Article Text

Concurrent PEDF deficiency and Kras mutation induce invasive pancreatic cancer and adipose-rich stroma in mice
  1. Paul J Grippo1,
  2. Philip S Fitchev2,
  3. David J Bentrem1,
  4. Laleh G Melstrom1,
  5. Surabhi Dangi-Garimella3,
  6. Seth B Krantz1,
  7. Michael J Heiferman1,
  8. Chuhan Chung4,
  9. Kevin Adrian1,
  10. Mona L Cornwell2,
  11. Jan B Flesche5,
  12. Sambasiva M Rao6,
  13. Mark S Talamonti2,
  14. Hidayatullah G Munshi3,
  15. Susan E Crawford2
  1. 1Department of Surgery, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
  2. 2Department of Surgery, NorthShore University Research Institute, Evanston, Illinois, USA
  3. 3Department of Medicine, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
  4. 4Section of Digestive Diseases, Department of Medicine, Yale University School of Medicine, West Haven, Connecticut, USA
  5. 5The Medical School at Ruprecht-Karls-Universität Heidelberg, Heidelberg, Germany
  6. 6Department of Pathology, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
  1. Correspondence to Dr Paul Grippo, Robert H Lurie Comprehensive Cancer Center, Northwestern University, 303 E. Superior St., Lurie 3-105, Chicago, IL 60611; p-grippo{at}


Background and aims Pigment epithelium-derived factor (PEDF), a non-inhibitory SERPIN with potent antiangiogenic activity, has been recently implicated in metabolism and adipogenesis, both of which are known to influence pancreatic cancer progression. Increased pancreatic fat in human pancreatic tumour correlates with greater tumour dissemination while PEDF deficiency in mice promotes pancreatic hyperplasia and visceral obesity. Oncogenic Ras, the most common mutation in pancreatic ductal adenocarcinoma (PDAC), has similarly been shown to promote adipogenesis and premalignant lesions.

Methods In order to determine whether concurrent loss of PEDF is sufficient to promote adipogenesis and tumorigenesis in the pancreas, the authors ablated PEDF in an EL-KrasG12D mouse model of non-invasive cystic papillary neoplasms.

Results EL-KrasG12D/PEDF deficient mice developed invasive PDAC associated with enhanced matrix metalloproteinase (MMP)-2 and MMP-9 expression and increased peripancreatic fat with adipocyte hypertrophy and intrapancreatic adipocyte infiltration (pancreatic steatosis). In support of increased adipogenesis, the stroma of the pancreas of EL-KrasG12D/PEDF deficient mice demonstrated higher tissue levels of two lipid droplet associated proteins, tail-interacting protein 47 (TIP47, perilipin 3) and adipose differentiation-related protein (ADRP, Pperilipin 2), while adipose triglyceride lipase, a key factor in lipolysis, was decreased. In patients with PDAC, both tissue and serum levels of PEDF were decreased, stromal TIP47 expression was higher and the tissue VEGF to PEDF ratio was increased (p<0.05).

Conclusions These data highlight the importance of lipid metabolism in the tumour microenvironment and identify PEDF as a critical negative regulator of both adiposity and tumour invasion in the pancreas.

  • Pancreatic cancer
  • PEDF
  • adipogenesis
  • TIP47
  • angiogenesis
  • pancreas
  • pancreatic disease
  • pancreatic fibrosis
  • pancreatic tumours
  • abdominal surgery
  • surgical oncology
  • cancer
  • endothelial cells
  • fatty liver
  • hepatocellular carcinoma
  • matrix metalloproteinase
View Full Text

Statistics from

Significance of this study

What is already known on this subject?

  • More than half of all pancreatic cancers have reduced or undetectable levels of pigment epithelium-derived factor (PEDF).

  • PEDF expression has been correlated with reduced liver metastases and improved prognosis.

  • PEDF is known as an antiangiogenic factor but also has implications in regulating adipogenesis.

  • Increased adipocyte density in pancreatic cancer is linked to aggressive tumour progression.

What are the new findings?

  • Confirmed reduced PEDF in human pancreatic cancer shows an inverse correlation with vascular endothelial growth factor.

  • Loss of PEDF in a murine model of mutant Kras-induced pancreatic neoplasia results in pancreatic ductal adenocarcinoma.

  • Mutant Kras/PEDF null mice have increased pancreatic adiposity and adipose marker expression.

  • Mutation of Kras alone decreases PEDF and adipose triglyceride lipase.

  • Mutant Kras/PEDF null mice have increased levels/activity of matrix metalloproteinase (MMP)-2 and/or MMP-9.

  • Knockdown of PEDF in pancreatic cell lines increases cellular motility.

How might it impact on clinical practice in the foreseeable future?

  • Restoration of PEDF has the potential to re-establish more normal regulation of adipogenesis and halt tumour progression, thereby serving as a potential therapy for pancreatic cancer.


Obesity, together with the associated metabolic syndrome, carries an increased risk for developing malignancy including pancreatic ductal adenocarcinoma (PDAC), the fourth leading cause of cancer-related death in the USA.1 ,2 Interestingly, nearly all the components of metabolic syndrome such as glucose intolerance, insulin resistance, diabetes and visceral adiposity are individually recognised as risk factors for PDAC.3 Metabolic syndrome can alter the local microenvironment of the pancreas by increasing adjacent fat density,2 a feature associated with tumour progression in both experimental models and human pancreatic cancer studies.4 ,5 These studies suggest that an adipose-rich environment may fuel tumour growth and progression; however, the mechanisms underlying the link between obesity and PDAC are unclear.

Obesity is characterised by increased adipogenesis and adipocyte hypertrophy with excessive storage of fat in the lipid droplets of adipocytes. It is also accompanied by ectopic accumulation of fat in lipid droplets (steatosis) within other tissues such as the liver.6 Lipid droplets are composed of a triglyceride-rich core surrounded by a phospholipid monolayer coated by lipid droplet-associated proteins, also known as PAT proteins, which include perilipin 1, adipose differentiation-related protein (ADRP) and tail-interacting protein 47 (TIP47).7 ,8 Whereas expression of perilipin is largely confined to adipocytes and steroidogenic tissues,9 expression of ADRP and TIP47 is ubiquitous.10 PAT proteins positively regulate adipogenesis by limiting the access of lipolytic enzymes, including adipose triglyceride lipase (ATGL) and hormone sensitive lipase, to their substrates in the lipid droplet core.7

Pigment epithelium-derived factor (PEDF) was originally recognised for having bioactivity in the eye,11 but its role has expanded to include many organ systems.12 It is a 50 kDa non-inhibitory SERPIN with potent antiangiogenic, neurotrophic and antiproliferative activities.13 ,14 PEDF circulates at relatively high levels in the sera of normal individuals.15 PEDF has recently emerged as an important factor in lipid metabolism. In metabolic syndrome, PEDF levels can be increased, presumably as a counterbalance to the adverse effects of glucose intolerance and insulin resistance.15 ,16 Although, one study using murine models of obesity suggests that PEDF contributes to obesity-induced insulin resistance,17 adult PEDF null mice exhibit increased liver triglycerides,18 an insulin resistant phenotype, and truncal obesity (unpublished observations; Chung, Crawford).

PEDF affects the homeostasis of the pancreas. Both normal epithelial and stromal cells in the pancreas express relatively high levels of PEDF.19 Similar to other tumours, PEDF levels are markedly decreased in PDAC.20 ,21 Moreover, restoration of PEDF via gene therapy can suppress growth of experimental pancreatic cancer.22 When PEDF was ablated in mice, pancreatic mass was increased with epithelial atypia;19 however, there was no evidence of neoplasia.

Ras, the most common mutation in PDAC, has been shown to promote adipogenesis.23 We hypothesised that PEDF deficiency in the setting of oncogenic Ras activity would promote tumorigenesis in the pancreas due to increased local adipogenesis and upregulation of lipid storage proteins, thus, acting as a source for energy to augment the proliferative process. In order to study this hypothesis, we crossed PEDF null mice, which have central obesity, with EL-KrasG12D mice, which develop non-invasive cystic papillary neoplasms (CPNs).24 We show that PEDF ablation in EL-KrasG12D mice can lead to increased expression of matrix metalloproteinase (MMP)-2 and MMP-9 and the formation of invasive PDAC that, in some cases, metastasize to the liver. Not only was there an increase in peripancreatic fat with adipocyte hypertrophy, which is an independent predictor of poor outcome in human disease,25 but there was also multifocal deposition of adipocytes within the pancreas (pancreatic steatosis). Interestingly, the stroma of the pancreas of EL-KrasG12D/PEDF deficient mice displayed increased expression of two lipid droplet associated proteins involved in adipogenesis, TIP47 and ADRP, and decreased levels of the lipolytic enzyme, ATGL, thus, suggesting that the pancreatic stroma may directly facilitate the high metabolic needs of tumour cells.

Material and methods

Generation of EL-Kras/PEDF null mice

PEDF knockout mice were previously generated by homologous recombination as previously described.19 EL-Kras transgenic mice were derived by microinjection of FVB/N fertilised mouse embryos as described.24

Tissue collection and preparation

Pancreas, spleen, liver and intestine were collected from cohorts of genetically engineered and control mice after sacrificing them at various time points according to approved IACUC protocol. Patient serum was collected at the time of surgery for pancreatic disease and stored frozen before analysis as provided by the Pathology Core Facility of Northwestern University approved by the Institutional Review Board at Northwestern University.

Histologic and immunohistochemical preparation and analysis

Following fixation in formalin, processed tissue was sectioned at 4 um, deparaffinised, and hydrated followed by H&E staining. Acinar ductal metaplasia (ADM) was determined in six animals in each of the four groups (table 1). For each animal, percentage ADM was expressed as percent metaplastic acini from total number of acini. CPNs, tubular complexes (TCs), foci of inflammation, and intrapancreatic adipocytes were determined in six animals per group with each of the lesions or adipocytes counted in a minimum of five representative low power (10×) fields per animal and represented as a mean number of lesions per 10 low power fields. For immunohistochemical staining, CK19 antibody (Abcam) and Amylase antibody (Santa Cruz) were used. Representative slides were also stained with an antibody directed against macrophages (KP1, Dako). Mean adipocyte area in peripancreatic adipose tissue was determined in EL-KrasG12D/PEDF wild type (wt) controls, non-mutant Kras/PEDF null controls and EL-KrasG12D/PEDF null mice. Immunohistochemical staining was performed on hydrated slides following antigen retrieval and scored by three observers (PSF, PJG and SEC).

Table 1

Pancreatic ductal epithelium disease in EL-KrasG12D/pigment epithelium-derived factor (PEDF) null mice (Kras/PEDF−/−) compared with EL-KrasG12D/PEDF wild type control mice (Kras/PEDF+/+)

Cell culture

The BxPC-3, Panc-1, MiaPaCa2, Capan-1 and CFPac-1 cell lines were originally purchased from ATCC and grown in DMEM supplemented with 5% FBS at 37°C and 5% CO2.

siRNA knockdown of PEDF mRNA expression

PEDF expression was transiently downregulated using the Amaxa-Nucleofector-System (Lonza, Allendale, New Jersey, USA) and a predesigned and validated Silencer small interfering RNA (siRNA) sequence specific for PEDF (Ambion siRNA IDs s10288). MiaPaCa2 cells were transfected with PEDF siRNA or control siRNA using the Nucleofector kit V (Amaxa/Lonza) and allowed to recover for 48 h. HPDE KRAS cells were transfected using the Nucleofector kit for mammalian epithelial cells (Amaxa/Lonza) and allowed to recover for 24 h. Changes in gene expression were analysed by real-time PCR prior to using transfected cells in other experimental methods.

Quantitative real-time PCR analysis

Reverse transcription of RNA to cDNA was performed using Taqman reverse transcription reagents (N808-0234) from Applied Biosystems. Quantitative gene expression was performed for PEDF and GAPDH with gene specific Taqman probes, TaqMan Universal PCR Master Mix and the 7500 Fast Real-time PCR System from Applied Biosystems. The data were then quantified with the comparative CT method for relative gene expression.26 ,27

Motility assay

Haptotactic motility was assessed as described previously.27 ,28 One thousand cells were plated onto thin layer type I collagen overlaid with colloidal gold. Cells were allowed to migrate for 18 h and phagokinetic tracks were monitored by visual examination using a Zeiss microscope and photographed using a Nikon camera. Relative migration was quantified using Image J software.

Western blot analysis

Protein lysates were collected from cells grown to ∼80% confluence in 60 mm dishes. Tissue lysates were collected from whole mouse pancreas. Proteins were separated by size and probed with primary antibodies (PEDF, Upstate and ATGL, Cayman). Signal was detected using chemiluminescence.

Gel zymography

About 10–20 μg of protein was electrophoresed and incubated in developing buffer. The amount of MMPs was detected as clear bands.


p Values were determined using the one-way ANOVA test or two-tailed t-test.


PEDF deficiency significantly increases pancreatic disease morphology in the epithelial compartment of EL-KrasG12D mice

We ablated PEDF in EL-KrasG12D mice by crossing them with PEDF knockout mice. Double mutant EL-KrasG12D/PEDF null mice and age-matched wt EL-KrasG12D/PEDF wt controls were analysed histologically in two age groups, at 6–12 months of age and at 13–22 months of age (table 1). In both age groups, the degree of ADM (figure 1A,B; left side) was significantly higher in EL-KrasG12D/PEDF null mice compared to controls (table 1). The presence of CK19 and amylase in ADM was demonstrated in a subset of these samples (online supplementary figure 1). The occurrence of CPN lesions (figure 1A,B; right side) was more than threefold higher in PEDF deficient mice compared to mice with retained PEDF (p<0.05, table 1).

Figure 1

Histologic features of the pancreas in EL-KrasG12D/pigment epithelium-derived factor (PEDF) wild type (wt) and null mice. (A) EL-KrasG12D/PEDF wt mouse at 13 months with acinar ductal metaplasia (ADM) (m; left), loose surrounding stroma (s; left) and cystic papillary neoplasms (CPNs) (right). (B) EL-KrasG12D/PEDF null mouse at 13 months with ADM (m and inset; left), stromal fibrosis (f; left) and CPN (arrow; right). (C) EL-KrasG12D/PEDF null mouse at 13 months with stromal deposition of adipocytes (a; left) and marked positivity of the stroma for TIP47 (arrow; right). (D) Peripancreatic adipose tissue of EL-KrasG12D/PEDF null mouse (left) with increased adipocyte size (*) compared to EL-KrasG12D/PEDF wt mouse (right) both at 15 months of age. (E) Pancreatic ductal adenocarcinoma (PDAC) from an EL-KrasG12D/PEDF null mouse at 13 months with PDAC with irregular ducts (arrow; left) infiltrating the stroma and PDAC metastatic to the liver (arrow; right) from a 21-month EL-KrasG12D/PEDF null mouse. (F) Cytologic atypia (arrows) similar to PanIN grade 3 within a tubular complex (TC; left) in EL-KrasG12D/PEDF null mouse at 13 months and higher power of PanIN grade 3 within a TC (right). Scale bars: (A left, B left, C, D) 100 μm; (A right, inset B left, B right, E, F left), 50 μm; (F right), 20 μm.

Loss of PEDF in EL-KrasG12D promotes adipogenesis in the pancreas and induces stromal expression of lipid droplet associated proteins TIP47 and ADRP while decreasing tissue levels of ATGL

Changes in the pancreatic stroma associated with ablation of PEDF in EL-KrasG12D mice displayed a mild increase in fibrosis (figure 1B; left side) and periductal inflammation. The inflammation was multifocal and composed predominantly of lymphocytes admixed with macrophages and occasional neutrophils. Histological analyses did not reveal a statistically significant increase in the number of inflammatory foci. The mean number of inflammatory foci for KrasG12D/PEDF null versus KrasG12D/PEDF wt mice at 6–12 months was 3.7 versus 1.2 (p>0.05) and 5.5 versus 2.3 (p>0.05) at 13–22 months of age.

A more important change in the stroma of EL-KrasG12D/PEDF null mice was the significant increase (p<0.001) in the number of intrapancreatic adipocytes in the 13–22 month age group (figure 1C; left side, and figure 2A) when compared to control animals. The number of intrapancreatic adipocytes in PEDF deficient mice progressed with age (p<0.05, figure 2A) unlike the control mice. Interestingly, this increased adipogenesis in the pancreas of EL-KrasG12D/PEDF null mice was even more abundant and accompanied by expression of the lipid droplet associated protein TIP47 in the CN lesions (online supplementary figure 1A) and surrounding parenchyma (online supplementary figure 1C) of a PEDF null mouse, which were both nearly absent in EL-KrasG12D/PEDF wt controls (online supplementary figure 1B,D). The highest expression of TIP47 was observed in the stroma (figure 1C; right side) of EL-KrasG12D/PEDF null mice. Both TIP47 and ADRP strongly immunolocalised to the adipocytes and to stromal fibroblasts, though most tumour and normal epithelial cells lacked ADRP staining (data not shown). Indeed, TIP47 and ADRP had increased levels of mRNA expression in the pancreas of PEDF null mice compared to wt controls (figure 3D,E). Western blot analysis of pancreatic tissue for ATGL, a key lipase in lipolysis, showed a step-wise decrease with the lowest levels detected in the KrasG12D/PEDF null mice (figure 3C), although pancreatic mRNA levels of ATGL were increased in PEDF null mice compared to those observed in control mice (figure 3F).

Figure 2

(A) Mean number of intrapancreatic adipocytes of EL-KrasG12D/pigment epithelium-derived factor (PEDF) wild type (wt) (Kras/PEDF+/+) and EL-KrasG12D/PEDF null (Kras/PEDF−/−) mice. There is a significant increase in the number of intrapancreatic adipocytes at 13–22 months of age in Kras/PEDF null mice as compared to age matched Kras PEDF wt mice. Intrapancreatic adipocytes significantly increased with age in PEDF deficient mice, but not in PEDF wt mice (NS, not significant). (B). Size (mean area) of adipocytes in peripancreatic adipose tissue at 14–16 months of age of EL-KrasG12D/PEDF wt (Kras/PEDF+/+), non-mutant Kras/PEDF null (PEDF−/−) and EL-KrasG12D/PEDF null (Kras/PEDF−/−) mice. Peripancreatic adipocytes of Kras/PEDF null mice were markedly increased in size with a significantly larger mean cross-sectional area (2125 μm2) than that of Kras/PEDF wt (mean area of 866 μm2) and PEDF null (mean area of 1210 μm2) mice, which was significantly larger than those of Kras/PEDF wt mice.

Figure 3

Pigment epithelium-derived factor (PEDF) and adipose triglyceride lipase (ATGL) expression in the pancreas of wild type (WT), Kras and Kras/PEDF null mice. Western blot analysis of PEDF, ATGL and β-actin protein expression (A) in the pancreas reveals an ∼50% reduction in PEDF (B) and ATGL (C) in Kras mutant mice compared to WT controls. Kras/PEDF null mice showed a further decrease in ATGL expression to <10% of that of WT controls (C). A nearly twofold increase in tail-interacting protein 47 (TIP47) (D), adipose differentiation-related protein (ADRP) (E) and ATGL (F) mRNA expression was observed in PEDF null mice compared to WT controls. In (A), 1. WT, 2. EL-Kras and 3. EL-Kras/PEDF null protein lysate samples.

Peripancreatic fat was also increased in EL-KrasG12D/PEDF null mice. Since the size of adipocytes can be increased in the setting of metabolic syndrome and larger adipocytes indicate altered lipid metabolism,6 we assessed the size of adipocytes in the peripancreatic fat and found that EL-KrasG12D/PEDF null mice had a wide range of adipocyte sizes, including a subset of greatly enlarged adipocytes (figure 1D; left side) compared to EL-KrasG12D/PEDF wt mice (figure 1D; right side). Analysis of peripancreatic adipocytes at 14–16 months of age (figure 2B) showed that EL-KrasG12D/PEDF null mice had a significantly greater mean cross-sectional area of adipocytes (2125 μm2) than that of peripancreatic adipocytes of age matched controls EL-KrasG12D/PEDF wt mice (mean area of 866 μm2, p<0.001). Peripancreatic adipocytes of non-mutant Kras/PEDF null mice although significantly larger than those of EL-KrasG12D/PEDF wt mice (mean area of 1210 μm2, p<0.05) were at the same time significantly smaller than those of EL-KrasG12D/PEDF null mice (p<0.001) (figure 2B).

PEDF deficiency induces invasive pancreatic cancer in EL-KrasG12D mice and is associated with TCs

EL-KrasG12D/PEDF null mice developed invasive PDAC in 6 out of 41 animals in the 13–22 months age group (figure 1E; left side) with a small subset of animals demonstrating metastases to the liver (figure 1E; right side). None of the control EL-KrasG12D/PEDF wt animals developed PDAC or metastases (table 1). PDAC in EL-KrasG12D/PEDF null mice was associated with TCs and multifocal dysplastic changes of the epithelium similar to those of mPanIN grade 3 (figure 1F). However, mPanINs grades 1 and 2 were absent. Comparing EL-KrasG12D/PEDF null mice to EL-KrasG12D/PEDF wt controls, we found an increase in the number of TCs per 10 low power fields in the 13–22 months age group (5.8 vs 2.4, p<0.01) but not in the 6–12 months age group (1.0 vs 0.9, p>0.05). These data suggest an alternative pathway leading to PDAC via TCs, supporting a recent model described by Esposito et al.29

PEDF deficiency increases MMP-2 and MMP-9 expression in pancreatic cells

Since MMPs have been implicated in promoting tumour invasion and metastasis in mouse models of pancreatic cancer, we analysed the tumours from EL-KrasG12D/PEDF null mice for evidence of increased MMP expression by real-time PCR and gelatin zymography. As shown in figure 4, there was increased MMP-2 and/or MMP-9 expression in pancreatic tissue harvested from most EL-KrasG12D/PEDF null mice compared to control EL-KrasG12D mice. Moreover, a similar inverse relationship between PEDF and MMP-9 expression levels was observed in human pancreatic cancer cell lines. PEDF levels could be detected in the secretions of several human PDAC cell lines including Panc-1 and MiaPaCa2 cells (figure 4C). PEDF levels, moreover, were inversely related to the level of MMP-2/-9 expression. Panc-1 cells displayed abundant MMP-2 expression while MiaPaCa2 cells (which contain the highest levels of PEDF out of 10 pancreatic cancer cell lines evaluated) did not. In the presence of an MMP-2 inhibitor, PEDF levels were increased in Panc-1 cells while this inhibitor had a minimal effect on MiaPaCa2 cells.

Figure 4

RNA and protein expression of matrix metalloproteinase (MMP)-2 and MMP-9 in EL-KrasG12D/pigment epithelium-derived factor (PEDF) null and wild type (wt) mice. Real-time PCR (A, represented as fold increase compared to wt control) and gel zymography (B) of MMP-2 and MMP-9 from protein lysates collected from wt and Kras/PEDF null mice at 16–18 months of age. PEDF protein expression via western analyses and gel zymography (C) in conditioned media following 24 and 72 h in serum-free conditions with and without the addition of the selective MMP-2 inhibitor ARP100 (2-[((1,1'-Biphenyl)-4-ylsulfonyl)-(1-methylethoxy)amino]-N-hydroxyacetamide) for Panc-1 and MiaPaCa2 cells. Equivalent loading is shown with Coomassie stain for gel zymography.

PEDF attenuates motility of MiaPaCa2 and HPDE-Kras cells

To further demonstrate the role of PEDF in pancreatic cancer development and progression, we examined the effect of modulating PEDF in pancreatic cells with relatively high levels of PEDF (MiaPaCa2 and HPDE-Kras cells). These cells were transfected with control siRNA or siRNA against PEDF, allowed to recover, and then plated onto collagen-coated surfaces overlaid with colloidal gold. As shown in figure 5A, there is >90% knockdown of PEDF with the PEDF specific siRNA. The PEDF siRNA also significantly increases motility of MiaPaCa2 and HPDE-Kras cells (figure 5B).

Figure 5

Motility of MiaPaCa2 and HPDE-Kras cells following knockdown of pigment epithelium-derived factor (PEDF). (A) MiaPaCa2 and HPDE-Kras cells were transfected with control siRNA and PEDF siRNA and changes in PEDF and GAPDH were analysed by quantitative real-time PCR. (B) The transfected cells were plated onto collagen-coated tissue plates covered with colloidal gold and allowed to migrate over 18 h. The phagokinetic tracks were monitored by visual examination using a Zeiss microscope and photographed (bottom panels) and quantified (top panels).

Cell lines from metastatic human pancreatic cancer have the lowest levels of PEDF

Serum-free conditioned medium was collected from commercially available human pancreatic cell lines cultured in vitro and analysed by western blot using anti-PEDF antibody. PEDF protein expression was evaluated relative to GAPDH (figure 6). CFPAC1 and CAPAN-1, two PDAC cell lines derived from liver metastases, demonstrated more than fourfold decrease in PEDF levels in the conditioned medium compared to the three cell lines derived from primary PDAC (MiaPaCa2, BXPC-3 and PANC-1). ASPC-1 cell line derived from ascites had an intermediate expression of PEDF compared to the other two groups. These data suggest that acquisition of a metastatic phenotype could necessitate degradation of PEDF and loss of its antitumour activity.

Figure 6

Expression of pigment epithelium-derived factor (PEDF) in human cell lines of pancreatic cancer, represented relative to expression of GAPDH. There is a more than fourfold decrease in expression of PEDF in cancer cell lines derived from liver metastases (CAPAN-1 and CFPAC1) compared to cell lines derived from primary tumour (MiaPaCa2, BXPC-3 and PANC-1). ASPC1 cells derived from ascites had an intermediate expression of PEDF.

Patients with pancreatic cancer have decreased serum and tissue levels of PEDF

We analysed PEDF protein levels in serum samples from 67 patients with pancreatic disease, 40 with pancreatic cancer and 27 with non-malignant pancreatic lesions (figure 6A). Patients with pancreatic cancer had a ∼75% decrease in PEDF levels with a mean PEDF serum level of 80.5 ng/ml (±41.4 ng/ml), as compared to 344.7 ng/ml (±145.3 ng/ml) for patients with non-malignant pancreatic disease. We also analysed serum levels of the proangiogenic factor vascular endothelial growth factor (VEGF), which demonstrated the reciprocal pattern of expression (figure 7A). Patients with pancreatic cancer had a 65% increase in VEGF levels (mean VEGF serum level of 440.5±61.3 pg/ml) compared to VEGF levels of 266±31.2 pg/ml in patients without evidence of cancer. Representing these data as a ratio of VEGF to PEDF revealed that the mean ratio of VEGF to PEDF (2.4) of patients with PDAC was three times higher than that of patients with non-malignant disease (VEGF to PEDF ratio 0.8). This suggests that PEDF may have potential use as a biomarker for PDAC in combination with VEGF. This is similar to other angiogenic-dependent diseases, where the net angiogenic activity (inducers/inhibitors) has greater clinical utility than the value of a single angiogenic factor.30 ,31

Figure 7

(A) Analysis of serum levels of pigment epithelium-derived factor (PEDF) and vascular endothelial growth factor (VEGF) in 67 patients, 40 with malignant disease and 27 with non-malignant pancreatic lesions. The mean PEDF serum level was decreased by ∼75% in malignant (80.5±41.4 ng/ml) versus non-malignant (344.7±145.3 ng/ml) lesions. Mean VEGF level was increased by ∼ 65% in cancer (440.5±61.3 pg/ml) versus non-cancer (266±31.2 pg/ml) patients. The means are depicted as a solid line within each grouping of data points (p<0.05). (B) Levels of PEDF and VEGF in human patients with pancreatic cancer. Immunohistochemical staining of 140 human pancreatic cancers for VEGF and PEDF. Tumours overexpressing VEGF (2+ or 3+ staining) have decreased expression of PEDF (p<0.05).

To determine if altered circulating levels of VEGF and PEDF were reflective of changes at the tissue level, we evaluated human pancreatic tumours (n=140) for PEDF and VEGF protein expression by immunohistochemistry (IHC). Similarly, we found an inverse relationship between VEGF and PEDF by IHC staining (figure 7B) with tumours expressing higher levels of VEGF (2+ or 3+ staining) and less expression of PEDF (p<0.05).


PEDF is a multifunctional glycoprotein with potent antiangiogenic and antiproliferative activities.13 ,14 ,19 ,32 We show that ablation of PEDF in EL-KrasG12D mice is sufficient to induce invasive PDAC, including metastatic disease to the liver, and this phenotype was associated with increased adiposity in the pancreas. Loss of PEDF in EL-KrasG12D mice promoted steatosis within the pancreas and around the organ with significantly larger adipocytes. A subset of adipocytes exhibited a massive increase in size or ‘supersized’ adipocytes. In support of altered lipid metabolism in the pancreas, the phenotype was accompanied by stromal expression of TIP47 and ADRP, two lipid droplet coat proteins which regulate cellular lipid stores by controlling the access of lipases to the lipid droplet.33 Although ADRP and TIP47 typically target adipocytes, limited studies suggest that they also play a critical role in the formation and turnover of ectopic lipid droplets outside adipose tissue.34 For example, intracytoplasmic lipid droplets have been identified in a variety of cancers with expression of ADRP and TIP47 noted in cancer cells.10 The current study expands the role of these lipid droplet associated proteins to non-adipocyte stromal cells in the pancreas and suggests that these cells may have a crucial role in supporting the metabolism of cancer cells by protecting a renewable source for energy to stimulate tumour progression.

We also observed a markedly decreased pancreatic expression of ATGL in EL-KrasG12D/ PEDF null mice. ATGL is a functional lipase involved in the lipolysis of triacylglycerols.9 It is best studied in the context of adipocyte biology; however, it is important in controlling several metabolic parameters, including triglyceride content in the liver through its interaction with PEDF.18 A deficiency in the lipolytic enzyme, ATGL, in the pancreas may promote pancreatic steatosis in these mice by creating an imbalance in triacylglyceride turnover, thus favouring a net increase in storage of lipid droplets.

In the setting of obesity, the adipocyte population can expand greatly; thus, fat-soluble factors can become increasingly important in determining the fate of tumour cells. In our model, loss of PEDF in the background of oncogenic Ras activity promotes adipogenesis in the pancreas and facilitates tumour progression. This model is similar to other mouse models of PDAC in which elevated adipocyte mass in the pancreas (in the setting of obesity) correlated with increased tumour growth and metastases.5 These models appear to recapitulate features of human PDAC. In one PDAC patient study, increased fat in the pancreas or pancreatic steatosis was associated with increased dissemination of the cancer.4

The pancreatic steatosis in EL-KrasG12D/PEDF null mice was accompanied by a marked increase in the size of individual adipocytes. Such an expansion in the size of adipocytes (adipocyte hypertrophy) is a feature of obesity.35 Adipocyte hypertrophy within the abdominal fat depots in humans is associated with increased risk of metabolic syndrome and its complications.36 ,37 In mice, differences in adipocyte size correlated with alterations in protein expression and insulin signalling.38 The ‘supersized’ adipocytes in our model may augment signalling and disrupt counter-regulatory mechanisms intended to normalise metabolism in the tumour microenvironment.

PEDF and Ras activity affect adipogenesis. Ras is the most common mutation in human PDAC and Ras oncogenes are sufficient to induce differentiation of 3T3-L1 fibroblasts into adipocytes.23 Ras mediates its effects through the Ras/mitogen-activated protein kinase (MAPK) pathway and the MAPK pathway plays a role in normal and abnormal adipogenesis.39 There is a possible convergence in their signalling since PEDF suppresses adipogenesis in 3T3-L1 adipocytes through inhibition of the MAPK pathway.40 Future studies are needed to confirm that PEDF deficiency promotes adipogenesis through the Ras/MAPK pathway.

We conducted our studies using the EL-KrasG12D model in which activation of Kras occurs in the acinar compartment of the pancreas under regulation of the elastase (EL) promoter. This leads to ADM and non-invasive CPNs, which do not progress to invasive PDAC.24 Expert evaluation of these mice concluded that they do not develop mPanINs unlike other models.41 In our study, ablation of PEDF in EL-KrasG12D mice increased the frequency, in older animals, of both CPNs and TCs. An unexpected finding in EL-KrasG12D/PEDF null mice was that invasive PDAC was associated with TCs and focal cytologic atypia similar to mPanIN grade 3 lesions. However, the typical progression from low to high grade mPanINs was absent. Such a link between TCs and PDAC has previously been reported in other mouse models of pancreatic cancer.41 ,42 In addition, a TC to PDAC progression has recently been suggested in human tumours when Esposito et al hypothesised that PanINs, the precursors to PDAC, can arise within TCs resulting from ADM.29 The absence of the typical progression to PDAC through mPanIN-1 to mPanIN-3 has also been described in another mouse model of pancreatic tumours using the EL promoter, the EL-transforming growth factor α (TGF-α) mouse model, in which TGF-α is overexpressed under the EL promoter. EL-TGF-α mice form non-invasive cystic papillary tumours and PDAC occurs following ablation of p53.43 Similar to EL-KrasG12D/PEDF null mice, PDAC in these mice was associated with TCs without any true mPanIN lesions.

Invasive disease may be related to loss of PEDF since there was increased expression of MMP-2 and MMP-9 in the EL-KrasG12D/PEDF null mice compared to EL-KrasG12D and control mice. MMPs modulate stromal angiogenesis through proteolytic degradation of angiogenic inhibitors and MMP-9 can promote invasion and liver metastasis of pancreatic cancer cells.44 PEDF serves as a substrate for MMP-2 and MMP-945 and in its absence, MMP-2 and MMP-9 may have greater activity with other substrates including those in the extracellular matrix. Interestingly, MMPs have also been implicated as regulators of adipogenesis.46 Both MMP-2 and MMP-9 have an inherent role in adipogenesis including increasing the survival of adipocytes in vivo, which may lead to abnormal extracellular matrix development47 and low grade inflammation.48 Furthermore, obese patients have increased plasma levels of MMP-2 and MMP-947 although it is not known if this correlates with an increase in local tissue levels.

Alterations in PEDF expression levels have been previously shown to have prognostic significance among patients with PDAC. One study found that higher PEDF expression in the carcinoma was an independent favourable prognostic factor, whereas absence of PEDF expression in the tumour carried increased risk of hepatic metastases and shortened survival.21 Our study expands the potential utility of PEDF in PDAC patients by demonstrating that mean serum levels of PEDF in pancreatic cancer patients were almost 75% lower than those with non-malignant disease of the pancreas. These differences were even more striking when net angiogenic activity was assessed by calculating mean VEGF to PEDF ratio in these patients. There was a nearly threefold increase of the VEGF to PEDF ratio in the serum of pancreatic cancer patients compared to those with non-malignant pancreatic disease. In accordance with this, the tissue VEGF to PEDF ratio assessed by IHC was significantly increased (p<0.05). There is a paucity of useful biomarkers in PDAC and it is possible that following serial PEDF levels may prove to add prognostic information for directing patient care. Moreover, therapeutic replacement of PEDF could represent a new strategy to overcome the tumour permissive properties of the stromal and epithelial compartments enriched by adipocytes in pancreatic cancer.


We wish to acknowledge the consultation with Drs Thomas Adrian and Richard Bell at the onset of this study. We would like to thank Dr Mohammad Reza Salabat and Carolyn Pelham for providing technical assistance and Drs Beth Plunkett and Jennifer Doll for editorial assistance.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • An additional figure is published online only. To view this file please visit the journal online (

  • Funding The research described herein was generously supported by funds provided by the Barnum Foundation and Zell Family Foundation at the Northwestern University, the Nathan and Isabel Miller Family Foundation, the IDP Foundation, NIH R01-CA126888 (HGM) and NIH R01-CA64239 (SEC).

  • Competing interests None.

  • Patient consent The Northwestern University IRB has consent forms signed by patients/guardians for the use of blood and tissue samples.

  • Ethics approval Northwestern University IRB.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.