a, We examined the secondary structure content for nine protein pairs using far-ultraviolet circular dichroism. The spectra of wild-type proteins are displayed as black dashed lines and those of mutants as continuous red lines. The measurements were taken at 25 °C. Each pair was compared using the same buffer conditions (Supplementary Table 3). Most of the mutants exhibited similar or identical circular dichroism profiles when compared with their wild-type counterpart, indicating that the content in secondary structure was identical or similar between them. Only the mutants 2WCV (E77Y), 1FRW (D170L/D173L/K175L/D176L) and 1D7A (K11L/E22L/E25L/D158L) showed major differences, but all showed a spectrum with negative ellipticity values in the 210–230 nm range, whereas positive values are expected for random coil. These data indicate that mutants retain a folded structure. b, Stability measurement curves of five wild-type and mutant pairs forming fibres in vivo. Protein stability was assessed for five pairs following the ellipticity at 220 nm from 20 to 85 °C at a heating rate of 1 °C min−1. Wild-type proteins are displayed as black dashed lines and those of mutants as continuous red lines. Each pair was compared using the same buffer conditions (Supplementary Table 3). None of the proteins fully unfolded in the temperature range probed. Thus, we measured the ellipticity of the samples in 2.5 M guanidinium chloride at 90 °C, which was taken as a relative unfolded state (maximal ellipticity, θmax). The ellipticity of the samples at 20 °C was taken as a relative folded state (minimal ellipticity, θmin). We show the normalized ellipticity (θnorm) defined as θnorm = (θT − θmin)/(θmax − θmin), where θT is the ellipticity measured at temperature T.