Prebiotic photoredox synthesis from carbon dioxide and sulfite

Abstract

Carbon dioxide (CO₂) is the major carbonaceous component of many planetary atmospheres, including the Earth throughout its history. Carbon fixation chemistry— that reduces CO₂ to organics—utilizing hydrogen as stoichiometric reductant usually requires high pressures and temperatures, and yields of products of potential use to nascent biology are low. Here we demonstrate efficient ultraviolet photoredox chemistry between CO₂ and sulfite that generates organics and sulfate. The chemistry is initiated by electron photodetachment from sulfite giving sulfite radicals and hydrated electrons, which reduce CO₂ to its radical anion. A network of reactions—generating citrate, malate, succinate and tartrate by irradiation of glycolate in the presence of sulfite—was revealed. The simplicity of this carboxysulfitic chemistry and the widespread occurrence and abundance of its feedstocks suggest that it could have readily taken place on the surfaces of rocky planets. The availability of the carboxylate products on early Earth could have driven the development of central carbon metabolism before the advent of biological CO₂ fixation.

Many CO₂ reduction reactions have been discussed in the context of prebiotic chemistry, but all are problematic in that they require very special conditions and/or materials that are simply rare on planetary surfaces. For example, reduction by hydrogenation of bicarbonate (HCO₃ ⁻) over a Ni-Fe alloy under hydrothermal conditions¹ requires high temperatures and pressures, and predominantly generates the C₁ product methane, a poor feedstock for elaboration into (proto)biomolecules. By separating H₂ and CO₂ with a thin Fe(Ni)S precipitate barrier across which there is a large pH difference, milder conditions enable reduction, but the product formate (HCO₂ ⁻) is only produced in trace amounts². Reduction of CO₂ using metallic Fe powder in water generates acetate, methanol, formate and pyruvate – the latter only transiently – but the widespread occurrence of Fe powder on rocky planets such as early Earth or Mars is unlikely³. Finally, UV photoreduction of CO₂ on colloidal ZnS semiconductor particles using hydrogen sulfide/hydrosulfide (H₂S/HS⁻) as a hole scavenger gives formate, acetate and propionate in low yield⁴, but these conditions are not likely to be common in a planetary context.

We previously demonstrated that hydrogen cyanide (HCN) can be reductively homologated using hydrated electrons (and/or hydrogen atoms derived therefrom by protonation) generated by UV irradiation of sulfidic anions in a process we termed cyanosulfidic chemistry^5–6. For this chemistry, we originally used H₂S/HS⁻ as stoichiometric reductant, but switched to using bisulfite⁷ (HSO₃ ⁻, pK_a ~7.2)/SO₃ ^2– because sulfur dioxide (SO₂) and H₂S are outgassed in a ~10:1 or greater ratio on Earth^8–9, and there is substantial evidence from the geological records of both Earth and Mars via the anomalous mass fractionation of sulfur isotopes that these sulfur species were important constituents of the early sulfur cycle^10–11. The Henry’s law constant for SO₂ is greater than that of H₂S and the first pK_a of hydrated SO₂ (~1.9) is far lower than that of H₂S (~7.1)¹², so dissolution and hydration of SO₂ in surficial water followed by dissociation would therefore have been greater than dissolution and dissociation of H₂S on early Earth and Mars. Based on reports that hydrated electrons generated by UV illuminating diamond surfaces reduce CO₂ to carbon monoxide (CO) in acidic aqueous solution¹³, and the aforementioned semiconductor UV photoreduction of CO₂, we now wondered if HSO₃ ⁻/SO₃ ^2– could serve as the source of hydrated electrons for CO₂ reduction by UV photodetachment¹⁴ (Supplementary Table 1). Given that alkaline lakes can simultaneously absorb atmospheric CO₂ and SO₂ to give HCO₃ ⁻ and SO₃ ^2– and a growing body of evidence that suggests that such lakes could have concentrated other prebiotically important species on early Earth and maybe Mars^15–16, we started to explore reduction chemistry at mildly alkaline pH.

Results

Photoredox CO₂ fixation reaction

We subjected an aqueous solution of the sodium salts of HCO₃ ⁻ 1 (50 mM) and SO₃ ^2– (100 mM) at pH = 9 to UV irradiation from Hg-lamps with principal emission at 254 nm in a standard laboratory UV photoreactor and analyzed the resultant mixture by ¹H-NMR spectroscopy, integrating signals relative to those of a subsequently added standard to quantitate products. After 4 hours irradiation, formate 2 (18 mM), hydroxymethanesulfonate 3 (200 μM), methanol 4 (200 μM), glycolate 5 (200 μM), acetate 6 (50 μM), tartronate 7 (600 μM), and malonate 8 (300 μM) had been produced alongside both rac- and meso-tartrate 9a (30 μM) and 9b (30 μM) (structures of products shown in Fig. 1, Supplementary Fig. 1). Sulfate was detected as a photoredox co-product¹⁴ by precipitation of barium sulfate upon addition of barium chloride under conditions where barium sulfite is soluble¹⁷. The bicarbonate-sulfite irradiation experiment was repeated using ¹³C-labelled HCO₃ ⁻ 1 to confirm that all the products were generated from the photoreduction of CO₂, and all product assignments were confirmed by spiking with authentic standards (Supplementary Fig. 1 and 2). Surprisingly, we were able to detect elemental hydrogen (H₂) by ¹H-NMR spectroscopy (δ = 4.5 ppm) if it was generated in situ by performing the irradiation experiment in a quartz NMR tube. This peak decreased/disappeared simply by shaking the NMR tube presumably because this accelerated degassing. The signal assignment for H₂ was confirmed by running an NMR spectrum of the products of mixing zinc with hydrochloric acid solution in an NMR tube (Supplementary Fig. 3). Taken together, these results show that HCO₃ ⁻ 1 is reductively converted to C₂, C₃ and (traces of) C₄ compounds as well as being reduced to other C₁ compounds in a process that also generates H₂ and SO₄ ^2–. If the initial concentration of HCO₃ ⁻ 1 was reduced to 5 mM and the concentration of SO₃ ^2– reduced to 10 mM, formate 2 (30 μM), glycolate 5 (20 μM), acetate 6 (10 μM), tartronate 7 (120 μM), and malonate 8 (30 μM) were observed by ¹H-NMR spectroscopy after 4 hours irradiation. The combined yield of organics in these experiments exceeded 10% demonstrating the remarkably high efficiency of this chemistry compared to other potentially prebiotic CO₂ fixation processes (Supplementary Fig. 4, Extended Data Fig. 1). The results were similar in experiments starting from carbon dioxide instead of sodium bicarbonate (Supplementary Fig. 5). In addition to the protiated products observed by ¹H-NMR spectroscopy, oxalate 10 was observed by ¹³C-NMR spectroscopy in yields as high as 11% (Supplementary Fig. 6). At higher concentrations of reactants, the yield of C₁ products, especially formate 2, went up relative to the yield of C_>1 products and after prolonged irradiation, a new C₃ product, β-hydroxypropionate 11 was identified (Supplementary Fig. 7).

Fig. 1. Carboxysulfitic photoredox reaction network starting from bicarbonate (HCO₃ ⁻) 1.

Starting in the top left, the reaction network starts with the addition of hydrated electrons (produced by photodetachment from sulfite) to CO₂ (blue) to give the carboxyl radical 12 after which point the network splits. Sequential reduction of the carboxyl radical 12 leads to the observed C₁ products (green) whilst dimerization of 12 to oxalate 10 initiates a path to C_>1 products (orange). Various reactions enable crossing between the C₁ manifold and the C_>1 manifold (the two manifolds are separated by a dashed line). The key oxidation of formate 2 back to the carboxyl radical 12 and the slow reduction of 2 that together divert flux from C₁ to C_>1 products are highlighted (fuchsia arrows). Photochemical reactions of oxalate 10, glyoxylate 18 and mesoxalate 23 (purple arrows) also contribute to the network.

Mechanistic study of the photoredox CO₂ fixation reaction network

We next investigated the photoreaction of the various products and some putative intermediates in the presence of SO₃ ^2– with a view to gaining information concerning the mechanism of the fixation chemistry. The results – summarized in Extended Data Fig. 2 (Supplementary Figs. 8 – 19) – can be rationalized by a reaction network based on photoredox radical chemistry (Fig. 1). Photodetachment of an electron from SO₃ ^2– gives a hydrated electron and a sulfite radical (^.SO₃ ⁻)¹⁴. At pH 9, both loss of hydroxide from HCO₃ ⁻ 1 and loss of water from its conjugate acid, H₂CO₃, furnish CO₂. The latter process is efficiently catalyzed by nucleophiles^18–20, especially sulfite²¹, so it is unlikely that the otherwise slow kinetics of equilibration limit the photoredox chemistry. Although the equilibrium concentration of CO₂ is very low in a solution containing HCO₃ ⁻ 1 at pH = 9 relative to the concentration of 1 (Supplementary Fig. 20a)²², the rate constant for reaction of CO₂ with hydrated electrons to give the carboxyl radical 12 is extremely high²³ and the rate greatly exceeds the rate for protonation of hydrated electrons by 1 giving hydrogen atoms²⁴. The carboxyl radical 12 can either be reduced by hydrogen atom transfer (HAT) from HSO₃ ⁻, which has a ~1% abundance relative to SO₃ ²⁻ at pH = 9, to give formate 2, or undergo dimerization to give oxalate 10, both directly and indirectly²⁵. Focussing on the chemistry of formate 2 first, one electron reduction, though relatively slow²⁶, gives the radical anion 13 and thence, through acid-base and hydration equilibria, the radicals 14 and 15 (although the latter is unfavoured relative to 13 and 14). The radicals 13 and 14 have two main fates; reduction by HAT from HSO₃ ⁻, or recombination with the carboxyl radical 12. Coupled with acid-base and hydration equilibria, the first process (shown only for 13), generates formaldehyde 16 and its hydrate 17 and the second (shown only for 14) generates glyoxylate 18 via its hydrate 19. Formaldehyde 16, in equilibrium with the bisulfite adduct 3, can be reduced to the radical 20 which gives methanol 4 by HAT and glycolate 5 by recombination with the carboxyl radical 12 ²⁷. Another major reaction of formate 2 is oxidation back to carboxyl radicals 12 by reaction with sulfite radicals. This is inferred from the observation that irradiation of formate 2 and SO₃ ^2– gives significant amounts of what appear to be products deriving from oxalate 10 in addition to C₁ products (Extended Data Fig. 2).

The other initial product of the carboxyl radical 12 – its dimer oxalate 10 – can be reduced by addition of a hydrated electron to give the radical anion 21 ²⁸. This reduction is much faster than the corresponding reduction of formate 2 (Ultrafast pump-probe experiments and discussion are in Method). The radical anion 21 can undergo HAT leading to glyoxylate hydrate 19, or recombination with another carboxyl radical 12 to give mesoxalate hydrate 22, which equilibrates with mesoxalate 23 ²⁹. Reduction of mesoxalate 23 by addition of a hydrated electron, or electron transfer from a carboxyl radical 12, followed by HAT gives tartronate 7 and deoxygenation of the latter followed by HAT gives malonate 8. In the same multistep way that formate 2 can be reduced to methanol 4, reduction of one of the carboxylate groups of malonate 8 leads to β-hydroxypropionate 11. Reduction of glyoxylate 18 (in equilibrium with the hydrate 19 and a bisulfite adduct)³⁰ and protonation of the initially formed radical anion³¹ leads to the key hydroxy-carboxymethyl radical 24 (pK_a~8.8)³² which can recombine with the carboxyl radical 12 to give tartronate 7, dimerize to give the tartrates 9, or undergo HAT to give glycolate 5. Deoxygenation of glycolate 5 gives the carboxymethyl radical 25, which by recombination with the carboxyl radical 12 can give malonate 8 ³³ and by HAT, acetate 6.

Finally, we identified a number of photochemical steps other than the photodetachment of electrons from SO₃ ²⁻, which initiates the reaction network. Norrish type I reactions of glyoxylate 18 and mesoxalate 23 generate radicals 12, 15 and 26 (similar photocleavage of malonsemialdehyde 27, en route to β-hydroxypropionate 11, would generate radicals 15 and 25) and photodetachment of an electron from oxalate 10 ^34–35 gives radical 28 which is thought to decarboxylate to the carboxyl radical 12. These additional photochemical steps set up futile cycles in the network, but also forge links from the C_>1 parts of the network to the C₁ part (Supplementary Figs. 21 – 26).

Based on the foregoing analysis, we thought that it might be possible to increase the amount of the C_>1 products by adding sulfite portionwise – this would ensure that at any one time, the concentration of HSO₃ ⁻ would be low, so reaction flux through oxalate 10 would be favoured, but overall, there would be more reduction capacity. In accordance with expectation, at the end of this experiment, the combined yield of C_>1 products (>25%) greatly exceeded C₁ products (<1%) and the combined yield of malonate 8 (16.2%) and acetate 6 (1.0%) was greater than twice that of tartronate 7 (6.6%) and glycolate 5 (0.8%). Unexpectedly, a new product, sulfoacetate 29 (0.8%) was formed in low yield, presumably through recombination of carboxymethyl radicals 25 with sulfite radicals. (Supplementary Fig. 27, Extended Data Fig. 1, Entry 6). The general features of the time course of the CO₂ reduction network were also revealed by this experiment. After 1 hour, formate 2 was the major product accompanied by traces of glycolate 5 and tartronate 7. After 2 hours, the amount of formate 2 had decreased and glycolate 5 and tartronate 7 were now the major products along with smaller amounts of acetate 6 and malonate 8. After further irradiation, the levels of formate 2, glycolate 5 and tartronate 7 stayed at about the same level and malonate 8 became the major product with minor amounts of acetate 6 and methanol 4. This time course behaviour can be understood from the reaction network (Fig. 1). At the outset of the experiment, the only carbonaceous species for the hydrated electrons to reduce is CO₂. Carboxyl radicals 12 thereby produced apparently undergo HAT from HSO₃ ⁻ faster than they dimerise, and so formate 2 increases. However, the conversion of carboxyl radicals 12 to 2 is reversible, and so after some time a sufficient amount of oxalate 10 is produced for it to be reduced by the hydrated electrons as well. The reduction of oxalate 10 is much faster than the reduction of formate 2 (investigated by ultrafast pump-probe spectroscopy and further discussed in Methods), so 2 is consumed at the expense of making reduction products of 10. The appearance of the radical anion 21 opens up a new path for consumption of carboxyl radicals 12, including recombination to give mesoxalate hydrate 22 that is rapidly converted to tartronate 7 and a new path for the consumption of HSO₃ ⁻, namely HAT to 21 giving glyoxylate 18 and thence, through rapid further reduction, glycolate 5. The opening of these new reaction paths reduces the level of formate 2 to a steady state where its consumption is balanced by continuous production from CO₂ via carboxyl radicals 12. Eventually, the deoxygenation of tartronate 7 coupled to the slow reduction of malonate 8 means that the latter becomes the predominant product. At higher initial concentrations of sulfite, the early formate 2 pulse lasts longer and produces higher early amounts of 2, but eventually the paths to C_>1 products start to operate and levels of formate 2 drop. Even if formate 2 is reduced, the reversibility of the downstream steps to C₁ products and other paths from the C₁ part of the network to the C_>1 part means that products more complex than 2 eventually accumulate.

It has been reported that CO₂ can be reduced to CO with hydrated electrons produced by UV irradiation of diamond in the presence of bisulfite as hole scavenger at pH = 3.2¹³. The carboxyl radical 12 is formed as in our chemistry, but absent the chromophoric sulfite acting to absorb UV, it is photolyzed³⁶ to CO and O^.−, the conjugate base of the hydroxyl radical, which is then reduced by bisulfite. Hydrogen was not observed as a product because the reaction was carried out under a sufficiently high concentration of CO₂ to out-compete H⁺ in reaction with electrons.

Glycolate photoredox reaction

As we investigated the photoreactions of the products and putative intermediates of the CO₂ reduction network with SO₃ ^2–, the photoredox chemistry of one product – glycolate 5 – stood out. Acetate 6 (16.2 mM), malonate 8 (0.1 mM), sulfoacetate 29 (7.8 mM), citrate 30 (0.2 mM), rac-tartrate 9a (1.5 mM), meso-tartrate 9b (1.0 mM), malate 31 (2.9 mM), succinate 32 (1.1 mM) and hydroxycitrate 33 (0.19 mM) along with C₁ products were detected by ¹H-NMR spectroscopy after 6 hours irradiation of glycolate 5 (50 mM) and SO₃ ^2– (100 mM) (Supplementary Fig. 12, Extended Data Fig. 2, Entry 4). Particularly noteworthy is the fact that citrate 30, malate 31 and succinate 32 are key constituents of the Krebs cycle – a major cycle of central carbon metabolism, the consequences of which are discussed below. Remarkably, when the concentration of glycolate 5 was reduced to 5 mM and the concentration of SO₃ ^2– reduced to 10 mM, after 2 hours irradiation, C₁ products were no longer detected but the higher products were still formed in a comparable overall yield albeit with a different relative abundance distribution (Supplementary Fig. 13, Extended Data Fig. 2, Entry 5). The chemistry that generates acetate 6, malonate 8 and the tartrates 9 is the same as some of that of the CO₂ fixation reaction network, but additional reactions now contribute to the detectable products (Fig. 2). Abstraction of a hydrogen atom from glycolate 5 by a sulfite radical generates the hydroxy-carboxymethyl radical 24 whilst redox compensatory reduction of 5 generates the carboxymethyl radical 25. Dimerization of 24 produces the tartrates 9 whereas dimerization of 25 produces succinate 32 as well as acetate 6 and glycolate 5 ³⁷. Recombination of radicals 24 and 25 provides one route to malate 31, a second would be from reduction of 9. Similar reduction of malate 31 would give a second path to succinate 32. Oxidation of the tartrates 9 and malate 31 to the corresponding hydroxyalkyl radicals 34 and 35 followed by recombination of these radicals with radicals 24 or 25 would give dihydroxycitrate 36, hydroxycitrate 33 and citrate 30. Reduction of 36 would constitute another reaction channel to hydroxycitrate 33 and further reduction of 33, another channel to citrate 30. In contrast to the reaction network starting from CO₂ where all products are reduced relative to the starting material, the network starting from glycolate 5 is more subtle and contains both carbon oxidations and reductions. Thus malate 31 and citrate 30 are at the same oxidation level as glycolate 5, succinate 32 and acetate 6 are more reduced and the tartrates 9 and hydroxycitrate 33 are, on average, more oxidized.

Fig. 2. Carboxysulfitic photoredox reaction network starting from glycolate 5.

Glycolate 5 (blue) can be both oxidized by sulfite radicals and reduced by hydrated electrons to give the radicals 24 and 25 which then react further to give the observed products (orange). Recombination of two C₂ radicals (24 and/or 25) gives C₄ products (tartrate 9, malate 31 and succinate 32). Tartrate 9 and malate 31 can be oxidized by sulfite radicals giving C₄ radicals 34 and 35 and recombination of C₂ and C₄ radicals gives C₆ compounds (including the products hydroxycitrate 33 and citrate 30). Reduction of dihydroxycitrate 36 and tartrate 9 provides additional reaction pathways to the C₆ and other C₄ products.

Photoredox reactions under lower photon flux

We also evaluated the bicarbonate reduction chemistry using a less intense broadband UV source, StarLab³⁸ - an in-house constructed photoreactor designed to deliver UV radiation with a wavelength distribution representative of that from the Sun incident on the surface of early Earth, at a ~100 fold higher intensity than the Sun in a quiescent state and ~10 fold lower intensity than that during maximum flaring. After irradiation for 7 days in this apparatus, an aqueous solution of the sodium salts of HCO₃ ⁻ 1 (5 mM) and SO₃ ^2– (50 mM) at pH = 9 gave a mixture of protiated products similar to that obtained upon higher intensity irradiation in the standard laboratory photoreactor (at 254 nm for shorter time intervals) plus ethanol 37, confirming the utility of using 254 nm UV light to study this chemistry (Supplementary Fig. 28, Extended Data Fig. 1, Entry 7). Oxalate 10 was also detected in a similar experiment using ¹³C-labelled bicarbonate (Supplementary Fig. 29). Ethanol 37 could plausibly be obtained via dimerization of the hydroxymethyl radical 20 giving ethylene glycol 38, dehydration of 38 through radical 39 and the enoloxy radical 40 ³⁹ to acetaldehyde 41 and reduction (Fig. 1). Alternatively, acetate 6 could be reduced to acetaldehyde 41 and thence ethanol 37.

We then investigated the carboxysulfitic photoredox chemistry of glycolate 5 in the StarLab photoreactor. After 8 hours irradiation of glycolate 5 (50 mM) and SO₃ ^2– (100 mM) with this less intense light source, acetate 6 (1.7 mM), sulfoacetate 29 (0.2 mM), rac-tartrate 9a (0.4 mM), meso-tartrate 9b (0.4 mM), malate 31 (0.2 mM), and succinate 32 (trace) along with C₁ products were detected by ¹H-NMR spectroscopy (Supplementary Fig. 14, Extended Data Fig. 2, Entry 6). Longer irradiation of more dilute samples of glycolate 5 (5 mM) and SO₃ ^2– (50 mM) in the StarLab photoreactor resulted in higher yields of the same species and additionally produced malonate 8 and hydroxypropionate 11 (Supplementary Fig. 30, Extended Data Fig. 2, Entry 7).

Discussion

Planetary relevance

An important aspect of this chemistry is that the conditions and materials necessary to foster carboxysulfitic carbon fixation (short wave UV light, CO₂ and SO₂ derived from volcanism, and bodies of standing and flowing water on the crust) are mild, widespread, and expected to be common on rocky planets. Notably, there is geological evidence from the rock records of Earth and Mars that these conditions were met early in their history. Oxygen isotope ratios from Hadean zircons^40–41 and sedimentological observations from the earliest sedimentary record⁴² indicate abundant surface liquid water. Silicate weathering reactions occurred that sourced the alkalinity necessary to enable the dissociated hydrates, bicarbonate and sulfite, to partition from the atmosphere and accumulate in bodies of water in contact with the atmosphere⁴³. Moreover, the anomalous fractionation of multiple sulfur isotopes in the early geological record¹⁰ provides a direct measure of SO₂ photochemistry that establishes a valuable atmospheric correlate of the aqueous carbon fixation processes described herein. Finally, each of these observations for the early Earth that illustrates the plausibility of this chemistry occurring now has its complement in the Mars geological record^{11, 44–47}. Thus, the ingredients and basic conditions for carboxysulfitic chemistry to take place would have been present on both Earth and Mars. Depending on conditions, carboxylates such as formate 2, oxalate 10 or acetate 6 and malonate 8 are likely to have been the major initial products. Decarboxylation of malonate 8 to acetate 6 occurs on a short geological timescale in solution (~10 years at neutral pH and at 25°C)⁴⁸ whereas oxalate 10 (in the absence of ferric ions and light)⁴⁹, like acetate 6 is long-term stable and so it seems likely that these latter two products would have become the most abundant C_>1 organics on early Earth had life not emerged – they might still be the most abundant organics on Mars if life did not emerge there.

The linkage between carboxysufitic chemistry and cyanosulfidic chemistry

The case for conditions conducive to cyanosulfidic chemistry being present on both young planets has also been made⁵⁰. We note that for the full range of cyanosulfidic chemistry products to result, a scenario involving the mixing of bodies of water or flows (e.g. stream water) having subtly different reaction histories would probably be necessary. In locations where the basic conditions for cyanosulfidic chemistry were met, but the mixing of streams was absent or different, a limited set of products would have been generated and the first product of the restricted reaction network, glycolonitrile 42, would probably have been the most widespread. In addition, glycolonitrile 42 could have resulted from reaction of HCN with formaldehyde 16 rained in after production in the upper atmosphere by photoreduction of CO₂ ⁵¹. Hydrolysis of the nitrile group of glycolonitrile 42, however produced, generates glycolate 5, which could be converted by subsequent carboxysulfitic chemistry to the range of carboxylate products previously described. (Fig. 3) As the hydrolysis of glycolonitrile 42 generates ammonia in addition to glycolate 5, we also carried out the irradiation of 5 and sulfite in the presence of ammonia. Ammonia did not affect the outcome of the photoredox chemistry – the same set of products was formed with or without ammonia (Supplementary Fig. 31).

Fig. 3. Connections between environmental chemistry and the development of metabolism.

Progression from heterotrophy fed by photochemical products of inorganic carbon reduction to autotrophy as the available products of environmental chemistry become less complex. The preformed building blocks of RNA, peptides and lipids produced by cyanosulfidic chemistry provision the origin and early evolution of life^5,52, but gradually become depleted (fading of blue colour in timeline arrow) triggering the development of metabolism starting from simpler but more abundant products derived from glycolate 5 by cyanosulfidic chemistry (dark orange colour in timeline arrow). In turn, these materials become scarce (fading of orange colour in timeline arrow) and biology adapts to using carboxysulfitic products of CO₂ and eventually, CO₂ itself.

Biochemical relevance

Use of the products of cyanosulfidic chemistry as building blocks by nascent biology would eventually lead to their environmental depletion and biology would then be under evolutionary pressure to synthesize these building blocks from anything else that happened to be available and usable. Biology could either spread to encounter these materials in their place of synthesis, or fluvial advection could move them to the location of biology. It is fascinating that the majority of the carboxylate products deriving from the carboxysulfitic chemistry of glycolate 5 are key nodes of central carbon metabolism in extant biology and it seems likely that their synthesis by carboxysulfitic chemistry set the stage for the development of this metabolic network. At first glance, tartrate 9 seems to be somewhat an outlier, but its dehydration would lead through an enol to oxaloacetate⁵³ and its oxidation, to dihydroxyfumarate which spontaneously decarboxylates to give glycolaldehyde⁵⁴, a precursor of higher sugars.

With time, supply of most of the products of the carboxysulfitic chemistry of glycolate 5 would also dwindle and biology would have to evolve to make do with simpler, more abundant carbonaceous materials in the environment. The major long term stable products of the carboxysulfitic chemistry of CO₂ – formate 2, acetate 6 and oxalate 10 – could then provision central carbon metabolism through development of a pyruvate-formate lyase activity and the glyoxylate shunt of the Krebs cycle via reduction of oxalate 10 to glyoxylate 18. Finally, even oxalate 10 and acetate 6 would become depleted and biology would be under evolutionary pressure to use the only remaining abundant carbon source, namely CO₂.

Methods

General Methods

All reagents and deuterated solvents used for reactions and spiking experiments were purchased from Sigma-Aldrich and were used without further purification. All photochemical reactions were carried out in Norell Suprasil quartz NMR tubes purchased from Sigma-Aldrich using Hg lamps with principal emission at 254 nm in a Rayonet photochemical chamber reactor RPR-200, acquired from The Southern New England Ultraviolet Company. StarLab is an in-house constructed photoreactor that delivers broadband UV-Vis (from 220 nm to ~750 nm by using water as an optical filter) irradiation to a sample from a 75W xenon lamp manufactured by Horiba³⁸. A Mettler Toledo SevenEasy pH Meter S20 was used to monitor the pH, and degassed H₂O or D₂O was achieved by four rounds of freeze-pump-thaw cycling. ¹H-, and ¹³C-nuclear magnetic resonance (NMR) spectra were acquired using a Bruker Ultrashield 400 Plus or Bruker Ascend 400 operating at 400.1, and 100.6 MHz, respectively. Samples consisting of H₂O/D₂O mixtures were analyzed using HOD suppression to collect ¹H-NMR data. Chemical shifts (δ) are shown in ppm. Coupling constants (J) are given in Hertz and the notations s, d, t represent the multiplicities singlet, doublet, and triplet. The conversion yields were determined by relative integrations of the signals using a known amount of acetamide as internal reference in the ¹H-NMR spectrum.

General method of photoreaction of carboxylates with sulfite

Carboxylates and sodium sulfite (final concentrations were mentioned in Extended Data Fig. 1 and Extended Data Fig. 2) were dissolved in degassed H₂O/D₂O (9:1, 0.5 mL). After the pH was adjusted to the reported value with NaOH/HCl, the mixture was transferred to a quartz NMR tube which was sealed and irradiated for the reported time (Extended Data Fig. 1 and Extended Data Fig. 2). The resultant solution was analysed by ¹H- and/or ¹³C-NMR spectroscopy. The yield was calculated by spiking with 4,5-dicyanoimidazole (final concentration of 0.5 mM, 1 mM or 5 mM) and relative integration.

Preparing hydrogen gas in an NMR tube

Metallic zinc (~6 mg) was added to 0.5 mL HCl (0.1 M) aqueous solution. This solution was transferred to an NMR tube after being vortexed for 5 seconds, and was then analysed by ¹H-NMR spectroscopy.

Sulfate identification¹⁷

Sodium bicarbonate (21 mg, 0.25 mmol) and sodium sulfite (63 mg, 0.5 mmol) were dissolved in degassed water (10 mL) and the pH of the resultant solution was adjusted to 9 by adding NaOH/HCl solution. The mixture was then sealed in a quartz tube and irradiated with 254 nm light in the Rayonet photoreactor for 4 hours. 3 mL of the resulting solution was diluted to 20 mL with water and acidified to pH = 1 by the addition of concentrated HCl. The acidified solution was heated to nearly boiling for at least 30 min to remove all carbon dioxide and sulfur dioxide. Barium chloride solution was then added to the solution to give a precipitate which persisted upon boiling for another 30 min. The precipitate did not dissolve in dilute HCl solution.

Ultrafast pump-probe experiments

The general principles of pump-probe spectroscopy are described in the following references^55–57. The fundamental of the excitation pulses (800 nm) was generated by a Ti:Sa based laser-amplifier system (Solstice Ace by Spectra-Physics, Newport Co.) with a repetition rate of 1 kHz and a pulse duration of ~90 fs. The excitation pulses (251 nm) were generated in a nonlinear amplifier system (Topas Prime + NIRUVis, Light Conversion, Ltd.) and stretched by a 25 cm fused silica block (Corning) to ~ 1.7 ps to suppress two-photon ionization of the solvent. The excitation energy at the sample position was ~1 μJ and the spot diameter of ~250 μm (fwhm). For our microsecond ultrafast pump-probe spectroscopy (Supplementary Table 2) the broadband probe light (unpolarized) was generated, delayed, and detected in an EOS Fire system (Ultrafast Systems, LLC), with a nominal spectral range of 350 – 950 nm. For our picosecond ultrafast pump-probe spectroscopy (Supplementary Fig. 32) the broadband probe light was generated, delayed, and detected in a HELIOS Fire system (Ultrafast Systems, LLC), with a nominal spectral range of 400 – 750 nm. These spectral ranges are ideal for monitoring the broad absorption feature of the hydrated electron, which is centered near 700 nm. The experiments were carried out at a temperature of 23°C.

The transient pump-probe data were cropped to the spectral range 450 nm – 913 nm and 20 adjacent channels were averaged (Surface Xplorer, Ultrafast Systems, LLC). A global fitting analysis to determine transient lifetimes was performed^58–60.

Ultrafast pump-probe spectroscopic investigation of reduction chemistry

The rate constant reported in the literature for reaction of hydrated electrons with formate 2 (k = 2.4 x 10⁴ M^-1s^-1)²⁶ is considerably lower than that for reduction of oxalate (average of three values given in reference⁶¹ is 3.1 x 10⁷ M^-1s^-1). However, in our experiments, oxalate 10 is also prone to photoionization^34–35 and so it was not clear if reaction of hydrated electrons with oxalate in our experiments is, or is not, faster than reaction of hydrated electrons with formate 2. The rate constant for reaction of hydrated electrons with glycolate 5 (k = 8.2 x 10⁶ M^-1s^-1)³² is high, although the authors of this paper caution that this “unexpectedly high value may be due to trace impurity in the sample”. Furthermore, although the rate constant for reaction of solvated electrons with CO₂ is known²³, we do not know the concentration of CO₂ in our experiments, or whether the catalysis of its interconversion with carbonic acid and bicarbonate by sulfite affects this rate. Accordingly, we used ultrafast pump-probe spectroscopy both to confirm the photoionization of oxalate 10 and compare it to that of sulfite (Supplementary Fig. 32A), and to measure hydrated electron decay kinetics in mixtures representative of the mixtures used in our continuous irradiation experiments (Supplementary Table 3). These pump-probe experiments confirmed the photoionization of oxalate 10 and further revealed that bicarbonate, formate 2 and glycolate 5 react at similar rates with hydrated electrons in our experiments, and that oxalate 10 reacts considerably faster (Supplementary Fig. 32 and Supplementary Tables 2 and 3).

Extended Data

Extended Data Fig. 1. Product concentrations and percentage yields after UV irradiation of solutions of NaHCO₃ and Na₂SO₃.

Extended Data Fig. 2. Product concentrations and percentage yields after irradiation of individual bicarbonate reduction products (50 mM) and Na₂SO₃ (100 mM).

Data availability

All data generated or analysed during this study are included in the manuscript and the Supplementary Information.