Loss of tumor suppressor WWOX enhances RAS activity in pancreatic cancer

Pancreatic cancer is one of the most lethal cancers, due to late diagnoses and chemotherapy resistance. The tumor suppressor WWOX, spanning one of the most active common fragile sites in the human genome (FRA16D), is commonly altered in pancreatic cancer. However, the direct contribution of WWOX loss to pancreatic cancer development and progression is largely unknown. Here, we report that combined conditional deletion of Wwox and activation of KRasG12D in Ptf1a-CreERexpressing mice resulted in accelerated formation of precursor lesions and pancreatic carcinoma. In addition, these mice displayed enhanced MAPK and IL6/Jak/Stat3 signalling, combined with higher rates of macrophage infiltration and inflammation signature at sites of acinar to ductal metaplasia (ADM). Moreover, accumulated DNA double strand breaks were observed at ADM lesions. Finally, overexpression of WWOX in patient-derived xenografts led to reduction in KRAS MAPK activity and IL6/Jak/Stat3 signalling so diminishing their aggressiveness and in vivo tumor growth. Our data underscore the important role of WWOX in regulating RAS activity in pancreatic carcinogenesis.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) is currently the fourth leading cause of cancer death in the modern world, with a five year survival rate of 9% of the patients (Siegel et al., 2019). Approximately, 15-20% of patients are eligible for surgical resection, the rest of the patients are inoperable or have advanced metastatic disease with minimal benefit to chemotherapy treatment (Ryan et al., 2014). PDAC is believed to develop in a gradual manner over time by accumulating mutations in a number of oncogenes and tumor suppressors (Hruban et al., 2000). Constant activation of oncogenic KRAS is one of the most frequent mutations found in human intraepithelial neoplasm (PanINs) and PDAC. For example, Kras-G12D was found to be a driver mutation in pancreatic mouse models needed for PanIN initiation, maintenance, and progression to PDAC (MP & CD, 2013). Excessive activation of KRAS has a wide range of outcomes in pancreatic acinar cells including cell death, metaplasia, fibrosis development resembling chronic pancreatitis, and PDAC development (Logsdon & Ji, 2009). Nonetheless, mutations in tumor suppressor genes are also necessary for the progression of PanINs to PDAC, such as mutations in CDKN2A, TP53 and SMAD4 (J et al., 2001) , (Moskaluk et al., 1997) , (Wilentz et al., 1998) , (Oncology, 2002). Despite identifying a plethora of driver genes in PDAC, its prognosis is still the worst among human cancers (Jemal et al., 2010) , (Siegel et al., 2015). Therefore, an in-depth understanding of the molecular mechanisms of additional genes associated with PDAC formation and identifying the molecular targets and pathways involved in this process would enhance our understanding of PDAC initiation and progression.
Recent studies using genome wide exome sequencing has identified new mutations and alterations contributing to PDAC development (Waddell et al., 2015). Among these mutations, deletion of the WW domain-containing oxidoreductase (WWOX) gene, spanning one of the most active common fragile sites (CFSs) involved in cancer, has been identified (Bednarek et al., 2001). WWOX has been mapped to the unstable PDAC subtype, which is characterized by genomic instability and defects in DNA maintenance (Waddell et al., 2015). Importantly, PDAC tumors harbour reduction or, even loss of WWOX protein expression. In fact, all pancreatic cancer cell lines and 40% of primary tumors exhibit a significant reduction of WWOX protein expression (Kuroki et al., 2004). Interestingly, WWOX expression has been shown to gradually dwindle with progression of preneoplastic lesions (Nakayama et al., 2008). In addition to genomic aberrations, WWOX promoter hypermethylation is evident in some pancreatic cancer cell lines and primary pancreatic adenocarcinoma cases (Kuroki et al., 2004). Furthermore, restoration of WWOX protein expression into pancreatic cancer cell lines induces apoptosis and suppresses their tumorigenicity both in vitro and in vivo (Kuroki et al., 2004) , (Nakayama et al., 2008). More recently, it has been reported that PDAC patients harbouring a single nucleotide polymorphism at WWOX rs11644322 G>A display worse prognosis and reduced gemcitabine sensitivity (Schirmer et al., 2016). Despite these numerous studies, the direct role of WWOX in PDAC initiation and progression is largely unknown.
WWOX is an adaptor protein harbouring two WW domains that interact with prolinerich motifs of its partner protein, regulating its stability and functions (Tanna & Aqeilan, 2018). An impaired DNA damage response (DDR) is one of the hallmarks of pancreatic cancer (Macgregor-Das & Iacobuzio-Donahue, 2013). We recently demonstrated that WWOX enhances efficient DDR and DNA double strand break (DSB) repair through interacting and regulating the activity of ataxia telangiectasiamutated (ATM), a major DNA damage checkpoint protein (Abu-Odeh et al., 2014), which is also implicated in PDAC (Perkhofer et al., 2017). Regardless, WWOX's role in maintaining genome stability in pancreatic cancer is largely unknown.
In this study, we provide evidence that WWOX ablation leads to acceleration in acinar to ductal metaplasias (ADM), PanIN lesions and PDAC development. This enhancement takes place due to hyperactivation of Kras, resulting in inflammatory signalling and enhanced genome instability. Moreover, restoration of WWOX in patient-derived xenografts (PDXs) of PDAC suppressed tumor growth. Our findings indicate WWOX as a bone fide tumor suppressor in PDAC regulating RAS activity.
We next monitored and carefully analysed the pancreas in adult WC, KC and KWC mice at different time points. H&E stain of pancreata of these mice revealed enhanced ADMs and PanINs formation in KWC mice compared to KC mice; lesions appeared as early as one month in the KWC mice (Fig. 1c). No phenotypes were identified in pancreata of WC mice suggesting that Wwox deletion alone is not sufficient for ADM and PanIN lesions (Fig. S1d). Quantification of ADM and low grade PainIN lesions at one month post-tamoxifen injection revealed a significantly increased number in KWC mice compared to KC mice (Fig.1d). No differences were observed between KWC mice harbouring Wwoxf/+ or Wwoxf/f alleles. Moreover, tdTomato sorted cells from mice two months post-tamoxifen injection revealed an induction in Runx1, Onecut2, and Foxq1, genes shown to be associated with ADM reprogramming (Fig. S1f) (Schlesinger et al., 2020). In addition, there was an increased rate of proliferation at these lesions as assessed by Ki-67 (Fig. 1h, i). Remarkably, some lesions progressed with time to form PDAC tumors in the KWC but not in the KC mice (Fig. 1c). These findings suggest that WWOX loss accelerates Kras-mediated ADM and PanIN lesions and promotes progression to PDAC formation.

WWOX deletion accelerates PDAC formation.
Our previous observations prompted us to further follow aging KC and KWC mice to monitor PDAC formation. To this end, mice were injected with tamoxifen at the age of one month, then followed for another eight months. Interestingly, 16 out of 64 Wwoxf/+ or Wwoxf/f KWC mice formed tumors between the ages 4-8 months while none of the KC mice (n=14) did in this time range (Fig. 2a). Histological characterization of KWC tumors revealed aggressive and highly proliferative tumors, with tomato expression confirming acinar cells as the origin of the tumors (Fig. 2b). Intriguingly, Wwoxf/+ KWC mice developed pancreatic tumors at the same rate as Wwoxf/f KWC mice. We therefore stained for WWOX protein expression in tumors that developed in Wwoxf/+ KWC mice and found loss of WWOX staining at the tumor site, suggesting loss of Wwox heterozygosity (Fig. 2c). The tumors formed were rich in spindle-shaped cell morphology, positive for pancytokeratin marker and negative for amylase staining, indicating mesenchymal identity (Fig. 2c). In addition, the tumors were highly positive for pERK, pStat3, and EMT markers confirming their aggressiveness (Fig. 2c, e). Liver, lung, and spleen metastases were found in 3 out of the 16 KWC tumor-bearing mice.
These metastatic lesions were positive for tomato reporter, negative for WWOX IHC and highly expressed pancytokeratin, pERK, and pStat3 suggesting they were of PDAC origin (Fig. 2d). To exclude the possibility of Ptf1a promoter leakage to distant sites such as liver and lung, we stained for tomato in WT mice expressing tomato posttamoxifen injection. Liver and lung tissues (n=3) were negative for tomato whereas pancreatic tissues were positive, suggesting that the origin of the metastases is the primary tumors, and they were not due to leakage of Ptf1a expression (Fig. S2b).
Altogether, our findings indicate acceleration of Kras-mediated PDAC carcinoma formation upon deletion of Wwox.

WWOX accelerates formation of lesions by regulating Kras signalling.
To shed light on the molecular mechanism of WWOX in accelerating lesion formation, total RNA from 3 WT mice, 3 KC mice and 5 KWC mice (3 Wwoxf/+ and 2 Wwoxf/f) at two months post-tamoxifen injection was extracted from tomato-sorted acinar cells and RNA sequencing (RNA-seq) was performed. Bioinformatic analysis revealed 1381 genes that were differentially expressed in KWC and KC mice (Supplementary Table   1 One of the main pathways activated by oncogenic Ras in early precursor lesions is the Il6/Jak/Stat3 signaling pathway (Gruber et al., 2016). Staining for pStat3 (Y705) revealed elevated expression of pStat3 in ADM lesions in KWC mice compared to KC mice. In addition, RNA-seq data revealed downregulation of Socs3, a negative regulator of Stat3, in KWC mice relative to KC mice (Fig. S3a). A major outcome of Stat3 activation is inflammation, by promoting the expression of crucial inflammatory genes including Il6, Il10, Il11, Il17, Il23, Cxcl12, and Cox-2 (Yu et al., 2009) , (Sikka et al., 2014). We therefore tested for COX-2 protein expression as an inflammatory downstream effector of Stat3 and found an increased number of cells positively stained for COX2 in the ADM lesions of KWC pancreata (Fig. 3g, h, c). Furthermore, RNA-seq data demonstrated an induction of Cox2 and Il6 transcripts in KWC mice relative to KC mice (Fig. S3c, d). As macrophages can activate Stat3 signalling to promote Il6 secretion and hence trans-signal to activate Stat3 in pancreatic epithelial cells (Lesina et al., 2011), we performed IHC staining with F4/80, a surrogate marker of macrophages, and confirmed enrichment in the number of macrophages in KWC mice compared to KC mice (Fig. 3i, j). Taking these findings together, we conclude that WWOX ablation contributes to lesion progression through regulating hyperactivation of Kras signalling, which in turn activates the Il6/Jak/Stat3 inflammatory pathway.

WWOX ablation enhances DSBs and genome instability.
Oncogene activation such as KrasG12D enhances genome instability by inducing replication stress (Kotsantis et al., 2018). Moreover, perturbation in genes responsible for the DDR such as WWOX accelerates accumulation of DSBs and genome instability (Abu-Odeh et al., 2015) , (Abu-Odeh et al., 2014). Therefore, we hypothesised that WWOX ablation in our model could contribute to lesion progression through accumulation of DSBs. To test our hypothesis, pancreata from mice of the different genotypes one and two months post-tamoxifen injection were stained for γH2AX and 53BP1, surrogate markers of DSBs. As expected, WT mice had low foci count of DSBs with an average of ~0.4% and 0.6% foci/nuclei for γH2AX and 53BP1 respectively, at one and two-months post-tamoxifen injection . On the other hand, a clear accumulation of DNA damage in ADM lesions of KC and even more in KWC mice was observed ( Fig. 4a-d, Fig. S4a-d). These results suggest that WWOX ablation is critical to ADM lesions and that its loss promotes enhanced DNA damage.

KRAS activation induces WWOX expression.
The preceding observations suggest that WWOX loss enhances KRAS signalling, but does KRAS activation modulate WWOX levels or function? To address this, we examined WWOX protein expression in pancreata of KC mice. Upon Kras activation in KC mice, cytoplasmic expression of WWOX was gradually lost with lesion progression (Fig. 5a), as revealed by IHC staining. These findings were consistent with a previous observation of gradual loss of WWOX expression in human pancreatic tumor samples (Nakayama et al., 2008). As WWOX's critical function is linked to early ADM lesions (Fig. 1, 3 & 4), we determined WWOX expression upon Kras activation in early lesions. To this end, we stained for WWOX in pancreata of WT and KC mice at 1, 2, 4, and 8 weeks post-tamoxifen injection. Strikingly, WWOX protein expression was increased in acinar cells upon Kras activation compared to WT mice (Fig. 5b, c).
In addition, qRT-PCR analysis revealed an upregulation of Wwox RNA expression in KC mice from sorted tomato acinar cells at 8 weeks post-tamoxifen injection compared to acinar cells sorted from WT mice (Fig. 5d).
To further confirm the effect of Kras activation on WWOX levels using a different system, we utilised the 266-6-Tet on/off KrasG12D murine acinar cells. Treatment of these cells with doxycycline for 6, 24, and 48 h increased levels of pErk, as revealed by western blot (Fig. 5e). Importantly, Kras activity was accompanied by induction of Wwox protein levels and upregulation of mRNA expression (Fig. 5e, g). We conclude that Kras activation initially results in induction of WWOX levels followed by its downregulation and likely progression to advanced lesions.

Overexpression of WWOX in pancreatic cancer patient-derived xenografts (PDXs) suppresses their tumorigenicity.
To further validate the effects of WWOX using a human system, we assessed its expression in five PDX lines derived from advanced tumors (50, 92, 114, 134, and 139). As seen in Fig 6a, WWOX expression was found to be reduced in three out of the five PDXs [92, 134 and 139]. Interestingly, p-ERK amount was prominent in PDX139. We next transduced all PDXs using a lenti-WWOX vector or empty vector (EV) and generated stable lines, which were validated by immunoblotting (Fig 6a, b).
Interestingly, WWOX restoration in PDX92, 134, and 139, harbouring low WWOX expression, successfully reduced pERK and pSTAT3 expression. To validate the biological effect of WWOX restoration in vitro, we performed colony formation assays for PDX50, 92, 134, and 139 (WT, EV, and WWOX overexpression (OE)). We found that WWOX restoration in PDX92, 134, 139, successfully suppressed colony formation ( Fig. 6c, S6a,d). In contrast, WWOX overexpression had no effect on PDX50 and PDX114 (Fig. S6d). Next, we intraperitonially injected PDX139 EV and OE into NOD/SCID mice to reveal whether WWOX restoration has any effect on in vivo tumor formation. Indeed, PDX139 with WWOX overexpression resulted in smaller and lighter tumors compared to PDX139, and PDX139 EV (Fig. 6d, e). Staining tumors of PDX139, EV and OE, revealed higher expression of pERK and pStat3 in PDX139, and PDX139 EV compared to PDX139 WWOX OE, further confirming WWOX's function in Kras and Stat3 signalling (Fig. 6F). Finally, restoration of WWOX in PDX139 successfully reduced the number of γH2AX foci/nuclei to 29% compared to 74.7%, and 71.3%, respectively in PDX139 and PDX139 EV (Fig S6b, c). These findings further imply an important role of WWOX in pancreatic tumor development.

Discussion
Pancreatic cancer patients have the lowest survival rates among cancer patients, mainly due to late diagnoses and limited benefit of standard of care treatments (Jemal et al., 2010) , (Siegel et al., 2015).Therefore, understanding the molecular changes and pathways contributing to PDAC is critical to improve survival rates. In previous studies, WWOX has been reported to be lost in pancreatic cancer cell lines, either by deletion or promoter hypermethylation (Kuroki et al., 2004) , (R et al., 2010) , . In addition, restoration of WWOX expression to these cell lines supresses cell growth both in vitro and in vivo and induces apoptosis (N. S et al., 2008). In our current study, we found that somatic WWOX loss combined with oncogenic activation (RasG12D)  Recently, a number of specific genes have been proposed to be associated with the ADM process in pancreatic carcinogenesis including Onecut2 and Foxq1 (Schlesinger et al., 2020). Interestingly, these genes were upregulated in tomato-positive cells isolated from KWC mice, as revealed by RNA-seq data.
Little is known about WWOX's contribution to pancreatic cancer at the molecular level.
One study suggested WWOX to suppress pancreatic cancer through upregulating  (Hamarsheh et al., 2020). Patients with history of familial pancreatitis, involving germline mutation in genes such as PRSS1 and SPINK1, have 69-fold higher probability to develop PDAC (Kitajima et al., 2016). RAS can induce inflammation through enhancing secretion of cytokines, which will induce infiltration of myeloid cells and particularly macrophages that secret IL-6 to activate Stat3 inflammatory signature (Lesina et al., 2011). The effect of WWOX deletion on enhanced MAPK signalling recapitulates these events in KWC mice leading to higher infiltration of macrophages, which is accompanied by an increased inflammatory response, manifested by increased Cox2 and Il6/Stat3 signalling. As WWOX loss has been shown to enhance IL6/Stat3 signalling in triple-negative breast cancer (TNBC) cells (Chang et al., 2018), we assume that WWOX has an important function in regulating the inflammatory stress response.
Characterization of our new mouse model also revealed an in vivo function of WWOX in regulating DSB repair and genome stability. Continuous activation of oncogenes, such as KRAS, does not only regulate cell cycle entry but can also induce replication stress through stalling and collapse of the replication fork, leading to DSBs formation at preneoplastic and carcinoma sites (Gorgoulis et al., 2005) , (Bartkova et al., 2006) , (Halazonetis et al., 2008). Continuous accumulation of DSBs, present in an enormous number of tumors, may contribute to genome instability (Halazonetis et al., 2008). Upon replication stress, CFSs are expressed in metaphase chromosomes (L. S & X, 2020). Since WWOX spans one of the most active fragile sites FRA16D (Bednarek et al., 2001), it is possible that WWOX can be lost. Paradoxically, it has been shown that products of CFSs can play a direct role in the DDR (RI et al.  (Schrock et al., 2017). Our current in vivo data indeed reveal that WWOX levels are induced upon RAS activation, likely as a stress response to oncogene activation. Importantly, deletion of WWOX is associated with accumulation of DSBs at sites of preneoplastic lesions mainly at ADM sites. Increased WWOX expression was also recapitulated in an in vitro system (266-6 cells) upon activation of RAS and in human PDX lines. These findings imply that WWOX is involved in regulating genome stability in PDAC formation.
We suggest a model (Fig. 7) in which continuous RAS activation, an early event in PDAC, will lead to an unstable genome. As a result, WWOX expression, among that of other proteins, is induced to act as a barrier to effectively activate the DNA damage protein checkpoints and to repair the damage. Gradual loss of WWOX, possibly due to its genomic localization at FRA16D, could facilitate formation of preneoplastic lesions as indicated in our results. Once the WWOX barrier is lost in preneoplastic lesions, additional DSBs accumulate enhancing the oncogenic activity of RAS and its inflammatory signature. In essence, our data leads us to propose that deletion of WWOX synergizes with RAS activation to accelerate formation of the preneoplastic lesions, tumor development and metastasis. These findings further underscore the significance of tumor suppressor genes spanning common fragile sites and their active roles in carcinogenesis.  (Golan et al., 2021) were cultured in DMEM (Gibco) supplemented with 10% FBS, 1% glutamine and 1% penicillin/streptomycin. All cells were grown in 37°C with 5% CO2.

Colony formation assay
500 cells from each PDX cell line were plated in triplicates in 60 mm 2 plates for 10-14 days; cells then were fixed by 70% ethanol, stained with Giemsa, and counted.

Histology and Immunohistochemistry
Tissues were fixed with 4% formalin, followed by paraffin embedding and sectioning then stained with haematoxylin and eosin for histology. For immunohistochemistry (IHC), paraffin embedded tissues were deparaffinized followed by antigen retrieval with 25 mM citrate buffer pH 6 in a pressure cooker. Then the sections were left to cool for 25 min, followed by blocking of the endogenous peroxidase with 3% H2O2 for 15 min. To reduce non-specific binding of the primary antibody, tissues were blocked with blocking solution (CAS Block) (Invitrogen) followed by incubation of the primary antibody overnight at 4°C. Sections were washed with TBST followed by incubation of the secondary anti-rabbit or anti-mouse immunoglobin antibody for 30 min. The reaction was then performed using a DAB peroxidase kit (Vector laboratories: SK-4100), followed by haematoxylin stain for 40 s as a counter stain. IHC stains were manually counted, and for each slide at least five pictures were taken, and the average was calculated for each slide.

Immunofluorescence
Paraffin embedded sections were used to perform immunofluorescence (IF) staining.
Deparaffinization and antigen retrieval was performed as described previously for IHC.
Sections were then blocked for one hour using a blocking solution containing 5% normal goat serum, 0.5% BSA, and 0.1% Triton X-100 in PBS (Biological Industries).
Sections were then incubated with primary antibody at 4°C overnight. On the following day, sections were washed three times with PBS and then incubated with the secondary antibody and Hoechst solution, diluted in the blocking buffer, for one hour.
Slides were washed three times with PBS, dried, and mounted with coverslips using fluorescence mounting medium (Dako: S3023). IF stains were manually counted, and for each slide at least three pictures were taken, and the average was calculated for each slide.

Isolation of acinar cells and sorting
Mice were euthanized by CO2, followed by pancreas removal and washing with cold HBSS (Biological Industries). The pancreas was then cut into 1-2 mm pieces in cold HBSS and centrifuged at 600 rpm for 5 min at 4℃. For the first dissociation step, the pieces were resuspended with 0.05% trypsin-EDTA (solution C, Biological Industries) at final concentration of 0.02% for 15 min shaking at 37℃, followed by washing with 10% FBS/DMEM to stop the reaction. Then cells were centrifuged at 600 rpm followed by washing with 4% bovine serum albumin (BSA) (Mpbio) prepared in HBSS containing 0.1 mg/ml DNase Ι (Sigma). For the second dissociation step, cells were resuspended with 4% BSA containing 0.1 mg/ml DNase Ι, and 10 mg/ml collagenase P (Sigma) at final concentration of 1 mg/ml of collagenase P. Cells were incubated for 20-30 min at 37℃ with shaking, and they were pipetted up and down 10 times every 5 minutes of the incubation. Cells were then passed through a 100 m cell strainer (Corning) and washed with 4% BSA prepared in HBSS containing 0.1 mg/ml DNase Ι. The samples were then treated with red blood cell lysing buffer (Sigma Aldrich) and passed through a 40 m cell strainer (Corning) and submitted for tdTomato sorting using BD FACSAria ΙΙΙ or Sony SH800.

RNA extraction, cDNA preparation and qRT-PCR
TRI reagent (Biolab) was used as described by the manufacturer to prepare total RNA.
To prepare cDNA, 1 ug of RNA was used using the QScript cDNA synthesis kit (Quantabio). SYBR Green PCR Master Mix (Applied Biosystems) was used to perform the qRT-PCR. All measurements were done in triplicates and normalized to the levels of Ubc or Hprt genes.

RNA-sequencing
3 WT, 3 KC, and 5 KWC mice were euthanized by CO2, pancreata were disassociated and tdTomato acinar cells were sorted using BD FACSAria ΙΙΙ. Preparation of the libraries and sequencing were performed by the Genomic Applications Laboratory in the Hebrew University's Core Research Facility using standard protocol. RNA quality was checked using tape-station (Agilent Technologies), followed by poly-A clean-up and cDNA preparation. For preparing the cDNA, 1 g of RNA from each sample was processed using KAPA Stranded mRNA-Seq Kit with mRNA Capture Beads (Kapa Biosystems). Then sequencing was performed using NextSeq 500 (Illumina). For library quality control, measures were applied to raw FASTQ files, trimming low quality bases (Q>20), residual adapters and filtering out short reads (L>20). These measures were performed with the Trim-Galore (v0.4.1) wrapper tool. File assessment before and after quality control was performed with FastQC (v0.11.5). Mapping the mouse transcriptome was done using Salmon v0.13.1. Differential gene expression for the 11 samples was explored by R package (v1.16.1) and analysed with DESeq2 (v1.28.1).        Continuous activation of K-ras will eventually cause the escape of some DNA damaged cells, which will secret proinflammatory factors in order to recruit the immune system (mainly macrophages) and in return they will signal back IL6 to these cells leading to the activation of pSTAT3 and eventually the formation of ADM lesions.

Figure Legends
Lower panel-Deletion of WWOX under oncogenic activation of K-ras will accelerate the accumulation of DNA damaged cells and the formation of ADM lesions.           Figure S5