An FTIR Investigation of Flanking Sequence Effects on the Structure and Flexibility of DNA Binding Sites

Abstract

Fourier transform infrared (FTIR) spectroscopy and a library of FTIR marker bands have been used to examine the structure and relative flexibilities conferred by different flanking sequences on the EcoRI binding site. This approach allowed us to examine unique peaks and subtle changes in the spectra of d(AAAGAATTCTTT)₂, d(TTCGAATTCGAA)₂, and d(CGCGAATTCGCG)₂, and thereby identify local changes in base-pairing, base-stacking, backbone conformation, glycosidic bond rotation and sugar puckering in the studied sequences. The changes in flanking sequences induce differences in the sugar puckers, glycosidic bond rotation and backbone conformations. Varying levels of local flexibility are observed within the sequences in agreement with previous biological activity assays. The results also provide supporting evidence for the presence of a splay in the G₄-C₉ base pair of the EcoRI binding site and a potential pocket of flexibility at the G₄ cleavage site that have been proposed in the literature. In sum, we have demonstrated that FTIR is a powerful methodology for studying the effect of flanking sequences on DNA structure and flexibility, for it can provide information about the local structure of the nucleic acid and the overall relative flexibilities conferred by different flanking sequences.

During both replication and transcription, amino acids of the protein factor involved recognize and interact electrostatically with the base pairs of the specific DNA sequence containing the recognition site (direct readout) with remarkable specificity. Upon binding of a protein in this region, a bending of the DNA occurs which enhances the stability of the complex by increasing the number of favorable contacts made between the DNA and the protein ^{1; 2}. The four classifications of these interactions are: (i) van der Waals interactions between aliphatic amino acids and the C5-methyl groups of thymines or the C5-hydrogen groups of cytosines, (ii) hydrogen bonds between the nucleic acids and the amide and carbonyl groups of the peptide backbone, (iii) hydrogen bonding bridges constructed with water molecules between amino acids and base-pairs ^{3; 4; 5}, and (iv) interlacing of amino acid side-chains between two bases ⁶. Although these are the major interactions, direct contacts at the recognition site alone do not account for the conformation of the DNA-protein complex; base pairs on either side of the recognition site, referred to as flanking sequences, are also thought to play a role in the conformation of the DNA-protein complex (indirect readout) ^{1; 7; 8}.

In most cases, it is necessary for the protein to distort the DNA in order to achieve the transition state. To compensate for the energetic costs of deformation, the DNA-protein complex may make any modification that improves the binding free energy such as releasing water from non-polar surfaces and/or increasing the number of direct contacts made between the DNA and protein ⁹. DNA that requires less deformation to reach the conformation of the transition state will have a lower free energy of association ¹⁰. Similarly, sequences that have an inherently greater distortability will also have a lower binding free energy. Studies examining the effect of flanking sequences on DNA-protein binding have indicated that flanking sequences are involved in dynamics and confer flexibility or rigidity to a sequence ^{1; 7; 8; 11; 12; 13; 14}. Since flexible DNA has a greater distortability, it is able to compensate more effectively and thereby more efficiently recognize and bind proteins and drugs than a similar, but more rigid DNA sequence ^{15; 16}.

Various methods have been commonly used to examine the effect of flanking sequences on DNA including conformational free energy calculations ⁸, ¹H-NMR and ³¹P-NMR ^{11; 17; 18}, crystal structures ^{3,Shui, 1998 #36}, restriction enzyme digests ^{12; 13; 19}, molecular dynamics simulations ²⁰, UV melting ^{14; 19}, sequence dependent anisotropic bend-ability (SDAB) ²¹, and CD ^{14; 19; 22}. These techniques have revealed valuable information about the features in DNA that change when flanking sequences are modified including changes in the unwinding and overwinding of the double helix, formation of hydrogen bonds, twist and glycosidic angles, helical parameters, flexibility, and restriction endonuclease cutting rates ^{3; 11; 12; 20}. Although these methods are highly advanced, there is some disparity with regards to which sequences are most flexible and most rigid ^{11; 12; 21}.

In principle, Fourier transform infrared (FTIR) spectroscopy can detect local conformational changes with high sensitivity ^{16; 23; 24}. For example, FTIR has been used to study DNA structural changes caused by CpG methylation ¹⁶ and the structural effects of G-tract length ²². Since this technique is non-destructive and provides a “snapshot” of the nucleic acid's structure, we have chosen to apply FTIR along with a library of IR bands compiled by Banyay et al. ²³ to interpret major features and changes in nucleic acid structure, conformations and conformational sub-states as a result of different flanking sequences.

MATERIALS AND METHODS

Sample Preparation

The DNA dodecamers studied were obtained from Alpha-DNA (Quebec, Canada). Samples were desalted by centrifugation in Y300 Centricons (Millipore). After desalting, samples were lyophilized overnight and stored at 4°C until analysis.

Fourier Transform Infrared (FTIR) Experiments

Samples for FTIR analysis were prepared by diluting to a final concentration of 3.0 mM in 100 mM sodium phosphate buffer (pH 7.4). Freshly made samples were used for all FTIR measurements. FTIR spectra were recorded at room temperature on a ZnSe crystal using a Thermo Nicolet Avatar 370 ATR-FTIR spectrometer (Thermo Scientific, Massachusetts). A background (atmospheric plus pure buffer) was collected after which DNA samples were run. For each DNA sample, FTIR spectra were taken in triplicate to ensure reproducibility. Experiments were run with a resolution of 4 cm^-1 and 512 scans. Using EZ-OMNIC software (Thermo Scientific), all DNA spectra underwent baseline correction and 15 point smoothing before peaks were assigned. Derivative spectra were 25 point smoothed. Final spectra were normalized with respect to the symmetric PO₂^- stretching mode appearing at 1090-1080 cm^{-1 25; 26}. Spectral assignments were made using the Banyay et al library²³. This library provides wavenumber ranges of ~5-10 cm^-1 for many assignments so differences less than 10 cm^-1 may not be significant.

³¹P NMR experiments

DNA samples were obtained from Alpha DNA (Montreal, Canada). These were salted with 100 mM phosphate buffer at pH = 7.4 then lyophilized. The samples were re-hydrated in 98% D₂O, re-lyophilized and finally dissolved in 99.9% D₂O, immediately before use.

2D ¹H nuclear Overhauser effect (NOESY) experiments were performed with mixing times of 500 ms and a gradient pulse applied during the mixing time. 4096 data points were collected in the t₂ dimension and 512 increments in t₁. The matrix was zero-filled to 4096 × 1024 points before double Fourier transformation. Phase sensitive ³¹P-¹H HSQC were recorded using Echo/Antiecho-TPPI gradient selection^{27; 28; 29} and decoupling during acquisition. All HSQC were performed with 2048 points in the t₂ dimension and 256 t₁ increments using a relaxation delay of 4 s. The chemical shift of H₂O and the center of the DNA phosphorus resonances were set as the center of HSQC and NOESY spectra. A spectral width of 9 ppm was used for all 1D experiments and in both dimensions of the 2D experiments.

Since H2' and H2" are strongly coupled with H1', H3', H5' (sugar protons) and to the base protons, enabling the assignment of protons on the NOESY using a sequential walk³⁰. The strong correlations between ³¹P and H3' and H5'/H4' then enabled us to assign the HSQC using the proton assignments determined from the NOESY.

%BII was determined as a function of the ³¹P chemical shift using a linear relationship between the observed fast-exchange chemical shift and the %BII conformation ³¹:

$$\%B_{II}=\frac{1}{2.368-0.005110T}\delta P\left(ppm\right)+\frac{1.345-0.002031T}{2.368-0.005110T}$$

RESULTS

We studied DNA sequences containing the EcoRI binding site as this binding site has been extensively studied by a number of techniques. In particular, we used the well-characterized Dickerson Sequence³² as a control so we can compare our results with the wealth of structural and dynamical data available in the literature on this sequence. Flanking sequences were chosen based on previous research on the effect of flanking sequences on DNA structure, dynamics and protein-DNA interactions.^{11; 12; 18; 26} Table 1 lists the EcoRI sequences that were examined using FTIR. The spectral peaks were assigned using the Banyay Library of DNA infrared bands ²³.

Table 1.

Sequences of DNA samples studied. The bold lettering indicates the EcoRI binding site.

DNA Sample	Sequence
EcoRI AAA/TTT	d(AAAGAATTCTTT)2
EcoRI TTC/GAA	d(TTCGAATTCGAA)2
EcoRI CGC/GCG	d(CGCGAATTCGCG)2

Effect of Flanking Sequences on the Base Region (1800-1500 cm^-1)

Figure 1 shows the 1800-1500 cm^-1 region for EcoRI binding site containing sequences. This region reports on changes in the bases of nucleic acids.²³ The vibrations of base-paired pyrimidines occur between 1670-1630 cm^-1. The EcoRI TTC/GAA sequence has a broad range of peaks within this region including distinct peaks at 1666 and 1639 cm^-1. The peak at 1666 cm^-1 falls into the range for single stranded guanines (1673-1660 cm^-1). EcoRI AAA/TTT has similar breadth in this region but differing intensities. EcoRI CGC/GCG has a broad peak with a clear resonance at 1645 cm^-1. The strong resonance at 1645 cm^-1 is indicative of single stranded cytosines. The breadth and lack of resolution in this region for the EcoRI CGC/GCG sequence may be the result of the large number of CG base pairs in this sequence. A peak at between 1604-1600 cm^-1 was present in all samples (although only weakly in EcoRI CGC/GCG) and is also a marker for single stranded guanines. The vibrations of base-paired cytosines occur at 1525 cm^-1 in EcoRI TTC/GAA and shifts upfield to 1535 cm^-1 in EcoRI AAA/TTT. There is a low intensity, broad resonance in this region for EcoRI CGC, again likely due to the increased number of CG pairs. An intense peak at 1710 cm^-1 is evident EcoRI AAA/TTT and EcoRI TTC/GAA, but only weakly present in EcoRI CGC/GCG. The Banyay Library attributes this peak to guanines involved in Hoogsteen binding to a third strand, but this is unlikely in the short DNA sequences at low concentrations studied here. Therefore, these peaks are more likely the carbonyl stretches of bases in A-form ³³.

Figure 1.

Spectra for the 1800-1500 cm^-1 in-plane base vibrations; EcoRI AAA/TTT (top), EcoRI TTC/GAA (middle), and EcoRI CGC/GCG (bottom).

Effect of Flanking Sequences on the Base-sugar Region (1500-1250 cm^-1)

The spectrum for the 1500-1250 cm^-1 region demonstrating the effect of flanking sequence changes on the glycosidic bond rotation, the backbone conformation and the sugar puckering modes of nucleic acids is illustrated in Figure 2. In EcoRI CGC/GCG, sugar ring vibrations of purines occur as a weak, broad peak centered at 1490 cm^-1. This becomes two peaks at 1493 and 1486 cm^-1 in EcoRI TTC/GAA and weakly in EcoRI AAA/TTT. Changes in the position of this band reflect interactions at N7 sites of purines ²³. Similarly, the broad peak due to the vibrations of adenines in A-/B-form conformation at 1458 cm^-1 in EcoRI CGC/GCG, splits into two distinct peaks at 1465 and 1458 cm^-1 in EcoRI AAA/TTT. This peak shifts to 1442 cm^-1 in EcoRI TTC/GAA. In EcoRI AAA/TTT there is a peak at 1420 cm^-1, which corresponds to S-type sugars in B-form helices; this marker is weakly present in EcoRI TTC/GAA and EcoRI CGC/GCG. A broad peak encompassing the Z-form marker at 1434-1440 cm^-1 is present in EcoRI TTC/GAA. Peaks at 1373 and 1365 cm^-1 for purines in anti form are present in EcoRI AAA/TTT and a broad resonance encompassing this region is present in EcoRI TTC/GAA. A weak, broad resonance at 1371 cm^-1 is present in EcoRI CGC/GCG. This shift suggests modified glycosidic torsion angles ¹⁶. In EcoRI AAA/TTT and EcoRI TTC/GAA, a marker for thymines with S-type sugars and a glycosyl torsional angle χ in anti conformation is evident at 1328 and 1323 cm^-1, respectively, but absent in EcoRI CGC/GCG. A marker corresponding to C4NH2 vibrations of a cytosine occurs in all samples between 1297-1287 cm^-1. The CN3H bends of thymines with S-type sugars and anti conformations occurs in EcoRI AAA/TTT at 1282 cm^-1; however, this marker appears only very weakly in EcoRI TTC/GAA and not at all in EcoRI CGC/GCG.

Figure 2.

Spectra for the 1500-1250 cm^-1 base-sugar vibrations; EcoRI AAA/TTT (top), EcoRI TTC/GAA (middle) and EcoRI CGC/GCG (bottom).

Effect of Flanking Sequences on the sugar-phosphate Region (1250-1000 cm^-1)

Figure 3 illustrates marker bands between 1250-1000 cm^-1 that identify changes in the backbone conformation. All EcoRI sequences exhibited a marker at 1216 cm^-1, which corresponds to B-form DNA. All three samples also demonstrate symmetric PO^2- vibrations that are insensitive to the B-to-A transition between 1082-1076 cm^-1. The shoulder to this symmetric PO^2- peak at 1051 cm^-1 is evident in EcoRI AAA/TTT and EcoRI TTC/GAA but is subsumed into a broad resonance in EcoRI CGC/GCG. This vibration is attributed to the CO stretch of the backbone and is strongly enhanced in Z-form DNA.^{25; 26} In EcoRI TTC/GAA, a marker for furanose vibrations in Z-DNA is evident at 1025 cm^-1. Finally, all three sequences have a weak marker at 1172 cm^-1, an A-form marker due to C3' endo/anti puckering.

Figure 3.

Spectra for the 1250-1000 cm^-1 sugar-phosphate vibrations; EcoRI AAA/TTT (top), EcoRI TTC/GAA (middle) and EcoRI CGC/GCG (bottom).

Effect of Flanking Sequences on the Sugar Region (1000-800 cm^-1)

The effects of sugar conformation on EcoRI spectra provided in the 1000-800 cm^-1 region are depicted in Figure 4. Careful examination of this region, particularly between 900-800 cm^-1, was required because important structural differences often manifested as minor spectral changes.

Figure 4.

Spectra for the 1000-800 cm^-1 sugar vibrations; EcoRI AAA/TTT (top), EcoRI TTC/GAA (middle), and EcoRI CGC/GCG (bottom).

To aid in characterizing peaks contributing to the broad resonances observed, second derivative spectra for this region are provided in Figure 5 ¹⁶. A significant peak at 970 cm^-1 correlating to the CC stretch of the B-form backbones is present in EcoRI AAA/TTT and EcoRI TTC/GAA. EcoRI CGC/GCG and to a much less intensity, EcoRI AAA/TTT demonstrate a resonance at 993 cm^-1. The derivative spectra show that EcoRI AAA/TTT has breadth but all of its intensity is localized about the 970 cm^-1 peak. EcoRI TTC/GAA has a peak at 995 cm^-1 and a broader one at 969 cm^-1 with additional resolution.

Figure 5.

The second derivative spectra of the FTIR region reporting on the backbone and sugar conformations. EcoRI AAA/TTT (top), EcoRI TTC/GAA (middle) and EcoRI CGC/GCG (bottom).

EcoRI CGC/GCG has a narrow peak at 995 cm^-1 and a broad peak at 985 cm^-1. This suggests a large variety of backbone conformations in these samples and some dynamic averaging between them. The backbone dynamics of the EcoRI CGC/GCG sequence have been previously studied and demonstrate facile interconversion of the BI and BII states in the C3pG4 and C9pG10 steps.^{31; 34} Figure 6 shows the %BII as a function of backbone step for each sequence as determined by ³¹P NMR. This plot demonstrates that the EcoRI AAA/TTT samples mainly BI backbone conformations across the entire sequence whereas EcoRI TTC/GAA shows steps with dramatic %BII character. This is in agreement with the FTIR data showing a broad resonance at 970 cm^-1 for EcoRI AAA/TTT and resonances at varying backbone conformations for EcoRI TTC/GAA. The EcoRI CGC/GCG sequence also shows variation in backbone conformation, again agreeing with the variety and breadth of backbone resonances apparent in its FTIR spectrum.

Figure 6.

The variation in backbone conformation (reported as %BII) as a function of nucleotide step for EcoRI AAA/TTT (black), EcoRI TTC/GAA (red) and EcoRI CGC/GCG (black).

The marker for adenine-thymine base pairs in B-form helices occurs in all samples at 928-931 cm^-1 (Figure 4). In addition, the deoxyribose peak at 898-893 cm^-1 is evident in all samples but broadest in the EcoRI CGC/GCG sequence. Broad peaks centered between 839-835 cm^-1, arising from S-type sugar puckers, are evident in three samples. The derivative spectra (Figure 5) show an almost equal population into two S-types markers at 829 and 835 cm^-1 as well as peaks at 815 and 804 cm^-1, indicative of N-type sugars in EcoRI CGC/GCG. Again, these results demonstrating varying furanaose structure and the presence of furnanose dynamics are consistent with previous studies of the furanose structure and dynamics in the EcoRI CGC/GCG sequence. ^{34; 35; 36} The EcoAAA/TTT and EcoRI TTC/GAA sequences demonstrate peaks at 846, 836 and 804 cm^-1 and an O4' endo sugar puckering marker at 825 cm^-1.¹⁶

DISCUSSION

Our findings demonstrate that one can use FTIR and the library of infrared marker bands compiled by Banyay et al.²³ to determine the structural and dynamical changes conferred by different flanking sequences. This approach allowed us to examine unique peaks and subtle changes in the spectra of EcoRI binding sites that have been flanked by various tri- nucleotides, and thereby identify local changes in base-pairing, base-stacking, backbone conformation, glycosidic bond rotation and sugar puckering in the studied sequences. We included the analysis of the well-studied Dickerson dodecamer in this investigation so that structural features observed via FTIR could be directly compared with previously reported NMR, X-Ray and Molecular Dynamics investigations of this sequence.

Variations in the nucleotide sequence produce different vibrations within both the direct and indirect readout regions. Based on the peak assignments in the 1800-1500 cm^-1 region, most of the nucleotides in the EcoRI sequences are double stranded. However, in all dodecamers studied peaks between 1673-1660 and 1608-1600 cm^-1, corresponding to a single stranded guanine nucleotide, are observed. These peaks are even evident in the EcoRI AAA/TTT sample that contains only one guanine nucleotide adjacent to the EcoRI restriction endonuclease cutting site (AAAG↑AATTCTTT). This finding supports the presence of a splay in the G₄-C₉ base pair in the EcoRI cutting site, which was described by Drew et al. and seen in NMR and Molecular Dynamics studies ^{32; 35; 37} in the crystal structure of the B-form helix. We also observed a number of changes in base-stacking that are attributed to the identities and positions of the bases in the flanking sequences of these dodecamers. Moreover, they can be credited to vibrations from the flanking sequences affecting the nucleotides within the GAATTC binding site. For example, in EcoRI TTC/GAA, the vibrations of base-paired cytosines occur at 1525 cm^-1 and shift downfield to 1535 cm^-1 in EcoRI AAA/TTT, a change that may be attributed to a decrease in the percentage of cytosines in the dodecamer. This region does show significant variations in the base stacking patterns of samples with different flanking sequences. This variation in base stacking as a function of flanking sequence is in agreement with the FTIR results of Lindqvist and Graslund.²². It has been previously demonstrated that changes in base stacking effects are directly related to backbone conformations.^{38; 39; 40} The current FTIR results of the backbone resonance breadth and ³¹P NMR chemical shift data are thus in keeping with the variety of base stacking resonances we observe.

Changes in the endocyclic torsion angles of the furnaose ring are evident by studying the 1000-800 cm^-1 region of the FTIR spectra. Our results show that all sequences contain both N-type and S-type sugars and that the S-type sugars are well resolved into C2' endo and O4' endo peaks. The EcoRI CGC/GCG sequence has broad sugar markers that demonstrate a variety of conformations in the derivative spectrum. Such furanose dynamics have been directly observed in the EcoRI CGC/GCG sample using solid state ²H NMR.³⁴ All sequences have the B-form backbone marker at 970 cm^-1 with the marker in the EcoRI TTC/GAA and EcoRI CGC/GCG sequences being broad and possessing a backbone resonance at 995 cm^-1. The derivative spectra shown in Figure 5 demonstrate various backbone states in these samples. We attribute this to the presence of the BI and BII conformational substates of the backbone ^{41; 42}, The breadth of the peak in the EcoRI CGC/GCG sequence is in agreement with recent ³¹P dynamic NMR results in the EcoRI CGC/GCG sequence that demonstrate facile backbone dynamics at the C3G4 and C9G10 backbone steps.³¹

The region between 1250-1000 cm^-1 is sensitive to vibrations along the sugar-phosphate backbone, and these marker bands provide information about the overall conformation of the double helix ²³. With the vibrations we observed in this region, we have been able to conclude that all EcoRI dodecamers exist predominantly in the B-form conformation as they exhibit peaks between 1224-1221 cm^-1. Notably, these peaks are broad, again suggesting the sampling of different backbone conformations. Despite the fact that the sequences all favor the B-form as their primary backbone conformation, the EcoRI AAA/TTT and TTC/AAG sequences exhibit B-, A-, and Z-form conformation markers in the 1500-1250 cm^-1 region as well as N-type sugar puckering, in the case of the EcoRI AAA/TTT sequence. A mixture of these markers can be attributed to fraying of the helix near the ends of the sequences, since the two other regions of interest of the spectra (1250-1000 and 1000-800 cm^-1) also exhibit a dominant B-form. However, it is more likely that the flanking sequences of these dodecamers confer rigidity or flexibility to these sequences. For example, the Z-type marker at 1434-1440 cm^-1 is present in EcoRI TTC. The N-type sugar marker at 1406 cm^-1 in EcoRI AAA/TTT that became much weaker in EcoRI TTC/GAA and completely disappeared in EcoRI CGC/GCG may indicate that EcoRI AAA/TTT and EcoRI TTC/GAA tend to exhibit more of the rigid A-form and Z-form conformations whereas EcoRI CGC/GCG primarily favors the flexible B-form. The marker for purines in the anti conformation appears at 1373 and 1365 cm^-1. The strong presence of this peak in EcoRI AAA/TTT is likely due to the A-tract flanking sequence. The EcoRI TTC/GAA sequence has a broad range of peaks in this region encompassing the syn marker at 1323 cm^-1. The EcoRI CGC/GCG sequence shows far less resolution in this region, again suggesting a more dynamic sugar ring.

Our findings suggest that the EcoRI AAA/TTT and EcoRI TTC/GAA sequences are more similar to each other than to the EcoRI CGC/GCG dodecamer in the 1800-1500 cm^-1 and 1500-1250 cm^-1 base-reporting regions. The EcoRI CGC/GCG spectrum is much smoother with broader peaks and more poorly defined marker bands than the other sequences. This evidence, along with the observation that local nucleotides in the rigid A-form and Z-forms occur more frequently within the EcoRI AAA/TTT and EcoRI TTC/GAA sequences, suggests that EcoRI CGC/GCG is the most flexible sequence.

In fact, the FTIR regions reporting on the backbone and sugar moieties all show increased breadth for the EcoRI CGC/GCG sequence. This is in keeping with the previous ³¹P³¹ and ²HNMR³⁴ work on EcoRI CGC/GCG and the work of Windolph et al who demonstrated increased cleavage rates of the K130E mutant of EcoRI on sequences with `favorable' flanking sequences <sup>12</sup>. The co-crystal structure of EcoRI bound the EcoRI CGC/GCG sequence shows that there are no direct base contacts to the flanking sequences <sup>43</sup> so these sequences must instead modulate the DNA structure or flexibility. The Windolph study used known algorithms <sup>44; 45</sup> to study this modulation and found that there was no real difference in the stacking energies of their starting pool and their `best cut' or `worst cut' pools, indicating that stacking energies are not responsible for the changes in cutting rates. However, they did demonstrate increased flexibility of their best cut pool using the flexibility algorithms of Satchwell and Travers ⁴⁶. We have shown this flexibility in the EcoRI CGC sequence using ³¹P³¹ and ²H NMR.³⁴ Here, we directly demonstrate this increased flexibility as a function of flanking sequence using FTIR.

We have been able to conclude this using FTIR as this methodology provides a “snapshot” of the nucleic acid sequences where sharp resonances are the result of defined conformations while broad ones suggest dynamical averaging on the FTIR timescale. That flanking sequences are able to affect a sequence's flexibility is in keeping with those of Lefebvre et al.¹⁸ who used NMR to study the effect of modified flanking sequences on a CpG step and found that overall d(CATCGATG)₂ sequence was more flexible than the d(CTTCGAAG)₂ octamer. Similarly, Cordier et al.,¹¹ who used ¹H-NMR and ³¹P-NMR, found that A and T tracts of three base pairs on either side of the CRE sequence, d(GAAAACGTTTTC)₂, induced rigidity in this sequence as compared to flexibility observed in d(ATGACGTCAT)₂. This confirmation of differing flexibility due to flanking sequence has important biological ramifications. It is known from the cocrystal structure of the EcoRI restriction endonucelase in complex with its recognition sequence that this protein-DNA interaction utilizes an induced fit mechanism ⁴⁷. The neokinks, distortions in the DNA structure upon complexation, incur an energetic cost and this cost is presumed to vary based on surrounding sequences ⁹. Thus, the changes in cleavage rates observed by Windolph were suggested to be caused by altered DNA flexibility caused by flanking sequence effects. Here, we have provided direct experimental evidence supporting that hypothesis.

In conclusion, we have shown variations in FTIR spectra that efficiently provide preliminary information about the specific local structural response of the EcoRI binding sites to changes in the tri- nucleotides flanking these sites. In particular we found that different flanking sequences can (1) promote A- or Z- type structures in primarily B-form DNA sequences, (2) alter backbone and sugar conformations and (3) promote dynamical averaging of backbone and furanose conformations. While these results are consistent with previous research ^{11; 18}, we recognize that this is an underdetermined problem and we cannot confidently identify the source of IR band shifts. We plan to continue this investigation using our recent ³¹P dynamic NMR method ³¹ to provide the quantitative dynamic data FTIR cannot yield. However, the application of this method to the well-studied EcoRI CGC/GCG sequence provides confirmation for the usefulness of FTIR in demonstrating the flexibility of the DNA backbone. In addition, by using the EcoRI CGC/GCG sequence as a control, we have demonstrated that our FTIR results correspond to the structural features determined using other techniques.^{31; 32; 34; 35; 48} Thus, we have shown that FTIR provides a rapid and efficient method for examining a large number of DNA sequences in order to uncover trends and to determine which samples are worthy of further investigation by more quantitative techniques.

The obvious changes in FTIR spectra caused by altering flanking sequences confirm that flanking sequences have an important role in local structure and flexibility in DNA sequences. This is of tremendous importance because it suggests that the local environment of the DNA binding site determines its binding affinity. Understanding this phenomenon underscores the important role of indirect readout in protein-DNA recognition events. ^{15; 16; 42}