IntroductionCellular plasticity and epigenetic reprogramming are prominent features of many human cancers1. In epithelial cancers (also known as carcinomas), tumor cells can transition between glandular, basal, neuroendocrine, and mesenchymal cell states, each conferring distinct fitness advantages during tumor evolution and metastasis2. Relevant to this study, basal lineage features, characterized by transcriptional and/or histological resemblance to cells of stratified squamous epithelium (e.g., epidermis and esophagus)3,4,5,6, emerge in several human adenocarcinomas, including those of the bladder, breast, and pancreas. The presence of basal identity in these tumors is associated with inferior clinical outcomes and differential responses to both conventional chemotherapy and targeted therapies7,8,9,10,11,12,13. Functional experiments have further demonstrated that the activation of basal identity programs directly promotes more aggressive disease characteristics14,15,16. In addition, acquisition of basal identity can be a mechanism of evading oncogene-targeted therapies, such as inhibitors of KRAS and EGFR in lung adenocarcinoma17,18. For these reasons, the biochemical mechanisms that specify basal identity are of high interest to the cancer research field.Pancreatic ductal adenocarcinoma (PDAC) is an aggressive and heterogeneous malignancy that exemplifies the clinical significance of lineage identity. PDAC tumors exhibit epigenetic and transcriptomic plasticity, giving rise to distinct cellular states that correlate with histopathology and clinical prognosis13,19. Two prominent cellular identities in PDAC, termed the ‘classical’ and ‘basal-like’ states, have been well-validated and tend to exist in a mutually exclusive manner10,11,12,13,20. The classical state features high expression of endodermal transcription factors (TFs) (e.g., HNF4α) and lineage markers (e.g., MUC1 and MYO1A), whereas the basal-like state (also known as squamous) features high expression of the TF ΔNp63 and basal lineage markers (e.g., KRT5 and KRT17)21,22,23. Across several independent patient cohorts, the expression of basal lineage markers correlates with poor overall survival, associated with increased metastatic potential and resistance to cytotoxic chemotherapy13,24. In accord with these clinical correlations, laboratory studies have shown that activation of basal identity (e.g., by ectopic ΔNp63 expression) drives more aggressive PDAC tumors22,23,25.KLF5 is a zinc finger-containing TF oncoprotein that regulates epithelial lineage identity in the aerodigestive tract and in the epidermis26,27,28,29. While a strong KLF5 requirement exists to complete embryogenesis and for wound healing responses in adult tissues29,30,31, conditional knockout studies demonstrate that KLF5 is dispensable for homeostasis of epithelial cells in the adult lung, intestine, and pancreas32,33,34. In contrast, KLF5 is a critical dependency in multiple carcinomas, as shown in both human cancer cell lines and genetic mouse models35,36,37. In mouse models of PDAC, Klf5 is transcriptionally upregulated during the acinar-to-ductal cell fate transition triggered by inflammation and by mutant KRAS34,38. In addition, KLF5 has been found to maintain classical identity in human PDAC cell lines, but its role in basal-like PDAC remains undefined39.RUVBL1 and RUVBL2 are evolutionarily conserved AAA+ ATPases that form obligate hetero-hexamers, which couple ATP hydrolysis to the regulation of protein-protein interactions and protein folding40,41. For example, they function as core scaffolding subunits for the assembly of multiprotein complexes, including the R2TP protein chaperone complex and the INO80 family of chromatin remodeling complexes (INO80, SRCAP, and TIP60/p400)42,43,44,45. In addition, emerging evidence suggests that RUVBL1/2 may also participate in transcriptional regulation through interactions with RNA Polymerase II and TFs such as MYC46,47. AAA+ ATPase activity is required for several RUVBL1/2 functions (e.g., protein chaperone activity); however, evidence also exists for ATPase-independent functions of RUVBL1/243,44,48.Although RUVBL1/2 are essential proteins in all eukaryotic cells, independent laboratories have developed small molecule inhibitors of RUVBL1/2 ATPase activity, which have anti-tumor activity in pre-clinical cancer models at well-tolerated doses48,49. This therapeutic effect has been observed in diverse forms of cancer, including adenocarcinomas of the lung and pancreas, Ewing sarcoma, and hematopoietic malignancies47,48,49,50. These compounds engage an allosteric pocket at the RUVBL1/RUVBL2 interface, trapping the complex in a rigid, ATP-bound conformation48,49,51. This state constrains RUVBL1/2 flexibility, implicating conformational dynamics linked to ATP hydrolysis as critical for their function51. The observed therapeutic effects of RUVBL1/2 inhibitors in cancer-bearing mice raise the possibility that this AAA+ ATPase activity regulates oncoprotein function. However, the precise mechanisms by which RUVBL1/2 supports oncogenic pathways remain poorly defined47,48.Here, we identify KLF5 as a context-dependent master regulator of cell identity in PDAC, capable of activating both classical and basal transcriptional programs via a flexible cistrome. Using an integrated biochemical and genetic screening strategy, we reveal the RUVBL1/2 complex as a direct, ATPase-dependent coactivator of KLF5. Our findings suggest that RUVBL1/2 carries out this function independently of both the R2TP and INO80 family of protein complexes by binding to a disordered segment of KLF5, which enables recruitment to lineage-specific enhancers. Pharmacologic inhibition of RUVBL1/2 ATPase activity disrupts this interaction, impairs KLF5 function, and suppresses tumor cell proliferation. We also present evidence that RUVBL1/2 can function more broadly to support oncogenic TFs important in other cancer contexts. Taken together, our findings reveal a dual regulatory role for KLF5 in PDAC lineage specification and establish AAA+ ATPase-driven conformation dynamics as a previously unrecognized transcriptional coactivator mechanism with therapeutic potential.ResultsKLF5 is highly expressed in the classical and basal-like subtypes of human PDACWe previously carried out genetic screens in search of novel regulators of basal identity in three independent models of basal-like PDAC (T3M-4, BxPC-3, and KLM-1), using KRT5 staining as a marker52. In addition to validating ΔNp63 as a master regulator of this cell state, these screens nominated KLF5 as a requirement for basal identity in each of the three PDAC models (Fig. 1A, Supplementary Data 1)22,23. This finding was unexpected, as previous studies demonstrate that KLF5 activates classical PDAC lineage identity, but suggest that KLF5 is absent in ‘high-grade’ PDAC tumors and is downregulated during epithelial-to-mesenchymal transitions (EMT)39,53. To investigate a possible role for KLF5 in basal-like PDAC, we re-analyzed several bulk RNA-seq datasets of human PDAC samples10,11,12,13,20,54. We found that KLF5 was elevated in PDAC tumors and metastases relative to normal pancreas tissue, with similar levels in both classical and basal-like tumor subsets (Figs. 1B, C, S1A–E, Supplementary Data 2). Re-analysis of single-cell RNA-sequencing data from a genetically engineered PDAC mouse model55 further confirmed Klf5 upregulation during disease initiation, and demonstrated that Klf5 expression is sustained throughout disease progression (Figs. 1D, S1F)46,47. Moreover, we re-analyzed single-nucleus RNA sequencing (snRNA-seq) data from 224,988 PDAC cells isolated from 43 resected primary human tumors to evaluate the intra-tumoral heterogeneity of KLF5 expression (Figs. 1E, S1G)56. Classical and basal identities are present as distinct cell states across this set of tumors, which express HNF4A or TP63, respectively, at high levels (Figs. 1E, F, S1H). Importantly, KLF5 was expressed at comparable levels in both subtypes of PDAC cells and was increased relative to normal pancreatic epithelial cells (Fig. 1F). In human PDAC cell lines, we found that KLF5 was also highly expressed, particularly in models with strong classical or basal identity (Figs. 1G, S1I). Unlike primary human PDAC tumors, we found that many human PDAC cell lines weakly express both the classical and basal signatures, with a subset of such models expressing KLF5 at low levels (e.g., PANC-1 cells) (Figs. 1G, S1I, Supplementary Data 3). Taken together, these observations indicate that KLF5 is expressed at high levels in both the classical and basal-like subtypes of human PDAC tumors and cell lines, which prompted us to investigate further the function of KLF5 in these two lineage contexts.Fig. 1: KLF5 is highly expressed in the classical and basal-like subtypes of human PDAC.A Genome-wide KRT5 reporter CRISPRi screens in three basal-like PDAC cell lines (KLM-1, T3M-4, and BxPC-3)52. Beta scores and significance were calculated using MAGeCK85 (maximum likelihood estimation). Negative beta scores indicate enrichment in the KRT5low population. Average beta scores of 1599 transcription factors are shown, ordered alphabetically along the x-axis. TFs with a p-value −0.125, basal_score