Novel insights in cryptic diversity of snow and glacier ice algae communities combining 18S rRNA gene and ITS2 amplicon sequencing

Abstract Melting snow and glacier surfaces host microalgal blooms in polar and mountainous regions. The aim of this study was to determine the dominant taxa at the species level in the European Arctic and the Alps. A standardized protocol for amplicon metabarcoding using the 18S rRNA gene and ITS2 markers was developed. This is important because previous biodiversity studies have been hampered by the dominance of closely related algal taxa in snow and ice. Due to the limited resolution of partial 18S rRNA Illumina sequences, the hypervariable ITS2 region was used to further discriminate between the genotypes. Our results show that red snow was caused by the cosmopolitan Sanguina nivaloides (Chlamydomonadales, Chlorophyta) and two as of yet undescribed Sanguina species. Arctic orange snow was dominated by S. aurantia, which was not found in the Alps. On glaciers, at least three Ancylonema species (Zygnematales, Streptophyta) dominated. Golden-brown blooms consisted of Hydrurus spp. (Hydrurales, Stramenophiles) and these were mainly an Arctic phenomenon. For chrysophytes, only the 18S rRNA gene but not ITS2 sequences were amplified, showcasing how delicate the selection of eukaryotic ‘universal’ primers for community studies is and that primer specificity will affect diversity results dramatically. We propose our approach as a ‘best practice’.


Introduction
The phenomenon of snow and glacier surface discolorations caused by microalgal blooms during the melting season has long been known.On a global scale, this cryoflora is recognized to play an important role in primary productivity in otherwise hostile cryogenic environments (Williamson et al. 2019 , Hoham andRemias 2020 ).Mor eov er, colour ed snow and ice cause acceler ated melting due to biological albedo r eduction, whic h in turn affects sea le v el rise (Hotaling et al. 2021 ).Yet, there is no compr ehensiv e ov ervie w r egarding the biodiv ersity of suc h extr emophilic phototrophs at the species level, nor is there a standardized methodological pr otocol.Ther efor e, we pr esent her e a best pr actice a ppr oac h, with an optimized w orkflo w including consistent sampling, light microscopy-based morphological guidance, the generation of a ppr opriate r efer ence sequences and a final manual verification of taxonomic assignments .T he protocol was tested on 18 snow and ice samples from the Alps and the European Arctic.
Classical studies of cryoflor a (mainl y based on light micr oscopy) hav e pr ovided limited information due to poor mor phological differentiation of the predominant, often uniform algal cell types .In most cases , these are dominated by either red spherical algal cysts (in snow) or dark purple, short algal filaments (on ice).This led to early assumptions that few cosmopolitan species of green algae dominate worldwide (Kol 1968 ).At the end of the 20th century, Sanger sequencing using molecular markers impr ov ed this knowledge, and overturned older concepts of low phototr ophic biodiv ersity on snow and ice of glaciers and ice sheets.In the last decade, the advent of high-throughput sequencing (HTS; Oliv eir a et al. 2018 ), has enabled community characterization at the species le v el, making a molecular a ppr oac h feasible for various habitats and questions, and allowing c har acterizations of the entire biocenosis, including bacteria, archaea, fungi, and protozoa, whic h ar e inv ariabl y associated with pigmented snow and glacier ice algal blooms.
The first HTS study targeting snow-dominated cryo-habitats was carried out by Lutz et al. ( 2015a ) in samples from Iceland, using a section of the 18S rRNA gene marker for eukaryotes.They found that green alga of the genus Chloromonas were the most dominant in snow.Using the same marker in samples from Svalbard, Lutz et al. ( 2015b ) revealed differences in community composition betw een Ar ctic gr een and r ed snow, whic h wer e also reflected in differences in snow chemistry and metabolic profiles .T he first biogeographic HTS survey for these habitats compar ed r ed snow algal samples fr om Gr eenland, Sv albard, Iceland, and northern Sweden (Lutz et al. 2016 ).The communities were found to be lar gel y uniform with r espect to the most abundant taxa, regardless of different location-specific geochemical and mineralogical factors .Furthermore , Lutz et al. ( 2017 ) compared distinct habitats in the Arctic (gr een/r ed snow, biofilm, ice, cryoconite holes), and demonstrated the functional differences between communities of glacier and snow algae by combining amplicon sequencing, targeted metabolomic and physicoc hemical anal yses.Ho w e v er, most of the studies r elied solel y on the mor e gener al 18S rRNA marker gene, and no taxonomic refinements were made beyond automated species assignments against the SILVA database (extended by an additional 223 sequences of cryophilic algae; Lutz et al. 2016 ).Segawa et al. ( 2018 ) accomplished the first bipolar comparison of the red snow phenomenon.Using the 18S rRNA gene and ITS2 mark ers, the y concluded that in polar regions, these blooms consist mainly of cosmopolitan ' Sanguina ' phylotypes ( Chlam ydomonas ʾ-snow gr oup B).They also include endemic phylotypes ( Chlamydomonas ʾ-snow group A), which are distributed either in the Arctic or Antarctica.Davey et al. ( 2019 ) used 18S rRN A metabar coding to compare the taxonomic and metabolic pr ofiles of gr een and r eddish snow algal blooms in maritime Antarctica.The y re ported that coastal communities were dominated by Chloromonas spp.independent of the snow colour.Green sno w w as mor e pr otein-ric h, while r ed snow accum ulated mor e carotenoids and lipids, an observation also reported for green and red snow algae-rich samples from SE Greenland (Lutz et al. 2014 ).Luo et al. ( 2020 ) c har acterized cryoflor a comm unities on King George Island, Maritime Antarctica, using Illumina sequencing of the hypervariable V4 region of the 18S rRNA gene.In addition to the aforementioned genera, they found communities dominated by other green algae, namely red slush caused by Chlainomonas sp. and green snow by an undescribed Trebouxiophyceae.Similarly, Soto et al. ( 2020 ) investigated snow blooms on the same island, combining amplicon metabarcoding with photophysiological measurements .T hey found taxonomic differences between sampling sites depending on the dominant bloom colour (gr een, r ed, and particularl y y ello w-bro wnish for chrysophytes).Engstrom et al. ( 2020 ) and Yakimovich et al. ( 2020 ) performed amplicon metabarcoding on snow algae from mountain ranges in British Columbia, Canada.While they found ele v ation gradients between the dominant genera Chlainomonas , Sanguina, and Chloromonas , mutualism of bacteria and fungi with algae did not appear to be taxon specific.In contrast, Krug et al. ( 2020 ) found specific interkingdom connectivity, and thus, distinct algae-bacteria interactions in snow algal blooms in the Austrian Alps.Using a minimum entropy decomposition approach in ITS2, Br own and Tuc ker ( 2020 ) delineated the fine-scale geogr a phic and genetic population structure of Sanguina snow algae .T hey argued against a cosmopolitan distribution of a single species by finding a distinct r egional biogeogr a phic molecular structur es.In line with that, Pr oc házk ová et al. ( 2019 ) reported a large number of oligotypes in a global comparison of Sanguina ITS2 sequences, suggesting the existence of high (intra-specific) genetic variability of the widespr ead S. niv aloides (formerl y addr essed as cf.Chlam ydomonas nivalis ).
In contr ast, str eptophytic gr een algae of the genus Ancylonema ar e r esponsible for blooms on melting ice surfaces of glaciers and hav e onl y r ecentl y been r ecognized as important players in changing the albedo on glacier and ice sheet surfaces (Yallop et al. 2012, Cook et al. 2020, Williamson et al. 2020, Che vr ollier et al. 2023 ).The first amplicon sequencing study of these glacier ice algae (Lutz et al. 2018 ) documented their distribution along a 100 km long tran-sect on the western margin of the Greenland Ice Sheet, showing the presence of several Ancylonema -related oligotypes, with a sitespecific distribution.More recently, Winkel et al. ( 2022 ) found Ancylonema in communities on Iceland, using similar protocols.
Given the current state of the variability in amplicon sequencing protocols, it is clear that amplicon data have to be evaluated with caution.For example, Xiao et al. ( 2014 ) sho w ed for phytoplankton that light microscopy tentatively underestimates biodiversity (i.e.small and less abundant cells are neglected), and HTS can lead to misidentification at the species le v el (either due to limited database entries and/or insufficient length or genetic variability of the chosen marker sequence).To test the potential of the curr ent next-gener ation sequencing methods, w e sho w ed in Lutz et al. ( 2019 ), based on a case study of cryoflora in the Austrian Alps, that HTS outputs need to be thor oughl y c hec ked when the organisms are poorly represented in sequence databases.Crucial is the a ppr opriate c hoice of similarity thr esholds to cluster sequences into OTUs (or ASVs), as well as the delineation of species (Lutz et al. 2018 ).
To our knowledge, the workflow of amplicon metabarcoding of snow and glacier ice micr oalgae comm unities still lacks a 'best pr actice' pr otocol.To fill this ga p and impr ov e species-le v el assignments , we introduce , besides other optimisations, new ITS2 zygnematophycean-specific primers and show that these finally allow for a thorough taxonomic assessment.

Field sampling and microscopy
A total of 18 samples from different locations in the Austrian and Swiss Alps ('Alpine'), as well as from Svalbard and Greenland ('Arctic') wer e anal ysed.The geogr a phical origin, GPS position, ele v ation, date of harv est and type of cryoflor a bloom used for Illumina sequencing are summarized in Table 1 .Sites were selected based on logistical availability and representation of the most common bloom types in this geogr a phic r egion (gr een, orange, red, and golden-bro wn sno w).Field sampling, sample processing and stor a ge prior to anal yses wer e performed as pr e viously described (Lutz et al. 2019 ).Briefly, surface snow or ice was tr ansferr ed to sterile 50 ml plastic tubes and stored frozen at -20 • C prior until further pr ocessing.Separ ate sample aliquots were collected in 50 ml tubes for light microscopy.Algae were observed and classified dir ectl y in their meltwater by light microscopy using either a Leica 700, a Zeiss Axiovert 200 M or a Nikon Eclipse 80i microscope.

DNA Sanger sequencing
To obtain the first ITS2 r efer ence sequences of glacier ice algae and Trochiscia -like red cells, virtually monospecific field populations were collected, identified by light microscopy, and subjected to Sanger sequencing.while when < 20 mg wet biomass was available (DR74a and DRAnt023), DN A w as extracted using the Instagene Matrix Kit (Bio-Rad Laboratories , Hercules , C A, USA).ITS2 was amplified from DNA samples by pol ymer ase c hain r eaction (PCR) using existing primers and new reverse primers ( Table S1 ).Specifically, primer pairs of ITS5 + ITS4 and TW81 + AB28 were used to amplify ITS2 from DRAnt023 and DR74a, respectively.To generate a longer DNA fr a gment cov ering ITS1 partial + 5.8S + ITS2 + partial 26S rRNA of WP211 and WP167 the primer pairs of ITS1 + LR3 and Zyg_ITS_F + LR3, r espectiv el y, wer e used.For Illumina sequencing, 'ice primer' pairs were selected targeting ITS2 regions (together with part of the 5.8 S rRNA), tested on unialgal Ancylonema material (WP211 and WP167) and finall y de v eloped and optimized for streptophytic (zygnematophycean) algae: 5.8SbF2 (CGATGAA GAA CGCA GCG) (Mikhailyuk et al. 2008 ) and the new LSULP (AATTCGGCGGGTGGTCTTG (this study).The amplification reactions and PCR mix for WP167 and WP211 were the same as described in Pr oc házk ová et al. ( 2018a ).In case of DR74a and DRANT023, amplification reactions were as follows: each 20.52 μL PCR reaction for amplification of 18S rRNA and rbcL genes contained 1 μL of DNA isolates (diluted to concentration of 5 ng μL -1 ), 4.32 μL 5x MyTaq Red Reaction Buffer (Bioline, Meridian Bioscience, USA), 1.08 μL of each 10 μM primer, 13.82 μL sterile Milli-Q water, and 0.22 μL of 5 U μL -1 MyTaq HS Red DNA pol ymer ase (Bioline , Meridian Bioscience , USA); amplification reactions were performed using the follo wing c ycle parameters: initial denaturation for 3 min at 95

Processing of 18S rRNA gene sequences
18S pair ed-end r eads wer e quality filter ed, trimmed, and denoised into ASVs using dada2.The first 10 bp of each read were trimmed off.Forw ar d reads were truncated at 250 bp and reverse reads at 200 bp in order to r emov e low quality r egions.ASVs wer e annotated using a Naive Bayes classifier pre-trained on the full-length Silva (v.132) database.Sequences matching bacterial and archaeal DN A w er e r emov ed fr om the 18S dataset.The featur e table was rarified using the lo w est common number of sequences per sample (i.e.61 000) and only ASVs with a minimum frequency of 10 across all samples were retained.For a better overview of the algal community composition sequences annotated with 'Chloroplastida' and 'Oc hr ophyta' wer e extr acted.The 51 most abundant ASVs (ASVs containing > 400 sequences across all samples) were manually BLASTed against the NCBI nt database.

Processing of ITS2 snow sequences
ITS2 regions of forw ar d and r e v erse r eads wer e extr acted separ atel y using itsxpr ess ('trim-pair-output-unmer ged') and subsequentl y filter ed, trimmed, and denoised into ASVs using dada2.Reads were not further trimmed.The Qiime compatible UNITE database of all eukaryotes (v.8.0) was imported into Qiime2 and the r efer ence r eads and taxonomy wer e used to tr ain a Naiv e Bayes classifier.Reference sequences were annotated using the pr e-tr ained classifier.ASVs with a minimum frequency of 10 across all samples were filtered out.For a better overview of the algal community composition sequences annotated with 'Viridiplantae' and 'Chromista' were extracted and retained in a separate algal feature table.It was also noted that some algal sequences were annotated as 'Protista' or 'Unassigned'.Therefore, sequences with these annotations were extracted and the 51 most abundant ASVs in the snow and glacier ice algae datasets wer e manuall y BLASTed a gainst the NCBI nt database.Sequences matching algae were retained and added to the algal feature table.Again, the 51 most abundant ASVs in this table were manually BLASTed against the NCBI nt database.Sequences matching to algae were retained and added, and non-algal sequences were r emov ed until the 44 or 51 most abundant ASVs r epr esented algal sequences in the snow and glacier ice algae datasets, r espectiv el y.

Processing of ITS2 ice sequences and secondary structure
ITS2 ice paired-end reads were quality filtered and denoised into ASVs using dada2.All taxonomic assignments of r epr esentative ASVs were performed by blast search against the nucleotide database in NCBI (see the 'ASV assignment and ITS2 secondary structure' section).The methods of annotation and prediction of the secondary structure of the ITS2 regions were the same as in Lutz et al. ( 2019 ) and Remias et al. ( 2020 ).The secondary structure of nuclear rRNA ITS2 of were drawn using VARNA version 3.9 (Darty et al. 2009 )

ASV assignment and ITS2 secondary structure
For the most abundant ASV IDs in this study, ASV assignment was performed by blast searc h a gainst NCBI.In the case of 18S rRNA gene sequences, an identity threshold of ∼99.4% had to be passed in order to be considered as a database match (i.e. a maximum of 2 bp nucleotide difference in a 342 bp sequence was allowed; Lutz et al. 2019 ).Sequences below this threshold were recorded as 'no blast hit'.For ITS2, an identity threshold of ∼89% against a r efer ence was r equir ed to be consider ed as a database matc h (Lutz et al. 2019 ).Sequences below this threshold w ere recor ded as 'no blast hit'.

Haplotype network
The alignments of the 19 most abundant A. nordenskioeldii , A. alaskanum , and Ancylonema sp.ASVs and the 35 most abundant Sanguina nivaloides , Sanguina aurantia , Sanguina sp.DR74a, and Sanguina sp.H14 ASVs were used in the consensus secondary structure modelling and haplotype networks.In the case of Sanguina sp.H14, the gr a ph w as supplemented b y the se v en most closel y r elated published ITS2 sequences from the states of Colorado and Washington, USA.(KX063717, KX063721, KX063724, KX063726, KX063729, KX063731, and KX063742; Brown et al. 2016 ).Each of these most abundant ASVs contributed more than by 1% to the ITS2 sequence data in at least one of the samples.A haplotype was delineated at the 100% similarity threshold (i.e. two sequences belong to the same haplotype if they are identical).Haplotype netw orks w er e used to demonstr ate the intr aspecific div ersity (within a species) and geogr a phical distribution of each haplotype (Škaloud et al. 2015 ).The nuclear network for ITS2 was constructed using TCS softwar e v ersion 1.21 (Clement et al. 2000 ) and statistical parsimony with a connection limit of 95%.The final edit of the haplotype network was done in Inkscape.

Results
A total of 18 European Arctic and Alpine habitats were investigated to test the impr ov ed metabarcoding a ppr oac h.These showcase blooms included three red snows in the Alps, one red and one orange snow in Svalbard, one reddish bloom on a glacier in Sweden, and finally 10 locations of bare ice surfaces in the Alps, Sv albard, and Gr eenland.These 16 comm unities wer e selected as 'r epr esentativ e' of melting snow and ice ecosystems in this region.Two less common phenomena of facultative snow algae were also included, namely sub-surface green snow (caused by algae atypical for cry oflora) belo w white snow in the Austrian Alps, and golden-bro wn sno w slush caused b y c hrysophytes in Sv albard.

Total eukaryotic community compositions
Figure 1 visualizes the abundance of the major eukaryotic taxonomic groups in each sample of melting snow and glacier icer algae habitats, based on metabarcoding of the 18S rRNA gene sequence marker.Micr oalgae [Chlor ophyta, Char ophyta ( = Streptophyta)] were most abundant, followed by fungi and the SAR group (Cercozoa, Ciliophor a, and Dinofla gellata).Samples did not cluster by sites or habitat in the non-metric multidimensional scaling (NMDS) plots.Compared to all other samples, the Arctic goldenbrown bloom (WP203) was dominated by chrysophytes only and contained, furthermore, almost no fungi.Results for prokaryote communities can be found in the supplement ( Table S2, Fig. S14 ).

Snow and glacier ice algal community composition
The Alpine and Arctic communities were compared using two different primer pairs (18S rRNA gene, ITS2).In the case of ITS2, two different primers ('snow' and 'ice') had to be used to cover more members of the cryoflora.Figure 2 shows the NMDS plots and the corresponding relative read abundances for microalgae in bar charts.In all NMDS plots, samples were assigned to one of four habitat classes: 'alpine sno w', 'ar ctic sno w', 'alpine ice', and 'arctic ice'.Analysis of similarity (ANOSIM) for the 18S rRNA gene sho w ed a high ov erla p between eac h of the snow and ice habitats (R = 0.7262, P = 0.001), regardless of whether they w ere ar ctic or alpine.Using, ITS2 snow primers, diversity sho w ed no significant difference betw een Ar ctic and Alpine sites (R = 0.2236, P = 0.032) and habitat types (R = 0.3404, P = 0.01).In the case of the ITS2 ice primer, glacier ice algae habitats differed between Alpine and Arctic locations (R = 0.9074, P = 0.011).
Figur e 1. T he total eukaryotic community composition and similarities of snow and glacier surface communities based on amplicon metagenomics of the 18S rRNA marker.The samples were assigned to one of four habitat classes and accordingly labelled in the plot and above the bars: 'alpine snow' (orange cir cle), 'ar ctic sno w' (orange ring), 'alpine ice' (blue cir cle), and 'ar ctic ice' (blue ring).
In a subsequent step, the 18S rRNA gene and ITS2 marker sequences were used to perform a detailed description of the eukaryotic phototrophs for each site, and to assign the ASVs according to r efer ence databases .T he bar charts in Fig. 2 show the most abundant algal ASVs found in the comm unities.Mor e extensiv e lists of abundant ASVs for the 18S rRNA gene and for the two ITS2 markers can be found in the Supplementary Tables S3 -S5 .These are derived from manually optimized taxonomic assignments using blast against the NCBI database.In detail, 18S rRNA metabar coding sho w ed that se v er al species of Sanguina (6 ASVs) wer e r esponsible for r ed blooms in both the Eur opean Arctic and Alps, but the limited length of the marker fr a gment did not species identification.All ice surfaces were inhabited by glacier algae of the genus Ancylonema , with 4 ASV belonging to the filamentous A. nordenskioeldii and 1 ASV to the unicellular A. alaskanum (formerly ' Mesotaenium berggrenii v ar.alaskana ', r eferr ed to as 'A.alaskana' in Pr oc házk o vá et al. 2021 ).T he latter was most abundant in the samples from the Alps, Svalbard, Sweden, and on a valley glacier in Greenland, but had very few reads in the samples from the Greenland ice sheet.On Greenland's Mittivakkat glacier (MIT 12-9), a mixture snow algae ( Sanguina ), glacier algae, and a psyc hr ophilic ubiquitous alga of cold regions ( Raphidonema sempervirens ) was found.The sub-surface ('atypical') green snow in the Alps (WP119) was dominated by Chlorominima sp. and Koliellopsis inundata (Chlorophyceae), both of which were not found in any other sample.In detail, ITS2 secondary structur e anal yses sho w ed that the first alga was closel y r elated to Chlorominima collina CC-Cryo 273-06 ( Fig. S1 ), but it r epr esented an independent species.A Chlainomonas species (ASV read 100% identical to OTU008, and 99.7% identical to: Chloromonas rubroleosa CC1, Chlainomonas sp.RRD1, Chlainomonas sp.Bagley_NODE_5578, and Chlainomonas sp.190526Oze2R) was found in all glacier ice samples and in one Svalbard red snow sample (WP205), but only in low abundance ( < 5.2% algal reads).
Unicellular Chrysophyceae dominated in Arctic golden-brown snow slush (Svalbard, WP203), and 5 ASVs were affiliated with Hydrurus sp.(two of which dominated this site).On Vestre Bröggerbreen in the same archipelago (Sva 13-8), Ochromonas clone Esp29 was co-dominant with many other microalgae.In the other melt-ing habitats and in Alpine sites, Chrysophyceae were detected only in very low abundances.
The ITS2 marker amplicons were able to elucidate the diversity in more detail ( Figs.S1 -13 ).In most respects, the qualitative taxonomic r esults wer e similar to those obtained with the 18S rRN A gene.Ho w e v er, in the case of Sanguina , ITS2 accompanied b y CBC sear c h in the secondary structur e was able to assign ASV sequences at the species le v el (fr om the ne wl y gener ated r eference sequence or from the type material BlastN match), including S. nivaloides in the Alps (1 ASV), Sanguina sp.DR74a in the Arctic and Alps (6 ASVs), for S. aurantia only in Arctic habitats (5 ASVs), and for Sanguina sp.H14 in the Arctic and Alps (3 ASVs).Based on ITS2 secondary structure comparison ( Figs.S2 and S3 ), the latter also belongs to the genus Sanguina , but r epr esents a so far undescribed species (pr e viousl y r eferr ed to as uncultur ed 'Chlam ydomonas' clone H14 in Brown et al. ( 2016)).Within the glacier ice algae phototrophic communities distinct geographical patterns were found (Fig. 3 ): Arctic habitats (in Greenland and Svalbard) w ere colonized b y one to three genetically distinct glacier ice algal species (which was not visible in the 18S rRNA gene sequence data), A. nordenskioeldii, A. alaskanum , and Ancylonema sp. ( Figs .S4 -S7, Table S6 ).T he latter undescribed taxon was also found in the Swiss Alps and Sweden, but at m uc h lo w er abundances ( < 0.5% of reads) ( Table S5 ).LM suggested that this undescribed species of Ancylonema might be filamentous and morphologicall y distinct fr om A. nordenskioeldii .In contr ast, the unicellular A .alaskanum was more dominant in Alpine habitats but almost absent from the Greenland Ice Sheet, as indicated by the haplotype network: A. nordenskioeldii resulted in ten ITS2 interconnected haplotypes (Fig. 3 ).The haplotypes differed by one to twelve nucleotide changes out of 194 base pairs.For A. alaskanum , five haplotypes wer e r ecov er ed fr om the dataset, differing from two to eight nucleotide changes out of 197 base pairs.Haplotype networking was also performed for Sanguina (Fig. 4 ).In the case of Sanguina sp.DR74a, this resulted in six ITS2 interconnected ha plotypes, eac h with either an Alpine or Arctic distribution.The same pattern was observed for three haplotypes of Sanguina sp.H14.
Other members of the glacier algal communities were R. sempervirens ( Fig. S8 ), which occasionally contributed up to 45% of ITS2 r eads, namel y in the Arctic (Sv albard) and the Alps (Switzerland), and an alga related to Ploeotila sp.CCCryo086-99 ( Fig. S9 ), abundant at an Arctic site (Greenland Ice Sheet (GrIS)), and Limnomonas spitsbergensis CCCryo 217-05 ( Fig. S10 ), observed in the south-eastern GrIS.Chlainomonas r eads wer e not r ecov er ed in the ITS2 dataset, likely because its amplifiable fr a gment was too long (about 640 bp including priming sites), and thus pr e v enting it fr om being sequenced by the sequencing platform used (MiSeq 300 bp PairEnd).
In general, the plausibility of the molecular results regarding the abundances of phototrophs was roughly evaluated and confirmed by light microscopy (data not shown).Four exceptions were found.The first was an Alpine red snow sample (WP181), which sho w ed discrepancies betw een LM evaluation and ITS2 Illumina abundances: While the molecular data indicated ∼60% Chloromonas sp.CCCryo 261-06 (conspecific with Chloromonas cf.alpina CCCryo 033-99; Fig. S11 ) and only 18.9% Sanguina -like ASVs, the LM sho w ed > 90% Sanguina -like c ysts.In contrast, but less striking in terms of div er gent numbers, another Alpine r ed snow (WP117) contained to some extend Chloromonas cells in the LM, but the latter were absent in the ITS2 Illumina data.Thirdly, Alpine red snow WP127 contained a high number of Sanguina cysts, but these were absent from the ITS2 molecular results, which were instead dominated by sequences of snow-dwelling Chloromonaslike spp.instead, namely Scotiella cryophila , Chlorophyta clone ALBC6 ( Fig. S12 ), or an alga related to OTU375 ( Fig. S13 ).Finally, Figure 3. Geogr a phic distribution of glacial algae based on molecular ITS2 haplotypes detected in this study: (A) A. alaskanum , (B) A. nordenskioeldii , and (C) Ancylonema sp.Only the most abundant ASVs were used which accounted for > 1% of the ITS2 sequence data in at least one of the samples.Each haplotype network was constructed by a statistical parsimony method with a 95% connection limit.The geogr a phical origin was labelled with colours according to the legend (upper left).Each circle represents a haplotype (i.e.field samples which had identical Illumina ITS2 ASVID).The size of the circle is proportional to the number of sampling sites, which belong to that specific haplotype.Lines connect each haplotype with its most similar relative.Open dots r epr esent m utation steps between ha plotypes, one dot indicates a change of one base pair.
the Hydrurus -related Chrysophyceae were generally not amplified with both ITS2 primers used in this study, resulting in a distorted comm unity structur e in the corr esponding bar c hart shown in Fig. 2 .

Discussion
The application of this best practice protocol confirmed that the main members of the local cryoflora are distributed within three genera of green algae: Chloromonas , Sanguina , and Ancylonema (Hoham and Remias 2020 ) .In the case of the latter two, undescribed species were revealed using ITS2.The likely fourth frequent genus, Chlainomonas , which causes red sno w, w as onl y mar ginall y covered by this study, as European blooms of this genus are mostly restricted to the melting ice cover of some high-altitude mountain lakes (Pr oc házk ová et al. 2018b ).Since amplicon sequencing can also r ecov er low-abundance taxa, it was possible to detect Chlainomonas using 18S rRNA in all glacier ice samples from the Eur opean Alps, the Gr eenland Ice Sheet, and a Swedish glacier.This alga is identical to an alga pr e viousl y found in Antarctica (Segawa et al. 2018 ) and very closely related to the records from North America and New Zealand (Novis et al. 2023 ).
Golden-bro wn sno w caused b y unicellular Chrysophyta is closel y r elated to the genus Hydrurus (Remias et al. 2013b, Luo et al. 2020, Soto et al. 2020 ), which was confirmed in this study.Also in the case of c hrysophytes, cryoflor a comm unities ar e dominated by species with very close genetic relationship.Not unexpectedly, this leads to similar cell morphologies and a low variation in certain molecular markers , e .g. 18S rRNA gene or rbc L. As a result, HTS studies with snow and glacier ice algae are challenging in terms of r e v ealing true biodiv ersity or comparing habitats.For example, Nakashima et al. ( 2021 ) argued that partial 18S rRNA gene sequences were insufficient for species differentiation within Sanguina .In such cases of very close taxa, evaluation of the hyper-v ariable secondary structur e of the ITS2 marker is an alternative, as demonstrated here.Ho w ever, this screening for compensatory base exchanges is currently non-automated and labour-intensive (Segawa et al. 2018, Lutz et al. 2019, Yakimowich et al. 2021 ).
A k e y finding of this study was that HTS r e v ealed unexpected biodiversity in at least two cases: first, the impr ov ed pr otocols r evealed a third species of glacier ice alga, Ancylonema sp.This is consistent with morphological differences: LM indicated that this species is filamentous and can be distinguished from A. nordenskioeldii by smaller cell sizes (Pr oc házk ová et al. 2021 ).Furthermore, the ITS2 marker sho w ed 91% sequence identity between A. nordenskioeldii and A. alaskanum (at 100% sequence cov er a ge), and the secondary structures of ITS2 transcripts revealed a single CBC in the middle part of helix I ( Fig. S3 ).This provides good molecular support that both r epr esent two independent species.Secondl y, the use of the ITS2 marker was crucial for the species-le v el elucidation of red snow.The use of metabarcoding combined with ITS2 secondary structure comparison revealed the existence of undescribed species, the two most important being Sanguina sp.DR74a and Sanguina sp.H14.They differ ed fr om the described species S. nivaloides and S. aurantia in their ITS2 secondary structures .T his confirms that the red snow genus Sanguina contains more species than curr entl y r ecognized but their cysts may r emain mor phologically indistinguishable.
Exceptional results came from tw o Ar ctic glaciers (Sva 13-19, Sv a 13-50) wher e both c hlor ophycean and str eptophycean micr oalgae wer e co-occurring.This can be explained by natural succession during the melting season: two independent seasonal blooms occurred chronologically.First, chlamydomonadacean fla gellates thriv e in melting snow, and later, zygnematophycean species grow on the bare ice surface after snow melt (Takeuchi 2013, Lutz et al. 2014 ).T hus , snow and glacier ice algae can be found together on the same sites during the late melting season, but in our experience there is not necessarily a causal r elationship.Suc h aspects of successional de v elopment wer e not consider ed in this study due to high logistical demands, hence all sites were sampled only once per season.Moreover, golden-br own comm unities and gr een snows can be consider ed e phemeral, thri ving due to either good water or nutrient availability.In such cases, succession can lead to either r a pid complete snow melt or transformation into red snows and glacier ice algae blooms on bare ice .T he latter ones can be regarded as 'climax stages' until complete snow melt or the onset of the next winter snowfall.
While ' true' cry oflor a is thought to r epr oduce onl y in snow and on ice (Hoham and Remias 2020 ), ubiquitous microalgae that can cope well with low temper atur es and high irr adiance ar e also found in these communities .T his was particularly the case for Raphidonema and Koliella (Trebouxiophyceae).Under favourable conditions (high availability of liquid w ater, lo w irradiance, and high nitrogen input), species atypical of the cryoflora can produce visible blooms, as was the case for Alpine sample WP119.In addition, another dominant trebouxiophycean alga caused green and or ange patc hes in the snow in coastal Maritime Antarctica (Soto et al. 2020 , partial 18S rRNA gene: O TU1674, O TU1653, O TU3753).It was initially assigned to 'uncultured Chlorella ' (accession number AB903015) and is curr entl y assigned to K. inundata (accession number MT274431).This is another example of how the quality and constant updating of r efer ence database can be essential for species delimitation.
Snow blooms caused mainly by Chloromonas species were not cov er ed by this study, but they are w ell kno wn from mountainous and coastal polar habitats (Hoham and Remias 2020 ).Ho w e v er, cells of this genus were present in many of the samples in vestigated.T he ne wl y gener ated ITS2 r efer ence sequence of C. pol yptera, obtained fr om snow adjacent to penguin r oc keries in Maritime Antar ctica, w as used to test the hypothesis of its absence in the northern hemisphere: In fact, there was not a single database match for the Alps and the Arctic.Previous reports of this Antarctic species in HTS studies of the Northern hemisphere (e.g.Lutz et al. 2015a, Terashima et al. 2017 ) may be due to ambiguous species assignment, as using only a section of the 18S rRNA gene does not distinguish it from other closely related Chloromonas species (Lutz et al. 2019 ).
Rosetta cells ('ruby cysts' etc.) were formerly assigned to Chlamydomonas nivalis (Kol 1968 ), their morphology and phylogenetic position is under pr epar ation (Engstr om et Ra ymond-pers .comm.).In this study, in the three samples (WP181, 127, and WP117), up to 7% of 18S rRNA gene r eads wer e identical to the uncultured algal isolate 0935-5 (accession number LC371440), which morphologicall y a ppear ed as dark red ellipsoidal cells (based on single-cell Sanger sequencing-see Fig. S22 in Segawa et al. 2018 ).This finding suggests that some Rosetta -like cells can also be found in the European Alps .T his is another example of how the understanding of microbial composition is str ongl y dependent on the quality of r efer ence sequence databases.Indeed, samples WP127 and WP181 sho w ed a high proportion of 'no blast hits' for ITS2 (Fig. 2 ).
We cannot exclude a technical issue that could be responsible for the observ ed discr epancies between ITS2 and 18S rRNA gene sequence results in some samples , e .g. sequencing errors that lead to a m uc h higher r elativ e abundance of certain ASVs when compared to the actual abundance of the organism in the sample (due to incorr ectl y r ecov er ed homopol ymer segments and carry-forw ar d incomplete-extension err ors, Lüc kling et al. 2014 ).This can be mitigated by se v er al methodological measures.This includes the generation of OTUs based on phylogenies as monophyletic lineages and the exclusion of very long br anc hes with low abundances that are close to very abundant ASVs, as well as the inclusion of tec hnical r eplicates, the r emov al of spurious sequences and unr epr esentativ e O TUs , the use of high stringency clustering methods for out generation, the estimation of treatment effects at higher taxonomic le v els, the ada ptation of the unique molecular identifier and other ne wl y de v eloped methods to reduce PCR and sequencing errors, and the identification of true low abundance r ar e species.Combined, these can impr ov e r epr oducibility (Wen et al. 2017 ).Ne v ertheless, Illumina 18S rRNA data, as a proxy for genus-le v el determination, corr esponded well with LM observations for the majority of samples .T herefore , light microscopy guidance is crucial in microalgae community assessment to detect and e v aluate possible data inconsistencies.A second issue to gener all y impr ov e the r esults is the way the field sampling is done: ideall y, harv est into 50 ml tubes should not be done from a single spot, but rather pooled from e.g.five spots of the bloom of 10 ml each within a giv en squar e meter to avoid r andom sampling err ors.Ov er all, a unified method for snow and glacier ice algae sampling, sample pr eserv ation, extr action, sequencing and data analyses etc is advised when the aim is to compare data sets across locations.
In our samples, two blooms were caused by facultative cryoflor a: subsurface Alpine gr een snow (WP119), wher e a dominant alga w as sho wn to be an undescribed taxon, which is closel y r elated to C. collina .The latter has been described as a psyc hr ophile from the snow in Maritime Antarctica (Gálvez et al 2021 ).In addition, we present the another report of L. spitsbergensis from Sv albard (Sv a13-50), whic h was originall y isolated and pr e viousl y r eported fr om persistent snowfields at Spitsbergen (Tesson and Prösc hold 2022 ).The widespr ead occurr ence of these algae in snow is not surprising, as they follow different rules than the true cryoflora.T hey ma y 'accidentally' bloom in quite water-logged snow or be triggered by external nutrient inputs.In the context of golden-bro wn sno w caused b y c hrysophytes, mor e sampling will be needed to elucidate their total biodiversity, and the fact that they were not detected by the ITS2 snow primers leaves their diversity at the species-level unresolved.
In conclusion, the aim of this study was to target the different c har acteristic cryoflor a types of the Alps and the European Arctic in order to assess a deeper le v el of biodiversity at the species le v el.In the future, f urther community metabarcoding using the highresolution ITS2 marker will help to elucidate which haplotypes (r espectiv el y species) dominate under specific conditions.In gener al, an impr ov ed primer pair for ITS2 reading 'glacier ice algae' and 'snow algae' sequences will be mandatory to cover all relev ant gr oups of micr oalgae in a conv enient wa y.T he results of this and similar studies curr entl y point to the presence of both ubiquitous cosmopolites and local species with limited distribution.It remains to be seen whether the latter are either geographically isolated or whether their occurrence depends on unknown ecological conditions.

Figur e 2 .
Figur e 2. T he snow and glacial algal community compositions based on the 18S rRNA and ITS2 markers.Similarities between samples are shown in the NMDS plots and the most abundant taxa are represented in the corresponding bar plots.For visualization, up to 10 most dominant species (ASVs) were selected for each sample .T he NMDS plots are shown to the left and the according r elativ e molecular abundances in bar charts to the right.The samples were assigned to one of four habitat classes and accordingly labelled in the plot and above the bars: 'alpine snow' (orange cir cle), 'ar ctic sno w' (orange ring), 'alpine ice' (blue circle), and 'arctic ice' (blue ring).

Figure 4 .
Figure 4. Geogr a phic distribution of Sanguina species based on molecular ITS2 haplotypes detected in this study: (A) S. nivaloides, (B) S. aurantia , (C) Sanguina sp.DR74a, and (D) Sanguina sp.H14.In case of (D), the gr a ph w as supplemented b y the se v en closest r elated published ITS2 rDNA sequences, fr om Color ado and W ashington States, U .S.A. (KX063717, KX063721, KX063724, KX063726, KX063729, KX063731, KX063742; Brown et al. 2016 ).The geogr a phical origin was labelled with colours according to the legend (upper left).Each circle r epr esents a haplotype .T he size of the circle is proportional to the number of sampling sites, which belong to that specific haplotype.Lines connect each haplotype with its most similar relative.Open dots r epr esent m utation steps between ha plotypes, circle indicates a c hange of one base pair.

Table 1 .
Sampling sites of snow and glacier ice with sample id, location, region, habitat, date of harvest, geographic position, and ele v ation in meters above sea level.

ID location region ha bitat c haracteristics harvest GPS ele v a tion
Al, Alpine; Ar, Arctic