Blocking the road to de‐differentiation: HNF1A/KDM6A complex safeguards epithelial integrity in pancreatic cancer

Epithelial differentiation of normal and tumor cells is orchestrated by lineage‐determining transcriptional regulatory networks that enforce cell identity. Recent research by Kalisz et al (2020) in the EMBO Journal elucidates the molecular mechanisms by which a transcriptional differentiation program governed by HNF1A and KDM6A maintains acinar differentiation and the epithelial identity of pancreatic ductal adenocarcinoma (PDAC). Loss of function of either transcriptional regulator induces tumor progression to a poorly differentiated and highly aggressive PDAC subtype with a squamous transcriptome and poor prognosis.

New work uncovers a functional link between the transcription factor HNF1A and tumor suppressive histone demethylase KDM6A to maintain pancreatic epithelial cell identity.

Pancreatic ductal adenocarcinoma (PDAC) remains the most deadly human cancer type since decades despite maximal treatment. Especially undifferentiated, basal‐like tumors that have lost their epithelial identity display a particular poor prognosis and are highly resistant to the currently available standard‐of‐care chemotherapeutic treatment regimens (Aung et al, 2018; Chan‐Seng‐Yue et al, 2020). Therefore, it is of particular interest to understand the molecular basis that drives de‐differentiation and loss of epithelial cell determination in this specific PDAC subtype.

Molecular profiling of PDAC revealed predominate mutations of KRAS, TP53, SMAD4, and CDKN2A and a much lower prevalence of other genomic alterations, such as mutations of genes involved in the DNA‐damage response or epigenetic and transcriptional regulators (Collisson et al, 2019). Interestingly, the basal‐like undifferentiated PDAC subtype is associated with mutations in genes involved in chromatin modification, including the histone lysin N‐methyltransferases MLL2 and MLL3 as well as the lysine demethylase KDM6A (Collisson et al, 2019). In addition, it has been shown recently that oncogenic KRAS gene dosage variation and expression level have a strong impact on PDAC phenotypes and the de‐differentiation of pancreatic tumor cells (Mueller et al, 2018; Chan‐Seng‐Yue et al, 2020). However, it remains largely unclear why and how these tumors lose their epithelial identity and which genes and transcriptional networks are involved in the fate determination of pancreatic epithelial cells during oncogenesis.

Pancreatic ductal adenocarcinoma can arise from acinar cells undergoing acinar‐to‐ductal metaplasia (ADM), a process of plasticity and reprogramming of acinar cells to an embryonic ductal progenitor state, which further proceed to the carcinoma stage (Roy & Hebrok, 2015). Recent work demonstrates that this process and alterations of epithelial cell differentiation in established human PDAC are governed by cell‐type‐specific transcriptional regulatory networks (Diaferia et al, 2016; Martinelli et al, 2016; Milan et al, 2019). Transcription factors (TF) that control acinar fate and identity include FOXA family members, GATA6, NR5A2, and RBPJL. Importantly, deficiencies in these transcriptional networks have been also associated with PDAC development and linked with its differentiation status. For example, loss of GATA6 in PDAC has been previously implicated with undifferentiated tumors and expression of mesenchymal gene sets as well as repression of epithelial markers, such as CDH1 (Martinelli et al, 2016). The FOXA TFs, FOXA1 and FOXA2, play an essential role in pancreas development by activating the master regulator of pancreatic differentiation PDX1. Interestingly, FOXA1 and FOXA2 are jointly expressed in well‐differentiated epithelial tumor cells while they are dissociated in undifferentiated PDAC, where only FOXA2 is present, indicating a differentiation stage‐specific transcriptional reprogramming. Indeed, FOXA2 function depends on the availability of differentiation‐specific partner TFs regulating FOXA2 target site binding. In well‐differentiated tumor cells, FOXA2 interacts with HNF1B and AP‐1 maintaining epithelial differentiation, whereas the FOXA2 cistrome is regulated by HOXB8 in poorly differentiated PDAC cells (Milan et al, 2019). These data clearly show that rewiring of highly complex transcriptional regulatory circuits underlies and maintains the de‐differentiation and progression of human PDAC.

To understand the complex interplay of TFs and molecularly altered epigenetic regulators in pancreatic cell identity and plasticity, Kalisz et al (2020) performed elegant in vivo genetic gain and loss‐of‐function experiments to uncover a novel functional link between the TF HNF1A and the histone demethylase KDM6A, thereby the study interconnects for the first time the tumor‐suppressive function of KDM6A to a lineage‐specific TF and shows how both maintain epithelial cell identity in the pancreas.

HNF1A, which encodes for a homeodomain TF involved in a form of autosomal dominant diabetes, acts as a tumor suppressor in the pancreas in concert with KDM6A by maintaining acinar cell differentiation. KDM6A is a X chromosome‐encoded H3K27me3 demethylase, a component of the COMPASS (complex of proteins associated with set‐1)‐like transcriptional co‐regulatory complex, which confines active super‐enhancers (Andricovich et al, 2018). Loss of function of either Hnf1a or Kdm6a in combination with activation of oncogenic KRAS in mice leads to poorly differentiated sarcomatoid‐like PDAC tumors. Pancreas‐specific deletion of Hnf1a or Kdm6a is characterized by downregulation of acinar‐related genes, such as Ptf1a, Pla2g1b, Serpini2, and Ctrb1, and the upregulation of a mesenchymal gene expression program. Mechanistically, HNF1A recruits KDM6A to genomic binding sites and directly mediates the expression of target genes important for acinar identity, such as FOXA3 and DEPTOR (Fig 1). This remodels the acinar enhancer landscape, activates an acinar differentiation program, and suppresses oncogenesis and EMT‐related genes.

Figure 1. HNF1A/KDM6A transcriptional complex regulates epithelial acinar cell differentiation.

HNF1A interacts with KDM6A to form a transcriptional complex which controls acinar cell differentiation. In cooperation with oncogenic KRAS signaling, loss of HNF1A and/or KDM6A induces upregulation of mesenchymal gene sets and leads to a sarcomatoid differentiated PDAC.

Somatic mutations and genomic deletions of KDM6A as well as decreased mRNA expression levels of HNF1A and KDM6A are associated with a sarcomatoid gene expression program and enriched in the non‐classical de‐differentiated human PDAC subtype (Fig 1). These findings indicate that disruption of HNF1A‐ and KDM6A‐dependent programs not only promotes the formation of PDAC, but also defines PDAC subtypes. However, how defective HNF1A‐ or KDM6A‐dependent gene regulation support the development of an undifferentiated high‐grade PDAC cellular phenotype remains unclear and further studies are clearly needed to uncover the underlying molecular mechanisms.

Investigating transcriptional networks maintaining epithelial and mesenchymal phenotypes, as well as their plastic transitions, has important clinical implications. For example, loss of KDM6A in PDAC activates super‐enhancers that regulate ΔNp63 and MYC, which are markers for non‐classical PDAC differentiation (Fig 1). Interestingly, these tumors were selectively sensitive to bromodomain and extra‐terminal (BET) inhibition resulting in blocked sarcomatoid differentiation and downregulation of MYC (Andricovich et al, 2018). Thus, targeting epigenetic reprogramming and reverting differentiation states, for example through BET inhibition, provide a potential therapeutic option for non‐classical PDAC tumors. In future, it will be also important to discover vulnerabilities and design therapies for HNF1A/KDM6A‐proficient PDAC subtypes. The pioneering work of Kalisz et al (2020) opens new horizons for such attempts by providing the basis for a deeper understanding of the evolution of distinct PDAC subtypes and their associated therapeutic vulnerabilities.

The EMBO Journal (2020) 39: e104759