Tumor-induced immune dysfunction: the macrophage connection

Tumor-induced immune dysfunction: the macrophage connection Klaus D. Elgert, David G. Alleva, and David W. Mullins Department of Biology, Microbiology and Immunology Section, Virginia Polytechnic Institute and State University, Blacksburg, Virginia Abstract: Although macrophages (M s) mediate tumor cytotoxicity, display tumor-associated antigens, and stimulate antitumor lymphocytes, cancer cells routinely circumvent these host-mediated immune activities, rendering the host incapable of mounting a successful antitumor immune response. Evidence supporting a direct causal relationship between cancer and immune dysfunction suggests that the presence of neoplastic tissue leads to immunologic degeneration. Furthermore, substantial data demonstrate that tumor growth adversely alters M function and phenotype. Thus, although M s can serve as both positive and negative mediators of the immune system, the importance of M s in tumor-induced immune suppression remains controversial. This review focuses on the evidence that tumor-derived molecules redirect M activities to promote tumor development. Tumors produce cytokines, growth factors, chemotactic molecules, and proteases that influence M functions. Many tumor-derived molecules, such as IL-4, IL-6, IL-10, MDF, TGF- 1, PGE 2, and M-CSF, deactivate or suppress the cytotoxic activity of activated M s. Evidence that tumor-derived molecules modulate M cytotoxicity and induce M suppressor activity is presented. This information further suggests that M s in different in vivo compartments may be differentially regulated by tumor-derived molecules, which may deactivate tumor-proximal (in situ) M populations while concurrently activating tumordistal M s, imparting a twofold insult to the host s antitumor immune response. J. Leukoc. Biol. 64: 275 290; 1998. Key Words: cytokines growth factors chemotactic molecules proteases INTRODUCTION The observation that tumor growth activates immune defense mechanisms, yet neoplastic tissue eludes these mechanisms, defines the paradox of tumor immunology. Our studies [1 8] show a direct causal relationship between tumor presence and immune dysfunction, suggesting that the proximity of neoplastic tissue leads to immunologic degeneration. Although the importance of macrophages (M s) in tumor-induced immune dysfunction is controversial, recent research suggests that the cellular basis for suppression includes the generation of immunoregulatory M s that inhibit T cell responses and that are tumoricidally dysfunctional. Various investigators have established that M s can serve as both positive and negative mediators of the immune system [9 14] (Fig. 1). As positive effector cells, M s mediate direct antitumor cytotoxicity or presentation of tumor-associated antigens (TAAs), which are effective strategies for the eradication of tumors [15 19]. In contrast, tumor-induced M s demonstrate tumor cell growth-promoting abilities [20 23] that aid in tumor growth, and these M s suppress many T cell [7, 24, 25] and natural killer (NK) cell [26] antitumor responses. This dualism begs the question: what tips the balance toward one or the other M function during tumor assault. Therefore, an understanding of the regulatory mechanisms that control M functions during tumor growth is critical to directing the balance between health and disease. Whether M s are serving a positive or a negative function, M -derived cytokines and proinflammatory substances are critical regulators of M activities. The discovery that tumors also produce pro- and anti-inflammatory cytokines and other regulatory molecules provides clues as to how tumors may subvert M antitumor functions to favor tumor growth. To further our understanding of this rapidly evolving field, we will focus on the evidence that suggests tumor-derived molecules can redirect M activities to promote tumor development. M s PROMOTE TUMOR GROWTH Although M s were first investigated for their role in antitumor immune responses, later studies revealed a dark side to Abbreviations: CSFs, colony-stimulating factors; ECM, extracellular matrix proteins; GM-CSF, granulocyte-macrophage colony-stimulating factor; hnrnp, heterogeneous nuclear ribonucleoprotein; inos, inducible nitric oxide synthase; IFN, interferon; IL, interleukin; M, macrophage; M-CSF, M colonystimulating factor; MCP-1, M chemotactic protein-1; MDF, M deactivating factor; MSV, murine sarcoma virus; NO, nitric oxide; PGE 2, prostaglandin E 2 ; ROI, reactive oxygen intermediates; TGF, transforming growth factor; TAAs, tumor-associated antigens; TAMs, tumor-associated M s; TNF, tumor necrosis factor. Correspondence: Dr. Klaus D. Elgert, Department of Biology, Microbiology and Immunology Section, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061-0406. E-mail: kdelgert@vt.edu Current address of David G. Alleva: Department of Medicine, The University Hospital, Immunology Unit-E337, Boston University Medical Center, Boston, MA 02118-2393. Received February 19, 1998; revised April 22, 1998; accepted April 23, 1998. Journal of Leukocyte Biology Volume 64, September 1998 275

progression [reviewed in ref. 22]. Furthermore, M presence correlates with tumor growth and metastasis [34]; without M s to stimulate the growth and development of tumor-nourishing blood vessels, solid tumors rapidly die [22]. One M -derived molecule that promotes blood vessel growth in normal tissue but destroys blood vessels within solid tumor masses is tumor necrosis factor- (TNF- ) [35]. However, TNF- production by tumor-associated M s (TAMs) is inhibited [36], suggesting that M -derived angiogenic molecules other than TNF- affect blood vessel development in tumors [22]. M s also may promote tumor growth through involvement in the metastatic process [37]. Formation of aggregates between human breast carcinoma tumor cells and M s, in which the M s partly or completely surround the tumor cells, could impart to the tumor cells many of the properties necessary for tissue invasion, a normal M function. Furthermore, such an association would provide the tumor cells with M -derived growth factors and camouflaging from other cytotoxic immune effector cells. Fig. 1. The duality of macrophage function during tumor growth. Depending on their in vivo context, M s can serve either as positive (green lines) or negative (red lines) effector functions during tumor growth. As positive effector cells, M s can mediate direct antitumor cytotoxicity and presentation of TAA. In contrast, tumor-derived signals (dashed line) induce negative effector functions, including the production of tumor growth factors and suppression of lymphocyte responsiveness. tumor-bearing host (TBH) M s; cancer actually induces M functions that promote tumor growth [20 22, 27]. This finding revealed the dual nature of M s, which depending on their in vivo context can impart diametrically opposed activities. For example, although vital effectors of innate immunity, M s may contribute up to or more than half of a tumor s mass [22] and are actually required for the tumor to survive [20 22, 28]. To ensure survival, tumors actively recruit monocytes by producing chemotactic agents, including transforming growth factor- (TGF- ) [29] and M chemotactic protein-1 (MCP-1) [30], which promote M infiltration [21, 30]. The seemingly paradoxical activity by tumors to promote M infiltration actually benefits the tumor through a number of mechanisms. For example, M s produce growth factors such as L-arginine-derived polyamines [13, 31]. L-Arginine is the substrate molecule for M biosynthesis of the cytotoxic molecule nitric oxide (NO) through the activity of the inducible form of nitric oxide synthase (inos). However, some tumors markedly limit intratumoral M production of NO [32] by shunting L-arginine metabolism to favor the biosynthesis of ornithine [13], a precursor for polyamine tumor growth factors required for cell replication. Using this mechanism, tumors cause M s to bypass NO synthesis, thus decreasing M cytotoxic activity while concurrently increasing M growthpromoting activity at the tumor site. In addition to promoting tumor growth, M s are potent promoters of angiogenesis [22, 33]. Through the production of growth factors and cytokines [including granulocyte-macrophage colony-stimulating factor (GM-CSF), TGF- and -, interleukin (IL)-1, IL-6, IL-8, and prostaglandins], activated M s can potentially influence every phase of angiogenic TUMORS SUPPRESS M ANTITUMOR ACTIVITY Tumors use a variety of mechanisms to evade detection and destruction by the immune system, including the release of cytokines and effector molecules and alterations in Fas-FasL interactions normally associated with immune cells [38]. Tumor cells release elevated levels of inhibitory cytokines that upset the normal balance of the immune system, leading to altered M function and immunosuppression [5, 39 41]. The resulting tumor cell-derived cytokine expression and cellular response leads to many changes in immune function, including suppression of host antitumor immune responses. Table 1 provides a partial list of known tumor-derived molecules, including cytokines, prostaglandins, growth factors, chemotactic molecules, and TAAs, and describes their actions on M cytotoxic and suppressor activities. Not included in this table are certain tumor-derived proteases [31, 98], which also affect M activities. The location, phenotype, developmental stage, and activation state of the M determine the variable effects of these tumor-derived molecules. Tumor cells control M production of effector molecules Tumors produce substances that both down-regulate [7, 62, 99 102] and up-regulate [90, 91, 103, 104] M cytotoxic and effector molecule production. These M -derived cytotoxic molecules include TNF- [16, 17, 35, 105], NO [106, 107], H 2 O 2 [108], reactive oxygen intermediates (ROI) [16, 109, 110], IL- 1 [17], and specific proteases [31, 96]. Tumors can stimulate M s to produce cytotoxic molecules through tumor soluble or membrane-bound TAAs [16, 90, 91, 111], extracellular matrix proteins (ECM) [95], or receptor-mediated binding of Fc portions of antibody [16, 112, 113] attached to tumor cells [114, 115]. Although M production of TNF-, NO, and ROI imparts cytotoxic and suppressor activities, tumor growth also increases M production of the noncytotoxic suppressor mol- 276 Journal of Leukocyte Biology Volume 64, September 1998

TABLE 1. Tumor-derived molecules Tumor-Derived Molecules Modulate M Activities Resting M s Effect on M -mediated: Cytotoxicity a Activated M s Suppression b References IL-4 > < ND c [42 50] IL-6 > < ND [48, 51 55] IL-10 ND < < [7, 46, 48, 56 60] MDF ND < ND [61, 62] TGF- > < > [7, 8, 48, 63 69] M-CSF > < > [70 77] GM-CSF > >< > [71, 72, 76, 78 83] PGE 2 < < ND d [7, 84 87] p15e ND < ND [21, 30] MCP-1 ND ND ND [21, 30, 48, 88] TAAs > > > [89 94] ECM > > ND [95 97] a M -mediated cytotoxicity denotes both M cytostasis or cytolysis to a target tumor cell mediated by secretion of cytotoxic molecules. Arrows denote the ability of tumor-derived molecules to stimulate (>) resting M s for cytotoxicity or to up-regulate (>) or down-regulate (<) activated M cytotoxic activity. b M -mediated suppression refers to the ability of M s to decrease alloantigen-, autoantigen-, or mitogen-stimulated lymphocyte proliferation or activity. Arrows denote the ability of tumor-derived molecules to increase (>)or decrease (<)M -mediated suppression. c ND, not determined. d Whether tumor-derived PGE 2 can directly affect M -mediated suppression of T cell reactivity has not been determined; however, indirect data discussed throughout strongly suggest that it should. ecules PGE 2, TGF- 1, and IL-10. The dual in vitro cytotoxic and suppressor functions of TNF-, NO, and ROI are misleading when considering the in vivo functions of these molecules in tumor-burdened animals. The in vivo existence of wellestablished tumors suggests that production of TNF-, NO, and ROI does not necessarily lead to tumoricidal activity, but may instead lead to suppression of antitumor lymphocytes. TBH M s outwardly appear to be unactivated, but they are actually primed for tumoricidal activity [21, 116, 117] and production of cytokines [5, 118, 119], NO [7, 120], and prostaglandins [5, 118, 119, 121, 122]. TBH M s are considered primed because they constitutively express mrna for TNF- [5, 36, 118] and other cytokines. After activation, these TBH M s demonstrate superior capacity to kill tumors and to produce larger amounts of factors, as compared with normal host M s [16, 17, 116]. Tumor-derived molecules differentially regulate M functions in different compartments Despite an onslaught of infiltrating M s into a tumor mass and the propensity of tumors to activate M s, tumor cells escape M antitumor activities by subverting M functions to minimize antitumor effector function and favor tumor progression. Through the production of cytokines and effector molecules, tumors use the immune system s own communications network to undermine the host s antitumor responses. Depending on the M s resident tissue and proximity to the neoplasm, tumorderived cytokines can impart differential effects in various in vivo compartments (Fig. 2) by either the priming of resting M s or the suppression of activated M cytotoxic molecule production. For example, mrna expression of several cytotoxic and suppressor molecules is increased in tumor-distal M populations, including splenic or peritoneal M s [7]. Our current studies suggest that tumor-derived factors induce increased translocation of the multifunctional transcription factor NF- B in splenic M populations, leading to expression of various cytotoxic effector molecules [McConnell and Elgert, unpublished observations]. Simultaneously in the same host, TAM cytotoxic molecule production is largely abrogated [32, 123]. Tumors circumvent TAM-mediated cytotoxicity By differentially controlling expression of M -derived molecules in different in vivo locations, tumors exploit the suppressor activity of cytotoxic effectors such TNF-, NO, and ROI while escaping their cytotoxicity. Tumor-derived TGF-, IL-4, IL-6, M colony-stimulating factor (M-CSF), and GM- CSF (see Table 1) may activate tumor-peripheral resting M s and down-regulate in situ activated M s (see Fig. 2). Circulating tumor-derived cytokines may prime resting tumor-distal M s to produce cytotoxic and suppressor molecules such as TNF-, NO, H 2 O 2, and PGE 2 [7, 118 120, 124, 125]. During migration to the tumor site, tumor-primed M s encounter increasing concentrations of activation signals such as TAAs and disrupted ECM proteins and become activated. However, as the activated M s enter the tumor microenvironment, tumor-derived cytokines inhibit M production of cytotoxic molecules. In addition, these activation molecules may convert the M -stimulatory action of tumor-derived cytokines to a M -deactivating action. Therefore, tumor growth downregulates cytotoxic molecule production as monocytes migrate to the tumor site. In vitro, tumor-derived molecules induce M production of noncytotoxic suppressor molecules such as PGE 2 [8], suggesting that tumor-proximal M s could remain suppressive, but not cytotoxic. Although in vitro investigations have established that M s can kill tumor cells while leaving normal cells unharmed [16, 126, 127], M -derived cytotoxic molecules may not be effective during in vivo tumor growth [16, 127 130]. This finding suggests that tumors somehow interfere with the activity or production of cytotoxic molecules, leading to the reduction of M cytotoxic activity [13, 17, 101, 102]. Many tumor-derived molecules, such as IL-4, IL-6, IL-10, M -deactivating factor (MDF), TGF-, PGE 2, M-CSF, and p15e, deactivate or suppress activated-m cytotoxic activity. Although TBH M s normally are primed for enhanced cytotoxicity, tumor supernatants suppress activated M tumoricidal activity [5, 7, 40, 90, 91, 103, 104, 116 118, 131, 132], and tumor-derived cytokines and chemotactic molecules fail to stimulate M TNF- and NO production [35, 133]. In our own tumor model, lipopolysaccharide activation of isolated TAMs failed to induce NO production in vitro [Mullins and Elgert, unpublished observations]. Furthermore, many tumors (especially spontaneously arising neoplasms) have mechanisms to resist toxicity from one or more M -derived cytotoxic molecules [16, 115, 127]. The simultaneous action of several M -derived molecules is therefore required for lysis of many spontaneous tumors, suggesting that Elgert et al. Tumor-induced immune dysfunction: the macrophage connection 277

Fig. 2. Macrophage activities vary with compartment during tumor growth. Depending on the M s resident tissue and proximity to the neoplasm, tumor-derived cytokines may either suppress (red octagon) activated M cytotoxic molecule production (top) or prime resting M s for enhanced release of proinflammatory mediators (bottom). Tumor-proximal M s are rendered incapable of producing cytotoxic molecules (i.e., tumors are not killed) but still suppress lymphocyte function through PGE 2 production. Tumor-distal M s, in contrast, produce cytotoxic molecules that fail to impart antitumor activity but further suppress lymphocyte function. Dashed lines represent suppressed or abrogated production of the indicated factor. the inhibition of one type of cytotoxic molecule may be sufficient for the tumor to escape lysis [115]. Tumors activate distal M populations After cell-cell contact, certain tumor cell membrane constituents directly induce M TNF- [90, 111] and NO [91] production, and circulating tumor-cell membrane debris may activate distal M s. Tumor supernatants suppress M TNF- and NO production [7], but also induce M s to suppress lymphocyte proliferation, partly by stimulating PGE 2 production [8]. Membrane preparations of these tumor cell cultures stimulate production of cytotoxic TNF- and NO [Alleva and Elgert, unpublished observations]. The secretion of monocyte chemotactic substances [21, 30], along with the generation of soluble ECM proteins by neoplasms [96], may activate tumordistal M s [95, 96]. M binding of these proteins to adhesion receptors may induce monocyte and M TNF- secretion [95, 134]. Tumor growth may trigger tumor-distal M TNF- production and other suppressor molecules by ECM protein binding to M s. Some tumor-derived cytokines may enhance the effects of ECM proteins by increasing expression of tumor-distal M adhesion molecules that bind soluble ECM proteins [97]. TUMORS INDUCE M SUPPRESSOR ACTIVITIES During tumor growth, M s can adversely affect host antitumor responses and mediate lymphocyte suppression (Fig. 3) [70, 79, 89, 121, 135 137]. In spite of considerable research into M -mediated immune suppression, controversy still exists about the nature and function of suppressor M s in immunedysfunctional cancer hosts. Although the complex interactions of immune cells initially obscured the true identity of the cell that was causing or mediating the suppressive phenomenon, direct evidence that M s suppressed the immune reactivity during tumor growth was reported as early as 1974. Splenic adherent cells from murine sarcoma virus (MSV)-induced hosts suppressed mitogen- and viral antigen-induced blastogenesis of normal lymphocytes [138, 139]. Even more convincing evidence was provided by follow-up studies with MSV tumor- 278 Journal of Leukocyte Biology Volume 64, September 1998

probably are the main PGE 2 -producing subpopulation of M s [118, 119]. Increased numbers and the suppressor activity of TBH class II M s down-regulate TAA presentation by class II M s to T cells in the spleen [155, 156] and perhaps in other compartments [18]. Fig. 3. Tumor-induced suppression of T cell responsiveness. T cell responsiveness to various stimuli, including mitogens, antigens, autoantigens, and alloantigens, was assessed by measuring tritiated thymidine incorporation as compared to controls. Representative data are shown. To determine whether tumor growth compromised T cell reactivity, splenic CD4 T cells were isolated from either 10 or 21 1-day TBHs or normal hosts and activated with various stimuli. Tumors were induced by the injection of 4 10 5 transplanted Meth-KDE fibrosarcoma cells, as described [1]. Data are percent responsiveness of TBH T cell cultures compared with similarly treated normal host T cell cultures. To determine the cell type responsible for suppressing T cell reactivity, admixture experiments were performed in which normal host CD4 T cells were cultured with activating agents in the presence of either purified TBH CD4 T cells or TBH M s. Data are percent responsiveness of normal host T cells in admix cultures compared with purified, activated normal host T cell cultures receiving their normal host equivalents. bearing athymic mice, where suppression was linked to cells with M -like characteristics [140], and the same group [141] showed M -like cells with suppressor functions in a rat tumor model. Others [142] showed that M -like suppressor cells from tumor-bearing mice were capable of blocking in vitro immunization against transplantation antigens. Soon thereafter, M s were implicated as suppressor cells in a transplantable fibrosarcoma model [143] and rat splenic M s were shown to inhibit lymphoproliferation to TAAs and lymphoma cell lines [144]. Subsequently, evidence accumulated to suggest that more than one suppressor cell type could operate during tumor growth. Our investigations into the dualistic cellular [1, 2] and molecular [145, 146] mechanisms of tumor-induced suppression were immediately corroborated [147], and this M aspect is discussed throughout. Tumor growth affects immune cell regulation M s can up- or down-regulate lymphocyte and other immune cell responses [148]. For example, antitumor lymphocytes can be stimulated by TAA presentation, although the survival of tumors may depend on a decline in TAA presentation to antitumor T cells [18]. At the same time, tumor-induced M s suppress many T cell [1, 7, 63, 70, 87, 110, 121, 136, 149 153] and NK cell [26, 121] responses. Altered antigen presentation by TBH M s may further subvert antitumor host responses through altered antigen presentation [94]. During tumor growth, the numbers of TBH splenic suppressor M s, which are usually MHC class II, are increased [122, 154]. PGE 2 is the main inhibitory molecule produced by class II M s during tumor growth [118, 119, 122], and these M s Tumor-derived cytokines inhibit T H 1-type immune responses Cell-mediated responses tend to be optimal in the eradication of tumors [16, 157]. Tumor-derived IL-4, IL-10, TGF- [38, 64], and PGE 2 [158] suppress the necessary cell-mediated responses supported by T H 1-type cells during cancer, while supporting the largely ineffective humoral responses maintained by T H 2-type cells. Tumors may create environments that inhibit M cytotoxicity by either directly suppressing it or by inhibiting T H 1-type cells that induce M tumoricidal actions. Tumor-derived IL-10 blocks tumor-induced M -mediated suppression of lymphocyte proliferation [40], unlike TGF- 1 and other tumor-derived cytokines. Therefore, tumor-derived IL-10 blocks M cytotoxicity at the tumor site [7, 57, 159] and does not induce M suppressor activity. At the same time, IL-10 can directly suppress proliferation of alloantigen- or mitogenstimulated T H cells [40, 57]. Studies show that IL-10 preferentially down-regulates T H 1-type cell activity by inhibiting M accessory (costimulatory) activity, which is necessary for T H 1- type cell activation [57]. T H 1-type cells promote M activation by producing interferon- (IFN- ), whereas T H 2-type cells suppress M accessory and cytotoxic activities by producing IL-4 and IL-10 [159]. IL-10 hinders M -mediated suppressor activity by decreasing M NO, TNF- [40], and PGE 2 production [160] and may also reduce IFN- and IL-12 production, which is unfavorable for the generation of T H 1-type cells [161]. This suggests that M -derived IL-10 may play a significant role in controlling T cell and M activities during tumor-induced immunosuppression. Tumor-induced M s suppress lymphoproliferation through production of proinflammatory and cytotoxic cytokines and factors M s suppress lymphocyte and NK cell functions mainly by secreting the inhibitory molecules lipocortin [162], PGE 2 [7, 40, 118, 119, 163, 164], NO [7, 8, 164 166], H 2 O 2 [108], TGF- [63], IL-10 [167], and the ROIs [168]. Although these molecules directly restrain T cell proliferation, only the inhibition of lipocortin, PGE 2, ROI, and NO production blocks M -mediated suppression in the TBH. The increased M output of autoinhibitory TGF- 1 and IL-10 during tumor growth controls the production of other M suppressor molecules; PGE 2, NO, TNF- [8, 40], and enhancing molecules such as GM-CSF [6] are triggered through autocrine induction. For example, normal host M s slightly curb T cell proliferation to maintain homeostasis, and antibody-mediated neutralization of TGF- 1 and IL-10 relieves normal host M -mediated suppression of T cell proliferation [8, 40]. In contrast, TBH M s remain strongly suppressive even after TGF- 1 and IL-10 neutralization because their output of other M -derived suppressor molecules is increased. Activated, but not resting, M s synthesize TGF- 1 and Elgert et al. Tumor-induced immune dysfunction: the macrophage connection 279

IL-10 [40, 169], allowing these molecules to autocrinely control the production of other suppressor molecules released by activated M s. When tumor growth dysregulates M suppressor molecule production, increased concentrations of TGF- 1 and IL-10 cannot control M suppressor activities [8, 40]. Activated tumor-proximal M cytotoxic activity is downregulated by tumor- or M -derived TGF- 1 and IL-10 (and perhaps PGE 2 ), but tumor-proximal M s retain their suppressor activity. In addition, tumor growth enhances M production of TGF- 1, IL-10, and PGE 2 [8, 40, 63], cytokines that strongly inhibit T lymphocyte proliferation [46, 57, 64, 170]. Although levels are boosted, the cytokines have pleiotropic effects during immunosuppression and can down-regulate tumor-induced M suppressor molecule production [8, 40]. Perhaps the immune system, through feedback mechanisms, is trying to return itself to basal levels in the presence of continual tumor antigen exposure. This effort is not surprising because the control of harmful M reactions by immunosuppressive anti-inflammatory cytokines such as IL-10 and TGF- 1 limits the inflammatory consequences of immune responses, as evidenced by IL-10- and TGF- 1 -deficient mice that develop uncontrolled leukocyte activation and tissue injury [171, 172]. M -derived TNF- strongly suppresses T cell proliferation by autocrinely inducing M PGE 2 and NO production [5, 7]. TNF- induces peritoneal M s to mediate suppression but causes splenic M s to up-regulate T cell proliferation [118]. This TNF- -induced regulation is hindered in the TBH by increased class II splenic M PGE 2 production [118]. In fact, tumor-induced splenic M -derived TNF- autocrinely induces M suppression of autoreactivity through PGE 2 synthesis [173]. TNF- may play a suppressive role in vivo because peritoneal M s are the strongest TNF- producers [35, 120, 174], and TNF- would always be produced in the microenvironment of these M s. This suppressive role is confirmed by the administration of TNF- into normal murine hosts, which causes M -mediated suppression of lymphocyte proliferation [175]. The changes in splenic immune cell content after TNF- injection into normal mice [175] are similar to the tumorinduced changes observed in a TBH [1, 24]. In our nonmetastatic fibrosarcoma model, M TNF- production occurs even in tissues far-removed from tumor growth (such as peritoneal and splenic M populations), an observation that supports the possible in vivo suppressor role of TNF-. Tumors limit M production of immunostimulatory molecules As we have seen, many reports have documented a tumorinduced increase in M production of immunosuppressive factors. However, suppression during tumor growth also may arise from a decrease in production of a M -derived stimulatory molecule that affects other in situ immune cells. One candidate molecule may be IL-18, a M -derived cytokine that induces IFN- production and promotes lymphocyte-mediated immune responses [176]. Another molecule could be IL-12, which promotes T cell and lymphokine-activated killer cell proliferation and cytotoxicity [177] and favors the generation of T H 1-type cells [161]. Although IL-12 does not directly affect tumor growth, IL-12 reduces the metastatic potential of many tumor types by promoting immune cell infiltration of tumors [177 181]. Direct in situ administration of IL-12 increases the number of infiltrating tumoricidal M s and T cells [179], suggesting that IL-12 promotes tumoricidal responses and that tumor growth may compromise IL-12 production. Exogenous IL-12 restores TBH immunocompetence; therefore, tumorinduced M dysfunction may be manifested in reduced expression of IL-12. Dysregulation of IL-12 occurs among both tumor-proximal [182] and -distal [Mullins and Elgert, unpublished observations] immune cell populations. Both fibrosarcoma cells and suppressive M s produce significant amounts of IL-10 and TGF- 1 [6, 7], which may directly or indirectly block IL-12 synthesis. For example, NO induces M IL-12 gene expression [183]. However, tumor-derived factors, such as TGF- 1, reduce M NO production by inhibiting inos activity [184], and NO production by in situ [32] and tumor-distal [87] M populations is compromised. In the absence of autocrine stimulatory signals, which are blocked by tumor-derived factors, M s may be incapable of producing IL-12. As a result of the IL-12 deficiency, CD4 T cells may demonstrate poor responsiveness to activation cytokines. Furthermore, neutralization studies suggest that the inhibition of M IL-10, TGF- 1, and NO production significantly reverses M suppressor activity against T cells [7]. This partial obstruction of M suppressor activity may permit increased expression of IL-12, which would select for the growth of important antitumor T H 1-type cells. Tumors induce M hyporesponsiveness to induction signals Because of possible M proliferation in resident tissues, M s may increase cell numbers and cytokine concentrations during immune challenge [185]. Tumor growth may partly decrease immunocompetence through M hyporesponsiveness to proliferation signals, including GM-CSF. An up-regulatory molecule produced and used by M s and T cells during immunogenic challenge [186], GM-CSF activities during cancer are not well defined. GM-CSF normally enhances M activation [17] and accessory function [187] and increases MHC class II molecule expression [188]; however, these activities are inhibited by tumor growth [6, 81, 82, 189]. In the TBH, M s produce lower levels of, and are hyporesponsive to, GM-CSF [6]. GM-CSF increases normal host, but not TBH, splenic M MHC class II expression. Class II TBH M s become more suppressive in the presence of GM-CSF; in contrast, GM-CSF partly reverses suppression mediated by class II normal host M s. Because GM-CSF increases M accessory cell functions and class II molecule expression, tumor-induced decreases in GM-CSF production may account partly for decreased M accessory functions and reduced class II molecule expression. However, activation of splenic TBH class II M s with a high concentration of lipopolysaccharide induces GM-CSF synthesis at concentrations comparable to normal host class II M s [6]. The class II M subpopulation is significant because it is a potent suppressor population during tumor growth [4, 137]. GM-CSF increases T cell reactivity in the presence of normal host class II M s, whereas it further suppresses T cell reactivity in the presence of TBH class II M s [189]. Although GM-CSF 280 Journal of Leukocyte Biology Volume 64, September 1998

synthesis is unaltered, TBH class II M synthesis of the inhibitory molecule PGE 2 is increased significantly [3, 137]. This activity, in turn, significantly suppresses T cell function. Stimulation of TBH M s with GM-CSF does not induce proliferation that is comparable to the normal host M populations [6]. IL-10 contributes to the changes in tumor-induced M synthesis of, and responsiveness to, GM-CSF. Tumor growth heightens M susceptibility to IL-10-mediated inhibition of GM-CSF synthesis [6]. Low IL-10 concentrations significantly decrease the production of IL-1, IL-6, IFN-, and TNF- [190, 191]. IL-10 also decreases M synthesis of GM-CSF [6], although the suppression is not as great as that observed with TNF- synthesis [191]. TBH M s are highly susceptible to inhibition of GM-CSF-induced proliferation by IL-10. Although IL-10 significantly inhibits the synthesis of M -derived cytokines and ROI [190, 191], its importance as an inhibitor of M proliferation is unclear. Even though we did not screen for tumor-induced changes in other potential M growth factors, our data support the conclusion that tumor growth disrupts GM-CSF activities and that these changes compromise immune cell activities. Altered GM-CSF-mediated functions are compounded by tumor-induced increases to susceptibility to IL-10 suppression. The latter findings suggest IL-10 acts to deactivate suppressor activity during tumor growth. To our knowledge, no studies other than ours [6, 40] have found that IL-10 may serve as an inhibitory signal of tumor-induced M mediated immunosuppression. Further investigations are required to clarify the in vivo relevance of IL-10 during tumor-induced dysfunction. TUMOR GROWTH LEADS TO SHIFTS IN M SUBPOPULATIONS We have described how tumor-derived molecules can alter factor production by particular M populations. In addition, tumors can affect changes in M phenotype and thus favor a particular function. Using depletion [12, 136, 192, 193] and single [154, 193 195] and double-label [195, 196] fluorescent antibody studies, we showed a shift in M subpopulations, changes in M marker expression, and changes in M function during tumor growth. Tumor growth alters cell-surface marker expression Cell surface expression of Mac-1 (CD11b/CD18), Mac-2, Mac-3, and MHC class II molecules on peritoneal [136, 163] and splenic [192, 194] M s demonstrates identifiable shifts during tumor growth. More important, our studies correlated certain phenotypic and functional changes in these M populations. Tumor growth caused the phenotype of the peritoneal M subpopulation to shift from Mac-3 to Mac-2 [163], and this shift correlated with increased PGE 2 production [163]. In addition, the Mac-1 M subpopulation, an important downregulator of M -produced PGE 2 in the normal host, was absent in the TBH [136]. These studies [136], combined with those assessing the expression of peroxidase activity [197], suggested that TBH M s demonstrate an immature phenotype, and this conclusion is supported by the constitutive expression of c-myb [124, 198] and dysregulated cell-cycle kinetics [199, 200]. The immaturity of TBH M s may allow priming and activation stimuli to dysregulate proto-oncogene expression, which leads to over-production of suppressor cytokines. In TBH spleens, accessory M s, which are normally Mac-1, shifted to a Mac-1 phenotype [192]. Others show the induction of Mac- 1 2 suppressor M s by tumor-produced factors [79, 94] and a shift in alveolar M s from helper to suppressor phenotype during tumor growth [85]. Furthermore, during tumor growth, M s become more homogeneous and demonstrated a dramatic shift toward a small-sized population [154]. If tumor growth can alter M development and differentiation, leading to tumorinduced immunosuppression, mechanisms that induce M development may prove to be important in restoring immune function. Differences in phenotype and factor production between TBH splenic and peritoneal M populations correlate with their in situ functions Spleens normally have high numbers of class II M s, which produce far less TNF- and PGE 2 [118, 119], and perhaps NO, than do class II M s. As expected, splenic M s are much better antigen-presenting cells than are peritoneal M s because splenic M s express more MHC class II molecules [3, 137, 154, 196]. Antigen presentation is drastically compromised in the TBH spleen because of increased numbers and suppressor activity of class II M s [3, 137, 156]. There are sufficient numbers of antigen-presenting class II splenic M s in the TBH because suppression of T cell proliferation is reversed when suppressor molecule activity is blocked [201]. The investigations of M class II expression and factor production during tumor growth, using a nonmetastatic tumor located away from the peritoneum or spleen [154, 196], showed increased production of TNF- and NO by splenic M s that still remains less than that produced by peritoneal TBH M s [120]. Tumors that metastasize to the spleen, therefore, may have a better chance of survival than do those in inflammatory sites where exudate M s are active. We [3, 118, 156, 194] and others [122, 150, 202 204] show that MHC class II M s are the main suppressor and cytotoxic M subpopulation in the spleen. Tumor growth increases the numbers of splenic class II M s [196] and their suppressor and cytotoxic activities [118, 122, 202, 205]. Although increased splenic class II M suppressor and cytotoxic activities are mediated by PGE 2 and TNF- [118, 119, 122], respectively, NO and H 2 O 2 also may contribute to these activities, as do peritoneal class II M s [5, 7, 206] and monocytes [125]. NO suppresses M class II expression [207], suggesting that the tumor-induced increase in NO production maintains the class II M phenotype during tumor growth. Normally, TNF- counter-balances class II M suppressor activity by inducing M production of molecules that enhance T cell proliferation [118]. However, increased PGE 2 production by class II M s during tumor growth blocks the stimulatory action of TNF-, making these cells highly suppressive to Elgert et al. Tumor-induced immune dysfunction: the macrophage connection 281

splenic T cells [118]. The M -activating factor IFN- also induces class II M s to stimulate T cell proliferation through TNF- production [119]. Again, tumor-induced PGE 2 production stops IFN- -induced splenic M regulatory accessory activity [119]. In contrast to splenic M s, TNF- normally causes peritoneal M s to become more suppressive by inducing them to produce high levels of PGE 2 and NO [5, 7]. Suppressor activities are associated with class II M s In studies of tumor-induced changes in peritoneal and splenic M expression of Mac and MHC molecules, suppressor and cytotoxic activities were strongly associated with class II M s [3, 118, 122, 136, 150, 156, 194, 202, 208]. Tumor growth increased the number of class II M s in the spleen from roughly 30% to 70% [154, 196]. If class II M s are a prime cause of immunosuppression, regulation of class II expression may be a major mechanism through which tumors exert influence over the immune system. In fact, tumor growth suppresses class II protein expression by suppressing mrna through a decrease in M responsiveness to inducing agents, such as IFN-, and an increase in M sensitivity to suppressive agents, such as PGE 2 [209]. In TBH treated with indomethacin, an inhibitor of prostaglandin production, the percentage of class II M s increases [208]. This increase is associated with tumor regression and blockage of metastasis. Class II splenic M s are the main cytotoxic and suppressor population in the spleen through increased TNF- and PGE 2 production [118, 119]. However, peritoneal class II M s have stronger cytotoxic and suppressor activities than do their splenic counterparts, which correlate with the normally high percentage of class II peritoneal M s [154, 196]. For example, stimulated normal host splenic M NO production is almost undetectable, whereas normal host peritoneal M s produce much higher levels of NO [120]. In contrast, tumor growth causes splenic M s to produce high levels of NO [120]. Tumors may block M binding to extracellular matrix by down-regulating Mac-2 M s must bind to tumor cells to achieve cytotoxicity. One way that tumors may evade the cytotoxic effect is by modulating the binding of M to ECM proteins [210]. M lectins provide a mechanism for cellular interaction, which can be inhibited specifically by the simple sugars that are recognized by lectins [211]. Of particular interest is Mac-2, which we have correlated with tumor-induced M -mediated suppression [154, 196] and an increase in immature M s [136]. Mac-2, a galactosespecific animal lectin, was cloned [212, 213] and identified in M membrane, cytoplasmic, and nuclear fractions. The membrane form may affect IgE [213] and laminin [214] binding, whereas the intracellular form may affect cellular proliferation by associating with the heterogeneous nuclear ribonucleoprotein (hnrnp) complex [215, 216]. The drop in M membrane Mac-2 observed during tumor growth [210], linked with Mac-2 s laminin-binding ability [214, 217], suggests a reduced M ability to bind ECM. Tumor cells may survive by controlling the expression of the two forms of M Mac-2; the decrease in membrane Mac-2 in the TBH may impair the ability of M s to adhere to the tumor. Also, by increasing intracellular Mac-2 and inducing translocation, nuclear Mac-2 accumulates. An interaction in the nucleus between Mac-2 and hnrnp may cause TBH M s to continue to replicate while ignoring the tumor. MODELS OF TUMOR REGULATION OF M ACTIVITIES Tumor-derived molecules may regulate tumor-distal and tumorproximal M activities by the model depicted in Figure 4. The model shows how tumors might escape M antitumor activities by differential regulation of M cytotoxicity and suppressor activities. Through soluble priming and activating signals, tumors stimulate distal M s to produce cytotoxic molecules, such as TNF-, NO, and ROI, which also mediate lymphocyte suppression. Tumor-derived cytokines such as TGF- 1, IL-4, IL-6, colony-stimulating factors (CSFs), and MCP-1 travel through the circulation and prime distal resting M s for cytotoxic/suppressor molecule production. The activation of distal M s may occur when TAAs and ECM proteins from tumors bind to specific M receptors. Tumors may produce cytokines during early stages of their growth to prime M s for later activation by TAAs and ECM proteins. As tumors grow, more TAAs and ECM proteins may result from necrotic tumor debris and increased protease production. ECM protein dissemination is associated with protease activity of metastasizing tumors during later stages of tumor growth. Some tumor-derived M priming cytokines, such as TGF- 1 and MCP-1, elicit monocyte chemotaxis to the tumor. Cytokine-primed migrating monocytes will encounter increasing levels of TAAs and ECM protein activation signals as they approach the tumor site. Some tumor-derived cytokines activate resting M s but suppress activated M s, suggesting that the priming activity of tumorderived cytokines will switch to a M -deactivating activity as migrating monocytes encounter high levels of TAAs and ECM protein. That is, production of the cytotoxic/suppressor molecules TNF-, NO, and ROI by activated tumor-distal M s will be inhibited by tumor-derived TGF- 1, IL-10, PGE 2, IL-4, IL-6, and M-CSF as M s approach the tumor. Simultaneously, M s will be induced by local tumor-derived TGF- 1, IL-10, CSFs, and perhaps other molecules (see Table 1) to remain suppressive to T cells through noncytotoxic suppressor molecules. The tumor also may become apoptosis-resistant to M -derived membrane or soluble TNF-. It is unknown whether M s mediate killing of tumor cells by Fas-FasLmediated apoptosis and thus if Fas-FasL interactions can be altered. Furthermore, these tumor-proximal M s will be induced to synthesize tumor cell-growth factors and angiogenic factors. The autocrine and paracrine effects of molecules produced by tumor-activated M s also regulate M activities (Figs. 5 and 6). Tumors increase the number and suppressor action of MHC class II splenic (Fig. 5) and peritoneal (Fig. 6) M s that are distal to tumor growth. Tumor-derived CSFs may mediate M priming for TNF-, TGF-, IL-10, PGE 2, and NO synthesis and may increase the numbers of class II M s. Although tumor-derived CSFs increase the total M numbers in the 282 Journal of Leukocyte Biology Volume 64, September 1998

Fig. 4. Regulation of distal and proximal macrophage activities by tumor-derived molecules. Tumor-derived cytokines prime and tumor-derived TAAs and ECM protein activate distal M s for cytotoxic and suppressor molecule production. As monocytes migrate to the tumor site, tumor-derived cytokines inhibit production of M cytotoxic/suppressor molecules such as TNF-, NO, and ROI, but increase production of noncytotoxic suppressor molecules such as PGE 2. Tumor-derived cytokines also may induce M production of tumor cell growth factors and angiogenic factors required for tumor growth. spleen, tumor-induced M production of TGF- 1, IL-10, PGE 2, and TNF- may select for class II M s by down-regulating class II expression. Besides selecting and maintaining the class II suppressor M phenotype, tumor-induced M production of PGE 2, NO, TGF-, and IL-10 also suppresses T H -cell functions. Perhaps tumors only suppress T H 1-type cell responses because tumors produce cytokines such as IL-10 and PGE 2 that inhibit T H 1-type cell activity or through reduced M IL-12 production, which would decrease the generation of T H 1-type cells. Tumor-stimulated M s produce TNF-, which autocrinely enhances PGE 2 and NO production by both splenic and peritoneal M s (Figs. 4 and 5). TNF- autocrinely and paracrinely induces suppressor activity by peritoneal M s (Fig. 6), but not by splenic M s (Fig. 5). The paracrine action of the T cell-derived M activation factor IFN- induces suppressor action by peritoneal M s (Fig. 6) but not by splenic M s (Fig. 5). Both TNF- and IFN- stimulate splenic M s to enhance T H cell functions (Fig. 5). During tumor growth, however, increased M PGE 2 production blocks TNF- - and IFN- induction of splenic M enhancing effects on T H cells, and it sometimes induces suppressor activity by splenic M s. TNF- and IFN- -induced suppressor action in M s is normally controlled by the autocrine activity of IL-10 and TGF- 1 because M production of TGF- 1 and IL-10 usually down-regulates M synthesis of other suppressor molecules (Figs. 4 and 5). Tumor growth causes dysregulation of suppressor molecule production, rendering the control by TGF- 1 and IL-10 ineffective. CONCLUSIONS These data show that tumor-mediated regulation of M activities can favor tumor growth. The evidence presented suggests that tumor-derived molecules deactivate tumor-proximal M Elgert et al. Tumor-induced immune dysfunction: the macrophage connection 283

Fig. 5. Tumor regulation of distal MHC class II splenic macrophage activities. Tumor-derived CSFs increase splenic M numbers, and tumor- and M -derived TGF- and PGE 2 down-regulate MHC class II expression. These M s are the main cytotoxic and suppressor M s during tumor growth by producing TNF- and PGE 2, respectively. Tumor-derived TAAs and ECM stimulate class II M PGE 2 and TNF- production. T H 1 cell-derived IFN- normally induces class II M TNF- production, which stimulates T H cell proliferation. Tumor-induced class II M PGE 2 production blocks TNF- -induced up-regulation of T H cell proliferation, causing class II M suppressor activity. Activated M -derived TGF- and IL-10 normally down-regulate M suppressor molecule synthesis. During tumor growth, the controlling mechanisms of TGF- and IL-10 on M suppressor activity is dysregulated so those TBH M s have increased expression of suppressor molecules. populations while activating tumor-distal M populations. The current challenge is to determine how tumor-derived molecules specifically select M functions that benefit tumor growth. For example, studies should determine whether cancer cells produce cytokines that inhibit M cytotoxicity while inducing suppressor activity, and whether tumor growth induces M production of molecules that are both cytotoxic and suppressive, as reported here for NO and TNF-. Cytokine cdna transfection studies may be most useful in assessing the action of certain tumor-derived cytokines. Because different tumor cells expressing the same transfected cytokine elicit varying immune responses, different tumor types may produce unique factors to manipulate immune responses. More attention should be given to the factors that tumors produce so an association can be made between the production of cytokines and particular tumor types. cdna transfection of tumor-derived factors or gene-deletion in tumor cells could characterize the importance of various tumor molecules on M activities. The use of humanized antitumor-derived factor antibody may be effective in interfering with tumor signals that disrupt M cytotoxic and antigen-presenting activities. In fact, transfected phagedisplayed anti-cytokine antibodies may provide a blocking tool. These kinds of studies will provide valuable insights into what changes occur in M s during tumor growth. More importantly, these studies will suggest mechanisms for how tumor growth changes M s. Without the underlying knowledge of how tumor growth alters M function, we can only describe a phenomenon. As we learn more about tumor growth, we can determine the molecular and cellular origins of tumor-induced changes in the immune system. Understanding this impressive array of biological effects will open a vast number of opportunities for therapeutic intervention. ACKNOWLEDGMENTS K. D. E. thanks the many past and present members of his laboratory for experimental assistance and helpful discussions. We gratefully acknowledge the review of the manuscript by Dr. Carol J. Burger, Dr. Andrew D. Yurochko, and Michael J. 284 Journal of Leukocyte Biology Volume 64, September 1998