Microbial community variation in cryoconite granules on Qaanaaq Glacier, NW Greenland

Cryoconite granules are aggregations of microorganisms with mineral particles that form on glacier surfaces. To understand the processes by which the granules develop, this study focused on the altitudinal distribution of the granules and photosynthetic microorganisms on the glacier, bacterial community variation with granules size and environmental factors affecting the growth of the granules. Size-sorted cryoconite granules collected from five different sites on Qaanaaq Glacier were analyzed. C and N contents were significantly higher in large (diameter greater than 250 μm) granules than in smaller (diameter 30–249 μm) granules. Bacterial community structures, based on 16S rRNA gene amplicon sequencing, were different between the smaller and larger granules. The filamentous cyanobacterium Phormidesmis priestleyi was the dominant bacterial species in larger granules. Multivariate analysis suggests that the abundance of mineral particles on the glacier surface is the main factor controlling growth of these cyanobacteria. These results show that the supply of mineral particles on the glacier enhances granule development, that P. priestleyi is likely the key species for primary production and the formation of the granules and that the bacterial community in the granules changes over the course of the granule development.

Greenland Ice Sheet, albedo, cryoconite granule, glacier ecosystem, bacterial diversity, Phormidesmis priestleyi

INTRODUCTION

Cryoconite is a dark-colored sediment that forms on the surface of glaciers and consists of mineral particles from the surrounding environment together with biogenic organic matter from in situ microbial production on glacier and the surrounding environment (e.g. Stibal, Šabacká and Žárský 2012). Originally observed on glaciers in northwestern (NW) Greenland (Nordenskiöld 1872), cryoconite holes have been widely reported to occur on glaciers in polar (Mueller et al.2001) and temperate regions (Margesin, Zacke and Schinner 2002; Takeuchi, Nishiyama and Li 2010). Diverse cold-adapted microbes live in cryoconite (Mueller et al.2001; Hodson et al.2008) and play important roles in biogeochemical cycling in glacial ecosystems (Porazinska et al.2004; Telling et al.2011; Stibal, Šabacká and Žárský 2012). On some glaciers, small particles of ‘cryoconite’ are aggregated and formed hundreds of micrometer to a few millimeter scale granular structure called ‘cryoconite granules’. Filamentous cyanobacteria dominate as primary producers in the microbial community of glaciers where they promote the formation granular structures (Takeuchi, Kohshima and Seko 2001). The aggregation process is facilitated by the microbial production of extracellular polymeric substances (Langford et al.2010). Cryoconite granules contain layers which can grow in thickness by 0.2 mm per year. These cryoconite granule layers were observed to persist for an average of about 3.5 years on a glacier in western China (Takeuchi, Nishiyama and Li 2010).

Molecular biological analysis has revealed high microbial diversity in Arctic and Antarctic cryoconite (Christner, Kvitko and Reeve 2003; Segawa and Takeuchi 2010; Edwards et al.2011, 2013; Cameron, Hodson and Osborn 2012; Zarsky et al.2013; Grzesiak et al.2015; Stibal et al.2015; Cameron et al.2016). Microbial community composition has been found to be affected by chemistry (Grzesiak et al.2015), and differences in community structure have been reported from geographically distinct sites (Segawa and Takeuchi 2010; Cameron, Hodson and Osborn 2012; Cameron et al.2016; Stibal et al.2015). The cryoconite absorbs solar radiation and causes excessive melting of the surrounding glacier ice, which causes the cryoconite to sink into the ice vertically, forming cryoconite holes (Gajda 1958). Cryoconite holes provide a more stable habitat for microbial communities, perhaps because they are less subject to washing out by streams of supraglacial meltwater. Cryoconite holes are distributed in the ablation zone of the Greenland Ice Sheet, and their depths increase with increasing altitude on glaciers in Thule (Gajda 1958) and near Uppernavik (Gribbon 1979). However, Uetake et al. (2010) recently showed that the depths of cryoconite holes on Qaanaaq Glacier were lowest at intermediate altitudes of the glacier, where the highest biovolumes were observed. The cryoconite granules in shallower cryoconite holes may be frequently disturbed and dispersed by meltwater running along the surface of the ice.

The high degree of light absorbency of cryoconite is attributable to substances they contain, such as the substantial amounts of humic substances (Takeuchi, Kohshima and Seko 2001; Takeuchi 2002) and pigmented green algae and cyanobacteria that grow on the snow and ice surfaces (Yallop et al.2012; Takeuchi et al.2014). Takeuchi, Kohshima and Seko (2001) proposed that unicellular green algae are more easily washed out by streams of meltwater than cryoconite granules are, so cryoconite granules are more stable (Hodson et al.2010; Irvine-Fynn, Bridge and Hodson 2011) and therefore have more continuous potential to reduce albedo on the ice.

Cryoconite holes on Qaanaaq Glacier contain both types of darkening biological material—cyanobacteria that dominate cryoconite granules in the intermediate ablation zone and pigmented green algae that are dominant in the lower ablation zone (Uetake et al.2010). In the intermediate zone, cryoconite granules are abundant, contributing to accelerated melting.

Here we examine the spatial distribution of cryoconite granules, their microbial communities and the role of microbes in their formation on Qaanaaq Glacier in NW Greenland. We compared C and N contents and the biomass of photosynthetic microorganisms in cryoconite granules collected at five different altitudes, we tested how the abundance of photosynthetic microorganisms is affected by environmental factors such as topographical information, nutrients and minerals, and we compared the microbial communities between different size class granules, based on 16S rRNA gene sequencing.

MATERIALS AND METHODS

Study area and sample collection

Field research was conducted on the Qaanaaq Glacier, on an outlet glacier from the small Piulip Nuna ice cap in NW Greenland (Fig. 1, Table S1, Supporting Information) in July 2012 as part of the scientific project ‘Snow Impurity and Glacial Microbe effects on abrupt warming in the Arctic (SIGMA)’ (Aoki et al.2014). The elevation of the top of the ice cap is approximately 1110 m a.s.l. (Sugiyama et al.2014), and the entire area of the Qaanaaq Glacier is accessible by walking from the village of Qaanaaq. We collected samples at five sites (Table S1, Supporting Information) along a transect line of the glacier. At each site, surface ice including cryoconite granules from a 0.2 m × 0.2 m area was collected into sterile Whirl-Pak plastic bags (Whirl-Pak, Nasco, WI, USA). A stainless steel scoop was used to sample cryoconite for the granule size experiment (analysis of DNA, carbon and nitrogen mass), DNA was removed using a DNA AWAY wipe (Thermo Fisher Scientific, MA, USA) to remove DNA just before sampling, and the scoop was rinsed by scooping surface ice from a mock sampling site. Samples for chemical analysis were placed into another Whirl-Pak using another pre-cleaned stainless scoop. Samples for estimating biomass by microscopic observation were melted and then preserved as a 3% formalin solution in 30 mL clean polyethylene bottles. All samples were taken in five replicates per site, and slope angles were measured in 10 different locations per site (Table S1, Supporting Information).

Figure 1.

Satellite image of Qaanaaq Glacier and study sites in 30 July 2012. Images were taken by the World View-2 satellite.

Samples were kept cold on ice until reaching Qaanaaq Village. All samples except those for chemical analysis were then deposited into a portable deep freezer (SC-DF25, Twinbird, Niigata, Japan) and transported frozen at −40°C to −10°C and were stored at −20°C in the lab until just prior to analysis. All melted ice samples for chemical analysis were filtered through syringe filters (Advantec, Tokyo, Japan) on syringes (SS-10SZ, Terumo) and transported with other frozen samples as described above.

Microscopic observation

The samples were sonicated for 10 min to loosen sedimentary particles. Aliquots of 2–100 μL of the sample water were filtered through hydrophilic membrane filters (pore size 0.45 μm, Millipore JHWP01300) and then the number of algae on the filter was counted (one to three lines on the filter) under an optical microscope (OLYMPUS BX51, Japan). Three filter membranes taken from the same sample were counted for each sample. The algal cell concentration (cells mL⁻¹) of each sample was obtained from the mean cell count. Mean cell volumes were estimated by measuring the size of 5–50 cells from each taxa. These data were used to estimate the total algal biomass (mL m⁻²).

Size sorting of cryoconite granules

All sediments obtained from the glacier surface were filtered through six different mesh sizes (30, 100, 250, 500, 750 and 1000 μm) of 47 mm diameter nylon mesh filters (30 and 100 μm (NRS-030 and NRS-100, respectively, Nippon Rikagaku Kikai, Tokyo, Japan) 250, 500, 750 and 1000 μm (NMG66, NMG38, NMG27 and NMG20, respectively, SEMITEC, Osaka, Japan)) in a polysulfone filter holder (KP-47H; Advantec). After melting the samples, sediment and melted ice were poured onto 1000 μm mesh size filters. Filters were washed enough using ultrapure water (Milli-Q^® Advantage A10; Merck Millipore, MA, USA) to separate cryoconite granules larger than each mesh size. After sorting, filtrate water samples containing smaller grains were filtered through successively smaller mesh filters until reaching mesh size 30 μm, and were divided into six different cut-offs (size 30, size 100, size 250, size 500, size 750 and size 1000). Figure S1 (Supporting Information) shows stereomicroscopic images of representative cryoconite granules from each size cut-off. All residues were kept frozen in 8 mL plastic tubes until further analysis.

Carbon and nitrogen measurements

Total carbon (TC), total nitrogen (TN) and cryoconite granules mass (g m⁻²) were measured to compare levels of microbial production on different cryoconite granule sizes and distribution patterns on the Qaanaaq Glacier. A portion of the size-sorted cryoconite granules was lyophilized in a freeze-dryer under vacuum (FDU-810; EYELA, Tokyo, Japan). Freeze-dried cryoconite granules were weighed on a digital balance (XPE205, Mettler Toledo, Greifensee, Switzerland). TC and TN percentages were simultaneously determined by the dry combustion method using an NC analyzer (Sumigraph NC-22A, Sumika Ltd, Tokyo).

Ion concentrations

Because the concentrations of NH₄⁺, NO₃⁻, PO₄³⁻ and SiO₃²⁻ ions were below the detection limit of ion chromatographic methods, the ion concentrations in meltwater samples were determined colorimetrically using an Autoanalyzer II (Bran+Luebbe, Norderstedt, Germany).

DNA extraction and high-throughput sequencing of 16S rRNA genes

Microbial community structure in the samples was analyzed by sequencing the 16S rRNA gene using the MiSeq sequencer (Illumina, San Diego, CA). Genomic DNA was extracted using the FastDNA SPIN Kit for Soil DNA Extraction (MP Biomedicals, Santa Ana, CA). Partial 16S rRNA gene sequences including the V3 and V4 regions were amplified using the primers Bakt_341F and Bakt_805R with Illumina overhang adaptor sequences attached to their 5′ ends. PCR amplification of the 16S r RNA gene, reaction clean-up, index PCR and sequencing were performed following Illumina methods for ‘16S metagenomic sequencing library preparation’ (https://support.illumina.com/content/dam/illumina-support/documents/documentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf). Reactions to amplify sequences with Illumina overhang adaptors were performed in triplicate and the amplicons were pooled prior to index PCR using a Nextera XT Index kit (Illumina). The resulting DNA was mixed with Phi X control DNA in a ratio of 80:20 and used as a template for paired-end sequencing using a MiSeq Reagent Kit v3 (600 cycles) and the MiSeq Desktop Sequencer.

USEARCH (Edgar et al.2011) was used to cluster sequences into operational taxonomic units (OTUs) if they were at least 98% identical. Reads were assigned to OTUs with a closed-reference OTU-picking protocol in QIIME (Caporaso et al.2010). USEARCH was also used to check for chimeric sequences. Taxonomic assignment was conducted using representative 16S rRNA gene sequences obtained by sequencing these glacier samples (LC076700–LC076750) (Segawa and Takeuchi 2010) as blast queries against the Silva 119 database (Quast et al.2013). Sequences related to eukaryotes, chloroplasts and unknown organisms were eliminated. Weighted UniFrac distances were computed between all samples in each replicate. To visualize relationships among significant OTUs and sample types, β diversity among all samples from all sites was analyzed using weighted UniFrac.

A table of the relative abundances by genus and OTU was constructed and indicator genera and OTUs of the smaller (30–249 μm diameter) and larger (>250 μm diameter) cryoconite granules were identified using the indval function in the labdsv R package (R Core Team 2015). Genera and OTUs with indicator values greater than 0.6 and P < 0.05 were chosen as indicators for each granule size.

Datasets from high-throughput sequencing of 16S rRNA genes have been submitted to the NCBI Sequence Read Archive under accession numbers DRA004099–DRA004103 (http://trace.ddbj.nig.ac.jp/dra/index.html).

Statistical data analysis

The spatial distributions of cryoconite granules (percentage and mass) were analyzed by two-way ANOVA followed by Tukey's HSD packaged in the XLSTAT software (https://www.xlstat.com). Spearman rank correlations between the masses of large cryoconite granules and the abundance of each photoautotroph were calculated using XLSTAT software.

Multivariate statistics for detrended correspondence analysis and redundancy analysis (RDA) were performed using Canoco 5.0 software (ter Braak and Šmilauer 2012). For RDA, altitude, slope angle, nutrient concentrations (phosphate, nitrate and ammonium ion), amount of mineral particles in smaller fraction (30–249 μm) and mineral composition (Nagatsuka et al.2014) were selected as environmental factors (Table S1, Supporting Information) with potential effects on microbial growth, and were screened by forward selection: Monte Carlo permutation test with 499 permutations (Šmilauer and Lepš 2014.). The amounts of mineral particles in the smaller size fraction were estimated to subtract carbon weight from the total dry weight of cryoconite granules, as these fractions contain less carbon than more developed granules.

RESULTS

Altitudinal distribution of photosynthetic microorganisms

We identified six types of photosynthetic microorganisms including green algae (Ancylonema nordenskioldii, Mesotaenium berggrenii, Chloromonas sp., Cylindrocystis brebissonii and an unknown small round green alga) and Oscillatoriacean cyanobacteria. The species composition was very similar to that found in a previous study on the same glacier in 2007 (Uetake et al.2010). The average biomass of all photosynthetic microorganisms at all study sites ranged from 0.172 mL m⁻² (QA1) to 0.586 mL m⁻² (QA4) (Fig. S2, Supporting Information). The dominant photosynthetic microorganism was A. noldenskioldii, which comprised 44%–83% of total biomass (Fig. 2) and exhibited its greatest amount at QA2. Biomass of A. noldenskioeldii at QA2 was significantly greater than at QA1 (P = 0.030) or QA4 (P = 0.038). The second dominant species was M. berggrenii, which comprised 1%–36% of total biomass and exhibited its greatest amount at QA3. However, no significant differences in biomass amounts of M. berggrenii were observed among sites (Fig. S2, Supporting Information). Cyanobacteria in the Oscillatoriaceae comprised 1%–26% of total photoautotrophic biomass, with an especially large amount only at QA4, which was significantly different in terms of biomass from the other sites (P < 0.004).

Figure 2.

Altitudinal distribution of photosynthetic microorganism biomass (proportion) at five sites on the Qaanaaq Glacier.

Altitudinal distribution of carbon and nitrogen content in cryoconite granules

The average TC percentage of cryoconite granules ranged from 0.67% to 5.23% and that of TN ranged from 0.08% to 0.47% (Fig. 3, Table S2, Supporting Information) throughout all study sites. Altitudinal abundance of average cryoconite granule mass, TC and TN for each granule size cut-off are shown in Fig. 4 and summarized in Table S3 (Supporting Information). The total mass of cryoconite granules, TC and TN summed for all granule sizes were highest (cryoconite granule mass: 47 g m⁻² dry wt, TC: 0.73 g m⁻², TN: 0.079 g m⁻²) at QA4 and lowest (cryoconite granule: 6.2 g m⁻² dry wt, TC: 0.08 g m⁻², TN: 0.009 g m⁻²) at QA1. The masses of the larger cryoconite granule size categories were significantly larger at QA4 (Table S4, Supporting Information).

Figure 3.

Altitudinal distribution of carbon and nitrogen content in each of six size fractions of cryoconite granules from five sampling sites.

Figure 4.

Average mass of cryoconite granules, TC and TN in each of six size fractions from five sampling sites.

The average percentages of TC and TN in smaller cryoconite granules (30–100 μm) ranged from 0.81% to 1.19% and from 0.10% to 0.11%, respectively (Table S1, Supporting Information). The percentages of TC and TN in larger cryoconite granules (250–1000 μm) ranged from 2.10% to 4.54% and from 0.12% to 0.44%, respectively. TC and TN percentages of larger granules at all sampling sites were significantly higher (P < 0.001) than those of smaller granules.

Multivariate analysis of photosynthetic microorganisms communities and environmental factors

Because the gradient length of the first axis was less than 4.0 in detrended correspondence analysis, and detrended correspondence analysis is difficult to use for absolute amounts (ter Braak and Šmilauer 2012), we used the RDA method. RDA of the biomass of photosynthetic microorganisms was conducted with environmental factors listed in Table S1 (Supporting Information). After forward selection, only the amount of mineral particles in smaller fraction (30–249 μm) was significantly correlated (P = 0.006) with the biomass of photosynthesis microorganisms (Fig. 5). The biomass of Oscillatoriacean cyanobacteria was most closely related to mineral particle concentrations, particularly at sampling site QA4. Otherwise, the biomass of A. noldenskioldii, which is a major alga on snow at all sites, was conversely related to that of Oscillatoriaceae.

Figure 5.

RDA six types of photosynthetic microorganisms and environmental factors at five sites on the Qaanaaq Glacier. Only mineral amounts that are significantly different (P < 0.01) are shown in figure.

Bacterial communities change depending on cryoconite granule size

After clustering and removal of chloroplastic and mitochondrial gene sequences, bacterial 16S sequence counts ranged from 48 430 to 181 108 and the mean was 128 639 (Table S5, Supporting Information). Taxonomical classification by QIIME compared with the Silva 119 database shows 11 major phyla (present as more than an average of 0.1% of total sequence data) in cryoconite granules from different sites and size fractions (Fig. 6). The major phyla present included Acidobacteria (5.97%–68.4%), Cyanobacteria (0.01%–47.10%), Proteobacteria (7.44%–41.70%), Actinobacteria (0.84%–37.82%), Bacteroidetes (9.06%–32.51%), Chloroflexi (0%–16.90%), Armatimonadetes (0.003%–7.78%), WD272 (0.01%–4.41%), Deinococcus-Thermus (0.0%–2.87%), Planctomycetes (0.0%–1.13%) and Candidate_division_TM7 (0.03%–0.93%).

Figure 6.

Relative abundance (%) of bacterial phyla in each size of cryoconite granule from each site.

Genus-level classification for organisms representing more than an average of 0.1% of total sequence data shows the relative abundance of 43 major genera comprising 95.1%–99.3% of total sequence data (Fig. S3a, Supporting Information) and includes only two groups of cyanobacteria: Phormidesmis (0%–45.6%) and an unclassified bacteria within the WD272 clade (0%–1.9%). The filamentous cyanobacteria found in these samples included only one group. This result indicates lower diversity than previously found in the reports of cryoconite community structure from Antarctica (three species of Phormidium and one species of Leptolyngbya from Christner, Kvitko and Reeve 2003) and China (three species of Phormidium, one species of Leptolyngbya, one species of Geitlerinema, one species of Limnothrix from Segawa and Takeuchi 2010). Zeng et al. (2013) showed in a study from Svalbard the same level of diversity of simple filamentous cyanobacteria and the same species (formerly known as Phormidium priestleyi).

The top 50 major OTUs, with representative percentages of greater than 0.01%, are listed in Table S6 (Supporting Information). The coverage against total sequences range from 48% to 58% in each sample, but a higher percentage of the sequences of the top four OTUs (OTU1–OTU4), including Phormidesmis and cyanobacterial species, were covered (23% to 49%).

Most of the related uncultured bacterium clones in 28 of the top 50 OTUs, or 36%–95% of the 50 major OTUs (Fig. S3b, Supporting Information), were initially identified in studies of microbial communities in glacial environments. Among these, 18 OTUs are similar to the uncultured bacteria from Austre Lovenbreen, Svalbard (Zeng et al.2013), 6 OTUs were similar to uncultured bacteria from Gulkana Glacier, Alaska (Segawa et al.2010) and 1 OTU was similar to uncultured bacteria from surface material at the front of Mittivakkat Glacier, Greenland, or from basal ice from Greenland Ice Sheet, Muztag Ata Glacier (An et al.2010) or from the Qiangyong Glacier in China (GU246829).

α and β diversity of bacterial communities and indicator species analysis

Two metrics of α diversity (using Simpson's Reciprocal Diversity Index or Shannon's Diversity Index) in all 18 samples are shown in Table S5 (Supporting Information). The α diversities of both size 30 and size 100 granules from QA1, the lowest altitude site, were relatively higher than those of granules from other sites. Despite the large difference in carbon and nitrogen percentages between size 100 and size 250 granules, there are no clear differences in α diversities between smaller and larger cryoconite granules at QA3 and QA4.

Otherwise, β diversity calculated by weighted UniFrac dendrogram showed clear differences between large granules and small cryoconite granules, with less than 13.5% similarity (Fig. 7). After indicator species analysis at the levels of both genus and OTU, 7 and 44 genera and 7 and 18 OTU were identified in the small and large granule cluster, respectively (Table S7a and b, Supporting Information). Edaphobacter was dominant in the small granule cluster. However, Edaphobacter belongs to OTU1 in the present analysis (see OTU1 in Fig. S3, Supporting Information) and is mostly related to the genus Actinobacterium with 97.6% similarity due to database issue. The other indicator OTUs were <96% similar to known isolates and so their identity is unclear. In the large granule cluster, the genus Phormidesmis was dominant, and Phormidesmis belongs to OTU (see OTU2 in Fig. S3) and is mostly close to P. priestleyi with 99.8% similarity. Similarly only OTU 24 was highly similar to Rhodococcus sp., while other OTUs showed lower similarity to known species.

Figure 7.

UniFrac dendrogram showing the similarity of bacterial community structures between sample size and sites.

DISCUSSION

Relationship between photosynthetic microorganisms and cryoconite granules

Altitudinal distribution patterns of green algae and cyanobacteria biomass clearly differ within even in a single small glacier (Fig. 2, Fig. S2, Supporting Information). The green alga Ancylonema nordenskioeldii was the most abundant species at QA2, while Oscillatoriacean cyanobacteria were significantly more abundant at the intermediate altitude of the glacier (QA4). This pattern is nearly identical to that found previously at the same site (Uetake et al.2010). Therefore, we argue the spatial distribution of photosynthetic microbes is stable on this glacier.

Correlations between the masses of large cryoconite granules and biomasses of each photosynthetic microorganism revealed that only the biomass of Oscillatoriacean cyanobacteria was correlated with the amount of developed granules (R² = 0.900, P = 0.083). Therefore, this result suggests that Oscillatoriacean cyanobacteria contribute substantially to granule formation.

Possible environmental factors affecting Oscillatoriacean cyanobacterial growth

Environmental factors that could control the growth of photosynthetic microorganisms in glacier ecosystem include altitude, slope angle, nutrient concentration, mineral particle amounts or composition, distance from the glacier terminus and surface hydrology. Geographical position (altitude and latitude) is an important factor to determine the temporal period of snow cover. Winter snow will melt faster at lower altitude and latitude, and the altitude of the snow line in summer varies from year to year. This annual environmental variation causes differences in algal (Takeuchi 2001), bacterial (Segawa et al.2011) and fungal (Uetake et al.2012) abundance and species compositions at each altitude. Slope angle can be a significant controlling factor within the relatively flat bare ice zone on the Greenland Ice Sheet, because meltwater could wash away both microorganisms and nutrients required for growth. Stibal et al. (2012) and Edwards et al. (2011) also suggest that surface hydrology could explain the differences in bacterial diversity between different glaciers. Nutrients such as nitrogen and phosphorus are potential growth-limiting factors in glacier ecosystems (Stibal et al.2009; Telling et al.2011). Furthermore, inputs of carbon, nitrogen and phosphorus promote bacterial growth in cryoconite granule layers (Säwström et al.2007). Lastly, mineral particles can become sources of phosphorus upon dissolution by either biotic or abiotic agents (Welch, Taunton and Banfield 2002; Stibal et al.2009). Distance from the terminus of a glacier can be an important factor affecting microbial growth on the Greenland Ice Sheet due to variation in the supply of allochthonous organic carbon from adjacent deglaciated areas (Stibal, Šabacká and Žárský 2012). Because Qaanaaq Glacier is a small glacier surrounded by glacier foreland, the influx of allochthonous organic carbon is not likely to vary with the altitude of the sampling sites, and the effect of distance from the glacier terminus on microbial growth is likely negligible.

Among the six environmental factors examined, RDA shows that only the amount of mineral particles in smaller fraction (30–249 μm) was correlated to the biomass of photoautotrophs (Fig. 5). Mineral particles can be sources of phosphate in glacier ecosystems (Stibal et al.2012) and are generally less influenced by meltwater. Based on our RDA results, we assume that small mineral particles might be bases or footholds for the growth of filamentous cyanobacteria. Microscopic observation indicates that filamentous cyanobacteria are not free living like snow algae, but are often attached or covered with small mineral particles during the early stages of cryoconite granulation. The presence of minerals seems to affect the colonization by microorganisms. The types of mineral components present are likely related to the growth of microorganisms, but silicate-containing minerals measured by X-ray diffraction (Nagatsuka et al.2014) are not correlated with the presence or growth of photosynthetic microorganisms. However, other interactions between cyanobacteria and particular minerals present on glacier ice had not yet been tested in incubation experiments. Therefore, in the future, the relationship between glacier ice mineral composition and microbial growth should be tested by incubating microbial isolates under various controlled conditions and varying the amounts or types of minerals present.

Potential bacterial communities associated with granulation

There are two types of microorganisms that reduce the reflection of solar radiation from glacier surfaces: green algae with dark intracellular pigments (e.g. Ancylonema noldenskioeldii, Mesotaenium berggrenii, Remias, Holzinger and Lütz 2009; Remias et al.2012), and microbes living in dark-colored cryoconite granules and promoting their aggregation. In this study, QA1 and QA2 were dominated by the former, QA4 by the latter and QA3 was an intermediate, containing both types.

In contrast to the pigmented green algae, Oscillatoriacean cyanobacteria (Phormidesmis priestleyi) do not contain dark intracellular pigments. The dark color of cryoconite granules is caused by the combination of mineral particles and humic substances. Humification can occur due to residual or inherited humin, polycondensation of humic precursors or bacterial synthesis (Gobat, Aragno and Matthey 2004). Although the characteristics of humic substances in cryoconite granules on the Qaanaaq Glacier were not determined in this study, bacterial humification is presumed to increase with increase in carbon during granulation process.

Humification occurs within single cryoconite granules, where the interactions between microorganisms, minerals and nutrient cycles are likely more important than the effects of the external environment on aggregation.

To understand the granule development process, we focused on the difference between the microbial communities in smaller and larger granules. The difference between their β diversity and the higher number of indicator species in larger granules suggests some degree of selection of the bacterial species inhabiting the larger granules (Tables S5–S8, Supporting Information). The most notable phylum in large granules are cyanobacteria, which are present as four indicator genera. Both Phormidesmis and Leptolyngbya are essential filamentous cyanobacteria that can contribute to binding mineral, microorganisms and organic matter together by production of glue like extracellular polymeric substances and fix carbon by photosynthesis (Hodson et al.2010; Langford et al.2014). These are certainly the most important bacteria in large cryoconite granules, because these filamentous cyanobacteria enlarge the surface of granular structure slowly over time. On bare ice surface, flow movement of cryoconite granule is relatively slow and stable (Irvine-Fynn, Bridge and Hodson 2011), and large granules (millimeter scale) are formed over at least several years. Their growth is probably slow due to the multiple environmental stress factors present in the supraglacial ecosystem (e.g. strong UV radiation, low temperature and low nutrients). Oxygen concentrations in the inner parts of the granules may decrease, which can affect microbial carbon and nitrogen cycling in the granules (Segawa et al.2014). Bacteria distinct from those commonly found in cryoconite may assume roles in carbon and nitrogen cycling processes under these conditions. However, due to the low similarity of OTUs detected in this study to known isolates it is difficult to infer their functions in the system. To further examine the potential functions of bacteria in cryoconite granules, more analyses are required, such as metagenomics, metatranscriptomics and DNA microarrays, to detect functions in whole communities of microorganisms by applying the same sampling strategy with comparisons of multiple sites and granule sizes as used here.

SUPPLEMENTARY DATA

Supplementary Data.

The authors wish to thank members of the ‘Snow Impurity and Glacial Microbe effects on abrupt warming in the Arctic (SIGMA)’ project and the ‘Green Network of Excellence (GRENE)’ program and ‘Arctic Challenge for Sustainability (ArCS)’ program in 2012. Special thanks are due to Tetsuhide Yamasaki and Finn Hansen for general fieldwork support and Sakiko Daorana for logistics and accommodation in the village of Qaanaaq. The authors also thank Ayumi Akiyoshi and Mizuho Mori for assistance with laboratory experiments and two anonymous referees, whose comments and suggestions significantly improved this final version of manuscript. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

FUNDING

This work was partially supported by a Grant-in-Aid for Scientific Research (S) [No. ], and by an NIPR publication subsidy.

Conflict of interest. None declared.