Introduction

The low reactivity of RNA nucleotides has for long been the greatest hurdle that impeded experimental studies aimed at understanding how the first informational polymers could assemble from their monomeric precursors. To overcome this problem, various highly efficient synthetic methods have been invented for the chemical modification of nucleotides. Among them, conversion to phoshoimidazolides (Beaucage and Caruthers 1981; Hill et al. 1993; Schrum et al. 2009; Szostak 2012) represents the perhaps most efficient way for activation of the phosphate group, and, therefore, it has become a widespreadly used method to study non-enzymatic template-directed oligomerization of nucleotides. Likewise, a number of other studies have been reported that utilise carbodiimides for the same purpose (Zielinski and Orgel 1987; von Kiedrowski et al. 1991; Jauker et al. 2015).

Recent synthetic efforts aimed at reconstructing life’s origin on our planet suggest that 2′,3′ and 3′,5′ cyclic nucleotides could have accumulated in a sufficiently large amount on the early Earth to serve as the building blocks of the first RNA-oligomers (Powner et al. 2009; Saladino et al. 2012b). Incorporation of the phosphodiester bonds in a cyclic ring offers perhaps the simplest way for activation of the phosphate group for transphosphorylation reactions leading to intermolecular phosphodiester linkages (Saladino et al. 2012a; Sponer et al. 2015; Powner et al. 2011). Albeit this way of the nucleotide activation is highly plausible in a prebiotic context, it has been found to be less efficient than those methods which are based on chemical modification of the nucleotide precursors, like activation in the form of phosphoimidazolides or isoureates. Nonetheless, it has been found that cyclic phosphates exhibit a sufficiently high activity in the presence of amines, (Verlander et al. 1973; Usher and Yee 1979) which are compounds highly abundant among the products of the volcanic Miller-experiment (Parker et al. 2011). In addition, a recent study on the template-free non-enzymatic oligomerization of a mixture of 5′-AMP and 5′-UMP (Da Silva et al. 2015) suggests that ammonium cations can also contribute to the activation of the phosphate group of nucleotides.

In the current work, we try to provide a theoretical rationale of those template-free oligomerization scenarios that could stay behind the spontaneous emergence of the very first RNA-oligomers from chemically unmodified precursors, i.e. either acyclic or cyclic nucleotides. Thus, the primary scope of our study is to shed light on the role played by amines, protons and cations in the mechanism of these reactions, a question that for many years has stayed in the focus of attention of the origin of life research.

Crystal Structure May Control the Steric Conditions of Template-Free Non-enzymatic Polymerizations Performed Under Dehydrating Conditions

The historical papers by Verlander et al. (Verlander et al. 1973) and Usher and Yee (Usher and Yee 1979) report on the formation of short oligonucleotides from 2′,3′ cyclic nucleotides upon heating under dehydrating conditions (see Scheme 1). In both cases, the synthesis started from the H-form of the cyclic nucleotide and was conducted in the presence of diamines or amino alcohols in dry solid samples. Metal cations apparently blocked the polymerization reaction. These two synthetic strategies led to mixed 2′,5′- and 3′,5′-linked oligomers, the ratio of the 2′,5′ and 3′,5′-linkages being ca. 1:2 in the oligomers formed. Prevalence of the 3′,5′-linkages was explained by the more advantageous steric conditions of the in-line attack dictated by the crystal structure (Usher and Yee 1979).

Scheme 1
scheme 1

Oligomerization of 2′,3′ cyclic nucleotides in the presence of diamines or amino alcohols (Verlander et al. 1973; Usher and Yee 1979)

Full size image

As Verlander et al. (Verlander et al. 1973) show, 2′,3′ cyclic AMPs (hereafter 2′,3′cAMP) in acid form do not polymerize on their own. The only successful strategy to facilitate the polymerization is to shift the pH of the 2′,3′cAMP solution used for the preparation of the dry samples to basic. This is achieved by adding various diamines and amino alcohols to the reaction mixture. The same approach was followed also in the experiments of Usher and Yee (1979). Under these conditions, the charge compensating species are protonated amines instead of protons. In the acid form of the dry material, protons are either bound to the phosphate of the nucleotide or protonate the bases. As a result, packing of the nucleotides might be completely different for the acid and cationic forms of the crystals (Sundaralingam and Prusiner 1978). For example, crystal packing of 5′ deoxycytidine-monophosphate is completely different in the acid and sodium forms [CSD-codes: DOCYPO03 (Pearlman and Kim 1985) and CITMOH01 (Perras et al. 2012), respectively] due to protonation of N3 of cytosine in the acid form. Likewise, the N1 position of adenine has a great affinity to bind protons: crystal data available in CSD (Allen 2002) for various adenine-nucleotides demonstrate that in H-form the N1 position of adenine is always protonated; see e.g. AMADPH10 (Silverton et al. 1982), ADPOSD (Sundaralingam 1966), ADPOSM (Kraut and Jensen 1963). On the contrary, addition of diamines to the acid form of the 2′,3′cAMP solution stabilises adenine in its non-protonated form. Under these conditions, the charge compensating species are protonated amines both in solution and, consequently, in the dry material used for the polymerization experiment. Most likely the crystal packing of 2′,3′ cAMP in the presence of protonated amines is entirely different from the one in the H-form and, in contrast to the H-form, it is able to structurally support the transphosphorylation between the nucleotide precursors.

The Role of Amines and Amino Alcohols

The catalytic effect of 2-aminoethanol on the polymerization of 2′,3′cAMP was interpreted by the formation of a 2-aminoethyl phosphodiester intermediate (see Scheme 2), (Verlander et al. 1973) which then readily undergoes a transesterification by the 5′ hydroxymethyl group of another cAMP (see reaction 4 in Scheme 3) (Verlander et al. 1973). This essentially leads to a 3′,5′- or 2′,5′- linkage. The latter reaction proceeds in a weakly basic solution at pH = 9, i.e. exactly at the pKa of the amino group of 2-aminoethanol. Under these conditions, ca. half of the amino groups are in protonated form, while the other half is non-protonated. Therefore, in order to envisage the catalytic effect of 2-aminoethanol on the oligomerization of 2′,3′cAMP, we have considered two transphosphorylation scenarios, involving a 2-aminoethyl phosphodiester intermediate with protonated and non-protonated amino groups.

Scheme 2
scheme 2

Phosphodiester intermediate formation with 2-aminoethanol according to Verlander et al. (1973)

Full size image
Scheme 3
scheme 3

Model chemistries used to describe various non-enzymatic template-free polymerisation scenarios. Reaction between two acyclic nucleotides in anionic form, i.e. uncatalysed case (1); the same in acidic environment, i.e. assuming neutral reactants, relevant to Ref. (Da Silva et al. 2015) (2); reaction between two anionic 2′,3′ cyclic nucleotides from Refs. (Verlander et al. 1973) and (Usher and Yee 1979) (3); reaction between two 2-aminoethyl phosphodiesters (Verlander et al. 1973) assuming that the amino group of the substrate is protonated (4); transphosphorylation between a nucleotide and a 3′,5′ cyclic nucleotide, relevant to the ring-opening polymerisation mechanism described in detail in Ref. (Sponer et al. 2015) (5). On the left side: the complete model; on the right: simplified models used for computations. Note, that model 1 can also be used to describe the reaction between two nucleoside-3′-phosphates (not shown in the scheme), i.e. attack of a 5′-hydroxyl at a 3′-phosphate

Full size image

The thermodynamics of a transphosphorylation reaction leading to phosphodiester linkages between nucleotides can be evaluated based on the reaction free energy change that accompanies the formation of its trigonal bipyramidal intermediate (hereafter denoted as ΔG). This high-energy species is very close in energy to both neighbouring transition states (i.e. that associated with the attack of the 5′-hydroxymethyl group at the phosphorus, as well as the one leading to the product) (Sponer et al. 2015; Liu et al. 2005). Therefore, ΔG provides with a fairly good assessment of the kinetics of the reaction. We have found a dramatic difference between the ΔGs when considering transphosphorylation between the amino-protonated and non-protonated variants of two 2-aminoethyl phosphodiesters (for the amino-protonated variant see reaction 4 in Scheme 3). ΔG computed for the non-protonated variant (33.0 kcal/mol in gas-phase) was very close to the one computed for the reaction of methanol with methyl-phosphate, representing the transphosphorylation between two acyclic nucleotides in the non-catalysed case (33.9 kcal/mol in gas-phase, reaction 1 in Scheme 3). In contrast, considering N-protonated amines, the gas-phase ΔG is much lower, 24.4 kcal/mol. In the language of organic chemistry this means that localising a positive charge next to the anionic oxygen of the phosphate moiety increases the positive charge on the phosphorus and thereby makes the thermodynamics of the nucleophilic attack less demanding. Such proton transfer can take place only in a less polar environment. To illustrate this, we carried out three sets of additional calculations for reaction 4. In the first set, we computed ΔGs with the C-PCM dielectric continuum method (Barone and Cossi 1998), (Cossi et al. 2003) using the standard parametrisation for water, i.e. assuming a dielectric constant ε = 78.4. We performed additional computations choosing ε = 20.0 and ε = 5.0, while keeping all other parameters of the solvation model unchanged. The purpose of these calculations was to describe the situation when the incomplete water-shell around the reactants (relevant to gradual desiccation of the environment) compensates only partially for the electrostatic interactions. While ε = 20.0 qualitatively represents the electrostatic shielding exerted by many dipolar aprotic solvents (e.g. ethanol, propanol, acetone), ε = 5.0 captures the situation in the case of even lower compensation of the electrostatic interactions, roughly equal to that observed in apolar solvents (e.g. chloroform, benzene, toluene). Whereas data in the first and fourth rows of Table 1 show that at ε = 78.4 and ε = 20.0, the ΔGs computed for reactions 1 and 4 are pretty similar, at ε = 5.0 ΔG computed for reaction 4 is significantly lower (34.2 kcal/mol) than that obtained for the reference reaction 1 (42.1 kcal/mol). In other words, a perfect hydration shell around the solute stabilises the proton on the amino group rather than on the negatively charged oxygen of the phosphate, i.e. no positive charge is transferred onto the phosphorus that would make the thermodynamics of the transphosphorylation more favourable. This suggests that 2-aminoethanol is able to fill its catalytic function exclusively under dehydrating conditions.

Table 1 B3LYP/6-311++G** reaction-free energies (ΔG, kcal/mol) for various transphosphorylation scenarios as a function of the dielectric constant (ε) of the solvent
Full size table

Obviously, it is reasonable to expect that direct protonation of the anionic oxygen of the phosphate would have the same effect on ΔG as protonation of the 2-aminoethyl group connected to the phosphate as long as the crystal packing structurally supports the oligomerization (i.e. unless protonation of the base changes the crystal packing of the nucleotides, see above). Indeed, this has recently been observed in the mixed polymerization of 5′AMP and 5′UMP under simulated hydrothermal conditions that was conducted in an acidic environment at pH = 2 (see reaction 2 in Scheme 3) (Da Silva et al. 2015; DeGuzman et al. 2014; Rajamani et al. 2008). In the neutral form of the nucleotide, two protons are connected to phosphate oxygens: in the trigonal bipyramidal intermediate one of them is linked to an equatorial oxygen, while the other to an axial oxygen. The equatorial one serves exactly the same role as the ammonium group of the aminoethyl moiety of the phosphodiester intermediate formed with 2-aminoethanol: by transferring a partial positive charge to the phosphorus it makes the intermediate formation energetically more favourable. This is reflected by the markedly lower ΔGs computed for reaction 2 as compared to reaction 1 (compare the corresponding data in the first two rows of Table 1).

Cyclic Ring Formation as a Means of Nucleotide Activation

In addition to protonation, cyclic ring formation is exploited for activation of nucleotides for non-enzymatic oligomerization. We have considered two plausible scenarios: the historical works of 2′,3′cUMP (Usher and Yee 1979) and 2′,3′cAMP (Verlander et al. 1973) polymerization in the presence of diamines as well as the ring-opening polymerization observed for 3′,5′cGMPs (Costanzo et al. 2009; Sponer et al. 2015; Costanzo et al. 2012; Morasch et al. 2014). Since the latter mechanism involves anionic reactants, the role of solvent environment is tremendous at quenching the electrostatic repulsion between them. The possible effect of environments with various degrees of hydration on ΔG was tested again by reducing the dielectric constant of the solvent from the experimentally determined value for water ε = 78.4 through 20.0 to 5.0. As the third and fifth rows of Table 1 show assuming ε = 78.4 and ε = 20.0, the computed ΔGs are very similar both for the 2′,3′- and 3′,5′-cyclic rings. In contrast, at ε = 5.0 the data computed for the anionic mechanism (i.e. reaction 5) are markedly higher (46.5 kcal/mol) and similar to the corresponding data computed for the reference reaction 1 (42.1 kcal/mol). This illustrates that steric constraints imposed by cyclic ring formation have similar energetic consequences on the formation of the trigonal bipyramidal intermediate as protonation of the anionic oxygens of the phosphate does, supposed that the environment is able to reduce the possible electrostatic repulsion between the reactants. Note, that the crystal of 3′,5′cGMP is never fully dry, it crystallises with 16 water molecules/unit cell. (Chwang and Sundaralingam 1974)

Resolving the Mystery of Metal Cations

Using the same logic one might think that metal cations could fill the same role as protons do. This contrasts the experimental observation that except for the mixed polymerization of 5′AMP and 5′UMP no activity is observed in any of the reported prebiotically relevant template-free polymerization mechanisms in the presence of metal cations. With regard to solid-phase polymerization of 2′,3′cUMP, this was explained by the higher lattice energy and thus lower mobility of the cyclic nucleotides in the presence of sodium counterions (Usher and Yee 1979). In the case of the anionic ring-opening polymerization reported for 3′,5′cGMP in solution and under dehydrating conditions, the reason is more obvious: highly mobile cations neutralise the anionic centres formed at the 3′-oxygens of the sugars that maintain the oligomerization (Sponer et al. 2015). On the contrary, in strongly acidic solution, metal cations do not interfere with the polymerisation (Da Silva et al. 2015). In our opinion, this is because the anionic oxygens of the phosphate have higher affinity to bind protons than metal cations. In addition, by increasing the conductivity, metal cations may slightly increase the reaction rates, as was also observed experimentally in the mixed polymerization of AMPs and UMPs (Da Silva et al. 2015). A more pronounced catalytic effect was reported for NH4 + cations, not surprisingly, because they are excellent mediators of proton transfer processes also involved in the discussed transphosphorylation reactions.

Conclusions

In general, the above considerations illustrate that protonation or steric strain due to cyclic ring formation is a viable way of phosphate activation for transphosphorylation reactions leading to oligonucleotides. Our computations also show that the extent of activation is very similar both ways. Likewise, amines may facilitate oligomerization of chemically unmodified nucleotides. The role of amines in these processes might be direct or indirect, depending on whether they actively participate in the activation or just prevent nucleobase protonation, which would lead to crystal structures not compatible with the required transphosphorylation reactions. The fact that either simple amines or ammonia participate in all known non-enzymatic template-free oligonucleotide syntheses in some form suggests that amines could play a central role in the emergence of an RNA world, perhaps being biology’s first cofactors, buffers and charge compensating cations. This puts forward the idea of the existence of a primitive chemical pre-RNA world which, beyond phosphates, was based solely on the four basic ingredients of organic chemistry: C, H, N, O.