Current views concerning the influences of murine hepatic endothelial adhesive and cytotoxic properties on interactions between metastatic tumor cells and the liver

Since the "seed and soil" theory proposed by Stephen Paget, there has been a long history of research into the reasons for organ-specific cancer metastasis [17, 18]. The liver is a common site for metastasis of human cancer and a convenient target for experimental studies of metastasis. From the latter it is apparent that endothelial cell surface adhesion molecules have an extensive role in regulating cancer cell site-specific arrest, transendothelial migration and metastases formation [3, 19‐23].

Expression of various hepatic endothelial adhesion molecules has been demonstrated to be selectively inducible by cytokines, bacterial lipopolysaccharide (LPS) or arresting tumor cells in the liver microvascular bed. In turn, these adhesion molecules can be shown to regulate the arrest of circulating cancer cells in the hepatic sinusoids. For example, interleukin-1α (IL-1α) pretreatment of mice altered the melanoma cell (B16F1) arrest pattern from 32 μm beyond the sinusoidal inlet to larger terminal portal venules (TPV) observed by intravital videomicroscopy, suggesting increased adhesive interactions between endothelial and tumor cells following IL-1α stimulation [24]. Interleukin-18 (IL-18) has been demonstrated in vivo and in vitro to promote liver metastasis by enhancing melanoma cell adhesion to the hepatic sinusoidal endothelial cells via microvascular VCAM-1 (vascular cell adhesion molecule-1) expression [25‐27]. With a basal expression level of ICAM-1 (intercellular adhesion molecule-1), minimal expression of VCAM-1 and no expression of E-selectin or αv integrin in unstimulated mouse livers, 1 μg/g body weight of LPS i.p. selectively induced the expression of ICAM-1 (4–48 h), VCAM-1 (4–24 h) and E-selectin (2 h) on the sinusoidal lining cell surface, while αv integrin expression was unchanged [28, 29]. LPS did not significantly alter the expression of VLA-4 (very late antigen-4, counter receptor of VCAM-1) or LFA-1 (leukocyte functional antigen-1, counter receptor of ICAM-1) on melanoma cells either in vivo or in vitro [30]. Tumor necrosis factor (TNF) induced sustained VCAM-1 expression within 4 h in the lung, liver and kidney of mice [31], and increased the adhesion of highly metastatic murine carcinoma cell line H-59, and human colorectal carcinoma lines HM 7 and CX-1 to murine hepatic endothelial cells in the primary culture. This effect was completely abolished by a monoclonal antibody to murine E-selectin [21]. Mannose receptor-mediated endothelial cell activation also contributed to B16 melanoma cell adhesion and metastasis in the mouse liver [32].

The expression of sinusoidal adhesion molecules is affected by metastatic cells in the hepatic microenvironment. The arrest of B16F1 melanoma cells in the liver sinusoids (following mesenteric vein injection) induced focal expression of VCAM-1 and more diffuse expression of ICAM-1 around the melanoma cell arrest sites [30]. Similarly, the arrest of murine carcinoma line H-59 cells after intrasplenic injection induced E-selectin expression on the hepatic sinusoidal endothelium between 2–24 h [33]. The expression of ICAM-1, VCAM-1, E-selectin and αv integrin was all induced to different degrees by the growth of melanoma tumors in the peritoneal cavity without liver metastasis in the mouse [28]. In a study on progression of mouse melanoma (B16-BL6) spontaneous metastasis, organ specific induction of VCAM-1 was observed in the cardiac, hepatic and cerebral vascular beds 4 weeks following the resection of primary tumors when metastatic pulmonary burden was maximal [34]. Intrasplenically injected B16 melanoma (B16M) cells also increased the expression of VCAM-1 significantly on hepatic sinusoidal endothelial cells within the first 24 h, which correlated with the increased in vitro adhesion of B16M cells to hepatic sinusoidal endothelial cells isolated from B16M cell-injected mice [35].

The mechanisms and significance of the selectivity of adhesion molecule induction have not been fully described at this stage. However, the inducibility of various adhesion molecules on the hepatic endothelial cell surface by different microenvironmental stimuli has provided a potential diversity and flexibility for sinusoidal endothelial cells to participate in tumor defensive responses when intravascular metastatic cancer cells are present.

The micro-structural and functional heterogeneity in the liver across its functional unit of acinar zonation has been well-described in the literature [36‐39]. This hepatic zonal heterogeneity has also played a significant role influencing the patterns of induced adhesion molecule expression. Differential zonal expressions of certain adhesion molecules induced by LPS stimulation have been demonstrated [28]. With a weak expression around the terminal portal venule regions (acinar zone 1) under basal conditions, ICAM-1 was induced to a uniform strong expression (4–48 h) across each entire acinus in the liver following LPS administration. On the contrary, VCAM-1 and E-selectin both had minimal or no expression in unstimulated livers, but had significantly stronger expression in acinar zone 1 than zone 2 and 3 after LPS stimulation, with VCAM-1 expressed between 4–48 h and E-selectin 2–12 h [28]. LPS stimulation also increased the retention of B16F1 melanoma cells in the liver between 8–24 h, especially in the terminal portal venule region presumably through increased expression of adhesion molecules, ICAM-1 and VCAM-1 [30]. IL-1 zonal heterogeneity of mannose receptor-mediated ligand endocytosis in the mouse and rat liver was also observed using flow cytometry following LPS stimulation [40, 41]. In human studies, major differences have been noted in the composition of the portal tract and sinusoid with regard to endothelial and parenchymal cell expression of cell-cell and cell-matrix adhesion molecules during inflammatory reactions in human liver grafts [42]. Differential expression of various adhesion molecules has been reported between normal and inflamed livers, or livers rejected after transplantation in humans. The selectins ELAM-1 (endothelial leukocyte adhesion molecule) and CD62 (cluster of differentiation 62) were basally expressed and inducible on portal tract endothelia and central vein endothelia with acute and chronic human liver inflammation, although sinusoidal endothelia lack this mechanism even with severe inflammation [43]. Portal and sinusoidal endothelia showed a different expression and inducibility of VCAM-1, ICAM-1, ICAM-2, and LFA-3 (leukocyte functional antigen-3) in human livers [43].

In addition to the impact of hepatic zonal heterogeneity on adhesion molecule expression, alterations in the liver sinusoidal architecture also significantly change the endothelial cell surface molecule expression and tumor cell behavior patterns. Using a murine liver cirrhosis model, where the sinusoidal lumens were narrowed due to the formation of fibrous tissue, the expression of adhesion molecules ICAM-1 and VCAM-1 was found to be significantly increased (stronger in acinar zone 1) on the endothelial surface with E-selectin undetectable [44]. After injecting melanoma cells into the portal vein, melanoma cell retention in the cirrhotic liver terminal portal venule regions was also significantly increased in comparison with the control livers [44].

Direct and indirect evidence from the literature has supported the hypothesis that the hepatic sinusoidal microvasculature is toxic to metastatic tumor cells. Various experimental data obtained to date have indicated that the hepatic endothelium exerts its antitumor defensive effects through the release of NO and other reactive oxygen species (ROS) [4, 10, 15, 16, 45‐47]. As with adhesion molecule expression, the cytotoxic regulatory functions in the liver have also been demonstrated to be inducible by microenvironmental stimuli.

The original evidence that hepatic endothelium-derived NO is induced by intravascular metastatic tumor cells was obtained by applying electron paramagnetic resonance (EPR) NO-spin trapping technologies into a classic murine melanoma metastatic model [15, 48, 49]. By injecting fluorescent microsphere-labeled B16F1 melanoma cells into the portal circulation of C57BL/6 mice, a swift burst of NO was detected in liver samples within 5 minutes of cell injection. NO induced apoptosis in 20–30 % of the melanoma cells arresting in the liver after 4 h [15]. NO was identified and its cytotoxicity to melanoma cells was supported by finding that the nonselective NO synthase inhibitor L-NAME (N_G-nitro-L-arginine methyl ester) blocked NO production and melanoma cell apoptosis in the sinusoids. The ability of a short burst of NO to cause apoptosis was further confirmed by detecting the apoptotic DNA fragmentation and cell membrane damage in B16F1 cells exposed to a NO donor for 5 min in vitro [15]. The mechanisms of tumor cell specific induction of NO release at the site of cell arrest have not yet been identified but are suggested to be partly due to tumor cell-induced vascular wall shear stress with circumferential stretch and isometric contraction (Lower levels of NO are released following injection of inert microspheres with similar diameters to melanoma cells) [15, 50, 51]. Using an in situ liver perfusion system, the cellular origin of the NO release following B16F1 cell arrest in the liver has been identified as periportal endothelial and sinusoidal lining cells, and hepatocytes adjacent to the arresting B16F1 cells [52]. The endothelial and sinusoidal lining cells released NO in an eNOS (endothelial NO synthase)-dependent manner over a time of 500 sec, and hepatocytes over a longer period of time measured by fluorescent 4,5-diaminofluorescein diacetate (DAF-2 DA) used as the NO detection probe [52]. In addition to the immediate burst of endothelial eNOS-dependent NO production upon melanoma cell arrest in the liver, a delayed iNOS (inducible NO synthase)-dependent cytotoxic NO induction after 4 h of cell injection into the mesenteric vein has also been demonstrated, which was partially due to the shear forces generated by melanoma cell arrest in the sinusoids, and produced from both sinusoidal lining cells and hepatocytes detected by double-labeling immunohistochemistry [15] (Figure 1: A – C).

This evidence has been supported by findings from Umansky et al. [4, 16] and Rocha et al. [53]. Using a well-characterized ESbL-lacZ mouse T lymphoma model, the authors have shown that a significant increase in NO production detected in vitro from ex vivo isolated liver endothelial cells and Kupffer cells coincided with the plateau phase (tumor retardation phase) of primary tumor growth and a low level of liver metastasis. They have also demonstrated that the activated host liver endothelial cells play dual roles in metastatic processes by expressing adhesion molecules and producing NO from iNOS activation [4, 16, 53].

Edmiston et al. have shown that unstimulated murine sinusoidal endothelial cells produced ROS that were selectively toxic to weakly metastatic human colorectal carcinoma clone A cells, with the toxicity blockable by pretreatment with NO synthase inhibitor, superoxide dismutase or dexamethasone [54]. Coculture of ischemic liver fragments with human colorectal carcinoma cells killed more weakly metastatic clone A cells at 24 h than highly metastatic CX-1 cells because of the higher sensitivity to NO and ROS in clone A cells [47, 55]. NO also induced apoptosis in different human neoplastic lymphoid cells and breast cancer cell lines through caspase activation pathways [46, 56].

In addition to NO, other cytotoxic ROSs are released from the liver endothelium and also possess antitumor cytotoxicity. In vitro, sinusoidal endothelial cells release hydrogen peroxide (H₂O₂) which enhanced VLA-4 mediated melanoma cell adherence to the hepatic sinusoidal endothelium and caused tumor cytotoxicity after IL-1 treatment in mice [57, 58]. Superoxide anion (O₂^-) may be involved in the cytotoxicity of murine hepatic sinusoidal endothelial cells to weakly metastatic human colorectal carcinoma cells [47, 54, 55]. The important interplays between NO and other ROSs, such as O₂^-, in cancer development and progression have been reviewed [45]. The rapid death of most cancer cells after delivery to some target organs has also been demonstrated to be a consequence of their mechanical interactions within the microvasculature [12].

The accumulated evidence to date has directed us to recognize the existence of a host natural defensive mechanism network in the hepatic microvasculature through the production of NO and other ROSs from the sinusoidal endothelium to generate cytotoxicity to invading intravascular tumor cells to fight against cancer metastasis in the liver.

Similar to the inducible adhesion molecule expression under the influence of hepatic zonal heterogeneity, evidence suggests that the release of NO from the hepatic endothelium is restricted to specific anatomical zones. Using an in situ C57BL/6 mouse liver perfusion system, the levels of NO production without and with tumor cells in the liver were found to be much greater in acinar zone 1 than zone 2 and 3 by direct visualization of NO synthesis through deesterification and conversion of intracellular DAF-2 DA to DAF-2T [52]. In cirrhotic mouse livers with altered sinusoidal architecture, significantly lower levels of NO production were detected both under basal conditions (without tumor cells) and after tumor cell arrest by the same experimental system [44].

The detection of a host defensive mechanism existing in the hepatic endothelium has raised the question of whether similar defense mechanisms also exist in other metastatic target organs. Direct in vitro lysis of metastatic tumor cells by cytokine-activated murine lung vascular endothelial cells has been demonstrated. NO (detected by nitrite concentration in the culture medium) produced by interferon gamma and TNF-activated lung vascular endothelial cells played a major role in the lytic destruction of reticulum cell sarcoma [59, 60]. Rapid death of transformed metastatic rat embryo cells, occurred via apoptosis in the lungs 24–48 h after injection into the circulation of immune-deficient nu/nu mice, has been reported [14, 61].

Using EPR NO-spin trapping technologies, a significantly increased production of NO was detected in lung tissue samples between 20 min and 4 h after the tail vein injection of fluorescent microsphere-labeled B16F1 melanoma cells [49]. The EPR results were also supported in an isolated, ventilated and blood-free mouse lung perfusion model, where NO production in situ was observed in real time using intact organ microscopy techniques. Fluorescent NO signals (DAF-2T) increased rapidly at the site of tumor cell arrest in the lungs and continued to increase throughout 20 min thereafter [49, 62]. NO contributed to tumor cell apoptosis since 3-fold more B16F1 cells subsequently underwent apoptosis in the lungs of wild-type mice compared to animals in which NO production was inhibited, in particular, in eNOS-deficient mice and NOS inhibitor L-NAME-pretreated mice [49].

The identification of a similar antitumor defensive mechanism in the pulmonary microvascular bed has reinforced the concept that the host can release NO and other ROSs as cytotoxic effector molecules to fight against the invading metastatic tumor cells in the microvascular beds of the first-line cancer metastatic organs, such as the liver and lung.

The majority of reports indicate that the underlying molecular mechanisms for NO-induced tumor cell cytotoxicity are direct damage to DNA and the cell membrane, or activation of apoptosis-initiating caspases (cysteine proteases) causing tumor cell apoptosis and necrosis [10, 14, 15, 45‐47, 49, 54, 56, 59, 60]. In addition, preliminary evidence also suggests that NO may induce oxidative damage on proteins through NO-superoxide-peroxynitrite and NO-nitrogen dioxide-nitrite pathways to form nitrotyrosine. The latter is the footprint of potent short-lived reactive nitrogen species, peroxynitrite (ONOO^-) production in vivo, mediating NO-induced oxidative attacks on biological macromolecules [63‐65] (Figure 1: D–G). More observations need to be made to provide supportive evidence along this direction.

The importance of interplays between NO and adhesion molecules in the regulation of liver cancer metastasis has been recognized and addressed in recent years [10, 19, 20, 45]. The inducible murine hepatic microvascular adhesive and cytotoxic regulatory functions have been regulated by using LPS [28, 30]. With enhanced local expression of VCAM-1 and ICAM-1 around B16F1 cell arrest sites in the liver, LPS significantly increased the retention of melanoma cells in the liver, especially in the terminal portal venule regions between 8 and 24 h after intramesenteric injection of melanoma cells [30]. LPS also significantly increased the levels of iNOS expression and tumor cell induced-NO production at 8 h after administration and cell injection, and increased the rates of B16F1 cell apoptosis in the terminal portal venule region [30]. These data have been interpreted to indicate that LPS stimulated a synergistic interaction by inducing both the hepatic endothelial adhesion molecule expression and NO release in the terminal portal venular regions, resulting in higher levels of tumor cell killing in this region in the liver [30]. The dual roles of activated host liver endothelial cells in murine lymphoma metastatic process have also been reviewed [16]. On one hand, upregulation of the expression of particular adhesion molecules is considered to lead to the increased tumor cell binding and stimulation of angiogenesis, and on the other hand, endothelial cells can contribute to host anti-metastatic responses by producing the cytotoxic molecule NO from arginine with the help of iNOS [16]. Synergistic interactions between LFA-1/ICAM-1 and lymphoma progression phases with cytotoxic NO production have been described [4]. Interactions between cytokine IL-18, VCAM-1, H₂O₂ and hepatic sinusoidal endothelial cells have also been demonstrated [35]. Recombinant catalase administered in vivo completely blocked the increase of VCAM-1 expression induced by B16M cell arrest in the liver, and blocked in vitro B16M cell adhesion to sinusoidal lining cells isolated from B16M cell-injected mice [35]. Incubation of hepatic endothelial cells with nontoxic concentrations of H₂O₂ directly enhanced VCAM-1-dependent B16M cell adhesion in vitro without proinflammatory cytokine mediation [35].

In addition to synergistic interactions between NO and adhesion molecules, their counteractive interactions have also been identified. NO reduces tumor cell adhesion to isolated rat postcapillary venules in vitro [66]. Anti-adhesive roles of constitutively produced NO in inhibiting leukocyte rolling and adhesion in the microcirculation have been described [67, 68]. Oxidative stress in the liver can be caused by ischemia/reperfusion (I/R) injury when tumor cells entering the hepatic microcirculation obstruct hepatic sinusoids and temporarily occlude blood flow before the hepatic circulation is reestablished by either tumor cell death or invasion into the parenchyma [8, 55]. The counteractive roles of NO with adhesion molecules, such as decreasing P-selectin and ICAM-1 mRNA expression, attenuating neutrophil accumulation and liver damage in hepatic ischemia/reperfusion injury have been reviewed [69‐71]. IL-10 has also been shown to inhibit hepatic I/R injury by inhibiting the upregulation of iNOS expression following I/R injury [55]. The multifaceted roles and effects of NO and adhesion molecule interactions support the scenario that the host uses this flexible natural defensive mechanism to protect itself from a variety of disastrous oxidative injuries and tissue damages to the hepatic microvasculature.