ERK hyperactivation serves as a unified mechanism of escape in intrinsic and acquired CDK4/6 inhibitor resistance in acral lentiginous melanoma
Patients with metastatic acral lentiginous melanoma (ALM) suffer worse outcomes relative to patients with other forms of cutaneous melanoma (CM), and do not benefit as well to approved melanoma therapies. Identification of cyclin-dependent kinase 4 and 6 (CDK4/6) pathway gene alterations in >60% of ALMs has led to clinical trials of the CDK4/6 inhibitor (CDK4i/6i) palbociclib for ALM; however, median progression free survival with CDK4i/6i treatment was only 2.2 months, suggesting existence of resistance mechanisms. Therapy resistance in ALM remains poorly understood; here we report hyperactivation of MAPK signaling and elevated cyclin D1 expression serve as a mechanism of intrinsic early/adaptive CDK4i/6i resistance. ALM cells that have acquired CDK4i/6i resistance following chronic treatment exposure also exhibit hyperactivation of the MAPK pathway. MEK and/or ERK inhibition increases CDK4i/6i efficacy against therapy naïve and CDK4i/6i-resistant AM cells in xenograft and patient-derived xenograft (PDX) models and promotes a defective DNA repair, cell cycle arrested and apoptotic program. Notably, gene alterations poorly correlate with protein expression of cell cycle proteins in ALM or efficacy of CDK4i/6i, urging additional strategies when stratifying patients for CDK4i/6i trial inclusion. Concurrent targeting of the MAPK pathway and CDK4/6 represents a new approach for patients with metastatic ALM to improve outcomes.
Acral lentiginous melanoma arises on the palms, soles and nail beds, and it behaves like a different disease from the melanomas that appear on sun-exposed skin. Outcomes are worse. Only about 20% of cases carry the BRAF V600E/K mutation that makes targeted therapy work elsewhere, against roughly 50% in other cutaneous melanoma, and immune checkpoint blockade is less effective. Patients are left with few options.
More than 60% of ALMs carry an alteration in the CDK4/6 pathway, which made CDK4/6 inhibition an obvious thing to try. A phase II trial of palbociclib returned a median progression-free survival of 2.2 months. The drug is not missing its target — something is escaping it.
This is collaborative wet-lab work led by the Rebecca lab at Johns Hopkins. My contribution was the statistical analysis.
Selecting patients by sequencing does not work here
Before the resistance question, a result that changes trial design.
Trial eligibility for CDK4/6 inhibitors in ALM has rested on the genetic status of CDK4 pathway nodes — CDK4 gain, CCND1 gain, CDKN2A loss. Across a cohort of 75 ALM patients with survival information, none of those alterations was significantly associated with overall survival. In TCGA, across 89 melanoma patients with paired protein and copy number data, cyclin D1 protein correlated with neither CCND1 copy number nor mRNA.
The same disconnect held in the cell line panel: amplifications did not reliably produce elevated protein, and baseline protein did not predict sensitivity to the drug. Neither did the biomarkers proposed elsewhere — MCM7, SH2B3, LZTR1, CRKL.
So the sequencing result that gets a patient into the trial does not tell you whether the pathway is actually running. That gap is worth taking seriously before enrolling anyone on genotype alone.
What escapes
CDK4/6 inhibition does what it should. Palbociclib suppressed p-Rb, FOXM1, PLK1 and cyclin A dose-dependently across the panel. But the effect was mostly cytostatic — a subpopulation persisted even at non-physiological drug concentrations, and after about three weeks of continuous exposure resistant colonies were growing and EdU incorporation had returned.
The mechanism turns out to be a released brake. CDK4/6 inhibition lowers DUSP4, the phosphatase that restrains ERK. Within 72 hours phospho-ERK rises and cyclin D1 with it — restoring the very protein the drug was meant to shut down. This happened regardless of BRAF or NRAS mutation status and regardless of copy number, so it is not a consequence of an existing MAPK driver.
Two manipulations close the loop. Overexpressing DUSP4 abolished the ERK activation and the cyclin D1 rebound, and made the drug work better. Silencing DUSP4 made it work worse. Silencing cyclin D1 restored the arrest that CDK4/6 inhibition alone could not hold.
Acquired resistance lands in the same place
Cell lines carried through weeks to months of chronic palbociclib converge on the same node: elevated phospho-ERK relative to their parental lines.
What is notable is what was not found. No new mutations in AKT1, FGFR2 or NRAS, and no HER2 mutations of the kind reported in endocrine-resistant breast cancer. Resistance here is pathway plasticity rather than a newly acquired driver — which is encouraging, because a rewired pathway can be targeted with drugs that already exist.
That makes ERK hyperactivation the shared exit for both the fast adaptive response and the slow acquired one. Hence "unified".
Blocking the exit
Adding the MEK inhibitor trametinib resensitized both therapy-naive and resistant cells, in culture, in 3D spheroids and in mice — including a patient-derived xenograft from a metastatic ALM and xenografts grown from the acquired-resistant lines. The combination beat either single agent.
Reverse-phase protein arrays on tumors at endpoint showed 46 proteins differentially expressed under the combination, resolving into three coherent programs: DNA repair capacity down (CENP-A, PARP, RPA32) alongside more double-strand breaks, cell cycle machinery down (cyclin B1, PLK1, E2F1), and apoptosis up (BAK, BID, BIM, cleaved caspase 7, with BCL2A1 reduced). The cells are not merely arrested. They are dying.
One practical detail: the resistant tumors kept their vulnerability to MEK inhibition only while CDK4/6 inhibitor pressure was maintained. Dropping the first drug when adding the second would give up the effect.
Caveats
The cell line panel is small, which the authors say plainly — these results generate hypotheses for a larger study more than they settle the question. RB1 deletion appeared in one resistant line, so a second escape route may exist alongside the MAPK one. And the correlative human data is retrospective, so the case against genotype-based stratification needs a prospective test.