ERK^MAPK Activity as a Determinant of Tumor Growth and Dormancy; Regulation by p38^SAPK1

Tumor latency, best defined as the time between the carcinogenic event and the onset of progressive growth, culminating in a clinically detectable tumor mass, has been documented in experimental cancer models and in people exposed to catastrophic radiation (1). It is assumed that during this period, cells undergo genetic and epigenetic changes, and those with growth advantage are selected giving rise to progressively growing tumors (2). A seemingly parallel phenomenon exists in cancer patients who, although cured of their primary disease, show no signs of metastasis but develop tumors in secondary sites, often after prolonged periods of dormancy. Although it is easy to envision that progression from a normal to a fully malignant cell may take time, the inability of a fully malignant cell such as a cell that separates from a growing primary cancer to resume growth upon arrival in a new organ is more difficult to understand. However, the fact that disseminated but undetected cancer cells eventually progress to overt metastases indicates that solitary cell, or small groups of cells, can survive away from the primary until, through as yet unknown mechanisms, they resume growth. Because of their scarcity and the absence of detection methods, dormant cells are difficult to isolate, and the mechanisms that maintain dormancy and regulate the transition from dormancy to progressive growth remain largely unknown. It is possible that metastases can grow to a clinically detectable size, only after they acquire the ability to induce new blood vessel (neoangiogenesis; Ref. 3), but other mechanisms must also exist (4). It has been shown that large proportion of cancer cells in distant organs remains as single dormant cells (5, 6, 7). Because these conclusions were drawn from experiments in which cells were injected directly into the circulation, thus bypassing the selection process that takes place during cancer progression, it remains to be seen if they will apply to induction of dormancy in spontaneous metastasis.

We propose that even cancer cells that may have accumulated multiple mutations may use the surface receptors and ECM³ components to regulate signaling pathways that control cell cycle progression and/or arrest. This implies that a shift between growth and dormancy may be anticipated upon cancer cell arrival in a new organ, either because of changed ECM or changed expression or function of their surface receptors.

In progressively growing tumors, constitutive activation of the ERK pathway allows for G₀-G₁-S-phase transition and cell division (8). Although ERK is mostly involved in induction of proliferation (9) and in some cases differentiation (10), a high level of p38 activity is believed to be a negative growth regulator (11) that, depending on the stimulating signal, may suppress cell proliferation by inhibiting ERK (12), inducing G₀-G₁ arrest (13) or triggering senescence (14) or apoptosis (15). There is evidence that activation of p38 can reverse Ras-induced fibroblast transformation (16) and that in epithelial cells, inhibition of p38 by Ras is required for transformation (17). Moreover, published evidence shows that p53 phosphorylation by p38 is required for its tumor suppressing activity and that PPM1D, a specific p38 phosphatase, by dephosphorylating p38 and inhibiting its activity blocks the tumor suppressor function of p53 (18). Thus, it seems that many effectors can alter the balance of ERK and p38 and that such change may have profound consequence for tumor growth and survival.

We previously found that a balance that favors p38 activation over ERK in highly malignant human carcinoma cells (HEp3) can induce persistent growth suppression (dormancy) in vivo (12). Specifically, we found that the rapid growth of metastatic HEp3 carcinoma in vivo is regulated by high expression of uPAR that, by interacting and activating α5β1-integrins, initiates a signaling cascade that culminates in very strong and persistent ERK activation (19). This activation, which is mediated through ligand-independent activation of EGFR (20), is additionally enhanced by uPA binding to uPAR and cell binding to FN (19, 21). Consequently, uPAR overexpression and activation of ERK generates a positive feedback loop that maintains uPAR mRNA and high protein levels, keeping the signaling cascade for cell cycle progression on (12). Upon down-regulation of uPAR, ERK activity is lost, p38 becomes activated, and the balance is shifted in favor of p38 (12). Increased p38 activity has been previously shown to inhibit activation of ERK (22, 23) and to block cell proliferation in culture (14). We found, however, that HEp3 cells with low ERK/p38 activity balance, while fully capable of proliferation in culture, when inoculated in vivo on the chick CAM, or s.c. in nude mice, rapidly arrest in G₀-G₁ and remain viable but dormant for a prolonged period of time (12, 19, 24). This paradigm provides one of the first examples whereby down-regulation of a surface protease receptor, by altering the interaction between a cancer cell and the extracellular matrix, induces in vivo dormancy of an otherwise fully malignant cell. This additionally suggests that some cancer cells may not be fully growth autonomous and that, in the metastatic sites, either the necessary growth inducing stimuli may be absent or the solitary cells may transiently lack components necessary for the interpretation of the extracellular cues, giving rise to dormant metastases. We showed that in HEp3 cells, it was possible, by disrupting the expression or the mutual interactions between uPA, uPAR, α5β1, FAK, EGFR, or FN proteins that constitute the ERK activating complex, to favor p38 activity and dormancy (12, 19, 21). Here, we examine the generality of the above paradigm by characterizing the ERK/p38 ratios in a series of tumor cell lines representing cancers of different origin in the context of their in vivo behavior. Our results indicate that the majority of the cell lines studied conform to the paradigm established for HEp3 cells in that their ERK activity or ERK/p38 activity ratio is predictive of their in vivo behavior and that uPAR and α5β1-integrin play a major role in that behavior.

DMSO, Triton X-100, sodium orthovanadate, NaFl, protease inhibitors, DNase I, BSA, normal goat serum, collagenase type 1A, rhodamine-phalloidin conjugate, and human fibronectin were from Sigma Chemical Co. (St. Louis, MO). Aprotinin and trypsin were from ICN Biomedicals, Inc. (Aurora, OH). DMEM, OPTI-MEM medium, glutamine, antibiotics, and Lipofectin were from Life Technologies, Inc., Laboratories (Grand Island, NY). Fetal bovine serum was from JRH Biosciences (Lenexa, KS), COFAL-negative embryonated eggs were from Specific Pathogen-Free Avian Supply (North Franklin, CT); protein G-agarose beads were from Boehringer Mannheim (Indianapolis, IN). Polyvinylidene difluoride membranes and ECL detection reagents were from Amersham (Amersham Life Sciences, Little Chalfont, United Kingdom). PD98059, SB203580, and its inactive analogue SB202474 were from Calbiochem (San Diego, CA). The stock solutions were prepared in 100% DMSO. DAPI was from Hoechst Molecular Probes (Eugene, OR). Antibodies: antiphospho ERK 1/2 (antiphospho-Tyr 204; clone E4) was from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Anti-ERK1/2 (Clone MK12) and anti-p38 (Clone 24) mAbs were from Transduction Laboratories (Lexington, KY). Antiphospho-p38 polyclonal antibodies were from Biosource (Camarillo, CA). Normal mouse IgG, fluorochrome-labeled secondary antibodies, and antihuman fibronectin polyclonal antibody (F3648) were from Sigma. Goat antimouse, Alexa-546-conjugated IgG, was from Molecular Probes (Eugene, OR). Antihuman uPAR mAb R2 was kindly provided by Dr. Michael Ploug (Finsen Laboratory, Copenhagen, Denmark). Anti-uPAR polyclonal rabbit antibody (399R) was from American Diagnostica, and polyclonal rabbit anti-β1 antibody (Mab1952) and anti-α5β1 antibody (Clone HA5) were from Chemicon International, Inc. (Temecula, CA). Antimouse IgG HRP-conjugated mAb and mounting medium (Vectashield) were from Vector Laboratories (Burlingame, CA). Antirabbit IgG-HRP and anti-HA antibody (clone 12CA5) were from Boehringer Mannheim (Indianapolis, IN). All antibodies used in vivo or in culture were free of azide. The endotoxin content of antibodies used in culture or in vivo was tested using Pyrogen-Plus test from Biowhittaker (Walkersville, MD) and found to have <24 pg/ml.

Human epidermoid carcinoma HEp3 (T-HEp3), serially passaged on CAMs of chick embryos, was used as a source of tumorigenic cells (19, 25). The source of spontaneous dormant tumor cells (D-HEp3) were HEp3 cells passaged in culture 120–170 times (24) with uPAR level of only ∼20% of that in tumorigenic cells (19). PC3, MDA-MB-231, MDA-MB-453, MDA-MB-468, MCF-7, and HT1080 cells were obtained from American Type Culture Collection. M24met cells were kindly provided by Dr. Barbara Muller from Scripps Institute (La Jolla, CA). T-HEp3, D-HEp3, PC3, MCF-7, MDA-MB-453, MDA-MB-468, and HT1080 cells were cultured in DMEM and MDA-MB-231, and M24met cells in RMPI with 10% heat-inactivated fetal bovine serum, penicillin (500 units/ml), and streptomycin (200 μg/ml). PC3 cells were transfected with an empty vector (pSV-Neo) or with a construct encoding an HA-tagged active mutant of Mek1, HA-R4F-Mek, kindly provided by Dr. Natalie Ahn (University of Colorado, Boulder, CO). G418-resistant clones were pooled to avoid clonal variation, and R4F-Mek expression was determined by Western blot. PC3-Neo and PC3-R4F-Mek cells were routinely cultured in the presence of 400 μg/ml G418. HT1080 cells were stably transfected with an MKK6b(E) active mutant in pCDNA3.1 (kindly provided by Dr. Jihuai Han (Scripps Institute) or with pCDNA3.1 only, and G418-resistant clones were pooled to avoid clonal variation, and MKK6b(E) expression was determined by Western blot.

Cells were detached with 2 mm EDTA in PBS, washed, and inoculated on the CAMs of 9–10-day-old chick embryos. At different times, postinoculation CAMs were excised, enzymatically dissociated, and single cell suspensions counted (19). In addition, PC3 or MDA-MB-231 cells treated with 2 μm SB203580, PC3-Neo or PC3-R4F-Mek cells, or HT1080-vector, or HT1080-MKK6b(E) cells were also grown on the CAMs as indicated above. In addition M24met or HT1080 cells were pretreated in suspension at 37°C for 20 min, with 10 μg/ml of anti-uPAR antibody (R2) or left untreated, washed, and inoculated onto 10-day-old CAMs. After 2–4 days, tumor growth was evaluated as indicated above. Alternatively, for continuous monitoring of tumor growth, after measuring the nodules size, the tissue was minced and passaged onto new 9–10-day-old CAMs (25).

For IF analysis, cells grown on coverslips were fixed in 3% paraformaldehyde in PBS for 15 min. The coverslips were washed and either permeabilized with 0.1% Triton X-100 or left nonpermeabilized, washed, blocked with 3% normal goat serum in PBS (15 min), and incubated for 1 h at room temperature with antifibronectin (F3648, 1:400) antibodies in 0.1% BSA/PBS or with vehicle alone. After washing and blocking, the secondary antibodies in 0.1% BSA/PBS containing rhodamine-phalloidin conjugate (1:70), or DAPI were added. Coverslips were mounted in Vectashield and kept at −20°C. Standard epifluorescence was captured with a Nikon E-600 epifluorescence photomicroscope (Toyko, Japan) using Plan-Neofluar ×40 and ×100 (N.A. 1.5 Oil) lenses (Nikon) or Plan-Apochromat lens ×63 and ×100 (Nikon) through a Diagnostic Instruments, Inc. SPOT-RT digital camera using SPOT and Adobe Photoshop 6.0 software on a Macintosh G4 computer.

For uPAR, detection cell lysates of the different cell lines were prepared as previously described (19), centrifuged, and equal amounts (50 μg) of the supernatant proteins were used in Western blots with R2 anti uPAR mAb. FACS analysis was performed as described previously (12). Antibodies (HA5, anti-α5β1, or isotype-matched IgG) were added to 5 × 10⁵ cells at 10 μg/ml and incubated at 4°C for 30 min, followed by two washes and FITC-conjugated goat antimouse (1:1000 IgG) incubation. Cells were fixed in 5% formaldehyde in PBS and analyzed in FACS-SCAN equipped with laser 488 (Becton Dickinson, San Jose, CA). Data analysis was performed using Cell Quest software and a Macintosh G3 computer.

Cells grown on glass coverslips in DMEM with 10% FBS or 5–10% FN-depleted FBS, with or without 5–10 μg/ml human serum FN were fixed and stained for FN, F-actin, or DAPI, as indicated above. To test the effect of disruption of uPAR/integrin interaction on FN-fibril formation, cells were plated on coverslips in 5% FN-depleted FBS/DMEM for 1 h and treated with anti-uPAR antibody to domain III (R2, 10–20 μg/ml) isotype-matched IgG (15 μg/ml) antibody or peptide α3–25 that inhibits uPAR/integrin interactions (26) for 20 min at 37°C, incubated overnight with 5–10 μg/ml human FN, fixed, and stained for FN as indicated above. FN-fibrils were counted on 200–350 cells/treatment in triplicate experiments and expressed as the percentage of DAPI-positive nuclei.

To monitor ERK and p38 activity levels, we used a reporter system (Pathdetect, Stratagene) based on the use of fusion proteins that comprise GAL4 DNA binding domain fused to the activation domain of specific transcription factors that, in turn, drive the expression of luciferase reporter gene. An expression plasmid encoding a chimeric protein GAL4-Elk1 (pFA-Elk, Stratagene; ERK phosphorylation target) or expression plasmids coding for GAL4-CHOP (pFA-CHOP, Stratagene) or GAL4-MEF2A (p38 phosphorylation targets) (in pCDNA3.1 kindly provided by Dr. Jihuai Han (Scripps Institute; were cotransfected with a Photynus pyralis luciferase reporter gene controlled by 5 GAL4-binding sites linked 5′ to the herpes virus thymidine kinase (tk) promoter (pD700, kindly provided by Dr. Ari Melnick, Albert Einstein College of Medicine, New York, NY). All transient transfections were performed using Fugene reagent (3:1 Fugene/DNA ratio) and 1:5 ratio of the trans-activator and reporter plasmids. In some experiments, additional plasmids encoding a dominant negative mutant of p38 (12), an active mutant of MKK6 [MKK6b(E)], active or inactive mutants of Cdc42 (Cdc42QL and Cdc42N17, kindly provided by Dr. Silvio Gutkind, NIH, Bethesda, MD) were cotransfected. In some experiments, after overnight transfection with the reporter system cells were treated for 12–24 h with 5–10 μm p38 inhibitor SB203580 and/or 50 μm arsenic trioxide as a stress inducer or with 25–50 μm PD98059 Mek inhibitor. In all transfections, a plasmid-encoding Renilla luciferase (Clontech) was cotransfected (100–500 ng), and firefly luciferase activity was normalized to Renilla luciferase activity. After 48–72 h after transfection, cells were lysed using the passive lysis buffer from Promega, and luciferase activity was detected with a luminometer using the Dual Luciferase Reporter Kit from Promega following the vendor’s instructions. Controls lacking the transactivator constructs (Elk-, CHOP-, or MEF2A-GAL4) but including the GAL4-tk-luciferase construct alone, or in combination with the SV40-renilla luciferase construct or untransfected cells, were included in every experiment and showed no or very low luciferase activity that was not affected by treatments. In addition, cell lysates of subconfluent monolayers of the different cell lines, prepared as previously described (19), were centrifuged, and equal amounts (20–50 μg) of the supernatant proteins were used in Western blots to detect either active or total ERK and p38 levels using phospho-ERK (p42/p44) or ERK1 and phospho-p38 or p38 antibodies. The effect of p38-inhibition on ERK activation was tested by treating the cells with 1–10 μm SB203580 or its inactive analogue SB202474 or 0.05% DMSO for 5–20 min or 5–48 h in serum-free DMEM. The phospho-ERK and ERK levels were analyzed by Western blotting. In some experiments, cells pretreated with medium alone, control IgG (10 μg/ml), or R2 antibody (10 μg/ml) were plated on poly-l-lysine (5 μg/ml) or fibronectin (5 μg/ml), and after 20 min, the cells were lysed and assayed for ERK activation by Western blot. In all cases after SDS-PAGE and transfer, the membranes were blotted with the corresponding antibodies and the signal was developed using ECL.

Active (GTP-loaded) Rac or Cdc42 was detected using a Rac/Cdc42 activation kit from Upstate Biotechnology following the vendor’s instructions. Briefly, after removing the medium, cell monolayers were lysed directly with a 2% glycerol, 25 mm HEPES (pH 7.5), 150 mm NaCl, 1% Igepal CA-630, 10 mm MgCl₂, and 1 mm EDTA containing lysis buffer with protease and phosphatase inhibitors and snap frozen in liquid N₂. The lysates were cleared by centrifugation, and the supernatants (0.6–1.0 mg of protein) were incubated for 60 min with Sepharose beads conjugated with a GST-p21 binding domain of PAK-1 (10 μl of beads/10 μg of PAK1-PBD) fusion protein, which binds to GTP, but not GDP, loaded Rac or Cdc42. After incubation, the beads were spun down and washed twice with lysis buffer, resuspended in 2× Laemmli sample buffer with 100 mm β-mercaptoethanol. The Rac or Cdc42 proteins in the precipitates or cell lysates were detected by Western blot using anti-Rac (clone 102) or anti-Cdc42 (clone 44) mAbs from BD PharMingen (San Diego, CA) as indicated above.

Subconfluent monolayers were washed three times with cold PBS, the cells were incubated on ice for 20 min with 5 ml of 0.5 mg/ml sulfo-NHS-biotin (Pierce, Rockford, IL), and the reaction was stopped by aspirating and washing the cells twice with 10 ml of ice-cold PBS. The cells were then scraped in 1 ml of PBS containing protease inhibitors, spun at 4°C, and the pellets were lysed with a buffer containing 1% Triton X-100, 50 mm HEPES (pH 7.5), 150 mm NaCl, 1 mm CaCl₂, 1 mm MgCl₂, and protease inhibitors as described for radioimmunoprecipitation assay buffer. Immunoprecipitation and biotinylated proteins detection were performed as indicated below.

Untreated M24met and T-HEp3 cells or surface-biotinylated T-HEp3 and MDA-MB-231 cells were lysed and extracted for 1 h with a lysis buffer containing 1% Triton X-100, 50 mm HEPES (pH 7.5), 150 mm NaCl, 1 mm CaCl₂, 1 mm MgCl₂, 1 mm orthovanadate, 1 mm NaFl, and protease inhibitors. Triton X-100 soluble fraction or insoluble fractions (400 μg protein) extracted for 30 min with radioimmunoprecipitation assay lysis buffer, after preclearing with protein-G beads and isotype-matched IgG for 30 min, were incubated with 4 μg of anti-α5β1 (HA5), R2 mAb, or isotype-matched IgG overnight at 4°C, precipitated with protein G-agarose beads and washed twice. The beads were resuspended in 2× Laemmli sample buffer, heated to 95°C for 10 min, and analyzed by Western blotting using anti-α5β1 integrin polyclonal antibodies or anti-uPAR 399R polyclonal antibodies. Alternatively, immunoprecipitates of surface-biotinylated proteins were separated on SDS-PAGE, transferred to a membrane, blotted with streptavidin-HRP, washed, and the signal developed with ECL reagent.

We previously showed that the functional assembly of a multimeric complex formed by uPA, uPAR, and α5β1and EGFR and the formation of fibrillar FN on the surface of squamous carcinoma HEp3 cells are required to induce and maintain high ERK/p38 activity ratio. The high ERK/p38 ratio allows cell cycle progression and tumor growth. Disruption of this complex caused an inversion in the ERK/p38 ratio toward p38 signaling, a change that results in cell cycle arrest and dormancy in vivo (12, 19). Here, we set out to explore the generality of this paradigm by considering the role of ERK/p38 activity ratio as a predictor of in vivo behavior. We also tested whether direct targeting of ERK or p38 activities can shift the cancer cell phenotype between the states of dormancy and tumorigenicity.

Human tumor cell lines representative of head and neck carcinoma (T-HEp3, tumorigenic, and D-HEp3 dormant), fibrosarcoma (HT1080), metastatic melanoma (M24met), prostate carcinoma (PC3), and breast carcinoma (MCF-7, MDA-MB-231, MDA-MB-453, and MDA-MB-468) reported to have varying malignant potential (27) were examined for their ability to form tumors on the CAM of chick embryo. The cells were inoculated onto 9–10-day-old CAMs (5 × 10⁵ cell/CAM), and their growth was monitored by weekly passage of similar amounts of tumor minces onto fresh CAMs for up to 8 weeks. Fig. 1 A shows that M24met and HT1080 cells formed rapidly growing, large tumors, a behavior similar to that of T-HEp3 cells, in which tumorigenicity was shown to be uPAR and integrin dependent (12, 19). In contrast, PC3 cells and three breast cancer cell lines formed small tumor nodules, which, similarly to D-HEp3 cells (19, 24), did not increase in size for up to 8 weeks. An additional breast cancer cell line, MCF-7, and a prostate cancer cell line, LNCaP, behaved like the D-HEp3 cells (data not shown), even if the inoculum size was increased 4-fold (28). Microscopic examination of trypan blue stained minces of the small CAM nodules showed that even after 8 weeks in vivo, they contained live cancer cells. We previously showed (19, 25) that the in vivo dormancy of D-HEp3 cells results from reduced proliferation and not increased apoptosis. To distinguish between these two potential mechanisms of dormancy induction, sections of growing (T-HEp3) or dormant (MDA-MB-231, D-HEp3, and PC3) tumors were examined for apoptosis using TUNEL assay. These experiments revealed an apoptosis level of <5% that did not differ between the growing and the dormant tumors (data not shown), suggesting that as previously described (19, 25), dormancy most likely is the result of a failure to activate the proliferative signal.

Having identified cell lines that displayed in vivo behavior similar either to T-HEp3 or D-HEp3, we tested whether their uPAR and α5β1-integrin expression conformed to the previously established pattern. uPAR level was measured by Western blot and α5β1-integrin by FACS analysis. The uPAR-rich T-HEp3 cells and the uPAR-poor dormant D-HEp3 cells (19) served as markers for the range of uPAR expression. The two cell lines that formed progressively growing CAM tumors (Fig. 1,A) had uPAR protein levels similar to that of T-HEp3 (HT1080) or only slightly lower (M24met; Fig. 1,B). The level of surface α5β1-integrin was comparable with that of T-HEp3 in M24met cells and even higher in HT1080 cells (Fig. 1,C). HT1080 and M24 cells express levels of EGFR similar to that in HEp3 cells (data not shown and Refs. 29, 30, 31). In contrast, cells that did not form tumors (dormant) had uPAR levels lower than D-HEp3 cells (PC3 and MDA-MB-453) or almost undetectable (MCF7 and MDA-MB-468; Fig. 1,B) and LNCaP (results not shown). The surface integrin expression of the dormant cells varied from more than 5-fold lower than in T-HEp3 cells to undetectable. An exception was the MDA-MB-231 cell line, which was dormant in vivo, yet expressed levels of uPAR comparable with that of T-HEp3 or M24met cells (Fig. 1,B) and had high level of α5β1 (Fig. 1 C). Therefore, with the exception of MDA-MB-231 in the rest of the cell lines tested, the level of expression of uPAR and α5β1 integrin showed a perfect correlation with tumorigenicity or dormancy, confirming the validity of the paradigm established in HEp3 cells. This suggests that the presence of these two receptors may be useful in predicting the in vivo behavior of tumor cells.

We previously showed that when expressed at high levels, uPAR interacts with and activates α5β1-integrin. This results in the integrin being able to organize FN into insoluble fibrils, suppression of p38 activity, tipping of the balance in favor of ERK activity, and tumor growth. Reduction of uPAR level reverses the balance in favor of p38 (12). Should the paradigm be applicable to other cell lines, the prediction would be that HT1080, M24met, and MDA-MB-231 will form FN-fibrils and will have high ERK/p38 ratios. The prediction for cell lines with low or undetectable levels of uPAR and α5β1 integrin (Fig. 1, B and C) would be that they will form no FN-fibrils and will have a ratio of ERK/p38 activities favoring p38. Examination of FN-fibrils by IF using anti-FN antibody revealed 100% conformity between FN-fibril presence and in vivo behavior (Fig. 2,A and Table 1). Again, the MDA-MB-231cells, despite a robust uPAR and α5β1-integrin expression, did not form FN-fibrils (Fig. 2,A), a finding that conformed to their dormancy on the CAM (Fig. 1 A and Ref. 28).

The lack of FN-fibrils in MDA-MB-231 cells suggested that the highly expressed uPAR may not be capable of α5β1-integrin activation. Because we previously showed that uPAR/integrins association was a prerequisite for their activation in T-HEp3 cells, we tested the existence of this association in MDA-MB-231 cells. These and the T-HEp3 cells serving as positive control were surface biotinylated, lysed, and used to immunoprecipitate the α5β1-integrin. Fig. 2,B shows that as previously determined, the anti-α5β1 antibodies pulled down uPAR in T-HEp3 cell lysates, whereas no integrin-associated uPAR was detected in MDA-MB-231 cells, which express similar uPAR levels (Fig. 1,B) and have similar levels of biotinylated uPAR (Fig. 2 C). (The three prominent bands found in MDA-MB-231 lysates are not specific as they are present also in lysates incubated with preimmune IgG.)

The effect of uPAR-induced activation of integrins, or lack thereof, on the level of ERK activation was tested by Western blot analysis of cell line lysates using antibodies that recognize phosphorylated (active) and total ERK1/2 proteins (Fig. 3,A). To compensate for the differences in total ERK level between cell lines, ERK activity was determined by scanning the bands and calculating the ratios of phospho-ERK to ERK in each cell line. The ratios were expressed as a fraction of that in T-HEp3 cells, which was arbitrarily set as 1. ERK phosphorylation, which varied widely among cell lines (Fig. 3,A), provided an opportunity for testing their link to integrin activation and in vivo growth. The tumorigenic HT1080 and M24met cells, both with high uPAR and active integrins (as determined by their ability to form FN-fibrils; Fig. 2,A and Table 1), had activity similar to that of T-HEp3 cells (0.63 and 1.04, respectively; Fig. 3,A). We previously showed that in T-HEp3 cells, the uPAR/α5β1-integrin complex activated EGFR in a ligand-independent fashion and that EGFR served as a mediator of the uPAR-induced signal to ERK (20). Although, we did not specifically address this question here, it is likely that EGFR serves similar function in the HT1080 and M24met because both express EGFR (data not shown and Refs. 29, 30, 31). The phosphorylated ERK in the dormant cell lines, PC3, MCF-7, MDA-MB-453, and MDA-MB-468 ranged between 0.03 and 0.33 and was similar or lower than that of the dormant D-HEp3 cells, which was 0.27. Phosphorylated ERK in MDA-MB-231 cells was 0.49 (Fig. 3 A), a level that apparently was not sufficient to facilitate their in vivo growth. This fits our findings in HEp3 cells showing that a 2-fold reduction in ERK activity by experimental manipulation inhibits tumor cells growth in vivo (see below and Refs. 12, 19, 20). Overall, high basal ERK activity measured in vitro appears to be a good predictor of in vivo growth.

Because we have previously shown that dormancy of D-HEp3 cells was associated with high p38 activity, we wondered whether the same was true for other cell lines that did not proliferate in vivo. More importantly, we wanted to explore whether p38 level can be used to predict in vivo behavior of cancer cells. Total p38 and P-p38 were analyzed by Western blotting, the p38 activity was estimated from the P-p38 to p38 ratio and expressed as a fraction of activity in D-HEp3 cells (Fig. 3 B). We found that, although in some cells the p38 activity matched their in vivo behavior, in others, it did not. For example, whereas the dormant PC3 cells had low p38 activity, the activity in the tumorigenic M24met cells was high, suggesting that by itself p38 activity is of modest value as a predictor of in vivo behavior.

To show that the level of phosphorylated ERK or p38, measured by Western blots in lysates of cells, correspond to their in vivo activity, we measured their ability to phosphorylate and, thus activate, their downstream targets. Cells to be tested were transiently transfected with plasmids coding for GAL4-Elk1 (target of phosphorylation by ERK) or GAL4-CHOP/GADD153 [target of phosphorylation by p38 (32)] fusion proteins. Controls lacking the transactivator constructs (Elk-, CHOP-, or MEF2A-GAL4) but including the GAL4-tk-luciferase construct alone or in combination with the SV40-renilla luciferase construct or untransfected cells were included in every experiment. The activation of ERK or p38 was measured by transactivation of a luciferase reporter. To establish the parameters of sensitivity and responsiveness of p38, the CHOP-dependent trans-activation of luciferase was examined in several cell lines under basal conditions, after treatment with 50 μm arsenic trioxide [a known inducer of p38 activity (11)] with and without treatment with a specific inhibitor of p38α and β-isoforms, SB203580 (5 μm), and after transfection with constitutively active MKK6b(E) (the immediate specific upstream activator of p38; Fig. 3,C). In MCF-7 cells, arsenic treatment increased CHOP activity by ∼2.5-fold. SB203580 treatment reduced the basal CHOP activity by >50% and eliminated arsenic-induced CHOP activity. In addition, treatment of T-HEp3 cells with arsenic also caused a strong induction of CHOP activity (Fig. 3,C). In D-HEp3 cells, MKK6b(E) expression caused an increase in CHOP activity similar to that produced by arsenic in MCF-7 cells. Taken together, these results suggest that CHOP activity reflects accurately p38 activation (Fig. 3,C). Similarly, the correlation between phospho-ERK levels and Elk activation was tested in T-HEp3, D-HEp3, and MCF-7 cells. As shown in Fig. 3,D, T-HEp3 cells showed a ∼4-fold higher basal Elk activity than D-HEp3 and MCF-7 cells, a ratio similar to that determined by Western blot analysis (Fig. 3,A). The high basal ERK/Elk activity in T-HEp3 cells was eliminated almost completely by overnight incubation with 25–50 μm Mek1/2 inhibitor PD98059 (Fig. 3 D). These results show that Elk and CHOP trans-reporting system reproduces the basal levels of active ERK and p38 and that it reports with fidelity changes in their ratios.

To test whether the measure of functional activity ratios of ERK and p38 will improve the predictive value of P-ERK even further, we transfected the three tumorigenic and five dormant cells lines with Elk or CHOP trans-reporting system, measured their luciferase activity, and calculated the ratio of Elk to CHOP. As shown in Fig. 3,E, two of three tumorigenic cells (T-HEp3 and HT1080) had a ratio of Elk/CHOP > 1, whereas the rest of the cell lines, including the tumorigenic M24met, had Elk/CHOP ratios < 1, with some as low as 0.01 (MCF-7 and PC3). Thus, the Elk/CHOP ratio predicts the in vivo behavior as well as the phospho-ERK level alone, except that it segregates the MDA-MB-231 cells in which P-ERK level is borderline for growth (Fig. 3 A), conclusively into the dormant group. In contrast, unlike the P-ERK level, the ELK/CHOP ratio in M24met wrongfully predicts their inability to grow in vivo.

To examine the existence of a functional link between the ERK/p38 (Elk/CHOP) ratio and in vivo behavior, we used approaches previously verified in HEp3 cells (12, 19, 20) that shifted the balance of these two activities. HT1080 cells were chosen to examine the effect of ERK/p38 activity shift in favor of p38 on dormancy induction. We first established that a negative cross-talk between P-p38 and P-ERK, first discovered in HEp3 cells, also exists in HT1080 cells. Cells were transiently transfected with the Elk-luciferase reporter alone or cotransfected with a DNp38 expression vector, incubated with or without serum, lysed, and examined for luciferase activity. Fig. 4,A shows that, as evidenced by a ∼2-fold increase in Elk-activated luciferase activity, it is possible, by inhibiting p38 activity, to additionally increase the level of active ERK. The effect was found to be serum independent (Fig. 4,A). This suggests that even in highly tumorigenic cells such as HT1080 with high ERK activity, p38 can still exert a negative regulatory effect. This conclusion was additionally supported by an experiment in which stable transfection of HT1080 cells with a constitutively active MKK6b(E) mutant, strongly reduced the P-ERK level (Fig. 4,B). We next examined whether inhibition of P-ERK and a resulting shift in balance in favor of p38 will affect the in vivo growth of HT1080. We used anti-uPAR antibody R2, which interferes with the uPAR/α5β1 complex and lowers P-ERK (19, 20) or a direct inhibition of Mek activity with PD98059 inhibitor. To test the effect of R2 on ERK activity, cells were pretreated with preimmune IgG or R2 antibodies and plated on FN or PL, lysed 20 min later, and tested for ERK and P-ERK levels. As shown in Fig. 4,C, adhesion to FN induced a ∼4-fold increase in P-ERK level that was reduced by 56% upon treatment with R2 antibody. Similar pretreatment of HT1080 cells with R2 antibody before inoculation on the CAM significantly reduced tumor volume (Fig. 4,D) and tumor cell number (data not shown) after 4 days of in vivo growth. This indicates that inhibition of ERK and most likely a change in ERK/p38 ratio were responsible for a significant growth inhibition of these cells in vivo. We next tested whether inhibition of ERK activity by a specific pharmacological inhibitor of Mek1/2 (PD98059), an approach that completely bypasses the ERK inhibitory loop generated by interfering with the uPAR/α5β1 complex, will produce similar effect on growth. Pretreatment of HT1080 cells for 24 h with the 40-μm PD98059, although not affecting the viability of these cells in culture (data not shown), caused a significant growth inhibition after 4 days on CAMs (Fig. 4,E). Finally, we tested a direct effect of increased p38 activity on in vivo growth. HT1080 cells stably transfected with an active mutant of MKK6b(E) (Fig. 4,B), which grew in culture as well as vector-transfected cells (data not shown) upon inoculation on CAMs and after 7 days of incubation in vivo, showed significantly reduced growth (Fig. 4 F). Thus, regardless of the mechanism used, P-ERK inhibition that increases the relative contribution of p38 signaling reduces the in vivo proliferative ability of these cells.

The identification of a negative feedback loop between active p38 and ERK meant that blocking of p38 activity should simultaneously release the inhibition on ERK (12). Such intervention should shift the balance in favor of ERK. To inhibit p38 directly, we used treatment with SB203580, which has an effect similar to that obtained by expression of a dominant negative p38 (12). We also activated ERK directly by transfecting cells with an active mutant of Mek1 (R4F-Mek1), the immediate-specific upstream activator of ERK. Because, out of the cell lines tested, PC3 cells had the lowest level of active ERK (Fig. 3,A and Ref. 33) and they remain dormant on the CAM, they were chosen for these experiments. PC3 cells, treated for 24 h with 2 μm SB203580 or left untreated, were examined for P-ERK level by Western blots. The treatment caused a strong increase in P-ERK level (Fig. 5,A). This suggested that at least, in part, the lack of ERK activity in PC3 cells is because of negative feedback from p38. To test if an increase in ERK activity will affect their in vivo behavior, PC3 cells pretreated with 5 μm SB203580 or controls were inoculated on CAMs. While parental cells inoculated on CAMs produced only ∼0.5 population doublings in 6 days, cells pretreated with 5 μm SB203580 underwent 2.5 population doublings (Fig. 5 B).

To examine the effect of direct ERK activation on in vivo growth, pools of PC3 cells transfected with vector and HA-tagged R4F-Mek1 were selected for antibiotic resistance and examined for ERK activity. Overexpression of R4F-Mek1 increased P-ERK level and increased the expression of uPAR (Fig. 5,A), a downstream target of ERK (12, 34), indicating that the pathway was functional. A comparison of in vivo growth of vector and R4F-Mek1-transfected cells showed (Fig. 5,B) that R4F-Mek1-induced activation of ERK coincided with a 4-fold increase in the doubling of tumor cells on CAM. Moreover, although the vector-transfected cells remained dormant, the R4F-Mek1-transfected cells formed tumors of increasingly larger size after passage on CAM (Fig. 5,C). Similarly, 48 h of pretreatment of MDA-MB-231 cells with SB203580 altered their in vivo behavior from dormant to proliferative (Fig. 5 D). We conclude that the change of ERK/p38 activity balance achieved either by a direct Mek1 activation of ERK or by release of ERK inhibition through reduction of p38 activity can lead to interruption of the state of in vivo dormancy.

The tumorigenic M24met cell line, with high uPAR, activated integrins and FN-fibrils (Figs. 1,B and 2,A), was found to have high P-p38 levels (Fig. 3), high CHOP activity (Fig. 6,B) and a low Elk/CHOP ratio (Figs. 3,E and 6,B). How could the high ERK activity be maintained in presence of very high p38 activity? Is it possible that the negative feedback between p38 and ERK is disrupted in these cells? To test this, M24met cells were incubated with 5 μm SB203580. Neither very short treatment (5–20 min), nor prolonged treatment (5–24 h), resulted in an increase in ERK phosphorylation (Fig. 6,A), indicating lack of negative regulation by p38. This was not attributable to a general loss of sensitivity to regulation because treatment with a combination of 4 μm SB203580 and 40 μm PD98059 (Mek inhibitor) reduced the level of ERK phosphorylation (Fig. 6,A), showing response expected of the classical Mek-ERK pathway. Was the lack of negative feedback indicative of other alterations in the p38 pathway? To examine this, M24met cells were transfected with either GAL4-Elk or GAL4-CHOP reporter constructs alone or in cotransfection with MKK6b(E) or DNp38 mutants (Fig. 6,B). The results show that the high CHOP activity was unaffected by inhibition of p38 with DNp38 mutant (Fig. 6,B), suggesting that the basal CHOP activity was p38 independent. The expression of MKK6b(E) produced a weak increase in CHOP activity (Fig. 6,B) that did not correlate with a reduction in Elk, confirming the lack of cross-talk between p38 and ERK. To test whether some of P-p38 signaling capacity was retained, we examined the fate of another downstream target of p38α, MEF2A (35). Luciferase activity of M24met cells transfected with GAL4-MEF2A and luciferase reporter plasmid was only slightly increased when MKK6b(E) was expressed (probably because p38 was already maximally active) and strongly inhibited by cotransfection with DNp38 mutant (Fig. 6 C), suggesting that at least part of the p38 signaling pathway may have remained intact.

The experiments summarized in Fig. 6 show that in comparison to T-HEp3 cells, the p38 pathway in M24met cells is altered such that the p38 activity is high, there is a loss of negative feedback to ERK and uncoupling of some of its downstream signaling targets, yet the cells remain tumorigenic. These differences prompted our inquiry into the molecular interactions responsible for the tumorigenicity of M24met cells. As a test of physical uPAR/α5β1 association, a prerequisite for integrin and ERK activation by uPAR in HEp3 cells (12, 19) we showed that these proteins could also be coimmunoprecipitated with anti-α5β1-integrin antibody in M24met cells (Fig. 7,A). [The two cell lines express similar levels of surface α5β1-integrin (Fig. 1,C).] Moreover, FN-fibrils formed by these cells (Fig. 2,A) were sensitive to disruption of the uPAR/integrin association with anti-uPAR antibody R2 (12). Compared with the control IgG, the R2 treatment of cells grown on coverslips in presence of FN, reduced the percentage of FN-fibril containing cells by ∼50% at 7 h and by ∼70% after 18 h. (Fig. 7, B and C). Similar results were obtained when M24met cells were treated with peptide α3-25, (Fig. 7,B) shown previously to disrupt uPAR-integrin interaction (26, 36, 37). We then tested whether FN-fibril disruption by R2 antibody affects the in vivo growth of M24met. As shown in Fig. 7 D, when compared with isotype matched IgG, pretreatment with R2 antibody significantly reduced the in vivo proliferation of M24met cells, suggesting that high uPAR is responsible for the high P-ERK and the positive mitogenic in vivo signal. These results show that despite high p38 activity pathway, M24met cells maintain the molecular assembly of proteins that, similarly to T-HEp3 cells, are responsible for their in vivo growth.

We previously showed that when uPAR level is reduced, α5β1-integrin becomes inactive, ERK activity drops, and p38 activity increases (12). Because Rac and Cdc42 are downstream effectors of integrin-induced signaling and because Cdc42 was shown to be involved in p38 activation (13, 38, 39), we hypothesized that by regulating the state of integrin activation and the activation of the Rho family GTPases, uPAR may affect p38 activity. We measured Rac and Cdc42 activities in T-HEp3 (high uPAR and low p38) and D-HEp3 cells (low uPAR and high p38 activity) by pull-down assays using the p21-binding domain of PAK1 (PBD-GST), which only interacts with GTP-bound forms of Cdc42 or Rac. T-HEp3 cells had ∼3.5-fold more active Rac than D-HEp3 cells (Fig. 8,A) and only marginally activated Cdc42. Conversely, upon uPAR down-regulation (as in D-HEp3 cells) or by expression of uPAR antisense (AS24), the active Cdc42 level increased by ∼3-fold when compared with the parental T-HEp3 cells (Fig. 8 B). PBD pull-down assays performed on two additional cell lines, MDA-MB-231 with inactive α5β1-integrin and MCF-7 with very low level of α5β1-integrin expression, showed active Cdc42 levels as high as in D-HEp3 cells. All three of these cell lines had high p38 activity.

Using transient transfections and the CHOP trans-reporting system, we next explored whether activation of Cdc42 was functionally linked to p38 activation. Expression of an active mutant of Cdc42 (Cdc42QL) in high uPAR, T-HEp3 cells caused a ∼3-fold increase in CHOP activation, whereas a dominant negative mutant of Cdc42 (Cdc42N17) did not additionally reduce the low basal activity (Fig. 8,C). Transfection of the D-HEp3 cells with MKK6b(E) mutant increased additionally the already high p38 activity, whereas transfection of a dominant negative mutant, Cdc42N17, reduced the basal activity by ∼2-fold (Fig. 8 D). An even greater reduction in CHOP activity (85 and 60%, respectively) was found in MDA-MB-231 and MCF-7 cells transfected with Cdc42N17. These results suggest a role for Cdc42 in p38 activation in cells in which α5β1-integrins are either inactive (D-HEp3 and MDA-MB-231) or expressed at a very low level (MCF-7).

Taken together, our results support the notion that the regulation of ERK and p38 activities by the uPAR-integrin complex is important for the growth of several tumor types and that targeting this complex may be clinically beneficial. Moreover, even in cells that do not form uPAR/integrin complex, either because of missing components or because of changes in molecular interactions, altering the ERK/p38 ratio by direct targeting of either one of these kinases, shifts their in vivo behavior between proliferation and dormancy.

The results of our experiments led us to conclude that ERK activity level and ERK/p38 activity balance are valid general predictors of tumorigenicity and dormancy in vivo. On the basis of our previous work, we functionally defined dormancy as a state in which cancer cells inoculated in vivo persist in a viable state without leading to an increase in tumor volume. In our experimental model of human head and neck carcinoma HEp3 grown on CAM, dormancy was achieved by proliferation arrest and not through increased apoptosis, and it was reversible (19, 25). We now find that many cancer cell lines, and especially the ones maintained in culture for prolonged periods, are dormant on the CAM. The fact that cells such as PC3 or MDA-MB-231, known to produce tumors in nude mice, enter dormancy when inoculated on the CAM, may seem paradoxical. This discrepancy can, however, be reconciled by the observation that in the nude mice, inocula as large as 10⁷ cell often require >6 weeks to produce palpable tumors (40, 41, 42). The reason for this delay is believed to be scarcity of tumorigenic cells within the inoculum capable of in vivo proliferation. Our studies of the CAM tumors, where the cells are accessible for continuous examination, indicate that the entire tumor cell population enters dormancy and, after a period of time in vivo, reemerges to initiate progressive tumor growth (25). A similar mechanism of dormancy may exist in the nude mouse and may be responsible for the latency of many transplantable tumors.

We confirmed our finding that uPAR/integrin complex and FN fibrils are important regulators of ERK activation by showing that in 4 of the 10 cell lines tested that expressed high uPAR and α5β1-integrin, 3 had active ERK and were tumorigenic when inoculated on the CAM. In contrast, in 6 cell lines, which had low or no uPAR and/or low integrin level, the signal to ERK was greatly reduced. In these cells, the p38 activity was elevated, resulting in an ERK/p38 ratio that favored p38 and dormancy on CAMs. Over a period of several months, they consistently produced only small nodules that, despite the continuous presence of live tumor cells, did not increase in volume (Fig. 1 A and Ref. 12), a behavior conforming to our definition of dormancy.

As previously shown for T-HEp3 cells (12), we showed that changing the balance of ERK/p38 in HT1080 cells by inhibiting Mek or activating p38 by an active MKK6 mutant or disruption of uPAR/integrin complex by anti-uPAR antibody, resulted in reduction of ERK activity in HT1080 cells and inhibition of their in vivo growth. (These experiments were not of sufficient duration to conclude that persistent dormancy was established.) In a reverse approach, changing the ERK/p38 ratio in favor of ERK in dormant PC3 cells, using an inhibitor of p38, SB203580, or transfection with an active Mek-R4F mutant, led to resumption of in vivo growth. Thus, cancer cells remain sensitive to changes in their ERK/p38 activities and they respond by a profound shift in their in vivo behavior.

Although the molecular interactions leading to uPAR-induced positive signals to ERK are at least partially understood (19, 20, 21), the mechanism of p38 activation induced by loss of uPAR is unknown. Because uPAR appears to regulate integrin activation, we tested whether classical downstream transducers of integrin signaling such as Rac and Cdc42 GTPases are involved in the differential regulation of ERK and p38. We found that in high uPAR-HEp3 cells, activation of integrins corresponds with strongly activated Rac. These results are in agreement with the recently published data showing that uPAR transfection of fibroblasts caused Rac-dependent membrane ruffling (43). Rac has also been shown to be required for facilitating the Ras-ERK signaling and cell cycle progression (44). In contrast, we found that cells with low uPAR and inactive α5β1-integrins (D-HEp3, AS24), or cells with low uPAR and integrin levels (MCF-7) or even cells with high uPAR but inactive α5β1-integrin (MDA-MB-231, see below), had active Cdc42 and high p38 activity levels. Moreover, we showed that Cdc42 activation was functionally linked to p38 activation in all these cells. These results are in agreement with reports showing that Cdc42, by activating p38, induces p21/p27 and represses cyclin D1 expression, thus causing cell-cycle arrest (13, 14, 15, 45). Our attempt at identifying molecules that may be responsible for regulation of Cdc42 included a 53-kDa Cdc42GAP (46). However, testing its association with α5β1 or FAK did not provide clues of its role in the differential activation of Rac and Cdc42 in the uPAR-rich and uPAR-deficient cells (data not shown). Recent findings (47) suggest that analysis of spatial and temporal regulation of effectors of Rac or Cdc42 activity by GDIs may have to be undertaken to understand the differential regulation of p38.

Of the cell lines studied, two, the MDA-MB-231 and the M24met, warrant discussion. The MDA-MB-231 cell line, despite having high levels of uPAR and α5β1-integrin, does not form FN-fibrils and has a level of active ERK that makes its assignment to the tumorigenic or dormant group uncertain. The lack of fibrils suggests that despite high uPAR level, its α5β1-integrin is inactive. The finding of relatively high level of active Cdc42 and p38 supports this conclusion. We showed that a prerequisite for uPAR-activation of integrin is the association of the two proteins (19), yet whereas anti-α5β1-integrin antibody coimmunoprecipitated uPAR from lysates of T-HEp3 cells, no such coimmunoprecipitation was found in MDA-MB-231 cells. It is possible that in some cells, uPAR partitions with other membrane receptors and becomes unavailable to partner with α5β1-integrin. Chapman et al. (37) showed that in 293 cells in which uPAR was overexpressed by transfection, 90% of uPAR was found in a complex with α3β1 integrin. Because they found that the α5β1-integrin content exceeded that of α3β1 by ∼4–5-fold, yet uPAR interacted almost exclusively with α3β1, they suggested that association of uPAR with α3 was preferred. However, we find that both in T-HEp3 cells and in MDA-MB-231 cells uPAR may associate with the predominantly expressed integrin. In T-HEp3 cells, uPAR associates with α5β1, which is at 6-fold excess over α3 (results not shown), whereas it seems to associate with α3 in MDA-MB-231 cells, which have a 2-fold excess of this integrin (37). If, however, a preference for association with α3 exists, then it is possible that α3β1-integrin has to be sequestered by other membrane proteins to allow uPAR to interact with α5β1 or, as suggested by published evidence, to prevent it from exerting a trans-dominant effect on other integrins (48). Hemler et al. (49, 50) have shown that a tetraspanin CD151 (PETA) enters into specific, lateral interactions with the α3β1 integrin. Testa et al. (51) has shown that CD151 is overexpressed in metastatic HEp3 cells in which we showed the α5β1-integrin to be activated by frequent association with uPAR (19, 20). CD151 could thus be a regulator of α3β1 availability. We are in the process of identifying the mechanism that dictates the preferential association of uPAR with individual integrins.

The second exception to the rule that high uPAR/integrin content and their proper interaction lead to high ERK and low p38 activities is the melanoma M24met. This cell line forms extensive FN-fibrils and has very high ERK activity but also very high p38 activity. This creates an ERK/p38 ratio predictive of dormant cells, yet the cells are highly tumorigenic and metastatic. This finding poses a dilemma because we showed that in all other cell lines studied, p38, through a negative feedback, exerts an inhibitory effect on ERK. How is it then possible that in M24met cells high ERK activity can coexist with high p38? The mechanism responsible for the negative feedback from p38 to ERK remains unknown. One study (52) shows that ERK1/2 coimmunoprecipitates p38 and that this interaction with p38 prevents ERK phosphorylation. Regardless of the mechanism, published reports suggest that the p38 pathway may be altered or dysfunctional in melanomas (53, 54, 55, 56). In some melanoma cell lines, the downstream targets appear to be selectively uncoupled from p38 (57). An aberrant function of the p38 pathway was also identified in a rhabdomyosarcoma (58), which despite high myoD expression, known to induce differentiation in normal myoblasts, continue to proliferate without undergoing differentiation and, only upon hyperactivation of the p38 pathway by MKK6b(E), proceed to differentiation. In some of these tumors, p38 does not signal properly to the downstream targets (58), suggesting an escape mechanism from the high p38 activity. In confirmation of these results, we found no evidence that the high p38 activity in M24met cells is engaged in a negative cross-talk to ERK (Fig. 6,A). Moreover, we found CHOP activity, a transcription factor known to be specifically activated by p38 phosphorylation, to be largely independent of p38 signaling (Fig. 6,B), yet another downstream target, MEF2A, to be regulated by p38 (Fig. 6 C). Interestingly, a p38-independent, phosphorylated form of CHOP that functions in interleukin 1-induced interleukin 6 gene transactivation has been identified (56). These results and published data (56, 57) suggest that the network of cross-talk and the downstream targets may be altered in melanoma and that, despite high levels of p38 activity, these cancer cells continue to proliferate in vivo. Moreover, because p38 activity has been shown to play a role in mRNA stability, including genes important for cancer invasion (59), M24met cell line may define a subgroup of cancers that take advantage of p38 activity while at the same time maintaining and exploiting the proliferative advantage of high ERK.

Overall, based on the study of 10 different cancer cell lines, our results show that uPAR and α5β1 expression and activation, most likely in presence of EGFR (20), generate a high level of ERK activity and, with the exception of melanoma, low p38 activity, which is necessary for the in vivo growth of cancer cells. The high ERK activity feeds into a positive feedback loop that transactivates uPAR (and uPA) expression (12, 34, 60, 61). In turn, the high uPAR, by activating α5β1 maintains high ERK activity (19, 20). The signaling loop required for cancer cell proliferation can be interrupted by a reduction in uPAR level by cleavage of its domain 1, important for uPAR interaction and activation of α5β1 integrin (19, 20, 62), and by loss of uPA and/or FN. It is plausible that these changes take place during cancer cell spread and establishment of metastases causing their initial dormancy. Because we showed that multiple experimental interventions decrease the ERK/p38 ratio, they should be considered in induction of dormancy. Although p38 inducers must be used with caution because some tumor properties may be dependent on this activity (63), direct inhibition of the ERK pathway may be safe. Mek inhibitors have been tested in preclinical studies and showed tumor growth inhibition without toxicity (33). This approach as well as therapies aimed at disruption of the uPAR/integrin interactions to persistently lower the ERK/p38 ratio may induce a state of prolonged dormancy of cells that have already disseminated but not yet entered progressive growth.

We thank rotating PhD students Luciana Giono and Felix Lohmann for help with some of the experiments and Dr. Samuel Waxman for continuous encouragement and support. We also thank the following individuals for donating reagents: monoclonal anti-uPAR antibody R2 from Dr. Michael Ploug, (Finsen Laboratory); P-p38 antibody from Erik Schaefer (Biosource); pD700 plasmid from Dr. Ari Melnick, (Albert Einstein College of Medicine); Cdc42 constructs from Dr. Silvio Gutkind (NIH); HA-R4F-Mek from Dr. Natalie Ahn (University of Colorado); MKK6b(E) and MEF2A constructs from Dr. Jihuai Han (Scripps Institute).