Tumor counterattack: fact or fiction?

Abstract

Cancer development relies on a variety of mechanisms that facilitate tumor growth despite the presence of a functioning immune system. Understanding these mechanisms may foster novel therapeutic approaches for oncology and organ transplantation. By expression of the apoptosis-inducing protein CD95L (FasL, APO-1L, CD178), tumors may eliminate tumor-infiltrating lymphocytes and suppress anti-tumor immune responses, a phenomenon called “tumor counterattack”. On the one hand, preliminary evidence of tumor counterattack in human tumors exists, and CD95L expression can prevent T-cell responses in vitro. On the other hand, CD95L-expressing tumors are rapidly rejected and induce inflammation in mice. Here, we summarize and discuss the consequences of CD95L expression of tumor cells and its contribution to immune escape.

Introduction

Tumors employ multiple mechanisms to escape from immune-mediated rejection. These mechanisms facilitate cancer development in immune-competent hosts and prevent sucessful immunotherapy. Detailed understanding of these events on a cellular and molecular level may result in therapeutic strategies to overcome immune escape and may also be exploited to establish tolerance to transplanted organs. Immune-escape mechanisms of tumors range from impaired antigen presentation and apoptosis resistance to active strategies such as the production of immunosuppressive factors [41]. Moreover, tumors may adopt a killing mechanism from cytotoxic immune cells to eliminate the attacking anti-tumor lymphocytes—a concept called “tumor counterattack” [37, 48]. Thus, many tumors have been found to express the apoptosis-inducing cell surface molecule CD95L (CD178, FasL, APO-1L), enabling them to kill T cells that are sensitive to CD95-mediated cell death. In vitro CD95L expression can prevent the expansion of cytotoxic T cells in an antigen-specific manner. However, expression of CD95L on tumor cells or grafts does not confer immune privilege in vivo, but, instead, induces inflammation and accelerates rejection. Hence, presently it is not entirely clear whether tumor counterattack really is a relevant immune escape mechanism of tumors [70, 76].

The CD95/CD95L system

CD95 (APO-1, Fas) and CD95L are transmembrane proteins of the tumor necrosis factor (TNF), family of receptors and ligands, respectively. CD95 belongs to the subfamily of death receptors characterised by an intracellular domain, the death domain [49, 80]. Binding of CD95L to CD95 leads to oligomerization of receptors and recruitment of FADD/MORT-1 and the proform of the “initiator” caspase 8 (Fig. 1). These molecules together form a protein complex called the death-inducing signalling complex (DISC). In the DISC, procaspase-8 is cleaved autocatalytically and then activates further effector caspases directly or via a mitochondrial pathway. Active effector caspases cleave cellular substrates—e.g. structural proteins of the cell (actin, plectin, fodrin, etc.,), signalling proteins (cPLA2, Akt-1, PKCδ,etc.,) and ICAD, the inhibitor of the endonuclease CAD—leading to the biochemical and morphological changes characteristic of apoptosis. The CD95 signalling pathway is regulated by many pro- and anti-apoptotic factors (Fig. 1). A major class of regulatory proteins are the members of the Bcl-2 family which regulate apoptosis at the mitochondrial level [62, 98]. According to their function, Bcl-2 family members can be divided into anti-apoptotic and pro-apoptotic proteins. In addition, FLIPs (FLICE-inhibitory proteins) interfere with the initiation of apoptosis directly at the level of death receptors [50], and IAPs bind to and inhibit caspases [19].

Fig. 1.

Apoptosis signalling via CD95 (simplified). Binding of CD95L leads to the formation of the death-inducing signalling complex (DISC). In the DISC, the initiator caspase-8 is activated. Cleavage of Bid by caspase-8 activates the mitochondrial pathway and can be used to amplify the apoptotic signal. Activation of mitochondria leads to the formation of the apoptosome. At the apoptosome, caspase-9 is activated. Caspase-8 and −9 activate executioner caspases which in turn cleave the death substrates, eventually resulting in apoptosis. Apoptosis can be inhibited on different levels by anti-apoptotic proteins (shown in green): CD95L can be prevented from binding to CD95 by soluble CD95 (sCD95). FLICE-inhibitory proteins (FLIPs) bind to the DISC and prevent the activation of caspase-8; and inhibitors of apoptosis proteins (IAPs) bind to and inhibit caspases

CD95 is expressed in many tissues, whereas expression of the apoptosis-inducing ligand CD95L is restricted to a few cell types and is regulated on the transcriptional level by various transcription factors [58]. After an increase in the intracellular calcium concentration, NFAT proteins are dephosphorylated, bind to the CD95L promotor and thus induce CD95L. Activation of mitogen-activated protein kinases (MAPK), for example by changes in the redox status of cells, can lead to the activation of AP-1 transcription factors which then induce CD95L. Moreover, early growth response gene (Egr) 2 and 3, NFκB and the constitutive transcription factor SP-1 contribute to CD95L expression.

Three different forms of CD95L protein have been described: First, membraneous CD95L on the cell surface is the primary mediator of apoptosis by oligomerizing CD95 receptors on a target cell. Second, the membraneous form is stored in intracellular microvesicles which are secreted into the intercellular space in reponse to various physiological stimuli [12]. Third, a soluble form of CD95L is generated by cleavage of the membraneous protein by matrix metalloproteases [61]. Human soluble CD95L can be cytotoxic and its cytotoxicity is enhanced by binding to the extracellular matrix [3]. By contrast, murine soluble CD95L is not cytotoxic, but blocks CD95L-mediated apoptosis by competing with the membrane-bound form for CD95 binding. This anti-apoptotic function has also been described for human soluble CD95L.

Role of CD95L in the immune system

Apoptosis, triggered via the CD95 system, plays a fundamental role in the regulation of T-cell homeostasis in the periphery [49]. Challenge with antigen first leads to clonal expansion of T cells, followed by a down-phase of the immune response several days later, in which most antigen-specific T cells are eliminated in a process called activation-induced cell death (AICD) [20]. Whereas T cells are resistant to CD95-mediated apoptosis during the expansion phase, in the down-phase they become CD95-sensitive [47, 81]. T-cell activation leads to expression of CD95L, resulting in “suicide” or “fratricide” of activated CD95-sensitive T cells. CD95-dependent apoptosis contributes to peripheral tolerance by eliminating autoreactive T cells in the periphery [51, 54]. Humans or mice with deleterious mutations of CD95 (lpr mutation) or CD95L (gld mutation) display a phenotype of accumulation of aberrant T cells, leading to lymphadenopathy, splenomegaly and an autoimmune disease that closely resembles systemic lupus erythematosus [95].

CD95L may also be involved in the maintenance of immune privilege. In immune privileged sites, such as eye and testis, deleterious immune reactions are prevented by reduced MHC expression, secretion of immunosuppressive cytokines and physical barriers. Thus, in the eye, a high percentage of human corneal transplants is accepted without tissue matching or immunosuppressive therapy. CD95L is expressed in epithelial cells of the eye and in sertoli cells of the testes [25, 31, 91]. In a mouse model, for corneal allograft transplantation, CD95L-positive corneal grafts were accepted at a significant rate, whereas all grafts from mice with a mutated, nonfunctional CD95L (gld) were rejected. CD95L-positive grafts transplanted to CD95-negative (lpr) mice were also not accepted. CD95L-positive grafts contained apoptotic mononuclear cells after transplantation [91]. Moreover, inflammatory cells entering the anterior chamber of the eye in response to viral infection underwent apoptosis dependent on CD95/CD95L interactions and did not produce any tissue damage. By contrast, viral infection in gld mice, which lack functional CD95L, resulted in the inflammation and invasion of ocular tissue by cells without signs of apoptosis [31]. CD95-mediated apoptosis of lymphoid cells was necessary for tolerance induction following antigen injection into the anterior chamber of the eye [29, 32]. The proposed mechanism for the induction of immune deviation after antigen presentation in the eye involves the production of the antiinflammatory cytokine IL-10 by the apoptotic cells [29]. Similarly, murine testis grafts expressing wild type CD95L survived indefinitely when transplanted under the kidney capsule of allogeneic animals, whereas testis grafts from gld mice were rejected [9]. However, attempts to reproduce these observations by another group were not successful [2]. The role of CD95L for maternal tolerance to a semiallogeneic fetus during pregnancy is also controversial. In one model, CD95L was proposed to prevent leukocyte trafficking between mother and conceptus, although in a different mouse model disruption of the CD95/CD95L system had no adverse effect on the outcome of pregnancy [77]. So far, no fertility defects in humans or mice with a defective CD95 system have been observed. In addition, another death receptor/ligand system, the TNF-related apoptosis-inducing ligand (TRAIL)/TRAIL-R system, has been suggested to be involved in the establishment of placental immune privilege [73]. Since neurons and astrocytes also express CD95L [17, 25, 78], immune privilege in the central nervous system may also involve CD95L. However, this has been contested since lpr and gld mice did not show an exacerbated experimental allergic encephalomyelitis (EAE), a model of multiple sclerosis, but instead, showed ameliorated clinical signs of EAE [79]. The development of a Th1 response or inflammatory cell infiltration into the central nervous system was not affected.

CD95L expression of tumors

Human tumor cells of diverse origin have been reported to express CD95L. These tumors include hepatocellular carcinoma, lymphoma, melanoma, astrocytoma, esophageal carcinoma, gastric adenocarcinoma and others [10, 11, 30, 33, 34, 60, 78, 89, 94]. Because of unspecific staining of certain CD95L antibodies, some reports have to be interpreted with caution [76, 90]. Moreover, it has been reported that CD95L is not expressed by the tumor cells themselves, but by infiltrated T cells and monocytes [53]. However, in many papers CD95L⁺ tumor cells were also shown to kill CD95-positive apoptosis-sensitive cells in vitro, reliably demonstrating functional CD95L expression.

Stable CD95L expression can only be observed in apoptosis-resistant tumors, otherwise the tumor cells are eradicated by suicide or fratricide. Indeed, many tumor cells are resistant to apoptosis [40]. Apoptosis resistance is an important feature of cancer development, because it contributes to independence from physiological growth control mechanisms, to escape from the immune system and to drug resistance. Tumor cells can acquire resistance to apoptosis by the expression of anti-apoptotic proteins or by the downregulation or mutation of pro-apoptotic proteins. Alterations of the p53 pathway also influence the sensitivity of tumor cells to apoptosis. Moreover, most tumors are independent of survival signals because they have upregulated the phosphatidylinositol 3-kinase (PI3K)/AKT pathway.

Cancer treatment by chemotherapy and γ-irradiation kills target cells primarily by the induction of apoptosis [88]. One of the best-defined mechanisms by which therapy-induced cellular stress eventually leads to the death of tumor cells involves the CD95 system [23, 27, 28, 66, 67]. Chemotherapeutic drugs, such as the nucleotide analogue 5-fluoruracil, induce CD95 by a transcriptionally regulated, p53-dependent mechanism, particularly in liver tumor cells. They also engage the stress-activated protein kinase (SAPK, also known as JUN-N-terminal kinase or JNK) pathway. SAPKs, which are members of the MAPK family, can regulate the activity of AP-1 transcription factors, which eventually leads to the upregulation of CD95L. Moreover, oxidative stress, triggered by the production of reactive oxygen intermediates and glutathione depletion, can also induce CD95L expression. Upregulation of CD95 and CD95L then allows the cells to either commit suicide or kill neighbouring cells. However, treatment of apoptosis-resistant tumor cells can upregulate CD95L expression without inducing tumor cell death. Thus, chemotherapy of these tumors does not only fail, but instead may enable the tumor to counterattack the immune system [48].

T-cell elimination by CD95L-expressing tumor cells in vitro

To characterize the effect of CD95L expression of tumor cells on tumor-specific T cells, the tumor counterattack situation was analysed in vitro. Whereas naive and short-term activated T cells were resistant to CD95-mediated apoptosis, activated T cells in culture for more than 4 days became sensitive to apoptosis induction by CD95L⁺ tumor cells or dendritic cells (DCs) transfected with CD95L [13, 39].

Furthermore, T cells were stimulated in vitro with CD95L positive or negative cells (Fig. 2). CD95L expression of a murine tumor with a specific tumor antigen prevented the expansion of cytotoxic antitumor T cells in mixed lymphocyte tumor reactions [39]. Human T cells activated by CD95L⁺ allogeneic tumor cells were not cytotoxic against the respective CD95L⁻ tumor cells [22]. In adition, CD95L-transduced DCs down-regulated T-cell proliferation in allogeneic mixed lymphocyte cultures [13]. Moreover, a novel method of cell membrane modification has been reported providing an alternative for gene transfer approaches [97]. Using this method, cells were “decorated” with functional CD95L by biotinylation of the membrane and subsequent incubation with a chimeric protein consisting of streptavidin and the extracellular portion of CD95L. Such CD95L-“decorated” splenocytes blocked primary and secondary alloreactive responses in vitro. In all these systems, T cells may first be activated by the tumor or DC leading to sensitization of the T cells to CD95-mediated apoptosis. The sensitive T cells may then be killed by CD95L⁺ tumor cells. Alternatively, the concurrent signals via the TCR and CD95 may directly inhibit activation and proliferation of the T cells.

Fig. 2.

Simulation of tumor counterattack in vitro. Upper panel In vitro stimulation of T cells with CD95L negative tumor or DCs expressing an antigen (Ag) that can be recognized by the T cells leads to T cell proliferation. These T cells can kill target cells expressing the antigen, irrespective of CD95L expression of the target cell. Lower Panel In vitro stimulation of T cells with CD95L positive cells does not lead to expansion of the T cells, although the stimulator cells express an antigen that can be recognized by the T cells. Remaining T cells are not cytotoxic towards target cells expressing the respective antigen. Symbols: light circle T cell, dark circle tumor cell, blue T cell receptor, green tumor antigen, red CD95L

Interestingly, although CD95L⁺ tumor cells could kill activated T cells, CD95L expression of the tumor had no significant influence on the cytotoxicity of T cells against tumor cells in vitro (Fig. 2). T cells that were stimulated with alloantigen or tumor antigen lysed CD95L⁺ and CD95L⁻ tumor cells in vitro to the same extent [39, 56]. Alloantigen stimulated human T cells lysed CD95L⁺ and CD95L⁻ tumor cells in a 4-h assay in a similar fashion [22]. In general, in cytotoxicity assays in vitro high T cell/tumor cell ratios are used. In vivo the ratio is skewed towards high tumor cell numbers. In human solid tumors only around 4 % of tumor infiltrating cells are cytotoxic T cells [56]. Hence, the outcome of the T cell-tumor cell interaction in vivo may not directly be inferred from these in vitro findings.

Induction of tolerance by CD95L in vivo

Immunohistological analysis of sections from human CD95L⁺ tumors suggested that counterattack was also operative in vivo. Apoptosis of tumor infiltrating lymphocytes (TILs) has been found in situ within CD95L-expressing human melanoma, hepatocellular carcinoma, gastric adenocarcinoma and lymphoepithelioma-like cancer of the stomach [10, 34, 52, 89]. In certain tumors, such as esophageal cancer, the number of TILs was reduced concomitantly with increased TIL apoptosis within CD95L-expressing areas of the tumors [11]. These immunohistological analyses represent correlations, but do not provide functional proof. Therefore, various animal models have been used to demonstrate the ability of CD95L expressed on tumors to downregulate anti-tumor immune responses. Tumor growth of a subcutaneously injected CD95L-positive murine melanoma cell line was slightly faster in wild-type or gld mice than in mice with a mutated or downregulated CD95 receptor (lpr mice) [34]. However, the observation period was rather short (8 days), and the difference in tumor incidence was only 1–2 days. The rat histiocytoma AK-5 transiently upregulated CD95L when injected intraperitoneally into syngeneic rats. This coincided with depletion of the intraperitoneal NK cell population [46]. Another study in syngeneic mice showed that growth of tumors of murine CD95L-transfected cells was significantly better than that of control cells when implanted under the kidney capsule [68]. Immunosuppression in vivo was directly demonstrated in allogeneic mice injected with CD95L-transfected colon carcinoma cells. Alloantibodies and allospecific cytotoxic and helper T lymphocytes were reduced [4].

These findings stimulated the notion that CD95L may be used to render a transplanted tissue an immune privileged site that is not rejeted by the host’s immune system. Indeed, it was reported that cotransplantation of syngeneic myoblasts genetically engineered to express CD95L protected allogeneic islets of Langerhans grafts from immune rejection. Graft survival was not prolonged with composite grafts of unmodified myoblasts [55]. However, other authors have found exactly opposite results using a similar approach [44]. On the other hand, rat livers transfected with a CD95L-encoding plasmid were transplanted to allogeneic recipients and showed significantly prolonged survival times compared to control grafts. The CD95L-transfected liver allografts caused apoptotic cell death in infiltrating T cells [57].

Interestingly, DC transfected with CD95L, the so-called ‘killer DCs’, can induce antigen-specific immunosuppression in vivo. In this way, peptide-specific DTH reactions [63] and virally induced chronic inflammatory responses [99] were suppressed by the killer DCs loaded with the respective antigens. Even unresponsiveness to allogeneic grafts was achieved [64, 100]. Moreover, immunization with cells that were decorated with the Streptavidin-CD95L chimeric protein mentioned above blocked alloreactive responses in naive and presensitized rodents and prevented the rejection of allogeneic pancreatic islets [97]. In addition, the vasculature of murine hearts was decorated with CD95L [6]. Transplantation of these hearts into allogeneic murine hosts resulted in delayed allograft rejection as compared with control allografts. Allograft survival was further extended by imunization with CD95L-decorated donor cells.

Induction of inflammation by CD95L in vivo

A number of reports contradicting the tumor counterattack hypothesis have also been published. In contrast to the above findings, it has been reported that CD95L expression by grafts or by tumor cells led to rejection, neutrophil infiltration and induction of immunity. CD95L expression on pancreatic islets transplanted into allogeneic hosts resulted in islet destruction and in a massive infiltration of neutrophils. Similarly, transgenic mice expressing CD95L in pancreatic β-cells developed an infiltration of neutrophils and diabetes [45]. CD95L expressing islet β-cells transplanted under the kidney capsule of syngeneic or allogeneic animals were not protected from rejection [2, 42]. In addition, CD95L-expressing hearts from transgenic mice transplanted into sygneneic and allogeneic recipients were more rapidly rejected than control grafts and showed neutrophil infiltration as early as one day after transplantation [93]. In contrast to other reports, it has also been published that injection of CD95L⁺ DCs into mice primed the recipients to a weak antigen and caused neutrophil infiltration [13].

Furthermore, overexpression of CD95L in murine tumor cells resistant to CD95-mediated apoptosis did not affect growth in vitro, but resulted in delayed tumor formation or tumor rejection in vivo [8, 38, 71, 83, 84, 96]. CD8⁺ T cell-mediated protective immunity against subsequent challenge with the parental tumor cells was elicited. CD95L expression also led to antibody-dependent tumor immunity in a murine tumor model with weak immunogenicity [86]. Rejection of the CD95L-expressing tumor has even been observed in the study mentioned above demonstrating immunosuppression by the tumor in an allogeneic mouse model [4]. In addition, a “bystander rejection” of CD95L-negative cells was found, when control tumor cells were co-implanted with the CD95L-expressing cells at the same site [83, 84]. Bystander rejection was also observed when a mixture of fibroblasts engineered to express CD95L and tumor cells was implanted into syngeneic mice [21]. Moreover, in mice that were injected with CD95L-negative tumor cells, followed by treatment with CD95L-positive tumor cells one or three days later, tumor growth was delayed compared to control mice or totally abolished [84].

In most animal experiments, CD95L-transfected tumor cell lines have been used. These cells may express different, probably higher levels of CD95L than naturally occuring tumors. So, it has been suggested that overexpression of CD95L may lead to rejection by neutrophils, whereas physiological levels may not induce neutrophilic infiltration but may still suffice to eliminate anti-tumor lymphocytes. However, CD95L expression levels in human tumors have not been quantified and compared so far. Moreover, cell lines expressing different levels of CD95L constitutively grew more slowly in mice than CD95L⁻ cells and a cell line in which different levels of CD95L could be induced via the tet-system was rejected, irrespective of the level of expression [38]. Therefore, the rejection of tumors in mice is not an artefact of CD95L overexpression. In some animal experiments, slowly growing CD95L⁺ tumor cells had lost their cytotoxic activity when analysed ex vivo. Thus, the initial CD95L expression of the tumor cells may result in rejection or retarded growth of the tumor. After selection of CD95L⁻ cells, a tumor can outgrow [38].

Tumors growing in situ may express CD95L at late stages of tumorigenesis, e.g., after induction by chemotherapy [23, 26, 66]. Moreover, the sensitivity of T cells to CD95-mediated apoptosis varies considerably with respect to the activation status of the T cells [49]. Thus, naïve and freshly activated T cells are resistant to CD95-mediated apoptosis, whereas several days after activation, T cells become sensitive to CD95L. Therefore, it was speculated that the time-point of CD95L expression could directly influence tumor counterattack. However, induction of CD95L expression at later time-points after transplant in established tumors also led to tumor rejection similarly to constitutive expression [38]. In addition, infection of a subcutaneously growing CD95-negative tumor by an adenoviral vector encoding CD95L resulted in rapid elimination of the tumor [5].

Taken together, the majority of in vivo experiments in which CD95L is expressed in a tumor show that CD95L expression is unfavourable for tumor cells and that CD95L has an immunostimulatory function in mice. Moreover, some early papers containing evidence initially thought to support the tumor counterattack hypothesis have now been withdrawn or refuted [75].

Mechanisms of tumor rejection and induction of inflammation

Several mechanisms for induction of inflammation and the recruitment of neutrophils to CD95L-expressing grafts or tumors have been proposed. Two studies suggested that soluble CD95L is directly chemotactic for neutrophils implicating that soluble CD95L promotes rejection of CD95L-expressing grafts [72, 82]. However, others have not found a chemotactic activity of soluble CD95L [8, 85]. Moreover, tumor cells expressing only soluble CD95L did not elicit a neutrophilic response [36, 85]. On the other hand, tumor cells expressing a non-cleavable membrane-bound form of CD95L were rapidly rejected [36, 43, 85]. The extent of inflammation induced by the various transfectants seemed to correlate with the cytotoxic activity of CD95L.

Alternatively, many studies indicate an indirect mechanism for the recruitment of neutrophils by CD95L (Fig. 3). CD95L may act on resident or surrounding cells to induce the production of granulocyte chemoattractants or cell death. Thus, CD95L induces the processing and release of IL-1β that may be responsible for the infiltration by neutrophils [65]. CD95L may act on resident macrophages leading to increased production of IL-1β and macrophage inflammatory proteins [35]. Moreover, engagement of CD95 on DCs may induce the secretion of proinflammatory cytokines [74]. Furthermore, CD95L has been shown to induce DC maturation [86] and may also augment tumor-DC interactions [92], thus initiating a specific immune response. Induction of apoptosis in resident or surrounding cells may also induce a local inflammatory response by activating endothelial cells via oxidized phospholipids on the apoptotic cells [14]. The accessibility of human tumors growing in situ and transplant tumors in mice to infiltration by these cells mediating the proinflammatory signal may be different. Thus, it has been shown that the stroma of human tumors may serve as a physical barrier between tumor and immune cells [69, 87]. Moreover, co-expression or injection of the immunosuppressive cytokine TGF-β at the tumor site protected CD95L⁺ tumors against rejection [15]. In summary, these data support a proinflammatory function of CD95L and an indirect mechanism for neutrophil recruitment.

Fig. 3.

Potential mechanisms of tumor rejection and induction of inflammation (simplified). a A CD95L⁺ tumor cell might induce death of a neighbouring tumor cell that has been sensitized to CD95-mediated apoptosis. b CD95L⁺ tumor cells might also interact with other cells at the tumor site, such as macrophages (MΦ) or DCs and induce cytokine secretion or cell death. c These events lead to the infiltration of neutrophils into the tumor. d Neutrophils might then directly kill tumor cells. For a more detailed description of potential mechanisms please refer to the text. Symbols: yellow CD95, other symbols see legend of Fig. 2

The exact mechanism how CD95L-expressing tumors are rejected is still unclear. In most studies, neutrophils were made responsible for eradicating CD95L⁺ tumors [8, 38, 71, 83, 84, 96]. Thus, neutrophil infiltration has been observed in most of the rejected grafts. In addition, depletion of neutrophils prevented graft rejection in some studies. Experiments using immunodeficient mice showed that the rejection is not dependent on a functional adaptive immune system. Moreover, depletion of NK cells or macrophages did not affect tumor rejection [83]. However, in another study neutrophil depletion did not have any effect on tumor growth [39]. Furthermore, in mice with neutrophils deficient for important cytotoxicity mechanisms, tumors grew similarly as in wild type mice. It has also been published that CD95L expression on tumor cells mediates inactivation of neutrophils [16]. Therefore, a direct cytotoxic effect of neutrophils against the tumor seems unlikely.

CD95L⁺ tumors may also be destroyed by other mechanisms leading to subsequent infiltration by neutrophils that then phagocytose the dying cells. The most probable alternative rejection mechanism is suicide or fratricide of tumor cells. Although in most studies tumor cells were used that were resistant to CD95-mediated apoptosis in vitro, killing of neighbouring cells by CD95L⁺ tumor cells in vivo cannot entirely be excluded. Cells growing in vivo may have closer cell-cell contacts, so that a reduced CD95-sensitivity may still suffice for killing. In addition, it has been shown that some tumor cells were sensitized for CD95-mediated apoptosis in vivo [56]. However, CD95L⁺ tumors were even rejected without change of CD95 expression or sensitivity in vivo [38, 39]. Moreover, in several reports, apoptotic cells were immunosuppressive or even induced tolerance [24, 29, 32]. Further alternative mechanisms that may be responsible for reduced growth of CD95L⁺ tumors in mice might include differences in stroma, vascularisation or clonogenicity.

Conclusions

Research in tumor immunology has provided a wealth of information about the interactions between tumors and the immune system. Many of these interactions are now known on a cellular and molecular level. Surprisingly, despite considerable efforts from many laboratories, it has not been clarified convincingly yet, whether CD95L-expressing tumors can eliminate anti-tumor lymphocytes, escape the immune response and thus have a growth advantage in vivo. It rather turned out that the initial idea of tumor counterattack was over-simplified.

In the meantime, different functions have been ascribed to CD95L. It is not only solely an apoptosis-inducing molecule, but might also have pro-inflammatory effects, yield costimulatory signals to T cells [1, 59], induce motility of tumor cells [7], contribute to liver regeneration [18] and deliver growth stimulatory signals to neurons [18]. This situation is possibly comparable to TNF that was first recognized as a tumor-death-inducing agent and later turned out to be a central mediator of the immune system. The various functions of CD95L might lead to an overlay of multiple effects of CD95L expression in vivo which probably explains, for example, the strange result that tumor rejection is observed despite immunosuppression [4]. In addition, presence and function of soluble CD95L has not been critically evaluated in each study. Such factors contributed to the finding that the consequences of CD95L expression in vivo are highly dependent on the experimental system. Furthermore, technical problems, particularly unspecific staining of certain CD95L antibodies, also led to controversial results [76, 90].

Nevertheless, the complexity of the tumor counterattack problem should not prevent further investigations, because a better understanding of the events may be rewarded with important novel therapeutic strategies to overcome immune escape and to establish tolerance in organ transplantation.