Central nervous system involvement at diagnosis in KMT2A-rearranged infant leukaemia
also called CNS disease · CNS involvement · CNS1 CNS2 CNS3 · leukaemic meningitis · CSF blasts
Leukaemia cells commonly cross into the fluid around the brain and spinal cord, and in KMT2A-rearranged infant leukaemia about half of babies already have cells there on the day of diagnosis. That is why every child gets chemotherapy injected into the spinal fluid as well as into the bloodstream, and why a drug that cannot reach the brain leaves a hiding place behind.
The same thing, for a scientist
CNS status is assessed by cytospin of a diagnostic lumbar puncture and graded CNS1 (no blasts), CNS2 (< 5 WBC/microL with blasts) or CNS3 (>= 5 WBC/microL with blasts, or cranial nerve palsy or a visible mass). In COG AALL0631, KMT2A-rearranged intermediate-risk infants were CNS1 in 50%, CNS2 in 34% and CNS3 in 14%, and high-risk infants under 90 days were CNS1 in only 17%, CNS2 in 57% and CNS3 in 26% - so roughly half of the intermediate-risk and more than four fifths of the youngest infants have detectable CSF blasts at presentation. The mechanism is at least partly NG2/CSPG4-dependent: in the Interfant cohort high NG2 expression associated with lower event-free survival, more circulating blasts and more frequent CNS disease and relapse, and NG2 is markedly upregulated on blasts infiltrating the CNS in xenografts, with NG2 blockade almost abolishing engraftment.
How CNS status is defined and why the categories are contested1 deeper
| Category | Definition | Interpretation |
|---|---|---|
| CNS1 | no blasts on cytospin | no detectable CNS disease |
| CNS2 | < 5 WBC per microlitre with blasts present | low-level involvement; prognostic significance debated |
| CNS3 | >= 5 WBC per microlitre with blasts, or cranial nerve palsy, or a radiologically evident mass | overt CNS leukaemia |
The categories are cytological, which makes them cheap and universally available but insensitive and operator-dependent. In a small infant a traumatic lumbar puncture can introduce peripheral blasts into the sample - and these babies often have hundreds of thousands of blasts per microlitre in the blood (concept-hyperleukocytosis-emergency) - so a CNS2 call in an infant with hyperleukocytosis is genuinely ambiguous. That ambiguity is why CNS2 frequencies vary between consortia and why a molecular or flow-based CSF assay would change the field.
The actual frequencies
From COG AALL0631, in KMT2A-rearranged infants 1:
| Intermediate risk (90-365 days, n = 111) | High risk (< 90 days, n = 35) | |
|---|---|---|
| CNS1 | 50% | 17% |
| CNS2 | 34% | 57% |
| CNS3 | 14% | 26% |
| Any CSF blasts (CNS2+CNS3) | ~48% | ~83% |
The gradient with age is striking and runs in the same direction as the survival gradient: the youngest infants are both the most likely to have CNS disease and the least likely to survive (5-year EFS 17 +/- 7% versus 40 +/- 5%) 1. In reported congenital leukaemia, CSF blasts were present in 25.4% of cases 4.
Why these blasts get into the CNS: the NG2 connection1 deeper
NG2/CSPG4 (the '7.1 antigen') is a chondroitin sulphate proteoglycan expressed almost specifically on KMT2A-rearranged blasts, which is why it is used as an immunophenotypic flag for the rearrangement (see concept-flow-cytometry-immunophenotype-kmt2a-r). It is also a molecule with a migratory day job: it functions in melanoma metastasis and in brain development.
Prieto and colleagues connected the marker to the phenotype 2. In the Interfant cohort of KMT2A-rearranged infant B-ALL, high NG2 expression associated with lower event-free survival, a higher number of circulating blasts, and more frequent CNS disease and relapse. In serial xenografts of primary KMT2A::AFF1 leukaemias, NG2 turned out not to mark a leukaemia-initiating or CNS-initiating cell - it is a malleable marker, not a stem-cell marker - but it was strongly upregulated on blasts that had actually infiltrated extramedullary sites and the CNS, those blasts carried a migratory gene-expression signature, and specific NG2 blockade almost completely abolished engraftment.
The careful reading is: NG2 is a state, not an identity. It is turned on in the cells that are moving, and blocking it impairs the move. That makes it a candidate for reducing CNS disease and for making CNS-directed therapy less toxic, which is the explicit conclusion the Interfant investigators draw 23.
What follows for a menin-directed programme
If half of KMT2A-rearranged infants have CSF blasts at diagnosis, then CNS exposure is not an optional feature of a targeted agent, it is a requirement. A menin inhibitor with poor CNS penetration would clear the marrow and leave a compartment that reseeds it - which is the classic anatomy of on-treatment relapse, and two thirds of relapses in this disease happen on treatment. Conversely, an agent that does reach the CNS could in principle allow reduction of intrathecal and high-dose systemic methotrexate, which is the single most consequential contributor to neurocognitive late effects in a baby (concept-neurocognitive-late-effects-infant-leukemia). The therapeutic prize is not only better survival but the same survival with less brain-directed chemotherapy.
What CNS-directed treatment consists of
In infants, cranial irradiation is avoided entirely because of its effect on the developing brain. CNS prophylaxis and treatment therefore rest on intrathecal chemotherapy (methotrexate, with cytarabine and hydrocortisone in triple therapy where used) delivered at every marrow-directed phase, plus systemic agents with CSF penetration - high-dose methotrexate, high-dose cytarabine and dexamethasone. Interfant protocols deliver intrathecal therapy throughout induction, consolidation and the intensification blocks (concept-interfant-06-treatment-backbone). Even so, CNS relapse remains part of the relapse pattern of this disease, and the Interfant reviewers themselves flag NG2-directed therapy as a route to 'reduce CNS disease and relapse' and to 'provide safer CNS-directed therapies' 23.
| assessment | cytospin of diagnostic lumbar puncture, graded CNS1/CNS2/CNS3 |
|---|---|
| congenital leukaemia csf blasts PCT | 25.4 |
| molecular mediator | NG2/CSPG4 (gene CSPG4) upregulated on CNS-infiltrating blasts; migratory signature; blockade abolishes engraftment in xenograft |
| cranial irradiation | avoided in infants |
definitions
{
"CNS1": "no blasts on cytospin",
"CNS2": "< 5 WBC/microL with blasts present",
"CNS3": ">= 5 WBC/microL with blasts, or cranial nerve palsy, or radiological mass"
}frequencies cog aall0631 kmt2a r
{
"intermediate_risk_90_365_days": {
"CNS1_pct": 50,
"CNS2_pct": 34,
"CNS3_pct": 14
},
"high_risk_under_90_days": {
"CNS1_pct": 17,
"CNS2_pct": 57,
"CNS3_pct": 26
}
}cns directed therapy
[ "intrathecal methotrexate (or triple intrathecal therapy)", "high-dose systemic methotrexate", "high-dose cytarabine", "dexamethasone" ]
- The CNS is a pharmacological sanctuary: an agent that clears the marrow but not the CSF converts a systemic disease into a relapsing one, and two thirds of relapses in this disease already occur on treatment.
- NG2 is a malleable state rather than a stable marker, so NG2-directed therapy could select for NG2-negative migratory blasts rather than eliminating the capacity to migrate.
- Reducing CNS-directed chemotherapy on the strength of a targeted agent's CSF exposure is a one-way bet: if the targeted agent fails, the CNS prophylaxis that was removed cannot be retro-fitted, and CNS relapse in an infant is close to unsalvageable.
- CNS2 status is partly an artefact of hyperleukocytosis and traumatic taps, so any trial that stratifies or selects on CNS status inherits that noise and will need larger numbers than the point estimates suggest.
- Would flow-cytometric or molecular (fusion-transcript or genomic-breakpoint PCR) analysis of CSF reclassify a substantial fraction of CNS1 and CNS2 infants, and does molecular CSF positivity predict relapse better than cytology?
- Do menin inhibitors achieve therapeutic CSF concentrations in infants, and if they do, can intrathecal or high-dose methotrexate be safely reduced - converting a survival drug into a late-effects drug?
- Is NG2 upregulation on CNS-infiltrating blasts a direct output of the KMT2A fusion's transcriptional programme, or a response to the microenvironment the blast enters? The answer determines whether menin inhibition would itself reduce CNS tropism.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In COG AALL0631, CNS status among KMT2A-rearranged infants was CNS1 in 50% of the intermediate-risk group (age 90-365 days, n = 111) and 17% of the high-risk group (age < 90 days, n = 35), CNS2 in 34% and 57%, and CNS3 in 14% and 26%; 5-year EFS was 40 +/- 5% for the intermediate-risk and 17 +/- 7% for the high-risk group.
FLT3 inhibitor lestaurtinib plus chemotherapy for newly diagnosed KMT2A-rearranged infant acute lymphoblastic leukemia: Children's Oncology Group trial AALL0631. · Brown PA, Kairalla JA, Hilden JM, et al. · Leukemia 2021What this evidence cannot show
CNS2 is a cytological category that is vulnerable to contamination from peripheral blood in a patient with hyperleukocytosis and to traumatic lumbar puncture in a small infant, so the high CNS2 frequency in the under-90-day group (57%) is likely to include some false positives - and those infants also have the highest circulating counts, which is exactly the confound. CNS3 at 14-26% is the more robust figure. The paper reports these as baseline characteristics, not as a prospective CNS study.
Evidence that argues against this (1)
- CNS status categories are partly an artefact of the lumbar puncture. CNS2 requires fewer than five white cells per microlitre WITH blasts, so its frequency depends on cytospin technique, on how many cells were counted and on traumatic taps — and infants have small CSF volumes and difficult punctures. A CNS2 rate of 57% in the under-90-day group against 34% in the older group is therefore confounded with how hard the tap was. Leukemia 2003
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In the Interfant cohort of KMT2A-rearranged infant B-ALL, high NG2/CSPG4 expression was associated with lower event-free survival, a higher number of circulating blasts and more frequent CNS disease and relapse; in serial xenotransplantation of primary KMT2A::AFF1 leukaemias NG2 did not enrich for leukaemia-initiating or CNS-initiating cells, but NG2 was highly upregulated on blasts infiltrating extramedullary haematopoietic sites and the CNS, those blasts showed a migratory gene-expression signature, and specific NG2 blockade resulted in almost complete loss of engraftment.
NG2 antigen is involved in leukemia invasiveness and central nervous system infiltration in MLL-rearranged infant B-ALL · Prieto C, Lopez-Millan B, Roca-Ho H, et al. · Leukemia 2018What this evidence cannot show
The clinical association is retrospective within a trial cohort and NG2 expression correlates with circulating blast count, so the CNS association may be partly mediated by tumour burden rather than by NG2 itself. The functional work is in immunodeficient-mouse xenografts, where CNS infiltration and the blood-brain barrier are imperfect models of an infant. 'Almost complete loss of engraftment' after NG2 blockade is a striking result that, as far as this node's search found, has not been independently replicated by another group, and the paper's own serial-transplant data show NG2 is not a stable stem-cell marker.
Evidence that argues against this (1)
- The same paper's own xenograft data refute the simpler hypothesis that NG2 marks the CNS-initiating cell, so NG2-directed therapy would be attacking a migratory state rather than eliminating a reservoir population - and a state can be re-entered by NG2-negative cells. source 2018
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
CNS disease is one of the clinical characteristics used to identify high-risk KMT2A-rearranged infant ALL, is especially associated with KMT2A::AFF1 and a pro-B immunophenotype, and NG2 expression is associated with CNS infiltration such that NG2-directed therapy may reduce CNS disease and relapse.
KMT2A-rearranged acute lymphoblastic leukemia in infants: current progress and challenges · Stutterheim J, van der Sluis IM, Pieters R, et al. · Haematologica 2025What this evidence cannot show
A narrative review by the trial group; the NG2 recommendation is an aspiration based on the preclinical work at [E2], not on any clinical trial of NG2-directed therapy, of which none is reported. CNS disease is not part of the formal Interfant-06 high-risk definition (which uses age, WBC and prednisone response), so 'used to identify high-risk patients' describes clinical practice rather than protocol criteria.
Evidence that argues against this (1)
- A review restates associations from the cohorts that produced them and adds no independent observation. The NG2 step is the weak one: NG2 marks KMT2A rearrangement itself, so its association with CNS infiltration may be the fusion's association, and "NG2-directed therapy may reduce CNS disease" is a proposal with no clinical test behind it. Oncofusion knowledge graph (internal analysis) 2026
- E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.book
In reported congenital leukaemia, blast cells were present in the cerebrospinal fluid in 25.4% of cases, and infants under 2 years accounted for 35-45% of all patients with CNS involvement in childhood leukaemia.
Infant Leukemia (StatPearls) · StatPearls Publishing · NCBI Bookshelf 2024What this evidence cannot show
Tertiary source pooling published case reports of congenital leukaemia, with strong publication bias; the 25.4% is well below the COG cohort figures, which is the opposite of what the acuity of congenital disease would predict and probably reflects incomplete lumbar puncture in the sickest neonates. The 35-45% statement is a share-of-all-CNS-cases figure, not a prevalence, and is easy to misread.
Evidence that argues against this (1)
- The congenital-leukaemia literature is a collection of individually reported cases, not a cohort. Roughly 175 cases assembled from the world literature over decades carries publication bias toward the visually striking and the unusual, so every frequency derived from that denominator is a frequency among CASES SOMEONE CHOSE TO PUBLISH. A 25.4% CSF-blast frequency in congenital leukaemia and a 35-45% infant share of childhood CNS leukaemia come from different denominators assembled by different means, and neither is restricted to KMT2A-rearranged disease. Blood Cells, Molecules and Diseases 2010
- ← related toCNS disease at diagnosis in infants: a real excess, measured by a procedure that partly creates it
infant-specific rate and the procedural sources of its variance
presenting feature: CNS disease at diagnosis
both track with high tumour burden at presentation and are managed in the same first hours
the same NG2 molecule drives CNS infiltration; adhesion blockade in marrow may not impede CNS tropism
defers the CNS niche, now with a 2026 primary paper in an immune-competent KMT2A::AFF1 infant model
intrathecal therapy is delivered throughout every phase of this backbone
21.16% abnormal B lymphoblasts in CSF at diagnosis, clinically silent
- ← supportsSkin first, blood later: a three-day-old referred for purple spots, with KMT2A::MLLT3 neonatal ALL
CNS2a: blasts on CSF cytospin with fewer than 5 leucocytes/uL
subdural haemorrhage probably begun in utero, with leptomeningeal, perivascular and intravascular blasts at post-mortem
CNS3 status at diagnosis, normalised after three intrathecal injections; CNS2 with orbital and sinus disease at relapse
64% blasts in CSF at relapse after two years of marrow MRD negativity, cleared to CSF MRD-negative
several of the six relapses occurred in the central nervous system