Microbial communities in glacial lakes of Glacier National Park, MT, USA

Abstract

Glaciers are retreating, altering alpine ecosystems and creating new proglacial lakes. Compared to lakes fed by snowpack, glacial lakes are often enriched in nutrients and suspended solids that decrease light penetration. However, the microorganisms and biogeochemical conditions within these newly formed lakes are not well characterized. We describe the microbial communities in 14 glacial lakes in Glacier National Park, MT, USA using 16S rRNA gene amplicon sequencing and measurements of nutrient concentrations, water clarity, and other environmental properties. Microbial communities were distinct between lakes, including those connected to the same glacier, indicating the importance of site-specific biogeochemical and physical dynamics on these systems. Microbial community composition correlated with lake age (formation before or after the Little Ice Age) and conductivity but not with whether a lake was connected to a contemporaneous glacier > 0.1 km². Heterotrophic lineages found in other glacial systems were abundant and widespread, while cyanobacteria only reached appreciable abundances in shallow lakes where light reached the benthos. Relative abundances of ammonia and nitrite oxidizers correlated with concentrations of nitrate and nitrite, suggesting nitrification may help control nitrogen forms and concentrations in glacial lakes. We show that as glaciers recede, unique glacial lake microbial communities will be formed and lost with them.

Comparisons of microbial communities in glacial lakes in Glacier National Park, MT, USA using 16S rRNA gene sequencing showed that lake communities are structured by age, conductivity, and relative light penetration.

Introduction

“While standing upon that peak overlooking the terrain above the rim wall, we got the thrill of thrills, for there lay the glacier, shriveled and shrunken from its former size, almost senile, with its back against the mountain walls to the east of it, putting up its last fight for life. It was still what seemed to be a lusty giant, but it was dying, dying, dying, every score of years and as it receded, it was spewing at its mouth the accumulations buried within its bosom for centuries.”—Albert Sperry describing Sperry Glacier in Glacier National Park in 1938 (Sperry 1938, Johnson1980)

Glaciers are prominent features of mountain ecosystems, providing important hydrological and biogeochemical inputs to downstream watersheds and their inhabitants (e.g. Milner et al. 2017). The almost universal retreat of glaciers in recent decades has been predominantly attributed to anthropogenic warming (Marzeion et al. 2014, Roe et al. 2017, Zemp et al. 2019), with extensive loss of glaciers and snowpack occurring in the mountainous regions of North America (Mote et al. 2005, Hansen et al. 2006, Pederson et al. 2010, Pederson et al. 2011). Glacier melting is predicted to continue over the coming decades (Bosson et al. 2019, Bosson et al. 2023), with consequences for mountain ecosystems and their biota (e.g. Jacobsen et al. 2012, Giersch et al. 2017, Huss et al. 2017, Milner et al. 2017, Muhlfeld et al. 2020).

One consequence of glacial retreat is the formation of new proglacial lakes from meltwater impounded by glacial moraines or other barriers (Ashley 2002). Hundreds of new glacial lakes have appeared since the 19^th century, with large increases in both total number and size since the 1990s (Wang et al. 2014, Buckel et al. 2018, Shugar et al. 2020, Zhang et al. 2023). Glacial lakes have distinct abiotic conditions not found in lakes formed primarily by snowmelt or rain (Slemmons et al. 2013). Glacial erosion of bedrock delivers fine suspended particles, known as glacial flour, that reduce the downward flux of light through the lake's water column (Rose et al. 2014). These proglacial systems are also often enriched in nutrients, including dissolved inorganic nitrogen, relative to snowmelt or rainwater-fed lakes (e.g. Hood et al. 2009, Saros et al. 2010, Singer et al. 2012, Slemmons and Saros 2012, Sommaruga 2015; Warner et al. 2017, Vanderwall et al. 2024). How these differences in light and nutrients control lake biodiversity and biogeochemistry remains unclear.

Microorganisms (defined here as bacteria and archaea) are important contributors to biogeochemical cycling in glacial ecosystems (Anesio and Laybourn-Parry 2012, Boetius et al. 2015, Anesio et al. 2017). Microbial communities in glacial lakes are often distinct from lakes formed by other processes because of differences in physical and chemical properties (Sommaruga 2015, Peter and Sommaruga 2016). Glacial flour, changes in light penetration, and higher nutrient concentrations in glacial lakes can alter the composition of primary producers and their rates of primary production (Hylander et al. 2011, Slemmens and Saros 2012, Burpee et al. 2018, Navarro et al. 2018, Tiberti et al. 2020, Sejr et al. 2022), with elevated concentrations of suspended particulate material further influencing trophic interactions through the inhibition of filter-feeding organisms (Koenings et al. 1990, Sommaruga and Kandolf 2014). Together, these differences can alter the taxonomic diversity and ecosystem function of communities in glacial lakes relative to snowmelt-fed lakes (Peter and Sommaruga 2016). Microbial communities also vary with spatial and temporal influences in glacial input, showing apparent differences depending on distance to, magnitude of, and time since glacial connectivity (Schütte et al. 2009, Brankatschk et al. 2011, Göransson et al. 2011, Philippot et al. 2011, Fernández-Martínez et al. 2017, Wei et al. 2023, Guo et al. 2024). Thus, with increased warming, glacier-associated lakes will experience major changes in temperature, photosynthetically active radiation, and resource supply, altering the structure and function of their microbial communities (Bradley et al. 2014, Peter and Sommaruga 2016, Milner et al. 2017, Elser et al. 2020). However, the microbial diversity in glacial lake ecosystems and the drivers of community composition, especially across North America, have not been well documented.

Here, we evaluated microbial community composition in glacial lakes in Glacier National Park (GNP), Montana, USA. Known as the “backbone of the world” by the Blackfeet, “Ya·qawiswit̓xuki” (‘the place where there is a lot of ice’) by the Kootenai, and the “Crown of the Continent” by George Bird Grinnell, GNP is a World Heritage Site established in 1910 that has historically been home to hundreds of glaciers. Rare alpine species and indigenous communities have depended upon these mountains and their watersheds for thousands of years (Reeves and Peacock 2001, Muhlfeld et al. 2011, Craig et al. 2012, Giersch et al. 2017). GNP is representative of global patterns of deglaciation: of the estimated ∼150 glaciers present at the end of the Little Ice Age (LIA; ca. ∼1850), only ∼35% remained by 2005, with 90% of all glaciers losing more than half of their area (Martin-Mikle and Fagre 2019). In turn, widespread glacier loss across northwest Montana has created new glacial lakes of various ages and sizes with distinct environmental conditions and biological communities relative to snowpack-fed lakes, including elevated concentrations of phosphorus and nitrate and distinct and simplified zooplankton communities (Vanderwall et al. 2024). Here, we evaluated microbial community composition and its environmental drivers in 14 lakes in GNP using 16S rRNA gene sequencing and physical and chemical analyses. We asked a series of questions about GNP lakes: (i) How has lake size changed since the LIA and what are the current biogeochemical conditions? (ii) What are the abundant microorganisms in these lakes and are they similar to those in other cryosphere habitats? (iii) Do microbial communities differ between lakes and what environmental characteristics drive these patterns? (iv) Because inorganic nitrogen concentrations can be elevated in glacial lakes, to what extent are the distributions of nitrifying microorganisms (including ammonia and nitrite oxidizers) consistent with patterns in lake nitrogen pools across these habitats? Our data provide insight into the structure and function of microorganisms in glacial systems and the environmental factors that shape their distributions as new lakes begin their long-term ontogeny.

Methods

Watershed and glacial area extent

GNP covers an area of ∼4000 km² and as of 2015 was home to 26 active glaciers (Fagre et al. 2017, Martin-Mikle and Fagre 2019). The park was covered by glacial ice during the Last Glacial Maximum ∼20 000 years ago but was largely ice-free ∼11 000 years ago (Carrara 1986, Carrara 1989). Glaciers have been active for the last ∼6000 years, with their largest recent extent during the LIA (Carrara 1989, MacGregor et al. 2011, Munroe et al. 2012). Here, we studied 14 lakes within GNP that are present across nine basins that all contained glaciers during the LIA. Both current and LIA watershed, lake, glacier, and permanent ice and snow feature areas within each basin were calculated as previously described (Giersch et al. 2017, Muhlfeld et al. 2020). Briefly, watershed boundaries specific to each lake as determined by the terrain, along with the ice cover and lake areas within those boundaries, were digitized from National Agricultural Imagery Program and WorldView satellite imagery from 2003 to 2015 in ArcGIS v 10.2. Because some sampled lakes fall within a vertical series, with one lake potentially draining into the other, the lower lake by default contains the watershed area, ice cover, and lake area of the upper lake. Using this data, lakes were categorized based on whether they were connected to a glacier of a certain size. Lakes were considered connected to a contemporary glacier if they were in a basin with a glacier whose overall total size across the basin was >0.1 km², a threshold previously used to define glaciers in GNP (Martin-Mikle and Fagre 2019). Glaciers smaller than this threshold, along with their lakes, were considered “inactive.” We acknowledge that if a smaller threshold of 0.01 km² was used, all but two lakes would still be connected to a glacier. Lakes were also categorized based on age of formation using estimates of glacial extent during the LIA (Fagre and Martin-Mikle 2018, Martin-Mikle and Fagre 2019). Lakes that were completely covered by their glacier during the LIA were considered younger than the LIA (<150 years old), while those that were not completely covered by their glacier during the LIA were considered older (>150 years old). Normalized difference vegetation index (NDVI) of each lake's watershed was determined as previously described (Vanderwall et al. 2024).

Sample collection

Glacial lakes were sampled during July, August, and September of 2016–2018. The dataset represents a subset of lakes that were previously described by Vanderwall et al. (2024). Lakes were typically sampled once, although those associated with Sperry and Grinnell glaciers, two of the most well-studied and visited glaciers in GNP, were sampled across multiple years. Measurements and water samples were generally collected at the deepest point of each lake as estimated using a hand-held sonar depth finder. The bathymetry and morphometry of many of these lakes have not been extensively surveyed: thus, we refer to the observed maximum depth, while the actual maximum depth may be deeper. Profiles of temperature (°C), dissolved oxygen (mg O₂ l^-1 and % saturation), pH, and specific conductivity (µS cm^-1) were obtained using a Hydrolab MS5 multi-sensor sonde. Because lakes had variable depths, we standardized comparisons of several of the physical properties common to lakes by using measurements obtained at a depth of 1 m from within the mixed layer. Water transparency was assessed with a 20 cm diameter Secchi disk. Secchi depth as a percentage of the observed maximum lake depth was also calculated. Lakes were considered stratified if water temperature at the bottom of the lake differed by 0.5°C from the average temperature in the near-surface waters (between 0 and 3 m).

Depth-integrated (0–6 m depth) water samples were collected using a hose, except in 2016 when some water samples were obtained from the shoreline using an extendable sampling pole. If the lake was shallower than 6 m, water was collected from the surface until just above the bottom. Although this integrated sampling always included the mixed layer, this technique did not always reach the hypolimnion for all lakes. Water samples were collected in triplicate and are referred to as field replicates. Samples were processed in the field. Approximately 100–400 ml of water was filtered through prerinsed 0.45 µm nitrocellulose filters (Millipore) for nutrient and ion analyses. Filters were retained for subsequent analyses of chlorophyll a concentrations. Water samples and filters were held in snow-chilled containers and then in ice-chilled coolers near 4°C until returned to the laboratory where they were preserved, refrigerated, or frozen prior to analyses. Total organic carbon (TOC), total phosphorus (TP), soluble reactive phosphorus (SRP), total nitrogen (TN), nitrate+nitrite (NO₃^-+NO₂^-), ammonium (NH₄⁺), chlorophyll a, and cations and anions were processed and measured as previously described (Vanderwall et al. 2024). Inlet stream water for seven lakes that were not directly abutting the source glacier was also collected and processed for nutrients and cations and anions as described above. Pearson correlations were performed to compare chemical and environmental variables between lakes.

16S rRNA gene amplicon sequencing

Microbial community composition within lakes was determined using 16S rRNA gene amplicon sequencing. Approximately 100 ml of lake water was filtered onto either a 25 mm diameter, 0.2 µm pore size polyethersulfone filter (Supor, Pall Co., NY, USA; 2017–2018) or a 25 mm diameter, 0.2 µm pore size polycarbonate filter (2016) while in the field. Filters were submerged in RNAlater (2017–2018), immediately placed on snow and ice for transport back to the laboratory, and frozen at -80°C. DNA was extracted using a MasterPure DNA purification kit (Lucigen, WI, USA). The V4–V5 region of the 16S rRNA gene was amplified using the primers 515F-926R (Parada et al. 2016) in triplicate. Samples were pooled to equimolar proportions and sequenced in two libraries on an Illumina MiSeq at the University of Hawaiʻi at Mānoa and the University of Montana. Negative control extractions on blank filters were processed and sequenced in the same manner. Sequence data are publicly available under the National Center for Biotechnology Information (NCBI) Bioproject accession number PRJNA1206200.

Sequence data were processed using QIIME2 (Bolyen et al. 2019). Sequences were denoised, chimeras removed, and amplified sequence variants (ASVs) identified using DADA2 (Callahan et al. 2016). ASVs were classified against the SILVA 138 database (Quast et al. 2013). Further analyses were performed with the R (R Core Team 2024) package phyloseq (McMurdie and Holmes 2013). For downstream analyses, field replicate samples collected from the same lake on the same day were processed as individual samples and their results were averaged unless specified (Prosser et al. 2010, Lennon 2011). Sequences belonging to the domain Eukaryota were removed. Read depth averaged 15 335 reads per sample (minimum 5928, maximum 36 645 reads per sample). Sequences within the phylum Cyanobacteria identified as plastids were classified based on the highest blastn similarity against the PhytoREF database (Decelle et al. 2015) and their relative abundances were estimated. Plastid sequences were then removed prior to all further analyses. Following plastid removal, two samples had anomalously low read depths (Lake Y, samples S46 and S47) and were discarded. Samples were then rarefied to an equal sequencing depth of 5930 reads per sample.

Diversity metrics were calculated to compare microbial community composition across lakes and to determine their environmental drivers. Alpha diversity was calculated using the Chao1 index in phyloseq both with and without rarefaction: general patterns did not differ. Beta diversity analyses were performed using non-metric multidimensional scaling (NMDS) ordinations, correspondence analysis (CCA) in vegan (Oksanen et al. 2008), hierarchical clustering, and pairwise comparisons between lakes. Prior to beta diversity statistical analyses, samples missing environmental data were removed and ASV abundances and environmental data were averaged between field replicate samples collected on the same lake on the same day unless specified. To determine if averaging across field replicates was admissible, statistical analyses were also repeated by randomly sampling one field replicate from each lake on each date 100 times to generate 100 different sampling groups: results did not differ. NMDS ordinations, hierarchical clustering, and pairwise lake comparisons were performed using the Bray–Curtis dissimilarity index. CCA and analysis of variance (ANOVA) using the function anova.cca were performed to identify statistically significant environmental parameters. Beta diversity analyses revealed three broad groups of lakes based on age, conductivity, and relative light penetration: these groups are hereafter referred to as “lake types.” Permutational analysis of variance (PERMANOVA) with the function adonis2 was used to determine if the presence of a contemporary glacier > 0.1 km², lake age, or lake type influenced community composition. Kruskal–Wallis tests were performed to compare chemical and environmental variables between these different lake groupings.

We explored the distributions of taxa of interest across lakes in GNP. Core, abundant ASVs were identified as those present in at least 10 of the 14 lakes. To avoid certain lakes with higher sample numbers biasing ASV distribution estimates, rarefied samples were merged by lake and then rarefied again to an equal sampling depth across all lakes. To qualitatively evaluate the broader biogeography of abundant lineages, GNP ASVs were compared against the NCBI nucleotide database using the Basic Local Alignment Search Tool (BLAST; Altschul et al. 1990) and against the 16S rRNA gene database MicrobeAtlas 1.0 (https://microbeatlas.org/; Rodrigues et al. 2017) to identify similar taxa. Differentially abundant taxa between specific lakes of interest or lake types were determined using the package DeSEQ2 (Love et al. 2014). DeSEQ2 comparisons were performed on samples without rarefaction but with the removal of low-abundance sequences (McMurdie and Holmes 2014). Finally, because nitrogen concentrations can be significantly higher in glacial lakes relative to snowpack-fed lakes, we explored the distributions of known nitrifiers and their relationships with nutrient concentrations and other measured environmental properties. This included members of the ammonia-oxidizing families Nitrosopumilaceae and Nitrosomonadaceae and the nitrite-oxidizing families Nitrospiraceae and Gallionellaceae.

Results

Here we describe the microbial communities within 14 glacial lakes of GNP. While all lakes were originally formed by glacial activity, they encompass a gradient of current glacial influence (Fig. 1, Table 1, and Figs S1 and S2). Using the definition of a glacier as that with an ice surface area >0.1 km², the lakes studied here are associated with seven contemporary (Blackfoot, Chaney, Grinnell, Siyeh, Sperry) or recently inactive (Shepard, North Swiftcurrent) glaciers (Fagre et al. 2017, Martin-Mikle and Fagre 2019). Nine of the lakes are associated with glaciers >0.1 km², three lakes are in a watershed with an ice mass that falls slightly below that threshold (North Swiftcurrent Lake, Upper Shepard Lake, Lower Shepard Lake), and two lakes had glaciers that have completely disappeared over the last 100 years (Iceberg Lake, Sue Lake; Alden et al. 1914). High-resolution topographical analysis showed that ice extent within the watershed of every lake had decreased by at least 40% since the LIA (Fig. 1, Table 1, and Fig. S2), consistent with these glaciers losing more than half of their area overall (Gibson and Dyson 1939, Dyson 1948, Johnson 1980, Carrara and McGimsey 1981, Hall and Fagre 2003, Brown et al. 2010, Clark et al. 2017, Florentine et al. 2018, Lambert et al. 2020). Ten of the lakes have formed within the last ∼150 years, while four are older and predate the LIA (Cracker Lake, Grinnell Lake, Sue Lake, Iceberg Lake). Watershed area, glacial extent within the watershed, and lake area within the watershed were variable between lakes. When considering just the lakes that have formed since the LIA, lake area, lake area as a percentage of the total watershed, and lake area gained to glacial ice area lost were lowest in Blackfoot East Lake, Blackfoot West Lake, and Lake Y.

Figure 1.

Topographic relief maps of sampled lakes and their associated glaciers in Glacier National Park, Montana (MT), USA. The map area in green reflects Glacier National Park with the red inset box showing approximate locations of sampled lakes in figure panels A–G. Lakes and glaciers include (A) from left to right, Chaney Lake with Chaney Glacier, Sue Lake, and Upper and Lower Shepard Lakes with Shepard Glacier, (B) Iceberg Lake, (C) North Swiftcurrent Lake and North Swiftcurrent Glacier, (D) Upper Grinnell and Grinnell lakes with Grinnell, Salamander, and Gem glaciers, (E) Cracker Lake with Siyeh Glacier, (F) Ghost Lake, Lake X, and Lake Y with Sperry Glacier, and (G) Blackfoot East Lake and Blackfoot West Lake with Blackfoot Glacier. Colors reflect the area extent of lakes (blue), lake watersheds (purple), contemporaneous glaciers (orange), ice fields (yellow), and snow (white). Orange lines reflect Little Ice Age (LIA) glacial extent.

Table 1.

Lakes in Glacier National Park sampled in this study and their associated glacier, watershed, and lake surface areas.

Lake	Lake latitude (°N)	Lake longitude (°W)	Glacial basin	Total glacier area in basin (km2)	Lake area (km2)	Lake watershed area (km2)	LIA glacial area in lake watershed (km2)	Glacier + ice area in lake watershed (km2)	Glacier loss in lake watershed since LIA (%)	Lake % of watershed area	Lake area gained since LIA (%)	Lake area gained to glacier lost in watershed (%)
Blackfoot East Lake	48.6074	−113.6669	Blackfoot	1.74	0.01	1.66	1.44	0.50	65.62	0.45	100.00	0.78
Blackfoot West Lake	48.6069	−113.6707	Blackfoot	1.74	0.02	1.66	1.62	0.93	42.73	1.14	99.60	2.71
Chaney Lake	48.8531	−113.8297	Chaney	0.36	0.06	1.06	0.86	0.36	58.00	5.30	100.00	11.20
Grinnell Lake	48.7656	−113.7051	Grinnell	0.76	0.75 (0.33)	8.19 (5.56)	2.28 (0.79)	0.78 (0)	65.7 (100.0)	9.1 (4.1)	55.3 (0)	27.6 (0.0)
Upper Grinnell Lake	48.7570	−113.7304	Grinnell	0.76	0.41	2.63	1.49	0.78	47.30	15.71	100.00	58.90
Iceberg Lake	48.8134	−113.7471	Iceberg	0	0.49	2.31	0.41	0.00	100.00	21.00	46.33	55.19
North Swiftcurrent Lake	48.7904	−113.7659	North Swiftcurrent	0.07	0.01	0.55	0.22	0.07	67.71	2.60	95.85	9.22
Lower Shepard Lake	48.8665	−113.8553	Shepard	0.07	0.03 (.02)	0.63 (0.36)	0.44 (0.25)	0.09 (.02)	78.6 (91.1)	5.0 (2.7)	100.00 (100.00)	9.2 (7.2)
Upper Shepard Lake	48.8644	−113.8551	Shepard	0.07	0.02	0.27	0.18	0.07	61.60	5.51	100.00	13.20
Cracker Lake	48.7437	−113.644	Siyeh	0.20	0.17	6.34	0.52	0.27	48.87	2.62	0.00	0.00
Ghost Lake	48.6346	−113.7592	Sperry	0.83	0.03	0.23	0.16	0.03	79.09	11.93	100.00	21.51
Lake X	48.6311	−113.7629	Sperry	0.83	0.06	0.46	0.41	0.22	46.50	12.26	100.00	29.70
Lake Y	48.6284	−113.7702	Sperry	0.83	0.01	0.60	0.57	0.06	89.29	1.24	99.97	1.48
Sue Lake	48.8613	−113.8454	Sue	0	0.35	1.45	0.07	0.00	100.00	24.46	0.08	0.40

Notes: Lake watershed areas and the glacier area within them are specific to a given lake based on terrain. For lower lakes, the first value includes the summed areas of both the upper and lower lakes and the value in parentheses is the area specific to just the lower lake. The total glacier area in basin reflects the total size of the eponymous glacier in each basin even if part of the glacier is not within the watershed of a lake.

Physicochemical conditions

Lakes varied in their physical and chemical properties (Table 2 and Table S1). Lake depth ranged from 3 m (Blackfoot East Lake) to 32 m (Iceberg Lake; Fig. S3). Near-surface water lake temperatures ranged from a minimum of 0.6°C (Upper Grinnell Lake) to 13.5°C (Grinnell Lake). Vertical profiles of temperature revealed that some lakes were stratified at the time of sampling, while others appeared thermally homogeneous, consistent with top-to-bottom vertical mixing (Fig. S3). Because our water sampling was restricted to the upper 6 m of the water column, our analyses do not capture chemical or microbiological characteristics of the hypolimnion in these lakes (i.e. Cracker Lake, Grinnell Lake, Sue Lake). All water columns were oxic, with a minimum oxygen concentration of 8.8 mg l^-1 and a minimum O₂ saturation of 78.3%. Water clarity, as estimated based on Secchi depth, was lowest in Chaney Lake (0.3 m) and highest in Iceberg Lake (12.5 m). Secchi depth tended to be lower in lakes currently connected to glaciers >0.1 km² in size, but this difference was not statistically significant (P < 0.13; Fig. S4). Specific conductivity in the mixed layer ranged from 4.8 to 112.2 μS cm^-1 and was strongly correlated with concentrations of calcium (Ca²⁺), magnesium (Mg²⁺), reactive silica (SiO₂), sulfate (SO₄^2-), and sodium (Na⁺; P < 0.05; Fig. S5). Solute concentrations and specific conductivity were not significantly different between lakes connected to a contemporaneous glacier and those that have lost connection to them (Fig. S4). Solute concentrations were generally similar in lakes and their inlet waters. However, Cracker Lake had significantly higher solute concentrations relative to other lakes and its own inlet waters (Fig. S6). A positive correlation was observed between specific conductivity and maximum lake depth (r² = 0.43, P < 0.02; Fig. S5), with older lakes formed prior to the LIA generally having higher specific conductivity. Specific conductivity, Secchi depth, and chlorophyll a concentrations were not correlated, although Secchi depth normalized to maximum lake depth was generally lower in lakes with higher specific conductivity. NDVI, a proxy for glacial presence and amount of vegetation in the watershed, was lowest at Upper Grinnell Lake (-0.027) and highest at Grinnell Lake (0.127). Specific conductivity was positively correlated with NDVI (r² = 0.31, P < 0.04).

Table 2.

Lakes in Glacier National Park sampled in this study and their associated environmental and biogeochemical parameters.

Nutrient concentrations varied between lakes. NO₃^-+NO₂^- ranged from 18.5 ± 3.5 μg-N l^-1 (North Swiftcurrent Lake) to 141.8 ± 4.5 μg-N l^-1 (Upper Grinnell Lake). NH₄⁺ was variable and ranged between 4.7 ± 2.2 μg-N l^-1 (Lake X) and 13.4 ± 4.2 μg-N l^-1 (Blackfoot East Lake). NH₄⁺, NO₃^-+NO₂^-, and TN were not significantly different depending on whether a lake was connected to contemporaneous glacier or not (Fig. S4). SRP ranged from below detection (< 0.8 μg P l^-1; Lake X) to 5.2 ± 0.3 μg-P l^-1 (Chaney Lake). In general, lakes no longer connected to a glacier >0.1 km² in size had lower concentrations of SRP but not TP when compared with those still connected to a glacier.

Microbial community composition

We evaluated community composition within the glacial lakes using 16S rRNA gene sequencing. Prior to rarefaction and the removal of chloroplast sequences, 3853 ASVs were present. While the 16S rRNA gene may not be a quantitative marker for plastid-containing organisms due to differences in copy numbers between species, chloroplast sequences represented a large proportion of the communities, often constituting >10% and up to 80% of the sequence reads in some lakes (Fig. S7). Plastid sequences were generally unique between lakes. The most abundant sequences were related to members of the Chrysophyceae (Chromulinaceae) but also included sequences related to the Chlamydomonadales (Chlamydomonas, Oophila), Kolelliaceae, and Bacillariophyta. Sequences showed the highest similarity to those from other cold, polar habitats, including glaciers. Chloroplast relative abundance did not correlate with chlorophyll a concentrations (Fig. S7). Photosynthetic eukaryotes were visually apparent on some glaciers, snowfields, and meltwater pools surrounding the lakes (Fig. S1).

Following the removal of eukaryotic and chloroplast sequences and rarefaction, 3189 ASVs were present across all lakes at a sequencing depth of 5930 sequences per sample. The most abundant phyla included the Proteobacteria, Bacteroidota, and Actinobacteriota, each on average representing >10% of lake communities (Fig. 2). Other phyla of note included the Planctomycetota, Verrucomicrobiota, and Acidobacteriota, which each made up on average 2%–3% of the communities. Sixteen ASVs were present in at least 10 of the 14 lakes (Fig. 3, Table S2). Together, these ASVs represented ∼44% of the community in each lake, but in some cases accounted for over 75% of the community. The most abundant sequences were related to putatively heterotrophic members of the genera Flavobacterium (Bacteroidota), Polaromonas (Bacteroidota), hgcI clade (Actinobacteriota), Pedobacter (Bacteroidota), and Rhodoferax (Proteobacteria). These sequences were identical to those from other high-altitude mountain lakes, glacial ecosystems, and freshwater environments (Table S2; e.g. Simon et al. 2009, LLorens-Marés et al. 2012, Kang et al. 2017, Neuenschwander et al. 2018). In contrast, photosynthetic members of the phylum Cyanobacteria were relatively rare (<10% of communities), showing appreciable abundances in only three lakes: Blackfoot East, Blackfoot West, and Lake Y (Fig. 2 and Fig. S8). Abundant cyanobacterial sequences included those related to the filamentous and biofilm-associated genera Chamaesiphon, Planktothrix, Tychonema, and members of the family Leptolyngbyaceae. Sequences related to free-living, non-filamentous picocyanobacteria, such as Cyanobium, were rare and only reached abundances of >0.02% in Cracker Lake and Grinnell Lake. Cyanobacterial abundances were highest in shallow lakes where light penetration, as assessed based on Secchi depth, exceeded 50% of the maximum lake depth (r² = 0.28, P < 0.02; Fig. S8).

Figure 2.

The most abundant phyla present in glacial lakes of Glacier National Park. Lakes are organized by basin and relative location.

Figure 3.

Sixteen amplified sequence variants (ASVs) were present in at least 10 of the 14 lakes. Top: relative abundances of ASVs across lakes. ASVs are labeled at the family and genus level. Bottom: the total combined abundance of the 16 ASVs, summed down each column, within each lake. Lakes are organized by basin and relative location.

We compared the lakes using metrics of alpha and beta diversity. Alpha diversity was similar across all sites and generally was not different depending on whether a lake was currently connected to a contemporary glacier or not (Fig. 4 and Fig. S9). However, alpha diversity was noticeably higher in three lakes: the two Blackfoot lakes and Lake Y. Non-metric multidimensional scaling ordinations and pairwise comparisons based on Bray–Curtis dissimilarity and canonical correspondence analyses revealed that lake community composition was distinct depending on lake of collection (Fig. 4 and Figs S10 and S11; PERMANOVA, F = 6.66, r² = 0.95, P < 0.001). This was true even for lakes connected to the same glacier and when lakes had been sampled in >1 year (i.e., lakes connected to Sperry and Grinnell glaciers). Lake age, as characterized by formation before or after the LIA, was a statistically significant driver of community composition (PERMANOVA, F = 3.19, r² = 0.15, P < 0.002). In contrast, lakes did not cluster based on whether they were currently in a basin with a glacier with an ice area >0.1 km² (Fig. S9; PERMANOVA, P > 0.15). Lake community composition patterns followed a gradient from shallow lakes with high relative light penetration to older and deeper lakes with higher conductivity. Lake depth, specific conductivity, Secchi depth, Secchi depth as a percentage of maximum lake depth, lake area as a percentage of the total watershed area, and percentage of lake area formed since the LIA were all statistically significant predictors of community composition (ANOVA, P < 0.05; total inertia explained, 63%). The older lakes (Cracker, Grinnell, Iceberg, and Sue lakes) all appeared to further separate based on NDVI and temperature.

Figure 4.

Glacial lakes in Glacier National Park show differences in alpha diversity (A) and beta diversity based on NMDS ordinations (B; stress = 0.12) and canonical correspondence analysis (CCA; C). Arrows in C reflect statistically significant environmental data; only samples with all environmental data are shown. U, Upper; L, Lower; Lake % increase, lake area increase since LIA; Secchi % depth, Secchi depth as the percentage of maximum lake depth; Lake % watershed, lake area as a percentage of the total watershed area; Depth, lake maximum depth.

To identify microbial lineages that correlated with patterns identified in the alpha and beta diversity analyses, lakes were grouped based on age (lake formation since LIA), specific conductivity, depth, and relative light penetration (Secchi depth as a percentage of lake depth; Fig. S12). These “lake types” were broadly defined as three groups: (i) shallow and well-lit lakes (Blackfoot East Lake, Blackfoot West Lake, Lake Y) which had a maximum depth < 4.5 m, Secchi depths exceeding 40% of the lake depth, and had formed since the LIA but represented a small proportion of the watershed area, (ii) older, higher conductivity lakes (Cracker Lake, Grinnell Lake, Iceberg Lake, Sue Lake) which all had formed prior to the LIA and had specific conductivity exceeding 40 μS cm^-1, and (iii) all other lakes, which tended to be intermediate to the other two groupings. These groups were statistically different from one another based on beta diversity analysis (PERMANOVA; F = 4.05, r² = 0.32, P < 0.001). While only ∼2% of ASVs were present in all three groups, these shared ASVs constituted on average 46% of each lake community. Abundant, shared ASVs included members of the hgcI clade of the Actinobacteria (order Frankiales), members of the phylum Bacteroidota (genera Flavobacterium, Pedobacter, and Arcicella), and members of the Burkholderiales within the Proteobacteria (genera Polaromonas, Methylotenera, and Rhodoferax). Comparisons between lake types showed that shallower lakes with greater light penetration were enriched in members of the Cyanobacteria, including Tychonema, Leptolyngbyaceae, Chamaesiphon, and Calothrix (Fig. S13). Certain ASVs related to the genus Polaromonas were present and abundant in all lakes, albeit enriched in the shallow & lit lakes relative to the older ones. In contrast, the older, higher conductivity lakes had higher abundances (combined abundances in excess of 30%) of specific ASVs related to the hgcI clade of Actinobacteriota, along with an enrichment of the genera Dinghuibacter and Methylopumilus. The genus Flavobacterium was both abundant, in some cases reaching up to 60% of communities, and highly diverse, with distinct ASVs unique to each type of lake.

Glacial lakes can have higher concentrations of bioavailable nitrogen relative to snowpack-fed lakes. Therefore, we explored the diversity of nitrifying organisms across lakes in GNP as they may be an important control on N in these systems. Nitrifier relative abundances were altogether low across most lakes, typically reflecting <1% of the community (Fig. 5). Nitrifier abundances were positively correlated with both NO₃^-+NO₂^- and TN (r² > 0.4, P < 0.01), while relationships to NH₄⁺ concentrations and other variables were more ambiguous (Fig. 5 and Fig. S14). These patterns were driven in large part by Upper Grinnell Lake, which had abundances of nitrifiers that exceeded 5% of the community and included ASVs related to the ammonia-oxidizing archaea Nitrosopumilaceae, ammonia-oxidizing bacteria Nitrosomonadaceae, and the nitrite-oxidizing bacteria Nitrospiraceae and Gallionellaceae (Table S2). Upper Grinnell Lake has arguably the greatest current connectivity with its glacier, which plunges into the lake at one end, and sits at the head of a series of paternoster lakes that include lower Grinnell Lake. In contrast, Lake Y, Blackfoot East Lake, and Blackfoot West Lake, which represent some of the shallowest lakes in our study and had high light penetration, also had nitrifiers that reached abundances of ∼1%. Seasonal patterns of nitrifiers and concentrations of NH₄⁺, NO₃^-+NO₂^-, TN, and TOC in Upper Grinnell Lake, which was sampled four times over two years, showed the highest relative abundances and concentrations in July and lower values in August and September (Fig. S15). The elevated nitrifier abundances and nutrient concentrations in July coincide with periods when flow between Upper Grinnell Lake and lower Grinnell Lake are expected to be highest based on prior studies on seasonality in runoff (years 1959–1971 and 2004–2009; Johnson 1980). Other taxa of note which were differentially abundant between Upper Grinnell Lake and the other lakes sampled here included ASVs putatively involved in the cycling of C1 compounds. This included an enrichment in Upper Grinnell Lake of Methylotenera and the Acidobacteriota family Vicinamibacteraceae, which reached abundances of over 10% and 5%, respectively, while other relatively deep but warm lakes (such as lower Grinnell Lake and Cracker Lake) had notably higher abundances of Methylopumilus.

Figure 5.

Nitrifiers are abundant in Upper Grinnell Lake and positively correlate with nitrogen concentrations. (A) Relative abundances of families with known nitrifiers across lakes. Lakes are organized by basin and relative location. Nitrifier abundances and their correlation with (B) nitrate+nitrite (NO₃^-+NO₂^-), (C) total nitrogen (TN), and (D) ammonium (NH₄⁺). The grey lines depict least squares linear fits.

Discussion

Glaciers are rapidly retreating worldwide, altering adjacent ecosystems, their watersheds, and the organisms within them. Here, we describe the microbial communities in 14 lakes within GNP that span a gradient of glaciation. While similar taxa were present and abundant across lakes, microbial communities were distinct from one another, even when lakes were connected to the same glacier. These differences indicate that site-specific environmental dynamics, including glacial discharge, time since glacial retreat, lake morphometry, and surrounding landscape inputs, are important in structuring microbial communities. Overall, as glaciers disappear and their associated lakes transition towards clearwater conditions, the microbial communities that are unique to these habitats will be lost.

Severing hydrological connectivity between a glacier and its watershed alters water temperature, light availability (e.g. via input of suspended particles), and nutrient concentrations within downstream systems (e.g. Peter and Sommaruga 2016, Peter et al. 2018). Rather than showing trends based on their connection to a contemporaneous glacier (as defined by a size > 0.1 km²), glacial lake microbial communities were organized based on lake age, specific conductivity, depth, and apparent light penetration. Older lakes, which formed prior to the LIA and are likely >2000 years old (i.e. Cracker Lake; Munroe et al. 2012), had higher conductivity and were typically more distant from (Cracker, Grinnell) or no longer connected to (Sue, Iceberg) the glaciers feeding them. Because conductivity and solute concentrations are reflective of water–rock interactions (e.g. Thies et al. 2007, Salerno et al. 2016, Carling et al. 2017, Li et al. 2022), we interpret higher conductivity as a reflection of lake age through historical glacial and soil input, although this may also be a result of site-specific variations in glacial discharge or accumulation through longer water residence times. Overall, these findings are consistent with the observation that microbial communities in lakes, streams, and sediments are correlated with conductivity (Wilhelm et al. 2013, Liu et al. 2019, Kleinteich et al. 2022) and highlight the importance of time since formation—where lakes predating the LIA are distinct from those formed more recently—on structuring lake microbial community composition. While we did not observe differences based on current glacier presence, we emphasize that the lakes sampled here all have had active glaciers within the last 100 years: of the lakes defined as those without a contemporary glacier, North Swiftcurrent, Upper Shepard, and Lower Shepard lakes are still attached to an ice mass but that ice mass is smaller than 0.1 km² (Fagre et al. 2017), while Sue Lake and Iceberg Lake both had well-organized glaciers in 1914 that remained until at least the 1940s (Alden 1914, Dyson 1941). Therefore, continued input from glacial remnants to lakes or the possibility that lakes may change more slowly over time following deglaciation than forefield sediments (time scales of 10 s to 100 s of years; e.g. Nemergut et al. 2007, Kim et al. 2016, Peter and Sommaruga 2016, Pessi et al. 2019, Pothula and Adams 2022), may explain our findings. Comparisons against lakes that have been deglaciated for even longer (e.g. Vanderwall et al. 2024) or temporally following a lake before and after it loses glacial input would help further clarify how severing glacial connectivity impacts microbial community composition.

Glacial meltwater can increase turbidity through input of suspended solids, but also supply nutrients to lakes, thereby altering the types, abundances, and vertical distributions of photoautotrophs (Hylander et al. 2011, Slemmons and Saros 2012, Burpee et al. 2018, Navarro et al. 2018, Bourquin et al. 2025). Here, we found that relative abundances of picocyanobacteria in glacial lakes of GNP were generally low. Lakes that had the highest abundances of cyanobacteria (Blackfoot West, Blackfoot East, Lake Y) were shallow with near-complete light penetration and contained filamentous lineages rather than small, planktonic forms such as Cyanobium or Synechococcus. Filamentous, biofilm-forming, and surface-attached cyanobacteria are often abundant in and provide fixed organic carbon to shallow glacial streams and cryoconite holes (Mueller et al. 2001, Stibal et al. 2006, Pessi et al. 2019, Brandani et al. 2022, Brandani et al. 2023). The lakes enriched in cyanobacteria contained distinct communities and the highest alpha diversity in our dataset. These lakes make up relatively small proportions of both the total watershed and the lake area gained relative to glacier lost since the LIA, perhaps suggesting increased importance of snowmelt, rain, or groundwater rather than glacial input to these systems. Across all lakes, we found that instead of cyanobacteria chloroplast-related sequences belonging to picoeukaryotic photoautotrophs were abundant. Picoeukaryotes can be both numerous and diverse on glaciers and in high alpine lakes (Hylander et al. 2011, García-Descalzo et al. 2013, Slemmons et al. 2013, Kammerlander et al. 2015, Filker et al. 2016, Slemmons et al. 2017, Warner et al. 2017, Burpee et al. 2018, Winkel et al. 2022). Given wide variance in taxa-specific chloroplast abundances per cell, we acknowledge that the 16S rRNA gene is likely not a quantitative marker for eukaryotic lineages; nonetheless, our findings generally agree with the observation that photosynthetic eukaryotes may be more abundant than picocyanobacteria in glacial lakes, perhaps due to mixotrophic growth strategies or reflecting a cyanobacterial preference for higher and deeper-penetrating light fluxes (Hylander et al. 2011, Slemmons and Saros 2012). However, we note that some lakes, such as Iceberg Lake, were optically clear, as indicated by deep Secchi depths, and had high concentrations of NO₃^-+NO₂^- yet still had low abundances of cyanobacteria. Indeed, Saros et al. (2010) also report glacially influenced lakes that are clear, have high nutrient concentrations, but contain low chlorophyll concentrations, indicating light penetration is not the only control on phototroph abundances. A eukaryotic preference for higher N or P concentrations, limitation by micronutrients, or cold temperatures may all contribute to eukaryotes outcompeting picocyanobacteria in these systems and controlling overall abundance patterns (Tilman et al. 1986, Slemmons and Saros 2012, Warner et al. 2017, Ouyang et al. 2023). Future work should more specifically target the diversity, abundance, and photosynthetic potential of picoeukaryotic algae and their competitive abilities relative to cyanobacteria, including at depths below the mixed layer, to understand their functional roles as these turbid lakes transition toward clearwater systems.

The shared, core microbiome present within glacier-associated systems often contains many putatively heterotrophic lineages that are widely distributed across cryosphere habitats (Bourquin et al. 2022, Ezzat et al. 2025). Similarly, GNP's glacial lakes contained high abundances of heterotrophic taxa previously identified as abundant in glacially influenced environments, including members of the genera Flavobacterium, Polaromonas, hgcI clade (Actinobacteriota), Pedobacter, Rhodoferax, and Arcicella (e.g. Liu et al. 2006, Nemergut et al. 2007, Hell et al. 2013, Murakami et al. 2015, Peter and Sommaruga 2016, Peter et al. 2018, Tolotti et al. 2020, Miller et al. 2021). Glaciers have previously been considered geographically isolated habitats for microorganisms, showing differences at both the regional scale and between continents (Darcy et al. 2011, Franzetti et al. 2013, Gawor et al. 2016, Darcy et al. 2018). While glacier retreat threatens the loss of unique biodiversity (Milner et al. 2009, Jacobsen et al. 2012, Giersch et al. 2017), abundant prokaryotes in our study lakes showed identical 16S rRNA gene similarity to sequences from other aquatic habitats, suggesting the dominant taxa in GNP are widespread across the cryosphere and may be locally extirpated but not completely lost to deglaciation (e.g. Jungblut et al. 2010, Gokul et al. 2016). While appearing ubiquitous, these putatively heterotrophic lineages varied in abundance across lakes, suggesting niche differentiation; for example, older, higher conductivity lakes were enriched in ASVs related to the Actinobacteriota, while lineages related to Flavobacterium showed high diversity and specificity to certain lakes. Although TOC was not statistically different between lake types, one explanation for these distributions could be that the type or lability of carbon drives niche separation among these groups (e.g. Hood et al. 2009, Ghylin et al. 2014, Neuenschwander et al. 2018, Smith et al. 2018). Given that some representative members of abundant heterotrophs found in GNP contain genes for rhodopsins and aerobic anoxygenic photosynthesis (e.g. Sharma et al. 2009, Liu et al. 2021, Busi et al. 2022), energy generation via photoheterotrophy may supplement chemotrophic demand for organic carbon in glacial lakes. Further work characterizing and quantifying the forms and sources of organic matter to these systems, along with using culture-based and metagenomic analyses to identify preferred carbon substrates, will be needed to evaluate these hypotheses.

Glacially impacted lakes and streams can have nitrate concentrations significantly higher than those fed by snowmelt alone (Saros et al. 2010, Slemmons et al. 2017, Warner et al. 2017, Vanderwall et al. 2024). The source of this nitrate is enigmatic and likely stems from a combination of processes, including abiotic geochemical weathering, microbial nitrification, and atmospheric deposition (Wynn et al. 2007, Nanus et al. 2008, Montross et al. 2013, Colombo et al. 2019, Beard et al. 2022). Here, we show that elevated nitrate concentrations in glacial lakes were correlated with abundances of nitrifiers, consistent with ammonia and nitrite oxidation as potentially important sources of nitrate to these lakes. Nitrification and nitrifying organisms have been documented not only within glacial lakes (Peter and Sommaruga 2016, Tolotti et al. 2020) but also cryoconites, sediments, subglacial systems, and snow (e.g. Wynn et al. 2007, Baron et al. 2009, Boyd et al. 2011, Brankatschk et al. 2011, Ansari et al. 2012, Hell et al. 2013, Zarsky et al. 2013, Segawa et al. 2014, Lutz et al. 2015, Segawa et al. 2020, Wang et al. 2022, Yu et al. 2023). Upper Grinnell Lake, which had the highest abundances of nitrifiers when runoff was highest, is tightly coupled to Grinnell Glacier, which sits partially within the lake itself. While nitrifiers are typically found at depth in oceans and lakes and are thought to be inhibited by light (e.g. Karner et al. 2001, Merbt et al. 2012, Peoples et al. 2024), they were also present in the shallow and well-lit Blackfoot lakes and Lake Y. Given that nitrifiers have been found in glacial meltwater (Wilhelm et al. 2013, Kohler et al. 2020), one explanation is that nitrifiers and the nitrate produced as a byproduct of their metabolism may be associated with glaciers and then deposited via runoff into lakes. Observed monthly changes also highlight the potential importance of seasonality on microbial communities and nutrient concentrations in these systems (Hotaling et al. 2022, Winkel et al. 2022). Future work should explore whether nitrifiers are active in glacial lakes and quantify their impact on the form and distribution of inorganic nitrogen in these ecosystems, including in the hypolimnion of deeper, stratified lakes that may have been missed in our sampling. Such studies could include the use of N tracer and isotopic approaches to determine the source of N and estimate the flux between reduced and oxidized pools.

Conclusions

The number of active glaciers in GNP has decreased from >100 to <30 over the past century. As these glaciers disappear, they take with them unique habitats but leave behind lakes that are, at least for a period, influenced by the glacial remnants. While our study provides the first snapshot of the microbial communities associated with glacier-influenced lakes in GNP, future work should address these systems more thoroughly over both space and time. First, while our dataset spans a gradient of glacial influence, all lakes studied here have been influenced by a glacier at some point in the last 200 years. The incorporation of even older lakes (e.g. Vanderwall et al. 2024), together with those along a chronosequence of glacial connectivity and age, would help clarify community differences due to age and glacial input. Examples of lakes for further study include the paternoster lakes connected to Grinnell Glacier (Grinnell Glacier → Upper Grinnell Lake → Grinnell Lake → Lake Josephine → Swiftcurrent Lake) or Sperry Glacier (Sperry Glacier → Ghost Lake, Lake X, Lake Y → Avalanche Lake → Lake McDonald) relative to those that no longer have glacial activity at all (Upper Two Medicine Lake → Two Medicine Lake → Lower Two Medicine Lake). Second, while we sampled most lakes only once, glacial ecosystems are highly seasonal. Future work should evaluate how the highly compressed growing season experienced by these lakes influences the water temperature, biogeochemistry, and succession of microorganisms. Because GNP is home to regionally endemic species found nowhere else, our understanding of these systems and their associated microorganisms is not only important but time-sensitive as glaciers continue to disappear.

Author contributions

Logan Peoples (Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Visualization, Writing—original draft, Writing—review & editing), J. Joseph Giersch (Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Visualization, Writing—review & editing), Tyler H Tappenbeck (Data curation, Formal Analysis, Investigation, Methodology, Writing—review & editing), Joseph W Vanderwall (Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Writing—review & editing), John M Ranieri (Data curation, Formal Analysis, Investigation, Methodology, Writing—review & editing), Trista J. Vick-Majors (Data curation, Formal Analysis, Investigation, Methodology, Writing—review & editing), James Elser (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing—review & editing), and Matthew Church (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing—original draft, Writing—review & editing).

Conflict of interest

None declared.

Funding

Funding for lake sampling and analysis was provided by the Jessie M. Bierman Professorship (FLBS), the National Science Foundation (DEB 1951002), the U.S. Geological Survey (USGS) Northwest Climate Adaptation Science Center, and the USGS National Climate Adaptation Center.