PERSPECTIVES

that indicate that multiple events arerequired for malignancy. Nevertheless,recent studies, particularly of oncogene-dri-ven tumours in transgenic mice, have shownthat the outcome of primary oncogenicevents in epithelial cells can be significantlymodified by the nature of the surroundingnon-malignant cells15. So, it is becomingapparent that the microenvironment has animportant role in allowing the tumour toexpress its full neoplastic phenotype andthat non-malignant cells can be used astherapeutic targets.

The tumour microenvironment con-tains many resident cell types, such asadipocytes and fibroblasts, but it is alsopopulated by migratory haematopoieticcells, most notably macrophages, neu-trophils and mast cells. These haematopoi-etic cells have pivotal roles in the progressionand metastasis of tumours2,69, and thisreview will focus on one such class thetumour-associated macrophages (TAMs).I will argue that the tumour microenviron-ment, in a similar way to that seen in normal development (BOX 1), educatesmacrophages to perform supportive rolesthat promote tumour progression andmetastasis. In addition, because of their fun-damental role in mediating the inflamma-tory response, and the growing appreciationthat chronic inflammation is a significantcause of cancer (BOX 2), I suggest thatmacrophages can initiate and promotetumorigenesis.

TAMs: markers of poor prognosisThe observation of leukocytes in tumoursdates back to the middle of the nineteenthcentury 6. Until recently, however, they wereusually overlooked10. Nevertheless, it is nowappreciated that most solid tumours areabundantly populated with TAMs and thatthese cells can alter clinical outcomes11. Innormal tissues, pathogenic challenge orwounding results in the local expression of awide variety of growth factors colony-stimulating factor 1 (CSF-1; also known asmacrophage CSF), granulocytemacrophageCSF (GM-CSF), macrophage-stimulatingprotein (MSP) and transforming growth factor-1 (TGF-1) and chemokines(chemotactic cytokines), which includeCCL2, CCL7, CCL8 (monocyte chemoattrac-tant protein family-1-3), CCL3 (macrophageinflammatory protein-1 (MIP-1)), CCL4(MIP-1) and macrophage migrationinhibitory factor (MIF). These factors,together with the products of tissue break-down, recruit circulating monocytes andstimulate them to differentiate intomacrophages. Macrophages, in turn, mediateimmune responses, kill pathogens, stimulateangiogenesis and effect tissue repair12,13.

Macrophages TAMs are alsorecruited to tumours by a similar range ofgrowth factors and chemokines, which areoften produced by the tumour cells them-selves8,9,14. Clinical studies have, on balance,shown a correlation between an abundanceof TAMs and poor prognosis14. These data areparticularly strong for breast, prostate, ovar-ian and cervical cancers; the data for stomachand lung cancers are contradictory14, and in asmall study in colorectal cancer, their pres-ence was associated with good prognosis15.However, taking all reports into account regardless of method and sample number more than 80% show a significant correlationbetween TAM density and poor prognosis,whereas less than 10% associate TAM densitywith a good prognosis14. So, increased TAMdensity is usually associated with advancedtumour progression and metastasis.

Evidence from clinical and experimentalstudies indicates that macrophagespromote solid-tumour progression andmetastasis. Macrophages are educatedby the tumour microenvironment, so thatthey adopt a trophic role that facilitatesangiogenesis, matrix breakdown andtumour-cell motility all of which areelements of the metastatic process.During an inflammatory response,macrophages also produce manycompounds ranging from mutagenicoxygen and nitrogen radicals toangiogenic factors that can contributeto cancer initiation and promotion.Macrophages therefore represent animportant drug target for cancerprevention and cure.

Solid tumours comprise not only malignantcells, but also many other non-malignant celltypes. This produces a unique microenviron-ment that can modify the neoplastic proper-ties of the tumour cells. However, there hasbeen an almost exclusive experimental focuson the malignant cells that make uptumours. This is probably because of the suc-cess of studies that defined the requirementfor several mutational events in epithelialcells for the formation of malignanttumours; the isolation of transforming onco-genes, the epithelial-restricted expression ofwhich causes cancers in animals; and the factthat these experimental results are intellectu-ally compatible with epidemiological studies

NATURE REVIEWS | CANCER VOLUME 4 | JANUARY 2004 | 71

Tumour-educated macrophagespromote tumour progression and metastasis

Jeffrey W. Pollard

O P I N I O N

72 | JANUARY 2004 | VOLUME 4 www.nature.com/reviews/cancer

P E R S P E C T I V E S

overexpressed in tumours of the reproductivesystem, including ovarian, uterine, breast andprostate tumours7,18-21. In each case, overex-pression of CSF-1 correlates with poor prog-nosis7,20. In breast cancer, CSF-1 expressioncorrelated with high grade and poor prognosisand was also associated with a dense leukocyticinfiltrate in 90% of the tumours that wereanalysed21. These clinical data therefore providesignificant support for the theory that, in mosttypes of solid tumour, TAMs are involved intumour progression through their recruitmentto these sites by chemokines and CSF-1.

Cytokines alter TAM functionThe conventional wisdom about TAM func-tion is that they are recruited to reject thetumour, which has been recognized as foreignbecause tumours express unique antigens.However, there is a growing body of evidencethat the tumour microenvironment isimmunosuppressive22,23, perhaps as a result ofselection for such an environment aprocess recently termed immunoediting24.So, even when bona fide tumour antigens such as MUCl are expressed, there seemsto be an attenuated immune response to theseantigens25. Recent data indicate that TGF-1has an important role in suppressing theselocal responses and that inhibiting this mole-cule can result in tumour rejection26. It isnoteworthy that TAMs can both produceTGF-1 and process latent TGF-s to pro-duce their active forms27. In addition, the localcytokine milieu in the tumour tends to blockthe immunological functions of these newlyrecruited mononuclear phagocytes such asantigen presentation and cytotoxicity towards tumours, and diverts them towardsspecialized TAMs that are immunosup-pressed and trophic28 (FIG. 1). A principal com-ponent of this cytokine mixture is CSF-1,which locally blocks the maturation of den-dritic cells so that they are unable to presentantigens and promotes the development ofimmunosuppressed trophic TAMs. In sup-port of this hypothesis, studies have shownthat renal-carcinoma cell-line production ofinterleukin-6 (IL-6) and CSF-1 inhibits den-dritic-cell maturation. This effect can bereversed by cytokines such as IL-4 and IL-13,which divert the immune response to onethat favours cytotoxic T cells2831 (FIG. 1).

The cytokine profile of the tumourmicroenvironment is therefore extremelyimportant to the phenotype of the localmononuclear phagocytes. This is probablywhy, under certain circumstances, a high density of TAMs correlates with good prog-nosis. One can hypothesize that, in thesecases, the environment pushes the TAMs

tumours, and gliomas this last being theoriginal source of its purification and highlevels of CCL2 correlate with poor prognosisin breast, cervical and bladder cancer8,16,17.CSF-1 the main growth factor that isresponsible for the survival, proliferation, dif-ferentiation and chemotaxis of mononuclearphagocytes, such as macrophages is widely

It is also striking that, in the tumour typesin which high TAM density correlates withpoor prognosis, there is also a substantialbody of clinical literature that shows thatoverexpression of macrophage growth factorsor chemokines correlates with poor progno-sis. CCL2 is widely expressed in tumours,including ovarian, cervical, bladder and breast

Box 1 | Macrophages have important developmental roles

The immunological and repair functions of macrophages are well documented. It is known thatthey are among the first cells to arrive at sites of wounding and/or infection, where they performseveral functions. They produce cytokines and chemokines to orchestrate the recruitment andactions of other immune cells, and produce growth factors, angiogenic factors and proteases topromote tissue repair. They also kill pathogens through the production of reactive oxygen andnitrogen radicals and present foreign antigens to cytotoxic T cells. What is less well appreciated istheir important role in tissue morphogenesis during development. Nevertheless, analysis of micethat are deficient for macrophages and other mononuclear phagocytic cells (such as osteoclasts)because of a null mutation in the gene that encodes colony-stimulating factor 1 (Csf-1) showsthat these cells have a significant role in the morphogenesis of many tissues83,84. Developmentaldefects include osteopetrosis, dermal hypoplasia, aberrant development of the sex-steroid-hormone feedback response in the brain, delayed and aberrant pancreatic morphogenesis andimpaired branching morphogenesis of the mammary gland. In this last case85, as shown in thefigure, the ends of the developing ducts form a unique multi-laminate structure called theterminal end bud (TEB), which grows out through the mammary fat pad and gives rise to thebasal arborized ductal tree. As these TEBs form, they recruit macrophages; the absence of thesein the Csf-1-null mutant mouse results in delayed ductal development and a poorly branched,atrophic ductal tree. Similar experiments, in which macrophages are ablated by other means,show important roles in eye development86. Together, these experiments show that cells of themononuclear phagocytic lineage have important developmental roles through their remodellingand trophic functions.

It seems likely that tumours co-opt the normal developmental roles of macrophages topromote their own development and invasion through the surrounding stroma87. In contrast tonormal epithelia, however, tumour cells owing to intrinsic transforming mutations havelost positional identity, and so they continue to send out help me signals that result in invasioninto the vasculature.

Macrophage

Tumour invasion and metastasis

Common functionsProduction of: angiogenic factors proteases growth factors/cytokines

Ductal invasionduring mammarydevelopment

CSF-1Chemokines

TEB

P E R S P E C T I V E S

whether manipulation of microenvironmentscontaining these cytokines through therapiesusing GM-CSF, IFN- and IL-12 will be effec-tive means of promoting the immune rejec-tion of tumours. These issues are discussed indepth in another article in the same issue ofthis journal32, which indicates that suchapproaches might be used effectively to causeimmunological rejection of tumours.

TAMs potentiate tumour progressionThe clinical evidence described above indi-cates that, in most tumour types in which ithas been studied, an abundance of TAMsthat are matured in the right cytokine envi-ronment has a negative impact on patientsurvival. Several recent experiments withanimal models support this view by showingthat TAMs potentiate tumour progressionand metastasis. Lin et al.1 used mice carrying anull mutation in the gene that encodes Csf-1to prevent macrophages accumulating inmammary tumours that were induced by themammary-epithelial-restricted expression ofthe polyoma middle-T oncoprotein (knownas PyMT mice). In the macrophage-deficientmice, the incidence and initial rates ofgrowth of primary tumours were not differ-ent from those seen in normal mice, but therate of tumour progression was slowed andtheir metastatic ability was almost com-pletely abrogated when compared with micethat contained normal numbers ofmacrophages. Increasing the abundance ofTAMs in the null mutant mice by expressingCsf-1 in a restricted fashion in tumours,using transgenic technology, accelerated therate of tumour progression and restored therate of metastasis to wild-type levels.Interestingly, overexpression of Csf-1 inwild-type mice also accelerated tumour pro-gression and increased their metastatic rate.These data indicate that the clinical correla-tion between overexpression of CSF-1 andpoor prognosis might be due to the ability ofCSF-1 to recruit and modulate the behaviourof TAMs.

Consistent with these genetic experiments,when immunocompromised mice that hadbeen xenografted with either a human malig-nant embryonic tumour or human colon-cancer cells were treated with antisenseoligonucleotides directed against mouse Csf-1,the growth of the embryonic tumour wascompletely suppressed, and the growth rate ofthe colon cancer was halved, with an increasedsurvival rate in these mice. As the human cellsdid not express the CSF-1 receptor, and theantisense oligonucleotides did not inhibithuman CSF-1 expression, these data argue foran effect on the host, and not on the tumour.

either to be immunologically neutral, or todifferentiate to become active participantsin the immune response against tumoursthrough their presentation of tumour anti-gens to T cells or by direct tumour-cellkilling. In determining this balance, anothermacrophage colony-stimulating factor GM-CSF together with interferon-(IFN-), are likely to be important, as thesedirect macrophages towards more cytotoxicand antigen-presenting phenotypes32.Interestingly, even the type of CSF-1 that isexpressed seems to determine the nature ofthe immune response. CSF-1 is made bothas a cell-surface glycoprotein and as asecreted proteoglycan33. Glioma cells thatexpress the cell-surface form of CSF-1 wererejected when transplanted intracraniallyinto mice, and 80% of the mice survived. Bycontrast, 100% of mice that were implantedwith the same cells died if they had been

transfected with a complementary DNAencoding the soluble form of CSF-1. Otherstudies indicate that this difference isbecause macrophages lock onto the cell-surface form of CSF-1 and kill the tumourcells either directly, or indirectly throughantigen presentation to T cells3436. Inhumans, the soluble form predominates in those tumours that overexpress CSF-118,19.So, one hypothesis that explains the relativelack of immune response in tumours is thatthey produce soluble CSF-1 and cytokines,which suppress dendritic-cell maturationand recruit TAMs, the immune and killingfunctions of which are suppressed and thetrophic and remodelling functions of whichare enhanced (FIG. 1).It will be important toknow how other macrophage chemokines,such as CCL2 and CCL3 which are pro-duced in high concentrations by tumours affect local immune responses, and

NATURE REVIEWS | CANCER VOLUME 4 | JANUARY 2004 | 73

Box 2 | Cancer as an inflammatory disease

The concept of inflammation as a cause of cancer dates back to the work of Virchow in the 1850sand to the work of Fibiger and Yamagiwa in the early twentieth century, which showed thatchronic irritation could trigger cancer6. Although this topic has been largely overlooked in thepast, a growing body of evidence has recently indicated that this inflammatory process is acontributor, if not a cause, of a wide variety of neoplasms2. First, it is now recognized that at least15% of tumours worldwide have a direct infectious origin88. Generally, the pathogens that areresponsible establish chronic infections that cause a persistent inflammatory response. Perhapsthe best documented are the roles of Helicobacter pylori in stomach cancer, herpes viruses incervical cancer and schistosomes in bladder cancers8991. Continuous irritants, such as asbestos,silica and cigarette smoke, also increase the probability of developing bronchial cancer92. Colonicinflammation, such as that found in ulcerative colitis or Crohns disease, predisposes sufferers tocolorectal cancer93. Indeed, even the prototypical oncogenic retrovirus, Rous sarcoma virus,requires an inflammatory response to induce tumours at secondary sites in chickens94.Furthermore, the classical studies of skin carcinogens showed that a second, non-carcinogeniccompound, such as a phorbol ester, could increase the effects of low doses of a primarycarcinogen. These promoters set up an acute inflammatory response and cause alterations incytokine signalling pathways that are required for the carcinoma to form75.

Inflammatory responses recruit many immune cells, among which macrophages are keyplayers95. Macrophages produce angiogenic factors, proteases and growth factors, which result inan environment that stimulates epithelial-cell migration, survival and proliferation. They are alsokey signalling cells that help to organize the responses of other cells most notably mast cellsand neutrophils. Both of these cell types have, along with tumour-associated macrophages(TAMs), been shown to have causal roles in tumour progression1,2,96,97. The signals responsible forthis are thought to be immune cytokines, including tumour necrosis factor- (TNF-),interleukin-1 (IL-1), IL-6 and a plethora of chemokines, such as IL-8. Not only do these cytokinesand chemokines recruit immune cells to specific sites that stimulate tumour progression, but ithas also recently been shown that their receptors are often found on tumour cells themselves,where they can increase tumour growth and migration. For example, GRO (growth-relatedoncogene, which is probably the same as CXCL1) an IL-8-related chemokine stimulatesmelanoma migration and proliferation98. Such data indicate that chemokines might make apermissive environment for tumour-cell growth and migration. Consistent with this are recentstudies that show that the expression of specific chemokines defines the site of tumourmetastasis99. Indeed, it seems more and more likely that tumour cells might subvert the pathwaysthat leukocytes use to migrate to specific sites.

Altogether, the data show that inflammation creates a microenvironment that causesneoplastic transformation and potentiates the progression of cancers. Such a realization shouldalter therapeutic strategies, both for prophylaxis and for treatment of acute disease.

74 | JANUARY 2004 | VOLUME 4 www.nature.com/reviews/cancer

P E R S P E C T I V E S

Macrophages promote angiogenesis. It iswidely recognized that tumours require angio-genesis to grow beyond a certain size. Thisprocess involves a wide range of soluble medi-ators that are both stimulatory and inhibitory,including basic fibroblast growth factor(bFGF),VEGF, the angiopoietins (ANG1 andANG2), IL-1, IL-8, tumour necrosis factor-(TNF-), thymidine phosphorylase (TP; alsoknown as vascular-derived endothelial growthfactor), the matrix metalloproteinases MMP-9and MMP-2, and nitric oxide (NO)39,40. Thesemolecules, which are expressed in a coordi-nated spatial and temporal fashion, result inthe proliferation and migration of endothelialcells, matrix remodelling and the eventual for-mation of stabilized vessels41. Macrophages areperfectly designed to promote these processes,as their monocytic precursors can migrateinto sites where they differentiate intomacrophages, and these wandering cells cansynthesize the required angiogenic moleculeson demand in specific locations42.

Macrophages are important, although notthe only, producers of VEGF, which is a keycomponent of the process of angiogenesis intumours43,44. Studies by Harris and co-workersin human breast cancer showed that TAMscluster in hot spots in avascular areas45. Thesehot spots correlate with a high level of angio-genesis, and also with decreased relapse-freeand overall survival8. It was suggested thathypoxia or the cytokines that are producedin response to this condition is one of thelocal attractants for macrophages, and thathypoxia itself upregulates the transcriptionfactor hypoxia-inducible factor-2 (HIF-2)in macrophages. HIF-2, in turn, inducesVEGF expression8,43 (FIG. 2). VEGF is alsoupregulated by CSF-1 in macrophages46, andthe CSF-1 antisense experiments describedabove showed a reduction in VEGF expressionand an inhibition of angiogenesis37. VEGF isalso a chemoattractant for macrophages44,47

and, as such, this might result in a positive-feedback loop, providing rapid vascularizationto tumours.

TAMs also produce many other pro-angiogenic cytokines8,40. They are key pro-ducers of TNF-48, the expression of whichincreases in these cells as tumours becomeinvasive carcinomas48, and which upregu-lates TP in breast cancer cell lines49.Correlative studies indicate that it also doesso in breast tumours, and that TP expres-sion is significantly correlated with angio-genesis and poor prognosis in thesetumours50. TNF- also induces MMP-9expression, and this in turn can releasebioactive VEGF from its extracellular-matrix (ECM)-bound latent form6. TAMs

and metastasis and provide experimental sup-port for the clinical observations that increasedTAM density promotes tumour malignancy.

Functions of TAMsMacrophages are therefore multifunctionalcells, the phenotypes of which are modified bythe local environment, and have importantroles in the morphogenesis of tissues. Theytake on non-immunological functions, whichprovide trophic support to tissues (BOX 1). Isuggest that a similar education processwithin a tumour results in macrophages that,on balance, promote the progression of thetumour to a more malignant state. In addi-tion, in the context of an inflammatoryresponse, macrophage functions can result inthe initiation or promotion of tumorigenesis.These diverse functions are summarized inFIG. 2, and are discussed below.

This was confirmed by the reduction seen inCsf-1 messenger RNA levels and serum Csf-1concentrations, which correlated with areduction in the number of TAMs37.

In another set of experiments in mice, pri-mary-tumour-stimulated macrophages wereshown to increase the metastatic ability oftail-vein-injected tumour cells. In these exper-iments, the ability of the injected cells to colo-nize and grow in the host lungs only occurredin mice that also carried a separate primarytumour38. This study provided evidence thatmacrophages modified by their exposureto the primary tumour promoted theseeding and vascularization of the injectedtumour cells through the induction of matrixmetalloproteinase 9 (Mmp-9) and vascularendothelial growth factor (Vegf).

Taken together, these recent experimentssupport a role for TAMs in tumour progression

Mononuclearphagocytic progenitor

Immaturedendritic cell

Trophic and scavenging macrophage

Cytotoxicmacrophage

Mature dendritic cells

Tumourprevention

Tumourpromotion

Tumour cells Tumour cells

Soluble CSF-1

Soluble CSF-1

IL-4IL-12 IL-6

Cell- surfaceCSF-1 GM-CSF

IL-13

Figure 1 | Pro- and anti-tumorigenic properties of macrophages depend on the cytokinemicroenvironment in the tumour. Tumours are populated by macrophages and dendritic cells thatare derived from mononuclear phagocytic progenitor cells. In many tumours, a high concentration ofsoluble colony-stimulating factor-1 (CSF-1) educates macrophages to be trophic to tumours and,together with interleukin-6 (IL-6), inhibits the maturation of dendritic cells. This creates amicroenvironment that potentiates progression to metastatic tumours. By contrast, CSF-1 presentedin a transmembrane form on the tumour surface activates macrophages to kill tumour cells. This together with high concentrations of IL-4, IL-12, IL-13 and GM-CSF causes dendritic cells tomature, allowing the presentation of tumour antigens to cytotoxic T cells, with the consequentrejection of the tumour.

P E R S P E C T I V E S

TAMs also release other molecules that caninfluence angiogenesis. They produce IL-1,which through cyclooxygenase 2 (COX 2) upregulates HIF-1, resulting in an increasein the transcription of VEGF59.VEGF produc-tion is also increased by IL-1 in co-cultures oftumour cells and macrophages. IL-1 is synthe-sized by macrophages and, in these experi-ments, IL-1 was required for tumour-cellinvasiveness and angiogenesis60. TAMs alsorelease NO through the induction of theenzyme inducible NO synthase (iNOS)61.Expression of iNOS has been correlated withtumour grade62, and its ablation delaystumour progression in mouse models ofbreast cancer63. Increased NO is likely to resultin vasodilation and increased vascular flow.Interestingly, endothelin 2 another vasoac-tive molecule involved in inflammation andangiogenesis that is expressed in tumours isa chemoattractant for macrophages, whichindicates that it might recruit TAMs tohypoxic areas64.

In conclusion, migratory TAMs areequipped to enter areas of the tumourwhere vascularization is needed. Here, thesecells synthesize angiogenic regulators, whichresults in the formation of new vessels thatallows further tumour growth and access oftumour cells to the vasculature for escapeinto the circulation. At these sites, a com-plex mixture of factors ranging fromhypoxia to cytokines controls the expres-sion of these regulators. This is consistentwith the hypothesis that macrophages areeducated to perform specialized tasks atspecific sites.

Macrophages produce growth factors and pro-teases that enhance tumour progression. It isapparent from the discussion above thatmacrophages are important producers ofproteases, which range from uPA to a varietyof matrix metalloproteinases especiallyMMP-7 and MMP-9. In our detailed study of the progression of PyMT-inducedtumours7,65, we noted that at the time ofmalignant transition, leukocytic infiltrateswere present that coincided with areas ofbasement-membrane breakdown andtumour-cell egress (FIG. 3). It is not clearwhether these invading leukocytes cause theinitial breakage of the basement membraneor if this is a result of the activities of thetumour cells themselves. Regardless of the ini-tiating mechanism, approximately 50% of theleukocytes are TAMs, which produce pro-teases that degrade the basement membrane,so creating a portal through which tumourcells enter the stroma. This is a key step intumour metastasis.

also synthesize other proteins that influenceangiogenesis for example, urokinase-type plasminogen activator (uPA)51, whichis activated by CSF-1 receptor signalling inmacrophages52 and by TGF-1 (REF. 53).Expression of uPA and its inhibitor, plas-minogen-activator inhibitor type 1 (PAI-1),has clinical and prognostic value54. uPA andits receptor are upregulated in TAMs in breast cancer51. The complex of uPA and its receptor is fully active, and mightcontribute to the ECM breakdown that is

required for vascular invasion to occur.Consistent with this, the expression of theuPA receptor in TAMs has been clinicallycorrelated with high microvessel densityand poor prognosis54,55. In addition, a sig-nificant correlation exists between PAI-1expression, vessel remodelling, and nodestatus and tumour grade5658. These dataindicate that the uPA system is important inestablishing the vascular network intumours and that TAMs have an importantrole in its expression and regulation.

NATURE REVIEWS | CANCER VOLUME 4 | JANUARY 2004 | 75

Recruit otherhaematopoietic cells

Angiogenicfactors

Proteasesinduce invasion

Growth factorsand chemokines

Endothelial cells

Stromal cells

Neutrophil

Mast cell

Oestrogens Reactive oxygen and nitrogen radicals

Chemotacticfactors

Migrateto hypoxicareas

Figure 2 | Pro-tumorigenic functions of tumour-associated macrophages. Macrophages arerecruited to tumours by chemotactic factors and provide many trophic functions that promote tumourprogression and metastasis. These tumour-associated macrophages (TAMs) migrate to hypoxic areaswithin the tumour, where they stimulate angiogenesis by expressing factors such as vascularendothelial growth factor (VEGF), angiopoietin 1 (ANG1) and ANG2, and recruit other haematopoieticcells mast cells and neutrophils that can perform similar tasks. TAMs also promote tumourinvasion by producing proteases such as urokinase-type plasminogen activator (uPA), matrixmetalloproteinase 9 (MMP-9) and cathepsins that break down the basement membrane andremodel the stromal matrix. MMP-9 also contributes to angiogenesis. Various growth factors andchemokines epidermal growth factor (EGF), transforming growth factor- (TGF-), interleukin-8 (IL-8)and tumour necrosis factor- (TNF-) contribute to the migration of tumour cells towards vesselsand provide proliferative and anti-apoptotic signals to these cells. Macrophages that are attracted tosites of inflammation or tissue breakdown can also initiate or promote tumorigenesis through theirsynthesis of oestrogens and the generation of mutagens as a by-product of their production ofreactive oxygen and nitrogen-oxide radicals.

76 | JANUARY 2004 | VOLUME 4 www.nature.com/reviews/cancer

P E R S P E C T I V E S

and so further select for tumour cells thatcontain mutations that enhance apoptosisresistance and increase motility. The conse-quence of all these processes is a moremalignant and invasive phenotype.

Macrophages can initiate tumorigenesisThe presence of TAMs in a tumour providesan environment that enhances the survival,proliferation and migration of epithelialcells that have accumulated primary onco-genic mutations. However, macrophagesmight have an even more sinister role byplaying a significant part in establishing theprimary oncogenic events in epithelial cells(FIG. 2). There is a growing body of evidenceto indicate that inflammation as a resultof chronic infection, continuous exposureto irritants or genetic makeup is acausative event in many cancers (BOX 2).Macrophages comprise a key component ofthe inflammatory response and function askey regulators of the activities of many ofthe other cell types that are involved ininflammation. At these inflammatory sites,macrophages produce high levels of reactiveoxygen and nitrogen species and these through the formation of peroxynitite can react with DNA, resulting in mutagenicevents in epithelial and surroundingcells72,73. This continuous generation ofmutagenic compounds in response to per-sistent infection is thought to be the mecha-nism by which Helicobacter pylori causesstomach cancer (BOX 2). Similarly, somecytokines that are produced by TAMs andother immune cells, such as TNF- andMIF, might also contribute to the generationof chromosomal abnormalities. MIF, forexample, suppresses TP53 transcription intumour cells, which results in the lack of aDNA-damage-repair response and, conse-quently, the accumulation of mutations74.TNF- treatment of carcinogen-treatedfibroblasts renders them capable of tumourformation in nude mice65. Mechanistically,this may be through the induction of iNOSin the tumour cells, which results in NO pro-duction. Studies of mice that lack Tnf-showed that this cytokine is also required for7,12-dimethylbenz(a)-anthracene (DMBA)-induced, 12-O-tetradecanoyl-phorbol-13-acetate (TPA)-promoted skin carcinogenesis,and this is related to the reduced inflamma-tion that is seen in the absence of Tnf-,which decreases the incidence of de novocarcinogenesis75.

TAMs are abundant in breast cancers8.Recent studies indicate that these cells mightboth respond to and produce oestrogens76,77.TAMs in the breast cancer bed express

heterodimerize and homodimerize. Some,such as ERBB3, do not have an active kinasedomain and need to heterodimerize totransduce a signal. ERBB2, for which a lig-and has yet to be identified, is overexpressedin approximately 20% of breast cancers andis an effective target for therapy 67. Similarly,a high level of ERBB1 (also known asEGFR) expression in breast cancer is corre-lated with poor survival and has importantdiagnostic value68. TAMs have beenreported to be the most significant source ofEGF in tumours69 and are associated withEGFR expression44 and poor prognosis.EGF can promote tumour-cell proliferationand is also a potent chemoattractant ofbreast cancer cells in culture70. We haverecently shown that tumour cells respond tomacrophage-produced EGF ligands in vivoby chemotaxis and invasion. Thesemacrophages are often associated with ves-sels, which indicates that they providechemotactic signals that recruit tumourcells to blood vessels and enhance theiregress into the vasculature (J. Wycoff et al.,unpublished observations). The absence ofthese macrophage signals might be part ofthe reason why mice that lack macrophagesshow a low rate of metastasis1.

A unifying concept for TAM action. In ourstudies of PyMT tumours that were dis-cussed earlier, focal sites of leukocytic infil-tration were found at the point of transitionof tumours71. Such sites are also commonlyfound in human breast cancers71. So, a sce-nario can be proposed in which intrinsicmutations that accumulate in the tumourcells cause them to send out chemoattrac-tive signals that are similar to those pro-duced in the developing mammary gland(BOX 1), indicating that they want to breakout through the basement membrane andinto the stroma. These signals recruitmacrophages and other haematopoieticcells, which interact to cause focal break-down of the basement membrane (FIG. 3).Tumour cells consequently exit into thestroma and TAM-induced angiogenesisoccurs at the location of this escape. Thesevessels become sites of macrophage align-ment. TAMs then provide chemotactic sig-nals to the tumour cells to promote theirmigration towards vessels and cause theirintravasation. Growth factors that aresecreted by TAMs might also promote theviability of tumour cells by overcoming the apoptotic signals that are induced bytheir detachment from the basement mem-brane. The same growth factors might also promote tumour-cell proliferation locally,

TAMs also produce a wide variety ofgrowth factors that can stimulate the growthand motility of tumour cells8,66. Theseinclude fibroblast growth factor (FGF),hepatocyte growth factor (HGF), epider-mal-growth-factor receptor (EGFR)-family ligands, platelet-derived growth factor(PDGF) and transforming growth factor-s(TGF-s). Ligands of the EGFR family seemto be very important in cancer, particularlyin cancers of the breast and lung, and thereceptors that belong to this family form a complex group that are able to both

Endothelial cellRed blood cell

Mast cell

Lymphocyte

Granulocyte

Invasive tumour cell

Basement membrane

Macrophage

Fibroblast

Tumour

Stroma

Monocyte

Figure 3 | The leukocytic infiltration site as aportal for the exit of tumour cells. Manytumours, as they become invasive, also haveleukocytic infiltration sites that are abundantlypopulated by macrophages. So, a hypothesis toexplain the role of macrophages in tumourinvasion and metastasis is that these cells,through their proteolytic activity, break down thebasement membrane around pre-invasivetumours, thereby enhancing the ability of tumourcells to escape into the surrounding stroma.Macrophages also stimulate angiogenesis atthese sites of tumour egress and send comehither signals to cells, causing them to move outof the tumour mass towards blood vessels. So,tumour cells flow out from the ductally constrainedtumour mass into the surrounding stroma, therebygaining access to the vasculature, with theconsequent ability to colonize distant sites.

P E R S P E C T I V E S

1. Lin, E. Y., Nguyen, A. V., Russell, R. G. & Pollard, J. W.Colony-stimulating factor 1 promotes progression ofmammary tumors to malignancy. J. Exp. Med. 193,727740 (2001).

2. Coussens, L. M. & Werb, Z. Inflammation and cancer.Nature 420, 860867 (2002).

3. Coussens, L. M., Tinkle, C. L., Hanahan, D. & Werb, Z.MMP-9 supplied by bone marrow-derived cellscontributes to skin carcinogenesis. Cell 103, 481490(2000).

4. Iyengar, P. et al. Adipocyte-secreted factorssynergistically promote mammary tumorigenesis throughinduction of anti-apoptotic transcriptional programs andproto-oncogene stabilization. Oncogene 22, 64086423(2003).

5. Krtolica, A. et al. Senescent fibroblasts promote epithelialcell growth and tumorigenesis: a link between cancer andaging. Proc. Natl Acad. Sci. USA 98, 1207212077(2001).

6. Balkwill, F. & Mantovani, A. Inflammation and cancer:back to Virchow? Lancet 357, 539545 (2001).

7. Lin, E. Y., Gouon-Evans, V., Nguyen, A. V. & Pollard, J. W.The macrophage growth factor, CSF-1, in mammarygland development and cancer. J. Mammary Gland Biol.Neoplasia 7, 147162 (2002).

8. Leek, R. D. & Harris, A. L. Tumor-associatedmacrophages in breast cancer. J. Mammary Gland Biol.Neoplasia 7, 177189 (2002).

9. Brigati, C., Noonan, D. M., Albini, A. & Benelli, R. Tumorsand inflammatory infiltrates: friends or foes? Clin. Exp.Metastasis 19, 247258 (2002).

10. Normann, S. J. Macrophage infiltration and tumorprogression. Cancer Metastasis Rev. 4, 277291 (1985).

11. Mantovani, A. Tumor-associated macrophages inneoplastic progression: a paradigm for the in vivofunction of chemokines. Lab. Invest. 71, 516 (1994).

12. Lingen, M. W. Role of leukocytes and endothelial cells in thedevelopment of angiogenesis in inflammation and woundhealing. Arch. Pathol. Lab. Med. 125, 6771 (2001).

13. Nathan, C. Points of control in inflammation. Nature 420,846852 (2002).

14. Bingle, L., Brown, N. J. & Lewis, C. E. The role of tumour-associated macrophages in tumour progression:implications for new anticancer therapies. J. Pathol. 196,254265 (2002).

15. Nakayama, Y. et al. Relationships between tumor-associated macrophages and clinicopathological factorsin patients with colorectal cancer. Anticancer Res. 22,42914296 (2002).

16. Ueno, T. et al. Significance of macrophagechemoattractant protein-1 in macrophage recruitment,angiogenesis, and survival in human breast cancer. Clin.Cancer Res. 6, 32823289 (2000).

17. Saji, H. et al. Significant correlation of monocytechemoattractant protein-1 expression withneovascularization and progression of breast carcinoma.Cancer 92, 10851091 (2001).

18. Smith, H. O. et al. The role of colony-stimulating factor 1and its receptor in the etiopathogenesis of endometrialadenocarcinoma. Clin. Cancer Res. 1, 313325 (1995).

19. Kacinski, B. M. CSF-1 and its receptor in breastcarcinomas and neoplasms of the female reproductivetract. Mol. Reprod. Dev. 46, 7174 (1997).

20. Kacinski, B. M. CSF-1 and its receptor in ovarian,endometrial and breast cancer. Ann. Med. 27, 7985(1995).

21. Scholl, S. M. et al. Anti-colony-stimulating factor-1antibody staining in primary breast adenocarcinomascorrelates with marked inflammatory cell infiltrates andprognosis. J. Natl Cancer Inst. 86, 120126 (1994).

22. Elgert, K., Alleva, D. & Mullins, D. Tumor-induced immunedysfunction: the macrophage connection. J. Leukoc.Biol. 64, 275290 (1998).

23. Ohm, J. E. & Carbone, D. P. VEGF as a mediator oftumor-associated immunodeficiency. Immunol. Res. 23,263272 (2001).

24. Dunn, G. P. et al. Cancer immunoediting: fromimmunosurveillance to tumor escape. Nature Immunol. 3,991998 (2002).

25. Taylor-Papadimitriou, J. et al. MUC1 and theimmunobiology of cancer. J. Mammary Gland Biol.Neoplasia 7, 209221 (2002).

26. Gorelik, L. & Flavell, R. A. Immune-mediated eradicationof tumors through the blockade of transforming growthfactor- signaling in T cells. Nature Med. 7, 11181122(2001).

27. Chong, H., Vodovotz, Y., Cox, G. W. & Barcellos-Hoff, M. H.Immunocytochemical localization of latent transforminggrowth factor- activation by stimulated macrophages. J. Cell Physiol. 178, 275283 (1999).

aromatase, which converts androgens tooestrogens76. Interestingly, in a cohort ofwomen at risk of breast cancer who hadbeen given intraductal lavages to accessepithelial-cell morphology, it was foundthat more than 50% of the retrieved cellswere macrophages78. These cells might syn-thesize oestrogen, which could explain thefivefold higher concentration of oestrogensin the ductal fluid than in serum. As oestro-gen exposure is the primary risk factor forbreast cancer79, macrophages in both thepre-malignant and malignant breast couldhave an important role in increasing expo-sure to high concentrations of this steroidhormone. This, coupled with their muta-genic properties, which are describedabove, could significantly enhance the riskof breast cancer.

Therapeutic opportunitiesThe data indicate that continuous inflam-mation as a result of persistent infection orother irritants has a causal role in tumourprogression. Is it possible that cancer isoften a pathology of chronic infectious dis-eases (BOX 2)? If so, therapies that aredirected at reducing inflammation orinhibiting the function of inflammatorycytokines could reduce cancer risk.Consistent with this notion is the ability ofnon-steroidal anti-inflammatory drugs(NSAIDs) to significantly lower colon-cancer risk and, perhaps, to prevent lung,oesophageal and stomach cancers80,81. Thesedrugs target cyclooxygenase enzymes,which are involved in the metabolism ofprostaglandins. Prostaglandins have animportant role in inflammatory responsesand are produced abundantly by macro-phages. It will be important to determinewhether macrophages are the site of actionof these NSAIDs, as COX2 can also beexpressed in tumour cells.

I have argued in this article that themicroenvironment educates macrophagesto take on specific phenotypes that canincrease the risk of cancer and, once cancersare formed, promote their progression.Although there is a considerable amount ofdata on the phenotype of activatedmacrophages that are involved in immuneresponses, there are relatively few dataabout the macrophages that are found atdifferent sites in the body. However, recentadvances that allow macrophages to bemarked in vivo 82 will allow their isolation,in a relatively unperturbed state, from spe-cific sites. The analysis of these cells usingDNA or protein microarrays will thendefine their phenotypes. Although most of

the expressed genes will be common to allmacrophages, these experiments will, I sus-pect, show cohorts that are unique to partic-ular macrophage environments. TAMs willtherefore be shown to have a unique set ofexpressed genes. The intracellular regulatorsof these different pathways might thereforebe potential therapeutic targets. The experi-ments in mouse models that are describedabove in which tumour progression andmetastasis are reduced by the ablation ofCSF-1, either genetically1 or by the use ofantisense CSF-1 oligonucleotides37 sup-port this idea. In the future, new methods toreduce the continuous over-exuberant pro-duction of macrophage chemoattractants,combined with anti-inflammatory drugs,might result in a significant abatement of the aggressive phenotypes of epithelialneoplasias.

SummaryThe data described above indicate thatmacrophages affect many processes, rangingfrom tumour initiation to the acceleration oftumour progression and metastasis. Theabilities of these cells to move to specificsites, increase matrix remodelling andinduce angiogenesis are also essential duringnormal development and in normal physio-logical processes, such as wound healing andinflammation. This indicates that tumoursrecapitulate these developmental signals toco-opt the normal developmental roles ofmacrophages (BOX 1). However, unlike nor-mal epithelial structures, neoplastic cellshave lost their positional identity due tointrinsic mutations and, so, they continue toinvade and enter the vascular and lymphaticsystems. Furthermore, these transformedepithelial cells send out continuous andunrestrained signals that constantly recruitmacrophages and other haematopoietic cellsto perform these morphogenic roles. Thisresults in a state that is described by thephrase tumours are wounds that neverheal6. But this could perhaps be betterstated as tissues that never cease to develop.This continuous on signal indicates thatstrategies that are directed to dampen therecruitment and tumour-associated func-tions of macrophages could find an impor-tant place within the therapeutic arsenalagainst tumours.

Jeffrey W. Pollard is at the Center for the Study of

Reproductive Biology and Womens Health and the Albert Einstein Cancer Center,

Albert Einstein College of Medicine,New York, New York 10461, USA.

e-mail: pollard@aecom.yu.edudoi: 10.1038/nrc1256

NATURE REVIEWS | CANCER VOLUME 4 | JANUARY 2004 | 77

78 | JANUARY 2004 | VOLUME 4 www.nature.com/reviews/cancer

P E R S P E C T I V E S

81. Ryan, B. M., Russel, M. G., Langholz, E. & Stockbrugger,R. W. Aminosalicylates and colorectal cancer in IBD: anot-so bitter pill to swallow. Am. J. Gastroenterol. 98,16821687 (2003).

82. Sasmono, R. T. et al. A macrophage colony-stimulatingfactor receptor-green fluorescent protein transgene isexpressed throughout the mononuclear phagocytesystem of the mouse. Blood 101, 11551163 (2003).

83. Pollard, J. W. & Stanley, E. R. Pleiotropic roles for CSF-1in development defined by the mouse mutationosteopetrotic (op). Adv. Dev. Biochem. 4, 153193(1996).

84. Pollard, J. W. Role of colony-stimulating factor-1 inreproduction and development. Mol. Reprod. Dev. 46,5461 (1997).

85. Gouon-Evans, V., Rothenberg, M. E. & Pollard, J. W.Postnatal mammary gland development requiresmacrophages and eosinophils. Development 127,22692282 (2000).

86. Lang, R. A. & Bishop, J. M. Macrophages are required forcell death and tissue remodeling in the developing mouseeye. Cell 74, 453462 (1993).

87. Gouon-Evans, V., Lin, E. Y. & Pollard, J. W. Requirementof macrophages and eosinophils and theircytokines/chemokines for mammary glanddevelopment. Breast Cancer Res. 4, 155164 (2002).

88. Kuper, H., Adami, H. O. & Trichopoulos, D. Infections as amajor preventable cause of human cancer. J. Intern.Med. 248, 171183 (2000).

89. Rosin, M. P., Anwar, W. A. & Ward, A. J. Inflammation,chromosomal instability, and cancer: theschistosomiasis model. Cancer Res. 54 (Suppl.),S1929S1933 (1994).

90. Mostafa, M. H., Sheweita, S. A. & OConnor, P. J.Relationship between schistosomiasis and bladdercancer. Clin. Microbiol. Rev. 12, 97111 (1999).

91. Castellsague, X., Bosch, F. X. & Munoz, N. Environmentalco-factors in HPV carcinogenesis. Virus Res. 89,191199 (2002).

92. Shacter, E. & Weitzman, S. A. Chronic inflammation andcancer. Oncology (Huntingt.) 16, 217226, 229;discussion 230232 (2002).

93. Kornfeld, D., Ekbom, A. & Ihre, T. Is there an excess riskfor colorectal cancer in patients with ulcerative colitis andconcomitant primary sclerosing cholangitis? A populationbased study. Gut 41, 522525 (1997).

94. Sieweke, M. H., Stoker, A. W. & Bissell, M. J. Evaluationof the cocarcinogenic effect of wounding in Roussarcoma virus tumorigenesis. Cancer Res. 49,64196424 (1989).

95. Libby, P. Inflammation in atherosclerosis. Nature 420,868674 (2002).

96. Di Carlo, E. et al. The intriguing role of polymorphonuclearneutrophils in antitumor reactions. Blood 97, 339345(2001).

97. Coussens, L. M. et al. Inflammatory mast cells up-regulate angiogenesis during squamous epithelialcarcinogenesis. Genes Dev. 13, 13821397 (1999).

98. Haghnegahdar, H. et al. The tumorigenic and angiogeniceffects of MGSA/GRO proteins in melanoma. J. Leukoc.Biol. 67, 5362 (2000).

99. Muller, A. et al. Involvement of chemokine receptors inbreast cancer metastasis. Nature 410, 5056 (2001).

AcknowledgementsI would like to acknowledge the outstanding work of E. Lin indemonstrating roles for macrophages in mouse models of breastcancer in my laboratory, and thank A. Niklaus for helping with thefigures. This research was supported by the National Institutes ofHealth and the Albert Einstein Cancer Center. J. W. P. is the Bettyand Sheldon E. Feinberg Senior Faculty Scholar in CancerResearch. This article is dedicated to Barry Kacinski, who first dis-covered CSF-1 receptor overexpression in human tumours andwhose untimely death is deeply regretted.

Competing interests statementThe author declares that he has no competing financial interests.

Online links

DATABASESThe following terms in this article are linked online to:Cancer.gov: http://cancer.gov/bladder cancer | breast cancer | cervical cancer | colorectalcancer | lung cancer | ovarian cancer | prostate cancer |stomach cancer | uterine cancerLocusLink: http://www.ncbi.nlm.nih.gov/LocusLink/CCL2 | CCL4 | CCL7 | CCL8 | CSF-1 | ERBB1 | GM-CSF | IFN- |IL-1 | IL-4 | IL-6 | IL-12 | MIF | MMP-2 | MMP-7 | MMP-9 | MSP |TGF-1 | TNF- | VEGF

54. Foekens, J. A. et al. The urokinase system ofplasminogen activation and prognosis in 2780 breastcancer patients. Cancer Res. 60, 636643. (2000).

55. Hildenbrand, R. et al. Urokinase receptor localization inbreast cancer and benign lesions assessed by in situhybridization and immunohistochemistry. Histochem.Cell. Biol. 110, 2732 (1998).

56. Fox, S. B. et al. Plasminogen activator inhibitor-1 as ameasure of vascular remodelling in breast cancer. J.Pathol. 195, 236243 (2001).

57. Knoop, A. et al. Prognostic significance of urokinase-typeplasminogen activator and plasminogen activatorinhibitor-1 in primary breast cancer. Br. J. Cancer 77,932940 (1998).

58. Hildenbrand, R. et al. Urokinase plasminogen activatorreceptor (CD87) expression of tumor-associatedmacrophages in ductal carcinoma in situ, breast cancer,and resident macrophages of normal breast tissue. J.Leukoc. Biol. 66, 4049 (1999).

59. Jung, Y. J. et al. IL-1 mediated up-regulation of HIF-1via an NFB/COX-2 pathway identifies HIF1 as a criticallink between inflammation and oncogenesis. FASEB J.17, 21152117 (2003).

60. Voronov, E. et al. IL-1 is required for tumor invasivenessand angiogenesis. Proc. Natl Acad. Sci. USA 100,26452650 (2003).

61. MacMicking, J., Xie, Q. W. & Nathan, C. Nitric oxide andmacrophage function. Annu. Rev. Immunol. 15, 323350(1997).

62. Wink, D. A. et al. The multifaceted roles of nitric oxide incancer. Carcinogenesis 19, 711721 (1998).

63. Ellies, L. G. et al. Mammary tumor latency is increased inmice lacking the inducible nitric oxide synthase. Int. J.Cancer 106, 17 (2003).

64. Grimshaw, M. J., Wilson, J. L. & Balkwill, F. R.Endothelin-2 is a macrophage chemoattractant:implications for macrophage distribution in tumors. Eur. J.Immunol. 32, 23932400 (2002).

65. Arnott, C. H. et al. Tumour necrosis factor- mediatestumour promotion via a PKC- and AP-1-dependentpathway. Oncogene 21, 47284738 (2002).

66. Ogmundsdottir, H. M., Petursdottir, I. & Gudmundsdottir,I. Interactions between the immune system and breastcancer. Acta Oncol. 34, 647650 (1995).

67. Menard, S., Tagliabue, E., Campiglio, M. & Pupa, S. M.Role of HER2 gene overexpression in breast carcinoma.J. Cell Physiol. 182, 150162 (2000).

68. Nicholson, S. et al. Epidermal growth factor receptor(EGFR); results of a 6 year follow-up study in operablebreast cancer with emphasis on the node negativesubgroup. Br. J. Cancer 63, 146150 (1991).

69. OSullivan, C., Lewis, C. E., Harris, A. L. & McGee, J. O.Secretion of epidermal growth factor by macrophgesassociated with breast carcinoma. Lancet 342, 872873(1993).

70. Wyckoff, J. B., Segall, J. E. & Condeelis, J. S. Thecollection of the motile population of cells from a livingtumor. Cancer Res. 60, 54015404 (2000).

71. Lin, E. Y. et al. Progression to malignancy in the polyomamiddle T oncoprotein mouse breast cancer modelprovides a reliable model for human diseases. Am. J.Pathol. 163, 21132126 (2003).

72. Maeda, H. & Akaike, T. Nitric oxide and oxygen radicals ininfection, inflammation, and cancer. Biochemistry (Mosc.)63, 854865 (1998).

73. Fulton, A. M., Loveless, S. E. & Heppner, G. H.Mutagenic activity of tumor-associated macrophages inSalmonella typhimurium strains TA98 and TA100. CancerRes. 44, 43084311 (1984).

74. Hudson, J. D. et al. A proinflammatory cytokine inhibitsp53 tumor suppressor activity. J. Exp. Med. 190,13751382 (1999).

75. Moore, R. J. et al. Mice deficient in tumor necrosis factor-are resistant to skin carcinogenesis. Nature Med. 5,828831 (1999).

76. Mor, G. et al. Macrophages, estrogen and themicroenvironment of breast cancer. J. Steroid Biochem.Mol. Biol. 67, 403411 (1998).

77. Mor, G. et al. Interaction of the estrogen receptors withthe Fas ligand promoter in human monocytes. J.Immunol. 170, 114122 (2003).

78. King, B. L. et al. Immunocytochemical analysis of breastcells obtained by ductal lavage. Cancer 96, 244249(2002).

79. Travis, R. C. & Key, T. J. Oestrogen exposure andbreast cancer risk. Breast Cancer Res. 5, 239247(2003).

80. Ricchi, P., Zarrilli, R., Di Palma, A. & Acquaviva, A. M.Nonsteroidal anti-inflammatory drugs in colorectalcancer: from prevention to therapy. Br. J. Cancer 88,803807 (2003).

28. Menetrier-Caux, C. et al. Inhibition of the differentiation ofdendritic cells from CD34+ progenitors by tumor cells:role of interleukin-6 and macrophage colony-stimulatingfactor. Blood 92, 47784791 (1998).

29. Menetrier-Caux, C. et al. IL-4 prevents the blockade ofdendritic cell differentiation induced by tumor cells.Cancer Res. 61, 30963104 (2001).

30. Boggio, K. et al. Ability of systemic interleukin-12 tohamper progressive stages of mammary carcinogenesisin HER2/neu transgenic mice. Cancer Res. 60, 359364(2000).

31. Nanni, P. et al. Prevention of HER-2/neu transgenicmammary carcinoma by tamoxifen plus interleukin 12.Int. J. Cancer 105, 384389 (2003).

32. Dranoff, G. Cytokines in cancer pathogenesis and cancertherapy. Nature Rev. Cancer 4, 1122 (2004).

33. Price, L. K. H., Choi, H. U., Rosenberg, L. & Stanley, E. R.The predominant form of secreted colony stimulatingfactor-1 is a proteoglycan. J. Biol. Chem. 267,21902199 (1992).

34. Jadus, M. R. et al. Macrophages kill T9 glioma tumorcells bearing the membrane isoform of macrophagecolony stimulating factor through a phagocytosis-dependent pathway. J. Immunol. 160, 361368 (1998).

35. Jadus, M. R. et al. Macrophages can recognize and killtumor cells bearing the membrane isoform ofmacrophage colony-stimulating factor. Blood 87,52325241 (1996).

36. Williams, C. C. et al. Membrane macrophage colony-stimulating factor on MADB106 breast cancer cellsdoes not activate cytotoxic macrophages butimmunizes rats against breast cancer. Mol. Ther. 3,216224 (2001).

37. Aharinejad, S. et al. Colony-stimulating factor-1 antisensetreatment suppresses growth of human tumor xenograftsin mice. Cancer Res. 62, 53175324 (2002).

38. Hiratsuka, S. et al. MMP9 induction by vascularendothelial growth factor receptor-1 is involved in lung-specific metastasis. Cancer Cell 2, 289300 (2002).

39. Carmeliet, P. & Jain, R. K. Angiogenesis in cancer andother diseases. Nature 407, 249257 (2000).

40. Boudreau, N. & Myers, C. Breast cancer-inducedangiogenesis: multiple mechanisms and the role of themicroenvironment. Breast Cancer Res. 5, 140146(2003).

41. Hanahan, D. & Folkman, J. Patterns and emergingmechanisms of the angiogenic switch duringtumorigenesis. Cell 86, 353364 (1996).

42. Bando, H. & Toi, M. Tumor angiogenesis, macrophages,and cytokines. Adv. Exp. Med. Biol. 476, 267284(2000).

43. Lewis, J. S. et al. Expression of vascular endothelialgrowth factor by macrophages is up- regulated in poorlyvascularized areas of breast carcinomas. J. Pathol. 192,150158 (2000).

44. Leek, R. D. et al. Macrophage infiltration is associatedwith VEGF and EGFR expression in breast cancer. J.Pathol. 190, 430436 (2000).

45. Leek, R. D. et al. Association of macrophage infiltrationwith angiogenesis and prognosis in invasive breastcarcinoma. Cancer Res. 56, 46254629 (1996).

46. Eubank, T. D. et al. M-CSF induces vascular endothelialgrowth factor production and angiogenic activity fromhuman monocytes. J. Immunol. 171, 26372643(2003).

47. Barleon, B. et al. Migration of human monocytes inresponse to vascular endothelial growth factor (VEGF) ismediated via the VEGF receptor flt-1. Blood 87,33363343 (1996).

48. Miles, D. W. et al. Expression of tumour necrosis factor(TNF) and its receptors in benign and malignant breasttissue. Int. J. Cancer 56, 777782 (1994).

49. Eda, H. et al. Cytokines induce thymidine phosphorylaseexpression in tumor cells and make them moresusceptible to 5-deoxy-5-fluorouridine. CancerChemother. Pharmacol. 32, 333338 (1993).

50. Leek, R. D. et al. Association of tumour necrosis factor and its receptors with thymidine phosphorylaseexpression in invasive breast carcinoma. Br. J. Cancer77, 22462251 (1998).

51. Hildenbrand, R., Dilger, I., Horlin, A. & Stutte, H. J.Urokinase and macrophages in tumour angiogenesis. Br.J. Cancer 72, 818823 (1995).

52. Stacey, K. J. et al. Regulation of urokinase-typeplasminogen activator gene transcription by macrophagecolony-stimulating factor. Mol. Cell. Biol. 15, 34303441(1995).

53. Hildenbrand, R. et al. Transforming growth factor-stimulates urokinase expression in tumor-associatedmacrophages of the breast. Lab. Invest. 78, 5971(1998).