1.

Introduction

The canonical double-helical structure of B-form DNA is widely recognized. However, recent findings demonstrate that numerous, biologically significant, sequence-induced local structural deviations from this Watson–Crick global average are interspersed throughout the human genome.^{1 2} For example, repetitive DNA sequences constitute ∼30 of the human genome.¹ Single-stranded DNA can adopt unusual higher-order structure as well.^{1 2 3 4 5 6} Several studies indicate that 5^′-d(CGG)₇-3^′ forms hairpins and tetraplexes (Figure 1),^{7 8 9 10} whereas 5^′-d(CCG)₇-3^′ prefers hairpin formation.^{1 2 7 8 9 10} These hairpin formations create C–C mismatches, which are prone to methylation at CpG islands in vivo.¹ The link of this hypermethylation with massive triplet repeat expansion at the 5^′ untranslated region of the fragile X mental retardation type 1 (FMR-1) gene is correlated with fragile X syndrome.^{1 2}

Figure 1

Suggested higher-order structures of d(^m C ^m CG)₇, a hairpin (A) and d(^m CGG)_n, a tetraplex (B).

We have employed cadmium sulfide (CdS) semiconductor nanoparticles, 40–50 Å in diameter, as protein-sized photoluminescent probes of DNA structure.^{11 12 13 14 15} The photoluminescence of CdS is heavily influenced by adsorbates, and we found that intrinsically “kinked” DNA binds more strongly and more quickly to the curved CdS surface than does “straight” DNA of the same nominal length.^{11 12 13 14 15} Thus, DNA sequence, and by implication local DNA structure and/or flexibility, influences the binding of the DNA polymer to the nonspecific colloidal CdS.¹¹

In the present paper, we examine the interactions of single-stranded hypermethylated triplet repeats (all cytosines 5-methylated) with our protein-sized cationic nanoparticles. We had previously found that nonmethylated triple repeat DNA structures could be optically detected by our nanoparticles.¹¹ Methylation might be expected to perturb not only the DNA structure, but also the counterion atmosphere and water solvation shell around the DNA,¹⁶ and thus affect the interaction of our nanoparticle probes with these hypermethylated triplet repeat structures.

2.

Experiment

3.

Results

Figures 2 and 3 show the circular dichroism spectra and the melting temperature plots of d(^m C ^m CG)₇ and d(^m CGG)₇. Both sets of data suggest the formation of stable, higher-order structures at room temperature. The literature predicts that d(^m CGG)₇ forms a tetraplex regardless of whether the cytosines are methylated.⁷ The structure of d(^m C ^m CG)₇ is not as well documented, but according to the CD studies of nonmethylated d(CCG)_n, d(^m C ^m CG)₇ may form a hairpin-like structure.^{7 11 17 18 19}

Figure 2

Circular dichrosim spectra of d(^m C ^m CG)₇ (top) and d(^m CGG)₇ (bottom).

Figure 3

Melting curves for d(^m C ^m CG)₇ (top) and d(^m CGG)₇ (bottom).

Figure 4 shows the emission spectra of Cd(II)-rich CdS nanoparticles without DNA and in the presence of equal molar concentrations of duplex d(^m C ^m CG)₇, or d(^m CGG)₇ DNA. The presence of all three sequences contributed to photoluminescence quenching; however, the absolute change in photoluminescence was sequence dependent. Duplex or d(^m CGG)₇ DNA quenched the particle luminescence in nearly equal proportions, and d(^m C ^m CG)₇ resulted in more photoluminescent (less quenched) nanoparticle solutions. Buffer alone, which contained 1 mM EDTA, also contributed to a large degree of quenching (∼50–70 of the maximum with DNA) due to the affinity of EDTA to complex with activating Cd ²⁺ that produces the 460 nm emission band.^{7 8 9 10 11}

Figure 4

Photoluminescence spectra of Cd(II)-rich CdS nanoparticles containing no DNA and after addition of 0.20 mM (per nucleotide) duplex DNA, d(^m C ^m CG)₇, and d(^m CGG)₇ to separate equal-volume nanoparticle solutions.

We monitored the integrated intensity decrease of the 460 nm emission band with respect to DNA concentration (Figure 5) and fit the Frisch–Simha–Erich (FSE) isotherm, derived for a long polymer adsorbing in short segments to a locally flat surface,²⁰ to our data to obtain relative equilibrium binding constants (Figure 5):

Figure 5

Luminescence titrations of 400 μL Cd(II)-rich CdS nanoparticle solution with d(^m C ^m CG)₇ (top) and d(^m CGG)₇ (bottom). In all cases CdS photoluminescence is decreased. DNA concentrations were 0.0, 0.067, 0.13, 0.20, 0.26, and 0.32 mM (per nucleotide) during the titrations.

We have previously applied the FSE adsorption isotherm to luminescence data to understand the interactions among protein-sized CdS quantum-dot particles and other sequences of DNA.^{11 12 13 14 15} In earlier work, we showed that unmethylated d(CGG)₇ and d(CCG)₇ perturb the photoluminescence of 45 Å Cd(II)–CdS nanoparticles under conditions where duplex DNA does not bind to the particles above the buffer background.¹¹ Relative binding constants for these DNAs, as judged by the FSE theory of polymer adsorption were 13 000 M⁻¹ for d(CGG)₇ and 6100 M⁻¹ for d(CCG)₇. ¹¹ Competitive photoluminescence binding experiments confirmed this ordering.¹¹ Here, binding constants of 17 000 M⁻¹ were calculated for both d(^m C ^m CG)₇ and d(^m CGG)₇ DNA, while normal double-helical DNA bound with a K of 32 000 M⁻¹. This is the opposite of what we observe with the nonmethylated analogs of the triplet repeats.¹¹ Interestingly, compared to all of our other studies, d(^m C ^m CG)₇ is not capable of quenching all the photoluminescence from the 460 nm emission band of the nanoparticles (Figure 5). Thus, even though nanoparticle solutions containing d(^m C ^m CG)₇ emit more intensely than those containing d(^m CGG)₇, the binding constants come out the same.

4.

Discussion

Our results indicate that hypermethylated higher-order structures can be detected by protein-sized Cd(II)-rich CdS nanoparticles, although duplex DNA is a significant interferent. The FSE data suggest that there is no preferential binding of d(^m C ^m CG)₇ compared to d(^m CGG)₇, but the differences in absolute photoluminescence imply a different interaction for each structure with the nanoparticle substrate. The mechanism of nanoparticle photoluminescence quenching is not entirely well understood, but since it is reversible with salt¹⁵ and results in no permanent base damage after irradiation,¹² we think of the nanoparticle–DNA interaction as a simple donor–acceptor adduct, familiar from inorganic chemistry. The loss of the surface-associated Cd ²⁺ is one likely source of quenching, and in fact studies with double-stranded DNA suggest that counterion release from the nanoparticle–DNA interface is the thermodynamic driving force for the interaction.¹⁵ The salt dependence of the binding constants for long DNA to Cd(II)-rich CdS supports the notion that these nanoparticles are cationic in solution;¹⁵ we have not yet investigated the interaction of CdS nanoparticles capped with other surface groups with these DNAs.

With these thoughts in mind, we can begin to interpret the interaction of the hypermethylated single-stranded structures with the cationic nanoparticle surface. One additional difficulty is that the structures of these sequences are not completely known. Evidence supports formation of a tetraplex for d(^m CGG)₇, ^{7 8 9 10} but there seems to be no consensus on the higher-order structure of d(^m C ^m CG)₇. ⁷ However, electrophoretic gel mobility experiments show a fast moving band for a ^m C -rich strand, whereas the G-rich strand moves more slowly.⁷ This result suggests that the C-rich strand would form a less bulky, possibly hairpin-like structure. CD spectra of the individual strands concur with findings that indicate that G- and C-rich strands form tetraplex and hairpin-like structures.^{7 8 9 10 11}

If we imagine that both d(^m CGG)₇ and d(^m C ^m CG)₇ ball up into somewhat rigid higher-order structures that do not wrap about the nanoparticle’s curvature, then it is reasonable for their binding constants to the cationic nanoparticles to be similar, on a per-nucleotide basis. However, as noted above, d(^m C ^m CG)₇ is not able to quench the emission to the extent that duplex DNA and d(^m CGG)₇ do. This may reflect differences in the surface area of the DNA–nanoparticle interface for these higher-order methylated structures. The DNA backbone surface area exposed to the Cd(II)-rich CdS nanoparticle may be greater for the G-rich structure than for the C-rich structure, which would expel more counterions from the nanoparticle surface and produce more photoluminescence quenching compared to the C-rich strand. Salt-dependent equilibrium binding studies are needed to sort out the possibilities. The DNAs may also differentially induce aggregation of the nanoparticles, which may be another parameter to include in understanding the system, which may be accessible by dynamic light scattering studies.^{23 24 25} Nonetheless, the luminescent mineral nanoparticles we have described are a new kind of optical reporter for biological applications that may fruitfully be explored for other diagnostic assays.