DNA double-strand break end synapsis by DNA loop extrusion

Abstract

DNA double-strand breaks (DSBs) occur every cell cycle and must be efficiently repaired. Non-homologous end joining (NHEJ) is the dominant pathway for DSB repair in G1-phase. The first step of NHEJ is to bring the two DSB ends back into proximity (synapsis). Although synapsis is generally assumed to occur through passive diffusion, we show that passive diffusion is unlikely to produce the synapsis speed observed in cells. Instead, we hypothesize that DNA loop extrusion facilitates synapsis. By combining experimentally constrained simulations and theory, we show that a simple loop extrusion model constrained by previous live-cell imaging data only modestly accelerates synapsis. Instead, an expanded loop extrusion model with targeted loading of loop extruding factors (LEFs), a small portion of long-lived LEFs, and LEF stabilization by boundary elements and DSB ends achieves fast synapsis with near 100% efficiency. We propose that loop extrusion contributes to DSB repair by mediating fast synapsis.

Fig. 1. A model of DSB synapsis, mediated by DNA loop extrusion.

a After a DSB has occurred, the two DSB ends may separate. How the two DSB ends are constrained from diffusing too far apart and brought back into proximity for downstream repair is not well understood. b Overview of DSB end synapsis mediated by passive diffusion. c Overview of the loop extrusion model. Loop extruding factors (LEFs) extrude bidirectionally away from the loading site; the two motors of a LEF extrude independently: after one motor is stalled by a boundary element (BE), the other motor can continue extruding until encountering a BE on the other side. d 3D polymer simulation reveals a large discrepancy between the kinetics of synapsis mediated by diffusion alone and the synapsis kinetics determined experimentally. The shaded area around the cumulative probability curve (calculated from 2223 DSB events) represents the 95% confidence interval of the cumulative probability estimated with Dvoretzky–Kiefer–Wolfowitz inequality. The vertical dash lines indicate the experimentally determined synapsis time to reach 95% and the simulated synapsis time by passive diffusion alone to reach 95%, respectively. e Loop extrusion may facilitate DSB end synapsis in two ways: (1) the constraining LEF prevents the two DSB ends from diffusing apart after DSB; (2) Additional gap-bridging LEFs loaded within the loop extruded by constraining LEF can extrude sub-loops to bring the two DSB ends into proximity. However, if the constraining LEF falls off before the two DSB ends are brought into proximity by gap-bridging LEFs, the two DSB ends may diffuse apart. In our simulations, we assume LEFs cannot pass one another or DSB ends.

Fig. 2. Loop extrusion achieves faster synapsis than passive diffusion alone.

a Overview of 3D polymer simulation setup and representative 3D polymer conformation. A snapshot of part of the chromosome used in 3D polymer simulations is shown on the right. The inset in the red dashed square depicts the synapsis of two DSB ends. b loop extrusion dynamic parameters estimated from the Fbn2 locus. c Improved synapsis efficiency with loop extrusion compared with passive diffusion alone. Cumulative probabilities of synapsis time for synapsis with loop extrusion and with passive diffusion alone are calculated from 1376 and 2223 DSB events, respectively. d Higher synapsis efficiency at DSBs constrained by LEFs (941 events) than unconstrained DSBs (435 events). The shades in (c, d) around the cumulative probability curves represent the 95% confidence interval of the cumulative probability estimated with Dvoretzky–Kiefer–Wolfowitz inequality. e Schematic diagram of 3D polymer simulations with active extrusion versus frozen loop. In simulations with active extrusion, all LEFs may unload and reload prior to and after DSB occurrence. In simulations with frozen loops, all LEFs may unload and reload prior to DSB occurrence; after DSB occurrence, only constraining LEF may unload (but not reload), and all other LEFs are frozen in place. f Improved synapsis efficiency with active extrusion at constrained DSBs (blue bars) and reduced synapsis efficiency with active extrusion at unconstrained DSBs (green bars). In simulations with frozen loops, 750 DSBs were constrained and 357 DSBs were unconstrained. The error bars of the bar plot in (c, d, f) represent a 95% confidence interval of the mean using maximum likelihood estimation of the exponential distribution accounting for censored data. Cumulative probabilities of synapsis time for synapsis at constrained and unconstrained DSBs in frozen loop situations are shown in Supplementary Fig. 8b.

Fig. 3. Synapsis can be quantitatively predicted and mechanistically understood using an analytical theory.

a The probability of synapsis can be decomposed into the probability of being constrained and the conditional probability of gap-bridging given that the DSB was constrained. b P_constrained can be predicted by the ratio of processivity and separation. Heatmaps of predicted (left) and simulated (right; numbers in brackets show standard error of the mean, n = 3 independent 1D simulations, with 216–218 DSB events per simulation) fraction of constrained DSB sites with different combinations of processivity (y-axis) and separation (x-axis). Boundary strength = 0.5 was used in the simulations. c Three important timescales in DSB end synapsis. d P_{end-joining∣constrained} is determined by two relative timescales: the ratio of loading time and constraining time, and the ratio of extrusion time and constraining time. λ is LEF processivity, d is LEF separation, l is the average LEF loop length (a function of λ and d), and v is the extrusion speed in one direction (i.e., 1/2 the total extrusion speed). e Larger improvement on synapsis efficiency can be achieved by reducing τ_loading/τ_constrained than by reducing τ_extrusion/τ_constrained. The data points (circles) indicate the P_{end-joining∣constrained} at the separation and processivity indicated in the legend (no stabilization of LEFs at BE), whereas the line plots show how P_{end-joining∣constrained} varies with the ratio of loading time and constraining time (left panel) and the ratio of extrusion time and constraining time (right panel), while holding the other ratio constant at the values corresponding to the circle data points.

Fig. 4. LEF stabilization by either BEs or DSBs improves synapsis efficiency, as does the presence of long-lived LEFs and targeted loading of LEF at DSB.

a–c Schematic diagrams of the effects of stabilization of LEFs by BE, having a small portion of long-lived LEFs, and stabilization of LEFs by DSB (top), and the corresponding synapsis efficiency (bottom) predicted by theory (lines) or obtained from 1D simulations (squares; the error bars represent the standard error of the mean, n = 3 independent 1D simulations, with 216–218 DSB events per simulation). LEF separation of 125 kb and boundary strength of 0.5 were used. The inset in (c) shows the modest but statistically significant improvement in synapsis efficiency when fold stabilization of LEF at DSB ends increases from 1 to 4. d Schematic diagrams of the effects of targeted loading of LEF at DSB (left), the corresponding synapsis efficiency (middle) predicted by theory (lines) or obtained from 1D simulations, and the corresponding mean-synapsis time (right). Conversion of 1D simulation time steps to synapsis time assumes a total extrusion speed of 1 kb/s. The error bars represent the standard error of the mean, n = 3 independent 1D simulations, with 216–218 DSB events per simulation. LEF separation of 125 kb and boundary strength of 0.5 were used.

Fig. 5. Large-scale 1D simulations reveal a physiologically plausible parameter regime that achieves synapsis with ≥95% efficiency.

a Parameters scanned along each of the five dimensions. b The two relative timescales in Fig. 3d can cluster the 5D parameter scan data points based on synapsis efficiency. c Among all the parameter combinations that achieve a synapsis efficiency average ≥95% (n = 3 independent 1D simulations, with 216–218 DSB events per simulation), we recorded the minimum parameter along each of the five dimensions. d Significantly accelerated synapsis with loop extrusion parameters highlighted in (c). Cumulative probabilities of synapsis time for synapsis with loop extrusion parameters highlighted in (c) and with passive diffusion alone are calculated from 1313 and 2223 DSB events, respectively. The shades around the cumulative probability curves represent the 95% confidence interval of the cumulative probability estimated with the Dvoretzky–Kiefer–Wolfowitz inequality. The error bars of the bar plot represents a 95% confidence interval of the mean using maximum likelihood estimation of the exponential distribution accounting for censored data. e The five aspects of synapsis are ranked ordered based on the predicted reduction in synapsis efficiency upon knocking out the corresponding mechanism. The schematic diagram shows a plausible mechanistic basis for each of the five aspects of synapsis. The bar plot shows the average synapsis efficiency before and after knocking out each of the five mechanisms. The error bars represent the standard error of the mean (n = 3 different parameter combinations that achieved ≥95% synapsis efficiency), with the synapsis efficiency from each parameter combination overlaid as individual dots on the bar plot.