Infrared Spectroscopic Observation of a G–C⁺ Hoogsteen Base Pair in the DNA:TATA-Box Binding Protein Complex Under Solution Conditions

Abstract

Hoogsteen DNA base pairs (bps) are an alternative base pairing to canonical Watson–Crick bps and are thought to play important biochemical roles. Hoogsteen bps have been reported in a handful of X-ray structures of protein–DNA complexes. However, there are several examples of Hoogsteen bps in crystal structures that form Watson–Crick bps when examined under solution conditions. Furthermore, Hoogsteen bps can sometimes be difficult to resolve in DNA:protein complexes by X-ray crystallography due to ambiguous electron density and by solution-state NMR spectroscopy due to size limitations. Here, using infrared spectroscopy, we report the first direct solution-state observation of a Hoogsteen (G–C⁺) bp in a DNA:protein complex under solution conditions with specific application to DNA-bound TATA-box binding protein. These results support a previous assignment of a G–C⁺ Hoogsteen bp in the complex, and indicate that Hoogsteen bps do indeed exist under solution conditions in DNA:protein complexes.

Canonical Watson–Crick base pairs (bps) form the basis for many fundamental DNA processes involving DNA, particularly those involving the act of reading the genetic code written in the DNA double helix during transcription and replication. Alternative DNA bps known as Hoogsteen bps^[1] were discovered using NMR spectroscopy^[2] to exchange with their textbook Watson–Crick forms in relatively short, naked DNA duplexes. In Hoogsteen G–C⁺ bps (Figure 1A), the guanine (G) is flipped 180° from the anti to the syn conformation and the cytosine (C) acquires a proton to form a new hydrogen bond with the G N7. Hoogsteen bps constrict the DNA backbone, alter the pair’s electrostatics, as well as induce major-groove kinking of the double helix.^[3] Thus, Hoogsteen bps present a completely different chemical face of the DNA to cellular machinery.

Figure 1.

G–C⁺ Hoogsteen bps in the crystal structure of the HG-DNA:TBP complex. A) Chemical structures and numbering for G–C Watson–Crick and G–C⁺ Hoogsteen bps. B) Crystal structure of the HG-DNA:TBP complex (PDB: 1QN3) highlighting the DNA bending. C) Enhanced view of the G–C⁺ Hoogsteen bp observed in the HG-DNA:TBP complex between C9 and G20. Protein residues that enforce this Hoogsteen bp (Leu72 and Phe57) are in green. D) The tandem G–C⁺ Hoogsteen bp between C19 and G20 (sticks) flanking the G20–C9 Hoogsteen pair (lines) thought to be an artifact. E) DNA sequences for the AdMLP DNA and the HG-DNA used in this study. At positon nine in the HG-DNA, the A–T pair is changed to a G–C pair (red). Sites of phenylalanine intercalation by TBP are indicated (blue line).

Several examples show that Nature has utilized Hoogsteen bps to carry out unique cellular functions. Many cytotoxic DNA lesions, such as N1-methylated adenine, adopt a Hoogsteen conformation,^[4] and repair enzymes^[5] have evolved to specifically recognize and repair these damaged bases. Furthermore, a member of the Y-family of DNA polymerases, polymerase iota, utilizes Hoogsteen bps to replicate DNA, thus providing mechanisms for bypassing lesions at the Watson–Crick face.^[6] Furthermore, the DNA-binding peptide antibiotic echinomycin induces Hoogsteen bps in certain DNA sequences.^[7] Finally, crystal structures reveal Hoogsteen bps in DNA bound to the tumor suppressor p53,^[8] the MATα2 homeodomain,^[9] and the transcription initiator TATA-box binding protein (TBP, Figure 1B).^[10] Recent studies employing modified bases and binding measurements also support the existence of Hoogsteen bps in a DNA:p53 complex.^[11]

Distinguishing Hoogsteen bps from Watson–Crick is difficult, as detection requires methods sensitive to changes at the level of the chemical bond. While examples from X-ray crystallography of Hoogsteen bps in DNA:protein crystal structures are scarce,^[12] such studies are biased towards those sequences and conformations that result in crystal formation. Furthermore, electron density at the DNA bases is frequently ambiguous^[13] and two-fold disorder in pseudo-palindrome duplexes can make the assignment of base conformation challenging, opening the possibility that Hoogsteen bps are prevalent but unobservable with X-ray methods. On the other hand, several examples of A–T Hoogsteen bps have been observed in crystal structures,^[14] but were later shown to be Watson–Crick by solution-state NMR spectroscopy. Thus, the question of whether these Hoogsteen sightings in DNA:protein complexes are artifacts, or if they represent true, biologically relevant forms of DNA remains unanswered; and requires solution-state detection methods with molecular resolution to answer.

While NMR spectroscopy has successfully detected A–T and G–C⁺ Hoogsteen bps in small, naked DNA duplexes,^[2,3] the size of proteins renders Hoogsteen detection in DNA:protein complexes via NMR spectroscopy both challenging and costly. We recently developed infrared (IR) methods for the detection of G–C⁺ Hoogsteen bps within DNA duplexes.^[15] Vibrational methods are molecularly resolved in addition to being immune to issues arising from molecular size.^[16] Thus, IR spectroscopy can potentially detect G–C⁺ Hoogsteen bps in large DNA:protein complexes when in the physiologically relevant solution state.

One of the key examples of G–C⁺ Hoogsteen bps was revealed by a structure of TBP bound to a variant of the adenovirus major late promoter (AdMLP) TATA-box elements, Hoogsteen-DNA (HG-DNA)^[10] (Figure 1C–E). Interestingly, while the bound AdMLP DNA sequence (Figure 1E) contained all Watson–Crick bps, careful inspection of the electron density in the variant HG-DNA sequence(Figure 1E) showed the presence of two G–C⁺ Hoogsteen bps.^[10] The Hoogsteen bps in the HG-DNA sequence occur near a distorted part of the DNA where a conserved phenylalanine residue (Figure 1C) inserts into the duplex, resulting in base unstacking and formation of favorable contacts between the purine and Phe rings.^[10,17] One of these Hoogsteen bps appeared to be necessary for binding (Figure 1C), as modeling a Watson–Crick G–C bp produced a clash with a nearby residue. In contrast, the second Hoogsteen bp (Figure 1D) appeared to be artificially induced by crystal contacts.^[10]

We used IR spectroscopy to examine whether G–C⁺ Hoogsteen bps occur in solution in HG-DNA bound to the DNA-binding domain of TBP. For these studies, we expressed and purified TBP (Supporting Information, Figure S1A). As we did not remove the His-6 affinity tag from our TBP, we performed fluorescence anisotropy (FA) to ensure the binding activity for both sequences was similar to literature values. Previous studies showed that TBP binds DNA with very high affinity, with reported apparent K_ds between 1 and 15 nm^[18] at neutral pHs. Our FA results for TBP AdMLP and HG-DNA binding (Supporting Information, Figure S1B,C) show apparent K_ds of approximately 1 nm for the AdMLP DNA and approximately 7 nm for the HG-DNA, consistent with the previously reported values.^[18c,d]

To ensure our HG-DNA:TBP complex resulted in formation of a G–C⁺ Hoogsteen bp, this complex was also crystallized and its structure determined to 2.75 Å (Supporting Information, Table S1, PDB ID 6NJQ); a sufficient resolution to determine if the conformation of the G base is syn or anti in our complex. The structure is in agreement with those previously reported (Supporting Information, Figure S2A), with TBP making numerous minor-groove contacts and inducing a significant bend in the DNA (Supporting Information, Figure S2B) to conform to the saddle-like shape of the protein. We find the sterically enforced Hoogsteen bp reported previously^[10] is indeed in the Hoogsteen configuration in the present structure (Supporting Information, Figure S3A). Notably, we found that the G–C⁺ Hoogsteen pair, thought to be an artifact in the previous work,^[10] is in the Watson–Crick form (Supporting Information, Figure S3B).

Having confirmed that the DNA:TBP complex forms a Hoogsteen bp in its crystal structure, we next used IR spectroscopy to determine if it is also present under solution conditions using the same buffer conditions (see the legend of Figure 2 for more details) as used in the original crystallography study.^[10] The IR spectra of the solution-phase DNA:TBP complex (Supporting Information, Figure S4A) were subtracted from that of free TBP at low concentration (approximately 0.2 mM), which removed signals from the protein while preventing unwanted oligomer signals (TBP has been found to form higher-order structures at high concentrations^[19]). Subtraction of free TBP from bound yields the spectrum of the bound DNA (Supporting Information, Figure S4B), which is in turn subtracted from the IR spectrum of the free DNA (Supporting Information, Figure S4B). Comparison of the HG-DNA double subtraction spectrum (Figure 2, red) to a positive control consisting of a G–C⁺ Hoogsteen duplex (GC₉, purple, Figure 2)^[15,20] shows several key signals characteristic for a G–C⁺ Hoogsteen bp within the HG-DNA:TBP complex. There is an increase at approximately 1730 cm⁻¹ due to the formation of a G C6=O in a weak Hoogsteen hydrogen bond and potentially the formation of C⁺ tautomers^[21] (Figure 2A), accompanied by a decrease of approximately the same magnitude at 1710 cm⁻¹, which may be due to loss of the more strongly hydrogen, bonded G C6=_O in a Watson–Crick bp. Other signals for Hoogsteen bps include a decrease at approximately 1490 cm⁻¹ due to the formation of a new hydrogen bond to the G N7 (Figure 2B), increases at both approximately 1360 and approximately 1320 cm⁻¹ due to a syn G (Figure 2B), and an increase at approximately 1277 cm⁻¹ due to formation of a C⁺ (Figure 2C).^[15] Other signals are clearly observed in the HG-DNA:TBP IR spectrum (for example, the large intensity increase at approximately 1410 and approximately 1150 cm⁻¹) and are likely due to the approximately 80° bend induced in the DNA by TBP upon binding^[17a] (Figure 1B, see the Discussion in the Supporting Information).

Figure 2.

IR observation of a G–C⁺ Hoogsteen bp in the HG-DNA:TBP complex. A) The difference IR spectrum of a positive control for G–C⁺ Hoogsteen bps in a pH-induced G–C⁺ Hoogsteen DNA duplex, GC₉ (purple, Hoogsteen minus Watson–Crick duplex)^[15] is compared to the double-difference IR spectrum of HG-DNA when bound to TBP at pH 5.9 (red) and two negative controls for Hoogsteen, the AdMLP DNA:TBP double-difference IR spectrum (black), and the HGDNA:TBP complex double-difference IR at pH 7.78 (blue). The difference spectrum between replicates is shown to illustrate the signal to noise for the difference bands (noise, cyan). The mean of three replicate measurements is in bold and their standard deviation is in shade. An enhanced view showing an overlay of the 1730 cm⁻¹ band for all of the spectra is shown in the inset. B,C) show enhanced views of (A) to highlight the G–C⁺ IR marker bands. D) The chemical structure of an m¹G–C bp. The N1-methyl modification on the guanine (red) produces a steric collision with the C N3 and favors formation of the Hoogsteen m¹G–C⁺ bp. E,F) The difference IR spectrum of free m¹G-modified HG-DNA and free unmodified HG-DNA (purple), a positive control for Hoogsteen IR signals, is overlaid with the double-difference spectrum of the m¹G-modified HG-DNA bound to TBP (red) and the double-difference spectrum of HG-DNA bound to TBP (black). Dashed lines indicate the positions of Hoogsteen marker bands. Buffer conditions are similar to those used in the original crystallography study^[10] (40 mm MES pH 5.9 or 7.7, 60 mm KCl, 4 mm MgCl₂, 300 mm ammonium acetate, 2% v/v glycerol, 10 mm BME).

Slight changes in protein conformation between the free and bound forms may also contribute to the observed differences, which could overlap with the Hoogsteen marker bands; resulting in a false positive for Hoogsteen detection. If the appearance of the Hoogsteen marker bands are indeed due to the occurrence of G–C⁺ Hoogsteen and not other changes in the DNA or protein structure, we would predict the Hoogsteen bands to be diminished in the AdMLP DNA:TBP complex. This DNA sequence is not expected to contain any Hoogsteen bps,^[10] but is expected to form a complex that has a very similar structure to that adopted by the Hoogsteen-containing sequence. To test this prediction, the difference IR spectrum of the AdMLP DNA:TBP complex (Figure 2, black) was compared to that of the HG-DNA:TBP complex.

Indeed, the AdMLP DNA:TBP complex does not contain a signal at + 1730 cm⁻¹ (Hoogsteen G C6=O, Figure 2A). The signal at −1490 cm⁻¹ (G N7 hydrogen bond) is also very weak compared to the strong signal observed in the HG-DNA:TBP complex difference spectrum (Figure 2B), as is the signal at + 1277 cm⁻¹ (Hoogsteen C⁺, Figure 2C). The only difference between the AdMLP DNA and the HG-DNA is a single G–C bp (Figure 1E), and thus the lack of these signals in the bound AdMLP DNA supports the conclusion that these signals are indicators of G–C⁺ Hoogsteen formation. Meanwhile, a comparison of the bound AdMLP and HG-DNA IR difference spectra show several almost identical signals in the protein amide I and II region (1650–1550 cm⁻¹) that can be attributed to alterations between the free and bound protein, and differences in the phosphate-stretching region and DNA backbone (centered around approximately 1220 and approximately 1080 cm⁻¹, respectively) that can be attributed to DNA bending and unwinding (Supporting Information, Figure S5). Interestingly, the bands at + 1320 and + 1360 cm⁻¹, attributed to a syn purine, are also present in the AdMLP DNA:TBP complex. These vibrations are localized to the C1′−N9−C8−N7 bonds of purines.^[15] Their presence in the AdMLP DNA complex indicates that increases in these bands can likely arise from DNA bending and subsequent twisting of the purine glyosidic bonds, in addition to formation of syn purines. These results indicate that the DNA in both sequences is globally distorted in a similar fashion, supporting the earlier study.^[10]

G–C⁺ Hoogsteen formation can be suppressed by raising the pH,^[15,22] as protonation of C is required for formation of a hydrogen bond to the G N7. If the + 1730, −1490, and + 1277 cm⁻¹ signals are due to the formation −of C⁺ and a hydrogen bond to N7 in the HG-DNA:TBP complex, exchange into a high pH buffer should diminish these signals. The HG-DNA:TBP IR difference spectra in the crystallography buffer at pH 5.9 (Figure 2, red) and pH 7.78 (Figure 2, blue) show that the + 1730, −1490, and + 1277 cm⁻¹ signals are significantly weakened in the high pH buffer. Meanwhile, signals indicating a syn G (+ 1320 and + 1360 cm⁻¹, Figure 2B, blue) are still present at pH 7.78, as might be expected as the syn base conformation is sterically enforced by the protein. This supports the conclusion that these IR signals constitute evidence for the formation of a G–C⁺ bp within the solution state HG-DNA:TBP complex.

As a positive control for G–C + Hoogsteen formation, we site-specifically incorporated an N1-methylguanine (m¹G, Figure 2D) base into HG-DNA at G20. This chemical modification sterically enforces an m¹G–C⁺ Hoogsteen bp,^[2c,3b,22] and is a naturally occurring form of DNA damage.^[23] We first showed that TBP bound m¹G20-modified HG-DNA by FA (Supporting Information, Figure S6). Subtraction of the IR spectrum of the free m¹G duplex from that of the free HG-DNA duplex produces all of the Hoogsteen marker bands, indicating it is forming a Hoogsteen bp (Figure 2E, purple). When bound to TBP, the m¹G-modified HG-DNA recapitulates the IR signals observed in the unmodified HG-DNA:TBP complex (Figure 2E,D), indicating that the Hoogsteen bp is formed in both.

Here, we have provided the first solution-state observation of a Hoogsteen G–C⁺ bp in a DNA:protein complex using IR spectroscopy. We show that the IR marker bands for HG G–C⁺ pairs are robust and can be observed even in the complicated difference spectrum produced when subtracting free DNA from protein-bound DNA, and are distinct from signals that arise due to DNA bending. This approach can be immediately deployed to help resolve the nature of G–C base pairing in DNA:protein crystal structures in cases where the DNA electron density is ambiguous. This work establishes the existence of Hoogsteen G–C⁺ bps in DNA:protein complexes in solution and supports the conclusion that Hoogsteen bps are exploited by Nature to serve functions inaccessible to the textbook Watson–Crick bps. As Hoogsteen bps can provide a new information layer to the genetic code, further examination into their presence in physiological contexts is warranted.