Introduction

Pancreatic ductal adenocarcinoma (PDAC) is among the most lethal malignancies, with a 5-year survival rate of less than 12% [1]. Many patients succumb to the disease within the first 6 months, reflecting late diagnosis, metastasis, and therapy resistance [2, 3]. A distinctive feature of PDAC is the progression of early pancreatic intraepithelial neoplasia to an advanced unresectable disease [2, 3]. During this process, the histopathological phenotype of the tumor changes from low-grade, characterized by well-formed glandular epithelial structures, to high-grade, whereby epithelial cells are diffuse throughout the desmoplastic stroma [2, 3]. This process of dissemination has been linked to the activation of epithelial-mesenchymal transition (EMT), a cell-biological program that is essential for the early stages of embryogenesis and organ formation and is exploited by cancer cells [4]. EMT supports pathological events associated with the loss of epithelial behavior and the acquisition of mesenchymal features that enable invasion and metastasis [4].

In PDAC, patient analysis, in vitro assays, xenograft models, and genetically engineered mouse models (GEMMs) have suggested that a set of transcription factors (TF) enables EMT [5,6,7,8,9,10]. We now know that PDAC progression is not solely driven by the classic EMT-TFs Snail or Twist [6]; however, the loss of EMT-TF Zeb1 can result in the growth of low-grade tumors [8]. These EMT-TFs were originally considered part of a binary process, whereby cells transition into two distinct epithelial or mesenchymal states, the latter defined by the loss of expression of the epithelial protein E-cadherin and gain of expression of the mesenchymal protein vimentin [4]. This concept has now shifted to reflect a continuum of partial EMT states with overlap** transcriptional features found in multiple cancers, including PDAC [4, 11]. Cancer cells can also revert from partial EMT to epithelial states through mesenchymal-epithelial transition (MET) [9, 10]. This spectrum of heterogeneous EMT states may be associated with differences in epithelial plasticity and cell migration, thereby influencing tumor progression, metastasis, and response to treatment [11, 12]. The intrinsic mechanisms that drive the partial EMT or complete EMT remain unclear.

PDAC tumor cells reside within a dense desmoplastic stroma comprising cancer-associated fibroblasts (CAFs) and associated extracellular matrix (ECM), which constitutes more than 80% of the tumor mass [13]. The mechanism by which the stroma modulates the behavior of tumor cells, particularly the progression of EMT, remains unclear. A primary focus has been placed on the cytokine TGFβ1, which is potently secreted by CAFs [14], resulting in autocrine signaling and differentiation into myofibroblastic CAFs that promote the deposition of ECM proteins [15, 16]. Paracrine TGFβ1 signaling can also act on benign neoplastic epithelial cells, leading to cell cycle arrest or on neoplastic cells, to induce proliferation, motility, and EMT [11]. Similar to previous reports on patient-derived PDAC organoids [17, 44,45,46,47,48], we showed that solid murine organoids are often derived from high-grade tumors. Our proteomic analysis revealed that the morphology of PDAC organoids correlates with the continuum of EMT states, which has been previously reported for mammary organoids [49, 50]. We found that within organoids isolated from both primary and secondary tumors, glandular organoids were often in an epithelial state, whereas solid organoids often underwent partial EMT. Although solid morphology was a feature of acinar organoids in previously published work [51, 52], no enrichment for acinar signatures were detected in our solid organoid proteomics dataset, ruling out an acinar origin for these samples (Fig. S4A). We also found that partial EMT organoids give rise to different EMT spectra along a previously described trajectory of the epithelial, EMT-hybrid, and mesenchymal states [22]. Interestingly, following serial passaging and allograft generation, a subset of solid organoids with a partial EMT phenotype underwent MET, characterized by reversion to a glandular phenotype and re-expression of EpCAM. This phenomenon is not uncommon, and has previously been reported ex vivo, following the culture of partial EMT cell lines [11] and in vivo following the migration of cells from primary tumors to distant organs [9, 10].

Among the proteins identified within the partial EMT state in solid organoids is the S100 protein family, which consists of 25 members with a variety of intracellular and extracellular cellular functions, including calcium homeostasis, proliferation, and apoptosis [53]. S100 proteins have also been shown to interact with cytoskeletal proteins, thereby affecting cellular morphology and migration [43, 53]. We were intrigued by the S100 proteins, as they have previously been shown to be secreted by epithelial cells in response to tissue damage or inflammatory responses [28, 54]. We found that S100a14 was increased in the organoids derived from PDAC tumors compared to the normal organoids, which is consistent with the increased expression observed in murine cancer cells and patient cancer cell lines and was linked to poor overall survival [22, 38, 55]. S100a14 has been described as a mesenchymal marker because upregulation of the transcription factor Gli1 promotes EMT and increases s100a14 expression [56]. However, using murine models [55] and human cell lines [38], we and others have shown that s100a14 is associated with epithelial and not EMT states, as its expression is lower in glandular MOs, murine tumors, and patients with low-grade tumors.

We were also interested in the increased expression of s100a4, which is associated with partial EMT in solid organoids, because s100a4 has also been associated with poor survival in PDAC [57]. S100a4 has also been shown to play a role in EMT and metastasis and is highly expressed in murine PDAC mesenchymal tumor cells, patient cell lines, and PDAC patients with high-grade tumors, suggesting a clear association with EMT [22, 38, 56]. As s100a14 is also increased in tumor cells and is associated with poor prognosis, S100A4 may be a more useful biomarker for treatment of advanced disease since high S100A4 expression correlates with resistance to gemcitabine, the mainstay chemotherapy provided to patients with PDAC [58, 59], and EMT is known to alter the response to chemotherapy [60].

Secreted factors within the tumor microenvironment are known to influence EMT in PDAC [21]. We explored the contribution of TGFβ1 and found that recombinant TGFβ1 could transition a glandular organoid to a solid phenotype, which is associated with the induction of a classic EMT gene signature. As previously shown in PDAC patient-derived organoids [17], we observed that a branching phenotype could also be induced by recombinant TGFβ1 suggesting a transition to a complete EMT state. We were particularly intrigued by the observation that TGFβ1 could upregulate S100a4 consistent with other studies [7, 8]. It is thought that TGFβ1 may regulate S100a4 through the EMT-TF, Zeb1, as Zeb1 overexpression also resulted in an increase in S100A4 expression in cell lines, while S100A14 decreased, consistent with our observations [38].

Despite being an obvious therapeutic target to prevent the transition from early stage EMT to complete EMT, there is likely limited clinical benefit following targeting of TGFβ1 in cancer due to the high probability of on-target toxicities due to the complexities of TGF signaling and function within the immune system. For this reason, we explored the role of IL-6 family cytokines in EMT, which are readily targeted therapeutically, and which we and others have shown to be induced downstream of TGFβ1 signaling [18, 19]. However, we failed to detect changes in organoid morphology following stimulation with recombinant IL-6 family cytokines nor did we detect an increase in classic EMT markers, suggesting that this cytokine family alone does not promote EMT. Moreover, despite previous reports showing that IL-11 upregulates S100A4 and S100A14 in human PDAC cell lines [38], IL-11Rα1 expression was not detected in the organoid biobank. Other studies have also shown that the treatment of lung cancer cell lines with IL-6 alone did not induce EMT, although the presence of receptor components was not confirmed [61]. We also did not observe the induction of S100a4 following stimulation with IL-6 or LIF. However, we observed a significant induction of S100a14 suggesting that although IL-6 and LIF may contribute to tumor features, they do not directly contribute to EMT. Thus, the therapeutic inhibition of the IL-6 cytokine family alone is unlikely to prevent EMT. However, the activation of signal transducer and activator of transcription (STAT)−3, a pro-tumorigenic transcription factor downstream of the IL-6 family of cytokines, is significantly correlated with S100A4 expression in PDAC patient tissues [38] suggesting that cooperative induction of S100A4 is possible. Previous reports have also suggested that synergy between ZEB1 EMT-TF and IL-6 family cytokines results in the augmentation of S100A4 expression, which was not explored in this study [38].

Taken together, our observations suggest that both extracellular and cell-intrinsic complete EMT programs converge on S100A4 and may represent useful biomarkers for predicting disease progression and prognosis.

Materials and methods

Mouse strains

PdxCre; RosaYFP (CY) and PdxCre; KrasG12V; p53R172H; RosaYFP (CKPY) mice [5] were maintained on a C57BL/6 background and bred and maintained in a specific pathogen-free animal facility at WEHI. All experiments involving mice were approved by the WEHI Animal Ethics Committee (AEC approval #2019.015 and #2020.032). CKPY mice were aged between 12 and 27 weeks and were collected together with aged and sex-matched CY mice.

Generation of organoids

FACs sorted cells from murine tissue were resuspended in >90% v/v Matrigel (Corning) and grown in murine pancreatic organoid medium (MPOM) containing advanced DMEM/F12 (Gibco) containing 10 mmol/L HEPES (Gibco), 1X GlutaMAX (Gibco), 1x penicillin/streptomycin (Gibco), 5% v/v Rspo2-Fc conditioned medium (harvested from transiently transfected FreestyleTM-293F cells, Thermo Fisher), 5% v/v Noggin-conditioned medium (harvested from Noggin-expressing 239 cells obtained from Foundation Hubrecht Organoid Technology (HUB), Hubrecht Institute, Uterecht, The Netherlands), 10 mmol/L nicotinamide, 1% v/v B-27 supplement without vitamin A, 1 mmol/L N-acetyl-L-cysteine, 100 ng/mL rh FGF-10, 50 ng/mL rh EGF, 10 nmol/L rh [Leu15]-gastrin I, and 3 µmol/L prostaglandin E2. Following passaging, the organoids were cultured in MPOM with 10 µM Y-27632 (Sigma) and 5 µM GSK-3 inhibitor (Sigma) for the first three days, followed by culturing in normal MPOM for regular maintenance, as described previously [62].

Generation of allografts

Wild-type C57BL/6 mice were used to establish murine organoid-derived allografts. Wild-type C57BL/6 mice were bred and maintained in a specific pathogen-free animal facility at WEHI. All experiments involving mice were approved and monitored by the WEHI Animal Ethics Committee (AEC Approval #2017.033 and ##2020.032).

To generate allografts, one confluent well of organoids (approximately 50,000 cells) was resuspended in 100 µL of 50% v/v PBS/ 50% v/v Matrigel and subcutaneously injected into each flank of C57BL/6 mice (gender matched to the organoid). If no visible tumor was present 3 months after engraftment, engraftment was deemed unsuccessful and the mouse was euthanized.

Mass spectrometry-based proteomics

Organoids were lysed in RIPA buffer containing 100 mM NaCl, 10 mM Tris-HCl, 1% (v/v) glycerol, 50 mM NaF, 2 mM EDTA, 1% (v/v) Triton X-100, 1 mM Na3VO4, complete mini protease inhibitor cocktail, and complete mini phosphatase inhibitor, and 20 µg per replicate was prepared for proteomic analysis using the USP3 protocol previously described [63] with some minor modifications. Lysates were heated at 95 °C for 10 min in buffer containing 1% (v/v) SDS, 100 mM Tris (pH 8), 10 mM Tris (2-carboxyethyl) phosphine (TCEP), and 40 mM 2-chloracetamide. Magnetic PureCube Carboxy agarose beads (Cube Biotech) were added to all the samples along with acetonitrile (70% v/v final concentration) and incubated at room temperature for 20 min. Samples were placed on a 96-well magnetic rack, supernatants were discarded, and beads were washed twice with 70% ethanol and once with neat ACN. ACN was completely evaporated from the tubes using a CentriVap (Labconco) before the addition of digestion buffer (50 mm Tris) containing 0.8 µg Lys-C (Wako, 129–02541) and 0.8 µg Trypsin-gold (Promega, V5280). Enzymatic digestion was performed with agitation (400 rpm) for 1 h at 37 °C. Following digestion, the samples were transferred to pre-equilibrated C18 StageTips for sample clean-up. The eluates were lyophilized to dryness using CentriVap (Labconco) before being reconstituted in 60 µL of 0.1% FA/2% ACN ready for mass spectrometry analysis.

For DDA analysis, peptides (2 µL) were separated by reverse-phase chromatography on a C18 fused silica column packed into an emitter tip (IonOpticks) using a nano-flow HPLC (M-class, Waters). HPLC was coupled to a timsTOF Pro (Bruker) equipped with a CaptiveSpray source. Peptides were loaded directly onto the column at a constant flow rate of 400 nL/min with buffer A (99.9% v/v Milli-Q water, 0.1% v/v FA) and eluted with a 90-min linear gradient from 2 to 34% buffer B (99.9% v/v ACN, 0.1% v/v FA). The timsTOF Pro was operated in PASEF mode using Compass Hystar 5.1. Settings were as follows: Mass Range 100 to 1700 m/z, 1/K0 Start 0.6 V·s/cm2 End 1.6 V·s/cm2, Ramp time 110.1 ms, Lock Duty Cycle to 100%, Capillary Voltage 1600V, Dry Gas 3 l/min, Dry Temp 180 °C, PASEF settings: 10 MS/MS scans (total cycle time 1.27 sec), charge range 0-5, active exclusion for 0.4 min, Scheduling Target intensity 10000, Intensity threshold 2500, CID collision energy 42 eV.