Unveiling microbial communities and biogeochemical cycles in Antarctic colored snow Authors: Xiaohan Ai, Ze Ren, Changzi Liu, Chunmei Zhang, Huirong Li, Haitao Ding, Yong Yu, Wei Luo, Yonghong Bi Published: 2026 ABSTRACT Snow cover, the extensive terrestrial habitat in Antarctica, sometimes exhibits vivid coloration, yet the structure and function of its microbial communities remain poorly characterized. Using metagenomic sequencing of red snow (RS) and green snow (GS) from the Fildes Peninsula, we found that bacterial, eukaryotic, and archaeal relative abundances were 85.82%, 13.52% and 0.16%, respectively. β-Diversity differed significantly between RS and GS across these three domains (P < 0.05). Dominant bacterial phyla included Bacteroidota (RS: 62.61%; GS: 38.72%) and Pseudomonadota (RS: 32.80%; GS: 54.10%). Among eukaryotes, Chlorophyta (RS: 58.10%; GS: 52.98%) and Basidiomycota (RS: 14.80%; GS: 8.08%) were prevalent. Nanobdellota dominated archaea, with lower abundance in RS than GS. In the algal community, Sanguina, Gonium and Chloromonas were significantly enriched in red snow, while Chlorella and Micractinium were enriched in green snow (P < 0.05). Marker genes associated with carbon (C), nitrogen (N), phosphorus (P) and sulfur (S) cycles were identified in green and red snow. Aerobic respiration and phosphate regulation were significantly enriched in red snow, while CO oxidation, fermentation, and denitrification were significantly enriched in green snow. Key microbial genera associated with these functional pathways also varied. In the denitrification of red snow, Stutzerimonas was the most abundant genus, while Janthinobacterium was abundant in green snow. Nitrification-related genes were detected only in red snow based on the present metagenomic data. The network of the red snow microbial community was potentially more complex and resistant based on topology, which not only benefited its own long-term survival but might also have potentially influenced the positive feedback effect of snowmelt by maintaining a low-albedo snow surface. This provided an ecological implication under climate warming: the expansion of red snow patches showed the potential to the increase nitrate runoff export, which would affect nitrogen nutrient levels in coastal Antarctic waters. Overall, this study used metagenomics to compare the multidomain (bacteria, archaea and eukaryotes) composition and diversity between red snow and green snow, and directly linked key microbial taxa with functional genes of biogeochemical cycles. This study provided new insights into the biological characteristics and functional potential of Antarctic colored snow.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12866-026-05306-y. Introduction Surface snow in polar regions serves as a crucial link between the atmosphere and the underlying soil or ice, creating unique ecosystems for various organisms [1]. Through adaptation to extreme cold, radiation and nutrient limitation, snow microbes proliferate and produce conspicuous pigmentation that yields colored snow, including green and red, and even orange, pink and yellow. The occurrence and distribution of colored snow are closely associated with local environmental conditions [2]. Colored snow directly influences snow albedo, which refers to the reflectivity of snow surfaces. Climate models now increasingly incorporate this “biological albedo” effect due to its significant impact on the estimation of snowmelt rate. Snow albedo reductions of the order of 13% have been reported during the snowmelt season [2–4]. By lowering surface reflectivity, snow microbes can accelerate melting. For example, glacier algae have been estimated to account for as much as 75% of the variability in rapid thawing observed on the southwest Greenland ice sheet [5, 6]. Following snowmelt, algal biomass frequently dominates surface communities and represents a crucial source of nutrients [7–9]. The average carbon content of red snow was estimated at 1.2 kg/km2 on the Harding Icefield, Alaska [10]. The seasonal total dry biomass of green algae blooms along the Antarctic Peninsula amounted to approximately 1.3 × 10³ tonnes, covering a total area of 1.95 km2 [8]. It highlights the importance of colored snow in local nutrient cycling. Microbial communities in snow can influence nutrient transformations, such as algal autotrophic production and bacterial ammonia oxidation [11]. Therefore, it is important to gain a deeper understanding of how microbial diversity and function vary among different colored snows. Colored snow hosts diverse and metabolically active assemblages including algae, bacteria, archaea and fungi [12–14]. Green snow is mainly associated with nitrogen and amino acid metabolism, with the dominant genus being Chlorella in the algal community [15]. In contrast, red snow is primarily dominated by Chloromonas and associated with carbohydrate and fatty acid metabolism [15]. Algae provide organic carbon via photosynthesis to heterotrophic bacteria, while bacteria can supply minerals and growth factors that stimulate algal growth [16]. The bacterial community is linked with Bacteroidota and Pseudomonadota. Fungal community is predominantly comprised of Basidiomycota [17, 18]. The composition of snow algal communities varies among colors, although no significant disparities are observed in the composition of the associated fungal and bacterial communities, which also demonstrate variations in relative abundance [15, 18]. There have been relatively few studies on colored snow archaeal communities. The archaeal communities in the glacier of Svalbard, Nitrososphaeraceae is the most abundant [19]. Currently, most studies have primarily focused on the diversity and composition of bacterial, fungal, or archaeal communities separately. However, research on the associations and functions of the entire microbial communities including bacteria, eukaryotes, and archaea domains remains limited. Recent advancements in molecular biology and metagenomics have enhanced our understanding of functional genes and specific metabolic pathways in microbial communities [20–23]. Metagenomics, for example, has been applied to compare autotrophic carbon fixation pathways across different environments and to analyze variations in denitrification genes among microbial communities [24, 25]. Colored snow represents a ubiquitous yet underexplored habitat across Antarctica. Recent study [26] identified key genes and metabolic pathways related to carbohydrate metabolism in green and red snow. However, an in-depth understanding of the interactions among microbial domains (bacteria, eukaryotes, and archaea) and their coupling to specific biogeochemical pathways is still lacking. Investigating microbial communities in colored snow, such as community composition and element cycle-related functions, is vital for understanding their roles in nutrient cycling and primary production, which are key to elucidating the structure and function of Antarctic terrestrial ecosystems, especially amid environmental changes. Metagenomic analyses were conducted on seven green snow and six red snow samples collected from the Fildes Peninsula, Antarctica. Our research focuses on two main objectives: (i) examining the composition and co-occurrence network interactions of microbial communities in green and red snow in Antarctica, and (ii) investigating the similarities and disparities in biogeochemical cycling processes and microbes involved in green and red snow. Unlike previous studies that focused only on a single community or a single metabolic pathway, this work employed multidomain comparative metagenomics strategy to systematically reveal the differences in microbial community structure and functional potential between red snow and green snow, and laid the groundwork for potential ecological effects evaluation of colored snow in Antarctic ecosystems in the future. Materials and methods Sample collection Snow samples were gathered from 13 sites across the Fildes Peninsula (Figure S1) during December 2019 and January 2020. Based on the map of the Fildes Peninsula drawn by the Chinese Antarctic Center of Surveying and Mapping at Wuhan University in 2005, the map was re-drawn in vector format using Adobe Illustrator. Each colored snow sampling location was demarcated by a 50 × 50 cm2 area and was collected into sterile plastic bottles. Subsequently, the samples were chilled at 4 °C and transported to the laboratory for further experimental analysis. These samples were six red snow (RS) and seven green snow (GS). Upon arrival, the snow samples were allowed to naturally melt at 4 °C, following which 500 mL of meltwater was filtered through 0.2 μm polycarbonate membrane filters (Whatman, UK). The filters were then preserved by freezing at -80 °C in cetyltrimethylammonium bromide (CTAB) buffer until the commencement of the analytical procedures. DNA extraction, sequencing, analysis, and annotation For DNA extraction, the cells on the filter were washed with 1 mL CTAB buffer and crushed with glass beads in a Mini-BeadBeater-16 (BioSpec, USA). Total genomic DNA was extracted from samples using the E.Z.N.A.® Soil DNA Kit (Omega Bio-tek, Norcross, GA, U.S.) according to manufacturer’s instructions. The extracted DNA was fragmented to an average size of about 400 bp using Covaris M220 (Gene Company Limited, China) for paired-end library construction. Paired-end sequencing was performed on an Illumina NovaSeq (Illumina Inc., San Diego, CA, USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) using NovaSeq 6000 S4 Reagent Kit v1.5 (300 cycles). Metagenomic Raw Sequence Statistics were presented in Table S1. The raw data were filtered and the quality control was performed. Briefly, the paired-end Illumina reads were trimmed of adaptors, and low-quality reads (length < 50 bp or with a quality value < 20 or having N bases) were removed by fsatp (v0.23.0) [27]. Metagenomics data were assembled using MEGAHIT (v1.2.9) [28], contigs with a length ≥ 500 bp were selected as the final assembling result, and then the contigs were used for further gene prediction and annotation. Open reading frames (ORFs) from each assembled contig were predicted using Prodigal (v2.6.3) [29]. The predicted ORFs with a length ≥ 150 bp were retrieved and translated into amino acid sequences using the NCBI translation table. A non-redundant gene catalog was constructed using CD-HIT (v4.6.1) [30] with 90% sequence identity and 90% coverage. High-quality reads were aligned to the non-redundant gene catalogs using Salmon (v1.10.3), and gene abundances with 95% identity were calculated using TPM (Trans Per Million) [31]. Quality-controlled sequencing data statistics, assembly and gene prediction results were presented in Table S2-S4. The TPM method includes normalization of gene length and sequencing depth for each sample, making TPM values comparable between samples. Representative sequences of non-redundant gene catalog were compared (e-value cutoff of 1e− 5) against the December 2025 release of non‑redundant protein (NR) database using Diamond (v2.0.13) with the best-hit for taxonomic annotation of domains, including archaea, bacteria, and eukaryotes. In addition, microalgae were selected to create a gene set to be analyzed separately [32]. The species information annotated by the NR database, the abundance calculated after TPM standardization, and species-abundance annotation tables were used for subsequent analysis. Functional annotation was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database with Diamond (v2.0.13) [32]. Relative abundance of pathways associated with carbon (C), nitrogen (N), sulfur (S) and phosphorus (P) cycles in microbial communities was calculated with reference to reported relevant genetic markers (KOs) [33, 34]. Relative abundance for each pathway was determined based on individual cycles, as detailed in Supplementary Material Table S5. These results revealed the functional potential of pathway associated marker genes rather than quantified abundance differences. Non-redundant gene catalogs were constructed for each pathway, with species information obtained through annotation against the NR database. After TPM normalization, species abundance table for each pathway was generated. The microbial composition and relative abundance of each pathway at the phylum and genus levels were also visualized. Statistical analysis The species composition of bacterial, eukaryotic, archaeal, and microalgal communities was used to calculate alpha diversity indices (including richness, Shannon index and Simpson index) using the “vegan” package and visualized with the “ggplot2” R package. Differences between groups were determined using the “Wilcoxon rank-sum test” and denoted with an asterisk (*) symbol when P < 0.05. Principal Co-ordinates Analysis (PCoA) was conducted over taxonomic profiles employing the abund-Jaccard distance in the “vegan” package. Given that rare taxa in the metagenomic data were readily masked, and that comparisons across multi-domain communities were subject to potential technical biases, the abund-Jaccard distance based on both species presence/absence and abundance information, better reflected the true differences in community composition. Analysis of similarities (ANOSIM) was used to assess the difference of microbial communities between green snow and red snow. Additionally, the Venn diagram representing algal species distributions was computed using the basic R package and visualized using Cytoscape (v3.9.1). The “Hmisc” and “psych” R packages facilitated the computation of robust correlations, identifying relationships with Spearman’s correlation coefficient (ρ) > 0.8 and FDR-corrected P < 0.01. Network modularity was analyzed using the “microeco” R package and visualized via Cytoscape (v3.9.1) [35]. Node-level topological properties were further assessed with the “igraph” R package by calculating the within-module connectivity (Zi) and among-module connectivity (Pi). Zi-Pi graphs, which help classify node attributes based on their topological characteristics, were generated using the “microeco” R package. According to topological characteristics, nodes can be categorized into 4 types: module hubs (Zi > 2.5 and Pi ≤ 0.62), connectors (Zi ≤ 2.5 and Pi > 0.62), network hubs (Zi > 2.5 and Pi > 0.62), and peripheral nodes (Zi ≤ 2.5 and Pi ≤ 0.62). The first three categories, excluding peripherals, are typically considered critical nodes [36]. Network robustness was assessed by simulating the deletion of nodes to calculate natural connectivity [37]. Microbial relative abundance at phylum level involved in the C, N, S and P pathways was visualized with a heatmap using the "pheatmap" R package. All statistical analyses were carried out using R (v4.0.2). Results Microbial communities The composition and diversity of the colored snow microbial communities were presented in Figure S3. No statistically significant difference in α-diversity was observed between the red and green snow communities (P > 0.05). In contrast, PCoA and ANOSIM analyses revealed a significant difference in β-diversity (P < 0.05). Bacteroidota, Pseudomonadota and Chlorophyta were the dominant phyla in red and green snow microbial communities. At the domain level, the majority of gene sequences in the metagenomes of the 13 samples were attributed to bacteria, eukaryotes and archaea, accounting for 85.82%, 13.52% and 0.16%, respectively. Due to the large differences in relative abundances between bacteria, eukaryotes and archaea, we presented them separately to describe clearly the specific community composition. The bacterial community was dominated by Bacteroidota, Pseudomonadota, and Actinomycetota, with their relative abundances differing notably between snow types. Bacteroidota was more abundant in red snow (62.61%) than in green snow (38.72%), whereas Pseudomonadota showed the opposite trend (GS: 54.10%, RS: 32.80%). Among eukaryotes, Chlorophyta was the most abundant phylum and exhibited similar proportions in both green and red snow. Basidiomycota was notably more prevalent in red snow (14.80%) than in green snow (8.08%). Within archaea, Nanobdellota was the dominant phylum in both communities, though its relative abundance was higher in green snow (24.01%) than in red (20.97%) (Fig. 1A). The Simpson index for bacteria in green snow was significantly higher than that in red snow. However, analysis of eukaryotic and archaeal richness, as well as Shannon and Simpson indices, revealed no statistically significant differences between green and red snow (P > 0.05) (Fig. 1B). Based on abund-Jaccard dissimilarity at the species level, PCoA indicated significant differences between the microbial communities of green and red snow, with ANOSIM results affirming significant distinctions in bacterial (Global R = 0.327, P < 0.05), eukaryotic (Global R = 0.284, P < 0.05) and archaeal (Global R = 0.311, P < 0.05) communities (Fig. 1C). Fig. 1Community analyses on bacteria, eukaryotes and archaea in green and red snow. A relative abundance of bacterial, eukaryotic and archaeal communities at the phylum level (> 1%) in green and red snow. B comparison of diversity indices (richness, Shannon index and Simpson index) on bacteria, eukaryotes and archaea in green and red snow, “NS” denote no significant differences (P > 0.05) based on Wilcoxon rank-sum test. C community structuring based on abund-Jaccard distance Principal Coordinates Analysis (PCoA) and Analysis of similarities (ANOSIM) Community analyses on bacteria, eukaryotes and archaea in green and red snow. A relative abundance of bacterial, eukaryotic and archaeal communities at the phylum level (> 1%) in green and red snow. B comparison of diversity indices (richness, Shannon index and Simpson index) on bacteria, eukaryotes and archaea in green and red snow, “NS” denote no significant differences (P > 0.05) based on Wilcoxon rank-sum test. C community structuring based on abund-Jaccard distance Principal Coordinates Analysis (PCoA) and Analysis of similarities (ANOSIM) To further understand the microalgal community in colored snow, we specifically analyzed its community diversity and composition. The Shannon and Simpson indices for the green snow algae community were both significantly higher than those for the red snow (P < 0.05) (Fig. 2A). The PCoA indicated that green snow and red snow exhibited distinct microalgal communities, which was further supported by ANOSIM results showing significant dissimilarities (Global R = 0.319, P < 0.05) (Fig. 2B). A total of 12 phyla of microalgae were detected in the colored snow samples, but the relative abundances of dominant phyla from green and red snow differed (Fig. 2C). In colored snow, Chlorophyta (GS: 68.95%, RS: 80.37%) and Cyanobacteriota (GS: 8.88%, RS: 5.83%) were the dominant phyla. Differences in microalgal composition and the top 35 genera between green and red snow were shown in Figure S4. Sanguina, Haematococcus, and Coccomyxa were dominant genera, Sanguina, Gonium and Chloromonas were significantly enriched in red snow, while Chlorella and Micractinium were significantly enriched in green snow (P < 0.05). The composition of microalgae differed significantly between green and red snow, with 1864 shared species and 352 and 142 unique species identified in green and red snow, respectively (Fig. 2C). The majority of unique species belonged to Cyanobacteriota, Chlorophyta and Bacillariophyta. Fig. 2Algal diversity and composition in green and red snow. A comparison of diversity indices (richness, Shannon index and Simpson index) in green and red snow algal communities. B community structuring based on abund-Jaccard distance PCoA and ANOSIM. C relative abundance and Venn diagrams of green and red snow algae communities, with each node of the Venn diagram representing a species and colors classified by phylum Algal diversity and composition in green and red snow. A comparison of diversity indices (richness, Shannon index and Simpson index) in green and red snow algal communities. B community structuring based on abund-Jaccard distance PCoA and ANOSIM. C relative abundance and Venn diagrams of green and red snow algae communities, with each node of the Venn diagram representing a species and colors classified by phylum Microbial co-occurrence network Network analysis enhanced our perspective on complex mutualistic symbiosis or competitive patterns beyond mere microbial community diversity and composition. Accordingly, we constructed co-occurrence network at the genus level using taxa exceeding 0.01% relative abundance (Fig. 3). The co-occurrence network of green snow consisted of 9680 edges and 478 nodes organized into eight modules, while the red snow network exhibited 13,844 edges and 533 nodes distributed across six modules. In both green snow and red snow co-occurrence networks, phyla with relative abundances less than 1% were defined as “Others”. The majority of microorganisms in both networks were associated with Pseudomonadota, Bacteroidota and Chlorophyta. Positive edges accounted for 71.39% of the green snow network and 75.52% of the red snow network, indicating that positive associations predominated among microorganisms in colored snow. Fig. 3Co-occurrence network of microbial communities based on genus-level Spearman correlations. Each block represents a module, each node represents a genus, and the colors of the nodes represent different phyla. Spearman’s correlation coefficient (ρ) > 0.8 and significance P < 0.01, edge colors indicate positive (red) and negative (blue) correlations Co-occurrence network of microbial communities based on genus-level Spearman correlations. Each block represents a module, each node represents a genus, and the colors of the nodes represent different phyla. Spearman’s correlation coefficient (ρ) > 0.8 and significance P < 0.01, edge colors indicate positive (red) and negative (blue) correlations By calculating a set of network-level topological features, it was determined that degree was higher in red snow, the closeness centrality and eigenvector centrality were higher in green snow (Fig. 4A). The stability of green and red snow microbial networks was evaluated by removing nodes to change the natural connectivity. The outcomes revealed that, for an equivalent proportion of nodes removed, the decline in natural connectivity within the green snow network (slope: -0.8535) exhibited a more pronounced descent compared to the red snow network (slope: -0.6776) (Fig. 4B). It suggested that red snow microbial network was potentially more complex and resistant based on network topology. Zi-Pi plots identified six and nine genera as keystone taxa in green and red snow, respectively (Fig. 4C). Notably, Flavobacterium, Undibacterium, Rhodanobacter, Ralstonia, Catalinimonas, Barnesiella were recognized as key genera within the green snow microbial networks, whereas connector genera in the red snow network included Rhodanobacter, Massilia, Flectobacillus, Roseomonas, Marinoscillum, Acidisoma, Polycyclovorans, Blepharisma and Belnapia. Fig. 4Characterization of co-occurrence network of green and red snow. A node features (degree, betweenness centrality, closeness centrality and eigenvector centrality). B natural connectivity and (C) Zi-Pi plot of microbial network by green snow and red snow. The differences were assessed using Wilcoxon rank-sum test, NS: not significant and *** P < 0.001. The topological role of each point was determined and distributed based on within-module connectivity (Zi) and among-module connectivity (Pi). Module hubs (Zi > 2.5 and Pi ≤ 0.62), connectors (Zi ≤ 2.5 and Pi > 0.62), and peripheral nodes (Zi ≤ 2.5 and Pi ≤ 0.62) Characterization of co-occurrence network of green and red snow. A node features (degree, betweenness centrality, closeness centrality and eigenvector centrality). B natural connectivity and (C) Zi-Pi plot of microbial network by green snow and red snow. The differences were assessed using Wilcoxon rank-sum test, NS: not significant and *** P < 0.001. The topological role of each point was determined and distributed based on within-module connectivity (Zi) and among-module connectivity (Pi). Module hubs (Zi > 2.5 and Pi ≤ 0.62), connectors (Zi ≤ 2.5 and Pi > 0.62), and peripheral nodes (Zi ≤ 2.5 and Pi ≤ 0.62) Microbial taxa associated with biogeochemical processes Marker genes associated with C, N, S and P were selected from the identified KOs and the relative abundance of geochemical cycling pathways was calculated. Figure S5A illustrated the differences in genetic potential between red and green snow across each pathway within various cycles. Microbial communities at phylum and genus levels involved in each pathway were also explored (Fig. 5, S5B, S6). Notably, Pseudomonadota exhibited the highest abundance, closely followed by Bacteroidota. In the carbon cycle, related genes to fermentation and CO oxidation were significantly enriched in the green snow (P < 0.05). The relative abundance of aerobic respiration-related genes in red snow (85.581%) was significantly higher than that in green snow (73.363%). Among the genera associated with aerobic respiration, Lobochlamys, Hymenobacter and Dunaliella showed higher abundance in red snow, while Hymenobacter and Herbaspirillum were more common in green snow. The main pathways of the nitrogen cycle included nitrogen assimilation and nitrogen mineralization. Denitrification was significantly enriched in green snow (2.762%) compared with red snow (0.836%) (P < 0.05), and the involved genera also differed. Specifically, the genera with relative abundances exceeding 10% in green snow were Polaromonas, Rhodoferax, and Janthinobacterium, while those exceeding 10% in red snow were Stutzerimonas and Variovorax. Based on the present metagenomic data, genes associated with nitrification were detected only in the red snow and not in the green snow samples, and the annotated genera included Candidatus_Reddybacter, Bradyrhizobium, and Sinimarinibacterium. In terms of the sulfur cycle, most of the detected marker genes were involved in sulfur mineralization and assimilatory sulfate reduction. The abundances of these processes showed no significant differences between red and green snow (P > 0.05), but the microbial compositions differed. Among the top three genera associated with sulfur mineralization, Pseudomonas, Polaromonas, and Janthinobacterium were found in green snow, whereas Polaromonas, Hymenobacter, and Halpernia were found in red snow. The genera involved in assimilatory sulfate reduction differed between green and red snow: Ferruginibacter and Hymenobacter dominated in red snow, while Polaromonas and Flavobacterium dominated in green snow. In the phosphorus cycle, genes related to phosphate regulation were significantly enriched in red snow (P < 0.05). In the phosphate regulation pathway, Hymenobacter and Pedobacter were abundant in both red and green snow, Flavobacterium and Pseudomonas showed higher relative abundance in green snow, while Ferruginibacter and Halpernia were more abundant in red snow. Fig. 5Genetic potential of each step of the carbon, nitrogen and sulfur cycles in green and red snow calculated using normalized combinations of marker genes from identified KOs. The relative abundance of the main microorganisms at the genus level in each step of the pathways was shown (top 10 genera and unclassified genera were grouped as “Others”) Genetic potential of each step of the carbon, nitrogen and sulfur cycles in green and red snow calculated using normalized combinations of marker genes from identified KOs. The relative abundance of the main microorganisms at the genus level in each step of the pathways was shown (top 10 genera and unclassified genera were grouped as “Others”) Discussion Community composition and co-occurrence networks In previous analyses of Antarctic snow, eukaryotic content ranged from 0.77% to 14.97%, with archaea accounting for less than 0.5% and prokaryotes making up the majority [38]. These proportions are similar to those observed in this study. The relative abundance of archaeal sequences was low, though their significance within the overall microbial community has progressively gained recognition in recent years [39, 40]. There was significant difference in β-diversity among the eukaryotic, bacterial, and archaeal communities in green and red snow, which suggested notable variations in community composition between red and green snow. The prevalence of Bacteroidota and Pseudomonadota was consistent with previous studies [41, 42]. Basidiomycota were prevalent in polar snowfields and can adapt to extreme environments through specialized membrane lipid compositions and altered metabolism [41, 43, 44]. However, Chytridiomycota has been rarely reported in other snowpacks compared with colored snow. This was probably due to the ability of Chytridiomycota to parasitize algae and their abundance was related to algal biomass [45, 46], so it was more common in green and red snow with higher biomass of algae. This relationship might affect the carbon flow in colored snow — Chytridiomycota rapidly release photosynthetic products into the dissolved organic carbon pool by lysing algal cells, stimulating secondary production by heterotrophic bacteria and thereby accelerating the functioning of the microbial community in colored snow. Similar heterotrophic communities (bacteria and archaea) have been reported in colored snow [18]. The red and green snow prokaryotic communities had similar dominant phyla, differing in relative abundance. Previous studies have emphasized the crucial role of algal communities in influencing snow coloration [15, 18, 47]. β-diversity showed significant variation in the taxonomic composition of algal communities between green and red snow. Many studies have shown that Chlorophyta constitute a large proportion of colored snow, but the second most abundant Cyanobacteriota was often overlooked. Consistent with previous reports [12], Cyanobacteriota made up less than 1% of the bacterial community, often acting as pioneer organisms, were major primary producers in polar and extreme environments and play key ecological roles [48]. The Shannon and Simpson for the green snow algae community were significantly higher than red snow, several genera associated with snow color were significantly different. Notably, Gonium was significantly enriched in the red snow. Unicellular Chlamydomonas was thought to resemble the unicellular ancestor of multicellular Gonium [49, 50]. Sanguina produced cysts enriched in carotenoids with a high level of astaxanthin [51]. Chloromonas could produce cytoplasmic lipid bodies, filled with secondary carotenoids in their generative resting stages [52]. The shift of algae from vegetative growth to dormant reproduction is not only a transition between life cycle stages but also an adaptive response to extreme environmental stress. It could be concluded that algae community composition played vital role in snow color. Furthermore, future studies that incorporate additional bioinformatic or morphological analyses for taxonomic validation would provide deeper insights into the microbial community structure of colored snow. The analysis of microbial networks enhanced our understanding of the complex network of interactions and ecological processes within colored snow communities [53, 54]. Compared with green snow, the microbial co-occurrence network of red snow had more nodes and edges, and also exhibited a higher degree, indicating potentially more complex associations. Such complex network structure was generally considered the basis for maintaining stability under environmental stress. This was supported by stability analysis: the microbial network in green snow showed potentially weaker resistance to disturbances compared to the red snow network. This also indicated that the red snow community was not in a terminal stage of functional decline, but rather formed a stress-resistant, complex, and stable state, helping microorganisms maintain basic metabolism and reproduction under extreme environmental conditions. On this basis, the biological characteristics of the red snow community and the physical processes formed a synergistic effect. Studies have shown that red snow exhibits lower reflectance and higher radiative forcing compared with green snow, enhancing energy absorption and promoting ice crystal melting [55]. Therefore, the red snow community not only benefited its own long-term survival, but might also have potentially influenced the positive feedback effect of snowmelt by maintaining a low-albedo surface. Geochemical cycles and their ecological significance The identification of gene sequences, involved in various ecological processes linked to carbon, nitrogen, sulfur, and phosphorus in colored snow, along with the disparities in the relative abundance of each pathway, indicated that microbial functions of colored snow were intricately tied to variations in biogeochemical cycling. In this study, Pseudomonadota and Bacteroidota were the most abundant phyla in C, N, S and P cycles, and possessed the ability to rapidly degrade organic matter, potentially serving as a crucial link in the direct transfer of organic carbon from algae to bacteria [56]. The functional potential of C, N, S, and P pathways in the green snow microbial community showed a significant correlation with Pseudomonadota, whereas that of the red snow microbial community showed a significant correlation with Bacteroidota and Chlorophyta (Figure S7). These associations not only reveal the key roles of microorganisms in C, N, S, and P cycles, but may also reflect the ecological adaptation strategies of microbial communities under different environmental conditions. As to carbon cycling, aerobic respiration predominated in the colored snow and has also been reported to prevail in Antarctic lake ecosystems [57]. This indicated that aerobic respiration was a widespread metabolic strategy for degrading complex carbon and thriving under oligotrophic conditions. The significant enrichment of aerobic respiration related genes in red snow likely reflects microbial adaptation to nutrient poor environments: through high efficiency energy yielding pathways, microorganisms extract maximum energy from limited organic carbon to sustain survival. The relative abundance of genes involved in fermentation was significantly higher in the green snow microbial community than in the red snow. Green snow typically occurs in areas relatively enriched in organic carbon [2]. This environment might support the high expression of fermentation-related genes, enabling heterotrophic microorganisms to rapidly degrade dissolved organic matter through fermentation. In the nitrogen cycle, genes associated with denitrification were significantly enriched in green snow, and the related genera also showed obvious differences. Stutzerimonas was recognized as a highly efficient denitrifying bacterium [58], and was the most abundant genus in denitrification in red snow. Janthinobacterium was a key representative for studying denitrifying microbial response mechanisms under extreme low temperature, dryness, and high radiation conditions [59], and was abundant in green snow. Based on the present metagenomic data, genes associated with nitrification were not detected in the green snow. The higher concentrations of NO₂ and NO₃- in the red snow compared with the green snow (Figure S2) also corroborated the difference in nitrification-related functional potential between red snow and green snow. Under climate warming, red snow patches might continue to expand, and their nitrification potential may convert more ammonium nitrogen into nitrate nitrogen, which could be transported to surrounding marine areas via snowmelt runoff, affecting nitrogen nutrient levels. But this inference still requires validation by isotope tracing or field flux observations. It has been reported that red snow contains very little phosphate [60]. Genes related to phosphate regulation were significantly enriched in red snow, which may represent a survival strategy under extreme stress. Microbial communities might enrich these genes to regulate phosphorus metabolism, thereby enabling survival in extreme environments. In summary, the different environmental pressures experienced by green and red snow led to different microbial community compositions and metabolic strategies, resulting in distinct geochemical cycling functions. This phenomenon serves as a macroscopic manifestation of microbial communities’ environmental adaptation strategies. Conclusion The study described significant differences in diversity and metabolism of the microbial communities between green and red snow. Bacteria dominated the microbial community in colored snow, and green and red snow harbored distinct microbial communities. Co-occurrence network analysis indicated that red snow communities exhibited greater complexity and a higher capacity to maintain stability than green snow. Meanwhile, the Zi-Pi plots revealed key connector taxa in green snow (Flavobacterium, Undibacterium, Rhodanobacter, Ralstonia, Catalinimonas, Barnesiella) and red snow networks (Rhodanobacter, Massilia, Flectobacillus, Roseomonas, Marinoscillum, Acidisoma, Polycyclovorans, Blepharisma and Belnapia). These key taxa likely contribute to community stability and to the functional differences observed between red and green snow. Genes associated with carbon, nitrogen, phosphorus and sulfur cycling were identified in green and red snow, and were highly associated with Pseudomonadota and Bacteroidota. Red snow had a higher representation of aerobic respiration genes and a lower representation of fermentation genes, suggesting differences in carbon dioxide production from decomposition of organic matter. Nitrification genes were detected only in red snow. Phosphate regulation genes were significantly enriched in red snow, indicating higher phosphorus bioavailability. Overall, colored snow influenced functional potential by altering microbial composition within the relevant pathways. These results provide metagenomic evidence for potential microbial and functional differences between red and green snow, and offer a more comprehensive prediction of Antarctic ecological functions under global change. Supplementary Information Supplementary Material 1. Supplementary Material 1.