Metaproteome workflow and taxonomic assignment of protein groups

Proteins from red snow and dark ice samples collected in summer 2021 from the southern margin of the GrIS (Supplementary Fig. 1a-e; "Methods") were extracted in triplicate and analyzed with liquid chromatography with tandem mass spectrometry (LC-MS/MS). MS spectra were analyzed using a predicted protein database compiled from GrIS metatranscriptomes from samples collected during separate field campaigns. Our metaproteomics workflow was successful for both sample types, with a total of 7517 protein groups (PGs) identified, and 6256 PGs quantified (identified in at least two replicates). The snow and ice samples showed minimal overlap in shared peptides (12%, Fig. 1a) and quantified PGs (19%, Fig. 1b), indicating a strong distinction in their peptide and protein identities. In addition, the large and relatively comparable number of peptides and PGs identified in both sample types confirmed that the database used catered well for both sample types (Fig. 1a, b).

Taxonomic assignment of PGs revealed that the majority of the PGs (>2500) across all samples belonged to Viridiplantae in both sample types (Fig. 1c, d). Snow samples were predominantly composed of chlorophyte PGs (>2200 PGs, ~77% total abundance), while ice samples were dominated by streptophyte PGs (>2300 PGs, ~80% total abundance). In both sample types, ~150 of the Viridiplantae PGs were not assigned to a specific phylum ("Viridiplantae other", Fig. 1c, d). Based on 18S amplicon sequencing data, algal taxa in the snow samples were dominated by chlorophyte species, primarily from the Chloromonas genus, with minor contributions from Sanguina spp. (Fig. 1e, Supplementary Fig. 2a, b). In contrast, ice samples were dominated by the streptophyte Ancylonema spp. (~80%; Fig. 1e, Supplementary Fig. 2a, b). This finding was corroborated by imaging, cell counts, and biovolume quantification, allowing us to clearly distinguish and identify multiple cell types for the chlorophyte snow algae (indicating different life stages of the same or different species) and unicellular and filamentous Ancylonema spp. (Fig. 1f, Supplementary Fig. 3). The relative chlorophyte vs. streptophyte abundance from our 18S amplicon and biovolume datasets mirrors the trends found for protein in both sample types (Fig. 1g). Fungal PGs were also abundant in snow (~5%) vs. ~0.5 in ice, consistent with the general trend found with our 18S sequencing (Supplementary Fig. 2a, b). We also quantified prokaryotic PGs (with bacterial taxa identified via 16S amplicon sequencing, Supplementary Fig. 2c, d), as well as metazoan and other classified and unclassified PGs in both sample types (Fig. 1c, d). One amoebozoan PG was identified in ice, and two viral PGs were found in snow. We subsequently filtered for all chlorophyte PGs in snow and streptophyte PGs in ice to compare their functional proteome composition, and hereafter we refer to these as the chlorophyte proteome and the streptophyte proteome, with individual PGs expressed relative to the total amount of weighted spectra for each proteome.

We annotated all protein sequences with GO terms, KEGG pathways, and INTERPRO domains, allowing a functional comparison between proteomes. Between the chlorophyte and streptophyte proteomes, the relative abundance of high-abundance, essential functional groups such as the GO terms ATP binding and translation, and the KEGG pathways Biosynthesis of secondary metabolites and Carbon metabolism (Supplementary Figs. 4a, b and 5, Supplementary Data 2 and 3) did not significantly differ. Only a few of the quantified functional groups differed between the two taxonomic groups (Fig. 2, Supplementary Data 2 and 3, Supplementary Fig. 4a, b).

Purine synthesis PGs were more abundant in the chlorophyte proteome with the GO terms purine nucleobase biosynthetic process (~10× more abundant) and the purine nucleotide binding and IMP biosynthetic process (Fig. 2a), as well as functional domains (assigned with INTERPRO) such as NAD/GMP synthase and Adenylosuccinate lyase also being at least 10× more abundant (Fig. 3a). The metabolomes of the same samples (see "Methods") revealed that indeed the purines hypoxanthine and adenine accounted for ~4% and 2% of the snow metabolome, respectively, but much less than 1% each in the ice metabolome (Supplementary Fig. 6a). In contrast, PGs linked to RNA metabolism were enriched in the streptophyte proteome, indicated by an increase in PGs annotated with GO terms RNA helicase activity and RNA processing, and the KEGG pathway Spliceosome (Fig. 2c, d). Furthermore, individual PG annotations revealed an enrichment in GUCT domain-containing proteins typical for certain RNA helicases (Fig. 3b). We also found a high abundance of one PG (>0.1%) with the annotation RNA-induced silencing complex, nuclease component Tudor-SN in the chlorophyte proteome, homologous to a plant protein known to inhibit the degradation of specific mRNAs in response to stress (Fig. 3b). The streptophyte proteome also had a notable abundance of PGs linked to protein degradation (with GO term metallopeptidase activity and KEGG pathway Ubiquitin mediated proteolysis) and individual PG annotations (including the FtsH protein, involved in the turnover of photosystem II proteins; Fig. 2c, d, Supplementary Fig. 7a). PGs involved in endocytosis, and phagocytosis were also enriched in the streptophyte proteome (Fig. 2c, d). The PGs with the GO terms structural constituent of cytoskeleton and microtubule were almost ~10× more abundant in the streptophyte proteome (Fig. 2c), largely due to the contribution of tubulin (Supplementary Fig. 7b).

Our functional analysis also indicated that PGs linked to signal transduction were more abundant in the streptophyte proteome, with the enrichment of the GO terms regulation of DNA-templated transcription (over ~10× more abundant) and signal transduction, and the KEGG pathway MAPK signaling pathway-plant (~6× more abundant; Fig. 2c, d). We identified numerous transcription factors (TFs) in both proteomes (Supplementary Fig. 8a), including Cold Shock Domain TFs, with the TF annotation Zinc finger, CCHC-type enriched in the streptophyte proteome. One PG with an ice-binding domain was quantified in an ice sample, homologous to the Ancylonema ice-binding protein identified by Procházková et al.. However, this PG could not be classified beyond "cellular organisms" by MEGAN (accession 149312, Supplementary Data 1). Using reciprocal BLASTp, we identified multiple photoreceptors, including the UV-sensing UVR8 PG in the streptophyte proteome (Fig. 3c). A high-abundance light-sensitive rhodopsin PG (identified with INTERPRO) also represented >0.1% of the streptophyte proteome (Fig. 3c). The chlorophyte proteome had only one photoreceptor: phototropin.

The most abundant protein in all samples was the large RuBisCO subunit (rbcL), invariably classified as chlorophyte in the snow samples and streptophyte in the ice sample (Supplementary Fig. 9). Overall, photosynthetic PGs were not found to be enriched in either proteome (Supplementary Fig. 5a, b). Both proteomes had high abundances of different carbonic anhydrases (annotations Carbonic anhydrase and Limiting CO2-inducible protein B/C, beta carbonyic anhydrase domain) involved in the conversion of carbon dioxide into bicarbonate (Fig. 3d). Based on GO term analysis, Photosynthesis, light harvesting PGs, made up of Chlorophyll A-B binding proteins were over three times more abundant in the chlorophyte proteome (Fig. 2a), with other photosynthesis PGs also found to be ~10× more abundant in the chlorophyte proteome (Fig. 3d). In contrast, the streptophyte proteome had ~10× more RuBisCO small subunit (rbcS) PGs and PGs associated with the chloroplast GO term (Figs. 2c and 3d). Finally, total Reactive Oxygen Species homeostasis PGs with the GO term oxidoreductase activity did not significantly differ between proteomes (Supplementary Fig. 5a), although individual PG functions for oxidative stress response differed between taxa (Supplementary Fig. 7c).

Proteins involved in lipid and fatty acid metabolism were more abundant in the chlorophyte proteome, as indicated by the GO terms fatty acid biosynthetic process and O-acyltransferase activity, as well as the KEGG pathways Fatty acid metabolism, Biotin metabolism, and Glycerophospholipid metabolism (Fig. 2a, b). PGs with a Glycerol-3-phosphate dehydrogenase NAD-dependent domain were ~100× more abundant in the chlorophyte proteome (Fig. 3e). The chlorophyte proteome also had a >0.1% abundance of the putative mitochondrial glycerol-3-phosphate dehydrogenase activity (Alpha-glycerophosphate oxidase, C-terminal) (Fig. 3e). Glycerol-3-phosphate was also noticeably more abundant by a similar order of magnitude in the snow metabolome, with no clear difference in the amount of cellular glycerol (Supplementary Fig. 6b). The chlorophyte proteome also contained highly expressed PGs annotated as Wax synthase and Glucose-methanol-choline oxidoreductase, all linked to lipid and fatty acid metabolism (Fig. 3e). Glucose-methanol-choline oxidoreductases have been linked to hydrocarbon production in algae and photosynthetic efficiency in the cold. However, the snow metabolome was not enriched in the hydrocarbon metabolites covered in our analysis (Supplementary Fig. 6c). In terms of carbohydrate synthesis, a PG with the annotation GDP-mannose 4,6 dehydratase, involved in the production of fucose, was highly abundant (~0.5%) in the streptophyte proteome. In contrast, two PGs with the annotation dTDP-4-dehydrorhamnose reductase family, linked to the production of rhamnose, constituted 0.2% of the chlorophyte proteome (Fig. 3f). While a difference in fucose abundance in the metabolite data was not evident, we found that rhamnose was ~10× more enriched in the snow metabolome (Supplementary Fig. 6d). Four PGs with a FAS1 domain annotation represented 0.1% of the chlorophyte proteome, which are predicted to have a role in extracellular cell adhesion or be homologous to the chlorophyte water soluble astaxanthin-binding carotenoprotein (Fig. 3f). In terms of polysaccharide degradation, we found that over 1.5% of the streptophyte proteome was made up of glycosyl hydrolases responsible for the hydrolysis of O-glycosyl compounds (compared to 0.2% for the chlorophyte proteome; Fig. 2c).

PGs linked to organic and inorganic nutrient homeostasis were found in both proteomes. Both taxa were characterized by a similar abundance of transmembrane proteins (~10% of both proteomes, Supplementary Fig. 8b). However, we found a high PG abundance of the GO term transmembrane transporter activity in the streptophyte proteome (~0.65% of proteome, vs. ~0.2% in the chlorophyte proteome; Fig. 2c). Such enriched proteins included those with domains involved in the transport of inorganic ions (i.e., Potassium transporter and Ccc1 family) or organic compounds (i.e., Amino acid/polyamine transporter I; Fig. 3g). Proteins involved in nitrogen metabolism were enriched in the chlorophyte proteome, mainly linked to glutamine and glutamate (glutamic acid) synthesis (Figs. 2a and 3g). Our metabolite data showed that glutamate, glutamine, and pyroglutamic acid all contributed ~10× more to the chlorophyte metabolome (Supplementary Fig. 6e). Snow algae had an enrichment of the annotation Dihydropyrimidine dehydrogenase domain II, linked to pyrimidine breakdown. Consistent with these metabolome results, PGs with the domain Glutamate decarboxylase, responsible for the conversion of glutamate to γ-aminobutyrate, were only present in the streptophyte proteome (Fig. 3g). Further striking differences between the two proteomes include, for example a higher PG abundance linked to Tryptophan metabolism (Fig. 2d) in the streptophyte proteome, while the chlorophyte proteome had a large amount of ferritin PGs (ferritin PGs were quantified in the ice samples but were not associated to streptophyte algae; Fig. 3g, Supplementary Data 1). The chlorophyte proteome had a high abundance of PGs with a Pyrophosphate-energized proton pump domain (over 60x more abundant; Fig. 3g), as well as PGs linked to sulfur metabolism with the GO term sulfate assimilation (Fig. 2a), including three PGs with the domain Sulphate adenylyltransferase (Fig. 3g).

The primary photoprotective pigment produced by Ancylonema spp., purpurogallin, is a hydrolysable tannin. The synthesis pathway of purpurogallin has yet to be described, but the early steps of the shikimate pathway may be essential for its production. We identified enzymes of the shikimate pathway using reciprocal BLASTp, and found that most proteins were present in both proteomes (Fig. 4a). We found a >10× higher relative abundance of the second (DHQS, 3-dehydroquinate synthase) and third (DHQ/SDH, bifunctional 3-dehydroquinate dehydratase/shikimate dehydrogenase) enzyme of this pathway in the streptophyte proteome (Fig. 4a). We also document the high expression of one UDP-glucosyltransferases (UDPGT) in the streptophyte proteome at a similar abundance to DHQS and DHHQD/SD (Fig. 4a). UDPGTs are essential in the production of hydrolysable tannins in other plant species. Conversely, for the carotenoid pigments typical of chlorophyte snow algae, both GO and KEGG analyses indicated an enrichment for PGs involved in terpenoid and carotenoid biosynthesis in the chlorophyte proteome (Fig. 2a, b). Reciprocal BLASTp identified multiple PGs belonging to the terpenoid synthesis methylerythritol phosphate pathway and carotenoid synthesis in both proteomes, with many >10× more abundant in the chlorophyte proteome (Fig. 4b). PGs involved in carotenoid cleavage and the production of apocarotenoids, Carotenoid Oxygenases, were ~100× more abundant in the chlorophyte proteome (Fig. 4b). In contrast, the only apocarotenoid quantified in the metabolome, a Vitamin A fatty acid ester, was not enriched in the snow metabolome (Supplementary Fig. 6f).