding HCl). Aggregates so obtained were separated by mild centrifugation (2000 ϫ g for 10 min), discarding the supernatant. Aggregates were then resuspended in appropriate amount of phosphate buffer, so that all vortexed solutions had similar tyrosine fluorescence (which ensures matching final peptide concentrations). 10 l of ice-cold H 2 O 2 and 6 l of freshly prepared FeSO 4 were added simultaneously to 600 l of these peptide solutions. These solutions were then vortexed for 60 s. Fluorescence spectra were obtained immediately by exciting at 260 nm. Although tyrosine has a fluorescence peak at ϳ305 nm, the cross-linked tyrosine peak is in the visible region. A similar experiment was performed with a similar concentration of free tyrosine solution to estimate the extent of random cross-linking in solution. The spectra were normalized by dividing them by the fluorescence intensity obtained at 305 nm before the cross-linking step. To remove the contribution from random cross-linking, the normalized spectra obtained from free tyrosine were subtracted from these normalized spectra.

Results and Discussion
Aggregation kinetics of the different peptides were compared by monitoring the concentration remaining in solution after gentle centrifugation (2000 ϫ g for 20 min; Fig. 2A). The concentration is measured by recording the absorbance of the supernatant at 230 nm. The saturation concentration (C sat ) is the limiting value of the concentration remaining in the solution after the solution is incubated for a sufficiently long time Steric Crowding of the A␤ Turn DECEMBER 11, 2015 • VOLUME 290 • NUMBER 50 (32). S ⌬G25 was found to reach a C sat of 566 Ϯ 150 M in 7 days. S on the other hand attains a C sat of ϳ18 Ϯ 2 M over a similar period, which is close to the full-length A␤ 40 (ϳ12 Ϯ 4 M). Thus, deletion of a single residue Gly 25 caused the solubility of A␤ 18 -35 fragment to increase by more than 20 times under physiological buffer conditions. The same deletion in the case of the full-length peptide also leads to a 3-fold increase in solubility (A␤ ⌬G25 solubility ϭ 36 Ϯ 12 M; Fig. 2A). This increase is less than that of S ⌬G25 but still very substantial. We also examined two more peptide variants: S ⌬G29 and S G25P mutants of S . They were found to reach saturation concentrations of 9 Ϯ 1 and 20 Ϯ 4 M ( Fig. 2A), respectively. Therefore, both the nature and the location of a glycine residue are critical for peptide aggregation. If the turn favors aggregation, then we expect that initially the aggregation kinetics of S G25P may be slow (for the trans population), but with time the trans form may isomerize to the cis form (the kinetics of such a cis-trans isomerization in an ordered aggregate may be slow), and the solubility may finally become similar to that of the wild type. This is indeed corroborated by the data.
Next we probed the effect of Gly 25 deletion on the soluble oligomeric species of the S. The size of this species was determined by performing FCS measurements using tetramethylrhodamine-labeled peptides as probes. FCS, in different forms, has proved to be an effective tool for following the size of evolving protein aggregates (32)(33)(34). 100 M of both peptides (at 1000:1 unlabeled:tetramethylrhodamine-labeled peptide ratio) were incubated for 6 h under physiological buffer conditions. The soluble aggregates were obtained from the supernatant of these solutions by mild centrifugation at 2000 ϫ g at 6-h time points. S ⌬G25 is subsaturated under this condition and does not form insoluble aggregates. For S ⌬G25 , FCS data were recorded both before and after centrifugation at the 6-and 16-h time points. The FCS data were analyzed with the model-free MEM-FCS software (26) to obtain the distribution of size (specifically, the hydrodynamic radius, R H ) in the solution (Fig. 2B). The size distribution at 6 h has peaks at 0.96 and 1.5 nm for S and S ⌬G25 , respectively. Hydrodynamic radii of S and S ⌬G25 solution incubated for a longer time period at very low concentrations (500 nM), which is expected to allow the oligomers to dissociate to the smallest possible state at such concentrations (possibly monomeric), were found to be 0.66 Ϯ 0.01 and 0.93 Ϯ 0.02 nm, respectively (Fig. 2C). Therefore, the species observed in solution phase after 6 h of incubation are most likely dominated by small oligomers and not the monomers, of S and S ⌬G25 . This tells us that both peptides form metastable oligomeric species at the early stages, with S ⌬G25 forming somewhat larger ones. However, these larger S ⌬G25 oligomers do not further grow into fibrillar aggregates, until a very high concentration is reached. On the other hand, the size of the putative monomer of S is consistent with the monomeric form of full-length A␤ after taking their masses into account (the R H for a homogeneous spherical particle varies as the cube root of the mass). However, S ⌬G25 remains comparatively bigger in size. This implies that either this is a monomeric species, which is more open (i.e. stiffer) than A␤, or, even at such concentrations (500 nM) and such long incubation times (21 days), S G25 remains in a small oligomeric state.
If the conformation of the soluble S and S ⌬G25 species is similar to that of A␤, then they may be expected to interact with it and affect its aggregation kinetics. This is examined by a crossseeding study. The aggregation kinetics of 100 M A␤ in presence of 100 M S and S ⌬G25 respectively was monitored and compared with the aggregation kinetics of 100 M A␤ by itself. The saturation concentration of A␤ 40 is not significantly different in the presence of either S or S ⌬G25 (Fig. 3). However, S leads to a somewhat faster initial aggregation phase. This suggests that the soluble species of S have conformations similar to  that of A␤ 40 (20) and is able to perturb the aggregation kinetics. However, because both A␤ and S have very similar saturation concentrations, the saturation concentration of the mixture is not altered. On the hand, S ⌬G25 does not interfere with the aggregation of A␤ 40 , likely because its own solubility is very high and its conformation is rather different from that of A␤ 40 .
These results suggest that the small oligomeric species formed by S ⌬G25 differ at the molecular level from S and A␤. To study this aspect, the secondary structures of these small oligomeric species (initial concentration ϭ 100 M) were obtained by circular dichroism measurements on the supernatant solutions (Fig. 4A). The CD data show that secondary structure of the soluble oligomers is mostly random coil in both the cases. However, S oligomers contain considerable amount of ␤-sheet, whereas S ⌬G25 oligomers contain some amount of ␣-helix, as determined by Reed's method (S, 77% random coil and 23% ␤-sheet; S ⌬G25 , 82% random coil and 17.3% ␣-helix) (35). The CD spectrum of S ⌬G25 does not change with time, as shown by a spectrum recorded 16 h after preparation (Fig. 4B). As may be expected, thio-T binding studies revealed that a considerably richer ␤-sheet character develops in the larger aggregates of S compared with that of S ⌬G25 as the aggregation proceeds (Fig.  4C), starting with an initial concentration of 100 M. Of course, only S forms these larger aggregates at these concentrations.
It is interesting to ask exactly what kind of secondary structure S ⌬G25 forms. The high solubility of S ⌬G25 gives us an opportunity to directly probe the structure of these oligomeric species with solution state proton NMR. Fig. 5 shows amide proton region of 1 H-1 H ROESY (mixing time ϭ 300 ms) spectrum of a day-old 400 M S ⌬G25 solution recorded at 298 K (Fig.  5A) and 283 K (Fig. 5B). Most of the H ␣ -H N cross-peaks could only be observed at the lower temperature, indicating the flexible nature of S ⌬G25 units in these soluble oligomeric states. 1 H chemical shift assignment could be achieved for 13 of 17 amino acids in S ⌬G25 using 1 H-1 H COSY, total correlation spectroscopy, and ROESY spectra recorded at 283 K. Chemical shifts so obtained are listed down in Table 1 along with the H ␣ secondary chemical shift (⌬␦, the difference between observed and random coil chemical shift) values for each amino acid. Because only two amino acids, Asn 27 and Met 35 , showed ⌬␦ values greater than ͉0.1͉ ppm, it is evident that the S ⌬G25 peptide is mostly devoid of well ordered secondary structure in its soluble aggregated states (36).
These studies reveal that S (and not S ⌬G25 ) acquires ␤-sheet secondary structural elements at an early stage of aggregation, which are further propagated as the aggregates grow in size. Even though deletion of Gly 25 from this fragment does not prevent the formation of small oligomers in the solution phase, it completely alters the conformation of these oligomers. It is interesting to speculate what holds these oligomers together despite the lack of secondary structures. This is likely driven by the nonspecific intermolecular interaction of the hydrophobic residues, but our experiments do not probe this aspect.
Next we ask whether this peptide is capable of forming amyloid fibrils and whether these fibrils bear the same characteristics as that formed by S. For this, 1 mM solution of S ⌬G25 was incubated for 72 h. The precipitated fraction of this solution was probed by using TEM, and the results are shown in Fig. 6. The TEM images showed amyloid like fibrillary structures, very similar in appearance to the fibrils obtained from S (20). However, a closer look at the TEM images revealed a very significant difference between the S and S ⌬G25 fibrils. Although for S ⌬G25 the average fibril width is 7.2 Ϯ 0.2 nm (average of 50 positions recorded from 5 different images obtained from duplicate sample preparations), it was found to be only 3.0 Ϯ 0.7 nm for the S  fibrils (20). These results strongly favor the hypothesis that S ⌬G25 forms wider fibrils by opening up the hairpin (type II; Fig. 1). We note that this does not rule out the presence of a minor population of precipitates, which do not form proper fibrils and may have a hairpin turn. For straight and wider fibrils, the peptides may assemble in a parallel (type IIA; Fig.  1) or in an anti-parallel (type IIB; Fig. 1) orientation. Of course, this assumes that the fibrils are still made up of ␤-sheets. To resolve this issue, we further investigated the secondary and tertiary structural contents of the S ⌬G25 fibrils by ssNMR and IR studies.
ssNMR studies were performed on aggregates of 7S ⌬G25 peptide containing isotopically enriched Val 18 , Phe 19 , Ala 21 , Gly 33 , and Leu 34 residues. These studies revealed that S ⌬G25 fibrils are also rich in ␤-sheet conformations, similar to the amyloid fibrils formed by S and full-length A␤ (20). Fig. 7A shows selected regions of two-dimensional 13 C-13 C PARIS-xy (m ϭ 1, n ϭ 0.5) through space correlation spectrum recorded at 17-kHz MAS using a mixing time of 20 ms. Also shown in the figure is the sequence assignment for each isotopically enriched amino acid. Two structural conformers were identified for all but the Phe 19 amino acid. Weakly populated conformations, as adjudged from the peak intensities, are marked with a prime symbol. This shows the presence of structural heterogeneity at molecular level in S ⌬G25 fibrils, a prevalent feature of amyloid aggregates (37)(38)(39). Chemical shifts so obtained (Table 2) were used to calculate the secondary chemical shifts (⌬␦) of ␣, ␤, and carbonyl carbons for each conformer and are shown in Fig. 7B. A positive value for ␤ and negative values for ␣ and carbonyl carbons are indicative of ␤-sheet propensity of an amino acid. A comparison with S shows that although the terminal residue Val 18 is unstructured, Phe 19 is in a ␤-sheet conformation in both the peptide fibrils. Ala 21 has its major population in ␤-sheet conformation in both the cases, but the weaker one is structured only in the case of S ⌬G25 . Leu 34 conformers show ␤-sheet propensities in both cases. The only difference was observed in the case of Gly 33 , which shows enhancement in heterogeneity, as well as nonstructured propensities in S ⌬G25 . Overall, we can say that S and S ⌬G25 fibrils share strong similarities at the secondary structure level.
Surprisingly, long range contacts in S ⌬G25 were observed between amino acids Phe 19 and Leu 34 and between amino acids Phe 19 and Gly 33 in two-dimensional 13 C-13 C PARIS-xy (m ϭ 1, n ϭ 0.5) through space correlation spectrum recorded at 12-kHz MAS using a long mixing time of 320 ms (Fig. 7C). Through space contact between Phe 19 and Leu 34 is a hallmark feature of the parallel hairpin structure typically observed for full-length A␤ fibrils (6 -8, 40) and also for S (20). In fact the disruption of this contact completely changes the toxicity of A␤ (22). However, this appears to be at odds with the TEM results, which predict that the peptide remains in an open state in the fibril. However, the apparent contradiction can be resolved if S ⌬G25 monomers are arranged in an anti-parallel ␤-sheet, with the Phe 19 residue of one monomer juxtaposed to the Leu 34 residue of the other. Also, a long range contact between Phe 19 and   Gly 33 was observed in S ⌬G25 fibrils, which is not observed for S and A␤ fibrils (20). The presence of this new contact is consistent with the presence of an atypical structural conformation in S ⌬G25 fibrils, possibly an open and anti-parallel arrangement as depicted in Fig. 1 (type IIB). The hypothesis of an open conformation in an anti-parallel arrangement is further supported by the IR spectra. IR amide band frequencies are directly correlated to the secondary structure components. The amide I band (Fig. 8A) of S ⌬G25 (solid line) is wider than A␤ 1-40 (dashed line). The peak at 1626 cm Ϫ1 indicates the presence of a ␤-sheet conformation in the fibrillar state, and the peak at 1680 cm Ϫ1 is characteristic of anti-parallel ␤-sheet strands (41).
A more direct proof for this intermolecular antiparallel ␤-sheet arrangement comes from tyrosine cross-linking experiments. Wild type A␤ has a tyrosine at the 10th position. The parallel in-register arrangement of A␤ peptides in the fibril form places their tyrosine residues next to each other, which can easily be cross linked by photo-induced cross-linking of unmodified proteins (42,43) or simply by Fe (II)/H 2 O 2 treatment (44,45). The formation of the cross-linked tyrosine can be quantitatively monitored by measuring the red shifted fluorescence of dityrosine (46,47). Because S does not have any tyrosine residue in it, a tyrosine residue was added at the N terminus of both S (Tyr-S) and S ⌬G25 (Tyr-S ⌬G25 ). Aggregates that possess a parallel in-register arrangement similar to that of fulllength A␤ should have their tyrosine residues amenable to cross-linking. The results show a high amount of cross-linking in case of Tyr-S and A␤ 40 and negligible cross-linking for Tyr-S ⌬G25 fibrils (Fig. 8B), clearly showing that the Tyr residues in S ⌬G25 are not close to each other. This provides a strong corroboration of our model of antiparallel ␤-sheet conformation for the ⌬G25 mutant, in contrast with the parallel, in register ␤-sheet conformation in the wild type S and A␤. Interestingly, unlike S ⌬G25 , A␤ ⌬G25 fibrils were found to be folded (fibrillar width 7-10 nm; Fig. 9A), and the IR spectrum suggests that it contains a parallel ␤-sheet structure (Fig. 9B). These results indicate that absence of the glycine residue in the 25th position makes the fold costlier even for the full-length A␤, but the additional 22 residues ultimately induce the hairpin fold in the final fibrillar aggregates.
Together, the TEM, ssNMR, IR, and tyrosine cross-linking data show that S ⌬G25 fibrils consist of straight monomers assembled in an antiparallel intermolecular arrangement. We note that molecular dynamic simulations of the phosphorylated Ser-26 and the S26D mutants, which cannot form the salt bridge present in the wild type, also show a tendency to open up at the turn region (19). The "Iowa" mutant D23N, which does not form a salt bridge, also shows an antiparallel arrangement in the protofibrils (9). In summary, the major fraction of S ⌬G25 monomers seem to aggregate without adopting the most prominent structural feature of S and the full-length A␤ peptide, viz., the hairpin turn.

Conclusion
Our results show the central role that the turn region plays in determining the aggregation propensity of A␤. A deletion of a glycine residue may in general be expected to have a relatively small effect on a hydrophobically driven aggregation process. However, it leads to a steric crowding of the turn region, prevents the peptide from bending, and makes the peptide at least 20 times more soluble. Interestingly, small oligomers still form, perhaps driven by the nonspecific intermolecular interactions of the hydrophobic residues. However, the oligomers have random coil architecture and cannot easily grow in size, whereas the oligomers of S and full-length A␤ have at least some ␤-sheet character, and they easily grow to form mature fibrils (7,11,18). The final fibrillar architecture of S ⌬G25 , obtained only at very high concentrations, shows that ␤-sheets do form, but the hairpin opens up, and the ␤-sheet is anti-parallel in nature. Surprisingly, the distal contact between Phe 19 and Leu 34 is still present, suggesting that forming this contact provides considerable energy stabilization for aggregation. Candidate drug molecules with the ability to crowd the turn region or to disrupt the contact between Phe 19 and Leu 34 are expected to have a very strong effect on the aggregation of A␤.