Letter Doi:10.1038/Nature25031

Letter doi:10.1038/nature25031

The divergent fates of primitive hydrospheric water on Earth and Mars Jon Wade1, Brendan Dyck2,3, Richard M. Palin4, James D. P. Moore5 & Andrew J. Smye6

Despite active transport into Earth’s mantle, water has been surface basalts6,7 and the widespread distribution of oxidized hydrated present on our planet’s surface for most of geological time1,2. phyllosilicate minerals7. The disassociation of water during such Yet water disappeared from the Martian surface soon after its silicate-mineral hydration reactions and the subsequent loss of hydro- formation. Although some of the water on Mars was lost to space gen to space, for example, may account for the presence of oxidized via photolysis following the collapse of the planet’s magnetic field3–5, magnetite13 and maghemite in the Martian meteorite NWA7034 the widespread serpentinization of Martian crust6,7 suggests that and the Gusev crater rocks it resembles 14, while the source region of metamorphic hydration reactions played a critical part in the Martian SNC (shergottite–nakhlite–chassignite) igneous meteorites sequestration of the crust. Here we quantify the relative volumes is much more reduced14,15. of water that could be removed from each planet’s surface via the Aside from the proportion of iron, the terrestrial and Martian mantles burial and metamorphism of hydrated mafic crusts, and calculate have broadly similar chondritic major-element compositions 16,17. The mineral transition-induced bulk-density changes at conditions abundance of iron in planetary mantles—and hence erupted surface of elevated pressure and temperature for each. The metamorphic material—is primarily controlled by the prevailing oxygen fugacity mineral assemblages in relatively FeO-rich Martian lavas can of planetary accretion, which was higher during the formation of hold about 25 per cent more structurally bound water than those the Martian core than during terrestrial core formation. As a result, in metamorphosed terrestrial basalts, and can retain it at greater terrestrial basalts are relatively FeO-poor (about 7–10 wt%)18 compared depths within Mars. Our calculations suggest that in excess of to more FeO-rich (about 17 wt%)19 basalts derived from the Martian 9 per cent by volume of the Martian mantle may contain hydrous upper mantle. As bulk-rock composition is the primary control on mineral species as a consequence of surface reactions, compared to the mineral assemblages that stabilize during surface hydration and about 4 per cent by volume of Earth’s mantle. Furthermore, neither metamorphism at elevated pressure (P) and temperature (T) condi- primitive nor evolved hydrated Martian crust show noticeably tions, these compositional differences probably played a critical part in different bulk densities compared to their anhydrous equivalents, determining each planet’s bulk geochemistry and surface water budget in contrast to hydrous mafic terrestrial crust, which transforms over geological time. to denser eclogite upon dehydration. This would have allowed Here, we have quantified the balance between mantle hydration efficient overplating and burial of early Martian crust in a stagnant- and density in both terrestrial and Martian interiors using integrated lid tectonic regime, in which the lithosphere comprised a single thermal and petrological modelling. We show that metamorphosed tectonic plate, with only the warmer, lower crust involved in mantle primitive and evolved Martian basalts can hold about 25% more struc- convection. This provided an important sink for hydrospheric water turally bound water in hydrous minerals than terrestrial equivalents, and a mechanism for oxidizing the Martian mantle. Conversely, with the majority stable to greater depths within Mars (>90 km) than relatively buoyant mafic crust and hotter geothermal gradients on within Earth (about 70 km). In addition, time-integrated mass-balance Earth reduced the potential for upper-mantle hydration early in calculations show that the absolute volume of the Martian mantle that its geological history, leading to water being retained close to its has potentially been hydrated over geological time (>9%) is at least surface, and thus creating conditions conducive for the evolution twice as large as that of the terrestrial mantle (about 4%). of complex multicellular life. To model the surface evolution of Earth and Mars, we considered a Surface water has existed on Earth for the vast majority of geological mantle potential temperature (TP) of 1,650 °C for both the terrestrial 1 time , with its volume having remained approximately constant since and Martian primitive basalts (although this TP was reached at different the end of the Archaean (2.5 billion years (Gyr) ago) and initiation of times on each planet), and a lower bound of 1,400 °C for evolved rocks. subduction-driven plate tectonics2. However, despite evidence for it SNC meteorites indicate that the rate of secular cooling on Mars was having once possessed liquid water, the modern-day surface of Mars conservatively about three times faster than that of Earth20. We com- is essentially dry and lacks any widespread features of active plate bined these assumptions with a plate-cooling model, taking into tectonics 8. The loss of surficial water from Mars since its formation account secular cooling and radiogenic heat production to calculate has been attributed to both photolysis in the upper atmosphere and aerotherms and geotherms through the lithospheres of each planet. sequestration into the crust via metamorphic hydration reactions9. We then investigated the petrological changes that would occur in Although loss to space by atmospheric sputtering and hydrodynamic the end-member cases of primitive and evolved, fully hydrated and escape probably became increasingly viable after cessation of the fluid-undersaturated crust during each planet’s geological evolution Martian magnetic field at about 4.1 Gyr ago3–5, uncertainties in its using thermodynamic phase equilibrium modelling at conditions efficiency, the timing of field collapse, and the initial water inven- defined by these P–T gradients. For Mars, we used bulk compositions tory suggest that this process alone may not have depleted the entire for primitive (Fastball) and evolved (Backstay) basalts, and for Earth volume of the hydrosphere over the lifetime of the planet10–12. Instead, we used a high-MgO Archaean tholeiitic basalt and a modern-day petrological evidence for crustal hydration is present as serpentinized normal mid-ocean-ridge basalt (N-MORB) (see Methods). We

1Department of Earth Sciences, University of Oxford, South Parks Road, Oxford OX1 3AN, UK. 2Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, UK. 3Department of Earth Sciences, Simon Fraser University, Burnaby, British Columbia V5A 1S6, Canada. 4Department of Geology and Geological Engineering, Colorado School of Mines, Golden Colorado 80401, USA. 5Earth Observatory of Singapore, Nanyang Technological University, Singapore. 6Department of Geosciences, Pennsylvania State University, University Park, Pennsylvania 16801, USA.

a 2.5 2.5 High-MgO Archaean basalt Fastball 2.0 Modern-day 2.0 terrestrial N-MORB Fastball 1.5 1.5 Backstay

1.0 1.0

Backstay High-MgO 0.5 Archaean basalt 0.5 Modern-day terrestrial Mass of water in hydrated column (wt%)

Water content of hydrated protolith (wt%) N-MORB

0 0 35 40 45 50 55 60 65 70 75 80 85 90 12345678910 Depth within planetary interior (km) Absolute volume of hydrated planetary mantle (%) b 10 Figure 2 | Degree of mantle hydration on Earth and Mars as a function of dewatering of metamorphosed basalts. N-MORB is normal 8 mid-ocean-ridge basalt. 6

4 amphibole to temperatures in excess of 1,200 °C, where they are pre- dicted to become Ti-rich, matching observed compositions from Positively buoyan t 2 Martian meteorites22. Although hydration extends to similar absolute depths on both bodies, the shallower Martian aerotherm implies that 0 the volume fraction of silicate Mars that may be hydrated following –2 High-MgO crustal subduction is about double that of Earth (Fig. 2). Additionally, Archaean basalt the faster cooling of Mars20 progressively extends the depth of hydra- –4 Modern-day terrestrial N-MORB tion, exacerbated by the presence of more-evolved basalts overplating the Martian surface and the negative Clapeyron slope for –6 Fastball Density de cit of hydrated column, δ U (%)

Backstay Negatively buoyant hornblende dehydration. The effect that hydration has on bulk-rock –8 density heightens this effect; metamorphosed terrestrial basalts are all 35 40 45 50 55 60 65 70 75 80 85 90 significantly less dense—and hence buoyant—when fluid-saturated, Depth within planetary interior (km) and undergo densification upon dehydration (Fig. 1b). The relatively Figure 1 | Calculated petrophysical properties for each modelled Fe-rich Martian materials are, however, calculated to exhibit rela- basaltic protolith. a, Plot of the structurally bound water content tively little volume expansion during hydration, with more-evolved (in weight per cent, wt%) in metamorphosed primitive and evolved basalts showing minimal change in bulk-rock density, and significant basalt on Earth and Mars. b, The relative density deficit between a crustal densification (>5%) of a water-bearing restite, post-melting. Here, the column composed of fully hydrated and nominally anhydrous metabasalt main reaction products of high-pressure hornblende dehydration at − (δρ = (ρanhydrous ρhydrated)/ρanhydrous). the fayalite–magnetite–quartz buffer are clinopyroxene, plagioclase, garnet, magnetite and silicate melt. considered metamorphism at pressures of 8–26 kbar (Earth) and Hydrated Martian basalts generate relatively high degrees of melt 3–10 kbar (Mars), and temperatures of 500–1,200 °C. Hydration at (about 30%) at moderate temperatures (about 800–900 °C), leaving lower-temperature conditions is controlled by phases belonging to the a hydrous residual assemblage of neutral or negative buoyancy. This serpentine, phyllosilicate and clay mineral groups, with higher water suggests that, unlike on Earth, hydrated materials could have been contents. Phase diagrams for each rock type, considering fluid-satu- transferred into its mantle via sagduction or crustal delamination23,24 rated conditions, are presented in Extended Data Figs 1–4. (Fig. 3). Mars may therefore have undergone extensive hydration of its Calculated water contents and bulk-rock densities for each terres- mantle simply by magmatic overplating, burial and metamorphism trial and Martian protolith as a function of depth within the planetary of hydrated surface basalts. This mechanism may also have been interior are shown in Fig. 1. Unlike terrestrial modern N-MORB, responsible for delivering oxidized material to the source region of the higher water-carrying potential of metamorphosed high-MgO the Martian surface basalts14. On Earth, the increased buoyancy of Archaean terrestrial basalts is primarily a result of their more refrac- hydrated Archaean greenstones and steeper geotherm limited the tory nature, with the absence of quartz allowing water-bearing water-carrying capacity of overplated material. The secular evolution amphiboles to persist to higher temperatures 21, and thus greater of plate tectonics and the transition from early-Earth high-MgO basalts depths in the mantle (Fig. 1a). A similar trend characterizes Martian to modern-day N-MORB basalts (with an inherently lower water- materials, with metamorphosed primitive basalt (Fastball) being carrying capacity)25 further limits the mechanisms of water transport comprised mostly of hornblendic amphibole at subsolidus P–T con- into the terrestrial interior. Mars shows the opposite trend, as its secular ditions. Evolved Martian melts (Backstay-type compositions) contain cooling, high water-carrying capacity of surface basalts, and the neutral a smaller proportion of water-bearing minerals at these temperatures, buoyancy of hydrated materials acted to preferentially sequester water but retain their water to greater depths. In both cases, the net water- into its interior. bearing potential of these Martian materials exceeds those of their On Earth, surface water is transported into the upper mantle terrestrial counterparts. primarily by K-amphibole26, which typically forms in hydrated MgO- The decomposition of amphibole during prograde metamorphism rich metabasalt. In contrast, more ferruginous Martian basalts stabilize marks the terminus of hydrous mineral stability in both terrestrial iron-rich hydrous phases at equivalent pressure and temperature and Martian materials. Indeed, evolved Martian basalts may retain conditions, such as hornblende and biotite (Extended Data Fig. 5).

a Early Mars b Early Earth 500 500

45 35 Hornblendite 600 50 600 Hornblende 40 pyroxenite SOLIDUS 55 700 Olivine 700 Garnet metagabbro websterite 45 60 50 800 65 800 Hornblende SOLIDUS granulite 55 70 SOLIDUS

900 Depth (km) 900 Depth (km) SOLIDUS 60 Temperature (°C) 75 Temperature (°C)

1,000 80 1,000 Garnet Two-pyroxene websterite 65 granulite 85 Eclogite 70 1,100 Websterite 1,100 90 Websterite Eclogite 75 1,200 100 1,200 Nominally Minimally Nominally Minimally anhydrous uid-saturated anhydrous uid-saturated Figure 3 | Schematic cross-sections through primitive Martian and and Mars are based on calculated phase assemblages. Orange (a) and terrestrial mafic crust. a, Mars. b, Earth. Lithological components of yellow (b) details represent melt. metamorphosed hydrated and nominally anhydrous mafic crusts on Earth

The decomposition of hornblendic amphibole during metamorphism 1. Peck, W. H., Valley, J. W., Wilde, S. A. & Graham, C. M. Oxygen isotope ratios and is controlled by pressure, temperature and oxygen fugacity, with rare earth elements in 3.3 to 4.4 Gyr ago zircons: ion microprobe evidence for high δ18 O continental crust and oceans in the early Archean. Geochim. lower temperatures and reducing conditions specifically favouring its Cosmochim. Acta 65, 4215–4229 (2001). stability27. Owing to its smaller size, Martian aerotherms have higher 2. Korenaga, J. Thermal evolution with a hydrating mantle and the initiation of dT/dP gradients than geotherms on Earth, which suggests that hydrous plate tectonics in the early Earth. J. Geophys. Res. Solid Earth 116, http://doi. org/10.1029/2011JB008410 (2011). phases, such as amphibole, comprise a greater proportion of its mantle. 3. Lillis, R. J. et al. An improved crustal magnetic field map of Mars from electron Rock-hydration reactions may therefore be important both as a surface reflectometry: highland volcano magmatic history and the end of the martian water-loss mechanism and as a means of transporting oxidized material dynamo. Icarus 194, 575–596 (2008). 4. Acuña, M. H. et al. Global distribution of crustal magnetization discovered by to upper-mantle depths. However, as hydration reactions are typically the Mars Global Surveyor MAG/ER experiment. Science 284, 790–793 (1999). associated with a decrease in bulk-rock density, re-entrainment of 5. Dubinin, E. et al. Effects of solar irradiance on the upper ionosphere and crustal material into the mantle—either via burial or subduction—may oxygen ion escape at Mars: MAVEN observations. J. Geophys. Res. Space Physics 122, 7142–7152 (2017). be limited until dehydration has occurred. 6. Ehlmann, B. L., Mustard, J. F. & Murchie, S. L. Geologic setting of serpentine In the absence of subduction-driven plate tectonics early in each deposits on Mars. Geophys. Res. Lett. 37, http://doi.org/10.1029/2010gl042596 planet’s geologic history28, both Earth and Mars would have experi- (2010). 7. Carter, J., Poulet, F., Bibring, J. P., Mangold, N. & Murchie, S. Hydrous minerals enced significant basaltic surface overplating. Elevated mantle potential on Mars as seen by the CRISM and OMEGA imaging spectrometers: updated temperatures would have led to higher degrees of partial melting and global view. J. Geophys. Res. Planets 118, 831–858 (2013). more voluminous, low-viscosity magmas. Erupted29 or emplaced30 in 8. Yin, A. Structural analysis of the Valles Marineris fault zone: possible evidence the presence of water, they may be assumed to hydrate efficiently31. for large-scale strike-slip faulting on Mars. Lithosphere 4, 286–330 (2012). 9. Lammer, H. et al. Outgassing history and escape of the Martian atmosphere Our results show that the ability of a planet to retain surface water and water inventory. Space Sci. Rev. 174, 113–154 (2013). is not simply a function of its bulk composition or size, but also its 10. Wordsworth, R. D. The climate of Early Mars. Ann. Rev. Earth Planet. Sci. 44, accretion history, not least the oxygen fugacities prevailing during core 381–408 (2016). 11. Di Achille, G. & Hynek, B. M. Ancient ocean on Mars supported by global formation. By setting the bulk silicate FeO content, the mantle redox distribution of deltas and valleys. Nat. Geosci. 3, 459–463 (2010). state places a first-order control on the amount of amphibole produc- 12. Carr, M. H. & Head, J. W. Martian surface/near-surface water inventory: tion and stability, and hence on high-pressure water-carrying minerals. sources, sinks, and changes with time. Geophys. Res. Lett. 42, 726–732 (2015). 13. Chassefière, E., Langlais, B., Quesnel, Y. & Leblanc, F. The fate of early Mars’ lost An Earth-like planetary size allows a long-lived steep geotherm, which water: the role of serpentinization. J. Geophys. Res. Planets 118, 1123–1134 in turn promotes the removal of water from hydrated phases close to (2013). the planetary surface, and thus the efficient return of buried water to 14. Tuff, J., Wade, J. & Wood, B. J. Volcanism on Mars controlled by early oxidation of the upper mantle. Nature 498, 342–345 (2013). the surface (Fig. 1). In contrast, Mars exhibits evidence of long-lived 15. Wadhwa, M. Redox state of Mars’ upper mantle and crust from Eu anomalies 32 33 water reservoirs both as extant sub-surface ice and as hydrated crust . in shergottite pyroxenes. Science 291, 1527–1530 (2001). These, together with the elevated FeO contents of the Martian mantle 16. McDonough, W. F. & Sun, S.-s. The composition of the Earth. Chem. Geol. 120, 223–253 (1995). and the smaller planetary size, result in significant water sequestra- 17. Dreibus, G. & Wanke, H. Mars, a volatile-rich planet. Meteoritics 20, 367–381 tion into the Martian mantle via increased amphibole content and the (1985). production of hydrous restite of neutral or negative buoyancy. Taken 18. White, W. M. & Klein, E. M. in Treatise on Geochemistry 2nd edn, Vol. 4, 457–496 (Elsevier, 2014). together, the individual water-carrying capacities of erupted magmas 19. Squyres, S. W. et al. Pyroclastic activity at Home Plate in Gusev Crater, Mars. and their respective fates upon burial are key to understanding the early Science 316, 738–742 (2007). hydrological cycles on Earth and Mars. 20. Lessel, J. & Putirka, K. New thermobarometers for Martian igneous rocks, and some implications for secular cooling on Mars. Am. Mineral. 100, 2163–2171 Online Content Methods, along with any additional Extended Data display items and (2015). Source Data, are available in the online version of the paper; references unique to 21. Hartel, T. H. D. & Pattison, D. R. M. Genesis of the Kapuskasing (Ontario) these sections appear only in the online paper. migmatitic mafic granulites by dehydration melting of amphibolite: the importance of quartz to reaction progress. J. Metamorph. Geol. 14, 591–611 received 27 June; accepted 3 November 2017. (1996).

22. Watson, L. L., Hutcheon, I. D., Epstein, S. & Stolper, E. M. Water on Mars—clues 33. Usui, T., Alexander, C. M. O. D., Wang, J., Simon, J. I. & Jones, J. H. Meteoritic from deuterium/hydrogen and water contents of hydrous phases in SNC evidence for a previously unrecognized hydrogen reservoir on Mars. Earth meteorites. Science 265, 86–90 (1994). Planet. Sci. Lett. 410, 140–151 (2015). 23. Fischer, R. & Gerya, T. Early Earth plume-lid tectonics: a high-resolution 3D numerical modelling approach. J. Geodyn. 100, 198–214 (2016). Acknowledgements J.W. acknowledges receipt of an NERC Independent 24. François, C., Philippot, P., Rey, P. & Rubatto, D. Burial and exhumation during Research Fellowship NE/K009540/1. J.D.P.M. was supported by the National Archean sagduction in the East Pilbara Granite-Greenstone Terrane. Earth Research Foundation (NRF) of Singapore under the NRF Fellowship scheme Planet. Sci. Lett. 396, 235–251 (2014). (National Research Fellow award NRF-NRFF2013-04) and by the Earth 25. Palin, R. M. & White, R. W. Emergence of blueschists on Earth linked to secular Observatory of Singapore, the NRF, and the Singapore Ministry of Education changes in oceanic crust composition. Nat. Geosci. 9, 60–64 (2015). under the Research Centres of Excellence initiative. 26. Williams, Q. & Hemley, R. J. Hydrogen in the deep Earth. Annu. Rev. Earth Planet. Sci. 29, 365–418 (2001). Author Contributions J.W. conceived the idea in discussion with A.J.S. and 27. Gilbert, M. C. Synthesis and stability relations of the hornblende ferropargasite. performed initial calculations. R.M.P. and B.D. performed the petrological Am. J. Sci. 264, 698–742 (1966). modelling. Thermal modelling was performed by J.D.P.M. All authors 28. Carr, M. H. & Head Iii, J. W. Geologic history of Mars. Earth Planet. Sci. Lett. 294, contributed to writing the final manuscript. 185–203 (2010). 29. Hauck, S. A. & Phillips, R. J. Thermal and crustal evolution of Mars. J. Geophys. Res. Planets 107, http://doi.org/10.1029/2001JE001801 (2002). Author Information Reprints and permissions information is available 30. Viviano, C. E., Moersch, J. E. & McSween, H. Y. Implications for early at www.nature.com/reprints. The authors declare no competing financial hydrothermal environments on Mars through the spectral evidence for interests. Readers are welcome to comment on the online version of the carbonation and chloritization reactions in the Nili Fossae region. J. Geophys. paper. Publisher’s note: Springer Nature remains neutral with regard Res. Planets 118, 1858–1872 (2013). to jurisdictional claims in published maps and institutional affiliations. 31. Poli, S. & Schmidt, M. W. Petrology of subducted slabs. Annu. Rev. Earth Planet. Correspondence and requests for materials should be addressed to Sci. 30, 207–235 (2002). J.W. ([email protected]). 32. Mouginot, J., Pommerol, A., Beck, P., Kofman, W. & Clifford, S. M. Dielectric map of the Martian northern hemisphere and the nature of plain filling materials. Reviewer Information Nature thanks T. Usui and the other anonymous Geophys. Res. Lett. 39, http://doi.org/10.1029/2011gl050286 (2012). reviewer(s) for their contribution to the peer review of this work.

Methods nominally anhydrous case (Extended Data Fig. 6). We considered an open-system Petrological modelling procedures. All phase diagrams were constructed using scenario in this work, whereby melt generated during metamorphism can reach THERMOCALC version 3.45i34 and the internally consistent thermodynamic a critical escape threshold and leave the local environment, which is a realistic dataset ds62 (ref. 35; updated 6 February 2012). Modelling was performed in the approximation of a natural system. Differential degrees of melt loss lead to a density Na2O–CaO–K2O–FeO–MgO–Al2O3–SiO2–H2O–TiO2–O2 (NCKFMASHTO) inversion between terrestrial and Martian metabasalts at upper amphibolite-facies compositional system using the following activity–composition relations for or granulite-facies conditions. As shown by our calculations, partial melting begins solid-solution phases: tonalitic melt, clinoamphibole and augitic clinopyroxene36, earlier (that is, at lower temperature) in a hydrated N-MORB than in a nominally orthopyroxene, garnet, biotite, muscovite, chlorite and ilmenite–hematite37, plagio anhydrous equivalent, owing to the abundance of amphibole in the former. In clase and K-feldspar38, olivine and epidote35, and spinel–magnetite39. Pure phases addition, melt generation is more voluminous, and quartz and feldspar are prefer- included albite, rutile, sphene, quartz and water. The initial bulk-rock compositions entially consumed as part of the melt-producing reactions. The density inversion used for modelling of all protoliths were converted from wt% oxides, as reported in for hydrous versus nominally anhydrous N-MORB occurs at the point of first melt each original study (Extended Data Table 1), to mol% oxides, with fluid contents extraction, when the former becomes melt-depleted, and the residual phase assem- adjusted on an individual basis, as described below. The software and data files used blage is entirely anhydrous (Grt + Aug). As bulk-rock density was calculated for to generate the phase diagrams may be downloaded from http://www.metamorph. the entire petrological system (that is, solid plus melt, before extraction), melt loss geo.uni-mainz.de/thermocalc. from the initially hydrated protolith causes the remaining solid residue to become For modelling under water-present conditions (Extended Data Table 2), bulk- denser than the initially nominally anhydrous equivalent, which has not yet accu- rock fluid contents were individually fixed such that each lithology was mini- mulated sufficient melt to cause a drainage event. Density differences between mally saturated at the intersection of the relevant geotherm/aerotherm and its both studied Martian basalts are also a result of bulk-rock composition and melt solidus. Here, we define minimal saturation as the equilibrium-phase assemblage fertility, with the relatively high alkali content of the Fastball protolith producing containing 0.5 mol% of free water at the point of first melting. Additional water low-density feldspars in anhydrous conditions, but relatively dense amphibole in was present within the bulk-rock composition in each case as structurally bound a water-rich environment. water within hydrous minerals. For nominally anhydrous conditions, the absolute Thermal modelling procedures. The thermal model was constructed analytically amount of water within each bulk-rock composition was fixed at a total of 0.5 mol% as a solution to the one-dimensional heat flow problem in a plate57, including an (before melt loss), with these bulk compositions shown in Extended Data Table 3. additional source term to account for radiogenic heat production58,59 Invariably, free water was absent from calculated assemblages, with all bulk-rock ∂Tx(,t)(∂2Tx,)t  tt−  water at the solidus being structurally bound within hydrous phases. For internal = κ + A exp− 0  2 0   consistency and ease of comparison, calculations for all rock types used a molar ∂t ∂x  τr  3+ bulk-rock X(Fe ) = [Fe2O3/(Fe2O3+FeO)] value of 0.1 (see refs 40 and 41). where κ is the thermal diffusivity, t0 is today, and A0 is the present-day radiogenic Uncertainty on the absolute positions of assemblage field boundaries in pressure– heat production rate with time constant τr. We apply the boundary conditions temperature space generally do not exceed ±1 kbar and ±50 °C for low-variance equilibria at the 2σ level42,43. Such variation is largely a function of propagated Tx(0==,)t 0 uncertainty on end-member thermodynamic properties within the internally  tt− 0  consistent data set. However, as all phase diagrams were calculated using the same Tx(,==Lt)eTp xp−   τ  dataset and activity–composition relations, similar absolute errors associated with s dataset end-members cancel, and calculated phase equilibria are expected to be where Tp is the mantle potential temperature, τs is the secular cooling time constant, 42,43 relatively accurate to within ±0.2 kbar and ±10–15 °C. and L is the lithospheric thickness. Initial conditions at the time of lithospheric The petrological effects of open-system melting and melt loss were modelled generation take into account the secular cooling using manipulation of the read-bulk-info matrix44. We assumed that melt loss was a  −  cyclical process, with drainage events occurring each time a rock’s melt proportion == −tt0  45,46 Tx(,tT0) pexp  reaches a rheologically critical threshold. The melt-extraction threshold for  τs  intermediate and felsic magmas is around 20%–30%, and we used the upper limit of this range (30%) as our transition point. Of this melt 10% was assumed to remain with the same parameters as the boundary conditions above. The solution may in the source rock following each drainage event, given that a minimum propor- be expressed as a Fourier series tion of about 8% melt must be present in a rock in order to overcome the liquid − 45 −tt0 x  percolation threshold . From a computational standpoint, this was implemented Tx(,tT)e= τs   p   by reducing the calculated modal proportion of the melt phase in the read-bulk- L tt−  nn−−0 22π κt info matrix from 0.3 to 0.1 at each relevant pressure–temperature point along the ∞  A0r((−−1) 1)τ eeτr − 21nxπ  () relevant geotherm or aerotherm. + ∑ sin  π   −π22κτ Theory and application of petrological modelling. The petrological modelling n=1 n L  1 n r  (1) used here relies upon a thermodynamic-equilibrium model of metamorphism,  tt− nn−−0 22π κt  in which it is assumed that equilibrium on a local scale is maintained among Tpκ(1−−)eτs e  47–50 mineral, fluid and melt phases during the evolution of the rock . A model is ++−πnt22κ () Tpe 22  used to represent the thermodynamics of each phase, comprising properties for 1 −πn κτs   its end-members combined with activity–composition relations to describe their  51 mixing in solid solution . This allows calculation of the equilibrium composi- of which we evaluate the first 200 terms to produce each geotherm and aerotherm. tions and proportions of phases (mineral, melt and aqueous fluid) in the stable To explore the uncertainty on the radiogenic heat production and 52 assemblage, at given pressure, temperature and bulk composition . Equilibrium secular cooling rates, we sampled 3,240 geotherms and 3,780 aerotherms, conditions in such a system require the equality of chemical potentials of each with a range of cooling rates on Earth of 20 Gyr < τs < 45 Gyr and on component in all phases, and must conform to the constraints of mass balance. Mars of 10 Gyr < τs < 40 Gyr, and radiogenic heat production rates of −3 −3 Such calculations are frequently performed in metamorphic geology, with the help 0.1 μW m < A0 < 1.0 μW m . This resulted in a range of mantle potential 34,52 of appropriate software , to construct pseudosections: phase diagrams that map temperatures for early Earth of 1,500 °C < Tp < 1,600 °C, for late Earth of 20,60 out the thermodynamically stable assemblages that would form in a particular 1,300 °C < Tp < 1,400 °C, for early Mars of 1,600 °C < Tp < 1,700 °C and for late 50,53 61 bulk-rock composition across pressure–temperature space . This calculation Mars of 1,350 °C < Tp < 1,450 °C. We then took the mean and 99% confidence procedure and the activity–composition relations used are mature, and have been interval of 150 million-year-old lithosphere for each family of thermal models, demonstrably effective in both forward (predictive) and inverse (descriptive) petro- equivalent to a Tp of 1,550 °C/1,650 °C and 1,350 °C/1,400 °C for the primitive and logical modelling studies (see refs 53–55 for representative examples). All phase evolved rocks of Earth and Mars, respectively, as the input for the petrological abbreviations are from ref. 56, alongside ‘L’ or ‘Melt’ for silicate melt. modelling and error estimates. Calculated bulk-rock density profiles. Calculated bulk-rock densities considered To assess the validity of a one-dimensional model when used as an approxima- all phases in the equilibrium assemblage (solids, aqueous fluids and melt) at each tion for a spherical shell we compared the steady-state (τs → ∞, t → ∞) solution pressure–temperature point. Metamorphism of terrestrial N-MORB in a hydrous of equation (1) to the steady-state spherical shell solution system produces a relatively low-density metamorphic product, owing to the abundance of water leading to the formation of amphibole and biotite (Extended ()RL−−()Rr Tr()= Tp (2) Data Fig. 5). In contrast, relatively dense garnet and pyroxene form in the Lr

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. RESEARCH Letter where R is the planetary radius, L the lithospheric thickness, and r the radial 46. Rosenberg, C. L. & Handy, M. R. Experimental deformation of partially melted position. We found that the error obtained using a one-dimensional approach granite revisited: implications for the continental crust. J. Metamorph. Geol. 23, compared to a spherical shell is 0.9% for Earth and 1.77% for Mars, which is well 19–28 (2005). 47. Khorzhinskii, D. S. Physicochemical Basis of the Analysis of the Paragenesis of within the uncertainty on the other parameters defined above. Minerals (Consultants Bureau, 1959). Numerical simulations for the temperature structure were also calculated a 48. Thompson, J. B. Local equilibrium in metasomatic processes. Res. Geochem. 1, posteriori from the results of the petrological modelling to account for depth- 427–457 (1959). dependent variations in the thermal diffusivity κ due to density variations in 49. Guiraud, M., Powell, R. & Rebay, G. H2O in metamorphism and unexpected the bulk mineral assemblages. We found that the maximum deviation in the behaviour in the preservation of metamorphic mineral assemblages. < J. Metamorph. Geol. 19, 445–454 (2001). thermal structure due to compositional effects to be 1.5% for times longer than 50. Powell, R., Holland, T. & Worley, B. Calculating phase diagrams involving solid 10 million years and thus this could also be safely neglected, in agreement with solutions via non-linear equations, with examples using THERMOCALC. other recent work62. J. Metamorph. Geol. 16, 577–588 (1998). Data availability. All relevant data are included in the paper and the Extended 51. Connolly, J. A. D. Multivariable phase diagrams—an algorithm based on Data. generalized thermodynamics. Am. J. Sci. 290, 666–718 (1990). 52. Connolly, J. A. D. Computation of phase equilibria by linear programming: a tool for geodynamic modeling and its application to subduction zone 34. Powell, R. & Holland, T. J. B. An internally consistent dataset with uncertainties decarbonation. Earth Planet. Sci. Lett. 236, 524–541 (2005). and correlations. 3. Applications to geobarometry, worked examples and a 53. Palin, R. M. et al. High-grade metamorphism and partial melting of basic and computer program. J. Metamorph. Geol. 6, 173–204 (1988). intermediate rocks. J. Metamorph. Geol. 34, 871–892 (2016). 35. Holland, T. J. B. & Powell, R. An improved and extended internally consistent 54. Carson, C. J., Powell, R. & Clarke, G. L. Calculated mineral equilibria for thermodynamic dataset for phases of petrological interest, involving a new eclogites in CaO-Na2O-FeO-MgO-Al2O3-SiO2-H2O: application to the Pouébo equation of state for solids. J. Metamorph. Geol. 29, 333–383 (2011). Terrane, Pam Peninsula, New Caledonia. J. Metamorph. Geol. 17, 9–24 36. Green, E. C. R. et al. Activity-composition relations for the calculation of partial (1999). melting equilibria in metabasic rocks. J. Metamorph. Geol. 34, 845–869 (2016). 55. White, R. W., Powell, R., Holland, T. J. B. & Worley, B. A. The effect of TiO2 and 37. White, R. W., Powell, R., Holland, T. J. B., Johnson, T. E. & Green, E. C. R. New Fe2O3 on metapelitic assemblages at greenschist and amphibolite facies mineral activity-composition relations for thermodynamic calculations in conditions: mineral equilibria calculations in the system K2O-FeO-MgO-Al2O3- metapelitic systems. J. Metamorph. Geol. 32, 261–286 (2014). SiO2-H2O-TiO2-Fe2O3. J. Metamorph. Geol. 18, 497–511 (2000). 38. Holland, T. & Powell, R. Activity-composition relations for phases in petrological 56. Kretz, R. Symbols of rock-forming minerals. Am. Mineral. 68, 277–279 calculations: an asymmetric multicomponent formulation. Contrib. Mineral. (1983). Petrol. 145, 492–501 (2003). 57. Parsons, B. & Sclater, J. G. An analysis of the variation of ocean floor 39. White, R. W., Powell, R. & Clarke, G. L. The interpretation of reaction textures in bathymetry and heat flow with age.J. Geophys. Res. 82, 803–827 (1977). Fe-rich metapelitic granulites of the Musgrave Block, central Australia: 58. Driscoll, P. & Bercovici, D. On the thermal and magnetic histories of Earth and constraints from mineral equilibria calculations in the system K2O-FeO-MgO- Venus: influences of melting, radioactivity, and conductivity.Phys. Earth Planet. Al2O3-SiO2-H2O-TiO2-Fe2O3. J. Metamorph. Geol. 20, 41–55 (2002). Inter. 236, 36–51 (2014). 40. Berry, A. J., Danyushevsky, L. V., O’Neill, H. S. C., Newville, M. & Sutton, S. R. 59. Michaut, C. & Jaupart, C. Secular cooling and thermal structure of continental Oxidation state of iron in komatiitic melt inclusions indicates hot Archaean lithosphere. Earth Planet. Sci. Lett. 257, 83–96 (2007). mantle. Nature 455, 960–963 (2008). 60. Herzberg, C., Condie, K. & Korenaga, J. Thermal history of the Earth and its 41. Christie, D. M., Carmichael, I. S. E. & Langmuir, C. H. Oxidation states of petrological expression. Earth Planet. Sci. Lett. 292, 79–88 (2010). Midocean Ridge basalt glasses. Earth Planet. Sci. Lett. 79, 397–411 (1986). 61. Filiberto, J. & Dasgupta, R. Constraints on the depth and thermal vigor of 42. Powell, R. & Holland, T. J. B. On thermobarometry. J. Metamorph. Geol. 26, melting in the Martian mantle. J. Geophys. Res. Planets 120, 109–122 155–179 (2008). (2015). 43. Palin, R. M., Weller, O. M., Waters, D. J. & Dyck, B. Quantifying geological 62. McKenzie, D., Jackson, J. & Priestley, K. Thermal structure of oceanic and uncertainty in metamorphic phase equilibria modelling; a Monte Carlo continental lithosphere. Earth Planet. Sci. Lett. 233, 337–349 (2005). assessment and implications for tectonic interpretations. Geosci. Front. 7, 63. Kato, Y. & Nakamura, K. Origin and global tectonic significance of Early 591–607 (2016). Archean cherts from the Marble Bar greenstone belt, Pilbara Craton, Western 44. White, R. W. Tutorial for Using the Readbulkinfo (rbi) Script. http://www. Australia. Precambr. Res. 125, 191–243 (2003). metamorph.geo.uni-mainz.de/thermocalc/tutorials/#rbi (Johannes Gutenberg 64. Gale, A., Dalton, C. A., Langmuir, C. H., Su, Y. J. & Schilling, J. G. The mean University of Mainz, 2010). composition of ocean ridge basalts. Geochem. Geophys. Geosyst. 14, 489–518 45. Vigneresse, J. L., Barbey, P. & Cuney, M. Rheological transitions during partial (2013). melting and crystallization with application to felsic magma segregation and 65. McSween, H. Y. et al. Alkaline volcanic rocks from the Columbia Hills, Gusev transfer. J. Petrol. 37, 1579–1600 (1996). Crater, Mars. J. Geophys. Res. Planets 111, E09S91 (2006).

Extended Data Figure 1 | Palaeo-Archaean high-Mg basalt pressure– on these values. Some small fields are not labelled for clarity, and others temperature pseudosection. Pressure–temperature pseudosection are numbered and contain assemblages listed to the right of the phase calculated for the bulk-rock composition of Palaeo-Archaean high-Mg diagram. Phase abbreviations are as follows: Ab, albite; Act, actinolite; basalt sample 02MB25663. Some small fields are unlabelled for clarity. Aug, augite; Bt, biotite; Chl, chlorite; Ep, epidote; Gln, glaucophane; Bold dashed line labelled ‘Early-Earth geotherm’ represents the pressure– Grt, garnet; Hbl, hornblende; H2O, aqueous fluid (water); Ilm, ilmenite; temperature profile calculated via thermal modelling, and is enveloped Ms, muscovite; Mt, magnetite; Ol, olivine; Opx, orthopyroxene; by short-dashed lines representing upper and lower confidence intervals Pl, plagioclase; Qtz, quartz; Rt, rutile; Spn, sphene.

Extended Data Figure 2 | N-MORB pressure–temperature modelling, and is enveloped by short-dashed lines representing upper pseudosection. Pressure–temperature pseudosection calculated for the and lower confidence intervals on these values. Some small fields are not bulk-rock composition of N-MORB64. Some small fields are unlabelled labelled for clarity, and others are numbered and contain assemblages for clarity. Bold dashed line labelled ‘Modern-day terrestrial geotherm’ listed to the right of the phase diagram. Phase abbreviations are as listed represents the pressure–temperature profile calculated via thermal for Extended Data Fig. 1.

Extended Data Figure 3 | Fastball (Mars) pressure–temperature enveloped by short-dashed lines representing upper and lower confidence pseudosection. Pressure–temperature pseudosection calculated for the intervals on these values. Some small fields are not labelled for clarity, and bulk-rock composition of Fastball19. Some small fields are unlabelled for others are numbered and contain assemblages listed to the right of the clarity. Bold dashed line labelled ‘Early Martian aerotherm’ represents the phase diagram. Phase abbreviations are as listed for Extended Data Fig. 1. pressure–temperature profile calculated via thermal modelling, and is

Extended Data Figure 4 | Backstay (Mars) pressure–temperature enveloped by short-dashed lines representing upper and lower confidence pseudosection. Pressure–temperature pseudosection calculated for the intervals on these values. Some small fields are not labelled for clarity, and bulk-rock composition of Backstay65. Some small fields are unlabelled for others are numbered and contain assemblages listed to the right of the clarity. Bold dashed line labelled ‘Late-stage Martian aerotherm’ represents phase diagram. Phase abbreviations are as listed for Extended Data Fig. 1. the pressure–temperature profile calculated via thermal modelling, and is

Extended Data Figure 5 | Hydrated terrestrial and Martian basalt aerotherms. Vertical dashed lines represent pressure–temperature points at mineral proportions. Calculated mineral proportions, bulk-rock which melt extraction events occurred (see Methods). Phase abbreviations densities, and water contents during metamorphism of hydrated terrestrial are as listed for Extended Data Fig. 1. and Martian basalts along their respective planetary geotherms and

Extended Data Figure 6 | Nominally anhydrous terrestrial and Martian aerotherms. Vertical dashed lines represent pressure–temperature points at basalts mineral proportions. Calculated mineral proportion and bulk- which melt extraction events occurred (see Methods). Phase abbreviations rock densities during metamorphism of nominally anhydrous terrestrial are as listed for Extended Data Fig. 1. and Martian basalts along their respective planetary geotherms and

Extended Data Table 1 | Bulk compositions of mafic crustal components used in this work (wt% oxides)

N.R., not reported. Data from ref. 63 for sample 02MB256 and from refs 19, 64 and 65 for the Backstay ‘preferred composition’.

Extended Data Table 2 | Bulk compositions used in phase-equilibrium modelling (mol% oxides) under minimally fluid-saturated conditions

3+ 3+ tot Oxygen contents were calculated assuming a bulk-rock X(Fe ) = Fe2O3/(FeO + Fe2O3) = 0.1, where X(Fe ) = (2 × O)/FeO . Stage numbers refer to the number of melt-extraction events, and the associated bulk compositions for numbers ≥1 represent increasingly melt-depleted residua (that is, 0 = initial undepleted bulk composition; 1 = residual bulk composition following one melt-loss event, 2 = residual bulk composition following two melt-loss events and so on).

Extended Data Table 3 | Bulk compositions used in phase-equilibrium modelling (mol% oxides) under nominally dry conditions

Oxygen contents were calculated as for Extended Data Table 2; stage number nomenclature is the same as for Extended Data Table 2.

reinfection at sites where pathogens will often need to be fully defined. Therapeutic targeting 1. Akondy, R. S. et al. Nature 552, 362–367 (2017). 6 2. Youngblood, B. et al. Nature 552, 404–409 first enter the body . Whether these different of the DNA-modification machinery, such as (2017). types of memory T cell arise through the same Dnmt3a, might prove to be a useful strategy to 3. Kaech, S. M. & Cui, W. Nature Rev. Immunol. 12, pathway remains to be determined. increase vaccine efficacy. ■ 749–761 (2012). These two studies suggest that a goal of 4. Winter, D. R., Jung, S. & Amit, I. Nature Rev. Immunol. 15, 585–594 (2015). vaccine design should be to stimulate a large, Kyla D. Omilusik and Ananda W. Goldrath 5. Gray, S. M., Kaech, S. M. & Staron, M. M. Immunol. robust response from the effector T cells from are in the Division of Biological Sciences, Rev. 261, 157–168 (2014). which memory T-cell populations can arise. University of California, San Diego, La Jolla, 6. Chang, J. T., Wherry, E. J. & Goldrath, A. W. Nature Immunol. 15, 1104–1115 (2014). However, the conditions that best promote California 92093, USA. effector cells to become memory cells still e-mail: [email protected] This article was published online on 13 December 2017.

PLANETARY SCIENCE a distinctly higher content of iron(ii) oxide (about 17% by weight) than do typical basalts on Earth (about 7–10% by weight), which suggests that oxygen fugacity was higher during Martian water stored Martian core formation than it was during Earth’s core formation. Wade and colleagues show that this compositional difference, underground along with the different geotherms of Mars and Earth, has a key role in the storage and Why did Mars lose so much of its surface water, whereas Earth retained its? transportation of planetary surface water in Models of the evolution of minerals on the two planets suggest one explanation: the crust and mantle of the two planets. the Martian water was drawn into the planetary interior. See Letter p.391 Hydration processes generally expand the crustal volume, making the crust less dense. By contrast, Wade and colleagues’ thermodynamic TOMOHIRO USUI The four rocky planets of the Solar System modelling indicates that the iron-rich Martian (Mercury, Venus, Earth and Mars) are thought basalts undergo small volume expansions dur- bservations of Mars and its geology to have been formed by the accretion of simi- ing hydration. Furthermore, hydrated iron-rich suggest that the red planet once had lar planetary building blocks. This resulted in Martian basalts tend to melt at lower tempera- an Earth-like hydrological cycle that their mantles having broadly similar composi- tures (about 800–900 °C) than does anhydrous Oincluded large lakes or oceans. In contrast to tions of all the major elements, except for iron. mafic rock (about 1,200 °C), and this melting this ancient wet environment, the surface of Metallic iron partitions into the metallic cores leaves relatively dense hydrated residues in Mars today is cold and dry. The transition to of the rocky planets, whereas iron(ii) oxide the mantle. In the apparent absence on Mars this present state is closely linked to the fate accumulates in silicate-rich planetary mantles. of tectonic processes that recycle crust mate- of the planet’s surface water, which is poorly A thermodynamic property known as oxygen rial into the mantle, the authors propose that understood. A substantial amount of surface fugacity, which is a measure of the amount of successive burial of hydrated crust might have water escaped to space from the atmosphere, oxygen present in a mixture, controls iron con- induced hydration of the mantle — such burial in part because of Mars’ relatively low gravity1. tent in planetary basalts. Martian basalts have would have gradually increased the tempera- However, atmospheric-escape models ture and pressure applied to hydrated account for only some of ancient Mars’ crustal basalts, causing them to melt, surface water. On page 391, Wade and therefore to leave hydrated residues et al.2 propose that much of the surface 2.0 in the mantle. water was sequestered underground. Hydration of the Martian mantle

Erupted lava had the chance to 1.5 would also lead to it becoming more interact with surface water to form oxidized. However, geochemical analy- hydrated crust on both Earth and Earth Mars sis of meteorites (known as shergottites) ancient watery Mars. Wade et al. 1.0 formed from young Martian basalts examine the thermodynamic proper- suggest that their source in the mantle ties of the water-bearing ‘mafic’ crusts is less oxidized than Earth’s mantle3. (which largely consist of the rock 0.5 Moreover, the shergottite source is basalt) of each planet, and show that depleted in water (less than 50 parts Martian basalt can hold more water per million)4 relative to Earth’s man- than terrestrial basalt, and can effec- basalt (wt%) content of hydrated Water 0 tle (typically about 100–200 p.p.m.)5. tively transport it to a greater depth 40 50 60 70 80 90 How do these observations fit into below the surface (more than 90 kilo- Depth within planetary interior (km) Wade and colleagues’ suggestion that metres; Fig. 1). The authors also com- the Martian mantle contains more pute the stability of water-containing water than Earth’s? Experiments have Figure 1 | The water content of basalt on Earth and Mars. Wade 6 minerals in hydrated crusts and their 2 shown that the shergottite source is bulk-rock densities along both planet al. report modelling of the surface evolution of Earth and Mars, located at a depth of approximately ets’ geotherms (which describe how which allowed them to estimate the amount of water that can be 100 km, which is near the base of sequestered by basalt rock in the planets’ interiors. They find that temperature varies with depth). They the Martian basalts can store more water (shown as percentage by the hydrated mantle column pro- conclude that the burial of hydrated weight), and at greater depths, than their terrestrial counterparts. posed by Wade and co-workers. The crusts progressively hydrates the inte- This suggests that much of the water thought to have existed on the lower oxidation and hydration of rior of Mars, but that this process does surface of ancient Mars was sequestered underground. (Adapted the shergottite source might there- not work effectively for Earth. from Fig. 1a of ref. 2.) fore be representative of (or place

©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Natu21/28re. All ri gDECEMBERhts reserved. 2017 | VOL 552 | NATURE | 339

RESEARCH NEWS & VIEWS a cap on the state of) the uppermost region volume of the crustal water reservoir. Institute, Tokyo Institute of Technology, of the unhydrated mantle, unless it represents Ground ice might also account for the miss- Tokyo 152-8551, Japan. a local phenomenon. ing water reservoir on Mars8–10. Subsurface e-mail: [email protected] The global surface-water inventory of Mars radar-sounder measurements8 have detected 7 1. Lammer, H. et al. Space Sci. Rev. 174, 113–154 was originally estimated to be about 2 × 10 an anomaly in an electrical property of rocks (2013). 8 3 to 2 × 10 km (ref. 7, and references therein), in the planet’s northern hemisphere, which 2. Wade, J., Dyck, B., Palin, R. M., Moore, J. D. P. & on the basis of the size of the putative ancient implies that massive ice deposits are embed- Smye, A. J. Nature 552, 391–394 (2017). oceans. But estimates of the total water loss ded among or between layers of sediment and 3. Herd, C. D. K. et al. Geochim. Cosmochim. Acta 66, 7 3 2025–2036 (2002). to space are much smaller (less than 10 km , volcanic materials at a depth of 60–80 m. The 4. Usui, T., Alexander, C. M. O’D., Wang, J., Simon, J. I. 7 based on atmospheric-escape models ). The ground-ice model has also been proposed on & Jones, J. H. Earth Planet. Sci. Lett. 357–358, discrepancy between these estimates hints the basis of analyses of hydrogen isotopes in 119–129 (2012). at the existence of a ‘missing’ water reservoir Martian meteorites9 and of crater morphol- 5. Saal, A. E., Hauri, E. H., Langmuir, C. H. & Perfit, M. R. 10 Nature 419, 451–455 (2002). beneath the surface. The widespread distri- ogy . The crater study indicates that the 6. Filiberto, J. & Dasgupta, R. J. Geophys. Res. Planets bution of hydrated materials on the surface subsurface water ice has a volume of about 120, 109–122 (2015). of Mars also implies the existence of a crustal 3 × 107 km3, which is comparable to the size 7. Kurokawa, H. et al. Earth Planet. Sci. Lett. 394, 179–185 (2014). water reservoir, but conventional spectro- of the ancient oceans. Subsurface exploration 8. Mouginot, J., Pommerol, A., Beck, P., Kofman, W. scopic observations are able to see only the will be required to test both the hydrated- & Clifford, S. M. Geophys. Res. Lett. 39, L02202 surface veneer. Wade and colleagues’ thermo- crust and ground-ice theories, and therefore (2012). dynamic modelling, together with remote- to shed light on the evolution of the Martian 9. Usui, T., Alexander, C. M. O’D., Wang, J., Simon, J. I. & Jones, J. H. Earth Planet. Sci. Lett. 410, 140–151 ■ sensing observations, offers a means to work water inventory. (2015). out the depth profile of hydrated materials, 10. Weiss, D. K. & Head, J. W. Icarus 288, 120–147 and to calculate a reasonable estimate of the Tomohiro Usui is at the Earth-Life Science (2017).

CANCER metastasis, such as melanoma and breast cancer, express high levels of VEGF-C and contain a dense network of lymphatic vessels6. Moreover, an increase in VEGF-C expression in tumours Tumour lymph vessels is highly correlated with metastasis to lymph nodes, and with a reduction in survival in indi- viduals with different tumour types, including boost immunotherapy skin, breast and lung cancers7. Does an increase in tumour lymphangio A high level of expression of the growth-factor protein VEGF-C is associated with genesis always promote tumour growth? tumours that have extensive lymph vessels and poor prognosis. It emerges that Fankhauser and colleagues addressed this such tumours are highly susceptible to immunotherapy. question by analysing the role of VEGF-C- driven lymphangiogenesis in tumour growth. They studied mice that express high levels CHRISTINE MOUSSION of these two branches of the vascular system of VEGF-C and provide a model system & SHANNON J. TURLEY by secreting proteins that are members of the for studying cancer. When the researchers vascular endothelial growth factor (VEGF) treated these animals with VEGF-C-blocking dvances in our understanding of family4. antibodies, they observed a surprising tumour immunology have led to The protein VEGF-A promotes the gen- result: although high VEGF-C expression immense clinical interest in harnessing eration of new blood vessels (angiogenesis). is associated with poor tumour prognosis, Aimmune cells to target cancer1. The immune Decades of work aimed at impairing tumour VEGF-C inhibition in the context of tumour system’s killer (cytotoxic) T cells can search for growth by blocking angiogenesis has led to immunotherapy resulted in increased tumour and destroy abnormal cells that they encoun- the approval of drugs that inhibit signalling growth. This suggested that VEGF-C might ter while patrolling tissues. However, in the by members of the VEGF family, with most have another, previously unsuspected role in suppressive microenvironment of a tumour, approaches4 target- limiting tumour growth by boosting patient such cells often enter a dysfunctional state. A “An increase in ing VEGF-A. responses to immunotherapy. goal of cancer immunotherapy is to trigger or the expression VEGF-C promotes The authors investigated the basis of this revitalize the antitumour immune response of a protein lymphatic-vessel for- phenomenon. They found a strong positive either through vaccination strategies or by that promotes mation (lymphangio- correlation between VEGF-C levels and the using an approach called checkpoint-blockade genesis) by signalling strength of a tumour-specific cytotoxic T-cell therapy that dampens immuno-inhibitory tumour growth through the VEGFR3 response after immunotherapy in their model signals, such as those of the PD-1 or CTLA-4 can also make receptor5. Lymphatic mice. They also observed a similar associa- pathways2. Writing in Science Translational a tumour more vessels can pro- tion between high VEGF-C expression and Medicine, Fankhauser et al.3 show that an responsive to mote the process of boosted immunotherapy responses (Fig. 1) increase in the expression of a protein that pro- immunotherapy.” metastasis — tumour- when analysing data from people who have motes tumour growth can also make a tumour cell spread beyond a metastatic form of skin cancer called mela- more responsive to immunotherapy. the primary tumour growth site — by provid- noma. Together, these results suggest that Tumours need a vascular supply to grow. ing a route for cancer cells to exit a tumour and VEGF-C and the lymphatic system might Blood vessels bring nutrients and oxygen to reach nearby structures termed lymph nodes. have unexpected roles in aiding cancer immu- a tumour, whereas lymphatic vessels remove This process enables tumour cells to grow in notherapy. Perhaps VEGF-C levels could be fluid and waste. A tumour and the cells in its lymph nodes, and to disseminate from there to used as a biomarker to identify patients who surrounding milieu stimulate the development other locations. Many tumours associated with are likely to respond to immunotherapy.

340 | NATURE | VOL 552 | 21/28 DECEMBER©2017 Mac mi ll2017an Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.