Oncofusion

start / graph / gene-il3ra

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q45-immunophenotype

IL3RA / CD123 - not just a myeloid target: the quartile of CD123 expression that carries the KMT2A rearrangements is the quartile that dies

also called IL3RA · CD123 · IL3R · Interleukin-3 receptor subunit alpha · IL-3Ra · P26951

CD123 is the receptor a blood cell uses to hear the growth signal interleukin-3, and leukaemia cells often carry far more of it than normal cells do. In children with AML, the quarter of patients with the most CD123 on their blasts were also the patients most likely to carry a broken KMT2A gene - and the quarter most likely to relapse and die.

The same thing, for a scientist

IL3RA (UniProt **P26951**, IL3RA_HUMAN, **378 aa**, single reviewed human record, synonym IL3R) encodes the alpha chain of the interleukin-3 receptor, known clinically as **CD123**. This graph names it in 9 entries (as CD123) and 5 (as IL3RA), and the only entry covering it, `target-cd33-cd123-myeloid-directed-kmt2a-r`, **covers it jointly with CD33** - the summary-entry pattern the PI's standing instruction forbids. `gene-cd33` and this entry split it. **What CD123 means here beyond 'a myeloid target'.** COG AAML1031 prospectively measured CD123 protein by multidimensional flow on **1,040 diagnostic paediatric AML marrows** and split them into quartiles. The top quartile (Q4, n = 260) had a **higher prevalence of high-risk KMT2A rearrangements and FLT3-ITD (P < .001 for both)** and a lower prevalence of low-risk t(8;21), inv(16) and CEBPA mutation. Q1-Q3 were indistinguishable from each other on outcome; **Q4 alone had relapse risk 53% vs 39%, EFS 49% vs 69% and OS 32% vs 50%, all P < .001**, and CD123 stayed independently significant in multivariable Cox regression including every contemporary high-risk cytogenetic and molecular marker [E1]. **Three things follow that the joint target entry cannot say.** 1. **CD123 is a threshold effect, not a gradient.** Three quarters of patients are outcome-equivalent and the signal lives entirely in the top quartile. A trial that enrols "CD123-positive" patients is not enrolling the population in which CD123 carried information. 2. **The co-segregation with KMT2A-r and FLT3-ITD is the mechanism candidate.** This graph already carries FLT3 as a cooperating lesion and `gene-cspg4` now records NG2 driving ligand-independent FLT3 signalling. CD123 sits in the same cytokine-receptor neighbourhood, and JMJD2/KDM4 demethylases are required for *Il3ra* expression and AML cell survival - so CD123 is an epigenetically maintained gene in this disease, not merely a surface convenience [E2]. 3. **In B-lineage KMT2A-rearranged disease, CD123 goes the other way.** In 1,382 B-ALL, KMT2A-rearranged cases had significantly **lower** CD123 mean fluorescence intensity than KMT2A-germline cases [E3]. CD123 is a t(1;19)/TCF3-PBX1 predictor in B-ALL flow scoring systems, not a KMT2A one [E4]. **The same antigen points in opposite directions in the two lineages of the same rearrangement.**

Go deeper
What this means for a CD123-directed agent in KMT2A-rearranged disease

The therapeutic case for CD123 in this disease rests on the AML side. Tagraxofusp (IL-3 fused to diphtheria toxin), CD123 CARs and CD123 bispecifics all exist, and the COG data give a defensible enrichment argument: target the quartile that carries the KMT2A rearrangements and the bad outcomes.

The case does not transfer to infant B-ALL, where the same rearrangement is associated with lower CD123 3. An agent selected on AML data and applied to infant KMT2A-r B-ALL would be aiming at an antigen the disease expresses less of than its comparator - and if the escape from B-lineage-directed therapy is a switch to the myeloid phenotype, the antigen arrives only after the relapse it was meant to prevent.

The toxicity is the standard CD123 problem and it is worse in infants. CD123 is expressed on normal haematopoietic stem and progenitor cells, on basophils and on plasmacytoid dendritic cells; the class signal for IL-3-receptor-directed agents is capillary leak. Nothing in the COG expression paper addresses toxicity, and nothing in this beat found a CD123-directed agent tested in infants.

Cell type, cell state, localisation, epigenetics
  • Cell type. Myeloid progenitors, basophils, plasmacytoid dendritic cells, and a fraction of normal CD34-positive HSPCs. On blasts: high in a quartile of paediatric AML enriched for KMT2A-r and FLT3-ITD 1; relatively low in KMT2A-rearranged B-ALL 3.
  • Cell state. A cytokine-receptor level, so it is a signalling-state marker rather than a lineage identity marker - which is consistent with it being a continuous quantity whose top tail is what matters.
  • Localisation. Plasma membrane, type I transmembrane, 378 residues; functions as the alpha chain of a heterodimer with the common beta chain CSF2RB 5.
  • Epigenetics. Jmjd2/Kdm4 demethylases are required for expression of Il3ra and for AML cell survival 2 - the cleanest link in this beat between the graph's chromatin layer and a surface marker other than CSPG4 and PROM1.
Facts and figures
uniprotP26951
gene symbolIL3RA
length AA378
cd numberCD123
b lineage directionKMT2A-rearranged B-ALL shows LOWER CD123 MFI than KMT2A-germline B-ALL (PMID 38468826); CD123 is a t(1;19)/TCF3-PBX1 predictor in B-ALL flow scoring, not a KMT2A one (PMID 30730614)
accession resolutiongene_exact:IL3RA AND organism_id:9606 AND reviewed:true, fields include gene_primary and length; one row; primary name read back as IL3RA; length 378 asserted for verify-accessions.py.
cog aaml1031 quartiles
{
  "n": 1040,
  "q4_enriched_for": [
    "KMT2A rearrangement",
    "FLT3-ITD"
  ],
  "q4_depleted_for": [
    "t(8;21)",
    "inv(16)",
    "CEBPA mutation"
  ],
  "q4_vs_q1_3_relapse_risk_pct": [
    53,
    39
  ],
  "q4_vs_q1_3_efs_pct": [
    49,
    69
  ],
  "q4_vs_q1_3_os_pct": [
    32,
    50
  ],
  "nct": "NCT01371981",
  "pmid": "34855461",
  "title": "CD123 Expression Is Associated With High-Risk Disease Characteristics in Childhood Acute Myeloid Leukemia: A Report From the Children's Oncology Group.",
  "authors": "Lamble AJ, Eidenschink Brodersen L, Alonzo TA, et al.",
  "venue": "J Clin Oncol",
  "year": 2022,
  "title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
How this could bite us back
  • CD123 is on normal haematopoietic stem and progenitor cells, basophils and plasmacytoid dendritic cells, and the class toxicity of IL-3-receptor-directed agents is capillary leak syndrome. The COG expression paper that makes the enrichment argument reports no toxicity data at all, and nothing found in this beat tests a CD123-directed agent in infants.
  • The prognostic signal lives entirely in the top quartile. A trial that enrols on 'CD123-positive by flow' is enrolling three quartiles in which CD123 carried no outcome information, and will dilute any effect it might have had. Any CD123 strategy must state its threshold and where that threshold came from.
  • A CD123 agent selected on AML data and applied to infant KMT2A-rearranged B-ALL is aiming at an antigen that genotype expresses LESS of than its comparator. If the myeloid phenotype only appears at lineage switch, the antigen arrives after the relapse the agent was meant to prevent.
  • CD123 co-segregates with FLT3-ITD in the high-expression quartile, and this graph carries FLT3 as a cooperating lesion with its own inhibitors. A CD123-directed and a FLT3-directed strategy are competing for the same patients, and no study in this beat compares them.
Still open
  1. Is CD123 in the top quartile a marker of the KMT2A-rearranged clone or of the FLT3-ITD subclone within it? Both co-segregate with Q4 and the COG analysis does not separate them.
  2. Does CD123 rise at lineage switch in KMT2A-rearranged B-ALL? If it does, it is a switch-detection marker rather than a diagnosis-time target, which is a different and possibly more useful role.
  3. Is IL3RA a direct KMT2A-fusion target gene? CSPG4 and PROM1 both are; nothing in this beat places the fusion at the IL3RA locus, and the contrast is informative either way.
  4. What is the absolute CD123 site density on infant KMT2A-rearranged blasts? Every number in this entry is a quartile or an MFI, and no CAR or ADC design can be built on either.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In 1,040 diagnostic bone marrow specimens from children with newly diagnosed AML on Children's Oncology Group trial AAML1031 (NCT01371981), CD123 protein was prospectively measured by multidimensional flow cytometry at a central clinical laboratory and patients were divided into expression quartiles of 260 each. The highest quartile had a higher prevalence of high-risk KMT2A rearrangements and FLT3-ITD mutations (P < .001 for both) and a lower prevalence of low-risk t(8;21), inv(16) and CEBPA mutations (P < .001 for all). Patients in quartiles 1-3 had similar relapse risk, event-free survival and overall survival, whereas quartile 4 patients had significantly higher relapse risk (53% versus 39%), lower event-free survival (49% versus 69%) and lower overall survival (32% versus 50%), all P < .001. CD123 maintained independent significance for outcomes when all known contemporary high-risk cytogenetic and molecular markers were incorporated into multivariable Cox regression.

    PMID 34855461DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **Quartile cut-points are defined within this cohort**, so 'top quartile' is not a transferable threshold and no absolute antigen density is given in the abstract. This is a prospectively collected but retrospectively analysed correlative study on a single trial; the multivariable model's covariate list is not enumerated in the abstract. **Paediatric AML, not infant B-ALL**, and the abstract does not report an infant subgroup. Association with outcome is not evidence that CD123 causes the outcome or that removing CD123-positive cells improves it.

    Evidence that argues against this (1)
    • In B-lineage disease the direction reverses: KMT2A-rearranged B-ALL showed significantly LOWER CD123 mean fluorescence intensity than KMT2A-germline B-ALL in 1,382 cases. A CD123 enrichment argument built on AML cannot be carried into the B-lineage form of the same rearrangement. Bisegna ML et al. 2024
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Using Jmjd2a/Jmjd2b/Jmjd2c conditional triple-knockout mice, Jmjd2/Kdm4 H3K9- and H3K36-specific demethylase activity was shown to be required for MLL-AF9-translocated AML in vivo and in vitro. Expression of the interleukin-3 receptor alpha subunit (Il3ra, also known as Cd123) is dependent on Jmjd2/Kdm4 through removal of H3K9me3 from the Il3ra promoter, and ectopic expression of Il3ra in Jmjd2/Kdm4 knockout cells alleviates the requirement for Jmjd2/Kdm4 in AML cell survival, establishing Il3ra as a critical downstream target.

    PMID 27257215DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **Mouse MLL-AF9 AML**, not human and not the index B-lineage fusion, so the H3K9me3-at-Il3ra mechanism is established in the myeloid form of a different KMT2A partner. The rescue experiment is the strong part - it establishes Il3ra as functionally downstream rather than merely co-regulated - and it is a single genetic rescue in one system. The DOI in this block was initially written from memory and was WRONG (10.1101/gad.278762.116); it was corrected to 10.1101/gad.280495.116 after re-fetching the record in the same session.

    Evidence that argues against this (1)
    • An epigenetic requirement for Il3ra expression in AML cells does not make CD123 a KMT2A-fusion target gene. Unlike CSPG4 and PROM1, no evidence in this beat places the KMT2A fusion at the IL3RA locus, and the KDM4 dependency is a general AML finding. Godfrey L et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Among 1,382 newly diagnosed B-ALL cases, KMT2A-rearranged B-ALLs showed significantly lower mean fluorescence intensity of CD34, CD22, TdT and CD123 and significantly higher MFI of CD45, CD38 and CD58 than KMT2A-germline cases.

    PMID 38468826DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. Single-centre, mixed adult and paediatric referral cohort. **MFI is instrument- and reagent-dependent and is not antigen density**; 'lower CD123 MFI' cannot be converted into a positivity rate or a target-density threshold for an antibody-drug conjugate or CAR. The comparator is KMT2A-germline B-ALL of all subtypes, which includes the CD123-high TCF3-PBX1 cases [E4], so part of the difference is the comparator's composition.

    Evidence that argues against this (1)
    • Because the comparator group contains TCF3-PBX1 cases, in which CD123 is one of the four markers used to predict t(1;19), the 'KMT2A-r is CD123-low' finding is partly a statement about the comparator. The honest claim is that CD123 does not discriminate KMT2A-r upward in B-ALL, not that KMT2A-r cases are CD123-negative. Tsagarakis NJ et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 377 patients with B-ALL, flow cytometric scoring systems built by ROC analysis used CD34, cytoplasmic mu, CD123 and CD66c to predict t(1;19)(q23;p13.3), and NG2, CD10, CD15, CD34 and CD20 to predict t(11q23). CD123 appears in the t(1;19) score and not in the t(11q23) score. The t(11q23) score achieved sensitivity 84.21% and specificity 99.04%.

    PMID 30730614DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **Retrospective, single-centre, n = 377, and the scores were derived and evaluated in the same cohort** - the authors themselves call for prospective evaluation, so 84.21%/99.04% are optimistic. That CD123 is absent from the t(11q23) score is an outcome of the ROC selection and is weaker evidence than a direct measurement that CD123 is uninformative for KMT2A-r.

    Evidence that argues against this (1)
    • Marker selection by ROC within one cohort can drop a genuinely informative marker because a correlated marker in the same panel absorbed the signal first. CD123's absence from the t(11q23) score is consistent with it being uninformative and also with it being redundant given NG2. Tsagarakis NJ et al. 2019
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IL3RA is UniProt P26951 (IL3RA_HUMAN), Interleukin-3 receptor subunit alpha, 378 amino acids, a single reviewed human record with primary gene name IL3RA and gene synonym IL3R.

    What this evidence cannot show

    Fetched 2026-09-20 with `gene_exact:IL3RA AND organism_id:9606 AND reviewed:true`, requesting `gene_primary` and `length` per rules 62 and 65; primary name read back as IL3RA; exactly one reviewed row. **The gene symbol and the CD number do not match** (IL3RA / CD123), so a literature search on one form misses the other - PubMed's own querytranslation maps CD123 and IL3RA to the same supplementary concept, which is why the search in this beat was run on the CD form. IL3RA is in the pseudoautosomal region and has a paralogue relationship with CSF2RA; the length 378 is asserted here so `verify-accessions.py` can catch a swap.

    Evidence that argues against this (1)
    • CD123 functions only as the alpha chain of a heterodimer with the common beta chain CSF2RB, so an IL3RA-only account of the receptor is incomplete, and a therapeutic that engages the alpha chain does not necessarily engage the signalling competent complex. Nothing in the accession bears on that. UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
Connected to