Time-resolved (ultra) SAXS (ID02, ESRF):

Small Angle X-Ray Scattering (SAXS)21 measurements were performed on the Time-Resolved Ultra Small-Angle Scattering beamline ID02 at the ESRF, Grenoble, France. The incident X-ray energy was 12.46 keV and two sample-detector distances were employed: 1.5 m (SAXS) and 10 m (USAXS). SAXS data were acquired using the Rayonix MX-170HS detector with exposure times between 0.01 and 0.03 seconds at room temperature (20 °C).

Pre-hydrolysed TEOS was added to 10 ml of 1 mg/ml TTCF solution at 1:50 (v/v) ratio at pH 7 ex situ, initiating the ensilication reaction. Using a sterile syringe, 1 ml of this ensilication mixture was injected into a quartz capillary (figure S2) after which the beamline hutch was checked for safety and closed for the start of measurement. We measured the delay on hutch closure to be approximately 1 minute. As a result, the total time delay between the start of the reaction and the first measurements was between 1 and 2 minutes.
The measured 2D patterns, after normalisation by incident flux, sample transmission, and solid angle, were azimuthally averaged to obtain the 1D static scattering profiles as a function of the magnitude of the scattering vector q=4π/λ sin(θ/2),  with λ the incident X-ray wavelength (=0.996 Å-1) and θ the scattering angle. This gave two overlapping q-ranges of 0.0008≤q≤0.008  and 0.006≤q≤0.5 Å-1. The scattering background in each case was measured using Tris buffer and the normalised background subtracted data are represented by I(q). 

Fitting of SAXS data:

Data fitting was done using several models within SASview to probe the various changes observed in the scattering signal. Good residual fits were found using a combination of power law, ellipsoid, broad peak and mass fractal models at different stages of the ensilication process. The ellipsoid model provided shape information (the polar and equatorial radii, Rpolar and Requatorial respectively) on the protein and the initial growth of its silica coating. The broad peak model gave  a characteristic length scale for scattering consistent with the particle sizes from the ellipsoid fits, and was utilised as a transition model towards the mass fractal growth. The latter provided the fractal radius of silica particulates and fractal dimension, Rfrac and Df respectively, which are indicative of reaction type. All models were assessed on χ2 as a goodness-of-fit indicator. Detailed information about the fitting parameters are presented in the Supplementary Information.

in situ time-resolved SAXS (i22, Diamond Light Source):

Small Angle X-Ray Scattering (SAXS) measurements during ensilication in situ were performed on the Time-Resolved Small-Angle Scattering beamline i22 at Diamond Light Source, Didcot, UK.
TTCF at 1 mg/ml in (50 mM Tris pH 7.0) buffer, 25 ml volume, was circulated using a peristaltic pump at 2 ml / min through Teflon tubing with an internal diameter of 1.6 mm. The sample solution passed through a 1.5 mm capillary flow cell in a loop before addition of hydrolysed TEOS via a syringe injector. Both pump and injector were remotely controlled (figure S3). 
SAXS frames were taken at 1 frame/sec for 120 seconds. Pre-hydrolysed TEOS was added to the sample at 1:50 (v/v) ratio after 3 seconds from start of measurement, so that data on the native protein could be acquired before ensilication began. The incident X-ray energy was 12.4 keV. SAXS data were acquired using the Pilatus P3-2M detector at 2.2 m sample distance with 0.8 seconds exposure time. The collected 2D data were processed using a pipeline setup in the Data Analysis WorkbeNch (DAWN) software27. The pipeline was set up with detector calibration, SAXS mask, Poisson error, time with flux and transmission correction. Azimuthal integration produced 1D data in I vs q, where q=4π/λ sin(θ/2), with λ the incident X-ray wavelength (0.9998 Å) and θ the scattering angle. The q range was between 0.008≤q≤0.75 Å-1. 
The experiment was performed at room temperature (20 °C). Background subtraction scattering was done using a double subtraction method (figure S4 & table S1). Empty capillary scattering was subtracted from Tris buffer scattering. The sample SAXS was then processed by subtracting Tris buffer (minus capillary scattering) and capillary scattering only (table S1).