Inhibition of endogenous SPARC enhances pancreatic cancer cell growth: modulation by FGFR1-III isoform expression

Background: Secreted protein acidic and rich in cysteine (SPARC) is a multi-faceted protein-modulating cell–cell and cell–matrix interactions. In cancer, SPARC can be not only associated with a highly aggressive phenotype, but also acts as a tumour suppressor. The aim of this study was to characterise the function of SPARC and its modulation by fibroblast growth factor receptor (FGFR) 1 isoforms in pancreatic ductal adenocarcinoma (PDAC). Methods and results: Exogenous SPARC inhibited growth, movement, and migration. ShRNA inhibition of endogenous SPARC in ASPC-1 and PANC-1 cells resulted in increased anchorage-dependent and -independent growth, transwell migration, and xenograft growth as well as increased mitogenic efficacy of fibroblast growth factor (FGF) 1 and FGF2. Endogenous SPARC expression in PANC-1 cells was increased in FGFR1-IIIb over-expressing cells, but decreased in FGFR1-IIIc over-expressing cells. The up-regulation of endogenous SPARC was abrogated by the p38-mitogen-activated protein kinase inhibitor SB203580. SPARC was detectable in conditioned medium of pancreatic stellate cells (PSCs), but not PDAC cells. Conditioned medium of PDAC cells reduced endogenous SPARC expression of PSCs. Conclusion: Endogenous SPARC inhibits the malignant phenotype of PDAC cells and may, therefore, act as a tumour suppressor in PDAC. Endogenous SPARC expression can be modulated by FGFR1-III isoform expression. In addition, PDAC cells may inhibit endogenous SPARC expression in surrounding PSCs by paracrine actions.

Secreted protein acidic and rich in cysteine (SPARC) or osteonectin is a 32 -33 kDa calcium-binding glycoprotein shown to associate with the cell membrane and membrane receptors (Yan and Sage, 1999). It belongs to a family of matricelluar proteins and is divided into three modules that exert various functions (Bradshaw and Sage, 2001). It functions not only to modulate cell -cell and cell -matrix interactions, but also has de-adhesive and growth inhibitory properties in non-transformed cells . In cancer, SPARC may exert divergent actions reflecting the complexity of this protein (Clark and Sage, 2008;Podhajcer et al, 2008). Thus, in certain types of cancers, such as melanomas and gliomas, SPARC is associated with a highly aggressive tumour phenotype, whereas in others, mainly ovarian, neuroblastomas, and colorectal cancers, SPARC may function as a tumour suppressor . These functions are thought to be exerted in part by its ability to counteract effects of several growth factor families including fibroblast growth factors (FGFs) (Yan and Sage, 1999;Francki et al, 2003;Motamed et al, 2003).
Modulation of FGF actions by SPARC is mediated by FGF receptor (FGFR) 1, because FGFR1 was reported to be indispensable for SPARC-induced inhibition of FGF signalling . The presence of several FGFR1 isoforms generated by alternative mRNA splicing makes the analysis of FGFR1 and SPARC interactions difficult. FGF actions strongly depend on the presence of specific FGFR isoforms and can be modulated by changes in isoform expression (Ornitz et al, 1996;Kornmann et al, 2001;Liu et al, 2007a). Alternative splicing of the second half of Ig-like domain III generates particular important isoforms, because this region determines ligand-binding specificity (Ornitz et al, 1996;Plotnikov et al, 2000). As a consequence, overexpression of FGFR1-IIIc in pancreatic cancer promoted tumourigenesis, whereas over-expression of FGFR1-IIIb inhibited the malignant phenotype (Kornmann et al, 2002;Liu et al, 2007b). In addition, several FGFs, signalling through FGFR1-IIIc, are highly over-expressed in human pancreatic ductal adenocarcinoma (PDAC) (Yamanaka et al, 1993). Several recent studies reported increased SPARC levels in human PDAC (Sato et al, 2003;Guweidhi et al, 2005) associated with poor prognosis (Infante et al, 2007). SPARC expression was absent in most of the cancer cells, but instead present at high levels in the peritumoural tissue harbouring fibroblasts and pancreatic stellate cells (PSCs) (Guweidhi et al, 2005;Infante et al, 2007).
Despite these recent efforts, the functions of SPARC and its associations with FGFR1-III isoforms in PDAC remain unclear. Therefore, the aim of this study was to elucidate SPARC functions and endogenous SPARC regulation depending on FGFR1-III isoforms expression in PDAC cells and its interaction with PSC cells.

Establishment of cell clones over-expressing FGFR1-III variants
The establishment of the FGFR1-IIIb and -IIIc PANC-1 clones over-expressing the full-length cDNA of human FGFR1-IIIb or -IIIc expressed in a modified pSVK3 vector under the control of the simian virus 40 early promotor in a stable manner was described earlier (Liu et al, 2007b;Chen et al, 2008).

Establishment of cell clones expressing SPARC shRNA
Validated SureSilencing human SPARC shRNA and control plasmids were from SuperArray Bioscience Corp. (Frederick, MD, USA). ASPC-1 and PANC-1 cells were transfected in a stable manner using lipofectamine (Invitrogen, Carlsbad, CA, USA) following the manufacturer's protocol. Transfected cells were selected with G418 (0.8 and 1.2 mg ml -1 for ASPC-1 and PANC-1, respectively) for 14 days before isolation of individual clones.

Immunoblot analysis
Total cell lysates were prepared and followed by immunoblot analysis as described (Liu et al, 2007b). A rabbit polyclonal antibody (SPARC, sc-25574, from Santa Cruz) was used (1 : 200) to detect SPARC protein. Bound antibodies were visualised using enhanced chemiluminescence. To confirm equal loading, membranes were stripped for 30 min at 501C in buffer containing 2% SDS, 62.5 mM Tris (PH 6.7), and 100 mM 2-mercaptoethanol and reprobed with an anti-b-actin antibody to show equal loading.

RT -PCR
Total RNA (1 -2 mg) was reverse transcribed using a SuperScript pre-amplification kit (Invitrogen). Primers were based on the sequences reported on Genebank (NM 003118). SPARC sense sequence was 5 0 -GTGGGCAAAGGGAAGTAACA-3 0 and SPARC anti-sense sequence 5 0 -GGGAGGGTGAAGAAAAGGAG-3 0 . The expected product size of SPARC cDNA was 514 bp. PCR amplification was performed in 25 ml reaction volumes containing 0.2 mM dNTPs, 20 pmol of each oligonucleotide primer, and 0.2 U Tag polymerase in PCR buffer. cDNA was amplified on a PCR thermal controller with an initial denaturation at 951C for 5 min, followed by cycles of 951C for 1 min, 651C for 1 min, and 721C for 1 min, 27 cycles, and a final extension step of 721C for 10 min. The amount of starting cDNA was adjusted using b-actin intensity.

Anchorage-dependent growth and functional assays
The effects of exogenous SPARC (from murine parietal yolk sac cells, S5174 Sigma, St Louis, MO, USA) on proliferation were determined by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. The indicated amount of cells was seeded in 96-well plates and propagated for 24 h in complete medium before further analysis. To assess basal growth and the effect of SPARC alone, 5000 cells per well were incubated for 48 h in complete medium in the absence or presence of SPARC. To assess the effects of SPARC on FGF-induced proliferation, cells (10 000 per well) were propagated for 48 h in serum-free medium including the indicated factors. FGF1 (recombinant human FGF acidic, R&D Systems) and FGF2 (recombinant human FGF basic, R&D Systems) were added in the presence of heparin (1 mg ml -1 ).
To analyse N-linked protein, glycosylation cells were incubated in the absence or presence of tunicamycin (5 mg ml -1 , T7765 Sigma) for 24 h as described (Liu et al, 2007a). To assess the effect of kinase, inhibitors cells were incubated in the absence or presence of the p38-mitogen-activated protein kinase (MAPK) inhibitor SB203580 (Calbiochem, Darmstadt, Germany) and the c-Jun N-terminal kinase (JNK) inhibitor SP600125 (Calbiochem) for 24 h (Liu et al, 2007a).

Single cell movement
Cells (50 000 per well) were seeded onto fibronetin-coated (5 mg ml -1 in PBS) six-well plates and grown for 20 h. Cells were then incubated in the presence or absence of SPARC (5 mg ml -1 ) for 24 h. Cell movement during that period was monitored by an Olympus IX81 motorised inverted microscope taking pictures every 10 min (Liu et al, 2007b). The total distance of individual cells covered within 24 h was determined using the ImageJ 1.32 program (NIH, Bethesda, MD, USA).

Cell migration assay
The ability of cells to migrate through filters was measured using a BioCoat Matrigel invasion chamber (BD Biosciences, San Jose, CA, USA). Cell culture inserts with an 8 mm pore size PET membrane were used according to the protocol of the manufacturer. The bottom chamber included medium (0.75 ml) containing 10% FCS, whereas cells (1.0 Â 10 5 suspended in 0.5 ml of medium containing 1% FCS) were seeded into the upper chamber and incubated overnight at 371C in a humidified atmosphere containing 5% CO 2 . Cells were then incubated in the absence or presence of exogenous SPARC (5 mg ml -1 ) for another 24 h. Remaining cells on the upper surface were mechanically removed. Membranes were then washed, fixed, and stained by Diff-Quik (Medion Diagnostics, Düdingen, Switzerland). The number of cells that migrated to the lower surface of the filters was determined by counting stained cells under a light microscope.

Anchorage-independent growth assay
Basal anchorage-independent cell growth was assessed by a double-layer soft-agar assay as described (Kornmann et al, 2002). Briefly, cells were suspended in complete medium containing 0.3% agar and seeded in triplicate in six-well plates onto a base layer of complete medium containing 0.5% agar. One mililiter of complete medium containing 0.3% agar was added every 5 days. After 14 days, 300 mg MTT per well was added to stain vital colonies for 24 h before counting by microscopy.

In vivo tumourigenicity assay
To assess the effect of expression of SPARC on xenograft formation, 10 6 cells per site were injected s.c. into two sites of 4-to 6-week-old female athymic (nude) mice. Animals were monitored for tumour formation every 4 days. Tumour size was measured in three dimensions. Tumour volume was determined by the equation vol ¼ l Â w Â d Â 0.5, where l is the length, w is the width, and d is the diameter. Animals had to be killed 12 weeks after injection according to our animal protocol (#718) if neither tumour volume (42 cm 3 ) nor skin ulcerations prompted earlier termination.

Immunohistochemistry of xenograft tumours
SPARC expression in control-transfected (n ¼ 10) and FGFR1-IIIb over-expressing (n ¼ 10) xenograft tumours (Liu et al, 2007b) A N-9 A 2-7 A 3-3 P N-21 P 1-15 P 1-20 Basal cell growth (% of wild type) was determined by immunohistochemical analysis. SPARC immunohistochemistry was performed using formalin-fixed and paraffin-embedded sections using a rabbit polyclonal antibody detecting human SPARC (1 : 2000, sc-25574 from Santa Cruz) in combination with a secondary biotynylated goat anti-rabbit antibody and a Vectastain Elite ABC kit (Vector Lab, Burlingame, CA, USA) according to the protocol of the manufacturer.

Detection of SPARC in conditioned medium
Indicated cells were grown in complete medium to 70% confluency in 10 cm dishes. After washing twice with PBS, cells were incubated for 48 h in 10 ml of serum-free medium containing proteinase inhibitors as described (Kornmann et al, 1997). For immunoblot analysis, conditioned medium of five dishes was collected and incubated at 41C overnight after adjusting the pH to 7.4 and adding 50 ml slurry heparin sepharose (CL-6B, Pharmacia Biotech, Piscataway, NY, USA) to pull-down calcium-binding proteins (Kornmann et al, 1997). The beads were collected by centrifugation, washed three times with 0.45 M NaCl/20 mM Tris -HCl (pH 7.4), and resuspended in 2 Â Laemmli buffer. Samples were boiled for 5 min and subjected to immunoblot analysis.

Expression of SPARC and its effects on proliferation and migration in cultured PDAC cells
The effects of exogenous SPARC (5 mg ml -1 ) on single cell movement and migration were investigated in PANC-1 and MIA PaCa-2 cells. SPARC reduced the distance covered within 24 h by 38% and 28% in PANC-1 and MIA PaCa-2 cells, respectively ( Figure 1C). Cell migration was inhibited in the presence of SPARC by 58% and 26%, respectively ( Figure 1D).

Effects of FGFR1 expression on SPARC modulation of FGF1 and FGF2
Modulation of FGF actions by SPARC was reported to depend on FGFR1 expression . It is not known, whether there are differences among the existing FGFR1 variants in mediating SPARC-modulated FGF actions. Therefore, we next investigated SPARC modulation of FGF-depended proliferation in regard to FGFR1-III isoform expression. Wild-type PANC-1, control-transfected (PN5), FGFR1-IIIb over-expressing (PF4), and FGFR1-IIIc over-expressing (PFc51) cells were incubated with FGF1 and FGF2 in the presence and absence of SPARC. As reported earlier, over-expression of FGFR1-IIIc (PFc51) resulted in enhanced FGF1-and FGF2-induced proliferation in comparison with wild-type, control-transfected, and FGFR1-IIIb (PF4) overexpressing cells (Figure 4). FGF2 effects were not markedly altered by exogenous SPARC. Irrespective of the FGFR1-III isoform and the degree of FGF1-induced proliferation, the inhibitory effects of SPARC on FGF1-induced proliferation seemed to be slightly more pronounced in FGFR1-III over-expressing clones compared with cells expressing lower levels of FGFR1 (Figure 4). This observation did not reach any significance.
This was confirmed in a xenograft model in vivo. Immunohistochemical analysis of SPARC protein expression of tumours over-expressing FGFR1-IIIb (Liu et al, 2007b) revealed that SPARC protein was up-regulated in FGFR1-IIIb over-expressing tumours ( Figure 5B).

Interactions of PDAC cells and stromal cells
Human PSCs expressed high levels of SPARC ( Figure 1A). SPARC was also detectable in the pull down of heparin-binding proteins of conditioned medium of PSCs, but not of COLO-357 and PANC-1 pancreatic cancer cells ( Figure 6A). To investigate whether SPARC expression can be altered by paracrine mechanisms, PSCs were incubated with conditioned medium harvested from PSCs and pancreatic cancer cells ( Figure 6B). Endogenous SPARC expression in PSCs was reduced by conditioned medium of COLO-357 and PANC-1 cells added to PSCs for 48 h, but not by conditioned medium of PSCs ( Figure 6B). Conditioned medium of PSCs did not alter SPARC expression in PANC-1 and ASPC-1 cells (data not shown).

DISCUSSION
SPARC is a matricellular protein with antiproliferative and counteradhesive functions (Yan and Sage, 1999). Its function in cancer is discussed in a very controversial manner (Clark and Sage, 2008;Podhajcer et al, 2008;Tai and Tang, 2008). In certain types of cancers, such as melanomas and gliomas, SPARC is associated with a highly aggressive tumour phenotype. In others, mainly ovarian, neuroblastomas, and colorectal cancers, SPARC may function as a tumour suppressor . We showed in this study that exogenous SPARC can inhibit cell proliferation, single cell movement, and migration of cultured PDAC cells. These exogenous SPARC functions were independent of endogenous SPARC expression. As reported (Sato et al, 2003;Guweidhi et al, 2005), the majority of the cell lines did not express endogenous SPARC, probably as a result of aberrant hypermethylation (Sato et al, 2003;Cheetham et al, 2008). Inhibition of endogenous SPARC in cultured human PDAC cells by small hairpin RNA enhanced cell proliferation, migration, colony formation, and xenograft formation. These results indicate that endogenous SPARC may act as a tumour suppressor in PDAC cells, a function SPARC also has in ovarian, neuroblastomas, and colorectal cancers . This function of a tumour suppressor is also supported by our finding that inhibition of endogenous SPARC resulted in a doubling of the mitogenic activity of FGF1 and FGF2, two ligands of the FGF family known to be over-expressed in the majority of pancreatic cancers and to correlate with poor prognosis (Yamanaka et al, 1993).
The presence of FGFR1 is important for mediating SPARC functions in endothelial cells and skeletal myoblasts . All cell lines used in our study expressed various levels of FGFR1 (Kornmann et al, 2002). Inhibition of FGF1-induced proliferation by exogenous SPARC was enhanced in FGFR1-III over-expressing cells compared with controls. This effect was independent of the FGFR1-III isoform. Thus, exogenous SPARC actions depend on the FGFR1 levels, but are independent of the domain III isoform expression.
Down-regulation of SPARC in pancreatic cancer cells is believed to depend on DNA methylation (Sato et al, 2003;Guweidhi et al, 2005;Infante et al, 2007;Cheetham et al, 2008). Another reason contributing to the down-regulation of SPARC in human pancreatic cancer cells may be the over-expression of FGFR1-IIIc. FGFR1-IIIc is the predominant FGFR1 isoform expressed in PDAC (Kornmann et al, 2002). FGFR1-IIIb usually expressed in cells of epithelial origin is almost absent in PDAC (Kornmann et al, 2002). Re-or over-expression of FGFR1-IIIb in PDAC cells resulted in a marked up-regulation of endogenous SPARC in vitro and in vivo. In contrast, an additional over-expression of FGFR1-IIIc further lowered endogenous SPARC expression. We recently showed that over-expression of FGFR1-IIIb in PDAC cells reverted the malignant phenotype inhibiting proliferation, single cell movement, and migration in vitro, as well as xenograft formation and growth in vivo (Liu et al, 2007b). On the other hand, it is well known that over-expression of FGFR1-IIIc enhances the malignant phenotype of PDAC (Wagner et al, 1998;Kornmann et al, 2002).
FGFR1-IIIb over-expression in PDAC cells resulted in strong p38-MAPK and JNK activitation (Liu et al, 2007a, b). In this study, we also investigated the effects of specific inhibitors of these kinases on endogenous SPARC levels. Our results showed that endogenous SPARC levels could be down-regulated by inhibition of p38-MAPK, but not of JNK. Therefore, our observations suggest that modulation of endogenous SPARC expression may be one of the mechanisms resulting in the different phenotypes seen for the FGFR1-III domain variants and that the observed FGFR1-IIIbinduced induction of endogenous SPARC is mediated through p38-MAPK.
Recent studies investigating SPARC expression in human pancreatic tissues reported high levels of SPARC in the surrounding stromal tissue harbouring fibroblasts and PSCs, whereas SPARC was often absent in the cancer cells (Guweidhi et al, 2005;Infante et al, 2007). High SPARC expression in the stroma portended a poor patient prognosis (Infante et al, 2007). We showed in our study that PSCs express higher levels of endogenous SPARC than cultured PDAC cells and that SPARC is detectable in the conditioned medium of PSCs. Our study also revealed that conditioned medium of pancreatic cancer cells down-regulated endogenous SPARC expression of PSCs. In contrast, co-culture of fibroblasts in the presence of PDAC cells augmented SPARC expression in fibroblasts (Sato et al, 2003), suggesting that high SPARC expression in the tumour stroma may be mainly a result of augmented SPARC expression in stromal fibroblasts.
In summary, we showed that inhibition of endogenous SPARC enhances the malignant phenotype of PDAC cells and showed that endogenous SPARC expression is regulated by FGFR1 domain III isoform expression. On the basis of these observations, we conclude that endogenous SPARC levels can contribute to the reversion of the malignant phenotype and may, therefore, act as a tumour suppressor in human PDAC cells. Future studies in human pancreatic cancer could aim at the design of treatment strategies specifically targeting SPARC -FGFR1 interactions.