2019 Volume 34 Issue 3 Pages 278-292
Hydrothermal systems, including terrestrial hot springs, contain diverse geochemical conditions that vary over short spatial scales due to progressive interactions between reducing hydrothermal fluids, the oxygenated atmosphere, and, in some cases, seawater. At Jinata Onsen on Shikinejima Island, Japan, an intertidal, anoxic, iron-rich hot spring mixes with the oxygenated atmosphere and seawater over short spatial scales, creating diverse chemical potentials and redox pairs over a distance of ~10 m. We characterized geochemical conditions along the outflow of Jinata Onsen as well as the microbial communities present in biofilms, mats, and mineral crusts along its traverse using 16S rRNA gene amplicon and genome-resolved shotgun metagenomic sequencing. Microbial communities significantly changed downstream as temperatures and dissolved iron concentrations decreased and dissolved oxygen increased. Biomass was more limited near the spring source than downstream, and primary productivity appeared to be fueled by the oxidation of ferrous iron and molecular hydrogen by members of Zetaproteobacteria and Aquificae. The microbial community downstream was dominated by oxygenic Cyanobacteria. Cyanobacteria are abundant and active even at ferrous iron concentrations of ~150 μM, which challenges the idea that iron toxicity limited cyanobacterial expansion in Precambrian oceans. Several novel lineages of Bacteria are also present at Jinata Onsen, including previously uncharacterized members of the phyla Chloroflexi and Calditrichaeota, positioning Jinata Onsen as a valuable site for the future characterization of these clades.
A major theme of environmental microbiology has been the enumeration of microbial groups with the capacity to exploit the diverse chemical potentials (i.e. chemical disequilibria) that occur in nature (e.g. 7, 25, 111). Hot springs, with their varied chemical compositions, provide reservoirs of novel microbial diversity; environmental and geochemical conditions in these environments select lineages and metabolic pathways that are distinct from other Earth-surface environments (e.g. 3, 9, 26, 126, 127). In addition to their value as sources of microbial diversity, hot springs are valuable test beds for understanding microbial community processes driven by different suites of metabolism (e.g. 50). This in turn allows these systems to serve as process analogs and also to provide a window into biosphere functions during early times in Earth history, for example when the O2 content of surface waters was low or non-existent. Most surface ecosystems today are fueled entirely by oxygenic photosynthesis by plants, algae, and Cyanobacteria; in contrast, hot spring microbial communities are commonly supported by lithotrophic or anoxygenic phototrophic organisms that derive energy and electrons for carbon fixation by oxidizing geologically sourced electron donors such as Fe2+, sulfide, arsenite, and molecular hydrogen (e.g. 34, 63, 69, 105, 126). These communities may therefore provide insight into the flow of energy and mass that characterized microbial communities on the early Earth or even other planets, in which oxygenic photosynthesis may be absent or less significant and anoxygenic photosynthetic or lithotrophic metabolisms may play a larger role, resulting in lower overall rates of primary productivity (e.g. 12, 63, 97, 131–133).
Here, we present a geomicrobiological characterization of a novel Precambrian Earth process analog site: Jinata Onsen on Shikinejima Island, Tokyo Prefecture, Japan. While a small number of metagenome-assembled genomes have previously been recovered from Jinata (125, 127), we describe here the first overall characterization of the geochemistry and microbial community of this site. This site supports sharp gradients in geochemistry that in some ways recapitulate spatially environmental transitions that occurred temporally during Proterozoic time. The modern, sulfate-rich, well-oxygenated ocean that we see today is a relatively recent state, typical of only the last few hundred million years (e.g. 76). For the first half of Earth history, until ~2.3 billion years ago (Ga), the atmosphere and oceans were anoxic (52), and the oceans were rich in dissolved iron but poor in sulfur (120). Following the Great Oxygenation Event ~2.3 Ga, the atmosphere and surface waters accumulated oxygen, and the oceans became more strongly redox stratified with oxygenated surface waters and anoxic deeper waters, rich in either dissolved iron or sulfide (88). At Jinata Onsen, this range of geochemical conditions is recapitulated over just a few meters, providing an ideal space-for-time analog to test hypotheses of how microbial diversity and productivity may have varied as environmental conditions changed through Earth history.
At Jinata hot spring, anoxic, iron-rich hydrothermal fluids feed a subaerial spring that flows into a small bay, and mixes with seawater over the course of a few meters. Over its course, the waters transition from low-oxygen, iron-rich conditions analogous to some aspects of early Proterozoic oceans, toward iron-poor and oxygen-rich conditions typical of modern coastal oceans. In upstream regions of the stream where oxygenic Cyanobacteria are at very low abundance, biomass is visibly sparse; however, downstream, biomass accumulates in the form of thick microbial mats containing abundant Cyanobacteria. Visible differences in accumulation and appearance of biomass across the temperature and redox gradient establish the hypothesis that microbial community composition, as well as the magnitude and metabolic drivers of primary productivity, varies along the spring flow. To begin testing this hypothesis and to provide a baseline description of the geochemistry and microbiology of this site in support of future investigation, we performed geochemical measurements, 16S rRNA gene amplicon sequencing, and genome-resolved metagenomic sequencing to recover draft genomes of diverse novel microbial lineages that inhabit Jinata Onsen.
Jinata Onsen is located at 34.318 N, 139.216 E on the island of Shikinejima, Tokyo Prefecture, Japan. Shikinejima is part of the Izu Islands, a chain of volcanic islands that formed in the last few million years along the northern edge of the Izu-Bonin-Mariana Arc (56). Shikinejima is formed of Late Paleopleistocene- to-Holocene non-alkaline felsic volcanics and Late-Miocene to Pleistocene non-alkaline pyroclastic volcanic flows, with Jinata Onsen located on a small bay on the southern side of the island (Fig. 1).
Location of Jinata Onsen on Shikinejima Island, Japan, and inset overview sketch of the field site with sampling localities marked.
Five sites were sampled at Jinata Onsen: the Source Pool, Pool 1, Pool 2, Pool 3, and the Outflow (Fig. 1 and 2). During the first sampling trip in January 2016, two whole community DNA samples were collected from each site for 16S rRNA gene amplicon sequencing. During the second sampling trip, additional DNA was collected from the Source Pool and Pool 2 for shotgun metagenomic sequencing along with gas samples for qualitative analysis. Samples for quantitative gas analysis were collected in October 2017 and April 2018.
Representative photos of Jinata. A) Panorama of the field site, with the Source Pool on the left (Pool 1 below), Pools 2 and 3 in the center, and the Outflow to the bay on the right. B) Undistorted view north up the canyon. C) Undistorted view south towards the bay, overlooking Pool 2. D) The Source Pool, coated in floc-y iron oxides and bubbling with a gas mixture containing CO2, CH4 and trace, potentially variable, H2. E) Pool 2, with a mixture of red iron oxides and green from Cyanobacteria-rich microbial mats. F) Close up of textured microbial mats in Pool 3. G) Close up of the Outflow, at which hot spring water mixes with ocean water. Reprinted with permission from (127).
Samples were collected as mineral scrapings of loosely attached, fluffy iron oxide coatings from surfaces and clasts upstream (Source Pool and Pool 1) and as samples of microbial mats downstream (Pools 2 and 3 and Outflow) using sterile forceps and spatulas (~0.25 cm3 of material). Immediately after sampling, cells were lysed and DNA was preserved with a Zymo Terralyzer BashingBead Matrix and Xpedition Lysis Buffer. Lysis was achieved by attaching tubes to the blade of a cordless reciprocating saw (Black & Decker, Towson, MD, USA) and operating for 1 min. Aqueous geochemistry samples consisted of water collected with sterile syringes and filtered through a 0.2-μm filter. Gas samples were collected near sites of ebullition emerging from the bottom of the Source Pool; collection was performed into serum vials by water substitution that were then sealed underwater to prevent contamination by air.
Geochemical analysisDissolved oxygen (DO), pH, and temperature measurements were performed in situ using an Extech DO700 8-in-1 Portable Dissolved Oxygen Meter (FLIR Commercial Systems, Nashua, NH, USA). In upstream regions at which water temperatures were higher than the operating temperature of the oxygen probe (50°C), DO concentrations were estimated by filling a 50-mL Falcon tube with spring water, sealing the tube to prevent gas exchange, and allowing the water to cool to acceptable temperatures before reopening the tube and immediately measuring oxygen concentrations, thereby minimizing the dissolution of atmospheric O2. Iron concentrations were measured using the ferrozine assay (110) following acidification with 40 mM sulfamic acid to inhibit iron oxidation by O2 or oxidized nitrogen species (67). Ammonia/ammonium concentrations were measured using a TetraTest NH3/NH4+ Kit (TetraPond, Blacksburg, VA, USA) following the manufacturer’s instructions, but with the colorimetry of samples and NH4Cl standards quantified with a Thermo Scientific Nanodrop 2000c spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) at 700 nm to improve sensitivity and accuracy. Anion concentrations were measured via ion chromatography on a Shimadzu Ion Chromatograph (Shimadzu, Kyoto, Japan) equipped with a Shodex SI-90 4E anion column (Showa Denko, Tokyo, Japan).
The presence of H2 and CH4 in gas samples was initially qualitatively assessed by comparison to standards with a Shimadzu GC-14A gas chromatograph within 12 h of collection in order to minimize the oxidation of reduced gases. Subsequent gas samples were analyzed according to previously described methods (112). In brief, samples were analyzed using a gas chromatograph (GC-4000; GL Sciences, Tokyo, Japan) equipped with both a pulsed discharge detector (PDD) and thermal conductivity detector (TCD). The GC was equipped with a ShinCarbon ST packed column (2 m×2.2 mm ID, 50/80 mesh) connected to a HayeSepo Q packed column (2 m×2.2 mm ID, 60/80 mesh) to separate O2, N2, CO2, and light hydrocarbons. Temperature was maintained at 40°C for 6 min before ramping up to 200°C at 20°C min−1. This temperature was held for 6 min before ramping up to 250°C at 50°C min−1 before a final hold for 15 min. The values of standard errors (SE) were obtained by replicate measurements of samples. The detection limit was on the order of 1 nmol cc−1 for H2 and CH4.
Water samples for dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC) concentration measurements were collected with sterile syringes and transferred after filtering through a 0.2-μm filter to pre-vacuumed 30-mL serum vials that were sealed with butyl rubber septa and aluminum crimps.
DIC and DOC concentrations in water samples were analyzed by measuring CO2 in the headspace of the sampled vials after the reaction of samples with either phosphoric acid for DIC or potassium persulfate for DOC with a Shimadzu GC-14A gas chromatograph. Sodium bicarbonate standards and glucose standards were used to make calibration curves for quantifying DIC and DOC concentrations, respectively.
16S rRNA gene amplicon and metagenomic sequencing and analysesThe sequencing and analysis of 16S rRNA gene amplicon data followed previously described methods (126). After returning to the lab, bulk environmental DNA was extracted and purified with a Zymo Soil/Fecal DNA extraction kit. The V4–V5 region of the 16S rRNA gene was PCR amplified using the archaeal and bacterial primers 515F (GTGCCAGCMGCCGCGGTAA) and 926R (CCGY CAATTYMTTTRAGTTT) (14). DNA was quantified with a Qubit 3.0 fluorimeter (Life Technologies, Carlsbad, CA, USA) following the manufacturer’s instructions after the DNA extraction and PCR steps. The successful amplification of all samples was verified by viewing on a gel after initial pre-barcoding PCR (30 cycles). Duplicate PCR reactions were pooled and reconditioned for five cycles with barcoded primers. Samples for sequencing were submitted to Laragen (Culver City, CA, USA) for analysis on an Illumina MiSeq platform. Sequence data were processed using QIIME version 1.9.1 (13). Raw sequence pairs were joined and quality-trimmed using the default parameters in QIIME. Sequences were clustered into de novo operational taxonomic units (OTUs) with 99% similarity using the UCLUST open reference clustering protocol (24). The most abundant sequence was selected as the representative for each de novo OTU (121). Taxonomic identification for each representative sequence was assigned using the Silva-132 database (90) clustered separately at 99 and 97% similarities. Singletons and contaminants (OTUs appearing in negative control datasets) were removed. 16S rRNA gene sequences were aligned using MAFFT (60) and a phylogeny constructed using FastTree (89). Alpha diversity was estimated using the Shannon Index (96) and Inverse Simpson metric (1/D) (46, 100). Sampling depths were estimated using Good’s Coverage (35). All statistics were calculated using scripts in QIIME and reported at the 99 and 97% OTU similarity levels. Multidimensional scaling (MDS) analyses and plots to evaluate similarities between different samples and environments were produced in R using the vegan and ggplot2 packages (91, 135) (Oksanen, J., F.G. Blanchet, M. Friendly, et al. 2016. Vegan: community ecology package. R package version 2.3-5. R Foundation, Vienna, Austria).
Following initial characterization via 16S rRNA gene sequencing, four samples were selected for shotgun metagenomic sequencing: JP1-A and JP3-A from the first sampling trip, and JP1L-1 and JP2-1 from the second sampling trip. Purified DNA was submitted to SeqMatic LLC (Fremont, CA, USA) for library preparation and 2×100-bp paired-end sequencing via Illumina HiSeq 4000 technology. Samples JP1-A and JP3-A shared a single lane with two samples from another project, while JP1L-1 and JP2-1 shared a lane with one sample from another project.
Raw sequence reads from all four samples were co-assembled with MegaHit v. 1.02 (75) and genome bins constructed based on nucleotide composition and differential coverage using MetaBAT (57), MaxBin (136), and CONCOCT (Alneberg, J., B.S. Bjarnason, I. de Bruijn, M. Schirmer, J. Quick, U.Z. Ijaz, N.J. Loman, A.F. Andersson, and C. Quince. 2013. CONCOCT: clustering contigs on coverage and composition. arXiv preprint arXiv:1312.4038.) prior to dereplication and refinement with DAS Tool (99) to produce the final bin set. Genome bins were assessed for completeness, contamination, and strain-level heterogeneity using CheckM (85), tRNA sequences found with Aragorn (71), and the presence of metabolic pathways of interest predicted with MetaPOAP (130). Coverage was extracted using bbmap (Bushnell, B. 2016. BBMap short read aligner. University of California, Berkeley, California. URL: http://sourceforge.net/projects/bbmap.) and samtools (74). Genes of interest (e.g. coding for ribosomal, photosynthesis, iron oxidation, and electron transport proteins) were identified from assembled metagenomic data locally with BLAST+ (10) and were screened against outlier (e.g. likely contaminant) contigs as identified by CheckM using tetranucleotide, GC, and coding density contents. Translated protein sequences of genes of interest were aligned with MUSCLE (23), and alignments were manually curated in Jalview (134). Phylogenetic trees were calculated using RAxML (106) on the Cipres science gateway (80). Node support for phylogenies was calculated with transfer bootstraps by BOOSTER (72). Trees were visualized with the Interactive Tree of Life viewer (73). Since the sequencing depth of each sample in the full metagenome was uneven, the relative abundance of genes of interest between metagenomic datasets was normalized to the coverage of rpoB genes in each raw dataset as mapped onto the co-assembly. Similar to the 16S rRNA gene, rpoB is a highly conserved, vertically-inherited gene that is useful for the taxonomic identification of organisms, but has the added advantage that it is only known to occur as a single copy per genome (15) and is more readily assembled in metagenomic datasets (e.g. 127). The presence and classification of hydrogenase genes was performed with HydDB (103). The taxonomic assignment of MAGs was made based on placement in a reference phylogeny built with concatenated ribosomal protein sequences following methods from Hug et al. (48), and confirmed using GTDB-Tk (86). The optimal growth temperatures of MAGs were predicted based on proteome-wide 2-mer amino acid compositions following previously described methods (Li, G., K.S. Rabe, J. Nielsen, and M.K. Engqvist. 2019. Machine learning applied to predicting microorganism growth temperatures and enzyme catalytic optima. BioRxiv. doi: https://doi.org/10.1101/522342).
The source water of Jinata Onsen emerges with low DO concentrations near our limit of detection, is iron-rich, and gently bubbles gas from the spring source (Fig. 1 and 2, and Table 1). Temperatures at the source are ~63°C. Water emerges into the Source Pool, which has no visible microbial mats or biofilms (Fig. 2D). Surfaces are instead coated with a fluffy red precipitate, likely a poorly ordered or short range-ordered ferric iron oxide phase such as ferrihydrite. Flow from the source is, at least partially, tidally charged, with the highest water levels and flow rates occurring at high tide. At low tide, flow rates drop and the water level of the Source Pool may decrease by decimeters, with portions of the Source Pool potentially draining during spring low tides. Spring water collects downstream into a series of pools (Pools 1–3) (Fig. 2C, E, and F), which cool sequentially (Fig. 3 and Table S1). Pool 1 contains iron oxides, similar to the Source Pool, but also develops macroscopic microbial streamers that are coated in iron oxides and thin veil-like layers of microorganisms overlaying iron oxide sediments—structures similar to those typically made by marine iron-oxidizing Zetaproteobacteria (e.g. 33). Streamers are very fine (mm-scale) and delicate (break apart on contact with forceps) but can reach several centimeters in length. Cyanobacteria in Pools 2 and 3 display high levels of photosynthetic activity as revealed by high DO concentrations (~234 μM), low DIC concentrations, and the accumulation of visible O2 bubbles on the surface and within the fabric of the mat. Downstream pools (Pools 2 and 3) mix with seawater during high tide due to wave action; however, this seawater influence does not appear to influence the Source Pool or Pool 1. Samples were collected and temperatures were measured at high tide, reflecting the lowest temperatures experienced by microbes in the pools; at low tide, hot spring input is dominant and temperatures rise (observed range at each site in Table S1). Subaqueous surfaces in Pools 2 and 3 are covered in thick microbial mats. In Pool 2, the mat is coated in a layer of fluffy iron oxide similar to that in the Source Pool, with a dense microbial mat below (Fig. 2E). Pool 3 contains only patchy iron oxides, with mostly exposed microbial mats displaying a finger-like morphology. These “fingers” were 0.5–1 cm in diameter, up to ~5 cm long and were closely packed and carpeting surfaces of Pool 3 below the high tide line, potentially related to turbulent mixing from wave action during high tide (Fig. 2F). The Outflow is the outlet of a channel connecting Pool 2 to the bay. Its hydrology is dominantly marine with small admixtures of inflowing spring water (Fig. 2G).
| T | 63°C |
| pH | 5.4 |
| DO | 4.7 μM |
| Fe2+ | 261 μM |
| NH3/NH4+ | 87 μM |
| Cl− | 654 mM |
| SO4− | 17 mM |
| NO3− | <1.6 μM |
| NO2− | <2.2 μM |
| HPO4− | <1 μM |
Summary of geochemical and microbiological trends along the flow path of Jinata Onsen. Top: Panoramic view of Jinata Onsen, with the Source Pool at the left and the flow of spring water toward the bay at the right, with sampling locations indicated. Middle: Geochemical transect across the spring, showing temperature (°C, left axis) and dissolved Fe(II) and O2 (μM, right axis). Bottom: Stacked bar chart of the relative community abundance of dominant microbial phyla as assessed by 16S rRNA amplicon sequencing. Sequence data were binned at the phylum level and duplicate samples at each site were averaged. Reads that could not be assigned to a phylum were discarded; all phyla that do not make up more than 2% of the community at any one site have been collapsed to “Other”. Near the source, the community is predominantly made up of iron- and/or hydrogen-oxidizing organisms in the phyla Proteobacteria and Aquificae. As hot spring water flows downstream, it equilibrates with the atmosphere and eventually mixes with seawater, resulting in downstream cooling, the accumulation of oxygen, and loss of dissolved iron due to biological and abiotic processes. Oxygenic Cyanobacteria become progressively more abundant downstream. Putative hydrogen- and iron-oxidizing lithotrophs dominate near the source, whereas phototrophic Cyanobacteria dominate downstream. Additional community diversity is found in Table S4.
Jinata hot spring was visited twice for observations and community DNA sampling in 2016 (January and September), and again for observations and gas sampling in October 2017 and April 2018. These visits corresponded to a range of tidal conditions, including a spring low and high tide in September 2016. The general features of the spring were consistent across this period (including the abundance and distribution of iron minerals and microbial mats), differing primarily in apparent tidal dependence in the flow rate and water level of the spring and the extent of the seawater influence on Pool 3. These differences in flow and mixing led to variations in water temperatures of 3–10°C (Table S1). At high tide, the flow rate of the spring increases, as does seawater influx to Pool 3. During the spring low tide, the spring flow stagnated and the water level of Source Pool and Pool 1 dropped by decimeters, with some portions draining entirely. During less extreme low tides observed on other dates, the spring flow was low but nonzero and the water level of the Source Pool did not significantly drop. While marked variability was observed in the flow rate from the spring based on tides (and resulting shifts in water level and temperature), the overall geochemistry of the source water and microbial community mostly appeared to be similar between expeditions.
GeochemistryGeochemical measurements along the flow path of Jinata Onsen revealed a major shift from hot, low-oxygen, high-iron source water to cooler, more oxygen-rich water with less dissolved iron downstream. Geochemistry measurements of Jinata source water are summarized in Table 1, while geochemical gradients along the stream outflow are shown in Fig. 3 and Table S1. Source waters were slightly enriched in chloride relative to seawater (~23.2 g L−1 in Jinata source water versus ~19.4 g L−1 in typical seawater) and depleted in sulfate (~1.6 g L−1 in Jinata versus ~2.7 g L−1 in seawater), but approached seawater concentrations downstream as mixing increased. Water emerging from the source was 63°C, very low in DO (~4.7 μM), at pH 5.4, and contained substantial concentrations of dissolved iron (~250 μM Fe2+). DOC in the water of Pool 1 was high (~1.31 mM). It is unknown whether this is produced in situ or if the source water emerges with high DOC. DOC and DIC both decreased along the outflow of the spring (Table S1). After emerging from the source, the spring water exchanges gases with the air due to mixing associated with water flow and gas ebullition, and DO increased to 39 μM at the surface of the Source Pool. As water flows downstream from the Source Pool, it cools slightly, exchanges gases with the atmosphere, and intermittently mixes with seawater below Pool 1.
H2 and CH4 were both qualitatively detected in bubbles from the Source Pool following initial sampling in September 2016. However, during subsequent analyses to quantify the gas composition in October 2017 and April 2018, the gas was found to contain CO2, CH4, and N2 (Table S2). This subsequent non-detection of H2 may be related to temporal variability in the gas composition at Jinata (e.g. following tidal influence; significant variability was observed in the CO2:N2 ratio between two sampling dates, Table S2) or may reflect oxidation of H2 between sampling and analysis. The detection limit of H2 for these later measurements was ~1 nmol cc−1 (in the gas phase of our quantitative gas analyses, or ~1 nM in the aqueous phase [2]), well above the energetic and ecological limits for hydrogenotrophic metabolisms (e.g. 51) leaving open the possibility of biologically significant H2 fluxes at Jinata around the time of sampling. The oxidation of H2 coupled to O2 reduction is a thermodynamically favorable process even at very low substrate concentrations (e.g. ΔrG’ <−375 kJ mol−1 with substrate concentrations of 0.1 nM H2(aq) and 0.1 μM O2(aq), below our limit of detection) (30). Consistent with this thermodynamic favorability, biology has been shown to make use of this metabolism in environments such as hot springs with H2 concentrations near our detection limits (19) and in Antarctic soils where microbes rely on uptake of trace atmospheric H2 at concentrations of ~190 ppbv (51). Therefore, the trace amounts of H2 that may be present in the source water at Jinata may be sufficient to support lithoautotrophy near the hot spring source in organisms possessing the genetic capacity for hydrogen oxidation, as discussed below. Metagenomic data suggest that organisms with the genetic capacity for hydrogenotrophic metabolisms are abundant in upstream regions of Jinata (discussed below), and so it is also possible that low dissolved H2 concentrations may be due to active biological consumption near the hot spring source. Improved quantification of H2 concentrations, measurements of hydrogenase activity, and the productivity of hydrogenotrophic microbes will be necessary to determine the relative contribution of hydrogen oxidation to productivity at Jinata.
16S rRNA gene amplicon and genome-resolved metagenomic sequencing16S rRNA gene amplicon and metagenomic sequencing of microbial communities at Jinata Onsen revealed a highly diverse community. In total, 16S rRNA gene amplicon sequencing recovered 456,737 sequences from the 10 samples at Jinata (Tables S3, 4, and 5). Reads per sample following filtering for quality and the removal of chimeras ranged between 2,076 for Pool 3 Sample B and 96,268 for Pool 1 Sample A (median 32,222, mean 35,479, and standard deviation 26,014). On average, 65% of the microbial community was recovered from Jinata samples at the 99% OTU level based on the Good’s Coverage statistic of the 16S rRNA gene (ranging between 50% coverage in Outflow Sample A and 80% in Pool 1 Sample A) and 82% at the 97% OTU level (between 69% for Pool 2 Sample B and 93% for Pool 1 Sample B). MDS analysis (Fig. S1) demonstrates that samples from the same site were highly similar, and adjacent sites (e.g. Source Pool and Pool 1, Outflow and Pool 3) also showed a high degree of similarity. However, a substantial transition was observed in microbial community diversity between the most distant samples (e.g. Source Pool and Outflow).
Shotgun metagenomic sequencing of four samples from Jinata Onsen recovered 121 GB of data, forming a 1.48-Gb co-assembly consisting of 1,531,443 contigs with an N50 of 1,494 bp. Nucleotide composition and differential coverage-based binning of the co-assembly via multiple methods followed by dereplication and refinement resulted in a final set of 161 medium- or high-quality metagenome-assembled genomes (MAGs) following current standards (i.e. completeness >50% and contamination <10%) (6). These MAGs are from diverse phyla of Bacteria and Archaea (Fig. 4); metagenome and MAG statistics with tentative taxonomic assignments for recovered MAGs are shown in Table S6, while MAGs of particular interest due to their potential contribution to primary productivity at this site or due to substantial genetic or metabolic novelty are discussed in depth below and shown in phylogenetic trees alongside reference strains in Fig. 5, 6, and 7.
Phylogeny of Bacteria and Archaea based on concatenated ribosomal proteins. Numbers in parentheses next to phylum labels refer to the number of MAGs recovered from Jinata Onsen. Labels for phyla with two or fewer MAGs recovered from Jinata were omitted for clarity. The reference alignment was modified from Hug et al. (48). A full list of MAGs recovered is available in Table S6.
Phylogeny of Zetaproteobacteria, rooted with Alphaproteobacteria, built with concatenated ribosomal protein sequences. Data from (77), (82), (101), and other draft genomes available on Genbank. Transfer bootstrap expectation (TBE) support values as calculated by BOOSTER (72) shown for internal nodes. In cases for which reference genomes have a unique strain name or identifier, this is included; otherwise Genbank WGS genome prefixes are used.
Phylogeny of Calditrichaeota, rooted with Bacteroidetes, built with concatenated ribosomal protein sequences. Transfer bootstrap expectation (TBE) support values as calculated by BOOSTER (72) shown for internal nodes. Data from (68) and other draft genomes available on Genbank. In cases where reference genomes have a unique strain name or identifier, this is included; otherwise Genbank WGS genome prefixes are used.
Detailed phylogeny of the phylum Chloroflexi, with class-level clades highlighted in gray, built with concatenated ribosomal protein sequences. The large basal class Dehalococcoidia, which was not observed in 16S rRNA or metagenome data from Jinata, is omitted for clarity. The phylogeny contains MAGs reported here, members of the phylum Chloroflexi previously described (16, 20, 38, 41, 43–45, 62, 70, 83, 104, 122, 123, 127–129), and members of the closely related phylum Armatimonadetes as an outgroup (21, 125). MAGs described here are highlighted in green, and MAGs previously reported from Jinata Onsen are highlighted in pink. Transfer bootstrap expectation (TBE) support values as calculated by BOOSTER (72) shown for internal nodes. In cases for which reference genomes have a unique strain name or identifier, this is included; otherwise Genbank WGS genome prefixes are used.
As Jinata spring water flows from its source to the ocean, it transitions from hot, low-oxygen, high-iron water to cooler, iron-depleted, oxygen-rich water in downstream regions (Fig. 3). Following this geochemical transition is a major shift in the composition of the microbial community—from a high-temperature, putatively lithotrophic community which produces little visible biomass upstream, to a lower temperature community with well-developed, thick microbial mats downstream. This shift in community composition is summarized in Fig. 3, with complete diversity data in the Supplemental Information (including OTU counts per sample in Table S4 and relative abundance binned at the class level in Table S5). Below, we discuss the overall physiological and taxonomic trends across the spring sites as inferred from diversity and genomic analysis.
Potential for lithotrophic iron and hydrogen oxidationThe hot spring water emerging at the Source Pool at Jinata contains abundant dissolved Fe2+ and trace H2 (though measurements of gas content varied, as discussed above) (Table 1). Although rates of carbon fixation were not measured, the appearance of (presumed) zetaproteobacterial veils and streamers together with molecular evidence for lithoautotrophic microbes suggest that these electron donors may fuel productivity and determine the microbial community upstream at the Source Pool and Pool 1, where microbial mats were not well developed. The low accumulation of visible biomass in upstream regions of Jinata is similar to other microbial ecosystems fueled by iron oxidation (e.g. Oku-Okuhachikurou Onsen [126], Fuschna Spring [43], and Jackson Creek [92]), in which lithotrophic communities appear capable of accumulating less organic carbon than communities fueled by oxygenic photosynthesis (including those in downstream regions at Jinata).
The results of 16S rRNA gene amplicon sequencing indicate that the most abundant organisms in the Source Pool are members of the Aquificae family Hydrogenothermaceae (32% of reads in the Source Pool and 11.5% of reads in Pool 1). Members of this family are typically marine thermophilic lithotrophs capable of iron and hydrogen oxidation as well as heterotrophy (114); at Jinata, they may be utilizing Fe2+, H2, or DOC. The seventh most abundant OTU in the Source Pool samples was a novel sequence that was 89% similar to a strain of Persephonella observed in an alkaline hot spring in Papua New Guinea. Persephonella is a genus of thermophilic, microaerophilic hydrogen-oxidizing bacteria within Hydrogenothermaceae (36). Despite their abundance as assessed by 16S rRNA gene amplicon sequencing (Fig. 3), only four partial Aquificae MAGs were recovered from Jinata, of which only one (J026) was reasonably complete (~94%). Two Aquificae MAGs recovered Group 1 NiFe hydrogenase genes, which could support hydrogenotrophy; the absence of hydrogenases from the other MAGs may be related to their low completeness or could reflect a utilization of iron or other electron donors and not H2 in these organisms.
The other most abundant organisms near the source are members of Zetaproteobacteria, a group typified by the neutrophilic, aerobic iron-oxidizing genus Mariprofundus common in marine systems (27). Zetaproteobacteria accounted for 24% of 16S rRNA gene sequences in the Source Pool and 26.5% in Pool 1. All Zetaproteobacteria characterized to date are obligate iron- and/or hydrogen-oxidizing lithoautotrophs (82), suggesting that these organisms play a substantial role in driving carbon fixation in the Source Pool and Pool 1.
Members of Mariprofundaceae have been observed to have an upper temperature limit for growth of 30°C (28), while Zetaproteobacteria at Jinata are found at temperatures up to 63°C. This currently represents a unique high-temperature environment for these organisms. In particular, the third most abundant OTU in the Source Pool and Pool 1 sample A is an unknown sequence that is 92% identical to a sequence from an uncultured zetaproteobacterium from a shallow hydrothermal vent in Papua New Guinea (79). This sequence likely marks a novel lineage of high-temperature iron-oxidizing Zetaproteobacteria.
The relative abundance of Hydrogenothermaceae decreases to less than 1% of sequences in areas where microbial mats are well developed downstream of Pool 1; however, Zetaproteobacteria continue to account for ~1–4% of reads in Pools 2 and 3 in which dissolved iron concentrations are still significant (Fig. 3). This relative abundance change may be due more to the increased abundance of other organisms rather than a decrease in the number of Zetaproteobacteria or their ability to make a living oxidizing iron. This hypothesis awaits confirmation by a technique such as qPCR. In contrast, the absence of Hydrogenothermaceae downstream may be a real signal driven by the rapid disappearance of trace H2 as an electron donor. However, in both cases, a drop in relative abundance is likely related to the increasing total biomass (i.e. number of cells) downstream as Cyanobacteria become more productive, leading to sequences from Hydrogenothermaceae and Zetaproteobacteria being diluted out by increased numbers of Cyanobacteria, Chloroflexi, and other sequences.
Four MAGs affiliated with Zetaproteobacteria were recovered from Jinata with completeness estimates by CheckM ranging between ~80 and ~97% (J005, J009, J030, and J098). While these MAGs did not recover 16S rRNA genes, RpoB- and concatenated ribosomal protein-based phylogenies illustrated that members of this group at Jinata Onsen do not belong to the characterized genera Mariprofundus or Ghiorsea, but instead form separate basal lineages within Zetaproteobacteria (Fig. 5). Despite their phylogenetic distinctness, these MAGs largely recovered genes associated with aerobic iron oxidation as expected based on the physiology of other Zetaproteobacteria. These include a terminal O2 reductase from the C-family of heme copper oxidoreductases for respiration at low O2 concentrations and Cyc2 cytochrome genes implicated in ferrous iron oxidation in Zetaproteobacteria and other taxa (e.g. Chlorobi) (39, 40, 59). Hydrogenase catalytic subunit genes (neither [NiFe] nor [FeFe]) were not recovered in zetaproteobacterial MAGs even at high completeness, suggesting that these organisms are not hydrogenotrophic. Consistent with the obligately autotrophic lifestyle of previously characterized Zetaproteobacteria, J009 and J098 encode carbon fixation via the Calvin cycle. However, J005 and J030 did not recover genes for carbon fixation via the Calvin cycle such as the large and small subunits of rubisco, phosphoribulose kinase, or carboxysome proteins. The high completeness of these MAGs (~94–97%) makes it unlikely that these genes would all fail to be recovered (MetaPOAP False Negative estimates 10−5–10−7). The absence of carbon fixation pathways from these genomes together with the availability of abundant DOC in Pool 1 (~1.3 mM) suggest that these organisms may be heterotrophic, a lifestyle not previously observed for members of Zetaproteobacteria.
Seven MAGs were recovered from the enigmatic bacterial phylum Calditrichaeota (J004, J008, J042, J070, J075, J140, and J141) (Fig. 6). While few members of Calditrichaeota have been isolated or sequenced, the best known of these is Caldithrix abyssi (81); this taxon was characterized as an anaerobic thermophile capable of lithoheterotrophic H2 oxidation coupled to denitrification and organoheterotrophic fermentation (1, 78). The Calditrichaeota MAGs reported here are up to 97% complete (J004) and contain members with variable putative metabolic capabilities, potentially including aerobic hydrogen- or iron-oxidizing lithoautotrophy. In the Calditrichaeota MAGs recovered from Jinata Onsen, aerobic respiration via A-family heme copper oxidoreductases could potentially be coupled to autotrophic hydrogen oxidation (via the Group 1d NiFe hydrogenase in J042) or iron oxidation (via the pioA gene in J075); however, C. abyssi appears incapable of aerobic respiration despite encoding an A-family heme copper oxidoreductase (68). A MAG from a member of Calditrichaeota has previously been recovered from Chocolate Pots hot spring in Yellowstone National Park (32); together with the data presented here this suggests that this phylum may be a common member of microbial communities in iron-rich hot springs. Unlike previously described Calditrichaeota which are all heterotrophic (78), most of the Calditrichaeota MAGs reported here possess a putative capacity for carbon fixation via the Calvin cycle. J004 is closely related to C. abyssi, while the other MAGs form two distinct but related clades (Fig. 6).
Oxygenic photosynthesisCyanobacteria are nearly absent from the Source Pool (<0.15% relative abundance), but are observed at low numbers in Pool 1 and become abundant starting in Pool 2. The most abundant Cyanobacteria present are predominantly members of Nostocales. This group includes Leptolyngbya and Phormidium, genera of filamentous non-heterocystous Cyanobacteria that are present in other hot springs of similar temperatures (e.g. 5, 94, 126). Diverse cyanobacterial MAGs were recovered, including members of the orders Pleurocapsales (J083), Chroococcales (J003 and J149), and Oscillatoriales (J007, J055, and J069). In Outflow samples, chloroplast sequences related to the diatom Melosira were abundant.
Cyanobacteria are sometimes underrepresented in 16S rRNA gene amplicon sequencing datasets as a result of a poor DNA yield or amplification biases (e.g. 84, 118); however, the low abundance of Cyanobacteria near the Source Pool was confirmed by fluorescent microscopy, in which cells displaying cyanobacterial autofluorescence were abundant in downstream samples, but not in samples from the Source Pool (Fig. S2). Thick microbial mats initially appear in Pool 2 when Cyanobacteria become abundant, suggesting that oxygenic photosynthesis fuels more net carbon fixation than lithotrophy in these environments.
It has been suggested that high ferrous iron concentrations are toxic to Cyanobacteria, and that this would have greatly reduced their productivity under ferruginous ocean conditions such as those that may have persisted through much of the Archean era (113). The abundant Cyanobacteria observed to be active at Jinata under high iron concentrations suggest that Cyanobacteria can adapt to ferruginous conditions, and therefore iron toxicity might not inhibit Cyanobacteria over geological timescales. Indeed, the soluble iron concentrations observed at Jinata are higher (150–250 μM) than predicted for the Archean oceans (<120 μM, 47) or observed at other iron-rich hot springs (~100–200 μM, 87, 126), making Jinata an excellent test case for determining the ability of Cyanobacteria to adapt to high iron concentrations. Culture-based physiological experiments may be useful to determine whether Jinata Cyanobacteria utilize similar strategies to other iron-tolerant strains (e.g. by those in Chocolate Pots Hot Spring, 87, or the ferric iron-tolerant Leptolyngbya-relative Marsacia ferruginose, 8) or whether Jinata strains possess unique adaptations that allow them to grow at higher iron concentrations than known for other environmental Cyanobacteria strains. This will in turn provide insight into whether iron tolerance is due to evolutionarily conserved strategies or whether this is a trait that has evolved convergently multiple times.
Diverse novel Chloroflexi from Jinata OnsenIn addition to the primary phototrophic and lithotrophic carbon fixers at Jinata, 16S rRNA gene and metagenomic data sets revealed diverse novel lineages within the phylum Chloroflexi. Twenty-three Chloroflexi MAGs were recovered, introducing substantial genetic and metabolic diversity that expands our understanding of this group. While the best known members of this phylum are Type 2 Reaction Center-containing lineages such as Chloroflexus and Roseiflexus within the class Chloroflexia (e.g. 117), phototrophy is not a synapomorphy of the phylum Chloroflexi or even the class Chloroflexia (e.g. 122) and most of the diversity of the phylum belongs to several other classes primarily made up of non-phototrophic lineages (127). The bulk of Chloroflexi diversity recovered from Jinata belongs to “subphlyum I”, a broad group of predominantly non-phototrophic lineages that was originally described based on the class- or order-level lineages Anaerolineae and Caldilineae (137), but also encompasses the related groups Ardenticatenia, Thermoflexia, and Candidatus Thermofonsia (20, 61, 127).
16S rRNA gene analysis indicated that members of Anaerolineae and Ca. Thermofonsia (annotated by Silva and GTDB-Tk as the order SBR1031) were fairly abundant at Jinata, with Anaerolineae at ~3% relative abundance in the Source Pool and Pool 1 and Ca. Thermofonsia at ~3.5% relative abundance in Pools 2 and 3. Three MAGs recovered from Jinata (J082, J097, and J130) are associated with the Anaerolineae class, as determined by RpoB and concatenated ribosomal protein phylogenies, along with seven associated with Ca. Thermofonsia (J027, J033, J036, J038, J039, J064, and J076). Of particular interest among these is J036, a close relative of the phototrophic Ca. Roseilinea gracile (66, 115, 116). J036 contains a 16S rRNA gene that is 96% similar to that of Ca. Roseilinea gracile, and two-way AAI estimates (93) showed 73.6% similarity between the two strains, indicating that these strains are probably best classified as distinct species within the same genus. Unlike other phototrophs in the phylum Chloroflexi that are capable of photoautotrophy via the 3-hydroxypropionate bicycle or the Calvin cycle (65, 98), J036 and Ca. Roseilinea gracile do not encode carbon fixation and are likely photoheterotrophic. Previous studies suggested that the Roseilinea lineage belongs to Anaerolineae (66) or Thermofonsia (127). However, our updated phylogeny presented here places J036 and Roseilinea in a separate lineage along with J033 and J162, diverging just outside of the Anaerolineae+ Thermofonsia clade. This suggests that these strains may instead be yet another class- or order-level lineage within the broader “subphylum I” of Chloroflexi (Fig. 7), an interpretation supported by analysis via GTDB-Tk that places these genomes outside of characterized clades (Table S6).
The Chloroflexi class Ardenticatenia was first described from an isolate from an iron-rich Japanese hydrothermal field (61) and has since also been recovered from sulfidic hot springs (128). Members of Ardenticatenia were present at up to 1.2% relative abundance in Pool 3 in 16S amplicon data. One MAG recovered from Jinata Onsen, J129, was closely related to Ardenticatena maritima. While A. maritima 110S contains a complete denitrification pathway (45), MAG J129 did not recover any denitrification genes. This may be related to the relatively low completeness of this MAG (~70%); however, MetaPOAP (131) False Negative estimates for the probability that all four steps in the canonical denitrification pathway would fail to be recovered in J129 given their presence in the source genome is less than 0.8%. This suggests that most, if not all denitrification genes are absent from the J129 genome and that the capacity for denitrification is not universal within members of Ardenticatena. This would be consistent with broad trends in the apparently frequent modular horizontal gene transfer of partial denitrification pathways between disparate microbial lineages to drive rapid adaption and metabolic flexibility of aerobic organisms in microoxic and anoxic environments, for reasons that are still not well established (17, 109).
Members of the Chloroflexi class Caldilineae were present at up to 0.5% abundance at Jinata in the 16S rRNA gene dataset. Members of Caldilineae have previously been isolated from intertidal hot springs in Iceland (55) and Japanese hot springs (95). Characterized organisms in this class are filamentous, anaerobic, or facultatively aerobic heterotrophs (37, 55, 95); therefore, these taxa may play a role in degrading biomass within low-oxygen regions of microbial mats at Jinata. Three MAGs were recovered that form a deeply branching lineage within the Caldilineae class (J095, J111, and J123), sister to the previously characterized genera Caldilinea and Litorilinea. Like other members of Caldilineae, these strains encode aerobic respiration via A-family heme copper oxidoreductases and both a bc complex III and alternative complex III, and are therefore likely at least facultatively aerobic. J095 also encodes carbon fixation via the Calvin cycle as well as a Group 1f NiFe hydrogenase, suggesting a potential capability for hydrogenotrophic lithoautotrophy, expanding the known metabolic diversity of this class and the Chloroflexi phylum as a whole.
MAG J114 branches at the base of subphylum I of Chloroflexi, potentially the first member of a novel class-level lineage. The divergence between Anaerolineae and Caldilineae has been estimated to have occurred on the order of 1.7 Ga (98). The phylogenetic placement of J114 suggests that it diverged from other members of subphylum I even earlier, and it may be a good target for future investigation to assess aspects of the early evolution of the phylum Chloroflexi. J114 encodes aerobic respiration via an A-family heme copper oxidoreductase and an alternative complex III like many other non-phototrophic Chloroflexi lineages (e.g. 122, 127) as well as a Group 1f NiFe hydrogenase and carbon fixation via the Calvin cycle, suggesting the capacity for aerobic hydrogen-oxidizing autotrophy—a lifestyle not previously described for members of Chloroflexi.
To our knowledge, this is the first overall geomicrobiological characterization of Jinata Onsen, providing baseline descriptions of geochemistry and microbial diversity in order to establish a series of testable hypotheses that can be addressed by future studies. We have also provided genome-resolved metagenomic sequencing of this site focusing on members of the microbial community predicted to be responsible for the bulk of primary productivity in this system along with other organisms belonging to novel or under-characterized lineages. However, this is just a subset of the diverse microbial populations at Jinata Onsen; many more MAGs from across the tree of life were recovered than are discussed in detail here but which may be of use to others (Fig. 4 and Table S6).
The diversity of iron-oxidizing bacteria at Jinata is different than in other Fe2+-rich springs and environments. For example, in freshwater systems such as Oku-Okuhachikurou Onsen in Akita Prefecture (126) and Budo Pond in Hiroshima Prefecture, Japan (58), iron oxidation is primarily driven by the activity of chemoautotrophs such as members of Gallionellaceae. In contrast, at Chocolate Pots hot spring in Yellowstone National Park, USA, iron oxidation is primarily abiotic, driven by O2 produced by Cyanobacteria, with only a small contribution from iron-oxidizing bacteria (32, 119). The distinct iron-oxidizing community at Jinata Onsen may be related to the salinity of the spring water, or biogeographically by access to the ocean, as Zetaproteobacteria are typically found in marine settings, particularly in deep ocean basins associated with hydrothermal iron sources (28). Despite the taxonomically distinct iron oxidizer communities between Jinata and Oku-Okuhachikurou Onsen, both communities support only limited visible biomass in regions dominated by iron oxidizers (126), perhaps reflecting the shared biochemical and bioenergetic challenges of iron oxidation incurred by diverse iron-oxidizing bacteria including Gallionellaceae and Zetaproteobacteria (4, 28, 126). Future work focused on isolation and physiological characterization of these microbes, quantification of rates and determination of microbial drivers of carbon fixation and aerobic and anaerobic heterotrophy, and carbon isotope profiling of organic and inorganic species along the flow path of the hot spring will be necessary to fully characterize the activity of microbes at Jinata and to fully compare this system to other areas with high dissolved ferrous iron concentrations (e.g. Oku-Okuhachikurou Onsen [126], Fuschna Spring [42], Jackson Creek [92], and Chocolate Pots Hot Spring [32, 119]).
Future work aimed at a more complete understanding of the geochemistry of Jinata and its impact on microbial communities will be valuable. For example, a quantitative determination of budgets and variability of gas chemistry and DOC will be beneficial. DOC may stimulate heterotrophic activity by the microbial community at Jinata, coupled to aerobic or anaerobic respiration (such as dissimilatory iron reduction, as observed in other iron-rich hot springs, e.g. 31), resulting in the drawdown of DOC downstream. Since the source of this DOC is unclear, future work will be necessary to determine whether DOC is present in the source water or if it is produced in situ by the microbial community in the Source Pool and Pool 1. Future work is also needed to evaluate the potential for dissimilatory iron reduction and other anaerobic metabolisms including sulfate reduction and methanogenesis.
Throughout Earth history, the metabolic opportunities available to life, and the resulting organisms and metabolisms responsible for driving primary productivity, have been shaped by the geochemical conditions of the atmosphere and oceans. The modern, sulfate-rich, well-oxygenated oceans we see today reflect a relatively recent state—one typical of only the last few hundred million years (e.g. 76). For the first half of Earth history, until ~2.3 Ga, the atmosphere and oceans were anoxic (52), and seawater was rich in dissolved iron but poor in sulfur (120). At this time, productivity was low and fueled by metabolisms such as methanogenesis and anoxygenic photosynthesis (12, 64, 131). Following the expansion of oxygenic photosynthesis by Cyanobacteria and higher primary productivity around the Great Oxygenation Event ~2.3 Ga (18, 29, 124, 133), the atmosphere and surface ocean accumulated some oxygen, and the ocean transitioned into a state with oxygenated surface waters but often anoxic deeper waters, rich in either dissolved iron or sulfide (11, 53, 54, 88). At Jinata Onsen, this range of geochemical conditions is recapitulated over just a few meters, providing a useful test case for probing the shifts in microbial productivity over the course of Earth history. In particular, the concomitant increase in visible biomass at Jinata as the community shifts from lithotrophy toward water-oxidizing phototrophy (i.e. oxygenic photosynthesis) is consistent with estimates for greatly increased primary production following the evolution and expansion of Cyanobacteria around the GOE (18, 97, 102, 124, 131, 133).
The dynamic abundances of redox-active compounds including oxygen, iron, and hydrogen at Jinata may not only be analogous to conditions on the early Earth, but may have relevance for potentially habitable environments on Mars as well. Early Mars is thought to have supported environments with metabolic opportunities provided by the redox gradient between the oxidizing atmosphere and abundant electron donors such as ferrous iron and molecular hydrogen sourced from water/rock interactions (e.g. 49), and production of these substrates may continue today (22, 107), potentially supporting past or present life in the Martian subsurface (108). Understanding the potential productivity of microbial communities fueled by lithotrophic metabolisms is critical for setting expectations for the presence and size of potential biospheres on other worlds and early in Earth history (e.g. 131–133). Uncovering the range of microbial metabolisms present under the environmental conditions at Jinata, and their relative contributions to primary productivity, may therefore find application to predicting environments on Mars most able to support productive microbial communities.
Data availabilityRaw 16S rRNA gene amplicon data, raw metagenomic sequence data, and MAGs have been uploaded and made publicly available on NCBI under Project Number PRJNA392119 (genome accession numbers are found in Table S6).
LMW acknowledges support from NASA NESSF (#NNX16AP39H), NSF (#OISE 1639454), NSF GROW (#DGE 1144469), the Lewis and Clark Fund for Exploration and Field Research in Astrobiology, the ELSI Origins Network, and an Agouron Institute postdoctoral fellowship. WWF acknowledges the support of NASA Exobiology award #NNX16AJ57G and the Simons Foundation Collaboration on the Origins of Life. MN and YU acknowledge the support of the Japan Society for the Promotion of Science (JP17K14412, JP17H06105)
