Journal of Cellular Biochemistry 101:805–815 (2007)

Tumor Microenvironment: The Role of the Tumor Stroma in Cancer

Hanchen Li, Xueli Fan, and JeanMarie Houghton* Division of Gastroenterology, Department of Medicine and Cancer Biology, University of Massachusetts Medical School, Worcester, Massachusetts 01605

Abstract The tumor microenvironment, composed of non-cancer cells and their stroma, has become recognized as a major factor influencing the growth of cancer. The microenvironment has been implicated in the regulation of cell growth, determining metastatic potential and possibly determining location of metastatic disease, and impacting the outcome of therapy. While the stromal cells are not malignant per se, their role in supporting cancer growth is so vital to the survival of the tumor that they have become an attractive target for chemotherapeutic agents. In this review, we will discuss the various cellular and molecular components of the stromal environment, their effects on cancer cell dynamics, and the rationale and implications of targeting this environment for control of cancer. Additionally, we will emphasize the role of the bone marrow-derived cell in providing cells for the stroma. J. Cell. Biochem. 101: 805–815, 2007. ß 2007 Wiley-Liss,Inc.

Key words: tumor microenvironment; tumor stroma; cancer-associated fibroblasts; neovascularization; tumor- associated inflammation; bone marrow-derived stem cells; mesenchymal stem cells

Tumor cells and their stroma co-evolve. The molecules. While none of these cells are them- logistics of which cell initiates and which cell selves malignant, due to their environment, responds to begin the process of cancer has not their interactions with each other, and directly been worked out. Research in this area aimed at or indirectly with the cancer cells, they acquire uncovering early events in carcinogenesis an abnormal phenotype and altered function. will undoubtedly broaden our understanding This abnormal interplay consisting of cell–cell of cancer formation and provide novel targets contact and active molecular crosstalk further for prevention and early eradication of lesions. drives the cancer stroma phenotype, and may What we do know about the tumor stromal result in permanent alterations in cell function. environment comes mainly from the investiga- Growth factor and chemokine production by tion of established tumors in humans, and fibroblasts and immune cells is altered leading in studies of genetically manipulated animal to direct stimulation of tumor cell growth models. The stroma consists of a compilation and recruitment of precursor cells, which of cells, including fibroblasts/myofibroblasts, themselves respond with abnormal growth glial, epithelial, fat, immune, vascular, smooth and proliferation. Malformed tumor vessels muscle, and immune cells along with the contribute to tumor hypoxia, acidosis, and extracellular matrix (ECM) and extracellular increased interstitial fluid pressures. The tumor in turn responds with a unique repertoire of gene expression which in turn acts to alter cell growth, invasion, and ultimately metastasis. The unique interplay between the various aspects of the tumor and the microenviroment Grant sponsor: NIH/NCI; Grant number: 1 RO1 CA119061; has been the recent target of molecular strate- Grant sponsor: Funderberg Gastric Cancer award. gies for tumor treatment. In order to under- *Correspondence to: JeanMarie Houghton, Division of stand the complex interplay of the cells and the Gastroenterology, Department of Medicine and Cancer Biology, LRB room 209, University of Massachusetts non-cellular stroma, we will discuss each Medical School, Worcester, MA 01605. major component separately prior to a global E-mail: [email protected] summary of the interactions between the Received 7 September 2006; Accepted 8 September 2006 components, followed by a brief discussion DOI 10.1002/jcb.21159 on targeting the stroma for therapy. Where ß 2007 Wiley-Liss, Inc. 806 Li et al. appropriate, we will focus on the contribution mucosa including fibroblast growth factor of bone marrow-derived cells (BMDC) to the 2 (FGF2), platelet-derived growth factor stroma. (PDGF), epidermal growth factor (EGF), and transforming growth factor-beta (TGF-b) [Zeisberg et al., 2000] are responsible for the CELLS OF THE STROMA conversion of a resting fibroblast to an activated Cells within the stroma include fibroblasts, fibroblast. vascular, glial, smooth muscle, epithelial, and Where do these activated fibroblasts come fat cells, and cells of the immune system. The from and where do they go? Activated fibro- most widely studied to date are the fibroblasts, blasts are largely believed to originate from a immune cells, and the vascular cells, which we pool of local fibroblasts recruited for tissue will focus on in this review. repair, however, the precise source of these cells has not been fully elucidated. After successful tissue repair, the number of acti- Fibroblasts vated fibroblasts which can be recovered in the Fibroblasts within normal tissue. In area of healing dramatically decreases. It is order to understand the role of fibroblasts in unclear, however, if the activated fibroblasts cancer, it is crucial to understand the role and revert to a non-activated status and remain the function of the normal fibroblast. In normal within the tissue, or if the activated fibroblast is tissue, fibroblasts are the predominant cell type eliminated and replaced by resting fibroblasts in the connective tissue stroma and are the from adjacent normal tissue. Data regarding primary producers of the non-cellular scaf- the reversibility or permanency of the activated folds—the extracellular matrix (ECM). Fibro- phenotype are inconclusive, but the answer to blasts are responsible for the deposition of the this question has substantial implications when fibrillar ECM—type I, type III, and type V extrapolating to tumor-associated fibroblasts collagen and fibronectin—and contribute to the and when targeting this population for anti- formation of the basement membrane by secret- tumor therapy. ing type IV collagen and laminin. The environ- Fibroblasts within tumors. Fibroblasts ment is not static; connective tissue and the are the main cellular component of tumor ECM are continually remodeled through a stroma comprising an integral component of dynamic process of ECM protein production the tumor. In some cancer types, fibroblasts and degradation by fibroblast-derived matrix constitute a larger proportion of cells within the metalloproteinases (MMPs). The turnover is, tumor than do the cancer cells. Fibroblasts however, well regulated and restrained. within tumors have an activated phenotype, During wound repair, fibroblasts are respon- and as such resemble fibroblasts in wound sible for orchestrating healing, and in order to healing. These cancer-associated fibroblasts do so become ‘‘activated,’’ with increased pro- (CAFs) are functionally and phenotypically liferation and alterations in both phenotype and distinct from normal fibroblasts that are in the secretory capacity. Production of alpha-smooth same tissue but not in the tumor environment. muscle actin (a-SMA) allows cells to migrate The distinction between these and physiologi- into areas of damage and contract for tissue cally activated fibroblasts is that they are restitution. Fibroblasts serve as a scaffold perpetually activated, neither reverting to a and secrete increased levels of ECM proteins, normal phenotype nor undergoing apoptosis growth factors, and chemotactic factors, and elimination. CAFs are identified within thereby coordinating the influx of inflammatory tumor stroma by their spindyloid appearance cells and vascular progenitor cells as well as and the expression of a-SMA; characteristics supplying the scaffold structure for cell growth shared by activated fibroblasts in wounds. and proliferation. Fibroblast activation during There are several theories put forth on the wound repair involves a dynamic crosstalk origin of CAFs including activation of a tissue between the fibroblast and the injured epithe- resident fibroblast, local cancer cells or epithe- lium. Direct contact with infiltrating immune lial cells undergoing epithelial-to-mesenchymal cells via the adhesion molecules ICAM1 and transition (EMT), or the migration and activa- VCAM1 [Clayton et al., 1998] and response tion of a marrow-derived cell. The origin and to factors secreted directly from the injured the mechanism of activation may be different in Tumor Microenvironment 807 different tissues, and may include combination few possibilities. A minority of cancer cells and of cell types and signals. When reviewing the stromal cells may share a common origin (the data supporting the activation of cells from the minority of cells demonstrating similar genetic local fibroblast pool, the signals responsible for alterations) supporting EMT of cancer cells for a conversion of a normal fibroblast to a CAF are subset of CAFs [Kurose et al., 2002; Petersen unclear. Experimentally, factors such as TGF-b et al., 2003; Tuhkanen, 2004]. More likely, these can induce normal fibroblasts to express a-SMA genetic mutations represent an independent [Ronnov-Jessen and Petersen, 1993], but it is response of the CAF cell to the cancer environ- not clear if these experimentally activated cells ment. While EMT of cancer cells to CAFs is acquire other characteristics of CAFs, and if the unlikely to account for the majority of cells phenotype is stable. These data suggest, but do within a tumor, an alternate suggestion is that not prove that factors secreted by tumor cells EMT of surrounding normal epithelial cells may themselves may drive the phenotypic and be an additional source of cells for CAF forma- functional alterations in the local fibroblast tion [Iwano et al., 2002]. pool. Recent studies have shown surprising Another theory involves EMT—the notion degrees of plasticity for BMDC and increasing that tumor cells detach, assume a more evidence for BMDC participation in the forma- mesenchymal cell like phenotype, and acquire tion of cancer. In a permissive environment, invasive and migratory ability. If EMT of cancer BMDC directly form gastric adenocarcinoma cells were the predominant source of CAFs, then [Houghton et al., 2004], skin [Aractingi et al., the genetic alterations found in the cancer cells 2005] and vascular tumors [Peters et al., 2005]. would be expected to be found within the Additionally, there is mounting evidence that stromal cells. Interestingly, studies evaluating bone marrow-derived precursor cells invade the genetic makeup of both compartments do tumors and function as CAFs [Iwano et al., find genetic alterations present in both the 2002; Direkze et al., 2004] (Fig. 1). It is not clear cancer cells and the CAFs—however, these if these cells become activated as a result of the alterations are rarely identical, suggesting a tissue environment, or if they are a subset of

Fig. 1. BMDC contribute to tumor formation. Marrow-derived cells expressing CXCR4 (pink cells) chemotax to areas of hypoxia and inflammation in response to SDF-1 (secreted by CAFs and inflammatory cells), and incorporate into vessels as ECs, into the stroma as CAFs, and as tumor cells. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.] 808 Li et al. cells within the marrow with an already by tumor cells attract hematopoietic bone activated phenotype, preferentially recruited marrow (HBM)-derived cells in a vascular to the site of tumors. MSC have been reported to endothelial growth factor-R1 (VEGF-R1) be inherently mutagenic, undergoing genetic dependent fashion. The type and quantity of alterations as a result of forced divisions factors elaborated seem to determine the [Serakinci et al., 2004; Rubio et al., 2005]. The organ to which the HBM cells will be attracted trait of accumulating genetic mutations as a to. The expression of VLA-4 allows adhesion result of proliferation would explain the high of these BMDC and local cluster formation, degree of genetic alterations found in these a prerequisite for subsequent retention of cells, distinct from the alterations found in the malignant cells. Elaboration of MMP-9 from surrounding tumor cells. An additional con- local fibroblasts and the BMDC alters the local sideration is whether these marrow-derived environment, releases Kit-ligand and VEGF-A cells fuse with peripheral cells or with cancer which in turn support the growth and develop- cells to acquire the CAF phenotype or if they ment of c-kit expressing cancer cells [Kaplan function independent of fusion. Fusion followed et al., 2005]. In the model used, metastasis did by a reductive division could potentially not occur in the absence of BMDC recruitment, explain some of the findings of EMT—with the stressing the role for BMDC in orchestrating simultaneous expression of epithelial and the growth and aggressive behavior of cancer. mesenchymal markers, and the acquisition of Immune Cells of the Stroma similar genetic damage, however, the pheno- menon of fusion between BMDC and cancer Monocytes/macrophages, neutrophils, and cells has not been fully addressed. lymphocytes are recruited to and reside in How CAFs function to assist cancer growth. the tumor stroma. Monocytes are actively The role for CAFs in promoting cancer has been recruited into tumors along defined chemotactic demonstrated in established tumors, carci- gradients. Once in the tumor, they differentite noma in situ, the initiation of cancer, and in into tumor-associated macrohages (TAMs). orchestrating metastasis. Indirect data exam- TAMs appear to be preferentially attracted to ining fibroblasts from peripheral sites in and retained in areas of necrosis and hypoxia patients with a variety of cancers show these where they become phenotypically altered peripheral fibroblasts have more of an activated and upregulate hypoxia-induced transcription phenotype when compared with fibroblasts factors (see below). Macrophages also release a from patients without cancer, with an increase number of factors that influence endothelial cell in proliferation and a reduction in growth behavior including VEGF, HGF, MMP2, IL-8. requirements [Kopelovich, 1982; Schor, 1986]. Neutrophils are recognized as stimulators of These data suggest, but do not prove the angiogenesis, via their release of VEGF, HGF, association between activation of fibroblasts MMP2, and IL-8. Additional immune cell and an increased predilection for cancer. The populations have a less well-documented role use of mouse models of cancer employing in carcinogenesis, and are not consistent resi- fibroblasts engineered to overexpress hepato- dents of the stroma with their presence cyte growth factor (HGF) or TGF-b has shown restricted to specific types of tumors. These that activated fibroblasts can initiate cancer at include myeloid suppressor cells, which have divergent sites including stomach and prostate the phenotypic characteristics of both macro- [Bhowmick et al., 2004; Kuperwasser et al., phages and granulocytes and the diverse effects 2004]. Additionally, once a lesion is initiated, of immune suppression, production of MMP9 CAFs have been shown to assist in proliferation and VEGF, and the additional ability to directly and progression of cancer through the produc- incorporate into vessel walls [Serafini et al., tion of growth factors and chemotactic factors, 2004; Yang et al., 2004]. Mast cells are asso- angiogenesis factors, and MMPs (see below, ciated with angiogenesis and tumor induction in non-cellular components of the stroma), allow- skin cancer models [Coussens et al., 2000; ing invasion and spread of cancer cells. Crivellato and Ribatti, 2005], while eosinophils, Creation of the metastatic niche. Perhaps associated with a host of solid tumors including the most intriguing data regarding the role of colon, cervical, lung, breast, and ovary as well as the stroma in cancer have been the discovery some lymphomas, can influence angiogenesis of the pre-metastatic niche. Factors elaborated via VEGF production. Tumor Microenvironment 809

Vascular Cells/Tumor Angiogenesis including budding of existing vascular net- works, recruitment of vascular progenitor cells Tumors require the formation of a complex to form new vascular channels, or ‘‘vascular vascular network to meet the metabolic and mimicry’’—a process by which tumor cell-lined nutritional needs for growth. VEGF is the main channels contribute to the blood tributaries factor involved in the formation of tumor supplying the tumor [Ribatti et al., 2003]. vessels. It is secreted by the tumor cells directly Budding of existing vessels occurs in response and by fibroblasts and inflammatory cells in the to local growth signals emanating from tumor stroma and is responsible for the ‘‘angiogenic cells or from entrapped bone marrow-derived switch’’ where new vasculature is formed to hematopoietic cells (HCs) within the tumors supply the tumor with nutrients. Tumor vessels [Takura, 2006], CAFs, and inflammatory cells. formed as a result of VEGF are abnormal; they Heparin sulfate proteoglycans on the tumor cell are non-uniformly distributed and irregularly surface and in the extracellular environment shaped, inappropriately branched and tortu- [Stringer, 2006] and the SDF-1/CXCL12 ous, often ending blindly. They do not have the axis [Genis et al., 2006] are critical signaling classic hierachical arrangement of arterioles, pathways for growth of new vessels from pre- venules, and capillaries and often form arterio- existing vasculature (Fig. 2). venous shunts. These vessels are variably In addition to sprouting from pre-existing fenestrated and leaky leading to high inter- vessels, progenitor cells are recruited from stitial pressures, further exacerbating tissue more distant sites for neovascularization. HCs hypoxia and stimulating additional VEGF and endothelial cells (ECs) share a common production [Carmeliet, 2005; Dvorak, 2005]. ancestry during embryogenesis, the hemagio- Under the influence of VEGF, tumor vessels blast. Postnatally, CD133, CD34, and VEGFR-2 are formed by one of several mechanisms positive cell subsets in the bone marrow,

Fig. 2. CAFs drive tumor angiogenesis. CAFs secrete multiple factors which drive angiogenesis including VEGF, MMP, SPARC, SDF-1. VEGF is the primary growth factor driving the abnormal tumor vessel growth. MMPs have a direct effect on angiogenesis through degradation of basement membrane and ECM proteins as well as an indirect effect via proteolytic cleavage of angiogenic factors such as SPARC. SDF-1 secreted by fibroblasts as well as other cell types in the stroma is chemotactic for EPC. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.] 810 Li et al. peripheral blood, and cord blood maintain the TGF-b, cancer cells either lose this growth functional activity of the hemangioblasts in that inhibitory response or usurp the pathway such they are able to differentiate into both HCs and that TGF-b stimulates proliferation of the ECs [Ribatti et al., 2002]. The existence of this cancer cell. Having lost or altered their own adult hemangioblast-like cell along with addi- response, many tumors gain the ability to tional in vivo data upholds the concept that express TGF-b which then acts in an autocrine endothelial progenitor cells (EPCs) of bone fashion to further stimulate growth, facilitate marrow origin exist, circulate, and contribute tumor invasion, and to protect the tumor to adult angiogenesis. To date, the best marker from attack by the immune system. Effects on of these vascular progenitor cells is the ex- the vascular system and tumor angiogenesis pression of CD133. VEGF in the circulation include migration of vascular progenitor cells along with hypoxia mobilize marrow-derived and growth control. CAFs are stimulated EPCs [Takahashi et al., 1999] which are then to proliferate, increase ECM production, and recruited to tumor sites via a host of chemokines secrete cytokines. Additionally, TGF-b induces such as VEGF, angiopoietin, stromal cell- the EMT in cancer cells aiding local invasion as derived factor (SDF)-1a [Hattori et al., 2001; well as facilitating metastatic spread [Siegel Iwaguro et al., 2002; Yamaguchi et al., 2003] and Massague, 2003]. secreted by the tumor cells and cells of the Mesenchymal and marrow-derived stromal stroma. Once recruited, the EPC cells recruit cells constitutively secrete SDF-1. Expression is additional EPCs through their own production upregulated in inflamed tissues [Houghton of VEGF, HGF, G-CSF, and GM-CSF. The et al., 2004], in wounds and in cancer, where it contribution of bone marrow-derived EPCs acts to attract cells expressing the receptor, (BMD-EPC) to tumor vasculature is usually CXCR4. CXCR4 is expressed by an array of low. However, the actual percentage is depen- cancer cells, and the receptor ligand interaction dant upon the tumor type, and in some functions as a mitogen for tumor cells and instances, these cells can contribute a sub- induces migration in a gradient-specific fashion stantial portion of ECs. resulting in local invasion as well as enabling cells to metastasize to distant sites expressing EXTRACELLULAR MOLECULES SDF-1 such as the bone marrow and peripheral organs. Marrow-derived EPC also chemotax to a Cytokines and Growth Factors SDF-1 gradient, and are recruited to tumor sites Cytokines and growth factors are secreted by for neovascularization [reviewed in Berger and cells of the stroma as well as by cancer cells into Kipps, 2006]. the stroma. These factors are numerous, but for SPARC, known as osteonectin or BM-40, is this review we will concentrate on the factors secreted by fibroblasts and inflammatory cells of which have conclusively and consistently been the stroma. SPARC is secreted in a bioactive shown to directly impact tumor behavior. These form and can undergo further processing by the include TGF-b, SDF-1, secreted protein acidic MMPs (see below) to release additional bio- rich in cysteine (SPARC), MMP, VEGF, and active fragments. Functions of SPARC include HIF-1-a. cell–cell deadhesion, inhibition of proliferation, TGF-b is best known as a growth inhibitor regulation of angiogenesis, and modulation of and a potent immunosuppressive factor. Under ECM production and composition [Sangeletti normal physiological conditions, TGF-b main- et al., 2003]. SPARC functions to recruit macro- tains tissue homeostasis through its effects on phages, leukocytes, and BMDC to tumors; these proliferation and apoptosis, effectively controll- cells in turn contribute to the functioning of the ing the growth of epithelium, endothelium, stroma through direct cell–cell interactions and neuronal tissue, and HCs. TGF-b also has release of soluble factors. strong inhibitory effects on the immune Degradation and Remodeling Enzymes response, including reducing T-cell prolifera- tion, inhibiting natural killer cell function, and In order for tumor cells to invade normal interfering with antigen presentation. The net tissue they must lose their connection to each effect is to induce tolerance and prevent other, and invade and integrate into the immune rejection of tissue. While proliferation surrounding normal structures. Normal tissue of normal epithelium is strongly inhibited by is designed to resist such breaches, making Tumor Microenvironment 811 invasion an active process on the part of the basement membranes and allow cell movement cancer cell. Tumor angiogenesis requires active [Genis et al., 2006]. Increased expression of remodeling and integration of new cells into MT1-MMP is seen in glioma, papillary thyroid, existing structures. To facilitate restructuring, breast cancer, and gastric cancer suggesting a fibroblasts, macrophages, and ECs of the role, however, as the absence of MT1-MMP stroma express and secrete MMPs. Secretion causes severe skeletal defects and early death, of these molecules by the stromal cells is the in vivo cancer models are lacking. MMP-1, result of a complex tumor-stroma crosstalk, derived from stromal fibroblasts, functions involving multiple ligands and cellular signal- to cleave and activate PAR1 creating PAR1- ing pathways [Stuelten et al., 2005]. As a family dependent Ca2þ signaling for growth, invasion, of compounds, the MMPs act to hydrolyze and migration [Pei, 2005]. MMP-2, also fibro- the extracellular proteins of the surrounding blast derived, inhibits bFGF-induced angio- tissue which include collagen, laminin, elastin, genesis with resulting alterations in tumor fibrinogen, fibronectin, and vitronectin [Stern- growth. Mice deficient in MMP2 have markedly licht and Werb, 2001; Boire et al., 2005]. From reduced tumor angiogenesis and a delay the earliest work addressing the role of MMPs in tumor progression. Additional targets for and cancer, there has been a clear connection MMP2 include connective tissue growth factor between certain members of the MMP group, (CTGF) galectin-1, osteopontin, death receptor, degradation of the ECM, and local cancer cell heat shock protein 90 (HSP90), procollagen C invasion [Liotta et al., 1980], but the association proteinase enhancer protein (PCPE), and the with metastatic disease has been less clear. In membrane bound chemokine, fructalkine. addition to the ECM, MMPs have additional Fractalkine is released from its membrane target proteins which include other proteinases, location by MMP2 as a fully functional domain. proteinase inhibitors, clotting factors, chemo- Additionally, MMP-2 cleaves fractalkine at kines and chemotactic factors, growth factors, a positions 4–5, releasing an antagonist form variety of cell surface receptors, and cell matrix [Overall and Dean, 2006] stressing the some- adhesion molecules [Coussens et al., 2002]. times opposing actions of the MMPs. Macro- MMPs have also been implicated in initiating phage-derived MMP9 is responsible for VEGF the EMT and in promoting genomic instability mobilization from matrix stores—the primary [Radisky et al., 2005], affording them a signal for neovascularization. MMP9 is also prominent role in both tumor progression and critical for the recruitment and engraftment prevention. In light of the diverse functions of of BMDC into tumor vasculature [Jodele the MMP molecules, mouse models of cancer et al., 2005]. The contribution of BMDC to the examining knockout or knockdown of indi- vasculature differs greatly between tumor vidual members of the MMP family give varying types, and this contribution will help determine phenotypes, dependent upon which MMP is the impact that MMP9 inhibition has on tumor altered and what cancer model is used. We angiogenesis. Macrophage elastase (MMP12) briefly describe the prominent members of this has been less studied, but appears to inhibit the family and their association with cancer. growth of cancer. In a Lewis lung cancer model a Membrane type matrix metalloprotineases deficiency of MMP12 suppressed the growth of (MT-MMPs) are a subclass of the MMPs which lung metastasis, but not the number of meta- can be expressed on cells of the stroma, and by static foci [Houghton et al., 2006]. Less is known cancer cells themselves. While most of the about the function of MMP12 in other models. MMPs are secreted proteins, the MT-MMPs As a group, MMP expression is prominent in are membrane bound with a cytoplasmic tumor stroma, and plays a vital role in the domain important for cell signaling. MT1- local growth, invasion, and migration of cancer MMP is the best characterized of this group cells. and has been shown to activate MMP2, display Tissue inhibitors of metalloproteinases its own proteolytic activity against the ECM, (TIMPs) also play a role in local growth and and to play a role in bone degradation/forma- spread of tumors. TIMP-1 has been shown to tion. During angiogenesis, quiescent ECs have an anti-apoptotic effect on bone marrow become activated and migrate to areas of stromal cells in culture mediated through the neovascularization. These cells release MMPs, PI3-kinase and JNK signaling pathways, and is specifically MT1-MMP, in order to degrade independent of the effects of TIMP on MMP 812 Li et al. activity [Guo et al., 2006]. These findings MMPs also cleave cytokines with variable suggest an effect of the TIMPs on cells within effects. For example, cleavage of IL-1b leads to the tumor environment which are of marrow both activation and inactivation depending origin and include bone marrow-derived tumor upon the cleavage site, and processing of fibroblasts, bone marrow-derived ECs within monocyte chemotactic protein (MCP) by the tumor vasculature, and bone marrow-derived MMPs leads to its downregulation. Direct tumor cells. release of HGF, IGF, and fibroblast activation VEGF is secreted by CAFs and is implicated protein (FAP) drives tumor cell growth while in many aspects of cancer growth including also functioning as immune modulators angiogenesis, ECM remodeling, generation of [reviewed in Silzle and Randolph, 2004]. inflammatory cytokines, and hematopoietic stem cell development. It is a potent vascular growth factor and the main stimulus for HYPOXIA AND THE tumor angiogenesis (see section on Tumor TUMOR MICROENVIRONMENT Angiogenesis). VEGF acts in the generation of Hypoxia in solid tumors is the result of inflammatory cytokines and is responsible structurally and functionally inadequate ves- for the upregulation of several other pro- sels and high metabolic demand, and is further angiogenic molecules and prometastatic mole- aggravated by cancer-induced anemia. Hypoxia cules. VEGF plays a crucial role in recruiting correlates with aggressive behavior of tumors, VEGF-R1 positive HBM progenitor cells to and resistance to therapy. Hypoxia-inducible peripheral sites to initiate the pre-metastatic factors (HIFs) are cellular transcription factors niche [Kaplan et al., 2005]. The main stimulus involved in the response to environmental for production in tumor stroma is hypoxia, stress. Important targets of the HIF system resulting in the recruitment of endothelial which are relevant to cancer biology include progenitor and mesenchymal stem cells to sites MDR-1 [Comerford et al., 2002], IGF-2 [Feldser of ischemia [Okuyama et al., 2006], which in et al., 1999], telomerase [Nishi et al., 2004], and turn secrete additional VEGF. CXCR4/SDF-1 [Ceradini et al., 2004]. The HIFs orchestrate neovasularization, glucose metabo- IMMUNE REGULATION BY THE STROMA lism, survival, and tumor spread in response to hypoxia and are major regulators of tumor cell Through the secretion of cytokines, chemo- adaptation to hyoxic stress. In addition, cells kines, and other factors, stromal cells are with genomic and proteomic changes favoring instrumental in creating the unique environ- survival under hypoxic conditions will pro- ment of chronic inflammation and immune liferate, thereby increasing metabolism and tolerance, allowing cancer cells exposure to demand for nutrients, exacerbating hypoxic growth factors while avoiding immune- conditions which in turn will lead to the mediated elimination. The individual compo- selection and expansion of more aggressive nents of the environment may differ between clones. VEGF production, induced by hypoxia, tumor types and models studied, with immune leads to inadequate neovascularization, inade- enhancing and immune suppressing pathways quate tissue oxygenation, and a cycle of addi- simultaneously activated. However, in order for tional VEGF production (see sections on Tumor the tumor to survive, the net result needs to be Angiogenesis and Extracellular Molecules). suppression of any immune response directed toward the tumor cells. Stromal cells are the main source of thrombospondin-1 (TSP-1) CONTRIBUTIONS OF THE which has both positive and negative effects MICROENVIRONMENT TO GENETIC on angiogenesis, and effects on tumor-tissue INSTABILITY AND EPIGENETIC CHANGES mediated immune suppression via activation of TGF-b and direct interaction with immune cells As tumors progress, the cells display [reviewed in Silzle and Randolph, 2004]. MMP increased genetic instability. The number of cleavage products of several proteins such as mutations found in tumor cells cannot be fibronectin and collagen are chemotactic for accounted for by the rate of mutations occurring leukocytes, modulate their proliferation and in somatic cells, leading to the notion that cytokine release [Barilla and Crsons, 2000]. cancer cells develop a mutator phenotype. Tumor Microenvironment 813

Conditions within the tumor microenvironment structures, hypoxia, and pH alterations. These including oxidative stress, hypoxia, nutrient challenges should not dampen our enthusiasm deprivation, and low pH contribute to genetic for targeting the stroma, but successful appro- instability through the induction of increased aches will require identifying appropriate tar- DNA damage enhanced mutagenesis and gets and designing efficient delivery methods. impaired DNA damage pathways [for a full Examples of effective anti-stromal therapy review of this topic see Bindra and Glazer, include targeting cancer-associated inflamma- 2005]. tion through the use of non-steroidal anti- inflammatory drugs (NSAIDs). NSAID use is associated with a decrease in the incidence of TARGETING THE MICROENVIRONMENT colorectal cancer [Sandler et al., 2003] and the FOR CANCER CHEMOTHERAPY precursor lesion, adenomas [Baron et al., 2003]. The tumor stroma, consisting of cells, struc- Prevention of other types of cancer has been tural proteins, and signaling molecules, is inferred but not yet confirmed by large trials recognized as playing a central role in tumor [Schreinemachers and Everson, 1994; Castelao initiation, progression, and metastasis (Fig. 3). et al., 2000; Thun et al., 2002], adding support to The level of aggression is greatly influenced by targeting inflammation for cancer prevention. this environment, providing multiple targets The strong association between MMP expres- for anti-cancer therapy. Targeting the stroma sion and cancer spurred the development of poses several obstacles, however. The fibro- biologically active MMP inhibitors, and subse- blasts, ECs, and inflammatory cells are them- quent clinical trials were designed to test the selves not malignant, therefore successful efficacy in a range of tumor types. Results from therapy needs to aim at phenotypic changes these trials have been disappointing but not unique to this population, while avoiding entirely unexpected when one considers all the normal cells elsewhere. Additionally, delivery diverse functions of the various MMPs and of agents to the stroma can be problematic realizes that only non-selective MMP inhibitor because of insufficient and defective vascular drugs entered trials. Also, MMP inhibition was employed at late stages of disease, where as pre- clinical data suggest the most efficacious time for therapy may be early in disease progression. As the net outcome of global MMP inhibition is dependent upon the interplay of multiple factors including the tissue where the tumor arises as well as the stage during which the drugs are given, the current push is for the development of highly selective MMP- inhibitors targeting single family members, and choosing the study population based on known alterations within specific cancer micro- environments. Present efforts targeting molecules unique to the tumor microenvironment will provide strategies for modifying the tumor environ- ment, while avoiding interrupting normal tis- sue homeostasis. The success of VEGF blockade [Presta et al., 1997] has encouraged blocking Fig. 3. The tumor stroma contributes to genetic instability of cancer cells. Hypoxia, inadequate delivery of nutrients, additional unique pathways, such as HIF-1 and decreased pH and free radical formation created within the HIF-1 target genes, in the hope that modifying stroma as a result of insufficient blood supply, and factors released the stroma alone or used in combination with by inflammatory cells and activated fibroblasts drive DNA conventions like radiation and chemotherapy to mutations and the suppression of DNA repair machinery. The increase efficacy. It is through understanding mutator phenotype of cancer cells allows the creation and selection of aggressive clones with greater metastatic potential. the complex interactions of the stroma with the [Color figure can be viewed in the online issue, which is available tumor that these therapies can be devised, and at www.interscience.wiley.com.] clinical trials developed which will allow the 814 Li et al. agents to be tested in appropriate clinical Coussens LM, Fingleton B, Matrisian LM. 2002. Matrix situations. metalloproteinase inhibitors and cancer: Trials and tribulations. Science 295:2387–2392. ACKNOWLEDGMENTS Crivellato F, Ribatti D. 2005. Involvement of mast cells in angiogenesis and chronic inflammation. Curr Drug This study was supported by grants from Targets Inflamm Allergy 4:9–11. NIH/NCI grant (1 RO1 CA119061 to G.H.), Direkze NC, Hodivala-Dilke K, Jeffery R, Hunt T, Poulsom R, Oukrif D, Alison MR, Wright NA. 2004. Bone marrow Funderberg Gastric Cancer award to J.H. and contribution to tumor associated myofibroblasts and the Our Danny Cancer Fund Fellowship award fibroblasts. Cancer Res 64:8492–8495. to H.L. Dvorak HF. 2005. Angiogenesis: Update. J Thromb Hae- most 3:1835–1842. REFERENCES Feldser D, Agani F, Iyer NV, Pak B, Ferreira G, Semenza GL. 1999. Reciprocal positive regulation of hypoxia Aractingi SJ, Kanitakis S, Euvrard C, Le Danff I, Peguillet inducible factor 1 alpha and insulin like growth factor K, Khosrotehrani K, Latz O, Carosella ED. 2005. Skin 2. Cancer Res 59:3915–3918. carcinoma arising from donor cells in a kidney transplant Genis L, Galvez BG, Gonzalo P, Arroyo AG. 2006. MT1- recipient. Cancer Res 65:1755–1760. MMP: Universal or particular player in angiogenesis? Barilla ML, Crsons SE. 2000. Fibronectin fragments and Cancer Metastasis Rev 25:77–86. their role in inflammatory arthritis. Semin Arthritis Guo LJ, Luo XH, Xie H, Zhou HD, Yuan LQ, Wang M, Liao Rheum 29:252–265. EY. 2006. Tissue inhibitor of matrix metalloproteinase-1 Baron JA, Cole BF, Sandler RS, Haile RW, Ahnen D, suppresses apoptosis of mouse bone marrow stromal cell Bresalier R, McKeown-Eyssen G, Summers RW, Roth- line MBA-1. Calcif Tissue Int 78:285–292. stein R, Burke CA, Snover DC, Church TR, Allen JI, Hattori K, Dias S, Heissig B, Hackett NR, Lyden D, Tateno Beach M, Beck GJ, Bond JH, Byers T, Greenberg ER, M, Hicklin DJ, Zhu Z, Witte L, Crystal RG, Moore MA, Mandel JS, Marcon N, Mott LA, Pearson L, Saibil F, van Rafii S. 2001. Vascular endothelial growth factor and Stolk RU. 2003. A randomized trial of aspirin to prevent angiopoietin-1 stimulate postnatal hematopoiesis by colorectal adenomas. N Engl J Med 348(10):891–899. recruitment of vasculogenic and hematopoietic stem Berger JA, Kipps TJ. 2006. CXCR4: A key receptor in the cells. J Exp Med 193:1005–1014. crosstalk between tumor cells and their microenviron- Houghton J, Stoicov C, Nomura S, Rogers AB, Carlson J, Li ment. Blood 107:1761–1767. H, Goldenring J, Fox JG, Wang TC. 2004. Gastric cancer Bhowmick NA, Chytil A, Plieth D, Gorska AE, Dumont N, originating from bone marrow-derived cells. Science Shappell S, Washington MK, Neilson EG, Moses HL. 306:1568–1571. 2004. TGF-b signaling in fibroblasts modulates the Houghton AM, Grisolano JL, Baumann ML, Kobayashi oncogenic potential of adjacent epithelia. Science 303: DK, Hautamaki RD, Nehring LC, Cornelius LA, Shapiro 848–851. SD. 2006. Macrophage elastase (matrix metalloprotei- Bindra RS, Glazer PM. 2005. Genetic instability and the nase-12) suppresses growth of lung metastases. Cancer tumor microenvironment: Towards the concept of micro- Res 66:6149–6155. environment-induced mutagenesis. Mutat Res 569:75– Iwaguro H, Yamaguchi J, Kalka C, Murasawa S, Masuda 85. H, Hayashi S, Silver M, Li T, Isner JM, Asahara T. 2002. Boire A, Covic L, Agarwal A, Jacques S, Sherifi S, Endothelial progenitor cell vascular endothelial growth Kuliopulos S. 2005. Par1 is a matrix metalloprotease-1 factor gene transfer for vascular regeneration. Circula- receptor that promotes invasion and tumorigenesis of tion 105:732–738. breast cancer cells. Cell 120:303–313. Iwano M, Plieth D, Danoff TM, Xue C, Okada H, Neilsonet Carmeliet P. 2005. VEGF as a key mediator of angiogenesis EG. 2002. Evidence that fibroblasts derive from epithe- in cancer. Oncology 69:4–10. lium during tissue fibrosis. J Clin Invest 110:341– Castelao JE, Yuan JM, Gago-Dominguez M, Yu MC, Ross 350. RK. 2000. Non-steroidal anti-inflammatory drugs and Jodele S, Chantrain CF, Blavier L, Lutzko C, Crooks GM, bladder cancer prevention. Br J Cancer 82:1364–1369. Shimada H, Coussens LM, Declerck YA. 2005. The Ceradini DJ, Kulkarni AR, Callaghan MJ, Tepper OM, contribution of bone marrow derived cells to the tumor Bastidas N, Kleinman ME. 2004. Progenitor cell traffick- vasculature in neuroblastoma is matrix metalloprotei- ing is regulated by hypoxic gradients through HIF-1 nase-9 dependent. Cancer Res 65:3200–3208. induction of SDF-1. Nat Med 10:858–864. Kaplan RN, Riba RD, Zacharoulis S, Bramley AH, Vincent Clayton A, Evans RA, Pettit E, Hallett M, Williams JD, L, Costa C, MacDonald DD, Jin DK, Shido K, Kerns SA, Steadman R. 1998. Cellular activation through the Zhu Z, l Hicklin D, Wu Y, Port JL, Altorki N, Port ER, ligation of intercellular adhesion molecule-1. J Cell Sci Ruggero D, Shmelkov SV, Jensen KK, Rafii S, Lyden D. 111:443–453. 2005. VEGFR1-positive haematopoietic bone marrow Comerford KM, Wallace TJ, Karhausen J, Louid NA, progenitors initiate the pre-metastatic niche. Nature Montalto MC, Colgan SP. 2002. Hypoxia inducible 438:820–827. factor-1 dependent regulation of the mutlidrug resistance Kopelovich L. 1982. Genetic predisposition to cancer in (MDR1) gene. Cancer Res 62:3387–3394. man: In vitro studies. Int Rev Cytol 77:63–88. Coussens LM, Tinkle CL, Hanahan D, Werb Z. 2000. MMP- Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, 9 supplied by bone marrow derived cells contributes to Carey L, Richardson A, Weinberg RA. 2004. Reconstruc- skin carcinogenesis. Cell 103:481–490. tion of functionally normal and malignant human breast Tumor Microenvironment 815

tissues in mice. Proc Natl Acad Sci USA 101:4966– Sangeletti S, Stoppacciaro A, Guiducci C, Torrisi MR, 4971. Colombo MP. 2003. Leukocyte rather than tumor- Kurose K, Gilley K, Matsumoto S, Watson PH, Zhou XP, produced SPARC determines stroma and collagen type Enget C. 2002. Frequent somatic mutations in PTEN and IV deposition in mammary carcinoma. J Exp Med 198: TP53 are mutually exclusive in the stroma of breast 1475–1485. carcinomas. Nat Genet 32:355–357. Schor SL. 1986. Occurrence of a fetal fibroblast phenotype Liotta LA, Tryggvason K, Garbisa S, Hart I, Foltz CM, in familial breast cancer. Int J Cancer 37:831–836. Shaffie S. 1980. Metastatic potential correlates with Schreinemachers DM, Everson RB. 1994. Aspirin use and enzymatic degradation of basement membrane collagen. lung, colon, and breast cancer incidence in a prospective Nature 284:67–68. study. Epidemiology 5:138–146. Nishi H, Nakada T, Kyo S, Inoue M, Shay JW, Isaka K. Serafini P, De Santo C, Marigo I. 2004. Derangement of 2004. Hypoxia inducible factor 1 mediates upregulation immune responses by meyloid suppressor cells. Cancer of telomerase (hTERT). Mol Cell Biol 24:6076–6083. Immunol Immunother 53:64–72. Okuyama H, Krishnamachary B, Zhou YF, Nagasawa H, Serakinci NP, Guldberg JS, Burns B, Abdallah H, Schrod- Bosch-Marce M, Semenza GL. 2006. Expression of der H, Jensen T, Kassem M. 2004. Adult human vascular endothelial growth factor receptor 1 in bone mesenchymal stem cell as a target for neoplastic marrow derived mesenchymal cells is dependent on transformation. Oncogene 23:5095–5098. hypoxia inducible factor 1. J Biol Chem 281:15554–15563. Siegel PM, Massague J. 2003. Cytostatic and apoptotic Overall CM, Dean RA. 2006. Degradomics: Systems biology actions of TGF-b in homeostasis and cancer. Nature 3: of the protease web. Pleiotropic roles of MMPs in cancer. 807–820. Cancer Metastasis Rev 25:69–75. Silzle T, Randolph GJ. 2004. The fibroblast: Sentinel Pei D. 2005. Matrix metalloproteinases target protease- cell and local immune modulator in tumor tissue. Int activated receptors on the tumor cell surface. Cancer J Cancer 108:173–180. Cells 7:207–208. Sternlicht MD, Werb Z. 2001. How matric metalloprotei- Peters BA, Diaz LA, Polyak K, Meszler L, Romans K, nases regulate cell behavior. Annu Rev Cell Dev Biol 17: Guinan EC. 2005. Contribution of bone marrow-derived 463–516. endothelial cells to human tumor vasculature. Nat Med Stringer SE. 2006. The role of heparan sulphate proteo- 11:261–262. glycans in angiogenesis. Biochem Soc Trans 34:451– Petersen OW, Nielsen HL, Gudjonsson T, Villadsen R, 453. Rank F, Niebuhr E, Bissell MJ, Rønnov-Jessen L. 2003. Stuelten CH, DaCosta Byfield S, Arany PR, Karpova TS, Epithelial to mesenchymal transition in human breast Stetlerr-Stevenson WG, Roberts AB, Anita B. 2005. cancer can provide a nonmalignant stroma. Am J Pathol Breast cancer cells induce stromal fibroblasts to express 162:391–402. MMP-9 via secretion of TNF-alpha and TGF-beta. J Cell Presta LG, Chen H, O’Connor SJ, Chisholm V, Meng YG, Sci 118:2143–2153. Krummen L, Winkler M, Ferrara N. 1997. Humanization Takahashi T, Kalka C, Masuda H, Chen D, Silver M, of an anti-vascular endothelial growth factor monoclonal Kearney M, Magner M, Isner JM, Asahara T. 1999. antibody for the therapy of solid tumors and other Ischemia and cytokine induced mobilization of bone disorders. Cancer Res 57:4593–4599. marrow derived endothelial progenitor cell for neovascu- Radisky DC, Levy DD, Littlepage LE, Liu H, Nelson CM, larization. Nat Med 5:434–438. Fata JE, Leake D, Godden EL, Albertson DG, Nieto MA, Takura N. 2006. Role of hematopoietic lineage cells as Werb Z, Bissell MJ. 2005. Rac1b and reactive oxygen accessory components in blood vessel formation. Cancer species mediate MMP3-induced EMT and genomic Sci 97:568–574. instability. Nature 436:123–127. Thun MJ, Henley SJ, Patrono C. 2002. Nonsteroidal anti- Ribatti D, Vacca A, Nico B, Ria R, Dammacco F. 2002. Cross inflammatory drugs as anticancer agents: Mechanistic, talk between hematopoiesis and angiogenesis signaling pharmacologic, and clinical issues. J Natl Cancer Inst pathways. Curr Mol Med 2:537–547. 94:252–266. Ribatti D, Vacca A, Dammacco F. 2003. New non-angiogen- Tuhkanen H. 2004. Genetic alterations in the peritumoral esis dependent pathways for tumor growth. Eur J Cancer stromal cells of malignant and borderline epithelial 39:1835–1841. ovarian tumors as indicated by allelic imbalance on Ronnov-Jessen L, Petersen OW. 1993. Induction of a- chromosome 3p. Int J Cancer 109:247–252. smooth muscle actin by transforming growth factor-b1in Yamaguchi J, Kusano KF, Masou O, Kawamoto A, Silver quiescent human breast gland fibroblasts. Implications M, Murasawa S, Bosch-Marce M, Masuda H, Losordo for myofibroblast generation in breast neoplasia. Lab DW, Isner JM, Asahara T. 2003. Stromal cell-derived Invest 68:696–707. factor 1 effects an ex vivo expanded endothelial progeni- Rubio D, Garcia-Castro J, Martin MC, de la Fuente R, tor cell recruitment for ischemic neovascularization. Cigudosa JC, Lloyd AC, Bernad A. 2005. Spontaneous Circulation 107:1316–1322. human adult stem cell transformation. Cancer Res Yang L, DeBusk LM, Fukuda K. 2004. Expansion of 65:3035–3039. myeloid immune suppressor GrþCD11bþ cells in tumor Sandler RS, Halabi S, Baron JA, Budinger S, Paskett E, bearing host directly promotes tumor angiogenesis. Keresztes R, Petrelli N, Pipas JM, Karp DD, Loprinzi CL, Cancer Cell 6:409–421. Steinbach G, Schilsky R. 2003. A randomized trial of Zeisberg M, Strultz F, Muller GA. 2000. Role of fibroblast aspirin to prevent colorectal adenomas in patients with activation in inducing interstitial fibrosis. J Nephrol 13: previous colorectal cancer. N Engl J Med 348:883–890. S111–S120.