The urokinase receptor (u-PAR)—a link between tumor cell dormancy and minimal residual disease in bone marrow?

Abstract

Minimal residual disease (MRD) is hypothesized to be the major cause of tumor recurrence and metastasis even years and decades after primary cancer diagnosis and curative solid tumor resection. In these patients disseminated tumor cells reflecting MRD can be detected in the bone marrow years after treatment. It is to be assumed that genetic determinants and a complex interplay between the disseminated tumor cells and their microenvironment in the bone marrow are responsible for tumor cell dormancy and the final reactivation towards metastasis. The urokinase receptor (u-PAR), a critical regulator of invasion, intravasation, and metastasis, is found to be a key player in regulating the shift between single cell tumor dormancy and proliferation. This has mainly been attributed to a regulation by u-PAR of integrins, and the ability of the latter to propagate signals from fibronectin through the EGF-receptor, ERK, and p38 signaling. Interestingly, u-PAR is found in disseminated tumor cells in the bone marrow of solid cancer patients, and is associated with the expansion of these cells and clinical prognosis. Here we summarize and discuss findings on disseminated tumor cells in the bone marrow, MRD and the role of u-PAR in tumor biology, especially focusing on its specific role in providing a switch between tumor cell proliferation and dormancy. Finally, we discuss the hypothesis that u-PAR might be an essential molecule in bone marrow disseminated tumor cells for long-term survival during dormancy, and/or reactivation of their proliferation years after primary treatment.

INTRODUCTION

There is increasing agreement that a combination of early tumor cell dissemination and dormancy essentially accounts for the devastating fact that, several years or even decades after complete and curative resection of the primary tumor, recurring metastases can arise (1, 2). The concept of minimal residual disease (MRD), which is certainly well accepted for hematologic malignancies, has been established also for solid cancers. Even in very early stages of solid carcinomas such as breast, gastric, colon, and lung cancers, numerous studies over the past two decades have shown that single disseminated tumor cells can be detected in compartments such as lymph nodes or bone marrow. These initial studies were based on immunocytochemistry (ICC) using antibodies directed against cytokeratins, especially cytokeratins 18 to 20 (3–6). Using ICC, different groups detected disseminated tumor cells within the mesenchymal compartment of the bone marrow in up to 55% of cases of breast, gastric, or colorectal cancer (7–23). Furthermore, more sensitive monitoring techniques for detecting MRD, and for a molecular profiling of disseminated tumor cells, have been developed. These include quantitative PCR techniques, or single-cell comparative genomic hybridization (CGH), enabling a first systematic analysis of such cells (1, 2). Quantitative PCR techniques started to provide potentially important information about the effectiveness of treatment and the risk of recurrent disease as shown by MRD analysis in patients with various malignant diseases (24-26). For example, CK-19 mRNA-positive cells in blood and bone marrow could be effectively targeted by trastuzumab (27) in breast cancer patients. However, whereas the clinical prognostic impact of DTCs in lymph nodes is well accepted as a prognostic factor (e.g. (22)), the clinical and prognostic significance of single DTCs in the bone marrow is more controversial due to contrasting results regarding the prognostic impact of the presence of these cells at the time of primary surgery in diverse tumor entities. For example, whereas in some tumor entities such as breast cancer the detection of these cells at primary surgery was associated with prognosis, a prognostic impact of these cells at surgery was not seen in studies of gastric cancer, and for colorectal cancer, studies detecting a prognostic significance contrast with our own unpublished observations (8, 10, 11, 13, 18, 19, 28-33).

DO DISSEMINATED TUMOR CELLS IN THE BONE MARROW CORRELATE WITH MINIMAL RESIDUAL DISEASE?

The bone marrow has been suggested to be an anatomic filter, and a common homing organ, for disseminated tumor cells recruited from the blood stream in carcinomas of the breast (6, 23, 34-37), colon (8, 38), lung (39-41), prostate (42-44), and ovaries (45). The majority of studies and the largest studies on disseminated tumor cells in the bone marrow have been conducted in breast cancer patients. For example, in a recent study (36) over 800 patients were investigated for disseminated tumor cells at the time of primary breast cancer surgery by immunocytochemistry. In this study, significantly higher tumor recurrences and deaths from breast cancer were observed in patients who had disseminated tumor cells in the bone marrow, as compared to patients negative for bone marrow DTCs (13, 19, 31, 46). In this and also other studies on breast cancer (23, 36, 37), the mere detection of tumor cells in the bone marrow was a significant prognostic factor.

However, in other tumor entities, such as gastric and colorectal cancer, the significant prognostic impact of disseminated tumor cells in the bone marrow, as judged from diverse studies, was less clear (8, 10, 11, 13, 14, 18, 19, 28-31). The biological relevance of these cells in the bone marrow, in terms of representing a “true” minimal residual disease component, is still unclear.

One major concern has been that disseminated tumor cells in the bone marrow can be detected in high percentages of patients with cancers that rarely develop manifest bone metastases (e.g. gastric cancer) (29, 31)). This observation suggested that— although clinically relevant—establishment of minimal residual disease at sites where it never metastasizes may represent a transit anatomic compartment. Further, although the interaction between the disseminated tumor cells and the microenvironment in this case is conducive to tumor cell survival, it does not result in manifest tumor recurrence or metastasis. However, it may serve as a reservoir for cells to further disseminate and resume growth at their final destination. It is speculated that in tumors, such as gastric cancer, the microenvironment within the bone marrow might be sufficient to enable tumor cell persistence. It is possible that the bone marrow supports a dormant state, or enables tumor cell persistence despite chemotherapy (47), but it might not be sufficient to support critical proliferative or proangiogenic signals to propel metastatic growth (47-50). Furthermore, there are still ongoing discussions as to whether disseminated tumor cells detected within the bone marrow in early stages of solid cancers indeed indicate a subclinical systemic disease component determining the patient's fate. Alternatively, some of these cells might represent non-viable, biologically irrelevant cells that may even have been passively shed during the process of primary surgery, or due to shedding processes associated with the bloodstream. Support for the latter hypothesis came from studies suggesting that a high percentage of initially found disseminated tumor cells in the bone marrow can disappear over time during clinical follow-up of cancer patients (51). To further investigate this phenomenon we performed a clinical study with serial bone marrow biopsies during long-term follow-up in curatively resected patients with gastric cancer (29, 31, 52). We found a striking association between the quantitative development of disseminated tumor cells in the bone marrow over time and the later clinical outcome of the patients. We observed that the majority of the patients having late tumor recurrences revealed either a clear increase or a constantly high and persistent number of disseminated tumor cells in the bone marrow. In contrast, patients without clinical recurrence were characterized as having either complete clearance of tumor cells, or continuously negative or low cell counts (31, 52). This study showed that the dynamic development of disseminated tumor cells in the bone marrow over time might be the crucial parameter for measuring the development of a clinically relevant minimal residual disease component in solid cancer, leading to later recurrence. This study supported the hypothesis that there may be a considerable phenotypic heterogeneity of disseminated tumor cells detected in the bone marrow of these patients, their interaction with the microenvironment being reflected by the quantitative development of the cells over time. However, the mere detection of these cells in the bone marrow at the time of primary surgery most likely cannot reflect this biological heterogeneity and, in consequence, the individual potential of these cells to cause metastases.

PHENOTYPES OF DISSEMINATED TUMOR CELLS

Increasing numbers of studies have addressed the phenotypic characteristics of disseminated tumor cells in the bone marrow. For example, several groups have reported numerical chromosomal aberrations in cytokeratin-positive tumor cells in the bone marrow and blood (53, 54). Gangnus et al. (55) found chromosomal copy number changes in disseminated tumor cells from early-stage breast cancer patients by single-cell comparative genomic hybridization, with a significant intercellular heterogeneity and differences from the corresponding primary tumors. Furthermore, different works have implicated a heterogeneous proliferative potential of disseminated tumor cells in the bone marrow of solid cancer patients (11, 54). Recently, Gangnus et al. (55) were able to select disseminated tumor cells with proliferative potential by short-term culture. Klein & Hoelzel (56) suggested that a high frequency of dormancy occurs in such tumor cells, and that this might initially be driven by epigenetic rather than genetic alterations, since they observed only a few genetic alterations in early disseminated tumor cells in the bone marrow by comparative genome hybridization (CGH). Still permanence in this anatomic location may subsequently select for specific genetic changes (56). Other studies suggested a loss of MHC class I molecules as a mechanism of “immunological escape” and protection from cytotoxic T-cells (57-59), and further studies have shown the presence of proliferative or prometastatic molecules, such as HER2/neu, u-PAR, and extracellular matrix metalloprotease inducer in blood cells (6, 40, 60). In this context, one recent study suggested that disseminated tumor cells in blood or bone marrow can be successfully eliminated with trastuzumab (Herceptin), an antibody targeting the HER2-receptor as a successful therapeutic concept in breast cancer (27). Another study detected cyclooxygenase-2 expression in disseminated tumor cells in the bone marrow of colorectal cancer patients (61). Thus, the availability of markers that detect DTCs, and also markers that may serve as targets, is important to determine the nature of the DTCs, and whether they can be targeted therapeutically.

THE UROKINASE RECEPTOR (u-PAR) AS A PREDICTIVE MOLECULAR MARKER FOR DISSEMINATED TUMOR CELL DEVELOPMENT AND CLINICAL PROGNOSIS

Our investigations of disseminated tumor cells in the bone marrow and potentially decisive phenotypic characteristics of these cells revealed that the urokinase receptor was strikingly associated with the quantitative development of disseminated tumor cells in the bone marrow over time. u-PAR was also associated with an unfavorable prognosis of solid cancer patients with evidence of these cells in the bone marrow. u-PAR has been gaining attention as one of the most relevant invasion- and metastasis-related molecules in solid cancer (see below). Several studies showed the biological relevance of u-PAR during invasion and metastasis as well as for tumor recurrence and clinical prognosis (29, 52). Thus, we investigated the u-PAR-gene expression on disseminated tumor cells at the time of primary cancer surgery, and its association with disseminated tumor cell burden in the bone marrow during postoperative follow-up, and with clinical prognosis. Our studies on gastric cancer patients revealed that u-PAR protein as measured by double immunocytochemistry on disseminated tumor cells was significantly correlated with increasing tumor cell counts in the bone marrow. In fact, the patients with evidence of u-PAR on disseminated tumor cells at the time of primary surgery most often showed a clear increase in tumor cell counts in the bone marrow during clinical follow-up (29, 31, 52). In contrast, the majority of patients without evidence of u-PAR in disseminated tumor cells at primary surgery showed an association with a decrease of tumor cells in the bone marrow, or even complete elimination of the tumor cells. Cells with u-PAR significantly increased in number during follow-up (52). In addition, patients exhibiting u-PAR in disseminated tumor cells in the bone marrow had a higher risk of clinical tumor recurrence and early tumor-related death. In contrast, cases without evidence of u-PAR in these cells were associated with decreased tumor recurrence. These results suggested that u-PAR expression is an important phenotypic characteristic critical for the development of clinically relevant systemic disease in early solid cancer (31, 52). Moreover, u-PAR has been demonstated to be a molecular predictor of the ability of disseminated tumor cells to increase in number over time, and potentially to escape microenvironmental mechanisms that would lead to their elimination.

In an independent and larger clinical study, we showed that, in a prospective series of over 150 gastric cancer patients, the u-PAR is an independent predictor of poor clinical prognosis when detected in disseminated tumor cells in the bone marrow at primary surgery (63). This finding is in contrast to the mere detection of disseminated tumor cell numbers without immunophenotyping at surgery, which had no prognostic impact on recurrence or survival in the same patients. This again supported the hypothesis that u-PAR is a promising molecular marker to define a critical subpopulation of disseminating tumor cells potentially important for the establishment of a clinically relevant MRD component in solid cancers.

THE UROKINASE RECEPTOR—FUNCTIONS ASSOCIATED WITH INVASION AND METASTASIS

Explanations for the link between u-PAR and the interruption of asymptomatic MRD may be found in the diverse functions of u-PAR, as a molecule involved in migration, invasion, and metastasis.

The urokinase receptor (u-PAR) interacts with a specific ligand, urokinase-type plasminogen activator (u-PA), a serum protease of 55 kDa, which, as bound by the receptor, converts plasminogen to active plasmin (for a comprehensive review see (64)). A further cascade of proteolytic activities is initiated, its concerted action leading to the efficient degradation of extracellular matrix components such as fibrin, vitronectin, proteoglycans and, as the main molecules in basement membranes, laminin and collagen IV (38, 65, 66). The u-PA-/u-PA-R-complex is bound by specific inhibitors, plasminogen activator inhibitors 1 and 2 (PAI-1 and PAI-2), respectively. The u-PA-/u-PA-R/PAI complex is internalized into the cell (67-71), allowing the free u-PA to be recycled to the cell surface.

The u-PA-receptor consists of three similar domains, the last of which is anchored to the cell membrane via glycosylphosphatidylinositol (72). This GPI anchor enables a high intramembrane mobility of the molecule (72, 73). Physiologically, the u-PAR gene is expressed in peripheral leukocytes, inflammatory activated monocytes (29, 74, 75), and endothelial cells and keratinocytes at the leading edge in reepithelializing wounds (76). Correspondingly, u-PAR gene expression plays an important role in inflammation, tissue remodeling, and wound healing (32, 77-79). Furthermore, studies on u-PAR knock-out mice have suggested a role of the u-PAR in chemotaxis, since granulocytes and monocytes of these mice are severely impaired in their migratory and chemotactic capacity towards inflammatory sites (80). A decisive function in cytoskeletal rearrangements and migration has been suggested from studies showing that the u-PAR is co-localized with integrins and acts as a coreceptor for vitronectin (80-82). The association with integrins has also been proposed to mediate u-PAR strong mitogenic function in tumor cells (see below).

Certainly, the most abundant evidence has been given respecting the important relevance of u-PAR for invasion, intravasation and development of metastasis in human cancer. However, the exact mechanism that may be driving metastatic growth has only recently been appreciated. Numerous studies have shown an overexpression of the u-PAR gene in diverse human malignant tumors as compared with corresponding normal tissue (62, 83–86). Direct evidence has implicated u-PAR in tumor invasion and metastasis. For instance, overexpression of human u-PAR cDNA increased matrigel invasion without altering cell migration in a human osteosarcoma cell line (87). The invasion of chicken embryo chorioallantoic membrane (CAM) required u-PA catalytic activity and the invasive efficiency of tumor cells was reduced by 75% when tumor u-PA activity or tumor u-PA production was inhibited (88). Other studies revealed that the expression of an antisense u-PAR cDNA in HEp3 squamous carcinoma cells decreased invasion and intravasation in a modified CAM assay (89, 90). Similar results were observed in transformed fibroblasts in in vitro invasion assays (91). In glioblastoma, an anti-u-PAR monoclonal antibody blocked invasion effectively in a Matrigel assay, in which inhibition of invasion ranged between 20% and 57% for human glioblastoma cell lines (92). Additional studies using various anti-u-PAR strategies have revealed similar results in several cancer models (93-96).

All these studies defined u-PAR as a critical molecule for invasion, intravasation and metastasis. However, it is not clear how these experimental data can fully explain several clinical studies that found elevated levels of members of the u-PA system in metastases as compared to primary tumors (31, 97, 98). Moreover, prospective studies on diverse cancers involving large numbers of patient have demonstrated a correlation of high u-PAR (and/or u-PA-/PAI-1) expression with short survival times and advanced tumor stages. Thus, the u-PAR and/or u-PA/PAI-1 have already been shown to be significant prognostic factors in many cancers, including breast (31, 98), lung (99), colon (100, 101), esophageal and gastric cancer (62, 102), and some of these studies reported an independent impact on survival probability in multivariate analysis, which was recently confirmed in independent patient series in our own studies (103).

Taken together, many clinical studies already suggest a strong clinical relevance of u-PAR for diagnosis, prognosis, a more precise high-risk prediction, and also response to therapy. In the past, this striking prognostic and clinical impact of the u-PAR and the u-PA-system has largely been attributed to its functions in invasion, migration, intravasation and metastasis. However, regarding the aforementioned association of u-PAR with the establishment, or maintenance, of MRD, and the prognostic impact of u-PAR found on disseminated tumor cells in the bone marrow, we speculate that the clinical relevance of u-PAR might be explained by those functions related to its role as a regulator of mitogenic signaling that allows tumor cells to switch between tumor cell proliferation and quiescence-induced dormancy.

THE u-PAR AS AN ESSENTIAL PLAYER IN THE DECISION BETWEEN PROLIFERATION AND QUIESCENCE-INDUCED DORMANCY

Tumor recurrence can occur a very long time after removal of the primary tumor (specifically, breast cancer can recur even decades after seemingly successful primary treatment) (110). Further, disseminated tumor cells in the bone marrow of solid cancers can be found years after tumor surgery in some patients without clinical evidence of relapse. These findings are inconsistent with the original concept of a constant proliferation of tumor cells. Dormancy was originally defined as the ability of a certain cell or organism to be in a condition of biological rest or suspended animation during which it is not active, but capable of become active again. “Cancer cell dormancy” describes the prolonged quiescent state in which tumor cells are present, but not biologically or clinically apparent (110). More specifically, an arrest in G₀/G₁-transition has been suggested as one possible mechanism of tumor cell dormancy (104). Pioneering studies suggested that dormant micrometastases can occur while the primary tumor is developing, and the nascent tumor mass is unable to recruit blood vessels (105). In this model the tumor mass is dividing actively but the lack of vessels causes cell death, which counterbalances proliferation. Other studies have shown that after dissemination, single tumor cells can enter a state of dormancy despite an active vasculature nurturing the target organ (50, 106-108). Thus, there might be two general instances where dormancy is manifested, an earlier one defined by cellular quiescence and a later one defined by angiogenic dormancy (105, 109). Single cell tumor dormancy might become activated after the tumor cells that survive dissemination are unable to cope with the cellular stress induced by this process, the new microenvironment, or additional stresses such as chemotherapy. While a majority of tumor cells might die, some may be able to survive this stress, but without the initial capacity to resume proliferation (110). This notion is supported by the observation that the large fraction of disseminated tumor cells found in the bone marrow has been found negative for proliferation markers such as PCNA, suggesting that these cells are in G₀/G₁-arrest (50).

Solitary tumor cell dormancy is assumed to happen at the single cell level; that is before a tumor mass is established. Increasing evidence suggests that an interplay betwen surface receptors and the tumor cell microenvironment may be a critial determinant of the shift between proliferation and dormancy of solitary tumor cells, in particular, the crosstalk between the urokinase receptor (u-PAR) and integrins, and how they regulate mitogenic (e.g. ERK), and stress (e.g. p38) signallling pathways. These may be major determinants of the shift from proliferation to G₀/G₁-arrest and consequently dormancy of single tumor cells (106, 110).

It has been shown in earlier studies that the blockade of u-PAR can induce apoptosis in some cases such as brain tumors, but that in other cases it can induce a state of dormancy (49), characterized by G₀/G₁ arrest. It was shown that the ability of u-PAR to regulate tumor growth and prevent dormancy was in part due to its ability to activate integrins, which in a fibronectin-dependent manner were able to transduce through FAK and the EGFR mitogenic signals (111). First, evidence suggesting that u-PAR can regulate tumor growth came from different studies in HEp3 squamous carcinoma cells. In this model, an antisense mRNA blocking the expression of u-PAR caused a significant reduction in invasion in vivo, but also led to a complete loss of tumorigenicity in vivo (112, 113). Low u-PAR-expressing HEp3-cells continued to form small nodules that remained dormant for several months, and interestingly this loss of tumorigenicity was not due to a decreasing proteolytic activity, enhanced apoptosis or lack of vasculature, but to an overall decrease in proliferation (113). Further studies demonstrated that high u-PAR-expressing HEp3-cells show an association of the u-PAR molecule with the fibronectin (FN) receptor α₅β₁ integrin (49). It was observed that the interaction of u-PAR with α₅β₁-integrin caused an increase in integrin activity, strong adhesion to fibronectin, and enhanced fibronectin fibrillogenesis (49, 106, 108). Strikingly, this interaction resulted in a permanent fibronectin-dependent activation of the Ras-ERK pathway, supporting rapid tumor cell proliferation in vivo. In contrast, when the same cells were downregulated in u-PAR by an antisense strategy, the interaction between u-PAR and α₅β₁ integrin was significantly reduced, and the ERK pathway was deactivated (49). Subsequently, we observed an arrest of the cells in the G₀/G₁ phase and dormancy (106). Further experiments revealed that a minimal functional complex consisting of u-PAR, α₅β₁ integrin, FAK and the EGFR was required to transduce u-PAR and fibronectin-dependent signals to the ERK pathway (106, 111, 114), since blocking expression or function of all of these molecules resulted in decreased ERK activation and dormancy.

These data suggested that ERK inhibition might be sufficient to induce dormancy. However, other pathways may be needed to force single disseminated tumor cells into dormancy. Correspondingly, in further studies it was found that u-PAR-mediated signaling induced tumorigenicity not only by activating ERK, but also by inhibiting the p38^SAPK pathway, which is necessary to maintain G₀/G₁ arrest (106, 111). We observed that in cells with high u-PAR expression the u-PAR-α₅β₁-integrin complex supported the assembly of fibronectin fibrils. The fibronectin fibrils in turn, through a mechanism requiring the inactivation of Cdc42, maintained the growth suppressive p38^SAPK pathway in a low activation state. In contrast, when the interaction between u-PAR and integrin or fibronectin fibrils was disrupted, an activation of the p38^SAPK pathway was observed, causing growth arrest in vivo. Taken together, these observations suggested that the expression level of u-PAR can provide a switch between tumorigenicity and dormancy in vivo, especially by regulating the balance between the ERK and p38^SAPK pathway (115). The activation ratio of these two pathways was found to be predictive for the in vivo behavior of various tumor types, such as prostate, breast or fibrosarcoma, implicating that the phenomena observed are not specific for the HEp3-model (115).

These observations were confirmed by other studies showing that spontaneously dormant cells showed a low expression of u-PAR, and a low ERK/p38^SAPK signaling ratio (116). Re-expression of the u-PAR, or constitutive activation of ERK in these cells, was sufficient to restore the u-PAR-α₅β₁ integrin complex, and high ERK activity, which was followed by an interruption of dormancy in these cells in vivo (116). Furthermore, both a pharmacological p38 inhibitor (SB203580) and also a dominant-negative p38a resulted in the interruption of dormancy, and restoration of in vivo growth. The inhibition of p38 resulted in ERK-dependent induction of u-PAR gene transcription, an assembly of the u-PAR-α₅β₁-integrin-complex, and restored ERK activation. This suggests that u-PAR expression can establish a positive feedback loop by maintaining ERK activation, which in turn induces u-PAR gene transcription. However, this loop may be inactivated if the ERK/p38 ratio is reversed and p38 signaling becomes dominant inhibiting ERK activation (115, 117). Taken together, these studies support the urokinase receptor and its interaction with α₅β₁ integrin, fibronectin and EGF-R as an essential switch between proliferation and quiescence-induced dormancy in tumor cells, regulating the balance between ERK and p38 signaling.

u-PAR AS AN ESSENTIAL MOLECULE IN DISSEMINATED TUMOR CELLS IN THE BONE MARROW AS A LINK BETWEEN INVASION, METASTASIS, PROLIFERATION, AND DORMANCY?

Considering the evidence from experimental models and that gathered in MRD studies we consider u-PAR in disseminated tumor cells to be a very likely candidate not only for predicting the metastatic phenotype of disseminated tumor cells in bone marrow, but especially for providing a switch between proliferating and dormant minimal residual tumor cells in the bone marrow. It is possible that the mechanisms described in experimental systems on u-PAR-dependent regulation of integrins, EGF-R, and ERK/p38 may serve as a switch from quiescence to proliferation. This hypothesis is especially tempting since the bone marrow is known to be rich in fibronectin (118). Thus, inhibition of u-PAR might be able to induce or maintain dormancy. This may be important after stress conditions such as chemotherapy. Thus, interrupting the u-PAR/fibronectin/integrin interaction and activating a growth arrest pathway through p38-disseminated tumor cells might help cells survive in a dormant state. Re-establishing the u-PAR/integrin/fibronectin interaction and strong proliferative signals might also re-establish a proliferative and invasive potential. It is interesting to speculate that dependence on the levels of u-PAR and its interactions through integrins with fibronectin in the bone marrow might be suitable for controlling disseminated tumor cells, to keep them in the non-proliferative dormant state, and potentially prevent recurrence. Further, it may be important to identify the survival pathways that maintain the viability of these DTCs as their inhibition might allow eradication of dormant cells without affecting their growth arrest. Although this will remain a challenging objective, the molecular characterization of the growth arrest of DTCs will without doubt provide novel targets for therapy. For example, recently Chaurasia and co-authors mapped a region within domain III of u-PAR (240–248 aa in human u-PAR), which is indispensable for ERK-pathway activation. Sequence alignment revealed four positions that are conserved among different vertebrates within this region (G240, C241, S245 and C247) (119). These data support the notion that especially this site could be a candidate site for targeted cancer therapy, for example, to re-target single disseminated tumor cells that have escaped primary treatment, and are still detectable and u-PAR positive after primary therapy.

Certainly, such speculative strategies still need to be investigated in appropriate preclinical and in vivo models. In analogy, additional inhibitors involving molecules of the u-PAR associated complex, such as EGF-R inhibitors (120), or integrin inhibitors, should be tested in appropriate models for their ability to potentially overcome ERK-dependent tumor cell proliferation in vivo. Most importantly, however, there is a clear need to provide ultimate experimental proof in appropriate in vivo models, and again bone marrow material of solid cancer patients, for our hypothesis that the u-PAR, its interactions with integrins, fibronectin and EGF-R, and the resulting ability to switch between ERK and p38 activation, is indeed one of the essential mechanisms for deciding between proliferation and quiescence-induced dormancy of disseminated tumor cells in the bone marrow of cancer patients. An important change will also have to occur in standard treatment, as strategies to maintain an asymptomatic MRD will require the treatment of patients during stages of asymptomatic disease. Further, the identification of biomarkers of dormancy vs recurrence will be essential to better administer such therapies. Several questions remain. For example, in gastric cancer, bone marrow DTCs with u-PAR are poor prognostic indicators, but these cells hardly ever manifest as metastases in bone. So are these cells proliferating and leaving the BM for the liver or other sites, or is this a phenomenon that reflects what is happening in the real target organ? In addition, studies suggest that even when u-PAR is low, DTCs in the bone marrow still survive. Thus, there are other mechanisms (u-PAR-independent) driving basal survival, or different thresholds of u-PAR levels that control survival and then proliferation. Finally, the data reviewed here support the notion that monotherapy will not be successful in inducing or maintaining dormancy. In fact, at least three molecules might have to be targeted for induction of dormancy (i.e., u-PAR, inte-grins, EGFR/ERBB2). This is supported by recent data showing that in DTCs in breast cancer, u-PAR and ERBB2 are coamplified, suggesting that a natural coupling of adhesion and RTK signaling is essential to drive growth of disseminated tumor cells (121).