Association of plasminogen with dipeptidyl peptidase IV and Na+/H+ exchanger isoform NHE3 regulates invasion of human 1-LN prostate tumor cells.

Binding of plasminogen type II (Pg 2) to dipeptidyl peptidase IV (DPP IV) on the surface of the highly invasive 1-LN human prostate tumor cell line induces an intracellular Ca2+ ([Ca2+]i) signaling cascade accompanied by a rise in intracellular pH (pHi). In endothelial cells, Pg 2 regulates intracellular pH via Na+/H+ exchange (NHE) antiporters; however, this mechanism has not been demonstrated in any other cell type including prostate cancer cells. Because the Pg 2 receptor DPP IV is associated with NHE3 in kidney cell plasma membranes, we investigated a similar association in 1-LN human prostate cancer cells and a mechanistic explanation for changes in [Ca2+]i or pHi induced by Pg 2 in these cells. Our results suggest that the signaling cascade initiated by Pg 2 and its receptor proceeds via activation of phospholipase C, which promotes formation of inositol 3,4,5-trisphosphate, an inducer of Ca2+ release from endoplasmic reticulum stores. Furthermore, our results suggest that Pg 2 may regulate pHi via an association with NHE3 linked to DPP IV in these cells. These associations suggest that Pg has the potential to simultaneously regulate calcium signaling pathways and Na+/H+ exchanges necessary for tumor cell proliferation and invasiveness.

The growth of many tumors, including invasive breast cancer, non-small cell lung cancer, and prostate carcinoma, is associated with changes in the tumor microenvironment leading to hypoxia and decreases in extracellular pH (1)(2)(3). Typically, the pH in the extracellular environment is 0.5 pH units or lower than in normal tissues, whereas intracellular pH (pH i ) in tumors is either similar or more alkaline than normal tissues (4,5). A key factor in prostate tumors is the rapid fluctuations in cytosolic free Ca 2ϩ concentrations ([Ca 2ϩ ] i ) that result from regulated signal transduction events associated with cell proliferation (6). One of these events occurs when plasminogen type II (Pg 2) 1 interacts with dipeptidyl peptidase IV (DPP IV) on the surface of human prostate tumor 1-LN cells (7), producing a rapid increase in [Ca 2ϩ ] i associated with expression of metalloproteinase-9, which in an in vitro model enhances the invasiveness of these cells (7). Using this model, we recently demonstrated that a Pg fragment containing kringles 1-3 (angiostatin) is able to inhibit the Ca 2ϩ signaling cascade induced by Pg via competition for binding to DPP IV (8).
In renal brush-border membranes, DPP IV exists in multimeric complexes with NHE3, a member of the NHE antiporter family (9,10). Both appear to follow a common apical secretion pathway (11,12). In humans there are nine NHE isoforms that are members of a gene family called SLC9A (13). The NHE family can be divided into plasma membrane and intracellular organellar isoforms. The established plasma membrane isoforms include NHE 1-5 (13). The plasma membrane isoforms are further divided into those that cycle to and from the recycling endosomes/plasma membrane including NHE3 (14) and NHE5 (15) and those that permanently reside on the plasma membrane and include NHE1, -2, and -4 (16,17). The organellar isoforms include NHE6, NHE7, NHE8, and NHE9 (13). NHE1-3 are present in all gastrointestinal organs, and NHE6, although predicted to be ubiquitous, is present only in pancreas and liver. NHE4 together with NHE8 and NHE9 is present only in liver (13). NHE5 is present only in brain (18). NHE2 is also expressed in prostate tissue (19).
We investigated a possible association of DPP IV with NHEs and a mechanistic explanation for changes in [Ca 2ϩ ] i or pH i induced by Pg 2 in 1-LN cells. We found that interaction between Pg 2 and its receptor induces activation of phospholipase C (PLC), promoting the formation of inositol 3,4,5-trisphosphate (IP 3 ), which in turn stimulates release of Ca 2ϩ from endoplasmic reticulum stores. We also found that Pg 2 may regulate pH i via an association with NHE3 linked to DPP IV in these cells. These associations have the potential to regulate simultaneously calcium signaling pathways and Na ϩ /H ϩ exchanges necessary for tumor cell invasiveness. Proteins-Human Pg was purified from human plasma by affinity chromatography on L-lysine-Sepharose (20) and separated into its two major classes of glycoforms, Pg 1 and Pg 2, by affinity chromatography on concanavalin A-Sepharose (21).
Antibodies-Rabbit IgG against DPP IV was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Antibodies against the 16-amino acid sequence (Val 275 -Val 290 ) of NHE3 conjugated to keyhole limpet hemocyanin (15) were prepared in rabbits by Covance (Denver, PA). The IgG fraction specific to NHE3 was purified by immunoaffinity on a resin containing the NHE3 peptide conjugated to carboxyhexyl-Sepharose. Affinity-purified anti-NHE1 IgG raised in rabbits against a 22amino acid domain of human NHE1 (23), affinity-purified anti-NHE2 IgG raised in rabbits against a 20-amino acid domain common to both human and rat NHE2 (24), and affinity-purified anti-NHE3 IgG raised in rabbits against a 22-amino acid domain of human NHE3 (25) were all purchased from Chemicon International Inc. (Temecula, CA).
Cell Cultures-The human prostate tumor cell line 1-LN was grown in RPMI 1640 supplemented with 10% (v/v) fetal bovine serum, 100 units/ml penicillin G, and 100 g/ml streptomycin as described previously (7). The 1-LN cell line was a kind gift from Dr. Phillip Walther, Department of Urology, Duke University Medical Center.
In Vitro Invasion Assays-Invasive activity in vitro was assessed by determining the ability of 1-LN cells to invade Matrigel® as described previously (26). Polycarbonate filters (8-m pore size; BD Biosciences) were coated with Matrigel® (12 g/filter) and placed in a modified Boyden chamber. Cells (1 ϫ 10 5 ) were added to the upper chamber in serum-free RPMI 1640 medium containing Pg 2 (0.1 M) in the absence or presence of anti-NHE3 IgG or a control nonimmune rabbit IgG and incubated for 24 h in a humidified atmosphere. Following incubation, noninvading cells were removed from the upper chamber with a cotton swab, and filters were excised and stained with Cyto-Quik® (Fisher). Cells on the lower surface of the filter were enumerated using an ocular micrometer. Five high-powered fields were counted. Each experiment was performed twice with triplicate samples.
Measurements of Intracellular Free Ca 2ϩ Concentrations and Cytosolic pH-1-LN [Ca 2ϩ ] i was measured by digital imaging microscopy using the fluorescent indicator Fura-2/AM (27). For measurements of pH i , 1-LN cells were incubated overnight in RMPI 1640 medium supplemented with 10% (v/v) fetal bovine serum on glass coverslips and then washed with Hanks' balanced salt solution (HBSS) containing 0.1 M sodium bicarbonate, pH 7.1. The cells were incubated for 20 min with 2 M bis(carboxyethyl)-carbonyl fluorescein in HBSS, rinsed three times with buffer, and placed on the fluorescent microscope stage. pH i was measured by a digital video imaging technique in cells stimulated by the ligands, which were added after obtaining a stable base line (28).
Measurement of IP 3 in 1-LN Cells-Confluent 1-LN cell monolayers grown in 6-well culture plates in inositol-free RPMI 1640 medium containing 10% fetal bovine serum were radiolabeled with 8 Ci/ml myo-[2-3 H]inositol in 2 ml of fresh medium for 16 h at 37°C as described previously (29,30). After this period, the cells were washed five times with HBSS containing 10 mM LiCl, 1 mM CaCl 2 , and 1 mM MgCl 2 , pH 7.4. Then the cells were exposed to Pg 2 (0.1 M) in HBSS for varying times at 37°C. The reaction was terminated by aspirating the medium and adding a volume of 6.5% ice-cold perchloric acid. The cells were scraped and transferred to tubes containing 5 mM EDTA and 1 ml of octylamine. Then the suspension was centrifuged at 5,600 ϫ g for 20 min at 4°C. The upper phase was applied to a 1-ml packed Dowex resin column (AG1-X8, formate, Bio-Rad) and sequentially eluted in a batch fashion with 50, 200, 400, and 800 mM and 1.2 and 2.0 M ammonium formate containing 0.1 M formic acid, respectively. An aliquot of this mixture was used for determining radioactivity as described previously (29,30).
Measurement of Cell-surface NHE3 Using a Cellular Enzyme-linked Immunosorbent Assay (CELISA)-Confluent monolayers of 1-LN cells grown in 96-well culture plates were gently rinsed in HBSS and then fixed for 5 min at 4°C in freshly prepared 0.25% glutaraldehyde in PBS. Excess glutaraldehyde was washed away with PBS. The plate was then incubated for 1 h at 22°C with 2% bovine serum albumin (200 l) to saturate protein binding sites on the plastic. Wells in triplicate were then incubated with specific rabbit anti-NHE3 (Val 275 -Val 290 ) IgG or nonimmune rabbit IgG in PBS (200 l of a 0.5 g/ml IgG solution) at 22°C for 90 min followed by three rinses with PBS. Next, the cells were incubated with an alkaline phosphatase-conjugated anti-rabbit IgG (200 l) of a solution containing 250 ng/ml secondary IgG). Bound IgG was monitored by hydrolysis of the alkaline phosphatase substrate p-nitrophenylphosphate (2 mg/ml) at a wavelength of 405 nm using an Anthos Labtec® kinetic plate reader. Cell-surface NHE3 was expressed as ⌬A405 nm/min. Immunofluorescence Microscopy-1-LN cells were plated at 5 ϫ 10 5 cells/ml on glass coverslips and allowed to adhere overnight. Cells were incubated at 4°C for 1 h in PBS containing 1% bovine serum albumin with anti-DPP IV, anti-NHE3, anti-actin (negative control), or preimmune IgGs. Cells were washed and incubated at 4°C with goat anti-rabbit IgG conjugated to Alexa Fluor® 488 dye purchased from Molecular Probes-Invitrogen before washing and fixing in 4% paraformaldehyde. Immunofluorescence microscopy was performed by using an Olympus BX-60 microscope (Olympus, Lake Success, NY).
Immunoprecipitation of DPP IV and NHE1, NHE2, and NHE3 from Plasma Membranes-1-LN cell monolayers from 5 ϫ 150 cm 2 were gently detached and washed twice with HBSS. After centrifugation at 5,000 rpm for 5 min, the cell pellet was suspended in HBSS buffer containing 10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 10 M benzamidine, and 10 M leupeptin. The cell suspension was transferred to chilled homogenizing tubes, and plasma membranes were isolated as described previously (31). Cell plasma membrane lysates were prepared by suspending the membranes in the above buffer containing 0.5% Nonidet P-40 and passing the suspension through a syringe (22.5-gauge needle) six times followed by centrifugation at 5,000 rpm for 5 min at 4°C to remove particles in suspension. The supernatant was divided in five aliquots (200 l), and each one was incubated with 50 l of protein A-Sepharose beads for 1 h at 22°C followed by centrifugation and removal of the resin. The supernatant was then incubated with 10 g of IgG anti-DPP IV, anti-NHE3 (Val 275 -Val 290 ), or specific anti-NHE1, -NHE2, or -NHE3 in the presence of 30 l of protein A-Sepharose beads overnight at 4°C with shaking. Then the mixtures were centrifuged at 5,000 rpm for 5 min at 4°C. The beads were washed three times in solubilization buffer and then prepared for SDS-PAGE and immunoblotting.
SDS-PAGE and Immunoblotting-Protein samples were solubilized in SDS sample buffer, and proteins were separated by SDS-PAGE using 7.5% polyacrylamide gels according to Laemmli (32). The dye-conjugated molecular weight markers (Bio-Rad) used were myosin (M r ϳ220,000), ␤-galactosidase (M r ϳ116,000), phosphorylase b (M r ϳ97,000), bovine serum albumin (M r ϳ84,000), and ovalbumin (M r ϳ66,000). For immunoblottings, proteins were transferred to nitrocellulose membranes according to the Western blot method (33). Sheets of nitrocellulose membranes containing transferred proteins were incubated first in a solution containing 5% nonfat dry milk and 0.1% Tween 80 in PBS for 1 h at 22°C to block nonspecific binding of antibody followed by incubation overnight at 4°C with specific rabbit anti-DPP IV or anti-NHE3 IgGs (1 g/ml) in PBS containing 0.1% Tween 80 (PBS-Tween). The membranes were then washed three times in PBS-Tween and incubated for 1 h with horseradish peroxidase-conjugated anti-rabbit IgG. Bound antibody was detected by incubating the membranes in PBS containing the horseradish peroxidase substrate 4-chloro-1-naphtol (1 mg/ml) and 0.01% Plasminogen Binding to Immobilized NHE3-Solubilized membrane aliquots (100 l) were incubated with 10 g anti-NHE3 IgG raised in rabbits against a 22-amino acid domain of human NHE3 (25) in the presence of 30 l of protein A-Sepharose beads as described above. After separation of the proteins by SDS-PAGE and blocking of nonspecific sites, nitrocellulose membranes containing individual lanes were incubated with human Pg 2 (1 M) in the absence or presence of 100 mM 6-aminohexanoic acid (6-AHA) in PBS-Tween at 4°C overnight. After extensive rinsing in PBS-Tween, the membranes were incubated with an affinity-purified goat anti-human Pg IgG (1 g/ml) at 22°C for 3 h. A membrane containing a control lane was reacted with an anti-NHE3 (Val 275 -Val 290 ) IgG. The membranes were then washed three times in PBS-Tween and incubated for 1 h with horseradish peroxidase-conjugated anti-goat IgG. Bound antibody was detected as described above.  (Fig. 1A). These experiments suggest that Pg 2 induces calcium release from intracellular calcium stores, which are activated by signaling pathways initiated by cell-surface Pg 2 receptors. Because IP 3 is responsible for the release of calcium from calcium stores in the endoplasmic reticulum, we studied the effect of Pg 2 receptor activation on IP 3 generation. The results (Fig. 1B) indicate that there is a severalfold increase in IP 3 formation in 1-LN cells stimulated with Pg 2 (0.1 M). Next we assessed the effect of inhibitors of PLC, which hydrolyzes PIP 2 into IP 3 and diacylglycerol, on the rise of [Ca 2ϩ ] i induced by Pg 2 on 1-LN cells. U-73122, a potent inhibitor of PLC activity, and generation of newly formed IP 3 (35) inhibit the rise of [Ca 2ϩ ] i induced by Pg 2 (0.1 M) (Fig. 3), whereas U-73343, a close analog that does not suppress PLC activity (24), does not influence the Pg 2 effect (Fig. 1C). Taken together, these data suggest that receptor activation leads to synthesis of IP 3 followed by a rise in [Ca 2ϩ ] i , primarily from intracellular calcium stores under our experimental conditions.

Changes in [Ca 2ϩ ] i in Response to
Identification of the Mechanisms Involved in the Rise of pH i Induced by Pg 2 in 1-LN Cells-Pg 2 (0.1 M) added to 1-LN cells induced a rise in pH i , which was continuous for 500 s (Fig.  2). In endothelial cells, a decrease in pH i induced by angiostatin, a Pg fragment, is mediated by NHE antiporters, which can be easily inhibited by amiloride analogs (36). Because DPP IV is a common receptor for Pg 2 and angiostatin on 1-LN cells (8), we hypothesized that their effects on pH i may be mediated by the same NHE receptor. Preincubation of 1-LN cells with amiloride (100 nM) prior to addition of Pg 2 (0.1 M) inhibits the rise in pH i (Fig. 2), suggesting that Pg 2 may regulate pH i in 1-LN cells via mechanisms involving NHE receptors. The association between DPP IV and NHE3 exists in kidneys (11)(12); however, a similar association has not been reported in other cell types. Our goal then was to determine whether NHE3 in association with DPP IV also serves as an additional receptor for Pg 2 on the 1-LN cell surface.
Synthesis of an Antibody against a Cell-surface Epitope on NHE3-A hidden Markov model was used to search for the transmembrane topology corresponding to NHE3 (37, 38). The membrane topology model predicts three outside loops in NHE3 (39), including the amino acid segments Asn 133 -Thr 142 , Thr 275 -Phe 291 , and Asp 427 -Leu 436 . The sequence including Val 275 -Val 290 (Fig. 3B) in the second outside loop of NHE3 contains one L-lysine residue (Lys 280 ) flanked by two hydrophilic amino acids (Thr 279 and His 281 ) suggested a candidate binding site for Pg 2. A comparison of this primary amino acid sequence in NHE3 with similar sequences in other members of the NHE family (Fig. 3A), shows segments in NHE1, -2, and -5 with primary structural homologies greater than 60%, whereas NHE4 and NHE6 show only 50 and 35% homologies, respectively. A topology model comparing the second outside loop of NHE3 with homologous structures in NHE1, -2, -4, and -5 (Fig.  3B), predicts outside loops only in NHE2 and -5. However, only the outside loop in NHE5 contains a candidate binding site for Pg 2, L-Lys 273 . These comparisons suggest NHE3 as a unique receptor for Pg 2 in prostate cells because NHE5 is expressed solely in the brain (18).
Identification of NHE3 on 1-LN Cell Surface by CELISA and Immunofluorescence Microscopy-The NHE3 peptide Val 275 -Val 290 was synthesized and antibodies were prepared and purified as described under "Experimental Procedures." The anti-NHE3 IgG was then used as a reagent to assess the presence of NHE3 on the surface of 1-LN cells by CELISA and its possible association with DPP IV using immunoprecipitation techniques. Confluent cultures of 1-LN cells in 96-well culture plates were titrated with anti-NHE3 IgG and a control preimmune rabbit IgG (Fig. 4). The anti-NHE3 IgG binds to these cells in a dose-dependent manner, whereas very little specific binding was observed with the nonimmune IgG, suggesting the presence of NHE3 in the 1-LN cell membrane. This experiment was validated by immunofluorescence microscopy studies in non-permeabilized 1-LN cells (Fig. 5). The cells were incubated first with anti-DPP IV (Fig. 5A), anti-NHE3 (Val 275 -Val 290 ) (Fig. 5B), anti-actin (negative control) (Fig. 5C) or preimmune IgGs (Fig. 5D)  plasma membranes from 1-LN cells with anti-NHE3 IgG and analyzed the precipitate for the presence of DPP IV by Western blotting and reaction with an anti-DPP IV IgG. The co-precipitation of a protein with a molecular mass of 120 kDa corresponding to a monomer of DPP IV (Fig. 6, lane 1) indicates association with NHE3. An immunoprecipitation of solubilized cells with an anti-DPP IV IgG and then Western blotting and reaction of the precipitate with anti-NHE3 IgG shows proteins of molecular mass of 84 kDa corresponding to a monomer of NHE3 (Fig. 6, lane 2), again demonstrating an association between DPP IV and NHE3.
Association between Pg 2 and NHEs on 1-LN Cell Surface-We examined the effect of anti-NHE3 IgG on the pH i changes induced by Pg 2 on 1-LN cells. Incubation of cells with anti-NHE3 IgG (10 g/ml) in serum-free RPMI 1640 followed by addition of Pg 2 (0.1 M) shows almost no changes in pH i (Fig. 8), suggesting that the antibody blocks Pg 2 access to NHE3 on the cell surface. To assess the interaction of Pg 2 with NHEs we monitored binding of Pg 2 directly to NHEs precipitated with anti-NHE3 (Val 275 -Val 290 ) IgG separated by electrophoresis and electroblotted to nitrocellulose membranes. We show that NHE3 (Fig. 9, lane 1) reacts efficiently with Pg 2 (Fig. 9, lane 2), whereas a membrane incubated with Pg 2 in the presence of 6-AHA (100 mM) shows no specific binding, thereby suggesting an L-lysine-dependent binding of Pg 2 to NHE3. This interaction was confirmed when we monitored the binding of Pg 2 directly to the Val 275 -Val 290 peptide conjugated to carboxyhexyl-Sepharose. When Pg 2 (2.5 mg) in PBS was filtered through this resin, almost 90% of the Pg 2 is bound (Fig.  10), thereby confirming that Pg 2 may bind to NHE3.
Effect of Anti-NHE3 IgG on the Pg 2-induced Stimulation of 1-LN Cell Invasiveness-We determined previously that Pg 2 enhances the ability of 1-LN cells to penetrate the synthetic basement membrane Matrigel® (7). To determine whether anti-NHE3 IgG influences Pg-induced cell invasiveness, we incubated cells with increasing concentrations of anti-NHE3 IgG and a single concentration of a control nonimmune rabbit IgG in the presence of Pg 2 (0.1 M). We show ( Table I) that anti-NHE3 IgG decreases the ability of Pg 2 to stimulate 1-LN cell-invasive activity, whereas a control IgG does not affect this activity.

DISCUSSION
In a previous report (7), we demonstrated that Pg 2 in association with DPP IV is pivotal in the invasive capacity of 1-LN prostate cancer cells. This association induces a Ca 2ϩ signaling cascade, which regulates the expression of matrix metalloproteinase-9, essential for invasiveness of these cells in an in vitro model (7). Pg 2 binds to DPP IV via its carbohydrate chains (7,40), and this interaction is easily inhibited by angiostatin, which competes with Pg for the DPP IV binding site on 1-LN cells (8). On endothelial cells, angiostatin induces a decrease in pH i (36), whereas Pg 2 induces a sustained increase in pH i (42). Conversely, angiostatin does not induce any changes on [Ca 2ϩ ] i on 1-LN cells (8), whereas Pg 2 induces a rise in [Ca 2ϩ ] i either on endothelial cells (42) or 1-LN cells (7,8). In the present study, we investigated the mechanisms involved in the changes of [Ca 2ϩ ] i and pH i induced by Pg 2 on 1-LN cells.
Our results suggest that Pg 2 induces a rise in [Ca 2ϩ ] i as a result of activation of PLC, which forms IP 3 , a direct inducer of Ca 2ϩ release from endoplasmic reticulum stores. A similar signaling pathway has been demonstrated for the epidermal growth factor receptor in DU-145 prostate tumor cells (43). Cells transfected with a dominant-negative fragment of PLC-␥, which induced a decrease in IP 3 generation, were significantly less invasive than control cells (43).
Intracellular pH plays a central role in the regulation of many aspects of cell physiology, and protons may function as a second messenger in a manner similar to that of Ca 2ϩ (44). pH is also one of the factors thought to control the rate of cell proliferation and transformation (45). In many tumor cell lines pH i is more alkaline than in normal cells (46). Because cell alkalinization precedes cell proliferation, a correlation between pH i and cell cycle has been proposed (46). Intracellular alka-  linization is an early event in malignant transformation (47), and the acid extrusion via NHE1 plays a key role in this process (48). Our present data are consistent with these observations. Pg 2 binding to 1-LN cells induces a rise in pH i and promotes cell invasiveness. Hypothetically, Pg 2 binds to DPP IV via its carbohydrate chain and to NHE3 via its L-Lysine binding sites. This ability gives Pg the capacity to act simultaneously on mechanisms involving changes in [Ca 2ϩ ] i or pH i as suggested by inhibition of the Pg 2-induced changes in [Ca 2ϩ ] i by inhibitors of PLC or changes in pH i by an anti-NHE3 IgG against the epitope including the amino acid sequence V 275 TRFTKHVRIIEPGFV 290 . This epitope is unique because the presence of Lys 280 facilitates binding of Pg 2, which, as we demonstrated previously, binds additionally to DPP IV via its carbohydrate chains (7,40).
Several reports demonstrate a cross-communication between pH i and [Ca 2ϩ ] i, where pH i is able to affect [Ca 2ϩ ] i homeostasis and contribute to the length, magnitude, and frequency of the Ca 2ϩ signal through the modulation of voltage-dependent or -independent plasma membrane Ca 2ϩ channels and/or through regulation of the mobilization of Ca 2ϩ from internal stores (44, 49 -51). On the other hand, [Ca 2ϩ ] i has been shown to induce changes in pH i (41,52).
In summary, our findings support the conclusion that Pg 2 binds to receptors on 1-LN cells that affect signaling pathways critical for cell motility and prostate tumor cell invasiveness, thereby suggesting these receptors as potential therapeutic targets to limit tumor progression.  tumor cell invasiveness 1-LN cells in 300 l of serum-free RPMI 1640 medium (at a cell density of 1 ϫ 10 5 ) were added to a modified Boyden chamber containing an 8-m pore filter coated with Matrigel (12 g/filter) and incubated with either Pg 2 (0.1 M) in the absence or presence of increasing concentrations of anti-NHE3 IgG at 37°C for 24 h. A control experiment was performed with nonimmune rabbit IgG. After this period, filters were excised, noninvading cells were removed from the top surface of the membrane, and staining was carried out with Cyto-Quik®. Invading cells were enumerated by using an ocular micrometer and counting a minimum of five high-powered fields. Data shown represent the mean Ϯ S.D. from experiments performed in triplicate.