A tour de force on the double helix: Exploiting DNA mechanics to study DNA-based molecular machines

Viral DNA packaging

The exact derivation of DNA extension as a function of force makes it possible to convert movement in distance into changes in base pairs, thereby providing exquisite insight into the operating mechanism of protein machineries that translocate on DNA, such as RNA polymerases⁵² and viral DNA packaging motors⁵³. In the latter case, dsDNA genomes are pumped into preformed protein capsids during viral assembly. The ring-shaped packaging motors are among the most powerful molecular machines found in nature. To compact the stiff, highly charged dsDNA to near-crystalline densities into a small capsid, the packaging motor needs to generate a large amount of force—as high as 60 pN—in order to overcome major energetic barriers. The packaging motor of bacteriophage φ29, one of the best-characterized molecular machines thus far, is a homopentameric ring ATPase that packs a 19.3-kbp genome into a capsid 50 nm in height and 40 nm in diameter⁵⁴. Using a dual-trap optical tweezers instrument, the Bustamante group was able to detect discrete DNA translocation cycles of the ring motor (Fig. 2A). Under low external forces (<10 pN), DNA is translocated in 10-bp cycles, each consisting of an ATP-binding dwell phase and a DNA-translocating burst phase (Fig. 2B). Under high external forces (30–40 pN), each 10-bp burst is decelerated and can be shown to be composed of four 2.5-bp steps. Each step is powered by the release of an inorganic phosphate molecule produced from ATP hydrolysis. Subsequent experiments using nucleotide analogs and ATPase mutants mapped specific chemical transitions (such as ATP hydrolysis and ADP release) onto the dwell-burst pathway, revealing intricate, clocklike coordination among the five ring subunits (Fig. 2C)^55–57.

The non-integer 2.5-bp step size is surprising. This result argues against any mechanism in which every motor subunit makes identical chemical contacts with the DNA during translocation. It further suggests that only four of the five subunits participate in DNA translocation per cycle. Indeed, it was shown that the translocating subunits make nonspecific contacts with the DNA during the burst phase⁵⁹, and that the ring symmetry is broken through specific electrostatic interaction with DNA backbone phosphates, bestowing a special regulatory role upon the contacting subunit^{56, 57}.

It was proposed⁶⁰ and experimentally observed⁶¹ that DNA inside the capsid is organized into a spool, which may require rotation of the DNA in order to relieve the torsional strain. In addition, the small difference between the 10.0-bp burst size of the motor and the 10.4-bp helical pitch of B-form DNA entails that the DNA may need to rotate relative to the motor to make crucial electrostatic contacts at the beginning of each cycle. To directly probe DNA rotation during translocation, a third “rotor bead” was introduced to the standard two-bead optical tweezers assay, which allowed angular changes of the DNA around its helical axis to be monitored concomitantly with its linear translocation⁵⁸. A similar setup was previously used to measure the twist elasticity of DNA⁶². The DNA was shown to rotate on average by ~1.5 degrees per base pair in a left-handed direction. Thus, the packaging motor can simultaneously generate force and torque, both of which can reach values high enough to denature DNA. The quantitation of DNA rotation also suggests that the same subunit makes the specific DNA interactions cycle after cycle, thus significantly constraining the possible models for the identity of the special subunit. After a 10-bp burst, the DNA backbone winds by 346 degrees. Thus a 14-degree rotation is required to re-align the DNA with the subunit that makes contacts in the previous cycle, thereby yielding an average rotation density of 1.4 degree/bp.

DNA packaging has long been anticipated to slow down as DNA fills up the capsid due to the mounting internal pressure working against the motor. This was directly proven by the single-molecule packaging assay showing that the velocity of the φ29 packaging motor drops from an initial value of >100 bp/s to essentially zero when the entire genome length is internalized⁶³. Subsequent high-resolution measurements revealed that the internal pressure affects multiple aspects of the mechanochemical cycle of the motor, including slowing ATP binding during the dwell phase and DNA translocation during the burst phase⁵⁸. These results led to an estimation of the final internal pressure to be ~20 atm—a remarkably large number that is consistent with predictions from analytical modeling and numerical simulation studies^64–66. Furthermore, the φ29 packaging motor was observed to take smaller bursts per cycle (10 bp at low filling vs. 9 bp at high filling) and smaller elementary steps (2.5 bp at low filling vs. 2.3 bp at high filling). An accompanying change in the DNA rotation density was also observed (~1.5 degree/bp at low filling vs. ~5 degree/bp at high filling). These concerted adjustments ensure that the distinct functions and coordination of the ring subunits are preserved even in the face of drastically different operating conditions⁵⁸. In addition, a recent simulation study suggested that, instead of being a passive substrate, the DNA itself is an active component of the packaging machinery, driving its own translocation by undergoing cyclic conformational distortions inside the viral portal channel⁶⁷. Overall, this model system showcases how knowledge in DNA mechanics allows for a detailed mechanistic dissection of the 3D trajectories of molecular machines that track on DNA.

DNA replication

Many DNA-based cellular processes involve interconversion between dsDNA and ssDNA. DNA duplex is unzipped into two separate strands by a multitude of helicases or degraded into ssDNA by exonucleases. On the other hand, DNA polymerases copy ssDNA templates into duplexes. The differential elasticity between ssDNA and dsDNA has been exploited to follow the progression of these biochemical reactions in real time without having to fluorescently label the enzymes. This was first used to study bacteriophage T7 DNA polymerase by optical tweezers²⁸ and magnetic tweezers⁶⁸. Interestingly, it was found that at high forces (>40 pN), the nucleolytic activity of the polymerase is greatly stimulated, effectively converting the polymerase into an exonuclease²⁸. Moreover, monitoring the enzymatic behavior as a function of applied tension unveiled intermediate states in the kinetic pathway that are related to the proofreading activity of the polymerase⁶⁹.

van Oijen and coworkers developed a single-molecule DNA flow-stretching assay that can monitor many molecules at the same time, thus affording a higher throughput than the optical tweezers assay. In this setup, bacteriophage λ genomic DNA is anchored to a coverslip surface on one end and conjugated to a micron-sized bead on the other end. The DNA is then stretched by a hydrodynamic flow (Fig. 3A). This setup was first utilized to study the activity of phage λ exonuclease, which converts dsDNA into ssDNA⁷⁰. Later, this setup was applied to the phage T7 DNA replication system⁷¹. Here a bead is attached to the parental end of a forked DNA substrate. At a typical stretching force of ~2 pN, ssDNA is much shorter than dsDNA (Fig. 1B). Leading-strand synthesis results in a shortening of the tether length due to accumulation of the lagging-strand ssDNA (Fig. 3B). In the presence of lagging-strand synthesis, gradual shortening of the DNA followed by sudden lengthening was observed, which was interpreted as formation and subsequent fast release of replication loops in the lagging strand. These looping events were also observed in single-molecule FRET⁷² and magnetic tweezers assays⁷³. Further analysis of the loop sizes and lag times between loops suggested that the initiation of primer synthesis (“signaling” mechanism) and the encounter with a downstream Okazaki fragment (“collision” mechanism) can both serve as a trigger for loop release. Such a dual-trigger mechanism ensures timely reset of the enzymatic apparatus at the replication fork after the completion of each round of Okazaki fragment synthesis⁷⁴. Thus, this simple and elegant assay, chiefly utilizing DNA elasticity, yielded key insights into the coordination between continuous leading-strand synthesis and discontinuous lagging-strand synthesis.

The same group later developed a more sophisticated assay to simultaneously follow T7 leading-strand synthesis and lagging-strand loop formation by attaching two beads to two arms of the forked DNA⁷⁷. This assay revealed that the looping events in the lagging strand contain both ssDNA loops formed during priming (priming loop, more frequent) and ss-ds ones that support Okazaki fragment synthesis (replication loop, less frequent). This study also showed that lagging-strand polymerases are often released from the replisome to complete Okazaki fragment synthesis behind the fork. These results depict a highly plastic replisome that can access multiple reaction pathways to achieve efficient and robust replication.

The DNA flow-stretching assay has also been applied to the E. coli replisome to investigate the processivity of the DNA polymerase III holoenzyme and its regulation by the DnaB helicase and DnaG primase⁷⁸. More recently, it was demonstrated in a eukaryotic system from S. cerevisiae⁷⁵, which utilizes dedicated leading- and lagging-strand polymerases and more regulatory factors. Meanwhile, a number of other single-molecule assays based on fluorescence imaging of stretched DNA (for example, rolling circle and DNA curtains) have also been developed to study various aspects of DNA replication, such as initiation, lesion bypass, and polymerase exchange^79–82.

Force-induced conversion between dsDNA and ssDNA

Using a single-molecule instrument that combines confocal fluorescence microscopy, dual-trap optical tweezers, and automated microfluidics, Wuite, Peterman and coworkers directly visualized the structural transitions of DNA during overstretching³³. They used the intercalating dye YOYO to stain dsDNA that was linked to optically trapped beads via the 3’ end of each strand (Fig. 4A). While the dsDNA molecule was stained along its full length under low forces, unstained segments appeared and grew when the tension was raised above 65 pN (Fig. 4B). These segments preferentially initiated from free DNA ends and internal nicks. To test whether these segments correspond to ssDNA regions, the authors then used fluorescently labeled ssDNA-binding protein (SSB), which wraps around ssDNA⁸⁸. SSB foci were observed at edges of YOYO-labeled dsDNA segments and became brighter as the force increased. These results were interpreted as SSB binding to relaxed ssDNA unpeeled from free ends and nicks. To study overstretching of DNA without preferred nucleation sites, the authors also designed DNA substrates that were linked to beads via both ends of each strand to prevent peeling from free ends (Fig. 4C). Notably, these torsionally constrained constructs displayed an overstretching plateau at much higher forces (~110 pN), in agreement with an earlier report⁸⁹. Single-stranded regions were generated above 110 pN, as indicated by the binding of RPA, another ssDNA-binding protein that can bind to ssDNA under tension (unlike SSB). This observation can be interpreted as the formation of single-stranded “melting bubbles”. Overall, these results demonstrated force-induced melting of dsDNA into ssDNA during overstretching regardless of the DNA attachment geometry.

In a follow-up study⁹⁰, the same group focused on the competition between three processes during overstretching: strand unpeeling, localized base-pair breaking (melting bubbles), and S-DNA formation (strand unwinding with base-pairing maintained) (Fig. 4D). The authors found that all three mechanisms are at work. In topologically closed but torsionally relaxed DNA where free ends and nicks are lacking, melting bubbles form preferentially at AT-rich regions at low ionic strength (Fig. 4E). The same type of construct was also used by another group to show that strand unpeeling is not a requirement for the overstretching transition at 65 pN⁹¹. High ionic strength, by contrast, stabilizes the double helix and inhibits melting-bubble formation. Instead, a different structural form results that cannot by stained by either RPA or Sytox (a dsDNA-intercalating dye). The authors interpreted this form, likely base-paired, as evidence for S-DNA. Increasing ionic strength also suppresses unpeeling and promotes S-DNA formation for topologically open DNA. Thus, the balance between these different processes during overstretching is dependent on DNA sequence, topology, and salt concentration (Fig. 4F). An accompanying paper employed magnetic tweezers to study the same transitions and reached essentially the same conclusion⁹². By investigating the temperature effect, the latter study calculated the entropic contribution from each of the three processes. The unpeeling and bubble-melting transitions are hysteretic with a positive entropy change of 17 cal/(K·mol), similar to thermal melting, whereas the B-to-S transition is nonhysteretic with a small negative entropy change of −2 cal/(K·mol). Moreover, a recent study adopted concurrent fluorescence polarization imaging and force manipulation to show that base pairs in S-DNA are substantially more tilted than those in B-DNA⁹³.

In these types of studies, it is worth keeping in mind the potential perturbation of DNA structure introduced by the intercalating dyes and ssDNA-binding proteins^{94, 95}. Thus, a low concentration of these reagents is recommended whenever possible. It has also been reported that the fluorescence intensity of intercalating dyes can be used as a sensor for the local tension in dsDNA⁹⁶.

Formation of nucleoprotein filaments

Many essential genomic transactions, such as DNA replication, repair, and recombination, involve the generation of ssDNA. Single-molecule studies of these reactions are aided by various experimental strategies to produce long ssDNA templates, such as force-induced duplex melting^{97, 98}. Using combined fluorescence and force spectroscopy, Ha and coworkers showed that SSB can rapidly diffuse along ssDNA, which facilitates its redistribution⁹⁹. The force dependence of SSB movement suggests that it migrates on DNA via intersegment transfer¹⁰⁰.

Besides SSB and RPA, there are other proteins that bind to ssDNA—notably RecA for prokaryotes and Rad51 for eukaryotes. These ATPases form nucleoprotein filaments on ssDNA in an ATP-dependent manner and play critical roles in homology search and pairing during homologous recombination¹⁰¹. In an earlier study, the Bustamante group interrogated the mechanical properties of RecA-ssDNA and RecA-dsDNA filaments at various nucleotide states¹⁰². It was found that RecA significantly stiffens and elongates DNA (both RecA-ssDNA and RecA-dsDNA filaments are ~1.5 times longer than bare B-form DNA). Combining fluorescence imaging and flow stretching, the Kowalczykowski group monitored the nucleation and bidirectional growth of RecA-dsDNA filaments¹⁰³ and examined the inhibitory effect of SSB on RecA-ssDNA filament assembly¹⁰⁴. The same group also investigated the process of homology search by RecA-ssDNA filaments on dsDNA¹⁰⁵. They found that the number of available DNA conformations, which decreases as the end-to-end distance of the target dsDNA is increased, exerts a major influence on the rate of homologous pairing. Using a dual-molecule manipulation setup combining optical tweezers and magnetic tweezers, the Dekker group also studied the process of homology recognition¹⁰⁶. By controlling the supercoiling state of the target dsDNA, the authors found that homologous pairing is strongly enhanced by underwinding of the target DNA, which facilitates transient engagement of the incoming duplex by a secondary DNA-binding site in the RecA filament. Both of these studies pointed to a mechanism that harnesses nonspecific and weak interactions between the RecA filament and target dsDNA for rapid homology search.

Similar to RecA, Rad51 binding also extends the DNA by about 50%¹⁰⁷. The disassembly of Rad51 nucleoprotein filaments occurs at the filament terminus one monomer at a time upon ATP hydrolysis, and is sensitive to tension, stalling at forces above 50 pN¹⁰⁸. By investigating the competition between RPA and Rad51 for DNA binding, Greene and coworkers showed that free RPA in solution inhibits Rad51 filament nucleation, but the elongation of Rad51 filaments on RPA-coated ssDNA is not significantly impacted¹⁰⁹. Interestingly, it was recently reported that RecA/Rad51 polymerizes faster on S-DNA than on B-DNA¹¹⁰, implying potential physiological functions of S-DNA.