Cancer Immunoediting: From Immuno-Surveillance To Tumor Escape

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

R EVIEW

Cancer immunoediting: from immuno-


surveillance to tumor escape
Gavin P. Dunn1, Allen T. Bruce1, Hiroaki Ikeda1, Lloyd J. Old2 and Robert D. Schreiber1
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

The concept that the immune system can rec- The cancer immunosurveillance hypothesis
ognize and destroy nascent transformed cells Paul Ehrlich was one of the first to conceive the idea that the immune
system could repress a potentially “overwhelming frequency” of car-
was originally embodied in the cancer immuno-
cinomas1. The idea of immune control of neoplastic disease was not
surveillance hypothesis of Burnet and Thomas. vigorously pursued, however, until the midpoint of the twentieth cen-
This hypothesis was abandoned shortly after- tury. To a large extent, the revisiting of the Ehrlich proposal had to
wards because of the absence of strong experi- await the maturation of the developing field of immunology. In the
1950s, the work of Medawar and colleagues clarified the critical role
mental evidence supporting the concept. New for cellular components of immunity in mediating allograft rejection2.
data, however, clearly show the existence of can- This work cast doubt on data that were used to argue for the existence
cer immunosurveillance and also indicate that of tumor antigens. Specifically, although it was generally accepted that
it may function as a component of a more gen- the immune response was capable of recognizing and destroying trans-
planted tumors derived from noninbred strains of mice, it soon became
eral process of cancer immunoediting. This clear that the underlying mechanism was one of allograft rejection
process is responsible for both eliminating rather than tumor-specific rejection. With the availability of inbred
tumors and sculpting the immunogenic pheno- strains of mice, the idea that tumors were immunologically distin-
guishable from normal cells could be critically tested. The demonstra-
types of tumors that eventually form in
tion that mice could be immunized against syngeneic transplants of
immunocompetent hosts. In this review, we will tumors induced by chemical carcinogens, viruses or other means
summarize the historical and experimental established the existence of “tumor-specific antigens”3,4 and provided
basis of cancer immunoediting and discuss its a key cornerstone of the cancer immunosurveillance hypothesis.
Clearly, there could be no tumor immunosurveillance if there were no
dual roles in promoting host protection against distinctive structures on tumor cells that could be recognized by the
cancer and facilitating tumor escape from immune system.
immune destruction. These emerging discoveries were incorporated into the formal
hypothesis of “cancer immunosurveillance” proposed by Sir
Since the formalized introduction of the cancer immunosurveillance Macfarlane Burnet and Lewis Thomas. In 1957, Burnet stated5:
concept, the idea that the immune system may have a protective role in
tumor development has been hotly debated. Recent work, however, has It is by no means inconceivable that small accumulations of tumour
lent new support to the idea’s central principle that the immune system cells may develop and because of their possession of new antigenic
can indeed prevent tumor formation. At the same time, this work has potentialities provoke an effective immunological reaction with regres-
shown that the immune system also functions to promote or select sion of the tumour and no clinical hint of its existence.
tumor variants with reduced immunogenicity, thereby providing devel-
oping tumors with a mechanism to escape immunologic detection and At about the same time, Thomas suggested that the primary function
elimination. These findings have led to the development of the cancer of cellular immunity was in fact not to promote allograft rejection but
immunoediting hypothesis, a refinement of cancer immunosurveillance rather to protect from neoplastic disease, thereby maintaining tissue
that takes a broader view of immune system–tumor interactions by homeostasis in complex multicellular organisms6. These speculations
acknowledging both the host-protecting and tumor-sculpting actions of formed the evolutionary framework that ultimately resulted in the devel-
the immune system on developing tumors. Here we discuss first the his- opment of the immunosurveillance concept, which was defined by
tory that led to formulation of the original cancer immunosurveillance Burnet as follows7,8:
concept and then review the data that prompted the wide-scale aban-
donment of the hypothesis. We continue with a summary of the data In large, long-lived animals, like most of the warm-blooded vertebrates,
underlying the development of the cancer immunoediting concept and inheritable genetic changes must be common in somatic cells and a pro-
present a unifying model that proposes the molecular and cellular portion of these changes will represent a step toward malignancy. It is
dynamics of the process. We conclude by highlighting the critical an evolutionary necessity that there should be some mechanism for
implications that immunoediting may have on the development and eliminating or inactivating such potentially dangerous mutant cells and
treatment of cancer in humans. it is postulated that this mechanism is of immunological character.

1
Department of Pathology and Immunology, Center for Immunology,Washington University School of Medicine, 660 South Euclid Avenue, St. Louis, MO 63110, USA. 2Ludwig
Institute for Cancer Research, New York Branch at Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA. Correspondence should be addressed to R. D. S.
(schreiber@immunology.wustl.edu).

www.nature.com/natureimmunology • november 2002 • volume 3 no 11 • nature immunology 991


R EVIEW

Both Burnet and Thomas speculated that lymphocytes acted as sen- In hindsight, however, there are important caveats to these experi-
tinels in recognizing and eliminating continuously arising, nascent ments. First, it is now clear that nude mice are not completely immuno-
transformed cells7. compromised. Although nude mice possess fewer T cells than wild-
type mice, they have detectable populations of functional αβ T cell
Challenging the immunosurveillance hypothesis receptor–bearing lymphocytes26–28. Therefore, it is not possible to pre-
The introduction of the immunosurveillance hypothesis was rapidly dict the functional effect of these T cells on tumor formation when
followed by many experiments aimed at testing its logical predic- compared with wild-type controls. Second, the strain of mice used in
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

tions—namely, that hosts with impaired immune systems would exhib- the Stutman experiments may have been highly sensitive to MCA-
it increased incidences of spontaneous or chemically induced tumors. induced tumor formation. MCA requires biotransformation from its
The earliest approaches explored whether tumor development in mice pro-form into its carcinogenic form by the aryl hydrocarbon hydroxy-
was influenced by experimental immunosuppression of the host9,10. lase enzyme system. It is now known that various inbred mouse strains
Whereas several groups used neonatal thymectomy to induce immuno- produce distinctive patterns of these enzyme isoforms with different
suppression in mice, they could not reach a consensus as to the effects specific activities29. The CBA/H strain mouse used in the Stutman stud-
of this treatment on the incidence of either chemically induced10–13 or ies expresses a form of the enzyme with a high specific activity30, and
spontaneous tumors10,14,15. Other groups who compromised the immune thus cellular transformation induced by MCA might occur in these
system of mice using either heterologous anti-lymphocyte serum or mice at a rate that overwhelms host immunological defense mecha-
pharmacologic methods obtained similar discordant results10. It also nisms. Third, even though the study of Rygaard and Povlsen used large
became clear that mice with induced immunodeficiencies showed a numbers of nude mice, monitoring periods of 3–7 months were proba-
high susceptibility to virally induced tumors and a greater tendency to bly too short to see spontaneous tumor formation in the face of fully
develop spontaneous lymphomas compared with immunocompetent functional intrinsic tumor suppressor systems (such as p53). Last, these
mice. The prevailing view about these observations, however, was that studies were carried out before the discovery of other lymphocyte pop-
they reflected the greater susceptibility of immunocompromised hosts ulations such as natural killer (NK) cells, which are not thymus depen-
to infectious agents, such as transforming viruses. The greater frequen- dent, and γδ T cells, a subset of which may develop extrathymically31.
cy of lymphomas could also be ascribed to chronic antigenic stimula- The lack of convincing support for the immunosurveillance concept
tion from defective control of bacterial or viral infections, resulting in left room for other theories on the possible functions of immune cells
increased lymphocyte proliferation, somatic mutation and eventually during the development of neoplasia. For example, Prehn proposed
the formation of lymphomas10,16. As a group, these studies were incon- that the immune system can actually promote the growth of tumors. In
clusive and therefore failed either to prove or disprove the immunosur- this “immunostimulation theory”32, the presence of tumor rejection
veillance hypothesis. antigens (TRAs) on tumors is explained by the argument that the
The identification and characterization of the athymic nude immune cells that recognize these TRAs provide a positive growth sig-
mouse17,18 enabled researchers to study tumor formation in a host with nal to the tumor cells. Although this theory was speculative rather than
a genetic immunologic impairment. The work of Osias Stutman repre- evidence based, it reflected the mood that the proponents of the
sents one of the most extensive uses of nude mice to explore whether immunosurveillance theory were overstating their case. Especially
immunosurveillance occurs in a physiologic setting. Stutman found after the Stutman experiments, enthusiasm for the original immuno-
that CBA/H strain nude mice did not form more chemically induced surveillance concept began to wane. Thomas later noted, “The great-
tumors compared with their wild-type counterparts, nor did they show est trouble with the idea of immunosurveillance is that it cannot be
a shortened tumor latency period after carcinogen injection. For exam- shown to exist in experimental animals”33.
ple, in one experiment, nude mice or immunocompetent mice that were During the ensuing years, little interest was paid to the possibility that
heterozygous for the nude mutation were injected subcutaneously with the immune system could prevent the development of nonvirally
0.1 mg of the chemical carcinogen methylcholanthrene (MCA) at birth induced tumors. By all intents and purposes, the cancer immunosurveil-
and were monitored for tumor incidence19. After 120 days, seven of 39 lance concept was considered dead by 1978, and the field of tumor
control mice formed tumors at the injection site with a mean time to immunology moved on to study other issues such as the definition and
tumor appearance of 95 days. Of the nude mice tested, five of 27 molecular nature of mouse and human tumor antigens, the immune
formed tumors with a mean time to tumor appearance of 90 days. The response to known tumor antigens and the development of immunother-
similarity between immunocompetent and nude mice was maintained apeutic strategies to treat cancer. This view relegated immunosurveil-
in subsequent experiments that employed mice of different ages, dif- lance to the historical dust bin and was clearly echoed in a major review
ferent doses of carcinogen and even in experiments where the observa- that appeared in 2000, which listed the six critical hurdles that a devel-
tion period was extended out to 420 days20,21. At least one other group oping tumor must circumvent to grow and survive34. No significant men-
corroborated these results22. In addition, Stutman showed that there tion was made of the natural immune response against tumors.
were no statistically significant differences in the incidence of sponta-
neous nonviral tumor formation between nude and wild-type mice23. The renaissance of cancer immunosurveillance
These findings were supported by a study of Rygaard and Povlsen that Between the mid-1970s and 1990s, researchers made several experi-
showed no differences in spontaneous tumor formation in 10,800 nude mental attempts to resurrect the cancer immunosurveillance concept.
mice over a study period of 3–7 months24,25. The results using nude mice The discovery of NK cells led to a considerable amount of enthusiasm
were thus more conclusive than those based on experimental immuno- over the possibility that they functioned as the effectors of immunosur-
suppression and failed to uphold the central predictions of the immuno- veillance35. This enthusiasm was dampened, however, when a precise
surveillance concept. Based on the limited understanding of the definition and understanding of these cells was difficult to obtain.
immunologic defects in the nude mouse available at that time, these Subsequently, others repeated the MCA induction experiments of
results were highly convincing and thus led to the abandonment of the Stutman using nude mice on a BALB/c background36. When they inject-
immunosurveillance hypothesis. ed nude and control mice with different doses of MCA and monitored

992 nature immunology • volume 3 no 11 • november 2002 • www.nature.com/natureimmunology


R EVIEW

them for tumor development, nude mice formed more tumors than con- The definitive work supporting the existence of a cancer immuno-
trols. These data were limited, however, in their statistical power surveillance process that is dependent on both IFN-γ and lympho-
because of the small experimental group sizes and the magnitude of the cytes came through the use of gene-targeted mice that lack recom-
differences. Similar suggestive results were obtained when tumor for- bination activating gene 1 (RAG-1) or RAG-2. Like DNA-PK, these
mation induced by MCA was compared between groups of wild-type enzymes are involved in the repair of double-stranded DNA breaks,
BALB/c mice and immunodeficient CB-17 severe-combined immunod- but unlike DNA-PK, they are expressed exclusively in the lymphoid
eficiency (SCID) mice37. The latter lack a functionally active subunit of compartment. Mice deficient in either of these genes fail to
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

the DNA-dependant protein kinase (DNA-PK) enzyme (DNA-PKcs) rearrange lymphocyte antigen receptors and thus completely lack
that is required to generate rearranged antigen receptors in lympho- natural killer T (NKT), T and B cells48. In contrast, DNA repair
cytes38, and thus they are unable to develop antigen-specific immune mechanisms in the nonlymphoid compartment of these mice are
responses. However, DNA-PKcs (and the intact DNA-PK enzyme) is completely normal.
normally expressed in all cells and is important in repairing double- For the first time, inbred strains of mice were available that carried
stranded DNA breaks39, which includes those that can potentially be genetically defined mutations affecting specific tissues that led to an
transforming. Thus, it was not possible to interpret the basis for the absence of lymphocytes bearing antigen receptors. These mice enabled
increased tumor incidence observed in SCID mice because of an inabil- researchers to carry out carcinogenesis experiments that could unequiv-
ity to differentiate between tumor suppressor roles for the immune sys- ocally be interpreted. After MCA injection, 129/SvEv RAG-2–/– mice
tem or for the DNA-PK enzyme itself. developed sarcomas more rapidly and with greater frequency than
Between 1994 and 1998, two key findings incited renewed interest genetically matched wild-type controls49. After 160 days, 30 of 52
in the process of cancer immunosurveillance. First, endogenously pro- RAG-2–/– mice formed tumors, compared with 11 of 57 wild-type mice.
duced interferon γ (IFN-γ) was shown to protect the host against the In addition, RAG-2–/– mice aged in a specific pathogen free mouse facil-
growth of transplanted tumors and the formation of primary chemical- ity formed far more spontaneous epithelial tumors than did wild-type
ly induced and spontaneous tumors. Using tumor transplantation mice housed in the same room49 (and unpublished data). Specifically,
approaches, researchers found that immunogenic fibrosarcomas grow 26 of 26 RAG-2–/– mice aged 13–24 months developed spontaneous
faster and more efficiently in mice treated with neutralizing mono- neoplasia, predominantly of the intestine; eight of these mice had pre-
clonal antibodies specific for IFN-γ40. Overexpression of a dominant- malignant intestinal adenomas, 17 had intestinal adenocarcinomas and
negative mutant of the IFN-γ receptor α subunit (IFNGR1) in sarco- one had both an intestinal adenoma and a lung adenocarcinoma. In con-
mas such as Meth A (BALB/c derived) or MCA-207 (C57BL/6 trast, only five of 20 wild-type mice aged 13–24 months developed
derived) completely ablated tumor sensitivity to this cytokine, and the spontaneous neoplasia, which was predominantly benign. Three of the
tumors displayed enhanced tumorigenicity and reduced immunogenic- wild-type mice developed adenomas of the Harderian gland, lung and
ity when transplanted into naïve syngeneic hosts40. Experiments based intestine, respectively, whereas the other two developed an adenocarci-
on models of MCA-induced tumor formation showed that 129/SvEv noma of the Harderian gland and an endometrial stromal carcinoma.
mice that were lacking either the IFN-γ receptor or signal transducer Thus, lymphocytes in a mouse not only protect the host against forma-
and activator of transcription 1 (STAT1, the transcription factor that is tion of primary sarcomas that are chemically induced but also prevent
important in mediating IFN-γ receptor signaling) were found to be the development of spontaneous epithelial tumors.
approximately 10–20 times more sensitive than wild-type mice to the The overlap between the tumor suppressor pathways that depend on
tumor-inducing capacity of MCA, developed more tumors than their IFN-γ and lymphocytes was defined by comparing tumor formation in
wild-type counterparts and showed a shortened tumor latency period41. 129/SvEv mice lacking IFN-γ responsiveness (IFNGR1 receptor–/– or
Subsequent independent experiments using mice on a different genet- STAT1–/– mice), lymphocytes (RAG-2–/– mice) or both RAG-2 and
ic background that lacked the gene encoding IFN-γ confirmed these STAT1 (RkSk mice)49. Each of the four lines of gene-targeted mice
results42. Similarly, mice lacking the genes encoding the tumor sup- formed three times more chemically induced tumors than syngeneic
pressor p53 and the IFNGR1 subunit of the IFN-γ receptor formed a wild-type mice when injected with 0.1 mg of MCA. Because no sig-
wider spectrum of tumors as compared with IFN-γ-sensitive mice nificant differences were detected between any of the groups of defi-
lacking only p5341. In addition, IFN-γ–/– mice on a C57BL/6 back- cient mice, the researchers concluded that the two extrinsic tumor sup-
ground showed an increased incidence of disseminated lymphomas pressor mechanisms heavily overlapped. RkSk mice, however, addi-
despite the presence of a normal p53 tumor suppressor gene, and IFN- tionally developed spontaneous breast tumors that were not observed in
γ–/– mice on a BALB/c background showed a low incidence of lung wild-type or RAG-2–/– mice, therefore demonstrating that the overlap
adenocarcinomas43. between the two tumor suppressor pathways was not complete. Similar
The second key finding was the observation that C57BL/6 mice lack- findings were made in carcinogenesis experiments with mice that
ing perforin (perforin–/–) were more prone to MCA-induced tumor for- lacked perforin, IFN-γ or both, in which a small increase was observed
mation compared with their wild-type counterparts. Perforin is a com- in tumor induction in the doubly deficient mice compared with mice
ponent of the cytolytic granules of cytotoxic T cells and NK cells that lacking only one of the two components42.
is important in mediating lymphocyte-dependent killing of many dif- Additional studies have used other inbred mouse lines with targeted
ferent target cells including tumor cells44. After challenge with MCA, disruptions in genes encoding critical components of the immune sys-
perforin–/– mice developed significantly more tumors compared with tem. They not only support the importance of immune system control of
perforin-sufficient mice treated in the same manner42,45,46. Untreated tumor formation but also suggest the involvement of both the innate and
perforin–/– mice also showed a high incidence of spontaneous dissemi- adaptive immune compartments in cancer immunosurveillance.
nated lymphomas, which was greater on a p53+/– background47, and a Specifically, genetic, immunochemical or functional ablations of NKT,
low incidence of spontaneous lung adenocarcinomas43. Taken together, γδ T cells, NK cells, αβ T cells, IFN-γ or interleukin 12 (IL-12) all lead
these observations showed that components of the immune system to increased susceptibility of the host to tumors (Table 1). One of these
were involved in controlling primary tumor development. studies has shed light on the specific subsets of lymphocytes expressing

www.nature.com/natureimmunology • november 2002 • volume 3 no 11 • nature immunology 993


R EVIEW

T cell receptors (TCR) that are important in tumor surveillance. The rel- concept as originally envisaged by Burnet and Thomas—namely, that
ative contributions of αβ and γδ T cells in blocking primary tumor for- the unmanipulated immune system is capable of recognizing and elimi-
mation were explored in αβ T cell–/– (lacking the TCR β-chain) or γδ T nating primary tumors and that lymphocytes and the cytokines they pro-
cell–/– (lacking the TCR δ-chain) mice50. MCA treatment of either type duce are important in this process.
of mouse increased the incidence of fibrosarcomas and spindle cell car-
cinomas as compared with wild-type controls. These data showed that Cancer immunosurveillance in humans
both T cell subsets are critical for protecting the host in this particular If indeed cancer immunosurveillance exists in mice, does it exist in
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

model of tumor development. In an initiation and promotion model of humans? The logical prediction from the immunosurveillance hypoth-
skin tumorigenesis induced by 7,12-dimethylbenzanthracene (DMBA) esis is that immunodeficient or immunosuppressed humans should
and 12-O-tetradecanoylphorbol-13-acetate (TPA), however, γδ T cell–/– show greater incidences of cancer. Early follow-up studies of transplant
mice showed a greater susceptibility to tumor formation and a higher patients who were immunosuppressed51 and individuals with primary
incidence of papilloma-to-carcinoma progression than wild-type mice, immunodeficiencies52 showed that they had a significantly higher rela-
whereas αβ T cell–/– mice did not. This result suggests that immunosur- tive risk for cancer development. Based on long-term studies of
veillance may be a heterogeneous process requiring the actions of dif- patients, it is clear that some of this higher risk was due to the devel-
ferent immune effectors in a manner that is dependent on the tumor’s opment of tumors of viral origin53. For example, in the transplant reg-
cell type of origin, mechanism of transformation, anatomical localiza- istries from Cincinnatti53, Scandinavia54, and Australia and New
tion and mechanism of immunologic recognition. In sum, the large Zealand55, particularly high incidence ratios have been observed for
amount of recent data obtained by many independent groups over- non-Hodgkin’s lymphoma, Kaposi’s sarcoma and carcinomas of the
whelmingly supports the basic tenets of the cancer immunosurveillance genitourinary and anogenital regions. Many of these cancers have a
viral etiology and are linked to infection
with Epstein-Barr virus, human her-
Table 1. Enhanced susceptibility of immunodeficient mice to formation of chemically pesvirus 8 and human papilloma virus,
induced and spontaneous tumors respectively56. These malignancies are
now occurring with increasing frequency
Phenotype or depletion Immunodeficiency Tumor susceptibility in AIDS patients56. The increased inci-
dence of virally induced tumors in
RAG-2–/– T, B and NKT cells MCA-induced sarcomas49
Spontaneous intestinal neoplasia49 immunocompromised patients represents
RAG-2 × STAT1
–/– –/–
T, B and NKT cells MCA-induced sarcomas 49 one immunosurveillance function that is
(RkSk) Insensitive to IFN-γ and IFN-α/β Spontaneous intestinal and not contested: natural protection against
mammary neoplasia49 infectious organisms. These data, howev-
BALB/c SCID T, B and NKT cells MCA-induced sarcomas100 er, do not invalidate the possibility that the
Perforin–/– Lack of perforin MCA-induced sarcomas42,45,46 immune system protects humans against
Spontaneous disseminated development of tumors of nonviral origin.
lymphomas43,47 Clearly, the difficulty in assessing can-
TCR Jα281–/– Subset of NKT cells MCA-induced sarcomas42,46,100 cer immunosurveillance in immunodefi-
Anti-asialo-GM1 antibody NK cells and activated macrophages MCA-induced sarcomas100 cient humans arises from their greater sus-
Anti-NK1.1 antibody NK and NKT cells MCA-induced sarcomas 46,100 ceptibility to endemic viruses and other
Anti-Thy1 antibody T cells MCA-induced sarcomas46,100
pathogens as well as viral reactivation.
This can confound the likelihood of find-
αβ T cell–/– αβ T cells MCA-induced sarcomas50
ing spontaneous tumors of nonviral origin,
γδ T cell–/– γδ T cells MCA-induced sarcomas50
which develop slowly because of the
DMBA/TPA-induced skin tumors50
shortened life spans and other intercurrent
STAT1–/– Insensitive to IFN-γ and IFN-α/β MCA-induced sarcomas41,49
Wider tumor spectrum in
medical problems of these patients.
STAT1–/– × p53–/– (ref. 41) Greater relative risk ratios have, however,
IFNGR1 receptor –/–
Insensitive to IFN-γ MCA-induced sarcomas 41,49 been observed for a broad subset of
Wider tumor spectrum in tumors with no apparent viral etiology.
IFN-γ receptor–/– × p53–/– (ref. 41) Approximately fourfold increases in the
IFN-γ–/– Lack of IFN-γ MCA-induced sarcomas42 incidence of de novo malignant melanoma
C57BL/6: Spontaneous disseminated after organ transplantation have been
lymphomas43 reported57,58. A review of data accumulated
BALB/c: Spontaneous lung by the Cincinnati Transplant Tumor
adenocarcinoma43 Registry from 1968 to 1995 found a
Perforin–/– × IFN-γ–/– Lack of perforin and IFN-γ MCA-induced sarcomas42 twofold greater risk in transplant patients
Spontaneous disseminated
43
for developing melanoma over that of the
lymphomas
–/– 46
general population59. In addition, whereas
IL-12 Lack of IL-12 MCA-induced sarcomas
only 0.3–0.4% of melanomas occur in the
WT + IL-12 Exogenous IL-12 Lower incidence of MCA-induced general pediatric population, the occur-
101
sarcomas
rence in pediatric transplant patients was
4%59. The same database showed that
Methylcholanthrene-treated wild-type (WT) mice were treated with IL-12 during tumor formation.
transplant patients were three times more

994 nature immunology • volume 3 no 11 • november 2002 • www.nature.com/natureimmunology


R EVIEW

likely to develop non-Kaposi’s


sarcomas60. When neoplasia
a b c
occurrence was assessed in 608
cardiac transplant patients at the
University of Pittsburgh be-
tween 1980 and 1993, the preva-
lence of lung tumors was 25-
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

fold higher than in the general


population61. In Australia and
New Zealand, the tracking of
925 patients who received
cadaveric renal transplants from
1965 to 1998 showed increased
risk ratios for development of
colon, pancreatic, lung and
endocrine tumors as well as
malignant melanomas55. In addi-
tion, assessment of 5,692 renal Figure 1. The three Es of cancer immunoediting. Cancer immunoediting encompasses three process. (a) Elimination
transplant patients from 1964 to corresponds to immunosurveillance. (b) Equilibrium represents the process by which the immune system iteratively selects
1982 in Finland, Denmark, and/or promotes the generation of tumor cell variants with increasing capacities to survive immune attack. (c) Escape is the
process wherein the immunologically sculpted tumor expands in an uncontrolled manner in the immunocompetent host. In a
Norway and Sweden showed and b, developing tumor cells (blue), tumor cell variants (red) and underlying stroma and nontransformed cells (gray) are shown;
higher standardized cancer inci- in c, additional tumor variants (orange) that have formed as a result of the equilibrium process are shown. Different lympho-
dence ratios for colon, lung, cyte populations are as marked.The small orange circles represent cytokines and the white flashes represent cytotoxic activi-
bladder, kidney, ureter and ty of lymphocytes against tumor cells.
endocrine tumors as well as
malignant melanomas as com-
pared with the general population54. Thus, individuals with severe that the immune system functions during tumor formation to select for
deficits of immunity indeed have a higher probability of developing a tumor variants that are better suited to survive in an immunologically
variety of cancers with no known viral etiology. intact environment, very much like it does with viruses, bacteria and
In addition to the supporting epidemiological data described above, parasites. Many studies have shown the immunoselective effects of
there is accumulating evidence showing a positive correlation between repassage of transplantable tumors through immunocompetent hosts
the presence of lymphocytes in a tumor and increased patient survival. and the generation of tumor variants with reduced immunogenicity.
Some of the most convincing evidence comes from the study of cuta- Two particularly informative studies used P815 mastocytomas73 and
neous melanomas. Three categories were established for tumor lym- ultraviolet-induced 1591 fibrosarcomas74. In contrast, few studies have
phocyte infiltration during the vertical growth phase of cutaneous compared the immunogenic characteristics of tumors originally gener-
melanoma (brisk, nonbrisk and absent)62. Sorting more than 500 ated in the presence or absence of a functional immune system. This
patients with primary melanoma who had 5-, 8- or 10-year follow-ups issue was difficult to study in the past because matched sets of tumors
into these categories and comparing their survival statistics showed that derived from isogenic immunodeficient and control immunocompetent
patients in the brisk tumor infiltrating lymphocyte (TIL) response cate- hosts were not generally available. A recent study, however, was carried
gory survived one and one-half to three times longer than patients in the out using many MCA-induced sarcomas obtained from either RAG-2–/–
absent TIL response group; patients in the nonbrisk response group had or wild-type mice on a pure 129/SvEv background49. When tumors iso-
intermediate survival times62,63. Researchers obtained the same prog- lated from wild-type or RAG-2–/– mice were transplanted into RAG-2–/–
nostic correlation when the presence of TILs in melanomas that had recipients, they all grew with similar kinetics, which indicates that there
metastasized to the lymph nodes was used as the criterion64. Similar were no inherent growth differences between tumors raised in the pres-
correlations between the presence of TILs and patient survival have ence or absence of an intact immune system. In addition, 17 of 17 sar-
also been made that involved more than 3,400 patients with cancer of comas originally isolated from wild-type mice were capable of estab-
the breast65, bladder66, colon67,68, prostate69, ovary70 or rectum71 and for lishing progressively growing tumors when transplanted into naïve
neuroblastoma72. In some cases, the correlation has been refined to immunocompetent 129/SvEv hosts. In contrast, 8 of 20 tumors origi-
show that CD8+ T cells are the relevant lymphocyte population that nally generated in RAG-2–/– mice were rejected when transplanted into
affect survival68. immunocompetent hosts, even when they were injected at a high cell
In sum, the data obtained from both mouse and human studies pro- number. Thus, tumors formed in the absence of an intact immune sys-
vide strong support for the existence and physiologic relevance of can- tem are, as a group, more immunogenic than tumors that arise in
cer immunosurveillance. These findings have rekindled generalized immunocompetent hosts.
interest in this process and have stimulated much work aimed at defin- Other more limited experiments have reached similar conclusions.
ing its immunologic components. Thus, the new millennium has wit- MCA-induced sarcomas derived from nude75 or SCID mice37 were
nessed the resurrection of this old and formerly controversial idea. rejected more frequently than similar tumors derived from wild-type
mice when transplanted into wild-type hosts. In addition, two MCA-
Immunologic sculpting during tumor development induced sarcomas derived from TCR Jα281–/– mice, which lack a major
Given the existence of cancer immunosurveillance, why then do can- population of NKT cells, grew more slowly when transplanted into
cers occur in immunocompetent individuals? It has long been thought wild-type hosts than did sarcomas originally isolated from wild-type

www.nature.com/natureimmunology • november 2002 • volume 3 no 11 • nature immunology 995


R EVIEW

mice46. In contrast, these tumors grew in a comparable manner when most tumors that are clinically apparent has already been modified to
transplanted into TCR Jα281–/– recipients. Finally, lymphomas derived some degree by their interactions with the immune system.
from perforin–/– mice grew avidly when transplanted into perforin–/–
mice, but most were rejected when transplanted into wild-type mice43. Cancer immunoediting
Taken together, these results show that tumors are imprinted by the Because the immune system exerts both host-protecting and tumor-
immunologic environment in which they form. This imprinting process sculpting effects on developing tumors, the term cancer immunosur-
can often result in the generation of tumors that are better able to with- veillance may no longer be appropriate to accurately describe the
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

stand the tumor-suppressing actions of the immune system by elimi- process because, in its original form, it was thought to function only to
nating tumor cells of intrinsically high immunogenicity but leaving protect the host and to act only at the earliest stages of cellular trans-
behind tumor variants of reduced immunogenicity (or that have formation. Rather, we have proposed the use of the broader term “can-
acquired other mechanisms to evade or suppress immune attack) that cer immunoediting” to describe more appropriately the dual host-pro-
have a better chance of surviving in the immunocompetent host. The tecting and tumor-sculpting actions of the immune system that not only
alterations that must occur during the immunologic sculpting of a prevent but also shape neoplastic disease. We envisage the scope of this
developing tumor are probably facilitated by the inherent genetic insta- process to be very broad such that it can promote complete elimination
bility of tumors76. Some of the likely targets of the immunologic of some tumors, generate a nonprotective immune state to others or
process that sculpts tumors are genes encoding tumor antigens, com- favor the development of immunologic anergy, tolerance or indiffer-
ponents of the major histocompatibility complex (MHC) pathways that ence. We envision important roles for components of both the innate
process and present antigens or components of the IFN-γ receptor sig- and adaptive immune systems in this process.
naling pathway. Tumors lacking antigen processing and presentation Much work is needed to define the molecular and cellular dynamics
components are commonly found (see Review by Khong and Restifo in of cancer immunoediting. A proposed basic framework for this process
this issue), and tumors have been identified that lack functional expres- is illustrated based on recent data from several groups (Figs. 1 and 2).
sion of at least three components of the IFN-γ receptor signaling path- Although it was not possible to discuss in detail the contributions from
way41 (unpublished data). It is likely that immunologic sculpting of each laboratory that were used in the construction of these models
tumors occurs continuously, but the major effects of this process are because of space limitations, we have incorporated key references in
most prominent early when the tumor is perhaps histologically—but the model descriptions to acknowledge the work of these groups. The
not clinically—detectable. It follows, then, that the immunogenicity of models are presented in the hope that they will stimulate additional

Figure 2. A proposed model a b


for the elimination phase of
the cancer immunoediting
process. (a) The initiation of
the response in which lympho-
cytes that participate in innate
immunity (NKT, NK and γδ T
cells) recognize transformed
cells that have accumulated
above a threshold that has yet
to be defined and are stimulated
to produce IFN-γ. (b) The initial
IFN-γ starts a cascade of innate
immune reactions that involve
(i) the induction of chemokines,
including the angiostatic chemo-
kines (CXCL10 (1P10), CXCL9
(MIG) AND CXCL11 (I-TAC))
c d
that block neovascularization in
the tumor and that also effect
the recruitment of NK cells,
dendritic cells, macrophages and
other immune effector cells to
the tumor site; (ii) an antiprolif-
erative action of IFN-γ on the
developing tumor and (iii) the
activation of cytocidal activity in
macrophages and NK cells
entering the tumor. These
events result in some tumor cell
death by both immunologic and
nonimmunologic mechanisms.
Dead tumor cells or tumor cell
debris (blue squares) are ingest-
ed by dendritic cells and are trafficked to the draining lymph node. (c) Tumor growth is kept in check by the cytocidal activities of NK cells and activated macrophages while
CD4+ and CD8+ T cells that are specific for tumor antigens develop in the draining lymph node. (d) Tumor-specific CD4+ and CD8+ T cells home to the tumor along a
chemokine gradient where they recognize and destroy tumor cells expressing distinctive tumor antigens.Tumor cells (blue); nontransformed cells (gray); dead tumor cells
(white to gray gradient circles surrounded by a dashed black line); lymphocytes, dendritic cells (DC) and macrophages (Mac) are marked and colored appropriately.

996 nature immunology • volume 3 no 11 • november 2002 • www.nature.com/natureimmunology


R EVIEW

research aimed at identifying the pathways that lead from cancer through genetic or epigenetic changes begin to expand in an uncontrolled
immunosurveillance to tumor escape. manner. This results in clinically observable malignant disease that, if left
We envisage cancer immunoediting as a result of three processes: unchecked, results in the death of the host.
elimination, equilibrium and escape. We call these the three Es of can-
cer immunoediting (Fig. 1). Immunosurveillance occurs during the Conclusions and implications
elimination process, whereas the Darwinian selection of tumor variants We have summarized here the events that led to the development of the
occurs during the equilibrium process. This in turn can ultimately lead concept of cancer immunoediting and described the evolution of the con-
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

to escape and the appearance of clinically apparent tumors. cept from its birth in immunosurveillance to its present form. The origi-
The elimination process encompasses the original concept of cancer nal formulation of the cancer immunosurveillance hypothesis by Burnet
immunosurveillance (Figs. 1a and 2). As such, when it is successful in and Thomas was indeed as prescient as it was powerful. Although it was
deleting the developing tumor, it represents the complete editing process based on an emerging understanding of the existence of tumor antigens
without progression to the subsequent phases. In the first phase of elim- and the laws of transplantation immunity, it came at a time when not
ination (Fig. 2a), once solid tumors reach a certain size, they begin to enough was understood about mouse models of immunodeficiency. This
grow invasively and require an enhanced blood supply that arises as a led to the unwitting use of imperfect animal models to critically test and
consequence of the production of stromagenic and angiogenic pro- prematurely abandon the concept. The concept was resurrected nearly
teins77. Invasive growth causes minor disruptions within the surrounding three decades later by the use of genetically defined mouse models of
tissue that induce inflammatory signals leading to recruitment of cells of complete immunodeficiency coupled with an enhanced understanding of
the innate immune system (NKT, NK, γδ Τ cells, macrophages and den- the molecular nature of tumor antigens. The incorporation of immuno-
dritic cells) into the site50,78,79. Structures on the transformed cells (either surveillance into the cancer immunoediting concept came about by the
expressed as a result of the transformation process itself or induced by finding that the immune system not only protects the host against tumor
the ongoing but limited inflammatory response) are recognized by infil- development but also can sculpt the immunogenic phenotype of a devel-
trating lymphocytes such as NKT, NK or γδ T cells, which are then stim- oping tumor. Thus, we now recognize a process that has both positive and
ulated to produce IFN-γ80–82. In the second phase (Fig. 2b), the IFN-γ negative effects on host–tumor relationships. Clearly, more work is need-
that was initially produced may induce a limited amount of tumor death ed to define the process on a molecular and cellular basis.
by means of antiproliferative83 and apopotic84 mechanisms. However, it There are four immediate implications of the cancer immunoediting
also induces the production of the chemokines CXCL10 (interferon- concept that directly apply to human cancer. First, we may need to
inducible protein-10, IP-10), CXCL9 (monokine induced by IFN-γ, examine whether immunity plays a role in the development of human
MIG) and CXCL11 (interferon-inducible T cell α chemoattractant, I- cancers associated with known abnormalities in genes encoding such
TAC) from the tumor cells themselves as well as from surrounding nor- proteins as p53, BRCA1, APC (adenomatous polyposis coli) Ras and
mal host tissues85–87. At least some of these chemokines have potent others. Second, it will be important to re-examine the carcinogenic
angiostatic capacities and thus block the formation of new blood vessels potential of many compounds that have never been tested in experi-
within the tumor, which leads to even more tumor cell death88–91. Tumor mental animals that are immunodeficient and thus may have been
cell debris formed as either a direct or indirect consequence of IFN-γ incorrectly labeled as noncarcinogenic. Third, the relation between
production at the tumor is then ingested by local dendritic cells, which immunologic defects and increased incidences of cancer associated
home to draining lymph nodes. Chemokines produced during the esca- with aging needs to be explored. Last, we need a way to assess the
lating inflammatory process recruit more NK cells and macrophages to extent to which a tumor has been edited.
the site. In the third phase (Fig. 2c), the tumor-infiltrating NK cells and Perhaps one of the most important contributions of the original
macrophages transactivate one another by reciprocal production of IFN- immunosurveillance concept was the furious experimentation that fol-
γ and IL-12, and kill more of the tumor by mechanisms involving tumor lowed in its wake. In this respect, we hope for a similar fate for the can-
necrosis factor–related apoptosis-inducing ligand, perforin and reactive cer immunoediting concept—new oeuvres framed by the three Es that
oxygen and nitrogen intermediates46,92–95. In the draining lymph node, the test every prediction of this hypothesis and, in eliciting a greater under-
newly immigrated dendritic cells induce tumor-specific CD4+ T helper standing of host–tumor interactions, will ultimately have profound
cells expressing IFN-γ (TH1 cells) that in turn facilitate the development effects on our understanding and treatment of cancer.
of tumor-specific CD8+ T cells96–99. In the fourth phase (Fig. 2d), tumor-
Acknowledgments
specific CD4+ and CD8+ T cells home to the tumor site, where the
Supported by grants from the National Cancer Institute (CA43059 and CA76464 to R. D. S.),
cytolytic T lymphocytes destroy the remaining antigen-bearing tumor
the Cancer Research Institute (to R. D. S., H.I and A.B.), the Ludwig Institute for Cancer
cells whose immunogenicities have been enhanced by exposure to local- Research (to R. D. S.), and the National Institute of Allergy and Infectious Diseases (to R. D. S.
ly produced IFN-γ49. and G. P. D.).We thank V. Shankaran, K. Sheehan,A. Dighe, D. Kaplan, R. Uppaluri, C. Koebel, J.
In the equilibrium process (Fig. 1b), the host immune system and any Bui, E. Stockert, E. Richards, M.White, C.Arthur and C. Brendel for their important roles in
developing the cancer immunoediting concept and for helpful comments during the prepara-
tumor cell variant that has survived the elimination process enter into a tion of this manuscript.
dynamic equilibrium. In this process, lymphocytes and IFN-γ exert
potent selection pressure on the tumor cells that is enough to contain, but 1. Ehrlich, P. Ueber den jetzigen stand der Karzinomforschung. Ned.Tijdschr. Geneeskd. 5, 273–290 (1909).
not fully extinguish, a tumor bed containing many genetically unstable 2. Silverstein, A. M. A History of Immunology (Academic, San Diego, CA, 1989).
3. Old, L. J. & Boyse, E. A. Immunology of experimental tumors. Annu. Rev. Med. 15, 167–186 (1964).
and rapidly mutating tumor cells. During this period of Darwinian selec- 4. Klein, G.Tumor antigens. Annu. Rev. Microbiol. 20, 223–252 (1966).
tion, many of the original escape variants of the tumor cell are destroyed, 5. Burnet, F. M. Cancer—a biological approach. Brit. Med. J. 1, 841–847 (1957).
6. Thomas, L. in Cellular and Humoral Aspects of the Hypersensitive States (ed. Lawrence, H. S.) 529–532
but new variants arise carrying different mutations that provide them with (Hoeber-Harper, New York, 1959).
increased resistance to immune attack. It is likely that equilibrium is the 7. Burnet, F. M.The concept of immunological surveillance. Prog. Exp.Tumor Res. 13, 1–27 (1970).
8. Burnet, F. M. Immunological factors in the process of carcinogenesis. Br. Med. Bull. 20, 154–158 (1964).
longest of the three processes and may occur over a period of many years. 9. Kaplan, H. S. Role of immunologic disturbance in human oncogenesis: some facts and fancies. Br. J.
In the escape process (Fig. 1c), surviving tumor variants that have Cancer 25, 620–634 (1971).
10. Stutman, O. Immunodepression and malignancy. Adv. Cancer Res. 22, 261–422 (1975).
acquired insensitivity to immunologic detection and/or elimination 11. Grant, G. A. & Miller, J. F. Effect of neonatal thymectomy on the induction of sarcomata in C57BL

www.nature.com/natureimmunology • november 2002 • volume 3 no 11 • nature immunology 997


R EVIEW

mice. Nature 205, 1124–1125 (1965). 60. Penn, I. Sarcomas in organ allograft recipients. Transplantation 60, 1485–1491 (1995).
12. Nishizuka,Y., Nakakuki, K. & Usui, M. Enhancing effect of thymectomy on hepatotumorigenesis in 61. Pham, S. M. et al. Solid tumors after heart transplantation: lethality of lung cancer. Ann.Thorac. Surg.
Swiss mice following neonatal injection of 20-methylcholanthrene. Nature 205, 1236–1238 (1965). 60, 1623–1626 (1995).
13. Trainin, N., Linker-Israeli, M., Small, M. & Boiato-Chen, L. Enhancement of lung adenoma formation by 62. Clark,W. H. Jr et al. Model predicting survival in stage I melanoma based on tumor progression.
neonatal thymectomy in mice treated with 7,12-dimethylbenz(a)anthracene or urethan. Int. J. Cancer J. Natl. Cancer Inst. 81, 1893–1904 (1989).
2, 326–336 (1967). 63. Clemente, C. G. et al. Prognostic value of tumor infiltrating lymphocytes in the vertical growth phase
14. Burstein, N. A. & Law, L.W. Neonatal thymectomy and non-viral mammary tumours in mice. Nature of primary cutaneous melanoma. Cancer 77, 1303–1310 (1996).
231, 450–452 (1971). 64. Mihm, M. C. Jr, Clemente, C. G. & Cascinelli, N.Tumor infiltrating lymphocytes in lymph node
15. Sanford, B. H., Kohn, H. I., Daly, J. J. & Soo, S. F. Long-term spontaneous tumor incidence in neonatally melanoma metastases: a histopathologic prognostic indicator and an expression of local immune
thymectomized mice. J. Immunol. 110, 1437–1439 (1973). response. Lab. Invest. 74, 43–47 (1996).
© 2002 Nature Publishing Group http://www.nature.com/natureimmunology

16. Klein, G. Immunological surveillance against neoplasia. Harvey Lect., 71–102 (1973). 65. Rilke, F. et al. Prognostic significance of HER-2/neu expression in breast cancer and its relationship
17. Flanagan, S. P. ‘Nude’, a new hairless gene with pleiotropic effects in the mouse. Genet. Res. 8, to other prognostic factors. Int. J. Cancer 49, 44–49 (1991).
295–309 (1966). 66. Lipponen, P. K., Eskelinen, M. J., Jauhiainen, K., Harju, E. & Terho, R.Tumour infiltrating lymphocytes as
18. Pantelouris, E. M. Absence of thymus in a mouse mutant. Nature 217, 370–371 (1968). an independent prognostic factor in transitional cell bladder cancer. Eur. J. Cancer 29A, 69–75 (1992).
19. Stutman, O.Tumor development after 3-methylcholanthrene in immunologically deficient athymic- 67. Nacopoulou, L., Azaris, P., Papacharalampous, N. & Davaris, P. Prognostic significance of histologic
nude mice. Science 183, 534–536 (1974). host response in cancer of the large bowel. Cancer 47, 930–936 (1981).
20. Stutman, O. in Proceedings of the International Workshop on Nude Mice Vol. 1 (eds. Rygaard, J. & 68. Naito,Y. et al. CD8+ T cells infiltrated within cancer cell nests as a prognostic factor in human col-
Poulsen, C.) 257–264 (Gustav Fischer, Stuttgart 1973). orectal cancer. Cancer Res. 58, 3491–3494 (1998).
21. Stutman, O. Chemical carcinogenesis in nude mice: comparison between nude mice from homozy- 69. Epstein, N. A. & Fatti, L. P. Prostatic carcinoma: some morphological features affecting prognosis.
gous and heterozygous matings and effect of age and carcinogen dose. J. Natl. Cancer Inst. 2, Cancer 37, 2455–2465 (1976).
353–358 (1979). 70. Deligdisch, L., Jacobs, A. J. & Cohen, C. J. Histologic correlates of virulence in ovarian adenocarcino-
22. Outzen, H. C., Custer, R. P., Eaton, G. J. & Prehn, R.T. Spontaneous and induced tumor incidence in ma. II. Morphologic correlates of host response. Am. J. Obstet. Gynecol. 144, 885–889 (1982).
germfree “nude” mice. J. Reticuloendothel. Soc. 17, 1–9 (1975). 71. Jass, J. R. Lymphocytic infiltration and survival in rectal cancer. J. Clin. Pathol. 39, 585–589 (1986).
23. Stutman, O. in The Nude Mouse in Experimental and Clinical Research (eds. Fogh, J. & Giovanella, B. C.) 72. Palma, L., Di Lorenzo, N. & Guidetti, B. Lymphocytic infiltrates in primary glioblastomas and recidi-
411–435 (Academic, New York, 1978). vous gliomas. Incidence, fate, and relevance to prognosis in 228 operated cases. J. Neurosurg. 49,
24. Rygaard, J. & Povlsen, C. O. Is immunological surveillance not a cell-mediated immune function? 854–861 (1978).
Transplantation 17, 135–136 (1974). 73. Uyttenhove, C.,Van Snick, J. & Boon,T. Immunogenic variants obtained by mutagenesis of mouse
25. Rygaard, J. & Povlsen, C. O.The mouse mutant nude does not develop spontaneous tumours. An mastocytoma P815. I. Rejection by syngeneic mice. J. Exp. Med. 152, 1175–1183 (1980).
argument against immunological surveillance. Acta Pathol. Microbiol. Scand. [B] Microbiol. Immunol. 82, 74. Urban, J. L., Holland, J. M., Kripke, M. L. & Schreiber, H. Immunoselection of tumor cell variants by
99–106 (1974). mice suppressed with ultraviolet radiation. J. Exp. Med. 156, 1025–1041 (1982).
26. Maleckar, J. R. & Sherman, L. A.The composition of the T cell receptor repertoire in nude mice. 75. Svane, I. M. et al. Chemically induced sarcomas from nude mice are more immunogenic than similar
J. Immunol. 138, 3873–3876 (1987). sarcomas from congenic normal mice. Eur. J. Immunol. 26, 1844–1850 (1996).
27. Ikehara, S., Pahwa, R. N., Fernandes, G., Hansen, C.T. & Good, R. A. Functional T cells in athymic nude 76. Lengauer, C., Kinzler, K.W. & Vogelstein, B. Genetic instabilities in human cancers. Nature 396,
mice. Proc. Natl. Acad. Sci. USA 81, 886–888 (1984). 643–649 (1998).
28. Hunig,T.T-cell function and specificity in athymic mice. Immunol.Today 4, 84–87 (1983). 77. Hanahan, D. & Folkman, J. Patterns and emerging mechanisms of the angiogenic switch during
29. Heidelberger, C. Chemical carcinogenesis. Annu. Rev. Biochem. 44, 79–121 (1975). tumorigenesis. Cell 86, 353–364 (1996).
30. Kouri, R. E. & Nebert, D.W. in Origins of Human Cancer (eds. Hiatt, H. H.,Watson, J. D. & Winsten, 78. Smyth, M. J., Godfrey, D. I. & Trapani, J. A. A fresh look at tumor immunosurveillance and immunother-
J. A.) 811–835 (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1977). apy. Nature Immunol. 2, 293–299 (2001).
31. Hayday, A. C. γδ cells: a right time and a right place for a conserved third way of protection. Annu. 79. Matzinger, P.Tolerance, danger, and the extended family. Annu. Rev. Immunol. 12, 991–1045 (1994).
Rev. Immunol. 18, 975–1026 (2000). 80. Yokoyama,W. M. Now you see it, now you don’t! Nature Immunol. 1, 95–97 (2000).
32. Prehn, R.T. Perspectives on oncogenesis: does immunity stimulate or inhibit neoplasia? 81. Cerwenka, A. et al. Retinoic acid early inducible genes define a ligand family for the activating
J. Reticuloenothel. Soc. 10, 1–16 (1970). NKG2D receptor in mice. Immunity 12, 721–727 (2000).
33. Thomas, L. On immunosurveillance in human cancer. Yale J. Biol. Med. 55, 329–333 (1982). 82. Diefenbach, A., Jensen, E. R., Jamieson, A. M. & Raulet, D. H. Rae1 and H60 ligands of the NKG2D
34. Hanahan, D. & Weinberg, R. A.The hallmarks of cancer. Cell 100, 57–70 (2000). receptor stimulate tumour immunity. Nature 413, 165–171 (2001).
35. Herberman, R. B. & Holden, H.T. Natural cell-mediated immunity. Adv. Cancer Res. 27, 305–377 (1978). 83. Bromberg, J. F., Horvath, C. M.,Wen, Z., Schreiber, R. D. & Darnell, J. E. Jr.Transcriptionally active Stat1
36. Engel, A. M. et al. Methylcholanthrene-induced sarcomas in nude mice have short induction times and is required for the antiproliferative effects of both interferon α and interferon γ. Proc. Natl. Acad. Sci.
relatively low levels of surface MHC class I expression. APMIS 104, 629–639 (1996). USA 93, 7673–7678 (1996).
37. Engel, A. M., Svane, I. M., Rygaard, J. & Werdelin, O. MCA sarcomas induced in scid mice are more 84. Kumar,A., Commane, M., Flickinger,T.W., Horvath, C. M. & Stark, G. R. Defective TNF-α-induced apop-
immunogenic than MCA sarcomas induced in congenic, immunocompetent mice. Scand. J. Immunol. tosis in STAT1-null cells due to low constitutive levels of caspases. Science 278, 1630–1632 (1997).
45, 463–470 (1997). 85. Luster, A. D. & Ravetch, J.V. Biochemical characterization of a γ interferon-inducible cytokine (IP-10).
38. Schuler,W. et al. Rearrangement of antigen receptor genes is defective in mice with severe com- J. Exp. Med. 166, 1084–1097 (1987).
bined immune deficiency. Cell 46, 963–972 (1986). 86. Liao, F. et al. Human Mig chemokine: biochemical and functional characterization. J. Exp. Med. 182,
39. Featherstone, C. & Jackson, S. P. DNA double-strand break repair. Curr. Biol. 9, R759–R761 (1999). 1301–1314 (1995).
40. Dighe, A. S., Richards, E., Old, L. J. & Schreiber, R. D. Enhanced in vivo growth and resistance to rejec- 87. Cole, K. E. et al. Interferon-inducible T cell α chemoattractant (I-TAC): a novel non-ELR CXC
tion of tumor cells expressing dominant negative IFN-γ receptors. Immunity 1, 447–456 (1994). chemokine with potent activity on activated T cells through selective high affinity binding to
41. Kaplan, D. H. et al. Demonstration of an interferon γ-dependent tumor surveillance system in CXCR3. J. Exp. Med. 187, 2009–2021 (1998).
immunocompetent mice. Proc. Natl. Acad. Sci. USA 95, 7556–7561 (1998). 88. Luster, A. D. & Leder, P. IP-10, a -C-X-C- chemokine, elicits a potent thymus-dependent antitumor
42. Street, S. E., Cretney, E. & Smyth, M. J. Perforin and interferon-γ activities independently control response in vivo. J. Exp. Med. 178, 1057–1065 (1993).
tumor initiation, growth, and metastasis. Blood 97, 192–197 (2001). 89. Sgadari, C., Angiolillo, A. L. & Tosato, G. Inhibition of angiogenesis by interleukin-12 is mediated by the
43. Street, S. E.,Trapani, J. A., MacGregor, D. & Smyth, M. J. Suppression of lymphoma and epithelial malig- interferon-inducible protein 10. Blood 87, 3877–3882 (1996).
nancies effected by interferon γ. J. Exp. Med. 196, 129–134 (2002). 90. Coughlin, C. M. et al. Tumor cell responses to IFN-γ affect tumorigenicity and response to IL-12
44. Russell, J. H. & Ley,T. J. Lymphocyte-mediated cytotoxicity. Annu. Rev. Immunol. 20, 323–370 (2002). therapy and antiangiogenesis. Immunity 9, 25–34 (1998).
45. van den Broek, M. F. et al. Decreased tumor surveillance in perforin-deficient mice. J.Exp.Med. 184, 91. Qin, Z. & Blankenstein, T. CD4+ T cell–mediated tumor rejection involves inhibition of angiogene-
1781–1790 (1996). sis that is dependent on IFN-γ receptor expression by nonhematopoietic cells. Immunity 12,
46. Smyth, M. J. et al. Differential tumor surveillance by natural killer (NK) and NKT cells. J. Exp. Med. 677–686 (2000).
191, 661–668 (2000). 92. Bancroft, G. J., Schreiber, R. D. & Unanue, E. R. Natural immunity: a T-cell-independent pathway of
47. Smyth, M. J. et al. Perforin-mediated cytotoxicity is critical for surveillance of spontaneous lym- macrophage activation, defined in the scid mouse. Immunol. Rev. 124, 5–24 (1991).
phoma. J. Exp. Med. 192, 755–760 (2000). 93. Trinchieri, G. Interleukin-12: a proinflammatory cytokine with immunoregulatory functions that bridge
48. Shinkai,Y. et al. RAG-2-deficient mice lack mature lymphocytes owing to inability to initiate V(D)J innate resistance and antigen-specific adaptive immunity. Annu. Rev. Immunol. 13, 251–276 (1995).
rearrangement. Cell 68, 855–867 (1992). 94. Ikeda, H., Old, L. J. & Schreiber, R. D.The roles of IFN-γ in protection against tumor development and
49. Shankaran,V. et al. IFN-γ and lymphocytes prevent primary tumour development and shape tumour cancer immunoediting. Cytokine Growth Factor Rev. 13, 95–109 (2002).
immunogenicity. Nature 410, 1107–1111 (2001). 95. Takeda, K. et al. Critical role for tumor necrosis factor-related apoptosis-inducing ligand in immune
50. Girardi, M. et al. Regulation of cutaneous malignancy by γδ T cells. Science 294, 605–609 (2001). surveillance against tumor development. J. Exp. Med. 195, 161–169 (2002).
51. Penn, I. Malignant Tumors in Organ Transplant Recipients (Springer-Verlag, New York, 1970). 96. Pardoll, D. M. Spinning molecular immunology into successful immunotherapy. Nature Rev. Immunol. 2,
52. Gatti, R. A. & Good, R. A. Occurrence of malignancy in immunodeficiency diseases. A literature 227–238 (2002).
review. Cancer 28, 89–98 (1971). 97. Gerosa, F. et al. Reciprocal activating interaction between natural killer cells and dendritic cells.
53. Penn, I. Posttransplant malignancies. Transplant Proc. 31, 1260–1262 (1999). J. Exp. Med. 195, 327–333 (2002).
54. Birkeland, S. A. et al. Cancer risk after renal transplantation in the Nordic countries, 1964-1986. Int. J. 98. Ferlazzo, G. et al. Human dendritic cells activate resting natural killer (NK) cells and are recognized
Cancer 60, 183–189 (1995). via the NKp30 receptor by activated NK cells. J. Exp. Med. 195, 343–351 (2002).
55. Sheil, A. G. R. in Kidney Transplantation (ed. Morris, P. J.) 558–570 (Saunders, Philadelphia, 2001). 99. Piccioli, D., Sbrana, S., Melandri, E. & Valiante, N. M. Contact-dependent stimulation and inhibition of
56. Boshoff, C. & Weiss, R. AIDS-related malignancies. Nature Rev. Cancer 2, 373–382 (2002). dendritic cells by natural killer cells. J. Exp. Med. 195, 335–341 (2002).
57. Hoover, R. N. in Origins of Human Cancer (eds. Hiatt, H. H.,Watson, J. D. & Winsten, J. A.) 369–379 100.Smyth, M. J., Crowe, N.Y. & Godfrey, D. I. NK cells and NKT cells collaborate in host protection
(Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1977). from methylcholanthrene-induced fibrosarcoma. Int. Immunol. 13, 459–463 (2001).
58. Sheil, A. G. Cancer after transplantation. World J. Surg. 10, 389–396 (1986). 101.Noguchi,Y., Jungbluth, A., Richards, E. & Old, L. J. Effect of interleukin 12 on tumor induction by 3-
59. Penn, I. Malignant melanoma in organ allograft recipients. Transplantation 61, 274–278 (1996). methylcholanthrene. Proc. Natl. Acad. Sci. USA 93, 11798–11801 (1996).

998 nature immunology • volume 3 no 11 • november 2002 • www.nature.com/natureimmunology

You might also like