RAD51D (O75771) - review notes

Human RAD51D / RAD51L3 / R51H3 / TRAD. RAD51 paralog, RecA/RAD51 family. Chr 17q11-12.
Trigger review: PMID:42682019 (Dwivedi et al., Biochem Soc Trans 2026).

Deep research: the falcon (HTTP 402), perplexity (not available) and openai (invalid key)
providers all failed on 2026-10-06. No deep-research file was generated; these notes are the
research record.

Complex membership: BCDX2 and X3CDX2

• RAD51D is a BCDX2 subunit and binds XRCC2 directly, forming the DX2 subcomplex.
  PMID:11751635
  PMID:10871607
• In the 2026 work, DX2 also pairs with CX3 to form X3CDX2 (XRCC3-RAD51C-RAD51D-XRCC2). That
  complex caps the 5' ends of RAD51 filaments and promotes homologous pairing, whereas the RAD51B
  complex (BCDX2) promotes ATP-hydrolysis-dependent filament assembly.
  [PMID:41196948 "the RAD51 paralogs assemble into two distinct heterotetrameric complexes"; "the XRCC3 complex stably caps the 5' termini of RAD51 filaments to promote homologous pairing"]
  [PMID:42020761 "the RAD51D-XRCC2 paralog complex remodels RAD51-X3C into a pentameric RAD51-X3CDX2 assembly"; "enhances RAD51-ssDNA filament assembly, and promotes strand exchange on RPA-coated ssDNA"]
• GO has no cellular component term for X3CDX2. Recorded as a suggested question.

ATP: binds, but does not hydrolyse in the complex

• Early biochemistry on isolated RAD51D (RAD51L3) reported ssDNA binding and DNA-stimulated ATPase
  activity. PMID:10871607 (abstract only)
• Structures show RAD51D bound to ATP but catalytically inactive: it lacks the catalytic glutamate
  that RAD51, RAD51B and RAD51C have.
  PMID:37344587
  PMID:41196948
• Deep mutational scan (preprint): the RAD51D-RAD51C interface regulates BCDX2 ATPase activity,
  and the authors propose that RAD51D slows the BCDX2 ATPase.
  PMID:41542474
• Walker motifs: a Walker B mutant is defective in partner binding and HR, while the Walker A
  (K113) results conflict between CHO and MEF systems.
  PMID:16717288
  PMID:16236763
• Conclusion: the GO:0008094 IDA (PMID:10871607) and IBA (RAD51D is the only listed donor) are
  both MARK_AS_OVER_ANNOTATED, not REMOVE. The isolated-protein assay is not dismissed, but a
  catalytic ATP-driven activity overstates RAD51D's role, given target-specific structural evidence
  that the catalytic glutamate is lost (RAD51C D159-T-E161 corresponds to RAD51D D206-S-V208,
  anchored on the Walker B aspartate). ATP binding is ACCEPTed.

Function in HR and fork biology

• BCDX2 acts as a RAD51 filament mediator (see the RAD51B notes) and works downstream of BRCA2 and
  upstream of RAD51 PMID:23149936.
• XRCC2-RAD51D catalyses homologous pairing in vitro and forms rings and ssDNA filaments.
  PMID:11834724
• The RAD51D-XRCC2 interaction is needed for HR. Cancer variants G96C and G107V disrupt it.
  Isoform 1 is the functional isoform.
  PMID:30836272
• Fork reversal and slowing require RAD51D (siRNA and KO in U2OS).
  PMID:32669601
  PMID:32669601
• DX2 regulates fork progression in an ATR-dependent manner (trigger review; XRCC2 S247
  phosphorylation). PMID:42682019
• XRCC2-RAD51D stimulates BLM disruption of 4-way junctions through a direct RAD51D-BLM
  interaction. PMID:12975363

Telomeres and centrosomes (non-core)

• RAD51D localises to telomeres. Its loss shortens telomeres and causes end-to-end fusions,
  including in telomerase-negative human cells.
  [PMID:15109494 "RAD51D was shown to localize to the telomeres of both meiotic and somatic cells"; "Inhibition of RAD51D synthesis in telomerase-negative immortalized human cells by siRNA also resulted in telomere erosion and chromosome fusion"]
• Centrosome co-localisation of HR proteins PMID:21276791 (abstract only; names XRCC2 and
  RAD51). The gamma-tubulin binding IDA from this paper cannot be checked against the abstract,
  so UNDECIDED.

Disease

• Germline loss-of-function variants confer ovarian cancer risk (RR about 6.3) and PARP-inhibitor
  sensitivity. PMID:21822267

Annotation decisions (summary)

• Protein binding IPI rows (87): all REMOVE (uninformative). Most come from high-throughput Y2H
  screens (HuRI and earlier). The paralog partners are captured by GO:0033063.
• GO:0140664 IEA and IMP (PMID:16717288): MARK_AS_OVER_ANNOTATED. The paper tests Walker motif
  requirements for partner binding and HR, not damage sensing or signalling.
• NEW GO:0031297 replication fork processing (IMP, PMID:32669601).
• NEW GO:0036297 interstrand cross-link repair (IMP, PMID:16717288; the RAD51C and XRCC2 reviews
  make the same addition).
• GO:0097435 MODIFY to GO:0000730, consistent with the sibling reviews.

2026-10-10 revision (PR #4409 review)

• Walker A conflict made explicit. PMID:16717288 (human RAD51D in hamster CHO rad51d cells) finds
  Walker A mutants complement efficiently and explicitly notes that the mouse study
  PMID:16717288.
  PMID:16236763 (mouse MEFs) reports
  PMID:16236763.
  Recorded as a DISPUTED finding_review on the PMID:16236763 reference (superseded_by PMID:16717288
  and the two cryo-EM papers that show RAD51D is not catalytic). The GO:0036297 NEW row now quotes
  the K113 result instead of the "ATP binding and hydrolysis" interpretation, and the GO:0008094
  reason now presents both Walker A results instead of only the CHO result.
• X3CDX2 core function: description explains why no MF is asserted (GO:0140619 assigned to XRCC3).
• Completed the truncated PMID:41196948 quote.