PML — domain architecture and residue coordinates

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• PML is a bona fide tumour suppressor and the organiser of stress-responsive nuclear bodies; disrupting the RING tetramer to attack the fusion would also dismantle wild-type PML bodies, whose loss is associated with impaired p53 responses.
• Arsenic already works on the PML moiety, so a second PML-directed agent has a high bar to clear and inherits the same on-target liability.

1. On the bcr1 break at residue 552 the SUMO-interaction motif at 556-562 misses the fusion by four residues. Is that near-miss functionally important, and does it differ between bcr1 and bcr3 in a way that maps onto the modest clinical differences between the two forms?
2. Where does the PML coiled coil actually end? UniProt says 253, the only RBCC structure stops at 256, and the literature implies ~360. A construct-boundary experiment or a longer structure would fix the single softest coordinate in this node.
3. PML RING tetramerisation is required for PML::RARA leukaemogenesis in mice. Does the tetramer interface differ enough between PML and other TRIM RINGs to be selectively druggable, given that arsenic already targets the same moiety by a different route?
4. How far does the PML coiled coil actually run? 359 is a floor set by the C-terminal-most residue anyone has mutated; Pfam says 570; no structure resolves any of it. Until someone solves the PML coiled coil, every coiled-coil bar in every figure of this protein is a guess with a citation attached.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   PML canonical (P29590-1, 882 aa) has a RING at 57-92, B-box 1 at 124-166, B-box 2 at 183-236, a coiled coil at 228-253, an NLS at 476-490 and a SUMO-interaction motif at 556-562, with PML::RARA translocation sites annotated at 394/395 and 552/553.

   UniProtKB entry P29590 (PML_HUMAN): reviewed sequence, length and positional feature annotations · UniProt Consortium · UniProtKB/Swiss-Prot 2026
   ▸What this evidence cannot show

   UniProt labels the two breakpoints 'type A APL' and 'type B APL', wording that does not match standard bcr1/bcr2/bcr3 usage; the coordinates are trustworthy, the labels are not. PML has twelve annotated isoforms and the canonical is PML-I.

   ▸Evidence that argues against this (2)

   • UniProt's coiled-coil feature at 228-253 is 26 residues long and excludes most of the experimentally validated PML oligomerisation surface. Mutating I279, and the partner-side residues L297/L298/V301/I326/V333/V345/L352/L356/L359, disrupts coiled-coil pairing and abolishes nuclear-body formation; crosslinking places coiled-coil contacts at T328-Y339. The functional coiled coil therefore runs to at least residue 359, and Pfam PF12126 models it as 240-570. The stored bar has been widened to 228-359 and the UniProt value is retained in the note. Tan Y et al. 2024
   • UniProt's B-box 2 at 183-236 overlaps its own coiled coil at 228-253 by nine residues, which cannot both be right as drawn. InterPro IPR000315 and PROSITE PS50119 end B-box 2 at 222, removing the overlap. We keep UniProt's 183-236 because the MBP-fusion crystal 8J25 was built on exactly that construct (SIFTS maps its PML chain to P29590 183-236), but the 222-236 stretch should be read as ambiguous between the two features. InterPro Consortium; EMBL-EBI 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The PML RING tetramerises through PML-specific sequences; tetramerisation is required for nuclear-body assembly and PML sumoylation, and impaired tetramerisation abolishes PML/RARA-driven leukaemogenesis in vivo and arsenic-induced differentiation ex vivo.

   RING tetramerization is required for nuclear body biogenesis and PML sumoylation · Wang P, Benhenda S, Wu H, et al. · Nature Communications 2018
   PMID 29599493DOI
   ▸What this evidence cannot show

   Crystallography plus SAXS on isolated RING constructs, with in vivo tests in mouse models; the in vivo readout is leukaemogenesis by a transgene, not treatment of established human disease. PDB 5YUF (title 'Crystal Structure of PML RING tetramer', verified) covers residues 49-99.

   ▸Evidence that argues against this (1)

   • A later cryo-EM structure of the whole RBCC finds the basic PML unit to be a DIMER, formed through the alpha6 helix rather than through the RING, and reports that coiled-coil point mutants impair PML self-association far more than RING/B-box mutants do (83.9%, 75.8% and 73.7% loss for L240R, L247R and I279R versus 18.5-45.2% for the RB-interface mutants), and that arsenic rescues nuclear-body formation by RB mutants but not by CC mutants. The authors reconcile the two models by proposing that coiled-coil pairing draws dimers together and the RING/B1 interface then supports higher-order assembly, but a flat statement that the PML RING forms a tetramer and does not dimerise no longer survives this paper. Tan Y et al. 2024
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.review

   APL progressed from a disease with a median remission duration of 11-25 months and ~35-45% cure on chemotherapy alone to ~90-95% complete remission with ATRA-based regimens, with arsenic trioxide improving outcomes further.

   Acute promyelocytic leukemia: from highly fatal to highly curable · Wang ZY, Chen Z · Blood 2008
   PMID 18299451DOI
   ▸What this evidence cannot show

   A historical review rather than a single trial; the percentages aggregate heterogeneous cohorts across decades and countries.

   ▸Evidence that argues against this (1)

   • Those numbers are trial numbers and they do not survive contact with an unselected population. In the Swedish Adult Acute Leukemia Registry, of 105 acute promyelocytic leukaemia patients diagnosed 1997-2006, 30 (29%) died within 30 days and 28 (26%) within 14 days, median 4 days; 41% of early deaths were haemorrhagic and 35% of the patients who died early NEVER RECEIVED ATRA at all. Among patients who survived the early window 97% achieved complete remission - so the 90-95% figure is conditional on surviving induction, and the registry's own conclusion is that early-death rates remain very high in an unselected population. A California Cancer Registry series of 772 patients aged 0-39 adds that 30-day mortality fell from 26% to 14% across the ATRA era while 7-day mortality did not improve at all (P = 0.229). Lehmann S et al. 2011
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   An AlphaFold model AF-P29590-F1 covers all 882 residues, and experimental structures exist for the RING (5YUF, 49-99), B-box 1 (2MVW/6IMQ, 120-168), B-box 2 (8J25, 183-236) and the RBCC dimer (8YTC, 46-256).

   AlphaFold Protein Structure Database entry for P29590 (model AF-P29590-F1) · Jumper J, Varadi M, et al.; EMBL-EBI/DeepMind · AlphaFold DB 2026
   ▸What this evidence cannot show

   No experimental structure covers PML beyond residue ~256 except short SIM-region peptides (4WJN, 6UYO series, 546-574). Everything from 260 to 882 in the AlphaFold model is low-confidence and largely predicted-disordered; UniProt annotates disorder at 1-48 and 467-589.

   ▸Evidence that argues against this (1)

   • 8YTC does not resolve the coiled coil. The authors state that the density for residues 236-256 was not visible and that the coiled coil 'might exhibit greater flexibility than previously anticipated', so the 46-256 SIFTS mapping is the construct range, not an observed one. No experimental structure of the PML coiled coil exists as of this check. Tan Y et al. 2024
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Recomputing the Ensembl PML-201 (ENST00000268058) coding exon structure places the end of coding exon 3 at codon 394 and the end of coding exon 6 at codon 552, matching the two UniProt-annotated PML::RARA breakpoints exactly.

   Ensembl transcript ENST00000268058 (PML): exon structure and coding-sequence coordinates, used to convert exon breakpoints into residue numbers · Ensembl; EMBL-EBI · Ensembl REST API 2026
   ▸What this evidence cannot show

   Computed by us; both exons end mid-codon (+1 nt), so the junction codon is chimeric and assigning it to PML is a convention. The agreement with two independently curated UniProt sites is what raises confidence above 'moderate'.

   ▸Evidence that argues against this (1)

   • PML exon numbering is the least stable of the genes in this set, so the exact agreement is a coincidence of transcript choice as much as a validation. The live Ensembl record for PML lists 36 protein-coding transcripts whose exon counts run from 2 to 9 - eleven with 8 exons, seven with 7, seven with 5, four with 6, four with 9 - encoding proteins from well under 641 aa up to 882, and two different transcripts share the canonical 882-aa length with 9 exons. A clinical report saying bcr1 or PML exon 6 gives no information about which of the 36 models was used. Ensembl; EMBL-EBI 2026
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   A 5.3 A cryo-EM structure of the PML RBCC shows that the basic unit is a DIMER formed through the alpha6 helix (V123/F177/L178), that the coiled coil projects away from that interface, and that nuclear-body assembly proceeds by ANTI-PARALLEL coiled-coil pairing between neighbouring dimers; the coiled-coil residues whose mutation abolishes nuclear-body formation are L240, L247 and I279, and the partner-side hydrophobic pockets involve L297, L298, V301, I326, V333, V345, L352, L356 and L359, with crosslinking-mass-spectrometry contacts at T328, S330, K337 and Y339.

   Cryo-EM structure of PML RBCC dimer reveals CC-mediated octopus-like nuclear body assembly mechanism · Tan Y, Li J, Zhang S, Zhang Y, Zhuo Z, Ma X, Yin Y, Jiang Y, Cong Y, Meng G · Cell Discov 10:118 2024
   PMID 39587079DOI
   ▸What this evidence cannot show

   FULL TEXT read (PMC11589706). Overall resolution is only 5.3 A (core ~4.4 A) on a sub-50 kDa MBP-fused fragment spanning residues 46-256, and the authors state plainly that the density for coiled-coil residues 236-256 was NOT visible; the anti-parallel coiled-coil geometry is an AlphaFold2-multimer prediction supported by BS3 crosslinks and by mutagenesis, not a solved coiled-coil structure. The cellular assays use GFP-tagged PML-IV (P29590-5) overexpressed in PML-knockout HeLa; PML-IV shares the RBCC with canonical PML-I so these residue numbers apply, but they are a nuclear-body-formation readout, not a measurement of helix extent.

   ▸Evidence that argues against this (1)

   • A crystallographic study of the same protein says the essential oligomer is a RING TETRAMER, and it has the in-vivo test the cryo-EM work lacks. X-ray crystallography plus SAXS show PML RING tetramerising through highly conserved PML-specific sequences required for nuclear-body assembly and for PML sumoylation; residues implicated in RING DIMERISATION in other TRIM proteins also contribute to tetramer stability, and - the load-bearing result - impaired RING tetramerisation abolishes PML/RARA-driven leukaemogenesis in vivo and arsenic-induced differentiation ex vivo. The 5.3 A cryo-EM map covers residues 46-256 with the 236-256 coiled-coil density absent, so it cannot exclude the tetramer as the functional unit. Wang P et al. 2018
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Member databases disagree with UniProt about PML's coiled coil by an order of magnitude and annotate a structured C-terminal domain that this node omitted: Pfam PF12126 / InterPro IPR021978 'PML-like, coiled-coil' spans 240-570 against UniProt's 228-253; InterPro IPR000315 and PROSITE PS50119 place B-box 2 at 184-222 against UniProt's 183-236, while CDD cd19770 gives 187-237; and Pfam PF25244 / InterPro IPR057617 define a 'PML, C-terminal domain' at 777-852 that has no UniProt feature.

   InterPro member-database domain matches for UniProt P29590 (Pfam, PROSITE, SMART, CDD, CATH-Gene3D, SUPERFAMILY) · InterPro Consortium; EMBL-EBI · InterPro 2026
   ▸What this evidence cannot show

   Profile-model match extents, not measurements, and PF12126 in particular is a long low-complexity coiled-coil model whose C-terminal extent to 570 is almost certainly an over-call in the other direction. Retrieved 2026-09-19 (database record, no full text to read). The 777-852 C-terminal domain is a recent Pfam addition and we found no experimental structure covering it; it is recorded as an annotation gap, not as a verified fold.

   ▸Evidence that argues against this (1)

   • The only crystal structure of PML B-box 2 agrees with none of the four profile calls. PDB 8J25 (Crystal structure of PML B-box2 mutant, X-ray, 2.6 A) is an MBP fusion whose PML segment aligns entity residue 375 to UniProt 183, with author numbering coinciding with UniProt numbering; the RCSB instance record shows auth 183-185 and 228-236 unobserved, so the OBSERVED B-box 2 is 186-227. That is 3 residues shorter at the N-terminus than UniProt and InterPro/PROSITE, 1 residue shorter than CDD, and 5 residues LONGER at the C-terminus than PROSITE's 222 - the measured extent falls outside every profile call in at least one direction. RCSB Protein Data Bank 2026

• related to →PML::RARA, t(15;17)(q24;q21)

  Counter-evidence added cycle 41: the essential PML oligomer may be a RING TETRAMER, not the cryo-EM dimer - impaired RING tetramerisation abolishes PML/RARA-driven leukaemogenesis in vivo and arsenic-induced differentiation ex vivo (Wang 2018, PMID 29599493). And the 90-95% remission figure is conditional on surviving induction: 29% of 105 unselected Swedish registry patients died within 30 days, 35% of them never receiving ATRA (Lehmann 2011, PMID 21502956).

• related to →PML::RARA, t(15;17)(q24;q21)

  Coordinate correction (Tan 2024, PMID 39587079): the PML coiled coil this node drew as 228-253 (UniProt) is a 26-residue stub; mutagenesis and crosslinking put the functional anti-parallel interface out to at least L359, so the oligomerisation module PML::RARA actually retains is ~130 residues, not 26. Both bcr1 (1-552) and bcr3 (1-394) keep all of it, so no fate call changes -- only the figure does.

• related to →Hypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

  Counter-evidence to a small-interface premise: PML self-association is dominated by a long anti-parallel coiled coil with at least three separate hydrophobic pockets (L240/L247/I279 against L297-L359), and single point mutants lose only 74-84% of self-interaction. A distributed interface of that size is hard to drug and is not obviously a single mutable hot spot.

• related to →RARA — domain architecture and residue coordinates

  Dispute recorded rather than resolved: the PML RING tetramer model (Wang 2018) and the RBCC dimer + coiled-coil-pairing model (Tan 2024) disagree about what the basic PML unit is. Anything downstream that assumes a fixed stoichiometry for PML::RARA should treat the oligomeric state as open.

• part of →PML::RARA, t(15;17)(q24;q21)

  5-prime partner domain architecture; RING, B-boxes and coiled coil are retained

• fuses with →RARA — domain architecture and residue coordinates

  PML::RARA, t(15;17)

• supports →Hypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

  the retained PML coiled coil oligomerises the fusion, a second non-KMT2A instance of the mechanism

• supports →Why PML::RARA became curable — and what KMT2A::AFF1 does not share with it

  RING tetramerisation is required for PML/RARA leukaemogenesis and for arsenic-induced differentiation

• supports →Breakpoint anatomy, and an audit of the prognosis claim

  bcr1 (PML 1-552) and bcr3 (PML 1-394) differ by the entire 395-552 region including the NLS

• ← fuses withRARA — domain architecture and residue coordinates

  t(15;17)(q24;q21)

• ← related toHypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

  Structural counterweight carried over from cycle 32 (PMID 39587079): PML self-association is a long anti-parallel coiled coil with at least three hydrophobic pockets, and single point mutants lose only 74-84% of self-interaction. Hard to escape with one mutation, and equally hard to occupy with one molecule -- the two halves of this hypothesis trade off.

• ← related toExperiment (proposed) — cross-family interface point-mutant panel: is the obligate interface real in each fusion?

  Design risk added (Tan 2024, PMID 39587079): PML self-association runs through a coiled coil in which each dimer's CC helix engages a neighbouring dimer's, and single point mutants lose only 74-84% of self-interaction. A panel that reads 'program survives one interface mutation' as redundancy will misread partial disruption; combinatorial mutants and matched abundance controls are needed.

• ← related toHypothesis — forced-oligomerisation interfaces are the mutation-resistant drug targets the fusion field has been ignoring

  Distributed interface, low-resolution basis (Tan 2024, PMID 39587079): PML self-assembly is a multivalent octopus-like set of anti-parallel inter-dimer CC contacts, from a 5.3 A cryo-EM map plus AlphaFold2 and XL-MS - poor material for compact single-site occupancy.

• ← related toWhy PML::RARA became curable — and what KMT2A::AFF1 does not share with it

  PML is the drug target, not just the dimeriser (Zhang 2010, PMID 20378816): arsenic binds RBCC zinc-finger cysteines and INDUCES oligomerisation, driving UBC9 recruitment, SUMOylation and degradation.

• ← co-occurs withHIRA - the chaperone that puts new H3.3 into transcribed gene bodies, and therefore the machine that runs this graph's clearance clock

  HIRA, UBN1 and ASF1A localise to PML nuclear bodies, requiring the Sp100A SIM

• ← related toDAXX - the strongest of five selective dependencies among 254 unwritten interactors but it FAILS Bonferroni correction (p=0.0128, rank 4 of 254), and not one paper connects it to this disease

  PML organises the domains where ATRX/DAXX deposit H3.3

• part of →PML::RARA, t(15;17)(q24;q21)

  5' partner domain coordinates

• ← related toPML nuclear bodies and PML::RARA — the best-worked case of a fusion destroying a nuclear compartment, and what it actually predicts for KMT2A

  PML nuclear-body assembly is the structural context for the fusion's mislocalisation

• ← related toPML / TRIM19 (P29590) - the scaffold of PML nuclear bodies, co-purified with all seven MLL fusions, and never once studied as a KMT2A-fusion partner

  same protein, domain-level entry written for the PML-RARA beat