start / graph / finding-q149-three-kmt2a-germline-infant-all-cases-are-irx1-high-so-irx1-high-is-not-a-fusion-phenotype

AnalysismoderateQ149-irx1-expression-across-partners

Three of seventeen KMT2A-GERMLINE infant ALL cases are IRX1-high, one at 10.37 - so a high IRX1 in an infant leukaemia is not by itself evidence of anything fusion-related

also called q149 the germline control arm · IRX1 in KMT2A-wild-type infant ALL

Seventeen babies in the same study had leukaemia without any KMT2A fusion. Three of them still had IRX1 switched on, one of them strongly. So IRX1 being on does not mean the fusion put it there.

The same thing, for a scientist

**The most useful arm in GSE68720 is the one that has no fusion.** Seventeen of its 97 infant ALL patients are `mll_par: --`, cross-checked as the same 17 the independent `mll_ind:` field calls `MLL_G` - **KMT2A germline**, same trial, same institution set, same array, same normalisation. **IRX1 is ON in 3 of those 17 (17.6%), and the highest value is 10.37 log2** - within one unit of the highest KMT2A::AFF1 case in the cohort (11.37) and above both IRX1-high KMT2A::EPS15 cases (10.11, 11.07). **Three consequences, and they run in different directions:** 1. **It is the null that makes the AFF1 result meaningful.** 21 of 48 AFF1 cases (43.8%) against 3 of 17 germline (17.6%) - a 2.5-fold enrichment, though **Fisher two-sided p = 0.0798** at these n, i.e. not significant against the germline arm. The AFF1 signal is significant only against the **non-AFF1 KMT2A-r** arm (p = 0.0047). **The germline arm is too small to anchor the AFF1 claim, and saying so is the honest reading.** 2. **It removes the non-AFF1 KMT2A-r result entirely.** 3 of 26 non-AFF1 (11.5%) against 3 of 17 germline (17.6%), Fisher two-sided **p = 0.666**. Outside AFF1, having a KMT2A rearrangement does not change IRX1 status relative to infant ALL with no rearrangement at all. 3. **It caps what any single IRX1-high sample can mean.** A KMT2A::EPS15 blast at IRX1 10.11 is a value that also occurs in a fusion-free infant leukaemia at 10.37. **So the EPS15 observation is two draws from a background that exists, not a fusion-associated state** - until a paired occupancy measurement says otherwise. **The HOXA control behaves the same way and that is the warning.** HOXA10 is ON in 7 of 17 germline cases (41.2%) against 25 of 48 AFF1 (52.1%). **HOXA is not a KMT2A-r-specific marker in infant ALL by this readout either.** A reader who treats HOXA-high as diagnostic of a KMT2A rearrangement in infant ALL would be wrong 7 times in 17.

Go deeper

What the 17 germline cases are, and the limit on this entry

What the deposit says: mll_ind: MLL_G, mll_par: --, and 10 of the 17 have karyotype: --. The deposit does not say what drives these leukaemias. Infant ALL without a KMT2A rearrangement is a heterogeneous group - it includes the NUTM1-rearranged, PAX5-altered and fusion-negative cases that later work separated out, none of which was definable in 2012.

So this entry cannot say which infant leukaemias carry IRX1 without a KMT2A fusion. It can only say that 3 of 17 do. Identifying them would need either the P9407 clinical annotations or a re-analysis of the array data for the subtypes now recognised - both cheap, neither done.

And the arm is small. 3 of 17 has a Wilson 95% interval of roughly 6-39%. It overlaps the AFF1 interval (30.7-57.7%) and the non-AFF1 interval (4.0-29.0%) almost entirely. Every comparison against it is underpowered, and the entry's value is in the existence of the three cases rather than in the rate.

A NUTM1-rearranged possibility worth flagging. NUTM1 fusions in infant ALL drive a HOXA9-high programme, and the resource survey in this beat rejected three NUTM1 GEO series as off-beat. If any of the 7 HOXA-positive germline cases here are NUTM1-rearranged, the HOXA arm of this control is partly explained - and the IRX1-positive germline cases would still not be.

Facts and figures

cohort	GSE68720, the 17 mll_par '--' / mll_ind 'MLL_G' cases
n	17
irx1 on	3
irx1 rate	0.1765
irx1 max	10.37
hoxa10 on	7
hoxa10 rate	0.4118
prom1 on	4
irx2 on	0
karyotype missing	10
absence class	measured-not-analysed

comparisons

measurements

comparisons — the numbers behind the chart

{
  "AFF1_vs_germline": {
    "table": [
      [
        21,
        27
      ],
      [
        3,
        14
      ]
    ],
    "fisher_two_sided": 0.07981,
    "verdict": "NOT significant at these n"
  },
  "nonAFF1_vs_germline": {
    "table": [
      [
        3,
        23
      ],
      [
        3,
        14
      ]
    ],
    "fisher_two_sided": 0.666495,
    "verdict": "no difference"
  },
  "AFF1_vs_nonAFF1": {
    "table": [
      [
        21,
        27
      ],
      [
        3,
        23
      ]
    ],
    "fisher_two_sided": 0.004733,
    "verdict": "the only significant contrast in the cohort"
  }
}

measurements — the numbers behind the chart

[
  {
    "quantity": "proportion_IRX1_ON",
    "subject": "KMT2A-germline infant ALL, GSE68720",
    "value": 0.1765,
    "units": null,
    "n": 17,
    "universe": "GSE68720 GPL570, max-of-ten-silent-genes per-sample floor; the same 17 cases the independent mll_ind field calls MLL_G",
    "null": 0.4375,
    "note": "3 of 17, max value 10.37. The 'null' quoted is the AFF1 rate of 21 of 48; Fisher two-sided between them is 0.0798, i.e. NOT significant - the germline arm is too small to anchor the AFF1 claim"
  },
  {
    "quantity": "proportion_HOXA10_ON",
    "subject": "KMT2A-germline infant ALL, GSE68720",
    "value": 0.4118,
    "units": null,
    "n": 17,
    "universe": "GSE68720, same floor, HOXA10 probes 213147_at and 213150_at",
    "null": 0.5208,
    "note": "7 of 17, against 25 of 48 in AFF1. HOXA-high is NOT diagnostic of a KMT2A rearrangement in infant ALL by this readout"
  },
  {
    "quantity": "proportion_IRX2_ON",
    "subject": "KMT2A-germline infant ALL, GSE68720",
    "value": 0,
    "units": null,
    "n": 17,
    "universe": "GSE68720, same floor, 2 IRX2 probes collapsed by maximum",
    "null": 0.1875,
    "note": "0 of 17, against 9 of 48 in AFF1. IRX2 is the ONLY gene measured here that is clean against this control"
  }
]

How this could bite us back

If IRX1-high occurs in infant ALL without a KMT2A rearrangement at 17.6%, then an IRX1-directed agent tested in a KMT2A-r-only population is being tested in the wrong population - some of the patients who would benefit do not have the fusion, and some who have the fusion do not have the target.
The same applies to HOXA: 7 of 17 fusion-free infants are HOXA-positive here, so 'HOXA-high infant ALL' is not a synonym for 'KMT2A-rearranged infant ALL' and a menin-inhibitor eligibility criterion written on a HOXA signature would enrol the wrong children.

Still open

What drives the three IRX1-high KMT2A-germline infant leukaemias? If any is NUTM1-rearranged or PAX5-altered, IRX1 activation has a second route into infant ALL that nothing in this corpus records.
Could any of the 17 carry a cryptic KMT2A rearrangement missed by 2012 assays? Nothing can re-test them, which means the cleanest control arm in this beat has an irreducible doubt attached to it.
IRX2 is 0 of 17 in this arm and 9 of 48 in AFF1 - the only gene in this beat that is clean against the germline control. Is IRX2, not IRX1, the actual fusion-associated event?

Every source behind this entry (1)

E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Of 17 KMT2A-germline infant ALL cases in GSE68720, three have IRX1 transcript above a per-sample ten-gene silent-tissue floor with a maximum of 10.37 log2, and seven have HOXA10 above the same floor - giving IRX1 and HOXA rates of 17.6% and 41.2% in infant ALL with no KMT2A rearrangement at all.
Gene expression profiles predictive of outcome and age in infant acute lymphoblastic leukemia: a Children's Oncology Group study · Kang H, Wilson CS, Harvey RC, et al. · Blood 119:1872-1881 2012
PMID 22210879 DOI
What this evidence cannot show
My analysis of their deposited data. n=17 gives a Wilson 95% interval of roughly 6-39% on 3 of 17, overlapping every other arm in the cohort. The germline group's actual genetic drivers are unknown - 10 of 17 have no karyotype recorded and the 2012 subtype vocabulary predates NUTM1, PAX5-alt and the other infant ALL entities. HOXA10 substitutes for HOXA9 because GPL570 has no single-gene HOXA9 probe. Array, thresholded calls, threshold-sensitive.
Evidence that argues against this (1)
- MLL germline status in 2012 was established by the assays of the time - FISH, Southern blot or RT-PCR panels - which return negative for every partner they were not designed around. So some of these 17 may carry a cryptic KMT2A rearrangement, and if the three IRX1-positive ones do, this entry's whole point dissolves. This project's own rule is that a targeted panel returns negative for what it did not test, and there is no way to re-test these samples. Kang H et al. 2012

Connected to

part of →In infant ALL, IRX1 transcript is enriched in KMT2A::AFF1 (21 of 48) and present but rare outside it (3 of 26) - so the occupancy zero outside AFF1 is not an expression zero
the within-cohort KMT2A-germline null for that partner table
supports →IRX1 status is set by the cell of origin, and the fusion partner predicts it only because different partners transform different cells - so the partner supplies no IRX1 address
without AFF1, having the fusion does not change IRX1 status relative to fusion-free infant ALL
refutes →Two of three KMT2A::EPS15 infant ALL cases carry IRX1 at KMT2A::AFF1 magnitude, while the only KMT2A::EPS15 patient with occupancy data has no oncoprotein at IRX1 at all
a KMT2A-germline infant case reaches IRX1 10.37, so the two EPS15 values may be draws from a background
depends on →GSE68720 - 97 infant ALL patients from COG P9407 with the KMT2A partner named per sample (mll_par) and a GPL570 expression matrix, 74 partner-named KMT2A-r cases and 17 germline controls
the cohort this control arm comes from
related to →HOXA10 - named in 27 entries as part of the HOXA output and never written; a posterior HOX paralogue that is regulated with HOXA9 and tested far less often
HOXA10 is ON in 7 of 17 fusion-free infant ALL cases, so HOXA-high is not diagnostic of a KMT2A rearrangement