Exploration of the reaction conditions. In continuation of our investigations on peptide formation reactions in SO2 we aimed to expand the scope of prebiotically plausible reactions in SO2 to phosphorylation reactions. In a pressure apparatus SO2 was condensed on a mixture of adenosine (A) (100 mM) and phosphorous acid (H3PO3) (1.0 eq.) (Fig. 1a and Supplementary Fig. 1) and the reaction mixture was stirred for 7 d at room temperature. Analysis of the reaction mixture by capillary electrophoresis coupled to electrospray ionization Orbitrap mass spectrometry (CE-ESI-MS) or UV detection (Supplementary Fig. 2) showed the formation of the phosphorylated products including adenosine H-phosphonate (A-pIII) (constitution of potential isomers has not been determined; labels and abbreviations illustrate all phosphate/phosphonate binding modes and refer to the entirety of all formed isomers) (Supplementary Figs. 15, 16, 77, 78, Supplementary Tables 1, 2)52. Apart from the P(III) compound, adenosine monophosphates (AMPs) (5’ ,3’ and 2’ AMP) and traces of cyclic adenosine monophosphate (cAMP) were detected. As expected for a non-enzymatic approach, the regioselectivity is not controlled. Phosphorylation proceeds at mild reaction temperature in liquid SO2 and the P(V) products shows that the reaction medium enables P(III) oxidation at the same time. In addition to the monomeric compounds, adenosine diphosphate (ADP) (5’ ADP and other isomers), dinucleotide species and A dimers bridged by a phosphate group (A-pV-A) were obtained. The extracted ion electropherogram (EIE) of the dinucleotide species (pV-A-pV-A/ A-pV-pV-A) shows a discrete peak at 9.90 min and several signals at around 12.40 min. The A units can either be bridged by pyrophosphate (A-pV-pV-A) or by phosphate linkages (pV-A-pV-A). Since phosphate nucleophiles are superior to alcohol nucleophiles, the formation of A-pV-pV-A dimers is conceivable7. Tandem mass spectrometry (MS/MS) spectra corroborate that A-pV-pV-A dimers are the faster migrating species (Supplementary Fig. 138). The signals at higher migration times were assigned to pV-A-pV-A dimers (Supplementary Fig. 139). Co-injection of a 5’-adenylic acid-3’,5’-adenosine phosphate (5’-pV-A-3’-pV-5’-A) reference confirmed this result (Supplementary Fig. 3, 85). It has to be noticed, that in contrast to the monomeric P(III) phosphorylated reaction products, no P(III) bridged nucleosides (A-pIII-A) and P(III) phosphorylated dinucleotides (pIII-A-pIII-A/ A-pIII-pIII-A) were detected. This is in agreement with experimental data, showing that oxidation of phosphonate diesters is rapid compared to phosphonate monoesters31, 53. Furthermore, Peyser et al. reported rapid hydrolysis of phosphonate diesters49. With these very promising results using liquid SO2 not only as solvent but also as oxidant in P(III) phosphorylation reactions we comprehensively screened the reaction conditions. At first, we varied the H3PO3 concentration. An increase in the yield of 5’ AMP and 5’ ADP was observed if an excess of H3PO3 (3.0 eq.) was applied (Supplementary Figs. 17-20, Supplementary Tables 3, 4). Furthermore, a broadening of the reaction product spectrum was observed. In addition to the acyclic products obtained with stoichiometric amounts of H3PO3, traces of adenosine diphosphonate (A-pIII-pIII/ pIII-A-pIII), mixed diphosphorylated species (A-pV-pIII/ pIII-A-pV), adenosine triphosphate (ATP) and mixed (non-) cyclic adenosine diphosphate (pV-A-cpV/ A-cpV-pV) were unambiguously detected (Supplementary Figs. 79, 80). Further increase of the H3PO3 concentration (5.0 eq.) led to the formation of mixed H-phosphonate phosphate dinucleotides (A-pIII-pV-A/ pIII-A-pV-A) in trace amounts but did not alter the product range to a greater extent (Supplementary Figs. 81, 82). Rapid oxidation of mixed P(III) P(V) dinucleotides to pV-A-pV-A / A-pV-pV-A by liquid SO2 is presumed to protect them from hydrolysis. An interesting aspect in this context is, that oxidation rates of phosphonate diesters exceed those of phosphonate monoesters, which suggests that the oxidation could be the driving force in polymerization reactions according to studies by Lönnberg31, 32, 53, 54. However, while the increase of H3PO3 (5.0 eq.) led to the formation of trace amounts of mixed dinucleotides, the 5’ AMP and 5’ ADP yields slightly decreased at the same time (Supplementary Figs. 21-24, Supplementary Tables 5, 6). It has to be considered that product formation and hydrolysis are competing processes. Acceleration of the latter by acids is possibly the reason that yields do not further increase even when more phosphorylation agent is added55.
Next, we investigated whether the phosphorylation could be enhanced by the addition of urea (Fig. 1b). Urea is prebiotically plausible and well known to promote phosphorylation and other condensation reactions16, 19, 51, 56. Discussed enhancement mechanisms involve urea hydrolysis and formation of activated intermediate carbamoyl phosphate and phosphor amidate species16, 19, 48, 57. Starting from the optimized H3PO3 concentration, we found that the phosphorylation is indeed promoted by urea. Both the addition of 1.0 eq and 3.0 eq. affected the product mixture only slightly but led to a clear increase in 5’ AMP and 5’ ADP yields (Supplementary Figs. 25-32, 83, 84, 86, 87, Supplementary Tables 7-10). Yields of up to 26.7% for 5’ AMP and 2.2% for 5’ ADP were detected with a stoichiometric amount of urea.
Furthermore, we examined the robustness of the P(III) phosphorylation from a prebiotic point of view. Prebiotically plausible conversions require robust product formation under simple reaction conditions. In addition, extraordinary performance at low reactant concentration is essential since limited quantities are often assumed for emergence of life scenarios. Thus, we tested lower concentrations in A with stoichiometric amounts of H3PO3. Fig. 1c shows that 5’ AMP and 5’ ADP yields increased at lower A concentrations (Supplementary Figs. 15, 16, 33-42, Supplementary Tables 1, 2, 11-16). At the same time, a robust phosphorylation was observed over a wide concentration range. Although, at higher concentrations the product spectrum narrowed (Supplementary Figs. 77, 78, 88-93).
Time dependence and reaction pathways of the phosphorylation. Starting with the optimized reaction conditions, we studied the reaction progress over time. 7.3% 5’ AMP were already obtained after 1 d and the yield further increased within 7 d (Fig. 2a, Supplementary Figs. 43, 44, Supplementary Tables 17, 18). Apart from monomeric species, (mixed) P(V) and P(III) based dimers (A-pV-A, pV-A-pV-A/ A-pV-pV-A, pIII-A-pV-A/ A-pIII-pV-A) and diphosphorylated species (ADP, A-pIII-pIII/ pIII-A-pIII, A-pV-pIII/ pIII-A-pV, pV-A-cpV/ A-cpV-pV) were found within 1 d (Supplementary Figs. 94, 95). ATP formation was detected after 7 d and we were able to quantify the 5’ ADP yield within the same period (Supplementary Figs. 25-28, 83, 84, Supplementary Tables 7, 8). After 26 d, decreased 5’ AMP (18.7%) and 5’ ADP (1.1%) quantities were observed and less P(III) based products were detected (Supplementary Figs. 47-50, 98, 99, Supplementary Tables 21, 22). Enhanced hydrolysis by water seems to be a conceivable reason. Another explanation for the latter observation is oxidation to the corresponding P(V) compounds over time.
We then sought to explore potential reaction sequences leading to (dimeric) nucleotides (proposed reaction pathways for phosphorylated compounds are displayed in Fig. 2b). Previously detected P(III) species suggest nucleoside condensation with H3PO3 prior to an oxidation step. As expected, replacement of H3PO3 with P(V) compounds led to inferior product formation since nucleophilic attacks at P(V) are slower than at P(III) centres8. In reactions of A with phosphoric acid (H3PO4), only traces of AMP were obtained (Supplementary Figs. 100, 101). A mixture consisting of A, 5’ ADP and urea showed hydrolysis to 5’ AMP (3.4%) and small amounts of 5’ ATP (0.5%), which is proposed to be formed by phosphate transfer (Supplementary Figs. 57, 58, 105, 106, Supplementary Tables 27, 28). On the contrary, the reaction starting with 5’ AMP showed no product formation at all.
To further identify accessible pathways of the P(III) based reaction we investigated A nucleotides as reactants. Various ADP and ATP isomers were obtained by phosphorylation of 5’ AMP with H3PO3 in the presence of urea (Table 1, Supplementary Figs. 51, 52, 102, Supplementary Tables 23, 24). Although, yields of the 5’ substituted products (5’ ADP and 5’ ATP) are smaller than of products (5’ AMP and 5’ ADP) which are accessible via the same number of coupling steps in the reaction starting with A (Table 1). Detection of the mixed diphosphorylated species A-pV-pIII/ pIII-A-pV indicates that, similar to the nucleoside phosphorylation reactions, the phosphorylation of the mononucleotide proceeds via P(III) intermediates. Isomerization of 5’ AMP was not observed. Conversion of 5’ ADP with H3PO3 in the presence of urea led to intense hydrolysis of the diphosphate reactant (Table 1, Supplementary Figs. 53-56, Supplementary Tables 25, 26). The hydrolysis product 5’ AMP (68.8%) could be phosphorylated again. As a result, A-pV-pIII/ pIII-A-pV was detected. Further condensation and/or oxidation steps led to mixed triphosphorylated species (A-pV-pV-pIII/ pIII-A-pV-pV/ pIII-A(pV)-pV), ATP, A-pV-pV-A/ pV-A-pV-A and to the isomerization of 5’ ADP (Supplementary Figs. 103, 104). Although, A-pV-pIII-pIII/ pIII-A-pV-pIII/ pIII-A(pV)-pIII were not detected, condensation prior to oxidation is conceivable since phosphonate diester are rapidly converted to their P(V) analogues49.
Finally, investigation of potential reaction sequences showed that condensation steps are reversible under the presented reaction conditions. However, due to the reaction medium’s redox properties, oxidation steps are irreversible.
Reaction with other nucleosides. Next, we tested the applicability of the phosphorylation to the other canonical nucleosides (Table 1). Starting with ribonucleosides under the optimized conditions, we found 5’ NMP and 5’ nucleotide diphosphate (5’ NDP) yields for the reaction of cytosine (C) and uridine (U) which are comparable to the phosphorylation of A (Supplementary Figs. 59-66, Supplementary Tables 29-32). Furthermore, both product mixtures resembled the one observed in the reaction with A. N-pV-N was not detected neither for C nor for U, but in addition to the products obtained with A, mixed P(III) P(V) triphosphorylated species and in the reaction with C also mixed (non-)cyclic triphosphorylated compounds were observed (Supplementary Figs. 107-111). In analogy to phosphorylation of A, two signal sets were obtained for the dinucleotides. As in the case of A, the faster migrating species are the pyrophosphate linked dimers (N-pV-pV-N) whereas the slower ones could be assigned to dimers containing a phosphate linkage (pV-N-pV-N) (Supplementary Figs. 140-143). On the contrary to the phosphorylation of these nucleosides, the reaction with guanosine (G) yielded only 7.1% 5’ GMP and the amount of 5’ GDP was too small for quantification (Supplementary Figs. 67-69, Supplementary Tables 33, 34). In comparison to the other ribonucleosides, a smaller product mixture was obtained, which contained non-cyclic monophosphorylated species, P(V) and P(III) derived diphosphorylated compounds, several guanosine triphosphate isomers and P(V) based dinucleotides (Supplementary Figs. 112, 113). However, the latter (G-pV-pV-G/ pV-G-pV-G) was only detected in trace amounts and therefore assignment of the corresponding signals by MS/MS measurements was not possible.
In the case of the deoxyribonucleosides, a significant reactivity difference was observed for purine and pyrimidine based reactants. Reaction of both deoxyadenosine (dA) and deoxyguanosine (dG) provided only traces of monophosphorylated compounds. N-pIII and NMP species are the sole products (Supplementary Figs. 114, 115). In contrast, yields of the reaction with pyrimidine based deoxyribonucleosides even exceeded those of the reactions with ribonucleosides (32.6% 5’ deoxycytidine monophosphate (5’ dCMP) and 31.7% 5’ deoxythymidine monophosphate (5’ dTMP)) (Supplementary Figs. 70-76, Supplementary Tables 35-38). Both product mixtures of deoxycytidine (dC) and deoxythymidine (dT) resembled the one derived from the reaction with A (Supplementary Figs. 116-120). In analogy to the ribonucleotide dimers, product peaks could be assigned to pyrophosphate (first signal set) and phosphate linked species (second signal set) by MS/MS measurements (Supplementary Figs. 144-147). In liquid SO2 phosphorylated derivatives of all canonical nucleosides are accessible, however product yield and chain length strongly depend on the particular nucleoside.
Phosphorylation of prebiotically relevant substrates. After nucleotide formation, we explored the application of the presented phosphorylation in a broader prebiotic context. Apart from nucleotide formation, the phosphorylation of sugars, alcohols, carboxylic and amino acids is essential from a prebiotic point of view, since the corresponding products participate in metabolic cycles or are essential for the formation of amphiphiles3, 58. Thus, we selected glycerol (Gly), glyceraldehyde (GA), D-ribose (Rib), sodium L-lactate (Lc) and L-serine (Ser) as test substrates. By phosphorylation in liquid SO2, we were able to obtain the acyclic monophosphates of all model compounds (Table 1). Co-injection of reference compounds confirmed the presence of both Gly monophosphates, O-phospho-Ser (2) and the formation of Rib-5-phosphate among other constitutional isomers (Supplementary Fig. 121, 124, 126). Cyclic monophosphates were only detected in the reactions with Gly and Lc (Supplementary Figs. 121, 125). Furthermore, H-phosphonates were observed for all substrates except Rib (Supplementary Fig. 124). Consequently, reaction sequences already proposed for the nucleoside phosphorylation seem also conceivable for these substrates. Different P bridged products were obtained from the phosphorylation of Gly, GA, Lc and Ser (Supplementary Figs. 122). However, GA is the sole substrate that showed the formation of H-phosphonate diester linked substrate units (GA-pIII-GA) and only Gly and Ser yielded dimers such as pV-Gly-pV-Gly/ Gly-pV-pV-Gly. The longest observed oligomer was Gly3 triphosphate. In addition, up to triphosphorylated substrates were detected in the reactions of Lc (triphosphonate) and Ser (triphosphonate and -phosphate).
Moreover, in the case of Ser the detection of Ser2 and Ser2 phosphate (4) are noteworthy (Supplementary Fig. 127). Peptide formation promoted by phosphates is well known in literature59, 60. However, the effect has not been described for solutions of liquid SO2 yet. Previous peptide formations in this reaction medium required the presence of a metal51. In analogy to a proposed mechanism for amino acid condensation enhanced by trimetaphosphate, we suggest the formation of a cyclic intermediate (3) by intramolecular condensation of 2 prior to the nucleophilic attack of a second Ser unit and concomitant ring opening which leads to 4 (Fig. 4a)61.
Phosphorylation of a nucleoside mixture. To mimic a more complex early Earth scenario, we investigated the phosphorylation of a mixture containing all canonical deoxy- and ribonucleosides. Mass spectrometric analysis of the product mixture after CE separation showed signals for many N-pV-N and N-pV-pV-N/ pV-N-pV-N combinations. m/z values of most dC, dT, C, G and U containing ribonucleoside, deoxyribonucleoside and mixed ribonucleoside-deoxyribonucleoside dimers could be unambiguously assigned to a single nucleoside pair (Fig. 4b, Supplementary Figs. 128-137). However, in the case of the remaining dimers identical m/z values and fragmentation patterns prevented unambiguous determination of the building blocks. In analogy to the phosphorylation of single nucleosides, EIEs of all detected dinucleotides displayed two signal sets. Thus, we assume that both pyrophosphate (N-pV-pV-N) and phosphate (pV-N-pV-N) linked dimers were formed. Although, low product concentration prevented confirmation by MS/MS analysis. Analogous to the phosphorylation reactions of single nucleosides, dA showed the worst performance. Only m/z values of dimers with C and G were detected. However, there are other dimers with identical m/z values (e. g. A-dC, dG-dC and A-dG, dG-dG, A-A based dinucleotides). Consequently, we were not able to state whether dA containing products were present. The same applies for dG and A containing products.
Apart from the RNA world hypothesis a heterogeneous RNA/DNA world scenario is discussed62. In such a hypothetical heterogeneous RNA/DNA world, transformation of the observed heterogeneous dinucleotides to homogeneous oligomers over time might be a conceivable scenario since duplexes with heterogeneous backbone structure are less stable than their homogeneous analogues63. Consequently, heterogeneous templates prevent template product inhibition, which is well known in non-enzymatic replication with homogeneous templates63, 64. Furthermore, heterogeneous templates prefer homogeneous substrates and Krishnamurthy et al. proposed that the sequence information of heterogeneous oligonucleotides might be heritable64.