Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

BCDX2–CX3 and DX2–CX3 complexes assemble and stabilize RAD51 filaments

Nature (2026) Cite this article

Subjects

Abstract

The repair of DNA double-strand breaks by homologous recombination is essential for genomic integrity, and its dysregulation is a hallmark of cancer1. Central to homologous recombination is the RAD51 recombinase, whose assembly into a nucleoprotein filament is governed by five RAD51 paralogues (RAD51B, RAD51C, RAD51D, XRCC2 and XRCC3)2. Mutations in any of these proteins predispose individuals to multiple cancers or genetic disorders3,4,5,6. These paralogues are thought to form two functionally separate complexes RAD51B–RAD51C–RAD51D–XRCC2 (BCDX2) and RAD51C–XRCC3 (CX3), that act independently at different stages of homologous recombination7,8,9,10,11. Here we demonstrate that all five paralogues can assemble into a single, ATP-dependent BCDX2–CX3–RAD51 supercomplex. The architecture of this assembly bound to single-stranded DNA reveals a contiguous filament where the CX3 module stacks atop BCDX2, creating a protofilament template for RAD51 filament formation. We further identify a novel, RAD51B-independent DX2–CX3 complex (RAD51D–XRCC2–RAD51C–XRCC3) functioning as a stable RAD51 anchor on single-stranded DNA, and we capture it in multiple states, including capping RAD51 filament segment. These distinct assemblies are differentially regulated by ATPase activity, defining a dynamic BCDX2–CX3 ‘loader’ and a stable DX2–CX3 ‘anchor’ that provide functional modularity to the homologous recombination machinery. This work provides a unifying mechanism for human RAD51 paralogue function and delivers an atomic blueprint for interpreting disease-causing mutations.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on SpringerLink
  • Instant access to the full article PDF.

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Biochemical and structural analysis of the BCDX2–CX3–RAD51 supercomplex.
The alternative text for this image may have been generated using AI.
Fig. 2: BCDX2 and CX3 cooperate in ssDNA binding and ATPase activation.
The alternative text for this image may have been generated using AI.
Fig. 3: Structural analysis of the DX2–CX3–RAD51 paralogue complex reveals multiple conformational states.
The alternative text for this image may have been generated using AI.
Fig. 4: Structure of DX2–CX3 capping a RAD51 filament segment and bound to a DNA bubble.
The alternative text for this image may have been generated using AI.
Fig. 5: Cellular interactions and genetic co-dependency of RAD51 paralogue complexes.
The alternative text for this image may have been generated using AI.

Similar content being viewed by others

Data availability

All cryo-EM maps have been deposited with the Electron Microscopy Data Bank (EMDB) and all models have been deposited with the PDB: BCDX2–CX3–RAD51–ssDNA (EMD-72154 and PDB 9Q2A); DX2–CX3–RAD51–ssDNA, active state (EMD-72152 and PDB 9Q28); DX2–CX3–RAD51–ssDNA, intermediate state (EMD-72148 and PDB 9Q23); DX2–CX3–RAD51–ssDNA, closed state (EMD-72150 and PDB 9Q25); DX2–CX3–RAD51 filament (EMD-72153 and PDB 9Q29); and DX2–CX3–RAD51–dsDNA (EMD-72155 and PDB 9Q2B). The proteomics data have been deposited with MassIVE (MSV000098718). Source data are provided with this paper.

References

  1. Prakash, R., Zhang, Y., Feng, W. & Jasin, M. Homologous recombination and human health: the roles of BRCA1, BRCA2, and associated proteins. Cold Spring Harb. Perspect. Biol. 7, a016600 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

  2. Sullivan, M. R. & Bernstein, K. A. RAD-ical new insights into RAD51 regulation. Genes 9, 629 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

  3. Loveday, C. et al. Germline mutations in RAD51D confer susceptibility to ovarian cancer. Nat. Genet. 43, 879–882 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Meindl, A. et al. Germline mutations in breast and ovarian cancer pedigrees establish RAD51C as a human cancer susceptibility gene. Nat. Genet. 42, 410–414 (2010).

    Article  CAS  PubMed  Google Scholar 

  5. Wadt, K. A. et al. Germline RAD51B truncating mutation in a family with cutaneous melanoma. Fam. Cancer 14, 337–340 (2015).

    Article  CAS  PubMed  Google Scholar 

  6. Park, J. Y. et al. Complementation of hypersensitivity to DNA interstrand crosslinking agents demonstrates that XRCC2 is a Fanconi anaemia gene. J. Med. Genet. 53, 672–680 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Masson, J. Y. et al. Identification and purification of two distinct complexes containing the five RAD51 paralogs. Genes Dev. 15, 3296–3307 (2001).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Wiese, C. et al. Interactions involving the Rad51 paralogs Rad51C and XRCC3 in human cells. Nucleic Acids Res. 30, 1001–1008 (2002).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Berti, M. et al. Sequential role of RAD51 paralog complexes in replication fork remodeling and restart. Nat. Commun. 11, 3531 (2020).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  10. Chun, J., Buechelmaier, E. S. & Powell, S. N. Rad51 paralog complexes BCDX2 and CX3 act at different stages in the BRCA1-BRCA2-dependent homologous recombination pathway. Mol. Cell. Biol. 33, 387–395 (2013).

    Article  CAS  PubMed  Google Scholar 

  11. Garcin, E. B. et al. Differential requirements for the RAD51 paralogs in genome repair and maintenance in human cells. PLoS Genet. 15, e1008355 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Wright, W. D., Shah, S. S. & Heyer, W. D. Homologous recombination and the repair of DNA double-strand breaks. J. Biol. Chem. 293, 10524–10535 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Sigurdsson, S., Trujillo, K., Song, B., Stratton, S. & Sung, P. Basis for avid homologous DNA strand exchange by human Rad51 and RPA. J. Biol. Chem. 276, 8798–8806 (2001).

    Article  CAS  PubMed  Google Scholar 

  14. Baumann, P., Benson, F. E. & West, S. C. Human Rad51 protein promotes ATP-dependent homologous pairing and strand transfer reactions in vitro. Cell 87, 757–766 (1996).

    Article  CAS  PubMed  Google Scholar 

  15. Sung, P. Catalysis of ATP-dependent homologous DNA pairing and strand exchange by yeast RAD51 protein. Science 265, 1241–1243 (1994).

    Article  CAS  PubMed  ADS  Google Scholar 

  16. Golmard, L. et al. Germline mutation in the RAD51B gene confers predisposition to breast cancer. BMC Cancer 13, 484 (2013).

    Article  PubMed  PubMed Central  Google Scholar 

  17. Lin, W. Y. et al. A role for XRCC2 gene polymorphisms in breast cancer risk and survival. J. Med. Genet. 48, 477–484 (2011).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  18. Greenhough, L. A. et al. Structure and function of the RAD51B-RAD51C-RAD51D-XRCC2 tumour suppressor. Nature 619, 650–657 (2023).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  19. Rawal, Y. et al. Structural insights into BCDX2 complex function in homologous recombination. Nature 619, 640–649 (2023).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  20. Longo, M. A. et al. RAD51C-XRCC3 structure and cancer patient mutations define DNA replication roles. Nat. Commun. 14, 4445 (2023).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  21. Taylor, M. R. G. et al. Rad51 paralogs remodel pre-synaptic Rad51 filaments to stimulate homologous recombination. Cell 162, 271–286 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Belan, O. et al. Single-molecule analysis reveals cooperative stimulation of Rad51 filament nucleation and growth by mediator proteins. Mol. Cell 81, 1058–1073.e7 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Roy, U. et al. The Rad51 paralog complex Rad55-Rad57 acts as a molecular chaperone during homologous recombination. Mol. Cell 81, 1043–1057.e8 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Liu, J. et al. Rad51 paralogues Rad55-Rad57 balance the antirecombinase Srs2 in Rad51 filament formation. Nature 479, 245–248 (2011).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  25. Akita, M. et al. Mechanism of BCDX2-mediated RAD51 nucleation on short ssDNA stretches and fork DNA. Nucleic Acids Res. 52, 11738–11752 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Taylor, M. R. G. et al. A polar and nucleotide-dependent mechanism of action for RAD51 paralogs in RAD51 filament remodeling. Mol. Cell 64, 926–939 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Landrum, M. J. et al. ClinVar: updates to support classifications of both germline and somatic variants. Nucleic Acids Res. 53, D1313–D1321 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Xu, J. et al. Cryo-EM structures of human RAD51 recombinase filaments during catalysis of DNA-strand exchange. Nat. Struct. Mol. Biol. 24, 40–46 (2017).

    Article  CAS  PubMed  Google Scholar 

  29. Short, J. M. et al. High-resolution structure of the presynaptic RAD51 filament on single-stranded DNA by electron cryo-microscopy. Nucleic Acids Res. 44, 9017–9030 (2016).

    CAS  PubMed  PubMed Central  ADS  Google Scholar 

  30. Lee, J. Y. et al. DNA recombination. Base triplet stepping by the Rad51/RecA family of recombinases. Science 349, 977–981 (2015).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  31. Kurumizaka, H. et al. Homologous-pairing activity of the human DNA-repair proteins Xrcc3.Rad51C. Proc. Natl Acad. Sci. USA 98, 5538–5543 (2001).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  32. Zaitseva, E. M., Zaitsev, E. N. & Kowalczykowski, S. C. The DNA binding properties of Saccharomyces cerevisiae Rad51 protein. J. Biol. Chem. 274, 2907–2915 (1999).

    Article  CAS  PubMed  Google Scholar 

  33. Liu, J., Ehmsen, K. T., Heyer, W. D. & Morrical, S. W. Presynaptic filament dynamics in homologous recombination and DNA repair. Crit. Rev. Biochem. Mol. Biol. 46, 240–270 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Appleby, R., Bollschweiler, D., Chirgadze, D. Y., Joudeh, L. & Pellegrini, L. A metal ion-dependent mechanism of RAD51 nucleoprotein filament disassembly. iScience 26, 106689 (2023).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  35. Krejci, L. et al. DNA helicase Srs2 disrupts the Rad51 presynaptic filament. Nature 423, 305–309 (2003).

    Article  CAS  PubMed  ADS  Google Scholar 

  36. Veaute, X. et al. The Srs2 helicase prevents recombination by disrupting Rad51 nucleoprotein filaments. Nature 423, 309–312 (2003).

    Article  CAS  PubMed  ADS  Google Scholar 

  37. Hu, Y. et al. RECQL5/Recql5 helicase regulates homologous recombination and suppresses tumor formation via disruption of Rad51 presynaptic filaments. Genes Dev. 21, 3073–3084 (2007).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Aihara, H., Ito, Y., Kurumizaka, H., Yokoyama, S. & Shibata, T. The N-terminal domain of the human Rad51 protein binds DNA: structure and a DNA binding surface as revealed by NMR. J. Mol. Biol. 290, 495–504 (1999).

    Article  CAS  PubMed  Google Scholar 

  39. Joudeh, L., Appleby, R. E., Maman, J. D. & Pellegrini, L. Structural mechanism of strand exchange by the RAD51 filament. eLife 14, RP107114 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

  40. Kurumizaka, H. et al. Region and amino acid residues required for Rad51C binding in the human Xrcc3 protein. Nucleic Acids Res. 31, 4041–4050 (2003).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Luo, S. C. et al. A RAD51-ADP double filament structure unveils the mechanism of filament dynamics in homologous recombination. Nat. Commun. 14, 4993 (2023).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  42. Greenhough, L. A., Galanti, L., Liang, C. C., Boulton, S. J. & West, S. C. Cryo-electron microscopy visualization of RAD51 filament assembly and end-capping by XRCC3-RAD51C-RAD51D-XRCC2. Science 391, eaea1546 (2025).

    Article  Google Scholar 

  43. Park, J. Y. et al. Breast cancer-associated missense mutants of the PALB2 WD40 domain, which directly binds RAD51C, RAD51 and BRCA2, disrupt DNA repair. Oncogene 33, 4803–4812 (2014).

    Article  CAS  PubMed  Google Scholar 

  44. Thrasher, J. G., Fagunloye, A. A. G., Justiniano, F. S., Bernstein, K. A. & Jensen, R. B. RAD51 paralogs and RAD51 paralog complexes BCDX2 and CX3 interact with BRCA2. Preprint at bioRxiv https://doi.org/10.1101/2024.10.10.617680 (2024).

  45. Arafeh, R., Shibue, T., Dempster, J. M., Hahn, W. C. & Vazquez, F. The present and future of the Cancer Dependency Map. Nat. Rev. Cancer 25, 59–73 (2025).

    Article  CAS  PubMed  Google Scholar 

  46. Pan, J. et al. Interrogation of mammalian protein complex structure, function, and membership using genome-scale fitness screens. Cell Syst. 6, 555–568.e7 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Kondrashova, O. et al. Secondary somatic mutations restoring RAD51C and RAD51D associated with acquired resistance to the PARP inhibitor rucaparib in high-grade ovarian carcinoma. Cancer Discov. 7, 984–998 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Ali, A. M. et al. Polymorphisms in DNA repair gene XRCC3 and susceptibility to breast cancer in Saudi females. Biomed Res. Int. 2016, 8721052 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

  49. Rein, H. L., Bernstein, K. A. & Baldock, R. A. RAD51 paralog function in replicative DNA damage and tolerance. Curr. Opin. Genet. Dev. 71, 86–91 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Rein, H. L. et al. Comprehensive RAD51C ovarian cancer variant analysis uncouples homologous recombination and replicative functions. Nat. Commun. 16, 6539 (2025).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  51. Saxena, S., Somyajit, K. & Nagaraju, G. XRCC2 regulates replication fork progression during dNTP alterations. Cell Rep. 25, 3273–3282.e6 (2018).

    Article  CAS  PubMed  Google Scholar 

  52. Somyajit, K., Saxena, S., Babu, S., Mishra, A. & Nagaraju, G. Mammalian RAD51 paralogs protect nascent DNA at stalled forks and mediate replication restart. Nucleic Acids Res. 43, 9835–9855 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  53. Henricksen, L. A., Umbricht, C. B. & Wold, M. S. Recombinant replication protein A: expression, complex formation, and functional characterization. J. Biol. Chem. 269, 11121–11132 (1994).

    Article  CAS  PubMed  Google Scholar 

  54. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Liu, J. et al. Srs2 promotes synthesis-dependent strand annealing by disrupting DNA polymerase δ-extending D-loops. eLife 6, e22195 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

  56. Wright, W. D. & Heyer, W. D. Rad54 functions as a heteroduplex DNA pump modulated by its DNA substrates and Rad51 during D loop formation. Mol. Cell 53, 420–432 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Rappsilber, J., Mann, M. & Ishihama, Y. Protocol for micro-purification, enrichment, pre-fractionation and storage of peptides for proteomics using StageTips. Nat. Protoc. 2, 1896–1906 (2007).

    Article  CAS  PubMed  Google Scholar 

  58. Budayeva, H. G., Ma, T. P., Wang, S., Choi, M. & Rose, C. M. Increasing the throughput and reproducibility of activity-based proteome profiling studies with hyperplexing and intelligent data acquisition. J. Proteome Res. 23, 2934–2947 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Cheung, T. K., Zhu, Y. & Rose, C. M. Offset mass carrier proteome improves quantification of multiplexed single cell proteomics. Mol. Cell. Proteomics 24, 100959 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. Erickson, B. K. et al. Active instrument engagement combined with a real-time database search for improved performance of sample multiplexing workflows. J. Proteome Res. 18, 1299–1306 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  61. Schweppe, D. K. et al. Full-featured, real-time database searching platform enables fast and accurate multiplexed quantitative proteomics. J. Proteome Res. 19, 2026–2034 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Huang, T. et al. MSstatsTMT: statistical detection of differentially abundant proteins in experiments with isobaric labeling and multiple mixtures. Mol. Cell. Proteomics 19, 1706–1723 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).

    Article  CAS  PubMed  Google Scholar 

  64. Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  65. Xu, J. et al. Mechanisms of distinctive mismatch tolerance between Rad51 and Dmc1 in homologous recombination. Nucleic Acids Res. 49, 13135–13149 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010).

    Article  CAS  PubMed  PubMed Central  ADS  Google Scholar 

  67. Adams, P. D. et al. PHENIX: building new software for automated crystallographic structure determination. Acta Crystallogr. D Biol. Crystallogr. 58, 1948–1954 (2002).

    Article  PubMed  ADS  Google Scholar 

  68. Pettersen, E. F. et al. UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82 (2021).

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We thank Genentech colleagues in the Biomolecular Research, Biochemical and Cellular Pharmacology, and Structural Biology departments, especially J. Sudhamsu, P. Hsu and S. Ro, for helpful discussions and feedback on the manuscript; B. Yauch for help with DepMap analysis and helpful discussions; S. Klaeger and H. Gurung for mass spectrometry analysis of native XRCC3 immunoprecipitation experiments; and A. Howes for preparing the illustration for the final model, for helpful discussions and for the feedback on the manuscript. Work in W.-D.H.’s laboratory was supported by NIH awards R01GM58015, R35GM157976, R01CA273911 and R01HD109322.

Author information

Authors and Affiliations

  1. Protein Sciences, Genentech Inc., South San Francisco, CA, USA

    Christopher W. Koo, Jiaqi Xiao  (肖嘉祺), Christine Yu, Caleigh M. Azumaya, Bobby Brillantes, Claudio Ciferri & Stanislau Yatskevich

  2. Discovery Oncology, Genentech Inc., South San Francisco, CA, USA

    Sebastien Coassolo

  3. Department of Microbiology and Molecular Genetics, University of California, Davis, CA, USA

    Jie Liu, Steven K. Gore & Wolf-Dietrich Heyer

  4. Proteomic & Genomic Technologies, Genentech Inc., South San Francisco, CA, USA

    Tommy K. Cheung & Christopher M. Rose

Authors
  1. Christopher W. Koo
    View author publications

    Search author on:PubMed Google Scholar

  2. Jiaqi Xiao  (肖嘉祺)
    View author publications

    Search author on:PubMed Google Scholar

  3. Sebastien Coassolo
    View author publications

    Search author on:PubMed Google Scholar

  4. Jie Liu
    View author publications

    Search author on:PubMed Google Scholar

  5. Christine Yu
    View author publications

    Search author on:PubMed Google Scholar

  6. Caleigh M. Azumaya
    View author publications

    Search author on:PubMed Google Scholar

  7. Steven K. Gore
    View author publications

    Search author on:PubMed Google Scholar

  8. Tommy K. Cheung
    View author publications

    Search author on:PubMed Google Scholar

  9. Bobby Brillantes
    View author publications

    Search author on:PubMed Google Scholar

  10. Christopher M. Rose
    View author publications

    Search author on:PubMed Google Scholar

  11. Wolf-Dietrich Heyer
    View author publications

    Search author on:PubMed Google Scholar

  12. Claudio Ciferri
    View author publications

    Search author on:PubMed Google Scholar

  13. Stanislau Yatskevich
    View author publications

    Search author on:PubMed Google Scholar

Contributions

S.Y. conceived the study. C.W.K. and S.Y. designed all the experiments. B.B. expressed all the proteins. C.W.K., C.Y., J.X. and S.Y. purified all the proteins. C.M.A., C.W.K. and S.Y. collected all the cryo-EM data. C.W.K., J.X. and S.Y. prepared the cryo-EM samples and determined the structures. C.W.K. performed all the SPR experiments. J.X. and S.Y. performed the pull-down, SEC, fluorescence anisotropy and ATPase assays. T.K.C. performed the mass spectrometry analysis. J.L. and S.K.G. planned and performed the D-loop assays. S.Y. performed the filament assay. S.C. and S.Y. performed all the cellular work. C.M.R., W.-D.H., C.C. and S.Y. supervised the work. C.W.K., J.X. and S.Y. prepared the manuscript, with input from all authors.

Corresponding authors

Correspondence to Claudio Ciferri or Stanislau Yatskevich.

Ethics declarations

Competing interests

S.C., C.Y., C.M.A., B.B., T.K.C., C.M.R. and C.C. are Genentech/Roche employees and own shares in the Genentech/Roche group. The other authors declare no competing interests.

Peer review

Peer review information

Nature thanks the anonymous reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Biochemical characterization of BCDX2-CX3 interaction.

a, Domain organization of human RAD51 and its five canonical paralogs. NTD, N-terminal domain. b, Schematic representation of the pull-down experiment. An 8xHis-tag is located at the C-terminus of XRCC2 protein, and BCDX2 complex was immobilized using Ni-NTA beads. For all pull-down experiments, BCDX2 complex contained 2xStrepII-SNAP-tag on RAD51C, CX3 complex had all tags cleaved, RAD51 contained an N-terminal 2xStrepII-SNAP-tag. c, Coomassie-stained SDS-PAGE gel of an 8xHis-BCDX2 pull-down assay incubated either with or without the CX3 complex and either with no nucleotide, or supplemented with 1 mM ADP or 1 mM ATP in all buffers. d, XRCC3 band quantification of pull-down assay shown in (c). Data are presented as mean ± s.d. for n = 3 independent biological replicates. e, Coomassie-stained SDS-PAGE gel of an 8xHis-BCDX2 pull-down assay incubated with the CX3 complex, RAD51 or CX3 and RAD51 in the presence or absence of 100 nt ssDNA. All reactions contained 1 mM ATP. f, XRCC3 and RAD51 band quantification of the assay shown in (e). Data are presented as mean ± s.d. for n = 3 independent biological replicates. g, Surface plasmon resonance (SPR) analysis of immobilized BCDX2, CX3 or BCDX2-CX3 complexes interaction with RAD51 (all tags removed). The inset on the top right shows a dose-response curve, which was used to determine equilibrium binding affinity. All experiments were performed in the presence of 1 mM ATP. MCK – multi-cycle kinetics. h, Size-exclusion chromatography (SEC) reconstitution using Superose 6 3.2/300 (Cytiva) with corresponding SDS-PAGE gels below. 19 nt ssDNA was used as ssDNA substrate. All runs were done in the presence of 1 mM ATP in all buffers. These experiments were repeated at least three times independently with the same result.

Source data

Extended Data Fig. 2 Biochemical validation of BCDX2-CX3 interface mutations.

a, Structural mapping of mutations used to disrupt BCDX2-CX3 complex assembly. The Q267 residue on RAD51C is at the interface with RAD51B protein, while RAD51B R8, K218, E248 contact RAD51C. The R150 and Q152K residues of XRCC3 are at the interface with RAD51. b, SEC reconstitution using Superose 6 3.2/300 (Cytiva), investigating the BCDX2-CX3 complex formation. The SEC profiles (top) and corresponding SDS-PAGE gels (bottom) show that either a mutation in RAD51C of the CX3 complex (CQ267RX3: RAD51C Q267R mutation) or mutations in RAD51B of the BCDX2 complex (BMCDX2: RAD51B R8E, K218Y, E248R mutations) abrogate the formation of the BCDX2-CX3 supercomplex. These experiments were repeated at least twice independently with the same result. c, Structural overview of the interface between XRCC3 and the first RAD51 protomer (RAD511), highlighting the docking of the RAD511 NTD and linker onto the XRCC3 RecA-like domain. d, Close-up views showing the conserved RAD51 linker residue F86 inserting into a hydrophobic pocket on the surface of XRCC3. The panel displays the electrostatic surface potential of XRCC3. e, Comparison of the XRCC3:RAD511 interaction (colored by subunit) with the canonical RAD51:RAD51 interaction within a filament (PDB 8BQ2)34, colored in grey). f, SEC reconstitution using Superose 6 3.2/300 (Cytiva), investigating the CX3-RAD51 complex formation. Mutations within the XRCC3 protein, either R150S (CX3R150S) or a patient-derived Q152K mutation (CX3Q152K), weaken the stable association of CX3 with RAD51. These experiments were repeated at least twice independently with the same result.

Extended Data Fig. 3 Structural and functional characterization of the BCDX2-CX3 complex.

a, Structural details of the ssDNA path through the BCDX2-CX3 complex represented as ribbon diagrams. b, SPR analysis of BCDX2-CX3 interaction with different DNA types and CX3 interaction with ssDNA. DNA was immobilized, and single-cycle kinetics (SCK) or multi-cycle kinetics (MCK) runs were performed to estimate the affinity constants. All experiments were performed in the presence of 1 mM ATP. c, Fluorescence anisotropy (FA) assays measuring ssDNA (25 nt, fluorescein labeled) binding of BCDX2, CX3, and BCDX2-CX3 complexes in the presence or absence of RAD51, as well as ssDNA binding by BCDX2 or CX3 mutants. All experiments were performed in the presence of 1 mM ATP. For the top left panel, data are presented as mean ± s.d. for n = 3 independent biological replicates. For all other panels, data represents mean value across two biological replicates.

Extended Data Fig. 4 Analysis and functional characterization of the ATPase activity.

a, Structural comparison of the Alvinella pompejana CX3 (apCX3) structure (PDB 8JGA20, left) with the human CX3 module bound to ADP (this work, middle). The right panel shows an overlay of both structures, with apCX3 colored in grey, showing identical amino acid and ADP positions. BeFx appears to substitute for the Mg2+ ion. b, Raw data of the kinetic EnzCheck ATPase activity assay measuring phosphate (PO42−) production. The BCDX2-CX3 supercomplex displays synergistic ATPase activity that is significantly higher than the individual BCDX2 or CX3 modules and is further stimulated by ssDNA and RAD51. Data represents the mean value across three biological replicates, with surface representing s.d. c,d, SEC reconstitutions using Superdex 200 3.2/300 (Cytiva), testing complex formation between BE144QCDX2 and CE161QX3 with CX3 and BCDX2, respectively. Wild-type BCDX2 and CX3 proteins had their tags cleaved during purification. Corresponding SDS-PAGE gels below confirm complex assembly. All runs were done in the presence of 1 mM ATP in the samples and running buffer. These experiments were repeated at least twice independently with the same result. e, FA assays measuring ssDNA (25 nt, fluorescein labeled) binding of BE144QCDX2 and CE161QX3 in the presence of either 1 mM ATP or 1 mM ADP. Data represent the mean value across two biological replicates.

Extended Data Fig. 5 Biochemical reconstitution of the DX2-CX3 and DX2-CX3-RAD51 complexes.

a, SEC reconstitution using Superose 6 3.2/300 (Cytiva), showing the formation of the DX2-CX3 complex. The SEC profile (top) and corresponding SDS-PAGE analysis (bottom) demonstrate that DX2 and CX3 form a stable, stoichiometric complex. All proteins had their tags cleaved during purification. These experiments were repeated at least three times independently with the same result. b, SEC reconstitution using Superose 6 3.2/300 (Cytiva), demonstrating the recruitment of RAD51 to the pre-formed DX2-CX3 complex. The SEC profile (top) and corresponding SDS-PAGE analysis (bottom) show that RAD51 stably associates with the DX2-CX3 complex. These experiments were repeated at least three times independently with the same result. c, Detailed comparison of the interactions at three distinct interfaces: the XRCC3:RAD51B:RAD51CCX3 interface in the BCDX2-CX3-RAD51 complex (left); the XRCC3:RAD51D:RAD51CCX3 interface in the DX2-CX3-RAD51 complex (middle); and the internal RAD51B:RAD51CBC interface from the BCDX2-CX3-RAD51 complex (right). d, Electrostatic surface representation of the DNA channel in the DX2-CX3 ‘active’ state with ssDNA shown as a molecular model. e, Comparison of the ssDNA conformation within the DX2-CX3 ‘active’ complex (left) and a canonical RAD51 filament (PDB 8BQ2)34, right)). Arrows highlight ssDNA splitting.

Extended Data Fig. 6 Biochemical characterization and cryo-EM data processing of the DX2-CX3-RAD51 filament.

a, (left) FA assays measuring ssDNA (25 nt, fluorescein labeled) binding of the DX2 or DX2-CX3 complexes in the presence of either 1 mM ATP or 1 mM ADP. Data are presented as mean ± s.d. for n = 3 independent biological replicates. (right) Quantification of the FA assay. Data are presented as mean ± s.d. for n = 3 independent biological replicates. b, SPR analysis of immobilized different DNA types with DX2-CX3 complex. The experiment was performed in the presence of 1 mM ATP. SCK – single-cycle kinetics. c, SPR analysis of immobilized DX2-CX3 complex with RAD51 protein, showing KD ≈ 98.9 nM. The experiment was performed in the presence of 1 mM ATP. The inset shows a dose-response curve, which was used to determine equilibrium binding affinity. MCK – multi-cycle kinetics. d, Raw data of the kinetic EnzCheck ATPase activity assay measuring phosphate (PO42−) production. All proteins and ssDNA (dT100 sequence) were used at 0.5 μM concentration. Data represents the mean value across three biological replicates, with surface representing s.d. e, Structural comparison of the DX2-CX3-RAD51 filament with the active DX2-CX3-RAD51 structure (left) and with RAD51 filament (right, PDB 8BQ2).

Source data

Extended Data Fig. 7 Molecular details of DX2-CX3 cap and recombination intermediate and RAD51 filament formation by BCDX2-CX3 and DX2-CX3 complexes.

a, Molecular model of the DX2-CX3-RAD51 filament with all NTDs displayed in space-filling representation. b, Comparison of the ssDNA conformation and binding within the DX2-CX3-RAD51 filament (left) and a canonical RAD51 filament (PDB 8BQ2)34; right)). Arrows highlight ssDNA splitting. c, Close-up view highlighting the specific docking of the rearranged RAD511 NTD onto the minor groove of the duplex DNA region. d, Detailed view of the molecular contacts between the RAD511 NTD and dsDNA. e, Interaction of the DX2-CX3 modules with the single-stranded portion of the DNA bubble, with key residues from RAD51D shown anchoring the ssDNA. f, The location of a structured ‘gate’ loop in XRCC3 (residues 221–227, purple) is positioned adjacent to the dsDNA, suggesting a role in organizing the DNA substrate. g, Representative cryo-EM micrographs of reconstituted RAD51 filaments with dT50 ssDNA in the absence or presence of paralog complexes. Insets show representative 2D class averages for each dataset. h, Quantification of the RAD51 filaments as observed in (g). Data are presented as mean ± s.d. for n = 3 independent biological replicates.

Source data

Extended Data Fig. 8 D-loop stimulation by BCDX2-CX3 and CX3 complexes.

a, Schematic representation of the D-loop assay design, in which RPA is added prior to RAD51. A fluorescently labeled ssDNA substrate was incubated with RPA, then with different paralog complexes, before the addition of RAD51 and a dsDNA substrate. b, Gel image showing D-loop formation by RAD51 with different paralog complexes in the presence of RPA inhibition. c, Quantification of D-loop yield in (d). Plotted are the means ± s.d. from n = 3 biologically independent experiments. d, Schematic representation of the D-loop assay design, in which RPA was added later to stimulate D-loop formation. A fluorescently labeled ssDNA substrate was incubated with different paralog complexes before the addition of RAD51, RPA, and a dsDNA substrate. e, Gel image showing D-loop formation by RAD51 with different paralog complexes when RPA was added later in the process. f, Quantification of D-loop yield in (e). Plotted are the means ± s.d. from n = 3 biologically independent experiments. g, Schematic representation of the D-loop assay design without RPA addition. A fluorescently labeled ssDNA substrate was incubated with DX2-CX3 paralog complex before the addition of RAD51 and a dsDNA substrate. The schematics in panels a,d,g adapted from ref. 55 under a Creative Commons licence CC BY 4.0. h, Gel image showing D-loop formation by RAD51 with DX2-CX3 and in the absence of RPA. i, Quantification of D-loop yield in (h). Plotted are the means ± s.d. from n = 3 biologically independent experiments.

Source data

Extended Data Fig. 9 Cellular interactions and genetic associations between RAD51 paralogs.

a, Immunoprecipitation (IP) experiments with nuclear cell extracts from A549 cells. Nuclear cell extracts were applied either to immobilized IgG or XRCC3 antibody, and XRCC3, RAD51B, or RAD51C were detected by immunoblotting after the IP. This experiment was repeated at least twice independently with the same result. b, Fractionation experiments using Superdex 200 (3.2/300) Increase size-exclusion column. Top: purified BCDX2-CX3, DX2-CX3, CX3, and BCDX2 complexes were separated by size and visualized on Coomassie-stained SDS-PAGE gels. Bottom: Expi293 nuclear lysates were separated by size, and immunoblotting was performed against the indicated components. This experiment was repeated at least twice independently with the same result. c, Scatter plot showing the strong positive correlation of gene essentiality scores between XRCC2 and XRCC3 across cancer cell lines from the DepMap database45. d, Violin plots illustrating the distribution of CRISPR gene dependency scores for RAD51 and its five canonical paralogs from the DepMap database45. e, BRCA1 and RAD51 foci visualization via immunofluorescence in A549 cells upon depletion of RAD51 paralogs. siCombo is co-depletion of RAD51B and XRCC3. Scale bar is 12 μm. f, Quantification of RAD51 foci in A549 cells upon paralog depletion. n = 3 independent biological experiments for each condition. Each bar represents the mean with the standard error of the mean. Statistical analysis was performed using ordinary one-way ANOVA. P-value is below 0.0001 for all statistical comparisons. g, Western blot analysis of paralog protein levels after depletion of corresponding proteins. This experiment was repeated three times independently with the same result.

Source data

Extended Data Table 1 Cryo-EM data collection, refinement and validation statistics
Full size table

Supplementary information

Supplementary Information (download PDF )

This file contains Supplementary Figures 1–11 and Supplementary Tables 1–3.

Reporting Summary (download PDF )

Peer Review File (download PDF )

Source data

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Koo, C.W., Xiao, J., Coassolo, S. et al. BCDX2–CX3 and DX2–CX3 complexes assemble and stabilize RAD51 filaments. Nature (2026). https://doi.org/10.1038/s41586-026-10314-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Version of record:

  • DOI: https://doi.org/10.1038/s41586-026-10314-z

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing