Pancreatic duct ligation reduces premalignant pancreatic lesions in a Kras model of pancreatic adenocarcinoma in mice

Pancreatic duct ligation (PDL) in the murine model has been described as an exocrine pancreatic atrophy-inducing procedure. However, its influence has scarcely been described on premalignant lesions. This study describes the histological changes of premalignant lesions and the gene expression in a well-defined model of pancreatic ductal adenocarcinoma by PDL. Selective ligation of the splenic lobe of the pancreas was performed in Ptf1a-Cre(+/ki); K-ras LSLG12Vgeo(+/ki) mice (PDL-Kras mice). Three experimental groups were evaluated: PDL group, controls and shams. The presence and number of premalignant lesions (PanIN 1–3 and Atypical Flat Lesions—AFL) in proximal (PP) and distal (DP) pancreas were studied for each group over time. Microarray analysis was performed to find differentially expressed genes (DEG) between PP and PD. Clinical human specimens after pancreaticoduodenectomy with ductal occlusion were also evaluated. PDL-Kras mice showed an intense pattern of atrophy in DP which was shrunk to a minimal portion of tissue. Mice in control and sham groups had a 7 and 10-time increase respectively of risk of high-grade PanIN 2 and 3 and AFL in their DP than PDL-Kras mice. Furthermore, PDL-Kras mice had significantly less PanIN 1 and 2 and AFL lesions in DP compared to PP. We identified 38 DEGs comparing PP and PD. Among them, several mapped to protein secretion and digestion while others such as Nupr1 have been previously associated with PanIN and PDAC. PDL in Ptf1a-Cre(+/ki); K-ras LSLG12Vgeo(+/ki) mice induces a decrease in the presence of premalignant lesions in the ligated DP. This could be a potential line of research of interest in some cancerous risk patients.

www.nature.com/scientificreports/ Interestingly, p53 mutation, assessed by nuclear overexpression of p53 protein, is a late event in the progression model of pancreatic adenocarcinoma 5,15 ( Figure S1). On the other hand, ultrastructural and immunohistochemical studies in normal and transgenic mice have demonstrated that after experimental pancreatic duct ligation (PDL), most of the acinar cells are rapidly and massively deleted by apoptosis-mediated exocrine atrophy. This process is also regulated by p53 active protein 13,[16][17][18][19] . In this regard, our group has demonstrated a complete acinar cell deletion both in short and long-term evolution with no sign of acinar regeneration after an efficient model of PDL in murine and also in pig normal pancreas 20,21 .
Taken these pieces of evidences in mind, our hypothesis is that selective PDL in Ptf1a-Cre (+/ki) ; K-ras LSLG-12Vgeo (+/ki) mice at an early stage (in the absence of p53 inactivation) can induce apoptotic-mediated exocrine atrophy, and may avoid progression to PDAC, overcoming pro-oncogenic stimulation. To study this, we examined the histological changes and the gene expression that take place in the distal and proximal pancreas after PDL.

Materials and methods
Study design. Twelve-week old Ptf1a-Cre (+/ki) ; K-ras LSLG12Vgeo (+/ki) mice referred as Kras mice and two control groups K-RasLSLG12Vgeo (+/ki) mice and Ptf1a-Cre (+/ki) mice referred to as G and Cre mice respectively 22,23 were used in this study. All animals were bred and raised until 8 Kras mice were divided in three experimental groups: non operated animals referred to as "control group" (n = 13), animals that were operated and surgically manipulated but did not undergo pancreatic ligation referred to as "sham group" (n = 8) and animals subjected to a pancreatic duct ligation referred to as "PDL group" (n = 26). Mice were sacrificed and analyzed 1, 3 and 6 months postoperatively (PO). Additionally, G (n = 12) and Cre mice (n = 11) were also subjected to PDL and were sacrificed at the same PO periods. Surgical technique. At 12 weeks of age, the peritoneal cavity was accessed through a midline laparotomy, under anesthetic induction with isoflurane at 4% and at 2.5% maintenance in 100% oxygen. The stomach, pancreas and spleen were mobilized as described previously 19,24 in all the animals subjected to surgery (sham and PDL groups). In PDL group mice, dissection and ligation of the splenic lobe of the pancreas were performed on the right side of the superior mesenteric vein, using 8/0 polyglycolic acid suture (Fig. 1). Special caution with Figure 1. Surgical ligation of splenic pancreatic lobe in Ptf1a-Cre (+/ki) ; K-ras LSLG12Vgeo (+/ki) mice. Pancreatic ligation was performed including the main splenic pancreatic duct adjacent to the superior mesenteric vein. Two pancreatic segments, proximal pancreas (PP) and distal pancreas (DP) were thus delimited and separately analyzed. www.nature.com/scientificreports/ this technique was observed with the preservation of the gastroepiploic pedicle so that any ischemic effect on the pancreas and the spleen is avoided. The pancreas and the abdominal viscera were finally relocated to their anatomical position, and the midline incision was sutured in two layers (aponeurotic plane with 4/0 absorbable continuous suture, and skin with 4/0 silk interrupted suture) in mice from the PDL and sham groups.
Postoperative care and necropsy. Animals were monitored for signs of stress and followed up for 1, 3 or 6 months. At each of these PO periods, some mice of each group were euthanatized with CO 2 inhalation to explore the peritoneal cavity. The pancreas, the spleen and a segment of the duodenum were simultaneously removed. The ligature, if present (see Fig. 2B,F), was noted and the splenic lobe was transected and referred as distal pancreas (DP) for further analysis. The other lobes (gastric and duodenal) were referred to as proximal pancreas (PP). The animals were euthanatized by inhalation of 70-90% CO 2 .
Histological study. Specimens were formalin-fixed and embedded in paraffin. Each pancreas was serially sectioned (3 µm). A total of 11,200 slides were considered and a selection of 568 Sections (8 slides per animal, 70 mice in total) was stained with hematoxylin and eosin and reviewed blindly by an expert pathologist. Every  Gene expression assays. Kras-PDL mice of different PO periods with enough quantity of atrophic distal pancreas to perform manual macroscopic dissection were selected for microarray analysis. The formalin-fixed, paraffin-embedded PP and DP of Kras-PDL mice were processed as previously described 27 . Briefly, total RNA was then isolated using the RecoverAll Total Nucleic Acid Isolation Kit (Life Technologies) from 40 serial sections with a thickness of 3 µm for each of the portions of the pancreas (PP and DP). RNA quantity and purity was determined on the ND-2000 Spectrophotometer (NanoDrop Technologies) and RNA integrity was assessed using Agilent 2100 Bioanalyzer (Agilent Technologies). As RNA integrity number (RIN) obtained from Bioanalyzer is not informative for FFPE samples, the percentage of RNA > 200nt was used as an estimation of RNA integrity for each sample.
For microarrays analysis, amplification, labeling and hybridizations were performed according to protocol GeneChip Pico Reagent kit (Affymetrix). The working conditions used were 50 ng input RNA, 12 cycles of pre-PCR and 16 h of IVT. After the sample processing it was hybridized to GeneChip Mouse Clariom S Array (Affymetrix) in a GeneChip Hybridization Oven 640. Washing and scanning were performed using the Expression Wash and Stain and the GeneChip System of Affymetrix (GeneChip Fluidics Station 450 and GeneChip Scanner 3000 7G).
Microarray data processing and gene expression analysis. Quality control was performed on the raw data CEL files corresponding to 5 paired Kras-PDL mice (DP and PP) samples with arrayQualityMetrics package 28 . Once quality checked, preprocessing of the data was executed with oligo package 29 . Normalization was performed using Robust Multi-array Average (RMA) algorithm which creates an expression matrix from Affymetrix data. The raw intensity values were background corrected, log2 transformed and then quantile normalized. Probes were annotated using annotateEset function and pd.clariom.s.mouse annotation data from the affycoretools package 30 . Next, a linear model was fit to the normalized data to obtain an expression measure for each probe set on each array. Before performing the differential gene expression analysis, the expression matrix obtained after RMA normalization was filtered by the Inter-quantile range (IQR) which provides a measure of the spread of the middle 50% of the intensity scores for each gene. Differential expression analysis was performed for paired samples using Limma 31 (Linear Models for Microarray Data) package. Normalized expression matrix was used as input for Gene Set Enrichment Analysis (GSEA). Genesets with an FDR < 0.25 were considered significant.
Human samples. Human pancreatic samples were obtained from patients who had undergone pancreatoduodenectomy (PD) with ductal occlusion of the remnant pancreas at the Parc de Salut Mar and had a primary diagnosis of PDAC or distal cholangiocarcinoma during the period 2012-2014. The protocol was approved by the Ethical Committee (CEIC-2012-130) of our institution (Hospital del Mar, Barcelona, Spain) and informed consent was obtained from all patients. All methods were performed in accordance with relevant guidelines and regulations for human subjects. In these selected cases, ductal occlusion of the main pancreatic duct of the distal remnant of the pancreas (without pancreatic anastomosis) was performed similarly the procedure described by Fromm et al. 32 These patients provided a unique opportunity to study morphologic changes of the pancreas after pancreatic duct occlusion according to the evaluation of appearance on CT 33,34 during the PO period. In addition, mutation analysis of exon 2, 3 and 4 of the KRAS gene was performed after manual or laser microdissection of precursor lesions in resected pancreas specimens.
Statistical analysis. Data are presented as mean and standard deviation (SD). Non-parametric tests (Chi-Square, U-Mann Whitney and Kruskal-Wallis) were employed for univariate analysis. For exposing purposes, premalignant lesions were evaluated independently in PanIN 1 and higher grade premalignant lesions (HGPL) including PanIN 2, PanIN 3 and AFL. Linear regression models were performed in order to determine the best-fit equation for the number of lesions during follow-up for PanIN1 and HGPL. In addition, factors found to be significant predictors in the univariate analysis were subjected to multivariate analysis (backward stepwise logistic regression analysis). Results were expressed as Odds Ratio (OR) with a 95% confidence interval (CI). Differences were considered to be significant at a threshold of p < 0.05. Statistical analyses were performed with statistical software (SPSS version 19.0; SPSS, Inc, Chicago, Ill).

Results
Model validation. All animals tolerated the surgical procedure well and recovered and ambulated promptly.
In the immediate PO period (4 days) the animals showed signs of stress and discomfort such as bristling hair or lack of grooming, as expected in a normal PO context. Later on, there were no signs of relevant disease in the PO follow-up and the Kras mice showed a normal growth curve, displaying no differences between groups.
As expected, premalignant lesions were not observed in control mice in which the conditional mutant Kras allele was not activated in the pancreas (Cre and G mice). In the Kras mice, we identified a majority of PanIN 1 lesions (n = 6290; 97%), followed in number by AFL (n = 160; 2.5%), PanIN 2 and PanIN 3 high-grade lesions (n = 32; 0.4% and n = 10; 0.1%, respectively) and frequent areas of focal lobulicentric atrophy (patchy atrophy) as already described in these models 26 . Therefore, hereinafter the results refer only to Kras mice. The time-course development of lesions is shown in Fig. 2.
Histologically, consistent with other PDL models 14,21,35 , the DP showed an intense pattern of atrophy with an increase of adipose and fibrosis over the PO periods. There also seemed to be an extension of adipose tissue www.nature.com/scientificreports/ surrounding the pancreatic tissue. More specifically, there was a complete replacement of exocrine acini by pseudoductal complexes, based on apparently novel duct-like structures composed of low cuboidal cells (Fig. 2F,K) (severe atrophy). There were also scattered infiltrates of lymphocytes and plasma cells in the fibrous boundary.
The number of PanIN 1 lesions in the ligated pancreas is reduced. All mice presented an increase of PanIN 1 lesions in the PP over time, without differences among groups (best-fit less squared function was obtained with a linear model; r = 0.57, p < 0.01). A similar increase was observed in the DP in the sham and control groups over the PO, but not in the PDL group. In fact, the number of PanIN 1 lesions in the PDL group was significantly lower in DP compared to its corresponding value in PP at each PO time being almost 24-fold lower at 6 months PO. Furthermore, the number of PanIN 1 lesion in DP was significantly lower (p < 0.01) in the PDL group at 6 months PO compared to the sham and control group (Fig. 3A). of HGPL in the PP usually increased over time in all groups and there were no significant differences between groups in any PO time. This increase was also observed in the DP of mice in the sham and control groups, which increased significantly over time following a linear model (r = 0.56, p < 0.01). Again, the number of HGPL in the PDL group was significantly lower in DP compared to its corresponding value in the PP at 3 and 6 months PO, being almost 22-fold lower at 6 months PO. Similarly to PanIN1 lesions, the number of HGPL in DP was significantly lower (p < 0.01) in the PDL group at 6 months PO compared to the sham and control group (Fig. 3B).

The risk of developing HGPL is reduced in the ligated pancreas. In the univariant and multivariant
analyses two variables demonstrated a significant influence on the presence of HGPL: postoperative time and experimental group. Thus, the increase of PO time was associated with 3 times more risk of HGPL (OR: 2.9, IC 95% 1.09-8.02, p = 0.03).
Having undergone PDL was protective for the development of HGPL: the risk of having HGPL was 6.87fold higher in the control group (IC 95% 1.21-39.17, p = 0.03) and 9.84-fold higher in the sham group (IC 95% 1.21-785.92, p = 0.028) than in the PDL group.
A statistically significant decrease in the number of HGPL and PanIN 1 was observed when severe atrophy was seen in the studied segment of the pancreas (Fig. 4). On the contrary, patchy atrophy was linked to a significant increase in the number of HGPL and PanIN 1 (Fig. 4).

Transcriptional analysis reveals downregulation of the acinar program in distal pancreas.
Differential expression analysis using Limma revealed 38 significant differentially expressed genes after p-value adjustment for multiple testing; 36 of those were downregulated and 2 upregulated in the distal portion of the ligated pancreas. The vast majority of downregulated genes corresponded to acinar enzymes or were related to the acinar function. (Table 1, Fig. 5A). Gene Set Enrichment analysis supported this downregulation of the acinar function in the distal pancreas of Kras-PDL mice. (Fig. 5B).
Human experience in the ligated pancreas. Finally, to assess whether these effects were relevant to patients, we examined the pancreas of four patients with primary tumors in whom the distal pancreas was ligated after pancreatoduodenectomy (PD) that also contained PanIN1 and HGPL (see a representative case in Fig. 6). In all cases, the distal remnant pancreas (body and tail) was characterized by an atrophic pattern based on the CT analysis as it has been described in previous studies after PD 33,34 . The mean rate reduction of the pancreatic parenchymal thickness measured at the cross-section of its body after 12 and 24 months was 25 ± 15% and 47 ± 11%, respectively. An important ductal dilatation of the main pancreatic duct was also observed. In two of the cases exon 2 mutated to p.G12D and p.G12V. These observations are highly reminiscent of the findings in mice.

Discussion
Many conventional therapies for pancreatic cancer aim to induce cellular damage to trigger apoptosis. In fact, apoptosis is an evolutionary designed method of preventing propagations of mutations or simply to eliminate unwanted or useless cells 36 . Mutated or useless cells are better dead and cell suicide through apoptosis may achieve this goal and this could be relevant in many diseases. For example, in acute pancreatitis, the extent of pancreatic acinar cell apoptosis has been shown to be inversely related to the severity of the disease and induction of apoptosis in animal models has been found to have a protective effect against acute pancreatitis 37 . In Figure 4. Types of atrophy in pancreas of Ptf1a-Cre (+/ki) ; K-ras LSLG12Vgeo (+/ki mice. When severe pancreatic atrophy (A.B) was present in the specimen (in DP of PDL-Kras mice) then PanIN 1 (C) and HGPL (D) were less frequently present than when there was no severe atrophy. On the contrary, when patchy atrophy (E,F) was observed in the specimen (in PP of control, sham and PDL-group and in DP of control and sham group) then PanIN 1 (G) and HGPL (H) were more frequently present than when patchy atrophy was absent. Bar errors including mean value and 95% confidence interval. *p < 0.05; **p < 0.01; # p = 0.059. Scale bars: 50 μm. www.nature.com/scientificreports/ other diseases like PDAC of the proximal pancreas, complete occlusion of the main pancreatic duct with the corresponding massive atrophy of the distal pancreas is frequent 38 (see Fig. 7). In that regard, it has usually been recognized that complete occlusion of the main duct of exocrine glands leads to the deletion of acinar cells by a massive and rapid apoptosis phenomenon. This has been demonstrated in the pancreas [16][17][18][19] and also in salivary glands 39 . We previously studied these phenomena with an efficient model of PDL in normal murine and normal pig pancreas in vivo 20,21 and we confirmed a massive deletion of exocrine cells through apoptosis while preserving the endocrine tissue. In the present study we triggered the same physiological mechanism but in Kras mice at an early stage (when p53 is still active) in order to wipe out these potentially pro-oncogenic cells with a long period of follow-up (up to 6 months) in comparison to several intergroup controls (control and sham group) and intragroup controls (proximal pancreas). Very recently, two reports 40,41 have described the influence of duct obstruction on the remaining distal pancreas in Kras mouse models at the cellular level but without quantitative analysis of preneoplastic lesions. Both of them described similar signs of strong atrophy and a complete loss of the acinar compartment with a ductal replacement but while Shi et al. 40 described less preneoplastic lesions in the acinar differentiation models in the distal pancreas, Cheng et al. 41 observed a higher number of proliferative non-mucinous cells of both acinar and ductal origins at 2 weeks after PDL, without severe nuclear atypia or tumoral infiltration after 5-month of follow up. Table 1. Gene name, fold change and p-value of differentially expressed genes between PP and DP following PDL in Ptf1a-Cre (+/ki) ; K-ras LSLG12Vgeo (+/ki mice. 38 genes were selected with a p-value < 0.05. www.nature.com/scientificreports/ In our study, after a long follow-up (up to 6 months) and an exhaustive quantitative analysis, both the global evaluation of premalignant lesions (PanIN 1 and HGPL) and the risk of the apparition of them in a single animal were found to be dramatically lower in DP than in PP (more than 20 times lower) in PDL group and significantly lower than in sham and control groups.
Interestingly, animals with PDL experienced a marked reduction of the DP (severe atrophy) with preservation of the proximal one and this phenomenon was linked to a lower number of PanIN 1 and HPGL. In fact, the signs of severe atrophy of the distal pancreas after ligation were already evident through gross examination of the specimens at 1, 3 and 6 months PO (see the small remaining tissue of DP in Fig. 2B) and matched well with the corresponding microscopic examination of a complete acinar dropout. A similar atrophic pattern was seen spontaneously after occlusion of the main pancreatic duct by a proximal tumor in patients 38 (Fig. 7) or over time in a very selective group of patients in which the remaining distal pancreas was ligated after pancreaticoduodenectomy based on the CT analysis (Fig. 6).
In the gene expression analysis of the animals, an expected reduction in pancreatic secretion was evident in DP samples of the PDL group. Interestingly, some top genes previously related to the development of preneoplastic lesions or PDAC had been upregulated in PP compared to DP like Sycn 42 , Dmbt1 43 , Klk1 44 and mainly Nupr1. This last gene cooperates with KrasG12D to induce PanIN formation and pancreatic ductal adenocarcinoma development in mice [45][46][47] (Fig. 5).
Our results are consistent with previous literature in which patchy atrophy accompanied by areas of acinoductal metaplasia was constantly described in Kras models and usually linked to premalignant lesions 6,25,48,49 . Patchy atrophy has also been related to the presence of PanIN lesions in patients with familial predisposition to PDAC 6,14 . Acinar to ductal metaplasia has been associated with PanIN lesions and many acini in the lobules of lobulicentric atrophy described earlier are often characterized by prominent acinar-ductal metaplasia 14,26 . However, it has been serendipitously stated that when the atrophy is severe (> 90% loss of acini and ducts) is unusual to see PanIN lesions at least in the resected human pancreas without adenocarcinoma 50 . These findings are also in agreement with the currently accepted origin of PanIN lesions: observations indicate that the ductal-like cells present in PanINs rarely arise by the transformation of normal duct cells. Instead, PanINs may originate either www.nature.com/scientificreports/ by transdifferentiation of acinar cells or by misdifferentiation of their precursors 23,51,52 . Therefore, the transformation to PanIN lesions of these acinar cells or their precursors could be interrupted when an early obstruction of the main pancreatic duct is performed even when Kras pro-oncogenic stimulation is present. This could be a potential research line of interest in some cancerous risk patients taken into account that: (1) Complete occlusion of the main pancreatic duct could trigger a physiological protective effect to avoid malignancies in pancreas harboring Kras mutations without the aid of any other potentially dangerous treatments as it has been seen before; (2) This procedure could be feasible both surgically and endoscopically 53 ; (3) Kras mutations are believed to be a common initiator of mucinous lesions and other cancer precursor lesions 54 and (4) More than 90% of pancreatic adenocarcinomas harbor activating mutations in Kras for which targeted therapies have been notoriously ineffective across all cancer types prompting the perception that RAS is undruggable 54 . However, some limitations of this research must also be addressed: All the experiences were done in mice models in which early pancreatic lesions are evaluated after duct obstruction. Therefore, the final effect of duct obstruction on established PDAC especially in humans are beyond the scope of this study and could be detrimental for its evolution.
In conclusion, Pancreatic duct ligation in Ptf1a-Cre (+/ki) ; K-ras LSLG12Vgeo (+/ki) mice induces a decrease in the presence of premalignant lesions in the ligated distal pancreas. This could be a potential line of research of interest in some cancerous risk patients.