Extended Data Fig. 3: Computational models of SPO11–TOP6BL complexes bound to DNA.

a, AlphaFold3 model colored according to pLDDT confidence score, including unstructured segment and C-terminal helix from TOP6BL. b, Details of SPO11 interaction with the TOP6BL C terminal helix, predicted by many of the AlphaFold3 models (Supplementary Discussion 2). The boxed region in the overview (upper left) is detailed at upper right. The sequence alignment below shows conservation of this portion of SPO11 (red dashed box), but not in S. cerevisiae. c, Range of distances between TOP6BL GHKL domains for the models in Supplementary Fig. 2. d,e, Reproducibility of DNA bending in models with DNAs of different sequence and/or length. Duplexes are superimposed for three models (d), and bend angles from all 25 models are summarized (e). Three models with minimally bent DNA are shown in gray. Although the bend position varies considerably (causing the poor DNA pLDDT score in panel a), the arm positions, bend angles, and local DNA deformation at the bend were all highly reproducible. f, Crystal structure of Topo VI holoenzyme from Methanosarcina mazei (pdb: 2q2e)¹⁴. g, Comparison of SPO11–TOP6BL and Top6A–Top6B (pdb: 2q2e) interfaces. h, Asymmetry of AlphaFold3 models. Each point indicates the distances in one AlphaFold3 model between the side chain oxygens of the two SPO11-Y138 residues and their respective “scissile” phosphates as defined by the center of the DNA bend and the offset needed to generate a two-nucleotide 5′ overhang. Distance 1 is the shorter of the two distances for each model. The cyan point is from the representative model shown in detail views throughout this paper. The dashed diagonal indicates expectation for perfect rotational symmetry. Values are omitted (off scale) for the three models with unbent DNA (gray circles in panel d).