Mariana serpentinite mud volcanism exhumes subducted seamount materials: implications for the origin of life

The subduction of seamounts and ridge features at convergent plate boundaries plays an important role in the deformation of the overriding plate and influences geochemical cycling and associated biological processes. Active serpentinization of forearc mantle and serpentinite mud volcanism on the Mariana forearc (between the trench and active volcanic arc) provides windows on subduction processes.  Here, we present (1) the first observation of an extensive exposure of an undeformed Cretaceous seamount currently being subducted at the Mariana Trench inner slope; (2) vertical deformation of the forearc region related to subduction of Pacific Plate seamounts and thickened crust; (3) recovered Ocean Drilling Program and International Ocean Discovery Program cores of serpentinite mudflows that confirm exhumation of various Pacific Plate lithologies, including subducted reef limestone; (4) petrologic, geochemical and paleontological data from the cores that show that Pacific Plate seamount exhumation covers greater spatial and temporal extents; (5) the inference that microbial communities associated with serpentinite mud volcanism may also be exhumed from the subducted plate seafloor and/or seamounts; and (6) the implications for effects of these processes with regard to evolution of life. This article is part of a discussion meeting issue ‘Serpentine in the Earth system’.


Introduction
and, if Fe-Ni alloys are present, more complex hydrocarbons [13]. The hydrocarbons released from these reactions are important for sustaining metabolic processes of microbial communities found in the mudflows from the Mariana serpentinite mud volcanoes [4,5,14,15].
Our recent findings suggest that the recycling of subduction channel constituents may include microbial communities from the subducting plate that survive sterilization because of possible temperature variations within a high-relief subduction channel. Towards the end of this paper, we present arguments discussing the environment of serpentinization within the suprasubductionzone mantle and its potential as a locale for the early Earth's development of life.

(b) Geologic setting
The western Pacific seafloor, east of the Mariana Trench, has hundreds of seamounts close to the trench [13] (figure 1). As the Pacific lithospheric plate and features on it approach the trench, lithospheric bending creates an outer trench bulge about 60-120 km east of the trench. This decreases the effective elastic plate thickness [16][17][18] so that normal faulting creates uplifted blocks with intervening troughs and/or trenchward, down-stepping half-grabens. These features are generally oriented parallel to the trench axis [19]. Small seamounts on the subducting plate are strongly affected by outer-rise deformation, but large ones remain essentially intact [20]. Impinging plate seamounts make first contact with exposed mantle at the inner slope of the Mariana Trench, which is the 'type' non-accretionary convergent plate margin [21]. Here, little or no sediment accumulates at the leading edge of the overriding plate and the suprasubductionzone mantle wedge is exposed below approximately 6 km on the inner trench slope. Igneous and sedimentary rocks of the forearc are exposed at shallower depths, sparsely capped by sediments. Uplift of the forearc in addition to along-and across-strike extension has created faults on many scales over the approximately 50 Myr history of convergence [22,23]. Deep faults, penetrating to the subduction channel [20], provide pathways for fluids from the down-going plate to hydrate and serpentinize the forearc mantle [4,5,[24][25][26]. Comminuted fault gouge, mobilized by the ascending fluids, creates a mush that rises through fault-controlled conduits to the sea floor, and debouches to form huge serpentinite mud volcanoes within about 100 km west of the trench [4,5]. The Mariana forearc serpentinite mud volcanoes episodically [3,4] tap the subduction channel at a range of depths and temperatures (table 1), providing windows into aspects of physical, chemical and biological processes that affect the subducted Pacific Plate as it moves downward and the overlying forearc lithosphere [5].

New observation of seamount subduction
In June 2016, we recorded the first observation of a large undeformed seamount that is exposed in the inner slope of the Mariana Trench. The only previous sighting of Cretaceous reefal material from a subducted seamount was in the late 1980s on a series of Nautile submersible dives in the deepest 20 km of the western inner slope of the Japan Trench [27]. This was adjacent to the impinging Pacific Plate's Daiichi Kashima Seamount. These observations were consistent with the suggestion that because of the strong deformation of subducting plates proximal to trenches, and because of loss of buoyancy, subducting Pacific Plate seamounts will be rotated and sheared as they enter a subduction channel [28]. This is not at all what we observed on the deep inner slope of the Mariana Trench.
We  Office of Exploration and Research (OER) 2016 'Deepwater Exploration of the Marianas' cruise [1,2]. On Dive 4 (at approx. 20.5°N) [1], we observed a 275-m-high sequence of exposed reef, starting at 5 995 m (figure 2a). These reef sequences are morphologically identical to those we observed on Dive 16 [2], during which we drove the ROV up a summit escarpment of Fryer Guyot (approx. 20.5°N), a Cretaceous-aged, flat-topped seamount immediately east of the trench (figure 2b,c). We interpret the Dive 4 exposure to be part of a Cretaceous reef complex of a seamount that is partially subducted beneath the edge of the overriding Mariana forearc. The reefal sequences on Dive 16 show numerous in situ fossils (figures 3a-c) including bivalves. White layers that have a rudist morphology (S. Stanley, 2017, personal communication) alternate with tan bands of other reef fossils (figure 3a-c). Similar sequences occur on Dive 4 (figure 3d-f ). Slump scars observed on Dive 4 indicate mass wasting, but there is no indication of shearing, rotation or any large-scale deformation. The ROV left the bottom at a depth of about 5 720 m on the inner trench slope to perform a mid-water survey, but the traverse of the subducting seamount ended without seeing the upper margin of the reef exposure.

Mariana forearc deformation and mud volcano distribution
In general, contact with the overriding plate edge at the initial subduction of high-relief features leads to uplift and erosion of the overriding plate and formation of adjacent deep basins or grabens [29][30][31]. Bathymetric relief in the Mariana forearc is significant within about 100 km of the trench [32]. Several broad ridges, shallower than the average regional depth along strike, cross the   entire forearc (figure 1). They lie collinear with the convergence direction of thicker Pacific Plate crust [33] and with clusters of Pacific seamounts (figure 1). Similar morphologic effects occur on other convergent margins where subduction of ridges drive thickening of forearc crust that is commensurate in dimensions with that of the impinging ridge (e.g. Cocos Ridge [34]). The uplifted Mariana forearc from 19.5°N to 21°N (figure 1) is shallower by 2-3 km than the general 4 km deep bathymetry along strike at about 100 km west of the trench axis. Both north and south of this ridge, the seafloor has deep basins (greater than 5 km) [35]. There are narrower ridges that extend across the forearc: one from approximately 17.5°to 19°N; one from approximately 16°t o 16.75°N; and an outer forearc ridge from approximately 15.25°to 15.5°N. All of these are bounded by anomalously deep seafloor (figure 1). Faulting related to vertical tectonic deformation, to east-west extension caused by the Pacific Plate's eastward roll-back from the trench, and to north-south extension caused by the forearc's increase in curvature with time, is extensive and distribution of serpentinite mud volcanism on the Mariana forearc is related to fault lineaments [4,[35][36][37]. These serpentinite mud volcanoes erupt episodically over millions of years [35]. It is likely their eruptions are triggered by release of slab-derived fluids in association with seismic events [4,5,36,37]. The mudflows are composed largely (greater than 90%) of clay-to sand-sized serpentine grains with pebbles and larger clasts of variably serpentinized forearc mantle and forearc crustal rocks [4,5,25].

Forearc mantle serpentinization
The potential degree of serpentinization of the Mariana forearc mantle by slab-derived fluids has been suggested to reach 100% over the approximately 50 Myr history of subduction at the Mariana convergent margin system [38] peridotites from the forearc show that many are only partially serpentinized [22]. Thus it is likely that the greatest degree of serpentinization is constrained to fracture-related pathways (e.g. [25,26]). The results of Ocean Drilling Program (ODP) coring of serpentinite mud volcanoes are consistent with this interpretation.
Drill cores on Conical Seamount included both summit and southeast flank sites and there were subtle systematic variations between them [42]. Of particular interest is the variation in degree of serpentinization in flank Site 779, which generally decreases down hole to a depth of approximately 125 m below sea floor (mbsf), but increases from approximately 170 mbsf to 235 mbsf. These types of variation are suggestive of differences in source regions for the mudflows, and analysis with X-ray diffraction of core samples from Hole 779A indicates at least six separate mudflow units [41]. The ultramafic clasts at the summit hole, which was drilled into the conduit of the mud volcano, are more highly serpentinized (61% to 100%) harzburgite, and have nearly uniform composition, but have an increase in degree of serpentinization and alteration with depth [42], suggesting a slow but continuous rate of percolation of pore fluids within the conduit. Very slow fluid emanation from carbonate chimney structures was noted in Alvin dives on Conical Seamount's summit [22].
The summit of South Chamorro Seamount (figure 1) was drilled at Site 1200 on ODP Leg 195 [43]. The ultramafic fraction of the rock clasts recovered in the serpentinite mudflow matrix was dominated by harzburgite with minor dunite. There was a single fragment of lherzolite. The degree of serpentinization (40% to 100%) of these rocks at Site 1200 and the range in textures were essentially similar to those observed on Conical Seamount, although there was little diversity in degree or type of serpentinization at Site 1200 [43]. All the Site 1200 drill holes were confined to a small region at the summit of the mud volcano in order to locate an optimal site for a cased hole. Post-cruise analyses of the rock clasts and mudflow matrix material indicated that the matrix was formed from the clasts by diminution within the conduit [44]. The matrix material reflects a dominant ultramafic composition with about a 10% addition of a subducted plate component [44].
There is little compositional difference, in general, between the Leg 195 ultramafic clasts versus those from Leg 125 with regard to Al 2 O 3 , Fe 2 O 3 , MnO and Na 2 O abundances, although the Leg 195 ultramafics have slightly higher loss on ignition (LOI), MgO, Cr and Ni [45]. The ultramafic rocks from both sites are interpreted to derive from the suprasubduction-zone mantle wedge and have experienced melt extraction at up to 25% to 30% [42][43][44][45][46].
Serpentinite mudflow matrix material from a wash core taken on ODP Leg 195 at Hole 1200B was wet-sieved for the 60-µm mesh fraction and had a far greater variety of metabasites (metamorphosed basalts) than were recovered in cores from ODP Leg 125 sites on Conical Seamount [43]. These amounted to a slightly greater total volume (approx. 12%) than those from Conical Seamount (approx. 10%) of this size fraction. Similar metabasites from dredged and cored (gravity and push cores from ROVs and submersibles) are consistent with low-to high-pressure, and low-to moderate-temperature origins [5,50].

IODP expedition 366 metabasites and reef limestones (a) Metabasite compositions
Analytical methods used on the analysis of the metabasite rocks from all three seamounts drilled on IODP Expedition 366 (12/2016 to 02/2017) [3,53] are given in the electronic supplementary material associated with this paper. The IODP Expedition 366 Sites are located closer to the trench than the ODP Sites on Conical and South Chamorro Seamounts. All of the drill sites recovered metamorphosed basalt, with on-board analyses indicating mid-ocean ridge basalt (MORB) and/or island arc tholeiite (IAT) provenance [3,53]. IODP Expedition 366 lavas from Asùt Tesoru and Fantangisña Seamounts (figure 1) have V/Ti ratios in the OIB lava field (figure 4 and table 2). The depletion of V relative to Ti is a function of the fugacity (f o 2 ) of a magma and its source, the degree of partial melting, and subsequent fractional crystallization [54]. Most IODP Expedition 366 metabasites are enriched in light rare earth elements relative to the heavy rare earths, but one is light-rare-earth depleted, similar to MORB or forearc basalt (figure 5a and table 3 [55]). MORB-normalized multi-element plots show that most samples have OIB-like patterns (figure 5b and table 3), with one similar to MORB (U1498B 21R-1) (compare with [56]). The negative Sr anomalies in the OIB-like samples indicate plagioclase fractionation, whereas the enrichments in fluid-mobile elements (K, Rb) indicate low-temperature alteration. These characteristics point to magma genesis from a locally enriched mantle source such as is abundant in Pacific Plate seamount provinces, but absent in the Mariana forearc [57,58]. The paragenesis of metamorphosed rocks of Pacific Plate origin collected from IODP Expedition 366 cores spans the gamut of metamorphic facies from zeolite [43], to greenschist [59], to blueschist [40,50,60] (figure 6).

(b) Limestone compositions
Limestone breccia and cherty-limestone fragments were also recovered from the IODP Expedition 366 cores [53]. Cores from Yinazao Seamount recovered the first pieces of reef material (figure 7) from a subducted Pacific Plate seamount from one of the forearc serpentinite mud volcanoes. The reefal cobbles recovered from Core U1491C-2H-CC-1-7 cm are, in fact, the first pieces of reef material ever collected from a subducted plate seamount in the world (figures 1 and 5a,b). The cobble analysed in detail is a Miogypsina rudstone with larger lithoclasts and coralline, red-algal grainstone matrix. The matrix includes echinoderm, bryozoan and decapod fragments, small benthic and a few planktonic foraminifera. The lithoclasts consist of packstone with large bioclasts such as a Halimeda plate or a Porites fragment. The latter is encrusted by a sessile foraminifer and     It is possible that the eastern half of the Mariana forearc mantle has been serpentinized over broad areas, lacks cohesion and has a low coefficient of friction when permeated by fluids at depth [62,63]. It may be faulted sufficiently so as to deform readily during incursion of a subducting edifice. Thus, the contact zone between subducted plate and forearc wedge must be lithologically complex and may have significant relief.

ICP-OES
The forearc west of the location of ROV Dive 4 is up to 4 km shallower than the average depth along strike. There is no multi-channel seismic (MCS) reflection data in this region, but MCS profiles across other areas of near-trench forearc uplift have failed to show evidence of underlying subducted seamounts [64], even where IODP Expedition 366 drill cores recovered reefal limestone clasts in serpentinite mudflows.

(b) Exhumed reef fragments
Assuming a Pacific Plate convergence rate of approximately 3.5 cm per year [65,66] for the southern Mariana forearc region, the hypothetical position of the seamount from which the reef cobble (in Core U1491C) derived was 400 to 700 km east-southeast of the trench in early to middle Miocene times when the equatorial Pacific carbonate compensation depth was lower than today [67]. Most likely its position was outside the outer-trench-rise. Using the current Pacific Plate convergence rate, and back-tracing approximately 550 km, the seamount would have been located in the area of today's Micronesia atolls where appropriate shallow water conditions still occur. Because the cobble from Core U1491C was recovered at a depth of over 4500 m below sea level, and from within a serpentinite mudflow, it must have been brought up from greater depth in the subduction channel. The depth to the subduction channel from MCS data at this location is 14 km below sea level [64]. The largest of the seamounts and guyots east of the Mariana Trench are approximately 4-5 km taller than the surrounding sea floor. Assuming that the seamount was subducted with minimal reshaping, the reefal cobble could have been derived from approximately 9-10 km beneath the forearc sea floor before being entrained in rising serpentinite mud. Temperatures and pressures at such a depth would be less than 80°C and approximately 0.3 GPa (table 1).
The exhumation of seamount lavas and unaltered reefal fragments containing shallow-water fauna thus indicates that materials from shallower in the subduction channel than the presumed abyssal depths of the subducting Pacific sea floor can be recycled through the forearc. Clearly, such materials would have experienced lower pressures and temperatures than the abyssal  [55] and an average ocean island basalt (OIB) [52]. Normalizing values from Sun and McDonough (1989) [55]. For data, see table 2.
seafloor during subduction and would likely be below the sterilization temperature and depth conditions for microbial communities extant within these materials.
The discovery of an intact expanse of a Cretaceous reefal section in the deep inner slope of the Mariana Trench makes incorporation of subducted seamounts into nonaccretionary forearc regions indisputable. The inclusion of pieces of oceanic plate and plate seamounts in serpentinite mudflows (figure 8) in the Mariana forearc indicates that rocks can be derived at any depth, from the contact with abyssal sea floor, from fault zones that border contact between subducted seamounts and adjacent forearc faults, or from regions where dismembered seamount fragments have been left behind as the main edifice is subducted.  overriding plate. Results of several modelling studies [71][72][73] and high-resolution bathymetric data [74], indicating forearc deformation linked with subduction of seamounts and ridges, have been argued both ways. If we examine the distribution of seismicity from 1900 until 2017 at depths less than 50 km in the Mariana forearc (figure 1), we see that seismic events are located near some serpentinite mud volcanoes, but not near others. Thus, the currently known spatial association of these events with known serpentinite mud volcanoes is not sufficient to identify a temporal relationship between serpentinite mud volcano activity and likely processes of seamount subduction. A temporal relationship between large clusters of earthquakes within the subduction channel and shallow events near a mud volcano has, at least in one locality, been demonstrated [4], however. The spatial relationship between forearc fault scarps or lineaments and the distribution of serpentinite mud volcanoes is well known [5,32,35,37,41,42,51,75]. At shallow levels in the forearc crust, multi-channel seismic data [76] show considerable faulting within forearc sediments in the 100 km proximal to the trench. A greater degree of faulting at millimetre to metre scales was observed during Deep Sea Drilling Project Leg 60 in cores at Sites 458 and 459 on the eastern Mariana forearc at approximately 18°N [77]. More recent multi-channel seismic data [75] provides evidence of faulting within the upper crustal sequences of the Mariana forearc, but there are no data that define faults within the forearc mantle wedge. An important approach to assessing the stress state of the subduction channel, on a fine scale, is by measuring the over-pressuring of fluids in the subduction channel as a forcing function to stimulate seismic activity. Within the Mariana forearc, over-pressure is transmitted in fluids released from the subducting plate initially through pore collapse and then by dehydration of hydrous minerals. Numerous studies have suggested a relation between seamount subduction and over pressuring of forearc fluids [28,29,31,68,70]. A borehole observatory emplaced in ODP Hole 1200C recorded pressure fluctuations that were triggered by two large (Mw 7+) earthquakes during a two-year monitoring period [78].
Including ODP Hole 1200C on South Chamorro Seamount, there is now a transect of borehole observatories on summits of four serpentinite mud volcanoes, each penetrating an active hydrologic zone. Once the IODP Expedition 366 cased boreholes are fitted with CORK-Lites [79] it will be possible to instrument the seamounts to monitor a range of physical and hydrologic properties, correlate them with regional seismicity, and conduct manipulative experiments to assess active processes within the conduits of the mud volcanoes and the underlying subduction  Rock clasts entrained in serpentinite mudflows provide the best evidence to evaluate the regional state of stress of the upper Pacific Plate in the Mariana subduction channel and the overlying forearc lithosphere. At a contact between a metabasite and a cherty limestone (figure 8), in core U1498B-21R-1 through U1498B-21R-2, we observed cataclastic faults and extensional structures (high-angle normal faults and extensional veins) (figure 8). These may be related to: (1) bending of the lower plate during subduction; (2) deformation within the subduction channel; and (3) accretion and vertical tectonism of pieces of the subducted plate and seamounts into the forearc region. Where seamounts are subducted, the likely heterogeneity of structural and stress environments may limit the dimensions of fracture propagation [29]. Stress events are likely frequent, leading to seismic creep within the zone of contact between subducting seamounts and the surrounding forearc.

(d) Morphology of the subduction channel
The admixture of fragments of shallow reef limestones, alkalic OIB lava fragments, and forearc lavas within serpentinite mudflow drill cores permits us to speculate that the contact between the subducting edifices and the surrounding forearc has higher relief than previously conceived. Although the maximum pressure-temperature ranges of the metabasites can be estimated based on distances from the trench axis, the broad range of metamorphic conditions of rock clasts from within each mud volcano indicates that metabasites can be plucked from considerable depth ranges along the conduit walls, including the abyssal depths of the subducted plate or the flanks of a seamount edifice subducting with it. In contrast, dredges that were collected on a scarp closer to the trench, near Conical Seamount, recovered low-pressure/low-temperature (greenschistfacies) metamorphosed basalts of these same provenances, as well as chert and limestones [47,81]. If the leading edge of the overriding plate is weak, the contact zone between the impinging edifice and the forearc will be the preferred zone of weakness for the development of a mud volcano conduit. Conduit development would be most likely to occur at the contact between a relatively strong edifice structure and a weak surrounding serpentinized forearc mantle wedge (figure 8).

(e) Implications for microbial exhumation
If the depth traversed during the exhumation of a mudflow unit in a given eruptive episode varies greatly, a conduit at the edge of a subducted seamount could tap into a range of pressure and temperature conditions at the boundary between forearc mantle and a subducting edifice. The largest of the Pacific Plate seamounts east of the trench are flat-topped guyots, as much as 100 km in diameter and, as the reefal material recovered indicates, were close to sea level in the Miocene. So, they would have been close to 4-5 km high. The region of the forearc wedge, on which the serpentinite mud volcanoes rest, is from approximately 13-19 km thick (table 1). A large subducting seamount could perturb the thermal structure of the outer forearc wedge into which it would have been subducted. We suggest that although the adjacent, deep subducting basaltic crust may experience temperatures in excess of the limits for life, shallower, cooler segments of the down-going plate, such as the flanks and summit of a subducted seamount, could still harbour viable microbial communities. Such communities, when they contact fluids associated with serpentinizing forearc mantle, would likely shift toward those in favour of organisms best suited to the high pH and reducing conditions in these fluids (e.g. [82][83][84][85]), and thus become dominant members of the community. These then would persist near the active summits of the serpentinite mud volcanoes today. This scenario is consistent with observations of the type of microbial communities observed thus far at the summit springs on several Mariana serpentinite mud volcanoes [13,15].
The results of IODP Expedition 366 show that the distribution of serpentinite mud volcanoes that erupt fragments of Pacific Plate and its seamounts extends over 600 km along-strike and over 90 km across-strike of the southern Mariana forearc. There are also serpentinite seamounts atop a ridge 50 km wide proximal to the Izu-Bonin Trench that extends for the entire length of that convergent margin, though they are apparently not currently active [42].

(f) Implications for the early earth environment
Although the Mariana serpentinite mud volcanoes are the only known currently active sites of serpentinite mud eruptions, deposits of serpentinite materials of similar type are observed world wide in convergent margin deposits with ages as old as Early Paleozoic (approx. 480 Ma) [4,5,86,87]. The occurrence of serpentinite deposits in the Isua formation of southwestern Greenland that are similar to the Mariana forearc serpentinite mudflows [88], are of particular note, and they are of Eoarchean (3.81 to 3.70 Ga) age. Thus, the phenomenon of convergent margin serpentinite mud volcanism has an apparently long and spatially broad history in the Earth system.
Ever since hydrothermal vents on mid-ocean ridge systems, teeming with life, were discovered, the focus on a chemosynthetic, hydrothermal origin for life on Earth has concentrated on the oceans. Over the past few decades, research into the environmental conditions conducive for life's origin has increasingly suggested these could have occurred in close association with serpentinization [89][90][91][92][93][94][95][96][97][98][99][100][101][102][103]. Many of these researchers have studied environments near mid-ocean ridges. Bada [104] has suggested this would be a problem if hyperthermophilic organisms do not form the root of the tree of life. Furthermore, it is established [105,106] that the high temperatures of hydrothermal systems would have precluded sufficient temporal stability of prebiotic macromolecules necessary for the development of living cells on the Hadean/Archean Earth. Most responses to this argument have been that, although at high temperature vents, conditions for the evolution of life would have been unlikely, only millimetres away the temperatures would have been cool enough to permit it. Even so, it is possible that the inherently episodic and ephemeral nature of hydrothermal systems [107] imposed a temporal constraint on the likelihood of life's beginning at Earth's hydrothermal systems. Furthermore, longevity and dispersal of the earliest life forms from hydrothermal vent environments would have been impinged upon by proximal hydrographic and regional geologic impediments, as they are today [108]. Some researchers [109][110][111][112] prefer models in which subaerial, shallow hydrothermal systems with fresh-water and drying/hydrating cycles would help to increasingly concentrate potential reactants. These too may suffer the difficulties with stability of progenitor molecules as in other hydrothermal systems. Relatively cooler, but temporally and spatially more stable environments in the early history of the Earth, where serpentinization was prevalent, would be a likely alternative. Ancient subduction zones are an obvious possibility, if some form of plate tectonics existed in antiquity.
Arguments continue regarding when plate tectonics first began on the Earth (e.g. see reviews [113,114]), and thus when plate margins (subduction zones and mid-ocean ridges) and associated volcanism first formed. Zones of lithospheric plate convergence, represented today by some ancient (Archean: 4-2.5 Ga) greenstone belts exist throughout the continents. Furthermore, the results of δ 18 O analysis of zircon crystals, with ages of up to 4.4 Ga [115,116], indicate that surface temperatures were cool enough in the late Hadean (4.4-3.82 Gya) for liquid water to have existed on Earth's surface and thus for life to have evolved [117]. Arguments regarding the possibility of initiation of plate tectonics in the Hadean (greater than 4 Ga) require, at minimum, surface water underlain by a dry mantle [113]. If subduction processes were occurring on the early Earth, serpentinization in suprasubduction channels would have been possible.
Today as we observe mid-ocean ridges, we see their segmentation varying on scales from tens to hundreds of kilometres [118,119]. Today's trenches stretch for thousands of kilometres (e.g. Mariana-Izu-Bonin, Peru-Chile and Aleutian Trenches) and have persisted for tens to hundreds of millions of years. They are the more temporally stable plate tectonic features, are cooler regimes and can provide a widespread contiguous serpentinization environment.

(g) Subduction zones and the origin of life
The Isua Formation of southwestern Greenland is garnering increasing interest as a suggested Eoarchean intraoceanic convergent margin [120,121] setting. The discovery of serpentinite deposits that are interpreted to indicate serpentinite mud volcanism in an Isua Formation forearc setting [88] pushes the formation of such mud volcanoes back to near the Late-Hadean/Early-Archean boundary. The chemical and physical characteristics of today's serpentinite mud volcanism in the Mariana forearc conform to many of the suggested requirements for optimal conditions of the development of life. The fluids emanating from these mud volcanoes are freshened relative to seawater [24]. The eruptive pattern is episodic at each edifice, and they persist over time scales of tens of millions of years [4,5,35]. The muds contained in the upper, active portions of the conduits are bathed in vent fluids during each eruptive cycle, but may undergo periods when there is no fluid flux. These dormant periods can concentrate precipitates within the porous mudflows. The conduit regions currently support active microbial communities [14,15]. As we have shown here, the conduits of all sampled thus far contain a broad range of clast lithologies. The rising muds, therefore, have the potential to contain a wide variety of reactants. Thus, if such mud volcanism can be considered a suitable serpentinite environment for the beginnings of chemosynthetic life on Earth, how early could this process have occurred?
Oxygen isotopic compositions of Hadean (4.4 Ga) zircons [115][116][117] imply that temperatures of the early Earth's surface were such that oceans could have existed. Data support fluid-flux melting of the mantle at least by Eoarchean (3.7 Ga) time [121]. The early formation of continental crustal masses [113][114][115][116][117][118][119][120][121], and of putative off-ridge plate seamounts formed by localized pressurerelease melting or hot-spot processes, and their collision at convergent plate margins argues for the potential subduction of high-relief features. If so, then serpentinization juxtaposed to fault bounded, protocontinental or seamount structures in ancient subduction-zone mantle could have been possible anywhere along thousands of kilometres of convergent plate margins at the Hadean/Archean boundary.
Subduction of high relief protocontinents or seamounts would have had the potential to form complex subduction channels, and to result in vertical tectonic deformation in forearc regions, deep penetrating faults, and contact between serpentinized mantle and compositionally complex features on the subducting plate. Thus, development of egress channels for deep-derived fluids, and the eruption of mudflows on the sea floor, composed of comminuted and remobilized serpentinite and fault gouge with fragments of subducted lithospheric lithologies, could have occurred. We agree that it is valid to consider serpentinization environments as ones likely to have promoted the transition from prebiotic conditions to the evolution of living cells. However, we suggest that envisioning such environments as limited to regions of hydrothermal systems on mid-ocean ridges, or in hot springs settings on land, is too narrow a view.

Conclusion
Our observation of an intact section of Cretaceous reef exposed in the Mariana Trench inner slope shows, for the first time, that large expanses of impinging seamounts can remain undeformed at first contact. It also confirms uplift of the surrounding forearc region. We observe that forearc relief is enhanced where thicker subducting crust or large seamounts impinge, thus adding to the vertical tectonic deformation of the forearc region. ODP and IODP drill cores from five serpentinite mud volcanoes contain a range of subducted plate and seamount lithologies. The mudflow matrix composition and the broad variation in degree of metamorphism of the rocks recovered suggest that the relief of the subduction channel beneath the Mariana forearc is likely much higher than generally thought.
High relief in the subduction channel would affect the temperature and pressure conditions, as well as mineralogical and pore fluid compositions of muds that erupt at the active Mariana forearc serpentinite mud volcanoes. A given eruption may involve materials plucked from conduit walls at a variety of depths. This can explain the wide range of degree of serpentinization of ultramafic clasts contained in the mudflows and the wide range of metamorphic facies and degree of deformation observed in metabasites derived from the down-going Pacific Plate and subducted seamounts. If subducting seamounts pass through the shallow to intermediate depths of the subduction channel relatively intact, microbial communities present in the sub-seafloor on their flanks or within the abyssal Pacific Plate's lithospheric slab could also be exhumed from varying depths. Serpentinization of forearc mantle fault gouge, in contact with rising fragments of the subducted plate, likely selects for those microbes that prefer the resulting high pH (up to 12.5) and reducing conditions of the serpentinizing pore fluids [13].
Researchers have suggested a variety of serpentinization environments that may be supportive of the earliest formation of life [82][83][84][85][89][90][91][92][93][94][95][96][97][98][99]. These mainly stress hydrothermal vent environments. The focus on mid-ocean ridges and off-ridge hydrothermal regions results from discovery of vast accumulations of vent fauna in the oceans today. The focus on subaerial hydrothermal pools near volcanoes stresses the hydration/dehydration potential for concentrating prebiotic molecules. However, hydrothermal activity is sporadic and ephemeral on geologic time scales along mid-ocean ridges, and is spatially miniscule by comparison with the 70% global expanse of the ocean floor. The environment of shallow (approx. 0-20 km) subduction at convergent plate margins is cooler, more contiguously widespread and supports optimum geochemical conditions over geologically greater timescales. So, perhaps we must ask what other geologic settings would have been conducive to enabling the transition from the milieu of probiotic molecules to the development of living cells. If we consider the possibility of the presence of an ocean on the early (4.5 Ga) Earth and the advent of cyclic crustal generation and consumption via some early form of plate tectonic activity, then the temporal continuity and spatial breadth of convergent plate margins is likely far greater than that of hydrothermal vent locations. As we ponder the most favourable environments for the development of chemosynthetic life on the basis of serpentinization, either on Earth or other planets, it would be well to consider processes other than those just of hydrothermal systems.
Even discarding speculations about ancient environments, the causes of and conditions during current, active episodes of escape/venting of material from the subduction channel in any phase of plate subduction remain uncertain. If seismic events trigger episodes of eruption at modern serpentinite mud volcanoes, is it possible to take the pulse of subduction? Cased boreholes left at the summits of four of the drilled ODP and IODP Sites at summits of four serpentinite mud volcanoes provide a regional laboratory for the study of geophysical, geochemical and geobiological processes associated with the subduction channel beneath the Mariana forearc. The planned and funded deployment of instrumentation in these seafloor observatory holes will permit further analysis of interrelationships between seismic activity, pore fluid flux, eruption mechanics and effects on microbial communities of the serpentinite mud volcanoes. The results will impact current paradigms of lithospheric deformation, mass cycling and physical conditions at the plate contact zone and help define potential constraints on the mechanism by which subsurface microbes could be exhumed from the subduction channel and 'recycled' without being subjected to sterilizing conditions. Data accessibility. The analytical methods text and an additional table with data supporting the major and trace element data in this article have been uploaded as part of the electronic supplementary material.
Authors' contributions. All authors who were shipboard participants of IODP Expedition 366 took active part over the 62-day IODP Expedition 366 in the descriptions, analyses and interpretations of the core materials collected aboard ship and wrote preliminary reports aboard ship for each site cored. P. interpretation of video observations. P. Fryer, D. Amon, D. Glickson and S. Pomponi described samples taken on board the NOAA R/V Okeanos Explorer during these legs of the Deep Water Exploration of the Marianas expedition and wrote final dive reports for the legs on which they participated. C. Kelley supervised the realtime public participation of scientists from around the world during both Legs of the NOAA Exploration of the Marianas expedition. P. Fryer is chiefly responsible for the interpretive aspects of the discussion section of the paper that deal with the subduction-related, early Earth ecosystem. All authors have been given the opportunity to review the text of this manuscript submission and provide comments as needed on the text.
Competing interests. We declare we have no competing interests.