|Home | About | Journals | Submit | Contact Us | Français|
The immune system of cancer patients recognizes tumor-associated antigens expressed on solid tumors and these antigens are able to induce tumor-specific humoral and cellular immune responses. Diverse immunotherapeutic strategies have been used in an attempt to enhance both antibody and T cell responses to tumors. While several tumor vaccination strategies significantly increase the number of tumor-specific lymphocytes in the blood of cancer patients, most vaccinated patients ultimately experience tumor progression. CD4+ and CD8+ T cells with an effector memory phenotype infiltrate human tumor microenvironments, but most are hyporesponsive to stimulation via the T cell receptor (TCR) and CD28 under conditions that activate memory T cells derived from the peripheral blood of the cancer patients or normal donors. Attempts to identify cells and molecules responsible for the TCR signaling arrest of tumor-infiltrating T cells have focused largely upon the immunosuppressive effects of tumor cells, tolerogenic dendritic cells and regulatory T cells. Here we review potential mechanisms by which human T cell function is arrested in the tumor microenvironment with a focus on the immunomodulatory effects of stromal fibroblasts. Determining in vivo which cells and molecules are responsible for the TCR arrest in human tumor-infiltrating T cells will be necessary to formulate and test strategies to prevent or reverse the signaling arrest of the human T cells in situ for a more effective design of tumor vaccines. These questions are now addressable using novel human xenograft models of tumor microenvironments.
The presence of tumor-infiltrating lymphocytes (TIL) in the microenvironment of spontaneous human tumors remains a histopathological variable that is notable and potentially exploitable, but not yet fully understood [1–4]. The type, density and location of immune cells within human tumors have been shown to be a valuable prognostic indicator and suggested to be superior to and independent of factors currently used to stage tumors . Moreover, there is compelling evidence demonstrating in both animals and humans that tumors are able to provoke adaptive anti-tumor immune responses mediated by B cells and CD4+ and CD8+ T cells and to initiate innate immune responses mediated by NK or NKT cells [6–12]. For example, the tumor-associated antigens NY-ESO-1 and LAGE-1 have been shown to be expressed on epithelial ovarian cancers and have been used in tumor vaccines to induce both anti-tumor antibody and tumor-specific CD4+ and CD8+ T cell responses in patients with ovarian cancer [13–15].
The fundamental question that remains is why human tumors progress in spite of the presence of a broad array of both adaptive and innate immune cells that are present within the tumor microenvironment. Multiple and diverse explanations have been offered to explain the failure of the immunocompetent cells to control tumor progression (some of which are reviewed in Fig. 1). These explanations fall largely into two categories. In the first category, the dysfunction of the TIL is suggested to be due to factors that are extrinsic to the TIL, many of which are thought to be orchestrated by the tumor cells themselves or by by-products secreted by tumor cells or surrounding tumor stroma. These include immunosuppressive cytokines (IL-10, TGF-β1), Fas-ligand-induced killing of TIL, immune editing of tumor, MHC loss or disruption of antigen processing, gangliosides, and co-regulatory molecules on the tumor. In the second category are factors both intrinsic and extrinsic to the TIL that are largely independent of the tumor. Examples include signaling defects in the TIL, dysfunction of cytolytic machinery, immune tolerance, immune exhaustion, regulatory T cells, myeloid-derived suppressor cell (MDSC), and myeloid dendritic cells. These and other so-called tumor escape mechanisms and TIL defects have been extensively reviewed elsewhere [16, 17].
While studies addressing the mechanisms responsible for the failure of the immunocompetent cells to control tumor progression have shed considerable light upon this conundrum, the majority of these studies have been conducted with mouse tumor models. At present not enough is known about which mechanism(s) are contributing to the regulation or inhibition of TIL in the microenvironment of human tumors. To be successful, cancer immunotherapy will require a broadening of our basic knowledge of the molecular and cellular events that contribute to the prevention or suppression of anti-tumor responses in humans [18, 19]. Currently, information is lacking about how the human immune system is regulated, and how T cells respond to human tumor cells and tumor stroma when they migrate to the tumor microenvironment. We and others have cautioned that the continuing confusion and lack of understanding of lymphocyte function in the human tumor microenvironment may be due in part to gaps between human and animal studies [18–20]. Human malignancies that arise spontaneously differ considerably from experimental transplantable mouse tumors  and significant differences between human and mouse immune systems exist and have been well documented .
In this review we begin with a histological, immunohistochemical and flow cytometric characterization of the human tumor microenvironment. This is followed by a summary and discussion of the literature on the hyporesponsiveness of the TIL and how this relates to the TCR signaling cascade. We will also review our data derived largely from the in vitro study of lymphocytes present in the microenvironment of human non-small cell lung tumors demonstrating that a significant portion of the tumor-infiltrating memory T cells are hyporesponsive or nonresponsive to activation via the CD3/CD28 signaling pathway and that this arrest is reversible. Next we review a diverse array of studies showing that fibroblasts have the ability to significantly alter T lymphocyte function and survival and we discuss the possibility that tumor-associated fibroblasts contribute to the regulation of T cells in the tumor microenvironment. In the final segment of this review, we discuss the design and use of human xenograft models of tumor microenvironments that are being used to study in vivo the dynamic interplay of human TIL with stromal cells and tumor cells in situ in human tumor microenvironments.
Human tumors are infiltrated by a wide variety of different leukocytes including macrophages, granulocytes, natural killer cells, myeloid dendritic cells, plasmacytoid dendritic cells, and lymphocytes . The vast majority of infiltrating leukocytes in advanced human solid tumors are lymphocytes including B lymphocytes, plasma cells and different subsets of T lymphocytes. Lymphocytes are found in juxtaposition to tumor cells and are often found to infiltrate directly into the tumor parenchyma (as indicated with the blue arrow in Fig. 2). Note in this figure the very dense accumulation of lymphocytes directly surrounding tumor nodules in both a human non-small cell squamous lung carcinoma and ovarian epithelial papillary serous adenocarcinoma (Fig. 2). This intimate association of the lymphocytes with the human tumor shown here is in contrast to several reports by others with animal tumor models suggesting that lymphocytes are restricted to the periphery of the tumor and fail to infiltrate tumor tissue.
Phenotypic analysis by flow cytometry of the lymphocytes derived from the tumors indicates that the majority of the CD45+ cells present within human lung non-small cell tumors are T cells with a phenotype that is consistent with effector memory T cells (i.e. CD45RO+, CD4+ or CD8+, CD44+, CXCR3+, CD28+ and negative for CD27, CD45RA and CD62L) . This same phenotype has been seen in CD45+ cells from ovarian tumors (Bankert et al., unpublished). The dominance of effector memory T cells in tumor microenvironments has been shown in a variety of different murine and human tumors [23–25], and these cells are the focus of this review. While these memory T cells can be further divided into phenotypically and functionally distinct subsets, including CD4+ cells that are distinguished by the cytokines they produce and CD8+ cytotoxic cells, they all appear to be sustained in the tumor microenvironment in a viable but hyporesponsive state.
The dominant presence of memory T cells within the microenvironment of human tumors raises at least three important questions. First, how can these cells persist in the face of continuous antigen stimulation? Next, why do these cells fail to attack and kill tumor cells? Finally, what is responsible for attracting and retaining these cells to and in the tumor microenvironment? Insights into the first two questions may be derived from murine studies where the ability of T cells to remain viable in a functionally inactive or hyporesponsive state in the face of persistent antigen stimulation has been shown . This T cell quiescence, termed adaptive tolerance , avoids excessive tissue damage and prevents the loss of antigen-specific T cells that may otherwise occur in the presence of persistent antigen stimulation. Such an antigen-driven senescence could ultimately lead to an unacceptable partial depletion of the T cell repertoire. The evidence for the hyporesponsiveness of the tumor-associated memory T cells and the several possible mechanisms responsible for the functional arrest of T cells are discussed later. The possible role of tumor-associated fibroblasts in preventing memory T cell apoptosis, thereby sustaining their presence in the tumor microenvironment, and the production of chemokines and cytokine receptors that are responsible for the specific attraction of effector memory T cells (both CD4+ and CD8+) to the tumor are also addressed below.
CD4+ CD25+ FOXP3+ regulatory T cells (Treg) represent another subset of T cells that are known to control all aspects of the immune response [27–31]. This subset of cells has been found in human solid tumors, tumor ascites fluids and hematological malignancies [32–34], and their presence correlated in human ovarian cancer with a poor prognosis [32, 33]. Indirect evidence suggests that Tregs are able to suppress tumor-specific immunity [34–42], but definitive evidence of the role of this T cell subset in the immunopathogenesis of human cancer is still lacking.
Although we know much about what inflammatory cells can do when removed and studied independently from the tumor microenvironment in vitro, it remains unclear how the cells may contribute to supporting or curtailing the development and spreading of spontaneously arising human tumors. It has been impossible to distinguish, by morphology, the inflammatory cell networks that contribute to tumor eradication from those that either have no effect or contribute to the promotion of tumor growth. The inflammatory infiltrates in tumors therefore remain a histopathological variable that is notable and potentially exploitable therapeutically, but not yet sufficiently defined. However, in spite of our incomplete knowledge with respect to how inflammatory cells influence tumor progression, a careful in situ analysis of tumor-infiltrating immune cells has become a valuable prognostic tool. By characterizing tumor-infiltrating immune cells in a large cohort of human colorectal cancers by gene expression profiling, and in situ by immunohistochemical staining, the type, density and location of immune cells within tumors were found to be a better predictor of patient survival than methods currently used to stage this cancer .
These studies are consistent with the notion that immunocompetent cells within the tumor may contribute to control of the growth and spread of mature tumors. Most studies in both mouse and humans have focused upon CD8+ cytotoxic T cells that kill tumors directly and TH1 CD4+ T cells (producing IL-2, and IFN-γ) that contribute directly or indirectly to tumor killing . While TH1 remains the dominant TH subset in tumors within the memory T cell pool, another T cell subset has recently been found in both mouse and human tumors that may play a significant role in the pathogenesis of tumors [44–46]. This subset is a CD4+ T cell that secretes IL-17A and has been labeled as the TH17 cell [47, 48]. These cells and the cytokines they produce (including IL-17A, IL-17F and IL-22) have been shown to play important roles in the pathogenesis of autoimmune diseases and inflammation [49, 50]. Their biological role in the tumor microenvironment is poorly understood, and at present suggested to be both beneficial and detrimental to the cancer patient . The generation and regulation of TH17 cells in ovarian cancer have been documented  and the possible function and clinical relevance of these cells in human tumor microenvironments reported . It has been postulated that TH17 cells contribute to protective human tumor immunity by inducing the production by tumor cells of two chemokines that attract and recruit effector T cells and effector memory T cells that express CXCR3, the receptor for the TH17 induced chemokines CXCL9 (MIG) and CXCL10 (IP10) . This could help to explain the dominance and sustained presence of the effector memory T cells in the microenvironment of human lung and ovarian tumors. However, as will be apparent from the discussion that follows of the complex and dynamic interactions of the memory T cells with the other inflammatory and stromal cells in the tumor microenvironment, it is not sufficient to recruit and retain memory T cells to control tumor growth. A significant portion of the effector memory T cells appear to undergo a functional arrest upon entering the tumor microenvironment [20, 52]. With an understanding of how this functional arrest occurs it should be possible to design approaches to reverse it and thereby recover full function of these cells. The T cells would then be able to orchestrate the killing of tumor cells locally and systemically. These important issues are the focus of the remainder of this review.
Our studies and those of others suggest that the inability of T cells to respond to and eradicate tumors is due to a functional arrest in the T cells present in the tumor microenvironment. This hyporesponsiveness of the effector memory T cells is caused by a regulatory arrest in the activation of signaling molecules in the TCR/CD28 pathways (Fig. 3).
Insufficient T cell function has been well documented among cancer patients and several studies have attempted to evaluate the molecular changes that occur with carcinogenesis. Using an in vivo murine colon carcinoma model, Mizoguchi and colleagues showed that the CD8 + T cells isolated from tumor bearing mice had impaired cytotoxicity, TNF-α production and granzyme expression . Furthermore, the T cells from tumor-bearing mice upon activation had a reduction in the phosphorylation of Lck, and a reduction in the expression of CD3γ and CD3ζ when compared to non-tumor bearing controls. Reduced expression or complete loss of the CD3ζ chain of the TCR complex has been reported in both murine  and human carcinoma . Rodriguez et al. further demonstrated that loss of CD3ζ resulted in a reduction in T cell proliferation and cytokine production, and that the loss of CD3ζ was due to a reduction in the levels of L-arginine, an essential amino acid that has been implicated in the regulation of T cell responses in cancer .
On the other hand, Agrawal et al. showed that the lack of activation of T cells in cancer could not be explained by defects in CD3ζ . By comparing the tumor-infiltrating T cells isolated from B cell non-Hodgkin’s lymphomas (NHL) with T cells from non malignant lymphoid organs these investigators found that in contrast to the T cells from the non-malignant lymphoid organs, the TIL failed to proliferate when stimulated with anti-CD3. However, by-passing the upstream TCR signaling cascade with PMA and CD28 enhanced proliferation.
In a murine adenocarcinoma model, Frey and colleagues  identified two populations of tumor-derived CD8 + T cells that differed in their ability to lyse target cells. The non-lytic TIL had deficiencies in the ability to flux calcium and tyrosine phosphorylation of the proximal TCR signaling molecules Zap-70 and LAT was significantly reduced when compared to the lytic TIL. The investigators specifically localized this proximal signaling defect to a lack of Lck activation that was associated with a localization of SHP-1 to Lck and an inability of Lck to be phosphorylated by ERK, which in the non-lytic TIL was sequestered away from Lck.
SHP-1, a tyrosine phosphatase, has been shown to directly dephosphorylate the subunits of the TCR, as well as Zap-70, PI3-K, PLC-γ and LAT thereby leading to defects in activation and subsequent downstream signaling . Its direct interaction with SLP-76 and the Vav-Grb2 complex suggests a role of SHP-1 in modulating T cell cytoskeletal structure and further downstream signaling. SHP-1 has also been implicated in the regulation of the Ras/MAPK pathway. Loss of SHP-1 enhances CTL function and increases the formation of CD8+ T cell-APC contact . In addition to its direct effects on TCR signaling, SHP-1 can associate with the suppressive co-regulatory molecule programmed death 1 molecule (PD-1, CD279), which also inhibits T cell activation .
Several recent reports in both human and murine models have identified the upregulation of PD-1 on antigen specific CD8+ T cells in chronic infection [60–62]. Blockade of PD-1 with an antibody to its ligand PD-L1 (B7-H1, CD274) resulted in an enhancement of the T cell response . PD-L1 and another PD-1 ligand called PD-L2 (B7-DC, CD273) are both expressed on a wide variety of cells such as DCs and macrophages , all of which are present in the tumor microenvironment. PD-L1 is present on tumors and its elevated expression has been correlated with poor prognosis [62, 63]. Our laboratory has recently identified both PD-L1 and PD-L2 on stromal tumor-associated fibroblasts in non-small cell lung carcinoma . One mechanism by which PD-1 and its ligands can inhibit T lymphocyte responses is by directly targeting the TCR pathway, although PD-1 ligation may also act indirectly by transducing a signal that leads to the recruitment of SHP-1 whose effects are described above.
Two independent investigators have observed that diacylglycerol kinase-α (DGK-α) is upregulated in anergic T cells [65, 66], indicating that there is a link between the TCR signaling molecule DAG and T cell responsiveness. As depicted in Fig. 3, upon TCR ligation PLC-γ generates diacylglycerol (DAG) and inosinitol-1,4,5-triphosphate (IP3) from phosphatidyl-inositol-4,5-bisphosphate (PIP2). DAG is responsible for the partial activation of the MAPkinase pathway and the activation of PKC-θ leading to NF-κB translocation. Defects in Ras have resulted in a reduction in the MAP kinases Erk, Jun and the subsequent trans-activation of AP-1. Therefore, DAG induced activation of PKCθ and Ras is absolutely essential for transcriptional activation of AP-1 and NF-κB. Co-stimulation with CD28 also greatly amplifies this signal. Phosphorylation of the hydroxyl group on DAG, catalyzed by DGK-α, converts it to phosphatidic acid. Koretzky and colleagues demonstrated, using microarray analysis, that DGK-α was upregulated in anergic cells and downregulated in normally responsive cells .
Other investigators have evaluated activation by assessing the translocation of transcription factors, NFAT and NF-κB. Uzzo et al. demonstrated that T cells isolated from human renal cell carcinomas (RCC) failed to translocate the transcription factor NF-κB after exposure to PMA and ionomycin which bypass the proximal tyrosine kinase events in the TCR pathway that are required for NFAT and NF-κB translocation . Gangioslides, which are commonly secreted from RCC, were found to be responsible for the suppression of NF-κB translocation.
We have also assessed the activation of memory T cells derived from the human lung tumor microenvironment by evaluating the translocation of NF-κB and NFAT. In most resting cells, NF-κB is sequestered in the cytoplasm bound to its inhibitor IκB. Upon CD3/CD28 stimulation there is a cascade of tyrosine phosphorylation of various signaling and adapter proteins [68, 69] that results in the ubiquitination and degradation of IκB (Fig. 3). Upon its dissociation from IκB, NF-κB enters the nucleus to initiate the transcription of genes involved in proliferation, survival, adhesion, migration and overall cellular function.
NFAT, like NF-κB, translocates from the cytoplasm into the nucleus with T cell receptor signaling (see Fig. 3). When calcium is released from the endoplasmic reticulum and/or from extracellular stores, it binds to calmodulin activating calcineurin , which dephosphorylates NFAT allowing its translocation from the cytosol into the nucleus.
We have demonstrated that unlike memory T cells derived from the peripheral blood of normal donors or from patient peripheral blood, T cells isolated from primary non-small cell lung tumors are unable to translocate both NF-κB and NFAT into the nucleus following CD3 and CD28 cross-linking. This signaling defect appears to be specific to the CD3/CD28 pathway as these TIL are responsive to TNF-α, which induces the translocation of NF-κB by an alternative signaling pathway .
In an effort to reverse the inability of the TIL to translocate these transcription factors we identified two molecules that play a role in regulating their activation. Transforming growth factor-β (TGF-β1) was found to be at least partially responsible for inducing the hyporesponsiveness  and interleukin-12 (IL-12) could reverse this TCR/CD28 signaling arrest .
TGF-β1 can regulate the development and function of all leukocytes, including T regulatory cells, dendritic cells (DC), NK cells and B cells . Both the differentiation state of the cells and the presence of immunomodulatory molecules, including co-stimulatory molecules and cytokines, influence the effect of TGF-β1 on T cell function.
The importance of TGF-β1 in regulating T cell function was first demonstrated in mice that had a T cell specific blockade in TGF-β1 signaling. These animals had widespread inflammatory leukocyte infiltration into multiple organs and died prematurely [72–74]. High levels of TGF-β1 are produced by various tumor types including breast, colon, stomach, liver  and fibroblasts . High levels of TGF-β1 have also been associated with poor prognosis . Since TGF-β1 is highly expressed in the tumor microenvironment and has been shown to suppress both the acquisition and expression of T cell effector function , it is likely that it may be at least partially responsible for the failure of TIL to control tumor progression .
Although the inhibitory effect of TGF-β1 on naïve and effector T cells has been established, very little is known about its effect upon human memory T cells particularly in the tumor microenvironment. One study showed that TGF-β1 suppressed the acquisition and expression of effector function of human memory T cells that were reactive against melanoma . Byrne et al. evaluated the effect of TGF-β1 on macrophages in the tumor microenvironment and found that TGF-β1 was responsible for the recruitment and retention of macrophages at the tumor and that this enabled the tumor to escape from the immune system .
These studies on TGF-β1 and T cell function suggest that TGF-β1 mediates a suppressive regulatory mechanism on memory T cells in the tumor. They also demonstrate that TGF-β1 has the potential to prevent the uncontrolled activation of T cells that may be present in an environment of chronic antigen stimulation.
Given the suppressive role of TGF-β1 in the tumor microenvironment, a significant question was whether TGF-β1 may be playing a role in the hyporesponsiveness of TIL to CD3/CD28 activation. Utilizing a TGF-β1 specific function blocking antibody we demonstrated that the hyporesponsiveness of the TIL could be reversed following TGF-β1 blockade . Neither TGF-β1 nor anti-TGF-β1 had any noticeable effect upon the activation of T cells derived from the peripheral blood of normal donors or cancer patients. Several cells within the tumor microenvironment have been shown to secrete TGF-β1. These include the immune and tumor cells, as well as the tumor-associated fibroblasts . However, it is not clear how tumor-derived versus other cell-derived TGF-β modulates the activation of the TIL. We further evaluated the effects of TGF-β1 by pulsing the TIL with a low pH buffer which should remove not only TGF-β1, but all molecules bound to the surface of the cell. After an overnight recovery period, the TIL cells treated with the low pH buffer responded normally to activation by TCR/CD28 induced cross-linking . We established that this was at least partially modulated by membrane-bound TGF-β1 by demonstrating that the hyporesponsiveness was re-established by the addition of exogenous TGF-β1 to the cell suspension, after low pH treatment.
In order to completely understand the direct role of TGF-β1 on T cell activation, it is necessary to evaluate and elucidate its effects on the TCR signaling components. Chen et al. demonstrated that TGF-β1 inhibited the phosphorylation and activation of Itk in activated CD4+ T cells . In addition, intracellular Ca++ was reduced and NFAT translocation was inhibited. TGF-β1 did not affect Lck and Zap70 phosphorylation, but did inhibit the phosphorylation of ERK . TGF-β1 has also been shown to induce IκBα and inhibit NF-κB translocation . It will ultimately be necessary to determine how the TGF-β1 signaling pathways intersect with and modulate the TCR signaling pathways.
Recent studies have evaluated anti-TGF-β1 therapies targeted to both the tumor, tumor microenvironment and other systemic effects related to tumorigenesis . Evidence from these studies strongly suggests that blocking TGF-β1 and/or TGF-β1 mediated signaling may be a therapeutic approach to treating cancer, although neither the cell(s) that are producing it nor the mechanism(s) by which TGF-β1 is able to inhibit TCR activation and signaling are completely clear. However, because membrane-associated TGF-β1 is suppressive to tumor development early on, complete loss of TGF-β1 could be detrimental .
Over the last few years it has become increasingly evident that IL-12 has the potential to be a powerful therapeutic agent or adjuvant for cancer immunotherapy. IL-12 is a cytokine that plays a critical role in both the innate and adaptive immune systems . IL-12 induces NK and T cells to proliferate and produce IFN-γ [81, 82]. Studies with IL-12 have validated its role in the reactivation and survival of memory CD4+ T cells [82, 83]. These studies have validated the multifunctional role of IL-12 and its potential utilization as an anti-cancer agent [82, 84, 85].
Until recently, the mechanism by which IL-12 provoked T cell mediated anti-tumor immunity was unknown. Studies in our laboratory have demonstrated that hyporesponsive human TIL can be reversed and the T cells reactivated by IL-12 in vivo. This IL-12 induced reactivation of the TIL resulted in their proliferation, production of IFN-γ and eradication of tumor cells [23, 86–88]. While it was evident that IL-12 could reverse the T cell hyporesponsiveness, the mechanism of this reactivation was unknown. We evaluated the effect of IL-12 on TCR responsiveness by incubating TIL in vitro with a brief pulse (3 h) of IL-12, followed by an overnight incubation without IL-12. Following overnight incubation, the T cells were stimulated with anti-CD3 and anti-CD28 cross-linking antibodies for 1 h, and the translocation of NFAT monitored by confocal immunofluorescence microscopy. The TIL pretreated with IL-12 responded normally to the TCR activation protocol whereas the control TIL (no IL-12 exposure) remained unresponsive to activation via CD3/CD28 . Therefore, IL-12 is able to reverse the arrest in the TCR signaling cascade of the TIL. The mechanism by which IL-12 reverses the TIL hyporesponsiveness has yet to be elucidated. Studies have demonstrated that one likely candidate that may be responsible for T cell hyporesponsiveness is the Tec family kinase Tec . IL-12 treatment of CD4 + T cells in peripheral blood has been shown to increase the expression of the Tec kinases . Studies in vivo have also demonstrated that IL-12 can enhance synapse formation and unresponsive T cells once treated with IL-12 were able to respond to weak peptides or self peptides . The ability to reverse the TCR signaling arrest in TIL derived from malignant tissues indicates that these cells are not in a state of permanent immunosuppression, and that they can be reactivated to respond to stimulation through the CD3/CD28 pathway. It will be critical to determine how other cells in the tumor microenvironment, such as tumor cells, regulatory T cells, myeloid suppressor cells, monocytes, dendritic cells, granulocytes and fibroblasts, are able to modify T cell activity within the tumor microenvironment.
Among the cells in the tumor stroma, T regulatory cells and myeloid-derived suppressor cells have received much attention in the literature for their role in downregulating immune responses to tumor cells, but many of the mechanisms by which these cells have been hypothesized to function within the tumor microenvironment have also been found in fibroblasts. The potentially immunosuppressive mechanisms which have been found on fibroblasts include the production of soluble products such as TGF-β1 or hepatocyte growth factor (HGF or scatter factor), hormones such as prostaglandin E2 (PGE2), the co-regulatory surface molecules like PD-L1 (B7-H1), and the tryptophan catabolizing enzyme indoleamine 2,3-dioxygenase (IDO). Fibroblasts are also known to produce a host of soluble factors that influence angiogenesis, such as vascular endothelial growth factor (VEGF), and chemokines that orchestrate the influx of immune cells (Fig. 4). Therefore, fibroblasts have the potential to be much more than inert structural cells and do dynamically interact with the other cell types within the tumor microenvironment. Myofibroblasts frequently make up the bulk of the tumor microenvironment yet research is only recently beginning to dissect the role of this prominent cell type in the pathogenesis of cancer as well as in other pathologic settings. Fibroblasts have been implicated as tumor cell-promoting through their ability to produce factors that support angiogenesis as well as their ability to support tumor cell metastasis (reviewed in [92–94]). Here, we review ways in which fibroblasts have been reported to interact with and influence the activation of T lymphocytes.
Fibroblasts are the heterogeneous mesenchymal cells of connective tissue and are identified morphologically by their cell processes giving them a spindle or stellate shape and by their adherence to cell culture plates in vitro. Molecular markers for fibroblasts include an 112 kD cell surface protein recognized by the D7-FIB monoclonal antibody (also called SM1214P) [64, 94], CD90 (Thy-1), fibroblast-specific protein 1 (FSP-1) , vimentin, and prolyl 4-hydroxylase (an enzyme for collagen synthesis). Myofibroblasts, which can be found in the tumor microenvironment, represent activated fibroblasts capable of increased levels of extracellular matrix (ECM) production and are characterized by their production of α-smooth muscle actin (reviewed in ).
However, no marker has thus far been identified that is found only on fibroblasts and expression of the current punative markers differs among fibroblast subsets. Mesenchymal stem cells (MSC) are fibroblast-like progenitor cells usually derived from the bone marrow stroma with the capacity for self-renewal and for adipogenic, chondrogenic, and osteogenic differentiation. Like fibroblasts, MSC can be isolated from the bone marrow based on D7-FIB antibody recognition of surface protein and these isolated cells are negative for CD45 while positive for CD90. MSCs differ from dermal fibroblasts in that MSCs express the phenotypic markers LNGFR, STRO-1, and HLA-DR . Based on their spindle-shaped morphology, adherence, and expression of D7-FIB antigen and CD90, MSCs represent a subset of fibroblasts with the capability of differentiation into other cell types. Like fibroblasts, MSC suffer from a lack of a specific marker, since even markers that distinguish MSC from differentiated fibroblasts are also found on other cell types. Thus, there is still considerable debate within the field as to what constitutes MSC and whether they represent true self-renewing, pluripotent stem cells.
Fibroblasts are responsible for the production of ECM components such as collagen, fibronectin, and laminin. Cell adhesion to the ECM is necessary for the organization of tissues and cell trafficking and is mediated primarily through integrins. Signaling via α2β1 integrin increases T lymphocyte IFN-γ production  and culturing activated T cells on ECM protein-coated plates has been shown to increase their proliferation [99, 100]. Thus, the fibroblast-produced ECM itself can serve as a means by which T lymphocytes receive co-stimulation and drive the inflammatory response.
Despite the known capability of fibroblasts to produce collagen and fibronectin, which have been shown to act as co-stimulatory molecules for T lymphocyte activation via the T cell receptor [100, 101], a number of studies have reported that MSC and mature fibroblasts derived from a variety of sources demonstrate immunosuppressive effects on T lymphocytes. Human MSC, rheumatoid arthritis synovial membrane fibroblasts, normal dermal fibroblasts, and articular chondrocytes have all been shown to inhibit proliferation of allogeneic peripheral blood T cells and reduce IFN-γ production [102–104]. MSC constitutively produce COX-2, PGE2, TGF-β1 and HGF and were able to suppress both MLR and concanavalin A driven T lymphocyte proliferation. PD-L1 and IDO expression could be induced and PGE2 and HGF expression increased by IFN-γ addition to MSC . Uccelli’s group found that this MSC-induced T cell unresponsiveness was a state of anergy that was reversible with administration of IL-2 . Similarly, human colonic myofibroblasts were shown to suppress the proliferation of anti-CD3/CD28 activated peripheral blood lymphocytes and IL-2 production via a contact-mediated mechanism and exogenous IL-2 could reverse this fibroblast-mediated inhibition .
The PD-1 ligands, PD-L1 and PD-L2, have been shown to be expressed constitutively on human fibroblasts and to a greater extent on myofibroblasts from normal colon biopsies . As mentioned previously, both are considered to be negative co-regulatory molecules of T lymphocyte function. IFN-γ has been shown to increase PD-L1 expression levels in both malignant  and normal fibroblasts . Since fibroblasts express PD-L1, they have the potential to influence PD-1+ T lymphocyte activation.
A number of laboratories have cocultured fibroblasts and lymphocytes as a means of investigating the interactions between these cell types. Most human coculture studies to date have utilized fibroblasts and peripheral blood lymphocytes derived from different patients. Since T lymphocytes recognize and respond to alloantigens, conclusions drawn from allogeneic coculture systems are open to question. The fact that in most studies T lymphocytes fail to respond to the alloantigens of MSC has added to the notion that MSC are immunologically inert or even immunosuppressive and some have extended this idea to include all adult human fibroblasts . However, it has been reported that fibroblasts activate neither autologous nor allogeneic T lymphocytes when cultured together , but either autologous or allogeneic resting T cells are able to activate transcription of several relevant genes in fibroblasts. Thus, results derived from experiments using allogeneic coculture systems may be as physiologically relevant as those from autologous systems.
Our experiments culturing anti-CD3/CD28 activated lung tumor-infiltrating T lymphocytes or patient peripheral blood T lymphocytes with tumor-derived fibroblasts indicate that fibroblasts have the ability to modulate T cell activation as monitored by IFN-γ production (Fig. 5). In most cases, IFN-γ levels are significantly increased in T cells activated with these tumor-derived fibroblasts over those T cells activated alone. In a few instances, suppressive fibroblasts were derived from non-small cell lung tumors and this suppression could be reversed by addition of PD-L1 blocking antibody . Fibroblasts were also shown to produce TGF-β. While these candidate molecules may explain the immunosuppression observed by three fibroblast lines, they cannot account for the much more frequently observed effect of an increase in IFN-γ levels found in coculture supernatants.
Fibroblasts produce many molecules with the potential to alter lymphocyte function and vice versa. For example, skin, bone marrow, and RA synovial fluid fibroblasts all spontaneously make IL-6 [110–112]. Increased levels of TNF-α, IL-6, IFN-γ  and IL-17A  were found when synovial fibroblasts were cultivated with T cells. Dermal fibroblasts also increase their level of IL-6 production when cultivated with T cells and this was even more pronounced with allogeneic cultures than when autologous T cells were used . TNF-α, lymphotoxin (TNF-β) [110, 116], IFN-γ , TGF-β1  and IL-17A  represent T lymphocyte-produced cytokines that have been shown to upregulate fibroblast production of IL-6. IL-6 is important in the differentiation of naïve T lymphocytes into TH17 cells that produce inflammatory IL-17A.
RA synovial fibroblasts have been shown to express surface IL-15 that could induce TNF-α, IFN-γ, and IL-17A in T lymphocytes . PGE2, which can be produced by fibroblasts, was shown to have anti-proliferative effects on T lymphocytes and to decrease IFN-γ levels, but more recent evidence has shown that PGE2 augments TH17 cell differentiation  and this may serve as another pathway by which fibroblasts could contribute to inflammatory disorders. Conversely, IL-17A has been shown to cause an increase in fibroblast PGE2 production .
Because fibroblasts represent a heterogeneous and mostly undefined cell type, different populations of fibroblasts derived from the same patient may be responsible for the suppression or enhacement of the T cell function. Alternatively, individual cells may have the capacity to produce both lymphocyte activation-enhancing molecules (such as collagen) and inhibiting molecules (TGF-β). The altered levels of IFN-γ or proliferation observed in T lymphocyte-fibroblast cocultures represent the sum of these conflicting actions.
While fibroblast-produced factors may be categorized as suppressive or enhancing to T cells, other fibroblast effects are less easily classified, but nevertheless modulate the immune response. For instance, human synovial fibroblasts and mouse lung fibroblasts prevent T lymphocyte apoptosis [102, 121].
Fibroblast produced type I interferons, particularly IFN-β, have been shown to rescue activated peripheral blood T lymphocytes from apoptosis when they are deprived of IL-2 common γ chain receptor cytokines (IL-2, IL-4, IL-7, IL-9, IL-15) [110, 122]. In our hands, we have been unable to detect IFN-β production by ELISA in the media supernatants of cocultured lung tumor fibroblasts and T lymphocytes.
Activation-induced cell death, a form of apoptosis occurring via the expression of Fas Ligand when T cells are restimulated through the TCR, can be decreased by fibroblast-produced PGE2, which inhibits the activation of caspase-8 . IL-6 can also prevent T cell activation-induced cell death through anti-CD3 stimulation by decreasing the expression of FasL on T lymphocytes . Components of the extracellular matrix, such as fibronectin, are also anti-apoptotic, particularly for memory T cells [99, 104]. By influencing apoptosis, fibroblasts have the ability to both mold the T cell repertoire and allow for inflammation to persist chronically.
Fibroblasts have been shown to increase the trafficking of T lymphocytes directly through their production of chemokines such as the CXCR3 ligands, CXCL9 (MIG) and CXCL10 (IP10), and the CXCR4 ligand, CXCL12 (SDF-1α), as well as indirectly through angiogenesis promoting factors or through increasing the adhesiveness of nearby endothelial cells to bind circulating lymphocytes. In addition to their role as chemoattractants for leukocytes, chemokines (including macrophage inflammatory protein-1α and β, RANTES, and MCP-1) can serve as costimulators of T cell activation to induce further proliferation .
Fibroblasts and endothelial cells are frequently in close proximity to one another in a number of different microenvironments. Thus, fibroblasts are in a prime position for altering the nearby vasculature. The manner in which fibroblasts alter the adhesiveness of endothelial cells has been shown to be in part dependent upon the milieu from which the fibroblast was derived. For instance, rheumatoid arthritis-derived synovial fibroblasts increased adherence of lymphocytes to endothelial cells while those taken from uninflamed microenvironments (dermal fibroblasts) actually decreased adherence of lymphocytes . Fibroblasts themselves express adhesion molecules with most expressing ICAM-1 particularly in the presence of T cells and with even greater levels of expression in the presence of activated T cells .
While fibroblasts are capable of producing chemokines, they are also regulated by them. For instance, CCL2 (monocyte chemotactic protein-1 or MCP-1), CCL5 (RANTES), and CXCL12 (SDF-1) have all been shown to enhance IL-6 and IL-8 production by RA synovial fibroblasts . IL-8 is a chemoattractant for neutrophils and thus a regulator of acute, innate inflammation, but also a promoter of angiogenesis. Cytokines differentially regulate synovial fibroblast production of chemokines. TGF-β1 induces CXCL12 production while TNF-α and IL-1β give rise to CCL5 production. All three of these cytokines lead to enhanced CCL2 expression . The cytokine milieu thus alters which cell types will be summoned to an inflammatory site, including the tumor microenvironment, based on which chemokines are released.
Through chemokines, cytokines, and modification of the vasculature, fibroblasts coordinate the movement of immune cells throughout the tissues. Fibroblasts can directly modify the activation and differentiation states of T lymphocytes through co-regulatory receptors, cytokine, and extracellular matrix production. Once T cell activation has occurred, fibroblasts influence which T lymphocytes remain and which go through apoptotic and survival pathways. All of these networks have been shown to exist within the tumor microenvironment and thus fibroblasts themselves may have much more impact on what constitutes the inflammatory infiltrate within tumors than has previously been recognized and thus warrant closer scrutiny.
Most of what has been learned and is discussed above with respect to the functional arrest of T cells in human tumors has come from in vitro studies of single cell suspensions of lymphocytes, stromal cells and tumor cells. While this knowledge could potentially help to design more effective immunotherapeutic strategies for cancer, it will first be essential to establish in vivo which of the immunosuppressive mechanisms identified in vitro are involved in the T cell arrest, and which of the cells and cell products contribute most significantly to the induced hyporesponsiveness of the TIL in situ. This must be achieved by monitoring events within structurally intact human tumor microenvironments. By implanting non-disrupted pieces of human tumor tissue into severe combined immunodeficient (SCID) mice, our laboratory has been able to establish xenografts in which the microenvironment of the human tumor tissue remained physically and functionally intact with a sustained presence of the inflammatory leukocytes, stromal fibroblasts and tumor cells [87, 126]. Initial studies with this first SCID-human tumor xenograft model confirmed that the TIL were hyporesponsive and that the functional arrest of the T cells could be reversed by the local and sustained release of IL-12 or IL-12 and GM-CSF delivered directly into the tumor by microspheres [23, 87] or liposomes . A major limitation of the early xenograft model was the presence of the host innate immune response that ultimately resulted in the invasion of host innate immune cells and the eradication of the xenograft . To overcome this pitfall, a number of different strategies have been used to selectively eliminate or block the innate immune cell-mediated disruption of the xenograft. One approach was to deplete the SCID mouse of natural killer (NK) cells with a polyclonal antibody to asialo-GM1 or a monoclonal antibody directed against the β chain of the IL-2 receptor . However, the effects of this treatment were transient and NK cells eventually were replenished and xenografts rejected. An alternative method for the elimination of NK cells was a further genetic alteration of the SCID mouse that was achieved by crossing the SCID mouse with the non-obese diabetic (NOD) mouse which has decreased NK cell function, lack of hemolytic complement and has functionally immature macrophages . While there was an initial improved engraftment of human tissues in the NOD-SCID mice, xenografts were eventually invaded and rejected by the host cells due to residual NK cell activity.
The initial limitations of the SCID and NOD-SCID mice for the implantation and long term survival of human normal and neoplastic tissues have now been overcome by the generation of a new generation of NOD-SCID mouse with a null mutation of the IL-2 common γ chain receptor [130–132]. These NOD-SCID IL2Rγnull mice are called NOG mice, and the IL-2R mutation results in a defect of the high affinity receptors for IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 that blocks the development of NK cells and further impairs innate immunity [130, 133]. The implantation of non-disrupted pieces of human tumor tissues into NOG mice results in the successful engraftment and long term survival of human tumor microenvironments with little or no invasion of the xenograft for up to 9 weeks post engraftment . The presence of viable cytokeratin positive tumor cells that are actively dividing (Ki67+) are observed in xenografts of human non-small cell lung tumor tissues for up to 12 weeks post-engraftment . The cytokeratin and Ki67 staining of a human squamous cell carcinoma xenograft established in a NOG mouse is shown in Fig. 6. The tumor is surrounded by stromal fibroblasts with a diffuse presence of inflammatory CD45+ leukocytes including CD3+ T cells and CD138+ plasma cells (Fig. 7). The majority of the T cells in the tumor xenografts are CD45RO+ and fewer CD20+ B cells and CD68+ monocytes are present in the xenografts. The entire xenograft tissue stains positively for HLA, indicating the tissue is of human origin without any noticeable infiltration of host murine cells . The T cells remain viable and responsive to IL-12 as shown by their proliferation and production of IFN-γ. The presence of human immunoglobulin in the sera of the xenograft bearing NOG mice suggests that B cells and plasma cells also remain viable and functional for prolonged periods following tumor tissue implantation.
The NOG mice have also been shown to be a reliable model to demonstrate and evaluate human NK cell mediated antibody-dependent cellular cytotoxicity (ADCC) directed in vivo to Hodgkin lymphoma . Another finding of considerable interest with this same lymphoma tumor model in the NOG mice was the ability of a monoclonal antibody directed against CCR4 (a chemokine receptor on Treg cells) to significantly reduce the number of Tregs that infiltrate the tumor xenograft .
The ability of tumor specific T cells to overcome the immunosuppressive effects of the tumor microenvironment was demonstrated in another NOG mouse tumor xenograft model by the direct injection of T cells (engineered to express a chimeric receptor with high affinity for a tumor antigen) into pre-established large tumor xenografts . The retargeted tumor-specific T cells survived and significantly reduced tumor burden, and, in some cases, resulted in the complete eradication of large pre-established mesothelioma xenografts . It was further shown in this tumor xenograft model that the incorporation of the CD137 signaling domain into the chimeric receptor, significantly enhanced the persistence of the engineered T cells in the tumor-bearing mice following intratumoral or intravenous administration.
The ability of NOG mice to support the long term engraftment of human tumor microenvironments and the demonstrated utility of this model to study the interaction of lymphocytes with stromal cells and tumor cells in situ provides an opportunity for the first time to study and determine which of the cells and factors shown to contribute to the arrest of T cell function in vitro are also found to be operating in vivo in the tumor microenvironment.
Immunotherapy either by itself or in combination with other treatments for cancer has great potential. However, to achieve full therapeutic benefits of immunotherapy, many important questions about how the cells and molecules within the tumor microenvironment interact with each other need to be addressed and answered. In particular, it will be essential to gain insights into how tumor stromal cells and TIL modulate each other’s function, and to identify potential molecular targets that can be exploited therapeutically to reverse the arrest of T cell signaling that occurs upon the entry of lymphocytes into the tumor microenvironment. While some of these issues have already been addressed by taking a reductionist approach in which different cell types from the tumor microenvironment are cocultured together in vitro, eventually it will be necessary to study the T cell and stromal cell interactions within the intact tumor microenvironment. Because of the differences between experimental mouse tumors and spontaneous human malignancies, and the increasing awareness of the differences between mouse and human immune systems, an increased emphasis needs to be placed upon the study of human tumor microenvironments in situ and in vivo. There are many obstacles both intrinsic and extrinsic to such human studies. Difficulties in obtaining human tissues, the genetic diversity of the study population, variability within and between tumor types and stage of tumor development, and the necessary limitations to the design and conduction of clinical trials to protect patients are just a few examples that have contributed to the impediments to conducting human studies. Nevertheless, these challenges must be addressed if we are to gain the insights into the complex cellular and molecular events occurring in human tumor tissues that are required to design ways to improve upon current immunotherapeutic approaches to cancer.
The use of new and improved immunodeficient mice for the establishment of intact human tumor microenvironment xenografts as discussed above has provided an opportunity to study human tumors in vivo for prolonged periods, and to test paradigms that develop from in vitro studies. It is expected that additional improvements in these xenograft models must and will be made to more closely approximate the native tumor microenvironment and the dynamic interactions between the tumor and tumor-associated cells that occur over time in patients’ tumors. An example of such improvements is the design of orthotopic tumor implantations that progress and undergo spontaneous metastases to other organ sites that is now in development (Bankert et al. unpublished). The generation of Class I and Class II HLA transgenic NOD-scid IL-2rγnull mice represents an exciting and potentially very important advance in humanized mice for studying T cell responses to tumor antigens in tumor microenviroments. By adoptively transferring human hematopoietic stem cells from an HLA-A2 donor into an HLA-A2 transgenic mouse it was possible to generate human MHC restricted antigen specific CD8+ T cells 7 days after infection with a Dengue virus . The use of these new humanized mice that are engineered to express both Class I and II HLA should ultimately provide an opportunity to pre-clinically evaluate tumor vaccination strategies in which both the generation of MHC restricted tumor-specific T cells and their effect upon the arrest of tumor growth can be monitored and quantified. Using these mice it will be possible to test the paradigm that upon entry into the tumor microenvironment from the periphery T cells become anergic, and subsequently to preclinically test strategies that are designed to reverse the signaling arrest of TIL and determine whether by reactivating these T cells the tumors are eradicated. The application of newer assay technologies such as two-photon laser-scanning microscopy will make it possible to visualize in tumor xenografts the dynamic changes in the movement, expansion, and contraction of lymphocytes, fibroblasts and tumor cells in vivo at the single cell level following perturbations to the tumor microenvironment that are designed to reactivate quiescent T cells [138, 139]. A combination of fine needle biopsies and real time PCR will make it possible to monitor changes in gene expression patterns and alteration in tumor progression without sacrificing the xenograft bearing mice.
The insights that will be gained from using human tumor xenografts to study the cellular and molecular dynamics in the tumor microenvironment are expected to lead to the design of novel approaches to enhance the therapeutic efficacy of cancer immunotherapies. These new approaches will ultimately need to be tested in clinical trials with cancer patients.
The discovery that IL-12 is able to reverse the hyporesponsiveness of T cells in the human tumor microenvironment provides a rationale for designing future clinical trials. While previous clinical trials with subcutaneous or intravenous IL-12 at doses of 500–1,000 ng/kg resulted in unacceptable toxicities , we predict (based upon the studies with human tumor xenografts) that ultra low and sustained levels of IL-12 delivered directly into the tumors will be effective at re-activating tumor-infiltrating T cells without the toxic effects that have been observed following large doses of this cytokine systemically. It is anticipated that by re-activating the T lymphocytes, these cells will orchestrate a cascade of events that result in the killing of tumor locally and that with the release of tumor antigens from the dead and dying tumor cells there will be an induction of a systemic anti-tumor immunity. This expectation has been confirmed in two animal tumor models [141, 142] and several clinical trials have shown that low doses of IL-12 injected locally into tumors show little or no toxicity (reviewed in ). An Investigational New Drug (IND) application for a local sustained delivery device for the introduction of IL-12 into tumor nodules has recently been approved by the FDA and a Phase I clinical trial that will address the safety and later the efficacy of this in situ tumor vaccination strategy is to be initiated shortly.
Future efforts must be focused upon determining how IL-12 is able to reverse the hyporesponsiveness of tumor-infiltrating memory T cells. Armed with the knowledge of the mechanisms by which this TCR signal transduction is arrested, it should be possible to design and exploit additional methods to reverse the T cell hyporesponsiveness and by understanding how stromal fibroblasts modulate T cell function within the tumor microenvironment we predict that novel therapeutic approaches will be developed that target the fibroblasts and the biologically active factors produced by these cells.
This work was supported by National Institute of Health grants R01CA108970, R01CA131407, and R56AI079188.