|Home | About | Journals | Submit | Contact Us | Français|
Fluorescent reporter groups have served for many years as sensitive probes of macromolecular structure. Such probes can be especially useful in comparative studies such as detection of conformational changes and discrimination among structural models. Spectroscopic methods such as fluorescence are attractive because they are rapid, require small amounts of material, are non-destructive, can be carried out with commonly available equipment, and are relatively inexpensive. In addition, there is a rich library of theoretical and practical materials available to aid in data interpretation. The intrinsic fluorescence of most nucleic acids is too low to be useful in structural studies. Thus, it is necessary to incorporate a suitable reporter group to utilize fluorescence methods involving polynucleotide structure. A highly fluorescent adenine analog, 2-aminopurine, has long served in this capacity. The present article describes our use of 2-aminopurine as a probe of loop structures in quadruplex DNA. In particular, we show how knowledge of the relative intensity of 2-aminopurine emission as well as its sensitivity to exogenous quenching molecules such as acrylamide can aid in comparing crystal and solution structures of an oligonucleotide model of the human telomere and in discrimination among models containing tandem repeats of the telomeric quadruplex.
Polynucleotides containing strings of contiguous guanine residues can associate to form quadruple stranded structures referred to as quadruplexes. There are many recent reviews of DNA quadruplexes that one may consult for details (e.g. (2–7)). For convenience, we provide a brief discussion of the structure of DNA quadruplexes below. These macromolecular assemblies consist of helical arrays of stacked planer, cyclic structures of four guanines linked by 8 hydrogen bonds involving the N1, N2, N7 and O6 atoms of the bases within an individual G-quartet. The O6 atoms within a quartet are directed toward the central cavity of the quadruplex and thereby form a coordination site that can accommodate a cation such as Na+ or K+ (Fig. 1).
Conformational diversity is one of the remarkable characteristics of quadruplexes. This plasticity results from a variety of chemical and physical variables. For example, the number of G-quartets comprising a stack may vary, the individual strands may be parallel, antiparallel or of mixed polarity, the loops connecting the quartets may be diagonal or lateral, the sequences flanking the stacks may differ, the N-glycoside bonds of the G residues may adopt syn or anti conformations, and the stabilizing cation may vary. The topological diversity of monomolecular quadruplex structures with three G-stacks is illustrated schematically in Fig. 2.
Although the structure of G-quartets was first described in 1962 (9), interest in quadruplex forming oligonucleotides has accelerated over the last twenty years. This may be attributed to the observation in 1989 (10) that telomeres (which provide a mechanism for complete chromosomal replication and protect chromosomal ends from degradation) consist of repeating tracts of G-rich sequences such as d(TTTTGGGG) in the protozoan Oxytricha nova and d(TTAGGG) in humans. In human chromosomal DNA, 200–300 bp of these sequences protrude as a single-stranded overhang from the 3′ end and studies with model oligonucletoides suggest that these sequences form a series of tandem quadruplexes (11). Subsequent investigators have shown that telomeres in vivo contain quadruplexes, and that quadruplex conformational changes may represent a key process in controlling chromosomal replication (12,13). T h e relatively facile ability to respond to environmental changes by undergoing conformational transformations (which stands in remarkable contrast to their high degree of thermal stability) suggests that quadruplex motifs may possess regulatory properties in other fundamental biological contexts, for example in cellular senescence and oncogenesis (14,15).
Bioinformatic analysis has revealed the presence of G-rich segments in parts of eukaryotic genomes in addition to those described above (16); several of these segments have been shown to be involved in controlling expression of a number of genes, including a variety of oncogenes as well as sequences involved in immunoglobulin switching (17). In addition, based on their putative regulatory roles, quadruplex-containing DNA segments are rational targets for drug development (7,18,19). Development of drugs targeted to specific quadruplex sequences requires an appreciation of subtle conformational differences likely to exist among different quadruplexes.
Interest in quadruplex oligonucleotides is also increasing in biotechnology. Some applications of quadruplex structures include molecular recognition (aptamers (20)), analytical applications (molecular sensors (21)), materials science (nanostructures), engineering (nanomachines), (22) and as drugs (19). Full development of the potential of each of these aspects of quadruplex science depends on understanding the conformational properties of quadruplexes and quadruplex assemblies, as well as their ability to undergo controlled conformational transitions.
X-ray crystallography and NMR methods continue to provide structures of a variety of specific quadruplexes (reviewed in (3,6)). However, these high resolution methods have proven not to be universally applicable. For example, NMR yielded an ensemble of conformers structures of a model of the 22-nt human telomeric sequence d[AGGG(TTAGGG)3] (HuTel22) in the presence of Na+ (1). This structure consists of an anti-parallel “basket” structure characterized by two lateral and one diagonal dTTA loops (Fig. 2). However, it was not possible to obtain a solution structure for HuTel22 in the presence of K+ (the more physiologically relevant cation), presumably because of the presence of several conformational isomers. Interestingly, HuTel22 could be crystallized in the presence of K+ (23). The resulting structure exhibits a parallel-stranded topology with the three dTTA loops packed against the quadruplex core (the “propeller” topology). However, several studies suggest that the all-parallel structure found in the crystal may not be the predominant solution structure ((8) and references therein). In fact, NMR solution studies of related sequences in the presence of K+ suggest the possibility that so-called hybrid-1 or hybrid-2 (3 parallel strands + 1 antiparallel strand) may be the preferred solution structure in the presence of this cation (24–26)
These examples provide a concrete basis for the importance of applying a variety of lower resolution techniques to test structural hypotheses, detect conformational changes and to compare unknown structures with known structures. Of these methods, the use of site-specific spectroscopic probes is particularly useful because of their ability to report specific rather than global structural differences brought about by mutation or changes in solution conditions. In general, fluorescent probes can serve as versatile, site-specific conformational reporters.
In contrast to polypeptides, for which the intrinsic fluorescence of the indole and phenolic side chains provide sensitive fluorescent reporters, the fluorescence quantum yield of the purine and pyrimidine building blocks of polynucleotides is too low to be analytically useful (except for the relatively rare Y base of found in t-RNAPhe). Thus, it is necessary either to substitute a fluorescent purine or pyrimidine analog at the residue of interest or attach a probe molecule to the oligonucleotide. The adenine analog, 2-aminopurine (Fig. 3), has been applied in a large number of studies as a site-specific fluorescent reporter for polynucleotide studies. This fluorescent purine exhibits a number of desirable properties that render it a good probe molecule for polynucleotides. Among these are: (1) 2AP can form Watson-Crick H-bonds with thymidine, uracil or cytidine, thus inducing little or no structural perturbation when substituted for adenine; (2) 2AP can be relatively inexpensively incorporated into oligonucleotides using common solid phase synthetic methods; (3) the quantum yield of fluorescence is highly sensitive to environmental conditions and thus local changes in conformation; (4) theoretical and practical aspects of the photochemistry of 2AP, especially in the context of oligonucleotide structure, are well established. In particular, 2AP fluorescence tends to increase with solvent exposure and to decrease with base stacking. The interested reader is referred to two excellent recent studies focusing on the application of site-specific 2AP probes of RNA folding (27,28). These papers provide a comprehensive outline of the properties of 2AP in relation to oligonucleotide structure as well as a guide to recent studies involving 2AP-containing oligonucleotides.
In the following, we describe our use of 2AP as a probe of quadruplex folding. In these studies, we used the susceptibility of 2AP fluorescence to collisional quenching by acrylamide as a probe of adenine exposure present in loops or between tandem quadruplexes. Taken in conjunction with other studies, these fluorescence quenching experiments help distinguish between conformational models of a single quadruplex as well as tandem two-quadruplex models of human telomeric DNA.
Many excellent texts (for example (29) provide good summaries of biological applications of fluorescence. Fluorescence is defined as the emission of a photon from a singlet excited state to the ground state subsequent to absorption of a photon by a fluorophore. The fluorescence properties of a system are characterized by a number of parameters: excitation wavelength, emission wavelength, quantum yield (q) and lifetime of the excited state (τ0). The quantum yield for fluorescence is defined as the ratio of photons emitted by fluorescence to photons absorbed. Alternatively, q = kf/∑ki, where kf is the rate of fluorescence emission and ∑ki represents the sum of rate constants for all mechanisms of loss of excitation energy. This expression shows explicitly the link between quantum yield and fluorescence lifetime.
Direct determination of q is generally not convenient; however, it is often sufficient to determine the relative quantum yield F, which is the emission intensity of the system measured under defined conditions. F is a function of the instrument (intensity of excitation source, sensitivity of the detection system etc.) as well as the conditions of the measurement (fluorophore concentration, temperature and solution conditions). F is most accurately obtained from integration over the wavelengths of the emission envelope; however, it may be adequate to determine F as the emission intensity at the emission wavelength maximum. Depending on the instrument, F may be given in photons/sec or in arbitrary units.
The excited state lifetime is the average time that a fluorophore spends in the excited state before emitting a photon. For a homogeneous system, fluorescence decay consists of a single exponential process. In general, for fluorophores in a heterogeneous environment the decay of fluorescence consists of multiple exponentials as defined by the relationship F(t)/F0 = ∑αi·exp(−t/τ0,i) where αi is the amplitude of the fluorescence intensity for species i and τ0,i is the lifetime for that species. Fluorescence lifetimes are generally in the sub-nanosecond to nanosecond time range and require specialized instrumentation for determination.
For many fluorophores, the measurable characteristics of fluorescence such as emission maximum, q and τ0 are sensitive to the local environment of the fluorophore. However, for 2AP, it turns out that the emission maximum is not very sensitive to its environment, but q and τ0 are environmentally quite sensitive. Often the structural basis of these variations in τ0 and q can be quite subtle. For example, 2AP fluorescence is enhanced by cooling (indicating the importance of conformational motion), quenched by base stacking, enhanced by energy transfer from neighboring adenine, and strongly quenched by collision with neighboring G residues. In addition, 2AP emission is quenched by solutes such as acrylamide that quenches by collision with the excited state fluorophore. As described below, the action collisional quenchers is generally described by the Stern-Volmer equation or modifications of it.
Classical Stern-Volmer quenching (also referred to as dynamic quenching) can be explained by a simple mechanism in which the quenching agent, Q, collides with an excited fluorophore F*. In this type of quenching, the excited state energy is lost when the excited fluorophore collides with Q and forms a transient encounter complex as depicted in scheme 1:
Formation of (F*·····Q) depends on the diffusion rate constant for F* and Q, their concentrations and the efficiency of the quenching process. For a single fluorescent species in a homogeneous environment, the degree of quenching increases with increasing [Q] according to the Stern-Volmer relationship:
where F0 is the fluorescence observed in the absence of quencher, FQ is the fluorescence observed in the presence of quencher at concentration [Q], KSV is the Stern-Volmer quenching coefficient, τ0 is the lifetime of the excited state in the absence of Q, kqis the bimolecular quenching constant, fQ is the efficiency of quenching, and k0 is the (diffusion-controlled) bimolecular collisional rate constant. From Eq. 2, a plot of F0/FQ versus [Q] is linear with a slope KSV from which f0 can be extracted if τ0 and k0 are known. The former may be available in the literature or measured directly and the latter can be estimated from the Smoluchowski equation (Note 1). Under defined conditions for a given quenching agent and a homogeneous fluorescent system, f0 is related to the accessibility of F* to Q. In the absence of known values of τ0 and k0, the relative accessibility of F* to Q can qualitatively be assessed by comparing values of KSV measured for the same fluorophore measured under different solvent conditions. Hence, under ideal circumstances, collisional quenching can provide insight into the relative degree of exposure of a fluorophore to solvent.
Experience has shown that the Stern-Volmer plot often is not linear but exhibits either a concave or convex relationship with respect to [Q]. A non-linear Stern-Volmer plot indicates a heterogeneous fluorescing system. This heterogeneity may arise either from fluorophores with different k0, f0 and/or τ0 values or from formation of static complexes with Q. In systems with a single fluorophore (such as the quadruplexes under consideration in this article) downward curvature is interpreted to indicate that the fluorophore experiences more than one micro-environment. Such a heterogeneous quenching system can in general be described by Eq. 3:
where KSV,i is the quenching constant for the ith species and fiis the fractional contribution of the ith species to the total fluorescence. Explicitly for a two-component system:
Eftink (30) has shown that if KSV,1 ≥ 4·KSV,2, the F0/FQ plot will exhibit visible downward curvature. Fitting the data to Eq. 4 by non-linear least squares allows determination of the optimal values of the parameters f1, KSV,1 and KSV,2 (although see Note 2).
Upward curvature is generally interpreted to indicate formation of both static and dynamic (collisional) complexes (Note 3). This case will not be treated here; the interested reader is referred to a standard text book for details regarding interpretation and quantitative analysis of upwardly curving Stern-Volmer plots.
The fluorescence properties of 2AP in water are as follows: emission involves a π*→ π transition, with an excitation maximum of ~305 nm and emission maximum at ~370 nm, q = 0.7; fluorescence decay is mono-exponential with τ0 = 12 ns (31). The emission maximum is relatively insensitive to solvent polarity and q is not affected by pH but is strongly dependent on temperature. In the context of an oligonucleotide, q is strongly influenced by nearest neighbors and by base stacking. Quenching of 2AP fluorescence by acrylamide gives a Stern-Volmer plot with slight upward curvature which has been attributed to a weak static quenching component at high [acrylamide]. KSV for acrylamide quenching of 2AP is ~28 M−1 at 10 °C and increases to ~58 M−1 at 70 °C.
Here we provide two examples of 2AP-acryamide quenching data used by our group to assess the degree of exposure of individual dA residues in two models of the human telomeric DNA sequence. In the first study, Li et al. (8) compared the acrylamide quenching curves of HuTel22 in sodium and potassium buffers. The purpose of this study was to compare the exposure of the loop adenines in solution with the predicted exposures assessed from the solution structure of HuTel22 in Na+ (antiparallel basket) determined by NMR and the structure of the same oligonucleotide determine in K+ by x-ray crystallography (all parallel propeller structure). The two topographical arrangements have very different loop geometries and hence different environments for the loop adenines. In the basket structure, A1, A13 and A19 are stacked upon G-tetrads; A7 is unstacked. This is consistent with the apparent quantum yields (F0 values) of the 2AP substituted oligonucleotides in Fig. 3 and summarized in Table 1: A7 > A19 > A1 ≈ A13. It is well established that 2AP fluorescence is strongly quenched by nearby G residues, probably via an electron-transfer mechanism. The close proximity of A1, A13 and A19 to Gs in the Patel structure may account for some or all of these quenching effects.
As described above, provided appropriate τ0 values are known, collisional quenching data can provide information about environmental heterogeneity at the positions of substitution. According to the Stern-Volmer model (Eq. 1), a single fluorophore with a uniform microenvironment exhibits a linear relationship between F0/F and [Q]. Differences in KSV between different single probe residues, each in a homogeneous microenvironment may result either from alterations in k0, τ0 or both (Eq. 1). Clearly, without prior knowledge of τ0 for different single 2AP substitutions, it is impossible to correlate KSV directly with the accessibility of the reporter group to solvent. Kimura et al. (32) have determined that in Na+ or K+, the fluorescence decay curves of HuTel22 with serial dA → 2AP substitutions, consist of a single exponential with lifetimes (τ0) of 0.54, 0.34 and 0.35 ns for AP7, AP13 and AP19, respectively (32,33). This implies that, to a close approximation, the observed alterations in KSV result from alterations in the k0, which varies with the accessibility of the excited state fluorophore to the quenching agent at constant temperature and viscosity.
The quenching curves shown in Fig. 3 B and D are clearly concave with respect to [acrylamide]. In the case of the Na+ structure the excited state 2AP must experience at least two distinct microenvironments: one that is relatively accessible to Q and one that is shielded from Q. The structural basis of these differences is unclear; however, given that NMR indicates a single global fold (the basket structure) the structural heterogeneity probably results from highly localized, nanosecond fluctuations in positioning of the adenine that are confined to the respective loop regions.
To put the analysis on a more quantitative basis, we fit the quenching curves to a two-state model (Eq. 3 with n = 2), even though we realized that such a model likely is an oversimplification. The resulting best fit parameters (KSV,1, KSV,2 and f1) are summarized in Table 1. KSV,1 varies from 6.8 M−1 to 12.8 M−1, KSV,2 varies from 0 to 0.2 M−1, and f1 varies between 0.52 and 0.86. This analysis suggests that in Na+, AP7 and AP19 are nearly homogeneous with respect to their accessibility to acrylamide quenching (f1 ≈ 0.86), whereas a significant fraction of the AP1 and AP13 bases are relatively inaccessible to acrylamide quenching (32% and 42%, respectively). If the KSV parameters in Table 1 in Na+ are divided by the τ0 values above, the bimolecular collisional constant k0 = 13.6, 20.1 and 21.6 M−1 ns−1 for AP7, AP13 and AP19, respectively for the solvent-exposed residues, and approximately 100-fold less for the inaccessible fractions. The results are consistent with our earlier conclusions derived from the NMR structure of HuTel22 in Na+ that A13 is packed within the diagonal loop at one end of the structure while A7 and A19 are in more exposed regions in the lateral loops at opposite ends of the structure. However, the heterogeneity apparent in the f1 values suggests some flexibility in the diagonal loop such that on average, about half of the residues are relatively solvent accessible.
For K+, all quenching curves also display downward curvature as shown in Fig. 3D, indicative of local environmental heterogeneity at the sites of substitution as described above with Na+. When these data were analyzed by the 2-state model (Eq. 4), only A13 exhibited nearly equal fractions of relatively accessible and inaccessible states (f1 = 0.55 and f2 = 0.45). For the other substitutions, only 10–20% of the 2AP residues are relatively inaccessible to acrylamide, suggesting a relatively homogeneous environment. Notably, the KSV parameters associated with the accessible population of states are all the approximately the same (KSV ≈ 9.0–9.5 M−1). Since the requisite τ0 values are not available in K+, it is not possible to directly relate these KSV values to residue accessibility. However, f1 values (for A13) that are markedly < 1.0 indicate that a significant fraction of the residues reside in a microenvironment that is inaccessible to acrylamide. This leads to our previous conclusion that the structure of HuTel22 in K+ solutions is not the propeller structure, in which all loop adenines (A7, A13 and A19) are in equivalent environments.
In a second example, Petraconne et al. (34) correlated emission intensities and acrylamide quenching data with solvent accessible surface areas calculated for several different models of a 50-nt oligonucleotide d(TTAGGG)8TT. This sequence models eight repeats of the TTAGGG human telomere sequence and consists of two tandem quadruplexes. Molecular dynamics was used to model various combinations of hybrid-1, hybrid-2 and propeller structures, and the solvent accessible surface area (SASA) of each dA residue was calculated from the final 4 ns of the molecular dynamics trajectory. The authors compared the SASA with the observed F0/F values at 0.6 M acrylamide for a series of oligonucleotides in which each of the eight loop dA residues was individually substituted with 2AP. They concluded that one combination of topologies (hybrid-1-hybrid-2) showed the best correlation between SASA and accessibility to acrylamide.
In summary, the two studies discussed above show how fluorescence emission and fluorescence quenching measurements can be used to distinguish unambiguously between closely related structural models of G-quadruplexes. In common with similar studies, it is possible to determine best fitting values of parameters such as KSV which may be related to solvent accessibility and hence can be correlated with various structural models. However, because KSV is a product of at least three independently variable physical parameters (τ0, k0 and f0), and because of the difficulty of determining all of them independently from quenching curves, KSV values should be interpreted cautiously with respect to specific structures. Nevertheless, we and others have demonstrated that even in the absence of a rigorous quantitative interpretation of these variables, when high resolution model structures are available from which predictions regarding accessibility of dA residues can be made, the fluorescence properties of 2AP can be a valuable aid in model discrimination, especially when taken in conjunction with other biophysical measurements such as sedimentation.
Supported by grant CA35635 from the National Cancer Institute.