A single-cell CRISPR screen of 318,032 A375 melanoma cells has identified 24 perturbations that confer resistance to vemurafenib and shown that most of them — 20 of 24 — push BRAF(V600E) melanoma cells toward a shared undifferentiated, drug-resistant state. The study, published in Nature on 15 April 2026 by Zihan Xu, Junyue Cao and colleagues at Rockefeller University, also demonstrates that triple inhibition of the transcription factors RREB1, SMAD3 and KLF5 cuts the growth advantage of resistant cells by a mean 3.1-fold under vemurafenib treatment.

The work rests on a new platform the authors call PerturbFate. It couples pooled CRISPR interference with multimodal single-cell profiling: chromatin accessibility, nascent and pre-existing RNA, and sgRNA identity, all captured from the same individual cell at a cost of less than $0.01 per cell. Existing Perturb-seq approaches typically read out one modality on commercial droplet platforms with limited throughput, the authors note, which constrains scale and obscures RNA kinetic dynamics.

What did the screen measure?

The team selected 143 candidate genes from published vemurafenib-resistance CRISPR screens, delivered a dual-sgRNA CRISPRi library at MOI 0.05 into dCas9–KRAB-expressing A375 cells, and treated cultures with 1 μM vemurafenib or DMSO. Vemurafenib at that concentration inhibited roughly 90% of A375 growth.

Single-cell profiling on day 6, plus parallel bulk screens on days 6 and 14, yielded several quantitative anchors:

• 218,678 cells with linked nascent/pre-existing transcriptomes plus chromatin accessibility; 99,354 with linked steady-state transcriptome plus accessibility
• 119 perturbations retained after quality filtering, averaging 438 tri-omic cells per perturbation under DMSO
• 52 perturbations significantly shifted trajectory position; 46 pushed cells toward undifferentiated states
• Correlation between perturbation-induced state shifts and day-14 growth advantage: Pearson r = 0.594 (P = 1.1 × 10⁻¹²)

Which regulatory programs drive the resistant state?

Reconstructed gene regulatory networks assign melanocytic identity to SOX10, in cooperation with MEF2C, while dedifferentiation runs through AP-1 family factors downstream of MAPK signaling and TEAD factors downstream of Hippo/YAP signaling. Knockdown of Hippo pathway YAP inhibitors — NF2, LATS2, PTPN14, TAOK1 — induced the strongest dedifferentiation. Mediator kinase module components (MED12, MED13, CCNC), SAGA complex genes and ubiquitin–proteasome genes produced the same phenotypic drift.

The vemurafenib-resistant state itself is marked by concurrent activation of FOSL1, TEAD1, SMAD2/3/4, TCF4, ZEB1, RREB1 and KLF5. YAP ChIP–seq data show enrichment of RREB1, KLF5, SMAD3 and FOSL1 at YAP binding peaks, linking these programs to Hippo signaling. In the TCGA skin cutaneous melanoma cohort, expression of these transcription factors correlated negatively with overall survival, and the regulons were activated in MAPK inhibitor–resistant patient tumors.

The authors caution that the combinatorial inhibitor experiments cannot exclude off-target or nonspecific additive fitness effects, so the convergence data should be read as supportive of the regulatory model rather than definitive proof of synergy.

How does the Mediator complex fit in?

Perturbing different Mediator modules produced divergent, condition-dependent outcomes. Kinase module knockdown (MED12, MED13, CCNC) activated inflammatory and Hippo/YAP signatures under DMSO, consistent with the kinase module restraining core Mediator activity at YAP-regulated enhancers. Tail module knockdown (MED15, MED24) caused the most severe global transcriptional impairment yet sustained YAP-dependent programs under vemurafenib.

MED12 knockdown stood apart: it uniquely promoted dedifferentiation under both conditions and closely mimicked SOX10 loss of function, suppressing MAPK inhibitor–induced SOX10 binding and target gene activation.

VEGFC emerged as the only factor consistently activated in dedifferentiated states under both treatments. VEGFC knockdown reduced AXL expression, suppressed growth under vemurafenib, and specifically cut the growth advantage of MED12, MED19, MED15 and MED24 perturbations.

Why it matters for dermatology

Resistance and relapse occur in nearly all BRAF-mutant melanoma patients within 6–12 months of vemurafenib, the first approved BRAF inhibitor. By showing that genetically diverse resistance mechanisms converge on one transcriptional program — the FOSL1/KLF5/RREB1/SMAD3 axis and the Mediator–YAP–VEGFC output — the study reframes combination targeting around that convergent network rather than individual mutant drivers. The authors position PerturbFate, with open protocols and code on GitHub and data under GEO accession GSE291147, as a generalizable framework for dissecting cell-fate regulation in other disease contexts, with the convergent transcription factor modules now the most concrete candidates for therapeutic testing in melanoma drug resistance.