Divalent cations modulate the integrin‐mediated malignant phenotype in pancreatic cancer cells

Abstract

We have previously demonstrated that pathophysiological shifts in the concentrations of extracellular Mg²⁺ and Ca²⁺ activate the α₂β₁ integrin‐mediated malignant phenotype on type I collagen in pancreatic cancer cells, as evidenced by increased adhesion, migration and proliferation. In the present study, we examined the integrin and divalent cation specificity of pancreatic cancer cell interactions with other physiologically relevant extracellular matrix proteins, including fibronectin, type IV collagen, laminin and vitronectin. Our results indicate that, like α₂β₁ integrin‐mediated interactions with type I collagen, β₁ integrin‐mediated adhesion to fibronectin, type IV collagen and laminin are promoted by Mg²⁺ but not by Ca²⁺. On vitronectin, cells attach via α_vβ₅ and β₁ integrins, and in the presence of either divalent cation. We also demonstrate that, like type I collagen, pancreatic cancer cell migration and proliferation on fibronectin, laminin and type IV collagen is maximal when Mg²⁺ is present at concentrations that promote optimal adhesion and Ca²⁺ is present at concentrations less than Mg²⁺. On vitronectin, Panc‐1 cell migration is maximal with decreased Mg²⁺ and increased Ca²⁺, but the reverse is true for BxPC‐3 cells. Both cell lines exhibited maximal proliferation with increased Mg²⁺ and decreased Ca²⁺, however. Together with evidence indicating that the in vivo local tumor microenvironment contains increased Mg²⁺ and decreased Ca²⁺, our studies demonstrate that such divalent cation shifts could activate the integrin‐mediated malignant phenotype in pancreatic cancer. (Cancer Sci 2008; 99: 1553–1563)

Abbreviations:

BSA

bovine serum albumin

DMEM

Dulbecco's modified Eagle medium

ECM

extracellular matrix

FBS

fetal bovine serum

Fn

fibronectin

IgG

immunoglobulin

Ln

laminin

PBS

phosphate‐buffered saline

TBS

Tris‐buffered saline

Vn

vitronectin

Pancreatic cancer is characterized by a hallmark desmoplastic response that includes upregulated expression of type I collagen, as well as other ECM proteins.^{( 1 )} Recent studies by our laboratory and others indicate that the α₂β₁ integrin‐mediated interaction of pancreatic cancer cells with type I collagen promotes a malignant phenotype, as defined by increased adhesion, migration and proliferation relative to other ECM proteins in vitro, and by increased tumorigenesis in vivo.^{( 1} , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ^{11 )}

A common characteristic of integrins, the heterodimeric transmembrane receptor proteins that mediate cell interactions with the ECM,^{( 12} , ^{13 )} is their absolute requirement for divalent cations, including Mg²⁺ and Ca²⁺, to function.^{( 12} , ¹⁴ , ¹⁵ , ¹⁶ , ^{17 )} Previous studies over the past three decades using immunological and biophysical approaches have indicated that divalent cations affect integrin receptor conformation, which results in changes in affinity and avidity for ligand.^{( 18} , ¹⁹ , ²⁰ , ²¹ , ²² , ^{23 )} We recently examined the regulatory role of divalent cations on α₂β₁ integrin‐mediated adhesion, migration and proliferation of pancreatic cancer cells on type I collagen.^{( 7} , ^{24 )} Our results indicated that cells attached to type I collagen maximally when Mg²⁺ was greater than approximately 1 mM, and that addition of increasing concentrations of Ca²⁺ reduced this Mg²⁺‐dependent adhesion. The effects of divalent cations were reversible and not detrimental to the cell, in that previous attachment to type I collagen in Mg²⁺ was subsequently reversed by the addition of Ca²⁺, and vice versa. We also demonstrated that cells migrated and proliferated on type I collagen maximally when Mg²⁺ was present at concentrations that promoted maximal adhesion and Ca²⁺ was present at concentrations less than Mg²⁺. Additionally, we demonstrated that the α₂β₁ integrin, bound to type I collagen by affinity chromatography in Mg²⁺, could be eluted with Ca^2+ (4). These data are consistent with previous studies of the α₂β₁ integrin, type I collagen and cell types involved in cutaneous wound repair, including keratinocytes, endothelial cells, fibroblasts and macrophages,^{( 25} , ²⁶ , ^{27 )} where strong parallels to the pancreatic cancer paradigm have been described.^{( 28} , ^{29 )}

During normal cutaneous wound healing in vivo, extracellular Mg²⁺ increases and Ca²⁺ decreases, activating the α₂β₁ integrin‐mediated migration of the wound healing cell types mentioned above on type I collagen.^{( 27 )} Similar shifts probably occur in pancreatic cancer as well, because solid tumors are distinguished by increased Mg²⁺ load, even at the expense of healthy adjacent tissue.^{( 30} , ^{31 )} Additionally, pancreas juice produced in the range of 1500–2000 mL/day, contains 1200‐fold more Mg²⁺ (137–244 µM) than Ca²⁺ (0.111–0.197 µM).^{( 32 )} As pancreatic ductal epithelial basement membranes have been shown to be discontinuous or absent in pancreatic cancer,^{( 1 )} pancreas juice could be expected to leak into the pancreatic cancer tumor microenvironment and further disrupt the divalent cation milieu. And while it is not yet clear whether divalent cation shifts in pancreatic cancer are intracellular, extracellular or both, these data collectively indicate that divalent cation shifts occur in the local tumor microenvironment in a manner similar to those found in wound healing, and could activate the α₂β₁ integrin‐mediated malignant phenotype on type I collagen in pancreatic cancer.

Other ECM proteins are also expressed in the desmoplastic pancreatic cancer tumor microenvironment,^{( 1 )} however, and little is known about their integrin‐ or divalent cation‐binding specificity. We therefore examined the integrin‐binding specificity of pancreatic cancer cells on other physiologically relevant ECM proteins, including Fn, type IV collagen, Ln and Vn. We then determined the effects of divalent cations on integrin‐mediated interactions with these substrates, including adhesion, migration and proliferation. Our results indicate that, like α₂β₁ integrin‐mediated type I collagen adhesion, β₁ integrin‐mediated cell adhesion to Fn, type IV collagen and Ln are promoted by Mg²⁺ but not by Ca²⁺. On Vn, cells bind using both α_vβ₅ and β₁ integrins, and in the presence of either cation. We also demonstrate that, like type I collagen, increased Mg²⁺ and decreased Ca²⁺ promotes maximal migration and proliferation on Fn, Ln and type IV collagen. On Vn, while BxPC‐3 migration was maximal with increased Mg²⁺ and decreased Ca²⁺ and the reverse was true for Panc‐1, both cell lines exhibited maximal proliferation on Vn with increased Mg²⁺ and decreased Ca²⁺. These data indicate that pathophysiological divalent cation shifts occurring in the tumor microenvironment of pancreatic cancer may be critical regulators of the integrin‐mediated malignant phenotype on the ECM generally, in a manner similar to type I collagen.

Materials and Methods

Cells.  Capan‐1, Colo‐357, AsPC‐1, BxPC‐3, MiaPaCa‐2 and Panc‐1 cells were from ATCC (Rockville, MD, USA). FG cells, a metastatic variant of Colo‐357 cells, have been described.^{( 5} , ⁶ , ^{8 )} Cells were cultured in DMEM (high‐glucose) supplemented with 10% FBS in a humidified atmosphere containing 5% CO₂ at 37°C.

Adhesion assays.  Adhesion assays were conducted as described^{( 7} , ¹⁷ , ^{24 )} using 3.5 × 10⁴ BxPC‐3, 2 × 10⁴ Panc‐1, 4 × 10⁴ AsPC‐1, 7.5 × 10⁴ Colo‐357, 4 × 10⁴ MiaPaCa‐2 or 3.5 × 10⁴ Capan‐1 cells/well in TBS containing 1 mg/mL BSA supplemented with titrations of either MgCl₂ or CaCl₂, and 96‐well plates that were previously coated with 5 µg/mL type IV collagen, 5 µg/mL Vn, 25 µg/mL Ln or 25 µg/mL Fn. These coating concentrations promote maximal adhesion for each cell line on each ECM protein.^{( 5 )}

Inhibition of adhesion assays.  Inhibition of cell adhesion assays were performed as described.^{( 4} , ⁵ , ⁶ , ²⁴ , ^{25 )} Briefly, 96‐well microtiter plates were coated with 5 µg/mL type IV collagen, 5 µg/mL Vn, 25 µg/mL Ln or 25 µg/mL Fn. Then, 5 × 10⁴ MiaPaC‐2 or Colo‐357 cells, or 3.5 × 10⁴ AsPC‐1, Panc‐1 or 2.4 × 10⁴ BxPC‐3 cells, were added to each well in serum‐free DMEM supplemented with 1 mg/mL BSA. Purified monoclonal antibodies were then added at a final concentration of 25 µg/mL. After 45 min at 37°C, attached cells were fixed, stained, solubilized and quantified as described.^{( 4} , ⁵ , ⁶ , ²⁴ , ^{25 )} The function‐blocking anti‐integrin antibodies used in these studies, P5D2 (β₁), LM609 (α_vβ₃), ASC‐3 (β₄) and P1F6 (α_vβ₅) (Chemicon International, Temecula, CA, USA), are the same IgG₁ isotype and have been described.^{( 33} , ³⁴ , ³⁵ , ^{36 )}

Migration assays.  Modified Boyden chamber migration assays were conducted as described (Neuro Probe, Gaithersburg, MD, USA).^{( 5} , ⁶ , ⁷ , ⁸ , ²⁴ , ²⁵ , ²⁶ , ^{27 )} Briefly, the chamber consists of two compartments separated by an ECM‐coated filter (5 µg/mL of each ECM protein^{( 5 )}), and migration was measured by counting the number of cells crossing the filter through 8‐µm pores. Lower chambers were filled with serum‐ and Ca²⁺‐free DMEM (Invitrogen, Carlsbad, CA, USA) supplemented with 1 mg/mL BSA. The normal Mg²⁺ concentration of the media was supplemented to a final concentration of 3.5 mM and Ca²⁺ was titrated at the indicated concentrations. Upper chambers were filled with 2.5 × 10⁴ BxPC‐3 or Panc‐1 cells that were serum‐starved 24 h prior to assay, in media consistent with that of the lower chamber. Lower chamber final volumes were 30 µL and the upper chambers were 50 µL. The entire apparatus was then incubated 18–22 h at 37°C. After the incubation period, the filters were fixed, stained and quantitated by counting five high‐powered fields (×100 magnification) per well using an inverted light microscope (Olympus BH 2; Olympus, Tokyo, Japan) as described.^{( 5} , ⁶ , ⁷ , ²⁴ , ²⁵ , ²⁶ , ^{27 )}

Immunofluorescence studies.  Immunofluorescence studies were conducted as described.^{( 7} , ^{26 )} Briefly, 13‐mm glass coverslips were coated with Fn (20 µg/mL) and blocked with 1 mg/mL BSA in PBS. Serum‐starved cells were then plated in serum‐free DMEM supplemented with 1 mg/mL BSA at 5 × 10⁵ cells/well for 24 h at 37°C. The cells were then fixed with 3% paraformaldehyde and permeabilized with 0.1% Triton X‐100 in PBS. After blocking with 1% normal goat serum/1% BSA in PBS, the coverslips were incubated with an anti‐E‐cadherin mAb (Chemicon International) at 10 µg/mL as described.^{( 6} , ^{7 )} After rinsing, coverslips were incubated with fluorescein‐isothiocyanate‐conjugated secondary antibody (goat antimouse IgG at 1:100) (Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Coverslips were rinsed, mounted and fluorescence microscopy was performed (Olympus BX‐60 microscope equipped with a Spot Digital Imaging package; Diagnostic Instruments, Burlingame, CA, USA).

Proliferation assays.  Proliferation assays were conducted as described.^{( 4} , ⁵ , ⁸ , ^{24 )} Briefly, 96‐well microtiter plates were coated with type IV collagen at 5 µg/mL, Vn and Fn at 10 µg/mL, or Ln at 25 µg/mL. These coating concentrations promote maximal proliferation of the cells used in these studies on the indicated ECM proteins.^{( 5 )} Twenty‐four hour, serum‐starved cells (5 × 10³/well) were cultured under serum‐free conditions on the indicated ECM proteins over a 3‐day time‐course. The normal Mg²⁺ concentration of the media was supplemented to a final concentration of 3.5 mM and Ca²⁺ was titrated at the indicated concentrations. Proliferation was quantified by measuring the absorbance at 450 nm and subtracting the value obtained for each cell line on each ECM protein at initial seeding using CellTiter 96 Aqueous One Solution Cell Proliferation Assay reagent according to manufacturer's instructions (Promega, Madison, WI, USA).

Statistical analysis.  Statistical significance was determined using two‐tailed Student's t‐tests.

Results

Integrin‐binding specificity of pancreatic cancer cells on the ECM.  Previous studies indicate that the β integrin subunit determines the divalent cation requirements for cellular interactions with the ECM^{( 17} , ^{25 )} and that pancreatic cancer cells express multiple β subunits, including β₁, β₃, β₄ and β₅.^{( 1} , ^{5 )} The specific integrins mediating pancreatic cancer cell interactions with type IV collagen, Fn, Vn and Ln have not been determined, however. Before examining divalent cation‐dependency, we first determined the β subunit‐binding specificity of pancreatic cancer cells on type IV collagen, Fn, Vn and Ln in adhesion assays with function‐blocking monoclonal antibodies directed against specific β subunits. As shown in Fig. 1(a–d), Colo‐357, AsPC‐1, Panc‐1, MiaPaCa‐2 and BxPC‐3 pancreatic cancer cells generally bind to Fn, Ln and type IV collagen in a β₁ integrin‐dependent manner. MiaPaCa‐2 cells also show statistically significant involvement of the α_vβ₅ integrin in Fn adhesion (~30% inhibition). On Vn, while the α_vβ₅ antibody was most effective at inhibiting adhesion of all five pancreatic cancer cell lines to Vn (~20–90%), the β₁ antibody was also active, though to a lesser extent (~0–30%). No inhibition with the β₁ antibody was observed with Panc‐1 cells on Vn. These data indicate that pancreatic cancer cells bind to Fn, Ln and type IV collagen predominantly via β₁ integrins. On Vn, cells bind predominantly via the α_vβ₅ integrin, with involvement from β₁ integrins as well.

Figure 1.

Effect of inhibitory anti‐integrin monoclonal antibodies on pancreatic cancer cell adhesion on fibronectin (Fn), vitronectin (Vn), laminin (Ln) and type IV collagen. Inhibition of cell adhesion assays were performed as described^{( 4} , ⁵ , ⁶ , ^{25 )} adding 5 × 10⁴ MiaPaC‐2 or Colo‐357, 3.5 × 10⁴ AsPC‐1 or Panc‐1, or 2.5 × 10⁴ BxPC‐3 cells/well of 96‐well microtiter plates coated with (a) 25 µg/mL Fn, (b) 5 µg/mL Vn, (c) 25 µg/mL Ln or (d) 5 µg/mL type IV collagen. Function‐blocking monoclonal antibodies were added at a final concentration of 25 µg/mL. Assays were quantified as previously described.^{( 4} , ⁵ , ⁶ , ^{25 )} Results are expressed as percentage control (no antibody), and represent the mean absorbance obtained for each cell line on each extracellular matrix (ECM) protein at 595 nm ± standard error of the mean from three (Fn, Ln, type IV collagen) or four (Vn) experiments conducted in triplicate. *P < 0.05.

Divalent cation‐dependent adhesion of pancreatic cancer cells on the ECM.  We next examined the divalent cation‐dependency of integrin‐mediated pancreatic cancer cell attachment to Fn, Vn, Ln and type IV collagen using adhesion assays. Fig. 2 indicates that AsPC‐1, Panc‐1, BxPC‐3 and Colo‐357 cell adhesion to Fn and type IV collagen was promoted by Mg²⁺, with maximal adhesion observed between 3–30 mM. By contrast, adhesion of these cell lines on Fn and type IV collagen was not promoted by Ca²⁺ at any of the concentrations tested through 30 mM. Similar results were observed with Capan‐1 and MiaPaCa‐2 cells, except that MiaPaCa‐2, which do not express the α₂β₁ or α₁β₁ type IV collagen‐binding integrins,^{( 1} , ⁴ , ^{5 )} did not attach to type IV collagen with either divalent cation (data not shown). Fig. 2 also demonstrates a similar Mg²⁺‐dependent promotion of integrin‐mediated cell attachment to Ln substrates, but not by Ca²⁺, though in AsPC‐1 and Panc‐1 cells, there was a slight (~10% of maximum) promotion of Ca²⁺‐dependent adhesion observed between approximately 1–10 mM. Similar results were also observed with MiaPaCa‐2 cells (not shown). BxPC‐3 and Colo‐357 cells, by contrast, exhibited no Ca²⁺‐dependent adhesion on Ln substrates. Similar results were also observed with Capan‐1 cells (not shown). On Vn, Fig. 2 demonstrates further that AsPC‐1 and Panc‐1 cells attach in both Mg²⁺ and Ca²⁺, with optimal adhesion occurring between approximately 3–30 mM. Similar results were also observed with MiaPaCa‐2 and Capan‐1 cells (not shown), though Capan‐1 cell adhesion in Ca²⁺ was minimal (~10% of maximum). By contrast, BxPC‐3 and Colo‐357 cells attach to Vn in Mg²⁺ but not in Ca²⁺. These data indicate generally that, like α₂β₁ integrin‐mediated adhesion to type I collagen,^{( 24 )}β₁ integrin‐mediated adhesion to Fn, Ln and type IV collagen is promoted by Mg²⁺, but not by Ca²⁺, whereas α_vβ₅ and β₁ integrin‐mediated cell adhesion to Vn occurs in the presence of either divalent cation.

Figure 2.

Effect of divalent cations on pancreatic cancer cell adhesion on fibronectin (Fn), vitronectin (Vn), laminin (Ln) and type IV collagen. (a) 4 × 10⁴ AsPC‐1, (b) 2 × 10⁴ Panc‐1, (c) 3.5 × 10⁴ BxPC‐3 or (d) 7.5 × 10⁴ Colo‐357 pancreatic cancer cells were added to each well of 96‐well microtiter plates previously coated with 25 µg/mL Fn, 5 µg/mL Vn, 25 µg/mL Ln or 5 µg/mL type IV collagen in Tris‐buffered saline supplemented with 1 mg/mL bovine serum albumin (BSA) in the presence of the indicated concentrations of CaCl₂ or MgCl₂ and incubated 1 h at 37°C. The remainder of the assay was conducted as previously described.^{( 4} , ⁵ , ⁶ , ^{25 )} Results are expressed as percentage maximum for each cell line on each extracellular matrix substrate and represent the mean absorbance observed at 595 nm for each cell line ± standard error of the mean from three (Panc‐1 and BxPC‐3) or four (AsPC‐1 and Colo‐357) experiments conducted in duplicate. (a) AsPC‐1 cells; (b) Panc‐1 cells; (c) BxPC‐3 cells; (d) Colo‐357 cells. () Mg²⁺; () Ca²⁺.

Divalent cation‐dependent migration of pancreatic cancer cells on the ECM.  In modified Boyden chamber migration assays, BxPC‐3 cells (Fig. 3a) were migratory on Fn, Ln, type IV collagen and Vn using concentrations of extracellular Mg²⁺ that supported maximal cell adhesion (3.5 mM). BxPC‐3 cell migration could be increased up to twofold, however, when both cations were utilized in combination, with Mg²⁺ present at 3.5 mM and Ca²⁺ present at concentrations less than Mg²⁺. As the Ca²⁺ concentration increased over that of Mg²⁺, BxPC‐3 migration declined, regardless of substrate. With Panc‐1 cells (Fig. 3b), while migration was essentially similar to BxPC‐3 cells on Fn, Ln and type IV collagen, some significant differences were apparent. First, maximal Panc‐1 cell migration was observed in the presence of 3.5 mM Mg²⁺ only. No Ca²⁺ was required to achieve maximal migration. In fact, on Ln and type IV collagen, addition of Ca²⁺ to Mg²⁺‐containing media reduced migration. Second, Panc‐1 cell migration on Fn was not significantly inhibited by adding Ca²⁺ to Mg²⁺‐containing media, except at the highest concentration (15 mM).

Figure 3.

Effect of divalent cations on BxPC‐3 and Panc‐1 pancreatic cancer cell migration on fibronectin, vitronectin, laminin and type IV collagen. (a) BxPC‐3 and (b) Panc‐1 cell migration was determined and quantitated as previously described^{( 5} , ⁶ , ⁷ , ⁸ , ²⁵ , ²⁶ , ^{27 )} using 2.5 × 10⁴ cells/well in 3.5 mM Mg²⁺ plus a titration of Ca²⁺ on extracellular matrix (5 µg/mL)‐coated filters. The results, expressed as percentage maximum for each cell line, represent the mean ± standard error of the mean of three experiments conducted in triplicate. For BxPC‐3 cells, 100% was equal to 318 cells/high‐power field. For Panc‐1 cells, 100% was equal to 111 cells/high‐power field.

The most striking difference in migration between the two cell lines was observed on Vn. As shown in Fig. 3(a), BxPC‐3 cell migration on Vn was maximal with Mg²⁺ present at 3.5 mM and Ca²⁺ present at concentrations less than Mg²⁺. By contrast, Panc‐1 cell migration in the presence of 3.5 mM Mg²⁺ alone was less than 50% of maximum, with optimal migration observed with Mg²⁺ at 3.5 mM and Ca²⁺ present between 1.88 and 7.5 mM (Fig. 3b).

Fig. 4(a,b) depicts the morphological differences in BxPC‐3 and Panc‐1 cell migration observed under different divalent cation conditions. In the presence of 3.5 mM Mg²⁺, BxPC‐3 cells were singular and well spread on all ECM proteins. Addition of 0.47 mM Ca²⁺ to 3.5 mM Mg²⁺‐containing medium increased BxPC‐3 cell migration nearly twofold on all substrates. Addition of Ca²⁺ at concentrations greater than Mg²⁺ (3.5 mM) resulted in increased BxPC‐3 cell–cell adhesion and decreased cell migration on all substrates. By contrast, Panc‐1 cells (Fig. 4b) showed increased or decreased cell migration on the various ECM proteins at the indicated divalent cation concentrations, consistent with Fig. 3(b). Of particular interest, note that with 3.5 mM Mg²⁺ and 3.75 mM Ca²⁺, qualitatively more Panc‐1 cell migration is seen on Vn compared to BxPC‐3 cells. Interestingly, no Panc‐1 cell–cell contacts were observed under any condition.

Figure 4.

Effect of divalent cations on the pancreatic cancer cell migratory phenotype on fibronectin, vitronectin, laminin and type IV collagen. Light photomicrographs shown are representative examples of (a) BxPC‐3 or (b) Panc‐1 cell migration conducted on extracellular matrix (5 µg/mL)‐coated filters at the indicated divalent cation concentrations. Photomicrographs were obtained using an Olympus BX‐60 microscope equipped with a Spot Digital Imaging package.

We have previously shown that Panc‐1 cells grow on three‐dimensional type I collagen scaffolds as single cells relative to the large aggregates seen with BxPC‐3, FG and Colo‐357 cells after 72 h in serum‐free culture.^{( 4 )} It may be that E‐cadherin is not functional in Panc‐1 cell lines. In support, Weinel et al. reported that Panc‐1 cells were negative for E‐cadherin by immunohistochemistry.^{( 37 )} We therefore examined the expression and localization of E‐cadherin in BxPC‐3 and Panc‐1 cells after 24 h in serum‐free culture on Fn‐coated glass coverslips. Results shown in Fig. 5 under low and high magnification indicate robust E‐cadherin expression and localization in BxPC‐3 cell–cell contacts, in agreement with our previously published results.^{( 7 )} Note the ‘stitch‐like’ cross‐striations in E‐cadherin localization in BxPC‐3 cell–cell contacts. This pattern of expression is similar to that previously shown with HaCaT keratinocytes on type I collagen.^{( 26 )} Panc‐1 cells, by contrast, showed only faint and disorganized E‐cadherin expression. It is therefore possible that differences in Ca²⁺‐dependent E‐cadherin function^{( 38 )} between Panc‐1 and BxPC‐3 cells may account for observed differences in divalent cation‐dependent cell–cell adhesion and cell migration.

Figure 5.

E‐cadherin expression and localization in BxPC‐3 and Panc‐1 cells after 24 h in serum‐free culture on fibronectin (Fn). BxPC‐3 and Panc‐1 cells were cultured on Fn (20 µg/mL)‐coated glass coverslips for 24 h, and immunofluorescence studies were conducted as described in Materials and Methods for E‐cadherin expression and localization. Representative photomicrographs taken under the exact same conditions for each cell line using low‐ and high‐power magnifications are shown.

Together, these data indicate that, like type I collagen,^{( 24 )} changes in extracellular Mg²⁺ and Ca²⁺ dramatically effect integrin‐mediated pancreatic cancer cell migration on Fn, Vn, Ln and type IV collagen. Generally, maximal cell migration was promoted when Mg²⁺ was present at concentrations that yielded maximal adhesion (3.5 mM) and Ca²⁺ was present at concentrations less than Mg²⁺ regardless of substrate, except for Panc‐1 cells on Vn, where Ca²⁺ concentrations greater than Mg²⁺ promoted maximal migration.

Divalent cation‐dependent proliferation of pancreatic cancer cells on multiple ECM substrates.  As shown in Fig. 6(a,b), Mg²⁺ alone supports cell proliferation over a 72‐h time‐course with both BxPC‐3 and Panc‐1 cells. The addition of Ca²⁺ increased BxPC‐3 cell proliferation on all ECM proteins tested by as much 35% compared to Mg²⁺ alone, as long as Ca²⁺ was present at concentrations less than Mg²⁺. As the Ca²⁺ concentration exceeded that of Mg²⁺, BxPC‐3 cell proliferation declined on all substrates. With Panc‐1 cells, proliferation on Fn, Ln and type IV collagen was maximal in Mg²⁺ only and progressively declined with increasing Ca²⁺ (Fig. 6b). On Vn, however, Panc‐1 cell proliferation was maximal from 0–7.5 mM Ca²⁺, after which proliferation significantly declined. The presence of an apparent spike in the proliferation of Panc‐1 cells on all ECM proteins tested in the presence of 3.5 mM Mg²⁺ and 7.5 mM Ca²⁺ should be noted. Further evaluation by Student's t‐tests, however, indicate that these data points are not statistically different from those obtained in 3.5 mM Mg²⁺ and 3.75 mM Ca²⁺ (Fn, P = 0.804; Vn, P = 0.697; Ln, P = 0.679; type IV collagen, P = 0.785). Taken together, these data indicate generally that cell proliferation is maximal when Mg²⁺ is present at concentrations that support maximal adhesion and Ca²⁺ is present at concentrations less than Mg²⁺, regardless of substrate.

Figure 6.

Effect of divalent cations on pancreatic cancer cell proliferation on the extracellular matrix. (a) BxPC‐3 and (b) Panc‐1 cell proliferation was determined and quantitated as previously described^{( 4} , ⁵ , ^{8 )} using 5.0 × 10³ cells/well in 3.5 mM Mg²⁺ plus a titration of Ca²⁺ on fibronectin (10 µg/mL), vitronectin (10 µg/mL), laminin (25 µg/mL) or type IV collagen (5 µg/mL)‐coated plates over a 72 h time‐course. The results, expressed as percentage maximum for each cell line, represent the mean ± standard error of the mean of three experiments done in duplicate. 100% is the maximum mean absorbance obtained at 450 nm for each cell line throughout the time‐course as described in Materials and Methods. () 24 h; () 48 h; () 72 h.

Discussion

Evidence derived from studies employing function‐blocking monoclonal antibodies directed against specific integrins or integrin subunits in inhibition of cell adhesion assays indicate generally that pancreatic cancer cells attach to Fn, Ln and type IV collagen in a strict β₁ integrin‐mediated manner. On Vn, by contrast, cells generally bind predominantly via the α_vβ₅ integrin, with a minor but significant contribution from β₁ integrins as well.

We also conducted adhesion assays designed to determine the divalent cation‐specificity of integrin‐mediated pancreatic cancer cell interactions with Fn, Ln, type IV collagen and Vn. Our data indicate generally that, similar to α₂β₁ integrin‐mediated type I collagen adhesion,^{( 4} , ⁷ , ^{24 )} β₁ integrin‐mediated cell adhesion to Fn, Ln and type IV collagen is promoted by Mg²⁺, but not by Ca²⁺, whereas α_vβ₅ and β₁ integrin‐mediated adhesion to Vn occurs in the presence of either divalent cation. These data are consistent with previous studies using primary cultures of human dermal fibroblasts and human epidermal keratinocytes in adhesion assays on type I collagen, laminin and fibronectin, in that β₁ integrin‐mediated adhesion to these ECM proteins was also shown to be promoted by Mg²⁺ but not by Ca²⁺.^{( 39 )} And though it has been previously shown with the human MG‐63 osteosarcoma fibroblast cell line that α_vβ₅ integrins organize in focal contacts on Vn in the presence of 1 mM Mg²⁺ or 1 mM Ca^{2 +   ( 40 )}, to the best of our knowledge, this is the first direct demonstration of the concentrations of these divalent cations required to promote optimal α_vβ₅ integrin‐mediated adhesion to Vn in any cell type.

In modified Boyden chamber migration assays, our results also indicate for the first time with any cell type that, like pancreatic cancer cells,^{( 7} , ^{24 )} fibroblasts,^{( 25} , ^{27 )} keratinocytes,^{( 26} , ^{27 )} endothelial cells^{( 27 )} and macrophages^{( 27 )} on type I collagen, the presence of Mg²⁺ at concentrations that promote maximal cell adhesion together with Ca²⁺ at concentrations less than Mg²⁺ promote maximal pancreatic cancer cell migration regardless of ECM substrate, except for Panc‐1 cells on Vn, where the presence of Ca²⁺ at concentrations greater than Mg²⁺ promotes maximal migration. And also similar to our recent results with pancreatic cancer cells on type I collagen,^{( 24 )} our proliferation data indicate for the first time with any cell type that the presence of Mg²⁺ at concentrations that promote optimal adhesion along with Ca²⁺ at concentrations less than Mg²⁺ promote maximal proliferation, regardless of ECM substrate. Taken together, these results indicate that changes in the relative concentrations of extracellular Mg²⁺ and Ca²⁺ affect the activation of integrin‐mediated adhesion, migration and proliferation of cancer cells on physiologically relevant ECM proteins. These data also indicate that the divalent cation concentrations in the desmoplastic tumor microenvironment, where other cell types are present, including fibroblasts, endothelial cells and inflammatory cells,^{( 1 )} could play a major role in the activation of the integrin‐mediated malignant phenotype in pancreatic cancer.

Previous mechanistic studies by our laboratory indicated that the use of serum‐free media containing 0.1 mM Ca²⁺ and 1.0 mM Mg²⁺ downregulated the overall expression as well as the localization of E‐cadherin in FG pancreatic cancer cell–cell contacts. This downregulation of E‐cadherin was accompanied by increased α₅β₁ integrin‐mediated migration on Fn compared to normal media conditions (1.8 mM Ca²⁺ and 1.0 mM Mg²⁺). Interestingly, Fn does not promote strong FG cell migration under normal media conditions.^{( 6} , ^{7 )} These data are consistent with previous studies by our laboratory examining keratinocyte migration on type I collagen, where lowered Ca²⁺ concentrations downregulated the overall expression levels and localization of E‐cadherin in cell–cell contacts, resulting in increased α₂β₁ integrin‐mediated migration.^{( 26 )}

Our present studies indicate that E‐cadherin is present in BxPC‐3 cell–cell contacts but not in Panc‐1 cells, where obvious cell–cell contacts are absent (Fig. 5). In the Mg²⁺‐only cultures on all ECM proteins tested, however, BxPC‐3 cells appear as single cells with no cell–cell contacts apparent (Fig. 4). These data are consistent with the well‐established Ca²⁺‐dependency of E‐cadherin‐mediated cell–cell adhesion.^{( 38 )} These data also indicate that, like HaCaT keratinocytes and FG pancreatic cancer cells,^{( 24} , ^{26 )} increasing Mg²⁺ and decreasing Ca²⁺ downregulates Ca²⁺‐dependent, E‐cadherin‐mediated cell–cell adhesion. This downregulation of E‐cadherin expression and localization is accompanied by increased integrin‐mediated pancreatic cancer cell migration on the ECM generally.

It has been previously shown by fluorescence‐activated cell sorting that alterations in the concentrations of divalent cations do not affect cell surface expression levels of β₁, α₁, α₂, α₅ or α₆ integrin subunits in primary cultures of human dermal fibroblasts or human epidermal keratinocytes.^{( 39 )} In agreement, β₁ integrin subunit expression levels have also been shown to be unaffected by alterations in divalent cation concentrations in primary cultures of human eosinophils.^{( 41 )} Preliminary examination of BxPC‐3 and Panc‐1 cell lysates by immunoblotting after 72 h in culture on Fn, Vn, Ln or types I or IV collagen, and in the presence of either 3.5 mM Mg²⁺/0 mM Ca²⁺, 3.5 mM Mg²⁺/0.5 mM Ca²⁺ or 3.5 mM Mg²⁺/3.75 mM Ca²⁺, indicate that there is no difference in the expression of the α₂β₁ integrin (John Grzesiak, personal communication, 2008). These data are consistent with results from a number of published studies over the past three decades, including immunological and biophysical approaches, which indicate that divalent cations modulate integrin activity by conformational alterations that affect receptor avidity and affinity for ligand.^{( 18} , ¹⁹ , ²⁰ , ²¹ , ²² , ^{23 )}

Under normal physiological conditions in vivo, Ca²⁺ is approximately 1.5 mM greater than Mg²⁺ in serum.^{( 25} , ²⁶ , ^{27 )} Our in vitro results indicate that pancreatic cancer cells are capable of a certain level of integrin‐mediated migration (3, 4), and proliferation (Fig. 6) on Fn, Ln, type IV collagen and Vn under these conditions. However, our data also indicate that the potential exists for at least a twofold increase in this activity. During normal cutaneous wound healing, such pathophysiological shifts in the concentrations of extracellular Mg²⁺ and Ca²⁺ occur locally and early in the process, activating the α₂β₁ integrin‐mediated migration of various wound healing cell types on type I collagen, including keratinocytes, fibroblasts, endothelial cells and macrophages.^{( 25} , ²⁶ , ^{27 )} It is not obvious whether sufficient shifts in the concentrations of extracellular Mg²⁺ and Ca²⁺ occur locally in the tumor microenvironment of pancreatic cancer in vivo.

As mentioned earlier and similar to wound healing, solid tumors are characterized by increased Mg²⁺ load, even at the expense of healthy adjacent tissue.^{( 30} , ³¹ , ^{42 )} Additionally, pancreas juice contains Mg²⁺ at a 1200‐fold excess relative to Ca^{2 + ( 32 )} and could be expected to leak from discontinuous or absent ducts which are characteristic of pancreatic cancer.^{( 1 )} Unlike cutaneous wound repair, where a wound bed is established and wound fluid can be collected and tested, it is not possible to directly determine the divalent cation concentrations of extracellular fluid in the pancreatic cancer tumor microenvironment. However, these data collectively indicate that changes in the concentrations of Mg²⁺ and Ca²⁺ similar to wound healing^{( 27 )} do occur in pancreatic cancer. Our data indicate that such changes in divalent cation concentrations could be a critical regulator of the integrin‐mediated malignant phenotype in pancreatic cancer.

A common problem encountered during resection of pancreatic cancer tumors is a high incidence of postoperative liver metastases, local recurrence and peritoneal tumor dissemination (38–86%, depending on the diagnostic methodology).^{( 43} , ^{44 )} Recent studies of the influence of divalent cations on colon cancer cell adhesion in a murine transplantable tumor model demonstrated that Mg²⁺ stimulated tumor formation to 96%, but Ca²⁺ reduced tumor formation to 56% with no apparent affect on serum Ca²⁺.^{( 45} , ^{46 )} Together with our results, these data collectively suggest that, during pancreatic cancer progression, a resulting increase in Mg²⁺ and decrease in Ca²⁺ in the local tumor microenvironment derived from pancreas juice or other physiological factors, and sharing many similarities to the wound healing microenvironment,^{( 28} , ^{29 )} could be a critical regulator of the integrin‐mediated malignant phenotype on the ECM generally.

Our data indicate that more than one integrin could be activated in response to divalent cation shifts in pancreatic cancer, and could potentially confound therapeutic strategies that target one particular integrin–ECM interaction. For example, targeting α₂β₁ integrin‐mediated interactions with type I collagen under altered divalent cation conditions could result in compensatory integrin–ECM interactions.^{( 47 )} In support of this hypothesis, it has been shown that the α₂ integrin subunit knockout mouse exhibits a normal cutaneous wound healing response, even though in vitro and explant studies of matrix contraction and wound healing suggest a critical role for the α₂β₁ integrin in the normal organization of the dermis and in wound healing.^{( 48 )} These data suggest that either α₂β₁ has no critical function in wound healing or that, in its absence, other integrin–ECM interactions can compensate for the loss.^{( 47 )} Our present studies suggest that shifted concentrations of divalent cations present in the local tumor microenvironment may activate several integrin–ECM interactions. These additional integrin–ECM interactions may also need to be addressed to achieve a positive outcome using anti‐integrin therapy for pancreatic cancer. Alternatively, the divalent cation milieu will also have to be considered when targeting integrin–ECM interactions therapeutically.