Zn²⁺ Blocks Annealing of Complementary Single-Stranded DNA in a Sequence-Selective Manner

Abstract

Zinc is the second most abundant trace element essential for all living organisms. In human body, 30–40% of the total zinc ion (Zn²⁺) is localized in the nucleus. Intranuclear free Zn²⁺ sparks caused by reactive oxygen species have been observed in eukaryotic cells, but question if these free Zn²⁺ outrages could have affected annealing of complementary single-stranded (ss) DNA, a crucial step in DNA synthesis, repair and recombination, has never been raised. Here the author reports that Zn²⁺ blocks annealing of complementary ssDNA in a sequence-selective manner under near-physiological conditions as demonstrated in vitro using a low-temperature EDTA-free agarose gel electrophoresis (LTEAGE) procedure. Specifically, it is shown that Zn²⁺ does not block annealing of repetitive DNA sequences lacking CG/GC sites that are the major components of junk DNA. It is also demonstrated that Zn²⁺ blocks end-joining of double-stranded (ds) DNA fragments with 3′ overhangs mimicking double-strand breaks and prevents renaturation of long stretches (>1 kb) of denatured dsDNA, in which Zn²⁺-tolerant intronic DNA provides annealing protection on otherwise Zn²⁺-sensitive coding DNA. These findings raise a challenging hypothesis that Zn²⁺-ssDNA interaction might be among natural forces driving eukaryotic genomes to maintain the Zn²⁺-tolerant repetitive DNA for adapting to the Zn²⁺-rich nucleus.

Introduction

Zinc is the second most abundant trace element (after iron) essential for all living organisms¹. Within the cells, zinc exists as a redox-stable divalent cation (Zn²⁺) which primarily coordinates with proteins as structural components or catalytic cofactors². The total number of Zn²⁺-binding proteins has been reported to be 2,800 and 2,042, respectively in the human³ and Arabidopsis⁴ genomes. It is known that dietary zinc ions are taken up rapidly by the nucleus⁵ of the mammalian cells that accumulate more than one third (30–40%) of the intracellular zinc ions¹. This nucleus-predominant distribution reflects the fact that many DNA synthesis/transcription-related proteins are Zn²⁺-dependent, e.g., DNA polymerase alpha-primase⁶ and RNA polymerase II⁷, each contains six and eight zinc atoms per molecule, respectively. Intranuclear free Zn²⁺ sparks caused by reactive oxygen species have been observed in mammalian cells under oxidative stress^8,9, but if these Zn²⁺ outrages could affect annealing of complementary single-stranded (ss) DNA, a crucial step in DNA synthesis, repair and recombination¹⁰ is yet unknown.

Previous studies on Zn²⁺-DNA interactions were mainly focused on double-stranded (ds) DNA. It has been shown that Zn²⁺ promotes renaturation of denatured dsDNA, destabilizes or stabilizes the double helix of specific DNA sequences^11,12, causes DNA bending¹³ and kinking¹⁴ and induces transition from the right-handed B-DNA to other unusual forms such as the left-handed Z-DNA¹⁵, Z-DNA-related *Z DNA¹⁶, triple-stranded *H-DNA¹⁷ and metal-DNA complexes so-called M-DNA¹⁸. The biological relevance of the Zn²⁺-induced conformational perturbations on dsDNA has not been well established, largely due to that the related studies often involve non-physiological conditions.

In this work, a simple low-temperature EDTA-free agarose gel electrophoresis procedure (LTEAGE) was developed and used to characterize the Zn²⁺-ssDNA interaction coupling with UV-Vis spectrum and fluorescence quenching analyses. The presented data indicate that Zn²⁺ interacts with ssDNA and blocks the annealing of complementary strands in a sequence-selective manner under near-physiological conditions and such blocking effects may impact DNA replication/repair-associated processes such as the end-joining of DNA double-strand breaks and the annealing of long stretches of ssDNA involved in homologous recombination. It also suggests that the repetitive junk DNA widely spread among eukaryotic genomes might provide annealing protection on the otherwise Zn²⁺-sensitive coding-DNA. The biological relevance of these findings is discussed while examining the distribution of Zn²⁺-tolerant junk DNA in representative chromosomes of humans and plants such as wheat.

Results

Zn²⁺ blocks annealing of complementary ssDNA with a preference for G+C-“rich” sequences

For initial tests, four sets of complementary short DNA oligos (25–33 nt) were synthesized based on the wheat actin gene (ACT1). ACT1-E25 (25 nt, G+C 60%) and ACT1-E26 (26 nt, G+C 50%) were designed to the second and third exon, respectively and ACT1-I26 (26 nt, G+C 42.3%) and ACT1-I33 (33 nt, G+C 30.3%) to the second intron of the ACT1 gene (Supplementary Table S1). LTEAGE tests (Fig. 1a) revealed that the annealing of the relatively G+C-“rich” ACT1-E25 or ACT1-E26 oligos was apparently reduced when the complementary ssDNA was pre-treated with 100 µm Zn²⁺ and completely blocked when the Zn²⁺ concentration increased to 250 µm, as indicated by the absence of dsDNA-specific bands and in contrast, the annealing of the A+T-rich ACT1-I26 or ACT1-I33 was only reduced by Zn²⁺ at ≥250 µm, as indicated by the gradually weakened dsDNA bands (Fig. 1b). The blocking effects were abolished when EDTA, which is a strong chelator of divalent metal ions¹⁹, was added to the annealing solutions (lanes 9), indicating that the blocking was indeed caused by Zn²⁺. Variations in salt concentrations within the ranges determined for the nucleus of eukaryotic cells (e.g., 10–30 mM NaCl)^20,21 did not change the annealing patterns significantly under the same Zn²⁺ treatments (data not shown). Three other divalent metal ions including Mg²⁺ and Mn²⁺ (Supplementary Fig. S1) as well as Ca²⁺ (data not shown) did not affect the annealing under the same conditions, or even at higher (up to 2 mM) concentrations. Incubation of pre-annealed dsDNA with Zn²⁺ did not cause disassociation of the double helix (Supplementary Fig. S2). These results indicated that the blocking of the annealing process was due to Zn²⁺-ssDNA interaction.

Figure 1

Zn²⁺ blocks annealing of complementary ssDNA in a sequence-selective manner.

(a) LTEAGE procedure. (b) Ethidium bromide (EtBr)-stained 2.5% agarose gels showing the annealing patterns. Control lanes were loaded with ssDNA (+ or – strand) without Zn²⁺ treatment. Lanes 9 were loaded with EDTA-treated (1 mM) samples. Double/single-stranded DNA-specific bands are indicated by black and gray arrows, respectively. M: 50-bp DNA ladder; (c) A295/A260 ratios plotted against metal ion concentrations. Arrows indicate the first points at which the ratio increased significantly. Numbers on the right of the strand symbols indicate the corresponding SD^Z values (at 250 µm). (d) and (e) Fluorescence quenching assay. Bars in (c) and (e) indicate standard deviations, most of which in (c) are masked by the squares representing the means; asterisks indicate significant differences (t-test, P<0.05).

UV-Vis spectrum analysis was done to detect Zn²⁺-induced conformational perturbations in ssDNA solutions. It was found that Zn²⁺ caused a significant increase in the A295/A260 ratio of all four tested ssDNA samples when added at concentration ≥ 500 µm with the maximum found at 750 µm (Fig. 1c). However, the minimal concentration causing an increase in the A295/A260 ratio differed: it was found at 100 µm for both the plus (+) and minus (-) strands of ACT1-E25 and for the plus strand of ACT1-E26 in contrast to 250–500 µm for the complementary strands of ACT1-I26 and ACT1-I33 (Fig. 1c, arrows). The observed A295/A260 ratio changes in the tested concentrations (e.g., 10–750 µm Zn²⁺) were not due to DNA precipitation as confirmed in separate experiments (data not shown); this was in consistence with a previous report that Zn²⁺ causes DNA precipitation with minimal concentrations at millimolar ranges, e.g., >1 mM²².

To enable a comparative analysis, the A295/A260 ratios (A*) were further normalized using the formula (A*₂₅₀- A*₀)/A*₀, in which A*₂₅₀ and A*₀ stands for the A295/260 ratio for the sample containing metal ions at 250 µm and the blank control, respectively. For convenience, the calculated value was represented by the symbol “SD^Z” to indicate “Single-stranded DNA Zn²⁺-binding potential”. This method was adapted solely for the comparative analysis of the UV-Vis spectrum data, not attempted to substitute for dissociation constant (K_d) (which cannot be determined in this study due to technical restraints). The calculated SD^Z for the +/− strands of ACT1-E25 (G+C 60%), ACT1-E26 (G+C 50%), ACT1-I26 (G+C 42.3%) and ACT1-I33 (G+C 30.3%) was 1.14/1.22, 1.00/0.75, 0.89/0.86 and 0.50/0.16, respectively (Fig. 1c, Supplementary Table S2). Neither Mg²⁺ nor Mn²+ caused significant changes in the A* values under the same conditions (Fig. 1c) or affected the annealing process of any ssDNA samples in the LTEAGE tests shown in Supplementary Fig. S1. These results confirmed that Zn²⁺ binds to ssDNA and the binding potential is likely determined by the G+C content of the targeted ssDNA molecules.

Fluorescence quenching assays were done to verify that the Zn²⁺ blocking event occurred independent of agarose gel electrophoresis conditions. Relative fluorescence units (RFU) were monitored in solutions containing the dual-labelled ACT1-E25 probe and its complementary strand pre-treated with metal ions (Fig. 1d). As shown in Fig.1e, the RFU started to decrease significantly when Zn²⁺ was added at a concentration as low as 50 µM and reached the lowest levels (5–10%) at 250-500 µM. Neither Mg²⁺ nor Mn²⁺ caused a RFU reduction; instead, Mg²⁺ appeared to enhance annealing efficiency at ≥250 µM. These data confirmed that Zn²⁺ was able to block the annealing of sensitive ssDNA like ACT1-E25 at an even lower concentration (50 µM).

Zn²⁺ does not affect annealing of simple repetitive DNA sequences lacking CG/GC sites

To determine how the individual nucleotides contribute to the Zn²⁺ binding potential of a given ssDNA, oligos composed of different simple repeats were synthesized (Supplementary Table S1) and subjected to LTEAGE and UV-Vis spectrum analyses. It was found that the annealing between all types of complementary simple repeats that lack CG/GC sites were not or only slightly affected when Zn²⁺ was added at a concentration up to 750 µm. These Zn²⁺-“tolerant” repetitive DNA included the mononucleotide repeats poly(dA)/(dT) and poly(dG)/(dC) (Fig. 2a), the dinucleotide repeats poly(dT-dG)/(dC-dA) and poly(dT-dC)/(dG-dA) (Fig. 2b), the trinucleotide repeats-like sequences such that associated with the wheat gypsy-type retrotransposon²³ and the hexanucleotide repeats such as the TTAGGG motifs found in the telomeric ends of all human chromosomes²⁴ (Fig. 2c). The other two types of dinucleotide repeats, i.e., poly(dC-dG)/(dG-dC) and poly(dA-dT)/(dT-dA) cannot be tested per se because of the self-complementarity. To demonstrate the vulnerability of the CG/GC sites to Zn²⁺ blocking, two sets of DNA oligos consisting of non-self complementary CG or TA repeats were designed: dCG-24 which contains six CG repeats separated by continuous G bases and dTA-24 which has two TA repeats flanked by continuous T or A bases (Supplementary Table S1). The annealing of dTA-24 was reduced but never completely blocked and in contrast, the annealing of dCG-24 was apparently reduced at 100 µm and almost completely blocked at ≥250 µm (Fig. 2d).

Figure 2

Annealing patterns and Zn²⁺ binding potentials of ssDNA composed of repetitive nucleotides.

(a–d). EtBr-stained 2.5% agarose gels showing the annealing patterns of mononucleotide (a), dinucleotide (b), tri- or hexanucleoride repeats (c) and the ssDNA containing CG or TA-repeats (d). Note that although the equal amounts of DNA were loaded, some ssDNA controls (e.g., lanes 1 and 2 in a–c) were not (or only faintly) visible because these CG/GC sites-lacking DNA oligos are not expected to form stable helical structures (as those shown in Fig. 3) that are preferentially targeted by EtBr, a well-known double helical DNA intercalator. (e and f). A295/A260 ratios plotted against metal ion concentrations as shown in Fig. 1. (g) Potential Zn²⁺ binding sites in the four types of nucleotides. The O and N atoms in the heterocyclic rings are highlighted in red. Hydrogen-bonding donors and receptors are indicated by gray and black arrows, respectively (adapted from ref. 10). Pink spheres representing the coordinated Zn²⁺-water complex (ref. 26). Note that adenine lacks O atom and the N3 atom of thymine may not be a preferred binding site for Zn²⁺ (dashed lines) due to its bonding with the proton.

UV-Vis spectrum data indicated that the A* values of poly(dA) remained unchanged at all Zn²⁺ concentrations and those of other three mononucleotide repeats increased significantly only at higher concentrations (≥250 µm) and the calculated SD^Z was all relatively low (≤1.0) (Fig. 2e, Supplementary Table S2). Similarly, the A* values were also found to increase significantly only at higher concentrations (≥250 µm) for all dinucleotide repeats except for poly(dT-dG) and dGC-24, both of which were found at 100 µm and the SD^Z was all low (≤1.0) except for poly(dT-dC) having SD^Z = 2.1 (Fig. 2f, Supplementary Table S2); These data suggested that the “tolerance” to Zn²⁺ blocking was due to the low Zn²⁺-binding potential (e.g., SD^Z <1) associated with one or both complementary strands of the simple repetitive DNA, e.g., dTA-24/dAT-24 (SD^Z = 0.0 vs. 0.13) and poly(dT-dC)/(dG-dA) (SD^Z = 2.1 vs. 0.5) (Supplementary Table S2). This was also true for the ACT1-derived random sequences, e.g., the Zn²⁺-tolerant ACT1-I26 and I-33 oligos had a low (0.17–0.89) SD^Z for both strands (Fig. 1c). But this rule does not apply for the dCG-24 oligos, which had lower SD^Z values for both strands (0.42 and 0.73, respectively) but annealed poorly (Fig. 2d, bottom). For the plus strand, this was likely due to the potential to form a G-quartet-like structure characteristic of repeated three-guanines separated by 3 to 5 T, A or G residues as described in previous studies²⁵. The ssDNA with this unusual secondary structure had a high A295/260 ratio as shown in the blank control (0.19 vs 0.02–0.12 for most oligos), leading to a lower SD^Z [calculated using the formula (A*₂₅₀- A*₀)/A*₀ as mentioned above]. Thus, the SD^Z values of the dCG-24-like ssDNA with G-quartet-like structures are not comparable with other CG/GC-sites-containing random sequences with respective to the annealing events. Different methods are needed to resolve this issue. However, the C base-lacking telomere end-specific oligo (Supplementary Table S1), which is also expected to form G-quartet-like structures, was found to be Zn²⁺-tolerant (Fig. 2c). This suggested that CG/GC sites (which are present in dCG-24 but not in the telomere end-specific oligo) are important for the observed Zn²⁺ blocking effects. This was further tested by site-specific mutational analysis (see below).

The Zn²⁺-binding potential of ssDNA was found to correlate with the chemical structures of compositional nucleotides. Adenine is the only one lacking an oxygen (O) atom; thymine is the only one lacking a deprotonated nitrogen (N) atom in the heterocyclic rings whereas guanine and cytosine have both O and one or more deprotonated N atoms (Fig. 2g). Previous studies have shown that divalent metal ions like Mg²⁺ and Zn²⁺ prefer to bind to G, C and T bases in a bidentate manner, i.e., bonded through an oxygen (O) and a nitrogen (N) atoms simultaneously²⁶. Thus, poly(dA) and poly(dT-dA) would have the lowest SD^Z because of the lack of bidentate sites; poly(dC) and poly(dT-dC) would have the highest SD^Z because cytosine has a bidentate site which is always accessible due to the lack of intramolecular base paring, but ssDNA consisting of only CG/GC repeats like the minus strand of the dCG-24 oligos (Supplementary Table S2) would have a low SD^Z because the frequent intramolecular C-G base paring would inactivate some bidentate sites (Fig. 2e–2g). On the other hand, in a random sequence consisting of all four types of nucleotides, the bidentate sites on G/C bases could remain open depending on the potential of C-G base pairing that is determined by the nucleotide sequences of a given ssDNA. Thus, for most random sequences, the higher the G+C content, the higher the Zn²⁺ binding potential (e.g., SD^Z >1) as shown for the ACT1-derived sequences (Fig. 1c, Supplementary Table S2).

Zn²⁺ blocks ssDNA annealing likely by targeting the CG/GC-like helix in the hairpin stem

Site-specific mutational analysis was performed to explore the structural basis of the Zn²⁺-specific blocking event. DNA folding analysis revealed that all Zn²⁺-“sensitive” ssDNA sequences are likely to form hairpins with a helical stem consisting of CG/GC or TG/GC external closing pairs. For example, in ACT1-E25, the last two bases (-CG₂₅) at the 3′ end would fold back to form a hairpin-like structure resulting in a 2-bp helix with C₁₈-G₂₅ as the 5′ external closing pair (Fig. 3a). In ACT1-E26, the last three bases (-CCG₂₆) would fold back giving rise to a 3-bp helix (TGG/GCC) with T₁₁-G₂₆ as the 5′ external closing pair through wobble paring¹⁰ (Supplementary Fig. S3a). Both ACT1-I26 and ACT1-I33 would form hairpin-like structures but the resulted helices all lack CG/GC-like base pairs at the 5′ end of the helical stem. ACT1-I26 would have a 3-bp helix (GGA/CCT) with G₁₁-C₂₄ as the 5′ external closing pair (Fig. 3b) and ACT1-I33 would have a 5-bp helix (ACTGC/GCATT) with A₁₇-T₃₂ as the 5′ external closing pair (Supplementary Fig. S3b). Site-specific mutations depriving the CG/GC-like helices in ACT1-E25 conferred Zn²⁺ “tolerance” (i.e., the annealing was not blocked) as shown for the ACT1-E25-G19C/G25C mutant (Fig. 3a) whereas the mutations introducing new hairpin stems with a 5′ end CG/GC-like helix led to Zn²⁺ “sensitivity” (i.e., the annealing was blocked) as shown for the ACT1-I26- G11C/C12G mutant (Fig. 3b) and also for the ACT1-I33-G20C/C21G mutant (Supplementary Fig. S3b). Mutations made to the TG/GC helix of ACT1-E26 did not change the Zn²⁺ sensitivity because the ACT1-E26-C7G/G26C mutant gained a new helical stem with 5′ GG/CC external closing pairs (Supplementary Fig. S3a). In addition, sequential mutations were made to the Zn²⁺-tolerant poly(dT-dG) which became Zn²⁺-sensitive only when the mutations had led to a hairpin stem with TG/GC as the 5′ external closing pairs (Supplementary Fig. S4). These results confirmed that the CG/GC (TG/GC or GG/CC) helices derived from the 5′ external closing pairs in the hairpin stems of ssDNA were likely targeted by Zn²⁺ in the blocking events. This model was consistent with a previous X-ray structure study which suggested that Zn²⁺ binds to G bases in the 5′ end external closing pairs of a hairpin DNA by coordinating with the N7 atom and an O atom of a phosphate group associated with the adjacent nucleotide²⁷.

Figure 3

DNA folding and site-specific mutational analyses reveal the hairpin structures potentially targeted by Zn²⁺ that prevent ssDNA annealing.

(a). Point mutations that eliminate the CG/GC helix in the Zn²⁺-sensitive (S) ACT1-E25 leads to a Zn²⁺-tolerant (T) mutant (ACT1-G19C/G25C). (b) Point mutations that introduce a TG/GC helix in the Zn²⁺-tolerant ACT1-I26 results in a Zn²⁺-sensitive mutant (ACT1-I26-G11C/C12G). One representative folding structure predicted to be stable (with smaller ΔG) was chosen for each ssDNA as shown at the top. Colors in circles indicate the probability (0.001/black - 0.999/red) for the coded bases to be single stranded upon folding of the ssDNA. Mutation sites are indicated by asterisks. The guanine bases potentially targeted by Zn²⁺ are indicated by red arrows. Hydrogen bonds between C and G bases are highlighted in red. EtBr-stained 2.5% agarose gels showing the annealing patterns are given below (see Figure 1b for related legends).

Zn²⁺ blocks the end joining of dsDNA fragments with short 3′ overhangs

The DNA oligo-based annealing assay mentioned above provided a simple way to evaluate the Zn²⁺-specific blocking effects on the annealing process, but it is not suitable for testing even shorter (e.g., <15 bp) sequences that cannot be resolved clearly on agarose gels. To overcome this drawback, a pair of dsDNA fragments was designed, in which each dsDNA fragment carried a 12-nt complementary 3′ overhang containing Zn²⁺-sensitive CG site (Supplementary Table S1), so that the annealing outcome can be easily monitored based on the success or failure of the end joining (EJ) between the two compatible dsDNA fragments in response to the Zn²⁺ treatment (Fig. 4a).

Figure 4

Zn²⁺ blocks the end joining (EJ) of double-stranded DNA fragments mimicking double strand breaks (DSBs).

(a) Procedures to generate DSBs. Oligos (L+/− and R+/−) are drawn in different colors with the 3′ overhangs highlighted in red or orange. Numbers (in nt/bp) indicate the size of DNA. (b) EtBr-stained agarose gels showing the end-joining patterns. Red asterisks indicate the point at which EJ was completely blocked. Note that the EJ event was not complete as seen in the control (lane 7), probably due to the lack of a DNA ligase (not included) for filling the gap at the EJ points. Control lanes were loaded with ssDNA (L+/− or R+/−) or DSB (L or R) alone without Zn²⁺ treatment. Lane 13 was loaded with EDTA-treated (1 mM) samples containing the two DSBs. End-joined and blocked DSBs-specific bands are indicated by black and gray arrows, respectively. M: 50-bp DNA ladder.

LTEAGE tests revealed that the EJ event was apparently reduced by Zn²⁺ at 250 µm and completely blocked at ≥500 µm (Fig. 4b). The 68-bp dsDNA product of the EJ event was well separated from the two dsDNA fragments (27 and 29 bp, respectively, excluding the 3′ overhang) and also distinguishable from the ssDNA samples consisting of the two complementary strands (27 to 41 nt) of the respective dsDNA fragment. The dsDNA fragment-specific bands but not the ssDNA-specific bands were clearly seen on the gel lanes loaded with the Zn²⁺-blocked samples (lanes 11 and 12), indicating that the absence of the EJ event was due to the failure of the annealing between the two complementary 3′ overhangs carried by the two dsDNA fragments, not the results of any Zn²⁺-induced disassociation of the double helix of each dsDNA molecule. Neither Mg²⁺ nor Mn²⁺ affected the EJ events under the same conditions (Supplementary Fig. S5). These results confirmed that Zn²⁺ was able to block the annealing between complementary ssDNA as short as 12 nt, which is within the size range of ssDNA intermediates involved in the double-strand breaks (DSBs)-associated DNA repair processes such as the microhomology-mediated end joining pathway²⁸.

Zn²⁺ blocks the renaturation of denatured dsDNA with a preference for coding sequences

To test the Zn²⁺ blocking effects on the annealing of even longer ssDNA without oligo synthesis limitations, DNA denture-renaturation tests were performed using dsDNA amplified by PCR as illustrated in Fig. 5a. Four dsDNA fragments (316–1,635 bp) were amplified from the wheat ACT1 gene. Two of them were amplified from the coding sequences with ACT1-E3 (420 bp, G+C 52.3%) from the third exon only and ACT1c (1,090 bp, G+C 52.9%) from the coding region involving three exons. The other two were amplified from genomic DNA with ACT1-I2 (316 bp, G+C 33.8%) from the second intron only and ACT1g (1,635 bp, G+C 44.7%) from the entire open reading frame (Supplementary Table S3).

Figure 5

Zn²⁺ blocks the renaturation of denatured dsDNA.

(a) Experimental procedures. C1, control 1 (intact dsDNA); C2, control 2, denatured DNA (kept on ice to preserve the single-stranded status). (b) EtBr-stained 1.0% agarose gels showing the renaturation patterns of the denatured DNA fragments indicated on the right (numbers in parentheses indicate the size, in bp). Red asterisks indicate the points at which annealing was completely blocked. Gels are labeled in a similar way as shown in Figure 1b. M, 100 bp (top two panels) or 1 kb (bottom two panels) DNA ladders. Numbers on the left indicate the positions of size markers.

LTEAGE tests revealed that the renaturation of the exonic ACT1-E3 (420 bp) was completely blocked by Zn²⁺ at 250 µm whereas that of the intronic ACT1-I2 (316 bp) was blocked by Zn²⁺ at 500 µm, twice higher than that of ACT1-E3 (Fig. 5b, lanes 6 and 7, top two panels).The renaturation of the coding DNA-derived 1,090-bp ACT1c, which was more than twice longer than ACT1-E3, was also completely blocked by Zn²⁺ at 250 µm (third panel). In contrast, the 1,635-bp ACT1g fragments, which contained three A+T-rich introns, renatured perfectly at the same concentration (bottom panel).These results indicated that minimal Zn²⁺ concentrations blocking the annealing of short (24–26 nt) DNA oligos (e.g., Fig. 1b and 2d) were sufficient to prevent the annealing of G+C-rich coding DNA with the size >10 times longer (420–1,090 bp) and more importantly, the Zn²⁺-tolerant intronic sequences provided annealing protection on the Zn²⁺-sensitive coding DNA when the two join together as compared with ACT1c and ACT1g (lanes 6 and 7, red asterisks). Interestingly, unlike the ACT1-I26 and ACT1-I33 DNA oligos, both of which were tolerant to Zn²⁺ blocking up to 750 µm (Fig. 1b), the 316-bp ACT1-I2 fragment which contained both ACT1-I26 and ACT1-I33 sequences from the same intron region became Zn²⁺-sensitive at 500 µm (Fig. 5b, the second panel from top). This was likely due to that the ACT1-I2 fragment actually contained short (13 and 10 bp, respectively) flanking exonic sequences incorporated into the PCR primers due to technical requirements (Supplementary Table S3) that would give rise a hairpin stem containing a GG/CC helix at the 5′ end as predicted by DNA folding analysis (data not shown); this de novo hairpin structure could have partially offset the Zn²⁺ tolerance of the original 293-bp intron sequence.

Discussion

Zn²⁺ has been a subject of studies on metal ion-DNA interactions since the 1960s¹¹ but much attention has been attracted to its ability to induce B- to Z-DNA transition¹⁵ and convert B-DNA to M-DNA by directly replacing the imino protons of thymine/guanine bases in the double helix¹⁸. While the influence of this well-known transition metal ion on the structure and physical properties of the DNA double helix is widely recognized, the biological relevance is still uncertain, largely because the related studies were often done under non-physiological conditions, e.g., the observed B- to Z-DNA transition involves the co-existence of certain organic compounds such as ethanol or ethylene glycol¹⁶ while the formation of M-DNA depends on alkaline conditions (pH = 8–9)¹⁸. Meanwhile, the interaction between Zn²⁺ and ssDNA has been largely untouched and the questions if and how such interaction could affect the annealing of complementary sequences have not been brought to attention in the field.

In the presented study, the Zn²⁺-ssDNA interaction was first investigated using a simple and sensitive LTEAGE procedure coupled with UV-Vis spectrum and fluorescence quenching analyses. The exclusion of EDTA, which is a strong chelator of divalent ions¹⁹ from the routine agarose gel electrophoresis was the key to success in this study and has not been reported in previous studies. This simple method should be generally applicable for future studies on metal ion-DNA interactions, especially those requiring high-throughput tests that may not be feasible for the application of sophisticated technologies such as X-ray absorption spectroscopy and nuclear magnetic resonance. More meaningfully, the Zn²⁺-ssDNA interaction was studied in solutions with near neutral pH (7.5) at body temperatures (37°C) that fit physiological conditions. The concentration of NaCl (e.g., 10 mM) tested in this study was much lower than that of physiological saline (~150 mM), however, it is just within the ranges as measured for the nucleus of eukaryotic cells as reported in previous studies^20,21.

Regarding the concentrations of Zn²⁺ in the annealing solutions, it is known that in routine lab stock solutions, free Zn²⁺ could be <0.1% of the calculated molar concentration due to natural oxidation^30,31, thus, the minimal Zn²⁺ concentrations required for blocking DNA annealing as reported in this study (50–100 µm) might be actually at nanomolar levels, i.e., 50–100 nM, which is close to the estimated concentrations of intracellular free Zn²⁺ at least in certain types of living cells as reported by several research groups specializing in zinc biology-related areas^29,30,31,32. The physiological concentrations of intranuclear free Zn²⁺ is largely unknown, but is expected to be significant as observed at least in the mammalian cells under oxidative stress^8,9. It is not unreasonable to assume that the observed Zn²⁺ blocking events might occur not only in vitro but also in the nucleus, which accumulates 30–40% of the total zinc in the human body¹. Certainly, this possibility would depend on a yet-to-be determined factor that the Zn²⁺-ssDNA interaction must be strong and quick enough to compete with protein-DNA interactions associated with various DNA synthesis/repair pathways^10,33,34. Nevertheless, there are reasonable chances for the “naked” ssDNA to be exposed, at least transiently, to intranuclear free Zn²⁺, e.g., upon the collapse of the DNA replication forks in which the replication proteins become disassembled and fall off the ssDNA templates, leading to the accumulation of ssDNA at the stalled replication forks^33,34.

What would be the biological consequences if the observed Zn²⁺ blocking event does occur in a living cell? Single-stranded DNA is an essential intermediate in DNA replication, repair, recombination and transcription and must be processed timely¹⁰. For example, double-stranded DNA breaks (DSBs) occur frequently in eukaryotic genomes due to the attacks of reactive oxygen species (ROS) and in order to repair the DSBs, the 3′-overhangs must be extended and subsequently annealed between complementary partners²⁸. Since ROS also causes free Zn²⁺ sparks in the nucleus⁸, it is not hard to image that ROS-released Zn²⁺ could block DSB repair if the 3′-overhangs contain Zn²⁺-sensitive sequences as shown in the end-joining tests mimicking this process (Fig. 4). As suggested by the denature-renaturation tests carried out in this study (Fig. 5), Zn²⁺ could block the annealing of ssDNA as long as 1,635 bp, which could be in the size ranges of ssDNA involved in homologous recombination-dependent DNA repair critical for the reactivation of stalled DNA replication forks caused by DNA damages common in living cells^35,36. A failure in rescuing the stalled replication forks could lead to catastrophic consequences including cell death and aberrant cell-division^33,34,35,36.

The presence of intracellular histochemically reactive free or labile Zn²⁺ (usually detectable at nano- to micromolar levels) is under the control of zinc homeostasis pathways^30,31,32. Disturbances on the balance of intracellular free Zn²⁺ have been linked to apoptosis/programmed cell death-related processes in several types of human cells^37,38 and in plant embryos³⁹. Excess free intracellular free Zn²⁺ has been associated with brain injuries and neuronal disorders such as Alzheimer's disease^30,38 and in several types of human cancer cells, especially those of the breast and lung cancers in which the free Zn²⁺ levels are significantly higher than in normal cells⁴⁰. A previous study has shown that Zn²⁺ inhibits DNA replication in the human pathogenic bacterium Salmonella enteric under phagocytic ROS attack in which the inhibition of DNA replication is always accompanied by intracellular Zn²⁺ mobilization⁴¹. Although these observations were likely due to the disturbance of Zn²⁺ on the functions of various pathway proteins, the data obtained in this study imply that the Zn²⁺-ssDNA interaction may be also responsible for cell death and/or tumor growth caused by a dysfunctional DNA replication/repair machinery.

The fact that the Zn²⁺-tolerant intronic DNA may provide annealing protection on otherwise Zn²⁺-sensitive coding-DNA (Fig. 5) implies further the biological relevance of the Zn²⁺-ssDNA interaction. Eukaryotic genomes often contain large amounts of repetitive DNA (>50% in humans and up to 80% in plants such as wheat), mainly transposable elements (TEs)^42,43. The justification for us to live with so much junk DNA⁴⁴ has never been resolved univocally^45,46. Eukaryotic genomes are usually organized as alternative blocks of euchromatin (relatively G+C-rich) and heterochromatin (all A+T-rich) and TEs are found predominantly in the euchromatin regions as shown for the human chromosome 1 (250 Mb) (Fig. 6a), in which all TEs contain short sequences (usually 20–50 bp) similar to those of Zn²⁺-tolerant repetitive DNA examined in this study, especially, poly(dA)/(dT) stretches (Supplementary Table S4). These repetitive elements are organized in concert so that the G+C-rich exonons/promoter-related DNA sequences (which, at least for some, have apparent homologues in bacteria) are all separated by TE-embedded intronic or intergenic regions containing Zn²⁺-tolerant segments almost every 100–200 bp (Fig. 6a). Some of TEs are well conserved in closely-related species (Fig. 6b) and were tested to be Zn²⁺-tolerant as shown for Alu (Fig. 6c) which takes up >10% human genome⁴⁷. Similar TE-dominant organization was also found in plant chromosomes such as those in wheat (Supplementary Fig. S6).

Figure 6

Zn²⁺-tolerant repetitive DNA associated with transposable elements (TEs) in human genome.

(a) Euchromatin, heterochromatin and euchromatin-associated TEs in human chromosome 1. Idiogram with G-bands (adapted from ncbi.nlm.nih.gov/projects/mapview) is shown at the top, a 200-kb region (positions 28.8-29 Mb of the annotation release 105) with an enlarged segment (20 kb) is shown below. Protein-coding genes are indicated by arrows/letters with numbers below the exons indicating the G+C content (%). RCC1: Regulator of chromosome condensation protein 1. (b) Alignments of major TEs in humans and other five primate species. A, T, G and C bases are shown in green, blue, red and orange background, respectively (complete alignments available upon request). Species abbreviations (with GenBank accession numbers): Homo sapiens/human (Hs), Macaca mulatta/Rhesus monkey (Mm), Chlorocebus aethiops/grivet monkey (Cha), Gorilla gorilla/gorilla (Gog), Nomascus leucogenys/gibbon ape (Nol), Pan troglodytes/chimpanzee (Pat), Pongo abelii/Sumatran orangutan (Poa). (c) Gel images showing the annealing patterns of DNA oligos representing the Alu sequence in B (thick or dashed lines). Note that Alu-50 which includes both Alu-25 (Zn²⁺-tolerant) and Alu-25b (Zn²⁺-sensitive) was Zn²⁺-tolerant (compare at red asterisks). Lane organization and labelling are the same as those in Figure 1b.

Do these observations mean we need junk DNA to be able to survive free Zn²⁺ attack while heavily relying on it to make numerous nuclear proteins to keep our genome alive? The human telomeric DNA all end up with a single-stranded tip (100–200 bp) consisting of exclusively the TTAGGG-like simple repeats²⁴ that are Zn²⁺-tolerant as shown in this study (Fig. 2c). This may be a direct example to address how smartly a linear chromosome has evolved in the Zn²⁺-rich nucleus of eukaryotic cells. For the whole genome of any eukaryotic organisms, it sounds quite reasonable that with some junk DNA sticking or jumping around, they may live with Zn²⁺ more safely despite having to pay for the evolutionary by-products, notoriously, the TE insertion-induced genetic diseases in humans⁴⁷ or the strange colors on kernels of cereal plants⁴⁸. Although much more need to be done both in vitro and in vivo before drawing any conclusions, the data obtained in this study have raised a challenging hypothesis that the Zn²⁺-ssDNA interaction might be among natural forces driving the eukaryotic genomes to maintain junk-like repetitive DNA for adapting to the Zn²⁺-rich nucleus.

Methods

DNA oligonucleotides

DNA oligonucleotides (24–50 nt) were designed manually and selected carefully against self-complementarity. Selected oligos were synthesized at Sigma Genosys (Sigma Aldrich) with desalting purification and prepared as 10 µm working solutions in nuclease-free water (Ambion-Invitrogen) and stored at −20°C. Sequences of the synthesized DNA oligos are given in Supplementary Table S1.

Annealing assays

Zn²⁺, Mg²⁺ and Mn²⁺ solutions were prepared from zinc sulfate (ZnSO₄), magnesium sulfate (MgSO₄) and manganese sulfate (MnSO₄), respectively, as 1 M stock solutions in ultrapure water obtained using the Milli-Q-Gradient purification system (Millipore) and stored at room temperature after autoclaving. The stock solutions were diluted to desired concentrations with ultrapure water just before use. For annealing assays, DNA oligo (2 µm) was suspended in 20 µl of solution (10 mM NaCl, 10 mM Tris pH 7.5) containing metal ions at a final concentration 0–750 µm, incubated at 37°C for 10 min. Metal ion-DNA solutions containing complementary ssDNA were then mixed at equal molar concentrations and allowed to anneal at 37°C for 30 min to 12 h. Three types of controls were included in all annealing assays, a DNA solution containing the plus (+) or the minus (−) strand of ssDNA alone without metal ions and an annealing solution containing a metal ion at 750 µm plus EDTA (1 mM). Annealed DNA samples were separated using the low-temperature EDTA-free agarose gel electrophoresis (LTEAGE) procedure, in which the ethidium bromide-containing gel (1–2.5%) was run at 4°C in Tris-acetate buffer (pH 8.0) without EDTA at 110 V for 25–30 min. Gel images were taken using the Gel Logic 200 System (Kodak).

UV-Vis spectrum analysis

Aliquots (100 µl) of solutions (10 mM NaCl, 10 mM Tris pH 7.5) containing ssDNA (1 µm) and metal ions (0–750 µm) or metal ion (750 µm) plus EDTA (1 mM) (control) were incubated in a 96-well clear microtiter plate (Corning) at 37°C for 10 min before scanning for absorbance at 260 nm (A260) and 295 nm (A295) using a Synergy H1 microplate reader (BioTek) with default settings for pathlength correction. A buffered solution without DNA was included as the blank control in all experiments. The A295/260 ratio (A*), which is an indicator of metal ion-DNA interactions¹⁶, was calculated and used to compare the relative Z²⁺ binding potential (SD^Z) of individual ssDNA using the formula (A*₂₅₀- A*₀)/A*₀ as described in the Results.

Fluorescence quenching assay

The plus strand was dual-labeled in a way adapted from a previous study⁴⁹ with Cy3 as the 5′ fluorophore and BQH2 as the 3′ quencher (Sigma). Aliquots (100 µl) of solutions (10 mM NaCl, 50 mM KCl, 10 mM Tris pH 7.5) containing the minus strand (0.2 µm) were pre-incubated with 0–500 µm metal ions in a 96-well black microtiter plate (Corning) at 37°C for 10 min. Probe was then added at an equal molar concentration and incubated at 37°C for 10 min. Fluorescence was detected using the microplate reader mentioned above with fluorescent settings for excitation/emission = 540/570 nm. Relative fluorescence unit (RFU) was calculated by normalizing FU of the annealing solution to the “blank” control which contained the probe only in presence of the metal ions.

DNA renaturation tests

The genomic and coding sequences of the wheat ACT1 gene (GenBank accession numbers KC775780 and KC775781, respectively) were amplified by PCR/reverse transcriptase PCR as described previously⁵⁰ and the PCR products were cloned into a pCR2.1 vector using the TOPO TA Cloning Kit (Invitrogen) and sequenced at Cornell University Biotechnology Resource Center. Double-stranded DNA fragments were amplified by PCR using primers given in Supplementary Table S3 and PCR products were purified using the QIAquick PCR Purification Kit (Qiagen). Purified dsDNA fragments were eluted in nuclease-free water with the final concentration adjusted to 100 ng/µl in 10 mM NaCl, 10 mM Tris pH 7.5 and stored at −20°C. For renaturation assay, dsDNA fragments were denatured by boiling for 10 min then chilled on ice for >10 min. Metal ions were then added to the denatured DNA samples (kept on ice) at a final concentration 0–750 µm and incubated for 10 min before allowing to renature at 37°C overnight.

Statistical analysis

Absorbance and fluorescence measurements were done on at least three replicates for each treatment and the readings were normalized to the blank control and further processed using the Gen5 2.0 Software (BioTek) (representative data sets including means and standard deviations are given in Supplementary Table S2) and subjected to t-test with the significant differences between treatments defined at P<0.05.

Sequence analysis

DNA hairpin structures were predicted using the mfold Web Server (http://mfold.rna.albany.edu)⁵¹ with default parameters for linear oligomers except the ionic condition which was set at 10 mM Na²⁺, 1 mM Mg²⁺. Predicted stable folding structures (e.g., ΔG <1) were examined and the representatives were chosen for analysis as shown in related figures. Nucleotide sequences of human chromosome 1 (complete) and wheat chromosome 3B (partial) and the representative transposable elements (TEs) from different species were obtained from GenBank databases. Genes on human chromosome 1 were annotated following on the NCBI annotation release 105 (http://www.ncbi.nlm.nih.gov/projects/mapview). G+C-rich (>60%) islands were identified using the CpG Island Searcher Server (http://www.cpgislands.com)⁵² and human TEs were identified by searching the Dfam database (http://www.dfam.org)⁵³ Sequence analysis and alignment were performed using the Lasergene software (DNASTAR).