Figure 2 . Examples of blue features. Blue mats and mineral precipitates are often found on surfaces of flowing water (a) , or along cracks in cave walls or ceiling, where water seeps in from the ground above (b-d) .
2.2 Sampling strategy
Sampling sites were kept minimal in size, with great care taken to touch as little of the cave walls as possible, so as not to disturb or contaminate the microbial mats, which have a slow growth rate due to the environmental conditions and nutrient availability, taking several decades to reach macroscopic size. Speleothems, which form on geological timescales, are frequently vandalized, for example, the breaking of stalactites and stalagmites for personal collections and home decoration. Though sometimes unavoidable when sampling, such as when bumping the ceiling or walls in tight spaces, great care was taken to minimize damage of the mineral formations.
Initial training by cave connoisseur Diana Northup at a summer school in 2016 in the Azores organized by the European Astrobiology Campus, followed by several expeditions with the same core team honed the following 5-step sampling strategy (see also Figure 3).
Step 1. Trace gas analysis, performed by one person to measure CO2, SO2,NH3, H2S and NOx, which have the potential to be dangerously high in caves.
Step 2. Microbial mat sampling, by scraping the mat off the rock under sterile conditions for DNA extraction and 16S rRNA gene sequencing analysis.
Step 3. Biogeochemical sampling, by chipping away a section of mat-covered rock with a geological hammer under sterile conditions for molecular analysis.
Step 4. Recording of the elemental composition of the rock below the scraped mats using a handheld XRF spectrometer.
Step 5. Geological sampling by chipping away the surface rock with a geological hammer for mineralogical analysis.
Figure 3. Sampling strategy. (a) Cave entrance to cave H, 12 m below ground. (b) Sampling in pairs: one sampler and one assistant handling cleaning and handing over of tools. (c)Biological sampling (step 2): scraping mats from the cave wall.(d) Handheld XRF measurements of cave wall (step 4).(e) Substrate sampling with a geological hammer and Falcon tube (steps 3 and 5).
2.2.1 Biomass sampling for DNA analysis
In order to minimize contamination, the second step in the sampling technique (after step one: ensuring the team’s safety with the trace gas analyzer) was the collection of biomass for DNA extraction and sequencing (Figure 3c). Sampling began by sterilizing sample tools, i.e. scoops and spatulas with cleaning acetone and 70% ethanol, eliminating potential organic contaminants. We wore nitrile gloves washed with 70% ethanol, and used scoops and sterile Eppendorf tubes to scrape the surface of the lava tube to collect the overlying biomass. The sampling was best done with two people, one doing the sampling and one assisting with cleaning and in handing over tools and sampling tubes (Figure 3b). Samples were collected in biological triplicate to ensure enough samples for any repeated analyses. Samples were then stored at 4°C onsite and for shipment and finally at -20°C until the DNA extraction protocol began.
2.2.2. Biogeochemical sampling for molecular analysis
The third step in sampling was the collection of geological samples covered in biomass. We sterilized tools using the same acetone and 70% ethanol-based sterilization technique as for DNA sampling and used sterile 50 mL Falcon tubes to catch the biomass-covered rock samples that we chipped off with the hammer, which was flame sterilized with a lighter in addition to the solvent cleaning (Figure 3e). Collected in triplicate, these samples were also stored at 4°C to ensure minimal alteration of biomolecules.
2.2.3 Portable X-ray Fluorescence (XRF)
A portable NitonTM XL3t GOLDD+ X-Ray Fluorescence Analyzer from Thermo Fischer gave us elemental data in situ (step 4). It can detect up to 30 elements from Mg to Bi in the standard range without helium or vacuum assistance. The measured concentration of the sample must be at least three times the standard deviation of the measurement (i.e., detection limit); the measurement confidence is 95% (two sigma). ”Mining mode” (best for analyzing raw or semi-processed mineral samples of varying density) was used for elemental quantification. The instrument was placed against the cave wall, measuring both the biofilm substrate post biological sampling as well as nearby uncolonized locations with an analysis time of 90 seconds.
2.2.4 Geological sampling of the lava tube substrate
Geological sampling was the fifth and final step in the sampling protocol. We chipped off rock samples from the cave wall with a geological hammer and stored them in ziplock bags under ambient conditions. The samples were then analyzed with SEM/EDS.
2.3 Characterization of microbial samples
To obtain phylogenetic data we performed DNA extractions, PCR amplification, and genetic sequencing on the microbial samples collected in step 2 of the sampling procedure.
2.3.1 DNA extractions Genomic DNA was extracted in triplicate from the biofilm samples using a method previously used with environmental rock samples with phototrophic communities (Stivaletta et al., 2012). First a washing step was performed, wherein 200 µL (or weight equivalent ~200 mg) of the samples were resuspended in 1.5 mL of TE pH 8 (10 mM Tris Hydrochloride pH7.4 + 1mM EDTA pH 8), centrifuged at 10000 rpm for 10 min, after which the supernatant was discarded, and the pellets resuspended in 400 µL of TE pH 8. Then, the solution was added to sterile tubes with glass beads (Lysing Matrix tubes, MP Biomedicals, 1.4 mm ceramic beads, 0.1 mm silica spheres and one 0.4 mm glass sphere) and subjected to a bead beater two times 60 s at 6 m/s (MP Biomedicals, FastPrep24).  Subsequently, 300 µL of phenol saturated with 0.1 M Tris Hydrochloride (tris-phenol, pH 7.4) and 300 µL of chloroform/isoamyl alcohol (24:1) were added to the tubes, and subjected to three 2-min cycles of heating at 60°C and vortexing for 30 s. After centrifugation (10 000 rpm, 5 min) the aqueous phase was extracted once with tris–phenol/chloroform/isoamyl alcohol (25:24:1); then 1/5 volume of TE buffer was added, and the pellet extracted again with phenol. Finally, the aqueous phases were extracted with chloroform/isoamyl alcohol (24:1) and nucleic acids precipitated overnight at -20°C with cold absolute ethanol and 0.3 M sodium acetate. After centrifugation (10 000 rpm, 5 min) and washing with 1 mL cold (-20°C) 75% ethanol, the pellets were dried out and resuspended in 30 µL nuclease free water. 2.3.2 PCR Amplification, Library Preparation and Sequencing
PCR amplification was made with PuRe Taq Ready-To-Go PCR Beads (GE Healthcare), 1 µL of each Illumina barcoded forward and reverse primers (16S rRNA gene V4 region primers 515F: GTGBCAGCMGCCGCGGTAA and 805R: GACTACHVGGGTATCTAATCC), 2 µL of DNA template, and 21 µL of nuclease-free water. The PCR cycling program used was 98°C for 30 s (denaturation), 35 cycles of 10 s at 98°C, 30 s at 60°C, 4s at 72°C, and 2 min at 72°C.
The multiplexed amplicon library was prepared by pooling equal amounts (20 ng) of each sample and paired-end sequencing was performed (at the SNP&SEQ Technology Platform, SciLife Labs, Sweden) on Illumina MiSeq with 300 base pairs (bp) read length using v3 sequencing chemistry. Raw sequences were processed with the R package DADA2 (version 1.18.0) (Callahan, Mcmurdie, et al., 2016). Due to low quality reverse reads affecting the merging of paired ends, we only used forward reads in the downstream analysis. Reads were truncated to 220 bp and assigned taxonomy using the Silva_132 database (Quast et al., 2013) and the RDP naïve Bayesian classifier (Wang et al., 2007). ASVs (Amplicon Sequence Variants) were normalized by Cumulative Sum Scaling. All samples with less than 100 reads were excluded from the analysis, and biological triplicate samples were merged. Visualization was done using the R packages phyloseq (v1.34.0) (McMurdie & Holmes, 2013) and ggplot2 (Wickham, 2016), and Krona plots were made with Krona (Ondov et al., 2011).
2.4 Characterization of geochemical and geological samples
To characterize the geological substrate and identify any molecular biomarkers therein, we analyzed the geochemical samples collected in step 3 using confocal Raman microscopy and the geological samples collected in step 5 using SEM/EDS. The geochemical samples were first pulverized with a mortar and pestle under sterile conditions. The geological samples were polished into thin sections without any prior treatment and coated with platinum or carbon (when carbon content was not analyzed).
2.4.1 Confocal Raman microscopy
Raman spectra were obtained with a confocal WITec alpha 300 system, at the DLR Berlin, consisting of a microscope equipped with a 10x long working distance objective with a 0.25 numerical aperture, a piezo-driven scan table, a UHTS 300 spectrometer with an ultrafast EMCCD detector and a frequency-doubled Nd:YAG laser. The excitation wavelength of the laser was 532 nm, the spot diameter at the sample was ~2.5 µm and the spectral resolution of the spectrometer was 4-5 cm-1. Integration time and laser power were varied according to the investigated sample (1-10 s and 0.1-7 mW respectively) to produce spectra with a sufficient signal-to-noise ratio, and to prevent sample damage/degradation and detector saturation.  2.4.2 SEM/EDS analysis of thin sections Secondary electron (SE) and backscattered electron (BSE) images of platinum and carbon-coated thin sections were obtained using a Zeiss GeminiSEM 450 Field Emission Gun Scanning Electron Microscope at Utrecht University, using 10-20 keV accelerating voltage, a 250-1000 pA probe current, and a 10 mm working distance. The SEM was coupled with a windowless Oxford Instruments Ultim-Extreme EDS detector to characterize the elemental composition of the minerals present in the thin sections. Overview element distribution maps (15-60 min counting time) and point ID measurements (30 s counting time) were acquired at voltage of 20 keV and 1 nA, using Oxford Instruments Aztec software v5.1. For improved spatial chemical resolution in finely zoned domains, an accelerating voltage of 10 keV was used, acquiring the L-alpha intensity of Cu.3 ResultsTo investigate cave biosignatures, we carried out fieldwork in Icelandic lava tubes, where we collected biological, biogeochemical, and geological samples, and in situ elemental data using a multi-step sampling protocol. We then characterized the samples using a variety of analytical techniques. Here we describe the results in sections based on the location of the collected samples, the type of sample, and the analytical technique used. 3.1 Surface & subsurface field observations 3.1.1 Sampling sites
The caves sampled include caves B and C in the southern Eldhraun lava field, and H and R in the northern Ódáðahraun lava field (Figure 1). The lava rock in the caves visited appears largely unaltered, and the presence of microbial mats is much scarcer than that seen by our team and others (Hathaway et al., 2014) in lava tubes in the Azores. Still, we observed alteration minerals and microbial mats of various, striking colors.
The cave entrances in the southern Eldhraun lava field are very shallow, and the entirety of the tubes is no more than two meters below the surface. The temperature was measured at 8°C, and much water condensation was observed on the cave walls, of which we measured a pH of 8.1 near the sampling sites. In cave B we observed white, red, yellow, and blue mats on top of the cave walls, while in cave C we found green, blue, red, and white mats. The cave entrance to cave H in the Ódáðahraun lava field was twelve meters below the surface, requiring four connected three-meter-long ladders, and additional propelling equipment for the team as a failsafe. The entrance of cave R was equally deep, but accessible by foot down a steep ash mound. Compared to the southern caves, the temperature in the northern caves was colder, measured at 3°C, the cave ceilings were higher, and the walls generally appeared much drier. We observed many blue and purple mats, orange and yellow mats, grey slime, black dots, and fluffy white deposits on the cave walls. While most microbial mats seemed to be randomly dispersed on the cave walls, the blue precipitates occurred only where water was present, such as along cracks in the cave walls (Figure 3). 3.1.2 Samples collected
A variety of mat colors were sampled, including white, yellow, orange, grey, purple, and blue. The blue mats were distinct from the others in their incredible vibrancy of color, and in their consistent occurrence in places of flowing water, near cracks in the cave wall or ceiling, or in puddles on the ground. Blue mat and rock samples were collected in caves B, C, and H (Figure 4). Blue features were found surrounding a crack in cave B (Figure 4a), on a water droplet on a ceiling in cave C (4b), and at two sample sites in cave H, one on the wall (4c) and one in a puddle on the ground (4d). While the puddle in Figure 4d was very wet, offering up a wealth of viscous material for biological sampling, not all of the sampling locations were so generous. The blue material on the cave wall in Figure 4c was fused to the rock and had to be scraped off with much patience.