(a) Proximity analysis assessed by excimer fluorescence in fibres carrying pyrene labels at single sites. Fibres were assembled at 25°C with 2.5μM NM (60 % labelled), either by gently rotating the samples (•), or by adding 4% sonicated seed () formed at 25°C. The ratio of excimer fluorescence/non-excimer fluorescence (I₄₆₅/I₃₇₅) is plotted.
(b) Excimer fluorescence in fibres assembled from mixtures of two different pyrene-labelled cysteine variants. Fibres were assembled at 25°C with gentle rotation, in reactions containing equimolar mixtures of each variant (25 % labelled and 75% unlabelled). Actual excimer values are provided in Fig. S3. Head and Tail region residues are marked.
(c) Effects of cross-links on amyloid assembly. Right, proteins cross-linked under denaturing conditions with either a 2Å disulfide bond (black bars) or a flexible 11Å BMB cross-link (grey bars) were diluted 200 fold into buffer to a final concentration of 2.5μM NM and assembled with rotation at 25°C. Left, thioflavinT, ThT, values for WT fibres, soluble NM, and a denatured standard (bovine serum albumin treated with DTT to induce denaturation). Values represent means ± SD (n=3 experiments)

a) Formation of a collapsed intermediate: An increase in acrylodan fluorescence at positions N21 (), Q38 (), G43 (), Y77 (), G96 (), Y106 () occurred very rapidly after proteins were diluted into buffer, before the formation of amyloid (also see Fig. S4). Amyloid formation was assessed by SDS insolubility because acrylodan and ThT fluorescence could not be analysed in the same reaction. Acrylodan fluorescence for mutants T158C () and K184C () in the M domain and ThT binding profiles of WT NM () are plotted for comparison.
(b) Susceptibility of different cysteines to spontaneous cross-linking during assembly under non-reducing conditions. Crosslinks were assessed by non-reducing SDS-PAGE.
(c) Addition of a single charge in the Head region severely impedes fibre assembly. Assembly of negatively charged iodoacetate () and uncharged iodoacetamide () labelled proteins was monitored by ThT fluorescence.
(d) Assembly kinetics of cysteine variants cross-linked with BMB at position N21C (▴), G25C (•), Q38C (▪), Wt NM (), G96 (), Y106C () and G43C () and Y73C () monitored by ThT fluorescence.

(a) NM sequence, showing residues mutated to cysteine in green, the five imperfect oligopeptide repeats shaded in grey, and the boundary between N and M as it is traditionally drawn. By aa composition, the boundary might equally well be drawn at aa 138.
(b) Accessibility of cysteine residues after assembly. Fibres, formed from single cysteine substitution mutants were labelled with pyrene maleimide for 3 hours. The ratio of labelled/unlabelled protein is plotted as a measure of accessibility. For all proteins labelling approached 100% in denaturant (6M GdmCl; data not shown),
(c) Denaturation profiles. Mutants were individually assembled at 25°C, using 75% unlabelled and 25% acrylodan-labelled protein. Each preparation was then equilibrated for 30 minutes at 24 different concentrations of GdmCl and fluorescence spectra were recorded.

One model for NM assembly is provided to illustrate the constraints that our data place on the nature of NM fibre structures. In the cooperatively folded amyloid core, Head residues in one NM molecule are in close proximity to Head residues of their neighbours; the same is true for the Tail (3). Central Core residues are sequestered from intermolecular interactions. M is largely unstructured, but the segment of M proximal to N becomes structured when amyloid forms. In the initial stages, NM molecules rapidly acquire a collapsed state (1) that retains a molten character until the Head regions of two molecules (2) come into proximity with each other and nucleate assembly. We do not know how fibrils are arranged within fibres (4).

a) Amyloid core length and stability differ in fibres assembled at 4°C and 25°C. Acrylodan-labelled fibres assembled at 4°C () and 25°C () were denatured as in Fig. 1b. D_1/2 values were obtained from full GdmCl denaturation profiles (see Fig 1c).
(b) Intersubunit interfaces change in fibres assembled at different temperatures. Excimer fluorescence of pyrene-labelled mutants assembled in rotated unseeded reactions at 25°C (), 4°C ( or in seeded unrotated reactions with 4% seed formed at 4°C (). Pyrene fluorescence is similar in seeded and unseeded reactions.
(c) BMB cross-links at different positions bias assembly towards distinct prion strains. NM proteins cross-linked in denaturant were diluted into buffer, assembled into fibres, and used to transform [psi⁻] cells to [PSI⁺] state. Left, controls: uncross-linked WT fibres assembled at RT or at 4°C are biased towards the production of weak or strong strains, respectively27. Soluble NM, buffer alone, and proteins cross-linked in the Central Core did not induce [PSI⁺] above background and were not scored. Values represent means ± SD (n=5 experiments)
(d) Strain biases created by cross-linking overcome the biases created by assembly at different temperatures. Whether they were assembled at 4°C or at RT, fibres cross-linked in the Head (G25C or G31C) produced mostly strong [PSI⁺] strains, and fibres cross-linked in the Tail (G96C and106C) produced mostly weak [PSI⁺] strains. Values represent means ± SD (n=5 experiments)