Oncofusion

start / graph / hyp-ig-mrd-fails-where-the-repertoire-is-proximally-biased

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q34-mrd-monitoring

Ig-based MRD fails specifically in the KMT2A-rearranged patients whose IGH repertoire is proximally biased and still recombining

also called proximal bias MRD failure hypothesis · IGHV6-1 assay failure hypothesis

The idea is that the leftover-disease test built on antibody genes fails in a predictable subset of babies - the ones whose leukaemia is still rearranging those genes and using the nearest available segment - rather than failing at random. If true, you could tell in advance which children need a different test.

The same thing, for a scientist

Interfant-06 describes IG/TR targets in this disease as *occasionally absent, oligoclonal, or less stable* and treats that as a property of the subtype [E1]. The repertoire data suggest it is instead a property of a **measurable subset**: KMT2A::AFF1 shows proximally biased IGHV usage with IGHV6-1 in up to 50% of samples, a median of three DNJH stems and short CDR3 [E2], and in adults high-degree IGH clonal evolution driven by ongoing DH/VH-DJH recombination is associated specifically with KMT2A rearrangement [E3]. The hypothesis is that these three features - IGHV6-1 usage, multiple DNJH stems, short CDR3 - identify in advance the patients in whom an ASO-PCR Ig target will be absent, unquantifiable, or discordant with the fusion target, and that the 19 Interfant-06 patients whose disease only the KMT2A target detected [E1] are enriched for them. **Prior probability 0.35.** The mechanism is plausible and each component is measured, but no link between repertoire feature and assay failure has been observed, and the competing explanation - that failure is driven by a pre-VDJ cell of origin producing no usable target at all [E4], which is a different and non-overlapping mechanism - is at least as well supported.

Go deeper
Predictions, in order of how cheaply they can be checked
  1. Among KMT2A-rearranged infants with both MRD targets, one-log discordance in the direction of a higher KMT2A value is more common in IGHV6-1 users than in users of distal segments. Testable on stored Interfant material (exp-paired-ig-and-breakpoint-mrd-in-kmt2a-aff1-archive).
  2. Patients in whom no IG/TR target could be established have fewer detectable expanded clones on diagnostic repertoire sequencing than those in whom one was. This separates the two competing mechanisms: absence predicts a polyclonal repertoire, instability predicts a present-but-evolving one.
  3. ASO primer design failure rate is higher in this subtype than in non-KMT2A paediatric BCP-ALL, and the excess is concentrated where CDR3 is short and V usage is shared. A laboratory-records question, answerable without any sequencing.
  4. Sequencing-based MRD with marker-DNJ-stem tracking recovers the discordant patients that single-target ASO-PCR missed 35.
  5. The effect is present in adults and older children with KMT2A::AFF1, where Ig/TCR MRD remains standard, and therefore predicts a measurable excess of Ig/TCR-negative relapses in that population.
What distinguishes this from the rival explanations

Three mechanisms could produce IG/TR MRD is unreliable in KMT2A-rearranged ALL, and they are separable by the predictions above.

mechanismpredictsdiscriminating observation
Proximal bias and ongoing recombination (this hypothesis)target exists but drifts or is hard to design againstdiscordance concentrated in IGHV6-1 users with multiple stems
Pre-VDJ cell of origin 4target frequently does not exist at allpatients with no obtainable target have polyclonal diagnostic repertoires
Regenerating-marrow false positivity 1excess IG/TR positives in remission, not negativesdiscordance direction reversed: IG/TR higher than KMT2A

The Interfant-06 data are weakly informative on the third: of 7 samples differing by at least one log at end of induction, 5 were higher by KMT2A and 2 by IG/TR 1, which is consistent with the first two mechanisms dominating but is 7 samples.

The hypothesis is not that Ig MRD should be abandoned in this subtype. Interfant already reached that conclusion for infants on other grounds 1. It is that the failure is predictable, and therefore that the population still monitored on Ig/TCR - adults and older children with KMT2A::AFF1 - can be triaged rather than switched wholesale.

Facts and figures
statementIg-based MRD failure in KMT2A-rearranged ALL is concentrated in patients whose diagnostic IGH repertoire shows proximal IGHV usage (IGHV6-1), multiple DNJH stems and short CDR3, and is therefore predictable at diagnosis rather than random.
originliterature-gap
prior probability0.35
predictions
[
  "one-log KMT2A-over-IG/TR discordance is enriched in IGHV6-1 users",
  "patients with no obtainable IG/TR target have polyclonal diagnostic repertoires (favours the rival pre-VDJ mechanism instead)",
  "ASO primer design failure rate is higher in KMT2A-rearranged than in non-KMT2A paediatric BCP-ALL",
  "sequencing-based MRD with marker-DNJ-stem tracking recovers the ASO-PCR-discordant patients",
  "adults and older children with KMT2A::AFF1 show an excess of IG/TCR-negative relapses"
]
experiments
[
  "exp-paired-ig-and-breakpoint-mrd-in-kmt2a-aff1-archive"
]
analyses
[
  "ana-join-breakpoint-archive-to-interfant-mrd-outcomes"
]
falsifiers
[
  "ASO-PCR failure and KMT2A-over-IG/TR discordance show no association with IGHV6-1 usage, DNJH-stem count or CDR3 length after adjustment for age and partner",
  "patients in whom no IG/TR target could be established have normally clonal diagnostic repertoires, which would locate the failure in primer design rather than in repertoire biology",
  "ASO primer design failure rates in KMT2A-rearranged infants are no higher than in non-KMT2A paediatric BCP-ALL",
  "the discordance direction in a larger sample is symmetric or reversed, implicating regenerating-marrow false positivity rather than target instability"
]
rival mechanisms
[
  "pre-VDJ cell of origin: the target never existed",
  "regenerating-marrow false positivity: the failure is excess positives, not missed disease"
]
How this could bite us back
  • If the hypothesis holds, the recommendation is to move a predictable subset of patients onto fusion-based MRD - which requires a breakpoint characterisation service, so the benefit accrues only to centres that already have one.
  • A triage rule based on diagnostic repertoire features adds a sequencing assay at diagnosis in order to decide which MRD assay to build. In an infant marrow with limited volume, that competes for the same material the MRD target itself needs.
  • If the rival pre-VDJ mechanism is correct instead, no repertoire-based triage rule can work, because the feature that would flag the patient is the absence of the very repertoire the rule reads. The two mechanisms imply opposite responses and the hypothesis is only useful if they can be told apart.
Still open
  1. Is the proximal IGHV bias reproducible on DNA as well as on RNA? The whole hypothesis rests on an RNA-seq read-count measurement, and a DNA-side replication is the prerequisite for treating IGHV usage as a patient-level covariate.
  2. Does the bias change between diagnosis and relapse in the same patient? If the clone's V usage shifts, that is direct evidence of target instability rather than of a static design problem.
  3. How large would the effect have to be to change practice? If Ig/TCR MRD fails in 10% of KMT2A-rearranged patients, a triage rule needs to identify most of them to be worth building.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Interfant-06 preferred the KMT2A genomic breakpoint over IG/TR targets because earlier studies obtained more reliable MRD data with it than with the occasionally absent, oligoclonal, or less stable IG/TR rearrangements and because false positivity of IG/TR PCR targets in regenerating marrow has been described; overall concordance between the targets was 94% of 223 samples with 7 of 90 end-of-induction samples differing by at least one log (5 higher by KMT2A, 2 by IG/TR), and at protocol cut-offs 19 patients were positive by KMT2A and not IG/TR with 9 relapses versus 12 the reverse with 1 relapse (P = .02).

    PMID 33405950DOI
    What this evidence cannot show

    Read in full text. The characterisation of IG/TR targets is the authors' summary of earlier work. The one-log discordance counts are small (7 samples) and the 19-versus-12 comparison is post hoc. Patients with no IG/TR target are excluded from all of these analyses by construction.

    Evidence that argues against this (1)
    • The trial's primary agreement analysis found no significant discrepancy between the two targets at end of induction (P = .60), so the failure this hypothesis seeks to explain is uncommon and may not be large enough to stratify on even if the mechanism is real. Stutterheim J et al. 2021
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 121 KMT2A::AFF1 BCP-ALL cases across three age classes, proximal IGHV genes were strongly favoured with IGHV6-1 significantly elevated, up to 50% of KMT2A-rearranged samples carried B-cell receptors involving IGHV6-1, the median number of different DNJH stems per sample was three, CDR3 regions were shorter than in most other BCP-ALL subtypes, and the authors state that clonal evolution interferes with reliable detection of MRD based on IG rearrangements.

    PMID 38650597DOI
    What this evidence cannot show

    Read in full text. Diagnostic samples only, RNA-seq read counts rather than DNA clone frequencies, no remission material and no MRD assay performed. The MRD statement is the authors' inference and is the starting point of this hypothesis rather than support for it.

    Evidence that argues against this (1)
    • Interfant-06 had already switched to the genomic breakpoint three years before this was published, so in the population where the disease is commonest the predicted assay failure has already been engineered around and the hypothesis is untestable prospectively there. Stutterheim J et al. 2021
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 280 adults with BCP-ALL analysed by high-throughput IGH sequencing with marker DNJ-stems, clonal evolution at diagnosis was present in one third, and among 167 subtyped patients a high prevalence and high degree of clonal evolution driven by ongoing DH/VH-DJH recombination was associated with the presence of KMT2A gene rearrangements.

    PMID 37143651DOI
    What this evidence cannot show

    Adults, small KMT2A-rearranged subgroup of unstated size, and clonal evolution inferred from sequence relatedness at a single timepoint rather than tracked through therapy.

    Evidence that argues against this (1)
    • VH replacement, the other clonal-evolution mechanism the same study describes, was enriched in Ph-like and DUX4 rather than KMT2A-rearranged disease, so the specific mechanism operating here may have different MRD consequences from clonal evolution generally. Darzentas F et al. 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Deep B-cell receptor repertoire sequencing of t(4;11)/MLL-AF4 infant BCP-ALL samples with more than 98% blasts found no significantly expanded VDJ-rearranged B-cell clones at diagnosis or relapse compared with healthy peripheral blood, in contrast to five non-MLL BCP-ALL patients who were all significantly clonal, supporting a pre-VDJ stem or progenitor cell of origin.

    PMID 30679323DOI
    What this evidence cannot show

    Read in full text. Four diagnosis-relapse pairs and five comparators; peripheral blood; a barcoded PCR repertoire method rather than a clinical MRD assay. It is the strongest available statement of the rival mechanism and it rests on very few patients.

    Evidence that argues against this (1)
    • A clinical study of 133 infants found IGH rearrangements detectable in 91%, which is difficult to reconcile with a complete absence of expanded clones and suggests the two assays measure different quantities - so the rival mechanism may be weaker than it appears. Jansen MW et al. 2007
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    High-throughput sequencing of IGH and TRG identified 38.7% more MRD-positive patients than flow cytometry at a 0.01% threshold in 619 paired paediatric B-ALL samples, and those discrepant patients had worse outcomes than flow-negative patients.

    PMID 29284596DOI
    What this evidence cannot show

    Paediatric B-ALL with no KMT2A-rearranged or infant subgroup, and the comparator is flow rather than ASO-PCR, so it supports the general proposition that sequencing recovers missed disease and not the specific prediction about ASO-PCR failure.

    Evidence that argues against this (1)
    • The same study found similar survival estimates by both methods at the same threshold, so recovering more positives did not improve the prognostic separation on average. Wood B et al. 2018
Connected to