Secretory Mucin 5AC Promotes Neoplastic Progression by Augmenting KLF4-Mediated Pancreatic Cancer Cell Stemness

Pancreatic cancer ranks as the third deadliest malignancy with a 5-year survival rate of 9.3%, due to asymptomatic progression, molecular heterogeneity, and early dissemination (1). The constitutively active Kras mutation, present in 95% of the patients with pancreatic cancer, is the major initiating event leading to the development of pancreatic intraepithelial neoplasia (PanIN). The PanIN lesions, manifested with cellular atypia and differential mucin expression, are the most common precursor lesions of pancreatic cancer that progress to invasive carcinoma (2). The dismal survival statistics and the aggressive nature of pancreatic cancer warrant the identification of molecular player(s) facilitating the progression of Kras-driven precursor lesions to invasive ductal carcinoma. The Kras activation in the pancreatic parenchyma leads to the neo-to-over expression of mucins, a hallmark of pancreatic cancer, which significantly contributes to pancreatic cancer pathobiology. Gel-forming mucin 5AC (MUC5AC), one of the top differentially expressed genes and the most abundantly expressed mucin in pancreatic cancer, appears de novo at PanIN-1A stage (3) and continues to express in tumor tissues while remaining absent in normal pancreas (4). The MUC5AC expression correlates with disease progression and has been reported as a potential diagnostic (5) and prognostic marker (6–8) in pancreatic cancer.

Despite ample clinical significance, very few studies have evaluated the mechanistic contribution of MUC5AC during pancreatic cancer pathology. The transcription factor, GLI1, has been shown to upregulate MUC5AC expression in pancreatic cancer cells, which disrupts E-cadherin/β-catenin interaction, leading to nuclear accumulation of β-catenin and upregulation of tumor-associated genes (9). Furthermore, MUC5AC has been shown to obviate TRAIL-mediated cancer cell death in a xenograft model (7). Studies have also demonstrated the role of MUC5AC in the metastatic process by increasing VEGF and matrix metalloproteinases (MMP) expression through PI3K/Akt pathways (6). However, all these studies have utilized implanted cell lines models, which seldom recapitulate the complexities of autochthonous tumors. Therefore, we genetically ablated Muc5ac in an autochthonous murine model of pancreatic cancer (Kras^G12D; Pdx-1cre, KC) to evaluate its contribution to pancreatic cancer initiation and progression. Our study demonstrated that knockout (KO) of Muc5ac resulted in a significant delay in the neoplastic onset and progression of pancreatic cancer precursor lesions, and abridged the cancer cell stemness as indicated by the decreased expression of Aldh1a1, Klf4, EpCAM, and CD133. The cancer stem cells (CSC) represent a small percentage of cells within the tumor with a self-renewing ability that primarily contributes to cancer initiation, progression, and metastasis (10–12). Our findings further suggested that MUC5AC interacts with integrin β5 in the presence of the extracellular matrix proteins, resulting in pSTAT3-mediated upregulation of Klf4, a transcription factor involved in CSC maintenance. Integrins are heterodimeric transmembrane molecules that create specialized niches for CSCs to reside and sense microenvironmental cues for proliferation, differentiation, and migration by activating numerous downstream signaling pathways (13, 14). Observations from this study have delineated that MUC5AC primes the integrin β5/pSTAT3/Klf4 signaling axis and thereby contributes to the pro-tumorigenic and self-renewal potency of the neoplastic cells to promote Kras-induced metaplasia and pancreatic cancer progression.

C57BL/6 Muc5ac^−/− mice (15), obtained from Dr. Christopher Evans at the University of Colorado (Boulder, CO), were crossed with Pdx1-Cre and LSL-Kras^G12D mice to obtain the genotype Kras^G12D; Pdx1-Cre; Muc5ac^−/− (KCM) in the mouse pancreas. The genotyping for Kras, Pdx-1-Cre, and Muc5ac was performed by PCR using the primer sequences listed in Supplementary Table 1. The Kras^G12D; Pdx1-Cre (KC) and Kras^G12D; Pdx1-Cre; Muc5ac^−/− (KCM) animals (male and female), were randomized and euthanized at 10, 20, 30, 40, and 50 weeks of age (3–6 animals/group/time point; Supplementary Fig. S1A).

The mouse pancreatic tissues were embedded in paraffin, cut into serial tissue sections of 5 μm thickness, and stained with hematoxylin and eosin (H&E). The H&E-stained slides of all the KC and KCM animals for all time points (10–50 weeks) were histologically profiled and blindly scored by the pathologist for the overall proportion of neoplastic lesions versus the normal parenchyma retained. The lesions were further histologically scored by the pathologist according to the established criteria of PanIN grades 1–3 (16). The occurrence of each of these PanIN grades was then calculated as a percentage of the total number of neoplastic lesions for each animal.

Pancreas of KC and KCM mice of 50 weeks were enzymatically digested with 0.012% (w/v) collagenase XI (Sigma) and 0.012% (w/v) dispase (GIBCO) in DMEM media containing 1% FBS (GIBCO), followed by embedding them in growth factor–reduced Matrigel (BD Biosciences). Upon establishment of the organoids in enriched medium (17), they were cultured for 5, 10, 15, and 20 days. Three independent replicates of organoid culture/time point/group were sectioned, and their histology was visualized using H&E staining and IHC for Muc5ac. At the end of culture, genomic DNA was isolated from the KC and KCM organoids and genotyped for the recombination of LSL-KrasG12D locus, as explained in (16).

FG/COLO357 (Scr/ShMUC5AC) cells of >95% viability were orthotopically implanted in the pancreas of randomized immunodeficient mice (male and female), at a concentration of 5 × 10⁵ cells/50 μL PBS/mouse, as per the established protocol (8 animals/group; ref. 18). All mice were sacrificed after 50 days of implantation, and the presence of metastasis to different organs (liver, spleen, lung, mesenteric lymph nodes, and on the peritoneal wall) was determined by thorough gross inspection and histologic analysis. Pancreatic tumors and organs with visible metastatic lesions were excised, weighed, measured, formalin-fixed, paraffin-embedded, and were sectioned into 5-μm thick sections. Five to six primary tumors (randomly selected) per group were stained for MUC5AC and Ki67. The percentage of Ki67⁺ tumor cells were blindly scored by the pathologist in both groups. Metastatic organs were stained for MUC5AC and presented as representative images. This experiment was repeated twice.

Increasing dilutions of FG-Scr and FG-Sh5AC cells (10⁵,10⁴,10³, and 10² cells in 100 μL PBS) were subcutaneously injected on left and right flanks (3–4 flanks/cell concentration/group) of randomized immunocompromised mice (male and female). The tumor onset was recorded in each group, and after 35 days postimplantation, the tumors were resected, and tumor weight was measured. All the animal experiments (autochthonous murine model and xenografts) were approved by the Institutional Animal Care and Use Committee at the University of Nebraska Medical Center (Omaha, NE).

The RNA sequencing (RNA-seq) was conducted on RNA isolated from age-matched KC and KCM pancreatic tissues. The raw sequences were mapped to the mouse genome (UCSC mm10), downloaded from iGenomes using TopHat (RRID: SCR_013035). The bam files were assessed with CuffLinks (RRID: SCR_014597), and a differential gene expression profile in the KC and KCM animals were generated using CuffDiff (RRID: SCR_001647; ref. 19). A volcano plot was made using R bioconductor (RRID:SCR_006442) to represent the differentially expressed genes (20). ConsensusPathDB (http://cpdb.molgen.mpg.de/; RRID: SCR_002231; refs. 21, 22) was then used to assess the gene ontology (GO) terms differentially regulated in KCM. The top hits from this pathway analysis were assessed using gene set enrichment analysis (GSEA; ref. 23).

The power analysis for animal numbers was performed using an alpha error probability of 0.05 and a power level of 0.8. Two-tailed independent sample Student t test was used for comparison of normally distributed data. Nonnormally distributed data were analyzed by the nonparametric Wilcoxon rank-sum test. All the in vitro experiments were performed independently at least three times with proper biological and technical replicates. A P value of <0.05 was considered statistically significant.

SW1990 and FG/COLO357 cells were obtained from ATCC and Dr. Ohsawa, University of Tokyo, (Tokyo, Japan), respectively. Both the cell lines used in the study were authenticated by short-tandem repeat DNA profiles as per standard guidelines every 6 months, were regularly confirmed to be free of Mycoplasma contamination, and were used at a passage number less than 10. The detailed description of in vitro assays, cell lines, and reagents [sequences of primers and shRNAs (Supplementary Table 1), sources of antibodies (Supplementary Table S2.1 and S2.2)] used for the study are provided in the Supplementary Materials and Methods.

To explore the role of Muc5ac in pancreatic cancer onset and progression, the KC and KCM mice were euthanized at 10, 20, 30, 40, and 50 weeks of age; pancreases were resected, weighed, and histologically assessed for the neoplastic changes (Fig. 1A). Pathologic scoring revealed a greater percentage of normal parenchyma retained in KCM mice at 20 (P < 0.05), 30 (P < 0.01), and 50 (P < 0.05) weeks of age, as compared with KC mice (Fig. 1B). Further analysis of pancreatic tissue at 20 weeks of age revealed a significantly lower frequency of neoplastic lesions in the KCM mice in comparison with KC, suggesting a delay in pancreatic cancer onset.

The PanIN lesions were graded at different weeks of progression (n = 3–6 animals/week/cohort). A consistent decrease in the percentage of medium (PanIN-II) and high-grade (PanIN-III) lesions were observed in KCM mice during different weeks of progression. At 30 weeks, KC mice showed the development of PanIN-3 (3.3% of the pancreas) besides extensive PanIN-I (76.6% of the pancreas) and PanIN-II (11.6% of the pancreas), whereas age-matched KCM mice predominantly comprised of low-grade PanIN-I (65%) and PanIN-II (1.6%), which was apparent at 50 weeks of age, with a 50% reduction in PanIN-2 and a 70% reduction in PanIN-3 lesions in the KCM group as compared with KC (Fig. 1C), suggesting that Muc5ac supports pancreatic cancer progression. In addition, the average weight of the pancreas isolated from the KC animals was greater as compared with age-matched KCM mice, especially at 50 weeks of age (Supplementary Fig. S1B).

Next, we generated 3D-organoids from age-matched KC and KCM mice pancreas and genotyped the same for recombination event at LSL-Kras^G12D locus (Supplementary Fig. S1C). Interestingly, a higher percentage of KCM organoids lost their viability after 15 days in culture, while the KC organoids persistently proliferated even at 20 days in culture. The quantitative assessment of organoid size showed a 50% reduction in KCM group compared with KC (Fig. 1D and E), which indicates the role of Muc5ac in the sustained survival of malignant cells. Furthermore, H&E-stained sections of KC organoids demonstrated complex papillary architecture, nuclear atypia, and budding of cell clusters in the ductal lumen, resembling the histology of advanced PanIN-2 to PanIN-3 lesions, as compared with the simpler architecture of the KCM organoids, resembling the histology of early PanIN lesions (Fig. 1F).

To identify the potential downstream targets of Muc5ac in facilitating pancreatic cancer progression, RNA-seq was performed on the pancreatic tissue from 50-week-old KC and KCM animals (GSE160029). The top significantly downregulated genes were associated with CSC maintenance and epithelial-to-mesenchymal transition (EMT; Fig. 2A). The qRT-PCR validation of RNA-seq. analysis showed significant downregulation of Aldh1a1, Dclk1, Klf4, CD133, CD44, Zeb1, and EpCAM in KCM as compared with KC, validating the RNA-seq analysis (Fig. 2B). Immunofluorescence (IF) studies revealed a decrease in Klf4⁺ ductal epithelial cells in the KCM pancreas (Fig. 2C), with a strong positive correlation (R² = 0.78) of Muc5ac and Klf4 expression in the PanIN lesions of KC mouse (Fig. 2D). Concurrently, there was a significant decline in Aldh1a1-positive (Supplementary Fig. S1D and S1E), and EpCAM-positive (Supplementary Fig. S1F and S1G) ductal cells in the PanIN lesions in KCM, as compared with the KC mice. Klf4 has been shown to induce ductal phenotype and promote PanIN progression in the presence of Kras^G12D mutation (24, 25). The strong positive correlation of Muc5ac and Klf4 in the ductal lesions, with a concomitant elevation of other CSC markers in the KC mouse, suggests that MUC5AC/Klf4 axis may be involved in the maintenance of self-renewing CSC population during murine pancreatic cancer progression.

Next, we asked whether the MUC5AC-mediated enrichment of cancer cell stemness exists in human pancreatic cancer cells. An evaluation of a panel of human pancreatic cancer cell lines demonstrated a high to a medium expression of MUC5AC in FG/COLO357, SW1990, and BxPC3, while telomerase-transformed normal pancreatic ductal epithelial cell line HPNE lacked MUC5AC expression (Supplementary Fig. 2A). As MUC5AC is a secreted mucin, its expression was validated in culture supernatant by enzyme-linked immunosorbent assay (Supplementary Fig. 2B), and IF analysis showed surface and diffused cytoplasmic staining (Supplementary Fig. 2C). MUC5AC was stably silenced in two pancreatic cancer cell lines, FG/COLO357 and SW1990 (Supplementary Fig. S2D–S2G) using two different shRNA sequences. The knockdown (KD) cell lines, when analyzed for CSCs by dye exclusion side population (SP) analysis, showed 87% and 55% decline in the percentage of SP cells in FG/COLO357 and SW1990 cells, respectively (Fig. 2E and F). This was further corroborated by a decrease in the expression of stemness markers, Aldh1a1, Sox9, Klf4, EpCAM, CD133, and CD44 in the KD cell lines (Fig. 2G).

Accumulating evidence supports the role of CSCs in tumorigenesis and metastasis of pancreatic cancer cells (11, 12). MUC5AC-silenced pancreatic cancer cell lines, FG/COLO357 (FG-Sh5AC) and SW1990 (SW-Sh5AC), demonstrated G₁/S arrest, with a 14% (P < 0.05) and 19% (P < 0.001) increase in the percentage of cells in G₁ phase, with a concurrent decrease of 12% and 19% of cells in the S-phase, respectively (Fig. 3A; Supplementary Fig. 2H). Expression of cell-cycle regulators responsible for G₁ to S transition like cyclin-D1 and cyclin-E were also downregulated upon MUC5AC silencing (Fig. 3B; Supplementary Fig. 2I). The MUC5AC-deficient cells formed significantly fewer colonies (P < 0.001) in a soft-agar assay, suggesting that MUC5AC assists in anchorage-independent growth of pancreatic cancer cells (Fig. 3C). In addition, KD of MUC5AC moderately reduced cell viability under nutrient-deprived condition, as assessed by MTT assay under reduced serum (1%) conditions (Supplementary Fig. S2J and S2K).

Orthotopic implantation of FG-Scr and FG-Sh5AC cells into the pancreas of nude mice showed a significant reduction (P < 0.01) in tumor weight (Fig. 3D) in FG-Sh5AC group (Mean ± SE, 303 ± 74.3 mg) as compared with FG-Scr control (Mean ± SE, 801 ± 195.5 mg). This was further supported by a distinct decrease in the percentage of proliferative Ki67⁺ tumor cells in the MUC5AC-silenced group (Fig. 3E; Supplementary Fig. 2L). The findings were corroborated in the autochthonous mouse model, which demonstrated a similar trend in the decline in the percentage of Ki67⁺ ductal cells in the KCM pancreatic tumors compared with KC across all weeks of progression (Fig. 3F; Supplementary Fig. 2M). To address whether MUC5AC's ability to enhance the tumorigenic potential and proliferative index of cancer cells is dependent on the maintenance of CSCs, we performed in vitro and in vivo limiting dilution assays. FG-Sh5AC cells demonstrated a significant decline in spheroid-forming ability under low attachment (Fig. 3G and H), followed by attachment in normal plate (Fig. 3I; Supplementary Fig. 3A) conditions. The reduction in colony size was more evident at lower dilution of cells, which was further corroborated upon subcutaneous implantation of serially diluted FG-Scr and FG-Sh5AC cells (Fig. 3J). Although 1 × 10³ FG-Scr cells could form detectable tumors after 35 days postimplantation, no tumor was detected with the same number of FG-Sh5AC cells (Fig. 3K), suggesting that the tumor-initiating CSC population is enriched in the presence of MUC5AC. The significant decline in the tumor weight of the MUC5AC-depleted group (Fig. 3L) further reinstates that MUC5AC-associated CSC maintenance provides protumorigenic growth advantage to the pancreatic cancer cells.

Furthermore, corroborating our RNA-seq analysis that demonstrates the downregulation of EMT-associated genes in KCM mouse, silencing of MUC5AC reduced the migratory propensity of pancreatic cancer cells by 85%, as observed in transwell migration assay (Fig. 4A), wound healing assay (Supplementary Fig. S3B), and by phalloidin staining–based actin polymerization (Supplementary Fig. S3C). Alongside, we observed a reduction in metastatic incidences to peritoneum, intestine, diaphragm, liver, lung, and mesenteric lymph nodes in the mice bearing FG-ShMUC5AC orthotopic tumors as compared with FG-Scr control (Fig. 4B; Supplementary Fig. S3D and S3E), which maintained MUC5AC expression in the metastatic lesions (Fig. 4C). The GSEA on the RNA-seq data showed significant enrichment for the genes involved in cell migration in the KC mouse pancreatic tissue (Fig. 4D), supporting the premise that MUC5AC-expressing pancreatic cancer cells have a higher metastatic propensity. MMP-2 and MMP-9 were among the top significantly downregulated genes in the KCM tumors. Moreover, a significant decrease in MMP-2 and MMP-9 expression was observed in the FG-Sh5AC tumor as compared with the FG-Scr group (Fig. 4E). Interestingly, enhanced activity of MMP-2 and MMP-9 was observed in gelatin gel zymography, where FG-Sh5AC cells were treated with MUC5AC-enriched conditioned media (CM) from FG-Scr cells (Fig 4F). Cumulatively, our in vitro and xenograft studies recapitulate the observation from the autochthonous murine model that MUC5AC potentiates CSC maintenance, which is further manifested by high tumorigenic and metastatic propensity of MUC5AC-expressing pancreatic cancer cells.

To identify the molecular players that are mechanistically involved in the MUC5AC-mediated pro-tumorigenic phenotype of pancreatic cancer cells, we performed GSEA analysis using our RNA-seq. data that indicated that loss of MUC5AC significantly impacted integrin-mediated signaling in the KCM tumors (Fig. 5A). Furthermore, Gene Ontology (GO)-based clustering from the RNA-seq data revealed integrin-mediated cell adhesion (GO: 0033632 and GO: 0033628) as the top differentially regulated pathways (FDR: 0.017, 0.001) with scores 4.4 and 3.0, respectively. Thus, we evaluated the expression of multiple integrins and observed a significant loss of integrins αv and β5 in MUC5AC-KD cells compared with controls (Fig. 5B). We investigated the phosphorylation of Src, a mediator of integrin signaling, and observed that KD of MUC5AC reduced Src phosphorylation at tyrosine 416 (Y416; ref. 26) compared with control cells (Fig. 5B). Downstream of activated Src cascade, MUC5AC silencing significantly reduced the phosphorylation of STAT3 at tyrosine 705 (Y705) compared with scramble controls (Fig. 5B). Quantitatively, there was a significant and consistent decline in integrin β5, pSrc, and pSTAT3 levels upon MUC5AC ablation (Fig. 5C). To address whether MUC5AC, a secreted gel-forming molecule, devoid of any known putative signaling domain, can enhance the tumorigenicity of pancreatic cancer cells through integrin αvβ5/Src/STAT3 axis, FG-Sh5AC cells were treated with conditioned media from FG-Scr cells, which substantially rescued integrin β5 expression in FG-Sh5AC cells (Fig. 5D and E). Interestingly, MUC5AC co-immunoprecipitated integrin β5 on fibronectin-coated plates (Fig. 5F), suggesting that integrin αvβ5 may undergo a conformational change upon interaction with its ligand (fibronectin), which facilitates its interaction with MUC5AC. In addition, FG-Sh5AC cells, which demonstrated a significantly reduced attachment and proliferation on fibronectin and vitronectin-coated plates, key ligands for integrins αv and β5, showed a drastic increase in attachment (Supplementary Fig. S3F) and proliferation (Fig. 5G) upon treatment with the FG-Scr media, almost comparable with that of Scr control. Furthermore, the transient KD of integrin β5 significantly impaired the migratory potential of MUC5AC-expressing pancreatic cancer cells, almost to the extent seen in MUC5AC-silenced cells. The combined loss of MUC5AC and integrin β5 completely abrogated motility of the pancreatic cancer cells (Fig. 5H; Supplementary Fig. S3G). IHC analysis of human pancreatic cancer tissues for MUC5AC and integrin αvβ5 showed their spatial coexpression in PanINs as well as in well-differentiated and poorly differentiated tumors (Fig. 5I), indicating a clinical relevance of this novel signaling axis. Owing to the difficulty in locating epithelial cells in the poorly differentiated tumor tissues, we focused on the PanINs and well-differentiated tumors for the quantitative analysis that suggests a strong positive correlation of MUC5AC and integrin αvβ5 expression (Fig. 5J).

To address whether MUC5AC/Integrin β5-mediated signaling cascade, through pSrc-Tyr 416 and pSTAT3-Tyr 705, regulates the stemness in pancreatic cancer cells, we transiently knocked down integrin β5 in MUC5AC-expressing and KD cells. The KD of integrin β5 in FG-Scr cells downregulated the phosphorylation of Src (Tyr 416) and STAT3 (Tyr 705), whereas the total Src and STAT3 levels remained unchanged (Fig. 6A). Further supporting our proposed mechanism, pSTAT3 expression was substantially less in the KCM pancreatic tumor tissues as compared with the KC animals (Fig. 6B), with a strong positive correlation (R² = 0.82) of MUC5AC and pSTAT3 expression in the KC mouse (Fig 6C). Interestingly, neoplastic cells that abundantly expressed MUC5AC were positive for pSTAT3 expression in their nuclei (white arrowhead, Fig. 6B), while cells with low MUC5AC showed lesser expression of pSTAT3 (white arrow, Fig. 6B). Next, treatment of FG-Scr and FG-Sh5AC cells with C188-9, a pharmacologic inhibitor of phosphorylation and dimerization of STAT3 (27), resulted in reduced Klf4 expression (Fig. 6D and E). The C188-9 (5 μmol/L, 10 μmol/L) treatment also significantly decreased the migratory propensity of MUC5AC-expressing cells (Supplementary Fig. S3H). Furthermore, we observed a strong positive correlation of MUC5AC expression with that of pSTAT3 (R² = 0.78) and Klf4 (R² = 0.75) in the orthotopic tumor tissues (Fig. 6F and G), further supporting the findings from biochemical experiments and autochthonous mouse model. To understand whether the MUC5AC-mediated CSC enrichment in pancreatic cancer is dependent on Klf4, we transiently knocked down Klf4 in FG-Scr and FG-Sh5AC cells and analyzed the expression of CSC markers and the percentage of SP population. Upon Klf4 KD, there was a decrease in the expression of Aldh1a1 (Fig. 6H), with a concomitant decline in SP in FG-Scr and FG-Sh5AC by 95% and 60%, respectively (Fig. 6I; Supplementary Fig. S3I). Furthermore, upon treatment with conditioned media from FG-Scr cells for 48 hours, there was an upregulation of Klf4 and Aldh1a1 in the FG-Sh5AC cells (Fig. 6H) with a drastic increase in the SP (Fig. 6I). However, the ablation of Klf4 in FG-Sh5AC cells blunted its response to the conditioned media treatment, as evidenced by a relatively marginal increase in SP and a moderate increase in Aldh1a1 expression.

The constitutive activation of Kras in pancreatic cancer has been shown to rewire the signaling pathways leading to the expression of mucins, a hallmark of pancreatic cancer (28). The predominant mucins upregulated in pancreatic cancer include MUC1, MUC4, MUC16, and MUC5AC. Among these, MUC5AC, the most overexpressed mucin, appears de novo in the early neoplastic stage (PanIN-I). In multiple malignancies, overexpression of MUC5AC has been correlated with disease progression and poor prognosis (8). Recently, survival analysis on 176 patients with pancreatic cancer revealed that MUC5AC expression is significantly correlated to poor survival (www.proteinatlas.org). In our study, we aimed to delineate the oncogenic functions and underlying molecular mechanism of MUC5AC during the early stages of pancreatic cancer using a novel mouse model with genetic ablation of Muc5ac (Muc5ac^−/−) in the presence of oncogenic Kras^G12D expression (Kras^G12D; Pdx1-Cre). Loss of Muc5ac (KCM) delayed neoplastic onset and progression of low-grade to high-grade PanINs with a significant decline in ductal proliferation as compared with KC mice. Guided by the RNA-seq analysis, which demonstrated a significant downregulation of several CSC-associated genes in KCM, and the findings from the xenograft and biochemical studies with human pancreatic cancer cells, we postulated that depletion of MUC5AC reduces the self-renewal propensity or cancer cell stemness, and thereby delays the neoplastic onset and disease progression in KCM. Accumulating evidence suggests the critical role of CSCs in tumor initiation and long-term clonal repopulation in many malignancies (12, 29–31). Apart from the tumorigenic potential, CSCs promote EMT and impart a metastatic phenotype to the tumor (32). Interestingly, from RNA-seq analysis, multiple EMT-associated genes were significantly downregulated in KCM. This was further corroborated by our xenograft study, which demonstrated a substantial decline in metastasis in MUC5AC-depleted group.

The GSEA on the RNA-seq data suggested a disruption of integrin-mediated signaling in KCM, which was further corroborated in human pancreatic cancer cells upon MUC5AC depletion. The expression of integrin β5 was reinforced in the MUC5AC-knockdown cells by ectopic treatment of MUC5AC-enriched culture supernatant. Integrins are cellular mechanosensors that can engage with extracellular matrix (ECM) components to sense microenvironmental cues, and thereby facilitate neoplastic cell growth, self-renewal, and motility by sustaining cancer stemness (13, 14). A perpetual deposition of dense ECM components by activated fibroblasts, cumulatively termed as fibrosis, is a hallmark of pancreatic cancer (33). ECM-induced tissue stiffness promotes tumor aggressiveness via integrin-associated mechanosignaling in the epithelial cells (34). It has been previously reported that bulky glycoproteins on the tumor glycocalyx mechanically cluster integrins into adhesion plaques to augment oncogenic attributes of tumor cells (35). Among the other putative downstream pathways of integrin activation, STAT3 gets phosphorylated, which upregulates genes involved in cancer progression (36). Alongside, the chronic inflammation associated with Kras^G12D-induced tumorigenesis activates STAT3 to sustain oncogenic growth signals and supports the early events of lung and pancreatic tumorigenesis (37), (38). A recent study precisely demonstrated a positive feed-forward loop wherein activated STAT3 in the tumor epithelial cells enhances stiff fibrotic deposition, which in turn, intensifies integrin-mediated mechanosignaling to promote tumor progression (39). Furthermore, the leukemia inhibitory factor/STAT3 axis promotes embryonic stem cell renewal by activating Klf4 transcription (40). Klf4 was previously shown to promote the onset and progression of PanINs in the presence of Kras^G12D mutation by upregulating early response genes and driving acinar to ductal reprogramming (24, 25). In the same context, STAT3 is essential to maintain the epithelial identity of neoplastic cells in the PanINs (37). Hence, we questioned whether MUC5AC-mediated maintenance of self-renewing CSC population during early events of Kras^G12D-driven pancreatic cancer is dependent on integrin β5/STAT3/Klf4 axis. Knockdown of integrin β5 suppressed phosphorylation of STAT3 at Tyr705 residue and inhibition of STAT3 phosphorylation led to diminished expression of Klf4 in the MUC5AC-expressing pancreatic cancer cells. Furthermore, abrogating Klf4 expression drastically reduced the stemness property of MUC5AC-expressing cells. Finally, the strong positive correlation of Muc5ac with KLF4 and pSTAT3 in the metaplastic cells, with a concomitant enrichment of CSC markers and efficient PanIN progression in the KC mouse, suggests that MUC5AC-expressing neoplastic cells experience an advantageous self-renewal–dependent growth potency over the nonexpressing cells via STAT3/Klf4 axis. Cumulatively, this study demonstrates that upon activation of oncogenic Kras, de novo–expressed MUC5AC amplifies integrin β5/pSrc/pSTAT3 signaling, which maintains Klf4-associated stemness or self-renewal propensity of neoplastic cells, thereby promoting Kras-induced metaplastic growth and hence culminating toward a faster pancreatic cancer progression (Fig. 6J). Future studies will be directed to elucidate the contribution of MUC5AC in the modulation of the tumor microenvironment and paracrine mechanisms during pancreatic cancer progression.

S.K. Batra is co-founder of Sanguine Diagnostics and Therapeutics, Inc. No disclosures were reported by the other authors.

K. Ganguly: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing-original draft. S.R. Krishn: Formal analysis, validation, investigation, methodology, writing-review and editing. S. Rachagani: Investigation, methodology. R. Jahan: Validation, investigation. A. Shah: Validation, investigation. P. Nallasamy: Validation, investigation. S. Rauth: Investigation. P. Atri: Software, formal analysis. J.L. Cox: Formal analysis, visualization, methodology. R. Pothuraju: Investigation. L.M. Smith: Software, methodology. S. Ayala: Investigation. C. Evans: Resources. M.P. Ponnusamy: Conceptualization, writing-review and editing. S. Kumar: Conceptualization, supervision, investigation, writing-review and editing. S. Kaur: Conceptualization, supervision, project administration, writing-review and editing. S.K. Batra: Conceptualization, supervision, funding acquisition, project administration, writing-review and editing.

We greatly appreciate the kind technical help of Ms. Kavita Mallya, UNMC (Omaha, NE). We also thank Dr. Jessica Mercer for her editorial contributions. The authors/work on this manuscript were supported, in parts, by grants from the NIH (R01 CA247471, RO1 CA210637, RO1CA206444, RO1 CA183459, UO1 CA200466, PO1 CA217798, R44 CA224619, R41 CA213718, and R41 CA235984).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.