NF1 in KMT2A-r leukemia — one homozygous deletion, one missense, and the only loss-of-function co-lesion in the disease

Two distinct NF1 observations exist in the St. Jude PCGP dataset, and they are easy to conflate.

1. A homozygous deletion in case INF018 1. Discussing the functional status of the pathway lesions, the paper states that the majority of the identified tyrosine kinase/PI3K/RAS mutations have been previously shown to be activating mutations — 'or inactivating mutations in the case of the homozygous deletion of NF1 in INF018'. This is the only lesion in the entire cohort that the authors flag as inactivating. Biallelic NF1 loss is a bona fide, fully penetrant route to RAS-MAPK activation, and in a single infant leukemia genome carrying ~1.3 non-silent mutations, a homozygous deletion of the RAS GAP is about as unambiguous a second hit as this disease produces.

2. NF1 listed as non-recurrent 1. In the same cohort, the non-recurrent kinase/PI3K/RAS mutations are FLT3, NF1, PTPN11 and PIK3R1 — 'non-recurrent' meaning observed once. NF1 was not among the recurrent lesions in the non-infant KMT2A-r cohort (which were FLT3, KRAS and NRAS at n = 3 each, with non-recurrent CBL, PIK3CD, PTPN11 and PPM1J), so NF1 appears to be an infant-cohort finding in this dataset.

3. A missense variant of uncertain significance 2. The public curation of the same study returns exactly one NF1 mutation in one of 93 sequenced samples: V1291I. V1291I is a conservative valine-to-isoleucine substitution outside the GAP-related domain (NF1's GRD spans roughly residues 1198-1530 in a 2,839-residue protein, so V1291 does sit within the GRD, but the substitution is chemically minimal). This node does not assert that V1291I is activating or pathogenic; no functional data for it in leukemia was located. It is most likely a passenger or a variant of uncertain significance, and it is probably a different event from the INF018 homozygous deletion — a copy-number lesion would not appear in a mutation table at all.

So the defensible statement is: NF1 is altered in roughly 1-2 of 47 infant KMT2A-r ALL cases, once by homozygous deletion and once by a missense variant of uncertain significance. Any higher frequency claim is unsupported.

The headline finding of this beat is that infant KMT2A-r ALL has essentially no recurrent loss-of-function lesions: no recurrent TP53, no recurrent CDKN2A/B (4/47, 8.5%), no recurrent IKZF1, no PTEN 1, 3. That absence is the structural reason the disease offers no synthetic-lethal partner (see comut-low-mutation-burden-infant-kmt2a-r).

NF1 is the exception that proves the rule. It is the one gene in the published co-lesion inventory that is inactivated rather than activated, and biallelically at that. Three consequences follow.

Mechanistically, NF1-null and RAS-mutant are not equivalent even though both activate RAS-MAPK. NF1 loss raises RAS-GTP across all three RAS isoforms simultaneously and is not relieved by intrinsic hydrolysis, so NF1-null cells are typically more MEK-inhibitor-sensitive than cells carrying a codon-12 RAS allele, and they are entirely insensitive to allele-selective RAS inhibitors. If a KMT2A-r patient were selected for MEK inhibition on genotype, an NF1-null case would be the best candidate in the cohort — and the trial designs that exist select on RAS point mutation, which would miss it 4.

Clonally, this is the one pathway lesion whose clonality can be inferred without an allele-frequency argument: a homozygous deletion detected by whole-genome sequencing and SNP array in a 92%-pure sample is clonal or near-clonal. That distinguishes it from the RAS point mutations, 65% of which sat below 30% mutant allele frequency 1. If clonal pathway lesions are the ones with prognostic weight — which is the most plausible reconciliation of the contradictory RAS prognostic literature (see comut-ras-pathway-kmt2a-r) — then NF1-null cases are the subgroup where the RAS-adverse signal should be strongest.

For synthetic lethality, NF1 loss is the only genuine loss-of-function scar on offer in this disease, and NF1-null cancers have a substantial synthetic-lethal literature in other contexts. Whether any of it transfers to a fetal B-progenitor carrying a KMT2A fusion is entirely untested, and with one or two cases per 47 it is not a tractable clinical hypothesis — but it is the only place in this beat where the question can even be asked.

NF1 was not reported as a recurrent lesion in the 124-case Interfant-99/06 cohort, where recurrent mutations were found exclusively in KRAS and NRAS 3. Combined with the PCGP finding of NF1 as non-recurrent at n = 47, the pooled picture is that NF1 alteration occurs in a low single-digit percentage of infant KMT2A-r ALL at most.

One caveat cuts the other way, though. Both negatives are mutation-centric. The Interfant cohort's headline burden metric was 'one genomic rearrangement and 2.5 non-silent SNVs per case' 3, and a homozygous deletion is neither — it is a copy-number event. Whether the Interfant analysis was powered to detect focal biallelic NF1 loss is not established from the abstract. Given that the PCGP finding is a deletion and was found by combined whole-genome plus SNP-array analysis, it is possible that NF1 loss is systematically under-called in cohorts analysed primarily for point mutations. The honest position is that the frequency of NF1 deletion in infant KMT2A-r ALL has been measured once, in 22 whole-genome cases, and found once.