Vascular Adhesion Protein-1 and ICAM-1 Support the. Adhesion of Tumor-Infiltrating Lymphocytes to Tumor. Endothelium in Human Hepatocellular Carcinoma.
Vascular Adhesion Protein-1 and ICAM-1 Support the Adhesion of Tumor-Infiltrating Lymphocytes to Tumor Endothelium in Human Hepatocellular Carcinoma Khong F. Yoong,1* Gillian McNab,* Stefan G. Hu¨bscher,† and David H. Adams2* T cell-mediated mechanisms are important in the defense against solid organ tumors. Why some tumors are more heavily infiltrated by T cells than others is poorly understood but is likely to depend upon adhesive interactions between circulating lymphocytes and tumor endothelium. In support of this hypothesis, the present study shows that primary human hepatocellular carcinomas (HCC) are more heavily infiltrated with T cells than colorectal hepatic metastases (CHM), and that their tumor vessels express high levels of several adhesion molecules. In HCC, an intense T cell infiltrate is observed within the tumor associated with strong expression of ICAM-1 and vascular adhesion protein-1 (VAP-1) on tumor endothelium. In contrast, fewer T cells infiltrated CHM and these tumors have little ICAM-1 and no detectable VAP-1 or VCAM-1 on tumor endothelium. T cells infiltrating both tumors are LFA-1 and very late Ag (VLA)-4 high. In vitro tissue-binding studies demonstrated that T cells bound readily to tumor endothelium in HCC, and Abs to ICAM-1, VAP-1, and to a lesser extent VCAM-1 could inhibit this binding. VAP-1 supported sialic acid-dependent adhesion under shear stress, suggesting that VAP-1 and ICAM-1 mediate, respectively, tethering and firm adhesion. In contrast, very few T cells bound to tumor vessels in CHM. Thus our data suggest that the VAP-1/VAP-1 receptor and ICAM-1/LFA-1 pathways are important in the recruitment of T cells to HCC. The strong expression of VAP-1 on tumor endothelium distinguishes HCC from CHM and supports our previous hypothesis that VAP-1 is an important hepatic endothelial adhesion molecule. The Journal of Immunology, 1998, 160: 3978 –3988.
T
cell-mediated mechanisms play an important role in controlling the growth of some types of tumor in both animals and humans (1). Tumor-specific CTL can be generated from several malignancies, including melanoma and head and neck squamous cell cancer, by culturing tumor-derived lymphocytes in high dose rIL-2 (2, 3). Furthermore, adoptive immunotherapy using rIL-2-expanded tumor-infiltrating lymphocytes (TIL)3 has resulted in objective responses in a proportion of patients with metastatic melanoma and renal cell carcinoma (4 – 6). Studies of TCR variable region gene expression show that accumulations of clonal T cells exist in many human solid tumors, suggesting T cell recognition of specific tumor Ags (1). Further evidence of a tumor-specific T cell response is provided by the identification of the tumor-associated Ags MAGE-1, MAGE-3, tyrosinase, MART-1, gp100, and gp75, all of which can stimulate T cell responses (1, 7).
*Liver Research Laboratories, Queen Elizabeth Hospital, Birmingham, United Kingdom; and †Department of Pathology, University of Birmingham Medical School, Birmingham, United Kingdom Received for publication February 10, 1997. Accepted for publication December 23, 1997. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 K.F.Y. was supported by a research fellowship from the Royal College of Surgeons of England. This work was supported in part by grants from the Wellcome Trust. 2 Address correspondence and reprint requests to Prof. D. H. Adams, Liver Research Laboratories, Clinical Research Block, Queen Elizabeth Hospital, Edgbaston, Birmingham B15 2TH, U.K. 3 Abbreviations used in this paper: TIL, tumor-infiltrating lymphocytes; HCC, hepatocellular carcinoma; CHM, colorectal hepatic metastasis; VAP-1, vascular adhesion protein-1; VEGF, vascular endothelial growth factor; MAdCAM, mucosal addressin cell-adhesion molecule-1; VLA, very late antigen; APAAP, alkaline phosphatase antialkaline phosphatase; hpf, high power field; HEV, high endothelial venule.
Copyright © 1998 by The American Association of Immunologists
To mount a successful immune response against a solid tumor, T cells must first enter the tumor tissue and then recognize and respond to tumor Ags. The mechanisms that regulate the recruitment of T cells to tumors are poorly understood. In general, T cells must first recognize and then adhere to endothelium before they can extravasate from the circulation into tissue (8 –10). This process is regulated by a sequence of molecular interactions involving cell adhesion molecules on both the T cell and the endothelium. The first step is a transient, tethering interaction classically mediated by selectins that bind to carbohydrate counterreceptors and induce the T cell to roll on the endothelium (11). Other molecules can also mediate tethering, including the a4 integrins and their ligands VCAM-1 and mucosal addressin cell-adhesion molecule-1 (MAdCAM-1) (12, 13). Strong, secondary adhesion is mediated by lymphocyte integrins such as LFA-1 and VLA-4 and their respective endothelial ligands, ICAM-1, ICAM-2, and VCAM-1 of the Ig superfamily. However, integrin-mediated adhesion requires activation for efficient engagement of integrin ligands, and it has been proposed that an additional step is required after tethering in which T cells come into contact with activating factors at the endothelial surface that trigger secondary, integrin-mediated adhesion (14, 15). These factors can be either cytokines, particularly those of the chemokine family, or cell surface molecules such as CD31 and CD73 (16). After secondary adhesion, the T cell extravasates across the vessel wall in response to local chemotactic factors (14, 17, 18). Thus, the combination of tethering (primary adhesion), triggering (integrin activation signal), integrin-mediated adhesion (secondary adhesion), and chemotactic factors are required to fulfill the combinatorial requirements for T cell recruitment (14). The involvement of particular combinations of molecules in the cascade can result in the recruitment of selected subsets of T cells to specific tissues. For example, the cutaneous lymphocyte Ag is expressed at high levels on T cells that migrate to the skin, where it mediates binding to dermal E-selectin (19) but not on T cells at 0022-1767/98/$02.00
The Journal of Immunology other inflammatory sites such as the liver (20). The integrin a4b7, which is expressed on low numbers of circulating T cells, is found on T cells that migrate to the gut, where it binds to a receptor, MadCAM-1, that is largely restricted to intestinal endothelium (21, 22). We have recently shown that the endothelial adhesion molecule vascular adhesion protein-1 (VAP-1), is constitutively expressed on liver vascular and sinusoidal endothelium, where it supports T cell adhesion, thus suggesting that it is involved in directing T cell recirculation to the liver (23). Although the precise mechanisms of T cell recruitment to tumors may differ from those involved in recruitment to inflamed tissue, it is likely that the same general principles apply. Thus, T cell recruitment to tumors will depend on the cellular adhesion molecules expressed by the T cell as well as the presence of appropriate endothelial ligands on tumor vessels. Once recruited to the tumor, the T cell must recognize and respond to tumor Ags (1). This process requires both recognition of tumor Ags presented by MHC molecules on the tumor cells and Ag-independent adhesive interactions that bring the T cell and tumor cells together and provide costimulation for T cell activation (24). Our immunohistochemical studies indicate that hepatocellular carcinomas (HCC) are more heavily infiltrated by T cells than colorectal hepatic metastases (CHM), suggesting that these tumors are candidates for adoptive immunotherapy with TIL. Support for this hypothesis comes from murine models in which adoptive immunotherapy has been successful (25). Novel therapies for HCC are much needed because of the very poor prognosis of these tumors with conventional therapy (26). In the present study, we have confirmed that HCC are more heavily infiltrated by T cells than CHM and looked for differences in the expression of adhesion molecules on tumor endothelium that might account for these findings.
Materials and Methods Tissues studied and patients’ characteristics Samples of fresh liver tumor and macroscopically normal liver tissue from elsewhere in the same specimen were obtained from 25 patients (14 males, 11 females) with a median age of 61 yr (range 21 to 79) who underwent liver resection for primary or secondary liver tumors at the Queen Elizabeth Hospital, Birmingham, U.K. Ten patients had well-differentiated HCC, including one fibrolamellar variant. Three of the HCC patients had background cirrhosis secondary to chronic hepatitis B, primary biliary cirrhosis, and primary sclerosing cholangitis. Fifteen patients had CHM (four well-, nine moderately, and two poorly differentiated tumors). Samples of tumor and macroscopically normal liver tissue (taken at a distance of more than 5 cm from the tumor margin of the same specimen) were snap frozen in liquid nitrogen and stored at 270°C until used for immunohistochemical analysis and tissue-binding assays.
Monoclonal antibodies The following mouse anti-human mAbs were used at saturating concentrations for flow cytometry and immunohistochemistry: UCHT-1 (antiCD3, IgG1; Dakopatts, Glostrup, Denmark); T4-4D7 (anti-CD4, IgG2a; Unipath, Bedford, U.K.); DK25 (anti-CD8, IgG1; Dakopatts); 63D3 (antiCD14, IgG1; a gift from Dr. S. Shaw, National Cancer Institute, Bethesda, MD); EBM11 (anti-CD68, IgG1; Dakopatts); 4KB128 (anti-CD22, IgG1; Dakopatts); ACT-1 (anti-CD25, IgG1; Dakopatts); UCHL-1 (antiCD45RO, IgG2; a gift from P. Beverly, University College, London, U.K.); MOC-1 (anti-CD56, IgG1; Dakopatts); L-78 (anti-CD69, IgG1; Becton Dickinson, Mountain View, CA); Ber-T9 (anti-CD71, IgG1; Dakopatts); F8/86 (anti-Factor VIII rAg, IgG1; Dako, Carpinteria, CA); CR3/43 (anti-HLA-Dr, IgG1; Dako); aIELb7 ACT-1 (anti-a4b7 integrin; IgG1; a gift from A. Lazarovits, University of Western Ontario, London, Canada); Ber-Act8 (anti-aIELb7, IgG1; Dako); Leu-8 (anti-CD62-L, IgG1; Becton Dickinson); HP/1 (anti-CD49d, IgG1; Coulter Immunology, Hialeah, FL); and 1.2B6 (anti-CD62E, IgG1; a gift from D. Haskard Imperial College, London, U.K.). The following mAbs were used at blocking concentrations of greater than 20 mg/ml of purified Ab for assessing binding in in vitro tissue-binding assays: 1B2 (anti-VAP-1, mouse IgG1 isotype) was
3979 a gift from Dr. Marko Salmi and Dr. Sirpa Jalkanen (Turku University, Turku, Finland); 84H10 (anti-CD54 (ICAM-1), mouse IgG1 isotype) was a gift from Dr. S. Shaw; and 2G7 (anti-CD106 (VCAM-1), mouse IgG1 isotype) was provided by Dr. W. Newman (Leukosite, Boston, MA). The following mAbs were used at a concentration of 5 mg/ml per 106 cells to block T cell adhesion molecules: 25.3.1 (anti-CD11a, mouse IgG1 isotype) used to block the a-chain of LFA-1 was purchased from Coulter, Luton, Bedford, U.K.; Kim249 (anti-CD11b, mouse IgG1 isotype) used to block Mac-1 was a gift from M. Robinson, Celltech, Slough, U.K.; R.15/7 (antiCD18, mouse IgG1 isotype) used to block the b-chain of LFA-1 was a gift from Dr. R. Rothlein (Boehringer Ingelheim, Hartford, CT), CN; MAB13 (anti-CD29, mouse IgG1 isotype) was kindly provided by Dr. K. Yamada (National Cancer Institute, Bethesda, MD).
Immunohistochemistry The phenotypic characterization of in situ TIL and the expression of endothelial adhesion molecules were studied by immunohistochemistry on 6-mm cryostat sections as described previously (20, 27). Tissue sections were fixed in acetone for 10 min at room temperature and then incubated with primary Ab, followed by secondary rabbit anti-mouse Ab, which was detected by an indirect alkaline phosphatase-anti-alkaline phosphatase (APAAP) technique, and the resulting enzyme complex developed with naphthol-AX and fast red. Sections were counterstained with Mayer’s haematoxylin. Incubations were done at room temperature for 45 min and sections were washed for 5 min with two changes of Tris buffer in between incubations. Normal tonsil sections were used as positive controls and sections stained with an irrelevant mouse primary mAb were used as negative controls. The enumeration of positive cells was conducted on 20 randomly selected high power fields (hpf) (magnification 3400) using an ocular grid on every section. The mean 6 SEM cells positive for each antigenic determinant was calculated for all patients in HCC and CHM. The intensity of endothelial staining on each section was graded from 0 to 3, where 0 5 absent, 1 5 weak, 2 5 moderate, and 3 5 strong staining. An overall mean score for the total number of patients in each group was then calculated for each endothelial adhesion molecule.
Lymphocyte isolation TIL were isolated from fresh tumor tissues removed at surgery as described previously (28). Tumor tissues were immediately cut into small pieces, washed, and digested using RPMI 1640 (Life Technologies, Paisley, U.K.) supplemented with 0.2% (w/v) collagenase type IV (Sigma, Poole, Dorset, U.K.) and 20% FCS (Life Technologies) for 2 to 3 h with continuous stirring at room temperature. The tumor digest was then passed through a nylon mesh to obtain a single cell suspension that was washed with PBS until the supernatant became clear. The single cell suspension was layered onto Ficoll-Hypaque (Lymphoprep; Nycomed, Oslo, Norway), centrifuged at 1600 rpm for 30 min at room temperature. TIL and tumor cells were recovered from the interface. Autologous PBL were isolated from heparinized venous blood obtained from the same patients immediately before surgery. PBL were separated by Ficoll-Hypaque centrifugation at 1600 rpm for 30 min at room temperature and then washed twice with PBS.
Culture of TIL and PBL TIL and PBL were cultured in RPMI 1640 supplemented with 10% (v/v) FCS, 100 U/ml penicillin, 100 mg/ml streptomycin, 250 ng/ml amphotericin, 2 mM L-glutamine, and 1000 IU/ml human rIL-2 (Chiron, Harefield, Middlesex, U.K.) at 37°C in a humidified atmosphere with 5% CO2. Cultures were started in 24-well plates at 0.5 3 106 cells/ml and when the cell concentration exceeded 2 3 106/ml, cultures were transferred to T75 flasks for further expansion.
Isolation and culture of hepatic endothelial cells Human intrahepatic endothelial cells were isolated from surplus liver tissue obtained from adult donor livers that had been reduced in size for transplantation into pediatric recipients as described previously (29, 30). Endothelial cells were isolated from approximately 150 g of human liver tissue after segmental perfusion with 5 mM calcium chloride, 0.05% collagenase H (Life Technologies), 0.025% dispase (Boehringer Mannheim, Mannheim, Germany), 0.0125% type 1-S hyaluronidase (Sigma), and 0.005% DNase (Boehringer Mannheim) in 10% HBSS. The cell suspension was filtered and centrifuged at 28 3 g to pellet hepatocytes and the residual supernatant centrifuged at 717 3 g to pellet nonparenchymal cells. The resulting pellet was washed and subjected to gradient centrifugation with 30% (w/v) metrizamide (Nycomed) before further separation using a JE-6B elutriator rotor (Beckman Instruments, High Wycombe, Buckinghamshire, U.K.) to isolate endothelial cells (31). The elutriated fraction was
3980
ADHESION MOLECULES IN HUMAN MALIGNANT LIVER TUMORS
then resuspended in mouse anti-human CD31 (NIH31-1 used at 10 mg/ml) and positively immunoselected using 107 sheep anti-mouse IgG1 (Fc)coated Dynabeads (Dynal U.K., Oslo, Norway). The immunoisolated cell fractions were resuspended in 2 ml of human endothelial basal growth medium, supplemented with penicillin and streptomycin (Life Technologies), 20% human AB1 serum (National Blood Transfusion Service, Edgbaston, Birmingham, U.K.), 10 ng/ml of vascular endothelial growth factor (VEGF), and 10 ng/ml hepatocyte growth factor (both from Bachem, Safron Walden, Essex, U.K.), plated onto collagen-coated six-well tissue culture plates (Costar, High Wycombe, Bucks, U.K.), and incubated at 37°C in a humidified atmosphere containing 3% CO2 in air. Once the cells had grown to subconfluence, they were characterized by immunocytochemistry and flow cytometry: .95% cells expressed the endothelial marker CD31 and .50% of cells expressed cytoplasmic VAP-1 (23). None expressed either the macrophage/leukocyte Ags (CD18 or CD45) or the fibroblast Ags as detected by mAb AS02 (Klein Fontenay 1; Dianova, Hamburg, Germany).
Flow cytometry The phenotypic composition of freshly isolated and rIL-2-expanded TIL and autologous PBL were analyzed by two-color flow cytometry using standard techniques (32). Single cell suspensions (106 cells/ml) were incubated with 5 ml of primary unconjugated mouse mAb followed by a 1 in 20 dilution of FITC-conjugated F(ab9)2 fragments of rabbit anti-mouse Ig (Dako). Thereafter, the cell suspension was incubated with normal mouse serum to saturate the free binding sites on the F(ab9)2 fragments before a final incubation with phycoerythrin-conjugated anti-CD3 (Dako). All incubations were conducted at 4°C for 30 min and cells were washed twice with PBS (0.02% w/v sodium azide and 2% v/v FCS) in between incubations. The cell suspension was fixed with 1% paraformaldehyde and analyzed using the FACS 440 (Becton Dickinson). A lymphocyte gate was set to exclude dead cells and debris and at least 10,000 cells were analyzed in each sample. Mouse isotypes IgG1 and IgG2a were used as controls.
Tumor tissue-binding assay We modified a previously published assay (23, 33, 34) to study the adhesion of rIL-2-expanded TIL and autologous PBL (cultured under the same conditions) to tumor endothelium on tissue sections. Cryostat sections (10 mm) of tumor tissue were cut onto poly-L-lysine-coated glass slides and fixed in acetone for 10 min before the study. TIL were preincubated with mouse anti-CD3 mAb for 30 min at 4°C and then washed twice with PBS and adjusted to a concentration of 107 cells/ml of medium (RPMI 1640 and 10% FCS). Sections were preincubated with control mAb or blocking mAb for 30 min at room temperature before addition of 150 ml of TIL suspension to each section with constant rotation (60 rpm) for 60 min at 4°C. Sections were gently washed with cold PBS to remove nonadherent lymphocytes and fixed in acetone for 10 min. Vascular endothelium and adherent lymphocytes were identified by anti-Factor VIII and anti-CD3 immunostaining, respectively, and developed with APAAP and fast red. The sections were counterstained with Mayer’s hematoxylin. The assays were done in the presence of a control mAb (S37) or blocking concentrations (.20 mg/ml) of the following mAb: 84H10 (anti-ICAM-1), 2G7 (antiVCAM-1), 1B2 (anti-VAP-1), 25.3.1 (anti-CD11a), R15/7 (anti-CD18), and KIM 249 (anti-CD11b). These mAbs have all been shown to block adhesion in binding assays (23, 30, 32, 35, 36). Four tissue sections were used for each mAb per patient and the number of lymphocytes adherent to vascular endothelium (detected with anti-Factor VIII) was counted in both HCC and CHM using an ocular grid on every section. In addition, the number of lymphocytes adherent to sinusoidal endothelium and to vascular endothelium in the HCC sections were counted separately. Twenty high power fields (magnification 3400) were randomly selected and counted on every tissue section. The number of lymphocytes adherent to endothelium in the presence of control mAb S37 defines 100% binding, and the number binding to vascular or sinusoidal endothelium in the presence of blocking mAb was expressed as a percentage of this. To determine whether VAP-1-mediated adhesion of TIL to HCC endothelium is carbohydrate dependent, the tissue-binding assays were repeated in three cases after pretreatment of tissue sections with neuraminidase to remove sialic acid residues as described previously (37). Briefly, tissue sections were incubated with 5 mU of Vibrio cholerae-derived neuraminidase (Sigma) in 50 mM sodium acetate buffer solution (pH 5.5 with 100 mM NaCl and 5 mM CaCl) for 30 min in a humidified chamber at 37°C. After neuraminidase digestion, the sections were thoroughly washed with distilled water to remove remaining enzymes and subsequently used for tissue-binding assays as described above. Control sections were incubated with buffer solution only in the first step and subsequently with the appropriate mAb.
Adhesion to liver-derived endothelium in vitro Human intrahepatic endothelial cells (between two and six passages) were plated out and cultured to confluence in 48-well tissue culture plates (for static assays) or 24-well plates for rotating assays in medium containing 100 U/ml of TNF-a for 24 h. The static adhesion assays were performed according to an established protocol as previously described (32). T cells were radiolabeled with 100 mCi of Na51CrO4, washed three times in PBS, resuspended in RPMI 1640 containing 0.2% BSA, and added to the confluent monolayers of endothelial cells at a final concentration of 2 3 106 cells/ml for 60 min at 37°C. At the end of the assay, nonadherent cells were removed by washing with PBS and the adherent cells lysed by incubation with 100 ml of 1% Igepal. Lysates were collected and analyzed for 51Cr activity. A modification of a previously described assay was used to determine lymphocyte adhesion to endothelial cells under conditions of shear stress (38). The assay was done as described for static adhesion except that lymphocytes were added to endothelial cells that were cultured in 24-well plates under constant rotation at 60 rpm for 2 h at 37°C. Nonadherent cells were then gently aspirated and the wells washed gently in PBS. Saturating concentrations of mAb to VAP-1 (1B2) or a control mAb A7 were added to block adhesion.
Statistics Results of positive cell enumeration are reported as mean 6 SEM per hpf. Differences between groups were analyzed by the nonparametric Wilcoxon’s rank sum test. The level of significance was taken at p , 0.05.
Results Characterization of the lymphocytic infiltrate in human malignant liver tumors There is a more intense T cell infiltrate in HCC when compared with CHM (Figs. 1 and 2). There were differences in the intensity and distribution of mononuclear cell infiltrates in HCC and CHM. CD31 cells were the predominant infiltrating cell type in HCC, and focal aggregates were often seen in close proximity to tumor vessels (Fig. 1A), whereas comparable numbers of CD31 T cells and macrophages were seen in CHM (Fig. 2). The T cell infiltrate was more intense in HCC, where there were significantly more CD31 T cells per hpf when compared with CHM (HCC mean 51 6 10 positive cells per hpf compared with CHM 19 6 4, p , 0.01; Fig. 2). Both tumors contained a predominance of CD41 cells in the mononuclear cell population (mean CD4/CD8 ratio 2.2 in HCC and 1.8 in CHM). In HCC, CD81 cells were found in the sinusoidal lumen and also in contact with tumor cells (Fig. 1C). In CHM, few CD31 cells were detected in the parenchyma of tumor tissue (Fig. 1B), but more intense T cell infiltrates were observed at the tumor periphery (Fig. 1D). There were few cells that were positive for the B cell marker CD22 in either tumor type (Fig. 2). Phenotype of freshly isolated TIL and PBL (Figs. 3 and 4) Two-color flow cytometric analysis revealed that more than 90% of freshly isolated TIL from both HCC and CHM were CD31 and CD45RO1 (Fig. 3) and a high proportion of TIL also expressed HLA-Dr and CD69, suggesting that they are activated memory T cells (Fig. 3). However, only a small percentage of the CD31 cells in either TIL or PBL were positive for the a-chain of IL-2 receptor (CD25) or the transferrin receptor (CD71) (Fig. 3). Very few freshly isolated TIL from either HCC or CHM expressed L-selectin (CD62) compared with autologous PBL (Fig. 4A). Comparable numbers of TIL and PBL expressed the a4b7 integrin; and up to 20% of CD31 TIL in both tumors expressed aIELb7 (Fig. 4A). In contrast, a higher proportion of CD31 cells in both TIL and PBL were positive for both the a- and b-chains of LFA-1 (CD11a and CD18, respectively) and, to a lesser extent, VLA-4 (CD49a and CD29, respectively) (Fig. 4B).
The Journal of Immunology
3981
FIGURE 1. Immunostaining of tumor cryostat sections with anti-CD3 (A, B, and D) and anti-CD8 (C) mAbs. Numerous CD31 cells are present in close proximity to tumor vessels (A) and CD81 cells in direct contact with tumor cells (C) in HCC. In CHM, few CD31 cells are detected in the central area of the tumor (B), but more are found in the peritumoral margin (D, n 5 uninvolved liver, T 5 tumor) (magnification 3240).
Phenotypic characteristics of IL-2-expanded TIL and PBL derived from HCC and CHM The phenotypic compositions of cultured TIL and PBL from both tumors used for in vitro adhesion assays are summarized in Table I. The phenotype of lymphocytes in culture was monitored serially by two-color flow cytometry until they reached phenotypic stability, which occurred after 4 wk of in vitro culture. All the early TIL and PBL cultures were enriched in CD41 T cells, but three of seven HCC TIL became CD81 enriched in the later stages. The expression of L-selectin was up-regulated in the early stages of culture but rapidly declined after 2 wk and remained low in longterm culture. CD31 cells from both TIL and PBL cultures showed higher expression of both the a (92% 6 4)- and b (90% 6 4)chains of VLA-4 in comparison to freshly isolated cells ( p , 0.01). TIL- and PBL-derived T cells used for adhesion assays expressed high levels of CD11a, CD18, CD29, and CD49a but low levels of L-selectin and CD11b according to flow cytometric analysis (Table I). Characterization of tumor endothelium: endothelium in HCC has an activated phenotype compared with CHM (Fig. 5)
adjacent nontumoral liver (Table II). In the autologous nontumoral tissues obtained from 15 patients without background cirrhosis, there was strong expression of VAP-1 on sinusoidal endothelial cells and moderate expression on hepatic arteries, and portal and hepatic veins. In contrast, Kupffer cells, hepatocytes, and biliary epithelial cells were all negative for VAP-1. A similar staining pattern was observed for ICAM-1 on hepatic endothelium, but in contrast to VAP-1, ICAM-1 was also strongly expressed on Kupffer cells although the parenchymal cells were negative. The pattern and intensity of both VAP-1 and ICAM-1 expression were consistent in all the sections studied. There was diffuse but weak expression of VCAM-1 on both the sinusoidal endothelium and Kupffer cells and variable, focal VCAM-1 expression on the endothelium of both hepatic and portal vessels. No E-selectin was detected on any structure in any of the sections studied. In primary HCC, the expression of VAP-1 and ICAM-1 on HCC sections was similar to that seen in nontumoral liver with a diffuse sinusoidal pattern of moderate to strong intensity in the tumor parenchyma in 9 of 10 cases of HCC (Table II). The exception was a fibrolamellar variant of HCC, in which the sinusoidal space has
Immunohistochemical study of the expression of endothelial adhesion molecules revealed differences between HCC, CHM, and
FIGURE 2. Quantitative immunohistochemical assessment of the phenotypic characteristics of mononuclear cells infiltrating HCC (n 5 10) and CHM (n 5 15). Twenty hpf (magnification 3400) were randomly selected for enumeration of positive cells in each section per patient. Results represent mean 6 SEM number of positive cells per hpf. There were significantly more CD31 T cells in HCC when compared with CHM ( p , 0.01).
FIGURE 3. Two-color flow cytometric analysis of activation molecule expression on freshly isolated CD31 TIL and autologous PBL from HCC (n 5 8) and CHM (n 5 15). There are significantly more CD45RO1 ( p , 0.001), HLA-DR1 ( p , 0.0001), and CD691 ( p , 0.0001) cells in the CD31 gated population in TIL from both tumors compared with their autologous PBL. Results are given as mean 6 SEM of percentage positive cells.
3982
ADHESION MOLECULES IN HUMAN MALIGNANT LIVER TUMORS Table I. Phenotypic characteristics of TIL and autologous PBL derived from HCC and CHM cultured for 4 to 6 wk in 1000 IU/ml IL-2 and analyzed by flow cytometry just prior to tumor tissue-binding assaysa HCC TIL HCC PBL (n 5 7) (n 5 7)
CD3 CD4 CD8 CD56 CD62L (L-selectin)b CD11ab CD11bb CD18b CD49db CD29b
CHM TIL (n 5 4)
CHM PBL (n 5 4)
91 6 3 86 6 4 93 6 4 41 6 12 58 6 6 57 6 7 41 6 14 37 6 7 26 6 4 14 6 3 16 6 4 28 6 4 561 18 6 4 862 c (5 6 2) (29 6 5) (7 6 2) 99 6 1 98 6 2 100 6 1 (115 6 6) (121 6 7) 22 6 6 ND 25 6 5 (18 6 4) (16 6 3) 99 6 2 100 6 1 98 6 3 92 6 3 82 6 2 92 6 5 90 6 3 80 6 4 92 6 4
83 6 3 50 6 9 34 6 5 20 6 3 21 6 4 (38 6 7) 99 6 3 ND 99 6 2 89 6 2 80 6 4
Data represent mean 6 SEM of percentage positive cells. Data represent percentage of positive cells in the CD3-gated cell population. c Values in parentheses represent mean 6 SEM of the differences between the median channel fluorescence of test mAb and that of control mAb. a b
FIGURE 4. Two-color flow cytometric analysis of cell adhesion molecule expression on freshly isolated CD31 TIL and autologous PBL from HCC (n 5 8) and CHM (n 5 15). A, A significantly lower percentage of CD31 cells in TIL from both tumors express L-selectin (CD62L) compared with autologous PBL (p , 0.0001). There is a small but significant percentage of TIL that are positive for aIELb7 mucosal lymphocyte antigen (MLA) compared with PBL (p , 0.001). In addition, a4b7 (ACT-1) is detected in up to 30% of CD31 cells in either TIL or PBL. B, The majority of CD31 cells in TIL and PBL express both the a- and b-chains of LFA-1 (CD11a and CD18) and VLA-4 (CD49d and CD29). Results are expressed as mean 6 SEM percentage positive cells in the CD31 gated population.
been replaced by fibrous tissue and immunoreactivity was detected only on the endothelium of tumor vessels. The staining intensity of the sinusoidal expression for VAP-1 was strong in 5 of 9 and moderate in 4 of 9 cases (Table II and Fig. 5F ). Similarly, the sinusoidal immunoreactivity for ICAM-1 was strong in 5 of 9 and moderate in 4 of 9 cases (Table II and Fig. 5E), although sections that were strongly positive for ICAM-1 did not show corresponding VAP-1 staining intensity. VAP-1 was detected on the endothelium of tumor vessels in all 10 HCC cases; in 1 case this was strong, in 7 cases moderate, and in 2 it was weak (Table II and Fig. 5C). ICAM-1 expression on tumor vessels in HCC was moderate in 3 of 10 and weak in 7 of 10 cases (Table II and Fig. 5B). In tumor sections from 2 HCC patients, there were focal areas of poorly differentiated tumor cells where the sinusoidal architecture was lost and VAP-1 and ICAM-1 expression were observed only on the endothelium of tumor vessels. In well-differentiated HCC, VAP-1 and ICAM-1 were consistently expressed on sinusoidal
lining cells and tumor endothelium. Weak VCAM-1 expression was detected on the endothelium of tumor vessels and sinusoids in 7 of 10 cases of HCC (Table II and Fig. 5, D and G). In contrast to the weak endothelial expression, macrophages lying in the sinusoidal lumen with appearance and distribution similar to Kupffer cells were strongly positive for VCAM-1. In contrast to HCC, the expression of endothelial adhesion molecules on CHM was very different. VAP-1 was absent from tumor vessels in 14 of 15 cases of CHM, and in the other case only very weak, focal expression was seen (Table II and Fig. 5J). ICAM-1 expression was absent on tumor endothelium in 13 of 15 and only weakly positive in the remaining 2 cases of CHM (Table II and Fig. 5I). Tumor vessels in CHM showed weak focal VCAM-1 expression in 5 of 15 but was absent in 10 of 15 cases (Table II and Fig. 5K). E-selectin was not detected on endothelial cells from either tumor type. T cells cultured from TIL and PBL bind to tumor endothelium in HCC via LFA-1/ICAM-1 and VAP-1 ligand/VAP-1 pathways (Figs. 6 –10) The ability of T cells derived from TIL and autologous PBL of HCC and CHM (cultured between 4 to 6 wk in 1000 IU/ml of rIL-2) to bind tumor endothelium was examined by an in vitro tumor tissue-binding assay conducted under rotary conditions (Fig. 6). Because sinusoidal endothelium and vascular endothelium represent two distinct endothelial cell types in HCC, we compared binding to vascular endothelium in HCC with binding to vascular endothelium in CHM; binding to sinusoidal endothelium was assessed separately (see below). HCC TIL-derived T cells bound readily to tumor vascular endothelium on tumor sections (15 6 4 cells/hpf; n 5 7). Autologous PBL-derived T cells cultured under the same conditions as TIL also bound to tumor endothelium although in lower numbers (9 6 3 cells/hpf; n 5 7) (Fig. 7A). When CHM TIL- and PBL-derived T cells cultured under the same conditions were tested for their ability to bind tumor vascular endothelium in autologous tumor sections, very few cells bound (Figs. 6E and 7A). In addition, the number of cells bound to vascular endothelium and sinusoidal endothelium in HCC sections was enumerated separately (Fig. 6, A and B). More cells bound to sinusoidal endothelium (TIL 34 6 5, PBL 22 6 5 cells/hpf) compared with vascular endothelium (TIL 15 6 4, PBL 9 6 3 cells/hpf), and T cells derived
The Journal of Immunology
3983
FIGURE 5. Immunolocalization of cell adhesion molecule expression on endothelial cells in human malignant liver tumors. ICAM-1, VAP-1, and VCAM-1 expression are detected on the vascular endothelium of tumor vessels (B, C, and D, respectively) and also on sinusoidal-type endothelium (E, F, and G, respectively) in HCC. In contrast, little or no ICAM-1 (I), VAP-1 ( J), and VCAM-1 (K) expression is detected on the endothelium of tumor vessels in CHM. Control sections for HCC (A) and CHM (H) are stained with an irrelevant mAb (S37). (C, E, F, G, magnification 3120; A, B, D, I, H, magnification 3240; J, K, magnification 3480)
from TIL bound more readily to either endothelium than those from autologous PBL (Fig. 7B). The ability of ICAM-1, VAP-1, and VCAM-1 expression on the vascular and sinusoidal endothelium of HCC to support TIL ad-
Table II. Expression of cellular adhesion molecules on different cell populations in macroscopically uninvolved liver HCC and CHMa
Uninvolved liver (n 5 15) Sinusoids Kupffer cells Portal vein Hepatic vein Hepatic artery HCC (n 5 10) Sinusoidal endothelium Vascular endothelium CHM (n 5 15) Vascular endothelium
VAP-1
ICAM-1
VCAM-1
2.43 (2–3) 0 1.87 (1–3) 1.67 (1–3) 1.53 (1–3)
2.57 (2–3) 2.57 (2–3) 0.40 (0–1) 1.30 (1–2) 0.27 (0–1)
1.14 (0–3) 1.43 (1–3) 0.57 (0–1) 0.57 (0–1) 0.50 (0–1)
2.56 (2–3) 1.90 (1–3)
2.56 (2–3) 1.30 (1–2)
0.70 (0–1) 0.70 (0–1)
0.06 (0–1)
0.13 (0–1)
0.33 (0–1)
Staining intensity was scored as follows: 0 5 absent, 1 5 weak, 2 5 moderate, and 3 5 strong. Values represent overall mean staining intensity with range of individual scores in parentheses. a
hesion was investigated by conducting the tissue-binding assays in the presence of blocking mAb alone and in various combinations (Figs. 6 and 8). Detailed analysis of binding between sinusoidal and vascular endothelium revealed that similar adhesion pathways were involved in binding to both structures. When used in isolation, mAbs to ICAM-1 and VAP-1 had the greatest inhibitory effect on binding to both types of endothelium (Fig. 6, G and F, respectively); anti-VCAM-1 inhibited binding to a lesser extent (Fig. 6, C and D). Approximately 60% of TIL adhesion to tumor endothelium in HCC could be inhibited by 84H10 (anti-ICAM-1 mAb) or 1B2 (anti-VAP-1 mAb) alone compared with S37 control mAb, whereas 2G7 (anti-VCAM-1 mAb) alone reduced adhesion by only 30% (Fig. 8). The combination of blocking mAb against ICAM-1, VCAM-1, and VAP-1 reduced binding by 90% (Figs. 6H and 8). Because LFA-1 (CD11a/CD18) and Mac-1 (CD11b/CD18) are both ligands for ICAM-1, we investigated the relative contribution of these molecules to T cell binding to HCC tumor endothelium. Flow cytometry revealed that TIL-derived T cells are CD11ahighCD11blowCD18high (Table I), and when tissuebinding experiments were repeated using blocking concentrations of mAbs to CD11a, CD11b, or CD18 alone, TIL binding
3984
ADHESION MOLECULES IN HUMAN MALIGNANT LIVER TUMORS
FIGURE 6. The effect of blocking mAb to endothelial adhesion molecules on TIL-derived T cell adhesion to HCC tumor endothelium under rotary conditions. T cells can be seen adhering to vascular endothelium and sinusoidal endothelium of HCC following pretreatment of tissue sections with control mAb S37 (A and B, respectively) and, to a lesser extent, antiVCAM-1 mAb 2G7 (C and D, respectively). There was little or no binding of T cells to vascular endothelium in CHM tissue section in the presence of control mAb S37 (E). Significantly fewer T cells bound to the vascular endothelium and the sinusoidal endothelium of HCC following pretreatment of sections with blocking concentrations (.20 mg/ml) of anti-VAP-1 mAb, 1B2 ( F); anti-ICAM-1 mAb, 84H10 ( G); or a combination of anti-VAP-1, -ICAM-1, and -VCAM-1 (H) mAbs. The adherent T cells were detected by anti-CD3 and endothelium by anti-Factor VIII mAb. Micrographs have been focused at the level of adherent TIL; hence, the indistinct underlying tissue. (A, C, E, F, G, H, magnification 3480; B, D, magnification 3240).
to tumor endothelium was not inhibited by anti-CD11b mAb. In contrast, anti-CD11a mAb reduced TIL binding by approximately 50% compared with control, and anti-CD18 mAb inhibited binding to a similar extent as anti-ICAM-1 mAb (Fig. 9). These results suggest that LFA-1 on TIL is the major ligand for ICAM-1 on HCC tumor endothelium. VAP-1 in normal tissue is heavily sialylated and the sialic acid residues are required for lymphocyte adhesion to VAP-1 in lymph node high endothelial venules (HEV) (37). To determine whether lymphocyte binding to VAP-1 on tumor endothelium is also dependent on sialic acid, we repeated the tissue-binding studies after treating the tumor tissue with neuraminidase. Adhesion was reduced by the presence of 1B2 alone when compared with the control mAb S37 (control, 19.3 6 1.9; 1B2, 10.5 6 9.4 cells/hpf), and neuraminidase alone also reduced adhesion (4.5 6 0.7 cells/hpf), but the combination of 1B2 and neuraminidase did not reduce adhesion further (3.9 6 0.6 cells/hpf) (Fig. 10). TIL bind to VAP-1 on human hepatic endothelial cells in culture under shear stress (Fig. 11) To determine whether VAP-1 supports shear-dependent adhesion, binding studies were conducted using monolayers of endothelial
cells isolated from human liver that expressed VAP-1 on the cell surface (data not shown). mAb 1B2 (anti-VAP-1) failed to inhibit adhesion to human hepatic endothelial cells under static conditions but reduced adhesion by up to 40% when assays were conducted under constant rotation. In contradistinction, 1B2 had no effect on either static or shear-dependent adhesion of the same TIL to the dermal microvascular cell line HMEC-1, which does not express VAP-1.
Discussion The extent of T cell infiltration varies between different solid tumor types and in many cases appears to reflect the potential susceptibility of the tumor to immunotherapy. Thus, malignant melanoma is heavily infiltrated by T cells and has been shown in preliminary trials to respond to immunotherapy whereas colorectal tumors are less heavily infiltrated and the response to immunotherapy has been disappointing (4, 28, 39 – 41). We observed that human HCC is frequently heavily infiltrated by T cells, suggesting that it too might be a suitable tumor for adoptive immunotherapy. In contrast, hepatic metastases from colorectal tumors contain fewer T cells, most of which are confined to the peritumoral liver
The Journal of Immunology
FIGURE 7. Results of T cell adhesion to liver tumors using an in vitro tissue-binding assay. T cell lines derived from TIL and PBL cultured in 1000 IU/ml of rIL-2 for 4 to 6 wk were added to tumor sections pretreated with control mAb (S37) and incubated at 4°C with rotation (60 rpm) for 60 min. A, A comparison of TIL- and PBL-derived T cells binding to vascular endothelium in HCC and CHM. B, In HCC, T cells adherent to vascular and sinusoidal endothelia were counted separately for both TIL and PBL. Autologous PBL-derived T cells cultured under the same conditions also bound but to a lesser degree than TIL-derived T cells. This figure shows the number of T cells binding per hpf; mean 6 SEM of seven experiments for HCC and four experiments for CHM using matched TIL and PBL and autologous tumor sections.
tissue rather than the tumor itself. Based on these initial observations, we hypothesized that qualitative and quantitative differences in endothelial adhesion molecule expression might account for the distinct inflammatory responses to primary and secondary liver tumors. In support of this, we have demonstrated that endothelium in HCC expresses functional adhesion molecules (including the liver-associated adhesion molecule VAP-1) that support T cell adhesion in vitro, whereas endothelium in CHM does not. Thus, lymphocytes are more likely to recognize and bind tumor endothelium in HCC, thereby promoting lymphocyte recruitment to the tumor. These observations suggest that HCC might be a suitable tumor for adoptive immunotherapy because adoptively transferred lymphocytes should be able to recognize and bind to the tumor endothelium in vivo. If TIL are to mediate antitumor responses, they must recognize and bind to tumor endothelium before migrating from the circulation into the tumor itself (42). Little is known about the expression of endothelial adhesion molecules in either primary or secondary liver tumors. Our results demonstrate that the phenotype of tumor endothelium in HCC differs markedly from that in metastatic colorectal carcinoma and that these differences are probably crucial in regulating T cell recruitment. With regard to adhesion molecules, endothelium in HCC is phenotypically and functionally similar to that in normal liver distant from the tumor. HCC con-
3985
FIGURE 8. TIL-derived T cells adhere to tumor endothelium in HCC via VAP-1-, ICAM-1-, and VCAM-1-dependent pathways. The binding of T cells to both vascular and sinusoidal endothelium could be inhibited by blocking mAbs 84H10 (anti-ICAM-1), 1B2 (anti-VAP-1), or, to a lesser extent, 2G7 (anti-VCAM-1) alone or in various combinations of mAb when compared with an irrelevant control mAb (S37). The number of T cells adherent to endothelium in the presence of control mAb S37 defines 100% binding. Binding of T cells to vascular or sinusoidal endothelium on tumor sections following pretreatment with blocking mAb is expressed as a percentage of binding in the presence of control mAb S37; data represent mean 6 SEM of three different patients.
tains two distinct types of endothelium; sinusoidal-like endothelial channels that run between the tumor cells and vascular endothelium on tumor vessels. Both of these endothelia stained strongly for ICAM-1 and VAP-1, two endothelial adhesion molecules that
FIGURE 9. LFA-1 on TIL-derived T cells is the main ligand for ICAM-1 on HCC endothelium. The binding of TIL-derived T cells to both vascular and sinusoidal endothelia in HCC tumor can be inhibited by mAb 25.3.1 (anti-CD11a), which blocks the a-chain of LFA-1 and mAb R15/7 (anti-CD18), but not mAb Kim249 (anti-CD11b). mAb R15/7, which blocks the b-chain of LFA-1, has approximately the same inhibiting effect on TIL-derived T cell adhesion as 84H10 (anti-ICAM-1). The number of T cells adherent to endothelium in the presence of control mAb S37 defines 100% binding. The binding of T cells is expressed as a percentage of binding in the presence of control mAb S37; data represent mean 6 SEM of three different patients.
3986
ADHESION MOLECULES IN HUMAN MALIGNANT LIVER TUMORS
FIGURE 10. Effect of neuraminidase on VAP-1-dependent adhesion to tumor endothelium. TIL-derived T cells binding to tumor endothelium in HCC were assessed in the presence of control mAb S37 or 1B2 (antiVAP-1 mAb) following treatment of liver tumor tissue sections with buffer alone (50 mM sodium acetate solution adjusted to pH 5.5 with 100 mM NaCl and 5 mM CaCl) or neuraminidase to remove sialic acid residues. Neuraminidase alone inhibited adhesion as efficiently as 1B2, and when 1B2 was used on sections that had been treated with neuraminidase, no further inhibition of binding was observed.
are found on sinusoidal and vascular endothelium in nonneoplastic liver (23, 43– 45). In contradistinction, endothelium in the CHM did not express VAP-1 and only expressed ICAM-1 weakly. We have previously demonstrated that VAP-1 is expressed constitutively in the liver where it can support binding of T cells to hepatic endothelium via an as yet unknown T cell ligand (23). The results of the present study, in which VAP-1 supported TIL adhesion to tumor endothelium in HCC, suggest that lymphocyte adhesion to endothelium in primary liver tumors and normal liver may be regulated by similar mechanisms. VAP-1 is a human endothelial adhesion molecule (34, 37), originally described by Salmi et al. (36) and Salmi and Jalkanen (37), which supports shear-dependent lymphocyte binding to HEV in lymph nodes. We observed that mAb to VAP-1 inhibited T cell adhesion both to the sinusoidal-type endothelium and to tumor
FIGURE 11. In vitro TIL-derived T cell adhesion assay using monolayers of liver EC (isolated human hepatic endothelial cells) vs skin EC (human microvascular endothelial cells). mAb 1B2 inhibits T cell binding to hepatic endothelial cells under rotating conditions but not in static assays. In contrast, 1B2 has no inhibiting effect on binding to skin EC. T cell binding to endothelial cells in the presence of control mAb A7 represents 100% binding. Data are expressed as % of control binding and represent mean 6 SEM of four individual experiments. EC, endothelial cells.
vessels in tissue sections of HCC. Adhesion could also be blocked using anti-ICAM-1 mAb and, to a lesser extent, anti-VCAM-1 mAb. Recent studies have proposed that VAP-1 mediates initial lymphocyte interactions with endothelium, suggesting that VAP-1 and ICAM-1 could play complementary roles in lymphocyte recruitment to HCC by mediating adhesion at different stages of the adhesion cascade (23, 46, 47). To pursue this, we tested TIL adhesion to monolayers of human liver-derived endothelial cells that express low but detectable levels of VAP-1 on their cell surface (29). VAP-1 supported adhesion to these endothelial cells only under conditions of shear induced by rotation; it had no effect in standard static adhesion assays. These observations are consistent with the recent report that VAP-1 mediates initial, primary interactions between lymphocytes and mesenteric venules (36). It has been suggested that lymphocytes can bypass the need for a tethering requirement in low flow systems such as the hepatic sinusoids. However, recent intravital studies suggest that lymphocytes interact with hepatic sinusoidal endothelium in a two-stage process (48), and we think it is likely that similar mechanisms will apply to the sinusoids in HCC. The failure of sinusoidal endothelium, both in normal liver and hepatoma, to express selectins (20, 48) suggests that VAP-1 might play a particularly important role in mediating primary adhesion to these specialized vessels. Thus, VAP-1 on HCC endothelium could bring the circulating lymphocyte into contact with the vessel wall, allowing subsequent secondary adhesion to be mediated via LFA-1 (which is expressed at high levels on the TIL in HCC) and ICAM-1, which is abundant on tumor endothelium. The weaker expression of VCAM-1 on tumor endothelium and the less marked inhibition of lymphocyte binding with mAb against VCAM-1 when compared with anti-ICAM-1 suggest that VCAM-1 is less important, although in vivo studies are required to confirm this. The lymphocyte receptor for VAP-1 is not known but previous studies have suggested that it acts independently of L-selectin and recognizes sialic acid residues on VAP-1 in tonsillar and lymph node HEV (36, 37). TIL binding to VAP-1 in HCC is also Lselectin independent because very little L-selectin is expressed by the TIL used for adhesion assays in this study. Furthermore, treatment of hepatoma tissue sections with neuraminidase abolished VAP-1-dependent adhesion, suggesting that adhesion to VAP-1 in hepatoma is also dependent on the presence of sialic acid residues (36, 37). In contrast to HCC, we were unable to demonstrate any appreciable TIL binding to tumor vessels in the hepatic metastases from colorectal carcinomas, a finding that is consistent with the very low levels of adhesion molecules expressed by vessels in these tumors. The inability of endothelium in CHM to support T cell adhesion may be responsible for the paucity of infiltrating T cells in this tumor. It is possible that tumor-derived factors actively suppress endothelial activation in CHM, thereby protecting the tumor from immune attack as has been proposed for other tumors (49). The expression of VAP-1 on HCC tumor endothelium is of great interest. VAP-1 is constitutively expressed on hepatic endothelium (23, 34), one of the few sites where VAP-1 expression is detected in the absence of inflammation. Although VAP-1 expression is up-regulated at inflammatory sites, including the gut and skin (50, 51), VAP-1 expression is very low in noninflamed mucosal vessels (50). In the present study, we detected strong expression of VAP-1 on the sinusoidal endothelium of nontumor autologous liver and also on the tumor vessels and sinusoids within the hepatomas. VAP-1 expression was absent in the colorectal metastases. This previously unreported finding provides further evidence that VAP-1 expression is characteristic of liver-derived endothelium. The fact that VAP-1 was expressed on endothelium in primary,
The Journal of Immunology liver-derived tumors but not in secondary tumors within the liver suggests that the anatomical position of the tumor within the liver is not sufficient in itself to induce VAP-1 expression. It is likely that the liver-cell origin of HCC is a crucial factor in determining the phenotype of the tumor endothelium. One factor that may be important in regulating the differentiation of hepatoma endothelium is VEGF, which we have recently shown is required to maintain VAP-1-expressing, human hepatic sinusoidal endothelial cells in culture (29). Other groups have reported that hepatoma cells express VEGF at the gene and protein levels (52), and our own unpublished observations show much stronger expression of VEGF in primary compared with secondary liver tumors (K. F. Yoong and D. H. Adams, unpublished observations). Thus, HCCderived VEGF could be responsible for the distinctive phenotype of HCC tumor endothelium. The alternative explanation that endothelial activation in HCC is a consequence of cytokines derived from the large numbers of infiltrating T cells seems unlikely because tumor endothelium showed strong expression of VAP-1 and ICAM-1 even in areas that were not heavily infiltrated by T cells. E-selectin was uniformly absent from both nontumoral and tumor tissue in both tumor types. Recent studies suggest that selectins are not involved in leukocyte interactions with inflamed hepatic endothelium (20, 48). However, E-selectin is expressed on hepatic vascular endothelium at sites of acute and chronic inflammation (20), and its absence from tumor endothelium might suggest active suppression by tumor-derived factors, as has been described for VCAM-1 and melanoma (49). TGF-b could be one such factor because several human tumors produce it, and it can inhibit the expression of E-selectin on human endothelial cell lines in vitro (53). We also looked for potential homing receptors on the TIL. T cells in both tumors were CD62Llow and CD45ROhigh, suggesting that they are memory cells. Two b7 integrins have been implicated in the recruitment and retention of T cells at mucosal sites within the gut and might therefore be involved in the recruitment of T cells to liver tumors. The a4b7 integrin mediates the binding of gut-tropic memory T cells to MadCAM-1, which is selectively expressed on mucosal endothelium in the intestines (21, 54). The aIELb7 integrin is expressed on intraepithelial T cells and mediates binding to E-cadherin on intestinal epithelial cells, thereby retaining T cells in the epithelium (55). However, b7 integrins were detected on a small percentage of TIL in either HCC or colorectal metastases. It thus seems unlikely that they play a major role in T cell recruitment to liver tumors. In the light of recent observations that aIELb7 is expressed on T cells within primary colorectal tumors, the lack of aIELb7 expression of TIL in CHM is surprising and may be related to down-regulation of E-cadherin on metastatic colorectal tumors (56, 57). In summary, the results of this study show that HCCs are more heavily infiltrated with T cells than hepatic metastases from colorectal carcinoma. This is possibly a consequence of the greater expression of functional adhesion molecules on endothelium in HCC, which is phenotypically similar to activated, nontumoral hepatic endothelium. In contrast, tumor endothelium in the hepatic metastases expressed low levels of adhesion molecules and failed to support T cell adhesion. The strong expression of VAP-1 on tumor endothelium in HCC supports our previous hypothesis that VAP-1 is an important hepatic endothelial adhesion molecule. We propose that TIL are recruited to hepatomas via interactions with VAP-1 (which mediates primary, tethering interactions) and ICAM-1 (mediating firm adhesion) on tumor endothelium. Why these T cells fail to suppress tumor growth is not known but will be important for the development of immunotherapy strategies for HCC. However, a better understanding of the mechanisms of T
3987 cell recruitment to tumor tissue will facilitate the generation of antitumor effector cells with the appropriate adhesion molecules to allow them to home to the tumor in adoptive immunotherapy. Such strategies will help to overcome the need for infusion of such large numbers of lymphocytes as is currently required to ensure that sufficient numbers of lymphocytes reach tumor deposits in melanoma (58).
Acknowledgments We are grateful to M. Salmi and S. Jalkanen for providing reagents and for critically reviewing the manuscript; Gillian McNab, Ann Williams, and Desley Neil for technical assistance; Dr. Simon Afford for helpful discussions; and our surgical colleagues P. McMaster, A. D. Mayer, and J. A. C. Buckels in Birmingham for providing the clinical specimens and allowing us to study their patients.
References 1. Robbins, P. F., and Y. Kawakami. 1996. Human tumor antigens recognised by T-cells. Curr. Opin. Immunol. 8:628. 2. Rivoltini, L., Y. Kawakami, K. Sakaguchi, S. Southwood, A. Sette, P. F. Robbins, F. M. Marincola, M. L. Salgaller, J. R. Yannelli, E. Appella, and S. A. Rosenberg. 1995. Induction of tumor-reactive CTL from peripheral blood and tumor-infiltrating lymphocytes of melanoma patients by in vitro stimulation with an immunodominant peptide of the human melanoma antigen MART-1. J. Immunol. 154: 2257. 3. Yannelli, J. R. 1991. The preparation of effector cells for use in the adoptive cellular therapy of cancer. J. Immunol. Methods 139:1. 4. Rosenberg, S. A., J. R. Yannelli, J. C. Yang, S. L. Topalian, D. J. Schwartzentruber, J. S. Weber, D. R. Parkinson, C. A. Seipp, J. N. Einhorn, and D. E. White. 1994. Treatment of patients with metastatic melanoma with autologous tumor-infiltrating lymphocytes and interleukin-2. J. Natl. Cancer Inst. 86:1159. 5. Escudier, B., A. Ravaud, M. Fabbro, J. Y. Douillard, S. Negrier, C. Chevreau, L. Mignot, D. Baume, T. Dorval, F. Vignal, F. Farace, and D. Maraninchi. 1994. High-dose interleukin-2 2 days a week for metastatic renal-cell carcinoma: a FNCLCC multicenter study. J. Immunother. 16:306. 6. Finke, J. H., P. Rayman, L. Hart, J. P. Alexander, M. G. Edinger, R. R. Tubbs, E. Klein, L. Tuason, and R. M. Bukowski. 1994. Characterization of tumorinfiltrating lymphocyte subsets from human renal-cell carcinoma: specific reactivity defined by cytotoxicity, interferon-gamma secretion, and proliferation. J. Immunother. 15:91. 7. van der Bruggen, P., C. Traversari, P. Chomez, C. Lurquin, E. De Plaen, B. van den Eynde, A. Knuth, and T. Boon. 1991. A gene encoding an antigen recognised by cytolytic T lymphocytes on a human melanoma. Science 2:1643. 8. Adams, D. H., and S. Shaw. 1994. Leucocyte endothelial interactions and regulation of leucocyte migration. Lancet 343:831. 9. Butcher, E. C., and L. J. Picker. 1996. Lymphocyte homing and homeostasis. Science 272:60. 10. Springer, T. A. 1994. Traffic signals for lymphocyte recirculation and leukocyte emigration: the multistep paradigm. Cell 76:301. 11. Lasky, L. A. 1992. Selectins: interpreters of cell-specific carbohydrate information during inflammation. Science 258:964. 12. Berlin, C., R. F. Bargatze, J. J. Campbell, U. H. Vonandrian, M. C. Szabo, S. R. Hasslen, R. D. Nelson, E. L. Berg, S. L. Erlandsen, and E. C. Butcher. 1995. Alpha-4 integrins mediate lymphocyte attachment and rolling under physiological flow. Cell 80:413. 13. Jones, D. A., L. V. McIntire, C. W. Smith, and L. J. Picker. 1994. A 2-step adhesion cascade for T cell endothelial-cell interactions under flow conditions. J. Clin. Invest. 94:2443. 14. Butcher, E. C. 1991. Leukocyte-endothelial cell recognition: three (or more) steps to specificity and diversity. Cell 67:1033. 15. Tanaka, Y., D. H. Adams, S. Hubscher, H. Hirano, U. Siebenlist, and S. Shaw. 1993. T-cell adhesion induced by proteoglycan-immobilized cytokine MIP-1? Nature 361:79. 16. Airas, L., J. Hellman, M. Salmi, P. Bono, T. Puurunen, D. J. Smith, and S. Jalkanen. 1995. Cd73 is involved in lymphocyte binding to the endothelium: characterization of lymphocyte vascular adhesion protein-2 identifies it as cd73. J. Exp. Med. 182:1603. 17. Mackay, C. R. 1996. Chemokine receptors and T-cell chemotaxis. J. Exp. Med. 184:799. 18. Hedrick, J. A., and A. Zlotnik. 1996. Chemokines and lymphocyte biology. Curr. Opin. Immunol. 8:343. 19. Picker, L. J., J. R. Treer, B. Ferguson-Darnell, P. A. Collins, P. R. Bergstresser, and L. W. M. M. Terstappen. 1993. Control of lymphocyte recirculation in man. II. Differential regulation of the cutaneous lymphocyte-associated antigen, a tissue-selective homing receptor for skin-homing T cells. J. Immunol. 150:1122. 20. Adams, D. H., S. G. Hubscher, N. C. Fisher, A. Williams, and M. Robinson. 1996. Expression of E-selectin (CD62E) and E-selectin ligands in human liver inflammation. Hepatology 24:533.
3988
ADHESION MOLECULES IN HUMAN MALIGNANT LIVER TUMORS
21. Berlin, C., E. L. Berg, M. J. Briskin, D. P. Andrew, P. J. Kilshaw, B. Holzmann, I. L. Weissman, A. Hamman, and E. C. Butcher. 1993. Alpha 4 beta 7 integrin mediates binding to the mucosal vascular addressin MAdCAM-1. Cell 74:185. 22. Schweighoffer, T., Y. Tanaka, M. Tidswell, D. J. Erle, K. J. Horgan, G. E. Luce, A. I. Lazarovits, D. Buck, and S. Shaw. 1993. Selective expression of integrin a4b7 on a subset of human CD41 memory T cells with hallmarks of gut-trophism. J. Immunol. 151:717. 23. McNab, G., J. L. Reeves, M. Salmi, S. G. Hubscher, S. Jalkanen, and D. H. Adams. 1996. Vascular adhesion protein-1 supports adhesion of T lymphocytes to hepatic endothelium: a mechanism for T cell recirculation to the liver? Gastroenterology 110:522. 24. Henkart, P. A. 1985. Mechanism of lymphocyte-mediated cytotoxicity. Annu. Rev. Immunol. 3:31. 25. Liu, D. L., C. H. Hakansson, and J. Seifert. 1994. Immunotherapy in liver-tumors. II. Intratumoral injection with activated tumor-infiltrating lymphocytes, intrasplenic administration of recombinant interleukin-2 and interferon-alpha causes tumor-regression and lysis. Cancer Lett. 85:39. 26. Shimizu, Y., S. Iwatsuki, R. B. Herberman, and T. L. Whiteside. 1990. Clonal analysis of tumour infiltrating lymphocytes from human primary and metastatic liver tumours. Int. J. Cancer 46:878. 27. Adams, D. H., P. Burra, S. G. Hubscher, E. Elias, and W. Newman. 1994. Endothelial activation and circulating vascular adhesion molecules in alcoholic liver disease. Hepatology 19:588. 28. Yoong, K. F., and D. H. Adams. 1997. Interleukin-2 restores CD3 zeta chain expression but fails to generate specific cytolytic activity in tumor infiltrating lymphocytes from patients with hepatic colorectal metastases. Br. J. Cancer. In press. 29. McNab, G., S. C. Afford, C. M. Morland, A. J. Strain, R. Joplin, and D. H. Adams. 1997. Cultured human hepatic sinusoidal endothelial cells express and secrete adhesion molecules and chemokines that are important for leukocyte recruitment to the liver. In Cells of the Hepatic Sinusoid, E. Wisse, D. L. Knook, and C. Balabaud, eds. The Kupffer Cell Foundation, Leiden, p. 123. 30. Morland, C. M., J. Fear, G. McNab, R. Joplin, and D. H. Adams. 1997. Promotion of leukocyte transendothelial cell migration by chemokines derived from human biliary epithelial cells in vitro. Proc. Assoc. Am. Physicians 109:372. 31. Knook, D. L., C. Barkway, and E. C. Sleyster. 1981. Lysosomal enzyme content of Kupffer and endothelial liver-cells isolated from germ-free and clean conventional rats. Infect. Immun. 33:620. 32. Adams, D. H., J. Yannelli, W. Newman, S. A. Rosenberg, and S. Shaw. 1997. Adhesion of tumour-infiltrating lymphocytes to endothelium: a functional and phenotypic analysis. Br. J. Cancer 75:1421. 33. Salmi, M., K. Granfors, M. Leirisalo-Repo, M. Hamalainen, R. MacDermott, R. Leino, T. Havia, and S. Jalkanen. 1992. Selective endothelial binding of interleukin-2-dependent human T-cell lines derived from different tissues. Proc. Natl. Acad. Sci. USA 89:11436. 34. Salmi, M., and S. Jalkanen. 1992. A 90-kilodalton endothelial cell molecule mediating lymphocyte binding in humans. Science 257:1407. 35. Shimizu, Y., W. Newman, T. V. Gopal, K. J. Horgan, N. Graber, L. D. Beall, G. A. van Seventer, and S. Shaw. 1991. Four molecular pathways of T cell adhesion to endothelial cells: roles of LFA-1, VCAM-1 and ELAM-1 and changes in pathway hierarchy under different activation conditions. J. Cell Biol. 113:1203. 36. Salmi, M., S. Tohka, E. L. Berg, E. C. Butcher, and S. Jalkanen. 1997. Vascular adhesion protein 1 mediates lymphocyte subtype-specific, selectin-independent recognition of vascular endothelium in human lymph nodes. J. Exp. Med. 186: 598. 37. Salmi, M., and S. Jalkanen. 1996. Human vascular adhesion protein-1 (vap-1) is a unique sialoglycoprotein that mediates carbohydrate-dependent binding of lymphocytes to endothelial-cells. J. Exp. Med. 183:569. 38. Spertini, O., F. W. Luscinskas, G. S. Kansas, J. M. Munro, J. D. Griffin, M. A. Gimbrone, Jr., and T. F. Tedder. 1991. Leukocyte adhesion molecule-1 (LAM-1, L-selectin) interacts with an inducible endothelial cell ligand to support leukocyte adhesion. J. Immunol. 147:2565. 39. Takagi, S., K. Chen, R. Schwarz, S. Iwatsuki, R. B. Herberman, and T. L. Whiteside. 1989. Functional and phenotypic analysis of tumor infiltrating
40.
41.
42. 43. 44.
45. 46.
47.
48.
49.
50. 51.
52.
53.
54.
55.
56.
57.
58.
lymphocytes isolated from human primary and metastatic liver tumours and cultured in recombinant interleukin-2. Cancer 63:102. Mackensen, A., G. Carcelain, S. Viel, M. C. Raynal, H. Michalaki, F. Triebel, J. Bosq, and T. Hercend. 1994. Direct evidence to support the immunosurveillance concept in a human regressive melanoma. J. Clin. Invest. 93:1397. Straten, P., J. C. Becker, T. Seremet, E. B. Brocker, and J. Zeuthen. 1996. Clonal T-cell responses in tumor-infiltrating lymphocytes from both regressive and progressive regions of primary human-malignant melanoma. J. Clin. Invest. 98:279. Whiteside, T. L., and R. B. Herberman. 1992. Extravasation of antitumor effector cells. Invasion Metastasis 12:128. Volpes, R., J. J. van den Oord, and V. J. Desmet. 1992. Vascular adhesion molecules in acute and chronic liver disease. Hepatology 15:269. Steinhoff, G., M. Behrend, B. Schrader, A. M. Duijvestijn, and K. Wonigeit. 1993. Expression patterns of leukocyte adhesion ligand molecules on human liver endothelia: lack of ELAM-1 and CD62 inducibility on sinusoidal endothelia and distinct distribution of VCAM-1, ICAM-1, ICAM-2 and LFA-3. Am. J. Pathol. 142:481. Adams, D. H. 1996. Lymphocyte-endothelial interactions in hepatic inflammation. Hepato-Gastroenterology 43:32. Friedman, S. L., J. S. Flier, F. Epstein, R. Glickman, and G. Scheele. 1993. Seminars in medicine of the Beth-Israel-Hospital, Boston: the cellular basis of hepatic-fibrosis: mechanisms and treatment strategies. N. Engl. J. Med. 328:1828. Kantele, J. M., H. Arvilommi, S. Kontiainen, M. Salmi, S. Jalkanen, E. Savilahti, M. Westerholm, and A. Kantele. 1996. Mucosally activated circulating human B-cells in diarrhea express homing receptors directing them back to the gut. Gastroenterology 110:1061. Wong, J., B. Johnston, S. S. Lee, D. C. Bullard, C. W. Smith, A. L. Beaudet, and P. Kubes. 1997. A minimal role for selectins in the recruitment of leukocytes into the inflamed liver microvasculature. J. Clin. Invest. 99:2782. Piali, L., A. Fichtel, H. Terpe, B. A. Imhof, and R. H. Gisler. 1995. Endothelial VCAM-1 expression is suppressed by melanoma and carcinoma. J. Exp. Med. 181:811. Salmi, M., K. Kalimo, and S. Jalkanen. 1993. Induction and function of vascular adhesion protein 1 at sites of inflammation. J. Exp. Med. 178:2255. Arvilommi, A. M., M. Salmi, K. Kalimo, and S. Jalkanen. 1996. Lymphocyte binding to vascular endothelium in inflamed skin revisited: a central role for vascular adhesion protein-1 (vap-1). Eur J. Immunol 26:825. Mise, M., S. Arii, H. Higashituji, M. Furutani, M. Niwano, T. Harada, S. Ishigami, Y. Toda, H. Nakayama, M. Kukumoto, J. Fujita, and M. Imamura. 1996. Clinical significance of vascular endothelial cell growth factor and basic fibroblast growth factor gene expression in liver tumor. Hepatology 23:455. Gamble, J. R., Y. Khew-Goodall, and M. A. Vadas. 1993. Transforming growth factor-b inhibits E-selectin expression on human endothelial cells. J. Immunol. 150:4494. Bargatze, R. F., M. A. Jutila, and E. C. Butcher. 1995. Distinct roles of L-selectin and integrins alpha-4-beta-7 and LFA-1 in lymphocyte homing to Peyers patchHEV in-situ: the multistep model confirmed and refined. Immunity 3:99. Cepek, K. L., S. K. Shaw, C. M. Parker, G. J. Russell, J. S. Morrow, D. L. Rimm, and M. B. Brenner. 1994. Adhesion between epithelial-cells and T-lymphocytes mediated by E-cadherin and the alpha(e)beta(7) integrin. Nature 372:190. DeVries, J. E., W. N. M. Dinjens, G. K. Debruyne, H. W. Verspaget, E. P. M. Vanderlinden, A. P. Debruine, M. M. Mareel, F. T. Bosman, and J. Tenkate. 1995. In-vivo and in-vitro invasion in relation to phenotypic characteristics of human colorectal-carcinoma cells. Br. J. Cancer 71:271. Gagliardi, G., O. Kandemir, D. Liu, M. Guida, S. Benvestito, T. G. M. Ruers, I. S. Benjamin, J. M. A. Northover, G. W. H. Stamp, I. C. Talbot, and M. Pignatelli. 1995. Changes in E-cadherin immunoreactivity in the adenomacarcinoma sequence of the large-bowel. Virchows Arch. A Pathol. Anat. Histopathol. 426:149. Pockaj, B. A., R. M. Sherry, J. P. Wei, J. R. Yannelli, C. S. Carter, S. F. Leitman, J. A. Carasquillo, S. M. Steinberg, S. A. Rosenberg, and J. C. Yang. 1994. Localization of indium-111-labeled tumor-infiltrating lymphocytes to tumor in patients receiving adoptive immunotherapy. Cancer 73:1731.
