Protein aggregates nucleate ice: the example of apoferritin

Biological material has gained increasing attention recently as a source of ice-nucleating particles that may account for cloud glaciation at moderate supercooling. While the ice-nucleation (IN) ability of some bacteria can be related to membrane-bound proteins with epitaxial fit to ice, little is known about the IN-active entities present in biological material in general. To elucidate the potential of proteins and viruses to contribute to the IN activity of biological material, we performed bulk freezing experiments with the newly developed drop freezing assay DRoplet Ice Nuclei Counter Zurich (DRINCZ), which allows the simultaneous cooling of 96 sample aliquots in a chilled ethanol bath. We performed a screening of common proteins, namely the iron storage protein ferritin and its iron-free counterpart apoferritin, the milk protein casein, the egg protein ovalbumin, two hydrophobins, and a yeast ice-binding protein, all of which revealed IN activity with active site densities > 0.1 mg−1 at −10 ∘C. The tobacco mosaic virus, a plant virus based on helically assembled proteins, also proved to be IN active with active site densities increasing from 100 mg−1 at −14 ∘C to 10 000 mg−1 at −20 ∘C. Among the screened proteins, the IN activity of horse spleen ferritin and apoferritin, which form cages of 24 co-assembled protein subunits, proved to be outstanding with active site densities > 10 mg−1 at −5 ∘C. Investigation of the pH dependence and heat resistance of the apoferritin sample confirmed the proteinaceous nature of its IN-active entities but excluded the correctly folded cage monomer as the IN-active species. A dilution series of apoferritin in water revealed two distinct freezing ranges, an upper one from −4 to −11 ∘C and a lower one from −11 to −21 ∘C. Dynamic light scattering measurements related the upper freezing range to ice-nucleating sites residing on aggregates and the lower freezing range to sites located on misfolded cage monomers or oligomers. The sites proved to persist during several freeze–thaw cycles performed with the same sample aliquots. Based on these results, IN activity seems to be a common feature of diverse proteins, irrespective of their function, but arising only rarely, most probably through defective folding or aggregation to structures that are IN active.