Knock down CDK13 and CDK12 side by side in lineage-matched KMT2A and non-KMT2A lines, and ask whether the DepMap signal is real
also called CDK13 knockdown experiment · CDK12 versus CDK13 in KMT2A-r
A public screen says KMT2A-rearranged cells need CDK13 more than other leukaemia cells do. This is the cheap experiment that would confirm or kill that: switch the gene off in matched pairs of cell lines and see whether the difference reproduces.
The same thing, for a scientist
The claim to be tested is from `hyp-cdk13-is-the-lineage-consistent-transcriptional-cdk-dependency`: CDK13 ranks 171st of 17,728 with -0.471 in the KMT2A-rearranged myeloid half and -0.492 in the lymphoid half against -0.278 in other leukaemia lines [E1]. Three confounds have to be removed and each has a specific arm. **Lineage:** run matched B-lineage and myeloid pairs rather than pooling. **Cyclin K competition:** run CDK12 knockdown in parallel and a cyclin K overexpression rescue, because if CDK13 loss acts by freeing or starving cyclin K for CDK12 the phenotype is not CDK13's [E2]. **Programme versus viability:** read HOXA9 and MEIS1 by RT-qPCR and Pol II CTD Ser2 by ChIP alongside viability, because a viability effect outside the fusion programme is a different and less interesting finding. A fourth arm separates KMT2A from the HOXA programme generally, by including a NUP98-rearranged line that carries the HOXA programme without a KMT2A fusion.
Design
Perturbation. Inducible shRNA or, better, dTAG/auxin degron on endogenous CDK13 and on CDK12 separately, with matched knockdown efficiency verified by immunoblot - not by mRNA, because these are large proteins with long half-lives.
Lines.
| arm | lines |
|---|---|
| KMT2A-r, B-lineage | SEM, RS4;11, KOPN-8 |
| KMT2A-r, myeloid | MOLM-13, THP-1, MV4-11 |
| non-KMT2A B-ALL | REH, NALM-6 |
| non-KMT2A AML | OCI-AML3, U-937 (this project's existing pathway-equivalent myeloid reference) |
| HOXA-programme control | a NUP98-rearranged line |
Readouts. Viability over 10 days; HOXA9 and MEIS1 transcript and protein; Pol II CTD Ser2 ChIP at HOXA9 and at a housekeeping control; and cyclin K immunoblot to check that knocking down one CDK does not change the other's partner availability.
Rescue. Cyclin K overexpression in the CDK13-knockdown arm. If it rescues, the phenotype belongs to cyclin K availability and CDK13 is not the target.
Pre-registered falsifiers are listed on the hypothesis node; the decisive one is that CDK13 knockdown is no more depleting in KMT2A-rearranged than in matched non-KMT2A lines at equal protein-level knockdown.
Why this cannot currently be done with a compound
Every molecule in the literature that inhibits CDK13 also inhibits CDK9, and often CDK12 and several HDACs as well; the most specific published example inhibits CDK9, CDK12, CDK13, HDAC1, HDAC2 and HDAC3 together 3. So a pharmacological test would be uninterpretable. This has to be genetic, which is also why it has not happened by accident.
| design | Inducible shRNA or degron depletion of CDK13 and CDK12 separately, at matched protein-level knockdown, across lineage-matched KMT2A-rearranged and non-KMT2A B-lineage and myeloid lines plus a NUP98-rearranged HOXA-programme control, with a cyclin K overexpression rescue arm. |
|---|---|
| readout | viability over 10 days; HOXA9 and MEIS1 transcript and protein; Pol II CTD Ser2 ChIP at HOXA9 versus a housekeeping gene; cyclin K immunoblot |
| controls | lineage matching rather than pooling; parallel CDK12 arm; cyclin K rescue; NUP98-rearranged line to separate KMT2A from the HOXA programme; knockdown verified by protein not mRNA |
| cost tier | med |
| timeline | 4-6 months |
| must be genetic | no selective CDK13 chemical probe with leukaemia data exists; published compounds co-inhibit CDK9, CDK12 and HDACs |
tests hypotheses
[ "hyp-cdk13-is-the-lineage-consistent-transcriptional-cdk-dependency" ]
related entries
[ "gene-cdk13", "gene-cdk12", "gene-cdk9" ]
- Even a clean positive would identify a target with no selective chemical matter, so the immediate translational value is low and the value is in understanding whether the CDK9 story is the right story.
- Is there a published dTAG or auxin degron allele for CDK13 in any human cell line that could be reused?
- Would a cyclin K degron be a cleaner first experiment, since it removes both kinases' partner at once and sets an upper bound on the combined phenotype?
- Does CDK13 depletion change Pol II pausing index genome-wide, and can that be compared against this graph's existing PRO-seq design at exp-pro-seq-pausing-index-kmt2a-aff1?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
CDK13 ranks 171st of 17,728 on index-minus-comparator difference in this project's DepMap scan, with -0.471 in six KMT2A-rearranged myeloid lines and -0.492 in four lymphoid lines against -0.278 in other leukaemia lines and -0.142 across all 1,208 lines.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API, re-queried for this beat on 2026-09-20 and split by lineage inside the KMT2A-rearranged arm; genome-scale ranks read from this repository's analyses/depmap-kmt2a-selective-scan-full/results.json (17,728 genes, 1,208 lines) · Broad Institute DepMap; Open Targets Platform; Oncofusion KG internal scan · Databases 2026What this evidence cannot show
Ten index lines; no permutation p; the comparator arm is itself deeper than the all-lines mean. This is the observation the experiment exists to test, not evidence that it is right.
Evidence that argues against this (1)
- CDK13 lies at 7p14.1 and appears among genes deleted in a dic(4;7) in a myelomonocytic leukaemia line, which is awkward for a dependency claim. Pan J et al. 2012
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Cyclin K binds both CDK12 and CDK13 to form two distinct complexes, so depleting one kinase changes how much cyclin K is available to the other; cyclin K inactivation in mice is embryonic lethal.
The Cyclin K/Cdk12 complex maintains genomic stability via regulation of expression of DNA damage response genes. · Blazek D, Kohoutek J, Bartholomeeusen K, Johansen E, Hulinkova P, Luo Z, Cimermancic P, Ule J, Peterlin BM · Genes Dev 2011What this evidence cannot show
Human cell lines plus a mouse genetic result; the competition for cyclin K is implied by shared partnership rather than demonstrated by a titration.
Evidence that argues against this (1)
- If CDK12 and CDK13 in fact cooperate over global Pol II processivity rather than dividing labour, a single-kinase knockdown may be partially buffered, which would make the effect sizes here underestimates. Bosken CA et al. 2014
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The most target-resolved published compound touching CDK13 in leukaemia inhibits CDK9, CDK12, CDK13, HDAC1, HDAC2 and HDAC3 together at low nanomolar concentrations, so no pharmacological test of CDK13 alone is currently possible.
Discovery of novel and bioavailable histone deacetylases and cyclin-dependent kinases dual inhibitor to impair the stemness of leukemia cells. · Liu Y, Wei Y, Wang X, Ma L, Li X, Sun Y, Wu Y, Zhang L, Wang J, Li M, Zhang K, Wei M, et al · Eur J Med Chem 2023What this evidence cannot show
One medicinal-chemistry paper; the absence of a selective probe is asserted from this beat's searches rather than from a systematic chemical-probe review.
Evidence that argues against this (1)
- A selective CDK12/13 probe may exist in the chemical-probe literature outside the leukaemia searches run here; this beat did not enumerate the chemical-probe portals. Broad Institute DepMap; Open Targets Platform; Oncofusion KG internal scan 2026
lineage-matched knockdown with a cyclin K rescue arm
the dependency this entry exists to confirm or kill