11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`P1: IKH
`10.1146/annurev.immunol.22.012703.104803
`
`Annu. Rev. Immunol. 2004. 22:329–60
`doi: 10.1146/annurev.immunol.22.012703.104803
`Copyright c 2004 by Annual Reviews. All rights reserved
`First published online as a Review in Advance on October 15, 2003
`THE THREE ESOF CANCER IMMUNOEDITING
`Gavin P. Dunn,1Lloyd J. Old,2 and Robert D. Schreiber1
`1Department of Pathology and Immunology, Center for Immunology, Washington
`University School of Medicine, St. Louis, Missouri 63110;
`email: schreiber@immunology.wustl.edu
`2Ludwig Institute for Cancer Research, New York Branch at Memorial Sloan-Kettering
`Cancer Center, New York, NY 10021; email: lold@licr.org
`
`Key Words
`immunosurveillance, tumor, lymphocytes, interferon, tumor sculpting
`■ Abstract After a century of controversy, the notion that the immune system regu-
`lates cancer development is experiencing a new resurgence. An overwhelming amount
`of data from animal models—together with compelling data from human patients—
`indicate that a functional cancer immunosurveillance process indeed exists that acts
`as an extrinsic tumor suppressor. However, it has also become clear that the immune
`system can facilitate tumor progression, at least in part, by sculpting the immunogenic
`phenotype of tumors as they develop. The recognition that immunity plays a dual role
`in the complex interactions between tumors and the host prompted a refinement of the
`cancer immunosurveillance hypothesis into one termed “cancer immunoediting.” In
`this review, we summarize the history of the cancer immunosurveillance controversy
`and discuss its resolution and evolution into the three Es of cancer immunoediting—
`elimination, equilibrium, and escape.
`
`INTRODUCTION
`The concept that the immune system can recognize and eliminate primary develop-
`ing tumors in the absence of external therapeutic intervention has existed for nearly
`100 years. However, the validity of this concept has, in the past, been difficult to
`establish. When first proposed in 1909 (1), the hypothesis could not be experimen-
`tally tested because so little was known at the time about the molecular and cellular
`basis of immunity. Later on, as the field of immunology developed and the concept
`acquired its name—cancer immunosurveillance (2, 3)—experimental testing be-
`came possible but failed to provide evidence for the process, using mice with spon-
`taneous mutations that rendered them immunocompromised but not completely
`immunodeficient (4). Only recently, with the development of gene targeting and
`transgenicmousetechnologiesandthecapacitytoproducehighlyspecificblocking
`monoclonal antibodies (mAb) to particular immune components, has the cancer
`immunosurveillance hypothesis become testable in unequivocal, molecularly de-
`fined murine models of immunodeficiency. Over the past ten years, the use of these
`329
`0732-0582/04/0423-0329$14.00
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 1 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`330
`
`DUNN ⌅ OLD ⌅ SCHREIBER
`
`improved in vivo cancer models has provided strong and convincing data that have
`rekindled interest in the cancer immunosurveillance hypothesis. Most recently,
`this conundrum has been further clarified by the demonstration that the immune
`system not only can protect the host against tumor development but also, by se-
`lecting for tumors of lower immunogenicity, has the capacity to promote tumor
`growth. These dual effects of the immune system on developing tumors prompted
`us to refine the cancer immunosurveillance hypothesis into one we termed cancer
`immunoediting (5, 6). We envisage that this process is comprised of three phases
`that are collectively denoted the three Es of cancer immunoediting: elimination,
`equilibrium, and escape. In this review, we first present data supporting the ex-
`istence of the elimination phase (i.e., cancer immunosurveillance) as it occurs in
`mice and humans and propose a model for the molecular and cellular events that
`underlie this process. Second, we provide evidence for a tumor-sculpting role of
`immunity and discuss the relationship between this function and the equilibrium
`and escape phases of cancer immunoediting. Third, we outline the implications of
`this concept for the understanding and treatment of human cancer.
`
`CANCER IMMUNOSURVEILLANCEIN MICE
`Historical Perspective
`The validity of the cancer immunosurveillance hypothesis has emerged only re-
`cently from a long history of heated debate (reviewed in 6). The notion that the
`immune system could protect the host from neoplastic disease was initially pro-
`posed by Ehrlich (1) and formally introduced as the cancer immunosurveillance
`hypothesis nearly 50 years later by Burnet and Thomas (2, 3, 7–9). Based on an
`emerging understanding of the cellular basis of transplantation and tumor immu-
`nity (10–15), Burnet and Thomas predicted that lymphocytes were responsible
`for eliminating continuously arising, nascent transformed cells. However, when
`this prediction was put to the experimental test using nude mice, which were the
`most congenitally immunodeficient mice available at the time (16, 17), no con-
`vincing evidence for such a process was obtained. Specifically, CBA/H strain nude
`mice neither developed increased incidences of carcinogen [methylcholanthrene
`(MCA)]-induced or spontaneous tumors nor did they show shortened periods of
`tumor latency compared with wild-type controls (4, 18–22).
`However, in retrospect, there are several important caveats to these experiments
`that could not have been appreciated at the time. First, the nude mouse is now
`recognized to be an imperfect model of immunodeficiency. These mice produce
`low but detectable numbers of functional populations of ↵ T cells (23–25) and
`therefore can manifest at least some degree of adaptive immunity. Second, the
`existence of natural killer (NK) cells (which are present and function normally
`in nude mice) was not well established at the time (26) and thus very little was
`known about their origins, actions, or roles in promoting innate immunity. In
`addition, the profound influence of innate immunity on adaptive immunity was
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 2 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`CANCER IMMUNOEDITING
`
`331
`
`not recognized (27). Thus, the residual adaptive immune system in the presence
`of a fully functional innate immune system may provide the nude mouse with
`at least some cancer immunosurveillance capacity. Third, the CBA/H strain mice
`used in Stutman’s MCA carcinogenesis experiments express the highly active
`isoform of the aryl hydroxylase enzyme that is required to metabolize MCA into its
`carcinogenic form (28, 29). Therefore, it is conceivable that MCA-induced cellular
`transformation in CBA/H strain mice occurred so efficiently that it masked any
`protective effect that immunity could provide. Nevertheless, since these caveats
`can only be appreciated in hindsight, the Stutman experiments were considered to
`be so convincing that by the end of the 1970s, the death knell had sounded for the
`cancer immunosurveillance hypothesis.
`
`THERENAISSANCEOF CANCER IMMUNOSURVEILLANCE
`IFN-, Perforin, and Lymphocytes in Tumor Immunity
`In the 1990s, two sets of studies incited renewed interest in cancer immunosurveil-
`lance. First, endogenously produced interferon- (IFN-) was shown to protect
`the host against the growth of transplanted tumors and the formation of primary
`chemically induced and spontaneous tumors (30–33). The injection of neutral-
`izing monoclonal antibodies specific for IFN- into mice bearing transplanted,
`established Meth A tumors blocked LPS-induced tumor rejection (30). In addi-
`tion, transplanted fibrosarcomas grew faster and more efficiently in mice treated
`with IFN--specific mAb. These observations were then extended to models of
`primary tumor formation. IFN--insensitive 129/SvEv mice lacking either the
`IFNGR1 ligand-binding subunit of the IFN- receptor or STAT1, the transcrip-
`tion factor responsible for mediating much of IFN-’s biologic effects on cells
`(34), were found to be 10–20 times more sensitive than wild-type mice to tumor
`induction by methylcholanthrene (31). Specifically, these mice developed more
`tumors, more rapidly, and at lower MCA doses than did wild-type controls. These
`results were subsequently confirmed by independent experiments using C57BL/6
`strain mice lacking the gene encoding IFN- itself (32). Similarly, in models of
`genetically driven tumorigenesis, mice lacking the p53 tumor suppressor gene
`and either IFNGR1 or STAT1 formed a wider spectrum of tumors compared with
`IFN--sensitive mice lacking only p53 (31). In addition, compared to their IFN--
`sufficient counterparts, IFN- / C57BL/6 mice showed an increased incidence
`of disseminated lymphomas, and IFN- / BALB/c mice displayed an increased
`incidence of spontaneous lung adenocarcinomas (33).
`Second, mice lacking perforin (pfp/) were found to be more susceptible
`to MCA-induced and spontaneous tumor formation compared with their wild-
`type counterparts (32, 33, 35–37). Perforin is a component of the cytolytic gran-
`ules of cytotoxic T cells and NK cells that plays an important role in mediating
`lymphocyte-dependent killing (38). Following challenge with MCA, pfp/ mice
`developed significantly more tumors compared with wild-type mice treated in the
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 3 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`332
`
`DUNN ⌅ OLD ⌅ SCHREIBER
`
`same manner (32, 35, 36). Untreated pfp/ mice also showed a high incidence of
`spontaneous disseminated lymphomas, which was accelerated on a p53/ back-
`ground (37). BALB/c mice lacking perforin also displayed a low incidence of
`spontaneous lung adenocarcinomas, which was not observed in wild-type mice
`(33). Taken together, these observations demonstrated that tumor development
`in mice was controlled by components of the immune system and stimulated a
`considerable amount of work aimed at better defining this process (Table 1).
`The definitive work demonstrating the existence of an IFN-- and lymphocyte-
`dependent cancer immunosurveillance process was based on experiments employ-
`ing gene-targeted mice that lack the recombinase activating gene (RAG)-2 (5).
`Mice lacking RAG-2 (or its obligate partner RAG-1) cannot rearrange lymphocyte
`antigen receptors and thus lack T, B, and NKT cells (39). Since RAG-2 expression
`is limited to cells of the immune system, the use of RAG-2/ mice provided an
`appropriate model to study the effects of host immunodeficiency on tumor devel-
`opment because, unlike other genetic models of immunodeficiency (such as SCID
`mice), the absence of RAG-2 would not result in impaired DNA repair in nonlym-
`phoid cells undergoing transformation. Following challenge with MCA, 129/SvEv
`RAG-2/ mice developed sarcomas more rapidly and with greater frequency than
`genetically matched wild-type controls (5) (Figure 1A). After 160 days, 30/52
`RAG-2/ mice formed tumors, compared with 11/57 wild-type mice. Similar
`findings were obtained in MCA tumorigenesis experiments that used RAG-1/
`C57BL/6 mice (40). Moreover, Helicobacter-free RAG-2/ 129/SvEv mice aged
`in a specific pathogen-free mouse facility and maintained on broad-spectrum an-
`tibiotics formed far more spontaneous epithelial tumors than did wild-type mice
`housed in the same room (5; A.T. Bruce & R.D. Schreiber, unpublished observa-
`tions) (Figure 1B). Specifically, 26/26 RAG-2/ mice ranging in age from 13–
`24 months developed spontaneous neoplasia, predominantly of the intestine; 8 of
`these mice had premalignant intestinal adenomas, 17 had intestinal adenocarcino-
`mas, and 1 had both an intestinal adenoma and a lung adenocarcinoma. In contrast,
`only 5/20 wild-type mice aged 13–24 months developed spontaneous neoplasia,
`which was predominantly benign. Three wild-type mice developed adenomas of
`the Harderian gland, lung, and intestine, respectively; one developed a Harderian
`gland adenocarcinoma; and one developed an endometrial stromal carcinoma.
`Thus, lymphocytes protect mice against the formation of both chemically induced
`and spontaneous tumors.
`The overlap between the IFN-- and lymphocyte-dependent tumor suppressor
`pathways was explored by comparing tumor formation in 129/SvEv mice lack-
`ing either IFN- responsiveness (IFNGR1/ or STAT1/ mice), lymphocytes
`(RAG-2/ mice), or both [RAG-2/ X STAT1/ (RkSk) mice] (5). Each of
`the four lines of gene-targeted mice formed three times more chemically induced
`tumors than syngeneic wild-type mice when injected with a single 100 µg dose
`of MCA (Figure 1A). Since no significant differences were detected between any
`of the gene-targeted mice, the conclusion was reached that the IFN-/STAT1
`and lymphocyte-dependent extrinsic tumor suppressor mechanisms were heavily
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 4 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`CANCER IMMUNOEDITING
`
`333
`
`TABLE 1 Enhanced susceptibility of immunodeficient mice to chemically induced and
`spontaneous tumors
`
`Technology
`
`RAG-2/
`
`Immune status
`
`Lacks T, B, NKT cells
`
`RAG-2/⇥STAT1/
`(RkSk)
`
`Lacks T, B, NKT cells;
`IFN-, ↵/-insensitive
`
`RAG-1/
`BALB/c SCID
`TCR/
`TCR/
`
`J↵281/
`LMP2/
`Anti-asialo-GM1
`
`Anti-NK1.1
`Anti-Thy1
`STAT1/
`
`IFNGR1/
`
`IFN- /
`
`Lacks T, B, NKT cells
`Lacks T, B, NKT cells
`Lacks ↵ T cells
`Lacks T cells
`
`Lacks NKT cell subset
`Lacks LMP2 subunit
`Lacks NK cells,
`mono-cytes/macrophages
`Lacks NK, NKT cells
`Lacks T cells
`IFN--, ↵/-insensitive
`
`IFN--insensitive
`
`Lacks IFN-
`
`GM-CSF/IFN- /
`
`Lacks GM-CSF, IFN-
`
`Pfp/⇥IFN- /
`
`Lacks Perforin, IFN-
`
`Pfp/
`
`TRAIL/
`Anti-TRAIL
`
`IL-12p40/
`Wt + IL-12
`Wt + ↵-GalCer
`
`Lacks Perforin
`
`Lacks TRAIL
`Blockade of TRAIL
`function
`
`Lacks IL-12
`Exogenous IL-12
`Exogenous NKT cell
`activation
`
`Tumor susceptibility
`relative to wild type
`" MCA-induced sarcomas;
`" spontaneous intestinal neoplasia
`" MCA-induced sarcomas;
`" spontaneous intestinal and
`mammary neoplasia
`" MCA-induced sarcomas
`" MCA-induced sarcomas
`" MCA-induced sarcomas
`" MCA-induced sarcomas;
`" DMBA/TPA-induced skin
`tumors
`" MCA-induced sarcomas
`" Spontaneous uterine neoplasms
`" MCA-induced sarcomas
`" MCA-induced sarcomas
`" MCA-induced sarcomas
`" MCA-induced sarcomas;
`wider tumor spectrum
`in STAT1/⇥p53/
`" MCA-induced sarcomas;
`wider tumor spectrum in
`IFNGR1/⇥p53/
`" MCA-induced sarcomas;
`B6: " spontaneous disseminated
`lymphomas;
`BALB/c: " spontaneous lung
`adenocarcinomas
`" Spontaneous lymphomas;
`" nonlymphoid solid cancers
`" MCA-induced sarcomas;
`" spontaneous disseminated
`lymphomas
`" MCA-induced sarcomas;
`" spontaneous disseminated
`lymphomas
`" MCA-induced sarcomas
`" MCA-induced sarcomas;
`" spontaneous sarcomas,
`disseminated lymphomas
`" MCA-induced sarcomas
`# MCA-induced sarcomas
`# MCA-induced sarcomas
`
`References
`
`(5)
`
`(5)
`
`(40)
`(40)
`(58)
`(58)
`
`(32, 36, 40)
`(169)
`(40)
`
`(36, 40)
`(36)
`(5, 31)
`
`(5, 31)
`
`(32, 33)
`
`(55)
`
`(32, 33)
`
`(32, 33, 35–37)
`
`(61)
`(60)
`
`(36)
`(62)
`(63)
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 5 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`334
`
`DUNN ⌅ OLD ⌅ SCHREIBER
`
`Figure 1 Increased incidence of chemically induced and spontaneous tumors in im-
`munodeficient mice. (A) Age- and sex-matched mice were inoculated with 100 µg
`MCA and monitored for tumor development for 160 days. (B) Mice housed in a
`specific pathogen-free facility were monitored for spontaneous tumor development
`between 13–24 months. Adapted from Shankaran et al. (5).
`
`overlapping. However, RkSk mice developed spontaneous breast tumors that were
`not observed in wild-type or RAG-2/ mice, therefore demonstrating that the
`overlap between the two pathways was incomplete (Figure 1B). Similar findings
`were made in carcinogenesis experiments employing mice that lacked either per-
`forin, IFN-, or both, where a small increase was observed in tumor induction
`in the doubly deficient mice compared with mice lacking only one of the two
`components (32).
`Identification of the Components of the Immmunosurveillance
`Network
`TUMOR CELLS AS KEY TARGETS OF IFN- The finding that endogenously pro-
`duced IFN- played a critical role in protecting mice against tumor develop-
`ment stimulated a search for the physiologically important cellular targets of this
`cytokine. Two approaches demonstrated that the tumor cell itself is an impor-
`tant IFN- target in tumor rejection. In the first, the effects of ablating IFN-
`sensitivity on the immunogenicity of IFN--sensitive tumor cells was assessed us-
`ing models of tumor cell
`transplantation (30). Meth A tumor cells, when
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 6 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`CANCER IMMUNOEDITING
`
`335
`
`engineered to be unresponsive to IFN- by overexpression of a dominant-negative
`IFNGR1 mutant (mgR.1IC) (41), grew more aggressively than mock-transfected
`controls when transplanted into na¨ıve syngeneic wild-type hosts and were resistant
`to LPS-induced tumor rejection (30, 41). Unlike their IFN--sensitive counter-
`parts, IFN--insensitive Meth A.mgR.1IC cells failed to prime na¨ıve recipients
`for development of Meth A immunity and were poorly recognized when injected
`into mice with pre-established immunity to the parental wild-type tumor cell line.
`Similar results were obtained with a second fibrosarcoma derived from a C57BL/6
`mouse(MCA-207).Thesecondapproachemployedanoppositestrategywherethe
`effects on in vivo tumor growth were assessed following restoration of IFN- sen-
`sitivity to tumor cells generated in IFN--insensitive IFNGR1/ mice (31). When
`transplanted into wild-type mice, IFNGR1-deficient RAD.gR.28 tumor cells were
`highly tumorigenic and formed progressively growing tumors even when injected
`at very low cell number (10–100 cells/mouse). In contrast, when RAD.gR.28 cells
`were rendered responsive to IFN- by complementation with wild-type IFNGR1,
`theresultingtumorcellline(RAD.gR.28.mgR)washighlyimmunogenicandfailed
`toformprogressivelygrowingtumorsinwild-typerecipientsevenwheninjectedat
`high cell number (5 ⇥ 106 cells/mouse). Demonstration that RAD.gR.28.mgR re-
`jection occurred via an IFN--dependent immunologic mechanism was evidenced
`by the observations that (a) rejection of RAD.gR.28.mgR cells in wild-type mice
`was inhibited by administration of IFN- mAb (31), (b) rejection was inhibited
`if wild-type mice were depleted of either CD4+ or CD8+ T cells (A.T. Bruce &
`R.D. Schreiber, unpublished observations), and (c) RAD.gR.28.mgR cells formed
`progressively growing tumors when injected into RAG-2/ mice (31). Thus, the
`effects of using IFN--insensitive tumor cells are the same as blocking IFN-
`availability in the intact mouse: Immune rejection of the tumor is inhibited. To-
`gether, these results formed the basis for the conclusion that the tumor cell is a
`physiologically relevant target of IFN- in the tumor rejection process.
`Subsequent studies have pointed to several effects of IFN- on tumor cells
`that could promote tumor elimination. IFN-’s capacity to enhance tumor cell
`immunogenicity by upregulating components of the MHC class I antigen process-
`ing and presentation pathway has been shown to be sufficient for tumor rejection.
`IFN--insensitive RAD.gR.28 tumor cells engineered for enforced expression of
`either TAP-1 (5) or H-2Db (A.T. Bruce & R.D. Schreiber, unpublished obser-
`vations) were rejected when transplanted into na¨ıve syngeneic recipients in an
`immunologic manner that was indistinguishable from that of IFN--responsive
`RAD.gR.28.mgR cells. In contrast, RAD.gR.28 cells engineered for expression of
`H-2Kbwerenotrejected(5).ThefindingthatenforcedexpressionofH-2Db,butnot
`H-2Kb, caused rejection of RAD.gR.28 corresponds to the H-2Db MHC restriction
`displayed by protective CD8+ T cells that arise naturally in mice immunized with
`RAD.gR.28.mgR (A.T. Bruce & R.D. Schreiber, unpublished observations). Thus,
`the capacity of IFN- to regulate tumor cell immunogenicity via enhancement of
`MHC class I pathway function is a physiologically relevant action that promotes
`tumor rejection.
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 7 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`336
`
`DUNN ⌅ OLD ⌅ SCHREIBER
`
`Other known IFN--dependent effects on developing tumors may also con-
`tribute to the rejection process; however, their physiologic relevance to the process
`has not yet been established. IFN- has profound antiproliferative and/or proapop-
`toticeffectsoncertaintumorcells.Intheformercase,IFN- caninduceexpression
`of cell cycle inhibitors such as p21WAF1/CIP1 or p27KIP1 that bind to and inhibit the
`cyclin-dependentkinaseCDK-2(42)orCDK-4(43),respectively.Inthelattercase,
`IFN- can induce expression of gene products such as caspase-1 (44, 45), Fas, and
`Fasligand(46)that,undertheproperconditions,canpromotetumorcellapoptosis.
`IFN- can also stimulate tumor cells to produce the chemokines CXCL-9 (Mig)
`and CXCL-10 (IP-10), which, in addition to having potent chemoattractant activity
`for CXCR3-expressing leukocyte populations, also function as powerful inhibitors
`of angiogenesis (47–52). Although all the aforementioned processes likely con-
`tribute in some way to the antitumor response, the relative importance and interre-
`lationships between the immunologic and nonimmunologic actions of IFN- on
`developing tumors in promoting tumor rejection requires further analysis.
`
`HOST CELLS AS POTENTIAL ADDITIONAL TARGETS OF IFN- Evidence has also
`been obtained supporting a role for IFN- and/or STAT1 at the level of the host
`immune system in the tumor rejection process. IFN--unresponsive mice lacking
`STAT1 failed to reject highly immunogenic P198 tumor cells that were completely
`eliminated in wild-type mice (53). Similar findings have also been made using the
`highly immunogenic RAD.gR.28.mgR fibrosarcoma cell line that was rejected in
`wild-type mice but grew progressively in STAT1/ mice (V. Shankaran & R.D.
`Schreiber, unpublished observations). In addition, T cells derived from STAT1/
`mice immunized with poorly immunogenic P1.HTR tumor cells in the presence
`of IL-12 failed to express cytolytic activity against the tumor. In contrast, T cells
`derived from similarly immunized wild-type mice developed potent cytocidal ca-
`pacity.MicelackingSTAT6,whichtendtopolarizetheirCD4+Tcellcompartment
`more easily into Th1 cells, spontaneously rejected poorly immunogenic P1.HTR
`tumor cells that grew progressively in wild-type mice (54). Thus, these stud-
`ies suggest that IFN-’s well-recognized STAT1-dependent promotion of CD4+
`T cell polarization into Th1 cells facilitates development of the appropriate type
`of cellular immune response needed for tumor rejection.
`Another study revealed a more indirect immunological action of IFN- at the
`level of the host in preventing tumor development (55). Both GM-CSF/IFN- /
`doubly deficient and GM-CSF/IL-3/IFN- / triply deficient mice were found to
`be highly susceptible to bacterial infection, displayed acute and chronic inflamma-
`tion in a variety of different organs, and developed high incidences of spontaneous
`lymphoma and nonlymphoid solid cancers. The incidences of infection, inflamma-
`tion, and neoplasia were much reduced in mice lacking GM-CSF alone, IL-3 and
`GM-CSF only, or IFN- alone. Tumor development in the IL-3/GM-CSF/IFN-
`triply gene-targeted mice was prevented or delayed by maintaining the mice on
`broad-spectrum antibiotics from birth. These results suggest a role for IFN-,
`in combination with GM-CSF, in controlling chronic infections that can lead to
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 8 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`CANCER IMMUNOEDITING
`
`337
`
`a chronic inflammatory state that ultimately may result in cancer development.
`Clearly,therelationshipbetweenbacterial/microbialimmunosurveillanceandcan-
`cer immunosurveillance warrants further analysis but must await the development
`of in vivo models that can unequivocally differentiate between the two processes.
`Finally, other studies have suggested that host cells of nonimmunologic origin
`may also be important targets of IFN- in the antitumor response (56, 57). These
`studies report that IFN- can induce angiostatic effects in tumors by targeting
`nontransformed host cells that are in close proximity to the tumor. It is possible
`that the underlying mechanism of this effect is similar to the one that has already
`been discussed in the context of the tumor cells themselves—the IFN--dependent
`induction in host stromal cells of the angiostatic chemokines IP-10 and Mig.
`
`THE CELLULAR EFFECTORS OF CANCER IMMUNOSURVEILLANCE Otherstudieshave
`begun to shed light on the specific lymphocyte subsets that are involved in can-
`cer immunosurveillance. Together, these studies have shown that components
`of both the adaptive and innate immune systems participate in the process.
`Girardi et al. (58) examined the relative contributions of different T-cell subsets in
`blocking primary tumor formation in mice lacking ↵ T cells (TCR/) and/or
` T cells (TCR/). MCA treatment of either type of TCR/ mouse led to
`an increased incidence of fibrosarcomas and spindle cell carcinomas compared
`with wild-type controls, thereby showing that both ↵ and T-cell subsets play
`critical and nonredundant host-protective roles in this particular model of tumor
`development. However, in an initiation/promotion model of DMBA- and TPA-
`induced skin tumorigenesis, TCR/ mice showed an increased susceptibility to
`tumor formation and a higher incidence of papilloma-to-carcinoma progression
`than wild-type mice, whereas TCR/ mice did not. This result suggests that im-
`munosurveillance may be a multivariable process requiring the actions of different
`immune effectors in a manner dependent on the tumor’s cell type of origin, mech-
`anism of transformation, anatomic localization, and mechanism of immunologic
`recognition.
`NK and NKT cells represent cellular populations of the innate immune com-
`partment that were shown to protect the host from tumor formation. C57BL/6
`mice depleted of both NK and NKT cells using the NK1.1 mAb were two to three
`times more susceptible to MCA-induced tumor formation than wild-type controls
`(40). In the same study, C57BL/6 mice depleted of NK cells following anti-asialo-
`GM1 treatment were two to three times more prone to developing MCA-induced
`tumors than control counterparts. Although anti-asialo-GM1 can also deplete ac-
`tivated macrophages, this study nevertheless supports the involvement of cells of
`innate immunity in blocking primary tumor development. A role for NKT cells in
`this process was implicated when J↵281/ mice, which lack a large population of
`V↵14J↵281-expressing invariant NKT cells, were found to develop MCA-induced
`sarcomas at a higher incidence than their wild-type counterparts (36).
`Additional evidence pointing to cells of innate immunity as critical effec-
`tors of cancer immunosurveillance comes from studies of the TNF-related
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 9 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`338
`
`DUNN ⌅ OLD ⌅ SCHREIBER
`
`apoptosis-inducing ligand (TRAIL). A member of the TNF superfamily that in-
`duces apoptosis through engagement of the TRAIL-R2 (DR5) receptor in mice,
`TRAIL is expressed constitutively on a subset of liver NK cells and is induced by
`either IFN- or IFN-↵/ in monocytes, NK cells, and dendritic cells (59). When
`injected with low doses of MCA, C57BL/6 strain mice treated with neutralizing
`antibodies to TRAIL (60) or lacking the TRAIL gene (61) developed fibrosar-
`comas at a higher incidence than wild-type controls. Moreover, C57BL/6 strain
`p53+/ mice treated with the same neutralizing TRAIL antibody exhibited a higher
`incidence of spontaneous sarcoma and disseminated lymphoma formation over a
`two-year span than control IgG-treated mice (60). Further study will be required
`to identify the specific innate cell subsets that manifest the TRAIL-dependent
`antitumor effects.
`Finally, evidence also exists showing that enhancing immune system activity
`leads to reduced primary tumor formation in models of MCA tumorigene-
`sis. Mice treated with either IL-12 (62) or the prototypic NKT cell activator ↵-
`galactosylceramide (↵-GalCer) (63) throughout the MCA carcinogenesis process
`had a reduced incidence of tumors after longer latency periods than control mice.
`In summary, using a variety of well-characterized gene-targeted mice, specific
`immune system activators, and blocking monoclonal antibodies highly specific
`for distinct immunologic components, a large body of work has now accumulated
`to support the statement that the immune system indeed functions to protect the
`murine host against development of both chemically induced and spontaneous
`tumors (Table 1).
`
`CANCER IMMUNOSURVEILLANCEIN HUMANS
`Given that there is significant evidence supporting the existence of a cancer im-
`munosurveillanceprocessinmice,doesasimilarprocessexistinhumans?Analysis
`ofindividualswithcongenitaloracquiredimmunodeficienciesorpatientsundergo-
`ing immunosuppressive therapy has documented a highly elevated incidence of vi-
`rally induced malignancies such as Kaposi’s sarcoma, non-Hodgkin’s lymphoma,
`and cancers of the anal and urogenital tracts compared with immunocompetent
`individuals (64–66). However, the study of the incidence of cancers of nonviral
`origins that may take many years to develop is confounded by the variety of viral
`and bacterial infections to which these immunodeficient/immunosuppressed pa-
`tients are susceptible and by the more rapid appearance of virally induced tumors.
`Nevertheless, one can draw upon three lines of evidence to suggest that cancer
`immunosurveillance indeed occurs in humans: (a) immunosuppressed transplant
`recipients display higher incidences of nonviral cancers than age-matched im-
`munocompetent control populations; (b) cancer patients can develop spontaneous
`adaptive and innate immune responses to the tumors that they bear, and (c) the
`presence of lymphocytes within the tumor can be a positive prognostic indicator
`of patient survival.
`
` Access provided by Cedars-Sinai Medical Center on 09/23/19. For personal use only.
`Annu. Rev. Immunol. 2004.22:329-360. Downloaded from www.annualreviews.org
`
`Genome Ex. 1046
`Page 10 of 37
`
`

`

`11 Feb 2004 18:54
`
`AR
`
`AR210-IY22-11.tex AR210-IY22-11.sgm
`
`LaTeX2e(2002/01/18)
`
`P1: IKH
`
`CANCER IMMUNOEDITING
`
`339
`
`Transplant Recipients Display Increased Incidences
`of Malignancies
`Increased relative risk ratios have indeed been observed in immunosuppressed
`transplant recipients for a broad subset of tumors that have no apparent viral ori-
`gin. Assessment of 5692 renal transplant patients from 1964–1982 in Finland,
`Denmark, Norway, and Sweden showed increased standardized cancer incidence
`ratios for colon, lung, bladder, kidney, ureter, and endocrine tumors compared to
`the general population (67). For example, the relative risks for colon cancer were
`3.2 for men and 3.9 for women. In addition, analysis of 925 patients who received
`cadaveric renal transplants from 1965 to