Can RNA polymerase I and CDK inhibitors combine to treat aggressive AML?

Why pair Pol I inhibition with CDK inhibition

Aggressive AML subtypes, including MLL-rearranged, FLT3-ITD and TP53-mutant disease, remain hard to cure, and the paper notes that agents such as Venetoclax are often ineffective in FLT3-ITD or TP53-mutant patients. CX-5461 was developed as a first-in-class inhibitor of RNA Polymerase I (Pol I), the enzyme that transcribes ribosomal RNA genes; blocking it triggers the nucleolar surveillance pathway (NSP), in which ribosomal proteins such as RPL5 and RPL11 accumulate, bind MDM2 and stabilize the tumor suppressor p53. CX-5461 also poisons Topoisomerase II alpha (TOP2A), which causes DNA damage.

The authors reasoned that a second drug acting on transcription elongation could widen the therapeutic window. CDK9 phosphorylates serine 2 of the RNA Polymerase II C-terminal domain, a step required for productive transcription elongation, and CDK9 inhibition rapidly lowers short-lived survival proteins such as MCL1. The hypothesis was therefore that blocking Pol I and CDK9 together would combine nucleolar stress with loss of survival signaling, and the paper tests this rather than assuming it.

An FDA-approved drug screen nominated CDK inhibitors

The team screened a library of FDA-approved compounds in A549 cells in which RPS19 had been knocked down to activate the NSP, then ranked compounds by increased p53 intensity and reduced cell number across concentrations from 5 micromolar to 40 nanomolar. Among the top 40 ranked compounds, CDK inhibitors formed one of the enriched classes (3 compounds, 7.5%), alongside topoisomerase, DNA/RNA synthesis, HSP90 and Aurora kinase agents.

The authors prioritized the pan-CDK inhibitors Dinaciclib and Flavopiridol for follow-up. Their stated reasons were practical and mechanistic: they had already reported synergy between topoisomerase inhibitors and CX-5461, HSP90 and Aurora kinase inhibitors carry dose-limiting toxicity, and all of the highest-ranked CDK inhibitors in the screen target CDK9. In the A549 system, CX-5461 at 175 nM combined with Dinaciclib at 20 nM or Flavopiridol at 200 nM produced additive increases in p53 protein, while proliferation was suppressed synergistically with Dinaciclib at 15 nM (Bliss synergy score 25.0) and Flavopiridol at 150 nM (Bliss synergy score 12.4).

Synergy across genetically different AML cell lines

Because Dinaciclib was described as more selective and less toxic than Flavopiridol, the rest of the study focused on the CX-5461-Dinaciclib pair. The combination was tested across six AML cell lines with different driver mutations and p53 status, using drug concentrations from 0.1 to 10 times each agent's IC50 and scoring interaction with the Chou-Talalay combination index, where a value below 1 indicates synergy.

Synergy appeared in five of six lines: MV4;11 (CI 0.64), NOMO-1 (CI 0.61), MOLM-13 (CI 0.68), KG-1 (CI 0.71) and SHI-1 (CI 0.70). OCI-AML3 did not meet the synergy threshold (CI 0.99). Because the responding lines differed in p53 status, the authors interpret the breadth as evidence that p53-independent stress pathways also contribute, not that p53 is irrelevant.

In MV4;11 cells, which carry MLL-AF4 and FLT3-ITD, CX-5461 and Dinaciclib reduced proliferation by 56% and 20% respectively, while the combination produced 88% arrest. At 24 hours the combination raised late apoptosis to 42.5%, against 11.7% for CX-5461 and 25.5% for Dinaciclib alone. In the p53-null KG-1 line the combination was also synergistic, but the effect came mainly from increased G2 cell-cycle arrest rather than extra apoptosis, a difference the authors attribute to p53-independent pathways.

The NSP is required for full p53 activation

Target engagement was checked directly. CX-5461 suppressed pre-rRNA expression, Dinaciclib alone had a variable and non-significant effect on Pol I transcription, and co-treatment suppressed pre-rRNA further than CX-5461 alone. Dinaciclib and the combination, but not CX-5461, reduced phosphorylated RPB1, confirming CDK9 inhibition at the doses used.

Both drugs stabilized p53, but through the same dependency: in Mdm2 C305F mutant mouse embryonic fibroblasts, which cannot bind ribosomal proteins, p53 stabilization by either drug was impaired, and knockdown of RPL5 or RPL11 likewise reduced it. The authors conclude that an intact NSP is needed for full p53 activation by either agent. They also observed more nucleolin dispersal, a marker of nucleolar disruption, with the combination than with either drug alone.

The two drugs still produced different cellular outcomes. Dinaciclib drove rapid apoptosis, while CX-5461 mainly promoted cell-cycle arrest. The paper presents this division of labor as the reason the combination clears tumor cells efficiently and delays the onset of resistance, an interpretation the authors draw from their own time-course and resistance data rather than a directly measured causal separation.

Survival gains in mouse models at lower CX-5461 exposure

In vivo, the combination significantly improved survival in two independent, genetically distinct murine AML models, MLL-ENL Nras G12D and MLL-AF9 Nras G12D. In the MLL-AF9 Nras G12D model, synergy more than doubled survival at a CX-5461 dose well below the 40 mg/kg monotherapy maximum tolerated dose, which the authors present as disease control at sub-genotoxic exposure. A third, non-MLL model, AML1-ETO9a Nras G12D, also showed improved tumor burden reduction with the combination.

Pharmacodynamic sampling in MLL-ENL Nras G12D mice showed on-target effects in both compartments. At 3 hours, CX-5461 lowered pre-rRNA in bone marrow, while Dinaciclib and the combination lowered it in spleen, which the authors attribute to rapid cell death rather than direct Pol I inhibition. Phosphorylated RPB1 was suppressed in bone marrow by Dinaciclib and the combination. By 10 hours, Dinaciclib and the combination markedly reduced spleen weight and bone marrow cellularity, and circulating GFP-positive tumor cells fell most in those two groups.

The kinetics differed in a way that matches the in vitro mechanism: Dinaciclib and the combination produced an early apoptotic peak, whereas CX-5461 induced delayed apoptosis, and G1 accumulation in bone marrow was most pronounced with the combination. These are short pharmacodynamic windows of 3 and 10 hours, so they establish target engagement and early tumor clearance rather than long-term safety.

Primary patient samples show the largest colony reduction

The most clinically relevant ex vivo test used bone marrow mononuclear cells from four independent non-MLL AML patients with distinct mutational profiles, cultured in semi-solid media with vehicle, CX-5461 at 25 nM, Dinaciclib at 15 nM, or the combination. Baseline colony-forming capacity varied between patients, but the response pattern was similar.

CX-5461 alone reduced colony formation by roughly 50%, the Dinaciclib response varied with patient and mutation status, and the combination reproducibly suppressed colonies by more than 90% relative to vehicle across all four samples. One of these samples carried a TP53 mutation, and the authors highlight this because TP53-mutant disease is among the high-risk groups where current options fail. The result is a colony-formation endpoint in a small number of patient samples, not a clinical response rate.

Lower CX-5461 dosing may limit TOP2A DNA damage

The paper identifies CX-5461's TOP2A-dependent genotoxicity as a recognized challenge for the drug. Its data show that combining CX-5461 with Dinaciclib allows lower CX-5461 dosing while maintaining strong efficacy and significantly reducing DNA damage. This matters because CX-5461's dual action is both the source of its NSP activity and a liability that could constrain dosing in patients.

The authors also probed which CDK is responsible. Dinaciclib inhibits CDK1, CDK2, CDK5 and CDK9, and CDK9 has been reported to phosphorylate TCOF1, a Pol I-associated factor, so the Pol I effect could be indirect. Two observations argue for CDK9: all top-ranked CDK inhibitors in the screen target CDK9, and the CDK9-specific inhibitors AZD4573 and Tambiciclib reproduced Dinaciclib's synergistic benefit in AML cells. The authors still describe Dinaciclib's contribution to nucleolar disruption as incompletely resolved and call for further mechanistic work.

What the preclinical evidence does and does not support

The case for Pol I-CDK9 co-targeting rests on convergence across systems: synergy in 5 of 6 cell lines regardless of p53 status, survival benefit in two aggressive murine models plus tumor burden reduction in a third, and over 90% colony suppression in four primary patient samples including a TP53-mutant case. The mechanistic thread is that both agents require an intact NSP for full p53 stabilization, yet act through different downstream programs, rapid apoptosis from CDK9 inhibition and sustained cytostatic arrest from Pol I inhibition.

The boundaries are equally specific. All efficacy data are preclinical: cell lines, mouse models and ex vivo colony assays, with no patient treatment or clinical endpoint. The primary-sample evidence covers four patients. The in vivo pharmacodynamic readouts cover 3 and 10 hours. The authors themselves note that p53-independent pathways contributing to synergy warrant further mechanistic investigation, and that Dinaciclib's own effect on Pol I transcription remains only partially explained.

For readers weighing this combination, the paper's own priority is the scenario split rather than a single average effect: the strongest rationale is in high-risk, hard-to-treat AML, including TP53-mutant and FLT3-ITD disease and MLL-rearranged models, where the combination suppressed colony formation and extended survival at reduced CX-5461 exposure. The unresolved question is whether that preclinical separation between efficacy and TOP2A-driven DNA damage holds in patients, which is the gap any clinical translation of Pol I-CDK9 co-targeting must close.

Sources used in this answer