Expression of granulocyte colony-stimulating factor and its receptor in epithelial skin tumors

Journal of Dermatological Science 25 (2001) 179– 188 www.elsevier.com/locate/jdermsci

Expression of granulocyte colony-stimulating factor and its receptor in epithelial skin tumors Koichiro Hirai a,*, Masanobu Kumakiri a, Shigeharu Fujieda b, Hiroshi Sunaga b, Li-Min Lao a, Yoshiaki Imamura c, Keiichi Ueda a, Masaru Fukuda c a

b

Department of Dermatology, Fukui Medical Uni6ersity, Matsuoka, Shimoaizuki 23, Yoshida, Fukui 910 -1193, Japan Department of Otorhinolaryngology, Fukui Medical Uni6ersity, Matsuoka, Shimoaizuki 23, Yoshida, Fukui 910 -1193, Japan c Department of Pathology, Fukui Medical Uni6ersity, Matsuoka, Shimoaizuki 23, Yoshida, Fukui 910 -1193, Japan Received 24 April 2000; received in revised form 10 July 2000; accepted 12 August 2000

Abstract Granulocyte colony-stimulating factor receptors (G-CSFR) have been observed on the surface of not only hematopoietic cells but also several cancer cells. In the present study, we investigated the expression of G-CSFR or G-CSF in epithelial skin tumors by immunohistochemical staining. The assessments were defined by the percentage of G-CSFR or G-CSF positive cells and expressed as G-CSFR and G-CSF scores. The G-CSFR score in SCC (77.6 920.0%) was significantly higher than that in Bowen’s disease (BD) (51.0 9 35.6%), actinic keratosis (AK) (49.3 9 34.6%) or normal skin (30.0 932.1%) (P =0.0004, P =0.0003, PB 0.0001, respectively). The mean G-CSF score in SCC (56.7 927.4%) or in BD (44.1 931.4%) was higher than that in normal skin (24.9 925.8%) (P = 0.0075, PB0.001, respectively). G-CSF expression in AK (29.8 9 31.2%) was lower than that in SCC (P= 0.0037). There was significant positive correlation between the G-CSFR score and the G-CSF score (k = 0.274, P= 0.0107) in skin tumors. These findings suggested that the assessment of G-CSFR expression might be associated with carcinogenesis of skin tumors. © 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Granulocyte colony-stimulating factor receptor; Granulocyte colony-stimulating factor; Bowen’s disease; Actinic keratosis; Squamous cell carcinoma

1. Introduction Granulocyte colony-stimulating factor (G-CSF) is a glycoprotein existing in various species that * Corresponding author. Tel.: + 81-776-613111, ext. 2328; fax: + 81-776-618112. E-mail address: [email protected] (K. Hirai).

regulates the growth and differentiation of granulocytic progenitor cells [1]. Stimulation with GCSF plays an important role in the local host defense response to inflammatory disorders or infection through granulocytic phagocytosis, chemotaxis and microbicidal activities [2,3]. However, a recent study showed that the effect of G-CSF was not limited to bone marrow cells or to

0923-1811/01/$ - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 9 2 3 - 1 8 1 1 ( 0 0 ) 0 0 1 3 1 - 6

180

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

hematopoietic derivation. It has been demonstrated that several malignant tumors, including bladder carcinoma [4], hepatocarcinoma [5], malignant mesothelioma [6], SCC of the oral cavity [7,8], eccrine carcinoma [9] and cell lines derived from skin carcinoma [10] secreted large amounts of G-CSF. Various effects of G-CSF are triggered by the binding of G-CSF to its receptor on the surface of the cells [11,12]. G-CSF receptors have been reported to be present on cells such as myeloblasts and mature neutrophils [13]. G-CSF receptors have also been demonstrated on a variety of cells, including human myeloid leukemic cells [14] and leukemic cell lines [15], human placenta and trophoblastic cells [16], human vascular endothelial cells [17], oral and mesopharyngeal carcinoma [18], bladder carcinoma [19], cell lines derived from human small-cell carcinoma of the lung [20] and cell lines derived from skin carcinoma [10]. In the present study, we investigated the expression of G-CSFR and G-CSF in Bowen’s disease (BD), actinic keratosis (AK), squamous cell carcinoma (SCC) and normal skin using an immunohistochemical method to elucidate the role of G-CSFR and G-CSF in human skin carcinogenesis.

2. Materials and methods

2.1. Patients sample collection We examined 41 patients with BD, 28 patients with AK and 31 patients with SCC who had undergone tumor-curative resections in the Department of Dermatology, Fukui Medical University between 1984 and 1998. The mean age of the patients was 73.2, 75.4 and 70.2 years, respectively (median age, 72.0, 81.5 and 73.0 years, respectively). Surgically resected tumor tissues were quickly fixed in 10% buffered formaldehyde for 24 h and embedded in paraffin.

2.2. Immunohistochemical staining Immunohistochemical staining was performed to detect G-CSFR in BD, AK and SCC tissues

using the traditional avidin-biotin-peroxidase complex technique. Paraffin-embedded blocks were cut into 5 mm thick sections and deparaffined by ethanol. After washing in distilled water and rinsing with PBS (pH 7.4), inhibition of endogenous peroxidase activity was accomplished by incubation in 0.3% H2O2 solution dissolved in absolute methanol at room temperature for 15 min. All specimens were washed in distilled water, rinsed in PBS and incubated with normal sheep serum (DAKO LSAB kit; DAKO, Carpinteria, CA) for 5 min at room temperature to block the background absorption of antiserum. Then, they were incubated with mouse antiserum G-CSFR mAb (LMM741 clone; PharMingen, San Diego, CA; diluted 1:50; G-CSFR mAb, [18]) at 4°C overnight. All specimens were treated with goat anti-mouse biotylated IgG (DAKO). Specimens were then rinsed with PBS and allowed to react with the avidin-biotin-peroxidase complex for 40 min at room temperature. After rinsing with PBS, peroxidase color visualization was carried out with 3,3-diaminobenzidine tetrahydrochloride solution (DAB; Dojin, Kumamoto, Japan; 30 mg dissolved in 150 ml of PBS and added to 10 ml of 30% H2O2 solution). Nuclear counterstaining was carried out with Harris hematoxylin for 30 s before mounting. The staining of G-CSF was similar to that for G-CSFR except for the Ab. Rabbit anti-G-CSF polyclonal Ab (Chugai Pharmaceutical Co., Tokyo, diluted 1:200; G-CSF Ab) was used. Goat anti-rabbit biotylated IgG (DAKO) were used as a second Ab. The staining of Ki-67 was similar to that for G-CSFR except for the Ab. Mouse antiserum Ki-67 mAb (MIB-1 clone; Immunotech, France, CA; diluted 1:50; Ki-67 mAb) was used.

2.3. E6aluation of specimens For microscope analysis of G-CSFR staining, we selected five high-powered fields, with each field containing more than 200 tumor cells, and counted both the number of positive and the total number of cancer cells. G-CSFR stainings were scored independently by two doctors (K.H. and H.S.) in a coded manner (without knowledge of the clinical parameters and outcomes). In total, at

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

least 1000 tumor cells were counted. We calculated the averaged of ten readings showing the percentage of G-CSFR-positive cells and expressed them as the G-CSFR score. Infiltrating neutrophils in the section were used as a positive control of the stain for the G-CSFR mAb, and mouse IgG or rabbit serum was used as a negative control instead of primary Ab for G-CSFR. The microscope analysis of G-CSF and Ki-67 was similar to that for G-CSFR.

2.4. Statistical analysis The comparison in G-CSFR, G-CSF and Ki-67 among the four groups (BD, AK, SCC and normal skin) was performed using the non-parametric Mann –Whitney U-test and non-parametric Spearman rank correlation coefficient test. A Macintosh personal computer system (Stat View software; Abacus’ Concepts Inc., Berkeley, CA) was used for all statistical analyses.

181

20.0% ranging from 0 to 90%, which was significantly higher than that in patients with BD or AK (P= 0.0004, P = 0.0003, respectively). Only one SCC patient did not express G-CSFR. Forty-nine of 100 samples had histologically normal adjacent tissue. The mean G-CSFR score in normal skin was 30.09 32.1%. G-CSFR score in normal skin was lower than that in BD, AK or SCC (P= 0.004, P= 0.0156, P B 0.0001, respectively, Fig. 2). Although, several normal skin samples had high G-CSFR score, the average G-CSFR score in normal skin was lower than that in corresponding tumor tissue. We had clinical outcomes about only 11 patients. The average G-CSFR score of patients who died of SCC (four cases) was higher than that of patients without recurrence (seven cases) (83.8 vs. 56.4%), but there was no significant difference between the two groups.

3.2. G-CSF expression 3. Results

3.1. G-CSFR expression The usual pattern of positive staining for GCSFR was cytoplasm and plasma-membrane of tumor cells. However, partial heterogeneity of staining was observed in each of these cases. There was no difference in the pattern of positive staining among BD, AK and SCC (Fig. 1). Even if tumor cells were negative for G-CSFR staining, infiltrated neutrophils in the section were positive and were used as positive controls of the staining for G-CSFR (data not shown, [18]). The mean G-CSFR score was 51.09 35.6% (mean9 S.D.) ranging from 0 to 90% in patients with BD (Fig. 2). Nine of 41 patients with BD (18.2%) had no tumors exhibiting G-CSFR. In patients with AK, the mean G-CSFR score was 49.3934.6% ranging from 0 to 85%. Eight of 28 AK (18.2%) had no expression of G-CSFR. The distribution and average G-CSFR in patients with AK was similar to that in patients with BD. Many patients with SCC had a high G-CSFR score. The mean GCSFR score in patients with SCC was 77.69

The same specimens that were examined for G-CSFR expression were also stained with antiG-CSF Ab (Fig. 3). The usual pattern of positive staining for G-CSF was in the cytoplasm of tumor cells. No difference was found in the pattern of positive staining among BD, AK and SCC (Fig. 3). The mean G-CSF score was 44.19 31.4% (mean9S.D.) ranging from 0 to 90% in patients with BD (Fig. 4). Six of 40 patients with BD (15.0%) had no tumors exhibiting G-CSF. In patients with AK, the mean G-CSF score was 29.89 31.2% ranging from 0 to 90%. Five of 22 AK (22.7%) had no expression of G-CSF. The G-CSF score in patients with AK tended to be lower than that in patients with BD, but there was no significant difference in G-CSF score between the two groups. In patients with SCC, the mean G-CSF score was 56.79 27.4% ranging from 0 to 90%, which was higher than that in patients with AK (P= 0.0037). Only one SCC did not express G-CSF. The mean G-CSF score in normal skin was 24.9925.8%. G-CSF score in normal skin was lower than that in BD, AK or SCC (P= 0.0075, P= 0.5134, PB0.001, respectively, Fig. 4).

182

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

Fig. 1. Immunohistochemical staining of G-CSFR. (a) Strong positive G-CSFR staining in patients with BD (G-CSFR score, 90%; ×132). (b) Negative G-CSFR staining in patients with BD (G-CSFR score, 0%; × 132). (c) Strong positive G-CSFR staining in patients with AK (G-CSFR score, 85%; × 132). (d) Negative G-CSFR staining in patients with AK (G-CSFR score, 0%; × 132). (e) Strong positive G-CSFR staining in patients with SCC (G-CSFR score, 90%; × 132). (f) Moderate positive G-CSFR staining in patients with SCC (G-CSFR score, 42%; × 132). The allow ( ’ ) indicates positive cells of G-CSFR. The symbol (*) indicates negative cells of G-CSFR. (g) Negative G-CSFR staining in patients with SCC (G-CSFR score, 0%; ×132).

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

3.3. Correlation between G-CSFR expression and G-CSF staining We examined the correlation between G-CSFR expression and G-CSF staining in all skin tumors studied. There was significant positive correlation between the G-CSFR score and G-CSF staining score using the non-parametric Spearman rank correlation coefficient test (k = 0.274, P =0.0107). Sequential preparation shows that SCC expressed both G-CSFR and G-CSF in same area (data not shown).

183

3.4. Ki-67 labeling index All specimens examined for G-CSFR expression were also stained with anti-Ki-67 Ab (Fig. 5). No difference was found in the pattern of positive staining among BD, AK and SCC (Fig. 5). The mean Ki-67 index was 51.4920.1% (mean9 S.D., median 50.0%) in patients with BD, 36.19 16.5% (median 35.0%) in patients with AK and 49.89 19.6% (median 45.0%) in patients with SCC. The Ki-67 index in patients with AK was lower than that in patients with BD or SCC

Fig. 2. G-CSFR score in BD, AK, SCC and normal skin. G-CSFR score of SCC is the highest among different types of lesions. The bar indicates the mean of the G-CSFR score.

184

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

Fig. 3. Immunohistochemical staining of G-CSF. (a) Strong positive G-CSF staining in patients with BD (G-CSF score, 90%; ×132). (b) Negative G-CSF staining in patients with BD (G-CSF score, 0%; × 132). (c) Strong positive G-CSF staining in AK (G-CSF score, 90%; × 132). (d) Negative G-CSF staining in patients with AK (G-CSF score, 0%; ×132). (e) Strong positive G-CSF staining in patients with SCC (G-CSF score, 90%; × 132). (f) Moderate positive G-CSF staining in patients with SCC (G-CSF score, 48%; ×132). The allow ( ’ ) indicates positive cells of G-CSF. The symbol (*) indicates negative cells of G-CSF. (g) Negative G-CSF staining in patients with SCC (G-CSF score, 0%; × 132).

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

185

Fig. 4. G-CSF score in BD, AK, SCC and normal skin. G-CSF score of SCC is the highest among different types of lesions. The bar indicates the mean of the G-CSF score.

(P = 0.0020, P= 0.013, respectively). There were no difference in Ki-67 index between patients with BD and patients with SCC. The mean Ki-67 index in normal skin was 5.592.2% (median 5.0%). Ki-67 index in normal skin was lower than that in BD, AK or SCC (P =0.0020, P B 0.0001, P B 0.0001, respectively).

3.5. Correlation between Ki-67 labeling index and G-CSF or G-CSFR expression We examined the correlation between Ki-67 labeling index and G-CSF or G-CSFR expression

in all skin tumors studied. Ki-67 labeling index was not correlated with the expression of GCSFR or G-CSF. 4. Discussion In the present study, G-CSFR expression increased in a step wise fashion from low levels in normal skin to high levels in BD, AK and SCC. The G-CSFR expression in SCC was significantly higher than that in BD and AK. Although, the physiological functions of GCSFR on the surface of skin tumor cells remain

186

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

unclear, expression of G-CSFR in oral and oropharyngeal SCC was associated with detrimental clinical outcomes [18]. The signal mediating G-CSFR possibly enhances the invasion activity of tumor cells and accelerates the metastatic potential of tumor cells. We have demonstrated that exogenous administration of G-CSF to a head and neck-derived SCC cell line with expression of G-CSFR augments the invasive potential with increased matrix metalloproteinase (MMP) [21]. MMP have been strongly implicated in the process of invasion [22]. In AK, a significant increase in MMP was reported compared with that in normal skin [23]. Additionally, the MMP in SCC was higher than that in AK [24,25]. The findings of the present study are consistent with the previous studies [21 –25] on G-CSFR expression. Concerning the prognosis for skin cancer, expression of epidermal growth factor (EGF) was closely associated with cellular proliferation and contributed to the clinical outcome [26]. Cancer cells traverse extracellular matrix barriers by mo-

bilization of proteolytic enzymes in response to epidermal growth factor (EGF)-EGF receptor (EGFR) interactions. EGF stimulation induced the elevation of MMP in cancer cells [27]. EGF is one of the factors capable of inducing the elevation of MMP in cancer cells. Thus, the expression of G-CSFR on skin tumors might reflect the potentiality of invasion to the basement membrane of the epidermis in a similar fashion to EGFR expression. Only 11 patients with SCC were followed-up for at least 3 years. G-CSFR expression in patients with poor prognosis tended to be higher than that in patients with good responder for the treatments. A large number of epidemiological study is now going on. G-CSFR-positive cancer cells have been shown to proliferate in vitro by exogenous G-CSF [10,19,28]. While G-CSF is generally produced by fibroblasts [29] and endothelial cells [30], we have shown that tumor cells in SCC and BD also produce G-CSF. These findings suggested that G-CSF and its receptor acts as an autocrine loop

Fig. 5. Immunohistochemical staining of Ki-67, (a) positive Ki-67 staining in patients with BD (Ki-67 index, 70%; × 132); (b) positive Ki-67 staining in patients with AK (Ki-67 index, 85%; ×132); (c) positive Ki-67 staining in patients with SCC (Ki-67 index, 80%; × 132).

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188

mechanism for the proliferation of skin tumors. This concept has previously been demonstrated [10,31]. Muller et al. [10] recently reported that all HaCaT keratinocyte clones express the G-CSFR and secrete G-CSF. G-CSF exhibited a stimulation of cell proliferation and migration of high grade malignant HaCaT cells. Additionally, both proliferation and migration of the high grade malignant cells were strongly inhibited by neutralizing antibody to G-CSF. Thus, they demonstrated that the functional role of G-CSF in high grade malignant skin cancer through an autocrine mechanism in vitro. In a human osteosarcoma cell line, transfection of human G-CSF expression vector into cell line made cells became autostimulatory and anchorage-independent colony formations [31]. However, the hypothesis that G-CSFR plays a role in the proliferation of skin cancer, was not be supported. Because, in the present study, expression of G-CSFR was not significant correlation with expression of Ki-67, a proliferation-associated protein. We have hypothesis that G-CSFR is one of critical signal in the invasion of skin cancer cells to the basement membrane of epidermis. Further studies are required to clarify many of these factors. rG-CSF is being used more frequently to treat patients with SCC after chemotherapy. Although, it is unclear whether G-CSFR in skin SCC is functional, we feel that care should be taken in the clinical use of rG-CSF in patients with skin SCC exhibiting G-CSFR. References [1] Nicola NA, Metacalf D. Binding of the differentiation inducer, granulocyte colony-stimulating factor, to responsive but not unresponsive leukemic cell lines. Proc Natl Acad Sci USA 1984;81:3765–9. [2] Asano S. Human granulocyte colony-stimulating factor its basic aspects and clinical applications. Am J Pediatr Hematol Oncol 1991;13:400–13. [3] Demetri GD, Griffin JD. Granulocyte colony-stimulating factor and its receptor. Blood 1991;78:2791–808. [4] Souza LM, Boone TC, Gabrilove J, Lai PH, Zsebo KM, Murdock DC, Chazin VR, Bruszewski J, Lu H, Chen KK, Barendt J, Platzer E, Moore MAS, Mertelsmann R, Welte K. Recombinant human granulocyte colony-stimulating factor: effect on normal and leukemic myeloid cells. Science 1986;232:61–5.

187

[5] Gabrilove JL, Welte K, Lu L, Castro-Malaspina H, Moore MA. Constitutive production of leukemia differentiation, colony-stimulating, erythroid burst-promoting, and pluripoietic factors by a human hepatoma cell line: characterization of the leukemia differentiation factor. Blood 1985;66:407– 15. [6] Demetri GD, Zenzie BW, Rheinwald JG, Griffin JD. Expression of colony-stimulating factor genes by normal human mesothelial cells and human malignant mesothelial cells lines in vitro. Blood 1989;74:940– 6. [7] Nagata S, Tsuchiya M, Asano S, Kaziro Y, Yamazaki T, Yamamoto O, Hirata Y, Kubota N, Oheda M, Nomura H, et al. Molecular cloning and expression of cDNA for human granulocyte colony-stimulating factor. Nature 1986;319:415– 8. [8] Tsukuda M, Nagahara J, Yago T, Matsuda H, Yanoma S. Production of granulocyte colony-stimulating factor by head and neck carcinomas. Biotherapy 1993;6:183– 7. [9] Suga T, Setoyama M, Kanekura H, Hamada H, Kanzaki T. Morphological, biochemical and molecular biological characteristics of a granulocyte colony-stimulating factorproducing human eccrine carcinoma cell line. J Dermatol Sci 1998;16:200– 7. [10] Mueller MM, Fusenig N. Constitutive expression of GCSF and GM-CSF in human skin carcinoma cells with functional consequence for tumor progression. Int J Cancer 1999;83:780– 9. [11] Fukunaga R, Ishizaka-Ishida E, Nagata S. Growth and differentiation signals mediated by different regions in the cytoplasmic domain of granulocyte colony-stimulating factor receptor. Cell 1993;74:1079– 87. [12] Nicholson SE, Oates AC, Harpur AG, Ziemiecki A, Wilks AF, Layton JE. Tyrosine kinase JAK1 is associated with the granulocyte colony-stimulating factor receptor and both become tyrosinephosphorylated after receptor activation. Proc Natl Acad Sci USA 1994;91:2985– 8. [13] Nicola NA, Metacalf D. Binding of 125I-labelled granulocyte colony-stimulating factor to normal murine hematopoietic cells. J Cell Physiol 1985;124:313– 21. [14] Begley CG, Metacalf D, Nicola NA. Primary human myeloid leukemia cells: comparative responsiveness to proliferative stimulation by GM-CSF or G-CSF and membrane expression of CSF receptors. Leukemia 1987;1:1– 8. [15] Park LS, Waldron PE, Friend D, Sassenfeld HM, Price V, Anderson D, Cosman D, Andrews RG, Bernstein ID, Urdal DL. Interleukin-3, GM-CSF and G-CSF receptor expression on cell lines and primary leukemia cells. Blood 1989;74:56– 65. [16] Uzumaki H, Okabe T, Sasaki N, Hagiwara K, Takaku F, Tobita M, Yasukawa K, Ito S, Umezawa Y. Identification and characterization of receptors for granulocyte colony-stimulating factor on human placenta and trophoblastic cells. Proc Natl Acad Sci USA 1989;86:9323– 6. [17] Bussolino F, Wang J, Defilippi P, Turriri F, Sanavio F, Edgell C-JS, Aglietta M, Arese P, Mantovani A. Granu-

188

[18]

[19]

[20]

[21]

[22]

[23]

[24]

K. Hirai et al. / Journal of Dermatological Science 25 (2001) 179–188 locyte-macrophage-colony stimulating factor induce human endothelial cells to migrate and proliferative. Nature 1989;337:471– 3. Tsuzuki H, Fujieda S, Sunaga H, Noda I, Saito H. Expression of granulocyte colony-stimulating factor receptor correlates with prognosis in oral and mesopharyngeal carcinoma. Cancer Res 1998;58:794–800. Tachibana M, Miyakawa A, Tazaki H, Nakamura K, Kubo A, Hata J, Nishi T, Amano Y. Autocrine growth of transitional cell carcinoma of the bladder induced by granulocyte colony-stimulating factor. Cancer Res 1995;55:3438– 43. Avalos BR, Gasson JC, Hedvat C, Quan SG, Baldwin GC, Weisbart RH, Williams RE, Golde DW, DiPerso JF. Human granulocyte colony-stimulating factor: biological activities and receptor characterization on hematopoietic cells and small cell lung cancer cell lines. Blood 1990;75:851– 7. Noda I, Fujieda S, Ohtsubo T, Tsuzuki H, Tanaka N, Sunaga H, Saito H. Granulocyte colony-stimulating factor enhances invasive potential of human head-and-neckcarcinoma cell lines. Int J Cancer 1999;80:78–84. Sato H, Takahisa T, Okada Y, Cao J, Shinagawa A, Yamamoto E, Seiki M. A matrix metalloproteinase expressed on the surface of invasive tumor cells. Nature 1994;370:61– 5. Tsukifuji R, Tagawa K, Hatamochi A, Shinkai H. Expression of matrix metalloproteinase-1, -2 and -3 in squamous cell carcinoma and actinic keratosis. Br J Cancer 1999;80:1087– 91. Wagner SN, Ockenfels SN, Wagner C, Soyer HP, Goos M. Differential expression of tissue inhibitor of metalloproteinases-2 by cutaneous squamous and basal cell carcinomas. J Invest Dermatol 1996;106:321–6.

.

[25] Airola K, Johansson N, Kariniemi AL, Kahari VM, Saarialho-Kere UK. Human collagenase-3 is expressed in malignant squamous epithelium of the skin. J Invest Dermatol 1997;109:225– 31. [26] Kikuchi A, Amagai M, Hayakawa K, Ueda M, Hirohashi S, Shimizu N, Nishikawa T. Association of EGF receptor expression with proliferating cells and of ras p21 expression with differentiating cells in various skin tumors. Br J Dermatol 1990;123:49– 58. [27] Rosenthal EL, Johnson TM, Allen ED, Apel IJ, Punturieri A. Role of plasminogen activator and matrix metalloproteinase systems in epidermal growth factor- and scatter factor-stimulated invasion of carcinoma cells. Cancer Res 1998;58:5221– 30. [28] Tachibana M, Miyakawa A, Uchida A, Murai M, Eguchi K, Nakamura K, Kubo A, Hata J-I. Granulocyte colonystimulating factor receptor expression on human transitional cell carcinoma of the bladder. Br J Cancer 1997;75:1489– 96. [29] Hamilton JA, Piccoli DS, Cebon J, Layton JE, Rathanaswani P, McColl SR, Leizer T. Cytokine regulation of colony-stimulating factor (CSF) production in cultured human synovial fibroblasts. II. similarities and differences in the control of interleukin-1 induction of granulocyte-macrophage CSF and granulocyte-CSF production. Blood 1992;79:1413– 9. [30] Bussolino F, Zieche M, Wang JM, Alessi D, Morbidelli L, Cremona O, Bosia A, Marchisio PC, Mantovani A. In vitro and in vivo activation of endothelial cells by colonystimulating factors. J Clin Invest 1991;87:986– 95. [31] Thacker JD, Dedhar S, Hogge D. The effect of GM-CSF and G-CSF on the growth of human osteosarcoma cells in vitro and in vivo. Int J Cancer 1994;56:236– 43.