Oncofusion

start / graph / ana-lamp5-the-published-mll-r-immunotherapy-target-sits-at-298-ntpm-in-basal-ganglia

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q138-kmt2a-aff1-surfaceome

LAMP5 was proposed as an immunotherapy target for MLL-rearranged leukaemia without a normal-tissue filter; two atlases independently put it above 290 in the basal ganglia, and it is not partner-specific either

also called LAMP5 normal tissue adjudication

A 2022 paper found LAMP5 by searching every surface protein in MLL-rearranged leukaemia and proposed it as a target for immune therapy. Checking the same protein in healthy organs, which that search did not do, shows it is one of the most brain-concentrated proteins in the body - and this leukaemia is treated in babies.

The same thing, for a scientist

LAMP5 is the output of the published MLL-rearranged surfaceome screen and is a demonstrated direct MLL-fusion target whose depletion blocks leukaemia growth [E1]. Applying the axis that screen did not apply: **HPA consensus 297.8 nTPM in basal ganglia; GTEx v8 323.89 TPM in nucleus accumbens, 134.31 in caudate, 85.09 in putamen, against a whole-body median of 1.21 TPM** [E2, E3]. Two independent atlases, same structure. Its normal-tissue profile does not track PTPRC (r = -0.30), so the brain signal is neural parenchyma and not resident leukocytes [E4]. Separately, LAMP5 **fails the fusion-specificity test this screen applies**: 2.68x a 13-line B-cell-precursor comparator (below the 4x threshold) and **1.18x** the KMT2A-rearranged-other-partner arm [E4]. That is not a criticism of the source paper, which explicitly sought a *lineage-independent MLL-rearranged* signature and found one. It does mean LAMP5 is a KMT2A-rearrangement antigen, not a KMT2A::AFF1 antigen, and it would not distinguish an infant with t(4;11) from a child with t(9;11).

Go deeper
Why this is a caution and not a refutation

Three things would have to be true for the brain number to be a real toxicity, and none is established here:

  1. The mRNA becomes protein in neurons. Likely, given the gene's name and its role in a subset of GABAergic neurons.
  2. That protein is at the plasma membrane, not in synaptic vesicles. UniProt lists LAMP5 at Cell membrane and at endosome, ERGIC, cytoplasmic-vesicle, synaptic-vesicle and recycling-endosome membranes, and notes it recycles to the plasma membrane 5. This is genuinely uncertain and it is the load-bearing step.
  3. The therapeutic reaches it. A CAR-T does cross into the CNS - that is why CD19 CAR-T works against CNS leukaemia - so this step is not a protection.

The honest statement is: the normal-tissue axis raises a specific, testable objection to LAMP5 that the proposing paper did not address, and the objection turns on surface presentation in neurons, which nobody has measured.

The pattern across both KMT2A-r surface antigens

Two antigens have now been proposed for this disease on fusion-driven grounds, by two groups, in two journals, three years apart:

antigenfusion relationshipnormal-tissue peakatlas agreement
NG2/CSPG4direct MLL-AF4 target geneblood vessel 124.5 nTPM / artery 246 TPMHPA and GTEx agree
LAMP5direct MLL-fusion target genebasal ganglia 297.8 nTPM / 323.9 TPMHPA and GTEx agree

Neither publication applied a normal-tissue filter, and both antigens fail one. That is the finding of this beat - not that either antigen is worthless, but that the axis has been systematically skipped, and this project skipped it too until now 6.

Facts and figures
scriptanalyses/kmt2a-aff1-surface-target-screen/run.py
statusrun
result summaryLAMP5: HPA basal ganglia 297.8 nTPM, GTEx nucleus accumbens 323.89 TPM, whole-body median 1.21 TPM; fusion-enrichment 2.68x B-precursor and 1.18x other KMT2A partners, failing both pre-registered thresholds.
two antigens same pattern
{
  "CSPG4": {
    "normal_peak": "blood vessel 124.5 nTPM / artery 246.21 TPM",
    "fusion_relation": "direct MLL-AF4 target gene"
  },
  "LAMP5": {
    "normal_peak": "basal ganglia 297.8 nTPM / 323.89 TPM",
    "fusion_relation": "direct MLL-fusion target gene, partner-agnostic"
  }
}
How this could bite us back
  • Raising a normal-tissue objection to somebody else's target on mRNA alone is exactly the over-reading this project warns about. The objection is stated as conditional on surface presentation in neurons, which is unmeasured.
  • If the field responds to this class of objection by abandoning fusion-driven antigens, it returns to lineage antigens - which are the ones this disease escapes by lineage switch. The normal-tissue filter can make a disease untreatable on paper while the clinic has no alternative.
  • LAMP5 is not AFF1-specific, so a LAMP5 trial enrolling on KMT2A-rearranged status would be correctly designed while a trial enrolling on t(4;11) would be over-narrow. Getting this backwards costs an eligible population.
Still open
  1. Is LAMP5 protein surface-exposed on human neurons? This is the single measurement that decides whether the brain mRNA is a toxicity or a red herring, and neither the proposing paper nor this beat has it.
  2. Do the two published KMT2A-r antigens represent a general pattern of skipped normal-tissue filtering in fusion-antigen discovery, or did this beat find the only two? A sweep of fusion-derived CAR targets against GTEx would answer it in a day.
  3. Would a LAMP5-directed agent with deliberately poor CNS penetration retain efficacy? KMT2A-rearranged infant ALL has frequent CNS disease, so the answer may be no, and the two requirements may be in direct conflict.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    LAMP5 was identified through an in silico human surfaceome filter over MLL-rearranged leukaemia expression data, is a direct target of the oncogenic MLL fusion protein, and was proposed as a novel immunotherapy target.

    PMID 33910331DOI
    What this evidence cannot show

    Abstract read in full; figures not read. The paper's specificity claim is against other leukaemias, which is the comparison its data supports.

    Evidence that argues against this (1)
    • The paper may discuss normal-tissue distribution in its text or supplement; this beat read only the abstract, so 'did not apply a normal-tissue filter' is a claim about the abstract's stated method, not a claim that the authors were unaware of brain expression. Haematologica 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    HPA consensus normal tissue puts LAMP5 at 297.8 nTPM in basal ganglia, its highest non-lymphohaematopoietic value.

    What this evidence cannot show

    Adult bulk brain mRNA.

    Evidence that argues against this (1)
    • Bulk tissue mRNA cannot distinguish a protein confined to intracellular vesicles from one displayed on the neuronal surface, and only the latter is a CAR target. UniProt 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    GTEx v8 independently puts LAMP5 at 323.89 TPM in nucleus accumbens (basal ganglia), 134.31 in caudate and 85.09 in putamen, with a whole-body median of 1.21 TPM across 54 tissues.

    What this evidence cannot show

    Adult post-mortem tissue, independent of HPA in sampling but overlapping in some upstream data sources, so the agreement is strong but not fully independent.

    Evidence that argues against this (1)
    • A 1.21 TPM whole-body median means LAMP5 is quieter outside the brain than almost any antigen in this screen, so for a non-CNS-penetrant modality it could have a better therapeutic index than CD19. GTEx Portal 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In this beat's screen LAMP5 is 2.68x a 13-line B-cell-precursor leukaemia comparator and 1.18x the KMT2A-rearranged-other-partner arm, failing the pre-registered fusion-enrichment thresholds of 4x and 2x; its normal-tissue profile correlates with PTPRC at r = -0.30.

    What this evidence cannot show

    Cell-line mRNA, two index lines in the B-arm, five lines in the other-partner arm. A 1.18x partner ratio computed over five lines is a weak measurement and is reported as such.

    Evidence that argues against this (1)
    • The source paper deliberately built a lineage-independent, partner-agnostic MLL-rearranged signature, so LAMP5 failing an AFF1-specificity test is the expected result and not a contradiction of it. Haematologica 2022
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt annotates LAMP5 at Cell membrane together with endosome, ERGIC, cytoplasmic-vesicle, synaptic-vesicle, growth-cone and recycling-endosome membranes, and records that it recycles from the endocytic recycling compartment to the plasma membrane.

    What this evidence cannot show

    Several location comments are ECO:0000250 by-similarity from the mouse orthologue rather than human experiments; the single human experimental citation supports the ERGIC, endosome and cell-membrane comments.

    Evidence that argues against this (1)
    • Multiple annotated compartments do not establish the steady-state distribution. A protein can be annotated in six places and be overwhelmingly at one of them. UniProt 2026
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across the whole surfaceome screen, 15 genes clear both fusion-enrichment thresholds and every one of them is killed or amber on normal tissue; CSPG4 and LAMP5 are the two with an independent published claim to being fusion-driven.

    What this evidence cannot show

    Pre-registered thresholds over cell-line mRNA. LAMP5 is in fact NOT among the 15, because it fails fusion-enrichment; it is included in this comparison because of its published status, not because the screen selected it.

    Evidence that argues against this (1)
    • Two antigens is a small base for a claimed pattern of skipped normal-tissue filtering, and publication norms in immuno-oncology often place normal-tissue analysis in a supplement that an abstract-level read would miss. Haematologica 2022
Connected to