ICOSLG on infant t(4;11) blasts converts T cells into regulatory T cells - the one immune-evasion route measured in this exact disease, in this exact age group
also called ICOSLG · ICOS ligand · B7-H2 · CD275 · Treg induction by t(4;11) blasts
Infant leukaemia cells with the t(4;11) rearrangement carry a surface protein that turns ordinary T cells into the kind that switch the immune response off. The more of it a child has at diagnosis, the sooner the leukaemia comes back.
The same thing, for a scientist
Almost everything in this beat is imported - from CML, from SCLC, from adult NPM1c AML, from murine MLL-AF9. This is not. Kulp et al. measured **ICOSLG expression at diagnosis in 43 patients with t(4;11) infant ALL** and found elevated levels associated with **inferior event-free survival**; a separate cohort of **18 patients at relapse** showed strongly increased ICOSLG; and co-culturing ICOSLG-high t(4;11) ALL cells with primary T cells **produced regulatory T cells**, an effect **impaired by a neutralising ICOSLG antibody** [E1]. That is an immune-evasion mechanism with a patient cohort, an outcome association, a cell-biological demonstration and a reagent that reverses it, all in the index fusion, the index lineage and the index age group. **It is also the only one.** The authors frame it as an explanation for relapse in a disease that relapses without accumulating secondary mutations - a chemo-resistance-and-immune-escape story rather than a genetic-evolution story, which fits this graph's `comut-low-mutation-burden-infant-kmt2a-r`.
Why this belongs at the top of the disease-specific list
The beat's main product is the distinction between immune evasion documented in KMT2A-rearranged disease and immune evasion documented in AML generally and assumed to transfer. The criteria that separate them are: index fusion, index lineage, index age, patient material, outcome association.
| route | fusion | lineage | age | patients | outcome link |
|---|---|---|---|---|---|
| ICOSLG / Treg induction | t(4;11) KMT2A::AFF1 | B-lymphoid | infant | 43 + 18 | yes, EFS |
| M-MDSC phenotype after lineage switch | KMT2A-r | myeloid after switch | paediatric | yes | no |
| NK resistance, inhibitory KIR ligands | KMT2A::AFF1 lines | B and myeloid | adult lines | primary blasts too | no |
| CNS T-cell exhaustion | KMT2A::AFF1 | B-lymphoid | infant model | murine + patient miR-93 | no |
| class I silencing / B2M loss | - | - | - | none in this disease | - |
| PD-L1 / checkpoint upregulation | - | - | - | none in this disease | - |
ICOSLG is the only row that fills every column. That is worth saying loudly, because the graph's immunotherapy layer has been built almost entirely on the bottom two rows' vocabulary while the top row went uncited.
What the mechanism is, and what a CD number does to it
ICOSLG is CD275, the ligand for ICOS (CD278) on T cells. ICOS signalling is required for follicular helper T-cell biology and for regulatory T-cell maintenance and expansion. A blast displaying ICOSLG is therefore providing a costimulatory input that biases the responding T cell towards a regulatory phenotype - the cell is not hiding, it is actively instructing the T cell to stand down.
Rule 69 note, because this beat exists partly to apply it. ICOSLG is the one-gene case: CD275 denotes ICOSLG and nothing else. But its alias B7-H2 sits in a family whose naming is a minefield - B7-H1 is PD-L1/CD274, B7-DC is PD-L2, B7-H3 is CD276, B7-H4 is VTCN1, and B7-H5/B7-H7 have been used inconsistently. A source saying 'B7-H2' should be resolved before use, and a token census counting B7-H2 in this graph would be counting a string, not a gene.
The mechanism is also distinct from the checkpoint biology the graph's 42 'checkpoint' mentions mostly refer to. ICOSLG is costimulatory, not co-inhibitory. Blocking it does not release a brake on a T cell; it prevents the generation of a suppressive cell. Those call for different reagents and different endpoints.
Limits, stated
- Cohort sizes. 43 at diagnosis and 18 at relapse, which is large for infant t(4;11) and small in absolute terms; an EFS association in 43 patients is hypothesis-generating.
- Expression, not causation. The outcome association is between ICOSLG level at diagnosis and EFS; the Treg induction is a co-culture. Neither shows that Tregs cause relapse in patients.
- No in vivo arm. The neutralising antibody impairs Treg development in culture; no animal model or clinical experiment is reported.
- One group. This beat found no independent replication.
Despite all of that it is the best-anchored immune-evasion finding in the index disease, which says as much about the state of the field as about the paper.
| gene | ICOSLG |
|---|---|
| cd antigen | CD275 |
| aliases warning | B7-H2 sits in a family with inconsistent B7-H numbering (B7-H1 = PD-L1/CD274, B7-H3 = CD276, B7-H4 = VTCN1); resolve before use |
| receptor | ICOS (CD278) on T cells |
| signal class | costimulatory, not co-inhibitory - blocking it prevents Treg generation rather than releasing a T-cell brake |
| reversal reagent | neutralising ICOSLG antibody, in vitro |
| why it ranks first | the only immune-evasion route in this beat that is in the index fusion, index lineage, index age group, in patient material, with an outcome association |
cohorts
{
"diagnosis": 43,
"relapse": 18,
"genotype": "t(4;11) KMT2A::AFF1",
"age": "infant",
"outcome_measure": "event-free survival"
}- If blasts actively generate regulatory T cells, then every T-cell-dependent therapy in this disease - blinatumomab, CAR T cells, donor lymphocytes, and the antigen-presentation strategies in this beat - is being deployed into a compartment the leukaemia is actively suppressing, and raising antigen presentation does nothing about it.
- ICOSLG is costimulatory, so a blocking antibody removes a signal rather than releasing a brake; the safety profile is not that of a checkpoint inhibitor and ICOS-axis manipulation has autoimmune consequences that have not been characterised in infants.
- A biomarker that predicts relapse at diagnosis invites intensification, and this graph's `risk-mrd-target-choice-changes-who-is-intensified` records what happens when a marker reassigns risk groups without a therapy attached to it.
- If ICOSLG rises at relapse, any therapy delivered at relapse - which is where menin inhibitors are currently used - is delivered into the most immunosuppressed state of the disease.
- Is ICOSLG a KMT2A::AFF1 fusion target, or a state marker? A degron or knockdown of the fusion with ICOSLG as a readout would settle whether this is fusion-driven biology or a property of the cell of origin.
- Does a menin inhibitor change ICOSLG? Everything else on the surface of these cells that has been looked at moves under menin inhibition - CLEC12A up, MHC class II up - and an inhibitory or Treg-inducing ligand moving the same way would be a bite-back on the whole strategy.
- Are regulatory T cells enriched in the marrow of infants with t(4;11) ALL at diagnosis and at relapse? The co-culture predicts it and the patient-level measurement has not been reported.
- Does ICOSLG blockade have any effect in vivo in a KMT2A::AFF1 model? The reagent exists and the experiment has not been done.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Elevated ICOSLG expression at diagnosis was associated with inferior event-free survival in a cohort of 43 patients with t(4;11) infant acute lymphoblastic leukaemia; a cohort of 18 patients with infant ALL at relapse displayed strongly increased ICOSLG expression; co-culturing ICOSLG-high t(4;11) ALL cells with primary T cells resulted in the development of regulatory T cells, and this was impaired by treatment with a neutralising ICOSLG antibody; the authors propose ICOSLG as a relapse-predicting biomarker and as a therapeutic target in a potential immune-evasion relapse mechanism, in a disease where relapse occurs despite ongoing chemotherapy and without the accumulation of secondary mutations.
The immune checkpoint ICOSLG is a relapse-predicting biomarker and therapeutic target in infant t(4;11) acute lymphoblastic leukemia. · Kulp M, Siemund AL, Larghero P, Dietz A, Alten J, Cario G, et al. · iScience 2022What this evidence cannot show
43 patients at diagnosis and 18 at relapse - large for this ultra-rare genotype, small in absolute terms, and an EFS association at that n is hypothesis-generating rather than established. The diagnosis and relapse cohorts are separate, so the increase at relapse is a between-cohort comparison, not paired sampling. The Treg induction is an in vitro co-culture with primary T cells; no in vivo model and no clinical intervention. Expression cut-points for 'elevated' are not given in the abstract. One group, no independent replication found. Read: abstract only; record and identifiers fetched 2026-09-20.
Evidence that argues against this (1)
- The framing that infant t(4;11) relapse needs an immune explanation because it occurs without secondary mutations is contestable on its own terms: this graph's `mechanism-low-mutation-burden-is-depth-bounded` records that the low mutation burden in this disease is a function of sequencing depth, and `cell-state-relapse-initiating-persister` and `mechanism-fusion-clonal-stability-diagnosis-to-relapse` offer non-immune accounts of the same observation. ICOSLG may be a marker of a persister state rather than a cause of immune escape. Kulp M et al. 2022
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
As of 2026-09-20 the immune-evasion literature in this disease is small and title-checkable: `(KMT2A OR MLL) AND leukemia AND "immune evasion"[tiab]` returns 11 records and `"immune escape"[tiab]` returns 11, with overlap; of these the records that concern KMT2A-rearranged disease specifically are the ICOSLG study in infant t(4;11) ALL, the M-MDSC study in lineage-switched KMT2A-rearranged AML, the CD19-negative myeloid escape report, a bleximenib mechanistic study in AML, a galectin-1 study in KMT2A-rearranged ALL cells and a CD36 study in MLL-AF9-driven AML, the remainder being about other diseases or other genotypes.
Q54 antigen-presentation beat: bounded counted-absence sweep, PubMed E-utilities esearch with querytranslation read back -- immune evasion and immune escape in KMT2A/MLL leukaemia, both nominalisations, title-checked · Q54-immune, Oncofusion Knowledge Graph · Oncofusion Knowledge Graph 2026What this evidence cannot show
Bounded enumeration. Rule 72 applied: 'immune evasion', 'immune escape' and 'immunoevasion' are three different strings and the first two were run; 'immunoevasion' and 'immune surveillance' were not, nor were 'immunosuppressive', 'immune suppression' or 'tolerance'. The KMT2A concept block is `KMT2A OR MLL` and does not include ALL-1, HRX or HTRX. Title checks were made on titles only, not full texts, so an on-topic paper with an off-topic title would be missed. The full enumeration lives in `ana-immune-evasion-routes-documented-in-kmt2a-r-versus-aml-generally`.
Evidence that argues against this (1)
- An enumeration built on the phrase 'immune evasion' systematically misses work that describes the same biology in other vocabulary - T-cell exhaustion, immunosuppressive microenvironment, Treg induction, NK resistance - and this beat found on-topic papers under every one of those phrasings that the immune-evasion query does not return. The number is a floor on a floor. Duguid A et al. 2026
- expressed in →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
ICOSLG elevated at diagnosis in 43 t(4;11) infants and strongly raised at relapse in 18
the one immune-evasion route measured in the index fusion, lineage and age group
proposed as the explanation for relapse without accumulating secondary mutations
route 1 of six documented in this disease