Targeted deletion of Wwox accelerates neoplastic lesions formation.
WWOX expression is reduced in human PanIN lesions9 suggesting that WWOX plays a direct role in PanIN formation and progression. To determine WWOX’s role of WWOX in PDAC formation, we generated a new genetically engineered mouse model harboring Wwox deletion and Kras activation in pancreatic acinar cells and examined their phenotypes. To achieve this, we bred mice carrying tamoxifen-inducible Cre (CreER) controlled by the Ptf1a specific acinar promoter (Ptf1a-CreER) with a conditional tdTomato reporter (Rosa26-LSL-tdTomato) to generate Ptf1a-CreER; Rosa26-LSL-tdTomato wild-type (WT) mice. These mice were then bred with conditional KrasG12D knock-in mice to generate Kras+/LSL-G12D; Ptf1a-CreER; Rosa26-LSL- tdTomato mice, hereafter referred to as KC mice. Wwoxfloxed mice (Wwoxf/+ or Wwoxf/f)17 were bred with KC mice to generate KWC mice (Wwox(f/f or +/f);Kras+/LSL-G12D; Ptf1a-CreER; Rosa26-LSL-tdTomato) and WC mice (Wwox(f/f)/(+/f); Ptf1a-CreER; Rosa26-LSL- tdTomato) (Fig. 1a). The models were validated by detecting tomato expression in frozen sections of the tamoxifen-induced pancreas (Fig. S1a), immunoblot analysis (Fig. S1b) and WWOX immunohistochemical staining (Fig. S1c) and immunofluorescence staining for tdTomato and amylase (Fig. S1d). Successful tamoxifen activation of Cre was assessed in mice one month after post-tamoxifen injection by validation of tomato expression using immunofluorescence and specific acinar ablation of WWOX and pERK using immunohistochemical staining (Fig. 1b).
Next, we compared the histological appearance of the pancreata between KWC and KC mice at different time points (1, 2, 3, 4, 5, and 6 months). In line with the patient data, WWOX deletion enhanced ADM lesions and PanIN formation in KWC mice compared to KC mice; lesions appeared as early as one month in KWC mice (Fig. 1c, black arrows). Moreover, some KWC mice started forming tumors as early as 4 months post tamoxifen injection, whereas none of the KC mice formed tumors at any of the experimental time points (Fig. 1c). However, no phenotypes were identified in the pancreata of WC mice, suggesting that Wwox deletion alone is insufficient for ADM and PanIN lesion formation (Fig. S1h). Quantification of ADM and low-grade PanIN lesions at one-month post-tamoxifen injection revealed a significant increase in the number of KWC mice compared to KC mice (Fig. 1d). Strikingly, no differences were observed between KWC mice harboring Wwoxf/+ and Wwoxf/f alleles. Therefore, we refer to these mice as identical throughout our study.
Furthermore, alcian blue staining revealed an increased number of ADM lesions in KWC mice compared to that in KC mice (Fig. 1e). Moreover, immunostaining of Ck-19 (a marker of ductal cells) revealed an increased number of ductal cells in KWC mice compared to KC mice (Fig. 1f-g). Additionally, a higher number of proliferative cells was observed in KWC mice than in KC mice (Fig. 1h-i). Taken together, our analysis revealed the acceleration of preneoplastic lesions in the KWC mice.
To dissect cell-autonomous versus non-cell-autonomous neoplastic lesion formation, ex vivo transdifferentiation of cells isolated from KC and KWC mice was performed. Consistent with previous results, cells isolated from KWC mice transdifferentiated faster than those isolated from KC mice (Fig. S1f-h). Moreover, tdTomato-sorted cells from mice two-months post-tamoxifen injection revealed an induction in the Runx1, Onecut2, and Foxq1 genes, which are associated with ADM reprogramming18 (Fig. S1i). These findings suggest that WWOX loss accelerates KRAS-mediated ADM and PanIN lesions by facilitating transdifferentiation of acinar cells into ductal cells.
KRAS activation induces WWOX expression.
It has been previously reported that WWOX levels are induced by stress conditions, such as DNA double-strand breaks (DSBs)19. Therefore, we determined whether WWOX levels increased following KRAS activation in normal non-transformed cells. To test this hypothesis, we induced KRAS expression both in vitro and in vivo, and monitored WWOX expression. Strikingly, when comparing WWOX protein expression in vivo in WT and KC mice at normal acinar sites, levels were found to increase following KRAS activation, as assessed by immunohistochemistry (IHC) staining at one- and two-weeks post-tamoxifen injection (Fig. 2a). Immunoblotting of WWOX in WT and KC mice one month after tamoxifen injection revealed the same results (Fig. 2b). Moreover, qRT-PCR analysis revealed an upregulation of Wwox RNA expression in KC mice from sorted tomato acinar cells at two-months post-tamoxifen injection compared to acinar cells sorted from WT mice (Fig. 2c).
These results are in contrast to those observed in patient samples, where WWOX expression has been shown to gradually decline upon PanIN formation9. To address this discrepancy, we examined WWOX protein expression in the pancreas of KC mice at different time points during lesion formation. Interestingly, we found that upon KRAS activation in KC mice, cytoplasmic WWOX expression gradually decreased with PanIN lesion progression (Fig. S2a-b). These results suggest differential expression of WWOX at short vs. extended time points, indicating a tumorigenic barrier role of WWOX upon oncogenic KRAS activation, which is elevated immediately after RAS activation and then declines with lesion progression.
To further validate these results in an in vitro system, 266-6-Tet on/off KrasG12D murine acinar cells were treated with doxycycline for 6 and 24 h. Following 6h of activation, immunoblotting of pERK (a downstream target of KRAS) and WWOX revealed mild activation of KRAS and slight induction of WWOX, respectively (Fig. 2d). Remarkably, after 24h of KRAS activation, increased levels of pERK and WWOX were observed (Fig. 2d). Moreover, qRT-PCR analysis revealed upregulation of Wwox after KRAS activation for 6 and 24h (Fig. 2e-f). Taken together, these results suggest that WWOX is induced upon KRAS activation, perhaps as a response to reduced oncogenic stress.
WWOX depletion enriches for EMT and cancer stem cells (CSCs).
To gain further insight into the molecular changes that accelerate lesion formation upon WWOX depletion, a comparative genome-wide transcriptional profiling was performed. Age-matched (two-months) WT (n = 3), KC (n = 3), and KWC (Wwoxf/+ n = 3 and Wwoxf/f n = 2) cells from tomato-sorted acinar cells were used for bulk RNA sequencing (Supplementary Table 1). A set of 612 differentially regulated genes was identified between KWC and KC mice (Supplementary Table 2). Principal component analysis (PCA) showed a clear separation among all groups (WT, KC, and KWC mice) (Fig. 3a). The differentially expressed genes were enriched for pathways known to be implicated in WWOX signaling and have implications in PDAC (Fig. 3b) (Supplementary Table 3).
Pathway analysis of the RNA-seq data revealed upregulation of TGF-β signaling upon WWOX deletion (Fig. 3b). Activation of the TGF-β signaling pathway is a common event in PDAC20. Furthermore, it has been reported that WWOX inhibits the TGF-β signaling pathway through direct interaction with SMAD314. TGF-β suppresses the growth of many cell types through its SMAD-dependent pathway. Nonetheless, one aspect in which tumors benefit from TGF-β signaling is the activation of the EMT process through SMAD-dependent or SMAD-independent pathway21. EMT, a crucial step in early embryonic morphogenesis22, can also play an important role in many types of cancer, such as pancreatic cancer23. To validate the enrichment of the EMT pathway in our KWC model, qRT-PCR for Ck-19, fibronectin, slug, and twist of independent aged-matched RNA (one-month) of KC (n = 3), and KWC mice (n = 3) revealed an upregulation of these markers in KWC mice compared to KC mice (Fig. 3c-f). Moreover, knocking out WWOX from 266-6 reduced amylase expression and enhanced Ck-19 expression (Fig. 3g-h). These findings indicate that WWOX loss, both in vitro and in vivo, induces a mesenchymal/ductal reprogramming phenotype, consistent with our previous findings (Fig. S1f-h).
Increasing evidence suggests that EMT acquisition results in cancer stem cell (CSC) characteristics that tend to invade tumors that are more resistant to therapies22. CSCs can self-renew; these cells exist in the tumor niche and account for the invasion of tumors24. Moreover, CSCs are a major cause of tumor recurrence in patients25. Interestingly, our RNA sequencing data revealed enrichment of stem cell-associated genes in KWC mice (Fig. 3i). To further validate our results, qRT-PCR analysis of CD44 levels in isolated acinar cells of age-matched (one-month) WT (n = 3), KC (n = 3), and KWC mice (n = 3) revealed upregulation of CSCs genes in KWC mice compared with KC mice (Fig. 3j). Moreover, immunochemical staining of CD44 revealed a higher number of CD44 positive cells in KWC mice than in KC mice (Fig. 3k-l). These results support a role for WWOX in the regulation of genes associated with cancer stemness.
WWOX ablation enhances EMT and CSCs phenotype through promotion of TGFβ/BMP2 signaling.
Growth factors such as TGFβ and BMPs are implicated in initiating the EMT process26,27. Gene set enrichment analysis (GSEA) of differentially expressed genes between KWC and KC mice identified hyperactivation of the TGF-β signaling pathway (Fig. 3b). To further strengthen our results, WT (n = 2), KC (n = 3), KWC Wwoxf/+ (n = 3), and KWC Wwoxf/f (n = 2) mice one-month post-tamoxifen injection were sacrificed then acinar cells were isolated based on tdTomato expression. The expression of WWOX in these mice was validated using immunoblotting (Fig. S3a). Interestingly, WWOX deletion led to an increase in pSMAD2/3, as revealed by immunoblotting in KWC cells compared to that in KC cells (Fig. 4a). Furthermore, upregulation of Angptl4, a downstream target of TGFβ and SMAD3, was observed in KWC compared to KC cells isolated from mice two months after post-tamoxifen treatment (Fig. S3b). To further validate these results, a new cohort of mice was sacrificed one-month post-tamoxifen injection, acinar cells were isolated from WT (n = 3), KC (n = 3), KWC Wwoxf/+ (n = 2), and KWC Wwoxf/f (n = 1) mice, and RNA was isolated. qRT-PCR analysis for Sperine1 and Angptl4, downstream targets of SMAD314, revealed their upregulation in KWC mice compared to KC mice (Fig. 4b-c). On the other hand, analysis of RNA-seq data for BMP2 revealed upregulation in KWC mice compared to KC mice (Fig. S3c). This finding was further substantiated by isolating Tomato positive cells from WT (n = 2), KC (n = 3), KWC Wwoxf/+ (n = 3), and KWC Wwoxf/f (n = 2) mice one-month post-tamoxifen injection and immunoblotted for BMP2 and pSMAD1/5/9 (Fig. 4d). Interestingly, WWOX deletion induced BMP2 expression and its downstream target, pSMAD 1/5/9, in KWC mice compared to KC mice (Fig. 4d). Moreover, Runx2, a downstream target of BMP228, was upregulated in KWC mice compared to KC mice one month after post-tamoxifen injection (Fig. 4e).
Several studies have described the crosstalk between ERK and SMAD signaling29,30 showing that TGF-β can induce ERK signaling through the PI3K/Pak2 pathway31. In contrast, BMPs can signal through the non-canonical pathway via different kinases32; for example, BMP4 can upregulate Dusp9 to attenuate ERK signaling in mouse embryonic stem cells33. To check the possibility that WWOX deletion affects the ERK MAPK pathway, WT (n = 2), KC (n = 3), KWC Wwoxf/+ (n = 3), and Wwoxf/f (n = 2) mice were sacrificed one month after post-tamoxifen injection, and acinar cells were isolated based on tdTomato expression and assessed by immunoblotting. Consistent with our previous results, WWOX deletion led to the induction of pERK in KWC mice compared to KC mice (Fig. S3a). Moreover, to validate the enrichment in KRAS signaling, pERK IHC staining in KC (n = 4) and KWC (n = 9) pancreata at one-month post-tamoxifen injection was performed. Strikingly, KRAS signaling was clearly enhanced in KWC mice compared to that in KC mice at this early time point (Fig. S3 d-e). To exclude the possibility that induction of KRAS signaling is due to the increased number of cells expressing KRAS because of the accelerated formation of preneoplastic lesions in KWC, immunofluorescence for pERK in KC (n = 4) and KWC (n = 5) pancreata at one-month post-tamoxifen injection was performed. Interestingly, pERK staining was more intense in KWC mice than KC mice (Fig. 4f-g). Altogether, these results imply that WWOX deletion enhances KRAS signaling through manipulation of TGF-β and BMP signaling.
WWOX physically interacts with SMAD3 and BMP2 in acinar cells.
WWOX acts as an adapter protein that regulates transcription, stability, and localization of its partners through physical interactions mediated by its WW1 domain16. For example, WWOX inhibits SMAD3 transcriptional activity via direct interaction in MCF10 cells14. To determine the occurrence of this interaction in pancreatic cells and whether WWOX interacts with the BMP family, we performed immunoprecipitation in 266-6 cells. To this end, inducible 266-6 KrasG12D murine acinar cells were activated for 24 h (+ dox), and immunoprecipitation of WWOX was performed to check possible interactions with either SMAD3 and/or BMP2 proteins. Interestingly, we found that WWOX physically interacts with both SMAD3 and BMP2 following KRAS activation (Fig. 4i). These results further imply an important role for WWOX in regulating EMT and CSCs by regulating both the SMAD3 and BMP2 pathways upon KRAS activation.
WWOX deficiency accelerates PDAC formation.
Reduction or loss of WWOX protein expression has been associated with PDAC formation9. Furthermore, ablation of WWOX accelerated the formation of preneoplastic lesions (Fig. 1). Taken together, these observations prompted us to further examine aging KC and KWC mice. To this end, mice were injected with tamoxifen at the age of one month, and then monitored and followed for another eight months. Interestingly, 16 out of 64 Wwoxf/+ or Wwoxf/f KWC mice formed tumors between the ages of 4–8 months while none of the KC mice (n = 14) formed tumors in this time range (Fig. 5a), which is consistent with the literature34,35, as activation of KRAS alone is not sufficient for the formation of neoplastic lesions at these time points. To validate that the tumors obtained from KWC mice originated from Ptf1a positive cells, we stained the tumors with tdTomato antibody and detected them by immunofluorescence. We found that the tumors were positively stained for tdTomato, confirming that the acinar cells were the cell of origin. Moreover, histological characterization (H&E) of KWC tumors revealed highly aggressive tumors (Fig. 5b). Interestingly, Wwoxf/+ KWC and Wwoxf/f KWC mice developed tumors at the same rate (Fig. 5a). Staining for WWOX protein expression in the tumors revealed that those developed in Wwoxf/+ KWC mice displayed reduced or lost WWOX expression at the tumor sites compared to the normal sites, suggesting a loss of Wwox heterozygosity at these sites (Fig. 5c). Furthermore, the tumors formed were rich in spindle-shaped cell morphology, positive for pancytokeratin markers, and negative for amylase staining, indicating a mesenchymal identity (Fig. 5c). To further validate tumor aggressiveness, we stained for pERK and pSTAT3, and found that the tumor sites were highly positive for these markers (Fig. 5c). Moreover, these tumors were highly positive for vimentin and weakly positive for E-cadherin (a marker of EMT), which may indicate their ability to metastasize (Fig. 5d). Collectively, these data suggest the aggressiveness of KWC PDAC tumors.
Enhanced progression and oncogenic signaling in KWC tumors.
It has been demonstrated that some BMPs components, such as BMP2, are upregulated in pancreatic cancer patients compared to normal pancreatic tissue36. Moreover, our results indicate an enhancement in BMP2 signaling upon WWOX ablation (Fig. 4d and S3c). IHC staining of KWC tumors for BMP2 indicated the induction of its expression compared to that in the adjacent spleen (non-metastatic) from the same mouse (Fig. 5e).
PDAC tumors have a high capacity to metastasize to the adjacent organs. Common sites of metastasis include the liver, peritoneum, spleen, and lungs37. Moreover, metastasis requires PDAC cells to switch their epithelial characterization to a mesenchymal state38, a phenomenon that was observed in KWC mice tumors (Fig. 5d). Therefore, we examined these organs in tumor-bearing KWC mice. Liver, lung, and spleen metastases were found in three of the 16 KWC tumor-bearing mice (Fig. 5f and S4c). To validate the origin of these metastatic lesions, they were stained for tdTomato and found to be positive for the tomato reporter, validating PDAC as the origin of metastatic lesions (Fig. 5f). Moreover, these metastatic lesions were negative for WWOX and highly expressed pancytokeratin, pERK, and pSTAT3, as assessed by IHC staining, further validating that PDAC tumors were the origin of these metastatic lesions (Fig. 5f). In contrast, normal KWC lungs were positive for WWOX and negative for pancytokeratin, pERK, and pSTAT3 by IHC staining (Fig. S4a).
To exclude the possibility of Ptf1a promoter leakage to distant sites, such as the liver and lung, we stained WT mice expressing tomato after post-tamoxifen injection. As expected, the liver and lung tissues (n = 3) were negative for tomatoes, whereas the pancreatic tissues were positive (Fig. S4b), suggesting that the primary tumor was the origin of metastases. Collectively, these data suggest the aggressiveness of KWC PDAC tumors and their ability to metastasize to the adjacent organs.
WWOX deficiency in PDAC patients.
WWOX expression was frequently altered in PDAC8 cells The KM plotter dataset (www.kmplot. com) to evaluate the prognostic value of WWOX expression in PDAC using the KM plotter dataset (www.kmplot.com). We found that patients with low WWOX expression displayed lower overall survival than those with high WWOX expression (Fig. 6a). To further validate the effects of WWOX using a human cell culture system, we assessed its expression in five PDX cell lines derived from advanced tumors39 (#50, 92, 114, 134, and 139). Immunoblotting of WWOX for the five PDX lines revealed a reduction in three of them relative to each other [#92, 134 and 139] (Fig. 6b). Transduction of all PDXs using a lenti-WWOX vector or empty vector (EV) was then performed to generate stable lines, which were validated by immunoblotting for WWOX (Fig. 6b-c). To validate whether WWOX restoration had any biological effects in vitro, colony formation assays were performed for all lines (WT, EV, and WWOX overexpression (OE)). Interestingly, WWOX restoration in PDX92, 134, and 139 cells harboring low WWOX expression successfully suppressed colony formation (Fig. 6d and Fig. S5a-b). In contrast, WWOX restoration in PDX50 or 114, harboring high WWOX expression had no effect on survival (Fig. S5c-d). To further examine the tumor suppressor function of WWOX, we used CRISPR/Cas9 to knock out WWOX in PDX114, which expresses high levels of WWOX, and assessed its effects on colony formation. PDX114 was transduced using either LentiCRISPR-V2 plasmid targeting WWOX exon 1 or empty vector-V2 plasmid and generated stable polyclonal lines, which were validated by immunoblotting and showed a significant reduction in WWOX levels by almost 25% in KO1 and 70% in KO2 (Fig. 6e). Interestingly, knocking out WWOX in PDX114 promoted tumorigenic traits of these cells, as evidenced by the enhanced colony formation ability (Fig. 6f). Moreover, we validated the effect of WWOX restoration on growth suppression in vivo, PDX139 WT, EV, and WWOX OE were intraperitonially injected into NOD/SCID mice. Indeed, PDX139 with WWOX overexpression resulted in smaller and lighter tumors compared to PDX139 WT and PDX139 EV (Fig. 6g-h), further confirming WWOX tumor suppressor activity. Staining of these tumors revealed higher expression of pERK in PDX139 and PDX139 EV than in PDX139 WWOX cells (Fig. S5e), which was consistent with the WWOX function observed in our mouse model (Fig. 4f-g & Fig. S3 d-f). These findings further indicated the role of WWOX in tumor suppression in human pancreatic cancer cells.