Classification of unique INP tandem arrays into either WO-coil or R-coil subdomains.

A) AlphaFold 2 model of PbINP coloured by domain. Purple: N-terminal domain, pink: flexible linker, green: water-organizing (WO) coils, blue: arginine-rich (R) coils, yellow: C-terminal cap. The inset shows a cross section through the solenoid coil. B) 16-residue tandem repeat forming one coil from the β-solenoid with positions numbered from N-to C-terminus. C) The number of repeats in the WO-coils and R-coils for each unique sequence. PbINP is included and indicated in orange. n = 121. D) Sequence logos constructed from each 16-residue repeat present in the dataset.

Ice nucleation activity of mutant INPs in which R-coils are incrementally replaced with WO-coils.

A) Diagram indicating the domain map of PbINP and from it, the design of the INP mutants. Sections of the WO-coils used in the replacement of the R-coils are indicated. wt: wild type. B) Ice nucleation temperatures measured by WISDOM for E. coli cells expressing PbINP and mutants with different numbers of R-coils. The temperature at which fifty percent of the droplets froze (T50) is indicated with a hollow circle, and its corresponding value written nearby. The shaded region indicates standard deviation. C) The same constructs assayed for ice nucleation using a nanoliter osmometer. The grey box indicates temperatures beyond the lower limit of the NLO apparatus for detecting ice nucleation in this experiment.

Ice nucleation activity of mutant INPs with entirely relocated or deleted R-coil subdomain.

A) Diagram indicating the design of the constructs. 11 of the 12 R-coil repeats were either moved within the construct or deleted. b) Freezing curves with T50 and number of unfrozen droplets indicated where applicable.

Site-specific mutagenesis of noteworthy motifs in the R-coil subdomain.

A) Diagram indicating the design of the constructs. Translucent bars indicate continuity of three conserved motifs along the length of the central repetitive domain. Neg. charge: Negative residues present in positions 11, 12, and 14 of the repeat. TxT: Thr-Xaa-Thr motif in positions 6-8 of the repeat. Y-ladder: An entirely conserved Tyr is position 3 of the motif. Three mutants were created in which these motifs were extended into the R-coils. K-coils: All arginines in the R-coils were replaced with lysines. B) Freezing curves with T50 and number of unfrozen droplets indicated where applicable.

A) A comparison of the nucleation temperatures of PbINP when assayed using intact E. coli cells and when assayed with filtered supernatant. B) A box and whisker plot showing the nucleation temperatures of filtered supernatant containing PbINP under different pH conditions. Boxes and bars indicate quartiles, with medians indicated by a centre line. Outliers are indicated by diamonds.

A) Measured freezing temperatures of heat-treated droplets containing either PbINP or PbINP with the first 32 repeats (counting from the N-terminal end) deleted. Wild-type PbINP freezing without heat treatment (kept at roughly 10 °C) is indicated-on the left. B) Fluorescent microscopy images of recombinant E. coli cells expressing PbINP tagged with GFP viewed under bright-field (BF) or fluorescent excitatory (GFP) light. Representative images are shown (n = 3). Note: cells that retain their fluorescence after 75 °C treatment are rarely observed.

A) Sites of N-terminal truncations to PbINP, indicating the location of the starting residue in the shortened construct. B) Alignment of representative INP C-terminal domains from the genus Pseudomonas. Mutated residues and their one-letter codes are indicated above. Symbols at the bottom indicate consensus (* for fully conserved,: for conservation of strongly similar chemical properties, for conservation of weakly similar chemical properties). C) Predicted location of mutated residues in the PbINP C-terminal cap with sidechains shown and predicted H-bonds for D1208 shown as dashed lines. D) The ice nucleation curves for the N-and C-terminal cap mutants.

Fibrous bundles observed by cryo-FIB and cryo-ET in in E. coli cells expressing INP.

A) Ion-beam image of a thin lamella containing E. coli cells expressing INP obtained from cryo-FIB milling. B) Zoomed-in view of a cryo-TEM image of the lamella in A). Boxes with dashed-lines indicate areas where tilt series were collected. C) and D) Snapshots from 3-D cryo-tomograms reconstructed from tilt series collected in the boxed regions in B) showing striking fibrous bundles (yellow arrowheads). The E. coli cell envelopes are indicated with thick dash-lines. E) Further examples of the fibrous bundles produced by INP-expressing E. coli. Size markers in A) is 10 μm, in B) is 2 μm and in C), D) and E) are 100 nm, respectively.

Filamentous multimer model for bacterial INPs.

A) A possible assembly of INP solenoids to form long fibres composed of antiparallel INP dimers (indicated by orange and yellow pairs). B) Dimers are formed along the tyrosine ladder, a previoiusly proposed dimerization interface. They are joined end to end by forming electrostatic interactions between negatively (red) and positively (blue) charged surfaces. All threonines are coloured light green, displaying the arrays of TxT WO-motifs. The termini of the INP solenoids are labeled N and C and coloured to match panel A. This illustration uses a manually flattened AlphaFold model of PbINP. C) Cross sections of the model at positions indicated in A. Monomers are rotated approximately 90° to each other and dimerized along their tyrosine ladders (purple). Toward their termini, a pair of dimers can be matched by oppositely charged electrostatic surfaces (teal).

AlphaFold shows high confidence in overall fold of the model.

A) Predicted location dependent difference test (pLDDT) shows the residue-by-residue confidence of the model generated by AlphaFold. Low values may indicate low-confidence or intrinsic disorder. B) Predicted aligned error (pAE) plots indicate the confidence in the relative orientation of the models. The x-and y-axes indicate residue position of the model, with low (blue) values indicating high confidence and high (red) indicating low. Rigid domains often appear as squares along the diagonal axis.

E. coli expressing INP mutant lacking R-coils show no fibre clusters as observed in those cells overexpressing wild-type INP.

A-D) Representative snapshots from 3-D cryo-tomograms showing cytoplasmic and extracellular features of various E. coli cells overexpressing an INP mutant in which all but the C-terminal R-coil have been replaced by WO-coils. All four images are in the same scale and the scale bar represents 100 nm.

Flowchart and quality control steps in sequence selection for bioinformatic analysis.

Ten known INPs from literature were used to generate a consensus sequence for WO-coils which was then used as a query in a BLAST against NCBI’s non-redundant protein database to identify INPs. NCBI E-Utils were used to generate a data set using only genes from long-read DNA sequencing data.