Evolutionary Importance of the Intramolecular Pathways of Hydrolysis of Phosphate Ester Mixed Anhydrides with Amino Acids and Peptides

Aminoacyl adenylates (aa-AMPs) constitute essential intermediates of protein biosynthesis. Their polymerization in aqueous solution has often been claimed as a potential route to abiotic peptides in spite of a highly efficient CO2-promoted pathway of hydrolysis. Here we investigate the efficiency and relevance of this frequently overlooked pathway from model amino acid phosphate mixed anhydrides including aa-AMPs. Its predominance was demonstrated at CO2 concentrations matching that of physiological fluids or that of the present-day ocean, making a direct polymerization pathway unlikely. By contrast, the occurrence of the CO2-promoted pathway was observed to increase the efficiency of peptide bond formation owing to the high reactivity of the N-carboxyanhydride (NCA) intermediate. Even considering CO2 concentrations in early Earth liquid environments equivalent to present levels, mixed anhydrides would have polymerized predominantly through NCAs. The issue of a potential involvement of NCAs as biochemical metabolites could even be raised. The formation of peptide–phosphate mixed anhydrides from 5(4H)-oxazolones (transiently formed through prebiotically relevant peptide activation pathways) was also observed as well as the occurrence of the reverse cyclization process in the reactions of these mixed anhydrides. These processes constitute the core of a reaction network that could potentially have evolved towards the emergence of translation.

evolutionary option, the elucidation of the potential evolutionary process through which aa-AMPs could have been introduced requires the identification of simple pathways capable of leading to these intermediates. A likely possibility is the reaction of a-amino acid N-carboxyanhydrides (NCAs) with inorganic phosphate 15 and its esters including adenylates that takes place spontaneously at moderate pH 16,17 (Fig. 2a). This possibility is supported by the role of NCAs deduced from the literature 2 and the disclosure of realistic abiotic pathways for their formation during the last decade 18,19 . Since the activation of the C-terminus in peptides has recently been identified as a plausible prebiotic pathway and involves the formation of 5(4H)-oxazolone intermediates 20 , it is reasonable that similar mixed anhydrides with phosphates involving acylated amino acids (acyl-aa-PEMAs) or peptides (peptidyl-PEMAs) could be formed by reaction of the energy-rich cyclic intermediate (Fig. 2b). The occurrence of abiotic pathways leading to aa-PEMA or peptidyl-PEMA must have preceded their involvement in chemical evolution. However, the low stability of these mixed anhydrides and the availability of highly reactive cyclic intermediates prone to polymerize more easily renders their role in early abiotic processes of peptide formation highly questionable.
The kinetic stability of aa-AMPs and of other aa-PEMAs has been studied in aqueous solution leading to contradictory results in the literature [21][22][23][24] . Of particular interest with regard to an evolutionary context is the description of a highly efficient CO 2 -catalyzed path of hydrolysis [21][22][23] . No definitive mechanism has been proposed but the intermediacy of NCAs is highly probable 2,25,26 since other activated amino acids (nitrophenyl esters, thioesters) proved to undergo conversion into NCAs in hydrogen carbonate buffers 25 . This analysis casts doubts on the possibility that aa-AMPs constitute efficient monomers for the abiotic formation of peptides in aqueous solutions 2,3,26 since most early Earth aqueous environments are likely to have contained CO 2 or HCO 3 2 . The present investigations were aimed at providing data on the efficiency of the CO 2 -promoted pathway ( Fig. 3a) in aqueous solution at neutral pH and in the presence CO 2 concentrations compatible with early Earth environments and at clearly identifying the NCA as an intermediate. They address both the issues of the stability of aa-AMPs and of other aa-PEMAs and that of the path of peptide formation. They demonstrate the prevalence of the CO 2 -promoted pathway in the hydrolysis of adenylates. More importantly, using model amino amide reactants, they additionally demonstrate that peptide bond formation takes place predominantly from the cyclic intermediates rather than directly from the mixed anhydrides ruling out any possibility of considering the latter as direct peptide precursors at early stages of chemical or bio-       chemical evolution. Lastly, considering NCAs as likely precursors of aa-AMPs and aa-PEMAs, the hypothesis of an abiotic formation of non-coded peptides through these mixed anhydrides becomes unnecessary. The evolution of translation must then have proceeded through a pathway independent from abiotic polymerization. This work also addresses the more general goal of understanding the stability of phosphate mixed anhydrides of amino acids and peptides in aqueous media at moderate pH. As a matter of fact, though Nacylation is an obvious way to prevent CO 2 participation, another intramolecular path of breakdown through 5(4H)-oxazolones is possible in the case of acyl-aa-PEMAs (Fig. 3b). Therefore, the issues of the importance of the NCA and 5(4H)-oxazolone pathways in the reactions of the corresponding mixed anhydrides (Fig. 3) are raised as well as that of the potential role of these cyclic intermediates as potential prebiotic precursors of these mixed anhydrides (Fig. 2). The consequences of these chemical pathways as factors determining early biological evolution of amino acid activation processes and their constraints on the contemporary biochemistry of adenylates will also be discussed.

Results
Experiments were carried out from model systems derived from Omethylated tyrosine 5 ( Fig. 4) likely to be representative of the reactivity of usual amino acid derivatives. The UV-absorption of the tyrosine side chain (l max 5 273 nm) was selected to monitor reactions by HPLC at a reasonably low (0.05-1 mM) concentration range in which activated intermediates have a lifetime sufficient for their behaviour to be determined. Furthermore, phenol methylation was introduced to simplify analyses by avoiding any side-reaction of this group. Reactions were carried out in non-nucleophilic MES or MOPS buffers at pH values of 6.5 or 7.5, respectively, whereas 50 mM phosphate or methyl phosphate buffers were used for studying the transient formation of mixed anhydrides. Analyses were performed to monitor the reaction progress of samples stored in the HPLC systems located in a room maintained at the temperature of 20uC. Fast reactions were monitored by withdrawing 1 mL samples from the reaction medium and the reaction was blocked by addition of a formic acid solution to bring the pH to a value below 4 (Supplementary information).
NCAs as intermediates of aa-PEMA reactions promoted by CO 2 .
The hydrolysis of methyl phosphate mixed anhydride 1b was studied in buffered solutions in the presence of varying contents of CO 2 / HCO 3 2 . The reaction rates were observed to strongly depend on the presence of CO 2 as shown by a c.a. 4 fold increase in rate using pH 6.5 MES buffers previously equilibrated with air as compared with a solution flushed with N 2 for 60 min (   , the polymerization into peptides thus proceeds through the NCA rather than directly from the starting material. An NCA intermediate was also observed to form rapidly at pH 7.5 in 100 mM MOPS buffers in the presence of added HCO 3 2 ( Supplementary Information, Fig. S1). This behaviour indicates that the formation of long peptides from adenylates reported in the literature 9,10 results probably from the polymerization of NCAs rather than from that of adenylates. The conversion of aminoacyl adenylates into NCA in the presence of CO 2 /HCO 3 2 was investigated starting from the Tyr(Me) derivative 1c ( Supplementary Information, Fig. S2). The conversion of 1c into NCA was observed to proceed with rates similar to that observed for mixed anhydride 1b. The release of AMP (r.t. 1.5 min, method A) accompanying the formation of NCA 3 could be detected by HPLC allowing the reaction to be monitored at 50 mM concentrations of reactant 1c (r.t. 6.8 min, method A). The lifetime of the adenylate decreased with increasing concentrations of CO 2 /HCO 3 2 (t 1/2 , 80 min, ,25 min, and ,2 min at pH 6.5 in N 2 -flushed buffer, air equilibrated buffer and in the presence of 500 mM HCO 3 2 , respectively). At pH 7.5 the lifetime of adenylate 1c was reduced to less than 1 min in the presence of 500 mM HCO 3 2 , which means that this mixed anhydride is likely to be converted into NCA within a few seconds at concentrations of CO 2 /HCO 3 2 above 2 mM and at pH value close to neutrality, which are representative of the present day ocean or physiological fluids. It is worth noting that this lifetime is not sufficient for peptides to be significantly formed by a direct reaction with adenylate so that any observation of peptide products under these conditions results for the most part from the intermediacy of NCAs.
At pH 4, the hydrolysis of mixed anhydride 1b was much slower (t 1/2 5 ca. 550 min) and CO 2 catalysis was not observed ( Supplementary Information, Fig. S3). This result is consistent with the results obtained by Kluger from alanyl ethyl phosphate 24 . The protonation of the amino group of 1b increases the electrophilic character of its acyl group and then the rates of nucleophilic attack, but it also prevents any possibility of reaction with CO 2 according the pathway of Fig. 3a. The hydrolysis of the acetylated mixed anhydride 2b was indeed observed to be slower (t 1/2 , 950 min at pH 6.5) and was not affected by addition of 10 mM NaHCO 3 (Fig. 6) in a way consistent with this explanation and with previously reported analyses 22 . However, it is important to emphasize that the CO 2 -catalyzed pathway does not only constitute a process leading to the deactivation and the hydrolysis of mixed anhydrides since peptide formation can be improved significantly by this means. As a matter of fact, with regard to peptide formation, the prevalence of the NCA pathway was demonstrated by studying the model reaction of 1 mM mixed anhydride 1b with 5 mM glycinamide either in a nitrogen-flushed sample or in the presence of 2 mM NaHCO 3 (Fig. 7). Importantly, less than 2 min were sufficient for the starting material to be exhausted in the presence of carbonate, whereas CO 2 removal increased the reaction times to much higher values (t 1/2 , 50 min) and reduced the final yield in dipeptide (Fig. 7). This reaction remained faster than that observed for the acetylated mixed anhyd-ride 2b (t 1/2 , 260 min) unable to undergo the conversion into NCA, but that will be demonstrated below to partly undergo cyclization into 5(4H)-oxazolones. These experiments carried out using glycinamide for mimicking a growing peptide chain show that the polymerization of adenylates and other aa-PEMA is improved in the presence of CO 2 by the occurrence of the NCA pathway owing to both the higher reactivity of the latter intermediate and its ability to suppress diketopiperazine formation.
The interconversion of 5(4H)-oxazolones and acyl-aa-PEMA and peptidyl-PEMA. The reaction of Ac-Tyr(Me)-OH-derived oxazolone 4 in methyl phosphate-buffered aqueous solution (pH 6.5) at 20uC was monitored by HPLC and compared with the hydrolysis of mixed anhydride 2b in MES buffers (Fig. 6). Comparable rates were observed and the intermediate of the 5(4H)-oxazolone 4 reaction was identified in situ by HPLC-ESI-HRMS (negative mode, calcd for C 13 H 17 NO 7 P 2 , 330.0743; found 330.0747) as the mixed anhydride 2b. A similar behaviour was observed from a reaction of inorganic phosphate ( Supplementary  Information, Fig. S5). The hydrolysis of mixed anhydride 2b was monitored by HPLC at 20uC in buffered solutions (Fig. 6). The reaction was also carried out in D 2 O to detect any hydrogen/ deuterium exchange resulting from the transient formation of 5(4H)-oxazolone 20,28 and compared to the product of a similar reaction of pure oxazolone 4 ( Table 1). The values obtained demonstrate the occurrence of an intramolecular pathway already suspected from the higher rate of conversion of acylated aa-AMPs compared to simple acyl-adenylates 29 . At pH values below 5, the hydrolysis of anhydride 2b ( Supplementary Information, Fig. S4) has been observed to become faster in a way similar to the observation made by Lacey's group for Ac-Phe-AMP 22 . The identification of an intramolecular pathway made in the present work strongly suggests that the acid catalysis of acyl-aa-PEMA hydrolysis is the consequence of a facilitated cyclization from a good neutral phosphate leaving group. However, the absence of H/ D exchange from the reaction of neither acyl-aa-PEMA 2b nor 5(4H)-oxazolone 4 at this pH (Table 1)  Similarly, we analyzed the degree of D/H exchange during the reaction of 2b with L-Ala-NH 2 in D 2 O at pH 6.5 ( Table 1). The observation of a partial deuteration of the two diastereoisomers of the dipeptide product demonstrates that even when a better nucleophile is present, the a-proton is exchanged to a significant extent before the subsequent reaction of the 5(4H)-oxazolone takes place. The fast reaction of acyl-aa-AMP 29 and other acyl-aa-PEMA results therefore, at least for a noticeable part, from a transient conversion into 5(4H)-oxazolones. Interestingly, the different degrees of deuteration of the two diastereomers indicate that the intramolecular path of Fig. 3b has a higher stereoselectivity as compared to the direct path (the reactants 2b and 4 were prepared under a racemic form 28 ).

Discussion
As regards aa-PEMA reactions, it is noteworthy that CO 2 catalysis proceeds through a pathway involving induced intramolecularity 30 . This kind of process shares one of the most important components of enzymatic activity, which corresponds to the utilization of binding energy to non-reacting portions of the substrate to bring about catalysis 31 . It was also proposed to constitute the easiest path for enzyme evolution under the name of uniform binding 32 and is moreover necessary for enzymes to exceed a physical limit 33 . Induced intramolecularity has also been used to drive highly stereoselective catalysis in organic synthesis 34,35 . The efficiency of this kind of catalysis relies on the rates of intramolecular reactions 36 . Carbon dioxide present at total concentrations of ca. 30-40 mM in pH 6.5 solutions equilibrated with air (as deduced from the Henry's coefficient of CO 2 37 and the pK a of carbonic acid) brings about a rate increase sufficient to render the catalytic pathway largely predominating, which is remarkable by considering a simple three-atom molecule compared to the efficiency of enzymes 38 . The ease of formation of 5-membered cycles from a-amino acid mixed anhydrides is also demonstrated by the conversion of acyl-aa-PEMA into 5(4H)-oxazolones.
These experiments demonstrating that the NCA path is prevailing at pH values close to neutrality in solutions equilibrated with air at present atmospheric levels of CO 2 (ca. 0.04%) suggest that the pathway must be overwhelming in natural environments with higher contents. The experiments at 2 mM HCO 3 2 are representative of present day ocean total concentration of dissolved carbonate 39 showing that the lifetime of aa-PEMA is expressed in tens of seconds in these media at pH 7.5. In biological media, with total carbonate concentrations approaching or exceeding 10 mM, the lifetime of mixed anhydrides would be even shorter. The early atmosphere had a CO 2 content that remains poorly constrained 40 but values similar to the present atmospheric levels 41 , or representing up to hundred times this value 40,42 , are often considered. Under these conditions, aa-PEMAs would be rapidly converted into NCA before any direct conversion into peptides could take place, which discards the earlier proposed contribution of aa-AMPs in the formation of prebiotic peptides [7][8][9][10][11] . Moreover, a less efficient polymerization ability of aa-PEMA and the diketopiperazine side-reaction make them improbable peptide precursors. The possibility that a very low content of CO 2 in the atmosphere could have transiently permitted mixed anhydrides to be stabilized 23 is made unlikely because it would have also required a very efficient removal of the most part of CO 2 in the whole ocean ($2 mM in HCO 3 2 ). On the contrary, the development of the activation pathway leading to translation must have occurred in an environment in which the role of NCA was unavoidable rather than in a local environment in which the mixed anhydrides were preserved from the presence of CO 2 and HCO 3 2 by any kind of geochemical processes. NCA can be considered not only as intermediates of the degradation pathway of adenylates but also as precursors of any kind of aa-PEMA mixed anhydrides including adenylates as well as precursors of peptides through a pathway suppressing diketopiperazine side-reaction. From this point of view, the catalysis by carbon dioxide may lead to a fast exchange among different energy-rich species capable of linking activated amino acids to phosphorylating species. This distribution of energy in a reaction network, that may have anticipated the role of ATP as an energy currency, ensured a global far from equilibrium situation that was essential even at early stages of chemical evolution 43   and nucleotide chemistries 44 the CO 2 -catalyzed pathway may then constitute a key-element in the systemic integration of the two subsystems 45 .
The fast conversion of adenylates, and more generally mixed anhydrides aa-PEMAs, into NCAs at low concentrations of CO 2 in water questions the way through which the biochemical amino acid activation evolved. As a matter of fact, aa-AMPs, possibly produced from ATP through ribozyme activity 46 , would rapidly be converted into NCAs impeding the evolution of translation. Conversely, the catalytic activity of aaRSs might have evolved by acting on the thermodynamically favourable reverse reaction of aa-AMPs (formed spontaneously from NCAs) as a primitive pathway to produce ATP 2,3 . One could argue that the NCA pathway of Fig. 3a is still active in living cells but this speculation is not supported by any experimental data. However, the mechanism of pretransfer editing of misactivated aaRSs (through which adenylates are hydrolyzed) remains uncertain 47 . Any possible release of adenylates from the active site to solution 48 during this step would lead to the formation of the corresponding NCA within seconds. Whatever NCA is actually or not a biochemical metabolite, the present results indicate that living organisms probably had to limit the importance of the release of adenylates into solution after translation evolved since a conversion into NCA would certainly lead to random aminoacylation of pending amino groups likely to be harmful to protein functional integrity. From this point of view, the N-formylation of methionine needed to initiate ribosomal peptide synthesis in bacteria might be considered as a remnant of a period in which NCA could be released in the cytoplasm. Therefore, we conclude that the potential formation of NCAs at least influenced the development of the translation apparatus and that of the aaRS family of enzymes in order to avoid random aminoacylation and that the NCA pathway must be taken into account in evolutionary studies.
Our analyses confirm the observations made by Lacey that CO 2 is a very efficient catalyst for the conversion of adenylates. However, taking into account the probable role of NCAs and the diversity of processes made available through their intermediacy leads us to the very different conclusion that the process could be favourable to the development and evolution of life rather than solely detrimental to the role of adenylates as intermediates of peptide formation. It is also worth noting that acyl-aa-PEMA that were considered by Lacey as blocked equivalents of aa-AMPs 22,23 does actually not constitute models of the reactivity of their parent compounds since they also undergo a spontaneous cyclization into 5(4H)-oxazolone. The transient formation of 5(4H)-oxazolone intermediates may be responsible for their efficiency in peptide formation 20 . The mixed anhydrides formed from free amino acids as well as peptide segments turn out to constitute unlikely precursors of peptides since their reactions are actually preceded by a very efficient cyclization into uncharged intermediates that thus constitute better electrophilic agents. This observation can be related to the evolutionary advantage of phosphate derivatives 49 that is partly related to their negative charge reducing spontaneous hydrolytic degradation with respect to their enzymepromoted reactions. From this perspective, their involvement required specific and efficient catalysts. However, the fact that NCA and 5(4H)-oxazolone also constitute precursors of mixed anhydrides through spontaneous processes provides a potential path through which these intermediates may have led for example to aminoacyl esters of RNA at predisposed locations 16,23,50 .