Optimization of Unnatural Base Pair Packing for Polymerase Recognition

Abstract

As part of an effort to expand the genetic alphabet, we have been examining the ability of predominately hydrophobic nucleobase analogs to pair in duplex DNA and during polymerase-mediated replication. We previously reported the synthesis and thermal stability of unnatural base pairs formed between nucleotides bearing simple methyl substituted phenyl ring nucleobase analogs. Several of these pairs are virtually as stable and selective as natural base pairs in the same sequence context. Here, we report the characterization of polymerase-mediated replication of the same unnatural base pairs. We find that every facet of replication, including correct and incorrect base pair synthesis, as well as continued primer extension beyond the unnatural base pair, is sensitive to the specific methyl substitution pattern of the nucleobase analog. The results demonstrate that neither hydrogen-bonding nor large aromatic surface area is required for polymerase recognition, and that interstrand interactions between small aromatic rings may be optimized for replication. Combined with our previous results, these studies suggest that appropriately derivatized phenyl nucleobase analogs represent a promising approach toward developing a third base pair and expanding the genetic alphabet.

1. Introduction

An expanded genetic alphabet, which includes a third base pair to supplement the natural base pairs formed between guanosine and cytosine and adenine and thymine, would allow for a wide range of biotechnology applications, such as site directed oligonucleotide labeling and in vitro selections with oligonucleotides bearing increased chemical diversity.¹ Additionally, a third base pair would lay the foundation for an organism with an expanded genetic code.² Efforts toward developing a third base pair have focused on nucleobase analogs designed to pair via orthogonal hydrogen-bonding (H-bonding), based on work of the Benner group,³ and more recently, on predominantly non-H-bonding analogs that pair via hydrophobic interactions, based on work of the Kool group.⁴ Pursuing the latter strategy, we⁵ and the Yokoyama and Hirao groups⁶ have shown that a wide variety of unnatural base pairs formed between identical nucleobase analogs (self pairs) or between different analogs (heteropairs) may be formed between hydrophobic nucleobase analogs that lack H-bonding potential and have little structural similarity to the natural purines or pyrimidines. In comparison to the stability of a natural base pair, the relative thermal stability of at least some of the predominantly hydrophobic unnatural base pairs was shown to result from a more favorable entropy change, implying that the classical hydrophobic effect contributes to their pairing.^5k As with proteins, hydrophobicity appears to be a suitable force to control molecular recognition within duplex DNA.

While hydrophobic interactions may be sufficient for base pair stability, it is unclear to what extent they can stabilize the transition states corresponding to polymerase-mediated DNA replication. In particular, the determinants of efficient and selective synthesis of an unnatural base pair, by insertion of the triphosphate opposite its cognate base in the template, and efficient continued primer extension of the nascent unnatural primer terminus, are largely unknown. While it is now apparent that hydrophobic base pairs may be designed that are efficiently synthesized by DNA polymerases, replication is commonly limited by the competitive insertion of dATP or dTTP opposite the unnatural base in the template and by the efficient continued primer extension after unnatural base pair synthesis. However, the potential contribution of interbase hydrophobic interactions to fidelity and extension has not yet been systematically explored.

Within duplex DNA, aromatic nucleobase analogs may interact either via face packing, where one nucleobase interstrand intercalates between its partner and a flanking nucleobase, or via edge-to-edge packing, in a manner topologically similar to a natural base pair. In general, pairs formed between large aromatic nucleobase analogs appear to pair via intercalative face-packing (D. Wemmer and F. Romesberg, unpublished results).⁷ After polymerase-mediated synthesis of the unnatural base pair, this mode of pairing at the nascent primer terminus is likely to induce structural distortions that might result in the observed poor extension rates. Thus, we hypothesized that nucleobase analogs with an aromatic surface area that is insufficient to cause interstrand intercalation, but that are optimized to interact edge-on, might form unnatural base pairs that are both efficiently synthesized and extended.

As a first step toward testing this hypothesis, we recently reported the systematic thermodynamic analysis of base pairs formed between twelve novel nucleotides bearing simple phenyl rings derivatized with methyl groups^5k (Figure 1). Surprisingly, we found that these rather simple unnatural nucleotides can form base pairs that are virtually as stable and selective as the natural pairs, despite lacking both H-bonding functionality and large aromatic surface area. We now report the complete kinetic analysis of these analogs as substrates for the exonuclease deficient Klenow fragment of E. coli DNA polymerase I (Kf). We observed that methyl group substitution has a large effect on all steps of DNA synthesis, with the different substitution patterns resulting in pairs that are synthesized and extended with rate constants varying over one to three orders of magnitude. In addition, polymerase mediated mispair synthesis by incorporation of a natural triphosphate opposite an unnatural nucleobase analog in the template reveals interesting trends that contribute to our understanding of unnatural base pair replication fidelity, as well as polymerase fidelity in general. Along with our previous studies, this data convincingly demonstrates that unnatural base pairs require neither H-bonds nor large aromatic surface for thermal stability and replication.

Figure 1.

Unnatural nucleobases used in this study.

2. Results

2.1 Efficiency of self pair synthesis

We have chosen to examine the effect of methyl group substitution in the context of self pairs. We have reported data on the synthesis of a wide variety of self pairs which serve as a reference point for the current studies.⁵ It should also be noted that self pairs do not limit the effort to expand the genetic alphabet, as a functional self pair would more than double the number of codons available for encoding proteins, and in fact, the use of self pair limits the potential mispairings with natural bases by a factor of two. The second order rate constants (i.e. k_cat/K_M), or efficiency, for Kf-mediated DM and TM self pair synthesis were previously reported as 2.8 × 10³ M⁻¹min⁻¹ and 2.2 × 10⁶ M⁻¹min⁻¹, respectively.^5b To further explore the effect of methyl group substitution, we examined the steady-state rates for BEN, DM2, DM3, DM4, and DM5 self pair synthesis (Table 1), which along with the DM and TM, provide a systematic analysis of the effect of double and triple methyl group substitution. Interestingly, the rate constants varied by over three-orders of magnitude, from below the limit of detection (<10³ M⁻¹min⁻¹) for the BEN self pair, to greater than 10⁶ M⁻¹min⁻¹ for the DM5 and TM self pairs. In general, there is no correlation between the rates and the extent of substitution. This is best illustrated by comparing the di-substituted self pairs, DM, DM2, DM3, DM4, and DM5, which are synthesized with rates that vary over three orders of magnitude. Remarkably, the DM5 and TM self pairs are synthesized only about 20-fold less efficiently than a natural base pair in the same sequence context.^5b The efficiency of DM5 and TM self pair synthesis results predominantly from a large k_cat, which is five to sixty-fold larger than that for the other self pairs, and only three to five-fold smaller than that for a natural base pair in the same sequence context.^5b The similar rates with which the DM5 and TM self pairs are synthesized suggest that substitution at the 2- and 4-positions is sufficient for efficient synthesis, while substitution at the 5-position is less important, highlighting the importance of the specific substitution pattern as opposed to the extent of substitution.

Table 1.

Rates of Unnatural Self Pair Synthesis^a

5′-d (TAATACGACTCACTATAGGGAGA)
3′-d (ATTATGCTGAGTGATATCCCTCTXGCTAGGTTACGGCAGGATCGC)
X	Triphosphate	kcat (min−1)	KM (µM)	kcat/KM (min−1M−1)
BEN	BEN	ndb	ndb	< 1.0 × 103
DM	DM	1.0 ± 0.1	359 ± 88	2.8 × 103 c
DM2	DM2	0.85 ± 0.2	47 ± 11	1.8 × 104
DM3	DM3	5.9 ± 0.6	54 ± 13	1.1 × 105
DM4	DM4	2.8 ± 0.6	25 ± 3	2.0 × 105
DM5	DM5	50 ± 4.6	25 ± 6	2.2 × 106
TM	TM	31 ± 4.6c	14 ± 3c	2.2 × 106c
dATP	dT	163 ± 7d	3.1 ± 1d	4.7 × 107d

^aSee Experimental Section for details.

^bReaction was too inefficient for k_cat and K_M to be determined independently.

^cSee Ref 5b.

^dSee Ref 5a.

A more detailed analysis of how the substituents impact base pairing during replication requires an assumption about how the nucleobase analogs are orientated with respect to the C-glycosidic linkage. In principle, the analogs may exist in either a syn- or anti-orientation, defined by the 2-position relative to the C-glycosidic linkage. However, when there is a methyl group at the 2-position, the syn-orientation is unlikely to be stable due to eclipsing interactions between the methyl group and the ribosyl oxygen lone pair of electrons. To avoid this potential structural ambiguity, it is instructive to compare the rates for DM2, DM3, and DM5 self pair synthesis. These analogs are all locked into the anti-orientation by the methyl group at the 2-position, and bear an additional substituent at one of the remaining three unique positions. Clearly, substitution at the 4-position is most favorable, followed by substitution at the 3-position, and then at the 5-position. The effects are substantial, resulting in a two-order of magnitude variation in the rate of self pair synthesis.

2.2 Efficiency of self pair extension

The step that consistently limits the synthesis of DNA containing unnatural self pairs (or heteropairs) is continued primer elongation after incorporation of the unnatural triphosphate opposite its partner in the template (i.e. extension). We thus examined the efficiency of extension of all twelve self pairs and found that methyl group substitution has a significant effect, with the second order rate constants varying over two-orders of magnitude (Table 2). The analogs can be roughly grouped into four categories. No extension could be detected with the MM2, DM, DM3, and TMB self pairs (k_cat/K_M <10³ M⁻¹min⁻¹). These nucleobase analogs all have a methyl substituent at the 3-position. For the BEN and MM1 self pairs, extension proceeded with a k_cat/K_M of ~1.5 × 10³ M⁻¹min⁻¹. The similar extension rate of the BEN and MM1 self pairs suggests that a single methyl substituent at the 2-position has little effect on extension. The MM3, DM2, and TM2 self pairs were extended with second order rate constants of approximately 5 × 10³ M⁻¹min⁻¹. Finally, the DM4, DM5, and TM self pairs were extended more efficiently than the other self pairs (k_cat/K_M of 2 to 5 × 10⁴ M⁻¹min⁻¹). DM4, DM5, and TM all have a methyl substituent at the 4-position in addition to at least one additional substituent at the 2- or 5-position. In comparison to the extension of a natural base pair, the extension of DM4, DM5, and TM is limited both by binding the next correct dNTP (K_M) and by reduced turnover of the complex (k_cat).

Table 2.

Rates of Correct Extension of Unnatural Self Pairs^a

5′-d (TAATACGACTCACTATAGGGAGAX)
3′-d (ATTATGCTGAGTGATATCCCTCTYGCTAGGTTACGGCAGGATCGC)
X	Y	kcat (min−1)	KM (µM)	kcat/KM (min−1M−1)
BEN	BEN	0.18 ± 0.02	112 ± 20	1.6 × 103
MM1	MM1	0.25 ± 0.07	173 ± 100	1.4 × 103
MM2	MM2	ndb	ndb	< 1.0 × 103
MM3	MM3	0.67 ± 0.38	110 ± 56	6.1 × 103
DM	DM	ndb	ndb	< 1.0 × 103
DM2	DM2	0.99 ± 0.08	154 ± 3	6.4 × 103
DM3	DM3	ndb	ndb	< 1.0 × 103
DM4	DM4	1.8 ± 0.02	92 ± 28	2.0 × 104
DM5	DM5	6.5 ± 1.1	161 ± 17	4.0 × 104
TM	TM	7.9 ± 1.4	152 ± 32	5.2 × 104
TM2	TM2	0.10 ± 0.002	25 ± 0.7	4.0 × 103
TMB	TMB	ndb	ndb	< 1.0 × 103

^aSee Experimental Section for details.

^bReaction was too inefficient for k_cat and K_M to be determined independently.

As with self pair synthesis, the second order rate constants for extension do not generally reflect hydrophobicity, as no extension could be detected with the most substituted analog (TMB). Analysis of the data reveals that substitution at the 4-position (compare BEN to MM3, MM1 to DM5, MM2 to DM4, DM2 to TM, and DM3 to TM2) is the most favorable, while substitution at the 3-position (compare BEN to MM2, MM1 to DM3, MM3 to DM4, and DM5 to TM2) is the most unfavorable. Substituents at the 4-position may pack favorably with one another, resulting in a self pair structure that is efficiently recognized and extended by Kf. Substituents at the 3-position may be oriented toward one another within the self pair interface and introduce a steric clash that results in the self pair adopting a geometry that is not well recognized by the polymerase. Generally these effects result predominantly from changes in k_cat, suggesting that the interactions are manifest at the interbase interface of the developing transition state. An exception is the DM4 self pair. DM4 differs from MM3 by the presence of a methyl group at the 3-position, but forms a self pair that is extended 3-fold faster than the self pair formed by MM3. This may be due to the DM4 analog adopting an orientation that positions the methyl substituent at the 3-position away from the interface with the pairing analog.

A methyl group at the 2-position had a more variable effect, depending on the specific substitution pattern of the nucleobase analog. Addition of a substituent at the 2-position of a ring that already possessed a methyl substituent at the 4- or 5-position increased the efficiency of extension (compare MM2 to DM2, MM3 to DM5 and DM4 to TM). DM4 again provided an exception. Addition of a methyl group at the 2-position of DM4, resulting in TM2, slightly decreased the efficiency of extension. Again, this may be due to the DM4 analogs adopting an orientation that positions the methyl group at the 3-position away from the interbase interface, while the 2-methyl group of TM2 forces the analogs into an anti orientation that positions the substituent at 3-position into the interbase interface, and results in a primer terminus geometry that is less well recognized by Kf. However, addition of a substituent at the 2-position of the parent phenyl ring (compare BEN to MM1) or a ring that already possessed methyl substituent at the 3-position (compare MM2 to DM3 and DM4 to TM2) resulted in no detectable change or a decrease in the extension rate.

2.3 Efficiency of polymerase-mediated mispairing of natural dNTPs with nucleobase analogs in the template

The efficiency of Kf-mediated mispair synthesis was examined by characterizing the rates at which the natural triphosphates were inserted opposite each of the twelve analogs in the template (Table 3). In general, dCTP and dGTP are the least efficiently inserted, with second-order rate constants less than 5 × 10³ M⁻¹min⁻¹. The second most efficiently inserted natural triphosphate was dTTP. Interestingly, in this case, the efficiencies are at least roughly correlated with the number of methyl groups that are expected to be presented to the incoming thymidine. BEN, MM1, MM2, MM3, DM, DM2, and DM5 can each present zero or one substituent to the incoming thymidine. Insertion of dTTP opposite each of these analogs proceeds with k_cat/K_M of 3.6 × 10³ M⁻¹min⁻¹ to 1.6 × 10⁴ M⁻¹min⁻¹. In the case of DM3, DM4, or TM, each analog may present two methyl groups to the incoming dTTP, and insertion proceeded with a k_cat/K_M of 3.0 × 10⁴ M⁻¹min⁻¹ to 9.0 × 10⁴ M⁻¹min⁻¹. Finally, the most substituted interfaces, presented by TM2 and TMB, template addition of dTTP with a k_cat/K_M of 1.5 × 10⁵ M⁻¹min⁻¹ and 4.1 × 10⁵ M⁻¹min⁻¹, respectively. This dependence of the second order rate constant on the number of methyl groups results predominantly from an increased k_cat. In fact, the most substituted analogs show k_cat values that are only four to six-fold reduced relative to that for a natural base pair in the same sequence context.^5b

Table 3.

Incorporation of Natural Triphosphates Opposite Unnatural Bases in the Template^a

5′-d (TAATACGACTCACTATAGGGAGA)
3′-d (ATTATGCTGAGTGATATCCCTCTXGCTAGGTTACGGCAGGATCGC)
X	Triphosphate	kcat (min−1)	KM (µM)	kcat/KM (min−1M−1)
BEN	A	21 ± 4.5	61 ± 11	3.4 × 05
	C	0.34 ± 0.01	328 ± 41	1.0 × 03
	G	0.09 ± 0.01	49 ± 13	1.8 × 103
	T	0.70 ± 0.10	192 ± 2	3.6 × 103
MM1	A	72.8 ± 18.1	25.4 ± 3.3	2.9 × 106
	C	0.18 ± 0.02	81 ± 69	2.2 × 103
	G	ndb	ndb	<1.0 × 103
	T	0.56 ± 0.15	135 ± 21	4.1 × 103
MM2	A	27.2 ± 12.4	56.5 ± 8.7	4.8 × 105
	C	ndb	ndb	<1.0 × 103
	G	ndb	ndb	<1.0 × 103
	T	1.3 ± 0.47	203 ± 34	6.4 × 103
MM3	A	14.2 ± 5.79	36.2 ± 2.98	3.9 × 105
	C	0.36 ± 0.04	303 ± 69	1.2 × 103
	G	ndb	ndb	<1.0 × 103
	T	1.28 ± 0.45	123 ± 14	1.0 × 104
DMc	A	1.1 ± 0.1	75 ± 15	1.5 × 104
	C	0.68 ± 0.03	307 ± 28	2.2 × 103
	G	ndb	ndb	<1.0 × 103
	T	2.9 ± 0.5	182 ± 30	1.6 × 104
DM2	A	42 ± 10	10 ± 2	4.2 × 106
	C	0.37 ± 0.17	83 ± 33	4.5 × 103
	G	ndb	ndb	<1.0 × 103
	T	1.5 ± 0.34	143 ± 6	1.0 × 104
DM3	A	9.9 ± 1.8	28 ± 5	3.5 × 105
	C	0.43 ± 0.08	297 ± 36	1.4 × 103
	G	ndb	ndb	<1.0 × 103
	T	8.4 ± 4.8	93 ± 2	9.0 × 104
DM4	A	9.1 ± 2.3	34 ± 8	2.7 × 105
	C	1.7 ± 0.36	258 ± 12	6.6 × 103
	G	0.06 ± 0.02	51 ± 7	1.2 × 103
	T	7.2 ± 1.7	113 ± 20	6.4 × 104
DM5	A	30 ± 1.8	20 ± 3.9	1.5 × 106
	C	ndb	ndb	<1.0 × 103
	G	0.22 ± 0.08	208 ± 7	1.1 × 103
	T	1.7 ± 0.39	169 ± 49	1.0 × 104
TMc	A	6.6 ± 0.2	26 ± 5	7.6 × 105
	C	0.18 ± 0.04	381 ± 35	4.7 × 102
	G	0.07 ± 0.01	140 ± 25	5.0 × 102
	T	6.9 ± 0.6	227 ± 42	3.0 × 104
TM2	A	16 ± 4.7	38 ± 14	4.2 × 105
	C	2.3 ± 0.19	300 ± 19	7.7 × 103
	G	0.13 ± 0.03	117 ± 49	1.1 × 103
	T	24 ± 0.61	156 ± 14	1.5 × 105
TMB	A	83 ± 11	29 ± 4.8	2.9 × 106
	C	ndb	ndb	<1.0 × 103d
	G	0.08 ± 0.01	54 ± 22	1.5 × 103
	T	42 ± 3.1	103 ± 20	4.1 × 105

^aSee Experimental Section for details.

^bReaction was too inefficient for k_cat and K_M to be determined independently.

^cSee Ref 5b.

^dDue to rapid mispair extension, the rate of dCTP insertion opposite TMB was measured with a template modified to contain a dA as the next 5’ base in the template.

The most efficiently inserted triphosphate opposite each unnatural base in the template is dATP. Interestingly, the insertion efficiencies generally fall into one of two groups, those in the range of 2.7 × 10⁵ M⁻¹min⁻¹ to 4.8 × 10⁵ M⁻¹min⁻¹ (BEN, MM2, MM3, DM3, DM4, TM and TM2) or 1.5 × 10⁶ M⁻¹min⁻¹ to 4.2 × 10⁶ M⁻¹min⁻¹ (MM1, DM2, DM5, and TMB). An exception is DM, opposite which dATP is inserted with a second order rate constant of 1.5 × 10⁴ M⁻¹min⁻¹. Comparing BEN to MM1, MM2, and MM3, it is apparent that the only single substitution that significantly affects dATP insertion is at the 2-position, which increases the second order rate constant by about an order of magnitude. This results from a large increase in k_cat and a small decrease in K_M.

Comparing BEN to DM indicates that a methyl group at the 3-position decreases the efficiency of dATP insertion. MM2 and DM4 are presumably oriented such that these methyl groups are directed away from the incoming nucleobase, and they likely template dATP in a manner similar to BEN. The effect of a methyl group at the 2- and 3-positions appears to approximately cancel with DM3. The similar rates at which dATP is inserted opposite MM1, DM2, and DM5 further support the suggestion that substituents at the 2- and 3-positions are generally the only ones that significantly impact dATP insertion within this scaffold. An exception is the insertion of dATP opposite TMB, where the additional two methyl groups (relative to DM3) increase the efficiency of dATP insertion by an order of magnitude. Remarkably, dATP is inserted opposite MM1, DM2, DM5, or TMB with a second order rate constant that is only approximately an order of magnitude reduced relative to the rate at which dATP is inserted opposite dT in the template. As with MM1 relative to BEN, the increased rates of dATP insertion opposite DM2, DM5, and TMB, relative to the other analogs, result mostly from an increase in k_cat and a small decrease in K_M. Remarkably, with each of these four analogs, the observed k_cat’s are within three to nine-fold of that for natural synthesis, and the K_M’s are within two- to seven-fold.^5b When these rates are compared to those for the insertion of the other natural dNTPs, it is apparent that MM1, DM2 and DM5 (and to a lesser extent, TMB), are functional mimics of dT, directing the efficient and selective insertion of dATP.

3. Discussion

Efforts to develop an orthogonal third base pair are expected to be facilitated by a general understanding of the determinants that underlie DNA duplex stability and recognition by DNA polymerases. With simple methyl-derivatized phenyl nucleoside analogs, we previously reported a systematic evaluation of the contribution of nucleobase shape and hydrophobicity to pairing stability.^5k Surprisingly, we found that despite a lack of H-bonding capacity and a significantly reduced aromatic surface area relative to a natural base pair, these small nucleobase analogs can form unnatural base pairs with high stability and selectivity. Using these same analogs, we now report the systematic evaluation of nucleobase shape and hydrophobicity to polymerase-recognition. We examined base pair synthesis in the context of self pairs, which are formed between two identical nucleobases. While this allows us to systematically address the effects of nucleobase modification, it is also of practical utility as self pairs offer the simplest route to the expansion of the genetic code, reducing the number of potential mispairs with the natural bases by a factor of two.

There are three facets to the synthesis of DNA containing an unnatural self-pair: first the pair must be efficiently synthesized by insertion of the unnatural triphosphate opposite the unnatural base in the template; second, insertion of a natural dNTP must not be competitive; and finally, the self pair must be efficiently extended. In this work we examined all of these facets in detail for each analog. We showed that methyl substitution has a pronounced effect on the rates of self pair synthesis. Overall, the specific methyl group substitution pattern appears to be more important than hydrophobicity. This is most evident from an examination of the rates at which the di-substituted analogs (DM, DM2, DM3, DM4, and DM5) form self pairs. Despite having similar surface area and hydrophobicity, the self pair synthesis rates vary over three-orders of magnitude. Substitution at the 2- and 4-positions appears to be sufficient for efficient synthesis. Substitution at the other positions has little or a deleterious effect. Synthesis of the DM5 and TM self pairs is only approximately an order of magnitude less efficient than a natural base pair in the same sequence context.^5b This is remarkable considering that these pairs do not form H-bonds, either between the nucleobase analogs or to the polymerase, and that they have significantly reduced aromatic surface area relative to a natural purine-pyrimidine base pair. In addition to being most efficiently synthesized, the DM5 self pair is also the most stable of this series of analogs in duplex DNA.^5k However, the rates for self pair synthesis do not generally reflect base pair stability as measured by the duplex melting temperature (see Supporting Information). Rather, the efficiencies result predominantly from large k_cat’s. Apparently, the methyl groups at the 2- and 4-positions form a stable interface between the analogs in the rate determining transition state, methyl groups at the 3-position are less stabilizing, presumably due to increased eclipsing interactions, and substituents at the 5-position are oriented away and do not contribute to the interface (Figure 2).

Figure 2.

(a) Model of pairing with hydrophbic base pars. (b) Predicted structure of DM5 self pair.^5k

Evaluation of the rates with which polymerases insert natural dNTPs opposite an unnatural base in the template allows for an assessment of unnatural base pair synthesis fidelity. Additionally, it allows for the systematic evaluation of the interactions that mediate natural base pair synthesis. Interestingly, the methyl group substitution pattern has a very different effect for each natural dNTP. Generally, dGTP and dCTP, the most hydrophilic of the natural dNTPs,⁸ are not inserted well opposite any of the twelve unnatural bases in the template. This suggests that desolvation may contribute to mispair synthesis. dTTP is inserted with a rate that at least roughly parallels the extent of methyl group derivatization, with less dependence on the specific substitution pattern. While the K_M’s vary by approximately a factor of two, the k_cat’s vary by 75-fold, suggesting that the increased hydrophobicity stabilizes the transition state rather than a Michaelis-like complex, perhaps by improving the orientation of the scissile phosphate bond relative to the nucleophilic 3’OH at the primer terminus.

Of the four natural triphosphates, dATP is consistently the most efficiently inserted triphosphate opposite the unnatural analogs in the template. However, unlike insertion of dTTP, these rates do not appear to be dependent on hydrophobic surface area, but instead are dependent on the specific substitution pattern. Generally, addition of a methyl group at the 2-position of the unnatural nucleobase favors insertion of dATP, while substitution at the 3-position disfavors it. The efficient insertions appear to result predominantly from an elevated k_cat, and to a lesser extent from a reduced K_M. Remarkably, dATP is inserted opposite MM1, DM2, DM5, or TMB with a second order rate constant that is only approximately an order of magnitude reduced relative to the rate at which it is inserted opposite dT in the template.^5b These results suggest that overall shape mimicry of dT (i.e. shape complementarity) is less important than the presence of a single appropriately placed substituent that is capable of packing with adenine in the developing transition state, likely at the hydrophobic methine moiety. This hypothesis is supported by data from the literature. For example, the efficient incorporation of dATP opposite the dT shape mimic, F, in the template is one of the central observations supporting the importance of shape complementarity (Figure 3).^4a,4b However, 2FB templates dATP with a similar efficiency,^5j suggesting that the fluorine substituent at the 2-position is sufficient for recognition of dATP. The importance of specific electrostatic and packing interactions, as opposed to shape complementarity, is also supported by recent studies that have shown that the efficiency of polymerase-mediated replication is better correlated with electronic properties of the nucleobase than with the shape.^5c,5h–j^,9

Figure 3.

(a) dA:F base pair and (b) dA:2FB base pair.

Mispairing with dT, dC, or dG does not significantly compromise self pair synthesis fidelity. However, dATP insertion opposite the unnatural base pairs in the template proceeds with rates similar to, or only marginally reduced relative to those for the self pairs. Thus, mispairing with dA compromises DNA synthesis fidelity with these self pairs. While methyl group substitution at the 2-position increases the rates at which both the self pairs and the mispairs with dA are synthesized, substitution at the 3- and 4-position favors self pair synthesis, but disfavors (3-position) or has little effect (4-position) on the rate at which the mispair is synthesized. Thus, orthogonality against dT, dC, and dG appears to be intrinsic to these small hydrophobic scaffolds, while orthogonality against dA may be achieved by judicious placement of methyl groups.

Perhaps most importantly from the perspective of the effort to develop unnatural base pairs, methyl group substitution was found to have a significant effect on the self pair extension rates. The observed rates do not parallel either the previously reported thermal stabilities (Supporting Information) or the hydrophobicity of the nucleobase analogs, suggesting that the specific substitution pattern is important. The most efficiently extended self pairs are those formed by DM4, DM5, and TM, all of which have a methyl group at the 4-position and at least one additional methyl group. While these self pairs are extended significantly less efficiently than are the natural base pairs, they are extended significantly faster than the other analogs, and only approximately 10-fold slower than the most efficiently extended unnatural base pairs identified to date.^5j These results suggest that with continued optimization, analogs based on this nucleobase scaffold may represent attractive third base pair candidates. For example, based on similar analogs, addition of a fluorine substituent to the 3-position of DM5 or TM is predicted to further increase the extension rates, as already reported for the parent phenyl scaffold.^5j

It is interesting to speculate about how nucleobase analogs with such limited surface area, and no ability to form H-bonds, can be recognized by DNA polymerases. While dipole and dipole-induced dipole interactions are all known to contribute to stable intrastrand stacking in natural DNA¹⁰, these analogs have neither large permanent dipole moments nor significant polarizability. Thus, it seems unlikely in these cases that these electrostatic interactions contribute to the differences in self pair synthesis, fidelity, or extension. Rather, specific inter-base pair packing interactions that develop in the transition state appear to underlie recognition of the analogs by Kf polymerase. This is consistent with previous studies of the DM5 self pair, which modeling studies predicted to be well accommodated in B-form DNA with the analogs positioned in the same plane and interacting with one another in an edge-on manner.^5k

It is reasonable that hydrophobic packing may replace inter-base H-bonding, as both the donors and acceptors are removed and thus no desolvation is required and no hydrophilic moieties are buried in the hydrophobic core of the duplex. However, interactions between the nucleobases and the polymerase seem more problematic as compensatory changes have not been made in the protein. These interactions might be expected to be especially important for base pair extension, since structural and biochemical studies have identified important H-bonds between polymerase H-bond donors and nucleobase H-bond acceptors in the developing minor groove (N3 of purines and O2 of pyrimidines).¹¹ For example, the contribution of this H-bond has been investigated by replacing dG or dA at the primer terminus with 3-deazaguanine or 3-deazaadenine, respectively.¹¹ These modified primer termini were extended with a steady-state efficiency of ~10⁵ M⁻¹min⁻¹. Remarkably, the DM5 and TM self pairs are extended with similar rates. Perhaps, these rates reflect the upper limits of Kf-mediated extension of a base pair that has suitable interbase interactions, mediated by either H-bonds or re-engineered with optimized packing interactions, but that does not engage the polymerase with an H-bond at the primer terminus. This suggests that the inclusion of a suitably positioned H-bond acceptor in the unnatural nucleobase scaffold might yield further improvements in unnatural base pair extension.^4d

We have previously shown that unnatural nucleotides bearing simple methyl-substituted benzene rings as nucleobase analogs may form base pairs that are virtually as stable as natural pairs, despite possessing neither H-bonds nor large aromatic surface area. We have now shown that these analogs may also be optimized for polymerase-mediated replication.⁵ These smaller nucleobases are not expected to induce distortions in duplex DNA, including those at the primer terminus that appear to limit replication of the larger unnatural base pairs. These nucleobase analogs will likely serve as scaffolds for further modification and optimization as unnatural base pairs. We are now focused on identifying combinations of methyl group substitutions and heteroatom derivatizations that will impart these smaller unnatural nucleobases with further improvements in stability and replication.

4. Experimental Section

General Methods

Chemical reagents were purchased from Sigma-Aldrich and used without further purification, unless otherwise stated. All reagents for oligonucleotide synthesis were purchased from Glen Research. ³¹P NMR spectra were recorded on a Bruker AMX-400 spectrometer. Coupling constants (J values) are reported in Hz. The chemical shifts are given in δ (ppm) using 85 % H₃PO₄ in D₂O for ³¹P NMR as an external standard. T4 polynucleotide kinase and Klenow fragment exo⁻ were purchased from New England Biolabs. [γ-³³P]-ATP was purchased from Amersham Biosciences.

Synthesis of oligonucleotides

All unnatural nucleosides and nucleotides used in this study were synthesized as previously reported.^5k Oligonucleotides were prepared by the β-cyanoethylphosphoramidite method on controlled pore glass supports (1 mmol) using an Applied Biosystems Inc. 392 DNA/RNA synthesizer as standard method. After automated synthesis, the oligonucleotides were cleaved from the support by conc. aqueous ammonia for 1 h at room temperature, deprotected by heating at 55 °C for 12 h, and purified by denaturing polyacrylamide gel electrophoresis (12–20%, 8 M urea). The primer oligonucleotides containing unnatural bases at the 3’-end were obtained using Universal Support, or 3’-phosphate CPG, which was treated with alkaline phosphatase after deprotection. The oligonuceotides were purified by PAGE, visualized by UV shadowing and recovered by electroelution. After ethanol precipitation, the concentration of oligonucleotides was determined by UV/Vis absorption.

General Triphosphate Synthesis Procedure

Proton sponge (1.5 eq.) and unnatural nucleoside (1 eq.) were dissolved in trimethylphosphate (final concentration ~0.3 M) and cooled to 0 °C. POCl₃ (1.05 eq.) was added dropwise and the mixture was stirred at 0 °C for 2 h. Tributylamine (5 eq.) was added, followed by a solution of tributylammonium pyrophosphate (5 eq.) in DMF (final concentration ~0.15 M). After 3 min, the reaction was quenched by addition of 1 M aqueous triethylammonium bicarbonate (10 vol. eq.) The resulting crude solution was stirred for 30 min at 0 °C and then lyophilized. The crude material was purified by reverse phase HPLC (C18 column, 1–35% CH₃CN in 0.1 M NEt₃-HCO₃, pH 7.5) followed by lyophilization to afford the triphosphate as a white solid. BEN triphosphate¹³ and DM triphosphate^5b were synthesized as described previously. DM2 triphosphate: ³¹P NMR (140 MHz, D2O) δ −5.91 (d, J=18.3 Hz), −10.59 (d, J=17.1), −22.16 (t, J=17.8 Hz). DM3 triphosphate: ³¹P NMR (140 MHz, D2O) δ −5.91 (d, J=17.2 Hz), −10.45 (d, J=17.2), −21.78 (t, J=17.8 Hz). DM4 triphosphate: ³¹P NMR (140 MHz, D2O) δ −5.91 (d, J=18.1 Hz), −10.58 (d, J=17.2), −22.16 (t, J=17.7 Hz). DM5 triphosphate: ³¹P NMR (140 MHz, D2O) δ −5.93 (d, J=18.5 Hz), −10.57 (d, J=17.1), −22.18 (t, J=17.8 Hz).

Gel-Based Kinetic Assay

Primers and templates were chosen according to which facet of replication was being examined, as shown in Table 1 and Table 2. Primer oligonucleotides were 5′-radiolabeled with T4 polynucleotide kinase and [γ-³³P]-ATP. Primers were annealed to template oligonucleotides in the reaction buffer by heating to 90 °C followed by slow cooling to ambient temperature. Assay conditions included 40 nM primer/template, 0.1–1.3 nM enzyme, 50 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 1 mM DTT, and 50 µg/mL acetylated BSA. The reactions were carried out by combining the DNA-enzyme mixture with an equal volume (5 µL) of 2x dNTP stock solution, incubating at 25 °C for 1–10 min, and quenching by the addition of 20 µL of loading dye (95% formamide, 20 mM EDTA, and sufficient amounts of bromophenol blue and xylene cyanole). The reaction mixtures were resolved by 15% polyacrylamide and 8 M urea denaturing gel electrophoresis, and the radioactivity was quantified by means of a PhosphorImager (Molecular Dynamics) and ImageQuant software. A plot of k_obs versus triphosphate concentration was fit to a Michaelis-Menten equation using the program Kaleidagraph (Synergy Software). The data presented are averages of three independent determinations.

Supplementary Material

1si20051106_11. Supporting Information.

Representative kinetic data and plot of kinetic data versus thermodynamic data. This material is available free of charge via the Internet at http://pubs.acs.org.