Oncofusion

start / graph / gene-b2m

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

B2M (beta-2 microglobulin) - the invariant light chain every MHC class I molecule needs, the classic class I escape route, and a gene this disease's literature has never once named

also called beta-2 microglobulin · beta2-microglobulin · B2MG · b2m

Every class I identity card on a cell's surface is built from two parts, and B2M is the small part that is the same on all of them - lose it and the whole display disappears at once. Nobody has ever reported measuring it in KMT2A-rearranged leukaemia, even though it sits inside a crystal structure this graph already describes.

The same thing, for a scientist

B2M (UniProt **P61769**, B2MG_HUMAN, **119 aa**, single reviewed human record, primary gene name B2M, no listed gene synonyms) is the non-polymorphic light chain that heterodimerises with every classical and most non-classical MHC class I heavy chains [E1]. Because one gene serves all three classical loci, **a single biallelic B2M inactivation removes surface HLA-A, HLA-B and HLA-C together** - which is why B2M loss is the canonical route by which a tumour becomes invisible to CD8 T cells while remaining perfectly capable of transcribing heavy chains. It is written here for three reasons. First, it is **structurally present in this graph already**: the seven HLA-B*07:02 deposits carrying the KMT2A 747-755 phosphopeptide are heavy-chain/B2M/peptide ternary complexes, and `interface-hla-b0702-kmt2a-phosphopeptide-747-755` records P61769 in its own payload [E2]. Second, it is the **measurement nobody makes**: no source in the antigen-presentation beat reports B2M under a menin inhibitor, so every claim that menin inhibition raises surface MHC class I is a claim about heavy chain with the light chain assumed [E3]. Third, it is a **bounded counted absence in this disease**: `(KMT2A OR MLL) AND leukemia AND B2M[tiab]` returns **0** while `B2M[tiab]` returns 2,102 and `B2M[tiab] AND leukemia` returns 120 [E4].

Go deeper
Why one small gene decides whether class I exists at all

The functional unit on the surface is a trimer: a polymorphic heavy chain, the invariant 119-residue B2M light chain, and an 8-11-residue peptide. Removing any one of the three removes the display. That gives B2M a property no heavy chain has - it is a single-gene, lineage-independent, allele-independent off switch for the entire class I system, and its loss is invisible to any assay that reads heavy-chain transcript.

The practical consequence for this beat: a flow panel using W6/32 or a Bw6 antibody reads the assembled complex and is therefore already a B2M-dependent readout, while an HLA-A qPCR is not. PMID 39704147 reports both, and the two can disagree in a B2M-deficient cell without anything looking wrong 3.

The absence, bounded

Measured 2026-09-20 with PubMed E-utilities, querytranslation read back on every query:

queryrecordsnote
B2M[tiab]2,102rule-67 control: the instrument returns ones
B2M[tiab] AND leukemia120the leukaemia literature on B2M exists
"beta-2 microglobulin"[tiab] AND leukemia AND (CAR OR "chimeric antigen receptor")5the CAR-escape literature exists
(KMT2A OR MLL) AND leukemia AND B2M[tiab]0
(KMT2A OR MLL) AND leukemia AND "beta-2 microglobulin"[tiab]0singular and spelled-out form, per rule 72

So the absence is real at the level of title and abstract and is not an artefact of a non-composing query. It is still bounded by vocabulary: a paper reporting an MHC class I loss-of-heterozygosity or antigen-presentation-machinery defect in KMT2A-rearranged disease without writing 'B2M' would not be retrieved, and the aliases searched were B2M and beta-2 microglobulin only - not the Greek-letter spellings, which PubMed does not index distinctly 4.

The two identifier traps on this three-letter token

B2M is one of the most-used qPCR reference genes in haematology. A token census that counts B2M across a corpus counts loading controls, exactly as rule 71's corollary found for ACTB (46 entries, a loading control) and PTH (37 entries, a serum analyte). A B2M hit in a methods section is not a B2M finding, and this entry should not be taken as licence to auto-queue every B2M mention.

And B2M is a serum analyte with its own clinical literature. Serum beta-2 microglobulin is a prognostic marker in myeloma and lymphoma, unrelated to tumour-intrinsic antigen presentation. Of the 120 B2M[tiab] AND leukemia records, a sampled title check found the majority in that register rather than in the antigen-presentation one. Record which B2M a source means: the serum analyte, the reference gene, or the class I light chain 4.

Localization, cell type, and why subnuclear is NOT APPLICABLE (Q87, 2026-09-21)

Subcellular. Secreted and cell surface (UniProt P61769, both terms ECO:0000269); HPA v25.1 IF, reliability Enhanced, Golgi apparatus and plasma membrane as main locations, cytosol additional.

Subnuclear: NOT APPLICABLE. B2M has no nuclear pool. Recorded as a category exclusion, not a bounded absence, so it does not inflate the count of proteins whose subnuclear address nobody has measured. This follows the treatment given to HLA-A and NRAS in the previous localisation tranche.

Cell type. Every nucleated haematopoietic cell, including leukaemic blasts; HPA records low tissue specificity. The informative variable is surface density, not the identity of the cell.

Shown for fusion. No - and in the strong sense that nobody has looked in any fusion.

Facts and figures
gene symbolB2M
gene primaryB2M
uniprotP61769
length AA119
uniprot entry nameB2MG_HUMAN
hgnc resolutionsingle reviewed human record returned by gene_exact:B2M AND organism_id:9606 AND reviewed:true; primary gene name read back 2026-09-20
roleinvariant light chain of every classical MHC class I heterodimer; one gene serving HLA-A, HLA-B and HLA-C
kmt2a linkstructural only - P61769 is a chain in the seven HLA-B*07:02 deposits carrying the KMT2A 747-755 phosphopeptide (see interface-hla-b0702-kmt2a-phosphopeptide-747-755)
token register warningB2M is a very common qPCR reference gene and a serum prognostic analyte; a token census counting B2M counts all three registers
subnuclear localizationNOT APPLICABLE - B2M is a secreted, ER/Golgi-routed, cell-surface protein with no nuclear pool. Recorded as a category exclusion rather than as a bounded absence, so that it does not inflate the count of proteins whose subnuclear compartment is genuinely unmeasured.
cell stateNo B2M-defined cell STATE is recorded here. The state-like reading in this disease - antigen-presentation-low immune escape - is a property of the class I pathway as a whole rather than of B2M's own compartment, and is written in the immune beat's pathway entries, not here.
shown for fusionNO for localisation, in the strong sense of NOBODY HAS LOOKED IN ANY FUSION. B2M[tiab] ANDed with the eight-term KMT2A/MLL/MLL-AF4/MLL-AF9/MLL-ENL/MLL-AF10/"t(4;11)"/"t(9;11)" clause and a ten-term localisation clause returns 0 records (PubMed esearch 2026-09-21). The question is also close to meaningless: B2M's compartment is a property of the secretory pathway, not of a fusion, and what a fusion could plausibly change is surface DENSITY, which is a different measurement.
gene synonyms
[]
counted absence
{
  "query": "(KMT2A OR MLL) AND leukemia AND B2M[tiab]",
  "count": 0,
  "date": "2026-09-20",
  "controls": {
    "B2M[tiab]": 2102,
    "B2M[tiab] AND leukemia": 120
  },
  "name_list": [
    "B2M",
    "beta-2 microglobulin"
  ],
  "tool": "PubMed E-utilities esearch"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.0523,
  "chronos_mean_other_leukaemia": 0.0325,
  "difference": 0.0198,
  "pan_essentiality_all_lines": 0.0326,
  "selectivity_rank": 11689,
  "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": "SECRETED and CELL SURFACE. UniProt P61769 carries exactly two location terms, Secreted (ECO:0000269 PubMed:1336137, PubMed:7554280) and Cell surface (ECO:0000269 PubMed:25356553), with the note 'Detected in serum and urine'. HPA v25.1 immunofluorescence, reliability Enhanced, gives main locations Golgi apparatus and Plasma membrane with Cytosol additional - the secretory route plus the surface destination, which is what a class I light chain should look like.",
  "subnuclear": "NOT APPLICABLE. B2M is not a nuclear protein at any point in its life cycle: it is translated with a signal peptide, folds in the ER with the class I heavy chain, traverses the Golgi and is displayed on the plasma membrane or shed into serum. There is no nuclear pool to have a subnuclear address, so this is a category exclusion and not an unmeasured gap - it belongs OUT of the denominator of proteins whose subnuclear compartment nobody has determined, in the same way as HLA-A and NRAS in the previous tranche.",
  "how_shown": "biochemical purification from serum and urine with isoelectric focusing and HPLC (PMID 1336137); cell-surface detection in macrophages (PMID 25356553); HPA antibody immunofluorescence",
  "not_applicable_reason": "extracellular / membrane-associated protein, no nuclear pool",
  "node": "localization-plasma-membrane-leukaemic-surface-antigens"
}
cell type
{
  "haematopoietic_answer": "ALL NUCLEATED HAEMATOPOIETIC CELLS, as the invariant light chain of every class I molecule, and therefore on leukaemic blasts. HPA v25.1 records low RNA tissue specificity, so the cell-type answer for B2M is 'everywhere', and the informative question is not WHICH cell but HOW MUCH assembled surface class I a given blast displays.",
  "why_the_usual_answer_is_a_trap": "data/symbol-collisions.json records B2M as one_symbol_three_registers - the MHC class I light chain, a haematology qPCR reference gene, and a serum prognostic marker in myeloma and lymphoma. A cell-type statement drawn from a token count would be counting the reference-gene register, which is the commonest one in a leukaemia corpus.",
  "expression_atlas": "HPA v25.1: RNA tissue specificity Low; IF reliability Enhanced."
}
localisation absence rebounded cycle 58
{
  "why": "The claim sat before a correction marker in scripts/check-stale-corrections.py's queue, and it was one of only three such assertions in an evidence item rather than in prose.",
  "alias_split_measured": {
    "B2M[tiab]": 2103,
    "\"beta-2 microglobulin\"[tiab]": 9485,
    "\"beta2-microglobulin\"[tiab]": 5485,
    "note": "the approved symbol is the MINORITY token, the same shape as the HCFC1 row added to the registry this cycle"
  },
  "intersections": {
    "B2M alone x (MLL OR KMT2A)": 2,
    "alias union x (MLL OR KMT2A)": 5,
    "alias union x localisation x leukemia": 1
  },
  "all_6_read_by_title": [
    "41443280 NK/T-cell lymphoma histology",
    "16843102 Korean myeloma cytogenetics",
    "15359102 dialysate solute assays",
    "12232763 retinoblastoma N-MYC",
    "8194803 polycystic kidney disease",
    "16644301 plasma cell dyscrasia FGF23"
  ],
  "verdict": "SURVIVES. No study reports B2M subcellular localisation in any KMT2A-fusion context."
}
How this could bite us back
  • Any therapeutic strategy that works by raising class I antigen presentation - a menin inhibitor, an EZH2 inhibitor, interferon-gamma - is defeated completely and silently by biallelic B2M loss, and the standard readouts for such strategies (heavy-chain transcript, HOXA/MEIS1 signature) cannot see it.
  • Selecting for B2M loss is a predictable consequence of any sustained MHC-restricted pressure, which means a successful TCR-based or transplant-based approach creates the resistance mechanism that defeats the next one.
  • Because B2M is invariant and a single gene, the escape is cheap: one lesion removes all three classical loci at once, with no fitness cost that has been demonstrated in leukaemia.
  • Conversely, B2M loss removes the inhibitory KIR ligands as well, which should make the cell MORE vulnerable to NK cells - so the bite-back has its own counter-move, and an NK-directed approach is the rational partner to any class I-raising strategy.
Still open
  1. What is the B2M status of the standard KMT2A-rearranged lines - SEM, RS4;11, MV4-11, KOPN-8, THP-1? One sequencing lookup and one flow panel would convert a complete blank into a fact, and would tell any future menin-inhibitor immunology experiment whether its readout can move at all.
  2. Does B2M change under a menin inhibitor? If the class I effect is bivalency-driven and B2M is not a bivalent gene, the light chain would be the rate-limiting component at high heavy-chain induction, which is a testable ceiling on the whole mechanism.
  3. Is B2M loss a route of relapse after allogeneic transplant in KMT2A-rearranged disease specifically, as it is in AML generally? The distinction matters because it decides whether TCR-based or MHC-independent approaches are the right second line in this subtype.
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 beta-2 microglobulin is P61769 (B2MG_HUMAN), primary gene name B2M, no listed gene synonyms, 119 aa, protein name 'Beta-2-microglobulin'; a gene_exact query returns exactly one reviewed human record.

    UniProtKB reviewed human entry P61769 (B2MG_HUMAN) · UniProt Consortium · UniProtKB 2026
    What this evidence cannot show

    Database fact. Fetched 2026-09-20 via rest.uniprot.org with fields accession, id, gene_primary, gene_synonym, protein_name, length, and the primary gene name was read back per rules 62 and 65. A single reviewed record means rule 62 does not fire here; the length of 119 aa is short enough that a length check discriminates B2M from essentially any other protein in this beat.

    Evidence that argues against this (1)
    • A single reviewed record is not a guarantee of unambiguous retrieval from the literature: the string B2M in an abstract is more often a qPCR reference gene or a serum analyte than the class I light chain, so the accession being clean says nothing about whether a paper naming 'B2M' is about this protein. Q54-immune et al. 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Beta-2 microglobulin (P61769) is a component of the HLA-B*07:02 ternary complexes that present the KMT2A 747-755 phosphopeptide EPR(pS)PSHSM, an HLA-B*07:02-restricted AML phosphoneoantigen recognised by the T-cell receptor TCR27, with seven deposited crystal structures of the peptide, phosphopeptide and synthetic analogues.

    Molecular mechanism of phosphopeptide neoantigen immunogenicity. · Patskovsky Y, Natarajan A, Patskovska L, Nyovanie S, Joshi B, Morin B, et al. · Nat Commun 2023
    PMID 37353482DOI
    What this evidence cannot show

    The paper is about phosphopeptide neoantigen immunogenicity in general; the KMT2A epitope is one of its subjects and this graph's `interface-hla-b0702-kmt2a-phosphopeptide-747-755` is the entry that analyses it. B2M's presence in the structures is a structural fact about how class I is assembled, not a finding about B2M in leukaemia. The epitope is HLA-B*07:02-restricted, so the whole route is available only to patients carrying that allele. Read: abstract and record fetched 2026-09-20; the structural detail is taken from this graph's existing interface entry, whose identifiers were re-fetched here.

    Evidence that argues against this (1)
    • The existence of a KMT2A-derived class I epitope is an argument for measuring the class I machinery in this disease, not evidence that it is intact. A B2M-null blast would present this epitope at zero regardless of how much KMT2A phosphopeptide it generates, and no study has checked B2M status in the cells the epitope was demonstrated on. Patskovsky Y et al. 2023
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    No source examined in this beat reports a beta-2 microglobulin measurement under a menin inhibitor: PMID 39704147 measures HLA-A transcript and pan-MHC class I surface protein, PMID 36635503 measures surface MHC-I with W6/32 and HLA-B-specific antibodies and profiles the HLA-B locus, and neither full text contains the token B2M.

    What this evidence cannot show

    This is a negative over two documents, not over the literature. It was measured by regex over the tag-stripped full text of each paper, fetched 2026-09-20 - Europe PMC fullTextXML for PMC7614190 and the PMC HTML page for PMC12130779 after fullTextXML returned HTTP 503. A supplementary table could contain a B2M row that the body text does not mention; supplementary files were not parsed. Tags were stripped before entity unescaping, per rule 61.

    Evidence that argues against this (1)
    • The absence of an explicit B2M measurement does not mean the surface readouts are uninformative about B2M. W6/32 and the Bw6 clone REA143 both detect assembled heavy-chain/B2M complexes, so a rise in W6/32 signal is evidence that B2M is present and not limiting in those cells. The gap is that the light chain is never reported as a variable, so nothing is known about whether it changes, or whether it is the limiting factor in the samples where class I did not rise. Sparbier CE et al. 2023
  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 B2M[tiab]` returns 0 records and `(KMT2A OR MLL) AND leukemia AND "beta-2 microglobulin"[tiab]` returns 0 records, against positive controls of 2,102 for `B2M[tiab]` alone, 120 for `B2M[tiab] AND leukemia` and 5 for `"beta-2 microglobulin"[tiab] AND leukemia AND (CAR OR "chimeric antigen receptor")`; querytranslation was inspected on every query and all concept blocks expanded normally.

    What this evidence cannot show

    Bounded absence. Name list searched: `B2M`, `beta-2 microglobulin`. NOT searched: `beta2 microglobulin` unhyphenated, `b2-microglobulin`, `HLA-I light chain`, and the retired-symbol space is empty because B2M has no listed synonym. The KMT2A concept block is `KMT2A OR MLL`, which does not include the retired symbols ALL-1, HRX or HTRX, so the count is a floor. A paper reporting class I antigen-presentation-machinery defects in this disease under a phrase like 'antigen processing machinery' without naming B2M is invisible to this query; note that `(KMT2A OR MLL) AND leukemia AND "antigen processing"[tiab]` also returns 0.

    Evidence that argues against this (1)
    • The zero is a measurement about the title and abstract layer only, and this beat has a worked example of that layer failing: PMID 39704147 measures HLA-A in its figures and says only 'MHC class I' in its abstract. The same could be true of B2M in any AML paper with a KMT2A-rearranged subgroup. Treat the zero as 'never a headline finding in this disease', not as 'never measured'. Hogeling SM et al. 2025
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P61769 places B2M in exactly two compartments - Secreted and Cell surface - and HPA v25.1 immunofluorescence (reliability Enhanced) gives Golgi apparatus and Plasma membrane as main locations with Cytosol additional. Neither source records any nuclear pool, so B2M's subnuclear localisation is NOT APPLICABLE rather than unmeasured.

    What this evidence cannot show

    Two database records, not a dedicated imaging study, and the two do not fully agree: UniProt does not list the Golgi and HPA does not list the secreted form, because each instrument sees a different part of one itinerary. An antibody-based IF call is also blind to whether the protein detected is assembled with a heavy chain - the W6/32 problem recorded as rule 74 - so 'B2M at the plasma membrane' by IF is not the same quantity as assembled class I.

    Evidence that argues against this (1)
    • A NOT APPLICABLE verdict is a claim that can be wrong, and the honest bound on it here is the protein's biosynthetic route plus two databases that record no nuclear term - not an exhaustive literature sweep, which was not run. And the very paper UniProt cites for the surface call argues against treating any of these compartments as fixed: M. tuberculosis ESAT-6 binds beta-2-microglobulin and DECREASES the amount found on the cell surface, so the surface register is itself condition-dependent. If a nuclear B2M pool were ever reported, this call would have to be withdrawn. Sreejit G et al. 2014
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    No study has reported B2M's subcellular localisation in any KMT2A-fusion context. Measured 2026-09-21: B2M[tiab] AND (KMT2A[tiab] OR MLL[tiab] OR "MLL-AF4"[tiab] OR "MLL-AF9"[tiab] OR "MLL-ENL"[tiab] OR "MLL-AF10"[tiab] OR "t(4;11)"[tiab] OR "t(9;11)"[tiab]) AND (localization[tiab] OR localisation[tiab] OR subnuclear[tiab] OR nuclear[tiab] OR nucleolar[tiab] OR speckle[tiab] OR speckles[tiab] OR "PML bodies"[tiab] OR chromatin-bound[tiab] OR cytoplasmic[tiab]) returns 0 records. **RE-BOUNDED CYCLE 58, AND IT SURVIVES.** The original search used the approved symbol, which is the minority token here: `B2M[tiab]` returns 2,103 records while `"beta-2 microglobulin"[tiab]` returns 9,485. Re-run on the alias union, `(B2M OR "beta-2 microglobulin" OR "beta2-microglobulin")[tiab] AND (MLL OR KMT2A)[tiab]` returns **5** where B2M alone returns 2, and the same union against localisation-and-leukaemia returns **1**. All **6** records were fetched and read by title, and every one is off topic - NK/T-cell lymphoma histology (41443280), Korean myeloma cytogenetics (16843102), dialysate solute assays (15359102), retinoblastoma N-MYC (12232763), polycystic kidney disease (8194803) and plasma-cell-dyscrasia FGF23 (16644301). The MLL hits are false positives from myeloma and dialysis contexts, not fusion biology. So the absence holds on a union that is four to five times the original operand, with every candidate read rather than counted.

    What this evidence cannot show

    A zero bounded by titles and abstracts only; supporting figures were not searched, and a surface-class-I flow panel in a KMT2A-rearranged cohort would not name B2M in its abstract (rule 80). Rule-67 positive control for this query SHAPE: the identical construction returned non-zero for seven of the sixteen genes run in this tranche on the same day - H2AZ1 1, EED 2, CHAF1B 1, CEBPA 3, TIFAB 1, SAMHD1 1, PROM1 1 - so the instrument can return a one. The querytranslation was read back on every run and confirms that the quoted cytogenetic terms expand to "t 4 11"[Title/Abstract] and "t 9 11"[Title/Abstract], not to a UID filter (rule 95). Bounded on titles and abstracts only, so a localisation panel inside a supplementary figure would not be seen, and preprints are not covered. Note also that PubMed normalises the hyphen: "beta-2 microglobulin" and "beta 2 microglobulin" return the identical count of 9,485, so they are one query and not two. A union count taken at retmax=300 over sets of 2,103 and 9,485 is TRUNCATED and no union size is quoted here for that reason - only the intersections, which are small enough to enumerate.

    Evidence that argues against this (1)
    • The absence is real and almost uninteresting, and stating it risks implying the underlying biology is unstudied. It is not: surface class I in leukaemia is measured constantly, by flow cytometry with pan-HLA reagents rather than by a localisation experiment on B2M. The commonest such reagent, clone W6/32, detects the ASSEMBLED heavy-chain/beta-2-microglobulin heterodimer and is therefore silently a B2M-dependent readout - so a great deal of B2M data exists in this disease under a label that never mentions B2M. Absence of the WORD is not absence of the MEASUREMENT, and a bounded absence written on a token search cannot see the difference. Sreejit G et al. 2014
Connected to