Cyclin-dependent kinases are serine/threonine enzymes that regulate fundamental cellular processes, including cell-cycle transitions and gene transcription. Humans carry 21 of them. The cell-cycle CDKs, such as CDK1, CDK2, CDK4 and CDK6, shepherd cells through G1, S, G2 and mitosis by partnering with cyclins, the proteins whose sequential binding drives progression through each phase. A separate group, comprising CDK7, CDK8, CDK9, CDK11, CDK12, CDK13 and CDK19, regulates transcription. CDK7, as part of the CDK-activating kinase module within TFIIH, phosphorylates serine 5 in the carboxy-terminal domain of RNA polymerase II to support transcription initiation. CDK9 binds cyclin T to form P-TEFb, the positive transcription elongation factor b, which phosphorylates serine 2 of the same domain along with DSIF and NELF, releasing promoter-proximal pausing and allowing productive elongation of RNA transcripts. CDK12 and CDK13 pair with cyclin K to regulate elongation and RNA processing, with CDK12 additionally supporting genes involved in DNA-damage responses.

The therapeutic credibility of the CDK family rests on an unambiguous clinical foundation. Palbociclib, the first CDK inhibitor approved by the FDA in 2015, inhibits CDK4 and CDK6 and was followed by ribociclib and abemaciclib, all now standard-of-care combined with endocrine therapy for hormone receptor-positive, HER2-negative breast cancer. That success stimulated development of inhibitors against other family members, but the path was littered with failures that the review interprets carefully. First-generation pan-CDK inhibitors such as flavopiridol, which blocked cell-cycle CDKs alongside transcriptional CDK7 and CDK9, showed potent anti-proliferative activity in vitro but limited clinical activity in solid tumors and significant toxicities, leading to program discontinuation. Second-generation compounds like dinaciclib, which potently inhibits CDK1, CDK2, CDK5 and CDK9, produced promising Phase I results and greater clinical activity than flavopiridol in chronic lymphocytic leukemia, yet Phase II results in solid tumors fell short and the program was halted.

The review’s central lesson is that clinical discontinuation of a CDK inhibitor does not necessarily mean the biological target failed. Limited efficacy can reflect inadequate target engagement, inability to reach efficacious exposure because of dose-limiting toxicity, or a pharmacokinetic profile that does not sustain inhibition long enough. Flavopiridol illustrates the point: optimizing the dosing schedule based on human pharmacokinetic data improved clinical activity, with responses correlating with unbound drug exposure. The authors argue that transcriptional CDK inhibitors should therefore be judged on free-drug exposure and pharmacodynamic target engagement rather than administered dose alone. Route of administration matters too, since intravenous dosing allows tightly controlled exposure while oral dosing suits the chronic, intermittent treatment that metastatic disease demands, as the development of fadraciclib from intravenous to oral formulation demonstrated.

Selectivity itself has been redefined as a biological question rather than a purely chemical one. Because CDK family members share considerable structural similarity, particularly in their ATP-binding sites, achieving high selectivity is difficult, and increased selectivity can reduce off-target toxicity but never eliminates on-target toxicity in normal tissues. A recent example of biologically guided selectivity is atirmociclib, designed for marked selectivity toward CDK4 over CDK6. The rationale came from the dependency of HR-positive, HER2-negative breast cancer cells on CDK4, whereas CDK6 plays a critical role in hematopoiesis; reducing CDK6 activity lowered neutrophil toxicity while improving CDK4 coverage and antitumor activity in preclinical models. Selectivity can also counter adaptive resistance: activation of the cyclin E-CDK2 axis is a well-established bypass of CDK4/6 inhibition, and selective CDK2 inhibitors such as INX-315, PF-07104091 and BLU-222 have shown in preclinical models that CDK2 can become an actionable dependency in resistant breast cancer, with translational studies indicating that dependency can shift from CDK4/6 toward CDK2 during resistance development.

Transcriptional CDKs hold particular appeal for cancers addicted to oncogenic transcriptional programs. Elevated expression of transcriptional CDKs has been reported across at least 15 cancer types, and high levels or activity of CDK7, CDK8, CDK12 and CDK19 are associated with tumor aggressiveness and poor clinical outcomes. CDK9 behaves differently: it is not consistently overexpressed, but cancer cells can become dependent on CDK9-mediated transcription through drivers such as MYC and short-lived survival proteins like MCL-1. MYC overexpression drives disease progression, epithelial-to-mesenchymal transition in breast cancer, biochemical recurrence in prostate cancer and metastasis in colorectal cancer, yet MYC itself is considered undruggable. CDK9 inhibition offers an indirect strategy against MYC-dependent programs. MCL-1 overexpression correlates with poor prognosis and therapy resistance, and MCL-1 levels fall after CDK inhibitor treatment in both preclinical models and clinical studies, making it a candidate pharmacodynamic and predictive biomarker. Functional studies also link CDK9 to metastatic phenotypes: in ovarian cancer, high CDK9 expression is associated with recurrence and distant metastasis, and CDK9 inhibition reduces cancer cell migration and invasion, potentially through regulation of MMP9, an enzyme that degrades extracellular matrix to facilitate invasion.

The clinical stakes of bone metastasis are enormous. Bone is the most common site of distant metastasis in breast cancer, with approximately 70 percent of metastatic patients developing bone metastases, and bone is the first distant site in roughly 25 to 40 percent of patients who develop metastatic breast disease. In advanced prostate cancer, over 90 percent of patients with metastatic disease have bone involvement. Bone metastases are rarely curable, partly because of the tumor-supportive microenvironment described by Paget’s seed-and-soil hypothesis. Tumor cells stimulate osteoclast-mediated bone resorption, releasing growth factors stored in the bone matrix that further fuel tumor growth, creating a self-reinforcing vicious cycle. The consequences include skeletal-related events such as pathological fractures, spinal cord compression, hypercalcaemia and the need for radiotherapy or surgery to bone, along with severe cancer-associated bone pain. Breast cancer metastases typically shift bone homeostasis toward excessive resorption, whereas prostate cancer metastases are frequently osteoblastic, forming pathological new bone.

The mechanistic case for CDK9 in bone disease extends beyond oncology. CDK9 is highly active in inflammatory models, and the inhibitor LDC067 reduces inflammatory responses and protects articular cartilage in osteoarthritis models by suppressing cytokines and matrix metalloproteinases including MMP3, MMP9 and MMP13. More directly, flavopiridol suppresses bone resorption in titanium particle-induced osteolysis and ovariectomy-induced osteoporosis models and inhibits RANKL-induced osteoclastogenesis in vitro. The signaling logic is now mapped: RANKL activates its receptor, NF-kappa-B signaling induces the osteoclastogenic transcriptional program through c-Fos and NFATc1, and P-TEFb/CDK9 contributes to productive transcription of NF-kappa-B-responsive genes, with acetylated RelA/p65 recruiting BRD4, which in turn recruits CDK9 to promote RNA polymerase II elongation. Pharmacological CDK9 inhibition suppresses RANKL-induced NF-kappa-B signaling, osteoclast differentiation and bone resorption, and the selective inhibitor LDC000067 inhibits NF-kappa-B/NFATc1 signaling and protects against osteolysis in vivo, with downstream osteoclast markers such as TRACP and cathepsin K suppressed.

Recent preclinical studies have brought these threads together in metastatic disease models. ORI-15412, a selective oral brain-penetrant CDK9 inhibitor, reduced metastatic burden across brain, bone, bone marrow, lymph nodes and lung in an intracardiac MV4-11 acute myeloid leukemia model, while suppressing phosphorylation of RNA polymerase II, MCL-1 and c-MYC, inhibiting osteoclast differentiation and reducing NF-kappa-B-dependent p52 expression; IND-enabling safety and toxicity studies identified no adverse events at efficacious doses. Additional work showed oral antitumor activity, reduced PD-L1 expression and enhanced combination efficacy in melanoma and triple-negative breast cancer models. OBP-004, described as a dual CDK9/13 inhibitor, reduced metastatic tumor burden in bone, brain, lung and lymph nodes in the same leukemia model, inhibited triple-negative breast cancer growth in bone, and in a castration-resistant prostate cancer bone metastasis model reduced metastatic growth, cancer-induced bone loss and bone pain. The authors propose two complementary mechanisms: direct suppression of tumor cell proliferation, invasion and survival, and suppression of osteoclastogenesis and pathological bone resorption within the microenvironment.

Important caveats temper the enthusiasm. To the authors’ knowledge, no clinical study has yet demonstrated that CDK9 inhibition specifically improves outcomes in bone metastases, partly because trials have rarely included bone-specific evaluation even in indications where skeletal spread is frequent, such as triple-negative breast cancer and castration-resistant prostate cancer. Measurement is itself a barrier, since bone lesions without a measurable soft-tissue component are non-measurable under RECIST 1.1, prompting calls for bone-specific endpoints including skeletal-related events, imaging, biomarkers and pain assessment, with the PCWG3 framework offering disease-specific guidance in prostate cancer. Beyond CDK9, CDK12 and CDK13 present context-dependent opportunities: CDK12-mutant prostate cancer showed a median time to metastasis of 34.9 months in one cohort, CDK12 loss can create synthetic lethality to CDK13 inhibition, and CDK12/13 inhibition sensitizes tumors to PARP inhibitors and DNA-damaging therapies. The review concludes that transcriptional CDKs represent a mechanistically differentiated strategy against both tumor dependencies and the bone niche, with future development guided by tumor genotype, pharmacodynamic engagement, exposure and tolerability rather than maximal selectivity alone.

Subject of Research: Transcriptional cyclin-dependent kinases, particularly CDK9, as therapeutic targets in bone metastasizing cancers

Article Title: Transcriptional cyclin-dependent kinases as therapeutic targets in bone metastasizing cancers

Article References: Kähkönen, T. E., Galabova, G., Thormann, M., & Halleen, J. M. (2026). Transcriptional cyclin-dependent kinases as therapeutic targets in bone metastasizing cancers. Frontiers in Oncology, 16, Article 1984003. https://doi.org/10.3389/fonc.2026.1984003

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1984003

Keywords: CDK9, transcriptional CDKs, bone metastases, cyclin-dependent kinases, RANKL, osteoclastogenesis, P-TEFb, MYC, MCL-1, CDK12, CDK13, drug development