Involvement of p38α Mitogen-activated Protein Kinase in Lung Metastasis of Tumor Cells*

To study the role of p38 mitogen-activated protein kinase (p38) activity during the process of metastasis, p38α+/- mice were subjected to an in vivo metastasis assay. The number of lung colonies of tumor cells intravenously injected in p38α+/- mice was markedly decreased compared with that in wild-type (WT) mice. On the other hand, the time-dependent increase in tumor volume after subcutaneous tumor cells transplantation was comparable between WT and p38α+/- mice. Platelets of p38α+/- mice were poorly bound to tumor cells in vitro and in vivo compared with those of WT mice. E- and P-selectin mRNAs were markedly induced in the lung after intravenous injection of tumor cells. However, the induction of these selectin mRNAs in p38α+/- mice was weaker than that in WT mice. Furthermore, the resting expression levels of E-selectin in lung endothelial cells and P-selectin in platelets of p38α+/- mice were suppressed compared with those of WT mice. The number of tumor cells attached on lung endothelial cells of p38α+/- mice was significantly reduced compared with that of WT mice. The transmigrating activity of tumor cells through lung endothelial cells of p38α+/- mice was similar to that of WT mice. These results suggest that p38α plays an important role in extravasation of tumor cells, possibly through regulating the formation of tumor-platelet aggregates and their interaction with the endothelium involved in a step of hematogenous metastasis.

cades of protein phosphorylation. There are three genetically distinct MAPKs in mammals, consisting of extracellular signalregulated kinase, c-Jun N-terminal kinase, and p38 MAPK (p38). All three members are activated by dual phosphorylation of the conserved TXY motif and then phosphorylate their respective substrates on serine or threonine residues (1)(2)(3). There are four mammalian isoforms of p38 (␣, ␤, ␥, and ␦). Among them, p38␣ and -␤ are expressed relatively ubiquitously, as shown by Northern blot analysis of adult tissues (4). Although targeted disruption of the p38␣ gene results in homozygous embryonic lethality because of defects in erythropoiesis and placental organogenesis (5,6), the p38␣ ϩ/Ϫ mouse is a useful tool for analyzing the in vivo role of p38 in disease models (7)(8)(9).
Tumor metastasis is a significant process resulting in unexpected death in cancer patients. Recent advances in molecular cancer research have clarified a variety of therapeutic targets, some of which are being tested in clinical trials (10). Very recently, the relationship between tumor metastasis and MAPKs has been investigated (11,12). It is reported that p38␣ is important for the maintenance of breast cancer with an invasive phenotype by promoting the stability of urokinase plasminogen activator and its receptor mRNA (13). It has also been demonstrated that p38 is involved in various metastatic processes (14 -17). On the other hand, MKK4, a common upstream prerequisite for c-Jun N-terminal kinase and p38 activation as an MAPK kinase, is characterized as a metastasis suppressor gene in human ovarian carcinoma (18,19). Likewise, loss of p38 activation leads to an increase in tumorigenesis because of a cell cycle defect (20). These findings clearly suggest that p38 activity in tumor cells regulates tumor progression and metastasis. However, there is no in vivo confirmation of a pathophysiological role of p38 in hosts during tumor metastasis. To elucidate this point, we used p38␣ ϩ/Ϫ mice to examine the in vivo role of p38␣ during tumor metastasis. Here, we showed that tumor metastasis is suppressed in p38␣ ϩ/Ϫ mice.

EXPERIMENTAL PROCEDURES
Experimental Animals-The use of animals in all of our experiments was in accordance with the guidelines for animal care of Chiba University and RIKEN. Female mice heterozygous for targeted disruption of the p38␣ gene (6) were crossed with C57BL/6J male mice (Saitama Experimental Animal Supply Co.) to generate p38␣ ϩ/Ϫ and p38␣ ϩ/ϩ (wild-type; WT) mice. Offspring (Ͼ8 generations) were genotyped by PCR analysis of tail-derived DNA. Multiplex PCR with three primers per reaction was used. The primers were as follows: A, 5Ј-CCCTA-TACTCCCTCTCTGTGTAACTTTTG-3Ј; B, 5Ј-CCCAAAC-CCCAGAAAGAAATGATG-3Ј; C, 5Ј-TTCAGTGACAACG-TCGAGCACAGCTG-3Ј. Using these primers for one cycle at 94°C for 5 min followed by 35 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 1 min, with an extension step of 7 min at 72°C at the end of the last cycle, produced 800-and 450-bp fragments from the mutant and WT alleles, respectively. 8 -12-week-old WT and p38␣ ϩ/Ϫ littermates were used for each experiment.
Experimental Lung Metastasis Model-B16 and LLC cells were trypsinized and recovered from the medium by brief centrifugation. They were suspended with physiological saline to a concentration of 10 7 cells/ml. Mice (WT and p38␣ ϩ/Ϫ ) were anesthetized with an intraperitoneal injection of avertin, and 10 6 cells (in a volume of 100 l) were systemically injected into the tail veins. In all experiments, injection of tumor cells was performed at a speed of 10 l/min using a syringe pump (model CFV-3200; NIHON KOHDEN). 3 and 4 weeks after injection of LLC and B16 cells, respectively, the lungs were dissected out and rinsed in phosphate-buffered saline (PBS). Then they were separated into individual lobes, and the number of surface metastatic foci was counted. In the case of using B16-␤-gal, the lungs were dissected out from WT and p38␣ ϩ/Ϫ mice 1 week after injection. Their lungs were immersed in 0.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4), followed by X-gal buffer (1 mg/ml 5-bromo-4chloro-3-indoyl-␤-D-galactopyranoside, 10 mM K 4 Fe(CN) 6 , 10 mM K 3 Fe(CN) 6 , and 2 mM MgCl 2 in PBS), and then embedded in paraffin and cut into sections 4 m in thickness. The sections were placed on poly-L-lysine-coated slides, and blue-dyed colonies were counted under a light microscope (Axioplan; Zeiss).
Tumor Growth Experiment-B16 and LLC cells were suspended with physiological saline to a concentration of 10 7 cells/ ml. Each cell type (10 6 cells in a volume of 100 l) was inoculated subcutaneously into the left flank of mice (WT and p38␣ ϩ/Ϫ ) under anesthesia with an intraperitoneal injection of avertin. Tumors were measured along the plane of the body cavity using a hand caliper, and tumor volume was calculated as (length ϫ width) 2 /2.
In Vitro Binding of Platelets to Tumor Cells-Platelets were prepared from WT and p38␣ ϩ/Ϫ mice by the method previously described (9). Then platelets were fluorescently labeled with calcein by incubation with 1 M calcein-acetoxymethyl ester (Molecular Probes) for 30 min at 37°C and washed twice with Hepes-Tyrode's buffer (10 mM Hepes, 137 mM NaCl, 2.68 mM KCl, 0.42 mM NaH 2 PO 4 , 1.7 mM MgCl 2 , 11.9 mM NaHCO 3 , 5 mM glucose, pH 7.2). B16, LLC, and HEK293T cells were seeded on 8-well chamber slides and incubated with fluorescently labeled platelets at a density of 10 7 /ml for 1 h at 37°C. In the case of activating platelets, 1 unit/ml thrombin and 20 g/ml collagen were added to each well just after the application of platelets. For elucidating the effect of heparin, 20 units/ml heparin (Wako Chemicals) was added to each well just after the application of platelets. After washing the chamber slides with Hepes-Tyrode's buffer twice, binding of platelets to tumor cells was examined by a fluorescence microscope (Axioplan; Zeiss).
In Vivo Binding of Platelets to Tumor Cells-B16 cells labeled with calcein were resuspended and used for a systemic injection into WT and p38␣ ϩ/Ϫ mice. After 30 min, the lungs were dissected out, fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4), and frozen in Tissue Tek OCT compound (Miles Inc.). Freshly cut lung sections (10 m in thickness) were placed on poly-L-lysine-coated slides. The sections were stained with a monoclonal anti-mouse CD41 (integrin ␣IIb) antibody (Pharmingen) in combination with a secondary antibody (Alexa Fluor 594-labeled goat anti-rat IgG (Molecular Probes, Inc., Eugene, OR)) and observed under a fluorescence microscope (Axioplan; Zeiss). The resulting fluorescence profile was analyzed by an Intelligent Quantifier (Bio Image) to quantify the fluorescence intensity of platelets adherent to calcein-labeled tumor cells.
Reverse Transcription-PCR Detection of P-and E-selectin mRNAs-B16 cells were injected into WT and p38␣ ϩ/Ϫ mice. As a negative control, B16-free PBS was injected into WT and p38␣ ϩ/Ϫ mice. The lungs were taken from these mice 4 h after injection. Total RNA was prepared from the lungs using ISOGEN (Wako Chemicals) according to the manufacturer's instructions. Single strand cDNA was synthesized from prepared RNA (3 g), with Moloney murine leukemia virus reverse transcriptase (Invitrogen) using an oligo(dT) primer (Invitrogen) in a total volume of 20 l. The resultant cDNA sample (1 l) was subjected to PCR for amplification of mouse P-or E-selectin cDNA using specific primers (for P-selectin, sense primer, 5Ј-TGCAGCTTTTCCTGTGATGAAGGC-3Ј; antisense primer, 5Ј-ATAGAGCCAACACCAAACTCTCCG-3Ј; for E-selectin, sense primer, 5Ј-GACCTTTCCAAAAATGGG-TCCAG-3Ј; antisense primer, 5Ј-AGAGCAATGAGGACGA-TGTCAGGAG-3Ј). As an internal control, mouse glyceraldehyde-3-phosphate dehydrogenase cDNA was amplified using specific primers (sense primer, 5Ј-GACCACAGTCCATGAC-ATCACT-3Ј; antisense primer, 5Ј-TCCACCACCCTGTTGC-TGTAG-3Ј). The settings of the thermal cycler were 30 cycles of 30 s at 94°C, 30 s at 59°C, and 1 min at 72°C for mouse E-and P-selectin and 25 cycles of 40 s at 94°C, 1 min at 60°C, and 1 min at 72°C for mouse glyceraldehyde-3-phosphate dehydrogenase. For detecting P-selectin mRNA in platelets, total RNA sample (0.25 g) was subjected to reverse transcription reaction with Moloney murine leukemia virus in a total volume of 20 l, and then the resultant sample (1 l) was subjected to PCR with 35 cycles. The amplified products were separated in 1.2% agarose gel and visualized with ethidium bromide staining p38␣ in Lung Metastasis under UV radiation. Specific amplification of the expected size (mouse E-selectin, 570 bp; P-selectin, 371 bp; and mouse glyceraldehyde-3-phosphate dehydrogenase, 453 bp) was observed.
Isolation of Lung Endothelial Cells-The lungs were dissected out from WT and p38␣ ϩ/Ϫ mice and washed with icecold PBS three times. The tissues were cut into small pieces and incubated in Dulbecco's modified Eagle's medium supplemented with 2 mg/ml collagenase (Worthington), 2 mg/ml hyaluronidase (Sigma), and 1 mg/ml dispase (Invitrogen) at 37°C for 30 min with shaking. Then the tissues were suspended well by a pipette, and the resultant suspension was passed through a 70-m nylon mesh filter (Falcon). The cells were collected by centrifugation at 400 ϫ g for 10 min. After washing twice with PBS, the cells suspended in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 50 units/ml heparin were put on the top of Percoll solution (50% lower and 30% upper) layers and centrifuged at 800 ϫ g for 30 min. The resultant interface fraction was collected and washed with PBS twice. The cells were suspended in 0.1% bovine serum albumin/PBS (Type V; Sigma) and subjected to MACS separation with rat monoclonal anti-PECAM-1 (BD Biosciences) and anti-rat IgG microbeads (Miltenyi Biotec.). The isolation of PECAM-1-positive endothelial cells was performed according to the manufacturer's protocol (Miltenyi Biotec.). The cells were seeded on a collagen-coated 6-well plate and maintained for 2 days in CS-C medium (Sigma) supplemented with 10% fetal bovine serum and EC growth supplement (Sigma). After confirming over 90% purity by the uptake of DiI-acetylated low density lipoprotein, the cells were used for experiments.
Flow Cytometry-Liver endothelial cells were isolated from WT and p38␣ ϩ/Ϫ mice by the method previously described (21) with modification. Cells were cultured for 2 days, and over 90% purity was confirmed by the uptake of DiI-acetylated low density lipoprotein and immunostaining with antibodies to VEcadherin and vascular endothelial growth factor receptor-2. Then cells were stained for P-selectin, E-selectin, or PECAM-1 (CD31), followed by fluorescein isothiocyanate-conjugated goat anti-rat immunoglobulin (BD Biosciences). The labeled cells were analyzed on a FACScan flow cytometer (BD Biosciences), and rates of positive cells were quantified using CELLQest software 2.1.1 (BD Biosciences).
Assay for Attachment and Transmigration of Tumor Cells-Primary cultured lung endothelial cells (10 5 cells) were seeded on 24-well cell culture inserts (apical chamber; Falcon) with pore size of 8 m and maintained in CS-C medium overnight. After replacing the medium with Opti-MEM I (Invitrogen), DiI-labeled B16 cells (3 ϫ 10 4 cells) were added to the apical chambers. At the same time, platelets (3 ϫ 10 6 ) freshly isolated from WT and p38␣ ϩ/Ϫ mice were added to the apical chambers for determination of the effect of platelets. As a negative control, DiI-labeled B16 cells with or without platelets were added to the endothelial cell-free apical chambers. All apical chambers were washed with Opti-MEM I twice for removing nonattaching cells 30 min after applying B16 cells to the apical chambers. Some apical chambers were fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for counting the number of B16 cells attached on endothelial cells by a fluorescence microscope (Axioplan; Zeiss). Some other apical chambers were further incubated to determine the transendothelial migration of B16 cells. After incubation for 8 h, nonmigrating cells were removed by scraping the apical side of the apical chambers with a cotton swab. Then the apical chambers were fixed and subjected to the examination of transmigrating B16 cells by a fluorescence microscope (Axioplan; Zeiss).

RESULTS AND DISCUSSION
Resistance of p38␣ ϩ/Ϫ Mice to Experimental Lung Metastasis of Tumor Cells but Not to Tumor Growth-Two transplantable tumor cell lines, B16 and LLC, were used in the present study, because both are highly metastatic to the lung (22,23). As shown in Fig. 1, A and B, there was a significant difference in surface pulmonary metastatic foci between WT and p38␣ ϩ/Ϫ mice following intravenous injection of B16 cells (WT, 91.8 Ϯ 21.0 cells/mouse; p38␣ ϩ/Ϫ , 9.3 Ϯ 3.2 cells/mouse). Also in the case of LLC cell injection, the formation of pulmonary metastatic foci was significantly decreased in p38␣ ϩ/Ϫ mice compared with WT mice (WT, 31.2 Ϯ 5.4 cells/mouse; p38␣ ϩ/Ϫ , 11.2 Ϯ 4.3 cells/mouse). To explore in greater detail whether the significant reduction in surface pulmonary metastases in p38␣ ϩ/Ϫ mice resulted from tumor growth, we established B16 cells stably expressing ␤-galactosidase (B16-␤-gal) in order to visualize a small focus of B16 cells. By using B16-␤-gal, we investigated the formation of pulmonary metastatic foci at an earlier time after injection of tumor cells. Also in this case, a significant reduction in pulmonary metastasis was observed in p38␣ ϩ/Ϫ mice (Fig. 1C). The size of metastatic foci dyed blue was comparable between WT and p38␣ ϩ/Ϫ mice, indicating that tumor growth is not impaired in p38␣ ϩ/Ϫ mice (data not shown). To further clarify this notion, we investigated the timedependent increase in tumor volume after subcutaneous tumor cell transplantation. Both B16 and LLC cells when injected subcutaneously formed tumors with 100% penetrance, in which the steady growth of tumors showed no difference between WT and p38␣ ϩ/Ϫ mice (Fig. 2). These results suggest that host p38␣ affects the metastatic potential of tumor cells but neither tumor cell growth nor tumor rejection.
Angiogenesis is well known to be an important process in tumor cell growth (24,25). The microenvironment of the local host tissue appears to be an active participant in exchanging cytokines and enzymes with tumor cells that modify the local extracellular matrix, stimulate migration, and promote tumor angiogenesis, proliferation, and survival. It has been demonstrated that p38 plays a role in angiogenesis via regulating the production of inflammatory mediators (26). Likewise, targeted disruption of the p38␣ gene results in homozygous embryonic lethality because of defects in placental organogenesis, in which p38␣ in Lung Metastasis DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 JOURNAL OF BIOLOGICAL CHEMISTRY 36769 angiogenesis plays a crucial role (5). Thus, the reduction of p38 intrinsic activity in the local host tissue by a single copy disruption of the p38␣ gene is thought to cause some impairment of angiogenesis. However, the comparable tumor growth between WT and p38␣ ϩ/Ϫ mice suggests that angiogenesis during tumor growth may not show impairment in p38␣ ϩ/Ϫ mice in the present experimental model.
Reduction of Tumor Cell-Platelet Interaction in p38␣ ϩ/Ϫ Mice-Tumor cells are frequently observed in the vasculature in complexes with platelets, and this association, together with the hypercoagulable state of malignant disease, appears to be essential for successful metastasis. The ability of tumor cells to induce platelet aggregation is widespread among cancers, including colon adenocarcinoma (27), lung carcinoma (28), melanomas (29), and others. The ability of tumor cells to induce platelet aggregation is of benefit for the survival of tumor cells for the following reasons: tumor cells coated with platelets acquire the ability to evade the body's immune system and blood flow's high shear forces; large tumor-platelet aggregate tends to embolize in the microvasculature; and tumor cells effi-ciently receive a number of growth factors from platelets (30). Thus, platelets possibly act to facilitate all of the intermediate steps of transvascular metastasis, including tumor cell retention and arrest, subendothelial interaction, and extravasation from the microvasculature. We recently demonstrated that the binding activity of activated platelets from p38␣ ϩ/Ϫ mice to fibrinogen, possibly mediated through integrin ␣IIb␤3, is suppressed compared with the case of platelets from WT mice (9). This clearly indicates that the function of platelets is attenuated in p38␣ ϩ/Ϫ mice compared with WT mice. Then we investigated whether p38␣ mediates the interaction of platelets with tumor cells in vitro and in vivo.
We labeled platelets freshly isolated from WT and p38␣ ϩ/Ϫ mice with calcein and applied them to B16 and LLC cells in vitro. We used HEK293T cells as a negative control and confirmed that resting and activated platelets from WT mice did not show the activity of rosetting on the cells. As shown in Fig.  3A, resting and activated platelets from WT mice rosetted on B16 and LLC cells, and this interaction was clearly reduced in the case of p38␣ ϩ/Ϫ mice. Heparin markedly inhibited the

p38␣ in Lung Metastasis
interaction of platelets from WT and p38␣ ϩ/Ϫ mice with tumor cells. Quantitative analysis showed that resting and activated platelets from p38␣ ϩ/Ϫ mice had significantly decreased activity of rosetting on tumor cells compared with those from WT mice. The stimulation of platelets from WT mice with thrombin and collagen increased the number of platelets rosetted on tumor cells. This increase was moderate in the case of p38␣ ϩ/Ϫ mice (WT, 1.8 times; p38␣ ϩ/Ϫ , 1.5 times), indicating that agonist-induced activation is reduced in platelets from p38␣ ϩ/Ϫ mice. The tumor cell-platelet interaction was significantly heparinsensitive (Fig. 3B). To examine the tumor cell-platelet interaction in vivo, we injected calcein-labeled B16 cells, and those that lodged in the capillaries of the lung were further stained with anti-CD41, a defined platelet marker. As shown in Fig. 4A, platelets almost covered B16 cells in the lung of WT mice. In contrast, platelets weakly and partly covered B16 cells in the lung of p38␣ ϩ/Ϫ mice. Quantitative analysis of the ratio of red fluorescence (CD41-like immunoreactivity) versus green fluorescence (B16) showed that the interaction of B16 cells with platelets in the lung was significantly decreased in p38␣ ϩ/Ϫ mice compared with WT mice (Fig.  4B). These results suggest that suppression of the interaction of platelets with tumor cells can affect the metastatic potential of tumor cells in p38␣ ϩ/Ϫ mice.
Involvement of p38␣ in Expression of P-and E-selectins-It is well defined that the specific interaction of P-selectin on platelets with mucin-type glycoproteins on tumor cells regulates tumor progression and metastasis via microemboli formation and is heparin-sensitive (31,32). It has also been clarified that endothelial P-selectin expression contributes to the development of hematogenous metastasis (33). In P-selectin-deficient mice, the interaction of platelets with injected tumor cells in the lung was rarely observed (34), which notably shows an obvious similarity to our present data (Fig. 4). Moreover, the reduction of heparin-sensitive tumor cellplatelet interaction in p38␣ ϩ/Ϫ mice (Fig. 3) suggests that low expression of P-selectin could be responsible for the reduced metastasis of tumor cells in p38␣ ϩ/Ϫ mice. On the other hand, E-selectin expressed on endothelial cells can drive the seeding of tumor cells into metastatic sites by recognizing both mucin and nonmucin ligands on tumor cells (32,35). Likewise, rapid induction of P-and E-selectin expression in the target organ was observed in tumor-bearing mice and might affect the metastatic potential of tumor cells (36,37). Then we investigated the expression of P-and E-selectin mRNA in the lung before and after tumor cell injection. As shown in Fig. 5, the expression  DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 of P-and E-selectin mRNA was markedly induced in the lung of B16-bearing WT mice. This induction was significantly suppressed in the lung of B16-bearing p38␣ ϩ/Ϫ mice, indicating that the metastatic tumor cell-induced expression of P-and E-selectin mRNA could be mediated by p38␣ activity. Although low resting expression levels of P-and E-selectin mRNA were observed in the lung of WT mice, those in p38␣ ϩ/Ϫ mice were below the detection limits in our experimental conditions. This suggests that p38␣ may also be involved in the regulation of basal P-and E-selectin mRNA, probably affecting spontaneous expression of their proteins on the cell surface. To elucidate this notion and further confirm the effect of p38 signal on P-and E-selectins expression, Western blot analysis by using lung endothelial cells and platelets from WT and p38␣ ϩ/Ϫ mice was performed. As shown in Fig. 6A, the expression level of p38 in lung endothelial cells from p38␣ ϩ/Ϫ mice was lower than that from WT mice. In accordance with this fact, basal activation of p38 and ATF-2, a specific substrate for p38, and expression of E-selectin in lung endothelial cells from p38␣ ϩ/Ϫ mice were lower than those from WT mice. The expression levels of PECAM-1 and ␣-catenin as internal controls were comparable between WT and p38␣ ϩ/Ϫ mice. The expression of P-selectin in lung endothelial cells was below the detection limits in our  Total RNA was prepared from the lung of WT and p38␣ ϩ/Ϫ littermates injected with or without B16 cells. PCR product samples were subjected to 1.2% agarose gel electrophoresis and visualized by staining with ethidium bromide. Similar results were confirmed in three independent experiments. FIGURE 6. Effect of p38 signal on resting expression of E-and P-selectins in lung endothelial cells and platelets. A, endothelial cells were freshly isolated from the lung of WT and p38␣ ϩ/Ϫ littermates and maintained for 2 days. Then the cell lysate was prepared, and 7.5 g of lysate was subjected to each Western blot analysis. B, platelets were freshly prepared from WT and p38␣ ϩ/Ϫ littermates. The cell lysate (15 g) and total RNA were prepared and subjected to each Western blot analysis and reverse transcription-PCR for P-selectin mRNA, respectively. Similar results were confirmed in three independent experiments.

p38␣ in Lung Metastasis
experimental conditions (data not shown). On the other hand, the expression of p38 and basal activation of p38 and ATF-2 were reduced in platelets from p38␣ ϩ/Ϫ mice compared with those from WT mice, which showed a good parallelism to the case with lung endothelial cells. In accordance with this, the expression of P-selectin in platelets from p38␣ ϩ/Ϫ mice was lower than that from WT mice. Likewise, the expression of P-selectin mRNA in platelets was detected, and its level in platelets from p38␣ ϩ/Ϫ mice was lower than that from WT mice. The expression level of CD41 as internal control was comparable between WT and p38␣ ϩ/Ϫ mice (Fig. 6B). The reduction in expression of E-selectin in endothelial cells from p38␣ ϩ/Ϫ mice was not restricted in lung. As shown in Fig. 7, the expression level of E-selectin on liver endothelial cells from p38␣ ϩ/Ϫ mice was significantly lower than that in WT mice. In the case of P-selectin, its expression was weakly detected and also reduced on liver endothelial cells from p38␣ ϩ/Ϫ mice compared with that from WT mice.
It has been demonstrated that p38 regulates the expression of E-selectin in vitro (38,39) and that high expression of both p38␣ and -␤ was found in endothelial cells (40), suggesting the possibility that p38 isoforms cooperatively play a role in the regulation of E-selectin. At least in vivo, however, disruption of a single copy of the p38␣ gene may be sufficient to affect E-selectin mRNA expression in endothelial cells in a resting state as well as in an activated state. Moreover, also in vitro, p38␣ heterogeneity affects the expression of E-selectin in endothelial cells in a resting state. As a result, the decrease in expression of E-selectin could be a rate-limiting factor for tumor cell adhesion to the endothelium in p38␣ ϩ/Ϫ mice. Indeed, the number of B16 cells lodged in the capillaries of the lung was decreased in p38␣ ϩ/Ϫ mice compared with WT mice (WT, 18.7 Ϯ 10 colonies/0.16 mm 2 ; p38␣ ϩ/Ϫ , 5.0 Ϯ 2.4 colonies/0.16 mm 2 ; 50 slides of lung sections prepared from calcein-labeled B16-bearing mice were examined). On the other hand, it was reported that tumor necrosis factor-␣ and lipopolysaccharide, which are known to be p38 activators, induce expression of the murine P-selectin gene in endothelial cells (41). Furthermore, in mice lacking ATF-2, a specific substrate for p38, lipopolysaccharide-induced P-selectin mRNA expression in the lung was markedly attenuated (42). These reports give reliability to our present data that the resting expression of P-selectin reduced in platelets from p38␣ ϩ/Ϫ mice correlates with the level of phosphorylated ATF-2. However, further study is needed to elucidate whether ATF-2 directly activates the P-selectin promoter. The question arises of whether the induction of P-selectin mRNA by metastatic tumor cells is restricted to endothelial cells. Circulating platelets do not have nuclei, and P-selectin constitutively stored in the ␣ granules of platelets can be functionally expressed on the membrane surface by the rapid redistribution in response to stimuli (43). However, circulating platelets retain mRNA for certain molecules from megakaryocyte (44). In fact, P-selectin mRNA was detected, and its level in platelets from p38␣ ϩ/Ϫ mice was lower than that from WT mice. Thus, the possibility that P-selectin mRNA derived from platelets interacting with tumor cells may accumulate in accordance with the number of tumor cells lodged in the capillaries of the lung is not ruled out.
Reduction of Tumor Cell-Lung Endothelial Cell Interaction in p38␣ ϩ/Ϫ Mice-Our present data suggest that the reduction of E-and P-selectins expression in endothelial cells and platelets, respectively, may lead the low metastatic potential of tumor cells for lung in p38␣ ϩ/Ϫ mice. To further elucidate this notion, attachment and transmigration of DiI-labeled B16 for lung endothelial cells in the absence and presence of platelets were investigated. As shown in Fig. 8A, attachment of B16 cells on endothelial cells from WT mice was increased by the addition of platelets from WT mice, and B16 cells formed large colonies (Fig. 8B). These profiles were not typical in case of using endothelial cells and platelets from p38␣ ϩ/Ϫ mice (Fig. 8D). Quantitative analysis showed that the numbers of B16 cells attached on endothelial cells with or without platelets from p38␣ ϩ/Ϫ mice were significantly decreased compared with those from WT mice, respectively. The application of platelets from WT mice significantly increased the attaching activity of B16 cells on endothelial cells from WT mice. This increase was significant but slight in the case of p38␣ ϩ/Ϫ mice (Fig. 8B). These results clearly support the fact that the number of B16 cells lodged in the capillaries of the lung was decreased in p38␣ ϩ/Ϫ mice compared with WT mice. As shown in Fig. 8C, the transmigration of B16 cells was comparable over all experimental conditions, indicating that a barrier function of lung endothelial cells from p38␣ ϩ/Ϫ mice is similar to that from WT mice. Transendothelial migration of tumor cells is a crucial step for cancer metastasis (45). However, these results suggest that the resistance of p38␣ ϩ/Ϫ mice to experimental lung metastasis of tumor cells does not result from the change in barrier function of lung endothelial cells in p38␣ ϩ/Ϫ mice.
In conclusion, the present study demonstrated that p38␣ plays an important role in a step of tumor cell metastasis, pos- sibly through regulating the formation of tumor-platelet aggregates and their interaction with the endothelium accompanied by selectin expression.