Guanine O⁶ and Thymine O⁴ Phosphitylation in Oligonucleotide Synthesis

Abstract

The guanine O⁶ phosphitylation side reaction, which can lead to the formation of the intermediates – abasic sites and O⁶ phosphate derivatives – responsible for internucleotide bond cleavage, and G-to-2,6-diaminopurine (D), or G-to-A after PCR amplification, substitution error, respectively, represents a significant barrier to the synthesis of long oligonucleotides (ONs) and limits the use of phosphoramidite capping in the synthesis of sensitive ONs. To confirm this side reaction and determine the degree to which it occurs, we synthesized the sequence 5′-T₆GT₁₈-3′, and subjected its 5′-capped version to repeated coupling and oxidation reaction cycles. Subsequent deprotection and cleavage with neat nBuNH₂ and LCMS analysis of the resulting products confirmed that both intermediates can form simultaneously, with quantities increasing with the increase of the number of the reaction cycles. In addition, after a sufficient number of the cycles, a substantial amount of thymine O⁴ phosphate derivative, which can cause T-to-C substitution errors and has not been reported in the literature, was also observed.

Graphical Abstract

LCMS analysis provides direct evidence of side-reaction intermediates responsible for internucleotide bond cleavage and troublesome sequence errors.

Introduction

Guanine O⁶ phosphitylation is one of the most common side reactions in oligonucleotide (ON) synthesis.^1–2 This side reaction has two distinct consequences for the ON product. One consequence is internucleotide bond cleavage. As a result, ON sequences rich in guanine are usually more challenging to synthesize and often give low yields. Gel electrophoresis usually reveals many ON molecules shorter than the target, and in many cases these shorter ONs appear as a smear. The pathway leading to this consequence begins with the O⁶ phosphitylated phosphite triester formed during the coupling step. In the oxidation step, it has been proposed that the iodinated tetra valent intermediate can coordinate with N7. Attack by water at O⁶ via the S_NAr mechanism followed by the loss of iodide gives an N7 phosphorylated intermediate. This intermediate is prone to depurination under acidic deblocking conditions. The resulting abasic site then undergoes internucleotide bond cleavage during deprotection and cleavage under basic conditions.^1–4

The other consequence of guanine O⁶ phosphitylation is G-to-A substitution, which is the most common error in chemically synthesized ONs.⁵ The pathway leading to this consequence also begins with the O⁶ phosphitylated phosphite triester.⁶ In this case, during the oxidation step, the phosphate group does not migrate to N7. Instead, a phosphate triester is formed at the O⁶ position. The phosphate may or may not be displaced by DMAP or NMI during capping via an S_NAr mechanism. After ON synthesis, deprotection with ammonium hydroxide displaces the phosphate, DMAP or NMI via another S_NAr reaction, giving 2,6-diaminopurine (D), which is recognized as adenine in subsequent PCR amplification or sequencing steps involving polymerases, and thus resulting in the G-to-A substitution error.⁵

Recently our research group has undertaken two projects. One is the synthesis of ultralong ONs such as the 800-mer GFP gene and the 1,728-mer Φ29 DNA polymerase gene.^7–8 For the synthesis of such long ONs, it is inevitable that many guanine residues are present in the sequence. Even if the occurrence of guanine O⁶ phosphitylation side reaction is low, the two consequences – internucleotide bond cleavage and G-to-A substitution errors – become serious limiting factors for long ON synthesis.

The other project is the synthesis of sensitive ONs, including oligodeoxynucleotides (ODNs) and oligoribonucleotides (ORNs).^9–13 For this purpose, the amino groups of adenine, cytosine and guanine have been protected with groups that can be removed under non-nucleophilic, nearly neutral conditions to avoid destruction of the sensitive functionalities such as N⁴-acetylcytidine (ac4C), esters, and α-haloamides. The amino protecting groups we developed include the 1,3-dithian-2-yl-methoxycarbonyl (Dmoc) and 4-((t-butyldimethylsilyl)oxy)-2-methoxybutanoyl (SoM) groups, which can be deprotected with NaIO₄/K₂CO₃ and Et₃N/HF, respectively. A significant challenge for sensitive ON synthesis is cap exchange, in which a small portion of the Dmoc or SoM group is replaced by an acetyl group in the capping step. Because the acetyl group cannot be removed under non-nucleophilic conditions, even exchange of a single Dmoc or SoM group within an ON molecule renders the entire molecule useless. For this reason, in sensitive ON synthesis using Dmoc or SoM as amino protecting groups, capping of failure sequences cannot be performed with Ac₂O. Instead, we used phosphoramidite such as NC(CH₂)₂OP[N(iPr)₂]₂ for capping, which is a known practice in the literature although usually for purposes other than preventing cap exchange for sensitive ON synthesis.^12,14–18 While this approach avoids cap exchange, the repeated exposure of the growing sequence to phosphitylation conditions makes the O⁶ guanine phosphitylation side reaction a serious concern.

For reasons associated with these two projects, we decided to revisit the O⁶ guanine phosphitylation side reaction in ON synthesis.^2,4,6 LCMS analysis of the ON products from the 5′-capped 5′-T₆GT₁₈-3′ sequence after repeated post-synthesis coupling-oxidation (PSCO) cycles indicates that this side reaction can indeed occur under the tested conditions, and both consequences – internucleotide bond cleavage and G-to-D substitution errors – can take place. In addition, we further found that O⁴ thymine phosphitylation,² which can cause T-to-C substitution errors and has not reported in the literature, can also occur. Fortunately, under the conditions used for our long ON synthesis and sensitive ON synthesis projects, these side reactions were not as severe as those observed under the conditions used in this study and therefore did not significantly limit the projects.

Results and Discussion

For ease of characterization of the ON products resulting from the O⁶ guanine phosphitylation side reaction, an ODN sequence containing only one G, 5′-T₆GT₁₈-3′ (1a, Scheme 1), was designed. This sequence was synthesized on a Mermade 6 DNA/RNA synthesizer on CPG under typical synthesizer manufacturer recommended conditions, which are provided in the Supporting Information. After synthesis, the 5′-trityl group was removed. The CPG was then divided into portions containing a theoretical 0.5 μmol ODN on the support. These portions were subsequently subjected to 0, 50, 100, or 200 PSCO cycles (Table 1). In the coupling step, NC(CH₂)₂OP[N(iPr)₂]₂ (2) was used as the phosphoramidite, and 4,5-dicyanoimidazole (DCI) was used as the activator. The molar ratio between 2 and DCI was maintained at 1:1 instead of the typical 1:2.5 used in the coupling step of ODN synthesis, for the reason discussed later. The oxidation step followed standard conditions. After completion of the PSCO cycles, the CPG was subjected to deprotection and cleavage using neat nBuNH₂ at 55 °C for 12 h.

Scheme 1.

Formation of the ODN side products via the O⁶ guanine phosphitylation reaction.

Table 1.

The post-synthesis coupling-oxidation (PSCO) cycle.

Step	Step name	Conditions
1	Wash	ACN, 7 times
2	Coupling	NC(CH2)2OP[N(iPr)2]2 (2, 0.1 M in ACN), DCI (0.25 M in ACN), adjust delivery volumes to achieve 1:1 molar ratio of DCI to 2, 100 sec × 3
3	Wash	ACN, 3 times
4	Oxidation	I2 (0.02 M in THF/pyridine/H2O), 40 sec × 2
5	Wash	ACN, 5 times (additional washes after the last cycle only)

For the portion that was subjected to zero PSCO cycles, after deprotection and cleavage, the ODN product should be 5′-HO-T₆GT₁₈-OH-3′ (1a). LCMS analysis gave a peak for this molecule at 15.8 minutes (Figure 1A). During the coupling steps for installation of the six T residues after the G was introduced in the ODN synthesis, there was always a slight chance of phosphitylation at the O⁶ position of G. As described earlier, there are two possible outcomes of this side reaction. In one scenario, the phosphite triester migrates to N7 during the oxidation step, resulting in the formation of 3a in the subsequent detritylation step. In the deprotection and cleavage step, 3a is converted into 4 and 5a (Scheme 1).² In the other scenario, migration of the phosphite triester does not occur, and 6a is formed instead. During deprotection and cleavage with nBuNH₂, an S_NAr reaction gives 7a.⁶ Therefore, we carefully analyzed the LCMS data and attempted to identify peaks corresponding to 4, 5a, and 7a. Only a very small peak corresponding to 4 was observed at 14.5 minutes (Figure 1A). The other species, 5a, which must have formed along with 4, was not detected, probably due to its reactive aldehyde group or its lower mass. A peak corresponding to 7a was not observed either. These results indicate that the O⁶ guanine phosphitylation side reaction can occur under normal ODN synthesis conditions, but its probability is very low.

Figure 1.

LC profiles of crude ODNs 1a and 1b. The ODNs was subjected to 0 (A), 50 (B), 100 (C) and 200 (D) post-synthesis coupling-oxidation (PSCO) cycles. LCMS conditions are in the experimental section. * The mass data such as m/z 5494/5495 were calculated and observed molecular masses, respectively. ** The mass data for 9a-h can be found in Scheme 2 and the Supporting Information. The identity of the peak at ~20 min in B and C could not be assigned.

For the portions that were subjected to 50, 100, or 200 post-synthesis coupling–oxidation (PSCO) cycles, the normal ODN product should be 1b. With repeated exposure of the guanine in the ODN sequence to phosphoramidite coupling conditions, the side products resulting from O⁶ guanine phosphitylation were expected to be much more and thus easier to detect. Moreover, in the coupling step, a 1:1 ratio of DCI to 2, which is lower than the typically used 2.5:1 ratio of DCI to nucleoside phosphoramidite, was employed. Because excess DCI could reverse the O⁶ guanine phosphitylation reaction, the lower ratio of DCI to 2 was expected to further increase the amount of side products. Specifically, at the end of the ODN synthesis and the repeated PSCO cycles, species 3b and 6b were expected to form (Scheme 1). After deprotection and cleavage with neat nBuNH₂, species 3b was expected to give 4 and 5b, while species 6b was expected to give 7b. The use of nBuNH₂ instead of the typically used concentrated NH₄OH or MeNH₂ was intended to facilitate distinguishing the side products from the normal product by MS. With nBuNH₂, the side product is 7b, which has a molecular weight of 7786. If NH₄OH were used, the corresponding 2,6-diaminopurine (DAP) side product would form, which has a molecular weight of 7730, whereas the normal ODN product 1b has a molecular weight of 7731. Evidently, if NH₄OH were used for cleavage and deprotection, the resulting DAP product would be difficult to be distinguished from 1b by MS.

The crude products after the PSCO cycles as well as deprotection and cleavage were subjected to LCMS analysis (Figures 1B–D). In all cases, side product 4, resulting from the internucleotide bond cleavage pathway of the O⁶ guanine phosphitylation reaction, was detected as expected. The ratio of this product, which appeared at 14.4 or 14.3 minutes in Figures 1B–D, to the normal ODN product 1b, which appeared at 18.4 minutes, increased as the number of PSCO cycles increased from 0 to 50, 100, and 200 (Figure 2). In addition to 4, we also detected 7b, which was resulted from the 6b S_NAr pathway (Scheme 1) and appeared at 20.5 or 20.4 minutes (Figures 1B–D). Again, the ratio of 7b to 1b increased as the number of PSCO cycles increased (Figure 2). These results confirm that both pathways of O⁶ guanine phosphitylation, one leading to internucleotide bond cleavage² and the other to G-to-D, or G-to-A after PCR, substitution error,⁶ as proposed in the literature, can operate simultaneously.

Figure 2.

Ratios of peak areas for ODN 4 to 1b (top) and 7b to 1b (bottom) versus the number of post-synthesis coupling-oxidation (PSCO) cycles performed on ODN 1b.

In addition to the O⁶ guanine phosphitylation side products, we also observed a large peak at ~22.5 minutes in the LC profile of the sample subjected to 200 PSCO cycles (Figure 1D). MS analysis of this peak revealed eight species with mass differences of ~55 Da, corresponding to the replacement of an -OH group with -NHnBu. Because the sequence contains only one guanine, the additional replacements must have occurred on some of the 24 thymines in the ODN sequence. The mechanism is the same as that for guanine, except that the pathway leading to an abasic site and subsequent internucleotide bond cleavage has essentially zero probability. Thus, O⁴ thymine phosphitylation followed by oxidation gives 8 (Scheme 2). It is also possible for O² thymine phosphitylation to occur, but according to the literature,¹⁹ this is less likely. The O⁴ thymine phosphitylation reaction can occur at any of the 24 thymine residues in the sequence. In addition, it may also occur in sequences in which guanine is not phosphitylated; however, the probability of this is lower because guanine phosphitylation is expected to be more likely according to the literature,^2,4,6 and O⁴ thymine phosphitylation has not been reported in the literature. Upon deprotection and cleavage with neat nBuNH₂, substitution of the phosphoramidate groups by nBuNH₂ via S_NAr produces 9a-h, depending on the number of thymines modified (Scheme 2).

Scheme 2.

Formation of the ODN side products via the O⁴ thymine phosphitylation reaction.

After observing 9a-h in the results obtained with 200 PSCO cycles (Figure 1D), we revisited the LCMS data from the experiments involving 50 and 100 PSCO cycles (Figures 1B–C). Peaks corresponding to the same species were also identified, although their intensities were much lower. We also found that the peak at 21.2 minutes in Figures 1B–D has a mass corresponding to the substitution of three -OH groups by -NHnBu (9b). Interestingly, species containing two -NhnBu groups (9a) were observed in peaks appearing after 22 minutes, indicating that the retention times of species 9a-h under the current RP HPLC conditions are not determined solely by the number of hydrophobic butyl groups. Instead, both the number of butyl groups and their positions within the ODN sequence contribute to the retention behavior.

A particularly interesting observation is the large difference in the intensity of the peaks appearing after 22 minutes, which correspond to 9a-h, among Figures 1B–D. With 50 PSCO cycles, the peaks after 22 minutes are minimal. With 100 PSCO cycles, their intensity does not change significantly. However, with 200 PSCO cycles, the peaks become much more prominent. Based on this observation, it is evident that the degree of ON modification leading to products 9a-h is not proportional to the number of PSCO cycles. One possible explanation is that, after some guanine and thymine residues are modified as shown in 8, the ON on the solid support becomes more hydrophobic, which may affect reagent transfer and washing and may even cause separation of components within a reagent. For example, as the ON molecules become more hydrophobic, water and iodine in the oxidation reagent may not reach the phosphite triester reaction sites of the O⁴ thymine phosphitylation product simultaneously. If water reaches the reaction sites later than iodine, the reversal of the phosphitylation reaction via hydrolysis would be less likely, resulting in a higher probability of forming 9a–h.

According to the above data, both projects mentioned in the introduction section of this paper would not be easy to accomplish. For the ultralong ON synthesis project, the synthesis of the 1,728-nucleotide Φ29 DNA polymerase gene required repetition of the phosphoramidite coupling step more than one thousand times. If O⁶ guanine and O⁴ thymine phosphitylation had occurred at frequencies close to those observed in the present study, the synthesis of such ultralong ODNs would have been impossible. The reasons that these side reactions did not occur, or occurred at much lower frequencies, may include the following. During the coupling steps for ultralong ON synthesis, a standard activator-to-phosphoramidite ratio was used, which is higher than the 1:1 ratio used in the present study. The excess activator may help convert the phosphite triesters (10a-b) back to guanine or thymine (Scheme 3). In addition, during ultralong ON synthesis, a capping step involving Ac₂O and NMI was performed after coupling and before oxidation. Both NMI and acetate nucleophiles can also help reverse the O⁶ guanine and O⁴ thymine phosphitylation reactions.⁴

Scheme 3.

The mechanism of the reversal of the O⁶ guanine and O⁴ thymine phosphitylation reactions by activators such as DCI.

Another important factor that enabled our success in ultralong ON synthesis was conducting the synthesis on smooth surfaces of materials such as glass beads and glass wool instead of within the pores of materials such as CPG.⁸ By performing synthesis on smooth surfaces, reagent transfer and washing are more efficient, which may help reverse the O⁶ guanine or O⁴ thymine phosphitylation reactions via the two mechanisms mentioned above. As described earlier, conducting synthesis on the smooth surfaces of glass beads and glass wool, as opposed to within the pores of materials such as CPG, may reduce chromatography-like intermolecular interactions between the reagents and nascent oligonucleotides, thereby reducing the likelihood of separation of components in reagents (e.g., oxidation solution) during delivery and minimizing side reactions.

For the sensitive ON synthesis project, the O⁶ guanine and O⁴ thymine phosphitylation side reactions also did not affect the outcome as much as expected based on the data presented in this article.^9–13 This may be due to the same reasons discussed for ultralong ON synthesis, with the following exceptions. In this case, capping was accomplished with phosphoramidites, which implies a greater chance for O⁶ guanine and O⁴ thymine phosphitylation, because these side reactions may occur during both the coupling and capping steps. In addition, capping with Ac₂O was not performed, meaning there was no opportunity for the reversal of the side reactions induced by acetate or NMI. Both factors should have made the sensitive ON synthesis project more difficult. However, we did not encounter significant challenges. The reasons may be twofold. One is that the higher ratio of activator to phosphoramidite may have played a significant role in reversing the phosphitylation side reactions. The other is that the ONs synthesized were relatively short, with the longest being a 49-mer.¹¹

Conclusions

In summary, by using the special sequence T₆GT₁₈ containing only one G, conducting capping with a phosphoramidite for multiple cycles after ON synthesis, and performing deprotection and cleavage with neat nBuNH₂ coupled with LCMS analysis, we were able to confirm the O⁶ guanine phosphitylation side reaction in actual ON synthesis. We further found that both pathways, which lead to internucleotide bond cleavage, and G-to-D, or G-to-A after PCR, substitution errors in ON synthesis, can operate simultaneously. In the literature, O⁴ thymine phosphitylation has been regarded as less likely,^2,4,6 and has not been documented; however, our data show that it can also occur. According to these data, our ultralong ON synthesis and sensitive ON synthesis projects, both of which involve more phosphoramidite coupling steps than conventional ON synthesis, would be expected to be challenging. Rationales for their success are provided. The knowledge obtained from this study is expected to provide guidance for developing new methods or optimizing existing methods for nucleic acid manufacturing.

Experimental

ODN synthesis and post-synthesis coupling-oxidation cycles.

The ODN 5′-HO-T₆GT₁₈-OH-3′ (1a) was synthesized at 1 μmol scale on CPG (500 Å) on an Mermade synthesizer. Deblocking: DCA (2.5% in DCM), 90 sec × 3. Coupling: iBu-dG, or T phosphoramidite (0.1 M in ACN), DCI (0.25 M in ACN), 100 sec × 3. Capping: NC(CH₂)₂OP[N(iPr)₂]₂ (2, 0.1 M in ACN), DCI (0.25 M in ACN), 100 sec × 3. Oxidation: I₂ (0.02 M in THF/pyridine/H₂O), 40 sec × 2. At the end of synthesis, the 5′ DMTr group was removed. Two 1 μmol syntheses were carried out. The ODN was divided into four portions with 0.5 μmol each. They were subjected to 0, 50, 100 and 200 post-synthesis coupling-oxidation (PSCO) cycles. Each cycle contained four steps. (1) Wash: ACN, × 7. (2) Coupling (i.e. capping with phosphoramidite): 2 (0.1 M in ACN), DCI (0.25 M in ACN), adjust delivery volumes to achieve 1:1 molar ratio of 2 to DCI, 100 sec × 3. (3) Wash: ACN, × 3. (4) Oxidation: I₂ (0.02 M in THF/pyridine/H₂O), 40 sec × 2. After the last PSCO cycle, the CPG was washed with ACN for 5 times.

Deprotection and cleavage.

To the CPG (0.1 μmol ODN) in a 1.5 mL centrifuge tube was added nBuNH₂ (750 μL). The tube was sealed and heated at 55 °C for 12 h. Volatiles were removed using a vacuum centrifuge evaporator. The residue was dissolved in nuclease-free H₂O (25 μL), vortexed, and centrifuged. The supernatant was used for LCMS analysis.

LCMS analysis.

LCMS analysis was performed on an Agilent system consisting of 1260 Infinity II Flexible Pump G7104C, Multisampler G7167A, and MCT G7116A; 1290 Infinity II DAD FS G7117A; InfinityLab LC/MSD SQ G6125C; and OpenLab CDS with MS Spectral Deconvolution. LC conditions: C-18 column (Agilent part# 659750–702, 2.1 × 50 mm, 2.7 μm); buffer A, 200 mM HFIP with 8.1 mM TEA in water; buffer B, methanol; gradient, 0–1 min 5% B, 1–21 min 5–22% B, 21–23 min 22–90% B, and 23–25 min 90% B; flow rate, 0.6 mL/min; column temperature, 65.0 °C; injection volume, 2 μL; detection, 260 nm with 360 nm reference and MSD. MSD conditions: ESI, negative mode, 300 °C gas temperature, 8.0 L/min gas flow, 35.0 psi nebulizer pressure, 3500 V capillary voltage, and m/z 750–2000 data range. MS deconvolution parameters: maximum charge, 40; minimum peaks in set, 3; MW agreement (0.01%), 5; absolute noise threshold, 500; relative abundance threshold (%), 50; MW algorithm, curve fit; MW algorithm threshold (%), 40; envelope threshold (%), 30; low molecular weight, 1500; high molecular weight, 15000; positive adduct, +H; negative adduct, −H.

Supplementary Material

^† Electronic supplementary information (ESI) available: LCMS data of ODNs

Data availability

The data supporting this article have been included as part of the Supplementary Information.