Oncofusion

start / graph / mechanism-serine-glycine-auxotrophy-mll-fusion

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q49-metabolism

Serine/glycine dependence in MLL-ENL leukaemia, reached through a DIRECT fusion target gene - and the reason a PHGDH counted absence in this graph would have been wrong

also called Ser/Gly starvation in MLL-fusion leukaemia · PTBP1-PKM2 glycolytic axis · serine biosynthesis dependency MLL-ENL

A gene the MLL fusion switches on directly, PTBP1, controls how the leukaemia cell splits sugar for energy and for building blocks. Knock it down and the cells use less glucose, make less lactate, and lose a key form of a glycolytic enzyme - and the leukaemia turns out to need serine and glycine, two amino acids the body can normally make for itself, so much that mice fed a diet without them survived significantly longer.

The same thing, for a scientist

This is the one mechanism in the beat where the chain from the fusion to the metabolic phenotype is fully written out with a **direct target gene** in the middle. Knockdown of the direct MLL-ENL target gene **PTBP1** was rate-limiting for proliferation and produced a metabolic phenotype of reduced glucose consumption and lactate production, accompanied by a reduction of the **PKM2** splice isoform of pyruvate kinase [E1]. Because PKM2 restricts glycolytic outflow to provide anabolic intermediates, the authors tested glucose, energy and Ser/Gly starvation and found that deoxyglucose, the energetic decoupler rotenone, and inhibition of serine biosynthesis by **CBR5884** all had significantly stronger effects on self-renewal and survival of transformed cells than on normal controls; serine-synthesis inhibition in particular caused accumulation of reactive oxygen species, DNA damage and apoptosis predominantly in leukaemic cells; depletion of exogenous Ser/Gly affected proliferation and self-renewal of murine and human leukaemia samples despite these being non-essential amino acids; the response correlated with glucose transport and did not involve AMPK; and **survival times in transplantation experiments were significantly extended by feeding recipients a Ser/Gly-free diet** [E1].

Go deeper
The methodological finding: this paper is invisible to a PHGDH symbol search

CBR5884 is a PHGDH inhibitor. This paper is therefore direct evidence that serine biosynthesis is a targetable dependency in MLL-fusion leukaemia - and it is unreachable from the symbol PHGDH.

Measured 2026-09-20 with the instrument described in ana-metabolic-enzyme-enumeration-kmt2a-bounded-absences: PHGDH AND (MLL OR KMT2A) AND leukemia returns 113 Europe PMC records, of which 0 have PHGDH in title or abstract together with KMT2A/MLL in the title. On the same instrument IMPDH2 returns 3, ACSL4 returns 2 and MAT2A returns 2, so the instrument works.

A counted absence on PHGDH would therefore have been accurate as a measurement and false as a conclusion - the paper frames the intervention at the level of the pathway ('inhibition of Ser biosynthesis by CBR5884') and never names the enzyme in its abstract. This is the same failure the brief's rule 64 documents for complex subunits, extended to a new class: a pathway framed by its inhibitor's name rather than its enzyme's name.

The transferable rule: before writing a counted absence on a metabolic enzyme, search the inhibitor names as well as the gene symbol. For this beat's candidates that means CBR5884 and NCT-503 for PHGDH, CB-839/telaglenastat for GLS, IACS-010759 for complex I, FK866 for NAMPT, brequinar and emvododstat for DHODH, and mycophenolic acid, mizoribine and ribavirin for IMPDH.

PKM2 is an isoform, not a gene, and the graph should say so

The paper's readout is a reduction of splice isoform-2 of pyruvate kinase M 1. There is no gene called PKM2: UniProt P14618 has primary gene name PKM at 531 aa, with PKM2 among its synonyms 2. PKM1 and PKM2 differ by mutually exclusive exons and have opposite effects on glycolytic outflow, so a PKM2 claim is an isoform claim that cannot be verified by an accession-plus-length assertion at all - both isoforms map to the same accession.

This matters for the mechanism: the whole anabolic argument depends on which isoform is present, and a reader who resolves 'PKM2' to a gene and asserts P14618/531 has recorded something true and useless.

What is strong here and what is not

Strong: a direct fusion target gene (PTBP1) sits in the causal chain; the comparator is normal cells rather than another cancer; three independent metabolic perturbations point the same way; the Ser/Gly effect is shown in murine and human leukaemia samples; and there is an in vivo dietary survival result 1.

Not strong: the model is MLL-ENL (KMT2A::MLLT1), a myeloid-capable fusion that is not this graph's index fusion; a Ser/Gly-free diet is not a human intervention, least of all in an infant; CBR5884 is a tool compound of modest potency and selectivity; and the mechanism attributes a metabolic phenotype to a splicing factor, which means a great many other splicing changes are also happening and only one was measured.

This graph already carries gene-ptbp1, written by an interactome beat from a domain-focused CRISPR screen. That entry does not cite this paper, which is the metabolic half of PTBP1's story and establishes it as a direct MLL-ENL target. Recorded here as a cross-beat gap, reported rather than edited.

Facts and figures
direction of causationfusion UPSTREAM of the metabolic phenotype, via a named direct target gene - the best-specified consequence chain in this beat
comparatornormal (non-transformed) control cells, not another leukaemia
in vivo resultsignificantly extended survival in transplantation recipients on a Ser/Gly-free diet
negative findingthe response did not involve activation of the AMPK energy homeostasis system
isoform notePKM2 is a splice isoform of PKM (P14618, 531 aa), not a gene; accession-plus-length assertions cannot discriminate PKM1 from PKM2
absence instrument lessonPHGDH returns 0 on a symbol-based KMT2A title check while this paper provides direct evidence for serine-biosynthesis dependency; searchable only via the inhibitor name CBR5884 or the phrase 'serine biosynthesis'
evidence lineageMLL-ENL / KMT2A::MLLT1 model plus murine and human leukaemia samples; not the index KMT2A::AFF1 fusion
chain
[
  "MLL-ENL (KMT2A::MLLT1)",
  "direct target gene PTBP1",
  "PKM2 splice isoform of pyruvate kinase",
  "restricted glycolytic outflow to anabolic intermediates",
  "serine/glycine biosynthetic demand",
  "ROS, DNA damage and apoptosis when serine synthesis is blocked"
]
perturbations tested
[
  "2-deoxyglucose",
  "rotenone (complex I)",
  "CBR5884 (serine biosynthesis / PHGDH)",
  "exogenous Ser/Gly depletion",
  "Ser/Gly-free diet in vivo"
]
How this could bite us back
  • Serine and glycine are required for nucleotide synthesis, glutathione and one-carbon units in every proliferating tissue. A dietary restriction that extends survival in a mouse would restrict growth in a human infant, in whom growth is the clinical priority second only to survival.
  • Serine restriction works here by causing ROS and DNA damage. Adding a DNA-damaging pressure to a leukaemia already treated with genotoxic chemotherapy is a route to therapy-related second malignancy, which is already a concern in this population.
  • The three perturbations in the source study are not interchangeable: rotenone inhibits complex I, and OXPHOS inhibition has been separately measured to PROMOTE leukaemia-initiating-cell emergence in MLL-rearranged B-ALL. A strategy assembled from this paper could pick the one perturbation that backfires in the lymphoid disease.
  • No approved PHGDH inhibitor exists, and the tool compound used is of modest potency, so the deliverable version of this mechanism does not currently exist.
Still open
  1. Does the Ser/Gly dependency hold in infant KMT2A::AFF1 B-lineage cells? The index fusion has never been tested and the lineage record in this beat is that myeloid results do not transfer.
  2. Is the dependency PHGDH-mediated or uptake-mediated? The response correlated with glucose transport, which points at flux into serine synthesis rather than at the enzyme, and a PHGDH knockout would separate them.
  3. PTBP1 is both a direct MLL-ENL target and, independently, the strongest RNA-layer dependency in KMT2A-rearranged lines by a domain-focused CRISPR screen. Is the dependency the metabolic phenotype, the splicing programme, or both?
  4. Can a Ser/Gly-restricted parenteral nutrition formulation be constructed that is safe in an infant? If not, this mechanism's only route to the clinic is a PHGDH inhibitor, and none is approved.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Knockdown of the direct MLL-ENL target gene PTBP1 was rate-limiting for cell proliferation and caused a metabolic phenotype of reduced glucose consumption and lactate production, accompanied by reduction of splice isoform-2 of pyruvate kinase M (PKM2); deoxyglucose, energetic decoupling with rotenone, and inhibition of serine biosynthesis by CBR5884 each had significantly stronger influence on self-renewal and survival of transformed cells than on normal controls; serine-synthesis inhibition caused accumulation of reactive oxygen species, DNA damage and apoptosis predominantly in leukaemic cells; depletion of exogenous serine and glycine affected proliferation and self-renewal of murine and human leukaemia samples although these are classified as non-essential amino acids; the response correlated with glucose transport but did not involve activation of AMPK; and survival times in transplantation experiments were significantly extended by feeding recipients a Ser/Gly-free diet.

    Exploitable metabolic dependencies in MLL-ENL-induced leukemia. · Garcia-Cuellar MP, Lawlor J, Bottcher M, ..., Slany RK · Blood Adv 2020
    PMID 32766858DOI
    What this evidence cannot show

    Abstract read this session; full text not retrieved, so the magnitude of the survival extension, the number of human samples and the CBR5884 concentrations were not checked. **The model is MLL-ENL (KMT2A::MLLT1), not the index fusion KMT2A::AFF1**, and the abstract does not report a lymphoid or infant arm. PTBP1 is a splicing factor: attributing its knockdown phenotype to the PKM2 isoform shift specifically is the authors' interpretation and many other splicing changes are expected. The dietary arm is a mouse chow manipulation with no human analogue in this population. CBR5884 is a modest tool compound and the abstract's claim is framed as 'inhibition of Ser biosynthesis', not as a PHGDH-specific result. Bounded absence: a Europe PMC sweep on 2026-09-20 found no independent replication of the Ser/Gly dietary result in any KMT2A-rearranged model.

    Evidence that argues against this (1)
    • In MLL-rearranged B-ALL, inhibiting oxidative phosphorylation blunts blast proliferation while promoting the emergence of leukaemia-initiating cells, whereas inhibiting hypoxia and glycolysis impairs them. Rotenone, one of the three perturbations used here, is an OXPHOS inhibitor, so the same manoeuvre that looks antileukaemic in this myeloid-capable model has been measured as pro-initiating in the lymphoid disease. The three perturbations in this study should not be treated as interchangeable. Morris V et al. 2022
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt reviewed human record P14618 (KPYM_HUMAN) has primary gene name PKM, length 531 aa, and lists PKM2 among its gene synonyms together with OIP3, PK2 and PK3; there is no separate human gene named PKM2, because PKM1 and PKM2 are mutually exclusive splice isoforms of the same locus.

    UniProtKB entry P14618 (KPYM_HUMAN), Pyruvate kinase PKM · UniProt Consortium · UniProt 2026
    What this evidence cannot show

    Resolved 2026-09-20 with gene_exact:PKM AND organism_id:9606 AND reviewed:true; exactly one reviewed row and gene_primary read back as PKM. The identifier lesson is the point of this item: an accession-plus-length assertion cannot discriminate PKM1 from PKM2 because both map to P14618, so verify-accessions.py offers no protection for any isoform-level claim. The related gene PKLR (liver/erythrocyte pyruvate kinase) is a distinct locus and is not what any of this literature means.

    Evidence that argues against this (1)
    • Because the two isoforms share an accession, a graph entry that records 'PKM2' as though it were a gene creates a claim that passes every mechanical check in this project while conveying the wrong biological object. The correct record is the isoform, stated as an isoform, with the gene given as PKM. Garcia-Cuellar MP et al. 2020
Connected to