Oncofusion

start / graph / gene-cd38

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q45-immunophenotype

CD38 - the one marker that is BRIGHTER in KMT2A-rearranged B-ALL, an antigen-targeted therapy that has already produced lineage switch, and an enzyme rather than a receptor

also called CD38 · ADP-ribosyl cyclase 1 · cADPr hydrolase 1 · T10 · P28907

CD38 is an enzyme on the surface of many blood cells that makes and breaks a calcium-signalling molecule, and it is the target of daratumumab, a drug already used in myeloma. It matters here for two reasons: KMT2A-rearranged leukaemia cells carry more of it than other B-ALL, and it is one of the four antigens after which leukaemias have been observed to change lineage entirely.

The same thing, for a scientist

CD38 (UniProt **P28907**, CD38_HUMAN, **300 aa**, single reviewed human record, no gene synonyms) is ADP-ribosyl cyclase / cyclic ADP-ribose hydrolase 1 - **an ectoenzyme, not a signalling receptor**, and the shortest protein in this beat. This graph names it in 21 entries and has never written it. **Two facts make it worth its own entry rather than a line in a panel.** *It is one of the very few markers that goes UP in this genotype.* In 1,382 newly diagnosed B-ALL, KMT2A-rearranged cases showed significantly **higher** mean fluorescence intensity of **CD45, CD38 and CD58** and significantly **lower** MFI of CD34, CD22, TdT and CD123 than KMT2A-germline cases, and these differences held within the same EGIL subgroup [E1]. Almost every other marker in this beat is defined by what KMT2A-rearranged blasts *lack* - CD10, CD34, CD22, CD123. CD38 is on the other side of the ledger. *It is already on the lineage-switch list.* Project EVOLVE collected **75 lineage switches after antigen-targeted immunotherapy against CD19, CD22, CD38 and CD7** - four antigens, and CD38 is one of them. Switch emerged at a **median of 1.5 months** after immunotherapy with 81.4% within six months, **64.3% of the B-ALL-derived cases carried KMT2A rearrangement**, remission rates after switch were **under 40%** and median overall survival was **4.8 months** [E2]. **The consequence is the one this beat keeps arriving at.** CD38 is attractive here precisely because it is bright in the genotype - and the genotype is the one that answers antigen pressure by changing lineage rather than losing the epitope. **Being the brightest antigen on a plastic blast is not the same as being a good target for it.** CD38 is also one half of the CD34/CD38 stem-cell grid: the fraction enriched for **normal** HSCs in patients with MLL-AF4, MLL-AF9 or MLL-ENL is **CD34+CD38-CD19-CD33-**, while the MLL-AF4 leukaemia-initiating fraction is **CD34+CD38+CD19+** [E3]. So CD38 positivity, in that grid, is what separates the leukaemia-initiating cells from the normal stem cells - an argument for CD38 targeting that the lineage-switch data immediately complicates.

Go deeper
Why CD38 is an unusual target: it is an enzyme with a job

Every other marker in this beat is a receptor, an adhesion molecule or a structural glycoprotein. CD38 is a catalytically active ectoenzyme that synthesises and hydrolyses cyclic ADP-ribose and consumes NAD+. Three consequences a panel line cannot carry:

  • Targeting it changes metabolism, not just marks a cell. CD38 is the dominant NAD+-consuming ectoenzyme in several tissues, and anti-CD38 antibodies alter NAD+ availability in the microenvironment as well as killing CD38-positive cells.
  • CD38 is on activated T cells and NK cells. An anti-CD38 antibody given alongside or before a T-cell-based therapy - a CAR, a BiTE - can deplete the effector cells. In infant KMT2A-rearranged disease, where blinatumomab and CD19 CAR-T are in use, the sequencing matters.
  • Expression is dynamic and drug-modifiable. CD38 surface levels fall under anti-CD38 antibody exposure, which is the standard myeloma escape route and which has been reported in daratumumab-treated patients.

None of this has been measured in infant KMT2A-rearranged B-ALL. The CD38-directed literature reached this beat almost entirely through T-ALL and myeloma.

Cell type, cell state, localisation, epigenetics
  • Cell type. Broad: plasma cells (brightest), activated T and NK cells, monocytes, a subset of B and myeloid progenitors, and non-haematopoietic tissue. CD38 is not lineage-restricted, which is why it is a target across myeloma, T-ALL and AML and why on-target toxicity is diffuse.
  • Cell state. An activation marker as much as a lineage marker; bright in KMT2A-rearranged B-ALL relative to other B-ALL 1; CD38-positivity distinguishes the MLL-AF4 leukaemia-initiating fraction from the normal HSC fraction within CD34-positive cells 3.
  • Localisation. Type II transmembrane ectoenzyme at the plasma membrane, 300 residues 4; the catalytic domain is extracellular.
  • Epigenetics. Nothing in this beat establishes CD38 as a direct KMT2A-fusion target gene. Its brightness in this genotype is unexplained, and that is itself a finding: the marker that goes up is the one nobody has a mechanism for.
Facts and figures
uniprotP28907
gene symbolCD38
length AA300
cd numberCD38
enzymaticADP-ribosyl cyclase / cyclic ADP-ribose hydrolase 1; NAD+-consuming ectoenzyme, catalytic domain extracellular
direction in kmt2a r ballHIGHER MFI than KMT2A-germline B-ALL, one of only three markers (with CD45 and CD58) that goes up; author within-subgroup confirmation omits CD38 (PMID 38468826)
lineage switch antigenone of the four antigens (CD19, CD22, CD38, CD7) after which Project EVOLVE collected 75 switches; per-antigen breakdown not given (PMID 40193715)
accession resolutiongene_exact:CD38 AND organism_id:9606 AND reviewed:true, fields include gene_primary and length; one row; primary name read back as CD38; length 300 asserted for verify-accessions.py.
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.053,
  "chronos_mean_other_leukaemia": 0.045,
  "difference": 0.008,
  "pan_essentiality_all_lines": 0.0303,
  "selectivity_rank": 10184,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • CD38 is one of the four antigens after which Project EVOLVE collected lineage switches, and KMT2A rearrangement was present in 64.3% of the B-ALL-derived cases. Adding CD38 to a targeting strategy in this genotype adds pressure on a blast whose documented answer to pressure is to stop being a B cell. Median survival after switch was 4.8 months.
  • CD38 is on activated T and NK cells. An anti-CD38 antibody given near a CD19 CAR-T or blinatumomab can deplete the effector population, and infant KMT2A-rearranged disease is treated with both. Sequencing is not a detail here.
  • CD38 is a catalytically active NAD+-consuming ectoenzyme with broad tissue distribution, so anti-CD38 therapy perturbs microenvironmental NAD+ metabolism as well as killing cells. No study in this beat characterises that in infant marrow.
  • CD38 surface density falls under antibody pressure without any mutation - the standard myeloma escape. A CD38-anchored MRD gate would therefore drift downward under the very therapy it is monitoring, in the same direction as a response.
  • The only quantitative claim behind 'CD38 is bright in KMT2A-rearranged B-ALL' is a group-mean MFI from one centre, and the authors' own within-subgroup confirmation omits CD38. This entry is deliberately not treating CD38 brightness as established enough to build a target argument on.
Still open
  1. How many of the 75 Project EVOLVE lineage switches followed CD38-directed therapy? The abstract names CD38 among the preceding antigens and does not break the cases down, and this is the number that would decide whether CD38 targeting is safe in this genotype.
  2. What is the CD38 antigen site density on infant KMT2A-rearranged blasts, in molecules per cell? Every CD38 claim in this entry is an MFI or a sorting coordinate.
  3. Why is CD38 brighter in this genotype? Unlike CSPG4 and PROM1 there is no proposed mechanism, and a direct-binding or H3K79 analysis at the CD38 locus in MLL-AF4 cells has not been reported.
  4. Does anti-CD38 antibody exposure before CAR-T manufacture or infusion impair the product in children? The apheresis-composition question has been raised for blinatumomab and inotuzumab and not for daratumumab.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Among 1,382 newly diagnosed B-ALL cases, KMT2A-rearranged B-ALLs showed significantly higher mean fluorescence intensity of CD45, CD38 and CD58, and significantly lower MFI of CD34, CD22, TdT and CD123, than KMT2A-germline cases. The study confirmed differences in CD45, CD34, CD22 and TdT MFI within the same EGIL immunological group, indicating no influence of the EGIL subtype on the KMT2A-rearranged immunophenotype.

    PMID 38468826DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **Single-centre, retrospective, mixed adult and paediatric referral cohort at one Rome institution.** The comparison is against KMT2A-germline B-ALL of all subtypes, so part of the CD38 difference reflects the comparator's composition. **MFI is instrument- and reagent-dependent and is not antigen site density**, so 'CD38-bright' here cannot be converted into molecules per cell, which is the quantity an antibody or CAR would need. Note also that the within-EGIL-subgroup confirmation sentence names CD45, CD34, CD22 and TdT - **not CD38** - so the claim that the CD38 difference is independent of EGIL subtype is not made by the authors and is not made here.

    Evidence that argues against this (1)
    • The authors' own within-subgroup confirmation lists CD45, CD34, CD22 and TdT and omits CD38, so the CD38 elevation may be confounded by the EGIL composition difference between the arms - KMT2A-rearranged cases are overwhelmingly pro-B and the comparator is not. The 'CD38 is brighter in this genotype' claim is therefore weaker than the other MFI findings in the same paper. Bisegna ML et al. 2024
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Project EVOLVE identified cases of lineage switch after a range of antigen-targeted therapies - CD19, CD22, CD38 and CD7 - and evaluated 75 cases: 53 (70.7%) B-ALL transforming to AML, 17 (22.7%) B-ALL to mixed phenotype acute leukaemia or acute leukaemia of ambiguous lineage, and 5 (6.7%) rare presentations such as T-ALL to AML. A further 10 cases with incomplete immunophenotypic change were described as lineage drift. Among the 70 B-ALL-derived cases, lineage switch emerged at a median of 1.5 months (range 0-36.5) after immunotherapy with 81.4% presenting within the first 6 months of the most proximal immunotherapy; 45 of 70 (64.3%) involved KMT2A rearrangements. Remission rates were under 40% and median overall survival after lineage switch diagnosis was 4.8 months.

    PMID 40193715DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read in this session (this graph records it as read in full elsewhere). **A retrospective international collection of reported cases with NO denominator** - it cannot give an incidence of lineage switch after CD38 targeting, or after anything else, only a description of the cases collected. The abstract does not break the 75 cases down by preceding antigen, so **the number of switches specifically after CD38-directed therapy is not established here.** That is the number this entry would most want and does not have.

    Evidence that argues against this (1)
    • A case collection over-represents the genotype people already expect to switch: a switch in a KMT2A-rearranged patient is more likely to be recognised, worked up and reported than the same event in a genotype nobody associates with plasticity. The 64.3% KMT2A figure is a property of the collection, not an estimate of risk. Kovach AE et al. 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In patients with MLL-AF4, MLL-AF9 or MLL-ENL, CD34+CD38-CD19-CD33- cells were enriched for normal haematopoietic stem cells with in vivo long-term multilineage repopulation capacity and polyclonal IGH rearrangement, while in MLL-AF4 patients CD34+CD38+CD19+ and CD34-CD19+ cells initiated leukaemia with clonal IGH rearrangement. Global gene expression profiling revealed CD9, CD32 and CD24 over-represented in MLL-AF4, MLL-AF9 and MLL-ENL leukaemia-initiating cells compared with normal HSCs, and in patient samples these molecules were expressed in CD34+CD38+ and CD34- leukaemia-initiating cells but not in CD34+CD38-CD19-CD33- HSCs.

    PMID 25538041DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **CD38 is used here as one coordinate of a sorting grid, not measured as a target antigen** - nothing in this study reports CD38 density or the effect of CD38-directed killing. The clean separation applies to the CD34-positive compartment only, and the MLL-AF4 leukaemia-initiating cells also include a CD34-negative fraction whose CD38 status is not stated in the abstract. Small numbers of primary patients per fusion.

    Evidence that argues against this (1)
    • A CD34+CD38+ versus CD34+CD38- separation between leukaemia-initiating and normal stem cells is an argument for sparing CD38-negative HSCs during CD38-directed therapy, and it is undercut by the breadth of CD38 expression outside that grid: plasma cells, activated T and NK cells and monocytes are all CD38-positive, so the sparing applies to one compartment and not to the patient. Silbert SK et al. 2025
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    CD38 is UniProt P28907 (CD38_HUMAN), ADP-ribosyl cyclase / cyclic ADP-ribose hydrolase 1, 300 amino acids, a single reviewed human record with primary gene name CD38 and no gene synonyms.

    What this evidence cannot show

    Fetched 2026-09-20 with `gene_exact:CD38 AND organism_id:9606 AND reviewed:true`, requesting `gene_primary` and `length` per rules 62 and 65; primary name read back as CD38; exactly one reviewed row; length 300 asserted for `verify-accessions.py`. Gene symbol and CD number coincide, so the alias hazard that affects CSPG4/NG2, IL3RA/CD123, MME/CD10 and PROM1/CD133 does not arise. **BST1 (CD157) is the paralogous ADP-ribosyl cyclase and is a different protein**, so a function-keyed rather than symbol-keyed search can confuse them.

    Evidence that argues against this (1)
    • CD38 surface expression falls under anti-CD38 antibody exposure - the standard myeloma escape route, reported in daratumumab-treated patients - so an accession-level identity claim says nothing about whether a treated patient's blasts are still targetable. Antigen density, not sequence, is the therapeutically relevant quantity and it is modulable. UniProt Knowledgebase, reviewed (Swiss-Prot) 2026
Connected to