The Paleoproterozoic snowball Earth: A climate disaster triggered by the evolution of oxygenic photosynthesis

The earliest evidence for glaciation comes from the late Archean and early Proterozoic, which suggests Earth at this time experienced global temperatures not much different from those today. The oldest known midlatitude glaciation, recorded in the Pongola Supergroup diamictite, occurred at 2.9 Ga (10). The period from 2.45 Ga until some point before 2.22 Ga saw a series of three glaciations recorded in the Huronian Supergroup of Canada (11) (Fig. 1). The final glaciation in the Huronian, the Gowganda, is overlain by several kilometers of sediments in the Lorrain, Gordon Lake, and Bar River formations (Fms.). The entire sequence is penetrated by the 2.22 Ga Nipissing diabase (12); the Gowganda Fm. is therefore significantly older than 2.22 Ga.

In its eastern domain, the Transvaal Supergroup of South Africa contains two glacial diamictites, in the Duitschland and Boshoek Fms. The base of the Timeball Hill Fm., which underlies the Boshoek Fm., has a Re-Os date of 2,316 ± 7 My ago (13). The Boshoek Fm. correlates with the Makganyene diamictite in the western domain of the Transvaal Basin, the Griqualand West region. The Makganyene diamictite interfingers with the overlying Ongeluk flood basalts, which are correlative to the Hekpoort volcanics in the eastern domain and have a paleolatitude of 11° ± 5° (14). In its upper few meters, the Makganyene diamictite also contains basaltic andesite clasts, interpreted as being clasts of the Ongeluk volcanics. The low paleolatitude of the Ongeluk volcanics implies that the glaciation recorded in the Makganyene and Boshoek Fms. was planetary in extent: a snowball Earth event (15). Consistent with earlier whole-rock Pb–Pb measurements of the Ongeluk Fm. (16), the Hekpoort Fm. contains detrital zircons as young as 2,225 ± 3 My ago (17), an age nearly identical to that of the Nipissing diabase in the Huronian Supergroup. As the Makganyene glaciation begins some time after 2.32 Ga and ends at 2.22 Ga, the three Huronian glaciations predate the Makganyene snowball.

In contrast to the Makganyene Fm., the three Huronian diamictites are unconstrained in latitude. Poles from the Matachewan dyke swarm, at the base of the Huronian sequence, do indicate a latitude of ≈5.5° (18), but ≈2 km of sedimentary deposits separate the base of the Huronian from the first glacial unit (19), which makes it difficult to draw conclusions about the latitude of the glacial units based on these poles. Low latitude poles in the Lorrain Fm. (20, 21), which conformably overlies the Gowganda diamictite, are postdepositional overprints (22).

The recent discovery of MIF of S isotopes in Archean and early Proterozoic rocks has provided a major constraint on atmospheric O₂. Large MIF of S (up to several hundred permil) is produced by photolysis of SO₂ to S by light of wavelengths of <200 nm, which would have been unable to penetrate to the lower atmosphere had O₂ levels been above a few percent of the present atmospheric level (23). To preserve MIF, multiple atmospheric S species must be maintained as partially isolated reservoirs rather than being homogenized by oxidation of reduced species like hydrogen sulfide and polysulfur, as would occur at greater than ≈10^–5 present atmospheric level O₂ (24). Mixing and dilution of the atmospherically fractionated component, both in the atmosphere and in the oceans, presumably yields the observed Archean MIF values of up to ≈10‰ (25). An active oceanic S cycle, as would exist at moderately high O₂ levels, also would likely prevent the preservation of MIF.

Before the deposition of the ≈2.32 Ga Rooihogite and Timeball Hill Fms. in the Transvaal Supergroup (13, 26) and the McKim Fm. in the Huronian Supergroup,† sedimentary sulfides often display MIF (8, 26, 27) over an order of magnitude larger than the largest values observed in modern sulfate aerosols (25, 28). One plausible interpretation of the diminished MIF observed in the Rooihogite, Timeball Hill, and McKim Fms. is a rise in atmospheric O₂. At present, only glaciers record the MIF sometimes present in sulfate aerosols; in the marine environment, with riverine input composing ≈90% of sulfate input and aerosols composing ≈10%, MIF is not preserved (25). The small range of MIF could reflect an environment in which atmospheric chemistry began to approach modern conditions, decreasing the magnitude of MIF, but in which aerosols formed the major component of marine sulfate input, allowing its preservation.

The small range of MIF also permits an opposite interpretation. Rather than a decrease in atmospheric fractionation, the diminution could be a product of increased continental input and ocean/atmosphere mixing driven by glacial conditions. The period of 2.5–2.2 Ga was a time of glaciations. Enhanced glacial weathering could have made unfractionated S from igneous sources a larger part of the marine S budget, whereas sharper thermal gradients would have driven homogenization of the S pool. Both effects would have decreased sedimentary MIF. If this interpretation is correct, only the final decrease of MIF after ≈2.09 Ga may reflect the rise of atmospheric O₂, and the oxygenation event may have been more rapid than the length of Farquhar and Wing's (8) MIF Stage II (2.45–2.09 Ga) indicates.

Both the Huronian and the Transvaal Supergroups contain features suggestive of, but not demanding, the first appearance of O₂ in the period between the Gowganda and Makganyene glaciations. As noted in ref. 8, MIF in the Huronian rocks examined so far is at diminished values compatible with either increased atmospheric O₂ or enhanced glacial weathering and ocean/atmosphere mixing. The Lorrain Fm. contains red beds, as well as a hematitic paleosol at Ville-Marie, Quebec, and the Bar River Fm. contains pseudomorphs after gypsum, consistent with an increase in sulfate levels from oxidative weathering of sulfide minerals on land. In the Transvaal Supergroup, the upper Timeball Hill Fm., like the Lorrain Fm., contains red beds.

Stronger evidence for O₂ appears after the Makganyene glaciation. The Ongeluk Fm. is overlain by the Hotazel Fm., which consists predominantly of banded iron formation (BIF). The basal half meter of the Hotazel Fm. contains dropstones (29) (see Supporting Text, which is published as supporting information on the PNAS web site), which suggests it was deposited toward the end of the glacial period. The Hotazel Fm. hosts a massive Mn member (29): a blanket of Mn deposition unmatched by any other known in the world, ≈50 m thick and with an erosional remnant, the Kalahari Mn Field, measuring ≈11 × 50 km in extent (30).

The Kalahari Mn Field indicates the release of large quantities of O₂ into the ocean and therefore suggests a highly active postglacial aerobic biosphere. In seawater at circumneutral pH, only NO₃^– and O₂ can chemically oxidize soluble Mn²⁺ to produce insoluble Mn(IV) (31) (Table 1). Although a Mn²⁺ oxidizing phototroph also could produce Mn(IV), no such organism has ever been identified. Given the high redox potential of the Mn(IV)/Mn(II) couplet, it is likely that no such organism exists. The unexcited P₈₇₀ reaction center in purple bacteria has a redox potential too low to accept electrons from Mn(II) (Table 1), but no known photosystem reaction center aside from the P₆₇₀ reaction center of photosystem II has a higher redox potential. Photosynthetic reduction of Mn(IV) therefore would likely require a two-part photosystem akin to that involved in oxygenic photosynthesis.

One plausible interpretation of the sequence of events leading up to the Paleoproterozoic snowball Earth is shown in Fig. 1 and is as follows.

For cyanobacteria to be directly responsible for triggering a planetary glaciation, they must have been able to produce enough O₂ to destroy the methane greenhouse before the carbonate–silicate weathering cycle could compensate. On sufficiently long timescales, global cooling would slow silicate weathering and carbonate precipitation, thereby allowing CO₂ to build up in the atmosphere (37). The response time of the carbonate–silicate weathering cycle is generally estimated at ≈1 My (38), although the time to replace the greenhouse capacity lost in a methane greenhouse collapse may be longer. To estimate the timescale for destruction of the methane greenhouse, we constructed a steady-state ocean biogeochemistry model based on the assumptions that biological productivity was controlled by P and N and that Fe(II) oxidation was the main inorganic O₂ sink (see Supporting Text).

The critical P/Fe flux ratio for net oxidation of the surface ocean increases with the burial rate of C (see Fig. 2a, which is published as supporting information on the PNAS web site). For a P flux similar to today's value of ≈8 × 10¹⁰ mol/y (39) and a burial fraction of 2 × 10^–2, similar to modern anoxic environments, the model results indicate that the critical value is ≈1/50. The current ratio of P flux to hydrothermal Fe flux is ≈1/2, whereas the ratio of P flux to hydrothermal and terrigenous Fe flux is ≈1/64 (1). Rare earth element patterns of BIFs (40) and the stability of terrestrial ferrous sulfides in an anoxic atmosphere suggest that the main source of reactive Fe in the Archean was hydrothermal, so the P/Fe ratio may have been closer to 1/2 than 1/64.

When the P/Fe ratio falls below the critical value, essentially all O₂ produced is captured by Fe²⁺. Above the critical P/Fe value, there is a net release of O₂ (Fig. 2a). Other dissolved O₂ sinks, such as Mn²⁺, H₂, and CH₄, will consume some O₂, but hydrothermal fluxes of these reductants are up to an order of magnitude less than the flux of Fe²⁺ (41). At today's levels, these are capable of consuming ≈7 × 10¹⁰ mol of O₂ per y, equivalent to 4 × 10^–10 bar/y. Although the flux of H₂S from vent fluids in today's high sulfate oceans is comparable to that of Fe²⁺, in an ocean where Fe²⁺ is the main electron donor, H₂S levels would be lower. A formaldehyde rainout flux of ≈10¹² mol/y (24) would be the largest O₂ sink, capable of adsorbing ≈6 × 10^–9 bar of O₂ per y, but would be nullified by H₂O₂ rainout once even a small amount of O₂ accumulated in the atmosphere. O₂ production in excess of ≈6 × 10^–9 bar/y (1 bar = 100 kPa) would therefore be sufficient to initiate CH₄ oxidation, and, once begun, net O₂ production in excess of ≈4 × 10^–10 bar/y would suffice to continue it. At the rates predicted by our model, destruction of a 1-mbar methane greenhouse, if it occurred at all, would likely occur within a few My, a timescale comparable with the carbonate–silicate weathering cycle. Therefore, either cyanobacteria did not evolve until shortly before the oxygenation event, or the nutrient flux did not reach sufficiently high levels at any point after the evolution of cyanobacteria until then.

If cyanobacteria were present during the Huronian glaciations, the increased P flux into the oceans generated by glacial weathering (42) should have triggered the oxygenation event. Instead, the oxygenation event seems to correlate with the later Makganyene glaciation; this finding suggests the evolution of cyanobacteria occurred in the interval between the Huronian glaciations and the Makganyene glaciation.

Whether N limitation could have delayed the destruction of a methane greenhouse‡ depends on the Fe demand of the N₂ fixers and the ability of cyanobacteria to capture Fe before it reacted with O₂ and sank beneath the photic zone as a ferric precipitate. With anoxic deep waters providing a large source of Fe to the photic zone, life for an early Proterozoic diazotroph would have been easier than for modern diazotrophs that must subsist off of Fe transported from the continents. C/Fe ratios in N₂-fixing populations of Trichodesmium range from <2,000 to 50,000 (43). If early cyanobacteria were inefficient at both capturing and using Fe (e.g., C/Fe = 2,000; capture efficiency = 1%), then N limitation could have protected the methane greenhouse, but under more optimistic assumptions (either a higher capture efficiency or a higher C/Fe ratio), it would not have done so (Fig. 2b). The geologic record, not computational models, must ultimately decide.

The interpretation presented here suggests that planetary oxygenation began in the interval between the end of the Huronian glaciations and the onset of the Makganyene glaciation and that the Paleoproterozoic snowball Earth was the direct result of a radical change in the biosphere. In the Archean and earliest Proterozoic oceans, life may have been fueled predominantly by Fe, with Fe(II) used as the electron donor for photosynthesis and Fe(III) as the main electron acceptor for respiration. The sediments therefore would be moderately oxidizing and the surface waters reduced (34). Because the redox potential for Mn²⁺ oxidation is much higher than that of Fe²⁺, Mn would have to be removed from the oceans in reduced form. The carbonates precipitated at this time contain up to ≈2% Mn (30), indicating that Mn²⁺ reached shallow waters and coprecipitated with Ca⁺²; oxidized Mn is extremely rare. The atmosphere was likely reducing. Astrophysical models predict the Sun was substantially dimmer than today, but a CH₄ greenhouse (44) produced by methanogens living in a reduced upper ocean would have kept the planet warm enough to allow for the presence of liquid water without leading to massive siderite precipitation (45).

This world would have been overthrown when cyanobacteria capable of oxygenic photosynthesis evolved, which molecular phylogenies indicate occurred later than the main bacterial radiation (46). The surface waters became oxidizing and more productive. Reduced C accumulated in the sediments; methanogenesis moved from the surface ocean to the deep ocean, where it was isolated from the atmosphere. Methanotrophy in the ocean and photochemistry in the atmosphere used O₂ to transform atmospheric CH₄ to CO₂, a less effective greenhouse gas. The rise of O₂ thus might have triggered a glacial interval without the ≈400 My delay assumed by others (35).

Phosphate flux into the oceans correlates with increased continental weathering during glacial intervals (42), so increased continental weathering produced by the glaciations may have increased nutrient availability above the threshold for net O₂ release into the atmosphere. Because of the relatively low global temperature, it would have taken only a trace of OH radicals from the oxygenic bloom produced when cyanobacteria appeared in a high-P ocean to damage the CH₄ greenhouse enough to bring average global temperatures to <0°C. During the at least ≈35 My (29, 47) it took to build up a sufficient CO₂ greenhouse to escape from the snowball, hydrothermal fluxes would have built up a rich supply of nutrients in the oceans. When the planet finally warmed, the oceans were ripe for a cyanobacterial bloom. The O₂ produced by the bloom cleared out tens of My worth of accumulated reductants and thus produced the Kalahari Mn field (29).

Despite the parsimony of cyanobacterial evolution occurring within a few My before the onset of the Paleoproterozoic snowball Earth, some organic biomarker evidence and indirect sedimentological and geochemical arguments have been used to suggest that the origin of cyanobacteria dates to far earlier times: at least 2.78 Ga and maybe as long ago as 3.7 Ga.

The critical piece of evidence placing the origin of cyanobacteria and locally oxic environments in the Archean is the discovery in bitumens from rocks as old as 2.78 Ga of organic biomarkers apparently derived from lipids used by cyanobacteria and eukaryotes in their cell membranes. Although Brocks et al. (48) concluded that the bitumens were likely syngenetic, they could not exclude the possibility that they were postdepositional contaminants. Even if the biomarkers are as old as their host rocks, however, the uniformitarian extrapolation of modern lipid distributions to the Archean should be viewed cautiously.

Hopanes. Among modern organisms, 2-methyl-bacteriohopanepolyol is produced predominantly by cyanobacteria and in trace quantities by methylotrophs like Methylobacterium organophilum (3). Hopanes derived from 2-methyl-bacteriohopanepolyol can be preserved in sedimentary rocks, where they have been used as tracers for cyanobacteria (3). Hopanol synthesis has traditionally been assumed, based on the understood modern distribution of hopanols, to occur only in aerobic organisms. Fischer et al. (79) recently demonstrated, however, that Geobacter sulfurreducens can synthesize diverse hopanols, although not 2-methyl-hopanols, when grown under strictly anaerobic conditions. Thus, fossil hopanes do not necessarily imply the presence of O₂-producing organisms, and nothing about 2-methyl-hopanols suggests that they are any different in this respect. Archean 2-methyl-hopanes also might have been produced by ancestral cyanobacteria that predated oxygenic photosynthesis.

Steranes. Produced by eukaryotes for use in the cell membrane, sterols are preserved in the rock record as steranes. Brocks et al. (4, 5) recovered steranes, along with 2-methyl-bacteriohopanepolyol, from 2.78 Ga Pilbara Craton shales. Because there is no known anaerobic sterol synthesis pathway, they used their discovery to argue for the presence of O₂.

Cholesterol biosynthesis in modern organisms is a long biochemical pathway that employs the following four O₂-dependent enzymes: (i) squalene epoxidase, (ii) lanosterol 14-α-methyl demethylase cytochrome P450, (iii) sterol-4-α-methyl oxidase, and (iv) lathosterol oxidase (49). These O₂-dependent enzymes perform reactions that, although not currently known to occur biochemically in anaerobic organisms, could feasibly occur without O₂. Moreover, the substitution of an O₂-dependent enzyme for an anaerobic step in a biosynthetic pathway appears to be a common evolutionary occurrence. Raymond and Blankenship (50) found that, of the 473 O₂-dependent enzymatically catalyzed reactions in the BioCyc database (www.biocyc.org), 20 have at least one O₂-independent counterpart that performs the same reaction. For instance, AcsF catalyzes an O₂-dependent cyclization step in the synthesis of chlorophyll and bacteriochlorophyll, a pathway that must have existed before the evolution of oxygenic photosynthesis. The O₂-independent enzyme BchE performs the same reaction as AcsF but uses vitamin B₁₂ in place of O₂ (50). The assumption that sterol synthesis is always O₂-dependent and always has been therefore merits close inspection.

Indeed, the Hamersley bitumens include their own cautionary message about the application of uniformitarian assumptions to fossil lipids. Dinosterane is generally accepted to be characteristic of dinoflagellates and is interpreted as a dinoflagellate biomarker in Phanerozoic rocks (51). Yet even though an Archean origin for dinoflagellates seems implausible, because it would indicate the Archean origin of modern eukaryotic group not known in the fossil record until at least the Paleozoic (52), Brocks et al. (5) found dinosterane. They interpreted the molecule as being produced by eukaryotes of unknown affinities, although an alternative explanation is that these modern-style putative Archean biomarkers are contaminants.

Although organic biomarkers may be difficult to interpret, they are a significant improvement on several other geological tracers that have been used to argue for the presence of cyanobacteria and environmental O₂, including microfossils, stromatolites, BIFs, and assorted isotopic fractionations.

Microfossils. Before 2 Ga, when diversified assemblages with affinities to major groups of cyanobacteria first appear in the fossil record, the microfossil record is murky (53). Some have interpreted filamentous forms in earlier rocks as cyanobacterial remains (54–56), but Brasier et al. (57) recently questioned the biogenic nature of these objects. Moreover, cyanobacteria cannot be identified solely by a filamentous form. Many nonoxygenic bacteria are also filamentous, including some mat-forming green nonsulfur and purple sulfur bacteria (58, 59) and a methanogenic archeon (60). The wide variety of filamentous prokaryotes highlights a problem in identifying fossil microbes lacking clear evidence of cell differentiation based on morphology: Any given form has probably arisen many times in Earth history, both in extant and extinct organisms.

Stromatolites and BIFs. Two types of late Archean rock Fms. have often been interpreted as indicating cyanobacterial activity: stromatolites and BIFs. Des Marais (7) argued that large stromatolite reefs indicate the presence of cyanobacteria and therefore a locally aerobic environment, but large reefs also can form under anaerobic conditions. Populations of anaerobic methane oxidizers, for instance, have built massive reefs at methane seeps in the Black Sea (61). In addition, the Archean and Paleoproterozoic oceans were likely more supersaturated with respect to calcite and aragonite than the modern oceans (62), which would have facilitated the precipitation of large reefs even without biological participation. Indeed, abiotic processes may have played a major role in the formation of many Precambrian stromatolites (63). Moreover, although the deposition of ferric iron in BIFs has traditionally been taken to imply the presence of free O₂ (40, 64, 65), BIFs also could have formed in a O₂-free environment, either by photooxidation (66) or by Fe(II)-oxidizing phototrophic bacteria (67, 68).

Isotopic Evidence. Rosing and Frei (6) argued based on isotopic evidence that >3.7 Ga metashale from West Greenland preserves signs of an aerobic ecosystem. They found organic C with δ¹³C values of –25.6‰ in the same sediments as Pb with isotopic ratios indicating that the samples were originally enriched in U with respect to Th and interpreted this finding as reflecting an environment in which U could be cycled between its reduced, insoluble U(IV) form and its oxidized, soluble U(VI) form. They concluded the light C was produced by oxygenic phototrophs and that biogenic O₂ had permitted the redox cycling of U. However, all biological C fixation pathways and some abiotic mechanisms can produce light C, and, just like the Fe(III) in BIFs, U(VI) can form in the absence of O₂. At circumneutral pH values, the midpoint potential of the U(VI)/U(IV) couplet is ≈0 V, similar to the Fe(III)/Fe(II) couplet and considerably less oxidizing than Mn(IV)/Mn(II), NO₃^–/N₂, or O₂/H₂O (see Table 1).

A strong negative δ¹³C excursion in organic C at ≈2.7 Ga has been interpreted as evidence for the evolution of aerobic methanotrophy (69). Such light C suggests repeated fractionation: first in the production of CH₄ subsequently oxidized to CO₂, then in re-reduction by a primary producer; similar fractionations are observed today in environments with methanotrophs. But although CH₄ oxidation was once thought to require O₂, geochemical measurements (70) and recent microbiological work (71, 72) have demonstrated that CH₄ oxidation also can occur with alternative electron acceptors, so O₂ is not needed to explain the isotopic excursions. Although the anaerobic methane oxidizing bacteria studied today rely on sulfate, which is unlikely to have been available in a high-Fe Archean ocean, thermodynamics permits a variety of electron acceptors, including Fe(III), to be used for CH₄ oxidation.

S isotope data indicate local spikes in sulfate concentration starting ≈3.45 Ga (73, 74). Canfield et al. (75) argued that these spikes were produced in high-O₂ environments. But given the low redox potential of the sulfide/sulfate couplet, local sulfate spikes can be explained by scenarios that do not involve O₂. Moreover, sedimentary sulfate deposits, which disappear in the rock record after ≈3.2 Ga (75), do not reappear until after the Huronian glaciations, which suggests that high sulfate conditions were rare.

None of these indirect lines of evidence necessitate oxygenic photosynthesis. The case for cyanobacteria and locally oxic environments existing before the disappearance of MIF of S isotopes and the massive deposition of Mn in the Kalahari Mn Field rests, at the moment, solely on steranes.

Our model for a Paleoproterozoic origin of cyanobacteria is testable by several methods. It suggests that sterols in Archean rocks, if they are original, were synthesized by anaerobic reactions. It therefore should be possible to find or engineer enzymes capable of synthesizing cholesterol under anaerobic, biochemically plausible conditions. In addition, a continuous biomarker record that stretches back from time periods when O₂ is definitely present into the Archean might reveal transitions in community composition. Current work with samples from recent drilling programs targeting the late Archean and the Paleoproterozic has begun this task. An intensive search for biomarkers with definite relationships to metabolism, such as those derived from the pigment molecules of phototrophic bacteria, also would produce a more convincing record. A search for cyanobacterial or eukaryotic fossils that predate 2.0 Ga yet have affinities to modern groups would complement the geochemical approaches.

With respect to the record of MIF of S, the timeline we propose for a rapid oxygenation scenario suggests that decreased fractionation during the interval of the Huronian glaciations may be a byproduct of enhanced glacial weathering and ocean/atmosphere mixing. If this scenario is correct, a similar phenomenon should have occurred in association with the Pongola glaciation at ≈2.9 Ga. Investigation of the Huronian deposits where low-MIF S is found should reveal these deposits to be sedimentologically immature, reflecting glacial input. Additionally, the syngenicity of the MIF values should be tested through techniques such as the paleomagnetic search for significantly postdepositional formation of sulfides.

Finally, our model predicts that the Makganyene snowball Earth was a singular event. Convincing paleomagnetic evidence (including positive syn-sedimentary field tests) that demonstrated the Huronian glaciations were low-latitude would contradict our model and instead support a protracted episode of planetary oxygenation with multiple snowball events not directly triggered by a singular event, the evolution of cyanobacteria.

Because of the importance of the evolution of cyanobacteria and the planetary oxygenation event in Earth history, it is particularly useful to consider multiple working hypotheses about these events. We propose a model that takes a skeptical attitude toward the evidence for Archean cyanobacteria and a protracted early Proterozoic planetary oxygenation. In our alternative scenario, an evolutionary accident, the genesis of oxygenic photosynthesis, triggered one of the world's worst climate disasters, the Paleoproterozoic snowball Earth. Intensive investigation of the time period of the Paleoproterozoic glaciations may reveal whether a novel biological trait is capable of radically altering the world and nearly bringing an end to life on Earth.

Author contributions: R.E.K. and J.L.K. designed research; R.E.K., I.A.H., and C.Z.N. performed research; R.E.K., J.L.K., and I.A.H. analyzed data; and R.E.K. and J.L.K. wrote the paper.

Abbreviations: Ga, giga-annum before present; My, million years; BIF, banded iron formation; Fm., formation; MIF, mass-independent fractionation.

We thank R. Adler, N. Beukes, R. Blankenship, J. Brocks, H. Dorland, A. Kappler, J. Kasting, A. Maloof, D. Newman, S. Ono, A. Sessions, D. Sumner, T. Raub, B. Weiss, and three anonymous reviewers for advice and discussion; R. Tada for help with fieldwork in the Huronian; A. Pretorius and Assmang Limited for access to Nchwaning Mine; and P. Hoffman for communicating this manuscript. This work was supported in part by the Agouron Institute and by a National Aeronautics and Space Administration Astrobiology Institute cooperative agreement with the University of Washington. R.E.K. was supported by a National Science Foundation Graduate Research Fellowship and a Moore Foundation Fellowship.