Oncofusion

start / graph / gene-hla-a

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q54-immune

HLA-A - the class I heavy chain a menin inhibitor was actually measured on, and the human orthologue of the MHC class I molecule that protects MLL-AF9 leukaemia stem cells from NK cells

also called HLAA · HLA class I histocompatibility antigen A alpha chain · H2-K1 (murine orthologue)

HLA-A is one of the three genes that make the identity card leukaemia cells show to killer T cells. It is the one gene a menin inhibitor has actually been shown to switch up in patients' leukaemia cells - and, confusingly, the same molecule in mice is something leukaemia stem cells need in order to hide from a different kind of killer cell.

The same thing, for a scientist

HLA-A (UniProt **P04439**, HLAA_HUMAN, **365 aa**, primary gene name HLA-A, synonym HLAA) is one of the three classical MHC class I heavy chains [E1]. It earns an entry here because it is the **only class I gene in this graph's evidence base with a named, measured response to a clinical-stage menin inhibitor**: bleximenib raised H3K4me3 and H3K27ac at the HLA-A locus and raised HLA-A mRNA in primary AML, with surface MHC class I rising in **9 of 25** samples [E2]. And it is the gene at the centre of the beat's sharpest tension: **H2-K1, the murine orthologue of human HLA-A, scored as a top in vivo dependency of MLL-AF9 leukaemia-initiating cells**, and knocking it down depleted leukaemia in bone marrow and spleen and extended survival - an effect **abolished by NK-cell depletion** [E3]. The same paper reports HLA-A upregulated on human AML blasts relative to normal haematopoietic stem cells [E3]. So in the index fusion family, the best-evidenced statement about a class I heavy chain is not that it is silenced but that it is **elevated and load-bearing**, as an NK brake.

Go deeper
The orthologue claim, and how far it can be pushed

PMID 39844759 states plainly that H2-k1 is 'the murine ortholog of human HLA-A' and supports the parallel with a human observation - HLA-A upregulated on AML blasts versus normal HSC 3. That is an author's framing and it is the usual one, but mouse and human class I are not one-to-one: the mouse has H2-K, H2-D and H2-L, the human has HLA-A, -B and -C, and the correspondence is by function and position rather than by descent from a single ancestral locus pair. A Ly49-KIR comparison is even looser - the inhibitory receptor families are structurally unrelated and convergently evolved.

So the transferable claim is at the level of the class, not the gene: a classical MHC class I molecule on an MLL-fusion leukaemia stem cell restrains NK-mediated clearance in vivo. Attributing it specifically to HLA-A in a patient is an extrapolation and is recorded as one.

What the menin-inhibitor result on HLA-A does and does not license

PMID 39704147's HLA-A arm is a chromatin result with a transcript and a protein confirmation, in primary human material, with a clinical-stage compound. That is as good as this literature gets. Its limits are worth stating exactly:

  • 9 of 25. HLA-A rose in a minority; HLA-DR rose in the majority. Both rose together in 5 of 25 2.
  • Not in KMT2A-rearranged patients. The functional immune experiments were done in NPM1c/DNMT3A-mutant AML; all four allogeneic responders and the whole autologous cohort were NPM1c 2.
  • No correlation with killing. The authors state the strength of the T-cell response was not directly correlated with the level of MHC upregulation, and one sample (AML54) gained HLA-DR, lost nothing on HLA-A, and became no more killable 2.

That last point is the one an enthusiastic reading loses. Raising the display does not by itself buy T-cell control.

The symbol is nearly unreachable from this disease's literature

(KMT2A OR MLL) AND leukemia AND (HLA-A[tiab] OR HLA-B[tiab] OR HLA-C[tiab]) returns 1 record as of 2026-09-20, and it title-checks to an alternative-donor transplant outcome study in lymphoma (PMID 25402415) - a false positive from HLA matching vocabulary 4. The control HLA-A[tiab] AND leukemia returns 741, so the instrument works.

This is rule 71 in its purest form: the query is accurate and the conclusion 'nobody has measured a class I heavy chain in this disease' would be false, because PMID 39704147 measures HLA-A in its figures and writes only 'MHC class I' in its abstract. The vocabulary of this field is the phrase, not the symbol. Any counted absence about class I in KMT2A-rearranged leukaemia must be run on "MHC class I" and "HLA class I" as phrases, and even then it will miss figure-level measurements.

Facts and figures
gene symbolHLA-A
gene primaryHLA-A
uniprotP04439
length AA365
uniprot entry nameHLAA_HUMAN
murine orthologue as stated by sourceH2-K1 (PMID 39844759); orthology is by function and position, not one-to-one descent
length discrimination warningHLA-A 365 aa vs HLA-B 362 aa - three residues; the length check cannot separate them
subnuclear localizationNOT APPLICABLE - HLA-A is a secretory-pathway type I membrane protein with no nuclear pool. Recorded as not-applicable, not as an unmeasured absence.
shown for fusionPARTIAL, and only for ABUNDANCE. Surface MHC class I has been measured by flow in primary AML under the menin-KMT2A inhibitor bleximenib (PMID 39704147, 9 of 25 samples up), and murine H2-K1 protects MLL-AF9 leukaemia stem cells from NK cells (PMID 39844759, already in this entry). Neither is a LOCALISATION experiment: no study has asked where HLA-A heavy chain sits, or what fraction completes ER-to-surface transit, in a KMT2A-rearranged cell. The murine result is MLL-AF9 specifically, not 'a KMT2A fusion'. The bounded localisation query returned 0.
gene synonyms
[
  "HLAA"
]
measured under menin inhibitor
{
  "drug": "JNJ-75276617 (bleximenib)",
  "readouts": [
    "H3K4me3 ChIP-seq",
    "H3K27ac ChIP-seq",
    "mRNA qRT-PCR",
    "pan-MHC-I flow"
  ],
  "result": "up in 9 of 25 primary AML samples",
  "pmid": "39704147",
  "title": "Bleximenib, the novel menin-KMT2A inhibitor JNJ-75276617, impairs long-term proliferation and immune evasion in acute myeloid leukemia.",
  "authors": "Hogeling SM, Lê DM, La Rose N, et al.",
  "venue": "Haematologica",
  "year": 2025,
  "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."
}
counted absence
{
  "query": "(KMT2A OR MLL) AND leukemia AND (HLA-A[tiab] OR HLA-B[tiab] OR HLA-C[tiab])",
  "count": 1,
  "on_topic_after_title_check": 0,
  "control": {
    "HLA-A[tiab] AND leukemia": 741
  },
  "date": "2026-09-20",
  "verdict": "accurate as a measurement, false as a conclusion - see rule 71"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.2739,
  "chronos_mean_other_leukaemia": -0.1673,
  "difference": -0.1066,
  "pan_essentiality_all_lines": -0.1724,
  "selectivity_rank": 997,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
localization
{
  "subcellular": "PLASMA MEMBRANE, single-pass type I membrane protein (UniProt P04439, ECO:0000269 PubMed:21263072, 25880248, 8805302), with an endoplasmic-reticulum-membrane pool during assembly (ECO:0000305, PubMed:8805302). HPA immunofluorescence gives main location Plasma membrane with Golgi apparatus additional, reliability Supported - the Golgi call is the secretory route between the two UniProt compartments and the three sources are consistent.",
  "subnuclear": "NOT APPLICABLE. HLA-A is a type I membrane glycoprotein of the secretory pathway; it has no nuclear pool and a subnuclear record for it would be an error, not a gap. This entry is recorded as not-applicable rather than as an absence so that it is not counted as a missing measurement.",
  "how_shown": "peptide-loading and assembly biochemistry plus surface immunostaining (UniProt's coded citations); antibody immunofluorescence in HPA's panel",
  "the_readout_is_the_identifier": "Brief rule 74 governs any surface claim here. W6/32 detects the ASSEMBLED heavy-chain/beta-2-microglobulin heterodimer, so a W6/32 readout is silently B2M-dependent and reports ASSEMBLY AND TRAFFICKING, not heavy-chain abundance; REA143 is anti-Bw6, a serological specificity carried by a subset of HLA-B alleles with Bw4 additionally on some HLA-A alleles. Record the clone, and say whether a surface claim was assembly-dependent or transcript-level.",
  "trafficking_is_the_regulated_step": "For this gene the interesting localisation question is not WHERE the protein is at rest but HOW MUCH of it completes the ER-to-surface journey. Loss of B2M, of TAP1/TAP2 or of tapasin retains heavy chain in the ER, and every immune-evasion mechanism in this disease acts on that step rather than on the resting address.",
  "node": "localization-plasma-membrane-leukaemic-surface-antigens"
}
cell type
{
  "haematopoietic_answer": "ALL of them. HPA Blood Atlas: Low immune cell specificity, Detected in ALL immune cell types; UniProt tissue comment: Ubiquitous. HLA-A is one of the few genes in this tranche with no informative cell-type restriction, and saying so is the honest record.",
  "the_variation_that_matters_is_not_cell_type": "Between haematopoietic cell types HLA-A is flat. What varies is (a) the ALLELE, which is a germline property and decides which peptides and which KIR/serological epitopes are presented, and (b) the SURFACE amount in a given leukaemic state under interferon or under a menin inhibitor. A cell-type field is the wrong instrument for this gene and this entry says so explicitly.",
  "cross_reference": "gene-nlrc5 holds the transactivator arm and gene-b2m the light chain; read the three together rather than treating HLA-A as a standalone."
}
cell state
{
  "state": "INTERFERON-RESPONSIVE and MENIN-INHIBITOR-RESPONSIVE surface abundance, not a resting state. See cell-state-interferon-herv-mhc2 for the interferon arm.",
  "measured_under_drug": "Pan-MHC-I flow rose in 9 of 25 primary AML samples under bleximenib (JNJ-75276617), already recorded in data.measured_under_menin_inhibitor (PMID 39704147). That is an ABUNDANCE measurement at the plasma membrane, not a localisation experiment, and the clone used for the pan-MHC-I flow is not recorded in this entry - brief rule 74 says it should be."
}
localization bounded absence
{
  "date": "2026-09-21",
  "tool": "NCBI E-utilities esearch, db=pubmed",
  "shape": "gene clause AND localisation clause AND KMT2A/MLL-fusion clause, one [tiab] tag per term (brief rule 94), every cytogenetic term quoted (brief rule 95)",
  "querytranslation_fusion_clause": "\"KMT2A\"[Title/Abstract] OR \"MLL\"[Title/Abstract] OR \"MLL1\"[Title/Abstract] OR \"MLL-AF4\"[Title/Abstract] OR \"MLL-AF9\"[Title/Abstract] OR \"MLL-ENL\"[Title/Abstract] OR \"KMT2A-rearranged\"[Title/Abstract] OR \"MLL-rearranged\"[Title/Abstract] OR \"t 4 11\"[Title/Abstract] OR \"t 9 11\"[Title/Abstract] OR \"MLL fusion\"[Title/Abstract]",
  "count": 0,
  "controls": {
    "fusion_clause_alone": 8743,
    "gene_clause_alone": 37125
  },
  "reading": "Counted in TITLES AND ABSTRACTS only; supporting figures were not searched (brief rule 80)."
}
How this could bite us back
  • A drug that raises HLA-A raises the ligand for inhibitory KIR and Ly49 receptors. In the MLL-AF9 model that ligand is a leukaemia dependency, so the drug would be reinforcing an escape route from NK cells while opening one for T cells.
  • The benefit is allele-dependent in a way no trial stratifies for: a KMT2A epitope presented on HLA-B*07:02 is available to the minority of patients carrying that allele, and the KIR-ligand consequences of raising class I depend on the patient's own Bw4/Bw6/C1/C2 groups.
  • Because class I raises the visibility of every self-peptide and not only tumour peptides, the same increase in the post-transplant setting is an increase in the substrate for graft-versus-host disease.
  • If the effect is a minority phenomenon (9 of 25) and does not predict killing, then any trial powered on MHC upregulation as a pharmacodynamic endpoint is powered on the wrong quantity.
Still open
  1. Is the NK brake saturated at baseline? If a leukaemia stem cell already expresses enough class I to engage inhibitory receptors maximally, raising it further costs nothing on the NK side and the menin-inhibitor effect is pure gain. If it is not saturated, the drug is trading one effector arm for another. Nobody has measured the dose-response.
  2. Does HLA-A rise under a menin inhibitor in a KMT2A-rearranged blast specifically? The measurement exists for AML as a whole and for NPM1c disease functionally; the index genotype has never been reported separately.
  3. Is the H2-K1 dependency reproduced in a KMT2A::AFF1 lymphoid model? Every NK-versus-class-I result in this fusion family is myeloid.
  4. Every MHC class I readout in this graph measures ABUNDANCE. Nobody has asked what fraction of HLA-A heavy chain in a KMT2A-rearranged blast is stuck in the ER rather than at the surface - which is the only measurement that distinguishes a transcriptional defect (NLRC5/KMT2A axis) from an assembly defect (B2M, TAP, tapasin). A W6/32 versus anti-heavy-chain pair on the same cells would separate them in one experiment.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The reviewed human UniProt record for the MHC class I A heavy chain is P04439 (HLAA_HUMAN), primary gene name HLA-A, synonym HLAA, 365 aa, protein name 'HLA class I histocompatibility antigen, A alpha chain'.

    UniProtKB reviewed human entry P04439 (HLAA_HUMAN) · UniProt Consortium · UniProtKB 2026
    What this evidence cannot show

    Database fact, fetched 2026-09-20 with gene_primary and length read back. The reviewed record is a reference sequence for an extraordinarily polymorphic locus - thousands of HLA-A alleles exist - so the accession identifies the gene product class, not a patient's allele. Any claim about peptide binding, epitope presentation or KIR ligand group is allele-specific and cannot be made from this accession. Length of 365 aa discriminates HLA-A from HLA-B (362 aa) by only three residues, so the length check in verify-accessions.py provides almost no protection against an A/B swap in this family; assert the primary gene name.

    Evidence that argues against this (1)
    • HLA-A at 365 aa and HLA-B at 362 aa are three residues apart, tighter than the CLL1/COLEC10 pair (12 aa) already logged in data/symbol-collisions.json and far tighter than rule 66's SMC1A/SMC3 example at 16 aa. The length defence is effectively absent for the class I heavy chains and nothing in the accession reads as wrong if the two are swapped. UniProt Consortium 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The menin-KMT2A inhibitor JNJ-75276617 (bleximenib) increased H3K4me3 and H3K27ac at the HLA-A locus in primary AML cells and OCI-AML3, raised HLA-A mRNA, and raised surface MHC class I protein in 9 of 25 primary AML samples, with both MHC class I and class II consistently upregulated in 5 of 25; the strength of subsequent T-cell cytotoxicity was not directly correlated with the level of MHC upregulation.

    Bleximenib, the novel menin-KMT2A inhibitor JNJ-75276617, impairs long-term proliferation and immune evasion in acute myeloid leukemia. · Hogeling SM, Le DM, La Rose N, Kwon MC, Wierenga ATJ, Van den Heuvel FAJ, et al., Schuringa JJ · Haematologica 2025
    PMID 39704147DOI
    What this evidence cannot show

    Adult AML only; the four samples showing enhanced allogeneic T-cell killing and the entire autologous cohort carried NPM1c mutations, so this is not a KMT2A-rearranged result. HLA-A is reported as transcript and as chromatin marks; the protein readout is pan-MHC class I flow and therefore does not resolve which heavy chain rose. HLA-B, HLA-C and B2M were not measured. Read: full text via the PMC HTML route (PMC12130779), 2026-09-20.

    Evidence that argues against this (1)
    • In sample AML54 the drug produced no HLA-A upregulation but strong HLA-DR upregulation and T-cell cytotoxicity was not enhanced - the AML cells expanded slightly in the presence of allogeneic T cells. A per-patient counterexample inside the paper's own dataset. Hogeling SM et al. 2025
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    H2-k1, described by the authors as the murine orthologue of human HLA-A, scored among the top in vivo dependencies in a CRISPR dropout screen of 961 cell-surface genes in MLL-AF9 leukaemia-initiating cells; H2-K1 was markedly elevated on c-Kit+ MLL-AF9 leukaemia cells compared with healthy bone marrow counterparts and HLA-A was upregulated on AML patient blasts versus normal haematopoietic stem cells; H2-k1 disruption did not affect ex vivo growth but strongly depleted leukaemia from bone marrow and spleen in vivo and extended survival; NK-cell depletion with anti-NK1.1 accelerated leukaemia and rescued the antileukaemic effect of H2-k1 knockdown, whereas macrophage depletion with clodronate liposomes did not.

    H2-K1 protects murine MLL-AF9 leukemia stem cells from natural killer cell-mediated immune surveillance. · Ghosh S, Rodriguez-Zabala M, Dushime GT, Reinbach K, Ramakrishnan R, Sitnicka E, Jaras M · Haematologica 2025
    PMID 39844759DOI
    What this evidence cannot show

    Murine MLL-AF9 model throughout; the human observation is an expression comparison, not a functional test. Published as a Haematologica letter with no abstract in the PubMed record, so the full text is the only source of the claims - retrieved via the PMC HTML page (PMC12130766) after the abstract field came back empty, which is the route-4 case the brief documents. MLL-AF9 is a myeloid model and this graph's index disease is infant KMT2A::AFF1 B-ALL; class I regulation and NK ligand repertoires differ between lineages. The screen was a dropout screen for surface genes, so the finding is a dependency, not a demonstration that raising class I further would worsen NK control.

    Evidence that argues against this (1)
    • A dependency is not the same as a dose-response. The experiment removes H2-K1; it does not raise it. Whether the NK brake is already saturated at baseline class I levels - in which case a menin-inhibitor-driven increase would add nothing to NK inhibition while still adding T-cell visibility - is not addressed anywhere in this literature and is the single most decision-relevant open question in this beat. Ghosh S et al. 2025
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    As of 2026-09-20, `(KMT2A OR MLL) AND leukemia AND (HLA-A[tiab] OR HLA-B[tiab] OR HLA-C[tiab])` returns exactly 1 PubMed record, which title-checks to an unrelated alternative-donor transplantation outcome study, against a control of 741 for `HLA-A[tiab] AND leukemia`.

    What this evidence cannot show

    Bounded absence and a demonstration that the bound is in the vocabulary rather than the query: the one hit is a false positive from HLA-matching language, and the true measurement (PMID 39704147, HLA-A) is unreachable because its abstract says only 'MHC class I'. Aliases searched: HLA-A, HLA-B, HLA-C. Not searched: HLA-I, MHC-I as symbols, allele-level strings such as HLA-A*02:01, or the murine orthologues. Rule 67 control run through the identical shape.

    Evidence that argues against this (1)
    • This absence is refuted, as a statement about the world, by the very paper the entry cites for HLA-A. It is retained because the measured number is still useful: it tells a future agent that a symbol-keyed sweep of class I in this disease will return nothing, and therefore that the sweep must be run on phrases. Hogeling SM et al. 2025
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P04439 places the HLA-A heavy chain at the cell membrane as a single-pass type I membrane protein, with an endoplasmic-reticulum-membrane pool, and HPA immunofluorescence gives main location Plasma membrane with Golgi apparatus additional (reliability Supported).

    What this evidence cannot show

    Both records describe a resting steady state in cells that are not leukaemic. The accession names the GENE while the serological and KIR-ligand properties that decide what an antibody in any cited figure can see are ALLELE properties - collision registry, HLA-A row. HLA-A (365 aa), HLA-B (362) and HLA-C (366) lie within four residues, so no length check can catch a swap between them.

    Evidence that argues against this (1)
    • A 'plasma membrane' record describes the successful OUTCOME of a pathway and silently excludes the population this disease is about. Point mutations in the alpha-2 domain that break the heavy chain's functional interaction with TAP change peptide loading and therefore surface delivery, so where HLA-A ends up is a property of the assembly line and not of the protein. Every immune-evasion mechanism relevant here acts on that line. Recording the resting address without the retained fraction makes the interesting variable invisible. Curr Biol 1996
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The Human Protein Atlas Blood Atlas records HLA-A as having low immune cell specificity and being detected in all immune cell types, consistent with UniProt's ubiquitous tissue call.

    What this evidence cannot show

    RNA, not surface protein, and normal blood rather than leukaemia. Because of the MHC alternative-haplotype scaffolds HLA-A carries several ENSG identifiers (collision registry), so any genome-wide join keyed on this symbol may silently drop it; the HPA row used here is the primary-assembly ENSG00000206503.

    Evidence that argues against this (1)
    • This entry's own cell-type answer is deliberately negative, and that negative can be over-read. 'Detected in all immune cells' is an RNA statement, and the layer that actually varies is post-transcriptional: glycans on tapasin and on the heavy chain have distinct, separable functions in class I assembly, so two cells with identical HLA-A transcript can present very different amounts of assembled heterodimer. Concluding 'cell type is the wrong instrument for this gene' is right for RNA and wrong if anyone ever measures assembled complex per cell type, which nobody has here. J Immunol 2011
Connected to