Exploring microbial diversity in Greenland Ice Sheet supraglacial habitats through culturing-dependent and -independent approaches

Abstract The microbiome of Greenland Ice Sheet supraglacial habitats is still underinvestigated, and as a result there is a lack of representative genomes from these environments. In this study, we investigated the supraglacial microbiome through a combination of culturing-dependent and -independent approaches. We explored ice, cryoconite, biofilm, and snow biodiversity to answer: (1) how microbial diversity differs between supraglacial habitats, (2) if obtained bacterial genomes reflect dominant community members, and (3) how culturing versus high throughput sequencing changes our observations of microbial diversity in supraglacial habitats. Genomes acquired through metagenomic sequencing (133 high-quality MAGs) and whole genome sequencing (73 bacterial isolates) were compared to the metagenome assemblies to investigate abundance within the total environmental DNA. Isolates obtained in this study were not dominant taxa in the habitat they were sampled from, in contrast to the obtained MAGs. We demonstrate here the advantages of using metagenome SSU rRNA genes to reflect whole-community diversity. Additionally, we demonstrate a proof-of-concept of the application of in situ culturing in a supraglacial setting.


Introduction
The Earth primarily consists of environments that remain cold ( < 5 • C) throughout the year, which contain communities of microorganisms adapted to these conditions (Margesin and Collins 2019 ).In recent years, the microbiome of this cryosphere is becoming better described (Bourquin et al. 2022 ).It is now widely accepted that glaciers and ice sheets are biomes that ar e lar gel y micr obiall y driv en, containing nic hes for micr obes ada pted to cold temperatur es, high UV r adiation, fr eeze-tha w cycles , and long, dark winters .T hese ecosystems are known to harbor a div erse r ange of microbes, including bacteria, fungi, eukaryotic algae, and viruses (Anesio et al. 2017 ), but are still widely understudied (Edw ar ds et al. 2020 ).Researc hing the div ersity of micr obes in the cryosphere is necessary not only to catalog the endemic species of an environment, but also to understand how these organisms may interact with each other and influence their own and adjacent habitats .For example , micr obes fr om the supr a glacial environment influence the biogeochemistry of downstream regions as they are transported off the ice sheet during the melt season (Ste v ens et al. 2022 ).
The Greenland Ice Sheet, being the northern hemisphere's largest body of ice (Stibal et al. 2015 ), is an important subject of study in this r egard.Se v er al studies described the microbial comm unities of supr a glacial envir onments on the Gr eenland Ice Sheet, such as ice and cryoconite holes (Stibal et al. 2015, Hauptmann et al. 2017, Perini et al. 2019, Mogr ov ejo et al. 2020, Poniecka et al. 2020, Lutz and Bradley 2021, Millar et al. 2021 ).On the ice surface, dark-pigmented eukaryotic algae are the main primary producers (Anesio et al. 2017 ).The ice surface is darkening due to blooms of these pigmented algae, leading to increasing glacial melt off (Cook et al. 2020 ).Cyanobacteria are the main primary producers in cryoconite holes (Anesio and Laybourn-Parry 2012 ), whic h ar e cylindrical holes filled with a layer of granular sediment that is a mix of biological and mineral material (Cook et al. 2016 ).Heter otr ophic bacteria ar e also common in these supr a glacial habitats (Lutz et al. 2017 ).Alpha-and Beta-Proteobacteria, Bacter oidetes, and Actinobacteria ar e commonl y r eported in these environments (Stibal et al. 2015, Hauptmann et al. 2017, Perini et al. 2019, Mogr ov ejo et al. 2020, Poniec ka et al. 2020, Lutz and Bradley 2021, Millar et al. 2021 ).
Most microbes are difficult to culture in vitro , and therefore the study of their diversity , physiology , and metabolism pr ov es c hallenging under contr olled conditions.Ho w e v er, cultur eindependent a ppr oac hes can be utilized to circumv ent this.In recent years , High-T hroughput Sequencing technologies (Reuter et al. 2015 ) have made it cheaper and easier to access genomic information from en vironmental microorganisms .Metagenome sequencing (MGS) can be used to obtain metagenome-assembled genomes (MAGs).This enables the study of the functional potential of microbes without the need of having the organisms in culture .T here are , ho w ever, still benefits of having microbes in cultur e, especiall y if investigating phenotypes and metabolic capabilities under controlled conditions is the goal (Poniecka et al. 2020 ).In order to target this uncultured "microbial dark matter," novel culturing methods have been developed in recent years (Lewis et al. 2020 ).One such method is in situ incubation, which involves physicall y separ ating micr obial cells in small cultur e c hambers and incubating them in the same environment that the sample was collected from (Kaeberlein et al. 2002, Jung et al. 2016, Liu et al. 2020 ).This allows the diffusion of growth factors from outside into the c hambers thr ough a semipermeable membrane.An example of this is the "isolation chip" or ichip , which is an array of separ ated cultur e c hambers, made by using a plate of metal or plastic with agar-filled holes (Berdy et al. 2017 ).These are inoculated with diluted sample and closed off on both sides using a membrane .T his allows the chip to be placed back into the original environment for culturing before being transported to the laboratory.Inter estingl y, it is found that eac h r ound of incubation in situ adapts the isolated strains further for growth under laboratory conditions through a domestication pr ocess.A pr ocess in whic h isolates are first grown in their natural en vironment, therefore , results in the isolation of more strains than using conventional tec hniques onl y (Berdy et al. 2017 ).Application of the ichip has successfully led to the discovery of a new antibiotic, teixobactin (Ling et al. 2015 ).The ic hip concept has been a pplied in permafr ost (Marcolefas et al. 2019 ) and ice wedge soil (Goordial et al. 2017 ).To our kno wledge, ho w e v er, it has ne v er been a pplied in a supr a glacial environment.
Se v er al studies hav e pr e viousl y inv estigated the micr obial comm unities of supr a glacial habitats of the Greenland Ice Sheet through either culturing or culturing-independent a ppr oac hes, or both (Musilova et al. 2015, Stibal et al. 2015, Cameron et al. 2016, Hauptmann et al. 2017, Perini et al. 2019, Mogr ov ejo et al. 2020, Poniecka et al. 2020, Millar et al. 2021 ).Most papers, ho w ever, do not focus on comparing cultured diversity to diversity that is observ ed thr ough sequencing, as was for instance done for Sv albard permafrost (Dziurzynski et al. 2022, Sipes et al. 2022 ) and Tibetan glaciers (Liu et al. 2022 ).
In this study, the micr obial div ersity of supr a glacial habitats of the Greenland Ice Sheet was assessed through a combination of culturing and culturing-independent a ppr oac hes.We demonstrate the po w er of MGS to visualize relative abundances of small subunit (SSU) rRNA genes of the entir e micr obial comm unity.Considering the scarcity of genomes reported from the Greenland Ice Sheet, we also investigate which part of the community present in these habitats is discov er able either as a MA G , through MGS, or as an isolate in axenic culture.As part of our culturing a ppr oac h, a novel in situ culturing method (here referred to as culture chips and culture chambers) was employed for the first time on the Greenland Ice Sheet.

Sample collection
Sample collection and in situ incubation were done in July-August 2021 during the Deep Purple ( https:// www.deeppurple-ercsyg.eu/ ) fieldwork campaign.The campsite was on the ice sheet in the south of Greenland, about 7.5 km from the margin, 61.10138895, −46.8481389, 617 m a.s.l.(Fig. 1 A, map generated using Open-Str eetMa p and ArcticDEM; Porter et al. 2023 ).
Environmental samples of the dark ice surface and cryoconite holes were taken across a 100 m × 100 m area (Fig. 1 B-D).Ice samples were taken by scr a ping a ppr oximatel y two v ertical cm of the dark ice surface with a field-sterilized ice axe (cleaned with 70% ethanol, and conditioned with similar ice) and stored in sterile 4 l Whirl-pak bags to melt at ambient temperature of 5-10 • C. Cryoconite sediment was collected with a polycarbonate aquarium pipette from 30 different cryoconite holes with various ways of hydrological connectivity within the aforementioned area.A total of 18 kg of surface ice and 3.5 kg of cryoconite sediment were collected.Eac h envir onmental sample was combined and homogenized in their own sterile sample bags.Subsamples of both the ice and cryoconite samples were subsequently taken in a 50 ml tube and k e pt cool during transport back to the lab.In total, three 4 ml cryotubes were filled with cryoconite sediment for DNA extr action.Biomass for DNA extr action fr om the ice sample was collected by filtering three technical replicates of 500 ml melted ice onto Sartorius cellulose nitrate filters (0.2 μm).Filters were rolled up and stored in 4 ml cry otubes.Tw o additional samples were a viscous suspended biofilm from a cryoconite hole and a red snow sample.Both were collected in 50 ml tubes and k e pt cool during tr ansport bac k to the lab.A subsample of the biofilm was taken for DNA extraction.All samples for DNA extraction (cryoconite, ice, and biofilm) were frozen in the field camp and k e pt at −20 • C thr ough tr ansport to Aarhus Univ ersity, Roskilde, Denmark.

Design
The culture chips used for in situ culturing were inspired by an existing protocol (Berdy et al. 2017 ) with minor alterations .T he 1-cm thick plate was made from polycarbonate and had an array of 96 holes that could be filled with solid medium.The 3-mm thin outer layers serve to protect the semipermeable membrane .T he plastic lay ers w ere sterilized b y autoclaving prior to use in the field.This stack was bolted down to stay together (Fig. 2 B, D, and F).The design of the culture chambers was similar but made with three autoclaved plastic washers that formed one bigger growth chamber when stacked together with membranes in between (Fig. 2 A,  C, and E).

In situ deployment
Cultur e c hips and c hambers wer e first half-assembled by gluing pol ycarbonate hydr ophilic membr anes (Cytiv a Nuclepor e, USA; 0.03 μm) to one side of the middle layer using aquarium silicone glue, applied to the plastic in a thin layer.
To pr epar e the inoculum, the homogenized cryoconite sediment and ice surface samples were serially diluted with in situ autoclaved (heated in pressure cooker to > 117 • C for 20 min) glacial stream w ater.Cry oconite sediment sample dilutions used w ere 10 −3 -10 −6 , and 10 −1 -10 −4 for ice; based on pr e viousl y r eported cell counts for both environments (Nicholes et al. 2019 ).A total of two cultur e c hips wer e made for eac h of the four dilutions.A volume of 400 μl of the dilutions were added to 50 ml tubes , in duplicate , for both the ice and cryoconite inocula.To one series of four dilutions for both cryoconite and ice samples, 500 μl nystatin ready-made solution (Sigma Aldrich, USA) was ad ded.In ad dition, 40 ml Reasoner's 2 agar (R2A) (Linde et al. 2000 ) (Alpha Biosciences), that was supplemented with 18% gl ycer ol to serve as a freezing point depressant in the solid medium, because of concerns for the agar gel deteriorating after freeze-thaw cycles on the ice, was added to each tube.After mixing, this medium was used to fill all wells of the culture chips, except for the last column of eight wells on each c hip, whic h was filled with R2A + 18% gl ycer ol without inoculum, as a negative control.A few mm headspace was left on top of each well.After the agar was solidified, another membrane was applied to the still open side, sealing the chip.The outer layers were then bolted on to finish the assembly.
For the culture chambers, 10 μl of the 10 −4 dilution of the cryoconite sample, or the undiluted ice sample, were added in triplicate to 2 ml tubes.A volume of 12.5 μl nystatin ready-made solution was added to two of these tubes.To one of the tubes with n ystatin, 2 μl str e ptom ycin (50 mg ml −1 ) was also added.This way, both the ice and cryoconite inoculum was mixed with nystatin, stre ptom ycin + nystatin, or no antibiotics.One ml R2A + 18% gl ycer ol a gar was added to eac h tube, this was used to fill the already half-assembled culture chambers.A negative control chamber with nonamended sterile agar was also made.
Once assembled, the culture chips and chambers were left to incubate in situ (Fig. 2 ) , inside a large cryoconite hole in which the chips and chambers containing the cryoconite inoculum were submerged (Fig. 2 C and D).The chips and chambers made with ice inoculum were deposited on a patch of dark ice for incubation (Fig. 2 A and B).The cryoconite and ice culture chips were left to incubate for 18 and 16 da ys , r espectiv el y.The cryoconite and ice cultur e c hambers wer e left for 17 and 15 da ys , r espectiv el y.After this time, the in situ culturing devices were removed from the environment and placed in sterile Whirl-pak bags , co vered in sterile glacial stream water and stored at about 5 • C during transportation back to the home laboratory.

Plates
A volume of 100 μl of cryoconite sediment and dark ice were plated onto R2A + 18% gl ycer ol a gar, supplemented with either 100 μg ml −1 stre ptom ycin, 12.5 μl nystatin, both, or no antibiotics.A volume of 100 μl of the snow and the biofilm samples were also plated on R2A + 18% gl ycer ol a gar plates without antibiotics.The agar plates were incubated at ambient temperature in the lab tent ( ∼ 5 • C) in parallel with the in situ culturing and at the end of the inoculation they wer e tr ansported bac k to Aarhus Univ ersity, Roskilde, Denmark at 5 • C. Back in the lab, the four environmental samples (100 μl) were plated onto R2A agar in triplicate, and incubated at 5 • C, 10 • C, and room temperature (20 • C).

Culturing of isolates
Data from the nearby PROMICE (Fausto et al. 2021 ) r e v ealed that during the in situ incubation (17 July 2021-4 August 2021) the air temper atur es r anged between 1.15 and 8.51 • C, with an av er a ge of 3.70 + / − 1.38 • C. Full weather station data can be found in the Supplementary Information S1 .Isolates were primarily cultured at 5 • C to stay close to this range.
After returning to the lab, the a gar fr om the cultur e c hambers was extruded and spread on R2A plates.Culture chips were first dried, opened, and examined under a dissection micr oscope.Eac h w ell w as inspected, and micr obial gr o wth w as streaked onto R2A plates with a sterile toothpick.Colonies were also observed on the outside of the membrane, and some were streaked onto R2A.
All plates were subsequently incubated at 5 • C until colonies a ppear ed (between 2 weeks and 2 months).Colonies were restr eaked onto ne w plates, selected based on differ ent mor phologies, until visible axenic cultures were obtained.Isolates picked from plates that initially were incubated at 10 • C or room temperatur e wer e e v entuall y also incubated at 5 • C for following r ounds of culturing.

Whole genome sequencing of cultures
Gl ycer ol stoc ks fr om −80 • C wer e used to inoculate overnight cultures in 2 ml R2B.Overnight cultures of each isolate were incubated in duplicate at both room temperature and 5 • C while shaking (200 r m −1 ), the fastest growing culture used for DNA extr action.DNA extr action was carried out using a Gentra Puregene kit (Qiagen, Hilden, Germany), according to the manufactur er's pr otocol, except for the DNA hydr ation solution.A total of 10 mM Tris, 50 mM NaCl pH 8.0 was used instead.Nextera XT kit (Illumina, San Diego, USA) was used for library pr epar ation.DN A w as resuspended in PCR water instead of resuspension buffer.The pooled genomes were sequenced on a NextSeq 500 using the MID output flow cell and the v2.5, 300 cycle c hemistry (Illumina), r esulting in 39 gigabases total output passing the Q30 threshold.The raw reads were processed through our automated whole genome sequencing (WGS) pipeline (Campuzano 2023 ).Briefly, the r aw r eads wer e trimmed, assembled, annotated, and compared to each other.Statistics regarding the quality of the reads, the completion of the assemblies and so on, were also calculated.Full specifications of the programs used, including their versions and options are listed in the Github repository ( https:// github.com/AU-ENVS-Bioinformaticshttps:// zenodo.org/badge/ latestdoi/ 546561474 ).

Amplicon sequencing of environmental samples
DN A w as extracted from three technical replicates of cryoconite sediment, filters with biomass from ice, and the biofilm using the DNeasy Po w erLyzer Po w er Soil kit (Qiagen).The follo wing univ ersal pr okary otic 16S rRN A gene and eukary otic 18S rRN A gene primers were used for amplicon library building.16S forw ar d 341F (TCGTCGGC A GCGTCA G A TGTGT A T AA GA GA CA GCCT A YGG GRBGCASC AG) and r e v erse 806R (GTCTCGTG GGCTCGGA GAT-GTGT A T AA GA GA C A GGGA CT A CNNGGGT A TCT A A T) (Hansen et al. 2012 ), and 18S forw ar d 528F (TCGTCGGC A GCGTCA G AT-GTGT A T AA GA GA C A GGCGGTAA TTCC AGCT CC AA) and r e v erse 706R (GTCTCGTG GGCTCGGA GA TGTGT A T AA GA GA C A GAATCCR A GAATTTC A CCTCT) (Cheung et al. 2010 ) all at 10 μM concentration.The amplicon library building was performed by a twostep PCR, as described by Feld et al. ( 2016 ) and Albers et al. ( 2018 ) with slight modifications.PCR r eactions wer e conducted on a Sim-pliAmp Thermal Cycler (Applied Bio-systems, Waltham, USA).In eac h r eaction of the first PCR, the mix contained 12.5 μl of 2x PCRBIO Ultra Mix (PCR Biosystems), 0.5 μl of forw ar d and r e v erse primer, 0.5 μl of bovine serum albumin (BSA) to a final concentration of 0.025 mg ml −1 , 6 μl of PCR-grade water, and 5 μl of template .T he r eaction mixtur e was pr eincubated at 95 • C for 2 min, follo w ed b y 33 c ycles of 95 • C for 15 s, 55 • C for 15 s, 72 • C for 40 s, with a final extension performed at 72 • C for 4 min.Samples were subsequently indexed by a second PCR.For this, amplification was performed in 28 μl reactions with 12.5 μl of 2x PCRBIO Ultra Mix (PCR Biosystems), 2 μl of indexing primers (P7/P5), 6.5 μl of PCRgrade water, and 5 μl of PCR1 product.The cycling conditions included initial denaturation at 98 • C for 1 min, follo w ed b y 13 c ycles of denaturation at 98 • C for 10 s, annealing at 55 • C for 20 s, and extension at 72 • C for 40 s, with a final extension performed at 72 • C for 5 min.The final PCR products were purified with 15 μl magnetic beads (MagBio Genomics, Gaithersburg, USA) according to the manufacturer's instructions and eluted in 27 μl buffer.Electr ophor esis 1% a gar ose gels and TapeStation D1000 DNA Screen-Tape (Agilent, Santa Clara, USA) were run to check the quality of the libr aries.Finall y, the concentr ations of the libr aries wer e measured on a Qubit 4.0 fluorometer (Invitrogen, USA), and these were then equimolarly pooled.The final pooled libraries from 16S and 18S were sequenced on an Illumina MiSeq using the V2 chemistry (Illumina) resulting in 2 × 250 bp reads.

Metagenome sequencing
The extr acted DNA fr om envir onmental samples was also used for shotgun MGS.Three replicates of dark ice and cryoconite sediments where used, but only one replicate of the biofilm was used for MGS.Ice and cryoconite samples were prioritized since they wer e mor e widel y used in the culturing a ppr oac h.Hence, in total se v en libr aries wer e pr epar ed using the Ultr a FS II DNA Libr ary Pr ep Kit for Illumina (Ne w England Biolabs , Ips wich, USA) following the manufacturer's protocol.The libraries were pooled equimolarly, and the pool was run in a TapeStation 4150 to c hec k for insert size distribution and the presence of adapter-primer dimers, using a D1000 DNA Scr eenTa pe.Dilution and denaturation of the library was performed according to Illumina's recommendations before sequencing on a NextSeq 500 using the high output flow cell, and the v2.5, 300 cycle chemistry.140 gigabases passing the Q30 threshold were obtained.

=
mapped contigs / ( gene length / 1000 ) / 1 , 000 , 000 .(1) All MAG taxonomy was identified using GTDB-Tk v 2.1.1 (Parks et al. 2018 ) and manuall y c hanged to match NCBI taxonomy at phylum le v el to most accur atel y compar e the MAGs to the cultures and amplicons.A full list of taxonomies can be found in Supplementary S2 .
Whole genomes of bacterial isolates were also mapped to the se v en meta genome assemblies to inv estigate the shar ed genetic material between the two methods.Bowtie2 (Langmead and Salzberg 2012 ) was used to index the whole genomes to the assemblies and reads per million was calculated (Equation 1).
Briefly, the r aw r eads wer e trimmed, then sorted into a "SSU bin," assembled into full length rRNA genes and annotated.Full specifications of the pr ogr ams used, including their versions and options are listed in the Github repository ( https: // github.com/AU-ENVS-Bioinformatics/TotalRNA-Snak emak e ).
Analysis of amplicon sequencing data was done using QIIME2-2021.8(Bolyen et al. 2019 ), with the Silva 138_99 database (Quast et al. 2013 ).Samples were not rarefied, as rarefaction curves ( Figures S6 and S7, Supporting Information ) sho w ed that sufficient sequencing depth was ac hie v ed for all samples.Taxonomy data from GTDB was man ually check ed and corrected to match NCBI taxonomy at phylum le v el.A full list of taxonomies is available in Supplementary Information S2 .Amplicon sequence vari-ants (ASVs) or 16S rRNA genes belonging to c hlor oplasts or mitoc hondria wer e discarded fr om amplicon and MGS r esults.Unclassified ASVs were pooled under "other." Ampvis2 (Andersen et al. 2018 ) was used to create heat maps and alpha diversity plots.Shannon Equitability Index was calculated according to Equation ( 2).
where H = Shannon diversity index and S = number of ASVs.Metacoder (Foster et al. 2017 ) was used for the creation of heat trees, showing diversity at multiple taxonomic levels at the same time.Both pac ka ges wer e used in R Studio 2021.09.01 (R Core Team 2021 ).Scripts are accessible in the Github repository ( https:// github.com/AU-ENVS-Bioinformatics/ GR21 _ Greenland _ Ice _ Sheet _ Microbial _ Diversity _ Data _ Handling ) (Jaarsma 2023 ).
All metagenome and bacterial isolate 16S rRNA genes were aligned using MAFFT v7.490 (Katoh et al. 2002 ) in Geneious Prime on RAxML (Stamatakis 2014 ).The GTR GAMMA model was used, under the "Rapid Bootstrapping and search for best-scoring ML tr ee" algorithm, using 100 bootstr a p r eplicates.Bootstr a p v alues lo w er than 80 were removed.

Sequencing output for amplicons and metagenomes
After quality c hec king, most mer ged r eads wer e pr oduced fr om the biofilm sample in both 16S and 18S rRNA gene sequencing, follo w ed b y cry oconite and ice (Table 1 ).Most ASVs w ere, ho we v er, obtained fr om cry oconite, follo w ed b y biofilm and ice (Fig. 3 ).Amplicon sequencing yielded 30 bacterial and 18 eukaryote phyla.Metagenome SSU rRNA gene observ ations mostl y came fr om ice, follo w ed b y cry oconite and biofilm.There w as a lar ge ov erla p in micr obial div ersity between the thr ee envir onments using these genes (Fig. 3 ).The metagenome SSU rRNA genes belonged to 24 bacterial and 12 eukaryote phyla.There were 133 high quality MAGs assembled from the metagenome data.The total number of MAGs varied by sample type .T he cryoconite metagenome gave the highest number of MAGs ( n = 89), while the ice surface and biofilm metagenome assemblies resulted in 32 and 12 MAGs, respectiv el y (Table 1 ).

Alpha di v ersity in supraglacial ha bitats
Shannon diversity, Shannon equitability, and inverse Simpson alpha diversity indices, for both amplicon sequencing and the rRNA genes from MGS, were calculated and grouped by sample type ( Figure S3, Supporting Information ).Based on the Shannon div ersity and inv erse Simpson indices of rRNA genes from the metagenomes, cryoconite had the highest diversity, follo w ed b y biofilm and ice.A similar trend was found for 16S rRNA gene amplicons, but the biofilm and ice were more similar here.18S rRNA gene amplicon alpha diversity was typically lo w er compared to those of the 16S rRNA gene .T he biofilm sample had the highest eukary ote diversity, follo w ed b y cry oconite and ice.Based on the Shannon equitability index, e v enness was highest in cryoconite, follo w ed b y biofilm and ice.

Metagenome-extracted SSU rRNA genes
The abundance of assembled SSU rRNA genes from the metagenomes is plotted as heat trees for each of the three main habitats studied here: ice , cryoconite , and biofilm (Fig. 6 ).The ice surface eukaryotic microbiome was dominated by Zygnematophyceae, a class of Streptophyte algae, and Chytridiomycota fungi, as determined from 18S rRNA gene sequences .T he bacterial community of the ice surface mainly consisted of Proteobacteria, Bacter oidetes, and Terr abacteria, the latter a taxon that encompasses phyla like Armatimonadetes , Firmicutes , and Actinobacteria.T he same bacterial taxa dominated the cryoconite and biofilm 16S rRNA genes .T he biofilm bacterial community was clearly dominated by Proteobacteria 16S rRNA genes .T he biofilm had a relativ el y higher total of eukaryote rRNA genes compared to bacterial rRNA genes than the cryoconite sample had.Glacier ice algae and Chytridiomycota 18S rRNA genes wer e a gain observ ed among the common eukaryotes in the biofilm sample.
For eukaryotic SSU rRNA, the Streptophyta were the most dominant on the ice surface, amounting to 44.6% of total microbial community SSU rRNA genes (Fig. 5 ).In contr ast, these Str eptophyta SSU rRNA genes were less dominating in the biofilm (0.3%) and cryoconite (0.6%) samples.In addition, in contrast to the 18S rRNA amplicon sequencing r esults, Chlor ophyta snow algae SSU    rRN A genes w ere also abundant in the dark ice sample (14.8%), y et they were less common in the biofilm (3.9%) and cryoconite (2.2%).
Similarly, ice contained 10% Basidiomycota SSU rRNA genes, while the biofilm and cryoconite both contained < 2%.Finally, Chytridiomycota are found in the ice surface sample with 9.6% relative abundance of SSU rRNA genes.Lower abundances were observed for the biofilm (4.3%) and cryoconite (2.2%) samples.A total of four different orders of Chytridiomycota were detected in the metagenome.Ice and biofilm both had a similar profile of chytrids, dominated by Chytridiales (67.0% and 71.4% of the total c hytrids sequences, r espectiv el y), whic h made up a smaller portion of the cryoconite chytrid community (37.6%).Chytrids found in cryoconite mainly belonged to Rhizophydiales (38.4%), which were less abundant in ice (26%) and biofilm (19.1%).Lobulomycetales made up 17.9% of the cryoconite chytrids community, but only 3.8% of that of ice, and 3.3% of that of biofilm.

MAGs
The 133 obtained MAGs were grouped by sample type, and their pr oportions ar e shown together with the meta genome and amplicon sequencing results (Fig. 4 ).The MAGs belonged to nine unique phyla (Table 2 ).Proteobacteria made up the majority of MAGs (63), follo w ed b y Bacteroidetes (26).The majority of MAGs were assembled from the cryoconite metagenome .T he 12 biofilm MAGs were made up of only 5.3% of the assembled contigs from the biofilm assembly.The 89 total cryoconite MAGs made up 6.4% (33 MAGS), 4.4% (28 MAGs), and 4.7% (28 MAGs) of the three cryoconite assemblies .T he 32 MAGs from the ice surface made up 3.5% (11 MAGs), 2.1% (8 MAGs), and 3.6% (12 MAGs) of the three assemblies from the ice sample.
Read mapping (see Equation 1) was used to determine if genomic material was shared between MAGs and sample types that they did not originate from.Since the MAGs were single sample assembled, the r eads fr om the individual samples did not mix and the MAGs that originate from one sample type can be confidently assumed to be present in that sample.To further c hec k for patterns in MAG occurrence across the three sample types, the r elativ e abundances of MAGs wer e cluster ed by Spearman r ank corr elation and depicted with a dendr ogr am.The se v en samples were ordered similarly (Fig. 7 ).Three cocorrelating clusters arose in the MAGs analysis, showing MAGs that were most abundant in the biofilm, cryoconite, or the ice surface assemblies (Fig. 7 A).The Spearman correlation clustering sho w ed three gr oups, widel y based on origin sample type.Cluster 1 is biofilm originating MAGs and cluster together because of their high abundance in the biofilm sample.Cluster two contains many MAGs that have abundance in all sample types but also clusters mainly in the cryoconite sample types.Lastly, cluster three is a mix of MAGs that were abundant everywhere and predominant in the ice surface samples.

Cultured isolates
Ther e wer e 160 isolates identified thr ough Sanger sequencing.Of the 88 bacterial isolates, 45 were cultured using in situ methods (cultur e c hips and c hambers), wher eas almost all the yeast isolates were obtained using in situ methods (Table 3 ).Only two out of 72 yeast isolates originated from plates .T he majority of isolates was obtained from cryoconite, followed by ice, snow, and biofilm.Whole genomes of 73 bacterial isolates were sequenced, after which full length 16S rRNA genes were extracted.BLASTn was used to identify the closest r elativ e in the NCBI database.A ta-ble of all isolate information, including closest r elativ es (highest % identity determined through Sanger sequencing and wholegenome sequencing), can be found in Supplementary S4 .The culturing method and sample origin of each isolate are also listed in this table.Only 33 of the isolated whole genomes sho w ed nonzer o r ead abundance in the meta genome libr aries, albeit low in general (Fig. 7 ).The only isolates to have reads mapped to the metagenome sample types belonged to the phylum Proteobacteria.Only two of the isolates that originated from the culture chamber method (from ice and cryoconite; Fig. 7 B) had a high read  mapping to the ice surface samples, which was nearly six times higher than the r eads ma pped fr om the other cultur es.Additionally, the isolates did not show any clustering pattern based on sample origin in the same way the MAGs did, but instead sho w ed a gener al pr esence or absence within the se v en meta genome libraries.
The full length 16S rRNA genes were also used to plot abundance of bacterial isolates in a heatmap (Fig. 4 ), whereas LSU Sanger sequencing results were used to plot the abundance of yeast isolates (Fig. 5 ).The 73 bacterial isolates from which a full genome was obtained belonged to 10 genera within Proteobacteria, Bacteroidetes, and Actinobacteria.Proteobacteria made up the majority ( n = 60) of isolates, with 43 isolates belonging to Pseudomonas .Herbaspirillum and Cryobacterium w ere tw o additional abundant genera, with nine and seven isolates, respectively.In total, 28 of the 73 bacterial isolates were cultured from the cryoconite sample, 25 from ice, 11 from snow, and nine from the biofilm.Yeasts were only obtained from the ice and cryoconite samples.A total of five yeast genera were observed, belonging to two phyla, Basidiomycota and Ascomycota.Mrakia ( n = 46) and Camptobasidium ( n = 23) made up the bulk of the isolates (Fig. 5 ).A phylogenetic tree of isolate and metagenome 16S rRNA genes is available in supplementary information S5 .

Discussion
In this study, we mapped the micr obial div ersity of supr a glacial habitats on the Greenland Ice Sheet.Using different sequencing methods, we identified different dominant community members.Subsequently, the combination of (metagenome) sequencing and the culturing a ppr oac h enabled comparison of obtained genomes (MAGs or isolate genomes) to these dominant community members.Using r ead ma pping, we visualized the genetic ov erla p between the obtained genomes and the metagenomes.

The Greenland Ice Sheet contains a variety of life
The Greenland Ice Sheet contains abundant microbial life in all its habitats (Anesio et al. 2017 ).With the culture-dependent and culture-independent work presented here, we expand this catalog with organisms from four environmental sample types: cryoconite , biofilm, surface ice , and snow.We documented and confirmed se v er al k e ystone species within these respecti ve habitats .For example , an important ecosystem engineer in cryoconite holes, the c y anobacterium Phormidesmis priestle yi (Gokul et al. 2019 ), constituted ∼5% of the MAGs acquired from cryoconite.The ice surface sample was dominated by eukaryote SSU rRNA genes, mainl y fr om the Str eptoph yte algae Zygnematoph yceae, which composed 45% and 69% of the metagenome and amplicon sequencing rRNA genes, r espectiv el y.This is not surprising, considering that these dark pigmented glacier ice algae can form large blooms on the ice surface over the summer (Williamson et al. 2020 ).While our data does not reflect abundance of live cells dir ectl y, the pr e viousl y described r elativ el y high counts of algal cells on the Greenland Ice Sheet (Holland et al. 2019 ) corroborates our high rRNA gene abundance data.Furthermore, it has been found that, on av er a ge, 13.5% of cells in supr a glacial meltwaters ar e lar ger than 10 μm, and ther efor e likel y belong to either c y anobacteria or eukary otic glacier ice algae (Ste v ens et al.

).
Although the role of glacier ice algae in the darkening and melt of the ice has been demonstrated (Chevrollier et al. 2022 ), their interactions with other members of the ice microbial community hav e r ar el y been described.Chytridiomycota have been suggested to be associated with algal blooms as par asites, perha ps pr oviding a top-down control on the glacial ice algae (Perini et al. 2022 ).The same interaction was documented in Alaskan cryoconite holes, and it was suggested that cryoconite holes could be hotspots for chytrid infections (K oba yashi et al. 2023 ).In the present study, we found chytrids to be abundant in the eukaryote rRNA gene data in both biofilm and cryoconite, but sur prisingl y to a lesser extent in the ice surface sample .T he chytrid communities also differed in the sample types, raising the question if chytrids in cryoconite holes parasitize different targets than those on the ice surface or in biofilms.
Furthermor e, pr edatory pr otists like Cercozoa and Ciliophora wer e observ ed in all sample types .T he Cer cozoa family Vamp yrellidae was 1.8 times more abundant in biofilm than cryoconite, and 2.7 times more abundant in biofilm than ice .T his algivore is known to perforate and feed on Zygnematophyceae and Chlorophyceae cells (Hess et al. 2012 ).Vampyrellidae have been observed in polar cryoconite before (Millar et al. 2021 ).It would be worthwhile to investigate further whether this protist is one of the biological controls of the glacier ice algae together with chytrids .T he biofilm was also rich in rRNA genes from Chrysophyceae, golden algae, whic h wer e m uc h less abundant in cryoconite (17.9 times less) and ice (3.9 times less).The biofilm observed in this study might thus be formed by detritus from ice surface eukaryotes that get washed into the cryoconite holes .T his is also supported by the br own-r ed color of the biofilm, which likely originates from algal pigment.Glacier ice algae SSU rRNA genes w ere scar ce in the biofilm.Further study is needed to confirm if they are consumed by chytrids or protists in the biofilm, which could explain the lack of observed genes due to their fast utilization as a food source.
The dominant bacterial phyla observed in this study match similar amplicon sequencing studies.Proteobacteria, Bacter oidetes, Actinobacteria, and Cyanobacteria ar e fr equentl y reported as abundant in supr a glacial habitats of the Greenland Ice Sheet (Musilova et al. 2015, Stibal et al. 2015, Perini et al. 2019, Millar et al. 2021 ).Cryoconite samples harbor the highest microbial diversity among the other supraglacial habitats sampled in this study.This is in correspondence with pr e vious studies that compare different supraglacial habitats, for instance on Tibetan glaciers (Liu et al. 2022 ).No significant differences in alpha diversity of ice and cry oconite w ere ho w ever observed b y others (Cameron et al. 2016 ).

Meta genome appr oach complements amplicon sequencing to in vestiga te microbial biodiversity
In the current study, we investigated the microbial diversity of supr a glacial habitats on the Greenland ice sheet using various culturing-dependent and -independent a ppr oac hes .T his allows for comparison of the effectiveness of these different methods for capturing this diversity.Using rRNA genes from MGS offers a few adv anta ges ov er using amplicon data because of pr ocedur al differences .T he choice of primer pair can influence the reflection of the diversity within short read amplicons (Lewis et al. 2020 ).Using the meta genome a ppr oac h, the intr oduction of primer bias in the data is, ho w e v er, avoided because the amplification step targets the adapter ligated DNA fragments, thus avoiding preferential binding.Another benefit is the fact that full length SSU rRNA genes are obtained through MGS and assembly.A higher resolution for phylogeny assignment is thus ac hie v ed compar ed to amplicon sequencing, whic h r elies on certain r egions of the rRNA genes, unless long read amplicons are used (Johnson et al. 2019 ).
A lo w er diversity of SSU rRNA gene taxa was observed in the metagenome than through amplicon sequencing, but they wer e mor e e v enl y distributed among the sample types in the metagenome.Despite a higher sequencing depth, the SSU rRNA genes make up only a small portion of the total metagenome (0.27%) and are thus relatively less covered.Low abundant organisms may, ther efor e, be ov erlooked.It is difficult to assess to which extent this has happened, and if sequencing depth should have been e v en gr eater.Total RNA sequencing can be consider ed as an alternative that requires lo w er sequencing depth, while still allowing r epr esentation of the whole community (Cottier et al. 2018 ).It has been pr e viousl y discussed that intr a genomic copies of 16S rRNA genes can lead to artificially high diversity when an ASV a ppr oac h is taken, falsely splitting them up into different taxa.Conv ersel y, closel y r elated but slightl y differ ent SSU rRNA genes ma y ha ve been assembled together in the meta genome a ppr oac h, leading to an underestimation of diversity (Johnson et al. 2019 ).
A classic amplicon a ppr oac h yields two different datasets, as eukary ote and prokary ote rRN A genes ar e amplified in separ ate PCR reactions.Rather than using two separate bar plots to visualize the r elativ e abundances of the 16S and 18S rRNA genes, which is a common way to present amplicon data, we demonstrate here the use of heat trees (Fig. 6 ) to visualize differences in prokaryote/eukaryote gene proportions in different en vironments .T his illustrates a po w erful benefit of MGS over amplicon sequencing, i.e .the direct comparison of rRNA gene r elativ e abundances of the entir e comm unity.This pr ovides v aluable information for the study of ecological relations between members of this community, although it should be k e pt in mind that the abundance of rRNA genes is a mere proxy for the abundance of active cells in the en vironment.T he ice surface sample contained more eukaryote than prokary ote SSU rRN A genes.Although cop y numbers of rRNA genes are generally higher for eukaryotes (Hori et al. 2023 ), this could indicate a higher number of eukaryotes on the ice surface.Despite gener all y higher cop y numbers in eukary otes, the biofilm and cryoconite samples contained mor e pr okaryote SSU rRNA genes.Based on this a ppr oac h, we demonstr ate the r elativ el y higher importance of the eukaryotic community compared to the pr okaryotic comm unity on the ice surface, while the opposite is true in cryoconite holes and the biofilm.

Distribution of MAGs
Assembling MAGs can r e v eal micr obial dark matter in the environment.It is, ho w ever, clear that most of the metagenomic assembl y is unused; onl y a small portion of contigs from the metagenome contributed to the assembly of high-quality MAGs.This can be partially explained by the fact that only high quality bacterial MAGs were assembled in this study, ignoring the eukary otic DN A in the assembl y.The MAGs that originated fr om a single sample type typically had some presence in the other two sample types.Spearman correlation of MAG read mapping abundances sho w ed that some of the MAGs were unique to their origin source, while others have some presence in the other sample types .T his emphasized that ther e ar e some or ganisms that are isolated to a unique environment through environmental filtering, whereas others may be more generic ice sheet microbes.In the heatma p (Fig. 7 ), ther e is a group of MAGs that maps back to both cryoconite and ice, regardless of which metagenome they were assembled from.This group includes the genera Granulicella , JAFAZD01 , Lacisediminihabitans , the families Sphingomonadaceae , Capsulimonadaceae , and the or der Sphingobacteriales .Examples of more specialist taxa include P. priestleyi, which was only acquired as a MAG from cryoconite.A collection of MAGs appears to be pr edominantl y associated with biofilm.This group includes the (genomo)genera Rhodoferax , Sediminibacterium , Methylotenera , Palsa-911 , Paucibacter , Arcicella , CAIQQQQ01 , and JAAFHG01.
Se v er al MAGs that could be classified to genomospecies le v el hav e been observ ed befor e .T hr ee F erruginibacter sp014377975 MAGs, assembled from the cryoconite metagenome but abundant in biofilm, have previously been obtained as MAG from a Greenlandic glacier (Genbank accession number GCA_014377975.1).Genome assemblies of Gemmatimonadaceae sp.AG11 sp014378185 (Genbank accession number GCA_014378185.1), and Undibacterium sp014376575 (Genbank accession number GCA_014376575.1) wer e r eported in the same study (Bioproject accession number PRJNA552582).Sphingomonas psychrolutea , abundant in biofilm, has previously been found in Tibetan glacier ice samples (Liu et al. 2015 ).
On Tibetan glaciers, the majority of genomes (either from isolates or MAGs) were obtained from cry oconite, follo w ed b y ice and snow (Liu et al. 2022 ).Mostl y Pr oteobacteria wer e observ ed, follo w ed b y Actinobacteria, Bacteroidetes , and Firmicutes .In contrast to our findings, it was observed that the majority of what was cultivated could also be obtained as MAGs (Liu et al. 2022 ).Isolates in that study were incubated at 4 • C in R2A broth.Another metagenome study of Greenland Ice Sheet cryoconite obtained 29 MAGs, 13 of whic h wer e 100% complete, and included a Cyanobacterium, Proteobacteria, Actinobacteria, Bacteroidetes, Acidobacteria, and Chloroflexi (Hauptmann et al. 2017 ).

Recovery of isolate genomes
Culturing a ppr oac hes ar e known to onl y ca ptur e a small r epr esentation of the microbes present in the natural environment, which leaves most of the organisms unknown (Lloyd et al. 2018 ).A recent study found that, of the microbial diversity of Svalbard permafrost observ ed thr ough amplicon sequencing, onl y 6.37% of bacterial and 20% of fungal genera observed through amplicon sequencing were cultivable (Dziurzynski et al. 2022 ).It is estimated that less than 1% of the total envir onmental micr obial div ersity can be cultivated under laboratory conditions (Kaeberlein et al. 2002 ).When isolates are the only source of information, most diversity is thus missing, as also demonstrated in this study.There is less diversity in isolates than there is in MAGs .T his is expected, as MGS circumv ents cultiv ation bias .T he observed MAGs might still be biased to w ar d more abundant taxa, despite high sequencing depth.The biofilm MAGs Ferruginibacter , Methylotenera , and Paucibacter for instance also appear abundant in the 16S rRNA extracted genes (Fig. 6 ).Only four genera overlap between MAGs and isolates: Sphingomonas , Rhodoferax , Rhodanobacter , and Lacisediminihabitans .
Man y cultur ed isolate taxa in this study matc h those pr e viousl y r eported.Pr oteobacteria, Actinobacteria, and Bacter oidetes, which encompass the isolates in this study are commonly found, albeit in varying ratios.In a review of 340 bacterial strains isolated from glacial habitats around the globe, Proteobacteria are the most dominant, follo w ed b y Actinobacteria, Firmicutes, Bacteroidetes , and Deinococcus-T hermus (Kim et al. 2022 ).The yeast genera Mr akia , Dothior a , Phenoliferia , and Camptobasidium have been pr e viousl y isolated from Greenland (Perini et al. 2019(Perini et al. , 2021 ) ).
Read mapping the abundance of the isolate genomes to the se v en meta genome assemblies shows that the cultur able or ganisms r epr esented a small pr oportion of the total envir onmental DNA (Fig. 7 B).Only 33 of the 73 isolate whole genomes had a nonzero amount of contigs map to the assembled metagenome reads .T he whole genomes that had presence in the metagenome assemblies were all from the phylum Proteobacteria, which was also the most abundant phylum in the MAGs .T his dominance of Proteobacteria could be because this phylum is the dominating microbe in many cryospheric environments (Bourquin et al. 2022 ), or because current methods in DNA identification are biased to w ar d cultur ed r epr esentativ es (Lloyd et al. 2018 ).When ther e ar e no ma pped r eads shar ed between the meta genome assembly and the (meta)genome's contigs, we can assume that there is little to no presence of that genome in the sample assembly.T he isolates , while belonging mostl y to the gener all y abundant Proteobacteria, seem rare in the supraglacial environment.In the phylogenetic tree of isolate and metagenome 16S rRNA genes ( Figure S5, Supporting Information ), the large group of Pseudomonas isolates clustered together with only one Pseudomonas 16S rRNA gene from the metagenome .T he amplicon data only yielded two Pseudomonas ASVs, both only found in ice.Similarly, the Mucilaginibacter isolates did not cluster together with metagenome 16S rRNA genes .Mucilaginibacter was , howe v er, observ ed in the amplicon sequencing results, with four different ASVs.While relatives of the other isolates were observed in the phylogenetic tree, most isolates are apparently not representing the most abundant genera in the environment.
Judging from the read mapping, the sample origin of the isolates does not always match the actual abundance in the meta genome fr om that habitat, wher eas this is mor e pr edictable for the MAGs .T here is thus a degree of randomness in observed isolates from the different samples.As a result, the obtained cultured isolates do not necessarily reflect the most abundant organisms in the habitat they were sampled from.Conflicting studies exist that cover the presence of distinct communities in habitats such as ice and cryoconite.In one study, spatial distribution across the ice surface seemed to be the main component of variability in bacterial communities, while differences between ice and cryoconite communities seemed less distinct (Cameron et al. 2016 ).Others ha ve , ho w ever, demonstrated that there can be significant differ ences in comm unity betw een the ice surface and cry oconite hole habitats (Musilova et al. 2015 ).Simultaneously, it is shown that many taxa are specific to a certain environment (Perini et al. 2019 ), which is also observed in the read mapping of the MAGs against the metagenomes in this study.
The habitats sampled in this study are in close spatial proximity to each other and connected through flow of liquid water in the melt season (Cameron et al. 2020 ).It has r ecentl y been suggested that microbes-and especially bacteria-can be tr ansported thr ough the weathering crust, for instance from and into cryoconite holes (Cook et al. 2016 ), in particular considering that surface meltwaters consistently contain about 10 4 cells ml −1 (Ste v ens et al. 2022 ).The controls on transport of microbial cells through the weathering crusts remains, ho w ever, one of the outstanding questions (Halbach et al. 2023 ).The standing theory for inoculation of the supr a glacial envir onments by bacteria is that the y are deli vered through aerial deposition (Šantl-Temkiv et al. 2018, Cameron et al. 2020 ).Once on the ice sheet, a significant portion of glacier surface microbes are subsequently found to be metabolicall y inactiv e (Br adley et al. 2022 ).This means certain taxa of microbes observed in this study through culturing might spring fr om tr ansient cells in the sample they wer e ca ptur ed fr om, rather than settled and thriving colonies.
Detection of these cells through sequencing would pr ov e difficult, since there is a minimum amount of biomass needed to detect their DN A. Ho w e v er, during culturing, e v en low abundant bacterial cells are given a chance to grow into a colony, after they ar e pr ovided a mor e stable medium to r eside in.This c hange in conditions may increase the fitness of some isolates that were otherwise less adapted to survive the harsh ice sheet conditions.Particularly in the in situ culturing methods used in this study, these cells ar e giv en mor e time, enabling stoc hastic awakening fr om dormanc y follo wing the "scout model" (Epstein 2009 ), but also potentiall y a mor e favor able set of envir onmental conditions that could induce awakening from dormancy according to the "comfort timing" strategy hypothesis (Laugier 2023 ).For the ca ptur e of diversity in genomes, it seems ther efor e beneficial to use a ( in situ) culturing a ppr oac h to access transient, dormant, or low abundant microbes if combined with metagenomics studies.

Ev alua tion of ( in situ ) culturing
The R2A plates inoculated in the field only yielded eight colonies, a very low amount compared to the 52 obtained through plating that was done under laboratory conditions once samples were returned from the field.The difference between the two a ppr oac hes was the presence of glycerol in the plates inoculated in the field, to match the medium that was used in the chips and chambers .T his study used gl ycer ol as this was essential as an antifreeze agent to protect the integrity of the agar under the field conditions .T herefore, we cannot exclude that if the gl ycer ol had a negative impact on growth of microbial isolates on the plates, it could also have done the same for the culture chips and chambers, since all chips and chambers contained the same concentration of glycerol.It is important to consider that choice of growth medium, as well as incubation conditions and sampling strategy, can influence the r esulting cultiv able or ganisms.
The cultur e c hips and c hambers k e pt their integrity while incubating on the ice sheet, despite weather conditions ranging from intense sun to heavy rain, hard wind, and freezing temperatures .Fungal o vergrowth and freeze-tha w cycles were foreseen as two challenges during the fieldwork planning phase.Fungal overgro wth w as not observed, meaning that the use of nystatin was unnecessary.To protect the structural integrity of the agar gel during r epeated fr eezing and thawing, gl ycer ol was added as an antifr eeze a gent.This pr ov ed to be effective, as the agar in the wells and chambers k e pt its structur e ov er the entir e period.Absence of liquid water, which can be an issue when incubating ichips in soils, was not a problem for the supraglacial environments used in this study.The c hips and c hambers incubating submer ged in the cryoconite hole wer e mor e similar to aquatic environments wher e cultur e c hips hav e been a pplied befor e (Jung et al. 2021 ).On the ice surface, the chips and chambers were k e pt above liq-uid water upon the weathering-crust but r eceiv ed plenty of exposure to hydration from rain and melt water.These natural processes were likely the conduits of nutrient transported through the membranes into the chips and chambers.Applying the chips and chambers on the ice sheet demonstrated that the range of use of in situ culturing can be extended further in the cryosphere.

Conclusion
A variety of complimentary methods (sequencing-and culturingbased) were used to investigate the microbial diversity of supr a glacial habitats on the southern area of the Greenland ice sheet.Classic amplicon sequencing a ppr oac hes wer e complemented with metagenome data.A clear advantage of the metagenomic a ppr oac h ov er amplicon sequencing was the ability to dir ectl y compar e pr okary ote and eukary ote SSU rRN A gene abundances, highlighting in particular the abundance of rRNA genes from eukaryotes over those from prokaryotes on the ice surface, and the opposite in cryoconite holes and biofilm.Two novel in situ culturing a ppr oac hes (cultur e c hips and c hambers) hav e been successfull y a pplied on the surface of the Greenland Ice Sheet.To our knowledge, this is the first time that these techniques have been applied to the challenging glacial en vironments , such as the ice surface and cryoconite holes .T his proof-of-concept pa ves the way for wider use of in situ culturing a ppr oac hes in the cryosphere.
We demonstrate that different genomic and culturing appr oac hes complement eac h other.If the objectiv e is to ca ptur e as m uc h genetic div ersity as possible, for instance in the search for biotec hnologicall y r ele v ant micr obes, MAGs complement the genomes obtained through culturing.The assembly of MAGs is biased by abundance in the metagenome, but the isolates obtained do not necessarily reflect abundant members of the community.The genomes obtained as MAG or through whole-genome sequencing thus originate from respectively more, and less established microbes.
These findings highlight the potential benefits of using multiple methods to study micr obial div ersity in complex en vironments , particularly for capturing genomic diversity.

F
igure 1. (A) Fieldw ork site on southw est mar gin of Gr eenland Ice Sheet.OpenStr eetMa p and Ar cticDEM (Porter et al. 2023 ) w er e used to cr eate ma p la yers .(B)-(D) examples of supr a glacial habitats investigated in this study; cryoconite (B, scale bar = 20 cm), biofilm in cryoconite hole (C, scale bar = 10 cm), and dark ice surface (D, scale bar = ∼ 120 m).

Figure 2 .
Figure 2. Cultur e c hips (B) and (D) and chambers (A) and (C) deplo y ed on ice (A) and (B) and submerged in cry oconite hole (C) and (D).Design of culture chambers (E) and culture chips (F) is based on a middle layer to form the chamber, which is closed off by two semipermeable polycarbonate membr anes, her e shown in blue.Two additional plastic outer layers protect the assembly.
Axenic cultures were grown in liquid culture (R2B, Alpha Biosciences, USA) at 5 • C on an orbital shaker (200 r m −1 ) until opaque (between a few days to 2 weeks), and subsequently transferred to cryotubes creating a stock in 18% gl ycer ol, flash fr ozen and stor ed at −80 • C. DN A w as extracted b y resuspending the pellet of 1 ml ov ernight cultur e in 100 μl milliQ water and boiling the suspension for 10 min.These extracts were stored at −20 • C. The V1-V8 region of the 16S (for bacteria) and the 5 terminal domain of the LSU rRNA genes (for yeasts) were targeted for amplification.The follo wing primers w er e used for pr okary otes 16S rRN A gene 27f (AGA GTT TGA TCM TGG CTC AG), 16S 1392r (ACG GGC GGT GTG TGT RC), and for eukaryotes (yeast) 26S rRNA gene LSU NL1 forw ar d (GCA T A T CAA GCG GA G GAA AA G), LSU NL4 r e v erse (GGT CCG TGT TTC AA C A CG G).PCR master mix contained 12.5 μl 2x Ultramix (PCR Biosystems, London, England) 0.5 μl of both primers (10 μM) and BSA (10 mg ml −1 ), 9 μl H 2 O, 2 μl template DNA.PCR protocol for 16S was as follows: 95 • C for 2 min, 30 cycles with 95 • C for 15 s, 55 • C for 15 s, 72 • C for 40 s, and final elongation at 72 • C for 4 min.LSU protocol was the same but using 45 s for denaturation, 30 s for annealing at 54 • C, and 2 min elongation.A volume of 5 μl of PCR product, together with 5 μl of forw ar d primer, w as sent for Sanger sequencing through the EZ-Seq service at Macrogen Europe (Amsterdam, the Netherlands).Geneious Prime version 2022.2.2 (Biomatters, Auckland, New Zealand) was used to c hec k the quality of sequencing results, trim ends, and BLAST the r esults a gainst the NCBI 16S and 18S ribosomal RNA databases.

Figure 3 .
Figure 3. Venn dia gr ams showing distribution of amplicon sequencing ASVs and meta genome SSU rRN A genes b y environment.

Figure 4 .
Figure 4. Heat map showing prokaryote r elativ e abundance data from metagenome and amplicon sequencing, abundance of MAGS and cultured isolates, all faceted by the habitat the sample came from.The top 19 most abundant phyla are shown.For the metagenome data, the r elativ e abundance is plotted for both the whole community and prokaryotes only.

Figure 5 .
Figure 5. Heat map showing eukaryote r elativ e abundance data from metagenome and amplicon sequencing, and cultured isolates, all faceted by the habitat the sample came from.The top 17 most abundant phyla are shown.For the metagenome data, the relative abundance is plotted for both the whole community and eukaryotes only.

Figure 6 .
Figure 6.Heat trees plotted using metagenome-extracted rRNA genes.Node size and color correspond to the relative abundance .T he top thirty most abundant taxa are labeled, and their text size also correlates with relative abundance.

Figure 7 .
Figure 7. Heatmap of metagenomic contigs mapped between (A) MAGs and (B) isolate genomes in unit of "reads per million" (Equation 1 ), which normalizes the sum of the mapped contigs by the length of the gene to one million.Hierarchical clustering depicts Spearman rank correlation of the (meta)genomes and the sample sites.Mapped reads are normalized with (A) abundance + 0.5 and (B) abundance + 1 to distribute values in a positive scale.Right side of panel (B) includes culture information; Origin: S = snow, C = cryoconite hole, I = ice, and B = biofilm.Method: P = plate (see methods) CC = culture chamber, C = culture chip.Antibiotics:-= none, N = nystatin, N + S = nystatin + stre ptom ycin.Heatmap de piction inspired by Rogers et al .( 2023 ).

Table 1 .
Sequencing output * only one biofilm replicate was used for MGS.

Table 3 .
Distribution of 160 identified isolates by culture method and environment.