Generation of Long RNA Chains in Water*

The synthesis of RNA chains from 3′,5′-cAMP and 3′,5′-cGMP was observed. The RNA chains formed in water, at moderate temperatures (40–90 °C), in the absence of enzymes or inorganic catalysts. As determined by RNase analyses, the bonds formed were canonical 3′,5′-phosphodiester bonds. The polymerizations are based on two reactions not previously described: 1) oligomerization of 3′, 5′-cGMP to ∼25-nucleotide-long RNA molecules, and of 3′,5′-cAMP to 4- to 8-nucleotide-long molecules. Oligonucleotide A molecules were further extended by reciprocal terminal ligation to yield RNA molecules up to >120 nucleotides long and 2) chain extension by terminal ligation of newly polymerized products of 3′,5′-cGMP on preformed oligonucleotides. The enzyme- and template-independent synthesis of long oligomers in water from prebiotically affordable precursors approaches the concept of spontaneous generation of (pre)genetic information.

The origin of informational polymers is not understood. The RNA polymerization process has been studied for five decades, the results showing that from preactivated precursors polymers of several tens can be obtained, as reviewed previously (1). These pioneering studies provide the proof-of-principle that RNA precursors can self-assemble yielding linear polymers. However, the prebiotic validity of a process based on complex preactivation procedures is limited (1,2), and the problem of defining a prebiotically plausible chemical and thermodynamic scenario for the synthesis and accumulation of informational polymers remains open. The core of the problem is the standard state Gibbs free energy change (3,4) stating that condensation reactions are very inefficient in water. Given that extant polymerizations occur in water, this is a major difficulty, only partially solved by the fact that these processes at present occur inside the active site of enzymes where water activity may be drastically reduced. The other part of the extant solution, fruit of evolution, is the use of biologically highly preactivated triphosphate nucleotides (3). In primordia, RNA molecules had no enzymes to catalyze their chain-wise growth, and highly activated precursors can be considered as prebiotic only with difficulty.
We reasoned that for a pre-enzymatic polymerization to occur the solution must have relied on a simple and robust process. Ideally, such a process should have been based on com-pounds that were reactive yet relatively stable, chemically not too elaborate to allow their efficient production, and not too dissimilar from the products of their polymerization to minimize the chemical cost of the process.
It was observed that phosphorylation of nucleosides occurs in formamide simply in the presence of a source of organic or inorganic phosphate at temperatures at which both the reactants and the products are stable (5). Phosphorylation occurs in every possible position of the nucleoside sugar moiety resulting, both for purine and pyrimidine nucleosides, in the production of 2Ј-, 3Ј-, 5Ј-, 2Ј,3Ј-cyclic, and 3Ј,5Ј-cyclic XMPs 3 (5). The phosphorylation reaction is faster for the open than for the cyclic forms, whereas higher stability of the cyclic forms at higher temperature favors their accumulation.
Coupled with the facile synthesis of all the nucleic bases from formamide (6) and with the formation of acyclonucleosides by TiO 2 -catalyzed formamide photochemistry (7), the nonenzymatic phosphorylation of nucleosides (5) shows that the formation of cyclic monophosphate nucleosides is chemically simple and prebiotically plausible. The formation of both 2Ј,3Ј-and 3Ј,5Ј-cyclic XMPs in water starting from nucleosides and an inorganic source was also observed (8).
The unsophisticated chemistry required for the formation of both open and cyclic nucleotides prompted us to investigate the possibility of their spontaneous polymerization. If so, nonenzymatic (pre)genetic polymerization could have taken place in warm little pond conditions, close to those imagined by Darwin (9).
Oligonucleotides-The oligonucleotides 5ЈA 24 3Ј, 5ЈC 24 3Ј, 5ЈA 12 C 12 3Ј, 5ЈA 12 U 12 3Ј, 5ЈU 24 3Ј, and 5ЈG 24 3Ј were purchased from Dharmacon and were provided unphosphorylated, at both the 5Ј and 3Ј extremities. solutions of the appropriate nucleotide (2Ј-AMP, 3Ј-AMP, 5Ј-AMP,  2Ј,3Ј-cAMP, 3Ј,5Ј-cAMP, 3Ј,5Ј-cGMP, 3Ј,5Ј-cUMP, and 3Ј, Acrylamide Gel Electrophoresis-Standard methodologies were used, with the following specifications: 1) 12% polyacrylamide was used in analyses encompassing the whole product of the polymerization reaction, from the 32 P-labeled monomer to the highest molecular weight fragments (Ͼ100 units), or 2) longer runs on 16% polyacrylamide gels were used for the analysis of low molecular weight polymers. With sequences allowing good resolution, the average chain length (N avg ) of the oligomers was determined by the equation N avg ϭ ⌺ i n i N i /⌺ i n i , where n i is the number of chain (in %) and N i is the length of RNA chains in nucleotides.

Polymerization Protocols and Analysis-Concentrated
Reference Ladders-The nucleotide ladders used as standard in the gel-electrophoretic analyses of the polymerization products consisted of partially hydrolyzed 24-mer Poly(G) or Poly(A) (Dharmacon), as appropriate. Products of combinatorial ligation of preformed oligonucleotides were also used as markers, obtained as detailed in a previous study (10). In practice, labeled 5ЈA 24 3Ј was mixed with unlabeled 5ЈA 15 3Ј yielding fragments of 39, 48, 69, 78, and 96 nucleotides in length.
For details, handling and analysis of the RNA hydrolytic products see Ref. 11. In brief, terminally labeled RNA oligonucleotides were hydrolyzed in water at 90°C for different time periods (between 0 and 24 h) and pre-analyzed on polyacrylamide gel.
Terminal Labeling of the Material Polymerized from Unlabeled Cyclic Nucleotides-The products of the polymerization reactions from cyclic nucleotides were ethanol-precipitated and dissolved in 44 l of water. For de-phosphorylation, 1 l of shrimp alkaline phosphatase (1 unit/l, MBI Fermentas) was added along with 5 l of 10ϫ shrimp alkaline phosphatase buffer, and the reaction was incubated at 37°C for 30 min, followed by phenol extraction and ethanol precipitation. Glycogen (1 l of stock 20 mg/ml) was added to facilitate precipitation. RNA was pelleted by centrifugation, then dissolved in 16 l of water and labeled at the 5Ј termini with 32 P. Phosphorylation was carried out by adding 1 l of T4 polynucleotide kinase (T4 PNK, 10 units/l, New England Biolabs), 2 l of 10ϫ PNK buffer and 0.5 l of [␥-32 P]ATP, followed by incubation at 37°C for 30 min. For gel electrophoresis, 10-l aliquots of the RNA samples were resuspended in 100% formamide and separated by electrophoresis on 12 or 16% polyacrylamide gels containing 7 M urea, along with the indicated markers.
Nuclease P1 from Penicillium citrinum (International Union of Biochemistry 3.1.30.1) is from Sigma (Cat N8630), specific activity of 200 units/mg of protein. It catalyzes the sequence nonspecific endonucleolytic cleavage of single-stranded RNA to yield nucleoside 5Ј-phosphates and 5Ј-phospho-oligonucleotides. Specific for 3Ј,5Ј-phosphodiester linkages, it is here typically used at 20 units/sample in 40 mM Tris-HCl, pH 5.4, 5 mM NaCl, 0.5 mM MgCl 2 , in 20-l assays. One unit liberates 1.0 mol of acid-soluble nucleotides from RNA per minute at pH 5.3 at 37°C. T1 from Aspergillus oryzae (EC 3.1.27.3) is a 3Ј-5Ј-specific ribonuclease. It cleaves with high preference at the 3Ј-end of G residues but at high concentration or at longer times will cleave also at other residues (12). One unit produces acid-soluble oligonucleotides equivalent to a ⌬A 260 of 1.0 in 15 min at pH 7.5 at 37°C in a reaction volume of 1 ml.

RESULTS
The oligomerization capacity of the cyclic forms (2Ј-3Ј or 3Ј-5Ј) of the four monophosphate nucleosides, guanosine, adenosine, cytidine, and uridine, was tested. The open nucleotides 5Ј-AMP, 3Ј-AMP, and 2Ј-AMP were also tested in water at temperatures between 40 and 90°C. A number of additional variables were analyzed: concentration, time, addition of formamide (from 0 to 100%), presence of several minerals known to catalyze phosphorylation (5) or to increase the half-life of nucleic polymers (11,(13)(14)(15), addition of Na 4 P 2 O 7 or Na 5 P 3 O 10 , and combinations thereof. Of all the conditions tested, the simplest proved to be the best: water between 40 and 90°C. Several pH values (3.2, 3.7, 5.0, 5.4, 6.1, 8.0, 8.2, and 8.4) were tested. The results observed were marginally different. The afforded polymers were 5Ј-terminally labeled with [␥-32 P]-ATP by T4 polynucleotide kinase, and the products were characterized by gel electrophoresis, allowing detailed evaluation of the lower sized oligomers.

Syntheses from Open Nucleotides
No product of polymerization was observed upon incubation of 2Ј-AMP or 3Ј-AMP in water (nor in any of the reaction variants listed above) at temperatures encompassed between 40 and 90°C for periods up to 400 h. Only degradation of the input nucleotides was observed (data not shown). 5Ј-AMP afforded only traces of oligomerized compounds whose total did not exceed 0.5% of the input (data not shown). The short half-life of 5Ј-AMP at 90°C (35 h) (11) is not compatible with the possibility of accumulating oligomers.

Syntheses from Cyclic Nucleotides
3Ј,5Ј-cGMP- Fig. 1 shows the products of polymerization obtained by treating 3Ј,5Ј-cGMP in water. The formation of oligomers is evident. 3Ј,5Ј-cGMP polymerized into RNA chains that reached a size of at least 25 nucleotides, the predominant oligomer being 8-mer. Panel A reports the synthesis obtained at 85°C as a function of the 3Ј,5Ј-cGMP concentration, showing that, above the optimal concentration of 1 mM, chain elongation is impaired and the preferentially formed 8-mer accumulates. Panels B and C show the syntheses obtained at the optimal 1 mM and at the highest possible (before aggregation) 100 mM concentration as a function of the temperature. In both cases the highest temperature tested was the most favorable for chain extension. Below 60°C the reaction rate dropped rapidly (data not shown).
The oligomers shown are the products of synthetic reactions lasting 1 h. In kinetic analyses it was observed that at the optimal concentration (1 mM) synthesis was fast, an N avg of 11.8 being reached during handling time (Ͻ1 min), followed by slow stepwise further growth. The kinetic constant of this further growth was determined by measuring the N avg of the oligonucleotide G chains formed as a function of time at 85°C with 1 mM 3Ј,5Ј-cGMP and was 0.4 ϫ h Ϫ1 .
3Ј,5Ј-cAMP-Under the same conditions of the 3Ј,5Ј-cGMP polymerization, 3Ј,5Ј-cAMP polymerized by a two-step mechanism. Fig. 2 shows the two steps observed in a 3Ј,5Ј-cAMP-fed growth experiment. First, a family of short oligomers was synthesized rapidly. The steady-state N avg of 5.32 (Fig. 2, lane 1) was reached by 60 min (50% of molecules formed in 20 min). The kinetic constant of the reaction leading to the formation of the short oligonucleotide A molecules (N avg 5.32) was determined at 85°C and was 2 ϫ h Ϫ1 . The short oligomers did not continue growing by slow ladder-wise addition, as for 3Ј,5Ј-cGMP, but extended their size forming a heterogeneous population (Fig. 2, lane 3) in which a rapidly formed 16-mer was prominent. Sequence extension lasted 200 h, forming molecules Ͼ100 nucleotides long (Fig. 2, lane 4). The distribution of the products of oligomerization beyond 28 nucleotides in length was size-discontinuous (see the numbering at the side of lane 4), comprising a complex series of fragments. Such heterogeneous numerical distribution is best interpreted as the result of ligation of shorter pieces. A model study (10) showed that mixing a limited number of different RNA oligomers in water yields a complex population of differently sized RNA fragments by nonenzymatic ligation. This second reaction, presumably based on ligation of the components of a heterogeneous population, is too complex to allow calculation of kinetic constants. By contrast, 2Ј,3Ј-cAMP yielded only short oligomers, up to tetramers (data not shown). Polymerization of 3Ј,5Ј-cUMP and 3Ј,5Ј-cCMP yielded only short fragments (N avg 5.49 and 5.45, respectively) at 85°C, which did not grow further.

The Bonds Formed, as Determined by RNase Analyses
The type of phosphate bond formed in the polymers derived from 3Ј,5Ј-cGMP and 3Ј,5Ј-cAMP was analyzed by enzymatic digestion with SVPD I (EC 3.1.4.1, a 5Ј-exonuclease cleaving 3Ј-5Ј and 2Ј-5Ј phosphodiester bonds from the 3Ј-extremity in a nonprocessive manner) and with P1 endonuclease (EC 3.1.30.1, a 3Ј-5Ј-specific ribonuclease). Treatment of the products of polymerization with 1 milliunit of SVPD I or of P1 for 20 min at 37°C completely converted the oligonucleotides into monomers, showing that the bonds formed are canonical 3Ј-5Ј phosphodiester bonds (data not shown). For details of these RNase assays, see Ref. 10. The type of phosphate bond formed in oligonucleotide G was further analyzed, as described below, confirming the formation of 3Ј-5Ј bonds.

On the Mechanism of Polymerization
Although detailed mechanistic aspects of the observed polymerization of cyclic nucleotides are beyond the aim of the present communication, the following facts elucidate the basics of the reaction: (i) the RNase digestion assays mentioned above show that the bonds formed by polymerization of 3Ј,5Ј-cyclic nucleosides are standard 3Ј-5Ј phosphodiester bonds. Given that the starting monomers are 3Ј,5Ј-cyclic phosphates, this is not unexpected. The combined SVPD I and P1 RNase analyses rule out the formation of 2Ј-5Ј bonds, of pyrophosphate bonds, or more complex alternatives. (ii) 3Ј,5Ј-cyclic nucleoside monophosphates hydrolyze in water yielding (in the temperature and pH conditions in which polymerization occurs) a mixture of 5Ј and 3Ј monophosphates, as verified by high performance liquid chromatography (data not shown) and as originally reported (16).
Thus, the polymerization could occur according to two different alternative models. Model A consists of the reactive species that is a 5Ј-XMP afforded by the opening of the 3Ј phosphodiester bond of the cyclic nucleotide. In this case, polymerization would occur via the 5Ј-phosphate reacting with the 3Ј-OH of another 5Ј-XMP, as indicated by the spark symbol in Fig. 3. The reactive species is a 3Ј-XMP, and the polymerization occurs via the 3Ј-phosphate reacting with the unphosphorylated 5Ј-extremity of another 3Ј-XMP molecule. Model A would lead to the phosphate group being on the top sugar molecule (as shown in Fig. 3), rather than on the lower sugar molecule (Model B, not shown).
The bias would be solved in favor of Model A if neo-formed oligonucleotide G, obtained as described in Fig. 1, would ligate to the 3Ј non-phosphorylated extremity of an acceptor oligonucleotide through 3Ј-5Ј phosphodiester bonds (as schematically described in Fig. 5). The experiments reported below (Figs. 4 -6) show that this is the case: the neo-formed oligonucleotide G ligated with 3Ј-5Ј bonds to the 3Ј-OH extremity of a 5ЈC 24 3Ј and of a 5ЈA 12 C 12 3Ј oligomer. Thus, Model A applies, as shown in Fig. 3.
In summary, in the presence of the thermodynamic driving force provided by stacking interaction, an isoenergetic phosphodiester exchange reaction is favored, affording the observed products. The possibility that the reaction occurs by general acid-base catalysis is disfavored by the observation that neither the 3Ј,5Ј-cAMP nor the 3Ј,5Ј-cGMP polymerizations are pHdependent (between pH 3.2 and 8.4, data not shown).
The fact that the order of the stacking potentials of the bases correlates with the corresponding polymerization rates (see below) establishes the relevance of stacking interactions in this reaction.

JOURNAL OF BIOLOGICAL CHEMISTRY 33209
key observation is that 3Ј,5Ј-cGMP actively reacted with the preformed oligonucleotide, affording longer fragments. In particular, a group of molecules with a number average (N avg ) of 42 formed in the presence of 3Ј,5Ј-cGMP (lanes 6 -8), that grew up to an observed length of Ͼ50 nucleotides in the presence of the higher concentration of cyclic nucleotide (as counted in the right corner inset, showing a lower exposure of the relevant gel position). A slower migration band is also observed in the upper part of the lanes 6 -8 (asterisk), probably representing a dimeric form of the extended sequence.
The N avg was calculated from graphical extrapolation of gel positions in the appropriate autoradiographic exposures. The band-compression effect characteristic of the C residues prevents a better resolution of high molecular weight oligomers and a more precise evaluation of fragment lengths. The system was explored with higher precision in 5ЈA 12 C 12 3Јpolymers (see below).
All the 5ЈC 24 3Ј fragments covalently reacted with oligonucleotide G oligonucleotides (lanes 7 and 8) and formed a new population reaching an average length of 42. This entails that in the solution in which the reaction takes place oligonucleotide Cs and oligonucleotide Gs interact, presumably by base-pairing, to form a double strand. Double strands withstand hydrolysis more than single strands. If this occurs also in our conditions and sequence set-up, a footprint of ϳ18 bases in length should be produced, which is actually observed (Fig. 4A,  The following are also noted: 1) The C stretch is highly sensitive to hydrolytic degradation (as already reported (17)); 2) 3Ј,5Ј-cAMP does not support polymerization growing on the 3Ј-extremity nor supports multimerization by ligation (as observed for 5ЈA 24 3Ј oligonucleotides (10)). Starting at 10 mM concentration, 3Ј,5Ј-cAMP enhances the hydrolytic degradation of the 5ЈC 24 3Ј oligonucleotide (lane 5). The same behavior was observed on Poly(A) 23 U, Poly(A) 24 , and on Poly(G) 24 (data not shown); 3) 3Ј,5Ј-cCMP and 3Ј,5Ј-cUMP are inert. Thus, only the reaction of oligonucleotide C with 3Ј,5Ј-cGMP was explored further. Fig. 4B shows the RNA-chain extension of 5ЈC 24 3Ј by 3Ј,5Ј-cGMP as a function of cyclic nucleotide concentration. Panel C shows selected examples of the same reaction on 5ЈA 12 C 12 3Ј. Consistent with the calculated N avg of the oligonucleotide G polymerized from 3Ј,5Ј-cGMP reported in Fig. 1 (in synthesis reactions in which the 8-mer was prevailing), the family of oligonucleotide Gs that polymerized from 3Ј,5Ј-cGMP in the presence of the 5ЈA 12 C 12 3Ј 24-mer and that ligated to its 3Ј C-extremity had an N avg of 8.75 (Fig. 4C). This N avg value was determined from the N avg calculated from the fragment sizes observed in the gel migration ladder (N avg ϭ 32.75) subtracting 24 (that is, the size of the acceptor 24-mer oligonucleotide).
The following is also noted: the footprint on the C 12 moiety is shorter relative to the one on the C 24 oligonucleotide, similar to the chains produced (N avg ϭ 32.75, corresponding to an extension of 8.75 on the 24-mer and to a footprint Ն8 residues, as indicated by dots) and as predicted in a model based on the Poly(C)-Poly(G) base-pairing in water. 3Ј,5Ј-cAMP, 3Ј,5Ј-cCMP, and 3Ј,5ЈcGMP did not support chain extension on the 5ЈA 12 C 12 3Ј (nor on the 5ЈC 24 3Ј; data not shown).
The pre-synthesized oligonucleotide G did not bind to (nor did 3Ј,5Ј-cGMP-fed polymerization occur on) pre-synthesized Poly(A) oligonucleotides (data not shown), thus excluding that the 5ЈA-extremity of the 5ЈA 12 C 12 3Ј molecule supported RNAchain extension on the Poly(C) oligonucleotides.
The fact that a footprint is observed, starting from the position in which sequence extension begins (i.e. the 3Ј-extremity) and is oriented in the specular direction, provides an assay for the presence of newly formed complementary sequences. No footprint is observed on the 5Ј-extremity, indicating that sequence extension only occurs on the 3Ј-OH extremity based on the 5Ј P-group from the incoming molecule, and not vice versa.
A quantitative evaluation of the RNA-chain extension occurring on 5ЈC 24 3Ј and on 5ЈA 12 C 12 3Ј as a function of the cyclic nucleotide concentration is reported in Fig. 4D. The plot shows that the growth of short segments occurring on 5ЈA 12 C 12 3Ј (N avg ϭ 8.75) levels off at lower concentration of 3Ј,5Ј-cGMP, relative to the growth on 5ЈC 24 3Ј (N avg ϭ 18).
The kinetic constant of the reactions leading to the formation of the extended monomers could not be determined, because the reaction was too fast even at the lowest concentration tested (200 nM 5ЈC 24 3Ј and 1 M 3Ј,5ЈcGMP), at 40°C. Reaction rates are given in Table 1.
In conclusion, 3Ј,5Ј-cGMP efficiently polymerizes in the presence of Poly(C) and is covalently bound to its 3Ј-extremity. Given that the 3Ј-extremity of the 5ЈA 12 C 12 3Ј oligonucleotide bears no phosphate but ends by OH in 3Ј, and given that the ligation occurred via 3Ј-5Ј phosphodiester bonds (see below, section on Characterization of the Bond Forma-  6 -8), 3Ј,5Ј-cCMP (lanes 9 -11), or 3Ј,5Ј-cUMP (lanes 12-14). Each group of three lanes contained 0.1, 1.0, or 10 mM nucleotide, respectively. The reaction was in Tris-HCl-buffered water (pH 5.4) at 60°C for 6 h. Lane 1: U, untreated; lane 2: no nucleotide. Inset: one-third autoradiographic exposure of the corresponding part of the gel. The model interprets the structure of the polymer in denaturing (left) and water condition (right). The polymer indicated by the asterisk is formed only in the presence of 3Ј,5Ј-cGMP. Its size is 84 Ϯ 3 nucleotides (as determined by band counting in the appropriate autoradiogram exposure and plot graphical extrapolation). This multimer is interpreted as a dimeric form of the extended monomer, possibly caused by oligonucleotide G-oligonucleotide G ligation. Oligonucleotide C-oligonucleotide C dimerization did not occur in the absence of G-based cofactors or G extensions. The band corresponding to the 23-mer is missing. For an explanation of the reduced hydrolysis of the last phosphodiester bond at the 3Ј-extremity (see Ref. 10). The 23-mer is produced in enzymatic degradations (see Fig. 6). B, RNA-chain extension by 3Ј,5Ј-cGMP as a function of concentration. The reaction was performed as above in the presence of the indicated concentration of 3Ј,5Ј-cGMP. C, 5Ј-labeled 5ЈA 12 C 12 3Ј reacted with 3Ј,5Ј-cGMP. tion), the observed chain-extension necessarily occurred by ligation through the 5Ј-phosphate group carried by the neopolymerized oligonucleotide G, as shown in Fig. 5.

The Rate-limiting Step
Is the rate-limiting step of the polymerization reaction the dinucleotide formation step or the extension reaction step? In the 3Ј,5Ј-cGMP system, to answer this question we tried to measure the kinetic constant of the dinucleotide formation by lowering the concentration of 3Ј,5Ј-cGMP down to the detection limit of the assay. The results, reported in Fig. 6, show that the shortest observed measurable chain is the G8 oligomer (N avg 8.75) and that its formation is immediate. The experiment also shows that the amount of elongated polymer formed depends on the concentration of 3Ј,5Ј-cGMP, not on a kinetically limiting step. Given that the kinetic constant of the elongation reaction, as determined in the same optimal conditions (85°C, 1 mM 3Ј,5Ј-cGMP), is relatively low (0.4 ϫ h Ϫ1 ) the limiting step is chain elongation.
As for the 3Ј,5Ј-cAMP system, the kinetic constant for chain elongation is 2 ϫ 10 Ϫ1 h. The kinetics of formation of the dimer was followed by high performance liquid chromatography analysis of the polymerized products. At no time point was the dimer observed to accumulate relative to the trimer, tetramer, etc. showing that in this system the limiting step is the dimer formation.

TABLE 1 Quantitative analysis of chain extension and terminal ligation test-systems results
Cyclic pyrimidine nucleosides 3Ј,5Ј-cCMP and 3Ј,5Ј-cUMP do not promote chain extension nor ligation (Ref. 8 and this study) (Fig. 4A and data not shown). a The chain extension rates were determined based upon densitometry measurements of autoradiograms of gel electrophoretic analysis of extension reactions (i.e., as in Fig. 4) and have been normalized with respect to the extension rate of 5ЈA 12 C 12 3Ј in the best observed conditions (6 h, 60°C, pH 6.2, 1 mM, 3Ј,5Ј-cGMP), which has been scaled to 10,000. b "Half max" indicates the concentration of cyclic nucleotide at which the rate of product yield is one-half of the maximum extension or ligation rate. c The terminal ligation rates were determined with the methodology described in footnote a, relative to the ligation rate of 5ЈA 24 3Ј, which has been scaled to 10,000 (see Ref. 8). d NA, not applicable. e Oligonucleotide C does not dimerize. In the presence of 3Ј,5Ј-cGMP it forms a multimeric form due to a more complex phenomenon (as described in the legend to Fig. 4A).

Characterization of the Bond Formed upon Ligation of the 5A 12 C 12 3 Oligonucleotide with the Neo-synthetized Oligonucleotide G
The 5ЈA 12 C 12 3Ј oligonucleotide was reacted with 3Ј,5Ј-cGMP (60°C, 6 h, 400 M 3Ј,5Ј-cGMP), then treated with T1 or SVPD I ribonucleases. Fig. 7 shows that the 5ЈA 12 C 12 G 8.75 3Ј is sensitive to the two nucleases, thus confirming the 3Ј-5Ј nature of the phosphodiester bonds formed, both in the oligonucleotide G and between the oligonucleotide G and the 5ЈA 12 C 12 3Ј oligomer.
A vast literature has accumulated on the preferential formation of the 3Ј-5Ј over the 2Ј-5Ј phosphodiester linkages or (more often and contrary to) in oligomerizations entailing nucleoside-5Ј-phosphorimidazolides and related phosphoramidates or in carbodiimide-mediated ligations (reviewed in Ref. 1). The syntheses may preferentially form one type of linkage (i.e. the 2Ј-3Ј linkage) (18) or the other (19) or both (20). For an in-depth review of this topic see (20,21). In summary of our RNases analyses: the linkage in the oligomers formed from 3Ј,5Ј-cGMP and 3Ј,5Ј-cAMP is 3Ј-5Ј. The discrepancy with the fact that the 2Ј,3Ј is the most commonly observed linkage in abiotic polymerizations from preactivated compounds may be explained simply by the fact that none of the previously reported syntheses was performed with 3Ј,5Ј-cyclic nucleotides in water, as in our case.

Increased Stability of RNA oligonucleotides in Water Is Caused by the Presence of Cyclic Monophosphate Nucleosides-
The half-life of RNA oligonucleotides in water has been a matter of detailed analyses (11,17). As expected, and based on a large body of previous studies, the observed half-life of RNA molecules depends on sequence composition, temperature, pH, and concentration. In the present analysis, the products of polymerization from 3Ј,5Ј-cGMP and 3Ј,5Ј-cAMP showed unexpectedly high t1 ⁄ 2 values. It was found that the increased life span of the oligonucleotides in water is induced by the presence of the free cyclic nucleotide, presumably due to interference with the hydrolytic degradation process by stacking interaction (Fig. 8).

DISCUSSION
How Was RNA Polymerization Started?-A key step missing in the reconstruction of the origin of living systems is an abiotically plausible synthesis of RNA. To fill this gap, for the robust synthesis and the simultaneous presence of all the necessary nucleic acid precursors (which is possible in principle (22)), an abiotic procedure for their activation and a thermodynamically sound polymerization mechanism are needed.
Using this logic we have analyzed nucleotide oligomerization in the conceivably simplest solvent and environment: water at temperatures between 40 and 90°C. Despite the limits set in principle by the standard-state Gibbs free energy change problem (3,4), we observed that the process does actually take place in water and report the nonenzymatic formation of RNA chains in water from 3Ј,5Ј-cyclic nucleotides.
We describe three mechanisms for nonenzymatic RNA generation: RNA polymerization from monomers, RNA ligation, RNA extension by polymerization on pre-existing oligomers, and ligation. RNA ligation was recently reported in a model study performed on Poly(A) oligomers (10).
We observe that 3Ј,5Ј-cGMP polymerized into RNA chains at least 25 nucleotides long (Fig. 1), the predominant oligomer being the 8-mer. At the optimal 1 mM concentration, synthesis was fast, a N avg of 11.8 being reached within 1 min, followed by slow stepwise further growth. Canonical 3Ј,5Ј-phosphodiester   NOVEMBER 27, 2009 • VOLUME 284 • NUMBER 48 bonds were formed, as determined by RNase sensitivity. 3Ј,5Ј-cAMP polymerized more slowly to oligomers that reached an N avg of 5.32 within 1 h. These oligomers expanded their size by inter-fragments ligation for a period of at least 200 h, yielding molecules Ͼ100 nucleotides long.

Nonenzymatic Polymerization of RNA
The Plausibility of 3Ј,5Ј Cyclic Nucleotides as Precursors in Nonenzymatic Polymerizations-Nonenzymatic polymerizations require preactivated monomers (3,4). The results obtained with the phosphoramidated nucleotides commonly used (23, 24 -30) show that the accumulation of polymerized forms is possible once suitable activated monomers are available. Although these studies provide useful data on the formation and properties of RNA oligomers formed by chemical synthesis, their prebiotic relevance was questioned (1,2). The action of several organic agents (31)(32)(33) and of inorganic polyphosphates (34) on polymerization in aqueous solution was reported.
An innovative nonenzymatic polymerization system was recently reported describing the lipid-assisted synthesis of RNA-like polymers from mononucleotides (35). Chemical activation of the mononucleotides was not required. Instead, synthesis of phosphodiester bonds was driven by the chemical potential of fluctuating anhydrous and hydrated conditions, with heat providing the activation energy. Chemical complexity prevented the full analysis of the RNA-like products of this otherwise promising system.
Cyclic nucleoside monophosphates were suggested as possible prebiotic compounds (36,37), the driving force for polymerization being their high reactivity and the large negative standard enthalpy of hydrolysis. The prebiotic relevance of these polymerizations was questioned, because efficient synthesis was observed with 2Ј,3Ј-but not with 3Ј,5Ј-cyclic forms.
In the possibly simplest activation system so far described, the phosphorylation of nucleosides by free phosphates or phosphate minerals in formamide was observed (5). The system affords 2Ј-AMP, 3Ј-AMP, 5Ј-AMP, 2Ј,3Ј-cAMP, and 3Ј,5Ј-AMP providing prebiotically plausible precursors to polymerization.
Nucleoside phosphorylation also occurs in water (8). Treatment of adenosine in water with 1 M KH 2 PO 4 afforded the five phosphorylated forms. A high concentration of phosphate donor is necessary and in optimized conditions (16 h, 1 M KH 2 PO 4 , 90°C, pH 6.1) the total amount of phosphorylated products reaches only the 7.3% of the input adenosine. In these conditions the half-lives of the open phosphorylated forms 2Ј-AMP, 3Ј-AMP, and 5Ј-AMP are 15, 23, and 35 h, respectively, whereas the 2Ј,3Ј-and 3Ј,5Ј-cAMP cyclic forms have half-lives of 165 and 450 h, respectively (11). Adenosine half-life in the same environments is 450 h. Thus, the formation of cyclic nucleotides also occurs in water, although not efficiently and at high temperature. Cyclic monophosphate nucleosides can be synthesized abiotically by a two-stage nucleobase assembly process on a sugar-phosphate scaffold, as shown for cytidine-2Ј,3Ј-cyclic phosphate (38).
The stability of cyclic monophosphate nucleosides and of their precursor is of concern when one attempts to retrace the route followed by initial nascent ribopolymers. A possible solution is provided by the observation that in monophosphate ribonucleotides the 3Ј-phosphate bond, the weakest bond in water, is stabilized upon polymerization (11). This property may endow the polymer with an evolutionary edge over the monomer, allowing accumulation of complex chemical information. Protective conditions, like inclusion in micelles, interaction with mineral surface (13) or inner strata (i.e. in clays (39, 40)), cycles of displacement into cooler surroundings, etc., might have played an important role in the formation and accumulation of activated precursors.
On the Mechanism of Polymerization-The observed polymerizations only occur with cyclic nucleotides and do not take place with noncyclic forms. Sizeable polymerization is observed only with 3Ј-5Ј cyclic nucleotides whereas the 2Ј,3Ј cyclic ones only afford very short chains (up to tetramers).
These facts help to focus on the possible mechanism, based on the formation of the internucleotide bonds requiring the opening of the cyclic phosphate bridge. The nonenzymatic joining of oligoadenylates on a polyuridylic acid template was reported (37). In that case 3Ј,5Ј-linked hexa-adenylic acid with a 2Ј,3Ј-cyclic phosphate terminus was shown to couple on a polyuridylic acid template in the presence of ethylenediamine, most often yielding a dodecamer.
Before that, syntheses of oligomers were obtained from 2Ј,3Ј-cAMP (36) upon polymerization on a poly(U) or from 2Ј,3Ј-cAMP evaporated from solution in the presence of catalysts such as aliphatic diamines (41). The self-polymerization afforded oligonucleotides of chain length up to at least 6. In both the reported reactions the opening of the phosphate cyclic bridge supposedly provided the necessary activation energy.
Nonenzymatic template-directed ligation of terminally preactivated oligonucleotides was reported (Refs. 19,20,23,42, and 43 and references therein). In these works the formation of the internal phosphodiester bond is attributed to the templatemediated proximity of the reactive groups. In contrast to these systems, the syntheses reported here require no special preactivation, no catalyst, and no dry chemistry, and polymerization spontaneously occurs in water.
A Role for Stacking Interactions-The observed polymerizations occur in solution. The question thus arises as to how nucleic bases interact, rapidly and not based on sequence complementarity, and pertains first to the conditions allowing stacking of nucleoside monophosphates in solution.
Stacking free energy profiles for all 16 natural ribodinucleoside monophosphates in aqueous solution were reported (44 -46). The potential of mean force calculations showed that the free energy profiles displayed the deepest minima and the highest barriers and, therefore, the highest stacking abilities, for the purine-purine dimers, especially for ApA and GpG. The free energy of stabilizing the stacked state were 2-6 kcal/mol higher for purine-purine dimers than for pyrimidine-pyrimidine dimers. Base combinations with different stacking potentials (ApA Ͼ GpG Ͼ UpU Х CpC) (45) show a corresponding order of decreasing polymerization rate (A Ͼ G Ͼ U Х C), reinforcing the explanation that the formation of oligonucleotides in solution relies on stacking for the passage from monomer to short oligonucleotides to occur.
The explanation for the formation of long sequences by terminal ligation (8) (Figs. 4 and 5) relies in the studies by Holcomb and Tinoco (47) and by Brahms et al. (48) who first described the double strand formation by ribo-Poly(A) and the relationship (48) between Poly(A) length and strand coupling. Poly(A) strands are held together by stacking (47,49), the double strands are parallel and ligate terminally (10) in the absence of enzymes, affording molecules with standard 3Ј,5Ј bonds in their entire length (10). The stacking-unstacking process is considered to be dependent on temperature (46), pH (46, 47), and fragment size (48) and, in general, favored by lower temperature and pH, and longer size. The study of the free energy profiles of stacking for all 16 natural ribonucleoside monophosphates based on potential of mean force calculations shows that many different conformations, with different degrees of stacking, are possible, revealing the gradual nature of the stacking phenomenon (45). This explains the variation in equilibrium constants and fraction stacking of ribodinucleoside monophosphates reported (50 -54) and predicts that various degrees of stacking may occur also in sub-optimal conditions, such as higher temperature.
Hence, we hypothesize that the oligomerization reactions from 3Ј,5Ј-cGMP and from 3Ј,5Ј-cAMP described in Figs. 1 and 2 rely on the stacking interaction of the purine moieties of the cyclic nucleotides, followed by the opening of the phosphodiester cyclic bond and the consequent formation of the inter-nucleotide phosphodiester bridge. This latter part of the reaction is favored by high temperature.
The ligation process involved in the formation of the long A stretches has been described (10). The sequence extension due to the terminal ligation reaction of Poly(G) on Poly(C) described in Figs. 4 and 5 need not be different from this type of ligation. Nevertheless, while the Poly(A) ligation occurred on parallel-bound double strands of A residues held by stacking, the latter occurred on antiparallel hydrogen-bonded base-paired double strands. The versatility of the set of nonenzymatic polymerization reactions leading to longer sequences (Fig. 9) is possibly the most relevant property of these self-polymerizing systems.