Robust IR Based Detection of Stable and Fractionally Populated G-C⁺ and A-T Hoogsteen Base Pairs in Duplex DNA

Abstract

Non-canonical G-C⁺ and A-T Hoogsteen base pairs can form in duplex DNA and play roles in recognition, damage repair, and replication. Identifying Hoogsteen base pairs in DNA duplexes remains challenging due to difficulties in resolving syn versus anti purine bases with X-ray crystallography; and size limitations and line broadening can make them difficult to characterize by NMR spectroscopy. Here, we show how infrared (IR) spectroscopy can identify G-C⁺ and A-T Hoogsteen base pairs in duplex DNA across a range of different structural contexts. The utility of IR-based detection of Hoogsteen base pairs is demonstrated by characterizing the first example of adjacent A-T and G-C⁺ Hoogsteen base pairs in a DNA duplex where severe broadening complicates detection with NMR.

Introduction

NMR relaxation dispersion studies^1--6 have shown that in canonical DNA duplexes, Watson-Crick A-T and G-C base pairs (bps) exist in dynamic equilibrium with their Hoogsteen^7,8 counterparts, which have populations on the order of 0.1% to 1% (Fig. 1). A-T and G-C⁺ Hoogsteen bps can form by flipping the purine base in a Watson-Crick bp and creating a new set of hydrogen bonds that require closer proximity of the two bases and protonation of cytosine N3 in the case of G-C. (Fig. 1). While Hoogsteen bps are a minor species in naked DNA duplexes, they can become the dominant species in DNA-drug^9--16 and DNA-protein^17--21 complexes where they are thought to play roles in DNA recognition,²² in damaged DNA where they are thought to play roles in damage induction, accommodation and repair;^23,24 and in the active site of Y-family polymerases such as human polymerase iota^25,26 where they are proposed to play roles in damage-bypass during replication.

Figure 1. Chemical structures for Watson-Crick and Hoogsteen bps.

Hydrogen bond acceptors are shown in red, and donors in blue. Green arrows denote the C1′ to C1′ distances between the Watson-Crick and Hoogsteen bps.

The detection and characterization of Hoogsteen bps can pose unique challenges to existing structural characterization techniques. There are now several studies documenting difficulties in distinguishing Hoogsteen from Watson-Crick bps by X-ray crystallography particularly for low-to-medium resolution maps. For example, a crystal structure of the Y-family DNA polymerase iota revealing Hoogsteen pairing during replication^25,26 was challenged²⁷ on the grounds that the weak electron density for the active site A-T bp makes it difficult to resolve a Watson-Crick from a Hoogsteen geometry. There have been other studies documenting difficulties in resolving Hoogsteen versus Watson-Crick bps in X-ray structures of DNA-homeodomain¹⁷ and DNA-p73²⁰ complexes. Because of these ambigu-ities,¹⁷ it is conceivable that Hoogsteen bps may have been mismodeled as Watson-Crick in existing X-ray structures of DNA complexes. The characterization of Hoogsteen bps by X-ray crystallography can also be complicated by potential effects arising due to crystal packing forces. For example, several AT-rich sequences form DNA duplexes composed of Hoogsteen bps by X-ray crystallography^28--32 but predominantly form canonical duplexes composed of Watson-Crick bps when examined using solution NMR. While solution NMR can be used to unambiguously characterize Hoogsteen bps, size limitations hinder application to large protein-DNA complexes. There have also been documented difficulties^4,6,33 in detecting Hoogsteen bps by NMR due to severe line broadening of resonances arising either because a fractional population of Hoogsteen (>20%) exchanges with the Watson-Crick form on the microsecond-to-millisecond timescale or due to enhanced solvent exchange. Such broadening contributions can lead to a blind spot for NMR-based detection, even in small systems. Both NMR and X-ray also require significant amounts of nucleic acid material, and data collection and analysis can be significantly time-consuming.

IR spectroscopy can in principle provide a convenient approach for identifying Hoogsteen bps. With vibrational spectroscopy's inherently picosecond timescale,³⁴ IR is unburdened by molecular weight or chemical exchange. DNA has been extensively studied^35--37 with both Raman and IR spectroscopy. Early work comparing the Raman or IR spectrum to the X-ray structures of crystalline DNA, aided by computational methods and isotopic labeling,^38--50 allowed precise correlations between the geometry of short duplexes and their wavenumber positions;^51--53 as well as yielding marker bands diagnostic^54--60 for a number of DNA conformations and environments.

IR and Raman have previously been used to examine the Hoogsteen bps in 1-methyl thymine and 9-methyladenine co-crystals,^61,62 in triple helices,^63--66 G-quadraplexes,^67--70 and an all-Hoogsteen parallel duplex DNA.⁷¹ While there have also been a few Raman and IR studies of Hoogsteen bps in DNA,^72--74 these have primarily focused on G-C⁺ bps under non-physiological low pH conditions in which other features of the DNA structure could be affected due to protonation. More recently, IR studies have examined⁷⁵ the electronic excited state of G-C⁺ Hoogsteen bps in D₂O in a duplex containing nine alternating G-C bps GC₉. Fewer studies have examined A-T Hoogsteen bps within DNA duplexes with most studies focusing on non-duplex environments such as co-crystals of 9-methyladenine and 1-methylthymine monomers,^{61, 62} triplexes^{63--65, 76--78} and parallel duplexes.⁷¹ This is in part due to challenges in obtaining inducible or trapped A-T Hoogsteen bps in DNA duplexes. Thus, IR marker bands for G-C⁺ Hoogsteen bps in duplex DNA in H₂O under physiological pH conditions remain to be fully characterized, and IR marker bands for A-T Hoogsteen bps in anti-parallel DNA duplexes remain to be established. The robustness of these Hoogsteen specific IR marker bands under a variety of DNA duplex conditions has yet to be assessed and is key for establishing IR as a method for Hoogsteen detection.

Given the growing interest in Hoogsteen bps within the duplex environment, we developed an IR-based approach for the robust detection of G-C⁺ and A-T Hoogsteen bps under physiological conditions. We used NMR verified model duplexes containing various compositions of G-C⁺ and A-T Hoogsteen bps to identify and validate robust IR marker bands for G-C⁺ and A-T Hoogsteen bps. We show that these IR marker bands can be used to detect Hoogsteen bps in duplex DNA even in cases where they are NMR-invisible due to chemical exchange. The utility of IR-based detection of Hoogsteen bps is demonstrated by characterizing the first example of adjacent A-T and G-C⁺ Hoogsteen base pairs in a DNA duplex where severe line broadening complicates NMR-based detection under physiological pH values.

Methods

Sample preparation

All chemicals (NaCl, sodium phosphate, HCl, KOH, DCl, KOD and D₂O) used in both ddH₂O and D₂O buffer preparation were purchased from Sigma-Aldrich. Unmodified DNA samples were purchased as single-stranded oligonucleotides from Integrated DNA Technologies (IDT) with standard desalting purification. The m¹G and m¹A modified DNA constructs were purchased from Yale-Keck Oligo Synthesis Resource (W.M. Keck Foundation, New Haven, CT).

Samples were prepared by annealing at 95 °C for 5 min followed by 40 min of cooling at ambient temperature in pure ddH₂O, after which they were exchanged 3× into their respective buffers with Amicon Ultra-15 or Ultra-0.5 centrifugal filters with a 3 kDa MW cutoff (Millipore). The samples with N1-methyladenine were annealed at 65 °C for 5 min. Concentrations were determined via UV-vis. with a NanoDrop 2000 (ThermoFisher) using the supplier provided extinction coefficients at 260 nm. Purity was assessed with 20% denaturing PAGE (Fig. S1A).

For the D₂O experiments, samples were exchanged 3× into the D₂O buffers at pD 6.8 (15 mM sodium phosphate, 25 mM NaCl) or pD 5.4. The pD on all D₂O buffers was determined by adding 0.4 to the pH value after incubating the meter in D₂O. The D8 GC₉ isotope construct was prepared using a protocol described for RNA.^79,80 Approximately 1 μmole of GC₉ was re-solubilized in 99.9% D₂O to a concentration of 1 mM. Triethylamine (TEA) was then added to a concentration of 75 mM, incubated at 95 °C for 24 hours, and lyophilized overnight to remove D₂O and TEA. Finally, the D8 DNA was exchanged 3X into either the pH 3.1 buffer (1 mM glycine, 1 mM NaCl, pH 3.1) or a low NaCl, pH 7 buffer (1 mM Tris, 1 mM NaCl, pH 7).

DNA:echinomycin complexes were formed as previously described¹³ by adding 3× molar excess of echinomycin in methanol to the DNA duplex and shaking at 4 °C for 3 hours. The sample was then placed under a dry air flow until it was completely dehydrated, after which it was re-suspended in ddH₂O.

CD and UV-visible spectroscopy

CD spectra were measured on an Aviv 202 CD spectropolarimeter equipped with a Peltier temperature control unit and a recirculating, cooling water bath. The temperature was set at 15 °C and the sensitivity was 100 mdeg. Concentrations of the GC₉ construct were 30 μM for each buffer condition (25 mM NaCl, 15 mM sodium phosphate, pH 6.8; and 1 mM glycine 1 mM NaCl, pH 3.1). The average of three measurements were taken and the buffer blank subtracted out with a custom Python script. UV-vis. measurements were performed on a ThermoFisher NanoDrop 2000 using a 300 μL quartz cuvette from Starna Cells Inc.

NMR spectroscopy

NMR experiments were performed on a Bruker Avance III 600MHz spectrometer or 700-MHz spectrometer equipped with 5 mm triple-resonance cryogenic probes. Assignments of GC₉ and CA-DNA:echinomycin were determined using two dimensional (2D) ¹H-¹³H Heteronuclear Single Quantum Coherence (HSQC), 2D SO-FAST ¹H-¹⁵N Heteronuclear Multiple Quantum Coherence (HMQC) imino,⁸¹ and 2D ¹H-¹H Nuclear Overhauser Effect Spectroscopy (NOESY) experiments. GC₉ experiments were performed at 10 °C with 1 to 2 mM GC₉ at 10 °C, and CA-DNA:echinomycin experiments were performed at 25 °C at concentrations of 1 mM. Data were processed with NMRpipe⁸² and analyzed in SPARKY (http://www.cgl.ucsf.edu/home/sparky/).

IR spectroscopy

The IR measurements were performed with a PerkinElmer Frontier FTIR with a Pike Technologies MIRacle ATR (attenuated total reflectance) KSR5 coated diamond optic and a narrow band MCT detector. Measurements on buffer exchanged DNA were performed by taking the background scans with the flow-through buffer on the crystal. Typically, 128 scans and 4 cm⁻¹ resolution using approximately 1 to 4 μL of 3 to 4 mM sample was used; and measurements were performed at ambient temperature. CO₂ and water vapor was subtracted from the spectrum using the provided automatic background removal function. For the direct titration of GC₉, 32 scans and 6 μL was employed for the initial measurement, after which 0.5 μL aliquots of 48 mM HCl in ddH₂O was added to the droplet, aspirated with a pipette three times to ensure mixing, followed by acquisition. These were found to be comparable to results from GC₉ exchanged into buffers of known pH (Fig. S1B).

IR data processing was performed with the PerkinElmer Spectrum software. Spectra were normalized to the 970 cm⁻¹ band, which have been found in previous work to be invariant to multiple conformational changes and were insensitive to D₂O exchange;³⁷ and all IR intensities are reported in normalized ATR units. The interactive baselining function was used to add a multipoint baseline prior to normalization. Subtraction spectra were calculated using the difference function on the normalized and baselined spectra. For the high salt GC₉ IR experiments, the construct was exchanged into 1 mM glycine and 300 mM NaCl, pH 6.9, at approximately 8 mM GC₉. The Z-form IR spectra of GC₉ were taken at pH 7.0 in buffers with 4 M NaCl and 15 mM sodium phosphate. The IR measurement of TA-DNA and CA-DNA echinomycin complexes was performed in a 15 mM sodium phosphate, 25 mM NaCl buffer at either pH 6.8 or 5.4. To reduce spectral congestion from the echinomycin IR bands, the spectrum of echinomycin suspended in water was first subtracted from the spectrum of the DNA:echinomycin complex to remove the strong amide bands present in the echinomycin spectrum and to isolate the IR spectrum of the DNA itself.

Results

Characterizing a pH-induced Watson-Crick to Hoogsteen transition in GC₉ by UV-visible, CD and NMR spectroscopy

Prior Raman, UV-vis, and CD studies^72,75 showed that under low salt concentrations (1 mM NaCl), GC₉ undergoes a transition from a Watson-Crick duplex to a Hoogsteen duplex when the pH is lowered to below 4.5. This pH inducible Hoogsteen duplex provides an ideal opportunity to identify IR bands unique to G-C⁺ Hoogsteen bps. We first confirmed this transition using UV-visible and CD spectroscopy (Fig. 2). Consistent with prior CD^83,84 and vibrational studies,^72,75 reducing the pH to below 4.5 resulted in an increase in intensity at 280 nm in the UV-vis reflecting formation of cytosine cations (Fig. 2A, top) and the appearance of a weak positive signal at 290 nm and weak negative signal at 255 nm in the CD spectrum attributed to formation of G-C⁺ Hoogsteen bps (Fig. 2A, bottom).

Figure 2. UV, CD, and NMR spectra of GC₉.

(A) UV-visible (top) and CD spectra (bottom) of GC₉ at low (red) and neutral (blue) pH. (B) SOFAST ¹H 1D NMR spectra of GC₉ at pH 6.8 (blue), 3.9 (black) and 3.1 (red), (all in 1 mM glycine, 1 mM NaCl, 0.1 mM EDTA, and 10% D₂O buffers). (C-E) Aromatic (C), aliased C1′-H1′ (D) and C5-H5 (E) HSQC spectra. (F) NOESY of GC₉ at pH 3.1 (buffer conditions as in (B)). Labels for resonances belonging to Hoogsteen and Watson-Crick forms are shown in red and blue, respectively.

We used NMR spectroscopy to further verify formation of G-C⁺ Hoogsteen bps at low pH in GC₉. We observed unique NMR signatures of G-C⁺ Hoogsteen bps⁸⁵ at pH 3.1, including the imino resonances of C⁺ at approximately 15 p.p.m. and a resonance at approximately 11 p.p.m. reflecting the now solvent-exposed G N3 (Fig. 2B). We also observe two sets of resonances in the 2D C-H HSQC spectra in slow exchange on the NMR timescale; a major Watson-Crick species and minor (approximately 30% based on volume analysis) Hoogsteen species with characteristic^1,3,4,6 approximately 2 to 3 p.p.m. downfield shifted G C1′, G C8, and C C6 and approximately 1 p.p.m. upfield shifted C C5 resonances (Fig. 2C-E). The Hoogsteen resonances also exhibit the expected Hoogsteen-specific NOE-based distance connectivities,^{1,3,4,6,24,86,87} including an enhanced NOE cross peak between G-H8 and G-H1′ due to formation of a syn base (Fig. 2F). While the highly repetitive sequence makes it difficult to assign NOEs to specific nucleotide positions along the DNA, we do observe sequential NOEs between the Hoogsteen species, and between Watson-Crick species, but not between Watson-Crick and Hoogsteen species, consistent with cooperative formation of all Hoogsteen or all Watson-Crick bp helices. However, we cannot rule out the existence of species with mixed Watson-Crick and Hoogsteen populations that are not readily detectable or assignable based on the NOESY data.

IR marker bands allow detection of stable and fractional G-C⁺ Hoogsteen base pairs in duplex DNA

We used the pH inducible GC₉ duplex to identify unique IR bands that could potentially be used in the identification of G-C⁺ Hoogsteen bps in DNA duplexes in H₂O. We observe significant changes in the IR spectra of GC₉ when lowering the pH from 7.0 to 3.1 (Fig.s3 and S1B and C). The difference IR spectra of GC₉ show the all the expected variations due to formation of Hoogsteen base pairs^{63--65,67--70,77,78} when lowering the pH from 7.0 to 3.1 (Fig. 3). These include increased intensities in Hoogsteen-specific IR bands due to formation of a syn G (increases at 1730 cm⁻¹, between 1350 and 1365 cm⁻¹; and between 1315 and 1325 cm⁻¹), protonated C⁺ (increases at 1730, 1540, and 1282 cm⁻¹), and a decrease at 1498 cm⁻¹ due to Hoogsteen type H-bonding at G N7 (Fig. 3B and Table 1).

Figure 3. Hoogsteen IR marker bands in GC₉.

(A) IR spectra for a direct pH titration of GC₉, with the initial Watson-Crick GC₉ signal (blue) converting to Hoogsteen (red) through intermediates (grey). Initial buffer conditions were pH 7.0, 1 mM NaCl, 1 mM glycine. (B) Low - neutral pH IR difference spectra of the pH titration spectra in (A), red line is approximately pH 3.1. Hoogsteen IR marker bands are in red, Watson-Crick are in blue, and backbone are in black. Bands whose assignments are preliminary are in italics. Inset: Structures of Watson-Crick (blue) and Hoogsteen G-C⁺ (red) bps with highlights on the atoms which are major contributors to the IR marker bands. All spectra are normalized to the 970 cm⁻¹ band.

Table 1.

Watson-Crick and Hoogsteen IR G-C⁺ signals in H₂O.^57,72,110 Positions are in wavenumbers (cm⁻¹). Normal mode composition for the Watson-Crick form is given in parenthesis listed with the highest contributors first.

Watson-Crick	Hoogsteen	Notes
1708	1730	C+ and GC6=O (C6=O, C5-C6, C5-C4, N1-H, N1-C6 for G35)
1530	1540	C+ (skeletal out of phase ring mode37)
1498	1498	Decrease in intensity upon increased H-bonding to N7 (N7=C8, C8-N9, C4=C5, C4-C9, N-C8H of G37)
1375	1350	syn G (C1′-N9, N9-C1′H, N7=C8 of G111)
1330	1320	syn G (C2′-H, C1′-C2′-H, C3′-C2′-H of G111)
1296	1282	C+ (C4-NH2 of C)

Control pH titrations on a mixed sequence with GC steps did not yield these spectral changes (Fig. S1D) indicating that the IR markers assigned to Hoogsteen bps do not arise from more generalized protonation of DNA duplexes. We confirmed assignment of the 1730, 1540, and 1282 cm⁻¹ bands to C⁺ based on observation of isotope shifts when exchanging the sample into D₂O and collecting IR data at two pDs (Fig. S2A and B). The 1498 cm⁻¹ band is an important marker of Hoogsteen type H-bond at G N7.^60,88--94 Based on studies of triplexes and quadruplexes³⁷ this band is expected to appear between 1480 to 1490 cm⁻¹. We confirmed that the 1498 cm⁻¹ band in the DNA duplex has significant contributions from a mode involving the G N7 via deuterium labeling at the nonexchange-able G H8 sites (Fig. S2C and D); as previously described.^{38,56, 95}

The above Hoogsteen IR signals in GC₉ were assigned based primarily on prior Raman work.^72,75 We located two other IR markers which may be tentatively assigned to Hoogsteen in plane C=O stretching and NH₂ modes. The G NH₂ and C C2=O each lose a H-bond in the Hoogsteen bp, and thus bands with contributions from these functional groups are expected to increase in wavenumber. The negative band at 1620 cm⁻¹ and the positive band at 1640 cm⁻¹ are consistent with loss of H-bonding to a G NH₂,³⁵ while the loss at 1655 cm⁻¹ and gain between 1680 and 1670 cm⁻¹ are consistent with a similar loss of an H-bond to the C C=O.³⁵ Supporting this assignment is the large isotope shift these bands undergo in D₂O, particularly for the 1620 cm⁻¹ band (Fig. S2A). These spectral changes may also arise from alterations in the G N1-C6 and G N1-H ring modes^35,37 when in Hoogsteen bps, and which also have IR bands between 1600 and 1700 cm⁻¹.

Formation of Z-DNA under low pH conditions is not expected based on the low ionic strength conditions used and could be ruled out based on the NMR and CD studies above, and by lack of an increase in the IR spectrum at the IR marker bands for Z-DNA³⁷ at 1690 cm⁻¹ and 1215 cm⁻¹ (Fig. S3). Exchanging GC₉ into a 300 mM NaCl pH 3.1 buffer also resulted in the expected suppression⁷² of the Hoogsteen-specific IR signals (Fig. S3, bottom). Formation of non-duplex structures such as G-quadraplexes was additionally ruled out by native PAGE gel electrophoresis (Fig. S1A). Finally, base open states were ruled out since they are expected to blue shift the N7 H-bond marker (1498 cm⁻¹) band due to decreased H-bonding and a red shift is observed consistent with increased H-bonding to a G N7 (1498 cm⁻¹ in pH 3.1 GC₉).

Interestingly, during the above studies of GC₉, Hoogsteen bps could be detected using IR under conditions where they were undetectable by NMR due to extensive line broadening. In the NMR spectrum of GC₉, no Hoogsteen iminos were detected at pH 3.9, with weak imino signals only appearing at pH 3.1 (Fig. 2B, red line). In contrast, the IR G-C⁺ Hoogsteen signature is clearly visible in the difference spectrum over a full range of pH values, including pH 3.9 (Fig. 3B, red line; and Fig. S1B). This highlights one of the advantages of using IR in the detection of fractional Hoogsteen bps in DNA duplexes.

IR detection of a single m¹G-C⁺ Hoogsteen base pair in a Watson-Crick duplex under physiological pH

In studies of Hoogsteen bps in duplex DNA, one will often be interested in the detection of single Hoogsteen bps surrounded by Watson-Crick bps under physiological pH conditions. Such single Hoogsteen bps have been observed in DNA:drug^{9,12,33, 96--99} and DNA:protein complexes^{17--19, 21} and in DNA containing damaged nucleotides by both X-ray crystallography^{17,20, 25,26} and NMR^2--4,6,33 spectroscopy. We therefore examined the feasibility of using the IR markers of G-C⁺ Hoogsteen bps obtained for GC₉ at low pH to characterize single G-C⁺ Hoogsteen bps surrounded by Watson-Crick under physiological pH. This is also an important test of the sensitivity of the IR-based method and ability to resolve a few Hoogsteen bps from a background of Watson-Crick bps. For these studies, we trapped a single G-C⁺ Hoogsteen bp by incorporating N1-methylated guanine (m¹G) in the well characterized^1,4,6,100 d(GCCAAAAAATCG) duplex (A6-DNA^m1G, Fig. 4A). Prior NMR studies showed^3,6,24,86 that m¹G-C⁺ forms a stable Hoogsteen bp surrounded by minimally perturbed Watson-Crick bps.

Figure 4. Hoogsteen IR marker bands in m¹G-C⁺ base pairs.

(A) Structure of the m¹G-C⁺ bp and the A6-DNA sequence; m¹G site in red. (B) IR spectra of A6-DNA^m1G (red) and A6-DNA IR spectra (black) at pH 5.4 (top) and 6.8 (bottom). (C) Comparison of the A6-DNA^m1G - A6-DNA (red) and the GC₉ low - neutral pH (grey) difference spectra. Shaded regions and red labels indicate the positions of Hoogsteen G-C⁺ marker bands present in both spectra; black labels indicate backbone IR bands. (D) IR difference spectra of A6-DNA^m1G at pH 5.4 (light grey), pH 6.8 (black) and pH 9 (blue), all subtracted from A6-DNA at pH 6.8. Red labels indicate marker bands for Hoogsteen that are lost as the pH is increased. Buffers consisted of 25 mM NaCl and 15 mM sodium phosphate. All spectra are normalized to the 970 cm⁻¹ band.

The IR spectrum of A6-DNA^m1G and A6-DNA (Fig. 4B) were subtracted, and their difference spectra compared to the difference spectra from low versus neutral pH GC₉ (Fig. 4C). The 1730 cm⁻¹ signal and other assigned signature G-C⁺ Hoogsteen bands are clearly observed in the A6-DNA^m1G difference spectra (Fig. 4C, Table 1). Exchange into pH 9.0, 6.8, and 5.4 buffers resulted in the expected increases/decreases under acidic or basic H₂O or D₂O conditions (Fig.s 4D and S4) with controls demonstrating these bands to be distinct from IR signals due to generalized duplex protonation (Fig. S4). These results show that the IR G-C⁺ Hoogsteen bp marker bands can be used to detect single G-C⁺ Hoogsteen bps surrounded by Watson-Crick bps under physiological pH.

IR marker bands of A-T Hoogsteen base pairs

Studies have shown^1,3,5,28--32 that A-T Hoogsteen bps are energetically more favorable than G-C⁺ Hoogsteen bps and are more likely to form in duplex DNA. Fewer studies have examined IR and Raman bands of A-T Hoogsteen bps as compared to G-C⁺ in part due to the absence of model duplexes for inducing Hoogsteen such as GC₉. While formation A-T Hoogsteen bps involves smaller changes in H-bonding and charge as compared to G-C⁺ Hoogsteen bps, a more subtle A-T Hoogsteen IR signature is still expected. Based on calculations and by analogy to the syn adenines in Z-DNA,^{39,42, 55,101} A-T Hoogsteen IR marker bands are expected to lie at approximately 1337 cm⁻¹ (shifted from 1344 cm⁻¹ for anti A), between 1315 and 1325 cm⁻¹ (shifted from approximately 1335 cm⁻¹ for anti A), between 1365 and 1355 cm⁻¹ (shifted from approximately 1375 cm⁻¹ for anti A), and between approximately 1437 and 1442 cm⁻¹ for a syn A (shifted from 1426 cm⁻¹ anti A bands). Other bands for Hoogsteen A-T bps are expected^35,40 to lie between 1650 and 1708 cm⁻¹ (H-bond alterations at A N7, T C4=O, and A N6-H₂), and approximately 1309 cm⁻¹ (A N9-C8 and N3=C2).

We used m¹A to trap an A-T Hoogsteen bp in two previously studied¹ DNA duplexes; d(GCCAAAAAATCG) (A6-DNA) and d(GCCAATTGATCG) (A2-DNA) (Fig. 5A). We also investigated IR marker bands for A-T Hoogsteen bps using a d(ACACCTACGTGT) (TA-DNA, Fig. 5A) duplex in complex with the antibiotic echinomycin^12--16 that forms A-T Hoogsteen bps. Echinomycin binds preferentially to GC steps⁹ separated by spacer bps (Fig. 5A, blue). NMR,^12,13 X-ray,^14,15 and footprinting¹⁶ have shown that for spacers with a TA step, complex formation results in tandem A-T Hoogsteen bps.

Figure 5. Persistent IR Signals for A-T Hoogsteen base pairs in different contexts.

(A) Chemical structure of m¹A (top), and A6-DNA^m1A, A2-DNA^m1A, and TA-DNA sequences. Hoogsteen base pairs are in red and echinomycin binding site in blue. (B) Overlay of A6-DNA^m1A - A6-DNA (top), A2-DNA^m1A - A2-DNA (middle), and TA-DNA:drug - drug - free TA-DNA (bottom) difference IR spectra. Bands due to Hoogsteen G-C⁺ present in both spectra are shown in red; shaded regions indicate the range of positions seen for these marker bands. Buffers consisted of 25 mM NaCl and 15 mM sodium phosphate. All spectra are normalized to the 970 cm⁻¹ band.

Significant changes are apparent in the difference spectra when comparing A6-DNA^m1A and A2-DNA^m1A with their respective unmethylated duplexes (Fig. 5B and Table 2, and Fig. S5A and B) that are distinct from those seen for m¹G (Fig. S5A). This was also the case when comparing IR spectra of the TA-DNA with and without the echinomycin (Fig. S5C and D). We observe consistent changes across the three systems including the expected decrease in intensity in the N7 band (1490 cm⁻¹ in the TA-DNA:drug complex) due to Hoogsteen-type H-bonding to A N7³⁷ and the expected increases between 1325 and 1315 cm⁻¹, approximately 1337 cm⁻¹, between 1365 and 1350 cm⁻¹, and an increase at approximately 1437 cm⁻¹ due to formation of a syn A (Fig. 5B and Table 2). This agreement, particularly in these conformationally sensitive regions of the IR, across three separate Hoogsteen A-T bp contexts supports these bands as markers for A-T Hoogsteen formation.

Table 2.

Preliminary assignments for Hoogsteen IR A-T marker bands in H₂O observed in A6-DNA^m1A - A6-DNA, A2-DNA^m1A - A2-DNA, and the TA-DNA:drug complex - drug - free TA-DNA difference IR spectra. Positions are in wavenumbers (cm⁻¹). Normal mode composition for the Watson-Crick form is given in parenthesis listed with the highest contributors first.

Watson-Crick	Hoogsteen	Notes
1498	1498	Decrease in intensity upon increased H-bonding to N7 (N-C8H, N7=C8, C2=N3 of A37)
1455	1436 to 1444	syn A (N1-C6, C6-N12 of A35)
1375	1350 to 1360	syn A (C1′-N9, C6-N6, N9-C1′H of A42)
1344	1335	syn A (N9-C8-H, C2-N3-H, C2=N3 of A42)
1344	1320	syn A (N9-C8-H, C2-N3-H, C2=N3 of A42)
1300 to 1308	1300 to 1308	A-N9=C8, decrease in intensity (N9-C8, N3-C2, C8-H, C2-H of A35)

Interestingly, the A6-DNA^m1A and A2-DNA^m1A difference IR spectra do not overlay perfectly with each other, with changes from the A6-DNA^m1A duplex significantly more intense in the double bond region than in A6-DNA^m1A (Fig. S5B). This increase may reflect formation of a more stable Hoogsteen m¹A-T bp in A6-DNA^m1A than in A2-DNA^m1A, as reported previously by NMR.¹ The two m¹A-T Hoogsteen containing duplexes and the complex do show differences between 1700 to 1550 cm⁻¹ region which may reflect more globally altered base pairing in the complex as compared to the m¹A duplexes (Fig. S5D).

Application of IR in detecting tandem G-C⁺/A-T Hoogsteen bps

Thus far, echinomycin^{33, 96--99} and the related molecule triostin A⁹ are the only two molecules that have been shown to induce Hoogsteen bps in duplex DNA. These compounds have been shown to induce Hoogsteen bps in a manner strongly dependent on DNA sequence. For example, when the spacer sequences between two CG binding sites are TA/AT, they form tandem Hoogsteen bps, however Watson-Crick bps are observed⁹⁹ when the spacer sequences are AT/TA. NMR studies⁹⁹ of DNA:echinomycin and DNA:triostin complexes have encountered difficulties in the detection of Hoogsteen bps due to severe line broadening. These challenges could potentially be overcome with the IR-based method which combined with the lower cost and ease of data collection and analysis might provide an efficient method to screen for compounds that induce Hoogsteen.

Hoogsteen bps often occur in tandem as adjacent TA/AT and CG/GC Hoogsteen bps in DNA-drug,^9--16 and DNA-protein^{17--19, 21} complexes. Interestingly, adjacent A-T and G-C⁺ Hoogsteen bps have not been observed to date. As an application of the IR-based detection of Hoogsteen bps, we examined whether tandem A-T and G-C⁺ Hoogsteen bps could be formed by placing a CA spacer sequence between the two echinomycin binding sites in the d(ACGTGCGT) duplex (CA-DNA, Fig. 6A). Due to line broadening, NMR spectra of the CA-DNA:drug complex at pH 6.8 were inconclusive as to whether Hoogsteen bps were present (Fig. 6A). However, at pH 5.4 we observed NMR signatures characteristic of both G-C⁺ and A-T Hoogsteen bps (Fig. 6A). These included resonances from C⁺ iminos; as well as downfield shifted purine C8 (Fig. S6A) and C1′ (Fig. S6B) chemical shifts. To our knowledge, this represents the first observation of a G-C⁺ and A-T Hoogsteen bps adjacent to one another.

Figure 6. IR evidence for tandem G-C⁺/A-T echinomycin-induced Hoogsteen base pairs under physiological pH.

(A) CA-DNA sequence and SOFAST ¹H 1D NMR spectra of the pH 6.8 (blue) and pH 5.4 (red) CA-DNA:drug complex in 15 mM sodium phosphate, 25 mM NaCl, 0.1 mM EDTA, and 10% D₂O buffers. (B) The 1700 to 1400 cm⁻¹ region for A6-DNA^m1A - A6-DNA (top), A6-DNA^m1G - A6-DNA (bottom), and the CA-DNA:drug - drug - free CA-DNA spectra at pH 6.8 (middle, red). (C-F) Enhanced views of the 1480 to 1460 cm⁻¹ (C), 1370 to 1345 cm⁻¹ (D), 1330 to 1300 cm⁻¹ (E), and 1300 to 1270 cm⁻¹ (F) IR difference spectra; labels as in (B). Bands due to the tandem formation of G-C⁺ and A-T Hoogsteen bps are in red; shaded regions indicate the range of positions seen for these marker bands. Buffers for the IR experiments consisted of 25 mM NaCl and 15 mM sodium phosphate. All spectra are normalized to the 970 cm⁻¹ band.

We used IR to examine whether the tandem G-C⁺ and A-T Hoogsteen bps persist under physiological pH. IR difference spectra of the CA-DNA:drug complex was compared to the G-C⁺ Hoogsteen bp in A6-DNA^m1G as well as A6-DNA^m1A (Fig. 6B - F). Unlike NMR, the CA-DNA:drug IR difference spectrum shows changes consistent with both A-T and G-C⁺ Hoogsteen bps at neutral pH (Fig. 6B). Weak bands due to G-C⁺ appear at 1730 cm⁻¹, 1540 cm⁻¹, and 1280 cm⁻¹ (C⁺), between 1436 and 1441 cm⁻¹ (syn A), between 1360 and 1355 cm⁻¹ (syn A and G), approximately 1335 cm⁻¹(syn A), and between 1325 and 1315 cm⁻¹ (syn A and G). The CA-DNA:drug complex shows both sets of purine syn bands, indicating the presence of both A-T and G-C⁺ Hoogsteen bps. As expected, the G-C⁺ Hoogsteen signals increased upon lowering the pH (Fig. S6C), consistent with the NMR data. These results again highlight how IR can be applied to identify Hoogsteen bps under conditions where detection via NMR is significantly hampered due to line broadening from solvent exchange.

Discussion

The results presented in this work indicate that IR marker bands that are unique to Hoogsteen bps can be used to robustly detect Hoogsteen bps in duplex DNA under a wide range of conditions and contexts; including low pH duplexes with all G-C⁺ Hoogsteen bps, single m¹A-T and m¹G-C⁺ Hoogsteen bps surrounded by Watson-Crick bps; and AT/TA and AT/GC Hoogsteen bps in DNA:drug complexes. Our study also shows that IR can be used to identify Hoogsteen bps that are only fractionally populated (>10%). Unlike NMR, IR is not burdened by chemical/solvent exchange line broadening. There are now several cases suspected to involve exchange with Hoogsteen in which the key resonances are broadened precluding NMR-based assessment of Watson-Crick versus Hoogsteen.^4,6,33 The added convenience of IR and lower sample requirements makes it an ideal method for exploring Hoogsteen bps in a variety of contexts such as in different damaged bases, when bound to different small molecules, under different physiochemical conditions (crowding, pH, salt concentration, etc.).

Our IR studies of these different systems also provide new insights into the structure of Hoogsteen bps. It is clear from the IR difference spectra of both GC₉ and m¹G that backbone signals near 1220 and 1088 cm⁻¹ are altered between the G-C⁺ Watson-Crick and Hoogsteen. The approximately 1088 cm⁻¹ region in particular has contributions from complex modes involving C-O and P-O bonds,⁴⁸ and is sensitive to backbone parameters such as sugar pucker and torsion angles. GC₉ shows near-complete overlap with Z-DNA at 1060 cm⁻¹, indicating similarities between Z and Hoogsteen backbone geometries. A6-DNA^m1G shows a large intensity increase at 1075 cm⁻¹, an intermediate value between 1088 cm⁻¹ (B-form) and 1060 cm⁻¹ (Z-form), which is likely⁴⁸ due to ζ and α torsion angles away from away from 0° and towards 180°. This is consistent with a recent study.¹⁰² which found similar alterations to these angles in a structure-based survey of the PDB.

The P-O backbone band at approximately 1225 cm⁻¹ is also altered in A6-DNA^m1G. In A6-DNA^m1G, this band exhibits a marked decrease compared to the unmethylated duplex, with a decrease at 1216 cm⁻¹; which may indicate a transition either towards a BII-like form,¹⁰³ or a decrease in A-from-like backbone character.³⁷ This is in line with the observed¹⁰³ decrease at 1050 cm⁻¹, which is also a B-DNA backbone marker. This points to Hoogsteen G-C⁺ bps exerting an influence over the distribution of DNA conformational substrates¹⁰³ and may have implications for protein recognition of DNA by indirect readout. As with A6-DNA^m1G, there are large changes to the backbone bands at approximately 1225 cm⁻¹ and 1060 cm⁻¹ in both m¹A constructs. This increase at 1060 cm⁻¹ is observed in Z-DNA⁹¹ and m¹A, and thus m¹A containing DNA may contain some Z-DNA backbone character.

Previous studies have performed theoretical investigations of the Raman^104--107 and IR spectra^106,107 of G-C⁺ Hoogsteen bps.¹⁰⁷ Early calculations¹⁰⁴ typically found shifts in the opposite direction to those observed in experiment. This prompted a few studies^105--107 that investigated the effects of anharmonicity on the vibrational density of states of Watson-Crick and Hoogsteen bps. These anharmonic corrections^105,106 to the potential provided better predictions of the Raman experimental value for the Hoogsteen protonated cytosine at 1260 cm⁻¹ and which we observe at 1280 cm⁻¹ in the IR measurements. The anharmonic corrections were particularly important for the G-C⁺ Hoogsteen bp modes involving the extra proton. Additionally, these studies^105,106 list several high frequency IR modes (between 1740 and 1770 cm⁻¹, depending on the potential used) due to carbonyl stretching and N-H bending modes that are not observed in our experimental spectra. In addition, these theoretical studies did not predict the change in position of the cytosine ring mode which we observe at around 1540 cm⁻¹ in Hoogsteen (1530 cm⁻¹ in Watson-Crick). The data provided in this work may help guide future theoretical studies of Hoogsteen IR bands.

As IR is unaffected by molecular weight, it holds great promise in the identification of Hoogsteen bps within DNA-protein complexes. The spectral congestion in larger complexes will present challenges due to overlap with Hoogsteen marker IR bands; however these problems may be overcome using isotope editing strategies^108,109 already being used for RNA and nucleic acid:protein complexes.

Conclusions

In conclusion, we have identified robust Hoogsteen IR bands that can be used in the identification of G-C⁺ in duplex DNA and made preliminary assignments of bands for A-T Hoogsteen bps. The IR spectra of pH titrations on mixed sequence content DNA (A6-DNA and A2-DNA) showed some changes in the IR, but were much smaller in magnitude than those seen for the Hoogsteen GC9. Significant changes were also observed for A-T Hoogsteen base pairs in both N1-methylated adenines and in DNA-drug complexes. We used these marker bands to identify the first example of tandem A-T and Hoogsteen bps in a DNA:drug complex. The persistence of these different IR marker bands across a range of different sequence contexts and environments is a promising sign for future studies examining larger DNA complexes for the presence of Hoogsteen bps, as well as analysis of structural perturbations that occur in large DNAs and RNAs containing modified bases.