CD274 (PD-L1) - named in zero entries of this graph until now, a direct MLL1-H3K4me3 target gene in another tissue, and a cell-intrinsic dependency of MLL-AF9 leukaemia-initiating cells
also called PD-L1 · PDL1 · B7-H1 · B7H1 · PDCD1L1 · PDCD1LG1 · programmed death ligand 1
PD-L1 is the brake leukaemia cells press to switch off attacking T cells, and it is the target of a whole class of cancer drugs. In this family of leukaemias it does something else as well - the leukaemia stem cells need it to keep dividing, whether or not any T cell is watching.
The same thing, for a scientist
CD274 (UniProt **Q9NZQ7**, PD1L1_HUMAN, **290 aa**, primary gene name CD274, synonyms B7H1, PDCD1L1, PDCD1LG1 and PDL1, CD antigen CD274) is the ligand for PD-1 and the target of the largest drug class in modern oncology [E1]. Before this beat it was **named in zero of this graph's 1,600 entries**, which is remarkable for a graph documenting three immunotherapy failures. Two results justify an entry rather than a bounded absence. **First, KMT2A regulates it directly, in another tissue**: H3K4me3 is enriched at the cd274 promoter in pancreatic tumour cells, **MLL1 binds the cd274 promoter directly and catalyses H3K4me3 to activate PD-L1 transcription**, MLL1 inhibition or silencing reduces both, and MLL1 inhibition plus anti-PD-L1 or anti-PD-1 suppressed tumour growth in a FasL- and CTL-dependent manner [E2]. **Second, in the index fusion family it is a cell-intrinsic dependency, not a checkpoint**: in an MLL-AF9 AML model, CD274 was much higher on Mac-1+/c-Kit+ leukaemia-initiating cells than on normal HSCs, CD274-null leukaemia cells gave dramatically extended survival on serial transplantation with reduced LIC frequency and G1 arrest, and the mechanism ran through direct association with JNK and Cyclin D2 rather than through PD-1 [E3].
The pattern this makes with HLA-A
Two 'immune' molecules have now been tested by loss of function in MLL-fusion leukaemia-initiating cells, and both turned out to be leukaemia dependencies:
| gene | model | loss-of-function effect | mechanism |
|---|---|---|---|
| H2-k1 (class I, HLA-A orthologue) | MLL-AF9 | leukaemia depleted in vivo, survival extended | NK-cell-dependent - rescued by NK depletion |
| Cd274 (PD-L1) | MLL-AF9 | survival dramatically extended, LIC frequency down, G1 arrest | cell-intrinsic - JNK/Cyclin D2, rescued by Cyclin D2 overexpression |
One is immune-mediated and one is not, but the direction is the same: in this fusion family the immune-surface molecules are higher on the leukaemia stem cell than on the normal one, and the leukaemia needs them. That is the opposite of the standard 'tumours silence immune genes to hide' model, and it is the reason analogy-ezh2-driven-mhc-silencing-in-fus-erg-and-cbfa2t3-glis2 flags that the FUS::ERG analogy may invert here.
A consequence worth stating. If PD-L1 is a cell-intrinsic proliferative signal in MLL-fusion LICs, then PD-1/PD-L1 blockade in this disease might work partly for reasons that have nothing to do with T cells - or might fail for reasons that have nothing to do with T cells, because an antibody that blocks the PD-1 interface need not block a cytoplasmic JNK association.
The MLL1 connection is a direct promoter result and this graph has never cited it
PMID 28131992 is a KMT2A paper. MLL1 binds the cd274 promoter and deposits H3K4me3 there; inhibiting or silencing MLL1 lowers both the mark and PD-L1 protein; and the therapeutic proposal is MLL1 inhibition plus checkpoint blockade 2. It is in pancreatic cancer, which is why a leukaemia-scoped query never returns it: (KMT2A OR MLL) AND leukemia AND (PD-L1[tiab] OR CD274[tiab]) returns 15 records and title-checking finds none about PD-L1 in KMT2A-rearranged leukaemia 4 - while this paper, which is squarely about the KMT2A-PD-L1 axis, is excluded by the word leukemia.
That is a rule-71-shaped failure with the disease term rather than the gene term doing the damage. The lesson generalises: when the question is does the fusion's parent gene regulate X, scoping the query to the disease removes exactly the papers that answer it.
What does not transfer. The pancreatic work uses MLL1 inhibition, and this graph's concept-a-wdr5-result-is-not-automatically-a-kmt2a-axis-result and concept-kmt2-family-attribution-problem both warn that an MLL1-axis result obtained with a WIN-site or menin-site inhibitor may be a COMPASS result. Whether the fusion protein - which has lost the SET domain and cannot deposit H3K4me3 - regulates PD-L1 the same way is a different question entirely, and an open one.
Aliases, and why a token census found nothing
PD-L1 is the name everyone uses; CD274 is the symbol; B7-H1 is the original name and sits in the B7-H numbering family that mechanism-icoslg-drives-regulatory-t-cells-in-infant-t-4-11-all warns about (B7-H2 = ICOSLG, B7-H3 = CD276, B7-H4 = VTCN1). A census over this graph's node files for PD-L1 returned 0 and for CD274 returned 0 as of the start of this beat.
That is not a symbol-resolution failure. It is a genuine content gap in a graph that carries 42 mentions of 'checkpoint' and none of the most-studied checkpoint ligand in medicine. The most likely cause is that the checkpoint mentions arrived through therapeutic-landscape prose rather than through any beat that went looking for checkpoint biology in this disease - and, correctly, found almost none to report.
| gene symbol | CD274 |
|---|---|
| gene primary | CD274 |
| uniprot | Q9NZQ7 |
| length AA | 290 |
| uniprot entry name | PD1L1_HUMAN |
| cd antigen | CD274 - the one-gene case under rule 69 |
| alias warning | PDCD1LG1 (this gene) versus PDCD1LG2 (PD-L2, CD273) differ by one character; B7-H numbering is inconsistent across the literature (B7-H2 = ICOSLG, B7-H3 = CD276, B7-H4 = VTCN1) |
gene synonyms
[ "B7H1", "PDCD1L1", "PDCD1LG1", "PDL1" ]
kmt2a axis
{
"finding": "MLL1 binds the cd274 promoter and catalyses H3K4me3 to activate PD-L1 transcription; MLL1 inhibition lowers both",
"tissue": "pancreatic cancer",
"pmid": "28131992",
"transfer_caveat": "the mechanism is SET-domain catalysis, which every KMT2A fusion has lost - so this concerns the retained wild-type allele",
"title": "The MLL1-H3K4me3 Axis-Mediated PD-L1 Expression and Pancreatic Cancer Immune Evasion.",
"authors": "Lu C, Paschall AV, Shi H, et al.",
"venue": "J Natl Cancer Inst",
"year": 2017,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}cell intrinsic dependency
{
"model": "MLL-AF9 murine AML",
"effect": "CD274-null extends survival on serial transplantation, lowers LIC frequency, arrests G1",
"mechanism": "direct association with JNK raising Cyclin D2; rescued by Cyclin D2 overexpression",
"pmid": "27855694",
"title": "CD274 promotes cell cycle entry of leukemia-initiating cells through JNK/Cyclin D2 signaling.",
"authors": "Fang X, Chen C, Xia F, et al.",
"venue": "J Hematol Oncol",
"year": 2016,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}graph census before this beat — the numbers behind the chart
{
"PD-L1": 0,
"CD274": 0,
"checkpoint": 42
}counted absence
{
"query": "(KMT2A OR MLL) AND leukemia AND (PD-L1[tiab] OR CD274[tiab])",
"count": 15,
"on_topic_after_title_check": 0,
"date": "2026-09-20"
}depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": 0.2529,
"chronos_mean_other_leukaemia": 0.1667,
"difference": 0.0863,
"pan_essentiality_all_lines": 0.1798,
"selectivity_rank": 16554,
"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."
}- If PD-L1 is a cell-intrinsic proliferative signal in leukaemia-initiating cells, checkpoint blockade could leave the dependency untouched while producing immune-related toxicity, which is the worst combination of outcomes.
- If MLL1 SET-domain activity drives PD-L1 transcription, then any agent that restores or mimics KMT2A H3K4me3 activity raises the brake as well as the display.
- A drug that raises MHC class I and does not lower PD-L1 produces a better-presented cell with an intact off switch, and the combination of the two would need to be measured together - no published experiment does.
- PD-1/PD-L1 blockade in children carries autoimmune risks in a population whose immune system is still developing, and the efficacy case in this genotype is currently zero published patients.
- Does a menin inhibitor lower PD-L1? The pancreatic result says PD-L1 transcription needs KMT2A SET-domain activity; a menin inhibitor evicts KMT2A from active genes. If PD-L1 falls while MHC rises, the drug improves immune visibility twice over, and nobody has measured it.
- Is PD-L1 a cell-intrinsic dependency in KMT2A-rearranged human leukaemia, as it is in the murine MLL-AF9 model? If so, the right modality is a degrader, not a blocking antibody, and that has never been tested.
- What is PD-L1 expression on infant KMT2A::AFF1 blasts at diagnosis and relapse? There is no published measurement, and this graph had no entry naming the molecule at all.
- Does the fusion protein, lacking the SET domain, regulate PD-L1 at all - or is the axis entirely a property of the retained wild-type allele?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The reviewed human UniProt record for programmed death ligand 1 is Q9NZQ7 (PD1L1_HUMAN), primary gene name CD274, synonyms B7H1, PDCD1L1, PDCD1LG1 and PDL1, 290 aa, protein name 'Programmed cell death 1 ligand 1' with alternative names PD-L1, B7 homolog 1 and B7-H1, CD antigen CD274.
UniProtKB reviewed human entry Q9NZQ7 (PD1L1_HUMAN) · UniProt Consortium · UniProtKB 2026What this evidence cannot show
Database fact, fetched 2026-09-20 with gene_primary and length read back per rule 62; one reviewed human record. Note the alias load: five synonyms, of which PDL1 and PDCD1LG1 are commonly confused with PDCD1LG2 (PD-L2, CD273), a different gene with a different ligand biology. The B7-H numbering family is inconsistently applied across the literature and should be resolved before use.
Evidence that argues against this (1)
- PD-L1 protein is regulated post-transcriptionally and by glycosylation and degradation as much as by transcription, so a transcriptional result at the CD274 promoter is not a reliable predictor of surface PD-L1, and an accession and a transcript measurement do not settle what a T cell sees. Lu C et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
H3K4 trimethylation is enriched at the cd274 promoter in pancreatic tumour cells; MLL1 directly binds the cd274 promoter to catalyse H3K4me3 and activate PD-L1 transcription in tumour cells; inhibition or silencing of MLL1 decreases H3K4me3 at the cd274 promoter and decreases PD-L1 expression; and inhibition of MLL1 combined with anti-PD-L1 or anti-PD-1 antibody immunotherapy effectively suppressed pancreatic tumour growth in a FasL- and cytotoxic-T-lymphocyte-dependent manner.
The MLL1-H3K4me3 Axis-Mediated PD-L1 Expression and Pancreatic Cancer Immune Evasion. · Lu C, Paschall AV, Shi H, Savage N, Waller JL, Sabbatini ME, Oberlies NH, Pearce C, et al. · J Natl Cancer Inst 2017What this evidence cannot show
Pancreatic cancer, mouse models and human specimens; no leukaemia and no KMT2A rearrangement. The MLL1 manipulation is by histone methyltransferase inhibitor and RNAi, and this graph's standing warning applies: an inhibitor targeting the WIN site or the menin interface perturbs COMPASS complexes generally, so the attribution to KMT2A specifically rests on the RNAi arm. Most importantly, the mechanism is H3K4me3 deposition by the SET domain - which every KMT2A fusion has lost - so this result speaks to the wild-type allele, not to the fusion. Read: abstract only; record and identifiers fetched 2026-09-20.
Evidence that argues against this (1)
- If PD-L1 activation requires KMT2A SET-domain catalysis, then a KMT2A fusion cannot drive it and the axis is about the retained wild-type allele - which means a menin inhibitor, by evicting both the fusion and the wild-type protein from chromatin, would be expected to LOWER PD-L1. That is a testable and therapeutically attractive prediction which no published experiment has examined in either direction. Sparbier CE et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In an MLL-AF9-induced acute myeloid leukaemia model in wild-type and CD274-null mice, CD274 was expressed at much higher levels on Mac-1+/c-Kit+ leukaemia-initiating cells than on normal haematopoietic stem cells; CD274-null leukaemia gave dramatically extended survival on serial transplantation, a significant decrease in leukaemia-initiating-cell frequency and arrest in G1; CD274 level was inversely correlated with overall survival of AML patients in silico; and mechanistically CD274 associated directly with JNK and raised Cyclin D2, with Cyclin D2 overexpression fully rescuing loss of CD274 - while CD274 is not required for maintenance of the normal haematopoietic stem cell pool.
CD274 promotes cell cycle entry of leukemia-initiating cells through JNK/Cyclin D2 signaling. · Fang X, Chen C, Xia F, Yu Z, Zhang Y, Zhang F, Gu H, Wan J, et al. · J Hematol Oncol 2016What this evidence cannot show
Murine MLL-AF9, myeloid; the human arm is an in silico survival correlation. The cell-intrinsic mechanism is demonstrated in a genetic knockout, which removes the whole protein - an antibody that blocks the PD-1 binding interface would not be expected to reproduce it, so this is not evidence that checkpoint blockade would work here. The normal-HSC sparing is attributed to the authors' previous study rather than shown in this one. Read: abstract only; record and identifiers fetched 2026-09-20.
Evidence that argues against this (1)
- A cell-intrinsic requirement for CD274 in leukaemia-initiating cells means the therapeutic reading of a PD-L1-high blast inverts: PD-L1 becomes a dependency to degrade rather than an interaction to block, and the enormous clinical infrastructure built around blocking antibodies is the wrong modality for it. No degrader or knockdown approach to PD-L1 has been tested in this disease. Fang X et al. 2016
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
As of 2026-09-20, `(KMT2A OR MLL) AND leukemia AND (PD-L1[tiab] OR CD274[tiab])` returns 15 PubMed records, and title-checking all 15 finds none reporting PD-L1 biology in KMT2A-rearranged leukaemia - the set comprises AML studies in other genotypes (TP53, splicing-factor mutant, p53 activation, BET inhibition), myeloproliferative and paroxysmal nocturnal haemoglobinuria studies, a serum soluble PD-L1 assay paper, a herbal pharmacology review, an oesophageal squamous carcinoma study, the MLL-AF9 leukaemia-initiating-cell study and the pancreatic MLL1-PD-L1 study; and before this beat, PD-L1 and CD274 each appeared in zero of this graph's node files.
Q54 antigen-presentation beat: bounded counted-absence sweep, PubMed E-utilities esearch with querytranslation read back -- PD-L1 and CD274 in KMT2A/MLL leukaemia, all 15 titles checked, plus a regex census over the graph's node files · Q54-immune, Oncofusion Knowledge Graph · Oncofusion Knowledge Graph 2026What this evidence cannot show
Bounded absence. Name list: PD-L1 and CD274. NOT searched: B7-H1, B7H1, PDCD1LG1, PDL1 unhyphenated, or 'programmed death ligand'. The disease term `leukemia` is itself the binding constraint here - it excludes PMID 28131992, the one paper about the KMT2A-PD-L1 axis - which is the point the entry makes. Title checks only.
Evidence that argues against this (1)
- The zero is a zero about KMT2A-rearranged leukaemia and not about the axis. Scoping out the disease term recovers a direct MLL1-cd274 promoter result, and scoping out the checkpoint framing recovers a cell-intrinsic MLL-AF9 dependency. Both are in the same 15 records or one query away, which makes this a vocabulary result rather than an empty field. Lu C et al. 2017
- depends on →KMT2A / MLL1 — the gene that breaks
MLL1 binds the cd274 promoter and catalyses H3K4me3 to activate PD-L1 - in pancreatic cancer, by SET-domain activity the fusion has lost
H3K4me3 enriched at the cd274 promoter; MLL1 inhibition lowers both the mark and PD-L1
- regulates →The stem-like self-renewal state — a limited stem-cell module grafted onto a committed cell
cell-intrinsic: CD274-null MLL-AF9 leukaemia has lower LIC frequency and G1 arrest via JNK/Cyclin D2
the checkpoint ligand that is a bounded counted absence in this disease