The ZNF384-fusion immunophenotype - weak CD10 with aberrant CD13/CD33 - and the one marker that separates it from KMT2A-rearranged disease runs the OPPOSITE way to expectation
also called ZNF384r immunophenotype · B/myeloid ZNF384 · CD10 weak CD13 CD33 aberrant
Leukaemia cells carrying a ZNF384 fusion look half B-cell and half myeloid: they carry myeloid badges (CD13, CD33) that B-cells should not have, and they are faint for the usual B-cell badge CD10. That mixed look is what first made pathologists suspect the subtype existed before anyone had sequenced it.
The same thing, for a scientist
ZNF384-rearranged leukaemia has a reproducible B/myeloid immunophenotype: **weak or absent CD10 plus aberrant CD13 and/or CD33**, first established as a common feature of the whole fusion family by Hirabayashi et al. in 291 children [E1] and confirmed in the 218-patient Ponte di Legno series, where CD13 ran 38-100% and CD33 78-100% positive and 14-51% of cases were CD10-negative depending on partner [E2]. The phenotype is variable enough to cross the diagnostic boundary: **48% of paediatric B/myeloid MPAL cases in the St Jude ALAL cohort carried a ZNF384 rearrangement**, and the gene-expression profiles of ZNF384-rearranged B/myeloid MPAL and ZNF384-rearranged B-ALL were **indistinguishable** [E3]. **The comparison this graph most needs, against KMT2A-rearranged disease, runs in the direction a reader will not predict:** in 43 adult ZNF384-rearranged cases against 29 KMT2A-rearranged, CD10 was **higher** in ZNF384-r (median 36.16% of cells, MFI 4.50) than in KMT2A-r (4.53%, MFI 2.06; p < 0.0001 for both), even though ZNF384-r is CD10-low relative to BCP-ALL generally [E4]. **So 'CD10-weak, myeloid-marker-positive B-ALL' describes both subtypes, and CD10 does not distinguish them by being low - it distinguishes them by being less low.**
The founding observation and its numbers
Hirabayashi et al. screened 291 children with BCP-ALL enrolled in a single clinical trial and found ZNF384-related fusions in 4.1%. Across 22 fusion-positive patients:
| marker | ZNF384-fusion cases | TCF3-PBX1 controls | ZNF384-wild-type controls |
|---|---|---|---|
| CD10 | mean 19.44 +/- 18.23% (range 0.39-67.34) | 98.66 +/- 0.89% (P < 0.001) | 77.71 +/- 25.79% (P < 0.001) |
| CD13 | mean 10.70 +/- 10.37% | 0.26 +/- 0.33% | 2.76 +/- 7.9% |
| CD33 | mean 59.86 +/- 34.01% | 0.59 +/- 0.76% | 6.63 +/- 1.72% |
and 7 of 22 (31.82%) and 17 of 22 (77.27%) exceeded 20% expression of CD13 and CD33 respectively 1.
Note the spread, not just the mean. CD10 in the fusion-positive group runs from 0.39% to 67.34%. A single case can be CD10-bright and still carry the fusion, which is exactly what the one adult TAF15::ZNF384 case report emphasised about itself: "Unlike most of the cases where CD10 expression was absent or weak, in our case CD10 was highly expressed" 5. The phenotype is a population tendency, not a rule about a patient.
Against KMT2A - and why this is the comparison the graph was missing
This corpus already holds concept-flow-cytometry-immunophenotype-kmt2a-r, mechanism-kmt2a-mpal-mixed-phenotype and concept-immunophenotype-as-a-genotype-surrogate. The gap it had was any head-to-head comparison of a second fusion family against KMT2A on the same flow panel.
Wang et al. provide one, in adults: 43 ZNF384-r, 45 BCR-ABL1, 29 KMT2A-r, 44 other BCP-ALL 4. The result that matters here:
- Against non-ZNF384 BCP-ALL as a whole, ZNF384-r has significantly lower CD10 and higher CD33, CD13 and CD123.
- Against KMT2A-r specifically, ZNF384-r has significantly higher CD10: median 36.16% (3.63-94.79) vs 4.53% (0.03-21.00), MFI 4.50 (0.86-32.26) vs 2.06 (0.87-4.04), p < 0.0001 for both.
Both statements are true and they point in opposite directions, because the reference group is different. A beat that remembers only 'ZNF384-r is CD10-weak' will predict the wrong sign when asked to separate it from KMT2A-r. The paper's own answer is a composite score - CD10%, CD33 MFI, CD13%, CD123 MFI - reaching 76.74% sensitivity / 91.53% specificity in the analysis cohort and 87.50% / 91.30% in validation, which is the honest reading: no single marker does it.
CD123 is reported as higher in ZNF384-r than in all three other groups (p < 0.01), stated as a first report 4. That is a genuinely new discriminator and, if it holds, it is the one that is not shared with KMT2A-r.
Crossing into MPAL - and the finding that the diagnosis, not the biology, moves
Alexander et al. centrally reviewed 159 potential paediatric cases of acute leukaemia of ambiguous lineage and confirmed 115 by WHO criteria, of which 35 were B/myeloid MPAL. Rearrangement of ZNF384 was present in 48% of B/myeloid MPAL cases, involving TCF3 (n = 8), EP300 (n = 5), TAF15 (n = 1) and CREBBP (n = 1) 3.
And then the decisive sentence:
"Analysis of a diverse range of acute leukaemias, including AML, showed that the gene expression profiles (GEPs) of ZNF384r B/M MPAL and B-ALL were indistinguishable"
"Thus, ZNF384r defines a distinct subtype of acute leukaemia with a variable immunophenotype ranging from B-ALL to B/M MPAL."
This is the same conclusion mechanism-kmt2a-mpal-mixed-phenotype reaches for KMT2A - that the immunophenotypic heterogeneity is largely independent of genetics and the entity is defined by the lesion, not the surface markers. Two unrelated fusion families, the same answer. That convergence is the comparative result, and it is stronger than either case alone.
| diagnostic score | Wang et al. composite of CD10%, CD33 MFI, CD13% and CD123 MFI; sensitivity 76.74% and specificity 91.53% in the analysis cohort, 87.50% and 91.30% in validation |
|---|
markers up
[ "CD13", "CD33", "CD123" ]
markers down
[ "CD10 (relative to BCP-ALL overall, but HIGHER than in KMT2A-rearranged)" ]
measurements — the numbers behind the chart
[
{
"quantity": "percent_positive",
"subject": "CD10 in ZNF384-fusion-positive childhood BCP-ALL",
"value": 19.44,
"units": "% of blasts",
"n": 22,
"universe": "22 ZNF384-fusion-positive patients among 291 children with BCP-ALL in a single Japanese clinical trial; SD 18.23, range 0.39-67.34",
"null": 77.71,
"note": "the null is the ZNF384-wild-type BCP-ALL arm of the same study (77.71 +/- 25.79%), not zero; P < 0.001"
},
{
"quantity": "percent_positive",
"subject": "CD33 in ZNF384-fusion-positive childhood BCP-ALL",
"value": 59.86,
"units": "% of blasts",
"n": 22,
"universe": "same 22 fusion-positive patients; SD 34.01",
"null": 6.63,
"note": "null is the ZNF384-wild-type arm (6.63 +/- 1.72%). 17 of 22 patients exceeded 20% CD33."
},
{
"quantity": "percent_positive",
"subject": "CD13 in ZNF384-fusion-positive childhood BCP-ALL",
"value": 10.7,
"units": "% of blasts",
"n": 22,
"universe": "same 22 fusion-positive patients; SD 10.37",
"null": 2.76,
"note": "null is the ZNF384-wild-type arm (2.76 +/- 7.9%). Only 7 of 22 exceeded 20% CD13, so CD13 is the weaker of the two myeloid markers here despite being the one usually quoted first."
},
{
"quantity": "percent_positive_median",
"subject": "CD10 in adult ZNF384-rearranged BCP-ALL",
"value": 36.16,
"units": "% of blasts (median)",
"n": 43,
"universe": "43 adult ZNF384-rearranged BCP-ALL, single centre; range 3.63-94.79",
"null": 4.53,
"note": "THE NULL HERE IS THE KMT2A-REARRANGED ARM (n = 29, median 4.53%, range 0.03-21.00), and ZNF384-r is EIGHT-FOLD HIGHER, p < 0.0001. Against non-ZNF384 BCP-ALL as a whole the same marker is significantly LOWER in ZNF384-r. The sign of the difference depends entirely on which comparator you pick."
},
{
"quantity": "mfi_median",
"subject": "CD10 median fluorescence intensity in adult ZNF384-rearranged BCP-ALL",
"value": 4.5,
"units": "MFI",
"n": 43,
"universe": "same adult cohort; range 0.86-32.26",
"null": 2.06,
"note": "null is the KMT2A-rearranged arm, MFI 2.06 (0.87-4.04), p < 0.0001"
},
{
"quantity": "proportion",
"subject": "ZNF384 rearrangement among paediatric B/myeloid MPAL",
"value": 0.48,
"units": "fraction",
"n": null,
"universe": "B/myeloid MPAL cases within a centrally reviewed cohort of 115 WHO-confirmed paediatric acute leukaemias of ambiguous lineage, of which 35 were B/myeloid MPAL",
"null": null,
"note": "QUOTED AS PUBLISHED AND NOT RECONCILED. The paper names 15 ZNF384-rearranged B/M MPAL cases (TCF3 8, EP300 5, TAF15 1, CREBBP 1) and two figure legends give n = 15; 15/35 = 42.9%. The denominator behind 48% is not stated. n is left null deliberately."
}
]- CD19-directed therapy is the live risk. The Ponte di Legno authors write that 'immunotherapy, such as anti-CD19 chimeric antigen receptor T-cell therapy, should be used carefully, if at all, in patients with ZNF384 rearrangements because of the risk of lineage switch to myeloid leukemia as a cause of relapse', and report that lineage switch after CAR-T has been observed in a patient with TCF3-ZNF384. A subtype that is half myeloid before treatment has an obvious escape route from a B-lineage-directed agent, and this graph already records the same failure mode for KMT2A in risk-lineage-switch-cd19-directed-therapy.
- CD33-directed therapy has the mirror-image problem. CD33 is positive in 78-100% of these cases, which makes gemtuzumab-class agents superficially attractive, but the same plasticity that produces CD33 expression can remove it, and a CD33-negative B-lymphoid escape is as available as a CD19-negative myeloid one.
- A four-parameter flow score with 91% specificity is not a genotype. Wang et al.'s score misses roughly a quarter of ZNF384-rearranged cases at the analysis-cohort operating point, and this graph's own concept-immunophenotype-as-a-genotype-surrogate exists because that substitution has gone wrong before. Immunophenotype should trigger a molecular test, not replace one - and the ZNF384 case is sharper than most, because the last-exon indel class in mechanism-znf384-c-terminal-truncation-phenocopies-the-fusion is invisible to the molecular test as well.
- Is CD123 really a ZNF384-versus-KMT2A discriminator? Wang et al. report it as higher in ZNF384-r than in all three comparator groups and flag it as a first report, but the CD123 comparison is available to this beat only through an abstract and only in adults. CD123 is an established target in this graph's therapeutic space, so if it is genuinely higher in ZNF384-r than in KMT2A-r that is an actionable difference and it needs replication in children.
- Why does CD10 sit LOWER in KMT2A-rearranged than in ZNF384-rearranged disease when both are described as CD10-weak B/myeloid leukaemias? If both fusions arrest differentiation at a similar point, the marker should not separate them eight-fold. Either the arrest points differ, or the KMT2A-r arm in that cohort was partner-skewed toward KMT2A::AFF1, and the full text would settle it.
- Does the ZNF384 fusion protein occupy the CD10/MME, CD33/SIGLEC3 and CD123/IL3RA loci directly? A CUT&RUN or ChIP track over those three genes in a ZNF384-rearranged sample would convert an immunophenotype into a mechanism, and the data to do it already exist - Zhao et al. deposited ZNF384 CUT&RUN from a TCF3-ZNF384 PDX and an EP300-ZNF384 cell line. One paper already shows EP300-ZNF384 transactivating IL3RA, which is CD123; nobody has closed the same loop for CD10 or CD33.
- Lineage-assignment criteria changed between WHO revisions and the 2022 ICC. Since 48% of B/myeloid MPAL is ZNF384-rearranged and ZNF384-r B-ALL and ZNF384-r MPAL have indistinguishable expression profiles, how much of the recorded MPAL/B-ALL split for this subtype is a criteria artefact rather than a biological boundary? The same question is open for KMT2A in mechanism-kmt2a-mpal-mixed-phenotype and could be answered for both at once.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 291 children with BCP-ALL from a single clinical trial, 4.1% carried a ZNF384-related fusion, and the fusion-positive cases shared a characteristic immunophenotype of weak CD10 with aberrant CD13 and/or CD33: CD10 mean 19.44 +/- 18.23% (range 0.39-67.34%) versus 98.66 +/- 0.89% in TCF3-PBX1 and 77.71 +/- 25.79% in ZNF384-wild-type cases (P < 0.001 for both); CD13 mean 10.70 +/- 10.37% and CD33 mean 59.86 +/- 34.01%, with 7 of 22 and 17 of 22 patients exceeding 20% expression respectively.
ZNF384-related fusion genes define a subgroup of childhood B-cell precursor acute lymphoblastic leukemia with a characteristic immunotype. · Hirabayashi S, Ohki K, Nakabayashi K, Ichikawa H, Momozawa Y, Okamura K, et al., Kiyokawa N · Haematologica 2016What this evidence cannot show
Published in Haematologica 2017;102(1):118-129 with a 2016 epub date; PubMed records the publication date as 2016-09-15 and that is what is recorded here. The numeric marker values were read from the PMC rendering of PMC5210242 in this session via a page fetch, NOT from the abstract, which carries none of them - Europe PMC and NCBI efetch both refused the full XML for this publisher. n = 22 fusion-positive patients for the marker statistics. The control group labelled 'wild-type ZNF384' is BCP-ALL without a ZNF384 fusion, not healthy B-cell precursors.
Evidence that argues against this (1)
- The range is the counter-evidence to the rule. CD10 in fusion-positive cases runs 0.39% to 67.34%, so 'weak CD10' is a statement about a group mean with a 170-fold spread inside it, and individual cases sit on both sides of any threshold. An independently reported adult TAF15::ZNF384 case was explicitly CD10-bright. Any diagnostic use of CD10 weakness as a ZNF384 flag will therefore miss a substantial minority, which is precisely why Wang et al. built a four-parameter score rather than a CD10 cutoff. Georgakopoulos N et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 218 children with ZNF384-rearranged B-ALL assembled by 16 international consortia, the majority of cases expressed the myeloid-associated antigens CD13 (38-100% across partner groups) and CD33 (78-100%), and 14-51% of cases were negative for CD10.
Clinical characteristics and outcomes of B-ALL with ZNF384 rearrangements: a retrospective analysis by the Ponte di Legno Childhood ALL Working Group. · Hirabayashi S, Butler ER, Ohki K, Kiyokawa N, Bergmann AK, Moricke A, Boer JM, Cave H, Cazzaniga G, Yeoh AEJ, et al., Moorman AV, Manabe A · Leukemia 2021What this evidence cannot show
FULL TEXT READ via Europe PMC fullTextXML for PMC8550960. The ranges quoted are across partner-gene subgroups and the per-subgroup values live in Supplementary Table 3, which was not retrieved. expected_absent_terms names 'steroid' deliberately: this paper contains no steroid-response data at all - the words steroid, prednis* and dexamethasone return zero hits in the complete fetched text - which matters because the steroid-resistance finding for TCF3::ZNF384 comes from Hirabayashi 2017 and must not be attributed to this cohort. MPAL cases were EXCLUDED from this series by design, so it cannot speak to the B/myeloid boundary.
Evidence that argues against this (1)
- This series deliberately excluded MPAL cases 'because they are frequently treated off-study and can receive hybrid therapies', and excluded 18 cases identified at relapse or in refractory disease. Both exclusions remove exactly the patients whose immunophenotype is most aberrant, so the CD13/CD33/CD10 distributions reported here are conditioned on a case having been diagnosable and treatable as B-ALL. The marker ranges are therefore a floor on the aberrancy of the family, not an estimate of it. Hirabayashi S et al. 2021
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In a centrally reviewed paediatric cohort of 115 WHO-confirmed acute leukaemias of ambiguous lineage including 35 B/myeloid MPAL, ZNF384 rearrangement was present in 48% of B/myeloid MPAL cases, involving TCF3 (n = 8), EP300 (n = 5), TAF15 (n = 1) and CREBBP (n = 1); the gene expression profiles of ZNF384-rearranged B/myeloid MPAL and ZNF384-rearranged B-ALL were indistinguishable, and the cases clustered together without separation by phenotype.
The genetic basis and cell of origin of mixed phenotype acute leukaemia. · Alexander TB, Gu Z, Iacobucci I, Dickerson K, Choi JK, Xu B, et al., Mullighan CG · Nature 2018What this evidence cannot show
FULL TEXT READ via the PMC author-manuscript rendering of PMC6195459 (Europe PMC fullTextXML returns HTTP 500 for this PMCID). **THE 48% DOES NOT RECONCILE WITH THE STATED COHORT.** The paper gives 35 B/myeloid MPAL cases and names 15 ZNF384-rearranged ones (8+5+1+1), and two figure legends independently state n = 15 ZNF384r B/M MPAL. 15/35 is 42.9%, not 48%; 15/31 would give 48%. The paper does not state the denominator behind its own 48%, so this entry quotes the figure as published and records that its numerator/denominator pair cannot be recovered from the text. Do not construct a fraction from it.
Evidence that argues against this (1)
- Two internal tensions in this paper bear on the phenotype claim. First, the unrecoverable denominator above: 48% is quoted everywhere in the field, and the arithmetic in its source paper gives 42.9% from the numbers the paper itself prints. Second, the transcriptional-identity claim is not unqualified - KDM6A alterations were observed ONLY in ZNF384r MPAL and not in ZNF384r B-ALL, and GSEA found ZNF384r B/M MPAL arrested at a MORE mature stage than other B/M MPAL while showing enrichment of stem-cell and ETP-ALL programmes relative to B-ALL. 'Indistinguishable GEPs' and 'a recurrent genetic difference between the two presentations' are both in the same paper. Alexander TB et al. 2018
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In adult BCP-ALL, comparing 43 ZNF384-rearranged cases with 29 KMT2A-rearranged, 45 BCR-ABL1 and 44 other BCP-ALL, ZNF384-rearranged cases had significantly HIGHER CD10 than KMT2A-rearranged cases - median 36.16% (3.63-94.79) versus 4.53% (0.03-21.00) of cells, and MFI 4.50 (0.86-32.26) versus 2.06 (0.87-4.04), p < 0.0001 for both - while having lower CD10 and higher CD33, CD13 and CD123 than non-ZNF384 BCP-ALL overall. CD123 was higher in ZNF384-rearranged than in all three other groups (p < 0.01), reported for the first time.
Immunophenotypic characteristics of ZNF384 rearrangement compared with BCR-ABL1, KMT2A rearrangement, and other adult B-cell precursor acute lymphoblastic leukemia. · Wang YZ, et al. · Cytometry B Clin Cytom 2022What this evidence cannot show
ABSTRACT ONLY, and this is a hard limit, not a shortcut: Wiley returns HTTP 403 to every route tried, no PMC record exists, and Unpaywall reports the article CLOSED with no repository copy. Consequences: (a) the ONLY ZNF384-r versus KMT2A-r figure available is CD10, percentage and MFI; (b) CD34, CD19, CD15, CD65, NG2/CSPG4 and MPO are not mentioned in the abstract at all and this beat has no values for them; (c) the CD33/CD13 superiority statement in the abstract is explicitly 'compared with BCR-ABL1 and other BCP-ALL', i.e. KMT2A-r is excluded from that comparison and it must NOT be extended to KMT2A-r. This is an ADULT cohort; this project's population is paediatric. The study is retrospective and single-centre.
Evidence that argues against this (1)
- The comparison this item is cited for is the one the abstract is weakest on. Because the full text is unreachable, this beat cannot check the KMT2A-r arm's composition - KMT2A-r in adults is partner-heterogeneous and the CD10-negative phenotype is strongest in KMT2A::AFF1, so a KMT2A-r arm dominated by one partner would produce this contrast for a reason that has nothing to do with ZNF384. n = 29 in that arm. The direction of the CD10 difference is also the opposite of what the widely-quoted 'ZNF384r is CD10-weak' summary predicts, which is a reason to want the full text rather than a reason to doubt the abstract, but the claim should be carried as moderate until somebody reads the tables. Hirabayashi S et al. 2016
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
An adult TAF15/ZNF384-rearranged early pre-B ALL case is reported in which CD10 was highly expressed, explicitly contrary to the subtype's usual absent-or-weak CD10, alongside the expected co-expression of myeloid markers CD13 and/or CD33.
An Adult Patient with Early Pre-B Acute Lymphoblastic Leukemia with t(12;17)(p13;q21)/ZNF384-TAF15. · Georgakopoulos N, Diamantopoulos P, Micci F, Giannakopoulou N, Zervakis K, Dimitrakopoulou A, Viniou NA · In Vivo 2018What this evidence cannot show
n = 1, a case report, and indexed by PubMed as both Case Report and Review. The fusion was established by FISH showing rearrangement of ZNF384 and TAF15, not by sequencing the junction, so the reading frame and breakpoint are unknown for this patient. The title writes the fusion 'ZNF384-TAF15', which is chromosome order (12p13 before 17q12) and NOT a claim that ZNF384 is the 5' partner; see fusion-taf15-znf384 for why that naming trap matters.
Evidence that argues against this (1)
- A single CD10-bright case cannot refute a group tendency, and this item is cited only to fix the range, not to challenge the phenotype. The counter-consideration that does bite is ascertainment: an atypical case is more likely to be written up as a case report than a typical one, so the published case-report literature systematically over-represents phenotypic outliers and cannot be used to estimate how often CD10 is bright. The group-level answer has to come from cohorts, and there the Hirabayashi range (0.39-67.34%) already shows the spread. Hirabayashi S et al. 2016
weak CD10 with aberrant CD13/CD33 is common to the whole ZNF384 fusion family
ZNF384 rearrangement in 48% of paediatric B/myeloid MPAL; ZNF384r MPAL and ZNF384r B-ALL have indistinguishable expression profiles
two unrelated fusion families reach the same conclusion: the entity is defined by the lesion, not by the surface markers
head-to-head flow comparison: CD10 is EIGHT-FOLD HIGHER in ZNF384-r than in KMT2A-r, the opposite of the predicted direction
a four-parameter flow score reaching 91% specificity, and the lesion class it cannot see
the noncanonical lesions reproduce the immunophenotype but add CD66c, which is rare in canonical ZNF384r
all six founding cases had dull or negative CD10 with aberrant CD13 and/or CD33
the case was within the ZNF384-fusion group sharing weak CD10 with aberrant CD13/CD33