Normal pancreatic mesenchyme expresses stem cell factor (KITL), which is lost in the cancer-associated fibroblast state during tumorigenesis and restrains pancreatic inflammation, epithelial identity, and tumor growth.
Abstract
Components of normal tissue architecture serve as barriers to tumor progression. Inflammatory and wound-healing programs are requisite features of solid tumorigenesis, wherein alterations to immune and nonimmune stromal elements enable loss of homeostasis during tumor onset. The precise mechanisms by which normal stromal cell states limit tissue plasticity and tumorigenesis, and which are lost during tumor progression, remain largely unknown. In this study, we show that healthy pancreatic mesenchyme expresses the paracrine signaling molecule KITL, also known as stem cell factor, and identify the loss of stromal KITL during tumorigenesis as tumor promoting. Genetic inhibition of mesenchymal KITL in the contexts of health, injury, and cancer together indicates a role for KITL signaling in the maintenance of pancreas tissue architecture, such that the loss of the stromal KITL pool increased tumor growth and reduced survival of tumor-bearing mice. Together, these findings implicate the loss of mesenchymal KITL as a mechanism for establishing a tumor-permissive microenvironment.
Significance:
By analyzing transcriptional programs in healthy and tumor-associated pancreatic mesenchyme, we find that a subpopulation of mesenchymal cells in healthy pancreas tissue expresses the paracrine signaling factor KITL. The loss of mesenchymal KITL is an accompanying and permissive feature of pancreas tumor evolution, with potential implications for cancer interception.
Introduction
Although evidence that normal tissue components can restrain tumor progression dates back to the 1960s (1), specific tissue-level barriers to plasticity and tumor outgrowth remain largely unknown. Mechanisms maintaining tissue homeostasis and limiting tumorigenesis include epithelial–epithelial interactions, such as regenerative or competitive epithelial functions (2–4); epithelial–immune interactions, wherein innate (5, 6) or adaptive (7, 8) immune cells clear mutant cells or early preinvasive lesions; and epithelial–mesenchymal interactions, with evidence that mesenchymal elements such as normal tissue fibroblasts can restrain growth of transformed epithelial cells (9–11). These mechanisms coexist with, and likely interact functionally with, epithelial cell–intrinsic tumor suppressor gene products, together creating genetic, cellular, and tissue-level checks on cancer development. Although epithelial cell–intrinsic mechanisms of tumor suppression have been studied extensively, and we have advanced considerably in our understanding of antitumor functions of the immune systems, mechanisms underlying the tumor-restraining potential of normal mesenchyme largely have not been identified.
Fibro-inflammatory reactions create tissue contexts permissive to tumor progression (12, 13). Local or systemic cues, including paracrine signaling from transformed epithelial cells or diverse sources of tissue damage, cause alterations to resident mesenchymal cells such as transition from quiescent fibroblasts to activated myofibroblasts and changes to or accumulation of immune cells. This wound-healing reaction helps promote plasticity in the epithelial compartment and overcome intrinsic barriers to tumor formation and growth (14). Consistent with this notion, although normal primary fibroblasts can suppress hyperplastic growth of mammary epithelial cells in vivo, this outgrowth is supported by activated, myofibroblastic stroma (15, 16). Although the causal link between inflammation and cancer has been appreciated for some time (17), recent studies of patient tissues have begun to identify specific mechanisms by which inflammatory insults promote cancer development. For example, environmental pollutants result in an accumulation of IL-1β–producing macrophages in the lungs, and this inflammatory signaling drives plasticity in the lung epithelium to promote tumorigenesis (18). Further study of the specific signals engaged by healthy or inflamed tissue components to restrain or promote tumorigenesis, respectively, may point to new avenues for early cancer intervention.
The recent discovery that normal, adult human pancreas tissue harbors up to hundreds of KRAS-mutant preinvasive lesions impels the field to understand intracellular and heterocellular mechanisms restraining neoplastic progression in the pancreas (19, 20). To assess a role for mesenchymal cell state alterations in the transition from a homeostatic to tumor-permissive tissue context, we performed transcriptional profiling of healthy and cancer-associated pancreatic mesenchyme using an established fate-mapping mouse model (21). These experiments have focused on pancreatic stellate cells (PSC), tissue-resident mesenchymal cells that serve as the cells of origin for a subpopulation of pancreatic ductal adenocarcinoma (PDAC) cancer-associated fibroblasts (CAF). We found that this mesenchymal lineage in normal human and murine pancreas tissue expresses KITL—this lineage has lipid storage capacity and co-expresses the leptin receptor (LEPR), with parallels to LEPR-positive mesenchyme previously implicated in tissue homeostasis in the bone marrow (22) and brown adipose tissue (BAT; ref. 23). Mesenchymal KITL expression is lost during tumor evolution and acquisition of a CAF stromal phenotype, with functional significance for tissue state and tumorigenic potential.
Results
To assess stromal evolution during stepwise tumorigenesis, we applied a previously established fate-mapping approach (21) to analyze the contributions of PSCs to the stroma of normal pancreas tissue, pancreatic intraepithelial neoplasia (PanIN), and invasive PDAC. To this end, we generated a dual recombinase genetically engineered mouse model of the genotype KrasFSF-G12D/+;Trp53FRT/+;Pdx1-FlpO;Rosa26mTmG/+;Fabp4-Cre (Fig. 1A) and assessed the presence of GFP+ stroma, indicating a lipid-storing origin. Although GFP+ PSCs were found in normal pancreas tissue as expected, very few were positive for PDPN, a cell surface marker upregulated upon fibroblast activation in PDAC. We found GFP+PDPN+ and GFP+ α-SMA+ cells associated with low-grade PanIN lesions and invasive cancer in this model (Fig. 1B and C), with a significant increase in PSC-derived fibroblastic cells in the context of PDAC compared with preinvasive lesions. In normal pancreas tissue and in tumors, PSCs or PSC-derived CAFs had a spatial distribution similar to the reported tissue distribution of stellate cells in the liver, the other tissue in the body where these mesenchymal cells reside. Hepatic stellate cells (HSC) are found in perivascular regions in close proximity to endothelial cells and adjacent to neighboring parenchymal cells (24). We found PSCs in normal pancreas tissue similarly to localize in perivascular regions and in the tissue parenchyma spatially poised for cell–cell communication with epithelial cells (Fig. 1D). This spatial distribution was conserved upon differentiation to a CAF phenotype, as GFP+ CAFs were found both immediately adjacent to and distant from endothelial cells in the genetically engineered PDAC model (Fig. 1E). Similar results were observed in an orthotopic model wherein PDAC cells derived from the KrasLSL-G12D/+;Trp53LSL-R172H/+;Pdx1-Cre (KPC) autochthonous model were implanted into syngeneic Rosa26mTmG/+;Fabp4-Cre hosts (Fig. 1F). These observations indicate that PSCs contribute to the stromal microenvironment throughout pancreatic tumorigenesis and are spatially distributed to engage in direct cell–cell contact with both endothelial cells and epithelial cells in healthy and cancerous pancreas tissue.
Figure 1.
PSCs contribute to the stromal microenvironment throughout tumorigenesis. A, Genetic schema of the KrasFSF-G12D/+;Trp53FRT/+;Pdx1-FlpO;Rosa26mTmG/+;Fabp4-Cre murine model. B, Representative images (above) and quantification (below, n = 3) of IHC staining for GFP (green) and podoplanin (PDPN, magenta) among normal pancreas, PanIN lesions, and PDAC lesions. Scale bar, 10 μm. C, Representative images (above) and quantification (below, n = 3) of IHC staining for GFP (green) and α-SMA (yellow) on normal pancreas, PanIN lesions, and PDAC lesions. Scale bar, 20 μm. D, Representative images of IHC staining for GFP (green) and CD31 (magenta) within normal pancreas (n = 5). Scale bar, 10 μm. E, Representative images of IHC staining for GFP (green) and CD31 (magenta) within genetically engineered mouse model (GEMM) pancreata (n = 3). Scale bar, 20 μm. F, Representative images of IHC staining for GFP (green) and CD31 (magenta) within pancreata of KPC-derived orthotopically implanted PDAC in Rosa26mTmG/+;Fabp4-Cre mice (n = 3). Scale bar, 10 μm. For image quantification, at least four fields per tissue were analyzed and averaged. All data are represented as mean ± SEM. The one-way ANOVA test was used: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, not significant.
We next assessed alterations in expression of cell surface ligands or receptors in this mesenchymal lineage during pancreatic tumorigenesis. We reasoned that paracrine signaling factors important for normal tissue architecture may be lost from the mesenchyme during the transition from normal tissue homeostasis to cancer. To identify candidate paracrine factors associated with normal mesenchymal function whose loss may be tumor permissive, we analyzed the transcriptional profiles of PSCs and PSC-derived CAFs. To this end, we sorted GFP+ PSCs and GFP+PDPN+ CAFs from healthy pancreas tissue and PDAC, respectively, and performed single-cell RNA sequencing (scRNA-seq) to assess gene expression differences in this cellular compartment within and across tissue states. As expected, these cells in normal pancreas and PDAC were pervasively positive for mesenchymal markers Vim and Eng and almost all positive for pan-tissue fibroblast markers Pi16 or Col15a1 and universal fibroblast markers Col4a1 and Hspg2 (Fig. 2A; ref. 25). Although not all cells expressed one of these two universal fibroblast markers, we note that PSCs are not strictly fibroblasts albeit fibroblast-like. Although PSCs and PSC-derived CAFs are partially perivascular as described above, these cells lacked expression of classical pericyte markers such as Cspg4 (encoding NG2) and Rgs5 (Supplementary Fig. S1A and S1B) and, despite detectable transcript, expressed low-to-absent CD105 (encoded by Eng) at the protein level (Supplementary Fig. S1C). CD105 expression at the protein level was rather low and sparse with the models and reagents used here. We also noted the expression of established markers of perivascular reticular cells (26, 27), specialized fibroblastic cells of lymphoid tissues, including Cd34, Cd29/Itgb1, and Ly6a (Fig. 2A). A subpopulation of these cells in normal tissue also expressed adhesion molecules associated with endothelial cell identity, including Pecam1, which was also observed sporadically among HSCs in normal liver upon analysis of a previously published scRNA-seq dataset (Fig. 2A; Supplementary Fig. S1D). However, these cells did not express other endothelial markers such as Vwf (Fig. 2A; Supplementary Fig. S1D) and seem to express far lower levels of both CD31 and CD105 (encoded by Pecam1 and Eng, respectively) than bona fide endothelial cells, given the limited overlap of these markers with GFP at the protein level by IHC (Fig. 1C; Supplementary Fig. S1C). That said, these results may indicate a shared precursor for stellate cells and endothelial cells late in development. Interestingly, the subpopulation in normal pancreas tissue lacking universal fibroblast markers expressed Vim, Eng, and genes generally associated with a macrophage identity, such as Csf1r and Adgre1 (encoding F4/80; Supplementary Fig. S1E). However, when we stained for GFP and macrophages in pancreas tissues, we detected no overlap (Supplementary Fig. S1F), suggesting that this subpopulation of cells in normal pancreas tissue may be fibrocyte-like or otherwise express some macrophage-associated genes without assuming a macrophage identity. In the context of cancer, PSCs gained expression of immune-modulatory cytokines such as il6 and Il33 as expected for CAFs (28) and pervasively expressed extracellular matrix (ECM) components such as Col1a1 and Col1a2 (Fig. 2B). These results validate activation of PSCs to a CAF phenotype in PDAC.
Figure 2.
Mesenchymal KITL loss within PSCs accompanies pancreatic tumorigenesis. A, Uniform manifold approximation and projection (UMAP) visualization of expression of the indicated genes in normal PSCs and PSC-derived CAFs from scRNA-seq data (n = 2 replicates pooled from n = 5 mice per arm). B, UMAP visualization of Il6, Il33, Col1a1, and Col1a2 gene expression from normal PSCs and PSC-derived CAFs scRNA-seq dataset (n = 2 replicates pooled from n = 5 mice per arm). C, UMAP visualization of Kitl transcript expression in normal PSCs and PSC-derived CAFs scRNA-seq dataset (n = 2 replicates pooled from n = 5 mice per arm). D, Left, UMAP illustrating the cellular landscape of normal pancreatic (blue) and PDAC (red) mesenchymal cells, comprising 5,337 normal and 2,861 tumor cells. Harmony was used to integrate the datasets and correct for batch effects. Right, Monocle 3 trajectory analysis was used to depict expression of the Kitl gene along the inferred pseudotime trajectory. Cells are colored based on Kitl expression levels, with values ranging from low (black) to high (yellow), revealing the spatial and temporal expression patterns of Kitl (n = 2 replicates pooled from n = 5 mice per arm). E, Left, qRT-PCR analysis of Kitl in quiescent (day 0) and culture-activated (day 7, on plastic) primary PSCs. Right, Quantikine ELISA KITL measurement of supernatant collected from primary PSCs in preactivated (day 2) and activated (day 10) states after 48 hours of incubation with media change on day 8 to harvest for day 10 sample. Immortalized ImPSC-1 included as the reference point. Data represents biological triplicate. F, qRT-PCR analysis of Kitl and Pdgfrα in indicated cell populations. Data show representative technical triplicates. G, Representative brightfield images of primary human PSCs and other mesenchyme. Scale bar, 50 μm. H, qRT-PCR analysis of indicated genes in primary human PSCs and other mesenchyme, shown for three independent patient samples. I, qRT-PCR analysis of Kitl in quiescent (day 0) and conditioned media–activated (day 14) primary human PSCs. Data represent biological triplicate. J, Representative images (left) and quantification (right) of RNA FISH staining for Fabp4 (green) and Kitl (red) in murine normal pancreas (n = 3). Scale bar, 10 μm. Below, Representative RNAscope staining of GFP (green) protein and Kitl (red) mRNA in PDAC from the genetically engineered mouse model (GEMM) depicted in Fig. 1A (n = 3). Scale bar, 10 μm. K, Representative images and quantification of RNA FISH staining for VIM (green) and KITL (red) in human PDAC tissues between benign adjacent and PDAC regions (n = 3). Scale bar, 20 μm. L, Representative images and quantification of RNAscope staining for Kitl (red) mRNA expression, GFP (green), and CD31 (magenta) in murine normal pancreas from Rosa26mTmG/+;Fabp4-Cre mice (n = 3). Scale bar, 20 μm. M, Representative images and quantification of RNAscope staining for GFP (green) protein and Kitl (red) mRNA in GEMM low-grade PanIN (n = 3). Scale bar, 20 μm. N, Representative image of IHC staining for phospho-c-Kit (Y721, red) in normal murine pancreas (n = 3). Scale bar, 10 μm. O, Representative images of IHC staining for vimentin (VIM, green) and Phospho-c-Kit (Y703, red) in benign adjacent and human PDAC tissues (n = 3) Scale bar, 10 μm. For image quantification, at least four fields per tissue were analyzed and averaged. An unpaired t test was performed for comparisons between two groups. For comparisons more than two groups, significance was determined by the ordinary one-way ANOVA. Data are represented as mean ± SEM;*, P < 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; Benign Adj., benign adjacent; CM, conditioned media; ns, not significant; other mes., other mesenchyme.
We next focused on paracrine signaling factors expressed in healthy pancreatic mesenchyme and that were lost in PDAC, which may represent barriers to tumor progression. We noted the expression of Kitl (also known as stem cell factor or SCF) in normal pancreas tissue but lost in CAFs (Fig. 2C), supported by pseudotime analysis (Fig. 2D). KITL expression has not previously been reported in normal pancreas tissue and was of interest to us in light of the significance of KITL-positive mesenchyme in the perivascular niche of the bone marrow, wherein stromal KITL is crucial for normal tissue structure and function (22). Furthermore, HSCs in the developing liver are critical sources of KITL to support the hematopoietic stem cell niche (29), providing precedent for functionally significant KITL production by stellate cells. KITL-positive mesenchyme in the bone marrow expresses the LEPR, and we detected low levels of Lepr expression among normal PSCs by using scRNA-seq (Supplementary Fig. S2A). We validated these findings by isolating primary PSCs from healthy pancreas tissue and activating them to a CAF phenotype in culture; these cells expressed Kitl and Lepr in their normal tissue state but progressively lost expression of both factors upon activation to a CAF-like state upon exposure to a stiff growth substrate with or without recombinant TGF-β (Fig. 2E; Supplementary Fig. S2B–S2D). We next validated the expression of Kitl in intact murine pancreas tissue. To further understand Kitl expression patterns in the pancreatic mesenchyme, we used standard density centrifugation to isolate PSCs from the gradient interface (30) and also isolated PDGFRα+ pancreatic fibroblasts from the lower layers of the gradient (31). We found significantly higher Kitl expression in PSCs than in conventional tissue fibroblasts (Fig. 2F), which may present to the substantial contribution of these PDGFRα-expressing fibroblasts to early pancreatic neoplasia. To perform a similar comparison in human pancreas tissue, we collected benign pancreas from surgical resection specimens and performed density centrifugation to isolate PSCs from the gradient interface (32), then performed negative selection for EpCAM, CD45, and CD31 on the remaining cells to collect mesenchymal cells from the lower layers of the gradient. We analyzed gene expression across these freshly isolated fractions, plating a small sampling on coverslips to image and validate, as well as finding that human PSCs expressed higher levels of KITLG than non-PSC components of the pancreatic mesenchyme, although we note that the degree of difference varied among the patients (Fig. 2G and H). To further validate our findings from mice, we cultured human PSCs from these benign resection specimens and activated them to a CAF-like phenotype by exposure to a stiff growth substrate with or without PDAC cell conditioned media. Human PSC activation featured a significant reduction in KITLG expression (Fig. 2I), further supporting the notion that normal pancreatic mesenchyme expresses Kitl/KITLG but this factor is lost as these cells transition to a CAF state in the tumor microenvironment.
We next assessed the patterns of Kitl expression in intact pancreas tissues. By RNA ISH (due to lack of specific antibodies, using branched cDNA hybridization), we detected Kitl expression in mesenchymal cells of normal pancreas tissue, which share markers with PSCs, although Kitl expression was lost among CAFs in PDAC (Fig. 2J). We also combined RNA ISH for Kitl (here using RNAscope, compatible with protein co-staining) with IHC for GFP on pancreas tissue from Rosa26mTmG/+;Fabp4-Cre mice and confirmed Kitl expression in fate-mapped PSCs (Fig. 2J). Specificity of our Kitl probe was confirmed by reduction in mesenchymal Kitl signal in pancreas tissues from Kitlflox/flox;Fabp4-Cre mice (Supplementary Fig. S2E). We extended these analyses to human pancreas tissue, and performed RNA ISH for KITL and mesenchymal marker VIM. Although benign human pancreas harbored KITL-positive mesenchyme, CAFs within human PDAC showed reduced KITL expression, consistent with observations in mice (Fig. 2K), although we note that KITL/VIM frequency was highly variable in benign adjacent regions, and this difference did not reach statistical significance. Human pancreas tissue showed minimal expression of KITL in CD45-positive leukocytes (Supplementary Fig. S2F). As perivascular mesenchyme is a critical source of KITL in other tissues (22, 29), we assessed the spatial distribution of mesenchymal Kitl expression by combining Kitl RNA ISH with IHC for CD31 and GFP on pancreas tissues from Rosa26mTmG/+;Fabp4-Cre mice. We found that PSCs express Kitl in both perivascular regions and when not adjacent to endothelial cells (Fig. 2L), suggesting that KITL from PSCs is poised to signal to multiple neighboring cell types. To assess the stage of pancreatic tumorigenesis at which mesenchymal Kitl expression is lost, we combined RNA ISH for Kitl and IHC for GFP (to indicate PSCs and PSC-derived CAFs) on tissues from KrasFSF-G12D/+;Trp53FRT/+;Pdx1-FlpO;Rosa26mTmG/+;Fabp4-Cre mice and noted retention of Kitl expression among GFP-positive stromal cells associated with low-grade PanIN lesions identified by a pathologist (Fig. 2M), suggesting that the loss of stromal Kitl accompanies late stages of pancreatic tumorigenesis. The expression of Kitl by some GFP-negative cells was noted within these areas of low-grade PanIN as well. Consistent with production of KITL protein in healthy pancreas, small numbers of phospho-c-KIT–positive cells were identified in murine and human benign pancreas (Fig. 2N and O), only detectable on fresh-frozen and not formalin-fixed tissues (see “Methods” section). Together, these analyses revealed the expression of Kitl by a lineage of healthy pancreatic mesenchyme in mice and humans, which is lost upon transition to a CAF phenotype in invasive cancer.
We next addressed the functional significance of KITL in pancreatic mesenchyme and assessed the consequence of stromal KITL loss for tissue homeostasis. First, we questioned the cell-intrinsic impact of KITL signaling on pancreatic mesenchymal cells. To address this, we generated loss- and gain-of-function systems in the cell culture by knocking down or overexpressing Kitl in immortalized PSCs (33) using short hairpin RNA or by the introduction of the Kitl open reading frame, respectively (Supplementary Fig. S3A). PSCs in culture express low but detectable levels of Kitl (Fig. 2E), so we reasoned that gene expression changes observed with Kitl overexpression would reflect downstream transcriptional programs in healthy mesenchyme, whereas Kitl knockdown would reflect consequences of gene expression changes upon transition to a CAF state. In culture, PSCs express low but detectable levels of Kit (encoding c-KIT; Supplementary Fig. S3B), the paracrine signaling partner for KITL, such that PSC monoculture seemed to be a reasonable in vitro model to begin assessing how KITL signaling impacts pancreatic mesenchyme. To this end, we analyzed the transcriptional profiles of Kitl-knockdown and Kitl-overexpressing PSCs, together with appropriate controls, by using RNA-seq. We prioritized gene expression changes resulting from Kitl overexpression as this cell line is activated and therefore CAF-like, although Kitl knockdown was indeed achievable. Restoring Kitl expression caused the upregulation of genes involved in cell adhesion and ECM or collagen organization, including integrins, laminins, cadherins, and protocadherins (Fig. 3A and B), suggesting the potential involvement of stromal KITL in regulation of normal tissue architecture. Conversely, Kitl knockdown led to the upregulation of genes involved in inflammatory processes, including genes involved in complement or interferon signaling, together with downregulation of cell adhesion genes. Specific inflammatory and architectural genes modulated by KITL restoration were regulated in the opposite direction by Kitl knockdown, supporting a specific role for KITL (Fig. 3C). We note, however, that a substantial group of genes positively regulated by Kitl signaling were expressed at a very low level in control cells and were not significantly downregulated further upon Kitl knockdown (Supplementary Fig. S3C and S3D). These results suggest that mesenchymal KITL may maintain pancreas tissue homeostasis, prompting us to move into in vivo modeling of KITL function.
Figure 3.
KITL regulates PSC state and pancreas tissue homeostasis. A, Volcano plot of all upregulated, nonsignificant, and downregulated differentially expressed genes (DEG) as defined by the Wald test (P.adj < 0.05 and log2FC > 1) from Kitl overexpression (Kitl OE) ImPSC-1 bulk RNA-seq dataset with representative gene labels included. Data represent three biological repeats. B, Gene ontology (GO) analysis of upregulated and downregulated genes in immortalized PSCs (ImPSC-1) overexpressing Kitl. Top 10 enrichment categories ranked by adjusted P values plotted in each direction. C, qRT-PCR analysis of indicated DEGs in ImPSC-1 expressing stable Kitl knockdown (shKitl) or overexpression (Kitl OE). PLKO vector (shCtrl) and fluorescence overexpression (Egfp OE) serves as the respective control. D, Representative images of CODEX staining (left) and quantification (right) for GFP (green) and CD31 (red) in normal pancreas from Fabp4-Cre or Kitlfl/fl;Fabp4-Cre mouse model (n = 3–4 mice per arm). Scale bar, 100 μm. E, Representative images of CODEX composite staining (left) and quantification (right) for CD45 (white) in normal pancreas from Fabp4-Cre or Kitlfl/fl;Fabp4-Cre mouse model (n = 4 mice per arm). Scale bar, 50 μm. F, Representative images of IHC staining for GFP (green), c-KIT receptor (KIT, red), and VIM (magenta) in healthy murine pancreas from Rosa26mTmG/+;Fabp4-Cre mice. Scale bar, 10 μm. G, Representative images of IHC staining for CD31 (green) and c-KIT receptor (KIT, red) in healthy murine pancreas. Scale bar, 10 μm. H, Representative images of IHC staining for panCK (white) and c-KIT receptor (KIT, red) in healthy murine pancreas. Scale bar, 10 μm. I, Representative hematoxylin and eosin images between caerulein-treated Fabp4-Cre and Kitlfl/fl;Fabp4-Cre mice. Scale bar, 100 μm. J, Representative images of IHC staining (left) for panCK (green) and amylase (red) between caerulein-treated Fabp4-Cre and Kitlfl/fl;Fabp4-Cre mice. Scale bar, 10 μm. PanCK quantification on right (n = 3 mice per arm). K, Representative IHC images (left) and quantification (right) for CD45 staining between caerulein-treated Fabp4-Cre and Kitlfl/fl;Fabp4-Cre mice. Scale bar, 100 μm. L, Representative images (left) and quantification (right) for CD68 staining in caerulein-treated Fabp4-Cre and Kitlfl/fl;Fabp4-Cre mouse pancreas. The same co-stained tissue section is shown in K and L. Scale bar, 100 μm. For comparisons of two groups, an unpaired t test was used. For comparisons of more than two groups with one independent variable, the one-way ANOVA test was used. Data are represented as mean ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; FC, fold changes; ns, not significant; panc area, pancreas.
To assess the relevance of mesenchymal KITL signaling for pancreas tissue architecture, we analyzed the consequences of conditional Kitl loss using Kitlflox/flox;Fabp4-Cre mice compared with Fabp4-Cre controls in the settings of homeostasis and tissue injury. First, we analyzed these tissues under normal, homeostatic conditions and crossed in a Rosa26mTmG/+ allele to enable visualization of PSCs based on GFP expression in these tissues. Based on our transcriptional profiling results, we compared tissue microenvironments in Rosa26mTmG/+;Kitlflox/flox;Fabp4-Cre mice compared with Rosa26mTmG/+;Fabp4-Cre controls using co-detection by indexing (CODEX), a barcode-based, multiplexed imaging approach (34). Although the total VIM-positive and CD31-positive cell abundance was not different between genotypes (Supplementary Fig. S4A and S4B), we observed clear changes to the perivascular niche with the loss of mesenchymal Kitl, including an increase in GFP-positive mesenchyme adjacent to endothelial cells (Fig. 3D). We also observed an increase in CD45-positive leukocytes within normal pancreas tissue when stromal Kitl was perturbed (Fig. 3E). We also noted a trend toward decreased α-SMA–positive, VIM-positive cells with Kitl perturbation (Supplementary Fig. S4C); as fibroblasts are α-SMA negative in normal pancreas tissue, this likely reflects a reduction in contractility of vascular smooth muscle cells. To assess potential cellular receivers of mesenchymal KITL that participate in paracrine signaling, we stained pancreas tissues from Rosa26mTmG/+;Fabp4-Cre mice for GFP, VIM, and KITL receptor KIT. KIT-positive cells were found adjacent to GFP-positive mesenchyme, consistent with the potential for cell–cell communication (Fig. 3F). As PSCs are localized in perivascular regions and next to pancreatic epithelium, but KIT-positive cells are few in number in the pancreas tissue, we reasoned that acinar cells were unlikely to be the cellular source of KIT but that CD31-positive endothelial cells and cytokeratin-high ductal epithelial cells may be relevant KIT-positive cell populations. Consistent with this notion, IHC demonstrated KIT expression by subpopulations of ductal epithelial cells and few endothelial cells (Fig. 3G and H). To confirm these results, we analyzed KIT expression by flow cytometry with co-stains for CD45 (immune cells), CD31 (endothelial cells), or EpCAM (epithelial cells), reasoning that KIT-positive cells that are negative for these three additional markers represent KIT-positive mesenchyme. KIT-positive cells were found in the EpCAM-positive fraction, consistent with a ductal epithelial identity and were rarely but measurably positive for CD31 or CD45 (Supplementary Fig. S4D and S4E), consistent with our IHC. We also noted a KIT-expressing population that was negative for these markers, which may be a population of mesenchymal cells expressing KIT. We also note that the fairly high proportion of KIT-positive cells among live cells in our flow cytometry experiments likely reflects substantial acinar cell death during the preparation of single-cell suspensions, as acinar cells seem to be negative for KIT, and we have likely therefore enriched for KIT-positive cells. In light of the observed patterns of c-KIT expression, we treated several cell types including PDAC cells and endothelial cells with recombinant KITL in culture but did not observe changes in cell viability over time (Supplementary Fig. S4F). Together, these results suggest that paracrine signaling via KITL influences neighboring cell types and acellular features in an intact tissue context.
In light of measurable albeit modest differences to tissue structure upon the loss of mesenchymal Kitl, we assessed the consequences of this KITL pool in the setting of tissue damage. For this, we subjected Kitlflox/flox;Fabp4-Cre mice and Fabp4-Cre controls to acute pancreatitis by administering repeated injections of the cholecystokinin analogue caerulein or saline as a vehicle control. As expected, in control mice, caerulein induced a mild inflammation characterized by edema and leukocyte accumulation evident by hematoxylin and eosin staining (Fig. 3I). However, in Kitl conditional knockout mice, caerulein led to far more pronounced tissue inflammation, as well as greater alterations to the epithelial compartment, which we speculated may represent metaplasia or altered epithelial plasticity. To assess this, we co-stained tissues from caerulein-treated mice with amylase (acinar cell marker) and pan-cytokeratin (ductal cell marker), which indicated an increase in ductal marker expression in the inflamed Kitl conditional knockout mice (Fig. 3J) along with evidence of amylase/pan-cytokeratin co-staining of individual cells. Inflammation was more pronounced in the Kitl conditional knockout mice, evidenced by an increased abundance of CD45+ leukocytes and CD68+ macrophages in the pancreas compared with that in control mice (Fig. 3K and L). Together, these results implicate mesenchymal KITL in regulation of pancreas tissue homeostasis such that KITL downregulation promotes inflammation and perturbation of normal tissue architecture.
We next addressed the potential of stromal KITL to regulate pancreatic tumor growth. We performed orthotopic implantation of KPC-derived PDAC cells from a pure C57BL/6J background into syngeneic Kitlflox/flox;Fabp4-Cre mice or Fabp4-Cre controls. Despite the aggressive nature of this mouse model, we found that the loss of mesenchymal Kitl significantly accelerated tumor growth (Fig. 4A) and increased tumor weights and tumor burden at experimental endpoint (Fig. 4B and C). We then repeated these experiments using moribundity as an experimental endpoint instead of a fixed timepoint. Consistent with the tumor growth measurements, survival studies revealed that the loss of mesenchymal Kitl significantly shortened survival compared with mice in KITL-expressing hosts (Fig. 4D). We characterized the mesenchymal compartment of these tumors by staining for PDPN (pan-CAF marker) and α-SMA (myofibroblast-like CAF marker) and found similar CAF abundance in tumors across genotypes (Fig. 4E), which is consistent with the notion that mesenchymal KITL regulates tissue homeostasis but is lost in an established tumor microenvironment. Proliferation and apoptosis were also similar across tumor genotypes at experimental endpoint (Fig. 4F and G), further supporting a role for KITL in early tumor progression. A tumor-restraining role for stromal KITL was also observed in an independent PDAC model (Fig. 4H–J), except that tumor weights at humane endpoint were not different in this model. As implantable models involve the introduction of cells that have already undergone malignant transformation into pancreas tissue, these results suggest that mesenchymal KITL expression represents a tissue barrier to PDAC progression at least in part independent of epithelial cell–intrinsic tumor suppression mechanisms.
Figure 4.
Mesenchymal KITL restrains pancreatic tumor growth. A, Average tumor area (mm2) between Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 7 mice per arm). Data are represented as mean ± SEM. B, Tumor weights (grams) at experimental endpoint between Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 7 mice per arm). Data are represented as mean ± SEM. C, Representative hematoxylin and eosin images of Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5, at the same experimental endpoint. Scale bar, 1 mm. D, Kaplan–Meier plot depicting percent probability of survival between Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 7 mice per arm). Log-rank P value = 0.0072. E, Representative images of IHC staining and quantification of α-SMA (top) and podoplanin (PDPN, bottom) on tumors from Fabp4-Cre control mice and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 3 mice per arm). Zoomed-out image scale bar, 200 μm. Zoomed-in image scale bar, 50 μm. F, Representative image of IHC staining and quantification of cleaved caspase-3 (ccas3) on Fabp4-Cre control mice and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 3 mice per arm). Scale bar, 10 μm. G, Representative images of IHC staining and quantification of Ki67 on Fabp4-Cre control mice and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells 6419c5 (n = 3 mice per arm). Scale bar, 10 μm. H, Average tumor area (mm2) between Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells FC1199 (n = 5 mice per arm). I, Tumor weights (grams) at humane endpoint from Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells FC1199 (n = 5 mice per arm). J, Kaplan–Meier plot depicting percent probability of survival between Fabp4-Cre control and Kitlfl/fl;Fabp4-Cre mice injected with KPC-derived murine PDAC cells FC1199 (n = 5 mice per arm). Log-rank P value = 0.0019. All data are represented as mean ± SEM, apart from Kaplan–Meier plots. For comparisons of two groups, the unpaired Mann–Whitney test was used. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, not significant.
Discussion
In this study, we provide evidence that a cell population in normal pancreatic mesenchyme expresses KITL/SCF; stromal downregulation of KITL is an accompanying feature of pancreatic tumorigenesis, as CAFs derived from these KITL-positive cells retain a lineage label but do not retain KITL expression; and functionally, stromal KITL is a barrier to tumor progression in pancreas tissue. The recent reports of abundant, KRAS-mutant, preinvasive lesions throughout examined cohorts of PDAC-free human pancreas tissues (19, 20) compared with the relatively low frequency of PDAC across the general population indicates the pervasive relevance of tumor suppression mechanisms in the adult pancreas. These mechanisms likely include epithelial cell–intrinsic mechanisms promoting, among other things, genome stability and susceptibility to immune surveillance; functions of the immune system, potentially including clearance of highly mutated epithelial cells with tumorigenic potential; and functions of the nonimmune stroma. Within the nonimmune stroma, mesenchymal components—fibroblasts in particular—are broadly implicated in the maintenance of normal tissue structure or architecture as well as support of healthy tissue homeostasis via production of soluble factors, basement membrane, and ECM components. Perturbation of fibroblast phenotypes to an activated state is an anatomically conserved feature of many solid cancers and some inflammatory conditions (25), and although activated fibroblasts in disease states generally express ECM components and immune-modulatory factors, granular features of fibroblast activation programs are tissue and disease specific. Although activated fibroblasts in cancer carry out diverse functions to promote tumor progression, normal fibroblasts serve to restrain tumor formation in promoting the ordered tissue structure that must be overcome to enable cancer formation or progression. We propose KITL as a tumor-restraining stromal mechanism in the pancreas, raising the possibility that specific effectors downstream of KITL signaling may hold use for cancer progression. Although this study suggests that inhibition of effectors downstream of stromal KITL have potential relevance for cancer prevention, we speculate that a therapeutic window prevents the use of receptor tyrosine kinase inhibitors such as imatinib, which inhibits c-KIT, from having a meaningful negative impact on normal pancreas tissue homeostasis, consistent with the excellent safety profile of imatinib (35). Future efforts will aim to investigate the significance of KITL signaling in the specific context of low-grade PanIN lesions, as these lesions are found in adult human pancreas (19, 20).
Although our study was restricted to the pancreas, these findings fit within a broader context of prior studies implicating mesenchymal KITL and/or LEPR-positive mesenchyme as critical regulators of healthy tissue homeostasis and normal tissue function in diverse organ sites. As briefly discussed above, LEPR-positive mesenchymal cells in the bone marrow associate tightly with endothelial cells and form a niche critical for hematopoietic stem cells (22). Interestingly, upon tissue damage such as irradiation or chemotherapy requiring regeneration of hematopoietic stem cells, LEPR-positive mesenchymal cells differentiate into adipocytes which in turn produce KITL to enable a functional niche and support hematopoietic regeneration (36). Complementary mesenchymal and signaling components were recently shown to support normal tissue homeostasis and suppress inflammation in BAT; LEPR-positive mesenchyme supports adaptive thermogenesis and restrains inflammation in BAT (23), whereas endothelial cell–derived KITL/SCF signals to KIT on brown adipocytes to promote homeostatic lipid accumulation when thermogenesis is inhibited (37). As the stellate cells under investigation in our study are also lipid-storing cells, these studies raise the possibility that lipid-storing stromal cells engage KITL signaling to promote tissue homeostasis and limit inflammation more broadly across organs.
Methods
Human Tissue Samples
All experiments with human patient–derived material were performed with approval of the Oregon Health & Science University (OHSU) and Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Review Boards. Sections from formalin-fixed, paraffin-embedded (FFPE) tissue samples of patients with PDAC harboring benign adjacent pancreas tissue were donated to the Oregon Pancreas Tissue Registry program with informed written patient consent in accordance with full approval by the OHSU Institutional Review Board or with informed consent of biospecimen collection with full approval by the MSKCC Institutional Review Board. Prospective collection of benign human pancreas tissue was performed with full approval by the MSKCC Institutional Review Board.
Animals
All experiments involving mice were reviewed, approved, and overseen by the Institutional Animal Care and Use Committees at the OHSU and MSKCC in accordance with the NIH guidelines for the humane treatment of animals. Male and female mice were used for all experiments, with ages specified in the experimental sections that follow. Littermate controls were used whenever possible. Animals included in pancreatitis and PDAC experiments were assessed daily based on score sheets with criteria including body condition scoring and physical examination to ensure humane treatment. Orthotopic tumors were grown to a maximum diameter of 1.0 cm based on institutional guidelines. Maximal burden was not exceeded with any animal. The following mice were used in this study, all purchased from The Jackson Laboratory: C57BL/6J (000664), Rosa26mTmG (007676), Fabp4-Cre (005069), Kitlflox (017861), Trp53frt (017767), and KrasFSF-G12D (023590). The Pdx1-FlpO mouse strain was kindly provided by Dr. Michael Ostrowski (the Medical University of South Carolina).
Cell Lines
The 6419c5, FC1199, and FC1245 cell lines were derived from autochthonous PDAC in the KrasLSL-G12D/+;Trp53LSL-R172H/+;Pdx1-Cre genetically engineered mouse model of pure C57BL/6J background and were kindly provided by Dr. Ben Stanger (6419c5, the University of Pennsylvania) and Dr. David Tuveson (FC1199 and FC1245, the Cold Spring Harbor Laboratory). PSC cell lines mPSC-1 and ImPSC-1 were generated by our group as previously described (33). Cell line authentication was not performed. Mycoplasma testing was performed on cell lines monthly using the MycoStrip Mycoplasma Detection Kit (InvivoGen rep-mys-20).
Pancreatitis Induction
Acute pancreatitis was induced in male and female mice at 8 weeks of age by intraperitoneal injection of caerulein (80 μg/kg, Sigma-Aldrich, C9026) eight times per day with 1 hour between injections, on two consecutive days, consistent with prior studies (38). Mice were then euthanized 2 days after the final caerulein injection and pancreata collected.
Orthotopic Transplantation of PDAC Cells
Male or female mice at 8 to 10 weeks of age were anesthetized and orthotopically implanted with 5 × 104 (6419c5) or 5 × 103 (FC1199 and FC1245) PDAC cells in a 50% Matrigel solution into the body of the pancreas. Tumor progression was monitored longitudinally by using high-resolution ultrasound using the Vevo 2100 Imaging System. Mice were euthanized and tumors collected either when the first mouse of the experiment reached the humane endpoint, or at different timepoints when each individual mouse in the experiment reached humane endpoints.
scRNA-seq
Cell Isolation
To isolate healthy PSCs, pancreata were harvested from Rosa26mTmG/+;Fabp4-Cre mice at 6 to 9 weeks of age, trimmed to remove any associated adipose tissue, minced with scissors, digested with 0.02% pronase (Sigma-Aldrich), 0.05% collagenase P (Sigma-Aldrich), and 0.1% DNase I (Sigma-Aldrich) in Gey’s Balanced Salt Solution (GBSS; Sigma-Aldrich) at 37°C for 10 minutes. Pancreata were further mechanically dissociated using a serologic pipette before returning to chemical dissociation at 37°C for 5 minutes. The resulting cell suspension was filtered through a 100-μm cell strainer nylon mesh. Cells were washed with GBSS, pelleted, and subject to red blood cell lysis via ACK lysis buffer (Thermo Fisher Scientific) for 3 minutes at room temperature (RT). Then, cells were washed in cold FACS buffer (PBS containing 2% FBS), pelleted, and resuspended in FACS buffer. Cells were kept on ice as a single-cell suspension and then GFP-positive cells were isolated by FACS using a BD FACSAria III or BD FACSymphony S6.
To isolate CAFs, 8-week-old Rosa26mTmG/+;Fabp4-Cre mice were orthotopically implanted with FC1245 PDAC cells as described above. At day 21 after implantation, pancreata were harvested, and any apparent normal pancreas tissue was trimmed away from the PDAC specimen. Tumors were briefly minced, placed in digestion media (DMEM with 1 mg/mL collagenase IV, 0.1% soybean trypsin inhibitor, 50 U/mL DNase, and 0.125 mg/mL Dispase) and incubated at 37°C for 1 hour. Whole tissue digests were centrifuged at 450g for 5 minutes, then resuspended in 10 mL prewarmed 0.25% trypsin and incubated at 37°C for 10 minutes. Cold DMEM (10 mL) was added to the suspension, which was then passed through a 100-μm cell strainer. Cells were centrifuged as above, washed with DMEM containing 10% FBS and centrifuged again, then centrifuged as above and resuspended in 1 mL ACK red cell lysis buffer. Cells were incubated at RT for 3 minutes, then 9 mL FACS buffer was added, and cells were centrifuged as above. Pelleted cells were counted and resuspended at 1 × 107 cells/mL in FACS buffer, CD16/CD32 Fc block (BD 553141) added 1:20 and incubated at RT for 2 minutes, then Biotinylated PDPN antibody (BioLegend, 127404) was added 1:200. Cell suspensions were incubated on ice for 30 minutes. Cold FACS buffer was added, cells centrifuged at 300g for 5 minutes at 4°C, and cell pellets were resuspended in 500 μL cold FACS buffer containing 1:1,000 APC Streptavidin (BD 554067) and incubated for 30 minutes on ice protected from light. Cold FACS buffer (2 mL) was added, cells were pelleted as above and resuspended in cold FACS buffer containing SYTOX Blue Dead Cell Stain (Invitrogen, S34857). Cells were incubated for 30 minutes on ice, then washed with FACS buffer, pelleted, and resuspended in cold FACS buffer. GFP-positive PDPN-positive cells were isolated by FACS using a BD FACSAria III or BD FACSymphony S6.
Sequencing and Analysis
The isolated pancreatic mesenchymal cells were immediately used for single-cell RNA-seq library preparation. Single-cell capture and cDNA library generation were performed using the 10× Genomics Chromium Next GEM Single Cell 3′ library construction dual index kit v3.1 (1000215) according to the manufacturer’s instructions. Libraries were pooled prior to sequencing based on the estimated cell number in each library per flow cytometry cell counts. Sequencing was performed on the Illumina NovaSeq 6000 platform at the OHSU Massively Parallel Sequencing Shared Resource, sequencing 20,000 read pairs per cell.
We aligned the sequenced reads to the mm10 mouse reference genome, and the unique molecule identifier for each gene in each cell was counted using the Cell Ranger (10× Genomics). Then we imported the resulting gene expression matrices into R (version 4.0.3) and analyzed the data using the Seurat (39) pipeline (version 4.0.1). Genes had to be expressed in at least three cells to be considered for downstream analyses. Cells were filtered to retain those that contained at least 1,000 minimum unique genes expressed, no more than 5,000 unique genes, more than 200 total unique molecule identifiers, and less than 10% of counts mapped to the mitochondrial genome. Batch correction was performed to integrate the samples from different conditions using the reciprocal principal component analysis integration workflow (40) within Seurat. The first 30 principal components were selected for downstream analysis based on the elbow point on the plot of SDs of the principal components. Uniform manifold approximation and projection was generated using the RunUMAP function with the same first 30 principal components used in clustering analysis.
We performed pseudotime trajectory analysis to elucidate the differentiation pathways of normal pancreatic and cancerous cells using Monocle 3 (v.1.3.7; ref. 41). To achieve this, we first integrated our single-cell RNA-seq datasets using Harmony (v.0.1.1; ref. 42) to correct for batch effects, enabling a unified visualization of cellular heterogeneity across samples. Subsequent trajectory inference with Monocle 3 was conducted using default parameters to order cells in pseudotime, thus highlighting the dynamic progression of cellular states. To visualize gene expression patterns along the trajectories, we used the “plot_cell_trajectory” function, focusing on the expression of Kitl in the Harmony-adjusted dimensional space.
Mouse Mesenchymal Cell Isolation
Primary mPSCs were isolated from wild-type (WT) C57BL/6J (000664) mice from The Jackson Laboratory at 8 to 9 weeks of age. Our isolation protocol is adapted from previously described methods (30, 43) with some minor modifications. Healthy pancreatic tissues from eight male mice were pooled, trimmed, and digested in Hank’s Balanced Salt Solution (HBSS; Sigma-Aldrich, H8264) containing 0.5 mmol/L of magnesium chloride hexahydrate (MgCl2 × 6H2O; Sigma-Aldrich, M9272), 10 mmol/L HEPES (Cytiva, SH30237.01), 0.13% collagenase P (Roche, 11213873001), 0.1% pronase (Sigma-Aldrich, 10165921001), and 0.001% DNase (Roche, 04716728001) for 7 minutes in shaking water bath (120 cycles/minutes) at 37°C. The remaining connective and adipose tissues were removed before the second incubation at 37°C in a shaking water bath (80 cycles/minutes) for an additional 7 minutes. Digested tissues were then filtered through a 250-µm nylon mesh (Thermo Fisher Scientific, 87791) and centrifuged at 450g for 10 minutes at 4°C. The cell pellet was washed in GBSS containing 120 mmol/L salt (NaCl; Sigma-Aldrich, S3014) and 0.3% BSA (Thermo Fisher Scientific, BP9703100) before repeating the above centrifugation step. Upon removing the wash buffer, cells were resuspended in GBSS + NaCl containing 0.3% BSA, to which an equal volume of 28.7% solution of Nycodenz (ProteoGenix, 1002424) in GBSS + NaCl was added and mixed well. The cell suspension in Nycodenz was then gently layered beneath GBSS containing 120 mmol/L NaCl and 0.3% BSA using a long needle and subjected to centrifugation at 1,400g for 20 minutes at 4°C. Primary quiescent PSCs were carefully harvested from the interface using sterile pipette and washed with GBSS + NaCl containing 0.3% BSA. Cells were pelleted and plated into multiple wells of a six-well dish in Iscove’s Modified Dulbecco’s Medium (IMDM; Cytiva, SH30228.02) containing 10% FBS (VWR, 97068-085) and 1% Antibiotic-Antimycotic (Thermo Fisher Scientific, 15240-062). The cell culture was maintained in a humidified atmosphere at 37°C in 5% CO2 and subjected to verification using qPCR and immunostaining upon treatment with 20 ng/mL recombinant TGF-β1 (R&D Systems, 7666-MD-005) over the course of 7 days.
To isolate PDGFRα+ mesenchymal cells from healthy mouse pancreas, five normal pancreata from WT C57BL/6J were pooled and subjected to primary mouse PSC isolation as described above. The methods are described in the methods section titled “Flow Cytometry”.
Human Mesenchymal Cell Isolation
Normal benign pancreas tissue was obtained from patients undergoing surgical resection for pancreatic disease. Tissue samples were collected in ice-mold MACS Tissue Storage Solution (Miltenyi Biotec, 130-100-008) and processed immediately. PSCs were isolated by density centrifugation as described previously (32). The resected tissue was trimmed of adipose tissue, transferred to 20 mL of digestion solution containing collagenase P, pronase, and DNase in GBSS as described above for mouse PSC isolation, and incubated at 37°C for 7 minutes in a shaking water bath. After the initial incubation, the partially digested tissue was transferred to a dish, finely minced with scissors, and subject to an additional 7-minute incubation. The digested tissue was then transferred to a 50-mL Falcon tube, thoroughly pipetted, and filtered through a nylon mesh to obtain a single-cell suspension. The cell suspension was centrifuged at 450g for 10 minutes at 4°C, and the resulting pellet was washed in GBSS + NaCl containing 0.3% BSA. The wash and centrifugation steps were repeated as described. The pellet was resuspended in 9.5 mL of GBSS + NaCl containing 0.3% BSA, and 8 mL of a 28.7% Nycodenz solution in GBSS + NaCl was added and mixed thoroughly. A gradient was prepared by layering 6 mL of GBSS + NaCl with 0.3% BSA over the Nycodenz–cell suspension in a round-bottom polycarbonate centrifuge tube. The sample was centrifuged at 1,400g for 20 minutes at 4°C. PSCs were separated into a distinct fuzzy band just above the interface between the BSA solution and the Nycodenz layer. This band was carefully harvested using a pipette without disturbing the gradient layers. The harvested cells were washed in GBSS + NaCl with 0.3% BSA, centrifuged at 450g for 10 minutes.
To isolate additional pancreatic mesenchyme, the cells below the interface band from the gradients described above were harvested, resuspended in GBSS + NaCl with 0.3% BSA, and centrifuged at 450g for 10 minutes. Cells were resuspended in 1 mL of IMDM and blocked for 15 minutes with human FcR blocking reagent (Miltenyi Biotec, 130-059-901, 1:200 dilution). After blocking, the cells were washed with 1 mL of GBSS + NaCl with 0.3% BSA and centrifuged again at 450g for 10 minutes. The cell pellet was resuspended in 120 µL of IMDM (Cytiva, SH30228.01), and 20 µL each of anti-human CD45 (Miltenyi Biotec, 130-118-780), CD326/EpCAM (Miltenyi Biotec, 130-061-101), Cytokeratin (Miltenyi Biotec, 130-123-094), and CD31 (Miltenyi Biotec, 130-091-935) MicroBeads were added. The suspension was incubated for 30 minutes at 4°C. After incubation, the cells were washed with 1 mL of GBSS + NaCl with 0.3% BSA and centrifuged at 450g for 10 minutes. The pellet was resuspended in 1 mL of IMDM. A magnetic column was prepared using a MACS Separator (Miltenyi Biotec, 130-042-202) and washed with 2 mL of IMDM. The labeled cell suspension was applied to the column and allowed to flow through completely. The CD45-negative, Cytokeratin-negative, CD31-negative, and CD326-negative flow-through was collected and centrifuged at 300g for 10 minutes. Gene expression analysis was performed on freshly isolated cells, with a sampling of each fraction plated on coverslips and imaged to validate stellate cell and fibroblast morphology.
To culture-activate primary human pancreatic mesenchymal cells, the cells were seeded onto collagen-coated coverslips (Corning, 354089) immediately after isolation. Serum-free conditioned media was collected from MIA PaCa-2 human PDAC cells, concentrated using a Vivaspin 3 kDa concentration unit (Cytiva, 28-9323-58), and mixed 1:1 with fresh complete IMDM containing 20% FBS and 2% glutamine, and added to the plated cells. Media was replenished every 2 days for a total of 14 days.
Stable Kitl Knock down and Overexpression in PSCs
Kitl knockdown (shKitl) and overexpression (Kitl OE) mPSC-1 cell lines were generated using MISSION Lentiviral short hairpin RNA (MilliporeSigma, Clone ID: TRCN0000067872) and Kitl open reading frame lentivirus (GeneCopoeia, EX-Mm03868-Lv158), respectively. Vector pLKO.1 neo (shCtrl; Addgene, 13425) and Egfp open reading frame (Egfp OE; GeneCopoeia, EX-EGFP-Lv158) were included as controls. Immortalized mPSC-1 cells were transduced with specified lentiviral particles for 48 hours prior to selection with 1 mg/mL Geneticin (Thermo Fisher Scientific, 10131035) for 4 days. KITL protein and transcript expression were then quantified using qPCR and ELISA to assess silencing and OE efficiency. Stable cells were maintained in a humidified atmosphere at 37°C in 5% CO2 and routinely passed in DMEM (Thermo Fisher Scientific, 11965118) containing 10% FBS (VWR, 97068-085), 1 mmol/L sodium pyruvate (Thermo Fisher Scientific, 11360070), and 1% Antibiotic-Antimycotic (Thermo Fisher Scientific, 15240-062).
RNA-seq of shKitl and Kitl OE PSCs
Total RNA was isolated using RNeasy Micro Kit (Qiagen, 74004) per the manufacturer’s instructions and quantified using a NanoDrop microvolume spectrophotometer before submission for bulk RNA-seq. RNA library preparation, sequencing, and analysis were conducted at Azenta Life Sciences as follows. Total RNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies), and RNA integrity was checked using Agilent 4200 TapeStation System (Agilent Technologies). ERCC RNA Spike-In Mix (Thermo Fisher Scientific, 4456740) was added to normalized total RNA prior to library preparation by following the manufacturer’s protocol. Total RNA underwent polyA selection and RNA-seq libraries preparation using the NEBNext Ultra II RNA Library Prep Kit for Illumina using the manufacturer’s instructions (New England Biolabs). Briefly, mRNAs were initially enriched with Oligo(dT) beads. Enriched mRNAs were fragmented for 15 minutes at 94°C. First-strand and second-strand cDNA were subsequently synthesized. cDNA fragments were end-repaired and adenylated at the 3′-ends, and universal adapters were ligated to cDNA fragments, followed by index addition and library enrichment by PCR with limited cycles. The sequencing library was validated on the Agilent TapeStation (Agilent Technologies) and quantified by using the Qubit 2.0 Fluorometer (Invitrogen) as well as by qPCR (Kapa Biosystems). The sequencing libraries were multiplexed and clustered onto a flow cell on the Illumina NovaSeq instrument according to the manufacturer’s instructions. The samples were sequenced using a 2- × 150-bp paired-end configuration at an average of 30 million reads per sample. Image analysis and base calling were conducted by using the NovaSeq Control Software. Raw sequence data (.bcl files) generated from Illumina NovaSeq were converted into FASTQ files and de-multiplexed using Illumina bcl2fastq 2.20 software. One mismatch was allowed for index sequence identification.
After investigating the quality of the raw data, sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality. The trimmed reads were mapped to the reference genome GRCm38.91 (mm10) available on ENSEMBL using the STAR aligner v.2.5.2b. The STAR aligner is a splice aligner that detects splice junctions and incorporates them to help align the entire read sequences. BAM files were generated as a result of this step. Unique gene hit counts were calculated by using featureCounts from the Subread package v.1.5.2. Only unique reads that fell within exon regions were counted. The gene hit counts table was used for downstream differential expression analysis. Using DESeq2, a comparison of gene expression between the groups of samples was performed. The Wald test was used to generate P values and log2 fold changes. Genes with adjusted P values <0.05 and absolute log2 fold changes >1 were called as differentially expressed genes for each comparison. Volcano plot visualization of significant differentially expressed genes were performed in Galaxy (44) using the ggplot2 R package. Significant gene labels from top gene ontology categories were included.
Functional enrichment analysis was performed using Enrichr (45) on the statistically significant set of genes by implementing Fisher exact test (GeneSCF v1.1-p2). Significance of tests was assessed using adjusted P values defined by Enrichr. Enrichment bar plots were generated using srPlot (46) to include the top 10 upregulated and downregulated Gene Ontology (GO) categories. A similar approach was used for conserved functional analysis, with the significance determined by using P values <0.05, as defined by Enrichr. To screen for the conserved GO categories, we compared the upregulated GO terms in shKitl with the downregulated GO terms in Kitl OE, and conversely, the downregulated GO terms in shKitl with the upregulated GO terms in Kitl OE.
IHC, Immunofluorescence, and Lipid Staining
Mouse and Human Tissue Sample Staining
Standard protocols were performed for IHC. Briefly, tissue samples were fixed overnight in 10% neutral buffered formalin (Sigma-Aldrich, HT501128-4L) and submitted to the MSKCC Laboratory of Comparative Pathology or Molecular Cytology Core Facility for paraffin embedding, sectioning, and hematoxylin and eosin sectioning. Sectioned tissues were deparaffinized using Citrisolv (Thermo Fisher Scientific, 22-143-975) and rehydrated in ethanol series (DeconLabs, 2701) before undergoing antigen retrieval using citrate or tris-based antigen unmasking solution (Vector Laboratories, H3300, H3301). The slides were then blocked with 8% BSA (Fisher Bioreagents, BP9703100) for 1 hour at RT and incubated in primary antibodies at 4°C overnight. Primary antibodies for α-SMA (Cell Signaling Technology, 19245S), PDPN (eBio8.1.1, Invitrogen, 14538182), GFP (Thermo Fisher Scientific, A10262; Abcam, ab1218; and Rockland Immunochemicals, 600-101-215), pan-cytokeratin (Thermo Fisher Scientific, MA5-13156), CD31 (R&D Systems, AF3628 or Abcam ab7388), Biotinylated anti-c-kit (R&D Systems, BAF1356), vimentin (Cell Signaling Technology, 5741 D21H3 XP), CD45 (R&D Systems, AF114), CD68 (R&D Systems, MAB101141), NG2 (EMD Millipore, MAB5384), CD105 (R&D Systems, MAB1320), or pancreatic amylase (Thermo Fisher Scientific, PA5-25330) were diluted at 1:200 to 1:400 in 8% BSA in PBS. The next day, slides were washed with PBS (Biotium, 22020) and incubated in α-chicken Alexa Fluor 488 (Thermo Fisher Scientific, A32931), α-rabbit Alexa Fluor 647 (Thermo Fisher Scientific, A21245), α-Syrian hamster Alexa Fluor 647 (Abcam, ab180117), or α-mouse Alexa Fluor 555 (Thermo Fisher Scientific, A21424) secondary antibodies at 1:200 to 1:400 dilution for 1 hour at RT. Tissue slides were washed with PBS and mounted with VECTASHIELD mounting media containing DAPI (Vector Laboratories, H-1200-10).
For phospho-c-KIT staining, fresh-frozen, optimal cutting temperature (OCT)-embedded tissue sections were washed with TBS-T (Santa Cruz Biotechnology, sc-362311) to dissolve at OCT. Samples were then permeabilized with 1% FBS (Thermo Fisher Scientific, A4766801) in PBS-T, 0.1% Triton X-100 (Sigma-Aldrich, X100-500ML). The slides were blocked with 5% FBS in PBS for 30 minutes at room temperature, then incubated in primary antibodies for 2 hours at room temperature, before incubating further at 4°C overnight. Primary antibodies for human phospho-c-Kit (Y703; Thermo Fisher Scientific, 710762) and human vimentin (Thermo Fisher Scientific, PA1-16759) were diluted at 1:50 and 1:100, respectively, in 1% FBS in PBS. Primary antibody for mouse phospho-c-Kit (G-Biosciences, ITA0925) was diluted at 1:50 in 1% FBS in PBS. The next day, slides were washed with 1% FBS in PBS-T, 0.1% Triton X-100 and incubated in α-chicken Alexa Fluor 488 (Invitrogen, A78948), α-rabbit Alexa Fluor 594 (Invitrogen, A21207), or α-rat Alexa Fluor 488 (Invitrogen, A21208) secondary antibodies at 1:100 to 1:200 dilution for 2 hours at RT. Tissue slides were further stained with DAPI (Thermo Fisher Scientific, 62248) at 1:1,000 dilution for 30 minutes at RT, washed with PBS, and mounted with VECTASHIELD Vibrance Mounting Media (Vector Laboratories, H-1700-2).
All images were acquired on a Carl Zeiss LSM 880 laser scanning confocal inverted microscope using 20×, 40×, or 63× objective. Whole slide scans were completed at the MSKCC Molecular Cytology Core Facility. Image analysis was performed using QuPath quantitative pathology and FIJI/ImageJ open source software. Where applicable, co-localization analysis was performed using the JaCoP plugin in ImageJ.
Cell Staining and Imaging
Cells seeded in chamber slides were fixed in 4% paraformaldehyde for 15 minutes and permeabilized with 0.1% Triton X-100 for 10 minutes before undergoing blocking in 5% BSA for 1 hour at RT. Sample slides were then probed with α-SMA primary antibodies (Thermo Fisher Scientific, MA5-11547) overnight at 4°C followed by standard Alexa Fluor 647–conjugated secondary antibody (Thermo Fisher Scientific, A21235) incubation for an hour at RT. Upon repeating the standard washing steps, slides were mounted for imaging using VECTASHIELD mounting media containing DAPI (Vector Laboratories, H-1200-10). For lipid staining, cells seeded in chamber slides were stained using Nile Red (MedChemExpress, HY-D0718) at 1 µmol/L final working concentration for 10 minutes and counterstained with DAPI (Thermo Fisher Scientific, 62248). Nile Red signals were detected at excitation/emission wavelengths of 559 nm/635 nm.
Two-plex FISH
Transcript expression on tissues, except where RNAscope was indicated, was performed using the Thermo Fisher Scientific ViewRNA ISH Tissue Assay kit (two plex) for use on mouse and human tissue samples. Briefly, samples were first permeabilized with controlled protease digestion, followed by incubation with proprietary probe-containing solution, according to the manufacturer’s instructions. During incubation, samples had to remain fully submerged. After hybridization with the probe, samples were washed, followed by sequential hybridization with the preamplifier and amplifier DNA. In accordance with the manufacturer’s instructions, hybridizations were performed with the preamplifier, amplifier, and fluorophore. Mounting medium with DAPI (VECTASHIELD HardSet mounting media with DAPI) was used to mount samples.
RNAscope Combined with IHC
Paraffin-embedded tissue sections were cut at 5 μm and kept at 4°C. Samples were loaded into Leica BOND RX, baked for 30 minutes at 60°C, dewaxed with BOND Dewax Solution (Leica, AR9222), and pretreated with EDTA-based epitope retrieval ER2 solution (Leica, AR9640) for 15 minutes at 95°C. The probe mKitL (Advanced Cell Diagnostics, ready to use, no dilution, 423408) was hybridized for 2 hours at 42°C. Mouse PPIB (Advanced Cell Diagnostics [ACD], Cat# 313918) and dapB (ACD, Cat# 312038) probes were used as positive and negative controls, respectively. The hybridized probes were detected using RNAscope 2.5 LS Reagent Kit—Brown (ACD, Cat# 322100) according to the manufacturer’s instructions with some modifications (DAB application was omitted and replaced with Fluorescent CF594/Tyramide (Biotium, 92174) for 20 minutes at RT).
After the run was completed, slides were washed in PBS and incubated in 5 μg/mL DAPI (Sigma-Aldrich) in PBS for 5 minutes, rinsed in PBS, and mounted using MOWIOL 4-88 (Calbiochem). Slides were kept overnight at −20°C before imaging.
After the slides were scanned, the coverslips were removed and the slides were loaded into Leica BOND RX for double immunofluorescence staining. Samples were pretreated with EDTA-based epitope retrieval ER2 solution (Leica, AR9640) for 20 minutes at 100°C. Double antibody staining and detection were conducted sequentially. The primary antibodies against GFP (2 μg/mL, chicken, Abcam, ab13970) and CD31 (Abcam, ab182981) were incubated for 1 hour at RT. For rabbit antibodies, Leica BOND Polymer anti–rabbit horseradish peroxidase [HRP; included in the Polymer Refine Detection Kit (Leica, DS9800)] was used; for the chicken antibody, a rabbit anti-chicken (Jackson ImmunoResearch, 303-006-003) secondary antibodies were used as linkers for 8 minutes before the application of the Leica BOND Polymer anti–rabbit HRP for 8 minutes at RT. The Leica BOND Polymer anti–rabbit HRP secondary antibody was applied followed by Alexa Fluor Tyramide Signal Amplification reagents (Life Technologies, B40953 and B40958) were used for immunofluorescence detection. After the run was completed, slides were washed in PBS and mounted in MOWIOL 4-88 (Calbiochem). Slides were kept overnight at −20°C before imaging.
CODEX
Antibody Panel Development, CODEX Staining, and Imaging
To construct an antibody panel visualizing pancreatic architecture in FFPE mouse samples using CODEX (34), conventional IHC staining was performed to screen for antibodies-binding canonical markers of pancreatic epithelial cells [E-cadherin, Novus Biologicals, #NBP2-33006 clone 1A4/asm-1; α-amylase, Cell Signaling Technology, #3796 clone D55H10], endothelial cells (CD31, Cell Signaling Technology, #14472 clone 4A2), stromal cells (vimentin, Cell Signaling Technology, #70257 clone D3F8Q; α-SMA, Cell Signaling Technology #77699 clone D8V9E), leukocytes (CD45, Cell Signaling Technology, #46173 clone D21H3), and lineage reporter (GFP, Rockland Immunochemicals, #600-101-215 polyclonal). Identified antibody clones were then conjugated with oligonucleotide barcodes using the Antibody Conjugation Kit (Akoya Biosciences). Prior to CODEX imaging, each conjugated antibody was validated by following the manufacturer’s instructions, and tissue staining patterning was confirmed using published literature.
CODEX staining and imaging was performed as described in the user manual (https://www.akoyabio.com/wp-content/uploads/2021/01/CODEX-User-Manual.pdf). In brief, 5-μm FFPE pancreas sections were mounted onto 22 × 22 mm glass coverslips (Electron Microscopy Sciences) coated in 0.1% poly-L-lysine (Sigma-Aldrich) and stained with using CODEX Staining Kit (Akoya Biosciences). A cocktail of the above-conjugated antibodies were incubated with tissue overnight at 4°C. On the next day, fluorescent oligonucleotide-conjugated reporters were combined with Nuclear Stain and CODEX Assay Reagent (Akoya Biosciences) in sealed light-protected 96-well plates (Akoya Biosciences). Automated fluidics exchange and image acquisition were performed using the Akoya CODEX instrument integrated with a BZ-X810 epifluorescence microscope (Keyence) and CODEX Instrument Manager (CIM) v1.30 software (Akoya Biosciences). The exposure times are as follows: E-cadherin, barcode BX006, 600 ms; Amylase, barcode BX031, 250 ms; vimentin, barcode BX025, 300 ms; α-SMA, barcode BX052, 250 ms; CD31, barcode BX002, 350 ms; CD45, barcode BX007, 400 ms; and GFP, barcode BX041, 250 ms. All images were acquired using a CFI Plan Apo I 20×/0.75 objective (Nikon). The “High resolution” mode was specified using Keyence software to reach a final resolution of 377.44 nm/pixel.
Processing of CODEX Images and Analysis
Image stitching, drift compensation, deconvolution, z-plane selection, and background subtraction were performed using the CODEX Processor v1.7 (Akoya Biosciences) per the manufacturer’s instruction (https://help.codex.bio/codex/processor/technical-notes). Individual channel images were then imported into ImageJ v1.53t for analyses as described below.
Total pancreatic areas were annotated by the sum of Amylase+ and E-cadherin+ regions. Immune cells were defined by DAPI and CD45 double positivity, whereas the vasculature area was annotated by the CD31+ region. vimentin and α-SMA signal were used to mark total and activated fibroblast cells, respectively. GFP positivity was used to track PSC lineage-derived cells.
Flow Cytometry
To analyze c-KIT expression, normal pancreas tissues were harvested from WT C57BL/6J mice aged 6 to 9 weeks and digested as described above. After ACK lysis, cells were incubated with CD16/CD32 antibody (BD Biosciences, 553141) to block Fc receptors for 2 minutes at RT. Cells were then stained with the following for 30 minutes on ice: SYTOX Blue Dead Cell Stain (Invitrogen, S34857); Biotinylated m-SCF R/c-kit antibody (R&D Systems, BAF1356). Cells were then washed with cold FACS buffer, pelleted, then stained with PE/Cyanine7 Streptavidin (BioLegend, 405206), anti-mouse CD31 APC (Invitrogen, 17-0311-82), and anti-mouse EpCAM (CD326) FITC (Invitrogen, 11-5791-82) for 30 minutes on ice, before the cells were washed with FACS buffer, pelleted, then resuspended in cold FACS buffer for flow cytometry.
To analyze epithelial cells, immune cells, and c-KIT, pancreata from C57Bl/6J mice aged 6 to 9 weeks old were harvested and digested as described above. After ACK lysis, cells were incubated with CD16/CD32 antibody (BD Biosciences, 553141) for 2 minutes at RT. Cells were then stained with SYTOX Blue and a biotinylated c-KIT antibody on ice for 30 minutes on ice, washed with cold FACS buffer, pelleted, then stained with PE/Cyanine7 Streptavidin (BioLegend, 405206), anti-mouse CD45 PE-Cyanine5 (Invitrogen, 15-0451-82), anti-mouse EpCAM (CD326) FITC (Invitrogen, 11-5791-82) for 30 minutes on ice. Cells were washed with FACS buffer, pelleted, then resuspended in cold FACS buffer for flow cytometry.
To isolate PDGFRα+ mesenchymal cells from healthy mouse pancreas, five normal pancreata from WT C57BL/6J were pooled and subjected to primary mouse PSC isolation as described above. The isolated cells were first stained with Zombie NIR dye (Therno Fisher Scientific, 50-604-714) for 5 minutes at RT to exclude dead cells. Next, cells were blocked with CD16/CD32 (BD Biosciences, 553141) for 10 minutes at 4°C to prevent nonspecific antibody binding, and then incubated with anti-mouse PDGFRα PE (BioLegend, 135905) at a 1:100 dilution for 20 minutes at 4°C. Upon staining, the cells were washed with FACS buffer containing HBSS (Thermo Fisher Scientific, 14175095) and 0.5% BSA (Fisher BioReagents, BP9703100), sorted using a BD Biosciences FACSymphony S6 cell sorter, and then processed for RNA isolation.
Gene Expression Analysis by qPCR
Total RNA was isolated using the RNeasy Micro Kit (Qiagen, 74004) per the manufacturer’s instructions and quantified using a NanoDrop microvolume spectrophotometer. About 500 ng to 1 µg of RNA was reverse transcribed using iScript Reverse Transcriptase Supermix (Bio-Rad, 1708841) to produce cDNA. RT-PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific, 4367659). Gene-specific primer pairs were designed using the NCBI Nucleotide database or acquired from MilliporeSigma. Gene expression was normalized to the reference gene Rplp0. Primer pair sequences were as follows: Rplp0 Forward 5′-GTGCTGATGGGCAAGAAC-3′ Reverse 5′-AGGTCCTCCTTGGTGAAC-3′, mKitl Forward 5′-TTATGTTACCCCCTGTTGCAG-3′ Reverse 5′-CTGCCCTTGTAAGACTTGACTG-3′, mKit Forward 5′-GAGACGTGACTCCTGCCATC-3′ Reverse 5′-TCATTCCTGATGTCTCTGGC-3′, mActa2 Forward 5′-AGCCATCTTTCATTGGGATGGA-3′ Reverse 5′-CATGGTGGTACCCCCTGACA-3′, mPdpn Forward 5′-AGATAAGAAAGATGGCTTGC-3′ Reverse 5′-AACAACAATGAAGATCCCTC-3′, mPdgfrα 5′-GCAGTTGCCTTACGACTCCAGA-3′ Reverse 5′-GGTTTGAGCATCTTCACAGCCAC-3′, mPcdhga12 Forward 5′-ACAATGCCCCTGAAGTAGCC-3′ Reverse 5′-TCCAGTGCGAGGTGAGTTTC-3′, mCdh26 Forward 5′-CCTCGTCGTTGTTGTGGAGA-3′ Reverse 5′-CTCTGAGGGTGAAAGGCTGG-3′, mItga1 Forward 5′-GGCAGTGGCAAGACCATAAGGA-3′ Reverse 5′-CATCTCTCCGTGGATAGACTGG-3′, mIrf5 Forward 5′-CCTACAGAACCACTCTTGCCTG-3′ Reverse 5′-CCTTGTGGGTTGCTGATGGTGA-3′, mLrrc15 Forward 5′-TTCAGCCACCTGAACCAGTTGC-3′ Reverse 5′-GTCCTGTAGAGCATTGGTGTGG-3′, hFABP4 Forward 5′-ACGAGAGGATGATAAACTGGTGG-3′ Reverse 5′-GCGAACTTCAGTCCAGGTCAA-3′, hPDGFRA Forward 5′-GACTTTCGCCAAAGTGGAGGAG-3′ Reverse 5′-AGCCACCGTGAGTTCAGAACGC-3′, hCOL1A1 Forward 5′-CATGGAGACTGGTGAGACCT-3′ Reverse 5′-GCCATACTCGAACTGGAATC-3′, hKITLG Forward 5′-CTGGAGACTCCAGCCTACACTG-3′ Reverse 5′-CTGCCCTTGTAAGACTTGGCTG-3′.
ELISA Quantikine Assay
Immortalized parental and shKitl mPSC-1 cells were seeded into a six-well dish at 3 × 105 confluency in growth media containing DMEM (Thermo Fisher Scientific, 11965126), 10% Seradigm FBS (VWR, 97068-085), 1 mmol/L sodium pyruvate (Thermo Fisher Scientific, 11360070), and 1% Antibiotic-Antimycotic (Thermo Fisher Scientific, 15240-062). Primary PSCs were seeded in six-well dish at 1 × 104 confluency in IMDM (Cytiva, SH30228.02) containing 10% FBS (VWR, 97068-085) and 1% Antibiotic-Antimycotic (Thermo Fisher Scientific, 15240-062). Conditioned media were collected at indicated timepoints and concentrated using Vivaspin Turbo 20 3K MWCO concentrator (Cytiva, 28932358) in accordance with the manufacturer’s protocol. Concentrated supernatants were quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23225), and mouse KITL protein quantification was performed using the Mouse SCF Quantikine ELISA Kit (R&D Systems, MCK00) according to the manufacturer’s protocol.
Cell Proliferation Assay
Human and mice PDAC macrophage and endothelial cells were seeded into 96-well white-walled plates at a density of 5 × 103 cells per well. All cells were maintained in DMEM (Thermo Fisher Scientific, 11965126) containing 10% FBS (VWR, 97068-085) except for human endothelial cells, which were cultured in endothelial cell media (Cell applications, 213-500). The following day, cells were washed with PBS (Cytiva, SH30028.FS) and treated with recombinant KITL/SCF at the final concentration of 100 ng/mL: human SCF (R&D Systems, 255-SC-010) or mouse KITL/SCF (R&D Systems, 455-MC-010). Plates were replenished with recombinant KITL/SCF daily and collected for read every 24 hours for three continuous days. Proliferation assay was performed using CellTiter-Glo Luminescent Cell Viability Assay reagent (Promega, G7572) per the manufacturer’s protocol and read using a GloMax plate reader.
Statistical Analysis
No statistical methods were used to predetermine sample sizes. The experiments were not randomized. For animal studies, a minimal number of mice were selected based on preliminary studies, with an effort to achieve a minimum of n = 3, mostly n = 5–10 mice per treatment group for each experiment. Age-matched mice were selected for experiments. For histological staining quantification, analyses were performed in a blinded fashion. For batch-processed images, image analyses were done in an unbiased manner using image analysis software. Some Western blots and RT-qPCR assays were performed by a researcher who was blind to the experimental hypothesis. Animals were excluded if an animal had to be removed from an experiment early for reasons seemingly unrelated to tumor burden. All experiments were performed and reliably reproduced at least two independent times. GraphPad Prism 9 was used to generate graphs and for statistical analyses. Statistical significance was calculated for two unmatched groups by using unpaired t test with Welch correction or Mann–Whitney test. One- or two-way ANOVAs were used for more than two groups as specified, followed by Tukey multiple comparisons tests. Datasets are presented as mean ± SEM. P values under 0.05 were considered significant.
Data Availability
The data generated in this study are publicly available in the Gene Expression Omnibus under accession numbers GSE278857 (bulk RNA-seq) and GSE278162 (scRNA-seq). Previously published scRNA-seq data from liver are available under accession number GSE172492 (47). Code for the project is hosted on GitHub at the following link: https://github.com/Canping-Chen/Stromal_KITL/tree/main.
Supplementary Material
Supplementary Figure S1: scRNA-seq reveals gene expression programs in PSCs and PSC-derived CAFs
Supplementary Figure S2: Kitl is expressed by healthy pancreatic mesenchyme and reduced upon activation to a CAF phenotype
Supplementary Figure S3: Stromal KITL promotes regulation of pancreas tissue architecture
Supplementary Figure S4: Stromal KITL promotes pancreas tissue homeostasis
Acknowledgments
We thank Mark Berry, Wenfei Kang, and all members of the Sherman lab for helpful discussion of this work. This study was supported by NIH grants R01 CA229580 and R01 CA250917 (to M.H. Sherman), P01 CA244114 (to M.H. Sherman and Y. Hang), R01 GM147365 (to Z. Xia), and R01 CA283378 (to S.W. Lowe); a Lustgarten Foundation Therapeutics Focused Research Program grant (to M.H. Sherman); and a Silver Family Innovation Fund Award (to Z. Xia). S.W. Lowe is an Investigator of the Howard Hughes Medical Institute and the Geoffrey Beene Chair for Cancer Biology at the MSKCC. Authors at Stanford University (Z. Yan, Y. Hang, and S.K. Kim) were supported by NIH object storage grant S10 OD025082. We thank members of the OHSU Histopathology Shared Resource, Advanced Light Microscopy Shared Resource, Flow Cytometry Shared Resource, and Massively Parallel Sequencing Shared Resource for supporting this study, with financial support from the NIH P30 Cancer Center Support Grant CA069533. We also thank members of the MSKCC Flow Cytometry Core, Molecular Cytology Core, and Center of Comparative Medicine & Pathology for supporting this work, with financial support from NIH–NCI P30 Cancer Center Support Grant CA008748.
Footnotes
Note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).
Authors’ Disclosures
No disclosures were reported.
Authors’ Contributions
M.K. Oñate: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Oon: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Bhattacharyya: Conceptualization, data curation, formal analysis, investigation, visualization, methodology. V. Low: Investigation. C. Chen: Resources, data curation, formal analysis, investigation, visualization, methodology. X. Zhao: Resources, data curation, formal analysis, investigation, visualization, methodology. F. Arnold: Resources, formal analysis. Z. Yan: Resources, data curation, formal analysis, visualization, methodology. S. Pramod: Formal analysis, validation, investigation, visualization, methodology. Y. Hang: Resources, data curation, formal analysis, investigation, visualization, methodology. Y.-J. Ho: Data curation, formal analysis. S.W. Lowe: Supervision. S.K. Kim: Supervision. Z. Xia: Data curation, formal analysis, supervision, investigation, methodology. M.H. Sherman: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, writing–original draft, writing–review and editing.
References
- 1. Stoker MG, Shearer M, O’Neill C. Growth inhibition of polyoma-transformed cells by contact with static normal fibroblasts. J Cell Sci 1966;1:297–310. [DOI] [PubMed] [Google Scholar]
- 2. Brown S, Pineda CM, Xin T, Boucher J, Suozzi KC, Park S, et al. Correction of aberrant growth preserves tissue homeostasis. Nature 2017;548:334–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Pineda CM, Gonzalez DG, Matte-Martone C, Boucher J, Lathrop E, Gallini S, et al. Hair follicle regeneration suppresses Ras-driven oncogenic growth. J Cell Biol 2019;218:3212–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Gallini S, Annusver K, Rahman NT, Gonzalez DG, Yun S, Matte-Martone C, et al. Injury prevents Ras mutant cell expansion in mosaic skin. Nature 2023;619:167–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Blaisdell A, Crequer A, Columbus D, Daikoku T, Mittal K, Dey SK, et al. Neutrophils oppose uterine epithelial carcinogenesis via debridement of hypoxic tumor cells. Cancer Cell 2015;28:785–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Cui C, Chakraborty K, Tang XA, Zhou G, Schoenfelt KQ, Becker KM, et al. Neutrophil elastase selectively kills cancer cells and attenuates tumorigenesis. Cell 2021;184:3163–77.e21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Goto N, Westcott PMK, Goto S, Imada S, Taylor MS, Eng G, et al. SOX17 enables immune evasion of early colorectal adenomas and cancers. Nature 2024;627:636–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. DuPage M, Cheung AF, Mazumdar C, Winslow MM, Bronson R, Schmidt LM, et al. Endogenous T cell responses to antigens expressed in lung adenocarcinomas delay malignant tumor progression. Cancer Cell 2011;19:72–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Kaukonen R, Mai A, Georgiadou M, Saari M, De Franceschi N, Betz T, et al. Normal stroma suppresses cancer cell proliferation via mechanosensitive regulation of JMJD1a-mediated transcription. Nat Commun 2016;7:12237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Brügger MD, Valenta T, Fazilaty H, Hausmann G, Basler K. Distinct populations of crypt-associated fibroblasts act as signaling hubs to control colon homeostasis. PLoS Biol 2020;18:e3001032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Cukierman E. A reflection on how carcinoma-associated fibroblasts were recognized as active participants of epithelial tumorigenesis. Cancer Res 2021;81:4668–70. [DOI] [PubMed] [Google Scholar]
- 12. Mueller MM, Fusenig NE. Friends or foes - bipolar effects of the tumour stroma in cancer. Nat Rev Cancer 2004;4:839–49. [DOI] [PubMed] [Google Scholar]
- 13. Quail DF, Joyce JA. Microenvironmental regulation of tumor progression and metastasis. Nat Med 2013;19:1423–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Gupta PB, Pastushenko I, Skibinski A, Blanpain C, Kuperwasser C. Phenotypic plasticity: driver of cancer initiation, progression, and therapy resistance. Cell Stem Cell 2019;24:65–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, Carey L, et al. Reconstruction of functionally normal and malignant human breast tissues in mice. Proc Natl Acad Sci U S A 2004;101:4966–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Hu M, Yao J, Carroll DK, Weremowicz S, Chen H, Carrasco D, et al. Regulation of in situ to invasive breast carcinoma transition. Cancer Cell 2008;13:394–406. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Tlsty TD, Coussens LM. Tumor stroma and regulation of cancer development. Annu Rev Pathol 2006;1:119–50. [DOI] [PubMed] [Google Scholar]
- 18. Hill W, Lim EL, Weeden CE, Lee C, Augustine M, Chen K, et al. Lung adenocarcinoma promotion by air pollutants. Nature 2023;616:159–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Braxton AM, Kiemen AL, Grahn MP, Forjaz A, Parksong J, Mahesh Babu J, et al. 3D genomic mapping reveals multifocality of human pancreatic precancers. Nature 2024;629:679–87. [DOI] [PubMed] [Google Scholar]
- 20. Carpenter ES, Elhossiny AM, Kadiyala P, Li J, McGue J, Griffith BD, et al. Analysis of donor pancreata defines the transcriptomic signature and microenvironment of early neoplastic lesions. Cancer Discov 2023;13:1324–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Helms EJ, Berry MW, Chaw RC, DuFort CC, Sun D, Onate MK, et al. Mesenchymal lineage heterogeneity underlies nonredundant functions of pancreatic cancer-associated fibroblasts. Cancer Discov 2022;12:484–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Ding L, Saunders TL, Enikolopov G, Morrison SJ. Endothelial and perivascular cells maintain haematopoietic stem cells. Nature 2012;481:457–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Haberman ER, Sarker G, Arús BA, Ziegler KA, Meunier S, Martínez-Sánchez N, et al. Immunomodulatory leptin receptor+ sympathetic perineurial barrier cells protect against obesity by facilitating brown adipose tissue thermogenesis. Immunity 2024;57:141–52.e5. [DOI] [PubMed] [Google Scholar]
- 24. Mederacke I, Hsu CC, Troeger JS, Huebener P, Mu X, Dapito DH, et al. Fate tracing reveals hepatic stellate cells as dominant contributors to liver fibrosis independent of its aetiology. Nat Commun 2013;4:2823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Buechler MB, Pradhan RN, Krishnamurty AT, Cox C, Calviello AK, Wang AW, et al. Cross-tissue organization of the fibroblast lineage. Nature 2021;593:575–9. [DOI] [PubMed] [Google Scholar]
- 26. Prados A, Onder L, Cheng HW, Mörbe U, Lütge M, Gil-Cruz C, et al. Fibroblastic reticular cell lineage convergence in Peyer’s patches governs intestinal immunity. Nat Immunol 2021;22:510–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Cheng HW, Onder L, Novkovic M, Soneson C, Lütge M, Pikor N, et al. Origin and differentiation trajectories of fibroblastic reticular cells in the splenic white pulp. Nat Commun 2019;10:1739. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Donahue KL, Watkoske HR, Kadiyala P, Du W, Brown K, Scales MK, et al. Oncogenic KRAS-dependent stromal interleukin-33 directs the pancreatic microenvironment to promote tumor growth. Cancer Discov 2024;14:1964–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Lee Y, Leslie J, Yang Y, Ding L. Hepatic stellate and endothelial cells maintain hematopoietic stem cells in the developing liver. J Exp Med 2021;218:e20200882. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Apte MV, Haber PS, Applegate TL, Norton ID, McCaughan GW, Korsten MA, et al. Periacinar stellate shaped cells in rat pancreas: identification, isolation, and culture. Gut 1998;43:128–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Garcia PE, Adoumie M, Kim EC, Zhang Y, Scales MK, El-Tawil YS, et al. Differential contribution of pancreatic fibroblast subsets to the pancreatic cancer stroma. Cell Mol Gastroenterol Hepatol 2020;10:581–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Vonlaufen A, Phillips PA, Yang L, Xu Z, Fiala-Beer E, Zhang X, et al. Isolation of quiescent human pancreatic stellate cells: a promising in vitro tool for studies of human pancreatic stellate cell biology. Pancreatology 2010;10:434–43. [DOI] [PubMed] [Google Scholar]
- 33. Auciello FR, Bulusu V, Oon C, Tait-Mulder J, Berry M, Bhattacharyya S, et al. A stromal lysolipid-autotaxin signaling Axis promotes pancreatic tumor progression. Cancer Discov 2019;9:617–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Goltsev Y, Samusik N, Kennedy-Darling J, Bhate S, Hale M, Vazquez G, et al. Deep profiling of mouse splenic architecture with CODEX multiplexed imaging. Cell 2018;174:968–81.e15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Druker BJ, Talpaz M, Resta DJ, Peng B, Buchdunger E, Ford JM, et al. Efficacy and safety of a specific inhibitor of the BCR-ABL tyrosine kinase in chronic myeloid leukemia. N Engl J Med 2001;344:1031–7. [DOI] [PubMed] [Google Scholar]
- 36. Zhou BO, Yu H, Yue R, Zhao Z, Rios JJ, Naveiras O, et al. Bone marrow adipocytes promote the regeneration of stem cells and haematopoiesis by secreting SCF. Nat Cell Biol 2017;19:891–903. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Lee HJ, Lee J, Yang MJ, Kim YC, Hong SP, Kim JM, et al. Endothelial cell-derived stem cell factor promotes lipid accumulation through c-Kit-mediated increase of lipogenic enzymes in brown adipocytes. Nat Commun 2023;14:2754. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Collins MA, Bednar F, Zhang Y, Brisset JC, Galbán S, Galbán CJ, et al. Oncogenic Kras is required for both the initiation and maintenance of pancreatic cancer in mice. J Clin Invest 2012;122:639–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Hao Y, Hao S, Andersen-Nissen E, Mauck WM III, Zheng S, Butler A, et al. Integrated analysis of multimodal single-cell data. Cell 2021;184:3573–87.e29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM III, et al. Comprehensive integration of single-cell data. Cell 2019;177:1888–902.e21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Cao J, Spielmann M, Qiu X, Huang X, Ibrahim DM, Hill AJ, et al. The single-cell transcriptional landscape of mammalian organogenesis. Nature 2019;566:496–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Korsunsky I, Millard N, Fan J, Slowikowski K, Zhang F, Wei K, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods 2019;16:1289–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Zhao L, Cai B, Lu Z, Tian L, Guo S, Wu P, et al. Modified methods for isolation of pancreatic stellate cells from human and rodent pancreas. J Biomed Res 2016;30:510–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Galaxy Community . The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2022 update. Nucleic Acids Res 2022;50:W345–1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics 2013;14:128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Tang D, Chen M, Huang X, Zhang G, Zeng L, Zhang G, et al. SRplot: a free online platform for data visualization and graphing. PLoS One 2023;18:e0294236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Mederacke I, Filliol A, Affo S, Nair A, Hernandez C, Sun Q, et al. The purinergic P2Y14 receptor links hepatocyte death to hepatic stellate cell activation and fibrogenesis in the liver. Sci Transl Med 2022;14:eabe5795. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Figure S1: scRNA-seq reveals gene expression programs in PSCs and PSC-derived CAFs
Supplementary Figure S2: Kitl is expressed by healthy pancreatic mesenchyme and reduced upon activation to a CAF phenotype
Supplementary Figure S3: Stromal KITL promotes regulation of pancreas tissue architecture
Supplementary Figure S4: Stromal KITL promotes pancreas tissue homeostasis
Data Availability Statement
The data generated in this study are publicly available in the Gene Expression Omnibus under accession numbers GSE278857 (bulk RNA-seq) and GSE278162 (scRNA-seq). Previously published scRNA-seq data from liver are available under accession number GSE172492 (47). Code for the project is hosted on GitHub at the following link: https://github.com/Canping-Chen/Stromal_KITL/tree/main.