Tension induces a base-paired overstretched DNA conformation

Results

Fig. 1A (a) describes the DNA constructs, each consisting of a 60–64 base-pairs double-stranded “kernel” and single-stranded “handles,” the latter formed by terminal transferase extension of the 3′-ends prior to hybridization of the duplex. During the extension either digoxigenin- or biotin-modified nucleotides were incorporated for attachment to polystyrene beads coated with digoxigenin antibodies or streptavidin, respectively. Three kernel sequences were used in this study: OligoGC1 and OligoAT, differing in GC-content (60% and 30%, respectively), and OligoGC2 with the same GC content (60%) as OligoGC1 but differing in nearest-neighbor base composition (Fig. S1). OligoGC1 and OligoAT were designed so that the strands of the kernel could be covalently linked at either or both ends using click-chemistry [Fig. 1A (b–d)]. Here the 5′-clicked OligoAT (5′OligoAT) was studied along with three clicked versions of OligoGC1: 5′-clicked (5′OligoGC1), 3′-clicked (3′OligoGC1), and double-clicked (3′5′OligoGC1). Each inter-strand link connects two bases on opposite strands diagonally from one base to the 5′-neighboring base of its complementary strand. The linker connects the two modified thymine bases at their C5 positions, most likely with the linker residing in the major groove of the helix (Fig. S2). In the experimental setup one bead was held stationary at a pipette tip by suction, while the other was manipulated by an adjustable optical trap (for details on measurement of position and force, see SI Text). Force versus distance curves were recorded by stretching a tethered construct from a lower force limit, F_min, to an upper force limit, F_max, thereafter relaxing it to reduce the force from F_max back to F_min at the same rate. The force limits were set to ensure that the kernel attained B-form at F_min and was in the overstretched state at F_max (for details, see SI Text).

The thermal stability of the sealed oligonucleotides at zero force was assessed from denaturing polyacrylamide gel-electrophoresis varying the temperature between 22 °C and 74 °C (Fig. S3 and Table S1). As expected, by avoiding complete strand separation the clicked DNA exhibits enhanced thermal stability compared to unmodified duplex (Table 1). Thermal melting was studied in 7 M urea to shift melting to lower temperatures as cross-linking significantly increases melting temperature (29). Sealing either end stabilizes OligoGC1 by at least 20 degrees and sealing both ends by another 10 degrees. Sealing the 5′-end of OligoAT also raises the melting temperature by at least 8 degrees.

Table 1.

Overstretching and thermal melting data

Species		n	(pN)	(nm)	(%)	Tm(°C)
OligoGC1	60	49	63.3 ± 1.11	10.4 ± 0.46	49.3 ± 2.2	< 22	14 ± 2	2 ± 3
OligoGC2	60	33	62.0 ± 1.78	11.2 ± 0.57	53.1 ± 2.7	< 22	10 ± 2	−2 ± 3
3′OligoGC1	64	12	63.4 ± 1.36	11.1 ± 0.71	49.5 ± 3.2	44–48	15 ± 3	2 ± 4
5′OligoGC1	64	34	62.1 ± 1.18	11.5 ± 0.87	51.1 ± 3.2	44–48	13 ± 4	0 ± 5
3′5′OligoGC1	64	10	62.5 ± 1.21	11.3 ± 0.74	50.6 ± 3.3	54–57	14 ± 3	1 ± 4
5′OligoAT	64	17	61.5 ± 2.63	14.9 ± 1.24	66.7 ± 5.5	∼30	−2 ± 5	−19 ± 6

Open in a new tab

: number of base-pairs in kernel; n: number of studied molecules; : population average transition force, GC-rich oligonucleotides; : population average melting force, 5′OligoAT; : average extension during transition; : percentage average extension relative to B-form DNA at F_tr or F_m (see SI Text); T_m: melting temperature (in absence of force) from 7 M urea PAGE; : number of remaining base-pairs in peeling melting model of Fig. 1B (i), with tension in melted region carried by one strand; : number of remaining base-pairs for internal melting model of Fig. 1B (ii), with tension in melted region carried by both strands. Statistics reported as mean ± 1 s.d. All force-distance data refer to 1 M NaCl.

Fig. 2A shows a force-distance curve during a pull-relax cycle for (nonclicked) OligoGC1, illustrating the typical behavior when force is increased up to 70 pN and thereafter reduced. For both processes the optical trap is moved at a constant rate (50 nm/s). OligoGC1 exhibits a force-dependent bistability showing oscillations between the B-form and an overstretched state. Notably, in the trap-position control mode used here, a transition to the extended state is detected as an abrupt decrease in the force as the bead relaxes towards the center of the trap. The reverse transition back to the B-form increases the force as the contracting molecule pulls the bead away from the trap center. As seen from the inset of Fig. 2A, the molecule switches back and forth between the two states during both pulling and relaxation of the construct. The transition appears completely reversible with no detectable hysteresis. Fig. 2D shows a time trace of force when the optical trap is stationary. The OligoGC1-molecule is seen to oscillate between B-form (high force) and overstretched state (lower force) on a millisecond time scale. The histogram of the force data is well described by a double Gaussian fit, justifying the analysis of the mechanical overstretching as a two-state process. The data points are assigned to either of the two states depending on their proximity to fitted lines at low and high force, respectively (Fig. 2B). Fig. 2C shows the probability P_stretched of the overstretched state as a function of force obtained from 16 pull-relax cycles on the same molecule. The transition force F_tr is the force at which P_stretched equals 0.5, and the extension during the transition, Δx, is calculated by dividing the force difference between the two fitted lines at F_tr by the slope of the B-form fitted line (33, 34). Based on data from 49 OligoGC1 molecules, a mean transition force of (± 1 s.d.) was obtained, and for OligoGC2 (n = 33) . Corresponding extensions were 10.4 ± 0.46 nm and 11.2 ± 0.58 nm, respectively. Data are summarized in Table 1.

Fig. 3A shows typical force-distance data for 5′-clicked OligoGC1 (Scheme b, Fig. 1A) during one pull-relax cycle. During pulling (blue) the 5′-sealed duplex is seen to undergo the same force-induced oscillation between two states as the unmodified OligoGC1 (Fig. 2A). The 3′- and double-clicked OligoGC1 (c and d in Fig. 1A) exhibit virtually identical behaviors (see Fig. S4). Importantly, compared to nonclicked OligoGC1, our results (Table 1) show that within experimental uncertainty the overstretching transition occurs at the same force whether the ends are sealed or not. The transition lengths are also similar for clicked and nonclicked GC-rich DNA with an extension per base-pair of 0.17–0.19 nm.

In addition to the reversible transition behavior, the 5′-clicked OligoGC1 is also found to undergo a secondary, irreversible extension step at higher force at which the duplex melts. Fig. 3A (insets) shows a force-extension curve for a case when a 5′OligoGC1 molecule is first reversibly overstretched (inset i, blue curve) and then further extended by an additional 3.6 nm at about 74 pN (inset ii, red curve) when it melts during relaxation. The transition is irreversible in the sense that the reverse process, molten strands rehybridizing to form an overstretched state, does not occur. In order to regain the duplex, on the time scale of our experiment, the force had to be lowered to about 20–25 pN before the two strands could reanneal into B-form. Experiments with double-clicked 3′5′OligoGC1 (Fig. S4D) showed no melting behavior even at higher forces (120 pN), supporting the idea that cross-linking of both ends inhibits melting. Control experiments with 3′OligoGC1 revealed no melting (Fig. S4B), and the contrasting behavior compared to 5′OligoGC1 (Fig. 3A) indicates an asymmetry in the OligoGC1 sequence regarding force-induced melting. The nonclicked constructs were frequently observed to dissociate at forces close to or above the overstretching region, but it is hard to discern whether they are falling apart due to strand separation in the kernel or due to loss of a bead anchor, since in both cases the tether between the two beads is lost. These observations illustrate the value of single-clicked constructs in studies of forced-induced melting, since the covalent linker prevents even a fully molten kernel [Fig. 1B (iv)] from undergoing complete disassembly.

Fig. 3B shows AT-rich 5′-clicked OligoAT during a cycle of pulling and relaxation. At 62 pN, the molecule undergoes an irreversible melting transition. The melting process gives rise to a distinct hysteresis during relaxation (red curve) and the construct does not reanneal until force is lowered to about 23 pN (inset ii) when full B-form duplex recovers. Thus, the behavior of AT-rich 5′OligoAT is completely different from that of the GC-rich DNA. Table 1 shows ensemble-averaged forces F_m, at which melting occurs for 5′OligoAT, and increase in length during transition. The average degree of extension is about 67% for the AT-rich duplex; i.e., significantly higher than the 51% for GC-rich duplexes (Table 1), and closer to the 70% reported for long mixed-sequence DNA. Prior to the abrupt transition at F_m, the 5′OligoAT in Fig. 3B exhibits a brief bistability involving an intermediate state (inset i) seen in most, although not all, molecules studied. The intermediate state is on average 7.5 ± 0.7 nm (n = 12) longer than the B-form state, which corresponds to a relative extension of 34 ± 3%. This observation suggests that constructs less stable towards melting also may undergo a more transient reversible conversion to an overstretched state before melting.

Fig. S5 A–C show that when the salt is lowered from 1 M NaCl to 5 mM NaCl, the reversible transition is still observed for 3′OligoGC1 and 3′5′OligoGC1, whereas 5′OligoGC1 melts in a way similar to what was observed for the AT-rich 5′OligoAT in 1 M NaCl. At 5 mM, all measured 3′OligoGC1 melted from the stretched state of the reversible transition, whereas none did at 1 M NaCl. No double-clicked 3′5′OligoGC1 were observed to melt at 5 mM NaCl, confirming that cross-linking inhibits melting. The results indicate that lowering the salt concentration decreases the melting force, but also that high salt concentration is not a requirement for the reversible overstretching transition.

To examine the degree of base-pairing in the two different overstretching transitions, we exposed the 5′-clicked AT-rich oligonucleotide (5′OligoAT, n = 6) as well as double-clicked GC-rich construct (3′5′OligoGC1, n = 7) to 0.5 M glyoxal. Fig. 4A shows a 5′OligoAT molecule before and after the bases are exposed to react with glyoxal present in the medium. The blue curve shows the first pull cycle before glyoxal modification (t₀), where the unmodified construct melts at about 60 pN (inset i) just as it would in the absence of glyoxal. After the molecule melts, the force is reduced to about 40 pN and held at this tension to allow any exposed guanine bases of the molten strands to react with glyoxal. The force is then reduced to 20 pN to allow rehybridization. Within a minute of glyoxal exposure, the melting force is markedly reduced () due to base modifications in the rehybridized duplex. Additional exposure lowers the melting force even further. The double-clicked GC-rich 3′5′OligoGC1 was exposed to glyoxal in its reversible overstretched state by holding it at 70 pN in presence of the chemical agent. Fig. 4B shows the DNA before and after a long exposure time. The inset indicates that the reversible transition remains unaffected after exposure, indicating that the bases in the reversible overstretched state are indeed not accessible to reaction with glyoxal. Control experiments in which the single-clicked GC-rich 5′OligoGC1 is exposed to glyoxal while the DNA is in its molten state (Fig. S5D) confirm that the same sequence reacts with glyoxal if the strands are separated.

Discussion

The reversible transition observed for OligoGC1 at 63 pN (Fig. 2, Table 1) proves that a native DNA duplex as short as 60 base-pairs can be mechanically overstretched by 51% compared to its B conformation without strand separation. Four observations in particular prove that the stretched state of the reversible transition is not a molten helix but a structurally and thermodynamically well-defined form of DNA.

(i) While the melting temperature (T_m) at zero force increases when the ends of the duplex are prevented from fraying by click-seals, the four OligoGC1 variants (nonclicked, 5′-clicked, 3′-clicked, and double-clicked) all exhibit the same bistable behavior and approximately the same transition force and extension. The difference in behavior observed here between melting and stretching suggests that the reversible overstretching in the GC-rich duplex is not related to thermal stability.

(ii) A valid melting model must also be able to explain why nonsealed OligoGC1 survives forces above the overstretching region without falling apart. The possibility that the duplex only partially melts during the transition, DNA held together by regions that remain base-paired, can be dismissed based on the following analysis of overstretching length. The number of base-pairs (N_ds) that would be left paired upon melting can be estimated as:

[1]

where is the total number of base-pairs in the kernel, b_ss(F,T) the rise per base in the resulting single strand at the transition force F_tr or F_m, respectively, and temperature T; and b_ds(F,T) is the rise per base-pair of the remaining double-stranded regions at the same force and temperature (see SI Text for details). Table 1 presents the calculated number of remaining base-pairs for the melting models i and ii in Fig. 1B. The main difference between the two models is whether the tension in the molten region is supported by one (i) or both (ii) strands. Assuming that the putative base-paired fraction remains in a B-type structure, numerical estimates in Table 1 indicate that there would be no base-pairs left to hold the strands in register in the internal melting scenario of Fig. 1B (ii). Even if the duplex melts from the ends [peeling scenario Fig. 1B (i)], only 10–15 base-pairs in total would remain, a duplex too short to be stable when exposed to 63 pN at the temperature and ionic conditions of our experiments (35, 36). The same analysis for 5′OligoAT yields -19 remaining base-pairs for an internal melting scenario and -2 remaining base-pairs for a peeling scenario. These small (even negative) numbers contradict a partial melting model, and further support our conclusion that the irreversible transition leading to hysteresis is indeed a consequence of full melting of the double stranded region, as no base-pairs would remain to hold the duplex together. The result for the peeling scenario agrees with the expected response of a fully melted single-clicked structure, Fig. 1B (iv), where only the linker holds the construct together.

(iii) The behavior of 5′OligoAT (30% GC) is quite different from that of the GC-rich OligoGC1 and OligoGC2. The hysteresis for 5′OligoAT (Fig. 3B) pulled at the same loading rate as the GC-rich oligonucleotides strongly indicates that the duplex melts under force. This difference in behavior between the GC-rich and AT-rich oligonucleotides at 1 M NaCl demonstrates that sequence is an important parameter controlling the probability of transition from B-form to stretched and/or melted forms. Equally important, this different behavior also implies that the reversible transition observed for GC-rich oligonucleotides is not due to melting. The 5′-clicked OligoGC1 occasionally exhibits hysteresis (Fig. 3A), but then only after a secondary, irreversible transition. This observation provides independent evidence that the reversible overstretching seen in GC-rich duplexes (Fig. 2) and the force-induced melting seen in the AT-rich 5′OligoAT (Fig. 3B) are two completely different processes. They may occur in the same DNA molecule, but are different in origin. Measurements in 5 mM NaCl (Fig. S5 A–C) show that lowering salt concentration lowers stability towards melting, but our results show that the reversible overstretching also occurs at low salt. Note, furthermore, that 5′OligoGC1 in 5 mM NaCl melts without displaying reversible overstretching (Fig. S5C), similar to what was observed for the AT-rich 5′OligoAT in 1 M NaCl. These results imply that it is not the sequence per se that determines whether or not reversible overstretching occurs, but rather the stability of the sequence towards melting. If the stability is low enough, by high AT-content or low salt, for example, the overstretched state will be short-lived and melt.

(iv) Glyoxal probing of nonbase-paired guanines shows that the AT-rich duplex melts but that GC-rich duplexes remain base-paired when overstretched. Refolding of overstretched 5′OligoAT exposed to glyoxal for less than a minute is partially inhibited, and progressively more so as the exposure time is increased to near 3 min (Fig. 4A), showing that a substantial number of guanines have unpaired and reacted. At the same time, the force needed to melt the construct is gradually lowered as more and more glyoxal-modified guanine bases are introduced. By contrast, exposure of overstretched double-clicked 3′5′OligoGC1 to glyoxal for similar or even longer times does not affect the reversible overstretching transition (Fig. 4B). Taken together, these observations strongly support our conclusion that 5′OligoAT melts under force, leaving the guanine bases exposed to glyoxal, whereas the overstretched GC-rich oligonucleotides keep the majority of bases in an overstretched base-paired state, as originally suggested for the S-form DNA (1, 2).

The transition forces that we measure for reversible overstretching of GC-rich duplexes and for the irreversible melting of the AT-rich duplex are similar to forces observed with long natural DNA, indicating that the stretching transitions observed here are also present in those seen with longer molecules. Thus, it is justified to compare the processes despite the several orders of magnitude difference in molecular length. Fu et al. (13, 14) noted that overstretching could involve both a slow transition—associated with hysteresis as one strand peels off from the other strand—and a fast, reversible transition to what may be S-form DNA. The two types of mechanical processes observed in our synthetic oligonucleotides are thus also likely to be represented in the overstretching behavior of naturally occurring, longer DNA molecules. An important distinction, however, is that the two processes are accompanied by a very distinct extension change. The reversible overstretching seen in the GC-rich duplexes extends the molecule by only 50–53% (Table 1), whereas the melting AT-rich duplex at about the same force and for the GC-rich duplex at higher force, extends it by about 67%, much closer to the value typically reported for long natural DNA (70%). In the GC-rich duplexes, the reversible overstretching and force-induced melting are easily distinguishable as separate processes in the same DNA molecule (Fig. 3A): the reversible overstretching transition manifests itself as a simple two-state process (Fig. 2), whereas the force-induced melting is seen as an irreversible process that extends the DNA molecule further.

In the present study, it is shown that the mechanical behavior of short DNA (at 1 M NaCl) is mainly controlled by the overall percentage of GC base-pairs in the sequence. Interestingly, an approximately 50% extension of DNA is also observed upon interaction of recombinases RecA and (human) Rad 51 (37, 38). Based on flow linear dichroism and small-angle neutron scattering data in solution (39), showing that the DNA bases are oriented with their planes perpendicular to the fiber axis of RecA-DNA complexes, some kind of heterogeneous stacking was early proposed and later supported by modeling for these complexes (1, 10). Recently, in crystal it was found that the extension of DNA in the RecA complex is inhomogeneous, three base-pairs stacked together followed by a gap into which an intercalated residue from the protein is inserted (37). Hence, an interesting question is whether the value of 50% extension is a mere coincidence or if the conditions inside the filamentous recombinase complexes may stabilize the same kind of extended base-paired DNA structure that the molecule adopts when subjected to external force applied to its ends. Another question is whether the reversible extension of free double-stranded DNA is homogeneous or could undergo a “disproportionation” (analogous to Peierls instability), yielding a structure in which clusters of stacked nucleobases are surrounded by gaps. The hydrophobic environment of stacked bases is a prerequisite for the hydrogen bonds to exert their full recognition power. One may envisage two ways by which bases can remain stacked in a more extended structure: the bases may either remain orthogonal or tilt. In both cases, the stacking energy (proportional to the contact surface area between the bases) must decrease, thus making the extended DNA less stable, in the orthogonal case due to the occurrence of gaps and in the tilted base arrangement due to a smaller overlap area between adjacent bases (SI Text, Eqs. S5–S8). We note that structures in which the nucleobases are perpendicular to the DNA helix axis, with stacking heterogeneity as found in RecA-DNA with stacked base triplets, are logical for the function involving base recognition as in recombination.

Materials and Methods

Unmodified oligonucleotides were obtained from IDT; US and clickable oligonucleotides from ATDBio, UK (see SI Text, Fig. S6 and Table S2). The terminal transferase (New England Biolabs) reactions were performed using a dUTP-biotin/digoxigenin (Digoxigenin-11-dUTP, Biotin-16-dUTP, Roche) to dATP ratio of 1∶10. The covalent link was formed by the Cu⁺ catalyzed click-reaction (0.01eq oligonucleotide, 1eq CuSO₄·5H₂O, 7eq tris-hydroxypropyltriazole, 10eq Na ascorbate, overnight at 22 °C) (29, 40, 41). Streptavidin beads with a nominal diameter of 2.1 μm were used as obtained from Spherotech. AntiDIG-beads of the same size were prepared in-house by cross-linking antidigoxigenin polyclonal antibodies (Roche) onto proteinG-modified polystyrene beads from Spherotech using Dimethyl pimelimidate (Sigma). Measurements were performed in a buffered high (1 M NaCl) or low (5 mM NaCl) salt environment (10 mM Tris pH 7.4, 1 mM EDTA). In the glyoxal exposure experiments, 0.5 M glyoxal (Sigma-Aldrich) was included in the high salt buffer, and the pH was adjusted to 7.9. Force-distance data were sampled at a frequency of 1 kHz. The temperature was measured within the instrument close to the fluidics chamber. The mean temperature for all measurements was 22.3 ± 0.9 °C (± 1 s.d.).

Supplementary Material