Pancreatic ductal adenocarcinoma (PDAC) is often described as an immunologically “cold” tumor. Immune checkpoint blockade has shown limited efficacy, and defective antigen presentation is frequently invoked as a major barrier. However, while studying tumor cell subtypes in PDAC, we repeatedly encountered observations that did not fully fit this narrative.
The transcription factor GATA6 is known as a key regulator of the classical PDAC subtype. It maintains epithelial differentiation and suppresses epithelial-to-mesenchymal transition (EMT), and its loss is associated with the more aggressive basal-like phenotype. Initially, our interest in GATA6 was rooted in tumor plasticity and subtype biology. However, as we began integrating spatial and transcriptomic data, an unexpected pattern emerged: tumors with high GATA6 expression appeared to be immunologically distinct.
Multiplex spatial imaging revealed that GATA6⁺ tumor cells were more frequently found in proximity to CD8⁺ T cells. Transcriptomic analyses of microdissected epithelium and stroma showed enrichment of inflammatory and antigen presentation pathways in GATA6high tumors. Moreover, GATA6⁺ tumor cells expressed higher levels of MHCI. What began as a study of differentiation state increasingly pointed toward a connection between cell state identity and immune visibility.
This shift in perspective raised a new question: could therapeutic perturbation alter immune recognition in a GATA6-dependent manner?
MEK inhibition (MEKi), targeting the MAPK pathway downstream of KRAS, has been proposed as a strategy to enhance tumor immunogenicity. In our models, MEKi indeed increased MHCI expression, but strikingly, this effect was restricted to GATA6high tumor cells. At first glance, this seemed promising: MEK inhibition could potentially render classical PDAC tumors more visible to cytotoxic T cells.
However, a paradox soon became apparent. Although MEKi enhanced MHCI expression in GATA6⁺ cells, prolonged treatment simultaneously reduced GATA6 expression and promoted EMT-associated features. In spontaneous and orthotopic models, recurrent tumors displayed diminished epithelial markers. In patient-derived xenografts, MEKi increased MHCI but also drove a reduction in GATA6 alongside induction of mesenchymal markers.
The therapy was enhancing antigen presentation, but eroding the cell state that enabled this response.
To test whether GATA6 was required for MEKi-induced MHCI upregulation, we initially generated constitutive GATA6 knockout models. The results were perplexing. Instead of simply losing MHCI induction, these tumors exhibited strong compensatory responses, including markedly elevated interferon signaling and pSTAT1 activation. It became clear that developmental loss of GATA6 fundamentally rewired the tumor cells. The system had adapted, making it difficult to disentangle primary effects from compensatory mechanisms.
This was a pivotal moment in the project. We realized that if we wanted to understand the role of GATA6 in therapy response, we needed temporal control. We therefore engineered an auxin-inducible degron (AID) system to acutely degrade GATA6 protein only after tumors had formed. This approach allowed us to avoid developmental compensation and to interrogate GATA6 function specifically during therapeutic intervention.
The results were decisive. Acute degradation of GATA6 abrogated MEKi-induced MHCI upregulation, prevented CD8⁺ T cell infiltration, and eliminated MEKi-mediated tumor suppression in vivo. Importantly, these effects occurred without the confounding compensatory interferon surge seen in the knockout models. The AID system clarified what had previously been obscured: GATA6 is essential for the MEKi-induced anti-tumor cytotoxic response.
Nonetheless, the paradox remained: MEKi increased MHCI but reduced the GATA6⁺ population. If immune visibility depended on maintaining this classical cell state, could we stabilize it under therapeutic pressure?
Histone deacetylase inhibitors (HDACi) offered a potential solution. HDAC inhibition has been reported to promote epithelial gene expression and enhance antigen presentation. When combined with MEKi, class I HDAC inhibition restored GATA6 expression, counteracted MEKi-induced EMT features, and further augmented MHCI expression. In heterogeneous endogenous tumors, combination therapy preserved the GATA6⁺ tumor cell population while maintaining strong MHCI expression. Spatial analysis revealed that these GATA6⁺MHCI⁺ tumor cells were closely associated with granzyme B⁺ T cells and exhibited increased apoptosis.
Functionally, the anti-tumor effect of the combination was entirely dependent on CD8⁺ T cells. In their absence, tumor suppression was lost, even though MHCI expression remained elevated. Thus, stabilizing a differentiation state translated into enhanced immune-mediated tumor control.
Looking back, the central insight of this study is not simply that MEK and HDAC inhibition synergize. Rather, it is that immune recognition in PDAC is dynamically regulated by tumor cell state. Tumors are not uniformly invisible; instead, immune visibility depends on the stability of cell state programs under therapeutic stress. Treatments that inadvertently promote cell state transitions may undermine their own immunogenic benefits.
By preserving the GATA6⁺ classical phenotype during MEK inhibition, we were able to enhance antigen presentation and sustain cytotoxic T cell activity. These findings suggest that in highly plastic cancers such as PDAC, effective immunomodulation may require not only activating immune pathways but also stabilizing immune-permissive tumor cell states.
Many questions remain. How does GATA6 intersect with interferon signaling at the level of gene regulatory networks? Can selective epigenetic modulators achieve similar effects with reduced toxicity? And more broadly, do other tumor lineages harbor analogous immune-permissive states that are lost under therapy?
What began as an investigation into subtype biology ultimately became a study of how cell state identity shapes immune recognition. In PDAC, restoring immune visibility may not be about forcing tumors to become inflamed, but about preserving the cell states that already know how to be seen.