Volume 132, Issue 4, Pages 1447-1464 (April 2007) Periostin Creates a Tumor-Supportive Microenvironment in the Pancreas by Sustaining Fibrogenic Stellate Cell Activity  Mert Erkan, Jörg Kleeff, Andre Gorbachevski, Carolin Reiser, Tomas Mitkus, Irene Esposito, Thomas Giese, Markus W. Büchler, Nathalia A. Giese, Helmut Friess  Gastroenterology  Volume 132, Issue 4, Pages 1447-1464 (April 2007) DOI: 10.1053/j.gastro.2007.01.031 Copyright © 2007 AGA Institute Terms and Conditions

Figure 1 POSTN expression in normal tissues vs PDAC and its implications for patient survival. (A) Expression of POSTN mRNA analyzed by QRT-PCR is depicted on a logarithmic scale as copies per microliter complementary DNA normalized to housekeeping genes. Horizontal bars represent the median values. Immunoblot analysis of protein mixtures of 7 normal and 15 PDAC samples equally contributing to the final lysate mirrors the 42-fold increase of POSTN mRNA. γ-Tubulin was used as loading control. (B) Kaplan–Meier survival analysis of 10 patients with POSTN mRNA levels within the normal range and 31 patients above the normal range. Median survival times were 19 vs 12 months, respectively (P = .14). (C and D) The localization of POSTN in the normal pancreas and PDAC by immunohistochemistry. Negative controls are shown as insets. The rare interacinar POSTN staining of (C) the normal pancreas significantly increased in all (D) PDAC samples analyzed. Note the significant intensification of POSTN staining on the invading front of the tumor around degenerating acini (black arrows) compared with normal acini (white arrows; original magnification: C, 100×; D, 25×). Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 2 Expression analysis of α-SMA, POSTN, COL1, and fibronectin in PSCs as well as in primary and metastatic PDAC lesions. (A) Standard exposure time was 2000 milliseconds (250 milliseconds for DAPI) during immunofluorescence microphotography. (A1) Activated PSCs express cytoplasmic α-SMA filaments. (A2) Intracytoplasmic POSTN expression significantly decreased when cells were starved (inset, 5000 milliseconds). (A3) COL1 and (A4) fibronectin expression were not affected by serum starvation. (B) Immunohistochemical analysis of ECM proteins in primary PDAC lesions. Consecutive sections were probed by (B1) α-SMA, (B2) POSTN, (B3) COL1, and (B4) fibronectin antibodies. (C and D) Immunohistochemical analysis of ECM proteins in (C) liver and (D) lymph node metastases of a patient with PDAC. Consecutive sections were probed with (C1 and D1) α-SMA, (C2 and D2) POSTN, (C3 and D3) COL1, and (C4 and D4) fibronectin antibodies. Note the strong α-SMA expression in metastases but weaker POSTN, COL1, and fibronectin staining relative to primary lesions (original magnification: A1, A2, and inset, 400×; A3 and A4, 200×; B1, C1, and D1, 50×). Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 3 Site-specific accumulation of POSTN in PDAC. Immunohistochemical analysis of POSTN expression around (A) tubular complexes and (B) cancer structures. Notice the negativity of tubular complexes and malignant cells in contrast to strong ECM staining around them. (C and D) Immunohistochemical analysis of α-SMA and POSTN expression in PDAC. Consecutive sections of PDAC tissues were probed with (C) α-SMA and (D) POSTN antibodies. Notice the stronger (C, inset) α-SMA and (D, inset) POSTN expression of PSCs directly contacting cancer cells compared with PSCs that did not have physical contact. Also notice that the strongest POSTN staining at the interface between the cancer and the normal pancreas (D, arrows) does not overlap with the strongest α-SMA expression, seen around cancer structures (C). (Original magnification: A, B, C [inset], and D [inset], 100×, C and D, 25×.) Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 4 Factors that affect the synthetic and secretory functions of PSCs. (A) Stimulatory effects of cancer cells on PSCs. Conditioned medium was prepared by keeping serum-free medium on Panc-1, SU86.86, and T3M4 cell lines for 24 hours. Specific ECM protein content of the stellate cell SN was measured after 96 hours of treatment with conditioned medium by immunoblot analysis. (B and C) To assess the synthetic and secretory effects of growth factors on PSCs, 100% confluent PSCs were treated with growth factors in the absence of FCS. Cell lysate and SNs were collected at 96 hours. (B) Matching cell lysates and (C) SNs were analyzed by immunoblotting to evaluate the synthetic and secretory responses, respectively. All experiments were repeated 3 times, and the results of densitometric analyses are presented as fold change (mean ± SEM) compared with control. Representative blots were consecutively probed with POSTN, γ-tubulin, COL1, α-SMA, and fibronectin antibodies. Note the protein-specific effects of growth factors and amplified changes in the SNs. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 5 Effects of POSTN on PSCs. To analyze the effects of POSTN on PSCs, (A) cells were either treated with rPOSTN (100 ng/mL and 1 μg/mL) or (B) endogenous POSTN was silenced using specific siRNA molecules. SNs and cell lysates were collected at 96 hours. TGF-β1 was used as positive control. Although rPOSTN (75 kilodaltons) increased the background signal in the cell lysates (POSTN-CL) in a dose-dependent way when probed by anti-POSTN antibody, it was not detectable in the SNs (POSTN-SN). siRNA transfection experiments were repeated 3 times, and the results of densitometric analyses were presented as percent change (mean ± SEM) compared with control. Representative blots were consecutively probed with the POSTN, γ-tubulin, COL1, α-SMA, and fibronectin antibodies. (C and D) Effect of TGF-β1 induction on COL1 and α-SMA expression of PSCs. (C) Cancer cell SNs or (D) 1 μg/mL rPOSTN was preincubated for 1 hour with either nonimmunized rabbit IgG control (10 μg/mL) or anti–TGF-β1 neutralizing antibodies (10 μg/mL) before the conditioned media were added to cultured PSCs for 96 hours. Note the prevailing increased COL1 and α-SMA expressions after treatment with Panc-1 and SU86.86 SNs as well as 1 μg/mL rPOSTN even after immunodepletion with anti–TGF-β1 antibodies. −, control siRNA; +, POSTN siRNA; *P < .05. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 6 The effects of POSTN on growth and invasiveness of PCCs and PSCs. (A and B) To analyze the effects of rPOSTN (100 ng/mL) on growth under (A) serum deprivation (SM 1%) and (B) hypoxia, crystal violet staining was carried out after 5 and 7 days. Bovine serum albumin (100 ng/mL) was used as control. After the dye was solubilized with methanol, OD was read at 570 nm. Results are presented as percent change compared with controls (100%). (C) To assess invasiveness, PCCs and PSCs were seeded in BD Biocoat Matrigel invasion chambers with 8-μm pore size in the presence of POSTN (0, 100 ng/mL, 1 μg/mL) without creating a gradient between the upper and lower chambers. Similarly, the invasiveness of PSCs was assessed after silencing of POSTN (right graph). The assays were performed in duplicate and repeated at least 4 times. After staining, the whole Matrigel was scanned and the invading cells were counted using the software for Zeiss KS300 (Zeiss Axiocam HR). Results are presented as fold change (mean ± SEM) compared with control. *P < 05. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 7 Subcellular pathways activated by POSTN. To assess Akt and 44/42-MAPK phosphorylation, (A) Panc-1 and (B) PSCs were treated with PBS (control = 0), rPOSTN 100 ng/mL (upper row), and 1 μg/mL (lower row). Protein was extracted after 1, 2, 3, 5, 10, 30, 60, and 90 minutes. Representative blots shown were blotted with p-Akt, total Akt, p44/42-MAPK, and ERK-2 antibodies. Densitometric analyses of 2 independent experiments were expressed as fold changes (mean ± SEM) compared with control after correcting for the appropriate equal loading controls. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 8 Effects of COL1 on the chemoresponsiveness of PCCs. To assess the influence of COL1 on PCC responses to 5-fluorouracil and gemcitabine, PCCs were treated with gradient doses of chemotherapeutics on normal plates (NP) and COL1 coated plates (CP). Either 5-fluorouracil or gemcitabine was added 12 hours after seeding at gradient concentrations; 0.01% dimethyl sulfoxide and PBS were used as controls, respectively. After 48 hours of incubation, the MTT test was performed. Formazan crystals were solubilized with acidic isopropanol, and the OD was read at 570 nm by an enzyme-linked immunosorbent assay reader (Opsys MR). The median effective doses of chemotherapeutic drugs on cancer cells were calculated from 5 independent experiments normalized to matching day-0 observations. Results are presented as percent change compared with controls. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 9 PSC responses to chemoradiotherapy and radiotherapy and effects of irradiated PSC SN on PCC growth and invasiveness. (A) To assess the influence of COL1 on PSC responses to 5-fluorouracil and gemcitabine, PSCs were treated with gradient doses of chemotherapeutics on normal (NP) and COL1-coated plates (CP). Either 5-fluorouracil or gemcitabine was added 12 hours after seeding at gradient concentrations; 0.01% dimethyl sulfoxide and PBS were used as controls, respectively. After 48 hours of incubation, the MTT test was performed. Formazan crystals were solubilized with acidic isopropanol, and the OD was read at 570 nm by an enzyme-linked immunosorbent assay reader (Opsys MR). Because PSCs were practically unresponsive to either drug, it was not possible to calculate median effective doses. Their responses are presented as percent change compared with control. (B and C) Conditioned medium was prepared by applying SM 10% on 10-Gy irradiated, 50% confluent PSCs for 7 days (RT-SN). Control SN was prepared similarly from nonirradiated sister clones of PSCs. PCC growth is presented as percent change (mean ± SEM) compared with control (B). To assess invasiveness, PCCs were seeded in BD Biocoat Matrigel invasion chambers with 8-μm pore size either in RT-SN or control SN. The assays were performed in duplicate and repeated at least 4 times. After staining, the whole Matrigel was scanned and the invading cells were counted using the software for Zeiss KS300 (Zeiss Axiocam HR). Results are presented as fold change (mean ± SEM) compared with control. *P < .05. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions

Figure 10 In vivo effects of NeoRCT on PDAC tissues and in vitro effects of POSTN on PSC responses to radiotherapy. (A) Immunohistochemical analysis of α-SMA and POSTN expression in PDAC tissues after NeoRCT. Consecutive sections of PDAC tissues were probed with α-SMA and POSTN antibodies (original magnification 50×). Notice the residual α-SMA staining only around cancer structures and weak POSTN staining in the stroma (inset). (B) Immunohistochemical analysis of ECM proteins after NeoRCT in PDAC lesions. Consecutive sections were probed by (B1) α-SMA (original magnification 100×), (B2) POSTN, (B3) COL1, and (B4) fibronectin antibodies. Notice the abundant fibrillary stroma rich in collagen and fibronectin with scarce stromal cells. (C) Sister clones of PSCs were irradiated (0 vs 10 Gy) and treated with POSTN (0, 100 ng/mL, 1 μg/mL). SN and cell lysates were collected at 96 hours postirradiation. Blots were consecutively probed with the POSTN, γ-tubulin, COL1, α-SMA, and fibronectin antibodies in the written sequence. Densitometric analyses of 3 independent experiments are expressed as fold changes (mean ± SEM) compared with control. Gastroenterology 2007 132, 1447-1464DOI: (10.1053/j.gastro.2007.01.031) Copyright © 2007 AGA Institute Terms and Conditions