Robust enzymatic production of DNA G-quadruplex, aptamer, DNAzyme, and other oligonucleotides: Applications for NMR

Abstract

Single-stranded DNA (ssDNA) oligonucleotides are widely used for biological research, therapeutics, biotechnology, and nanomachines. Large-scale enzymatic production of ssDNA oligonucleotides forming noncanonical structures has been difficult. Here, we present a simple and robust method named “palindrome-nicking-dependent amplification” (PaNDA) for enzymatic production of a large amount of ssDNA oligonucleotides. It utilizes a strand-displacing DNA polymerase and a nicking enzyme together with input DNA and deoxynucleotide triphosphates at 55°C. Scaling up of PaNDA is straightforward due to its isothermal nature. The ssDNA products can easily be isolated through anion-exchange chromatography under non-denaturing conditions. We demonstrate applications of PaNDA to ¹³C/¹⁵N-labeling of various DNA strands, including a 22-nt telomere repeat G-quadruplex, a 26-nt therapeutic aptamer, and a 33-nt DNAzyme. The ¹³C/¹⁵N-labeling by PaNDA greatly facilitates characterizations of noncanonical DNA by nuclear magnetic resonance (NMR) spectroscopy. For example, the behavior of therapeutic DNA aptamers in human serum can be investigated.

Graphical Abstract

Single-stranded DNA (ssDNA) oligonucleotides are extremely versatile chemical tools. They are essential as polymerase chain reaction (PCR) primers in biological research and clinical diagnostics.¹ DNA aptamers are used to inhibit particular proteins for therapeutic purposes.² Catalytic ssDNA oligonucleotides called DNAzymes serve as biotechnological tools that can specifically process biomolecules.³ DNA computing, DNA nanostructures, and DNA nanomachines also utilize ssDNA oligonucleotides.^4,5 In this paper, we present a simple and robust method for enzymatic production of a wide variety of ssDNA oligonucleotides in a large quantity, which facilitates characterizations of ssDNA oligonucleotides by multidimensional heteronuclear NMR spectroscopy.

Although ssDNA oligonucleotides are typically produced through chemical synthesis, enzymatic production could be more useful in some applications. One such example is isotope-labeling. Enzymatic production of ¹³C,¹⁵N-labeled ssDNA using ¹³C,¹⁵N deoxynucleotide triphosphates (dNTPs) is more convenient than chemical synthesis because ¹³C,¹⁵N phosphoramidite is far more expensive and requires a DNA synthesizer. Although there are several enzymatic methods for ssDNA,^6–8 the existing methods severely suffer from a difficulty in separation of labeled ssDNA from its complementary strand, whose concentrations are identical. This separation is notoriously difficult for relatively long ssDNA in a milligram quantity. Furthermore, noncanonical DNA structure can inhibit DNA polymerases at physiological temperature.⁹ Due to the difficulty in isotope-labeling of ssDNA, even recent NMR structures of noncanonical DNA structures (e.g., G-quadruplexes and i-motifs) longer than 20 nucleotide (nt) residues were determined primarily using unlabeled DNA (e.g., Refs.^10–13). DNA isotope labeling is far rarer than RNA isotope labeling, as indicated by the statistical data (Tables S3 and S4 in the SI) of Biological Magnetic Resonance Data Bank (BMRB).

Our method, which we refer to as palindrome-nicking-dependent amplification (PaNDA), can enzymatically produce various ssDNA oligonucleotides and make their isotope-labeling far easier than other methods (comparison is described in the SI). Figure 1A shows the PaNDA reaction scheme. It requires three macromolecular components: a strand-displacing DNA polymerase, a nicking enzyme, and input DNA. This DNA serves as both the template and the primer and contains a replication template sequence followed by a palindromic primer sequence that forms a homodimer (or possibly, a monomeric hairpin). The polymerase extends the DNA strand and produces double-stranded DNA (dsDNA). The nicking enzyme recognizes a sequence within the palindrome and generates a nick at the junction between the original DNA and the extended portion. This nick allows the strand-displacing DNA polymerase to carry out the same extension process, displacing the prior product from the template. The ssDNA product is amplified through the cycle of the nicking, strand-extension, and displacement processes. The amplified ssDNA product is less than a half in charge compared to the extended dsDNA. Therefore, the product can easily be isolated from the reaction mixture through anion-exchange chromatography under non-denaturing conditions.

Figure 1.

Enzymatic ssDNA production through PaNDA. (A) Reaction scheme. The input DNA sequences used in our current study are shown in Table S1 in Supporting Information (SI). (B) Polyacrylamide gel electrophoresis (PAGE) for PaNDA of a 15-nt ssDNA oligonucleotide. DNA was visualized by SYBR Gold (Invitrogen). (C) Resource-Q anion-exchange chromatogram for the PaNDA reaction mixture to produce the 15-nt ssDNA.

For PaNDA, we used the Bst 2.0 DNA polymerase and the Nt.BstNBI nicking enzyme (New England BioLab, Inc.), whose optimal temperatures are 65°C and 55°C, respectively.^14,15 We used input DNA containing a template sequence followed by a 24-nt palindromic sequence, CTTGGACTCAGATCTGAGTCCAAG (underlined, the Nt.BstNBI recognition sequence). The complete sequences of the input DNA strands used in our current study are shown in Table S1 in the SI. The Nt.BstNBI enzyme generates a nick between the 4^th and 5^th nucleotide residues from the 3’-end of the recognition sequence. Importantly, since the Nt.BstNBI site is within the 24-bp palindrome, any sequence with no Nt.BstNBI recognition sequence can in principle be amplified through PaNDA.

The PaNDA reactions using Bst 2.0 and Nt.BstNBI were optimal at 55°C (Fig. S1 in the SI). In PaNDA reactions, we used 0.2 mM dNTP (each), 1 μM input DNA, 160 units/ml Bst 2.0 polymerase, 200 units/ml Nt.BstNBI enzyme, 20 mM Tris•HCl (pH 8.8), 10 mM (NH₄)₂SO₄, 8 mM MgCl₂, 250 mM KCl, and 0.1% Tween20 (see Table S2 in the SI). The reaction mixture was incubated at 55°C for 90 minutes. The condition with 250 mM KCl and 8 mM MgCl₂ was effective to suppress the 3’-overhang attachment by the Bst 2.0 polymerase (which has no 3’-to-5’ exonuclease activity) without compromising the yield (Fig. S2 in the SI). Typically, 0.5–1.5 mg of purified ¹³C,¹⁵N-labeled ssDNA was obtained from a 10-ml PaNDA reaction mixture. The yield can be increased by adding multiple doses of dNTPs when most dNTPs have been consumed in the reaction mixture (see Fig. S5 in the SI). Unlike PCR, which requires microtubes placed in a thermal cycler, PaNDA can be conducted in a single large volume at the constant temperature and is easy to scale up.

We first applied PaNDA to ¹³C,¹⁵N-labeling of a 15-nt ssDNA strand, 5’-CCAAAGCCATTAGGG-3’. This corresponds to a strand of the 15-bp DNA duplex used in our prior studies.^16–18 Previously, a PCR-based method¹⁹ was used to prepare the 15-bp DNA in which both strands are ¹³C,¹⁵N-labeled. The PCR-based method is not applicable for ssDNA labeling. Using ¹³C,¹⁵N-dNTPs in PaNDA, we labeled the 15-nt ssDNA strand and produced the 15-bp DNA duplex in which only that strand is ¹³C,¹⁵N-labeled. Comparison of the HSQC spectra for the current and previous samples confirmed successful ¹³C,¹⁵N-labeling by PaNDA (Figure 2A; see also Fig. S2 in the SI).

Figure 2.

PaNDA-based ¹³C,¹⁵N-labeleing of ssDNA for 15-bp DNA and the 22-nt telomere repeat DNA G-quadruplex (Tel22). (A) Comparison of the spectra recorded for the 15-bp DNA duplex in which only one strand is ¹³C,¹⁵N-labeled by PaNDA (red) and for the same DNA in which both strands are ¹³C/¹⁵N-labeled by PCR (black). (B) ¹H-¹⁵N and ¹H-¹³C HSQC spectra recorded for Tel22 in the pH 7.4 phosphate buffers with either 100 mM NaCl or 100 mM KCl.

Next, we applied PaNDA to produce a ¹³C,¹⁵N-labeled 22-nt telomere repeat DNA G-quadruplex AGGG(TTAGGG)₃. Through NMR investigations using unlabeled DNA, other groups previously showed that this DNA adopts a single G-quadruplex structure in a buffer of 100 mM NaCl²⁰ but exhibits structural polymorphism in a buffer of 100 mM KCl.²¹ With our ¹³C,¹⁵N DNA produced through PaNDA, such differences were immediately obvious in the ¹H-¹⁵N/¹³C HSQC spectra shown in Figure 2B. The NaCl sample exhibited the expected number of HSQC signals from G-quadruplex imino groups with ¹⁵N chemical shifts (142–146 ppm) that remarkably differ from those in Watson-Crick base pairs (~147 ppm). By contrast, the KCl sample exhibited about three times as many signals, demonstrating the presence of multiple G-quadruplex states in slow exchange (Figure 2B).

We applied PaNDA to ¹³C,¹⁵N-labeling of a therapeutic DNA aptamer, NU172 (Figure 3), as well. This aptamer inhibits thrombin in human blood and was used for clinical trials in Phase II for anticoagulation.²² The crystal structures of thrombin-NU172 complexes show a G-quadruplex structure involving two layers of G tetrad.²³ Since G-quadruplex typically involves three or more tetrad layers,²¹ one may wonder if NU172 can form the G-quadruplex in the absence of the protein. Figure 3A shows the G imino region of HSQC spectra recorded for a ¹³C,¹⁵N-labeled NU172 solution in a pH 7.4 buffer with 100 mM NaCl. Eight G imino signals were observed in the ¹⁵N chemical shift range characteristic for G-quadruplex. As described in the SI (see also Fig. S4), we assigned ¹H, ¹³C, and ¹⁵N resonances of NU172 and found that these imino signals were indeed from the two-tetrad G-quadruplex region. Thus, NU172 appears to form the G-quadruplex structure even in the free state.

Figure 3.

Application of PaNDA to a 26-nt therapeutic DNA aptamer NU172. (A) Guanine imino region in the ¹H-¹⁵N HSQC spectra recorded for the ¹³C,¹⁵N-labeled DNA aptamer NU172 at 12°C and 37°C in a buffer of 10 mM potassium phosphate (pH 7.4) and 100 mM NaCl, and 5% D₂O. (B) Corresponding region of the spectra recorded for NU172 in a human serum. (C) Decay of the NU172 HSQC signals due to degradation in the serum.

We recorded ¹H-¹⁵N HSQC spectra for 200 μM ¹³C,¹⁵N NU172 dissolved in a human serum at 37°C (Figure 3B). Since typical concentrations of thrombin in sera are up to ~1 μM,²⁴ the vast majority of NU172 in this sample is not bound to thrombin. The NMR signals were broad, presumably due to macromolecular crowding with 60–80 mg/ml serum proteins,²⁵ some of which may nonspecifically interact with the DNA aptamer. Nonetheless, some signals with ¹⁵N chemical shifts characteristic to G-quadruplex were observed, suggesting that NU172 in blood retains the G-quadruplex structure in therapeutic applications. The HSQC signals from NU172 in the human serum became progressively weaker (Figure 3C). The rate for this decay was consistent with the reported half-life of NU172 in a human serum,²⁶ suggesting that the decrease in NMR signal intensities is due to the nuclease activity in the serum. Thus, PaNDA facilitates characterization of therapeutic DNA aptamers in human fluids.

We also applied PaNDA to the 33-nt DNAzyme Dz5C (Figure 4),²⁷ which catalyzes a site-specific cleavage of ssRNA in the presence of Mg²⁺ ions. We confirmed the catalytic activity of the Dz5C oligonucleotide produced by PaNDA (Figure 4A). We recorded ¹H-¹³C HSQC spectra for ¹³C,¹⁵N-labeled Dz5C in the free state and for its complex with the substrate RNA (unlabeled) in the absence of Mg²⁺ ions. As shown in Figure 4B, the two spectra were remarkably different, reflecting the structural differences between the free state and the substrate-bound state. Isotope-labeling by PaNDA can readily facilitate characterizations of DNAzymes by ¹³C,¹⁵N heteronuclear NMR experiments.

Figure 4.

Application of PaNDA to Dz5C, a 33-nt DNAzyme. (A) Confirmation of RNA cleavage by PaNDA-produced Dz5C in the presence of Mg²⁺. Lane 1 is for a substrate 19-nt RNA; Lane 2, chemically synthesized Dz5C; Lane 3, Dz5C produced through PaNDA; Lane, 4, the mixture of Dz5C and the RNA substrate in the absence of any divalent ions; Lane 5, the reaction mixture of the RNA substrate and the chemically synthesized Dz5C; and Lane 6, the reaction mixture of the RNA substrate and the Dz5C produced through PaNDA. DNA and RNA were visualized by SYBR Gold. (B) Overlaid ¹H-¹³C HSQC NMR spectra recorded for ¹³C,¹⁵N-labeled Dz5C in the free state and in the complex with the RNA substrate in the absence of divalent metal ions.

As shown in Fig. S6, a 46-nt ssDNA strand was also successfully prepared using PaNDA. In this case, the reaction mixture showed the noticeable presence of shorter fragments, possibly due to star activity of the nicking enzyme or dissociation of the polymerase before reaching the 5’-end of the template. We were able to remove the shorter byproducts through anion-exchange chromatography. However, we should note that such byproducts will likely increase when PaNDA is used for longer ssDNA. The size limit for PaNDA is discussed in the SI.

In conclusion, PaNDA allows for simple and robust enzymatic production of various ssDNA strands in large quantity and greatly facilitates isotope-labeling of ssDNA oligonucleotides, including those that were very difficult to produce with other enzymatic methods. Since heteronuclear ¹H/¹³C/¹⁵N NMR spectroscopy is powerful for characterizations of nucleic acids,^28,29 we expect that PaNDA will promote biophysical characterizations of various ssDNA-based materials and thereby help develop DNA-based biotechnologies. For example, PaNDA may open an avenue for solid-state NMR studies of DNA nanomachines, in which only specific DNA strand is ¹³C,¹⁵N-labeled.