An ice-binding protein from the glacier ice alga Ancylonema nordenskioeldii

As plants began to colonize the land c. 470–450 million years ago, they had to overcome many abiotic stresses not experienced by their marine ancestors (Rensing, 2018). One such stress was freezing and thawing, which can damage plant cell walls. Bacteria, which had established their presence on land well before the arrival of plants, greatly aided the transition of plants to land by the horizontal transfer of key genes (Yue et al., 2012; Ma et al., 2020). Bacteria are likely donors of genes that can mitigate freeze–thaw injury, as proteins with ice-binding activity have been found in several bacteria (Raymond et al., 2007, 2008; Vance et al., 2018). Each of the proteins in those studies contains a c. 200-aa domain called DUF3494 that has a beta solenoid structure with one side that binds to ice crystals (Vance et al., 2019). At very low concentrations, these proteins can drastically prevent the recrystallization of ice that occurs during thawing, which is thought to damage cell walls. Similar proteins have been identified in hundreds of species of bacteria, although not all of them have been examined for ice-binding activity. Horizontally acquired genes of this type appear to be the source of freeze–thaw tolerance in a number of algae and fungi that live in icy habitats (Raymond & Kim, 2012).

Zygnematophyceae, being the closest known relatives of all land plants (Cheng et al., 2019), are of interest because of their remarkable ability to find solutions for the stresses encountered by early land plants (Kunz et al., 2024). In this vein, Bowles et al. (2024) recently obtained the metagenome of algae inhabiting the Morteratsch Glacier in Switzerland to investigate the adaptation of the early streptophyte alga Ancylonema nordenskioeldii to life in ice. Among the survival mechanisms investigated, the authors looked for genes encoding ice-binding proteins (IBPs). They found several candidates in the protein kinase superfamily, ATP-binding cassette protein family and heat shock protein family, although none were confirmed to have ice-binding activity. Apparently, they did not see our earlier paper in which we identified an IBP in A. nordenskioeldii (AnIBP) from the Morteratsch Glacier (Procházková et al., 2024). Here, we summarize the main findings of this paper, in which we showed that AnIBP has ice-binding activity and that this activity could be attributed to a protein of the DUF3494 family.

The Ancylonema species A. nordenskioeldii and A. alaskanum dominate the algae on Morteratsch Glacier, although their relative abundances can vary drastically from year to year (Procházková et al., 2024). Other chlorophyte algae are found in relatively small numbers.

Initially, a sample of A. nordenskioeldii was found to have ice-binding activity, that is, to cause distortions in the form of pitting and faceting in a growing ice crystal (fig. 3 of Procházková et al. 2024). The activity was destroyed by heat treatment, strongly suggesting that the active substance was a protein. Furthermore, the sample retained activity after three cycles of ice-affinity purification (Fig. 1).

We obtained a metagenome of the Morteratsch algae to search for the ice-active protein. A problem with finding genes for IBPs in a metagenome from an icy habitat is that such metagenomes likely include many IBP genes from bacteria and fungi. We found a few DUF3494-type genes, most of which could be attributed to bacteria or fungi. The source of one of these genes remained a puzzle until recently submitted sequence data from a Closterium species (Zygnematophyceae) (Sekimoto et al., 2023) unambiguously showed that it belonged to A. nordenskioeldii. The predicted protein (AnIBP, GenBank WHL30856) consisted of an N-terminal signal peptide, a DUF3494 domain, a beta sandwich domain and a C-terminal root cap domain. The root cap domain is a curious feature as it is mostly found in root cells of vascular plants, where they are involved in sensing and interacting with the environment (Kumpf & Nowack, 2015). The sequence of AnIBP based on the data of Bowles et al. is 100% identical to the one we obtained from our metagenome. The gene has a length of 515 aa and a predicted molecular mass of 54.5 kDa. No homologs of the DUF3494 domain of AnIBP were found in our metagenomes of A. nordenskioeldii and A. alaskanum, the metagenome of Bowles et al., or in other Zygnematophyceae in GenBank. Ancylonema alaskanum appears to use glycerol instead of an IBP as a cryoprotectant (Procházková et al., 2024).

The structures of the DUF3494 and beta sandwich domains were confidently predicted by AlphaFold3, as shown by low predicted local distance difference test (pLDDT) values. Here, we show the updated structures using AlphaFold4, which have somewhat lower pLDDT values but virtually the same structures (Fig. 2). The protein has an ice-binding side populated with an ordered array of hydrophilic threonine and serine residues whose spacings were close to the a-axis repeat distance in ice. The 3D structure of the protein can be seen more clearly in Supporting Information Video S1. Furthermore, a 2D electrophoretic gel of the ice-affinity-purified sample shown in Fig. 1 had a prominent spot of the expected molecular weight and pI of AnIBP (fig. 5 of Procházková et al. 2024).

A phylogenetic tree of DUF3494 domains shows AnIBP's relations to its most closely related proteins from archaea, bacteria, fungi and algae (Fig. S1). The closest algal protein is an IBP from a snow alga (Chloromonas brevispina), but the bootstrap values of all the nearest neighbors are low, leaving AnIBP's origin unclear. A bacterial origin seems more likely (see Discussion).

To examine the relative expression of AnIBP, the complete sequences of two common housekeeping genes, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and ribosomal protein L13a (RPL13a), were assembled from our metagenome data and the RNAseq and PacBio data of Bowles et al. (Notes S1). Five 100-nt sequences were obtained from each gene and used to query the RNAseq data of Bowles et al. (The conditions under which the RNAseq data were obtained were not stated, but it seems likely they were obtained under daytime glacier conditions.) These data showed that AnIBP was expressed over 12 times more strongly than GAPDH and over eight times more strongly than RPL13a (Table S1).

So far, over 1000 proteins with DUF3494 domains have been identified, virtually all of them in microorganisms. They can vary enormously in size, with some having two or more DUF3494 domains and many others, such as AnIBP, having other domains of unknown function. Many DUF3494 proteins are found in nonfreezing habitats, in which DUF3494 domains may bind to other substrates or even individual water molecules as a way of mitigating desiccation. The ice-binding sides of DUF3494 proteins have numerous TXT motifs (where X is any amino acid), which are known to bind water molecules (Graether & Sykes, 2004).

Some of the key genes that allowed the Zygnematophyceae to survive desiccation and other stresses on land were most likely acquired from soil bacteria (Cheng et al., 2019). This appears to be the case for the DUF3494 domain of AnIBP as it contains no introns and does not closely match any algal sequences in the databases. Its most closely matching sequences are from a lichen from Alberta, Canada (MCJ1363104), and a bacterium found on an Antarctic moss (ALG05169), both having only c. 50% amino acid sequence identities to AnIBP. The lichen gene also has no introns, suggesting that it was also acquired from a bacterium. On the other hand, the beta sandwich domain has no homologs in bacteria but dozens of homologs in Closterium algae, suggesting that it is a characteristic feature of Zygnematophyceae.

A number of IBPs have been identified in land plants, especially grasses (e.g. Kuiper et al. 2001). They also have the form of a beta solenoid, but one that significantly differs from the ones in DUF3494 proteins. Ancylonema nordenskioeldii, on the other hand, belongs to a group of streptophytes, the Zygnematophyceae, which are the closest known relatives of all land plants (Cheng et al., 2019). It thus seems likely that the grass-type IBPs developed later in the evolution of streptophytes, and that the best IBPs available early in the evolution of streptophytes were bacterial DUF3494 proteins.

None declared.

LP, DR and LN designed the original study and carried out the collections. JAR analyzed the IBP data and wrote the paper.

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.