Functional aggregation of cell-free proteins enables fungal ice nucleation

Significance Organisms have evolved efficient molecular strategies to control the nucleation and growth of ice. Although these strategies have developed independently across biological kingdoms, they all seem to use proteins to construct extended functional domains. While for bacteria, the use of large proteinaceous units to build superior ice-making complexes is established, the identity of fungal INs remains unknown. Here, we demonstrate that small extracellular proteins, of which over a hundred are capable of assembling in cell-free environments, make up the fungal INs that enable ice formation at warm temperatures. Our findings highlight that nature uses a common strategy, E pluribus unum (out of many, one), to enable high subzero ice nucleation temperatures by assembly of ice-nucleating proteins into large functional aggregates.


Filtration Experiments
Fig. S5 shows freezing curves of F. acuminatum IN solutions after subsequently passing them through 50 and 30 kDa molecular weight cut-off (MWCO) filters.Filtrations decreased the cumulative number of INs per gram of mycelium and shifted the initial freezing temperature towards lower temperatures.As shown in Fig. S5A, a very significant ice nucleation activity remained (Nm(T) ~−5 °C) even after passing through a 30 kDa filter.By comparing the Nm plots and the fraction of ice of the filtered and unfiltered solutions (Fig. S5B), we find that the number of Fusarium INs after passing through a 50 kDa filter, appears similar to the 10 -5 dilution of the initial sample.The similarity of filtered and diluted samples is further supported by dynamic light scattering experiments that show comparable hydrodynamic radii of ~200 nm for different dilutions of the Fusarium extracts and filtered samples (Fig. S5C).The finding that ice nucleation activity of Fusarium INs persists in the filtrates provides additional evidence that the aqueous extracts of Fusarium INs consist of smaller units which aggregate to larger icenucleating complexes in solution.However, it is important to note, that the molecular weight cut-off (MWCO) in filters are nominal classifications and not precise boundaries (1).Typical filters contain a broad range of pore sizes, making it impossible to achieve 100% retention of even very large molecules, and thus will detort conclusions drawn solely based on them.

Estimation of Protein Subunits in the Aggregates
We estimate the area of the 5.3 kDa protein assuming, for simplicity, that is has a globular fold.
The approximate radius of a 5 kDa protein is 1.1 nm (2), resulting in a projected area Am = π*1.1 2 nm 2 = 3.8 nm 2 for the monomer.The number of the required subunits that enable freezing is then calculated by dividing the value obtained through the HUB and HINT analysis by Am.This calculation assumes that the IN assembly consists of a single well-packed layer of the ice nucleating monomers.As such, and considering that we assumed that the IN are as good as ice at promoting ice nucleation and that the IN surfaces are square, the values reported are a lower bound for the actual number of monomers in the IN aggregates of F. acuminatum.

Dynamic Light Scattering
Dynamic light scattering (DLS) measurements were performed on an ALV spectrometer consisting of a goniometer and an ALV-5004 multiple-tau full-digital correlator (320 channels), which allows measurements over an angular range from 30° to 150°.A He-Ne Laser (wavelength of 632.8 nm) was used as the light source.Measurements were performed at 20 °C at 9 angles ranging from 30° to 150°.The hydrodynamic radii (Rh) of ~1 mg/mL filtered F. acuminatum solutions were determined using DLS.The Rh of the smaller and larger fractions of F. acuminatum was found to be similar.

Amino Acid Analysis.
Amino acid analysis was performed by the Molecular Structure Facility at UC Davies as described elsewhere (3) and the used samples were ice-affinity purified and filtered.S1.

Fig. S1 .
Fig. S1.Fraction of ice (fice) for different dilutions of aqueous extracts containing INs from spores and mycelial surfaces from F. acuminatum.The presented data corresponds to the Nm plot shown in Fig. 1A of the main text.

Fig. S2 .
Fig. S2.SDS PAGE gel of the crude Fusarium extract, ice-purified INPro and molecular weight markers.Lanes A-E show different purified and filtered Fusarium IN samples.The impurities that are visible in the crude extract were removed in the purified fractions and a band at <~10 kDa remains.

Fig. S3 .
Fig. S3.Secondary structure contents of the Fusarium INPro as determined by BeStSel (3, 4).BeStSel is a webserver for protein secondary structure prediction and fold recognition from circular dichroism spectra.The analysis shows that the Fusarium INPro have a ~29% antiparallel β-sheet and a ~12% helical content.

Fig. S4 .
Fig. S4.Cumulative number of INs per unit mass of F. acuminatum (Nm) for extracts containing INs from spores and mycelial surfaces.The red line represents the optimized differential spectrum obtained through the HUB-backward code assuming that it is a combination of one (A) two (B) or three (C) Gaussian subpopulations.Normalized distribution function that represents the corresponding differential freezing spectrum nm(T).The mean square error (MSE) between the experimental and predicted Nm(T) decreases from 14.4% to 3.4% to 1.1% as the number of subpopulations increase from one to three.

Fig. S5 .
Fig. S5.Size determination of fungal ice nucleators (INs) from F. acuminatum upon filtration (A) Cumulative number of INs per unit mass of F. acuminatum (Nm) for extracts containing INs from spores and mycelial surfaces and samples that were passed through 50 kDa (magenta diamonds) and 30 kDa (green circles) filters.(B) Fraction of ice for different dilutions of INs from F. acuminatum and the highest concentration of the sample that was passed through a 50 kDa filter.(C) Hydrodynamic radii for different dilutions of fungal INs and of samples passed through 50 kDa (magenta) and 30 kDa (green) filters.Radii were determined using dynamic light scattering and error bars represent the standard error of the measurements.

Fig. S6 .
Fig. S6.Freezing experiments of aqueous extracts containing fungal INs from F. acuminatum.(A) Cumulative number of INs per unit mass of F. acuminatum (Nm) for the largest (~660 kDa) SEC fraction.(B) Cumulative number of INs per unit mass of F. acuminatum (Nm) for the smallest (~12 kDa) SEC fraction.We estimate the concentrations of the SEC fractions to be ~1.6 mg/mL for the larger and 0.3 mg/mL for the smaller fraction.The differential spectrum nm(T) that best fits the smaller fraction has three subpopulations.The modes of the distributions are -6.4,-8.5, and -11.8 °C for the small (~12 kDa) SEC fraction and the mean square error (MSE) between the experimental and predicted Nm(T) is 1%.(C) The difference of the logarithm of    and the logarithm of    as a function of temperature.(D) The differential freezing spectrum with three subpopulations estimated using the HUB-backward code.

Fig. S8 .
Fig. S8.Impact of freeze-thaw cycles and sampling on the activity of IN from F. acuminatum.(A) Cumulative number of INs per unit mass of F. acuminatum (Nm) for the same sample after twelve freeze-thaw (FT) cycles.(B) Differential spectra (nm) corresponding to the distribution of ice nucleating temperatures of the IN along these FT cycles.The modes, spread, and weights of the subpopulations of these FT-treated samples are shown in TableS1.

Table S2 .
Minimum ice nucleating area number of 5.3 kDa protein units that would nucleate ice at a given temperature Thet.The minimal areas are computed with the numerical implementation of classical nucleation theory for finite surfaces implemented in the HINT algorithm, assuming that the IN binds ice as strongly as ice itself and that the IN surfaces are squares.The minimum number of protein units is estimated as the ratio of the minimum IN area by 3.8 nm 2 , the estimated cross section of a globular 5.3 kDa protein.