Cyanobacteria sustain microbial diversity and community stability in Tibetan glacial cryoconites

Abstract

Cyanobacteria play a key role in aggregating cryoconite granules on glacier surfaces, creating stable microhabitats that support diverse microbial communities, which influence glacier albedo and melting. However, their contribution to bacterial diversity and community stability is not well understood. This study explores their impact on bacterial diversity and interactions within three cryoconite-related environments: sediments and overlying water in cryoconite holes, and surface cryoconites across four Tibetan glaciers. Our study revealed that Cyanobacteria contributed the most (15–21%) to the differences in community compositions between these three habitats within each glacier. Cyanobacteria were abundant only in cryoconite sediments and surface cryoconites, accounting for 31–37% and 12–38% of all sequences, respectively, and contributing 6–10% and 5–9% to bacterial richness. Cyanobacteria genera such as Chamaesiphon and Pseudanabaena were key taxa, interacting closely with Bacteroidetes genera (e.g., Flavobacterium and Ferruginibacter) and Proteobacteria genera (e.g., Rhodoferax and Polaromonas). Metabolic analysis suggests that Cyanobacteria may provide essential nutrients to their heterotrophic bacterial partners through carbon and nitrogen fixation. The collaboration between Cyanobacteria and these bacteria contributes to community stability. These findings suggest that Cyanobacteria may act as ‘engineer taxa’, influencing bacterial diversity and the structural-functional stability of cryoconite microbial communities, and providing new insights into the potential responses of glacier ecosystems to climate change.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-025-00817-z.

Introduction

Cryoconites are recognized as "ice-cold hotspots" of biological activity and microbial diversity [1–3]. These unique ecosystems host complex microbial communities, including bacteria, archaea, fungi, and viruses, which not only survive but actively contribute to regional biogeochemical cycles through their metabolic processes [4–7] Some cryoconite granules disperse across glacier surfaces, forming surface cryoconites [8]. Surface cryoconites are more directly exposed to environmental factors, undergo greater movement, dispersion, and redistribution in response to external disturbances, making them dynamic features on the glacier surface. The dark color of the surface cryoconites absorbs more solar radiation than the surrounding ice, leading to localized melting and gradual deepening over time [9, 10]. Finally, the depression collapses to form a near-cylindrical holes in ice surfaces known as cryoconite holes, with cryoconite sediments accumulating at the bottom. The ice above cryoconite sediments melts, creating an overlying water layer during the summer, which typically refreezes in the winter [11].

The persistence of cryoconite habitats, such as cryoconite sediment and surface cryoconite, on temperate mountain glaciers is generally short-lived [9, 12]. This is due to the rapid flow, higher air temperatures, and significant diurnal temperature fluctuations characteristic of these glaciers. During the ablation season, cryoconite holes and surface cryoconites release debris, microorganisms, and organic matter into proglacial streams and lakes. This process has considerable downstream effects, such as the potential increase in anthropogenic dissolved organic matter in streams, which could heighten glacier sensitivity to climate warming [13]. Additionally, microorganisms from cryoconite habitats can significantly impact the biodiversity and community structure of glacier-fed streams [14]. Despite these ecological consequences, studies on microbial communities in the diverse cryoconite habitats of temperate mountain glaciers, particularly in cryoconite sediments and surface cryoconites, remain limited compared to research on polar glaciers.

Microorganisms interact with each other in various ways, including nutrient competition, cooperation through cross-feeding, communication via secretions, and sensing extracellular substances [15]. Co-occurrence networks provide a valuable method for assessing the degree of integration and interaction among different organisms. Co-occurrence patterns in microbial communities offer key insights into potential interactions between components and their environmental responses, which are often difficult to detect using traditional microbial community analyses [16–18]. Investigating microbial interactions in cryoconite habitats, where biodiversity and microbial activity are concentrated on glacier surfaces, helps us understand how microbes contribute to carbon and nitrogen cycles. Using co-occurrence networks, Cyanobacteria and Actinobacteria were identified as core organisms with the highest betweenness centrality metrics, which were tightly and negatively connected with each other in cryoconites on the Arctic ice cap [19]. However, the potential functional interactions between these organisms were not immediately clear. Recent study linking co-occurrence networks to genomic data have enabled functional predictions such as methane production in wetlands [20], this offers a new perspective to explore the potential functional interactions of core organisms in cryoconite habitats.

Cyanobacteria play a crucial role in cryoconite habitats, their filamentous morphology allows them to entangle debris, aiding in the formation of cryoconite granules [21]. Studies have shown a positive correlation between cyanobacterial abundance and granule size, suggesting their role in stabilizing and shaping granule structure [22]. Cyanobacteria are also primary producers in cryoconite habitats, dominating biomass accumulation and carbon fixation [10, 23]. Through photosynthesis, they generate organic substrates that sustain heterotrophic microbes within the larger microbial community [24]. Additionally, Cyanobacteria can fix nitrogen when bioavailable nitrogen is limited [25, 26], further contributing to ecosystem productivity. Their interactions with other microorganisms, such as Actinobacteria, also influence the development of cryoconite microbial communities [19]. While the structural and functional roles of Cyanobacteria have been already explored, quantitative assessments of their impact on bacterial community diversity and stability, especially in the context of Tibetan glacier cryoconite habitats, remain limited, presenting an important area for future research.

To address these gaps, we investigated the role of Cyanobacteria in shaping bacterial community diversity and co-occurrence networks within cryoconite holes and surface cryoconites on Tibetan glaciers. The Tibetan Plateau and its surrounding regions host the largest mountain glaciers in the world [27], which provide freshwater to over one billion people across East and South Asia [28]. However, accelerated glacier retreat in recent decades due to global warming [29] is likely to increase the frequency and intensity of cryoconite hole and surface cryoconite melting, as well as their redistribution across glacier surfaces. Understanding the role of key microbial taxa, particularly Cyanobacteria, in cryoconite habitats is therefore crucial. We hypothesize that Cyanobacteria play a central role in promoting bacterial community diversity and enhancing ecosystem stability in both cryoconite sediments and surface cryoconites on Tibetan glaciers.

Materials and methods

Study area and sampling

In order to consider geographic variability, four glaciers in different regions of the Tibetan Plateau were chosen in the present study: Jiemayangzong Glacier (JMYZ), Parlung No. 4 Glacier (PL4), Qiangyong Glacier (QY), and Tanggula Longxiazailongba Glacier (TGL) (Fig. S1). The JMYZ Glacier (30°13′N, 82°10′E; 5008 m a.s.l.), situated in the western Himalayas of the southwestern Tibetan Plateau. The PL4 Glacier (29°15′N, 96°56′E; 4700 m a.s.l.), located in the Kangri Garpo Range (southeastern Tibetan Plateau). The QY Glacier (28°53′N, 90°13′E; 5600 m a.s.l.) on the northern Himalayan slope of the southern Tibetan Plateau. The TGL Glacier (33°08′N, 92°04′E; 5560 m a.s.l.), positioned in the Tanggula Mountains of the central Tibetan Plateau.

Sampling was conducted across all four glaciers from July to September 2019 to collect cryoconite sediment (C), overlying water (W), and surface cryoconite (SC) samples (Table S1, Fig. S1). For each glacier, 6–11 cryoconite holes were systematically selected. A sterile syringe connected to a suction tube was used to collect overlying water from the small holes, ensuring minimal disturbance to the structure. The same method was applied to collect cryoconite sediment from below the overlying water, although a small amount of water was drawn in during this process. At the same time, 8–11 surface cryoconite samples were collected from peripheral zones near cryoconite hole clusters using sterile spatulas, after removing the top ~ 2 cm of the surface layer. All samples were stored in 500 mL Nalgene HDPE bottles, transported on ice within 24 h, and preserved at  − 20 °C until laboratory analysis. Detailed information on sample collection can be found in the Supplementary Methods.

Total carbon (TC) and Total nitrogen (TN) of cryoconite samples were measured using a Vario MAX elemental analyzer (Elementar Corp. Germany). For overlying water samples, dissolved organic carbon (DOC) and total nitrogen (TN) were determined using a TOC-L (Shimadzu Corp., Japan) after filtering through 0.45 μm PES syringe filters.

DNA extraction, 16S rRNA gene sequencing and metagenomic sequencing

In the laboratory, the cryoconite sediment mixture was left to stand overnight, allowing the sediment to naturally settle. The biomass of the overlying water (approximately 400 mL) was filtered onto 0.22 μm pore-size polycarbonate filters (47 mm diameter; Millipore). The filters were then cut into small pieces, soaked in 1.5 mL of lysis buffer (20 mg/mL Proteinase K, 0.1 M EDTA, and 10% SDS), and incubated at 55 °C for two hours before DNA extraction. Cryoconite samples were carefully thawed at 4 °C, and DNA was extracted from 0.5 g of cryoconite debris. DNA from all water and cryoconite samples was extracted using the PowerSoil DNA Isolation Kit (Qiagen, Germantown, USA), and an aliquot (50 ng) of purified DNA from each sample was used as a template for amplification.

DNA obtained from 101 samples (Table S1) was amplified using primers targeting the V4-V5 region of the 16S rRNA gene: 515F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 907R (5’-CCGTCAATTCMTTTRAGTTT-3’) [30]. The PCR cycle consisted of an initial denaturation at 95 °C for 3 min, followed by 30 cycles of denaturation at 95 °C for 45 s, annealing at 50 °C for 60 s, extension at 72 °C for 90 s, and a final extension at 72 °C for 10 min. Each sample was amplified in triplicate, and the PCR products were pooled for subsequent purification and Illumina HiSeq sequencing. Sequencing was performed at MAGIGENE (Shenzhen, China) on the Illumina HiSeq platform (Illumina Inc., San Diego, CA, USA) using a paired-end strategy (2 × 250 bp).

DNA from 12 cryoconite samples (Table S1) was fragmented using an ultrasonic disruptor (Bioruptor™ sonicator, Diagenode, Belgium), producing fragments ranging from 100 to 400 bp in length. The fragmented DNA was then used to construct a metagenomic library with fragment sizes of 300–500 bp using the KAPA HyperPrep Kit (Roche, Switzerland). After library construction, the library was quantified using the Qubit dsDNA HS Assay Kit (Invitrogen, USA) on the Qubit 2.0 fluorometer (Invitrogen, USA). Following quantification, the library was diluted to a concentration of 1 ng/μL, and the insert fragments were assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Germany) with the Agilent High Sensitivity DNA Kit (Agilent Technologies, Germany). If the insert fragment sizes fell within the expected range, the library's effective concentration was accurately determined using quantitative Polymerase Chain Reaction (qPCR), ensuring an effective concentration greater than 2 nM. The metagenomic sequencing for this study was conducted by MAGIGENE (Shenzhen, China) using the Illumina HiSeq 2500 platform (PE150) with 150-bp paired-end reads.

Processing of 16S rRNA sequence data

QIIME2 (v2022.02, https://qiime2.org/) and VSEARCH (v2.15, https://github.com/torognes/vsearch) were mainly used for the 16S rRNA gene sequence analyses. Firstly, paired-end reads were merged utilizing the—fastq_mergepairs command in VSEARCH. Subsequently, the removal of forward (GTGCCAGCMGCCGCGGTAA) and reverse primers (CCGTCAATTCMTTTRAGTTT) was performed using cutadapt (v3.5). After quality filtering, amplicon sequence variants (ASVs) of the 16S rRNA gene were identified using DADA2 to group similar sequences, with representative sequences occurring less than five times being removed. Finally, taxonomy classification was carried out in VSEARCH using the classify-consensus-vsearch command with reference to the SILVA 138.1 database [31]. And we removed ASVs that affiliated with mitochondria/chloroplast. To address uneven sequencing depth, all samples were randomly subsampled to the smallest library size, which was 17,188 reads.

Metagenomic data analysis

The quality control process was performed using the software Trimmomatic (v0.39) with the parameters (LEADING:3 TRAILING:3 SLIDINGWINDOW:5:20 MINLEN:50) [32]. After quality control, MEGAHIT (v1.2.9, https://github.com/voutcn/megahit) was used for de novo assembly of the high-quality sequences with K-mer settings as follows: k-min 35, k-max 95, k-step 20 [33]. Metagenomic assembled contigs, with a minimum length of 1000 bp were binned using MetaWRAP v1.3.2 with two most effective binning methods, MetaBAT 2 (version 2.12.1) and MaxBin 2 (version 2.2.7) [34]. The completeness and contamination of the medium to high quality Metagenome-Assembled Genomes (MAGs, completeness > 50%, contamination < 10%) were assessed with CheckM (v1.1.3) [35]. The MAGs were then clustered at the species level using the dRep software (v3.2.0), with a 95% ANI threshold and 30% aligned fraction, resulting in 559 dereplicated species-level genome bins (SGBs). Detailed information on the SGBs was provided in Supplementary Table S7.

Taxonomic classification was carried out with GTDB-tk (v0.3.2) using the Genome Taxonomy Database (GTDB). The SGB abundance was calculated using CoverM (v0.6.0) with the TPM method. The MAGs were annotated using the Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthologies (KOs) database (release 107.1) via the “annotate” function with Hidden Mark Models (HMMs, v 3.3.1) in EnrichM v0.5.0 [36]. And the Cross-validate using KofamScan at an e-value threshold of le-5 were also provided in the Supplementary Table S8. Furthermore, to avoid misannotation for nitrogenase (nifH) gene, we have constructed a phylogenetic tree based on nifH protein sequences in cyanobacterial MAG and the reference genomes downloaded from NCBI (Fig. S10).

Indicator ASVs and network analyses

To identify indicator taxa associated with each habitat, we applied the IndVal (indicator values) method using the “indicspecies” package in R (version 4.1.2) [37]. ASVs with significantly higher (P < 0.05) abundance in cryoconite sediment, water, and surface cryoconite samples were categorized as C-enriched ASVs, W-enriched ASVs, and SC-enriched ASVs, respectively. The remaining ASVs, which showed no significant differences in relative abundance, were classified as ‘Others’.

Co-occurrence networks were constructed to assess species coexistence across habitats, following the method described by Friedman and Alm [38]. Only ASVs with an average relative abundance greater than 0.05% were included. ASV correlations were calculated using the SparCC algorithm [38], implemented in the SparCC Python module, with 99 bootstrap cycles. Correlations with a coefficient (ρ) > 0.7 and FDR-corrected P-values < 0.01 were used to build the network with the "igraph" package in R. Network topology was described using indices such as edges (E), nodes (N), average degree (AD), clustering coefficient (CC), average path length (APL), and modularity (MD). Visualizations were generated using the Gephi platform (v0.9.2). Subnetworks for each habitat were derived from the overall network using the "induced_subgraph" function in R [39, 40]. Network robustness was defined as the proportion of remaining nodes after 50% random node removal [41, 42], and the effect of cyanobacterial node removal on subnetwork robustness was also assessed. Furthermore, the equivalent removal analyses for Cyanobacteria nodes and keystone nodes to compare the robustness among them were conducted.

Keystone taxa were identified based on among-module connectivity (Pi) and within-module connectivity (Zi) [43]. Nodes were classified into four categories: module hubs (low Pi ≤  0.62, high Zi> 2.5), connectors (high Pi > 0.62, low Zi ≤ 2.5), network hubs (high Pi > 0.62, high Zi > 2.5), and peripheral nodes (low Pi < 0.62, low Zi < 2.5), as described by Deng et al. [44].

Statistical analysis

The analysis was performed using R (version 4.1.2). Bacterial alpha diversity, specifically richness, was calculated using the “diversity” function from the “vegan” package [45]. We also assessed the contribution of Cyanobacteria to alpha diversity by calculating the ratio of cyanobacterial richness to bacterial richness. The Kruskal-Wallis test was used to compare richness across habitats (overlying water, cryoconite sediment, and surface cryoconite).

Non-metric multidimensional scaling (NMDS), based on Bray-Curtis dissimilarity [46], was applied to visualize differences in bacterial and cyanobacterial community composition across habitats. To identify the taxa contributing to habitat variation within each glacier, similarity percentage analysis (SIMPER) was conducted using the “vegan” package, with only the top 10 ASVs presented. Finally, permutational multivariate analysis of variance (PERMANOVA) was performed using the “adonis” function from the “vegan” package to assess variations in bacterial and cyanobacterial communities across habitats.

Results

Chemical characteristics of the cryoconite samples

Total carbon (TC) and total nitrogen (TN) of cryoconites and dissolved organic carbon (DOC) and total nitrogen (TN) of overlying water were significantly different among the four glaciers (P < 0.001, Fig. S9, Supplementary Table S9). For cryoconite sediments and surface cryoconites, TC contents were the highest in QY glacier with the mean values of 2.92 ± 0.29 and 3.15 ± 0.46 mg g-1, respectively, followed by TGL and PL4, and the lowest in JMYZ (Fig. S9). The trend of TN contents is similar with that of TC, with the maximum in glacier QY (mean 0.23 ± 0.02 and 0.28 ± 0.06 mg g-1 in C and SC, respectively). Furthermore, DOC and TN contents were the highest in TGL glacier with the mean values of 102.41 ± 43.72 mg L-1 and 8.10 ± 2.27 respectively, followed by QY and PL4, and the lowest in JMYZ.

Microbial community composition

Bacterial compositions across the three habitats of all four glaciers were distinct (Fig. 1). Cyanobacteria were the dominant phylum in both cryoconite sediments (C, 35%) and surface cryoconites (SC, 26%), followed by Bacteroidetes (29% in C, 25% in SC) and Proteobacteria (22% in C, 26% in SC). Specifically, in the JMYZ glacier, Cyanobacteria accounted for 37 and 20% of the total sequences in cryoconite sediments and surface cryoconites, respectively (Fig. 1a). Similar trends were observed in the other glaciers: 31 and 12% in PL4 (Fig. 1b), 35 and 38% in QY (Fig. 1c), and 36 and 37% in TGL (Fig. 1d). In contrast, the bacterial community in the overlying waters (W) was notably different, with Proteobacteria (53%) and Bacteroidetes (32%) as the two most abundant phyla, and Cyanobacteria (1%) presented at much lower levels compared to cryoconite habitats (Fig. 1).

Fig.1.

Relative abundance of the dominant bacterial phyla presents in the overlying water (W) and cryoconite sediment (C) and of cryoconite holes and surface cryoconites (SC) on a JMYZ, b PL4, c QY and d TGL glaciers, respectively. Only taxonomic groups with relative abundances > 1% were shown. JMYZ, Jiemayangzong, PL4, Parlung No. 4 Glacier; QY, Qiangyong Glacier; TGL, Tanggula longxiazailongba Glacier

Cyanobacteria dominated cryoconite sediments and surface cryoconites, with distinct genera observed across glaciers (Fig. 2). Loriellopsis and Phormidesmis were abundant in the cryoconites from the JMYZ glacier (Fig. 2a–b), while Nostoc and Phormidesmis were prominent in the cryoconites from PL glacier (Fig. 2a–b). Tychonema and Alkalinema were most abundant in the cryoconites from the QY and TGL glaciers (Fig. 2c–d).

Fig.2.

Relative abundance of the cyanobacterial genus in bacterial community present in the overlying water (W) and cryoconite sediment (C) and of cryoconite holes and surface cryoconites (SC) on a JMYZ, b PL4, c QY and d TGL glaciers, respectively. The taxonomic groups that can't be classified into genus are labeled as “Others”. JMYZ, Jiemayangzong, PL4, Parlung No. 4 Glacier; QY, Qiangyong Glacier; TGL, Tanggula longxiazailongba Glacier

Microbial biodiversity

Bacterial richness was assessed using the total normalized ASV table, revealing significant differences between cryoconite sediments, surface cryoconites, and overlying water (Fig. 3a–d). In all four glaciers, richness in cryoconite sediments or surface cryoconites was significantly higher than in overlying water (Kruskal-Wallis test, P < 0.05). Cyanobacteria made a significant contribution to bacterial diversity in cryoconite habitats, accounting for 6–10% in cryoconite sediments and 5–9% in surface cryoconites, which was higher than in water, where Cyanobacteria accounted for only 2–3% (Fig. S2a–d).

Fig.3.

Bacterial alpha diversity (a–d) and beta diversity (e–h) among the overlying water (W) and cryoconite sediments (C) of cryoconite holes and surface cryoconites (SC) samples. a–d Richness indexes of bacterial communities in a JMYZ, b PL4, c QY and d TGL glaciers, respectively. e–h NMDS based on Bray–Curtis dissimilarity of bacterial communities in e JMYZ, f PL4, g QY and h TGL glaciers, respectively. Symbols were colored by habitats. W, overlying water of cryoconite holes; C, cryoconite sediments of cryoconite holes; SC, surface cryoconites. JMYZ, Jiemayangzong; PL4, Parlung No. 4 Glacier; QY, Qiangyong Glacier; TGL, Tanggula longxiazailongba Glacier. Ellipses in the plots denote 95% confidence intervals

NMDS analysis based on Bray-Curtis dissimilarity revealed that both bacterial communities (Fig. 3e–h) and cyanobacterial communities (Fig. S2e–h) clustered according to habitat types. These distribution patterns were confirmed by PERMANOVA (P < 0.05, Tables S3 and S4). Communities from cryoconite sediments and surface cryoconites were more similar to each other but distinct from those in the overlying water, indicating a close association between cryoconite sediments and surface cryoconite communities (PERMANOVA, P < 0.05; Tables S3 and S4).

SIMPER analysis showed that the top 10 ASVs accounted for 34 to 52% of the variation among habitats within each glacier (Fig. S3). Among these, ASV1 and ASV26 (Tychonema and Loriellopsis genera, both within Cyanobacteria) contributed the most to variation in the JMYZ, QY, and TGL glaciers, with contributions of 12, 8, and 13%, respectively (Fig. S3). ASV20 (Phormidesmis genus) also contributed 5% to the variation in glacier PL4. All Cyanobacteria ASVs contributed 15–21% to habitat variation across the glaciers (Table S5). Furthermore, ASVs from Proteobacteria and Bacteroidetes were secondary contributors, with their contributions ranging from 2 to 9% (Fig. S3).

Species co-occurrence patterns

Among the total network and three subnetworks, the empirical networks exhibited a higher average clustering coefficient compared to the respective Erdös-Rényi random networks (Table S6), Based on correlation analysis, 1132 edges representing significant and strong pairwise correlations between ASVs (ρ > 0.7, P < 0.01) were identified among 162 nodes (ASVs). Indicator species analysis revealed 62, 34, and 34 ASVs significantly enriched in cryoconite sediments (C), overlying waters (W), and surface cryoconites (SC), respectively (Fig. 4a). The C-enriched and SC-enriched ASVs, which included a higher abundance of Cyanobacteria, exhibited significantly more interconnections within module 1 (Fig. 4, Table S10). In contrast, the W-enriched ASVs, primarily dominated by Proteobacteria and Bacteroidetes, were highly interconnected within module 3 (Fig. 4). The interactions between ASVs from different habitats also showed a relatively high proportion, and most of them were positively correlated (Fig. S4). These habitat-specific indicators formed the basis for constructing three empirical subnetworks, which exhibited higher average clustering coefficients compared to the respective Erdös-Rényi random networks (Table S6). The C and SC subnetworks exhibited enhanced structural cohesion, as evidenced by their higher average degree (SC: 7.75; C: 11.10 vs W: 4.65) and clustering coefficients (SC: 0.81; C: 0.73 vs W: 0.69) (Table S6). Their compact connectivity was further reflected in shorter average path lengths (SC: 1.38; C: 2.12 vs W: 2.37), suggesting more efficient interspecific communication in particulate habitats.

Fig. 4.

Co-occurrence networks (a–c), keystone taxa (d) and robustness of microbial subnetwork (e). The networks of ASVs pairs are based on significant sparse correlations (ρ > 0.7, P-value < 0.01). a Nodes colored by corresponding habitat (i.e., cryoconite sediment and water of cryoconite holes and surface cryoconites). b Nodes colored by the phylum/class-level taxonomy. c Nodes colored by the corresponding module. The size of each node (ASV) is proportional to the number of connections (i.e., degree). Cyanobacterial nodes are marked with numbers. 1: Alkalinema; 2: Anagnostidinema; 3–5: Chamaesiphon; 6: Leptodesmis; 7, 8: Loriellopsis; 9: Oculatella; 10–14: Phormidesmis; 15–16: Pseudanabaena; 17: Tychonema. d The keystone taxa in network based on their topological roles. Module hubs are identified as Zi > 2.5, Pi ≤ 0.62, connectors are identified as Zi ≤ 2.5, Pi > 0.62. Nodes colored by the phylum/class-level taxonomy. e The robustness of microbial subnetwork by three habitats. W, overlying water of cryoconite holes C, cryoconite sediments of cryoconite holes; SC, surface cryoconites

Within our network, 48 ASVs were identified as either module hubs (n = 2) or connectors (n = 46), with the majority enriched in modules 1, 2, and 3 (Fig. 4d). These keystone taxa were predominantly affiliated with the phyla Proteobacteria, Bacteroidetes, and Cyanobacteria (Fig. 4d). Specifically, within the Proteobacteria phylum, 17 connectors and 1 module hub (e.g., the genera Acidiphilium, Polaromonas, and Rhizobacter) were primarily associated with W-enriched ASVs, whereas 14 connectors and 1 module hub (e.g., the genera Ferruginibacter and Flavobacterium) from Bacteroidetes, along with 4 Cyanobacteria connectors (i.e., the genera Chamaesiphon, Loriellopsis, and Pseudanabaena), were predominantly linked to cryoconite sediments and surface cryoconites.

Robustness analysis revealed that the C and SC subnetworks maintained higher stability than the W subnetwork under random species loss (P < 0.01; Fig. 4e). The critical role of Cyanobacteria was further confirmed by targeted removal experiments: excluding cyanobacterial nodes reduced the C and SC robustness to 0.73 and 0.88, respectively (Fig. S5). Furthermore, the robustness of the entire network decreased more significantly after the removal of all cyanobacterial nodes compared to the removal of an equivalent number (17) of keystone nodes (i.e., module and connector nodes) (Fig. S6). To account for variability, we randomly selected 17 nodes from the 48 keystone nodes and repeated this process 100 times.

The essential metabolic potential of glacial cyanobacteria and their interacting bacteria

Cyanobacterial ASVs exhibited distinct habitat preferences, with higher proportions found in cryoconite sediments (16%) and surface cryoconites (12%) compared to overlying waters (3%) (Fig. 4a and b). This niche partitioning was reflected in the divergent interaction patterns observed in the stratified network analysis (Fig. S7). In cryoconite sediments, cyanobacterial nodes (affiliated with the genera Anagnostidinema, Chamaesiphon, Phormidesmis, and Pseudanabaena) formed positive, localized clusters with Bacteroidetes ASVs, including the genera Flavobacterium, Ferruginibacter, and Hymenobacter. In surface cryoconites, the positive interaction network was dominated by filamentous Cyanobacteria (affiliated with the genera Leptodesmis and Phormidesmis) (Fig. S7), which bridged nodes affiliated with Bacteroidetes genera (Flavobacterium, Ferruginibacter and Hymenobacter) and Proteobacteria genera (Rhizobacter and Polaromonas).

To explore potential functional interactions among microbial organisms, we investigated the energy-related metabolism and cold tolerance potential of the Cyanobacteria present in the network and their interacting bacteria (Fig. 5). In total, 23 cyanobacterial SGBs were reconstructed from 12 metagenomic samples, with a mean completeness of 87% and a contamination rate of 3% (Supplementary Table S7). The predominant genera included Microcoleus (genus Tychonema, 2 SGBs, 3%; Fig. S8), Chamaesiphon (4 SGBs, 2%), Alkalinema (3 SGBs, 2%), and Phormidesmis (4 SGBs, 2%).

Fig.5.

Metabolic potential of Cyanobacteria (23 SGBs) and their interacting bacteria (83 SGBs) in cryoconite sediments and surface cryoconite. The completeness of MAGs exceeds 80% of these genomes. The color gradient represents the proportion of MAGs within each genus that encode the gene of interest

All cyanobacterial genera (23 SGBs) harbored key enzymes of the Calvin-Benson-Bassham (CBB) cycle, including the ribulose-bisphosphate carboxylase large chain (rbcL) and phosphoribulokinase (prkB), indicating their capacity for CO₂ fixation (Fig. 5). Oxygenic photosynthetic marker genes (psaAB and psbAB) were identified in most cyanobacterial genera (9–23 SGBs). Nitrogen fixation potential, represented by the nifH marker gene, was limited to a single Nostoc (1 SGB). Genes responsible for assimilatory nitrate reductase (narB and nirA) were universally present in all cyanobacterial genera (22 SGBs), while dissimilatory nitrite reductase, represented by nirS, was only detected in Chamaesiphon (1 SGB). Cold adaptation was supported by the presence of cold-shock protein genes (csp) in five cyanobacterial genera (10 SGBs). Crucially, the capacity for exopolysaccharide (EPS) biosynthesis was widely distributed. The exopolysaccharide production protein (exoZ) and polysaccharide biosynthesis protein (pslH) were detected in Phormidesmis, Tychonema, Nostoc and Alkalinema (10 SGBs). Two genes encoding polysaccharide exporters (wza, cysE) were identified in all cyanobacterial genera (10 SGBs), while protein-tyrosine kinase (epsB) was detected in Phormidesmis (2 SGBs).

The metabolic profiles of cyanobacterial interacting bacteria (83 SGBs) were also analyzed (Fig. 5), with the majority belonging to the phyla Bacteroidetes (e.g., Ferruginibacter, Hymenobacter and Spirosoma) and Proteobacteria (e.g., Rhodoferax, Polaromonas and Rhizobacter). Among these, carbon fixation genes (rbcL and prkB) were detected in only four genera (Ferruginibacter, Rhodoferax, Polaromonas and Rhizobacter) (11–15 SGBs). Oxygenic photosynthesis marker genes (psaAB, psbAB) and nitrogen fixation genes (nifH) were absent in these bacteria. Genes for assimilatory nitrate reductase (nirB, nasA) and dissimilatory nitrite reductase (nirS) were found in Rhodoferax, Polaromonas and Rhizobacter (17–29 SGBs). Cold adaptation genes (csp) were universally present in cyanobacterial interacting bacterial genera (52 SGBs). Exopolysaccharide production proteins (exoZ, pslH) were predominantly found in Ferruginibacter, Hymenobacter, Rhodoferax, Rhizobacter, Arenimonas and Fimbriiglobus (37–59 SGBs), while genes encoding polysaccharide exporters (wza, cysE) were identified in most cyanobacterial-interacting bacterial genera (64–66 SGBs). Protein-tyrosine kinase (epsB) was detected in Rhodoferax, Polaromonas, Rhizobacter and Fimbriiglobus (25 SGBs).

Discussion

The dominance of cyanobacteria contribute significantly to bacterial biodiversity in cryoconite habitats

Our results demonstrate that cryoconite habitats and the overlying water harbor distinct bacterial communities, with Cyanobacteria predominating in the former (Fig. 1). This finding aligns with studies on temperate [47] and polar glaciers [19], where Cyanobacteria constitute 10–40% of the microbial communities. Their consistent presence across all four studied glaciers underscores their crucial role in cryoconite habitats.

Although Cyanobacteria dominate all cryoconite habitats, their abundance varies between surface cryoconites and cryoconite sediments (Fig. 2c–d). In the glaciers JMYZ and PL4, Cyanobacteria are more abundant in cryoconite sediments, likely due to the more dynamic nature of surface cryoconites. The higher retreat rates of glaciers JMYZ (5 m yr⁻¹, Yao et al. [27]) and PL4 (15 m yr⁻¹, Yang et al. [48]) further contribute to the instability of surface cryoconites, limiting both space for cyanobacterial growth. The dominant cyanobacterial genera identified in our study, including Phormidesmis, Tychonema, Nostoc, and Alkalinema, have also been reported in the cryoconites of polar and Asian glaciers [49]. Loriellopsis, originally described from caves in Spain and Greece [50]—environments typically classified as oligotrophic—has not been reported in other cryoconite habitats. Nevertheless, it has been observed in other glacial habitats, including suspended sediment in glacial streams [51] and the polar desert of the Canadian High Arctic [52]. Alkalinema was initially described from saline–alkaline lakes in the Brazilian Pantanal wetland, which can survive and produce biomass at a range of pH (pH 4–11) and be able to alter the culture medium to pH values ranging from pH 8.4–9.9 [53]. This suggests that Alkalinema may favor alkaline environments. In a previous study, the pH of the overlying water in cryoconite holes from Glacier TGL (Tanggula) was found to be alkaline, with a value of 9.52 ± 0.10 [54]. The widespread distribution of cyanobacterial genera across various glacial habitats suggests a global trend in the dispersal of these microorganisms [49]. This phenomenon is partly attributed to atmospheric diffusion, which facilitates the long-distance transport of bacteria via wind currents [55, 56]. Distinct bacterial organisms from the Indian subcontinent and the Indian Ocean are transported by the Indian monsoon and westerly winds, influencing the airborne bacterial communities over Tibetan glaciers [57]. Furthermore, the presence of Cyanobacteria in these environments may be linked to their adaptive mechanisms, such as light-harvesting systems, nutrient storage capacities, and tolerance to freeze–thaw cycles [58, 59]. Moreover, distinct cyanobacterial genera are observed across different glaciers (Fig. 2a–d). In glaciers JMYZ and PL4, the low nitrogen concentrations (Fig. S9) favor the dominance of Loriellopsis and Nostoc (order Nostocales), which are known for nitrogen fixation [60]. In contrast, glaciers QY and TGL, with higher nitrogen levels, support a greater abundance of Tychonema and Alkalinema, which do not fix nitrogen (Fig. 5). The abundance and community composition of Cyanobacteria in cryoconite habitats may be influenced by glacier dynamics, nitrogen availability, and climate warming, which together drive ecological shifts in cryoconite environments.

Our findings show that solid cryoconite habitats harbor more diverse bacterial communities than the overlying water (Fig. 3), which is consistent with previous studies. Sommers et al. [7] reported higher microbial alpha diversity in cryoconite sediments compared to the overlying water in the Taylor Valley, Antarctica. Cyanobacteria were key contributors to bacterial diversity, accounting for 6–10% of the richness in cryoconite sediments and 5–9% in surface cryoconites, significantly outnumbering their presence in the overlying water (Fig. S3a–d). As primary producers, Cyanobacteria exhibit greater diversity in cryoconite habitats than in other glacial environments, such as ice and snow [49, 61]. Their roles in carbon and nitrogen fixation [62] not only support their growth but also provide nutrients that promote greater diversity in the broader bacterial community.

Cyanobacteria as keystone taxa mediate structural–functional stability of cryoconite networks

Network analysis identified cyanobacterial ASVs as keystone taxa, exhibiting distinct enrichment patterns in cryoconite sediments and surface cryoconites compared to the overlying water systems (Fig. 4d). Their central role was further demonstrated by strong co-occurrence associations with Bacteroidetes (e.g., genera Flavobacterium, Ferruginibacter, and Hymenobacter) and Proteobacteria (e.g., genera Rhodoferax, Polaromonas, and Rhizobacter). This finding contrasts with a previous study, where cyanobacterial taxa were primarily associated with Actinobacteria taxa in the cryoconites of Arctic glaciers [19]. The differences between our findings and those of the previous study is likely attributed to the different environmental conditions in the cryoconite habitats. Actinobacteria, being gram-positive, non-motile, and possessing thick cell walls [63], may be better suited to the low-temperature environments of high Arctic ice caps. In contrast, under the intense radiation of the Tibetan Plateau, Bacteroidetes, which are gram-negative and capable of producing pigments [64, 65], may have a competitive advantage. Furthermore, Bacteroidetes are motile, which facilitates their ability to approach Cyanobacteria and acquire organic carbon.

Quantitative network robustness analysis highlighted the potential ecological significance of Cyanobacteria as key taxa influencing community stability. Cryoconite sediment and surface cryoconite subnetworks demonstrated higher structural resilience, with degree and clustering coefficients 0.7–1.4 and 0.4–0.6 times greater than those in overlying water communities (Fig. 4e). Node removal experiments revealed that eliminating Cyanobacteria caused a disproportionate breakdown of the network, with robustness of cryoconite sediments experiencing a 100% greater loss than that of surface cryoconites (Δ robustness index = 0.26 vs. 0.13 in C and SC; P < 0.05, permutation test, Fig. S5). This highlights the potentially critical, habitat-specific role of Cyanobacteria. In our study, the Cyanobacteria within the cryoconite network are filamentous. These filaments play an active role in forming ligand-metal complexes [66], enabling them to interact with and bind to clays and organic matter. As ecosystem ‘engineers’ in cryoconite habitats, Cyanobacteria contribute to both physical structuring and biochemical stabilization. They facilitate the aggregation of aeolian debris [10, 21] and stabilize granular substrates through their filamentous networks [67]. Moreover, Cyanobacteria and its interacting bacteria contain genes involved in extracellular polymeric substance (EPS) metabolism (Fig. 5). The coordinated action of these genes produces protective EPS matrices, forming biofilms that protect against cryoconite abrasion and environmental extremes while stabilizing microhabitats through porous EPS structures [68, 69]. This biogenic architecture provides thermal buffering and moisture retention [70], both of which are crucial for sustaining microbial life in polar environments.

As primary producers, Cyanobacteria provide other microorganisms with essential substances for survival through metabolic interactions [62]. Our network analysis revealed that Cyanobacteria exhibited strong co-occurrence relationships with heterotrophic partners from the phyla Bacteroidetes (e.g., genera Flavobacterium, Ferruginibacter, and Hymenobacter) and Proteobacteria (e.g., genera Rhodoferax, Polaromonas, and Rhizobacter). Metagenomic profiling further validated these interactions, showing that cyanobacterial MAGs encoded key genes involved in autotrophic processes, such as photosystem function (psaAB/psbAB) and carbon fixation (rbcL/prkA) (Fig. 5). These results align with previous studies, which identified Cyanobacteria as the dominant phylum responsible for carbon fixation via the Calvin-Benson-Bassham (CBB) cycle, rather than through other CO₂ fixation pathways, within the cryoconites of glaciers in polar and central Asia [71, 72]. Additionally, the nitrogen fixation marker gene (nifH) was found in cyanobacterial MAGs from Arctic cryoconite, specifically within the genus Nostoc [73]. In contrast, these Cyanobacteria interacting bacteria, were typically the dominant heterotrophic bacteria within cryoconite [73, 74], which lacked these functional modules involved in photosynthesis and nitrogen fixation. Beside nitrogen fixation, assimilatory nitrate reductase process was widespread in cyanobacterial and proteobacterial MAGs from cryoconite habitats, including Arctic and central Asia cryoconites [71]. This functional dichotomy underscores the role of Cyanobacteria as key primary producers in oligotrophic cryoconite habitats, facilitating autotrophic-heterotrophic coupling through the continuous supply of photoassimilates and bioavailable nitrogen [62, 75]. Indeed, the ecological interaction between photoautotrophs (such as Cyanobacteria or eukaryotic algae) and heterotrophs is common across diverse habitats [62, 76, 77]. Photoautotrophs convert CO₂ into organic carbon to fuel their growth and support their heterotrophic partners. In return, heterotrophs produce essential vitamins, such as B-vitamins [78], which are crucial for cyanobacterial growth. Additionally, heterotrophs contribute extra CO₂, provide protection against environmental stressors and predation, and often produce various secondary metabolites [79]. This mutualistic relationship is especially important in extreme environments, such as those facing desiccation, nutrient scarcity, or extreme salinity and temperature [80–82]. The exchange of metabolites allows heterotrophs to survive in conditions where organic carbon is limited. Moreover, the presence of Cyanobacteria promotes functional redundancy among heterotrophic microorganisms, thus enhancing the overall stability of the ecosystem [83].

These findings highlight the dual roles of Cyanobacteria in cryoconite habitats: as primary producers [72] and potential ecosystem engineers [9]. They may facilitate autotrophic-heterotrophic coupling, supporting microbial life in nutrient-limited environments. Additionally, their production of extracellular polymeric substances (EPS) could contribute to the stabilization of cryoconite networks, potentially enhancing resilience to environmental stress. By shaping their habitats, Cyanobacteria may create microhabitats that support diverse microbial communities, thereby playing a role in maintaining ecological stability and biodiversity in glacial ecosystems.

Conclusions

In summary, this study highlights the pivotal role of Cyanobacteria in shaping bacterial community diversity and structural-functional stability within cryoconite habitats. Our findings show that Cyanobacteria not only enhance bacterial richness in cryoconites but also drive beta diversity patterns, significantly influencing ecological niches across overlying water, cryoconite sediments, and surface cryoconite habitats. As core microorganism in these environments, Cyanobacteria may contribute to the stability of cryoconite microbial networks, as primary producers and potential ecosystem engineers in both cryoconite sediments and surface cryoconites. These findings emphasize the fundamental importance of Cyanobacteria in cryoconite ecosystems and deepen our understanding of microbial ecology in glacial systems. This research has important implications for predicting climate change impacts on high-altitude and high-latitude ecosystems, where cryoconite communities play vital ecological roles.

Author contributions

XZ: Writing—original draft, Methodology, Visualization, Formal analysis, Data curation. KL and YC: Writing—original draft, Writing—review & editing, Conceptualization. KL, YC, YL and PL: Writing—review & editing, Funding acquisition, Conceptualization. FW: Methodology, Data curation. ZZ: Data curation, Visualization.

Funding

National Natural Science Foundation of China,42330410,42406260,China Postdoctoral Science Foundation,2023M741487

Data availability

All raw 16S rRNA and metagenomic data from this study have been deposited in the NCBI Sequence Read Archive (SRA) under project IDs PRJNA1165965 and PRJNA813429, respectively. The key code has been uploaded to GitHub (https://github.com/Zhu19971130/R_code).

Declarations

Competing interests

The authors declare no competing interests.