The techniques applied in this work extend previous characterisations of microbial life from the Bratina meltwater pond landscape20,21,25,26,27, which mostly focused on the dominant bacterial diversity of the microbial mat communities. To characterise the eukaryotic communities living in the supraglacial meltwater pond landscape, sterol biomarkers were extracted, analysed, and converted to their fossil analogues via simulated diagenesis. This approach generates not only a would-be fossil assemblage for a type of organosedimentary system unlikely to be preserved in the fossil record12, but also a point-of-view from which low-abundance and/or potentially co-eluting sterols can be more readily detected and interpreted. Significant eukaryotic diversity remains under characterised in this Cryogenian analogue environment via modern genetic sequencing, underscoring the continued utility of steroid biomarker analyses of potential eukaryotic refugia.

18S rRNA gene eukaryotic communities in meltwater pond microbial mats

The eukaryotic 18S rRNA gene sequences analysed from the meltwater pond and Bratina Lagoon microbial mats were broadly classified into the major eukaryotic groups Amoebozoa, Archaeplastida, Centrohelida, Cryptophyceae, Excavata, Haptophyta, Opisthokonta, SAR, incertae sedis, and uncategorisable Eukaryota. Opisthokonta, along with SAR and Archaeplastida, comprised the greatest relative abundances of categorizable Eukaryota across the meltwater ponds. The high proportions of Opisthokonta detected likely reflect the presence of meiofauna and microfauna in the microbial mats, including nematodes, tardigrades, and rotifers. Due to the size difference between these meiofauna, microfauna, and protist and fungi cells, it is possible that the metazoa may bias the 18S rRNA gene survey results. It is for this reason that metazoa are commonly excluded from 18S rRNA gene surveys in polar environments20. The prevalence of 18S rRNA gene sequences which were not categorizable past Eukaryota and the broadest classification levels may be due to the use of short sequence reads, as well as the limited representation of polar microorganisms in reference databases22. It is also possible that these ponds may support cryptic or dark eukaryotic diversity that has yet evaded growing annotated environmental databases, as is currently the case with diverse, polyphyletic groups such as algae28 and SAR29.

The eukaryotic compositions of the meltwater ponds and lagoon varied greatly despite the close proximity and physicochemistries of the individual sites, a finding similarly observed during a previous 18S rRNA gene analysis of a small subset of these ponds19. The implication of this variation is that there is no archetypal microbial mat eukaryotic composition in these environments, though it is generally possible to expect the consistent presence of SAR, Opisthokonta, and Archaeplastida as the most abundant categorizable constituents. For example, both ponds Conophyton and Salt contain these three eukaryotic groups, but Conophyton Pond is dominated by Archaeplastida while Salt Pond is SAR-dominated (Fig. 6).

Potential controls, including environmental variables, have not been previously shown to majorly impact eukaryotic community composition20. However, the graphical representation of the PCA conducted in this work (Fig. 7b) revealed that salinity may play a more prominent role than previously determined. Ponds with more extreme principal component scores tended to have relatively low conductivities, while ponds closer to the origin, whose PC scores did not appear to be significantly influenced by the variables examined, were measured to have higher conductivities except for Castenholz. This potential subtle salinity control is more apparent in the sterane ternary (Fig. 4) and sterane PCA analysis (Fig. 7a) and is discussed further below.

Sterols

Sterol assemblages detected across the mats, although varied in their individual components and their distributions, broadly correspond to those produced by microalgae, protists, and potentially, metazoan meiofauna (Table 2).

Previous studies of Antarctic diatoms have recovered sterol distributions which closely overlap with those observed within the meltwater pond microbial mats, the components of which include 22-trans-24-norcholesta-5,22E-dien-3β-ol, cholesta-5,22E-dien-3β-ol, cholest-5-en-3β-ol, 24-methylcholesta-5,22E-dien-3β-ol, 24-methylcholesta-5,24(28)-dien-3β-ol, 24-methylcholest-5-en-3β-ol, 24-ethylcholesta-5,22E-dien-3β-ol, 24-ethylcholest-5-en-3β-ol, and 4α,23,24-trimethylcholest-22E-en-3β-ol30,31. Many of these compounds have been detected in other protists, such as dinoflagellates, and microalgae, including marine and freshwater eustigmatophytes, haptophytes, cryptophytes, chrysophytes, and prasinophytes23,32,33 Rampen et al.’s34 detailed classification of sterols from a variety of diatoms demonstrated that the most common sterols in this group were comprised of 24-methylcholesta-5,24(28)-dien-3β-ol, cholest-5-en-3β-ol, 24-methylcholest-5-en-3β-ol, and 24-ethylcholest-5-en-3β-ol, all of which are present in the majority of the microbial mats examined in this work34. Some sterols detected, such as 5α-cholest-22E-en-3β-ol and 27-nor-(24S)-methylcholesta-5,22E-dien-3β-ol, may derive from dinoflagellates35,36,37, while others, including the 22-trans-24-norcholesta-5,22E-dien-3β-ol detected in Brack Pond, and 24-methylcholesta-5,22E-dien-3β-ol, have been shown to be produced by both diatoms and dinoflagellates34,38,39. Certain Rhizaria, such as Cercozoa, have also been shown to produce 24-methylcholesta-5,22E-dien-3β-ol and (24S)-ethylcholesta-5,22E-dien-3β-ol32. Other sterols detected, including (24R)-methyl-5α-cholest-7-en-3β-ol, (24R)-ethyl-5α-cholestan-3β-ol and (24S)-ethyl-5α-cholest-7,22E-dien-3β-ol, have been detected in microalgae as well23,40,41. These potential overlapping sterol sources, including diatoms, dinoflagellates, and microalgae (Fig. 6) were all detected during the18S rRNA gene surveys of microbial mats from these ponds.

Certain sterols detected, including cholest-5-en-3β-ol, are commonly associated with animals, though they are widely produced by algae and other organisms24. It is likely that within the microbial mats, both algae and meiofauna contribute to the cholest-5-en-3β-ol abundances demonstrated, which is consistent with the 18S rRNA recovery of sequences from both groups. Seafloor metazoa and meiofauna may also contribute to the steroid assemblages as their remnants are distributed across the undulating ice shelf surface and line the base of each meltwater pond, and their steroids preserve for longer time periods in the environment than DNA (Fig. 2).

A mixture of 5α and 5β stanols were recovered in the microbial mats, possibly reflecting the natural reduction of sterols in a reducing setting42 as opposed to fecal inputs that might be expected in settings nearer to sites of human habitation24. Aged microbial mats, including the mat collected from the Skua Pond margin and the relict microbial mat, contained measurable 5β-cholestan-3β-ol, while microbial mats sampled from pond waters did not, suggesting that this molecule may be generated from the hydrogenation and epimerization of cholest-5-en-3β-ol, processes active particularly in dried, or decaying mats. A similar natural diagenetic process may have led to the production of the (24R)-ethyl-5β-cholestan-3β-ol detected in the Skua Pond margin microbial mat, which is derived from the reduction and subsequent epimerization of (24S)-ethylcholesta-5,22E-dien-3β-ol and (24R)-ethylcholest-5-en-3β-ol.

Steranes

The most abundant steranes generated by the simulated diagenesis of the microbial mat sterols corresponded to the C 27 , C 28 , and C 29 precursor sterols discussed in Sterols. For example, during catalytic hydrogenation/hydrogenolysis, the C 29 sterols (24S)-ethylcholesta-5,22E-dien-3β-ol, 24R-ethylcholest-5-en-3β-ol, and (24R)-ethyl-5α-cholestan-3β-ol all convert to 24-ethylcholestane as their double bonds are reduced and oxygens lost; these modifications mimic common chemical transformations that occur during early diagenesis. However, small amounts of other steranes, including norsteranes and C 30 4-desmethylsteranes were detected, likely reflecting contributions from the meltwater communities themselves as well as contributions from seafloor communities which have been transported to the ice shelf surface over time (Fig. 2)12.

Norsteranes have been detected in a variety of oils and their source rocks, including the C 26 21-, 24- and 27-norcholestanes43 and the C 27 norsterane 27-nor-24-methylcholestane44. These molecules are thought to represent contributions from diatoms and dinoflagellates36,38,43,44,45,46, though some norsteranes have been hypothesised to derive from sponges47,48. Hydrogenation of the microbial mat sterols in this work resulted in the production of the C 26 norsteranes, 24-norcholestane and 27-norcholestane, while 21-norcholestane, a possible product of thermal degradation, was notably absent43. In addition, a series of compounds that eluted closely following cholestane, 24-methylcholestane, and 24-ethylcholestane were observed, likely representing small quantities of isomers of analogous 27-norsteranes for C 27 , C 28 , and C 29 steranes. To eliminate the possibility that these additional steranes were generated as a byproduct of the simulated diagenesis conducted on the microbial mats, the hydrocarbon fraction of the 700-year-old relict microbial mat was also analysed for its sterane contents. The same norsterane series evident in the hydrogenated 700-year-old microbial mat was also present in its free hydrocarbon fraction (Supplementary Fig. 13), which had not been subjected to hydrogenation after total lipid extraction. Further, hydrogenation experiments of pure sterol standards have been previously shown to not generate 27-norsteranes49, further suggesting that 27-norsterol precursors are indeed present in the microbial mats. Given the abundance of microalgal sterols detected in the microbial mats in this landscape, two of which are precursors molecules to 27-norsteranes, it is possible that these norsteranes are indicative of small microalgal or diatom inputs which are undetectable against more abundant sterols and stanols, though it is possible that some of the 27-norsterol derived from seafloor life as seafloor sediments line the bases of the meltwater ponds and are scattered across the undulating ice. These findings are in line with Moldowan et al.’s43 inference that unlike the primary cholestanes, 24-methylcholestane and 24-ethylcholestane, which result from a wide variety of precursor molecules, norsteranes and other less common steranes may be more informative for analyses that aim to link fossil molecules with their biological precursor producers43.

In addition to the norsteranes, the C 30 4-desmethylsteranes 24-n-propylcholestane and 24-isopropylcholestane as well as 4-methylsteranes were detected in varying abundances across the mats. Precursor molecules to 24-n-propylcholestane, 24-n-propylidene-cholesterol and 24-n-propylcholesterol, have been detected in a variety of algae, including chrysophytes and pelagophytes32,50,51 Volkman et al.52 generated 24-n-propylcholestane from the hydrogenation of lipids extracted from a prasinophyte algae, though its C 30 sterol precursor was not detected during the sterol composition analysis of that organism in that study52. Since other works have shown the detection of its precursor molecules in other algae32, it is possible that the precursor sterol was present in the original algal culture examined but in quantities undetectable by conventional full scan GC-MS methods. Mild catalytic hydrogenation methods like the ones employed by Volkman et al. in 1994 and in this study have not been demonstrated to selectively methylate their steroid reactants. In addition to algal sources, foraminifera may also contribute 24-n-propylcholestane precursors53.

Sterol precursors of the hydrocarbon 24-isopropylcholestane are commonly attributed to demosponges54,55 and feasibly represent inputs from demosponges transported to the ice shelf surface. Sponge remains were ubiquitous on the undulating ice at the time of sampling and have been identified on the McMurdo Ice Shelf surface for over a century56 (Fig. 2), having been transported from the seafloor via the conveyor belt mechanism referenced by Hawes et al. in 201812, and with many species of Antarctic demosponges having been previously characterised57. Recent studies have proposed alternative origins for 24-isopropylcholestane precursors, including bacterial sponge symbionts, pelagophyte algae, and Rhizaria49,58,59. Others have suggested generation of 24-isopropylcholestane via the thermally-driven methylation of C 29 algal sterols60. Such a process is excluded at the near-freezing to below-freezing temperatures on the surface of the ice shelf and in the absence of geothermal heating. The ratios of 24-isopropylcholestane to 24-n-propylcholestane in the dataset vary by 2 orders of magnitude across the samples from which both molecules were detected after hydrogenation, and in some cases, 24-n-propylcholestane was absent where 24-isopropylcholestane was present (Table 3), potentially ruling out an algal source for 24-isopropylcholestane in the meltwater environments as both molecules would be expected from algal sources. Further, 26-methylstigmastane was not detected in any hydrogenated microbial mat lipid extract, eliminating Cercozoa from contention as producers of the 24-isopropylcholestane precursors. Cercozoa have previously been described to produce both 24-isopropylcholestane and 26-methylstigmastane49. The application of long-read sequencing for 18S rRNA gene metabarcoding, the characterisation of environmental genomes of polar aquatic communities, and even ancient DNA analyses may greatly assist the definitive source assignment of the C 30 4-desmethylsteranes in this environment, though the simultaneous presence of algae, sponge fragments, and Rhizaria in the environment may be always confounding.

Further, the clear identification of precursor molecules for these compounds in sterol fractions may ultimately aid in source-assignment. However, the complex nature of environmental samples, particularly those with such a diverse array of C 27 to C 29 steroids, makes the detection of low-abundance steroids exceedingly difficult via traditional full-scan gas chromatography-mass spectrometric methods using nonpolar GC capillary columns. It is possible that the precursor molecules for the norsteranes as well as the C 30 steranes detected, which must be present in the microbial mat sterol fractions, co-elute with other more abundant compounds or are present at levels below the instrumental detection limits during a full scan. However, once hydrogenated, they are readily detectable via dynamic multiple reaction monitoring, a technique which affords significant sensitivity and selectivity advantages compared to full scan mass spectrometry. Such is the case with the 4-methylsteranes, which were detected at measurable or trace abundances in all ponds studied via the 414 → 231 Da transition during multiple reaction monitoring (as in Fig. 5). Their precursor molecules are produced by dinoflagellates in the case of 4α,23,24-trimethylcholestanes and by prymnesiophyte algae in the case of 24-ethyl-4α-methylcholestanes54,61,62. However, these precursor molecules were similarly not detected within the sterol fractions analysed prior to hydrogenation except in the case of Skua Pond (Table 2), though a previous study of microbial mats in this environment successfully detected 4α,23,24-trimethylcholest-22E-en-3β-ol and its associated stanol26.

C 27 -C 28 -C 29 sterane ternary analysis of the steranes generated by the simulated diagenesis of the microbial mat sterols (Fig. 4) reveals a distribution largely consistent with open marine and estuarine/bay ecosystems63. This distribution of steranes synthesised from Bratina microbial mat sterols contrasts with sterane distributions of Neoproterozoic and early Cambrian marine samples from around the globe, which cluster near the C 27 or C 29 endmembers during ternary analysis47,64,65,66. However, there is noticeable similarity between the central placement of the steranes generated from meltwater pond microbial mats and the distribution of steranes recovered from Phanerozoic sediments and oils64. The dominance of C 27 or C 29 steranes observed in most of the mats aligns closely with sterane distributions previously reported for Quaternary-aged sediments collected from the western Ross Sea67, though in two ponds, Brack and Salt, the major sterane contained 28 carbon atoms. While most of the mats were collected from ponds with conductivities which correspond to freshwater or mildly saline environments, ponds Brack and Salt had conductivities of 6690 and 28,400 µS/cm, respectively (Table 1). While these values are still lower than those measured in seawaters, they may be associated with a greater abundance of C 28 sterol producers as demonstrated by sterane PCA analysis (Fig. 7). The sterane distributions align with previous research which suggested salinity as an abiotic factor influencing community composition in these environments, although that relationship was only demonstrable for prokaryotes20,21. Given the association observed between C 28 steroid abundances and conductivity, it may be possible to make inferences about the conductivities of now-dry ponds from the hydrocarbon distributions of their relict microbial mats. The decaying microbial mat from Skua Pond, which was collected from the dry pond margin and therefore may represent a precursor relict mat, plotted among the other ponds with low conductivities (Fig. 4). The conductivity of the pond from which it derived was 1920 µS/cm, a number considerably lower than those measured in the ponds with higher proportions of C 28 steranes. The 700-year-old relict microbial mat similarly plotted among mats with low conductivities, suggesting that it may have once grown within a freshwater or slightly saline pond.

Supraglacial meltwater oases as Cryogenian refugia

The detailed characterisation of the eukaryotic assemblages of supraglacial meltwater ponds is a prerequisite for their continued candidacy as possible Cryogenian eukaryotic refugia. In order to explain the ensuing expansion of eukaryotic and multicellular life in the Ediacaran as evidenced by the fossil record and molecular clock analyses8,68, Cryogenian refugia would have needed to support sufficient eukaryotic diversity; this is consistent with an emerging consensus that global glaciations did not impose an evolutionary bottleneck on eukaryotic life11,12.

The complementary application of 18S rRNA gene and sterol biomarker analyses revealed that microbial mats in supraglacial meltwater ponds are capable of supporting diverse eukaryotic assemblages. While 18S rRNA gene analysis provided information about living eukaryotic communities, sterol and sterane analysis enabled the assessment of the activities of current and former eukaryotic communities as well as of marine eukaryotic communities beneath the ice shelf whose remnants have accumulated atop the McMurdo Ice Shelf. A considerable proportion of the eukaryotic communities in the meltwater ponds was categorised as Archaeplastida and SAR through 18S rRNA gene surveys, consistent with the recovery of a diverse assortment of sterols commonly attributed to members of these groups. However, the abundance of uncategorisable eukaryotes, either due to methodological limitations or the lack of characterised reference genomes, potentially complicates analyses which attempt to examine physicochemical controls on 18S rRNA gene derived eukaryotic community compositions. Sterol and sterane assemblages can broadly describe eukaryotic communities in the absence of genetic data and revealed that eukaryotic assemblages in this environment vary with pond conductivity.