Decorin Protein Core Inhibits in Vivo Cancer Growth and Metabolism by Hindering Epidermal Growth Factor Receptor Function and Triggering Apoptosis via Caspase-3 Activation*

Decorin is not only a regulator of matrix assembly but also a key signaling molecule that modulates the activity of tyrosine kinase receptors such as the epidermal growth factor receptor (EGFR). Decorin evokes protracted internalization of the EGFR via a caveolar-mediated endocytosis, which leads to EGFR degradation and attenuation of its signaling pathway. In this study, we tested if systemic delivery of decorin protein core would affect the biology of an orthotopic squamous carcinoma xenograft. After tumor engraftment, the animals were given intraperitoneal injections of either vehicle or decorin protein core (2.5-10 mg kg-1) every 2 days for 18-38 days. This regimen caused a significant and dose-dependent inhibition of the tumor xenograft growth, with a concurrent decrease in mitotic index and a significant increase in apoptosis. Positron emission tomography showed that the metabolic activity of the tumor xenografts was significantly reduced by decorin treatment. Decorin protein core specifically targeted the tumor cells enriched in EGFR and caused a significant down-regulation of EGFR and attenuation of its activity. In vitro studies showed that the uptake of decorin by the A431 cells was rapid and caused a protracted down-regulation of the EGFR to levels similar to those observed in the tumor xenografts. Furthermore, decorin induced apoptosis via activation of caspase-3. This could represent an additional mechanism whereby decorin might influence cell growth and survival.

The growth of human cancer cells is often dependent or facilitated by the overexpression of receptor tyrosine kinase, such as the EGFR, 2 that provide a growth advantage to the growing and infiltrating neoplasms (1). To prevent the dire consequences of uncontrolled activation of EGFR, a number of negative feedback mechanisms, both extracellular and intracellular, have evolved (2,3). The prominent role of the EGFR as a crucial relay station among various inputs from the environment and cellular responses has raised the significance of this signaling-transducing receptor to a new level and offers new possibilities for therapeutic intervention (4). We have previously shown that decorin, a secreted small leucine-rich proteoglycan (5,6), is capable of suppressing the growth of tumor cells with various histogenetic backgrounds (7,8) by directly interacting with the EGFR (9 -11). Decorin evokes a protracted down-regulation of EGFR tyrosine kinase (12) and other members of the ErbB family of receptor tyrosine kinase (13) and causes an attenuation of the EGFR-mediated mobilization of intracellular calcium (12). Decorin induces expression of the endogenous cyclin-dependent kinase inhibitor p21 WAF1 (14,15) and a subsequent arrest of the cells in the G 1 phase of the cell cycle (7). These growthsuppressive properties of the soluble decorin and its protein core can also affect murine tumor cells (8) and normal human cells, such as endothelial cells (16) and macrophages (17). A number of observations point toward a key role for decorin in the control of cell proliferation. First, decorin expression is markedly induced in most normal diploid cells at quiescence, whereas its expression is absent in most transformed cells (18 -21). Second, although decorin null animals do not develop spontaneous tumors (22), double mutant mice, lacking both decorin and the tumor suppressor gene p53, develop lymphomas at accelerated rates as compared with the p53 null animals, indicating that the absence of decorin is permissive for tumor development (23). Third, transformation induced by the activating transcription factor-3 and the nuclear vSrc and vJun oncoproteins is associated with a marked suppression of decorin gene expression (24 -26). Fourth, decorin expression is differentially down-regulated in hepatocellular (27), lung (28), and ovarian (29) carcinomas, and reduced expression of decorin is associated with poor prognosis in invasive breast carcinoma (30). Fifth, gene therapy of established tumor xenografts using decorin-expressing adenovirus vectors causes a growth inhibition of various tumors (31)(32)(33) and prevents metastastic spreading of a breast carcinoma orthotopic tumor model (34).
In this study, we tested whether systemic delivery of decorin protein core would affect A431 cells grown as orthotopic skin tumor xenografts. Our results show for the first time that systemic delivery of decorin protein core suppresses in vivo tumorigenicity by specifically targeting EGFR-expressing tumor cells, thereby causing a significant inhibition of tumor metabolism and cell division and concurrent increase in apoptosis. These findings were corroborated by in vitro studies showing that decorin at very low concentrations (ϳ2 nM) caused apoptosis by activating caspase-3. Collectively, these data support a complex mode of action for decorin which culminates in tumor growth suppression and raise the possibility of an efficient protein therapy for cancer.
Animal Experiments and A431 Orthotopic Tumor Xenografts-Animal experiments were performed in accordance with the Guide for Care and Use of Laboratory Animals and the Institutional Animal Care and Use Committee of Thomas Jefferson University. Orthotopic tumor xenografts were established as described previously (31). Purification and characterization of biologically active decorin protein core was described before (36). Immunocompromized nu/nu mice (Charles River Laboratories) were examined every day until 2-3-mm tumors were visible. Tumor xenografts were measured (31) and treated every other day by intraperitoneal injections of 60 -200 g of decorin protein core (2.5-10 mg kg Ϫ1 ) per animal for 18 -38 days. Animals were sacrificed at the end of the experiment and the tumor, spleen, liver, lungs, kidneys, and heart were dissected.
Tumors and organs were snap-frozen in liquid N 2 and used for either immunohistochemical or biochemical analysis.
Positron Emission Tomography (PET) and Computerized Tomography (CT)-PET studies of control and decorin-treated animals were performed using the MOSAIC PET scanner (Philips Medical Systems). The scanner used 2 ϫ 2 x 10-mm 3 gadolinium oxy-orthosilicate crystals coupled to 19-mm diameter photomultiplier tubes via a continuous slotted light guide. The detectors were arranged to produce a transaxial field of view of 128 mm and an axial field of view of 120 mm. The absolute coincidence sensitivity was 1.3% for a point source, and the transverse resolution was 2.2 mm at full-width-half-maximum. Images were reconstructed into 0.5 ϫ 0.5 ϫ 0.5-mm 3 voxels using a three-dimensional RAMLA algorithm supplied with the camera. CT studies were performed using the MicroCATII CT scanner (Imtek Inc.). A 70 ϫ 100-mm phosphor screen was optically coupled to a 2048 ϫ 3072-pixel CCD camera. The x-ray source and detector were rotated around the subject to produce a transaxial field of view of 51.2 mm and an axial field of view of 76.8 mm. X-rays were generated at 80 kV peak and 500 A. Images were reconstructed into 0.2 ϫ 0.2 ϫ 0.2-mm 3 voxels using a Feldkamp cone beam reconstruction algorithm. To perform PET and CT imaging, mice were injected with 0.4 -0.5 mCi of [ 18 F]flourodeoxyglucose ( 18 FDG) and allowed 2 h for tracer distribution. Just prior to imaging, mice were anesthetized with an injection of Ketamine, Xylazine, Acetopromazine (200, 10, 2 mg kg Ϫ1 ) via an intraperitoneal injection and placed in a 50-ml tube to facilitate multimodality stereo tactic positioning. PET data were acquired in a single position for 15 min followed by CT data acquisition for 5 min. The images were registered with an internally developed automated mutual information rigid registration algorithm. Volumes of interest were defined by drawing multislice regions of interest on the PET images using 50% of the full-width-half-maximum of the tumor to determine the tumor boundary. PET regions were also defined on contralateral soft tissues and compared with the CT images where necessary. The images were normalized on the average uptake of contralateral abdominal regions.
Immunofluorescence Microscopy and Quantification-Frozen sections were dried for 1 h and fixed in ice-cold acetone for 5 min. After washing, the sections were blocked for 18 h with 5% (w/v) bovine serum albumin/phosphate-buffered saline at 4°C and subsequently subjected to standard immunofluorescence protocols with various antibodies, costained with 4Ј,6-diamidino-2-phenylindole, and mounted with Vectashield medium (Vector Laboratories, Inc., Burlingame, CA). Images were acquired using an Olympus BX51 microscope driven by SPOT advanced version 4.0.9 imaging software (Diagnostic Instruments, Inc.). To quantify the fluorescence of EGFR in sections of tumor xenografts, control sections were analyzed at different exposure times and gain settings and converted into grayscale (37). The distribution of the pixel intensity was studied using the histogram function of Adobe Photoshop 7.0 (Adobe Systems Inc., San Jose, CA). The final adjustment was an exposure time of 50 ms, and a camera gain of 8, which records images not saturated in fluorescence intensity. Digital images for control (n ϭ 11) and decorin-treated (n ϭ 8) tumors were acquired. For fluorescence visualization, RGD color images were converted to 8-bit grayscale images, and three-dimensional surface plot analyses were generated with the Surface Plot function of ImageJ 1.34 (http://rsb.info.nih.gov/ij) to show the intensity of the representative fluorescent signals (38).
Determination of Apoptosis and Mitotic Index in Tumor Xenografts-To analyze apoptosis we used two approaches; the TUNEL assay (BD Bioscience), which labels internucleosomal DNA fragmentation and the detection of the protein active caspase-3 on frozen sections. Fluorescence signal of TUNEL staining and active caspase-3 was detected and quantified as described above. To determine the proliferative (mitotic) index, frozen sections of tumor xenografts were stained for the proliferation-associated marker Ki67 and by collecting the total pixel density of 41 individual (200ϫ) fields (20 for control and 21 for decorin-treated tumors) from 8 mice.
Pulse-Chase and Dose-Response Experiments-For the pulsechase experiment, confluent A431 cells were serum-starved for 18 h, pulsed with 30 g ml Ϫ1 decorin protein core for 30 min at 37°C, washed on ice, and then chased with serum-free medium for various time points (0.5, 1, 2, 3, 4, and 6 h) at 37°C. For the dose-response experiment, confluent A431 cells were serumstarved for 18 h and pulsed with various concentrations (0, 1, 5, 10, and 30 g ml Ϫ1 ) of decorin protein core. After 30 min, cells were washed with ice-cold phosphate-buffered saline and chased for 120 min with serum-free medium. Cells were harvested, and lysates were subjected to immunoblotting using specific antibodies described above. Cells were analyzed for EGFR, ␤-actin, and PY20-horseradish peroxidase total phosphorylation. Several x-ray films were analyzed to determine the linear range of the chemiluminescence signals and were subsequently quantified with Scion Image alpha 4.0.3.2 and normalized on ␤-actin.
DNA Fragmentation Analysis by FACS and Active Caspase-3-Approximately 0.4 ϫ 10 6 A431 cells were seeded in 6-cm dishes, cultured under standard conditions overnight, and then treated for 24 h with different concentrations of decorin protein core (0.1 ng to 90 g ml Ϫ1 ) and 200 ng ml Ϫ1 EGF in medium containing 10% fetal bovine serum. As positive controls, cells were treated with etoposide (5 and 10 M), a topoisomerase-II inhibitor and an established cell cycle-specific DNA-damaging agent (39). Prior to DNA fragmentation analysis, the cells were trypsinized and fixed with 80% ethanol at 4°C for 1 h, washed twice with phosphate-buffered saline, and resuspended in 50 g ml Ϫ1 propidium iodide solution containing 0.5 g ml Ϫ1 RNase A (40). Cells were stained for 3 h at 4°C and DNA fragmentation was analyzed by flow cytometry using an Epics XL-MCL (Beckman Coulter). To evaluate if the kinase activity of EGFR was required for apoptosis, cells were treated with 30 g ml Ϫ1 decorin protein core and 1 M AG1478 and processed as described above. To corroborate the results obtained for A431 cells, HeLa cells were also used under the same conditions. Active caspase-3 activity was measured with the Caspase-Glo TM 3/7 kit (Promega, Madison, WI), a luminescence assay that measures caspase-3 and -7 activities. It is a mixture of a luminogenic substrate that contains the tetrapeptide sequence DEVD, in a reagent optimized for caspase activity, luciferase activity, and cell lysis. Following caspase cleavage, a substrate for luciferase is released, resulting in the luciferase reaction and the production of light that is proportional to the amount of caspase activity present. Approximately 5 ϫ 10 3 cells were cultured in 96-well plates for 18 h, following a 24-h dose-dependent treatment with decorin protein core (0.1, 1, 5, 10, 30, and 90 g ml Ϫ1 ) or 200 ng ml Ϫ1 EGF in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Both fragmentation and caspase-3 activation experiments were also performed in the presence of the pan-caspase inhibitor VAD (Biovison, Mountain View, CA), which preferentially blocks caspases-1-3 (41). Prior to each analysis, the culture medium was supplemented with 50 l of substrate solution and incubated for 1 h at 25°C. Luminescence measurements were carried out with a microplate reader 1420 Victor 3 (PerkinElmer Life Sciences). Results are given as means (n ϭ 3-5) with three independent measurements for each group. Statistical evaluation was done with an unpaired Student's t test using the Sigma Plot 9 statistical package. p Ͻ 0.05 was considered as significant.

Growth Inhibition of Tumor Xenografts by Systemic Delivery of Human Recombinant Decorin Protein
Core-Decorin protein core was purified to homogeneity from the secretions of human embryonic kidney 293 cells and migrated as a monomer of 46 -48 kDa on a silver-stained SDS gel (42). Before injecting into the animals, each batch of filtered decorin protein core was tested for biological activity by determining the suppression of basal EGFR phosphorylation in quiescent A431 cells (37). We evaluated the effects of systemic delivery of recombinant decorin protein core on the growth kinetics of established orthotopic A431 squamous cell carcinoma xenografts. The results showed that at low dosages of decorin protein core (4.5 mg kg Ϫ1 ), there was a significant ( p ϭ 0.025) growth inhibition of the tumor xenografts in two independent experiments (Fig.  1A). We performed a total of four in vivo experiments utilizing 40 animals, of which four were excluded because there was no tumor engraftment. The total amount of decorin protein core injected in each mouse was 540 g. The basis for the size difference at day 19 was primarily due to a growth rate advantage, since doubling time for the vehicle-treated tumors was significantly shorter than that of the decorin-treated animals (2.5 versus 3.75 days, respectively) (Fig. 1B). Similar growth kinetic was observed with another independent experiment using the same regimen with the exception of lower (2.5 mg kg Ϫ1 ) decorin protein core dosages (data not shown). Interestingly, when the dosage was increased to 10 mg kg Ϫ1 , there was a greater growth inhibition ( p Ͻ 0.001) that lasted for up to 38 days (Fig. 1C). In this case, a total amount of 3.8 mg of decorin protein core was injected in each mouse. While in the first 19 days of treatment the tumor doubling times were similar to those obtained with lower decorin dosages, the doubling time of the decorin-treated tumors was much greater (9 days versus 2.5 days) at a later time. Thus, these results indicate that systemic delivery of decorin protein core retards the growth of established orthotopic tumor xenografts in a dose-dependent manner.
Next, we determined whether decorin could inhibit in vivo tumor growth by affecting tumor metabolism. To this end, several animals from separate experiments were analyzed by CT and PET scan. This strategy allows for direct visualization and quantification of tissue metabolic activity via the administration of a radioactive sugar, 18 FDG, which proportionally distributes to metabolically active tissue. Animals were imaged toward the end of each experimental protocol to allow for maximal tumor visibility in control and treated groups. The results showed a marked inhibition of 18 FDG uptake in the decorin-treated tumor xenografts as compared with controls (Fig. 1D). PET scans were carried out using 200 -450 Ci of 18 FDG at different time points (16,18,34, and 36 days) including animals receiving both low and high dosages of decorin protein core, and in all cases there was a significant reduction in metabolic activity. Quantification of a mixture of animals (n ϭ 10) by normalizing the maximal signal in each tumor to that in the FIGURE 1. Intraperitoneal injections of decorin protein core suppress growth of established orthotopic tumor xenografts. A, growth kinetic of the tumors treated with (4.5 mg kg Ϫ1 ) decorin protein core starting at day 7. Approximately 1 ϫ 10 6 A431 cells were injected subcutaneously into the dorsal flank of nu/nu mice. Tumor xenografts were treated by intraperitoneal injection with vehicle (control) or decorin protein core every other day. After 18 days a significant growth inhibition was observed ( p ϭ 0.025, n ϭ 8 each). The data derive from two independent experiments. B, semilogarythmic plot of the data presented in A shows that the doubling time of the tumors is increased by decorin treatment (3.75 versus 2.5 days in control). C, in vivo experiment where tumor xenografts were treated with 10 mg kg Ϫ1 decorin protein core starting at day 4 for 38 days (n ϭ 5 each). D, longitudinal and sagittal views of tumor xenografts from a control and a decorin-treated mouse obtained by PET scan following injection of 18 FDG. Arrows indicate tumor location in the animal. Control tumors show a higher uptake of 18 FDG as compared with the decorin-treated xenografts. Location and dimension of the tumor xenografts were also established by CT scan (data not shown). E, quantification of the 18 FDG uptake shows a significant decrease evoked by decorin treatment ( p ϭ 0.016, unpaired Student's t test). The values represent the mean Ϯ S.E. abdomen of each animal showed a significantly reduced ( p ϭ 0.016) tumor metabolic rate (Fig. 1E). These findings were not due to differences in tumor "size" since we found a significant decrease in tumor metabolism even when comparing similar size tumors. Tumor identification and volumes were verified by concurrent CT scanning, and these values supported the results obtained by manual measurements (data not shown).
Therefore, PET scan image analysis indicates that tumor growth inhibition evoked by decorin treatment is at least in part due to reduced tumor metabolic rate.
Decorin Targets the Cancer Cells within the Orthotopic Tumor Xenografts-The central hypothesis of our research is that the functional receptor for decorin, i.e. the EGFR, needs to be present within a tumor cell population for decorin binding and suppression of the known oncogenic activity of the receptor. Having established that systemic delivery of decorin retards in vivo tumor growth, we investigated whether decorin would specifically target the EGFR-overproducing tumor cells. Using fluorescence microscopy and an anti-His antibody, which recognizes the N-terminal His tag on decorin, we discovered that decorin protein core specifically targets the tumor cells (Fig. 2). Specifically, decorin epitopes could be detected in the tumor cells proximal to the blood vessels (Fig. 2, B-D), with intervening areas lacking any reactivity. Decorin epitopes were patchy and primarily associated with the cell surface of the A431 squamous carcinoma cells (Fig. 2D), with a distribution similar to the EGFR. The gradient of the fluorescence signal obtained from staining with anti-His suggests that decorin diffuses through the vascular beds and specifically targets the tumor cells. In contrast, very little or no decorin was associated with normal organs, such as spleen and liver (see supplemental Fig. 1). Because decorin has been previously reported to inhibit in vitro angiogenesis by either interacting with thrombospondin-1 (43) or by reducing the endogenous tumor levels of VEGF (44), which is also achieved by neutralizing antibodies against the EGFR (45), we quantitatively analyzed the blood vessel density in control and decorin-treated tumor xenografts utilizing an antibody directed toward the C terminus of perlecan (35). The tumor blood vessels were specifically labeled by this antibody which recognizes both the human and mouse perlecan (Fig. 2,  A-D). It has been shown that the human-derived perlecan is incorporated into the newly formed basement membrane (46) and that in other tumor xenografts generated by implanting mouse RT101 tumor cells into rats, the newly formed basement membranes are hybrid structures comprising both rat and mouse perlecan (47). Quantification of the blood vessels showed that decorin treatment slightly reduced the tumor microvascular density, but the changes were not statistically significant (Fig. 2E, p ϭ 0.137, n ϭ 52 for the control and n ϭ 60 for the decorin-treated samples).
Collectively, these results indicate that decorin specifically targets the EGFR-overexpressing tumor cells in vivo, presumably through the leaky tumor endothelium, and that the growth-suppressive properties of decorin are not mediated by an indirect effect on angiogenesis. Systemically injected human decorin protein core targets the tumor xenografts. A and B, immunofluorescence views of a control and decorin-treated tumor following staining for blood vessels (perlecan, red) and His 6 tag (decorin, green). C and D, higher magnification views of different tumors showing the cell surface localization of decorin, predominantly on the tumor cell islands surrounding blood vessels. Lu ϭ lumen of blood vessels. Bar ϭ 100 m. E, quantification of fluorescence signals obtained from the blood vessel staining for perlecan. There is a minor reduction in blood vessels in the decorin-treated tumors; however, this reduction is not significant ( p ϭ 0.137, n ϭ 52). The values represent the mean Ϯ S.E.

Decorin Treatment Reduces Total EGFR and Attenuates EGFR Activity in Tumor Xenografts-We have previously
shown that tumor xenografts generated by co-injection of wildtype A431 cells and A431 cells genetically engineered to express decorin proteoglycan, grew at much lower rates than the tumor xenografts generated by wild-type A431 cells, and that the growth inhibition was directly proportional to the number of co-injected decorin-expressing cells (12). These findings suggest that the decorin-secreting cells would inhibit the growth of EGFR-overexpressing cells in both a paracrine and autocrine fashion. Thus, we hypothesized that systemic delivery of decorin protein core would also be able to affect in vivo EGFR function, perhaps by decreasing total receptor levels, thereby accounting for the reduced tumor growth. We utilized a mouse monoclonal antibody that specifically recognizes the humanderived EGFR without cross-reacting with the endogenous murine EGFR (Fig. 3A). To allow quantitative comparison of EGFR levels between control and decorin-treated tumors, immunofluorescence images of either sample were captured using a constant 50-ms exposure, a strategy that allows visualization of differences in fluorescence intensity using three-dimensional surface plots (38). The EGFR fluorescence intensity could be quantified, insofar as the section thickness, magnification, and microscopic filters were also held constant. Quantita- tive analysis of EGFR-stained 5-m tumor sections showed that systemically delivered decorin protein core caused a significant reduction ( p ϭ 0.014, n ϭ 46) in the amount of EGFR on the cell surface (Fig. 3, B-D). These findings were corroborated by a ϳ50% decline in total EGFR in the decorin-treated tumors ( p ϭ 0.002, Fig. 3, E and F). Notably, chronic decorin treatment did not affect EGFR levels in the liver (see supplemental Fig. 2), an organ known to have express relatively high levels of EGFR. Moreover, we found that there was a concurrent decrease in tyrosine phosphorylation (activation) in the treated tumors ( p ϭ 0.006, Fig. 3, E and G), as measured by using either a total phospho-Tyr antibody or a specific antibody to EGFR phospho-Tyr 1068 , an established autophosphorylation site (48,49). The fact that decorin protein core negatively regulates EGFR Tyr 1068 phosphorylation suggests that decorin would also interfere with the recruitment of the adaptor protein Grb2 which specifically binds to phospho-Tyr 1068 , thereby preventing the Ras-dependent mitogenic signal cascade. Importantly, Grb2 is increasingly drawing attention as a molecular target for anti-cancer therapy (50).
Taken together, these results demonstrate that systemic delivery of decorin protein core leads to a significant down-regulation of total EGFR and a concurrent attenuation of basal as well as EGF-induced receptor activity, specifically the phosphorylation of Tyr 1068 in the tumor xenografts. It is important to point out that the difference in EGFR phosphorylation is independent from changes in total EGFR, insofar as the data from control and treated samples were normalized on the respective total EGFR amounts.
Decorin Treatment Enhances Apoptosis and Reduces the Mitotic Index in Tumor Xenografts-Next, we determined whether the growth suppression observed in the tumor xenografts treated with decorin would be associated with an enhancement in programmed cell death. We utilized two approaches: TUNEL to detect fragmented DNA and an antibody that specifically recognized the active form of caspase-3, a key protease involved in apoptosis (51). We discovered that decorin treatment caused a significant enhancement of apoptosis (Fig.  4, A-F). Specifically, decorin treatment increased TUNEL staining by ϳ3- fold ( p ϭ 0.04, Fig. 4C) and active caspase-3 by ϳ3.6-fold ( p ϭ 0.002, Fig. 4F). To test whether A431 tumor xenograft growth was affected by decorin, we stained for proliferation-associated marker Ki67, a nuclear antigen uniquely expressed in all the phases of the cell cycle but absent in resting (G 0 ) cells (52). Control tumors exhibited intense labeling for Ki67 (Fig. 4G). In contrast, only scattered and faintly Ki67-positive cells were detected in decorin-treated tumor xenografts (Fig. 4H). Quantification of Ki67-positive cells in vehicle-and decorin-treated tumor xenografts showed that decorin caused a 68% inhibition of mitotic index ( p ϭ 0.005, Fig. 4I).
Collectively, these results indicate that decorin exerts synergistic effects on the tumor xenografts by inducing apoptosis and suppressing tumor cell division, two events that have been associated with the suppression of EGFR activity (4). Furthermore, these data support the results obtained by PET scan.
Decorin Is Internalized by A431 Cells and Causes a Protracted Down-regulation of the EGFR-To further investigate the mechanism of action of decorin, we exposed quiescent (serum-starved) A431 cells to decorin protein core for 30 min and chased for various periods of time (0.5-6 h). The results showed that decorin was rapidly internalized and degraded (T1 ⁄ 2 ϳ 20 min), but a significant (ϳ20%) proportion of intact decorin protein core remained associated within intracellular compartments for up to 6 h (Fig. 5, A and B). Notice that no intermediary species of decorin were detected during the chase time, suggesting that the majority (ϳ80%) of decorin is rapidly degraded, once internalized. No detectable decorin protein core was found in the chase media (data not shown). These biochemical results are in agreement with our recent discovery that decorin is internalized primarily via caveolar-mediated endocytosis and that it remains within perinuclear punctuate vesicles without being recycled to the cell surface (37). We estimated that the amount of cell-bound decorin protein core was 160 ng 10 Ϫ6 cells, which is equivalent to ϳ2.1 ϫ 10 12 molecules of decorin protein core. Notably, there are ϳ2.5 ϫ 10 12 EGFR molecules 10 Ϫ6 A431 cells (12), indicating a nearly 1:1 decorin/ EGFR ratio. Because we found that most of the cell surface EGFR follows the decorin-induced pathway of internalization (37), we quantified the amount of total EGFR and normalized its content on ␤-actin. The total EGFR was rapidly downregulated with kinetics similar to those observed for the internalized decorin protein core (T1 ⁄ 2 ϳ 18 min, Fig. 5C). However, once the decorin core was removed, there was a slow but gradual recovery of the total EGFR. By 2 h, the total EGFR levels were similar to control values and remained unchanged for the rest of the chase. These data suggest that a short exposure to decorin causes a reversible down-regulation of the EGFR with a slow recovery phase of ϳ2 h.
Next, we determined the kinetics of EGFR activation during the continuous presence of decorin. Decorin caused a transient activation of the EGFR which peaked at ϳ1 h, but subsequently the phosphorylation of the EGFR declined to reach base-line levels by 4 h and then remained suppressed for up to 8 h (Fig. 5, D and E). Interestingly, the total EGFR protein was reduced upon transient exposure to decorin and reached ϳ20% of control levels at 1 h (Fig. 5, D and F); however, it subsequently increased without any further activation, even in the constant presence of decorin. The EGFR antibody used is able to recognize activated forms of the receptor, ruling out a possible artifact. Most importantly, the EGFR levels were reduced by ϳ40% and reached a plateau at ϳ4 h of continuous exposure to recombinant decorin protein core. These data are in agreement with our previous studies in which we have shown that several independent clones of A431 cells stably expressing decorin contain ϳ40% less EGFR than wild-type cells and that the number of receptors, as determined by 125 I-labeled EGF and Scatchard analysis, was also reduced by FIGURE 5. Decorin is internalized by A431 cells and causes a protracted down-regulation of the EGFR. A, immunoblotting of total cell lysates from a time-course experiment of decorin protein core uptake by A431 cells. Quiescent, serum-starved A431 cells were exposed to decorin protein core (30 g ml Ϫ1 ) for 30 min and then chased in serum-free media for the designated time points. The blot was reacted with a polyclonal goat anti-decorin antibody that recognizes the entire protein core. B and C, quantification of intact decorin protein core and total EGFR, respectively, in pulse-chase experiments similar to that shown in A. The values are the mean Ϯ S.E. (n ϭ 3). D, effects of continuous exposure (0.25-8 h) to exogenous decorin on the EGFR phosphorylation and amount in A431 cells. Immunoblotting of total cell lysates probed with either PY20 (Tyr(P)) or anti-EGFR). E and F, quantification of activated (phosphorylated) EGFR and total EGFR, respectively, upon continuous exposure to decorin protein core. The values are the mean Ϯ S.E. (n ϭ 3).
40% without significant changes in affinity for EGF (12). Another important aspect is that decorin can prevent EGFR dimerization induced by EGF (37). Dose-response experiments with A431 cells and decorin protein core showed that even a low dosage of 0.1 g ml Ϫ1 decorin protein core significantly reduced the amount of EGFR (data not shown). Thus, decorin reduces the amount of EGFR in A431 cells to an extent that is very similar to that observed in the tumor xenografts in vivo and these effects are protracted. This attenuation of EGFR quantity and activity provides a mechanistic explanation for the in vivo growth suppressive properties of recombinant decorin protein core after systemic application.
Decorin Induces Apoptosis-Our in vivo results showed that decorin treatment caused an increase in apoptosis by activation of caspase-3. Therefore, we wanted to elucidate this mechanism of action by studying the in vitro effects of decorin protein core on A431 cells by FACS (DNA fragmentation) analysis following propidium iodide staining. Determination of the frac-tion of cells with hypodiploid (sub-G 0 ) DNA content, reflecting the nuclear changes seen in apoptosis, showed a significant increase in apoptosis evoked by decorin (Fig.  6A, p Ͻ 0.001, n ϭ 5). As a positive control we utilized 5 and 10 M etoposide for 24 h in full serum (Fig.  6B). Significant increase in DNA fragmentation was observed even at low decorin concentration, between 1 and 10 ng ml Ϫ1 . Notably, the induction of apoptosis was dose-dependent and reached a plateau at ϳ100 ng ml Ϫ1 (Fig. 6A). The effect of decorin protein core at low concentrations is important because these nanomolar levels could be easily reached in vivo after systemic administration. Moreover, the dose dependence of the effect on apoptosis by decorin supports the idea of a specific mechanism of action. These results make unlikely the possibility that our highly purified preparations of decorin protein core are contaminated with an unknown toxin. In fact, if this were the case, we would have expected to observe a dilution effect insofar as the decorin preparation was diluted ϳ900,000-fold (from 90 g ml Ϫ1 to 0.1 ng ml Ϫ1 ).
To further assess the specificity of decorin-induced DNA fragmentation, A431 cells were treated with either decorin (30 g ml Ϫ1 ) or EGF (200 ng ml Ϫ1 ) in the presence or absence of 1 M AG1478, an established inhibitor of the EGFR tyrosine kinase (53). We have previously shown that at this low concentration, AG1478 is highly specific for the EGFR kinase in A431 cells, whereas the PDGF receptor activity is not affected (9), and does not inhibit the uptake of decorin (37). The results showed that treatment with decorin or EGF induced DNA fragmentation to a similar extent, whereas AG1478 did not (Fig.  6C). Notably, concurrent addition of decorin protein core and AG1478, or EGF and AG1478, resulted in a significant inhibition in DNA fragmentation caused by either decorin or EGF alone (Fig. 6C). In contrast, AG1478 was not capable of blocking the etoposide-induced effects but rather showed a synergistic effect. We conclude that decorin induces apoptosis primarily via the EGFR. Similar experiments using HeLa cells showed also a marked induction of apoptosis in contrast to fibroblasts which were not reactive (data not shown). This is consistent with the observation that exogenous decorin is protective against apoptosis of mouse fibroblasts cultured in a three-dimensional collagen lattice (54). . Decorin protein core induces apoptosis in A431 cells. A and B, subconfluent A431 cells were exposed to different concentrations of decorin or etoposide as indicated for 24 h. Decorin protein core concentration is presented in logarithmic scale. Note that control value (i.e. no decorin added) was 0.41% (data not shown). FACS analysis was used to visualize DNA fragmentation. Values represent the percent of total cell number, mean Ϯ S.E. ( p Ͻ 0.001 for all values, unpaired Student's t test, n ϭ 5). C, subconfluent A431 cells were exposed to vehicle (control), etoposide (10 M), decorin protein core (30 g ml Ϫ1 ), EGF (200 ng ml Ϫ1 ), and EGFR kinase inhibitor AG1478 (1 M) either alone or together with decorin or EGF. Values represent the mean Ϯ S.E. (n ϭ 5).
Previous studies have shown that relatively high dosages of EGF inhibit A431 growth and induce apoptosis (55,56). Furthermore, it has been shown that EGF induces growth arrest (57) and apoptosis in A431 cells by increasing STAT1 and by up-regulating p21 WAF1 (58) and also by inducing persistent activation of p38 mitogen-activated protein kinase (59). Notably, decorin causes a protracted induction of p21 WAF1 and growth arrest (8,14), which is in line with the data presented above.
Decorin Induces Apoptosis via Activation of Caspase-3-Next, we determined whether decorin could induce the activation of caspase-3, a key enzyme in apoptosis (41,51). A431 cells were treated with decorin protein core or EGF for 24 h and then incubated for 5 min with a luminescent caspase-3-sensitive substrate. The results showed that a significant activation of caspase-3 was induced by recombinant decorin protein core or EGF ( p Ͻ 0.001 for both conditions, Fig. 7A). The induction of caspase-3 activity was completely blocked by the pan-caspase inhibitor VAD (Fig. 7A), which preferentially blocks caspases-1-3 (41). To further prove that caspase-3 was directly involved in the response to decorin, A431 cells were treated with decorin or EGF as above in the presence or absence of VAD and subjected to DNA fragmentation analysis. The results showed that apoptosis was similarly blocked by the pan-caspase inhibitor VAD (Fig. 7B).
These results indicate that both decorin and EGF activate caspase-3 and that this is the main event leading to DNA fragmentation.
Recent studies have shown that the EGFR is a substrate for active caspase-3 and that this mechanism is independent of internalization (60). The highly conserved C terminus (intracellular) domain of the EGFR is specifically cleaved by activated caspase-3, and this could be an additional way to shut down EGFR signaling during apoptosis (60). Since decorin treatment increases active caspase-3 both in vivo and in vitro, it is intriguing to hypothesize that decorin could similarly induce further degradation of the EGFR via caspase-3 activation, leading to inhibition of receptor signaling, in addition to its ability to induce EGFR internalization and degradation via caveolar-mediated endocytosis (37). This matter will be subjected to further investigation.
Conclusions-The results of our study show for the first time that systemic delivery of decorin protein core can suppress in vivo tumorigenicity by specifically targeting EGFR-overexpressing tumor cells. We show that decorin protein core blocks the EGFR pathway by inhibiting the phosphorylation at Tyr 1068 . Tyr 1068 is a primary autophosphorylation site induced by EGF and leads to activation of the mitogen-activated protein kinase pathway and clathrin-dependent endocytosis. We therefore hypothesize that one of the main in vivo actions of decorin is to "direct" the EGFR to distinct intracellular compartments via a caveolar-mediated endocytosis, destined for degradation as we have recently demonstrated to occur in vitro in A431 (37). How the dynamics of lipid-rafts, caveolin, and decorin might affect EGFR trafficking and cellular responsiveness will be an interesting topic for future investigations.
We propose a working model that could explain the basic mechanism of decorin in vivo action. This model involves an initial interaction of this protein with its functional cell surface receptor, the EGFR. Circulating decorin, following its absorption from the peritoneal cavity, would presumably reach the tumor cells via the leaky tumor microvasculature. This leads to EGFR down-regulation and consequent attenuation of the EGFR signaling cascade, as shown by quantitative decline in total EGFR and reduced activation of the receptor. As a consequence of this, the tumor xenografts grow slower, as shown by increased doubling time, and exhibit an enhanced apoptotic rate and a reduced mitotic index. These results can mechanistically explain the reduced metabolism of the tumor xenografts that is independent of tumor size but dependent on decorin treatment. Furthermore, decorin treatment induces apoptosis in a dose-dependent manner at nanomolar concentration by activating caspase-3. Notably, decorin-induced apoptosis can be blocked by AG1478, a specific EGFR inhibitor. Collectively, our data open the possibility of an efficient protein therapy for various forms of human cancers where EGFR plays a key pathophysiological role. . Decorin protein core induces active caspase-3 in A431 cells. A, subconfluent A431 cells were exposed to decorin protein core (30 g ml Ϫ1 ), EGF (200 ng ml Ϫ1 ), and the pan-caspase inhibitor VAD (20 M) either alone or together with decorin or EGF, for 24 h in 10% fetal bovine serum. The values represent the mean luminescence expressed as relative light units (RLU), derived from the caspase-3-mediated cleavage of the luminescent cell-permeable substrate. Note that active caspase-3 is equally increased by both decorin and EGF ( p Ͻ 0.001) and completely blocked by VAD ( p Ͻ 0.001). Mean Ϯ S.E. (n ϭ 5). B, cells were exposed to the same conditions as in A, with the exception that the cells were processed for FACS analysis as detailed in Fig. 6. Note that DNA fragmentation (apoptosis) evoked by either decorin or EGF is completely blocked by VAD ( p Ͻ 0.001). Mean Ϯ S.E. (n ϭ 5).