Arrangement of subunits and ordering of -strands in an amyloid sheet

Amyloid fibrils are associated with several disease states, but their structures have yet to be fully defined. Here we use site-directed spin labeling to explain some of the specific interactions that are formed between subunits when the protein transthyretin (TTR) assembles into amyloid fibrils, which are associated with both spontaneous and familial amyloid diseases in humans. The results suggest that fibrils are formed when a major conformational change displaces the terminal -strand from the edge of a -sheet in the native structure, exposing the penultimate strand. The newly exposed strand then allows a novel -sheet interaction to form between the TTR subunits. This interaction and another previously identified subunit association lead to a plausible model for the specific sequence of -strands in one of the indefinitely repeating -sheets of TTR amyloid, which is formed by a head-to-head, tail-to-tail arrangement of subunits.

Amyloidoses are disorders of unknown etiology that are characterized by the extracellular deposition of stable, insoluble rod-like fibrils formed from normally soluble proteins¹. The self-assembly of conformationally altered proteins into amyloids is closely associated with a variety of clinically significant neuro-degenerative conditions, including Alzheimer's disease, the prion-related transmissible spongiform encephalopathies and diabetic neuropathies^{2, 3}. There are >20 biochemically distinct amyloidogenic proteins, which show no significant sequence similarity but form fibrils that appear grossly similar. Amyloid fibrils from different sources share a common core structure believed to consist of continuous -sheets lying parallel to the long axis of the fibril, with the constituent -strands running perpendicular to this axis⁴. To understand amyloid assembly better, finer structural details are required. In particular, specific structural data are needed to clarify the -sheet interactions that occur between subunits to drive amyloid formation.

Figure 1. The three-dimensional structure and self-assembly of transthyretin. a, The native tetrameric structure of TTR, containing a primarily -sheet structure17, undergoes structural rearrangements that give rise to amyloidogenic intermediates that self-assemble into protofibrils. Current models of protofibrils depict a core of extended -sheets organized such that the sheets are parallel to the long axis of the protofibril, with the constituent -strands perpendicular to this axis4. An example of one of the extended intermolecular -sheets is shown here for clarity. b, The dimeric interface of the native tetrameric structure consists of two intermolecular eight-stranded -sheets, CBEFF'E'B'C' and DAGHH'G'A'D'. The F−F' interface has been shown to be preserved or recovered during the formation of fibrils13. The highlighted amino acid residues are the focus of site-directed spin labeling studies presented here, and residues 29, 31, 40 and 46 are labeled for reference (residue 33 is highlighted but not labeled). This figure and Fig. 3c were generated using PDB entry 2PAB17 and RIBBONS36. c, Electron micrographs of spin-labeled mutants 33C (left) and 40C (right) reveal extended linear fibrils typical of amyloid. Full Figure and legend (113K)

Various models have been proposed to describe some of the relatively large-scale features of transthyretin amyloid fibrils, such as the manner in which several extended -sheets of a protofibril might twist together¹². However, models differ with respect to the number of sheets in a given protofibril and the distinction between a parallel or antiparallel arrangement of strands. Moreover, how the subunits come together or what specific molecular interactions lead to continuous -sheets has not yet been made clear. A partial answer to this question was provided by a recent study¹³ that identified one of the subunit interfaces in TTR fibrils as similar to one of the natural dimeric interchain associations evident in the structure of the soluble protein (Fig. 1b). Although TTR is observed to dissociate into a mixture of monomers and dimers during in vitro amyloidogenesis, this native-like dimeric interface is apparently retained or reformed during fibril assembly. This arrangement, in which the F -strands (Fig. 1b) from two subunits are hydrogen-bonded to each other, provides the semblance of intermolecular -sheets consistent with the arrangement envisioned for amyloid fibrils. However, because this dimeric arrangement is symmetric — that is, 'head-to-head' — at least one additional subunit interaction ('tail-to-tail') would be required to build a filamentous assembly from such a dimeric building block. In this study, we have probed transthyretin amyloid fibrils for such additional interfaces between TTR subunits.

The present survey for subunit interactions in TTR fibrils focused on the strands near the edge of one of the native intermolecular -sheets (the B and C -strands) that, upon self-assembly, could give rise to an extended -sheet characteristic of amyloid fibrils. The likelihood that TTR subunits might associate through interactions between such edge strands has been pointed out by others^{7, 14, 15, 16}. However, the specific strands that would constitute the edge strands has not been explained. We addressed this question using site-directed spin labeling (SDSL) to probe for interactions between specific amino acid residues in the fibrillar state.

If like residues from different subunits come into proximity in the fibrillar state, as they would in a symmetric arrangement of -strands, then such an interaction should be apparent in SDSL experiments. When spin-labeled residues are sufficiently close to each other (separated by 5−25 Å), dipolar-coupled spins produce a drop in electron paramagnetic resonance (EPR) signal amplitude that can be used to estimate the distance between modified residues (see Methods). Using this approach, single-site mutants of TTR were derivatized with one discrete spin label on each subunit, thus allowing us to determine the distance between equivalent residues on different subunits of TTR. The EPR spectra of various spin-labeled single-Cys mutants of TTR (Fig. 1b) were recorded before and after fibril formation to identify new interactions formed specifically in the fibrillar state. A total of 14 single-site mutants were prepared, of which 7 could be satisfactorily modified with spin label and still form amyloid fibrils (Fig. 1c). Ultimately, five of the mutants showed EPR changes upon fibril formation and, therefore, were informative in the following analysis.

In control experiments on proteins in solution, the EPR spectra of spin-labeled mutants of TTR confirmed that the B -strands from different subunits are not in close proximity before fibril formation, as seen in the crystal structure (Fig. 1b). For example, mutant 29C has no substantial difference between the spectra obtained from fully labeled (black spectrum) and magnetically dilute (red spectrum) samples; there is no spin−spin interaction in the soluble state (Fig. 2a). This is consistent with the known structure of soluble TTR (Fig. 1b), in which residue 29 resides at widely separated locations in the oligomeric structure¹⁷, with a 30 Å 'intradimer' distance between residues on subunits that form a dimer through intermolecular -sheets, and a 35 Å 'interdimer' distance between residues on subunits from distinct dimers in the native tetramer. For residue 31, the closest separation in the native tetramer is 24 Å, and, again, no spin−spin interaction is measured in the soluble state (Fig. 2b). For residue 33, weak spin−spin interactions are detected, corresponding to a 21 Å interspin distance, which is consistent with its being closer to the natural dimer interface (Fig. 2c).

Figure 2. Site-directed spin labeling of B strand residues of TTR. The spin labeling data are shown for a, mutant 29C; b, mutant 31C; and c, mutant 33C. Here and in Fig. 3, the spectra from fully labeled (black spectra) and magnetically diluted (red spectra) samples were acquired at room temperature over 200 G and normalized relative to first derivative amplitudes. d, The fibrillar state internitroxide distances are most easily explained by a new intermolecular interface generated by an antiparallel arrangement between the B -strands from two different subunits of distinct building blocks. This new subunit interaction involves the strand B of one building block (X) hydrogen bonded to the strand B' of a different building block (Y). The coloring of subunits corresponds to that in Fig. 1b. Full Figure and legend (43K)

EPR spectra were then obtained from suspensions of fibrils prepared from these spin-labeled mutants to see if any mutants revealed a gain of spin−spin interaction in the fibrillar state. A comparison of the spectra obtained before and after fibril formation clearly shows that new interactions are gained upon fibril assembly. Although characterized by large interspin distances in the soluble state, spin-labeled mutant 29C self-assembles such that like residues from distinct subunits come within 12 Å of each other in the fibril (Fig. 2a). Similarly, the spectra for mutant 33C indicate a decrease in interspin distance to 14 Å during fibrillogenesis (Fig. 2c). The spectra for the single Cys mutant 31C are particularly striking. They suggest that the side chains of residue 31 from different subunits come within 8 Å of each other upon fibril formation (Fig. 2b). Because the B -strands from different subunits are widely separated in the soluble state, the close proximity of these strands in the fibrillar state identifies a new subunit interaction.

The SDSL results suggest that the B and B' -strands from two different building blocks come together during fibrillogenesis to create a new intermolecular -sheet contact. In the fibrillar state, the B -strand of one building block seems to be hydrogen-bonded to the B' -strand of a different building block (Fig. 2d). This new interface creates an antiparallel arrangement between the B and B' -strands, with the center of symmetry near residue 31. This subunit organization explains the short interspin distance in the fibrillar state for position 31, with slightly longer distances for positions 29 and 33.

Formation of a novel intermolecular interface involving the B and B' -strands (see above) would require a conformational change to expose these penultimate strands of the native -sheet. Such structural rearrangements would most likely involve the mutation-prone region of the C and D -strands^{14, 18} (described as the CD region). Spin labeling studies involving the CD region were undertaken to demonstrate that structural changes occur in this region upon fibril formation. We report the solution and fibrillar state EPR spectra of two CD region mutants with nitroxides incorporated on the C and C' -strands. As expected, the spectra for mutant 40C clearly illustrate strong spin−spin interactions before fibril formation (Fig. 3a). The measured inter-nitroxide distance of 11 Å in the soluble state correlates with the 'intra-dimer' separation of 12 Å observed in the known crystal structure¹⁷. In contrast, the spectra of the mutant in fibrillar form reveal a lack of dipolar-coupled spins, indicating that a significant increase in inter-residue distance (from 11 to >25 Å) at position 40 occurs upon fibrillogenesis. This result is consistent with earlier solution studies on amyloidogenic intermediates that have implicated significant rearrangements in the CD region during amyloidogenesis^{7, 15}.

Figure 3. Site-directed spin labeling of C strand residues of TTR. The spin labeling data are shown for a, mutant 40C and b, mutant 46C. c, Ribbon diagrams of TTR, indicating the suspected conformational change in the CD region (red) that results in exposure of the B and B' -strands and allows novel subunit associations to extend the native intermolecular -sheet. The perspective on the right, which is similar to that in Fig. 1b, arises when the dimer on the left is viewed from above. Full Figure and legend (54K)

A detailed analysis of the EPR spectra for TTR mutant 46C suggests that although a portion of the C and C' -strands from different subunits come into relative proximity during amyloidogenesis, this does not interfere with the formation of a new interface between the B and B' -strands. Before fibril formation, the spectra for the single-Cys mutant 46C indicate no spin−spin interaction (Fig. 3b), consistent with the closest distance being 33 Å between corresponding residues in the soluble tetramer. After self-assembly, spin-labeled mutant 46C results in fibrils with spins separated by 16 Å. Despite this modest approach of strands C from different building blocks, the more intimate association between residues 29−33 indicates that the the B strands form the intermolecular -sheet between different building blocks and that the C strands move away to enable this interaction (Fig. 3c). The approach of the C -strands from different building blocks probably occurs as part of an additional subunit interface rather than as part of the intermolecular -sheet identified here.

The new subunit interface identified in the present study involves the other edge of the intermolecular -sheet. Structural changes in the CD region, facilitated by either acid denaturation or mutagenesis, apparently cause the original edge strand C to move away and expose the penultimate strand B of the sheet (Fig. 4a). The newly exposed B strand leads to the second intermolecular association. When combined, these two distinct symmetric subunit interfaces give rise to a head-to-head and tail-to-tail arrangement of subunits and a continuous intermolecular -sheet (BEFF'E'B')_n of indefinite length (Fig. 4b). Our observation that the assembly of transthyretin subunits into amyloid is promoted by a loss of structure at the edge of a native -sheet is consistent with recent studies suggesting a similar event occurs during the self-assembly of ₂-microglobulin^{19, 20}. Whether displacement of a terminal -strand from the edge of a native -sheet is a general mechanism of amyloidogenesis remains to be seen.

Figure 4. Arrangement of subunits and ordering of -strands in TTR fibrils. a, The native tetrameric structure of TTR (shown here as a monomer for clarity) is perturbed in the CD region (red) during amyloidogenesis, resulting in exposure of the B or B' -strand and giving rise to an intermediate prone to self-assembly. b, Upon fibrillogenesis, interactions between the F and F' -strands (green dyads) and between the B and B' -strands (blue dyads) result in two distinct symmetric (head-to-head and tail-to-tail) subunit associations. According to the model, TTR subunits are arranged to produce a continuous intermolecular -sheet with strands arranged in the order (BEFF'E'B')n. c, A rendition of the proposed amyloid -sheet structure in which the different colors correspond to distinct subunits, each contributing three -strands (BEF) to the sheet. Several of such sheets are believed to compose the core of a protofibril Full Figure and legend (38K)

The SDSL results presented here do not completely rule out all other models for the arrangement of strands in the amyloid -sheet under investigation. Other models consistent with the experimental results could be fashioned by invoking major structural rearrangements relative to the native structure. However, a spin-labeling survey of residues along the central E and E' -strands suggests that these strands undergo little, if any, conformational change upon fibril formation (data not shown). This suggests that the antiparallel arrangement of the B, E and F -strands is probably maintained in TTR fibrils. One might also argue that the B and B' strands could come into close proximity without being hydrogen-bonded to each other as part of a -sheet — for example, through interactions between distinct protofibrils. Nevertheless, although recognizing that other scenarios might be possible, the model presented here seems to be the simplest one that is consistent with the SDSL data and with our current understanding of amyloid structure.

The present study on TTR fibrils also brings into focus the significance of studies of TTR in solution and in crystals. Recent experiments have highlighted the loss of interactions between -strands B and C in soluble amyloidogenic intermediates^{15, 21}. In the recent crystal structure of a highly amyloidogenic TTR mutant, Eneqvist and co-workers¹⁶ have reported a striking shift in register between the B and C -strands. The native interactions between strands B and C are easily disrupted, in agreement with the present finding that strand B becomes exposed during amyloidogenesis.

Although the present study provides a compelling model for one of the indefinitely long -sheets of TTR amyloid protofibrils, a description of other such sheets in the protofilament is not yet available. It is possible that a similar antiparallel arrangement may exist in the form of an extended -sheet consisting of strands (AGHH'G'A'). We have attempted to investigate such putative subunit interfaces that might arise from -strand interactions involving this other sheet in the native structure, but have been unsuccessful because of insufficient spin labeling (see Methods). Nonetheless, this study does provide experimental evidence to support a specific arrangement of strands in one extended -sheet of an amyloid protofibril.

Note added in proof: The findings presented here are consistent with a recent study by Richardson and Richardson³⁵, which emphasizes how strands at the edges of -sheets have evolved to prevent natural proteins from aggregating.

The sulfhydryl groups in the Cys mutants were derivatized with a methanethiosulfonate spin label (MTSSL) to form a paramagnetic side chain. Spin labeling of TTR mutants was carried out as described¹³ with 90% efficiency, which was determined by ESI-MS. However, Cys mutants along the H strands were unsatisfactorily labeled, most likely because of secondary steric constraints imposed by the dimer−dimer interface of native tetrameric TTR. The EPR spectra of the almost fully labeled TTRs (black spectra) were obtained, both before and after incubation at fibril-forming conditions, to identify dipolar-coupled nitroxide residues if present. First derivative absorption spectra were recorded on a Varian E-109 X-band spectrometer fitted with a loop-gap resonator. Spectra were collected at room temperature (20−22 °C) by signal averaging 16 scans over 200 G (1 G = 0.1 mT), using a microwave power of 2 mW and a modulation amplitude optimized to the natural line width of each individual spectrum.

	Top

1. Sipe, J.D. Annu. Rev. Biochem. 61, 947−975 (1992). | Article | PubMed | ISI | ChemPort |
2. Perutz, M.F. Curr. Opin. Struct. Biol. 6, 848−858 (1996). | Article | PubMed | ISI | ChemPort |
3. Kelly, J.W. et al. Adv. Protein Chem. 50, 161−181 (1997). | PubMed | ISI | ChemPort |
4. Sunde, M. et al. J. Mol. Biol. 273, 729−739 (1997). | Article | PubMed | ISI | ChemPort |
5. Westermark, P., Sletten, K., Johansson, B. & Cornwell, G.G. Proc. Natl. Acad. Sci. USA 87, 2843−2845 (1990). | PubMed | ChemPort |
6. Kelly, J.W. Curr. Opin. Struct. Biol. 8, 101−106 (1998). | Article | PubMed | ISI | ChemPort |
7. Lai, Z., Colon, W. & Kelly, J.W. Biochemistry 35, 6470−6482 (1996). | Article | PubMed | ISI | ChemPort |
8. Lashuel, H.A., Lai, Z. & Kelly, J.W. Biochemistry 37, 17851−17864 (1998). | Article | PubMed | ISI | ChemPort |
9. Olofsson, A. et al. J. Biol. Chem. 276, 39592−39599 (2001). | Article | PubMed | ISI | ChemPort |
10. Serpell, L.C. et al. J. Mol. Biol. 254, 113−118 (1995). | Article | PubMed | ISI | ChemPort |
11. Serpell, L.C. et al. J Mol. Biol. 300, 1033−1039 (2000). | Article | PubMed | ISI | ChemPort |
12. Blake, C.C. & Serpell, L.C. Structure 4, 989−998 (1996). | Article | PubMed | ISI | ChemPort |
13. Serag, A.A., Altenbach, C., Gingery, M., Hubbell, W.L. & Yeates, T.O. Biochemistry 40, 9089−9096 (2001). | Article | PubMed | ISI | ChemPort |
14. Serpell, L.C., Goldsteins, G., Dacklin, I., Lundgren, E. & Blake, C.C. Amyloid 3, 75−85 (1996). | ISI | ChemPort |
15. Liu, K., Cho, H.S., Lashuel, H.A., Kelly, J.W. & Wemmer, D.E. Nature Struct. Biol. 7, 754−757 (2000). | Article | PubMed | ISI | ChemPort |
16. Eneqvist, T., Andersson, K., Olofsson, A., Lundgren, E. & Sauer-Eriksson, A.E. Mol. Cell 6, 1207−1218 (2000). | Article | PubMed | ISI | ChemPort |
17. Blake, C.C., Geisow, M.J., Oatley, S.J., Rerat, B. & Rerat, C. J. Mol. Biol. 121, 339−356 (1978). | Article | PubMed | ISI | ChemPort |
18. Eneqvist, T. & Sauer-Eriksson, A.E. Amyloid 8, 149−168 (2001). | PubMed | ISI | ChemPort |
19. McParland, V.J., Kalverda, A.P., Homans, S.W. & Radford, S.E. Nature Struct. Biol. 9, 326−331 (2002). | Article | PubMed | ISI | ChemPort |
20. Hoshino, M. et al. Nature Struct. Biol. 9, 332−336 (2002). | Article | PubMed | ISI | ChemPort |
21. Liu, K. et al. J. Mol. Biol. 303, 555−655 (2000). | Article | PubMed | ISI | ChemPort |
22. Ekiel, I. & Abrahamson, M. J. Biol. Chem. 271, 1314−1321 (1996). | Article | PubMed | ISI | ChemPort |
23. Garzon-Rodriguez, W., Sepulveda-Becerra, M., Milton, S. & Glabe, C.G. J. Biol. Chem. 272, 21037−21044 (1997). | Article | PubMed | ChemPort |
24. Friedhoff, P., Schneider, A., Mandelkow, E.M. & Mandelkow, E. Biochemistry 37, 10223−10230 (1998). | Article | PubMed | ISI | ChemPort |
25. Enya, M. et al. Am. J. Pathol. 154, 271−279 (1999). | PubMed | ISI | ChemPort |
26. Liu, Y., Gotte, G., Libonati, M. & Eisenberg, D. Nature Struct. Biol. 8, 211−214 (2001). | Article | PubMed | ISI | ChemPort |
27. Janowski, R. et al. Nature Struct. Biol. 8, 316−320 (2001). | Article | PubMed | ISI | ChemPort |
28. Knaus, K.J. et al. Nature Struct. Biol. 8, 770−774 (2001). | Article | PubMed | ISI | ChemPort |
29. Padilla, J.E., Colovos, C. & Yeates, T.O. Proc. Natl. Acad. Sci. USA 98, 2217−2221 (2001). | Article | PubMed | ChemPort |
30. Sunde, M. & Blake, C. Adv. Protein Chem. 50, 123−159 (1997). | PubMed | ISI | ChemPort |
31. Rabenstein, M.D. & Shin, Y.K. Proc. Natl. Acad. Sci. USA 92, 8239−8243 (1995). | PubMed | ChemPort |
32. Altenbach, C., Oh, K.J., Trabanino, R.J., Hideg, K. & Hubbell, W.L. Biochemistry 40, 15471−15482 (2001). | Article | PubMed | ISI | ChemPort |
33. Pake, G.E. J. Chem. Phys. 16, 327−336 (1948). | Article | ISI | ChemPort |
34. Gross, A., Columbus, L., Hideg, K., Altenbach, C. & Hubbell, W.L. Biochemistry 38, 10324−10335 (1999). | Article | PubMed | ISI | ChemPort |
35. Richardson, J.S. & Richardson, D.C. Proc. Natl. Acad. Sci. USA 99, 2754−2759 (2002). | Article | PubMed | ChemPort |
36. Carson, M. Methods Enzymol. 277, 493−505 (1997). | Article | ISI | ChemPort |

	Top

Competing interests statement:  The authors declare that they have no competing financial interests.