Templated synthesis of peptide nucleic acids via sequence-selective base-filling reactions.

The templated synthesis of nucleic acids has previously been achieved through the backbone ligation of preformed nucleotide monomers or oligomers. In contrast, here we demonstrate templated nucleic acid synthesis using a base-filling approach in which individual bases are added to abasic sites of a peptide nucleic acid (PNA). Because nucleobase substrates in this approach are not self-reactive, a base-filling approach may reduce the formation of nontemplated reaction products. Using either reductive amination or amine acylation chemistries, we observed efficient and selective addition of each of the four nucleobases to an abasic site in the middle of the PNA strand. We also describe the addition of single nucleobases to the end of a PNA strand through base filling, as well as the tandem addition of two bases to the middle of the PNA strand. These findings represent an experimental foundation for nonenzymatic information transfer through base filling.

Template-directed nucleic acid synthesis is an essential component of living systems (and their early precursors) that allows genetic information to be passed down to subsequent generations. The desire to understand early nucleic acid replication systems has inspired the search for nonenzymatic routes to templated nucleic acid polymerization in the laboratory.(1) Previous examples of templated nucleic acid synthesis have coupled nucleotide monomers or short nucleotide oligomers by joining backbone functional groups in reactions including, or analogous to, phosphodiester formation (Figure 1a). While these reactions can yield the desired products, they often produce additional nontemplated products and can exhibit modest sequence specificity.(1b)

Figure 1

Two strategies for the templated polymerization of nucleic acids: (a) backbone ligation and (b) base filling.

We envisioned base filling, the addition of individual bases to an abasic backbone (Figure 1b), as an alternative route to the nonenzymatic templated synthesis of nucleic acids.[2,3] Compared with the backbone ligation approach, a base-filling approach offers at least two advantages: (i) the nucleobase monomers do not contain complementary functional groups and therefore are not self-reactive, potentially reducing nontemplated reactions; and (ii) the reaction site is as close as possible to the site of base pairing, potentially increasing sequence specificity. Here we describe the templated synthesis of both native and modified peptide nucleic acids (PNAs) through base filling.

Early hypotheses on the mechanism of prebiotic nucleic acid replication considered the addition of bases to an abasic backbone to be unlikely given the unfavorable equilibrium of N-glycosidic bond formation in water,[4,5] although Bartel and co-workers more recently demonstrated the ability of a ribozyme to catalyze the addition of uracil to activated ribose pyrophosphate.(6) Over the past few decades, however, researchers have introduced several nucleic acid analogues, including some for which prebiotic syntheses have been proposed, that may support base filling through reactions more favorable than N-glycosidic bond formation. One of the most well studied of these nucleic acid variants is PNA.(7) We considered PNA to be a promising candidate for replication by base filling since the abasic reaction site is a versatile secondary amine. Base filling with nucleobase acetic acids by amine acylation would provide native PNA, and base filling with nucleobase acetaldehydes by reductive amination would generate deoxyPNA (dPNA) (Figure 2).

Figure 2

Base-filling chemistries. (a) Base filling of DNA by N-glycosidic bond formation. (b) Base filling of PNA by reductive amination to generate dPNA. (c) Base filling of PNA by amine acylation to generate native PNA.

The incorporation of a single dPNA monomer in the middle of a PNA−DNA duplex has been found to dramatically lower Tm values.(8) When the modification is at the end of the strand, however, there is little effect on Tm,(9) likely due to the greater structural flexibility at the end of the duplex. In light of these observations, we designed and synthesized PNA templates and substrates 1−15 that collectively represent the base filling of each of the four nucleobases at a site in the middle of a PNA duplex, as well as at a site at the end of the duplex (Table 1). Dimethyl lysine was added at each terminus to improve water solubility and to prevent aggregation. Template and substrate strands were capped at their N-termini with a hexanoyl (Hex) or acetyl (Ac) group, respectively, to make the products and templates distinguishable by mass spectrometry.

Table 1

Effect of PNA Modifications on Melting Temperaturea

Conditions: 3 mM each PNA, 10 mM potassium phosphate buffer, pH 6.2.

The Lys(Me2) and capping groups are omitted for clarity.

With these PNA strands in hand, we conducted a series of Tm experiments and observed that the dPNA monomer lowered melting temperature significantly when present in the middle of a PNA−PNA duplex (ΔTm = −9.8 °C), but to a much lesser extent at the end of the duplex (ΔTm = −0.9 °C), consistent with the previous studies on DNA−PNA duplexes.[8,9] We found that incorporation of an abasic site had a parallel, though more dramatic effect on Tm (ΔTm = −15.7 °C when present in the middle of the duplex, and ΔTm = −3.8 °C when present at the end of the duplex) (Table 1).

Next we tested the ability of PNA bases to fill a single abasic site in the middle of a PNA strand, where the incoming base benefits from both Watson−Crick base pairing and base-stacking interactions. PNA strands 3, 4, 5, or 6 were combined with their complementary template (1 or 2) and either an equimolar mixture of all four aldehyde monomers 16a−d(10) under reductive amination conditions, or an equimolar mixture of all four carboxylic acid monomers 17a−d under amide bond-forming conditions (Figure 3), and the reactions were monitored by MALDI mass spectrometry. Representative MALDI spectra for the selective addition of guanine to 3 by either reductive amination or acylation are shown in Figure 4. All reactions were performed in duplicate, and five MALDI spectra obtained from each for a total of 10 measurements per reaction. Product yields were calculated from the average ratio of observed product and template ion counts using a standard curve generated from independently quantitated and synthesized authentic product (see Supporting Information).(11)

Figure 3

Nucleobase substrates for base-filling reactions.

Figure 4

MALDI spectra of sequence-specific base filling of a guanine nucleobase (16a or 17a) at the abasic site in 3 via (a) reductive amination and (b) amine acylation. See Table 2 for reaction conditions.

Table 2

Yields of base-filling reactionsa

			reductive amination		acylation
entry	PNA duplexb	matchedbase	yield(%, matched)	yield(%, mismatched)	yield(%, matched)	yield(%, mismatched)
a	1N′TAC TGC TAG ACGC′3C′ATG AC_ ATC TGCN′	G	91c	<2c	78d87e	<2d2(A)e
b	1N′TAC TGC TAG ACGC′4C′ATG ACG _TC TGCN′	A	78c	6(G)c	38d73e	5(G)d7(G)e
c	2N′CGT CTA GCA GTAC′5C′GCA GA_ CGT CATN′	T	81c	11(G), 4(C)c	8d<2e	<2d19(A), 13(G), 5(C)e
d	2N′CGT CTA GCA GTAC′6C′GCA GAT _GT CATN′	C	98c	<2c	66d16e	<2d26(G), 17(A)e
e	2N′CGT CTA GCA GTAC′7C′_CA GAT CGT CATN′	G	58c	13(A), 9(C)c	39e	16(A), 5(C)e
f	1N′TAC TGC TAG ACGC′8C′_TG ACG ATC TGCN′	A	47c	29(G), 11(T), 8(C)c	42e	26(G), 7(C)e
g	2N′CGT CTA GCA GTAC′9C′GCA GAT CGT CA_N′	T	2c	4(C), 3(G), 2(A)c	4e	15(C), 7(G), 6(A)e
h	1N′TAC TGC TAG ACGC′10C′ATG ACG ATC TG_N′	C	7c	4(G), 4(A), 3(T)c	28e	10(A), 9(G), 9(T)e
i	1N′TAC TGC TAG ACGC′13C′ATG AC_ _TC TGCN′	G−A	38c	7(G−G), 3(G−C),3(C−A)c	56e	23(G−G), 5 (G−C),4(C−A), 3(A−A)e
j	14N′TGT ACG CGA TCAC′15C′ACA TG_ _CT AGTN′	C−G	21c	6(G−G), 4(A−G)c	9e	16(G−G), 12(A−G),3(T−G)e

Yields shown each represent the mean of 10 measurements from two independent reactions and are accurate to ±10%.

Lys(Me2) and capping groups are omitted for clarity.

3 μM template, 2.5 μM reactant strand, 150 μM each 16a−d, 15 mM NaBH3CN, 10 mM phosphate buffer, pH 6.2, 5 °C, 24 h.

3 μM template, 2.5 μM reactant strand, 830 μM each 17a−d, 9 mM sNHS, 13 mM EDC, 100 mM MOPS buffer, pH 7.5, 5 °C, 24 h.

3 μM template, 2.5 μM reactant strand, 830 μM each 17a−d, 13 mM DMT-MM, 100 mM MOPS buffer, pH 7.5, 2 °C, 24 h.

Addition of a single base to an abasic site at the middle of the PNA strand using reductive amination proceeds with good to excellent yield (78−98%) and good selectivity (≥7:1 ratio of the matched product to each possible mismatch product) for each of the four bases (Table 2, entries a−d). Small amounts of mismatched products are observed in the additions of adenine and thymine, but the impurities in the adenine addition can be attributed to formation of a T:G wobble pair.

For middle-of-strand base filling by amine acylation (Table 2, entries a−d), we observed a much broader range of results than for reductive amination depending on the identity of the substrate base and the acylation reagents used. Using EDC and sNHS, we observed good selectivity for all four reactions but good yield (66−78%) only when adding guanine or cytosine. Adenine addition proceeds with a significantly lower yield (38%), and the yield for thymine addition is poor (8%). Switching the acylation reagent to DMT-MM increases the yield for purine addition but lowers the yields and the sequence specificity for pyrimidine additions.

These variations in acylation efficiency and selectivity likely arise from the different steric and π-stacking characteristics of the activated ester intermediates. Our results also suggest that, for middle-of-strand base filling, G:C base pairing supports more efficient product formation than A:T base pairing. Further, despite producing canonical PNA, acylation is generally less efficient and selective than reductive amination. A similar trend has been observed for other DNA-templated reaction substrates(1f) and is likely a consequence of the reversible formation of an iminium intermediate during reductive amination.(1d)

We next investigated base filling of a single abasic site at the end of the PNA strand (Table 2, entries e−h). These reactions are more challenging, since the base-stacking interaction between the incoming base and the template should be significantly weaker than the case in the middle of a strand. Under reductive amination conditions, we observed moderate yields for the additions of guanine and adenine (47−58%) and low yields for the additions of cytosine and thymine (2−7%). Under acylation conditions, we observed no significant addition of any of the four bases using EDC/sNHS, but DMT-MM provided moderate yields (39−42%) for the additions of guanine and adenine and poor yields (4−28%) for the additions of cytosine and thymine. The selectivities of all of the reactions were significantly lower than those observed for the middle-of-strand additions. The universally lower yields for the end-of-strand reactions in Table 2 implicate base-stacking interactions and the lower disorder of double helices away from strand termini as key determinants of base-filling reaction efficiency.

To explore the ability of base filling to support the transfer of multiple consecutive bases, we attempted the addition of two neighboring bases to the middle of the PNA strand (Table 2, entries i−j). We used PNA template 1 to test the tandem addition of guanine and adenine and introduced a new template (14) to test the tandem addition of cytosine and guanine. We observed significant yields of doubly base-filled product for both reductive amination (up to 38%) and amine acylation (up to 56%), although yields were lower than the analogous single-base addition reactions. Consistent with our observations above, purines added more efficiently than pyrimidines (comparing the addition of adenine+guanine with that of cytosine+guanine), and the reductive amination reactions proceeded with higher selectivities than the acylation reactions.

In conclusion, we have achieved sequence-selective templated nucleic acid synthesis through base filling. Using either reductive amination or acylation reactions, a single base can be sequence-specifically added to an abasic site on a PNA strand in the presence of all four nucleobase monomers in the reaction mixture. We observe that middle-of-strand addition is more efficient than end-of-strand addition and that purines add more efficiently than pyrimidines, highlighting the role of base-stacking interactions in promoting base-filling reactions. Base-pairing interactions may also play a significant role, as in general guanine adds more efficiently than adenine and cytosine adds more efficiently than thymine. Additionally, we observe that reductive amination generally provides higher yields and selectivities than the amine acylation chemistries tested, likely resulting from the reversible formation of a stable iminium intermediate in the former reaction. These findings lay an experimental foundation for the development of more complex information transfer systems, including the possibility of exponential replication of nucleic acids through base-filling reactions.