3.2.Fundamental physical and environmental controls on rock-hosted life

The thermal state of the crust constrains the habitable zone. The currently recognized temperature limits for metabolic activity range from −20°C for microorganisms trapped in Siberian permafrost (Rivkina et al., 2000) and −25°C for an aerobic, halophilic heterotroph in laboratory microcosm experiments utilizing ¹⁴C-labeled acetate (Mykytczuk et al., 2013) up to 122°C for a methanogen isolated from a deep sea vent plume (Takai et al., 2008). Based upon temperature alone, the habitable volume for Earth's continental and oceanic crust has been estimated to be ~2×10¹⁸ m³ (Heberling et al., 2010; Magnabosco et al., 2018a), using global heat flow and surface temperature maps and thermal conductivity estimates. In the case of the continental crust, the average depth to the 122°C isotherm is 4km; a maximum depth of 16–23km occurs in the Siberian Craton where mean annual temperatures and heat flow are both lower than average (Magnabosco et al., 2018a). Similar types of calculations for Noachian Mars indicate an average depth to the 122°C isotherm would have been 6–8km, and the corresponding habitable volume based on temperature constraints alone would also be ~10¹⁸ m³ (Michalski et al., 2017).

Temperature and possibly the ionic strength of the crustal fluids play a role in constraining the abundance and activity of subsurface life since cell concentrations appear to be inversely correlated with both parameters (Magnabosco et al., 2018a). Organic markers of biodegradation of petroleum suggest that the maximum temperature of the subsurface biosphere may typically be closer to 80–85°C and that salinity >50g L⁻¹ may inhibit low-energy metabolisms such as methanogenesis to even lower temperatures (Head et al., 2014). High concentrations (>220g L⁻¹) of chaotropic salts, such as MgCl₂, may even preclude life (Hallsworth et al., 2007).

The fluid-bearing (saturated or thin film) porosity that is accessible, that is, with pore throats that are greater than 0.1μm in diameter, also controls the habitable volume. On Earth, rock strata from 3 to 5km depth may have a matrix porosity of 0.5% to 1%, but their habitable volume could be as little as 0.05% to 0.002% (Supplementary Fig. S1; Supplementary Information available online at www.liebertonline.com/ast) due to compaction and cementation. On Mars the porosity is likely to be 10% at comparable depths due to the lower gravitational force, and as a result the subsurface habitable volume on Mars may be greater than that of Earth's.

Porosity and permeability also constrain the flux of nutrients and the degree of metabolic activity. For terrestrial life, a finite minimum quantum of energy determined by the reaction ADP+P → ATP must be available through catabolic redox or substrate-level reactions to be metabolically useful (Müller and Hess, 2017). To sustain life, the Gibbs free energy flux (energy per unit time per cell) (Hoehler, 2004; Onstott, 2004) must be equal to or exceed that required for a cell's (Hoehler and Jørgensen, 2013; Onstott et al., 2013) or a syntrophic community's (Scholten and Conrad, 2000) maintenance. Temperature is a principal control on the maintenance energy demand in part because the diffusivity of H⁺ through the cell membranes increases with temperature (van de Vossenberg et al., 1995). As a result, cell metabolic rates must increase with temperature to counteract these effects. The higher the temperature, the higher the nutrient flux needed to maintain a given subsurface biomass. The same may also hold true for salinity as microorganisms need to manufacture internal osmolytes to maintain osmotic pressure and osmolytic production exerts an additional energy requirement (Oren, 1999).

To the extent that higher rock permeability increases groundwater velocities that increase the rate at which reactants flow toward microorganisms hosted by rock strata, higher permeability can lead to a more prolific subsurface biomass by maintaining a nonzero Gibbs free energy (Marlow et al., 2014a). Three other abiotic processes that operate to enhance the local flux of nutrients are chemical gradients or boundaries, physical heterogeneity, and local abiotic and biotic recycling. First, the presence of high electron donor/acceptor chemical spatial gradients in rock units enhances local diffusive fluxes, which leads to higher microbial activity and biomass than in homogeneous units. On modern Earth the most dramatic examples are typically found at the contacts between organic-rich shale and sulfate-bearing sandstone (Krumholz et al., 1997) or oil and water (Bennett et al., 2013) producing higher rates of microbial metabolism than observed some distance away from these contacts. Other examples include fluid-rock redox interfaces at small scales along fractures, detailed below. On Mars an example might be the boundary between a serpentinized olivine-rich unit and an overlying sulfate-rich unit at Northeast Syrtis (Marlow et al., 2014a). Serpentinization of the olivine would generate H₂ that would then diffuse into the overlying aquifer where it could be microbially oxidized using sulfate, leading to a zone with potential to support high biomass at the boundary. Second, physical heterogeneity can also act to create favorable zones, as is observed in the high cell concentrations within highly fractured and brecciated rock of the Chesapeake Bay impact structure, compared to the overlying marine sediments (Cockell et al., 2012). In this case metabolically active microorganisms are constrained to the fractured rock where they draw down the energy substrates and increase the product concentrations. Though the surrounding massive rock has pore spaces too small for microbes, the diffusive flux of energy substrates from the massive rock into the fractured zone and the diffusion of the products from the fractured zone into the massive rock enhance the biomass residing in the fracture rock (see Section 3.3, Supplementary Fig. S1). Third, abiotic recycling reactions such as radiolysis can continuously generate H₂ from water and recycle metabolic waste products, such as HS^- back into sulfate, and sustain subsurface microbial communities without the need for fluid transport (Lin et al., 2006). “Cryptic sulfur cycling,” in which iron oxides abiotically oxidize sulfide to more oxidized sulfur species, can support organic carbon degradation in non-stoichiometric proportions with a relatively limited sulfur supply (Holmkvist et al., 2011). Finally, syntrophic interactions between different microorganisms also act to sustain subsurface communities by converting waste products back into reactants locally without the need for advective transport (Lau et al., 2016b). Thus, obligately mutualistic metabolism (Morris et al., 2013) may be a characteristic aspect of subsurface microbial communities as a means of avoiding extinctions (Gaidos et al., 1999) because Gibbs free energy will remain nonzero.

3.3.Subsurface biomass distribution

Magnabosco et al. (2018a) recently estimated the total living subsurface prokaryote biomass for Earth was 7–11×10²⁹ cells of which 2–6×10²⁹ cells occur in the continental subsurface, 2×10²⁹ cells exist the oceanic crust, and 3×10²⁹ cells reside in the subseafloor sediments. Permafrost-affected crust and continental ice sheets cover a large fraction of the continental area and are often considered terrestrial analogs to early Mars. In these, as a function of depth, biomass generally declines; but at all depths, biomass varies by orders of magnitude, depending on the sampling location, including substantial intra-site variation (Fig. 3A). The total cell concentrations for Siberian permafrost sediments can be quite high, ranging from 10⁷ to 10⁹ cells g⁻¹ (brown-filled diamonds in Fig. 3A) and diminish with depth up to 100m and with increasing permafrost age up to 2Ma (Gilichinsky and Rivkina, 2011). The cell concentrations within the Greenland ice sheet (light blue-filled circles in Fig. 3A) are much lower, on the order of 10⁵ cells cm⁻³, except at the very bottom where the ice sheet is in contact with the Precambrian bedrock and cell concentrations reach 10⁹ cells cm⁻³ as a result of H₂ generation at the rock-ice interface (Fig. 1). Antarctic permafrost (brown crosses) and subglacial sediments (brown-filled triangles) exhibit cell concentrations that are also greater than those of the adjacent and overlying ice sheets (open triangles). In general, cell concentrations in ice sheets (open squares, triangles, and circles) do not diminish as a function of depth and age as rapidly as observed for permafrost sediments and likely reflect a combination of airborne input flux and in situ metabolism (Chen et al., 2016). Cell concentrations are higher near rock-ice interfaces and in dust-rich ice, pointing to the importance of chemical and physical gradients.

Open in a separate window

FIG. 3.

(A) Cell concentrations versus depth for ice sheets, subglacial sediments, and permafrost. Open squares=Tibetan glacial ice sheets; brown-filled diamonds=Siberian permafrost; blue-filled diamonds=Siberian cryopeg; light gray-filled triangles=Antarctica ice sheets and lakes; brown-filled triangles=Antarctic subglacial sediments; brown crosses=Antarctic permafrost and subglacial sediment in New Zealand; orange crosses=Canadian High Arctic and Svalbard permafrost; light blue–filled circles=Greenland ice sheet; orange-filled circle=Greenland subglacial sediment. (B) Cell concentrations versus depth for rock and soil cores from nonpolar regions. Orange-filled circle=water-saturated sediments or sedimentary rock; orange open circle=vadose zone sediments or sedimentary rock; brown squares=Chesapeake Bay Impact sediments; pink squares=Chesapeake Bay Impact impactite; black-filled orange circle=oil-gas-coal-bearing sediment or sedimentary rock; gray-filled gray diamond=water saturated rhyolitic ash; open gray diamond=deep vadose zone rhyolitic ash; open black diamond=vadose zone basaltic rock; gray-filled black diamond=water-saturated basaltic rock, which includes recent Deccan Trap data from Dutta et al. (2018); red-filled diamond=Deccan Trap granite data from Dutta et al. (2018); purple square=metamorphic rock. Rest of data are from Magnabosco et al. (2018a). Blue open circles=Atacama desert soil from Connon et al. (2007) and Lester et al. (2007). Solid and dashed lines represent the best-fit power law for subseafloor sediments proposed by Parkes et al. (2014).

Unlike the ice cores, the cell concentrations from rock and sediment cores decline with increasing depth following a power law fit (Fig. 3B). Notable exceptions are the cell concentrations reported for the Chesapeake Bay Impact that increase at a depth of 1.5km where the highly fractured basement rock exists (brown-filled squares to pink-filled squares in Fig. 3B; Cockell et al., 2012). In soil zones the concentrations range from 10⁹ cells g⁻¹ down to as low as 10^5–6 cells g⁻¹ in the case of the Atacama Desert (blue open circles in Fig. 3B; Connon et al., 2007; Lester et al., 2007), considered by some to be a terrestrial analog site for Mars because of its aridity and low organic content. Below 10m depth, however, cell concentrations do not correlate with water saturation of the pore space (orange open circles versus orange-filled circles in Fig. 3B). For example, the cell concentrations within unsaturated volcanic ash deposits at a depth of 400m in central Nevada (deep vadose zone) range from 5×10⁴ to 5×10⁷ cells g⁻¹ (gray open diamonds in Fig. 3B; Haldeman and Amy, 1993; Haldeman et al., 1993), which is not significantly different from the cell concentrations reported for water-saturated post-impact sediments of the Chesapeake Bay Impact (brown-filled squares in Fig. 3B; Breuker et al., 2011; Cockell et al., 2012) or Atlantic Coastal Plain Sediments (orange-filled circles in Fig. 3B; Magnabosco et al., 2018a) or Deccan Trap basalts of similar depth (gray-filled diamonds in Fig. 3B; Dutta et al., 2018).

The cell concentrations also do not correlate with the rock type, that is, sedimentary versus igneous versus metamorphic, with the possible exception of salt deposits where cell concentrations are low, ranging from 0.02 to 10⁴ cells g⁻¹ (Schubert et al., 2009a, 2009b, 2010; Wang et al., 2016). Despite their paucity within salt, microorganisms exhibit remarkable preservation with viable cells being isolated from salt deposited tens of thousands to hundreds of millions of years in the past (Jaakkola et al., 2016), though the older claims remain controversial (Hebsgaard et al., 2005; Lowenstein et al., 2005).

Cell concentrations on fracture or cavity surfaces are often considerably higher than those of the surrounding fluid or matrix especially if the interface acts to focus redox fluxes. For example, in deep vadose zones cell concentrations up to 10⁷ to 10⁸ cells cm⁻² and prolific and pigmented biofilms exist on the surfaces of caverns in quartzite (Barton et al., 2014), basalt (Riquelme et al., 2015), and carbonate (Jones et al., 2016). Because of the difficulty of aseptically sampling water-saturated fracture surfaces at depth, only two studies of the cell concentrations on deep fractures have been published. Analysis of modern biofilms, occurring on fracture surfaces in 2.7Ga metavolcanic rocks at a depth of 2.8km, revealed 10⁵ cells cm⁻² with cells occurring in clumps of 2 to >20 (Wanger et al., 2006). Given the fracture width, such a concentration corresponds to a 100× enhancement of the living cell concentration relative to the fracture water. Much lower living cell concentrations, 40 to 2×10³ cells cm⁻², have been reported for 186m deep groundwater-fed fractures in granite (Jägevall et al., 2011). Although these two studies would suggest that deep fracture surfaces do not harbor high biomass concentrations, examination of buried Cretaceous hydrothermal veins reveals preserved organic remains of microbial colonies in mineral surfaces, which by mass would be equivalent to ~10⁹ cells cm⁻² (Klein et al., 2015). Similarly, observations of fossil microbial cells in veins in granite indicate a fossil biomass equivalent to 3×10⁷ cells cm⁻² (Pedersen et al., 1997). The higher cell concentrations in fossil biomass versus living biomass result from accumulation of necromass in biofilms on fluid-filled fracture surfaces over time, similar to what is observed in shallow subseafloor sediments (Lomstein et al., 2012).

McMahon et al. (2018) stated “the bulk of Earth's massive deep biosphere, and presumably also its fossil record, is a poor analog for any ancient or modern Martian equivalent which, in the absence of a productive surface biosphere, would be much smaller and dominated by chemoautotrophs, not heterotrophs.” This statement is not borne out by existing data. As described in Section 3.1, the cell concentrations in continental rock and groundwater do not exhibit any correlation with dissolved or particulate organic carbon concentrations and are dominantly inhabited by chemolithoautotrophs. This finding contrasts with the observations of shallow subseafloor sediments where cell concentrations do correlate with the organic photosynthate content (Lipp et al., 2008). The deep subsurface environments associated with Phanerozoic-age oil (Head et al., 2014) and coal (Kirk et al., 2015) deposits, where heterotrophic metabolisms would perhaps dominate, comprise only 10¹² m³, or 0.0001%, of the total habitable volume of Earth's subsurface biosphere. The overlap in cell concentrations of the continental rocks with those of deep subseafloor sediments suggests that access to organic photosynthate has little impact on deep subsurface biomass (Fig. 3). The cell abundance data and observed metabolisms do not support the claim that most of Earth's deep biosphere is sustained by heterotrophic metabolism of surface-derived photosynthate. Rather, fracture surface concentrations of 10⁵ to 10⁹ cells cm² are observed in Earth's chemolithoautotrophic communities in settings isolated from organic photosynthate and fueled by chemolithoautotrophy, which is an appropriate analog to Mars. (See Section 4 for a discussion of Earth's fossil record.)

Using simple assumptions from the chemical energy available from basalt weathering (10⁻¹³ kJ/g-yr), Jakosky and Shock (1998) estimated that over 4 billion years Mars could accumulate 10g cm⁻² of biomass from a 100m thick basalt layer. For comparison, Earth's continental crust is estimated to contain 0.006–0.02g of extant life cm⁻², integrated from 1m depth to the 122°C isotherm (Magnabosco et al., 2018a). The addition of fluid flow from an oxic surface to reducing subsurface, as where sulfate rocks are in contact with serpentinizing rocks, increases this estimate by an additional 10g of biomass cm⁻² by enhancing the delivery of reactants and removal products (Marlow et al., 2014a). Calculations of the energy flux from subsurface radiolytic reactions for Mars reveal an energy source comparable to that found on Earth, indicating that the subsurface biomass abundance should be comparable to that of Earth's (Onstott et al., 2006; Dzaugis et al., 2018), and allay the concerns raised about limited oxidant supply to the martian subsurface from the surface (Fisk and Giovannoni, 1999). Radiolysis releases energy into the rock at a rate of 10⁻⁹ kJ/g-yr based upon the parameters utilized by Onstott et al. (2006) of which some fraction is accessible for biomass production depending upon the porosity. This rate is greater than that estimated for weathering reactions and would be even higher on Mars during the Noachian when the radioactive parent isotopes were more abundant. These calculations suggest that adequate energy exists on Mars to support substantial biomass equivalent to that of the rock-hosted biosphere on Earth and that an inhabited Mars would accumulate organic matter over time in subsurface aquifers, completely independent of a habitable martian surface.

3.4.Biodiversity and geography

As is the case for biomass, the species richness of deep subsurface environments is highly variable. Subsurface organisms span the branches of the 16S rRNA phylogenetic tree, including deeply rooted lineages (e.g., Methanomada, Archaeoglobi, Korarchaeota, Thermotogae, and Synergistetes) (Magnabosco et al., 2018a). Reports of deep subsurface planktonic communities dominated by a single archaeal (Chapelle et al., 2002) or bacterial (Chivian et al., 2008) species are rare. More commonly, diversity estimates range from over a hundred (Marteinsson et al., 2013) to almost 100,000 (Bomberg et al., 2016) operational taxonomic units or OTUs (at 97% identity in the 16S rRNA gene) within a single fluid sample. To some extent this reflects the improvement in sequencing technology and the fact that the highly variable subregions of the 16S rRNA gene are targeted. It is not unusual to find a large number of OTUs that comprise <1% of the total population (Castelle et al., 2013; Magnabosco et al., 2014), referred to as the rare biosphere (Sogin et al., 2006). Additionally, single-cell genome sequencing of subsurface Candidatus Desulforudis audaxviator indicates significant differences in the genomes of single species (Labonté et al., 2015).

Species abundance does not correlate with its ability to influence the overall function of a subsurface community. For example, in continental subsurface environments, methanogens frequently comprise only 2% of the total community, but the primary gas phase is biogenic CH₄. In the case of one deep subsurface site in South Africa, this biogenic CH₄ was the principal carbon source for the remainder of the community (Simkus et al., 2015; Lau et al., 2016b). Electron microscopy (Kyle et al., 2008; Middelboe et al., 2011; Engelhardt et al., 2014), single-cell genomes (Labonté et al., 2015), and metatranscriptomic analyses (Lau et al., 2016b) have also revealed that viruses are abundant and actively infecting bacteria and altering their genomes (Paul et al., 2015) in subseafloor sediments and fractured rock aquifers. Because active viral populations can transfer genes between microbial species and control the population density, they may be a vital component of obligately mutualistic metabolic SLiMEs.

One might expect that subsurface diversity would resemble island-like behavior because of the lesser connectedness between subsurface habitats when compared to surface habitats where wind and surface water are transport agents. In island-like ecosystems, the number of species should increase with the size of the island (i.e., volume of groundwater sampled) (Locey and Lennon, 2015). However, Magnabosco et al. (2018a) did not find any such correlation. Species richness may instead correlate with greater heterogeneity of microenvironments, which are difficult to characterize in the subsurface. The lack of species richness versus habitat size could also reflect a surprisingly high degree of connectedness between habitats and motility. This is consistent with the presence of the same rare biosphere OTUs in fractures ranging from 0.6 to 3.0km depth and separated by hundreds of kilometers in South Africa and a general lack of a distance-decay relationship (Magnabosco et al., 2018a). The presence of the same OTUs in fracture water of different isotopic compositions also suggests that these species are actively motile (Magnabosco et al., 2014). The implication of these observations for an early martian subsurface biosphere is that impacts and volcanic activity could have produced zones of sterilized rock, but these zones would have quickly become recolonized by groundwater circulation.

3.5.Subsurface metabolic activity

Estimates of the in situ metabolic rates of subsurface ecosystems offer insight into their longevity and biomass turnover and provide a better sense of how they impact biogeochemical cycles on a planetary scale, though obtaining accurate estimates has proven challenging (Orcutt et al., 2013). Geochemical estimates of the electron production rate from marine subsurface in situ microbial activity range from ~5mol e^- L⁻¹ yr⁻¹ at seafloor hydrothermal vents (Wankel et al., 2011) to ~2 to 100pmol e^- L⁻¹ yr⁻¹ in oligotrophic subseafloor red clays (Røy et al., 2012) to 0.004 to ~4pmol e^- L⁻¹ yr⁻¹ for continental deep fractured rocks and consolidated sediments (Kieft and Phelps, 1997), estimates spanning more than 15 orders of magnitude. Recent metabolism-agnostic approaches utilize isotopic labeling and measurements of the D/L of aspartic acid of bulk subseafloor sediment (Lomstein et al., 2012) and of cells separated from deep continental fracture fluids (Onstott et al., 2013) to determine the bulk rate of growth and repair, often referred to as cell turnover. These analyses yielded cell turnover times ranging from 73,000 years for subseafloor sediments to 1.7–1.8 years for 60°C fracture water and imply that in situ metabolic rates are strongly temperature dependent (Xie et al., 2012; Onstott et al., 2013; Trembath-Reichert et al., 2017). In the case of the 60°C fracture water dominated by a single species of autotrophic sulfate-reducing bacterium, the cell turnover time corresponded to a metabolic rate of 2.6–2.8nM of sulfate per year or 21–22nmol e^- L⁻¹ yr⁻¹ (Onstott et al., 2013). The shorter cell turnover times observed in deep continental fracture water are also consistent with metatranscriptome and metaproteome observations from similar environments that revealed significant intracommunity recycling of metabolic waste products (Lau et al., 2016b). Recycling of biogenic CH₄, sulfide, and CO₂ suggests that initial geochemical approaches for estimating metabolic rates (Phelps et al., 1994) may have underestimated the rates of cell turnover and thus metabolic rates. Despite the challenges posed in accurately estimating the metabolic rates of subsurface microbial ecosystems, recent in situ approaches utilizing natural ¹⁴C (Simkus et al., 2015) suggest that a wide range can be found that correlates with the energy fluxes and temperatures. The longer turnover times in colder ecosystems raise questions about the nature and rate of evolution in subsurface ecosystems, which are also characterized by physicochemical conditions that are more stable over time than surface ecosystems and where microorganisms are exposed to low radiation dosage rates relative to those of surface ecosystems (Teodoro et al., 2018).

4.1.Subsurface life biosignatures in hydrothermally altered ultramafic rocks

Magma-poor paleocontinental margins expose large volumes of mantle peridotites to infiltration by seawater (Whitmarsh et al., 2001). Along the Lower Cretaceous Iberian margin, seismic data indicates 25–100% serpentinization-driven alteration of ultramafic crust to depths of 4km (Dean et al., 2000). The upper ~1km is most heavily altered with serpentinized rocks crosscut by calcite-brucite assemblages, for which isotopic data indicate precipitation at temperatures from 25–40°C. The contact zone is hypothesized to represent the deep plumbing of a Cretaceous “Lost City” type hydrothermal system (Kelley et al., 2005, 2015).

Mineralized veins in the contact zone at depths of ~750m are significantly enriched in organic carbon. Analysis for biosignatures revealed round to rod-shaped structures, ~2–200μm in diameter, which are consistent with the morphologies of microbial colonies. Analyses of these putative fossilized cells with Raman spectroscopy revealed them to be carbon-enriched, with C–H, –CH₂, and –CH₃ functional groups. Band positions are consistent with lipids, amino acid side chains of proteins and carbohydrates, and amide I bonds in proteins. Further analysis of lipid biomarkers revealed nonisoprenoidal dialkylglycerol diether lipids of bacterial origin and acyclic glycerol dibiphytanyl glycerol tetraether lipids of archaeal origin. Thus, hydrothermal activity ~750m beneath the seafloor at ~120Ma sustained very abundant archaeal and bacterial microbial communities, equivalent to ~10⁹ cells cm⁻², within fractures leaving behind morphologic fossils, organic carbon, and lipids (Klein et al., 2015).

Fossilized remains of microorganisms have also been described in carbonate or serpentine veins of ~1Ma ultramafic peridotite rocks in the Mid-Atlantic Ridge, and characterized by a combination of morphology, chemical composition, and the presence of organic matter, sometimes including specific complex amides usually characteristic of biopolymers (Ménez et al., 2012; Ivarsson et al., 2018). In these systems, complex organics are found in either aragonite veins or within a poorly crystalline mix of serpentine, magnetite, and hydrogarnet, and remnant orthopyroxene with chemical enrichments in Ni, Co, Mo, and Mn (Ménez et al., 2012; Ivarsson et al., 2018) that are different compared to microbial preservation in basalt-hosted systems, which are instead dominated by clay minerals or Fe oxides (see Section 4.2 below).

4.2.Fracture-filling fossilized complex subsurface chasmoendolithic and cryptoendolithic communities in igneous rocks and precipitated rocks

Basaltic rocks cored from modern-day continental groundwater circulation sites show that cells are strongly concentrated within clay and oxides assemblages in fractures and pore spaces (e.g., Trias et al., 2017, their Supplementary Fig. 15). Encrustation of biological matter by mineralization is a key means of preserving the structures over geologic time, as cataloged for rocks of multiple ages and types (Hofmann and Farmer, 2000; Hofmann, 2008). Investigations of ancient, now fossilized fracture surfaces in igneous rocks show these are zones of concentration of microbial activity, sometimes including complex communities of organisms with multiple trophic levels, preserved by mineralization.

In seamount basaltic lavas ranging in age from 48 to 81Ma, the fossilized remains of chasmoendolithic (fracture-dwelling) subsurface microorganisms, that is, coccoidal, filamentous or stromatolitic structures with elevated carbon concentration and organic matter such as lipids and rare chitin, are preserved in mixtures of clay, Fe oxides, and Mn oxides and carbonate and gypsum veins (Ivarsson and Holm, 2008; Ivarsson et al., 2009, 2012; Bengtson et al., 2014). The microstromatolitic structures found at 68–153m below the seafloor within the fractured basalt are interpreted as the result of Fe- and Mn-oxidizing bacteria, and based on mineral succession, appear to be the initial colonizers of subseafloor basalt (Bengtson et al., 2014; Ivarsson et al., 2015). These rocks are interpreted to preserve a syntrophic community of chemolithoautotrophs, hyphae-forming fungi, and microstromatolitic Frutexites. Most filamentous and coccoidal fossils have so far been interpreted on morphological characteristics and rare chitin as fungal hyphae and yeast growth stages, respectively (Ivarsson et al., 2012, 2015). This is probably not due to a dominance of fungi in subseafloor crust but instead due to fungi being more easily recognized than prokaryotic fossils. Microstromatolites and associated single-celled features with morphologies comparable to S-cycling archaea like Pyrodictium species suggest prokaryotic remains are present as well (Bengtson et al., 2014; Ivarsson et al., 2015). The organic micron-sized coccoidal shapes occur in concentrations equivalent to ~10⁷ cells cm⁻² on the vein walls with vein-containing tubular ichnofossils (see Section 4.3 below) and with saline fluid inclusions recording entrapment temperatures of ~130°C.

As a second example, in continental flood basalts of Miocene age (17–6Ma) in Oregon, secondary minerals formed within and near fractures preserve ~1μm sized coccoidal and rod-shaped microstructures and framboidal pyrite associated with kerogen (McKinley et al., 2000) in an aquifer where the present-day microbial communities are comprised of sulfate-reducing bacteria (Baker et al., 2003). Iron oxyhydroxides, smectites, zeolites, and silica within fractures and smectite veins were investigated because they contained framboidal pyrite, and the cell-like microstructures were discovered.

As a third example, fossilized remains of biofilms have also been found preserved within calcite-filled fractures of 1800Ma granite at a depth of 200m in the Fennoscandian shield using transmission electron microscopy (Pedersen et al., 1997). Stable isotope, fluid inclusion, and fission track analyses constrain the age of occupation and entrapment between 115ka and 400–300Ma. More recently, Drake et al. (2017a) have discovered fossil biofilms preserved in calcite veins dated at 355±14Ma that occur at 300m depth in a similar Swedish granite and core lipids indicative of sulfate-reducing bacteria extracted from calcite veins from 80 to 920m depth in Swedish granite and that are 358–394 million years in age (Drake et al., 2018). In the same granitic rocks, coupled bacterial sulfate reduction–anaerobic CH₄ oxidation paleoactivity is recorded by the −125‰ to +36.5‰ V-PDB δ¹³C values and diagnostic lipid biomarkers preserved in vein-filling calcite (Drake et al., 2015a) that formed at temperatures <50°C and the −54‰ to +132‰ V-CDT δ³⁴S values in pyrite lining open fractures (Drake et al., 2015b, 2018) over a depth range of 200–750m. These paleobiosignatures are consistent with the present-day observations of a coupled bacterial sulfate reduction–anaerobic CH₄ oxidation zone over similar depth ranges in fracture water from the granite, though in one instance the sulfate-rich zone is above the methane-rich zone (Pedersen et al., 2014), and in the other instance the opposite is true (Hallbeck and Pedersen, 2012).

Excellent preservation of fossilized fungal mycelia have also been reported from granites of the Fennoscandian shield with diagnostic morphologies like anastomosis between branches (Ivarsson et al., 2013). Drake et al. (2017b) have reported a fossilized anaerobic fungi–sulfate reducing bacteria consortium from a 740m deep fracture in granite located at the Laxemar site, Sweden. In particular, fossils of filamentous microorganisms of fungi are within impact-induced fractures in granite from the 89Ma Dellen impact and 458Ma Lockne impact (Ivarsson et al., 2013). Both sites are related to the subsequent hydrothermally formed mineralization, indicating that impact-generated habitats in igneous rock can be favorable for microbial colonization and preservation.

As a fourth example, some organisms, known as “autoendoliths,” play a more active role in the formation of rock edifices, whose precipitation can result directly from microbial metabolism and encapsulate the responsible microbial constituents (Marlow et al., 2015). For example, anaerobic methanotrophs oxidize methane, increase alkalinity, and produce bicarbonate that precipitates as carbonate rock at methane seeps (Peckmann et al., 2001). Metabolic activity continues from within the rock (Marlow et al., 2014b), and biosignatures of the entombed organisms can persist for hundreds of millions of years (Peckmann and Thiel, 2004).

Other examples of fracture-filling filamentous fabrics of rock-hosted life include those preserved in chalcedony and/or zeolite in tens of terrestrial volcanic rocks ranging in age from Tertiary to Mesoproterozoic (Hofmann and Farmer, 2000); filamentous fabrics of what are interpreted as Fe-oxidizing chemotrophic bacteria of Devonian age in calcite veins cross-cutting lacustrine sedimentary rock (Trewin and Knoll, 1999); complex mineralized filamentous structures in Devonian-age pillow basalt (Peckmann et al., 2007; Eickmann et al., 2009); and fungi-like mycelial fossils in vesicles and fractures of 2.4Ga basalt in South Africa (Bengtson et al., 2017). The interpretations that these fossils represent subsurface prokaryotes and fungi are consistent with observations of the present-day subsurface biosphere (see Section 3.1). Nonetheless, as the record is pushed backward, Earth's overprinting processes demand more sophisticated high-resolution analyses for biogenicity determination (see also Section 4.6).

4.3.Microbial trace fossils in recent and ancient glass

Complex, tubular structures that are sometimes found emanating from alteration mineral–filled fractures in basalts, basaltic glass, or impact glass may in some cases be trace fossils of microbial origin, called ichnofossils. These structures were first reported in Pleistocene volcanic glass in Iceland (Thorseth et al., 1992) and have since been documented globally in young seafloor pillow basalt glass (Thorseth et al., 1995; Furnes et al., 1996; Fisk et al., 1998; Fisk and McLoughlin, 2013) and in pillow basalt of ancient ophiolites and greenstone belts dating back 3.5Ga (Furnes et al., 2008). The mechanisms of formation of these features combine dissolution of glass with leaching of cations and formation of clay minerals, Fe and Mn silicates, and Fe and Ti oxides (Staudigel et al., 2008).

In modern rocks, basaltic glass samples from 1.3–1.8km depth from Hawaii Scientific Drilling Program core that were examined with Raman, deep UV fluorescence, and 16S rRNA sequencing show microorganisms are present in clays at the dissolution boundary with the glass near microtubular structures (Fisk et al., 2003). In modern seafloor basalts on the Mohns spreading ridge of the Norwegian Sea, tubular dissolution structures originate from the palagonite-glass interfaces, and evidence for bacterial processing includes the characteristic rounded and elongated, microbial-sized, 0.5–2μm, pores, enrichments of Mn on the rims of coccoid-shaped structures with elevated concentrations of C, N, and organic carbon with a depleted isotopic signature (Kruber et al., 2008; McLoughlin et al., 2011).

Determining biotic versus abiotic origin in fossil rocks requires careful observations of textural subtleties and paired morphological-chemical criteria (e.g., McLoughlin and Grosch, 2015). As with many exothermic inorganic chemical processes that are exploited by chemolithoautotrophic microorganisms, distinguishing structure formed by abiotic reactions from biologically mediated ichnofossils is challenging (Knowles et al., 2012; Grosch and McLoughlin, 2014; Wacey et al., 2017). We describe two of the best ancient examples involving rock-hosted microbial life associated with glass dissolution.

Ichnofossils like the modern examples are found in volcanic glass of the 92Ma Troodos ophiolite (Furnes et al., 2001). Tubular dissolution structures possess 3-D spiral morphologies with organic carbon and nitrogen enriched linings (Wacey et al., 2014). Careful analyses of the relationship between infilling clay minerals and organics shows that the organics formed first by microbial extracellular materials and were then mineralized by clays.

In the 14Ma Ries impact crater, microtubules related to microbial life are found in the impact-generated glass within heavily altered zones with clay minerals and Fe oxides. The carbon-bearing materials have C-H_x and amide I and II absorption bands from organic materials, not observed in tubule-free areas (Sapers et al., 2014). Further analysis with Raman spectroscopy showed quinoic compounds, and STXM coupled with NEXAFS showed Fe redox patterns in these areas consistent with microbially mediated dissimilatory Fe reduction (Sapers et al., 2015).

4.4.Lipid biomarkers of rock-hosted life

Neutral lipid biomarkers have been widely utilized as a biomarker of terrestrial life in sedimentary and petroleum deposits (see review by Brocks and Summons, 2005), and their application to Archaean marine sediments had a significant impact on the understanding of the evolution of prokaryotes and eukaryotes (e.g., Brocks et al., 1999). However, aliphatic and polycyclic lipids in the metamorphosed Archean sediments containing polyaromatic hydrocarbons were later shown to be drilling contamination (French et al., 2015). Unmentioned, and unresolved, is whether some of the low concentrations of bacterial lipids and archaeal isoprenoids found in the rock matrix could in fact have originated from extant and extinct subsurface microorganisms colonizing the rock mass over billions of years.

Analyses of lipid biomarkers in modern marine sediments under anoxic conditions document how rapidly the lipid biomarkers of marine planktonic biomass, both eukaryotic and prokaryotic, are quickly replaced within the water column and the surface seafloor sediment by the bacterial lipids of the subseafloor biosphere (Schubotz et al., 2009). The archaeal lipid half-lives appear to be longer on the order of hundreds of thousands of years (Xie et al., 2012). A geological test for both the age and preservation of subsurface bacterial lipids has been documented in Cretaceous-age marine sediments that were intruded by mafic dikes 3.4 million years ago (Table 1; Fig. 4). Profiles of the phospholipids (active bacteria) and glycolipids (extinct bacteria) both stratigraphically and as a function of distance from the intrusion indicate that the glycolipids result from the decay of phospholipids of subsurface bacteria. These lipids from subsurface bacteria predate the 3.4Ma intrusion and postdate the 44Ma age of burial sterilization (T_max=125°C) of the marine shale (Ringelberg et al., 1997; Elliott et al., 1999). This persistence indicates that study of lipid biomarkers of rock-hosted life warrants considerably more attention as the microbial record of ancient terrestrial rocks is interrogated.

4.5.Putative Archean subsurface versus surface microbial biosignatures

The earliest, commonly agreed upon, preserved microbial structures are in the 3.4–3.5Ga rock units of the Pilbara Craton, Australia. Some units contain microfossils and laminated sedimentary structures consistent with stromatolites and contain contentious carbonaceous biosignatures (e.g., Walter et al., 1980; Buick et al., 1981; Schopf, 1993; Brasier et al., 2006; Noffke et al., 2013) that are suggestive of photosynthetic microorganisms at this time, though their biogenicity has been questioned (e.g., Buick et al., 1981; Lowe, 1994; McLoughlin et al., 2008).

The 3.46Ga Apex chert represents hydrothermal dikes with silica-mineralization containing kerogenous microfossils involved in subsurface cycling of CH₄ (Schopf et al., 2018) and/or organic matter formed by abiogenic mechanisms, for example, Fischer-Tropsch-type synthesis or remobilization of other organics (Brasier et al., 2002, 2005, 2006; García-Ruiz et al., 2003). Other workers have suggested the −56‰ δ¹³C V-PDB value of CH₄ trapped in fluid inclusions of the same veins could be the earliest evidence of methanogenesis in the subsurface (Ueno et al., 2006).

Sulfur isotope fractionation between sulfides and barite in sediments and crosscutting veins (Shen et al., 2009) and within seafloor basalt and komatiite (Aoyama and Ueno, 2018) of the Dresser Formation suggest that sulfate-reducing bacteria were also present and metabolically active at the near surface and subsurface by ~3.46Ga. Microfossils preserved in chertified Strelley Pool arenite and pyrite with negative δ³⁴S V-CDT values also suggest the presence of subsurface sulfate-reducing bacteria at 3.43Ga (Brasier et al., 2015). Pyritic filaments preserved in the 3.24Ga deep sea volcanogenic massive sulfide deposits within the Sulphur Springs Group provide the most convincing evidence of early life in the form of thermophilic, chemotrophic prokaryotes living in hydrothermal systems beneath the seafloor (Rasmussen, 2000).

Putative microfossils in the form of mineralized tubular features in basaltic glass from the 3.47–3.46Ga upper Hooggenoeg Formation of the Barberton Greenstone Belt, containing isotopically light carbonate, have been proposed as subsurface biosphere fossils (Furnes et al., 2004; Banerjee et al., 2006). Titanite mineralized microtubules in 3.35Ga basaltic glass with a minimum age of 2.9Ga have also been suggested to represent evidence of biological processing (Banerjee et al., 2007). Very negative δ³⁴S V-CDT values from pyrite within these microtubules suggest that they were formed by microbial sulfate reduction (McLoughlin et al., 2012). More recently, however, Grosch and McLoughlin (2014) disputed the biogenicity of the microtubule textures suggesting they represent contact metamorphic textures associated with postdepositional intrusions (Table 1; Fig. 3).

Even more controversial are the earliest traces of life from 3.95–3.75Ga amphibolite-grade metamorphic rocks in Greenland and Labrador in the form of “biogenic graphite” (Mojzsis et al., 1996; McKeegan et al., 2007; Tashiro et al., 2017) and a single graphite inclusion in 4.1Ga zircons from Jack Hills, Australia (Bell et al., 2015) that yield negative δ¹³C V-PDB values comparable to those of modern life. Determining whether they represent primary organic matter versus secondary organic matter formed during later metamorphism is challenging (Papineau et al., 2011), and isotopic values do not determine whether they formed by biological fractionation or abiotic processes (van Zuilen et al., 2003; Sherwood Lollar et al., 2006), let alone whether they represent “chemofossils” of phototrophic or subsurface chemoautotrophic microbial biomass. Other evidence for Archean life is based upon textural evidence such as putative centimeter-scale stromatolites in 3.7Ga metacarbonates in Greenland (Nutman et al., 2016) and tens-of-micron-scale hematite filaments in 4.2Ga metasedimentary rock in Quebec (Dodd et al., 2017). The former features, however, were recently refuted as true stromatolites (Allwood et al., 2018).

In summary, the paucity of low metamorphic grade Archaean rock record hampers our ability to identify unambiguous biosignatures older than ~3Ga, which is approximately the time frame at which Mars' broad-scale habitability began to decline. Putative morphological biogenic structures, combined with C and S isotopic evidence, have been preserved in volcanics, dikes and quartzites that are consistent with subsurface life (methanogenesis and sulfate reduction) while those found in marine sediments are suggestive of photosynthetic life. Noteworthy is a quote from the work of Brasier et al. (2015), as follows: “Why are few cellular fossils found in rocks before 2.5Ga? For decades, the main search image has been cyanobacteria-like assemblages as silicified algal mats and stromatolites. Have we been looking for fossils in the wrong places?” (Brasier et al., 2015). In light of new insights on the magnitude of the rock-hosted biosphere on Earth, it seems clear that while we were not looking in the wrong places, the taphonomic windows and environmental settings investigated for the biomarkers of ancient life should be expanded. The same can be said, therefore, for the >50% of the surface of Mars that is older than ~3Ga. As it is unmetamorphosed relative to Earth, it represents a particularly compelling exploration frontier (see Section 5).

This work has greatly benefited from the discussions and intellectual contributions of the 20 participants in the Rock Hosted Life Workshop, February 8–10, 2017, at Caltech as well as the >100 participants in the four pre-workshop telecons open to the community. Thanks to Mary Voytek and Michael Meyer at NASA Headquarters for workshop funding and to NASA Astrobiology Institute director Penny Boston and the NASA Ames Research Center meeting support team for providing web-hosting for the telecons and their recording; all the information can be found at http://web.gps.caltech.edu/~rocklife2017. We thank partial support of T.C.O. through a subcontract supported by NASA Exobiology Award NASA NNX17AK87G to Andrew Schuerger of the University of Florida and partial support to B.L.E. by the NASA Mars Data Analysis Program award 80NSSC17K0444. The contribution of M.C. was carried out partly at the Jet Propulsion Laboratory (JPL), California Institute of Technology, under contract with the National Aeronautics and Space Administration (NASA), and via Grant NNA13AA94A issued through the NASA Science Mission Directorate and supported by the NASA Astrobiology Institute.