Microbial Communities and Their Predicted Metabolic Functions in Growth Laminae of a Unique Large Conical Mat from Lake Untersee, East Antarctica

In this study, we report the distribution of microbial taxa and their predicted metabolic functions observed in the top (U1), middle (U2), and inner (U3) decadal growth laminae of a unique large conical microbial mat from perennially ice-covered Lake Untersee of East Antarctica, using NextGen sequencing of the 16S rRNA gene and bioinformatics tools. The results showed that the U1 lamina was dominated by cyanobacteria, specifically Phormidium sp., Leptolyngbya sp., and Pseudanabaena sp. The U2 and U3 laminae had high abundances of Actinobacteria, Verrucomicrobia, Proteobacteria, and Bacteroidetes. Closely related taxa within each abundant bacterial taxon found in each lamina were further differentiated at the highest taxonomic resolution using the oligotyping method. PICRUSt analysis, which determines predicted KEGG functional categories from the gene contents and abundances among microbial communities, revealed a high number of sequences belonging to carbon fixation, energy metabolism, cyanophycin, chlorophyll, and photosynthesis proteins in the U1 lamina. The functional predictions of the microbial communities in U2 and U3 represented signal transduction, membrane transport, zinc transport and amino acid-, carbohydrate-, and arsenic- metabolisms. The Nearest Sequenced Taxon Index (NSTI) values processed through PICRUSt were 0.10, 0.13, and 0.11 for U1, U2, and U3 laminae, respectively. These values indicated a close correspondence with the reference microbial genome database, implying high confidence in the predicted metabolic functions of the microbial communities in each lamina. The distribution of microbial taxa observed in each lamina and their predicted metabolic functions provides additional insight into the complex microbial ecosystem at Lake Untersee, and lays the foundation for studies that will enhance our understanding of the mechanisms responsible for the formation of these unique mat structures and their evolutionary significance.


INTRODUCTION
Microbial mats are often considered to be modern day analogs of ancient organo-sedimentary structures such as microbialites and stromatolites, and the geological record indicates their origin as early as ∼3.5 Ga BP (Gerdes, 2010;Seckbach and Oren, 2010). Morphological diversity and longevity of stromatolitic formations from the Archaean and throughout the Proterozoic indicate its success as a growth strategy, and comparisons with extant examples of microbial mats can explain the possible reasons behind this success (Grotzinger and Knoll, 1999;Hofmann et al., 1999). Modern microbial mats often represent metabolically layered structures, capable of oxygenic photosynthesis at the top of a mat where light is most abundant (Foster and Green, 2011). In the deeper segments in the mats, reduced light and depleted oxygen enables the growth of organisms capable of a cascade of anaerobic metabolisms (Kühl and Fenchel, 2000). Such layering is often visible as color changes, with an orange surface from abundant lightprotective carotenoid pigments, transitioning to green lightharvesting phycobilin and chlorophyll pigments and pink layers corresponding to photosynthetic sulfur bacteria in anoxic zones. There are many examples of this spatial organization of microbial communities with defined metabolic roles in the literature, and such close proximity of interacting metabolic structures allows for diffusion to be an effective coupling mechanism that provides key advantages to this mat structure (Seckbach and Oren, 2010). Viewed as an integrated unit, the microbial mat shows metabolic coupling, facilitated by this one-dimensional, organized structure, that allows a more efficient utilization of resources than any one organism could do alone (Stahl, 1995;Paerl and Pinckney, 1996;Franks and Stolz, 2009). Indeed, some argue that the organized complexity and physiological cooperativity of a microbial mat is analogous to that of eukaryotic tissues (Webb et al., 2003;Guerrero and Berlanga, 2010).
These laminated microbial mats are often best developed under extreme conditions such as thermal springs, sulfur springs, high salinity and icy polar environments where conditions preclude large metazoan bioturbators that would otherwise disrupt mat formation (Seckbach and Oren, 2010). Microbial mats in ice-covered Antarctic lakes are thought to be particularly well developed as, in addition to the exclusion of higher eukaryotes through extreme conditions and isolation, they are also protected from physical perturbation -wind and wave forcing -by thick, perennial ice cover (Priscu and Foreman, 2009;Cavicchioli, 2015). To date, most of our understanding of microbial mats in the Antarctic continent has come from perennially ice-covered lakes in the McMurdo Dry Valleys (Priscu and Foreman, 2009;Jungblut et al., 2016). In particular, benthic photosynthetic cyanobacterial mats cohabited by heterotrophic bacterial communities have been implicated as the key biological components in maintaining the lake's productivity and nutrient cycling (Laybourn-Parry and Pearce, 2007;Quesada et al., 2008). However, in addition to the McMurdo Dry Valleys, a wide range of freshwater lakes have been described in the East Antarctic ice plateau, ice-free oases and foothills of the Transantarctic Mountains. Among these, Lake Untersee (71.34 • S, 13.45 • E) is the largest of the perennially ice-covered lakes in the interior of central Queen Maud Land (Loopmann et al., 1988;Wand et al., 1997Wand et al., , 2006. The lake presents an exclusively microbial ecosystem and consists of unique conical microbial mats that appear as smooth cones, rising to ∼0.5 m above the lake bottom, unlike anything reported thus far from other Antarctic or temperate lakes . Microbial community structures in ice-covered Antarctic lakes have been studied using a range of culture-dependent and culture-independent molecular techniques (Laybourn-Parry and Pearce, 2007;Laybourn-Parry and Wadham, 2014). For instance, microbial community compositions have been investigated in (1) ice core samples from Lake Vida by clone library, and denaturing gradient gel electrophoresis (DGGE) cultureindependent approaches targeting the 16S rRNA gene (Mosier et al., 2007); (2) water column samples from Lake Hoare by culturing chemoorganotrophic isolates on media (Clocksin et al., 2007) and molecular probe of SYBR gold (Thurman et al., 2012); (3) water samples from Lake Bonney by restriction fragment length polymorphism (RFLP) analysis (Glatz et al., 2006); and (4) benthic microbial mat samples from Lake Joyce by microscopic analysis . Besides microbial mats in marine and other tropical ecosystems (Mobberley et al., 2012;Louyakis et al., 2017), NextGen sequencing technology has been applied to other ecosystems in Antarctica including lakes, sediments, soils, endoliths, and hypoliths. Examples include: (1) benthic mats from Lake Fryxell using Illumina Miseq platform , (2) water samples from Ace Lake using pyrosequencing (Lauro et al., 2011), (3) water samples from Lake Fryxell and Bonney using pyrosequencing (Vick-Majors et al., 2014), (4) soils, endoliths and hypoliths of Victoria Valley in Eastern Antarctica using pyrosequencing (Van Goethem et al., 2016), (5) surface sediments of King George Island and Bransfield Strait of Northwestern Antarctic Peninsula by pyrosequencing (Franco et al., 2017), (6) sediments of freshwater inland lakes and estuarine environment of Byers Peninsula in Antarctica (Gugliandolo et al., 2016), and (7) ice and freshwater water samples from lakes in Schirmacher Oasis of East Antarctica (Mojib et al., 2009).
In this study, we report the taxonomic composition of the microbial community and their predicted metabolic functions of the top (U1), middle (U2), and inner (U3) laminae of a large conical mat from Lake Untersee. We have used NextGen sequencing technology and bioinformatics tools on the community DNA targeting the 16S rRNA gene. The objective of this study was to describe microbial taxa in each of the top three laminae and use a taxonomy-based bioinformatics tool with the appropriate validation method to predict the likely metabolic processes performed by distinct and concentrically distributed microbial communities. We believe the results from this study will help understand the mechanisms for the formation of these unique laminated conical mats in the ecosystem of Lake Untersee.

Description of Lake Untersee and Sample Collection
Located in the Gruber Mountains, Lake Untersee is 563 m above sea level, 6.5 km long and 2.5 km wide, with an area of 11.4 km 2 and dammed at its North end by the terminus of the Anuchin Glacier. The lake floor is mostly covered with photosynthetic benthic microbial mats with a conspicuous distribution of vertically rising pigmented large conical mats. Microbial pigments are concentrated near the tops of cones, which appear purple/pink color due to phycoerythrin . These mats are soft, composed of organicrich alternating clear and mineral-rich opaque ∼0.5 mm-thick laminations with a high inheritance of microbes, showing no evidence of lithification .
A single core of a conical mat from Lake Untersee (Figures 1A,B) was collected by scientific divers utilizing SCUBA by accessing the lake ecosystem through a hole made on the ∼3.5 m surface ice using methods described by Andersen (2007). The core was collected by gently inserting a 50 mm diameter sterile polycarbonate core tube through the top of a conical mat structure, then sealing it with rubber stoppers and returning it to the surface. The core was stored in Antarctic Logistics Centre International (ALCI), Cape Town, South Africa facility first at −20 • C in a walk-in freezer and then transported to UAB in dry ice and kept in −20 • C freezer until used. The top three laminae (herein U1, top lamina; U2, middle lamina; and U3, inner lamina) were separated from the core and selected individually (Figures 1C,D) for sequencing. Each lamina was carefully separated using sterile forceps and rinsed with distilled water to avoid carryover of microorganisms between laminae before DNA extraction.

Purification of Community DNA
Five samples (∼1 cm 2 each) from different segments of each lamina of the large conical mat were sliced with a sterile scalpel, transferred into separate microcentrifuge tubes consisting of beads (MO BIO Laboratories Inc., Carlsbad, CA, United States), and homogenized in a high velocity bead beater (BIO101/Savant FastPrep FP120 -Qiagen, Inc., Valencia, CA, United States). DNA was then purified using the PowerBiofilm R DNA isolation Kit (MO BIO Laboratories) (Abbasian et al., 2015). The large conical mats used in this study are unique, found only in Lake Untersee. Therefore, to minimize sample collection while obtaining sufficient DNA to investigate the microbial composition of the mat laminae, we used five spatially representative samples from each lamina, and pooled the purified DNA into single samples for amplicon sequencing. The concentration and the quality of the pooled DNA from each mat lamina was determined by using a Lambda II spectrophotometer (Perkin Elmer, Norwalk, Conn.) followed by agarose gel electrophoresis (1% wt/vol agarose in 1X Tris-Acetate-EDTA (TAE) buffer, pH 7.8) to confirm that each sample consists of mostly high molecular weight DNA (>2 kbp) (Ausubel et al., 1987). The DNA was then dried in a Savant Speedvac Evaporator SVC 100H and stored at 4 • C until used for sequencing.

Sequence Processing Using Bioinformatics Tools
All total of 15,999 raw barcoded pyrosequence reads from the bacterial and cyanobacterial 16S rRNA genes from U1, U2, and U3 laminae were merged into fna, qual, and mapping file formats (Table 1). Then, the split_libraries.py command was used to split the sequence libraries based upon the barcodes specified in the mapping file by using the Quantitative Insights into Microbial Ecology (QIIME, v1.8.0) pipeline (Caporaso et al., 2010b). Chimera sequence reads were removed using identify_chimeric_seqs.py module of USEARCH (Edgar, 2010). The poor-quality sequence reads that did not match sufficiently with the primer sequences, barcode sequences, or contained any ambiguous reads, or homopolymers of more than 8 bp, or had average quality scores of less than 28 were excluded from the rest of the analysis (Kunin et al., 2010) resulting into a total of 14,137 sequence reads ( Table 1). The quality of sequence reads in all samples were evaluated by PRINSEQ (Schmieder and Edwards, 2011). The result showed a normally distributed sequence length (mean sequence length value of 380.78 ± 72.60 bp) and GC content (mean value of 54.55 ± 2.24%). The majority of the sequences had a mean Phred quality score of 32-33, and no occurrence of ambiguous nucleotides (N) (Schmieder and Edwards, 2011). After the trimming processes, reads were clustered into a total of 1,630 unique Operational Taxonomic Units (OTUs) at a 3% distance using UCLUST (Table 1) (Edgar, 2010). Then, representative sequences were selected from each OTU and aligned using PyNAST (Caporaso et al., 2010a). The Ribosomal Database Program (RDP) classifier 2 , trained using the Greengenes (v13.8) database 3 (McDonald et al., 2012), was used to assign taxonomy for each representative sequence at a confidence threshold of 80% (0.8) (Wang et al., 2007). A single OTU table consisting of all OTUs, taxonomic and abundance information was used to generate stacked column bar graphs of relative abundance using Microsoft Excel software (Microsoft, Seattle, WA, United States). In order to normalize sequence reads across all samples, the "single_rarefaction.py" command in QIIME was used, which resulted in an even sampling depth of 4,460 reads per sample. The subsampled sequence reads were then used for downstream Alpha [observed OTUs, Shannon diversity index (Shannon et al., 1964), and Simpson diversity index (Simpson, 1949)], as well as Beta (principle coordinate analysis (PCoA) plot) diversity analyses. The real_edge table and real_node_table files generated from QIIME (v1.8.0) were exported into Cytoscape (v2.8.2) (Shannon et al., 2003) and an edge-weighted force-directed eweight layout was selected to generate OTU network (Pope et al., 2012). Top 25 most highly abundant taxa at the family level were selected, and then used to generate the heatmap using "heatmap.2" function in R package 4 .

Oligotyping Analysis
Oligotyping (v1.0) bioinformatics tools was used to assign an oligotype identity from similarly clustered sequence reads based on their nucleotide variation (Eren et al., 2013(Eren et al., , 2014Schmidt et al., 2014). In this study, family Cytophagaceae, Opitutaceae, and Xanthomonadaceae, and genus Phormidium and Leptolyngbya sequence reads were extracted separately from the "seq.fna" file generated through the QIIME pipeline (v1.8.0), and the five extracted sequence reads were used separately to calculate Shannon entropy (Shannon et al., 1964) using "entropyanalysis" command embedded in oligotyping (v1.0). Since the number of sequences from the five extracted sequence files were below 5,000 per reads, a value of the minimum count of the most abundant unique sequence in an oligotype (M) was set up as 5 for Opitutaceae, 3 for Phormidium, 2 for Cytophagaceae and Leptolyngbya, 1 for Xanthomonadaceae, and a value of the minimum actual abundance of an oligotype across all layer samples (A) was set up as 10 for Opitutaceae or 0 for Cytophagaceae, Xanthomonadaceae, Phormidium, and Leptolyngbya to conduct oligotyping (v1.0) (Menke et al., 2014). Oligotypes that did not meet these criteria were removed, and ≥80% reads were retained for each oligotyping results. After completing the analysis, each oligotype representative sequence was subjected to BLAST search in NCBI non-redundant (nr) database 5 .

PICRUSt Analysis for Predicted Metabolic Functions
Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt v1.0.0 6 ) (Langille et al., 2013) analysis was conducted to determine the predicted metabolic functions of the microbial communities in three large conical mat laminae. Cyanobacteria and heterotrophic bacterial sequences were separated from the seq.fna file and each was used for the PICRUSt analysis. As suggested in the PICRUSt (v1.0.0) protocol, OTUs were closed-referenced picked against the Greengenes (v13.5) database at a 97% identity from each sequence file. The  (Kanehisa and Goto, 2000;Kanehisa et al., 2014), using the "predict_metagenomes.py" command in PICRUSt (v1.0.0). The likely accuracy of the PICRUSt predictions was estimated using the Nearest Sequenced Taxon Index (NSTI) value, which is the average branch length separating OTUs in each sample from the reference genome (Langille et al., 2013). The NSTI value is considered to be the standard method for validation of the PICRUSt-predicted KEGG functional groups (Langille et al., 2013;Staley et al., 2014;Zeng et al., 2015;Hu et al., 2016;Lokesh and Kiron, 2016;Lopes et al., 2016;Yuan et al., 2016). Low NSTI values imply a close relationship to organisms in the known microbial reference genome databases, representing high accuracy of the predicted KEGG functional groups. A pairwise statistical comparison of the predicted metabolic functions in the three laminae (U1 vs. U2; U1 vs. U3; and U2 vs. U3) was carried out using STAMP (Parks et al., 2014), two sided G-test (w/Yates') + Fisher's statistical test with the DP: asymptotic-CC confidence interval method with the Benjamini-Hochberg FDR multiple test correction using a P-value of <0.05 (Mason et al., 2012;Li et al., 2014).

Taxonomic Distribution in Top Three Laminae
The relative abundances of microbial taxa identified to the most resolvable level (up to family or genus) in each of the three laminae are elaborated in Figure 2. The U1 lamina represented by the highest relative abundances of members from phylum cyanobacteria (91.2%). Within cyanobacteria, Phormidium (65.8%) was found to be a dominant taxon followed by Leptolyngbya (11.1%), and Pseudanabaena (1.8%) (Figure 2). Pyrosequencing using the bacteria-specific primers on the U1 lamina produced a negligible number of poor quality sequence reads (NCBI accession number: SRP104174). Additionally, the cyanobacteria-specific primers used on the same U1 sample 7 http://www.genome.jp/kegg/ko.html showed a low occurrence of sequences belonging to heterotrophs (<1%). The U2 and U3 laminae were dominated by heterotrophic bacterial taxa. Within the heterotrophs, family C111 and Microbacteriaceae, Opitutaceae, Cytophagaceae, and Xanthomonadaceae were found to be the highly abundant taxa in U2 and U3 laminae (Figure 2). At a higher taxa resolution (up to genus level) from within these families, Leifsonia, Iamia, Rhodobacter, and Opitutus were detected in both U2 and U3 laminae (Figure 2).

Differentiation of Taxa Using Oligotyping Method
To improve the resolution of the taxa up to species level, sequences extracted from Phormidium, Leptolyngbya, Cytophagaceae, Opitutaceae, and Xanthomonadaceae from the three laminae were subjected to oligotyping followed by BLAST analyses (Figure 3 and Supplementary Table 2). Total sequences, the number of oligotypes, and the number of shared oligotype for these taxa were elaborated in Table 2. Family C111 and Microbacteriaceae were excluded from oligotyping analysis due to the high variation among each of the representative sequences, as well as limited genera information available FIGURE 2 | Stacked column bar graph representing the relative abundance and distribution of the most abundant taxa up to genus level across the three laminae in this study. The figure shows that while phylum Cyanobacteria was mostly dominant in U1, the relative abundance of Cyanobacteria decreased through U2 and U3. At the highest resolution, Phormidium was found to be the most abundant taxa within the Cyanobacteria in the all laminae. The U3 had a higher abundance of Microbacteriaceae, Opitutaceae, Xanthomonadaceae, and Cytophagaceae as compared to U2, except for C111, the abundance of which was found to be decreased in U3. Sequence data analyses was performed using QIIME (v1.8.0), and the stacked column bar graph was generated using Microsoft Excel software (Microsoft, Seattle, WA, United States).
for these families within 16S rRNA reference taxonomic databases (e.g., Greengenes). The representative oligotypes, their diversity and relative abundances in U1, U2, and U3 laminae are listed in Figure 3, and in Supplementary Table 2. BLAST results of the representative oligotypes from the Phormidium taxon showed that Phormidium cf. uncinatum CYN108 and Phormidium sp. Ant-Orange were found to be the most highly abundant taxa in three laminae ( Figure 3A and Supplementary Table 2). Phormidium autumnale, Phormidium cf. uncinatum CAWBG523, and Phormidium spp. were also found in each of the three laminae ( Figure 3A and Supplementary Table 2). Leptolyngbya antarctica ANT and Leptolyngbya sp. were found to be the most abundant taxa in U1, U2, and U3 laminae ( Figure 3B and Supplementary Table 2). Uncultured Bacteroidetes, Cytophaga sp., and Flexibacter aggregans were found to be highly abundant in U2 and U3 laminae ( Figure 3C and Supplementary Table 2). There were no sequences assigned to Cytophagaceae in U1 lamina ( Figure 3C). All oligotypes identified within Opitutaceae in U1, U2, and U3 laminae closely matched with genus Opitutus spp. (Figure 3D and Supplementary Table 2). In addition, Pseudoxanthomonas sp., Pseudomonas sp., Xanthomonas sp., and Stenotrophomonas sp. belonging to Family Xanthomonadaceae were highly dominant in U2 and U3 laminae (Figure 3E and Supplementary Table 2). The identity and the e-value of each taxa resulting from the BLAST analysis have been listed in Supplementary  Table 2.

Alpha and Beta Diversity Estimation of the Microbial Communities
A heatmap was constructed to represent the microbial abundances and compositions of top 25 taxa at family level in U1, U2, and U3 laminae (Figure 4). The relative abundance for each bacterial taxon was shown by color intensity, as depicted by the legend elaborated in Figure 4.
The Shannon diversity index (Shannon et al., 1964) and Simpson diversity index (Simpson, 1949) showed microbial diversity in each of the three laminae ( Table 1). The weighted UniFrac generated OTU network ( Figure 5A) and PCoA plots ( Figure 5B) show distribution of the observed microbial taxa in U1, U2, and U3 laminae. The results showed that unique cyanobacterial OTUs clustered in the U1 lamina and a large number of common heterotrophic bacterial OTUs in U2 and U3 laminae (Figure 5A), which is also supported by the PCoA plot ( Figure 5B). The rarefaction curves of sequences from all three laminae were saturated with respect to the species richness and diversity (Supplementary Figure 2).

Predicted Metabolic Functions Using PICRUSt
Predicted metabolic functions of the microbial communities in U1, U2, and U3 laminae were investigated by using PICRUSt (v 1.0.0) (Figure 6). The KEGG level 2 results showed that FIGURE 3 | Oligotype profiles describing the distribution and abundances of each dominant microbial taxon in three laminae (U1, U2, and U3). Color bars within each (but not between) graph represent the oligotypes within (A) Phormidium, (B) Leptolyngbya, (C) Cytophagaceae, (D) Opitutaceae, and (E) Xanthomonadaceae. The gray bars above each color graph represent the abundance of sequence reads that were assigned to different bacterial taxa. Oligotyping analyses were performed using the open-source pipeline for oligotyping, available at http://oligotyping.org. Note that no Cytophagaceae and Xanthomonadaceae sequences were identified in U1 lamina, as shown in graphs (C,E). the U1 lamina had a high number of sequences assigned to energy metabolism such as carbon fixation in photosynthetic organisms; oxidative phosphorylation; photosynthesis; and methane, nitrogen, and sulfur metabolism. The U2 and U3 laminae depicted functional categories such as amino acid and carbohydrate metabolisms, and membrane transport proteins (Figure 6 and Supplementary  Figure 3 and Table 4). The functional categories collectively for the heterotrophic bacteria at KEGG level 3 showed a high number of sequences in transporters, ABC transporters, DNA repair and recombination FIGURE 4 | Heatmap generated by using the "heatmap.2" function in R package to show the microbial compositions of the top 25 taxa at family level in U1, U2, and U3 of a conical microbial mat from Lake Untersee, Antarctica.
The rows indicate the bacterial taxa and the columns represent the three mat laminae used in this study.
proteins, purine metabolism, phosphotransferase system (PTS), and a high number of sequences for ribosome and two-component systems (Supplementary Figure 3 and Table 4). Detailed PICRUSt results of additional metabolic functions at KEGG levels 2 and 3 have been elaborated in Supplementary  Figure 3 and Tables 3, 4. The NSTI value was determined for each lamina (0.10 in U1; 0.13 in U2; and 0.11 in U3) and the mean value calculated to be 0.11 ± 0.017 s.d. for all laminae. These values are either comparable or better than the mean NSTI value ranging from 0.10 ± 0.016 s.d. to 0.23 ± 0.02 s.d. reported for microbial communities of samples collected from diverse environments (Langille et al., 2013;Staley et al., 2014;Zeng et al., 2015;Hu et al., 2016;Lokesh and Kiron, 2016;Lopes et al., 2016;Yuan et al., 2016). These values indicate that the predicted metabolic functions displayed by the observed microbial taxa in each of the three laminae are close to the known microbial reference genome databases, and thus imply a higher accuracy of the predictions. Predictive microbial functions along with the microbial taxonomic classification of the microbial communities of U1, U2, and U3 laminae of the conical mat have been summarized in Figure 7.

DISCUSSION
The mats in Antarctic ice-covered lakes are comprised of complex microbial communities and are composed primarily of cyanobacteria and heterotrophs Mackey et al., 2015;Jungblut et al., 2016). The large conical mats of Lake Untersee form distinct laminae with each layer being ∼0.5 mm in thickness, and a top lamina dominated by distinctive pigmented photosynthetic cyanobacteria . Since these laminae are easily distinguishable, we have explored the taxonomic composition and predicted metabolic functions of microbial communities of the top three laminae (U1, U2, and U3) by using NextGen sequencing technology and bioinformatics tools. The results showed that cyanobacteria dominated the U1 lamina (Figures 2, 4, 5). More specifically, Phormidium sp. was the most prevalent cyanobacteria, followed by Leptolyngbya sp. and Pseudanabaena sp. in these laminae (Figures 2, 3). These cyanobacterial taxa have previously been reported to occur in other benthic microbial mats, including the low irradiance environments of perennially ice-covered Antarctic lakes (Oguni and Takahashi, 1989;Taton et al., 2003;Jungblut et al., 2010;Andersen et al., 2011;Vincent and Quesada, 2012;Zhang et al., 2015). Hawes et al. (2013) demonstrated the composition of cyanobacterial communities of the microbial mats in areas recently flooded due to rising lake levels at Lake Vanda in the McMurdo Dry Valleys. They showed that the upper 2-4 laminae were dominated by orange-brown pigmented myxoxanthophyll intertwined with Leptolyngbya trichomes. Up to six laminae below the top layer were inhabited by phycobilin-rich green/pink-pigmented Leptolyngbya, Oscillatoria and Phormidium morphotypes. In a separate study, Sutherland and Hawes (2009) explored whether the changes in the cyanobacterial community composition in annual growth layers could potentially indicate proxies of past growth performances of microbial mats in various depths of Lake Hoare of the McMurdo Dry Valleys. The results from this study showed that there was no apparent effect of the species composition of cyanobacterial communities along with organic matter and extracellular polysaccharides in mats from all depths. Hawes et al. (2016) described the development of microbial mats in response to variations in irradiance in Lake Hoare of McMurdo Dry Valleys. This study revealed that photosynthetic biomass and species composition in mats remain remarkably stable for long period. The accrual of laminae and their thicknesses were directly correlated with the intensity of irradiance passing through the ice cover of this lake. The results from these studies corroborated our observation of the overall distribution of high abundances of Leptolyngbya and Phormidium in the U1 lamina of Lake Untersee large conical mats.
Further oligotyping analysis within Phormidium sp. showed three dominant species, P. autumnale, P. cf. uncinatum and P. Ant-Orange in U1, U2, and U3 laminae (Figure 3A and Supplementary Table 2). Particularly, the heightened observance of P. autumnale in the U1 lamina in this study corroborated previously reported microscopic and other biochemical analyses methods . P. autumnale is a gliding photosynthetic cyanobacterium and is known to manifest horizontal to upward gliding movement along mucous-like extracellular polymeric substances (EPS), which is driven by changes in the microscale gradient of nutrients in their surrounding environment (Geurts, 1976;Freytet and Plet, 1996). Thus, the relatively high abundance of Phormidium, particularly P. autumnale in our study indicates perhaps their crucial role FIGURE 5 | (A) Operational Taxonomic Units (OTU) network generated through QIIME (v1.8.0) and Cytoscape (v2.8.2). The green and red edges from the peripheral white nodes are OTUs representing cyanobacteria and heterotrophic bacteria, respectively. The Edge-Weighted Spring-Embedded Eweights format was used to generate the OTU network map; (B) two PCoA plots generated by QIIME (v1.8.0) representing significant differences in bacterial community cluster among U1, U2, and U3 of the conical microbial mat from Lake Untersee, Antarctica.
in the formation of the vertically oriented, uniquely structured large conical mats in Lake Untersee. However, such functions need to be verified by future experiments, to demonstrate gliding behavior, growth rate and gene regulation (Hawes et al., 2013).
Oligotyping analysis within Leptolyngbya sp. revealed two abundant species, L. antarctica ANT and L. sp. CENA538 ( Figure 3B and Supplementary Table 2). Specifically, the dominant presence of L. antarctica ANT across all three laminae supports previous reports of this cyanobacterium occurring in the majority of benthic microbial communities of Antarctic lakes (Komárek, 2007;Zhang et al., 2015;Jungblut et al., 2016). The straight and irregularly coiled L. antarctica has been reported to form massively large mats particularly in the benthic environments of continually frozen lakes (Komárek, 2007). Their dominance in Lake Untersee's limited photosynthetically active radiation (PAR) and UV-B environment could be attributed to the previously reported study where exposing Leptolyngbyadominated microbial mats to increased UV-B relative PAR (400-700 nm) resulted in both significantly lower chlorophyll a concentrations and higher photochemical inhibition in the mats (George et al., 2001).
Lastly, the oligotyping analysis of the third most abundant cyanobacteria in our study, Pseudanabaena sp., revealed two species: Pseudanabaena catenata UAM 412, and Pseudanabaena minima CHAB705. Our observation corroborated previously reported studies of their widespread distribution in Antarctic aquatic ecosystems, but to a lesser abundance as compared to Phormidium and Leptolyngbya (Taton et al., 2006;Hawes et al., 2013).
Given the prominent abundance of the three aforementioned cyanobacteria in the U1 lamina, one can associate the notable observances of PICRUSt-predicted functional categories to these taxa as photosynthetic organisms, namely energy metabolism, cyanophycin, chlorophyll, photosynthesis protein, photosynthetic-antenna protein, and carbon fixation (Figures 6,  7, Supplementary Figure 3, and Tables 3, 4). Furthermore, our PICRUSt results support previous reports that have demonstrated that cyanobacteria are involved in photosynthetic activities (Couradeau et al., 2012) and cyanophycin production, as matbuilders in all microbialites (White et al., 2015).
As reported in other studies, the heightened abundance of Actinobacteria in Lake Untersee microbial mats could imply a possible role in recycling organic components from decomposed cyanobacteria from the top (U1) to the inner laminae (U2 and then U3) (Goodfellow and Williams, 1983;Glöckner et al., 2000;Rondon et al., 2000;Zwart et al., 2002;Cottrell et al., 2005;Warnecke et al., 2005;Ward and Bora, 2006;Ventura et al., 2007;Philosof et al., 2009). The sequence data of U1 lamina using the bacteria-specific primers revealed a negligible abundance of heterotrophs. Subsequent bioinformatics analyses were unable to process the sequence reads. The raw sequences reads for the U1 lamina representing bacterial taxa are available at NCBI under accession number SRP104174.
The oligotyping analysis within Verrucomicrobia showed the diverse psychrotolerant, obligatory alkaliphilic, polysaccharideutilizing ultramicrobacterium Opitutus spp., including Opitutus terrae, of family Opitutaceae (Figure 3D and Supplementary Table 2) (Wilhelm et al., 2011;Larose et al., 2013). Considering their preference to anaerobic habitats, and anaerobic nitrogen fixation metabolism reported in previous studies (Wilhelm et al., 2011;Larose et al., 2013), the observation of Opitutaceae in the inner laminae led us to predict an oxygendepleted microenvironment within the large conical mats of Lake Untersee. However, the existence of such a microenvironment and the nitrogen fixation properties of Opitutus terrae in the inner laminae requires further investigation.
Previous reports have suggested the ability of Cytophaga to enzymatically lyse cyanobacteria and Gram-positive bacteria, in addition to performing photooxidation through respiration and promoting nitrogen fixation by lowering the oxygen concentrations in the surrounding environment (Marshall, 1989;Madigan et al., 2008). Thus, the existence of Cytophagaceae in Lake Untersee mats in U2 and U3 laminae could demonstrate a possible mutually beneficial relationship with cyanobacteria, but such function needs to be verified by future studies.
PICRUSt analysis of the observed heterotrophic bacterial taxa showed increased predicted metabolic functions in U2 and U3 lamina, which have previously been reported to be common properties of heterotrophic bacterial communities (Figures 6, 7, Supplementary Figure 1, and Tables 3, 4) (Newton et al., 2011;Varin et al., 2012). Previous reports showed that Actinobacteria and Proteobacteria are especially capable of conducting carbon metabolism, membrane transport system including zinc transport, and the stress response regulatory system (Vikram et al., 2016), which supports our PICRUSt FIGURE 7 | Schematic diagram of U1, U2 and U3 laminae of a large conical mat from Lake Untersee, Antarctica. This diagram shows (A) predicted metabolic functions (PICRUSt) elucidating the metabolism of the carbon, sulfur, and nitrogen in the conical mat; and (B) microbial taxonomic classification of the microbial communities in the conical mat was analyzed by both QIIME and Oligotyping. The U1 lamina was dominated by cyanobacteria, which contributed to high energy metabolism. The U2 and U3 laminae were dominated by heterotrophic bacteria, which were involved in various metabolic activities, particularly membrane transport, amino acid and carbohydrate metabolism. The mean Nearest Sequenced Taxon Index (NSTI) value for all three laminae was 0.11 ± 0.017 s.d.
In summary, the large conical mats of Lake Untersee have thus far been considered unique amongst all reported perennially ice-covered lakes in the Antarctic continent . These mats offer distinct laminae harboring diverse taxonomic compositions. The distribution of microbial communities showed high abundance of photosynthetic cyanobacteria in the outer U1 lamina. A high abundance of P. autumnale with gliding properties along the EPS-enriched U1 lamina indicate this species as a likely contributor for the vertical structure of these large conical mats. Predicted metabolic functions of the phototrophic cyanobacterial community in the U1 lamina showed this community capable of producing nutrients by photosynthesis. The U2 and U3 laminae consisted of a high abundance of heterotrophic bacterial taxa, of which their predicted metabolic functions determined by PICRUSt are likely involved in the decomposition of organics and metabolites. Further shotgun metagenomics analysis would substantiate detailed metabolic traits in each of these microbial communities, which could serve as an ideal model system to better understand the microbial processes that possibly contribute to the formation of these structures analogous to the ancient stromatolites.

AUTHOR CONTRIBUTIONS
HK conducted the bioinformatics analysis of the NGS data. DA, YT, and IH conducted field work in Antarctica including the collection and processing of a conical stromatolite mat sample from Lake Untersee. NM conducted DNA extraction and pursued the DNA samples to Research and Testing Laboratory (Lubbock, TX, United States) for Pyrosequencing. AB and DA conceived the study. HK, JH, and AB wrote the manuscript and all authors contributed and commented wherever appropriate in the manuscript. AB oversaw the overall progress of this study.

FUNDING
Primary support for this research was provided by the Tawani Foundation of Chicago (6803099), the Trottier Family Foundation, NASA's Exobiology and Astrobiology Programs (NNX08AO19G), and the Arctic and Antarctic Research Institute/Russian Antarctic Expedition.