Long-lived pauses reveal tunable kinetic barriers during Holliday junction migration

Abstract

Holliday junctions (HJs) are key intermediates in homologous recombination, and their branch migration influences how recombination intermediates are positioned and processed. Although structural studies have shown that HJs interconvert between open and stacked-X conformations, and the stacked-X state has long been proposed to inhibit migration, it remains unclear whether such pausing creates a substantial kinetic barrier to long-range spontaneous branch migration. Here, we develop a magnetic-tweezer single-molecule assay that tracks spontaneous HJ migration over extended distances ( bp) immediately after junction formation. We find that migration is discontinuous, alternating between active movement and long-lived pauses lasting seconds to tens of seconds. Focusing on pauses of at least 1 s, we show that Mg strongly increases both pause duration and pause frequency, whereas higher temperature and applied tension suppress long-lived pausing. Under high Mg, low temperature, and low force, these pauses dominate total migration time and reduce effective migration several fold, establishing them as a major kinetic barrier. In addition, nucleoid-associated protein HU increases long-pause frequency and pause occupancy under conditions where intrinsic pausing is weak. Together, these results show that spontaneous HJ migration is governed by a tunable long-lived paused state consistent with a stacked-X-like migration-inhibitory conformation.

Graphical Abstract

Graphical Abstract.

Introduction

Homologous recombination is a central pathway for the repair of DNA double-strand breaks and for the generation of genetic diversity during meiosis [1–4]. A key intermediate in these processes is the Holliday junction (HJ), a four-stranded DNA structure formed when two homologous duplexes exchange strands [5]. The migration and resolution of HJs must be tightly controlled to ensure accurate chromosome segregation and long-term genome stability [6–8]. Because the position and fate of an HJ can influence crossover outcome, regulation of HJ dynamics is fundamental to both genome maintenance and genetic diversification.

HJ processing occurs within a multilayered protein network that directs recombination intermediates toward crossover or non-crossover outcomes. In bacteria, RuvAB drives branch migration and RuvC resolves junctions, whereas RecG provides an alternative motor and remodelling activity [9–11]. In eukaryotes, FANCM/Mph1 and the BLM–TOP3A–RMI1/2 dissolvasome remodel or dissolve HJs, while structure-selective nucleases including MUS81–EME1, SLX1–SLX4 and GEN1/Yen1 promote resolution when appropriate [12–15]. Such extensive enzymatic control suggests that spontaneous, diffusion-driven branch migration is unlikely to be entirely unconstrained, as unrestricted migration would erode positional information prior to regulated processing.

Structural and biochemical studies have long shown that HJs are conformationally dynamic [16–18]. Two principal conformations are commonly described: an open, square-planar form in which the four arms are splayed, and a compact stacked-X form in which pairs of arms undergo coaxial stacking to generate two quasi-continuous helices [19–22]. Divalent cations such as strongly favour the stacked-X conformation, as reflected by increased lifetimes of this state at higher concentrations in single-molecule Forster resonance energy transfer (smFRET) measurements [23–25]. Because coaxial stacking is expected to stabilise a compact junction geometry, the stacked-X state has long been proposed to inhibit spontaneous branch migration, whereas the open form is thought to be more permissive for base-pair exchange and enzymatic processing [17, 23–27].

Consistent with this view, bulk biochemical studies have shown that average branch-migration rates are highly sensitive to experimental conditions, increasing with temperature and decreasing strongly in the presence of Mg [28, 29]. However, ensemble measurements report only average behaviour and cannot resolve how frequently migration is interrupted, how long such interruptions persist, or how strongly they contribute to the effective migration rate. smFRET studies later identified brief pauses in the millisecond to sub-second regime during short-range migration, consistent with transient formation of the stacked-X state [23, 24]. Yet smFRET primarily captures local conformational transitions and short migration events, and is further limited by photobleaching. As a result, the relationship between pause statistics and effective branch migration over extended distances remains poorly defined.

A further gap concerns the role of physiologically relevant mechanical and DNA-binding factors. In vivo, HJs are embedded within chromosomes that are subject to tension and complex mechanical constraints. During meiosis, homologous chromosomes undergo programmed double-strand break formation, synapsis, telomere-led movements, and organization along proteinaceous axes [30–34]. Recombination intermediates can therefore mechanically couple homologs and are likely to experience forces across the junction. However, how such mechanical tension influences spontaneous branch-migration kinetics, and how these effects interplay with ionic conditions, remain largely unexplored.

In addition, the impact of DNA-binding proteins on HJ migration dynamics remains poorly understood. For example, HU, an abundant bacterial nucleoid-associated protein (NAP) in Escherichia coli, is known to bind both double-stranded DNA (dsDNA) and HJs [35–37]. However, whether such interactions influence spontaneous HJ migration has remained unclear. More broadly, the extent to which architectural DNA-binding proteins can modulate the intrinsic kinetics of branch migration has not been quantitatively characterized. Addressing this question could provide useful insight into how the local protein environment shapes the behaviour of recombination intermediates.

Here, we developed a magnetic-tweezer single-molecule assay that directly monitors HJ migration over hundreds of base pairs under defined force, ionic conditions and temperature immediately after junction formation. Using this approach, we find that spontaneous HJ migration is not continuous, but instead alternates between progressive migration and long-lived paused states. Because pauses in the second-timescale regime are more plausibly compatible with protein engagement than very brief interruptions [38, 39], we focused our main comparative analysis on pauses lasting at least 1 s.

We show that Mg strongly enhances this long-pause regime, whereas elevated temperature and applied tension suppress it. Under some conditions, these pauses account for a substantial fraction of the total migration time, demonstrating that they form a major kinetic barrier to effective branch migration. We further show that the NAP HU increases the occurrence of long-lived pauses under conditions in which intrinsic pausing is otherwise weak, indicating that protein interaction can reshape the pause landscape of spontaneous HJ migration. Together, these results establish spontaneous HJ migration as a chemically and mechanically tunable process governed by a pause-mediated barrier that can be further reinforced by external protein binding. The observed behaviour is consistent with stabilization of a stacked-X-like migration-inhibitory state.

Materials and methods

Single-molecule constructs

All DNA oligonucleotides used in this study were custom-synthesized by Integrated DNA Technologies (Coralville, Iowa) as listed in Table 1. The designed dsDNA construct for HJ formation was assembled using standard polymerase chain reaction (PCR), digestion, annealing, ligation, and purification protocols.

Table 1.

DNA oligonucleotide sequences used in this study

Oligo ID	Sequence (5 3)
T_Han_HJ	/5Phos/AATAAATTTAAGCAG
T_Handle_Thiol	/5ThioMC6-D/CGAATGATGTTAAATTTATTCATA
HP1_Bst_CATA	TTAACCACCATAATGGTGGGTCAGGTTGTTCTTTAGG
HP1_Bst_AAAA	TTAACCACAAAAATGGGCTCCCGGATTCAGTGATTTA
HP2_Bst_TTTT	TTAACCACTTTTATGGGCTCCCGGATTCAGTGATTTA
HP2_Bst_GTGG	TTAACCACGTGGATGGTGGGTCAGGTTGTTCTTTAGG
GCrich_Bst_CCAC	TTAACCACCCACATGGGAAGCGGTGATGCTGATAGA
GCrich_Bst_CAGC	TTAACCACCAGCATGGGATGTCAGCGCAGTCAAATG
B_Tag_bio	/5Phos/GCGCCTCGCGGCTCCGCCGGCTCCGCGCTGCCCGCG/3Bio/
B_Tag_bio_CAGC	/5Biosg/CGCGGGCAGCGCGGAGCCGGCGGAGCCGCGAGGCGCCAGC

The complete construct consists of five components: a 5 thiol-labeled top handle, inverted sequence 1, inverted sequence 2, a GC-rich bottom handle, and a 3 biotin-labeled bottom strand. Each DNA fragment was designed with unique sticky ends to ensure specific and efficient assembly into the intended configuration:

Thiol-labeled top handle inverted sequence 1 inverted sequence 2 GC-rich bottom handle biotin-labeled bottom strand

The thiol-labeled top handle and biotin-labeled bottom strand were generated by annealing the respective single-stranded DNA oligos with their complementary partners (e.g., ‘T_Handle_Thiol’ and ‘T_Han_HJ’ in Table 1). The 5 thiol tag enables attachment of the construct to magnetic beads, while the 3 biotin tag facilitates surface immobilization. Annealed products were aliquoted and stored at C.

The two inverted sequence fragments and the GC-rich bottom handle were amplified by PCR from -DNA template using the primers listed in Table 1 (HP1_Bst_CATA, HP1_Bst_AAAA, HP2_Bst_TTTT, HP2_Bst_GTGG, GCrich_Bst_CCAC, GCrich_Bst_CAGC). Each primer included a recognition sequence for the restriction enzyme BstXI, which was incorporated into the PCR products. Following PCR amplification, products were purified using a PCR purification kit (Invitrogen), digested with BstXI to generate sticky ends, repurified, and stored at C.

Finally, all five DNA fragments were combined and ligated using T7 DNA ligase. The ligation mixture was resolved by agarose gel electrophoresis (Supplementary Fig. S3), and the band corresponding to the full-length construct (1300 bp) was excised and purified using a gel extraction kit (Invitrogen). The purified construct was stored at C until use.

Single-molecule manipulation

Single-molecule measurements were carried out on a custom-built back-scattered vertical magnetic tweezers instrument with a spatial resolution of 1 nm and a temporal resolution of 200 Hz [40]. DNA constructs were immobilized on the coverslip through a biotin–traptavidin linkage, while the opposite end was tethered to an epoxy-coated superparamagnetic bead via a thiol–epoxy bond. All experiments were performed in a laminar flow channel. Force calibration was performed using the overstretching transition of dsDNA, assumed to occur at 65 pN. Temperature control during experiments was achieved with an objective heater (Bioptechs) applied directly to the objective lens.

Hidden Markov model for pause–migration detection

To segment HJ extension traces into pause and migration states, we implemented a custom analysis pipeline in Python using a hidden Markov model (HMM). Raw extension–time data from magnetic tweezer experiments were first smoothed with a centered rolling mean, then divided into fixed-width time bins in which local slopes of extension versus time were estimated by linear regression. A control dataset containing pause-only traces under the same experimental condition was used to parameterize the Gaussian distribution of slopes associated with the paused state, which served as the initialization for the HMM. Experimental traces were then processed to calculate standardized slope features, capped at the 99th percentile to minimize the influence of outliers. A two-state Gaussian HMM (pause versus migration) was initialized with the control-derived pause distribution and a separated migration distribution, and iteratively optimized by expectation–maximization. The trained HMM assigned each time bin to pause or migration, with state probabilities used to generate sample-level annotations (Supplementary Fig. S1). Post-processing filters removed artifactual pauses shorter than 0.99 s and excluded boundary segments (Supplementary Fig. S2), while migration events were characterized by their start/end times, duration, and net extension change. This approach enabled unbiased, probabilistic identification of dynamic transitions between paused and migrating states across single-molecule traces.

HU protein preparation and measurements

Recombinant HU protein was purchased from Abcam (Cambridge, UK; catalogue no. ab225611). The protein was handled according to the manufacturer’s instructions and stored in small aliquots at −20 C to avoid repeated freeze–thaw cycles. Immediately before use, an aliquot was diluted into the corresponding experimental buffer to the desired final concentration.

For HU measurements, the protein was introduced into the flow cell at a final concentration of 100 nM in the same base buffer used for the matched control condition (150 mM NaCl, 10 mM Tris, pH 7.4, with MgCl, temperature, and force as indicated in the relevant figure legends). After buffer exchange, the tether was allowed to equilibrate briefly before data acquisition. Control experiments were performed under identical solution and force conditions in the absence of HU. All HU-treated traces were analysed using the same HMM-based workflow and pause-definition criteria as the protein-free controls. For the baseline Mg, temperature, and force datasets, the tether was treated as the unit of replication for summary statistics and standard error of the mean (SEM), although individual repeated migration cycles are shown in the plots. For the HU dataset, because measurements were obtained from repeated cycles on a single tether, individual traces were treated as the unit of analysis for calculation of summary statistics and SEM.

Results

Development of a single-molecule assay to track long-distance HJ migration

An HJ is a four-way DNA junction formed between a pair of homologous duplexes that can undergo branch migration. In palindromic dsDNA, extrusion of a DNA cruciform generates two opposing hairpins whose crossover region is an HJ, so that the cruciform can be viewed as a specific realization of an HJ embedded in a longer duplex [41, 42]. Throughout this work, we refer to this central crossover simply as the HJ. Previous studies have shown that diverse DNA secondary structures, including cruciforms and four-way junctions, can be engineered for single-molecule manipulation either by sculpting structures from a single dsDNA molecule using force-induced structural transitions and sequence design [43] or by employing a ligation-free, multistep assay that assembles a wide range of DNA, RNA, and RNA–DNA hybrid configurations [44]. Building on these strategies, we developed a method to repeatedly generate a migrating HJ from a palindromic dsDNA substrate, enabling us to track HJ migration dynamics over an extended distance (290 bp).

The construct for the HJ migration assay is designed to enable tracking the migration of an HJ upon its formation over a long migration distance and duration using a sequence-specific approach. The HJ is generated through the force-dependent overstretching transition of an initial dsDNA at 65 pN [45]. Each strand of the dsDNA contains a 290-bp inverted sequence (61% AT, marked in red and blue, Fig. 1B) and is flanked by two dsDNA handles: a short, AT-rich (80% AT, 15-bp) top handle and a long, GC-rich (38% AT, 686-bp) bottom handle. This design ensures that during the overstretching transition, the bottom handle undergoes a B-to-S transition to S-DNA, where the strands remain base-paired, while the top handle and the inverted sequence region undergo strand separation [46–49]. Following this transition, a 290-bp hairpin forms in the dissociated relaxed strand. When the force is reduced below 13 pN, the same hairpin forms on the force-bearing strand, enabling the re-annealing of the separated top-handle strands back into dsDNA. This forms an HJ at the centre of the four-way junction (Fig. 1B).

Figure 1.

Single-molecule assay design for HJ formation and migration. (A) Schematic of two limiting HJ conformations: an open planar conformation that is permissive for migration and a stacked-X-like conformation that is consistent with migration inhibition. (B) Experimental cycle used to generate and monitor HJ migration in a magnetic-tweezers assay. A palindromic DNA construct is first overstretched at high force to induce strand separation within the inverted-repeat region. Upon force reduction, intramolecular hairpins form and re-annealing of the top-handle strands generates a four-way junction. The resulting HJ then undergoes force-biased branch migration, which is monitored as a change in bead height. The cycle can be repeated multiple times on the same tether.

Upon its formation, the HJ can undergo spontaneous migration in two directions: vertical and horizontal. If the migration occurs in the vertical direction, the hairpin structure progressively shrinks, while the top and bottom dsDNA regions elongate, causing progressive increase in the bead height. Conversely, migration occurring in the horizontal direction results in progressive decrease in bead-height. Due to the biasing force exerted in the vertical direction, the HJ would undergo directional branch migration in the vertical direction, increasing the bead height. In our experiments, forces of a few pN are applied, therefore the HJ migration is biased toward the vertical direction.

As illustrated in Fig. 2, immediately after HJ formation, the microbead reaches its lowest height relative to the coverslip surface. Subsequent biased vertical migration of the HJ leads to a progressive increase in microbead height until the tether returns to its original dsDNA state under the same applied force. This configuration allows continuous tracking of HJ formation and migration over long distances, corresponding to ∼290 bp of strand exchange. Once the tether reverts to the dsDNA state, the same procedure can be repeated to regenerate the HJ on the same DNA molecule, enabling multiple measurement cycles from a single construct. The experimental data obtained in this study under a specific condition were typically recorded from 4–6 independent tethers, each for 4–8 cycles.

Figure 2.

Representative magnetic-tweezers traces reveal repeated HJ formation and pause-and-run migration. Representative bead-height traces recorded during repeated cycles of HJ formation and migration. Junction formation is marked by a drop in bead height at constant force, followed by progressive height recovery during force-biased branch migration. Schematics illustrate the corresponding structural states during the experimental cycle. The migration trajectories display a characteristic pause-and-run pattern, with alternating periods of active migration and dwell-like interruptions.

HJs exhibit dynamic switching between migration and pause states

A prominent feature of the HJ migration trajectories is a characteristic ‘pause-and-run’ pattern (Fig. 2). In contrast to the brief pauses in the millisecond to sub-second regime reported in previous smFRET studies [23, 24], our measurements reveal longer dwell periods on the order of seconds. These observations extend earlier short-range measurements by identifying a longer-lived pause regime and show that spontaneous HJ migration over extended distances is not continuous, but instead alternates between progressive migration and interrupted states.

To quantitatively identify pause events, we applied an unsupervised HMM to each migration time course (see the ‘Materials and methods’ section and Supplementary Text). This analysis resolved two distinct kinetic states: paused phases, marked by orange intervals, and migratory phases, marked by green intervals (Fig. 3). Pause durations were obtained directly from the lengths of the orange intervals, allowing systematic quantification of pause lifetimes across conditions.

Figure 3.

HMM segmentation of HJ migration traces into pause and migration states. Representative HJ migration trace analysed using a two-state HMM. Raw extension data are shown together with the inferred pause state (orange) and migration state (green). The schematic annotations define the quantities extracted from each trace, including pause duration, pause frequency, pause-time ratio, net migration rate, and overall migration rate. Pauses shorter than 1 s were merged into the migration state during post-processing for the main analysis.

This analysis yielded several complementary measures of HJ migration dynamics, including pause duration distributions, pause frequency (number of pauses per migrated base pair), the fraction of total migration time spent paused, net migration rate during active phases (including short-lived pauses), and overall migration rate for each trace. Together, these metrics allowed us to separate changes in pausing behaviour from changes in effective branch migration.

For subsequent comparative analysis, we focused on HMM-identified pauses lasting at least 1 s. This threshold was used as a biologically motivated analysis window for long-lived pauses rather than as an absolute physical boundary between pause classes (see the ‘Discussion’ section). Sensitivity analysis showed that the major condition-dependent trends were similar with a 0.5 s cutoff, whereas a 2 s cutoff was substantially more restrictive and excluded a larger fraction of pause events (Supplementary Fig. S4).

To test whether the observed pauses arose from local sequence-dependent barriers, we examined the distribution of pause positions across repeated migration events from five different tethers. Pause locations were broadly distributed along the migration coordinate, with no pronounced hotspot enrichment at the resolution of this analysis (Supplementary Fig. S5). This argues against a dominant sequence-specific barrier and instead suggests that the pauses primarily reflect an intrinsic feature of junction dynamics.

Overall, the data are consistent with stochastic switching between a migration-competent state and a migration-inhibited state, as introduced in the Introduction. A stacked-X-like conformation provides a plausible structural interpretation for the latter, although the present assay reports on migration dynamics rather than junction structure directly. In this framework, the overall migration rate, defined as the total migrated distance divided by the total migration time, depends not only on the migration velocity during active phases but also on how often the junction enters the paused state and how long it remains there.

Environmental and mechanical conditions tune the long-pause regime of HJ migration

We next examined how external conditions reshape the long-pause regime of HJ migration by varying magnesium concentration, temperature and applied force. For consistency with the biologically emphasized analysis framework introduced above, the main comparisons in this section are based on pauses lasting at least 1 s.

We first tested the effect of magnesium under otherwise constant conditions (150 mM NaCl, 10 mM Tris, pH 7.4, 23 C,  pN; Fig. 4A–C). Increasing the Mg concentration strongly promoted long-lived pausing. Mean pause duration increased by approximately four-fold from 0 to 2 mM Mg (Fig. 4A). In parallel, pause frequency increased by nearly 38-fold, with 70% of traces in the absence of Mg showing no detectable long pauses (Fig. 4B). As a combined consequence of these increases in pause duration and pause frequency, the fraction of total migration time spent paused rose markedly, from 10% in the absence of Mg to 60% and 80% at 1 and 2 mM Mg, respectively (Fig. 4C). Thus, at magnesium concentrations above 1 mM, the junction spent most of its total migration time in the long-lived paused state.

Figure 4.

Effect of solution and mechanical conditions on the long-pause regime of spontaneous HJ migration using a 1 s pause cutoff. (A–C) Effect of MgCl concentration at 23C and 3 pN in 150 mM NaCl, 10 mM Tris (pH 7.4): pause duration (A), pause frequency per migrated base pair (B), and pause-time ratio (C). (D–F) Effect of temperature at 2 mM MgCl and 3 pN in 150 mM NaCl, 10 mM Tris (pH 7.4): pause duration (D), pause frequency (E), and pause-time ratio (F). (G–I) Effect of force at 2 mM MgCl and 23C in 150 mM NaCl, 10 mM Tris (pH 7.4): pause duration (G), pause frequency (H), and pause-time ratio (I). Dots represent individual traces and open diamonds indicate mean values with SEM calculated using tether-level averages. Violin width indicates relative data density and is not proportional to sample number. A value of zero in panels (C), (F), and (I) indicates that no pauses longer than the 1 s analysis threshold were detected in that trace. Increasing Mg enhanced both the duration and frequency of long-lived pauses, whereas increasing temperature or force suppressed the long-pause regime. Inset: mean duration of the full HMM-resolved pause population plotted against force and fitted with an exponential relation (). Interpreted within a Bell-like framework, this corresponds to an effective transition-state distance of  nm at 23 C.

We next examined the effect of temperature under fixed ionic and force conditions (150 mM NaCl, 2 mM MgCl, 10 mM Tris, pH 7.4,  pN; Fig. 4D–F). Raising the temperature from 23 C to 37 C strongly suppressed the long-pause regime. Mean pause duration decreased by approximately six-fold (Fig. 4D), and pause frequency decreased by >23-fold (Fig. 4E). Together, these changes reduced the fraction of total migration time spent paused several-fold, from 80% at 23 C to 17% at 37 C (Fig. 4F). These data indicate that elevated temperature both shortens the persistence of long pauses and reduces the probability of entering the long-lived paused state.

We then tested whether mechanical load similarly modulates HJ migration under fixed ionic and temperature conditions (150 mM NaCl, 2 mM MgCl, 10 mM Tris, pH 7.4, 23 C; Fig. 4G–I). Increasing the applied force from 3 pN to 10 pN strongly suppressed the long-pause regime. Mean pause duration decreased by more than five-fold (Fig. 4G), while pause frequency decreased by ∼24-fold (Fig. 4H). As a result, the fraction of total migration time spent paused fell from 80% at 3 pN to 19% at 10 pN (Fig. 4I). To examine the force dependence of pause escape more directly, we additionally analysed the full population of HMM-resolved pauses above the effective temporal resolution of the method (0.2 s). Mean pause duration for this full HMM-resolved pause population decreased approximately exponentially with force and was well described by a Bell-like fit (), consistent with force-facilitated escape from the paused state and yielding an effective transition-state distance of nm (see Supplementary Text) along the force-coupled coordinate (Fig. 4G, inset).

Together, these results show that the long-pause regime of spontaneous HJ migration is highly tunable across physiologically relevant environmental and mechanical conditions. Mg stabilizes this regime by increasing both pause duration and pause frequency, whereas elevated temperature and applied force have the opposite effect. These perturbations therefore strongly reshape the occupancy and persistence of the long-lived paused state.

Long-lived pauses impose a kinetic barrier to effective spontaneous HJ migration

To determine whether long-lived pauses materially suppress spontaneous HJ migration, we compared, for each migration trace, the net migration rate during active migration phases with the overall migration rate calculated over the full trace, including pauses (Fig. 5). This comparison distinguishes whether a given condition primarily affects the intrinsic speed of active migration or instead reduces effective migration by increasing the time spent in the paused state.

Figure 5.

Comparison of net and overall migration rates under different solution and force conditions. (A) MgCl dependence at 23C and 3 pN in 150 mM NaCl, 10 mM Tris (pH 7.4). (B) Temperature dependence at 2 mM MgCl and 3 pN in 150 mM NaCl, 10 mM Tris (pH 7.4). (C) Force dependence at 2 mM MgCl and 23C in 150 mM NaCl, 10 mM Tris (pH 7.4). Net migration rate was calculated from active migration periods only, whereas overall migration rate was calculated over the full migration trace including pauses. Dots represent individual traces and open diamonds indicate mean values with SEM calculated using tether-level averages. Violin width indicates relative data density and is not proportional to sample number. Conditions that favoured long-lived pausing produced a larger separation between net and overall migration rates, showing that pausing substantially suppresses effective HJ migration. Welch’s t-test was used for statistical comparisons. Significance is indicated as follows: ns, ; ; ; ; .

The difference between net and overall migration rates was strongly condition-dependent. Under conditions that strongly favoured the long-pause regime, the overall migration rate was substantially lower than the net migration rate, demonstrating that long-lived pauses act as a major kinetic barrier to effective branch migration. This effect was most pronounced under high-magnesium, low-temperature and low-force conditions. For example, at 2 mM MgCl, 23 C and 3 pN, the mean net migration rate was  bp s, whereas the overall migration rate was only  bp s, corresponding to a ∼5.4-fold reduction in effective migration. A similarly strong effect was observed at 1 mM MgCl, where the overall migration rate was reduced by ∼2.8-fold relative to the net migration rate. Across the force series, the suppressive effect of pausing remained strong at low force, with ∼4.1-fold and 3.1-fold reductions at 4 and 5 pN, respectively.

These large differences coincided with conditions under which the fraction of total migration time spent paused was highest (Fig. 4C, F, and I), including 1–2 mM MgCl at 23 C and forces below  pN. Under these conditions, long-lived pauses accounted for the majority of total migration time and therefore dominated the effective migration output.

In contrast, when the long-pause regime was weak, net and overall migration rates converged. At 23 C in the absence of MgCl, the difference between the two rates was not significant, consistent with the rarity of long-lived pauses under this condition. Likewise, at 37 C, where long-lived pauses were strongly suppressed, net and overall migration rates were again not significantly different. Along the force series, the difference progressively narrowed at higher force, although a modest but still significant reduction in overall migration rate remained detectable at 8 and 10 pN. Thus, as the occupancy of the long-pause regime declines, pausing makes a correspondingly smaller contribution to limiting effective large-scale spontaneous migration.

Together, these results show that the significance of the long-pause regime lies not merely in the presence of pauses, but in their capacity to substantially suppress effective spontaneous HJ migration when they are both frequent and persistent. This identifies long-lived pausing as a condition-dependent kinetic barrier and highlights why modulation of this regime could be biologically relevant for regulating HJ dynamics, particularly under conditions in which intrinsic pausing is strong or, conversely, when additional factors may be needed to reinforce it.

HU reinforces pause-mediated suppression of HJ migration under weak intrinsic pausing

To test whether interaction with a DNA-binding protein can reshape HJ migration under conditions in which intrinsic long-lived pausing is otherwise weak, we examined the effect of HU, a bacterial nucleoid-associated DNA-binding protein, under two such conditions: 2 mM MgCl, 23 C, 10 pN and 2 mM MgCl, 37 C, 3 pN (representative traces for 2 mM MgCl, 23 C, 10 pN before and after HU addition are shown in Supplementary Fig. S6). To relate the HU effect measured on a single tether to the broader protein-free condition, we first confirmed that the pre-HU baseline behaviour of that tether was consistent with the corresponding protein-free dataset compiled across multiple tethers (Supplementary Fig. S7). This justified using the broader baseline dataset as the reference for HU comparisons. Overall, HU shifted both conditions towards a more pause-dominated regime, with the strongest effects observed in pause frequency, the fraction of total migration time spent paused, and the separation between net and overall migration rates (Fig. 6).

Figure 6.

Effects of HU on spontaneous HJ migration under two baseline conditions with relatively weak intrinsic pausing: 2 mM Mg, 23C, 10 pN and 2 mM Mg, 37C, 3 pN. (A) Pause duration. (B) Pause frequency per migrated base pair. (C) Pause-time ratio. (D) Net and overall migration rates under baseline and 100 nM HU conditions. Dots represent individual traces and open diamonds indicate mean values. For the baseline dataset, SEM was calculated using tether-level averages; for the HU dataset, individual repeated traces from the measured tether were treated as the unit of analysis for mean and SEM. Violin width reflects relative data density and is not proportional to sample number. HU had little or modest effect on pause duration, but significantly increased pause frequency and pause-time ratio under both conditions, consistent with enhanced pause-mediated suppression of effective HJ migration. Welch’s t-test was used for statistical comparisons. Significance is indicated as follows: ns, ; , ; , ; , ; , .

We first examined the effect of HU on long-pause duration (Fig. 6A). At 23 C and 10 pN, the mean long-pause duration increased from 1.8 to 2.4 s, corresponding to a ∼1.4-fold increase, but this change was not significant. At 37 C and 3 pN, the mean duration increased from 1.6 to 2.4 s, corresponding to a ∼1.5-fold increase, and was significant. Thus, HU produced only a modest effect on the duration of individual long pauses.

In contrast, HU had a much stronger effect on long-pause frequency (Fig. 6B). At 23 C and 10 pN, the mean long-pause frequency increased from 0.0027 to 0.0076 bp, corresponding to a ∼2.8-fold increase. At 37 C and 3 pN, the increase was even stronger, rising from 0.0027 to 0.013 bp, corresponding to a ∼4.7-fold increase. These results indicate that HU markedly increases the probability of entering the long-pause regime, particularly under the condition in which intrinsic long pausing is weakest.

Consistent with this increase in pause frequency, HU also increased the fraction of total migration time spent paused (Fig. 6C). At 23 C and 10 pN, this fraction increased from 0.19 to 0.42, corresponding to a ∼2.2-fold increase. At 37 C and 3 pN, it increased from 0.17 to 0.50, corresponding to a ∼2.9-fold increase. These changes show that HU shifts the junction into a substantially more pause-dominated regime even when intrinsic long-lived pausing is otherwise weak.

We next asked whether these changes translated into stronger suppression of effective HJ migration (Fig. 6D). Under both tested conditions, HU widened the separation between net and overall migration rates. At 23 C and 10 pN, the net-to-overall rate ratio increased from 1.2 in the baseline condition to 1.7 with HU. Likewise, at 37 C and 3 pN, the ratio increased from 1.2 to 2.0. Thus, HU strengthened the contribution of pausing to the suppression of effective spontaneous HJ migration under both conditions, with the larger effect again observed at 37 C and 3 pN.

Together, these data show that HU reinforces the pause-mediated kinetic barrier to spontaneous HJ migration. Although its effect on individual long-pause duration is modest, HU strongly increases the frequency and occupancy of long pauses and enhances their impact on effective migration. This supports the idea that protein interaction can reshape the pause landscape of HJ migration and restore substantial pause-mediated suppression under conditions in which intrinsic long-lived pausing alone is reduced.

Discussion

Our single-molecule measurements show that spontaneous HJ migration is not continuous, but instead proceeds through alternating periods of active migration and long-lived pauses. These pauses range from seconds to tens of seconds, far exceeding the millisecond- to sub-second dwell times previously reported in short-range single-molecule FRET studies [23, 24]. The important point is not simply that pauses occur, but that under many conditions they dominate effective migration kinetics. In particular, the divergence between net and overall migration rates shows that long-lived pauses can suppress effective branch migration by several fold, establishing pausing as a major kinetic barrier rather than a minor fluctuation superimposed on otherwise smooth motion.

Biological significance of the long-pause regime

Our main comparative analysis focused on pauses lasting at least 1 s. This threshold was not intended as a strict physical boundary, but rather as a biologically emphasized analysis window for long-lived pauses. Pauses on the order of seconds are more plausibly compatible with productive interaction by HJ-processing factors than very brief interruptions. Protein–DNA association is ultimately diffusion-limited, with specific on-rates in aqueous solution of ∼– [50, 51]. Using these rates together with typical intracellular abundance estimates for the HJ-binding protein RuvA in E. coli [52] gives effective encounter times on the order of seconds, making 1 s a reasonable conservative window for pauses likely to be relevant for regulatory interaction with the junction. Importantly, the major condition-dependent trends were robust to moderate variation of this threshold: a 0.5 s cutoff produced qualitatively similar behaviour, whereas a 2 s cutoff was substantially more restrictive and excluded a larger fraction of pause events (Supplementary Fig. S4).

Across the explored parameter space, the dominant determinant of effective spontaneous migration was not simply the speed of active migration, but the frequency and persistence of pausing. In this framework, the fraction of total migration time spent paused is best understood as an integrative consequence of pause frequency and pause duration rather than as an independent mechanistic variable. Although spontaneous HJ branch migration is often described as a bidirectional random-walk process [53, 54], our results indicate that, over extended distances, its effective behaviour is strongly shaped by intermittent long-lived pausing. Such pausing could be biologically useful by preserving a temporal window for regulatory factors to engage recombination intermediates before spontaneous migration alters their position.

A stacked-X-like paused state is the leading mechanistic interpretation

The strongest structural interpretation of the paused state is that it corresponds to a compact migration-inhibitory conformation consistent with a stacked-X-like state. This view is strongly supported by the Mg dependence, which is in agreement with structural and single-molecule studies showing that divalent cations favour stacked-X conformations of HJs [17, 23–25]. In this sense, the magnesium series provides an important point of consistency with prior work: the same ionic conditions known to stabilize compact stacked junction conformations also strongly enhance the long-pause regime in our migration assay. Mechanistically, Mg promotes close, near-parallel alignment of duplex arms through electrostatic screening and stabilization of coaxial stacking interactions [16, 20].

Conversely, elevated temperature suppresses the long-pause regime, consistent with thermal destabilization of a compact paused state and more rapid escape back into a migration-competent configuration; temperature may also enhance base-pair opening and thereby accelerate migration once the junction is in a migration-competent state [55, 56]. This temperature dependence further supports the idea that the paused state represents a genuine energetic barrier rather than an arbitrary analytical classification, because both the occupancy and persistence of the pause regime are strongly reduced at higher temperature. It also raises a broader biological point. Unlike mammals, many bacteria and ectothermic organisms such as amphibians experience substantial variation in environmental or body temperature [57, 58], and such sensitivity could in principle alter the extent to which spontaneous branch migration is intrinsically restrained under different physiological conditions. In that sense, the pause-mediated barrier may not be a fixed property of the junction alone, but one whose effectiveness depends on thermal context.

At the same time, our data do not require that every long pause correspond to a single canonical stacked-X structure. Other transient migration-inhibitory conformations, including non-canonical junction geometries, may also contribute to the paused ensemble [16, 25]. Because the present assay reports on migration dynamics rather than junction structure directly, it cannot resolve these possibilities at the structural level. We therefore regard a stacked-X-like migration-inhibitory state as the leading interpretation of the long-pause regime, while not excluding contributions from other transient conformations that similarly impede branch migration.

Force provides additional mechanistic insight into this barrier. Increasing tension strongly reduced pause frequency, shortened pause duration and, consequently, reduced the fraction of total migration time spent paused. This behaviour is consistent with applied force destabilizing a compact migration-inhibitory state or lowering the barrier for escape from that state, while also disfavouring re-entry into the paused conformation. Consistent with this view, the mean duration of the full HMM-resolved pause population decreased approximately exponentially with force and was well described by a Bell-like fit (; Fig. 4G, inset). Interpreted within this framework, the fitted exponent corresponds to an effective distance to the transition state of ∼1.2 nm along the force-coupled coordinate (see Supplementary Text), compatible with a modest extension change during escape from a compact paused conformation. This parameter should be interpreted as an effective kinetic descriptor rather than a direct geometric measure of the full structural difference between stacked-X and open planar conformations.

The force dependence also underscores that the paused state is mechanically sensitive. Although the simple axial loads applied here do not reproduce the full complexity of chromosomal DNA, they show that relatively modest tension can shift the balance between pause entry, pause escape and active migration. In this sense, force acts here primarily as a physical probe of the underlying barrier, revealing that both access to, and escape from, the migration-inhibitory state are coupled to extension along the force-bearing coordinate.

Taken together, the coherent dependence on Mg, temperature and force strongly argues that the pauses reflect a physically meaningful conformational barrier coupled to branch migration rather than arbitrary stochastic delays. More broadly, these perturbations show that the effectiveness of this barrier is chemically, thermally and mechanically tunable.

DNA architectural protein HU reinforces the pause barrier

The HU experiments provide a biological extension of the intrinsic pausing mechanism. Under two conditions in which long-lived pausing is otherwise weak (2 mM MgCl, 23 C, 10 pN and 2 mM MgCl, 37 C, 3 pN), HU produced only a modest increase in long-pause duration, on the order of 1.4–1.5-fold, with statistical significance reached under one condition. By contrast, HU caused a much larger increase in long-pause frequency, of ∼2.8- and 4.7-fold. This pattern indicates that HU primarily promotes entry into the long-pause regime rather than strongly extending the lifetime of individual pauses.

Because HU interacts with both distorted DNA substrates and duplex DNA [36, 37], it may promote long-lived pausing either through direct engagement of the junction or, more likely, through association with nearby duplex arms that introduces an external barrier to migration. The observed increase in pause-entry frequency, coupled with the modest change in pause lifetime, is more consistent with a mechanism in which HU enhances the likelihood of transient kinetic trapping during branch migration—potentially by imposing local structural constraints or introducing friction-like resistance, rather than strongly stabilizing a specific paused HJ conformation. While our data do not allow us to assign a unique binding geometry, they suggest that the dominant effect of HU arises from its nonspecific dsDNA-binding activity.

Taken together, these results show that HU can reshape the pause landscape by selectively enhancing pause entry, thereby restoring substantial pause-mediated suppression under conditions in which intrinsic long-lived pausing is weakened. This biological extension is important because it suggests that intrinsic pausing need not act alone. Even when mechanical or thermal conditions weaken the long-pause regime, interaction with architectural DNA-binding proteins can re-establish effective suppression of migration. More broadly, this raises the possibility that additional factors, including chromatin-associated proteins or nucleosomes, may similarly regulate HJ migration by modulating the probability of pause entry rather than solely stabilizing paused states [59]. In this view, intrinsic conformational pausing and external protein-mediated constraints act together to control spontaneous migration and preserve recombination intermediates.

Methodological context and limitations

The magnetic-tweezer assay used here complements earlier approaches to HJ dynamics. Bulk biochemical assays provided ensemble-averaged migration rates [28], while structural studies revealed static junction conformations [17, 20, 22]. Single-molecule FRET studies provided valuable insight into fast local transitions between stacked and open conformations [23, 24], but are inherently limited in migration range and observation time. By contrast, our assay captures long-range migration and long-lived pausing over extended timescales.

Several limitations remain. The assay does not directly report junction conformation, so the stacked-X-like assignment is an interpretation based on condition dependence and prior structural knowledge rather than direct visualization. The HMM-based segmentation has a finite temporal resolution of ∼0.2 s, and our main analysis emphasizes pauses lasting at least 1 s, so very short interruptions are not fully captured and derived kinetic parameters should be interpreted as effective descriptors of the experimentally resolved pause landscape. In addition, the present experiments probe applied tension but not torsional constraints. In vivo, twisting, supercoiling, and topological confinement are likely to influence how mechanical load is transmitted to the junction [60–62].

Overall, our results show that spontaneous HJ migration is not a featureless diffusive process, but is governed by a tunable long-lived paused state that acts as a kinetic barrier to effective branch migration. Mg, temperature and force reshape this barrier in opposite directions, and external protein interaction can reinforce it when intrinsic pausing is weak. We therefore propose that spontaneous branch migration is intrinsically self-limiting through condition-dependent pausing, and that this pause-mediated barrier provides a physically grounded framework for understanding how cellular factors regulate the timing, extent and outcome of HJ processing.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the National Research Foundation, Singapore, through the Mechanobiology Institute under its Mid-Sized Grant (MSG) (NRF-MSG-2023-0001), the Ministry of Education under the Research Centres of Excellence programme through the Mechanobiology Institute at the National University of Singapore (ref. no. A-0003467-00-00), and the Singapore Ministry of Education Academic Research Fund Tier 2 (MOE-T2EP50123-0008). Open access publication was funded by the National Research Foundation, Singapore, through the Mechanobiology Institute under its Mid-Sized Grant (MSG) (NRF-MSG-2023-0001).

Data availability

Original data from the single-molecule experiments are available upon request.