|Home | About | Journals | Submit | Contact Us | Français|
The cell-surface receptor protein tyrosine phosphatase mu (PTPμ) is a homophilic cell adhesion molecule expressed in CNS neurons and glia. Glioblastomas (GBMs) are the highest grade of primary brain tumors with astrocytic similarity and are characterized by marked dispersal of tumor cells. PTPμ expression was examined in human GBM, low-grade astrocytoma, and normal brain tissue. These studies revealed a striking loss of PTPμ protein expression in highly dispersive GBMs compared to less dispersive low-grade astrocytomas and normal brain. We hypothesized that PTPμ contributes to contact inhibition of glial cell migration by transducing signals in response to cell adhesion. Therefore, loss of PTPμ may contribute to the extensive dispersal of GBMs. The migration of brain tumor cells was assessed in vitro using a scratch wound assay. Parental U-87 MG cells express PTPμ and exhibited limited migration. However, short-hairpin RNA (shRNA)-mediated knockdown of PTPμ induced a morphological change and increased migration. Next, a brain slice assay replicating the three-dimensional environment of the brain was used. To assess migration, labeled U-87 MG glioma cells were injected into adult rat brain slices, and their movement was followed over time. Parental U-87 MG cells demonstrated limited dispersal in this assay. However, PTPμ shRNA induced migration and dispersal of U-87 MG cells in the brain slice. Finally, in a mouse xenograft model of intracranially injected U-87 MG cells, PTPμ shRNA induced morphological heterogeneity in these xenografts. Together, these data suggest that loss of PTPμ in human GBMs contributes to tumor cell migration and dispersal, implicating loss of PTPμ in glioma progression.
Glioma tumors of the CNS represent a significant health concern, and their poor prognosis is due in part to a lack of effective screening and therapies. While the cellular origins of gliomas are not well understood, it is likely that the majority of brain tumors are derived from glial cells or their precursors that are retained in the adult CNS.1 Based on histological characteristics, gliomas can be classified into astrocytomas or oligodendrogliomas and assigned a histological grade (I–IV) according to the presence of features of malignancy.2,3 Grade II–IV astrocytomas generally are all considered malignant, but there are distinct differences in the prognosis, malignancy, and diagnosis of each histological subtype. The range of mean survival times is from 5 to 15 years for grade II low-grade astrocytomas to less than 1 year for grade IV glioblastoma multiforme (GBM) tumors.1 This discrepancy in prognosis is attributed to the extremely malignant character of GBMs. GBMs are significantly more proliferative, angiogenic, necrotic, and dispersive than are low-grade astrocytomas.4,5 This is in part evidenced by the diagnostic patterns seen in conventional imaging modalities. GBMs are typified by a mass effect that is enhanced in a T1-gadolinium MRI study. Low-grade gliomas, however, typically are nonenhancing and require T2 or fluid-attenuated inversion recovery signals on MRI for identification.6
Primary brain tumors rarely metastasize to other organs.4 Instead, malignant astrocytomas infiltrate the surrounding neuropil. Low-grade astrocytomas are locally infiltrative. This contrasts with the extensive infiltration of GBM tumors, where cells disperse to distal sites throughout the entire brain parenchyma.7,8 How this dispersal is regulated is unclear but is likely to depend on interactions between the tumor cells and the microenvironment of the brain. Identification of key regulatory signals that control cellular communication and migration will enhance our understanding of glioma dispersal and may identify new therapeutic targets for treating brain tumors.
Tyrosine phosphorylation controls cellular migration and functions as a molecular switch, turning on or off various protein–protein interactions that are controlled by protein tyrosine kinases and protein tyrosine phosphatases (PTPs). Upregulation and amplification of receptor tyrosine kinases (RTKs) are common features in GBM tumors.4,5 For example, the epidermal growth factor receptor (EGFR), a member of the RTK superfamily, is amplified in 40% of all GBMs, and a constitutively active mutant form of EGFR occurs in 20%–30% of GBMs, contributing to an aberrant cellular tyrosine phosphorylation status.9,10
The role of tyrosine phosphatases in GBM, however, has not been well studied. Receptor-like PTPs (RPTPs) have the unique ability to sense a signal in their extra-cellular environment via their extracellular segment and to transduce this signal intracellularly using their phosphatase activity.11 Some RPTPs have homology to the neural cell adhesion molecule N-CAM, which contains multiple immunoglobulin (Ig) domains and fibronectin type-III (FNIII) repeats, and function as homophilic cell–cell adhesion molecules.12 One such RPTP, PTPμ, is a member of the PTPμ-like subfamily, which includes PTPμ, PTPκ, PTPρ, and PCP-2.12 The extracellular segment of the PTPμ-like RPTPs contains motifs found in cell adhesion molecules, including a meprin/A5-protein/PTPμ (MAM) domain, an Ig domain, and four FNIII repeats. PTPμ, PTPκ, PTPρ, and PCP-2 (also called PTPλ) mediate adhesion via homophilic binding.13–17 These events occur when PTPμ on the extracellular surface of one cell binds to PTPμ on the surface of an adjacent cell, forming an adhesive contact. The intracellular segment of PTPμ contains a juxtamembrane sequence with homology to cadherins and two conserved phosphatase domains, of which only the most membrane proximal is catalytically active.11,12 There is evidence that both cell adhesion and tyrosine phosphorylation play a role in contact inhibition of cell movement.18,19 PTPμ, which mediates cell–cell adhesion, could directly transmit signals generated from cell–cell interactions via changes in tyrosine phosphorylation to regulate contact inhibition of movement.
Here we show the loss of PTPμ protein expression in highly dispersive human GBMs, whereas less dispersive tumors and the human glioma cell line U-87 MG express PTPμ. Short-hairpin RNA (shRNA)-mediated knockdown of PTPμ protein in U-87 MG cells induces a morphological change, migration, and dispersal in scratch wound and adult rat brain slice assays. Furthermore, we used intracranial injection of the U-87 MG cells in an in vivo mouse xenograft model to demonstrate that loss of PTPμ induced a morphological change and tumor cell heterogeneity in the U-87 MG cells in vivo. These data suggest that loss of PTPμ in human GBMs contributes to glial tumor cell migration and dispersal.
Human primary intraparenchymal brain tumor tissue was obtained from surgical resections in accordance with an approved protocol from the University Hospitals Case Medical Center Institutional Review Board. After the neuropathologists made an intraoperative diagnosis of glioma and obtained all the tissue necessary for diagnosis, the remaining tissue was cryopreserved for future studies. Tissue from patients undergoing cortical resections for intractable epilepsy was collected and cryopreserved for use as nontumor “normal” tissue. The U-87 MG and LN-229 human glioma cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). U-373 MG glioma cells were a gift from Dr. Saroj Mathupala (Wayne State University, Detroit, MI, USA). LN-229 and U-87 MG cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (HyClone, Logan, UT, USA) at 37°C, 5% CO2. U-373 MG cells were maintained in RPMI (HyClone) supplemented with 10% fetal bovine serum at 37°C, 5% CO2. Cell lysates were prepared from cell cultures grown to confluence. Normal rat astrocytes were prepared as described previously.20
Normal and tumor tissue lysates were prepared by adding 10 volumes of lysis buffer to tissue samples that had been frozen at the time of surgical resection. The lysis buffer contained 20 mM Tris-HCl (pH 7.0), 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 3 mM benzamidine, 5 μg/ml aprotinin, 5 μg/ml leupeptin, 1 μg/ml pepstatin A, 2 mM sodium orthovanadate, 2 mM EGTA, 5 mM EDTA, 30 mM sodium fluoride, and 40 mM β-glycerol phosphate. The tissue was disrupted using a 2-ml frosted Dounce homogenizer, incubated on ice for 30 min, and then sonicated and centrifuged at 3,000 rpm for 3 min. The supernatant was collected, and the protein concentration was determined using a bicinchoninic acid protein assay (Pierce, Rockford, IL, USA). Sample buffer (2× sodium dodecyl sulfate [SDS]) was added to the lysates. The lysates were then incubated at 95°C for 5 min. Each lane on an SDS-polyacrylamide gel was loaded with 30 μg protein. Lysates from glioma cell lines and cultured astrocytes were prepared similarly. PTPμ expression was analyzed by immunoblot using the intracellular antibodies SK-15 and SK-18 as described previously.13,21 An antibody against actin (JLA20) was obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA, USA). Antibodies recognizing N-cadherin and vinculin were from BD Biosciences (San Jose, CA, USA) and Sigma-Aldrich (St. Louis, MO, USA), respectively. Densitometry analyses were performed using either the Quantity One imaging software (version 4.6.7) of the Fluor-S Max MultiIm-ager system (Bio-Rad, Hercules, CA, USA) or ImageJ software (version 1.40g).22 PTPμ densitometry values were calculated by adding the densities of the full-length and cleaved bands.
Lentiviral shRNA constructs V2LHS_171008 (shRNA 1) and V2LHS_171013 (shRNA 2) targeting human PTPμ mRNA were purchased from Open Biosystems (Huntsville, AL). A control lentiviral shRNA construct was a gift from Drs. E. Johnson and R. Keri. Vesicular stomatitis virus glycoprotein–pseudotyped lentiviral particles were generated by triple transfection of shRNA constructs with the packaging plasmids pCMVΔR8.91 and pMD.G into 293T cells using Lipofectamine 2000 (Invitrogen) according to a previously described protocol.23 The particles were concentrated by ultracentrifugation and used to infect cells in the presence of 6 μg/ml polybrene. Cells were harvested or assayed at 3 days postinfection. Knockdown of PTPμ by lentiviral shRNA was verified by immunoblotting with antibodies to PTPμ. The infection efficiency with lentiviral shRNA constructs containing a green fluorescent protein (GFP) reporter was visualized by fluorescence microscopy.
RNA was isolated from U-87 MG cells expressing either control or PTPμ shRNA using Trizol according to the manufacturer’s protocol (Invitrogen). RNA concentrations from each condition were determined by reading the absorbance at 260 nm using a spectrophotometer and were normalized. Complementary DNA was prepared from 1 μg RNA using a ThermoScript reverse transcriptase (RT)-PCR system from Invitrogen. PCR was performed using the following parameters: 93°C for 30 sec, 58°C for 30 sec, and 72°C for 30 sec for 35 cycles, followed by a 5-min extension at 72°C. Primers designed against the 5' end of the mRNA of PTPμ-like subfamily members were as follows: PTPμ forward, GGGAGGAGGACCCAGGAC; PTPμ reverse, TGTACGTGTTGGGTCTCCAG; PTPρ forward, CCCTCAGCCTGCTCCTGA; PTPρ reverse, CCACCATTCACCTTCACGTA; PTPκ forward, AAACTCGGCATGGATACGAC; PTPκ reverse, GGCATCTCGGGTGGTAGATA; PCP-2 forward, CAGGCCCAGTACGATGACTT; and PCP-2 reverse, AGCTGAACTGCACACAGTGG.
U-87 MG cells were infected with control or PTPμ shRNA lentivirus. After 3 days, confluent monolayers of cells were scratched with a 20-μl pipette tip to induce a wound. The wounded edges were imaged using a Nikon (Tokyo, Japan) TE200 inverted microscope and Spot RT digital camera and image acquisition software (Diagnostic Instruments, Inc., Sterling Heights, MI, USA). Images were collected with a ×10 objective 0, 6, 24, and 48 h after wounding. Images were quantitated using MetaMorph software (version 7.1.7; Molecular Devices, Downingtown, PA, USA) by measuring the width of the wound at each time point. The difference between the width of the wound at the 0 time point and the time point being analyzed was divided by 2. This value reflects the distance migrated by the leading edge of one side of the wound during a given time interval. Replicates were averaged and plotted using Excel (version 12.2.0; Microsoft Inc., Redmond, WA, USA). Error bars indicate standard error. Data were analyzed for statistical significance using an unpaired Student’s t-test. Inhibition studies of the rho family GTPase Rac1 were performed using a Rac1-specific inhibitor from Calbiochem (Gibbstown, NJ, USA).24,25 For morphological studies, the inhibitor was dissolved in water and added at a final concentration of 100 μM just prior to scratching the cells. Cells were analyzed after 6 h. For migration studies, the inhibitor was dissolved in water and added at a final concentration of 50 μM just prior to scratching the cells. Cells were analyzed after 24 h.
U-87 MG cells were infected with control or PTPμ shRNA lentivirus and serum arrested with basal media at 24 h postinfection. Cells were either growth-arrested in serum-free media for 48 h or stimulated after 24 h with complete medium containing 10% serum for an additional 24 h. Propidium iodide labeling of cells for cell cycle analysis by flow cytometry has been described previously.26 Briefly, cells were trypsinized, washed three times with phosphate-buffered saline (PBS), and fixed in ice-cold methanol. Fixed cells were washed twice with PBS, treated with RNase, and labeled with 50 μg/ml propidium iodide. Cell cycle analysis was performed in two separate experiments using a Coulter EPICS XL-MCL flow cytometer (Beckman Coulter Inc., Fullerton, CA, USA) and analyzed using ModFit LT (version 3.1; Verity Software House, Topsham, ME, USA). Replicates were averaged and plotted using Microsoft Excel. Error bars indicate standard error.
All brain slice experiments were performed in accordance with an approved protocol from the Case Western Reserve University Institutional Animal Care and Use Committee. Organotypic brain slice cultures were prepared according to previously described protocols,20,27 with some modifications. Adult female Sprague-Dawley rats 10–12 weeks of age were used for the preparation of brain slices into which human tumor cells were injected. Animals were purchased from Harlan Laboratories (Indianapolis, IN, USA). Whole brains were dissected, embedded in 2% agarose in PBS, and sliced into 300-μm-thick coronal sections using a vibratome (Leica Microsystems, Inc., Bannockburn, IL, USA). Slices were collected in cold calcium/magnesium-free phosphate buffer (CMF) supplemented with 0.5% glucose and placed on membrane culture inserts (Millipore, Billerica, MA, USA) of six-well dishes. Slices were cultured in media containing 50% minimal essential medium, 25% horse serum, and 25% CMF supplemented with 0.5% glucose, 2 mM l-glutamine, and antibiotic/antimycotic solution (Invitrogen) at 37°C, 5% CO2.
U-87 MG cells were injected into prepared brain slices with modifications to a previously described protocol.28 U-87 MG cells were infected with lentiviral shRNA constructs encoding control or PTPμ shRNA containing a GFP reporter. Cells were trypsinized at 3 days post-infection and resuspended in DMEM containing 2 mg/ml Matrigel (BD Biosciences) at a concentration of 105 cells/μl. Cells were injected into the cortex of 1-day-old rat brain slices by creating a small indentation with a 2-μl micropipettor and implanting 0.5 μl of the cell suspension. Slices were incubated for 48 h and then fixed in 4% paraformaldehyde overnight. After washing in PBS, the slices were mounted on slides and then viewed under a fluorescence-equipped microscope to evaluate cell migration and dispersal into the brain tissue via GFP fluorescence. Images were taken with a ×10 objective using a Leica DMI 6000 B automated inverted microscope (Leica Microsystems GmbH, Wetzlar, Germany) attached to a Retiga EXi camera (QImaging, Surrey, BC, Canada).
We developed a quantitative method for determining the migration and dispersal of injected tumor cells in brain slices using MetaMorph software (version 7.5; Molecular Devices). Briefly, the area of the injection site was excluded, and the image was deconvolved to reduce flaring. Next, the area near the injection site was thresholded using parameters that highlighted only cells that had dispersed away from the injection site. Five regions of 200 × 200 μm were selected near the injection site. The number of highlighted pixels in each region was determined and used to calculate the thresholded area in square micrometers. Control and test brain slices were analyzed using identical parameters. The thresholded areas for each image were averaged for each cell type and plotted using Microsoft Excel. Studies were repeated in three separate experiments with at least 11 replicates. Error bars indicate standard error. Data were analyzed for statistical significance using an unpaired Student’s t-test.
Intracranial xenograft experiments were performed in accordance with an approved protocol from the Cleveland Clinic Institutional Animal Care and Use Committee. U-87 MG cells expressing either control or PTPμ shRNA were trypsinized and washed in complete medium followed by a PBS wash. Cells (2.5 × 104) were resuspended in 3 μl PBS to implant intracranially into the right hemisphere of nude mice. A small-animal stereotactic manipulator was used to inject the cell suspension 2 mm lateral, 0.5 mm anterior, and 3 mm inferior to bregma as a reference point. Six animals were injected for each cell type. After 2 weeks, animals were sacrificed. Whole brains were fixed in 10% paraformaldehyde, washed with PBS and PBS with 10% and 20% sucrose, frozen in Tissue-Tek optimal-cutting-temperature compound (Sakura Finetek, Torrance, CA, USA), and cryosectioned into 10-μm sections. Sections were either stained with hematoxylin and eosin or mounted in medium containing diamidinophenylindole (DAPI). Bright-field and fluorescent images were acquired with a ×40 objective using a Leica DM 5000 B automated inverted microscope (Leica Microsystems GmbH) attached to a Retiga-SRV cooled CCD camera (QImaging). Images were collected using ImagePro Plus (version 220.127.116.116; Media Cybernetics, Bethesda, MD, USA) and Turboscan Surveyor (version 18.104.22.168; Objective Imaging, Kansasville, WI, USA) software systems. Confocal fluorescent images were acquired with a ×63 objective using an upright Leica TCS-SP2 spectral laser scanning confocal microscope (Leica Microsystems GmbH) and Volocity 3D imaging software (version 4.3.2; Improvision, Inc., Lexington, MA, USA).
Astrocytomas resemble normal astrocytes histologically.2 Like many other cell types, normal astrocytes exhibit density-dependent contact inhibition of growth.29 PTPμ protein expression was examined in primary astrocytes grown at low and high cell density (Fig. 1). Primary rat astrocytes express both the full-length PTPμ (200 kDa) and proteolytically processed form of PTPμ (100 kDa). Cleavage of PTPμ occurs in the endoplasmic reticulum and results in two noncovalently bound, tightly associated fragments.30 More important, PTPμ protein is stabilized at the plasma membrane in primary astrocytes at high cell density (Fig. 1), consistent with previous data in other cell types.31 Therefore, PTPμ is a good candidate to regulate density-dependent events such as contact inhibition of growth and migration.
In an effort to understand the role of PTPμ in gliomas, we examined the PTPμ protein levels among human glioma tumors compared to control tissue. PTPμ protein expression was strikingly reduced in GBM. For example, five independent primary GBM samples had dramatically reduced levels of PTPμ compared to brain tissue obtained from epilepsy patients (Fig. 2). An additional five GBM samples showed loss of PTPμ expression (data not shown). In contrast to GBMs, low-grade astrocytoma samples maintained some level of PTPμ expression. Next, we assessed the expression of PTPμ protein in the commonly used human glioma cell lines U-87 MG, U-373 MG, and LN-229. U-373 MG and LN-229 cells are migratory in vitro32 and expressed very little PTPμ protein (Fig. 2). U-87 MG cells exhibit some migration in vitro but must be treated to stimulate substantial migration.32 In contrast to U-373 MG and LN-229 cells, U-87 MG cells expressed significant levels of PTPμ. As a control for the expression of another cell surface adhesion molecule, N-cadherin expression was examined in these samples. Somewhat unexpectedly, GBM samples exhibited a variable decrease in the level of N-cadherin expression. Vinculin expression was used as a control for protein loading among the tissue samples. Densitometry analyses of the 200- and 100-kDa bands on the immunoblots indicate that there is a complete reduction of PTPμ expression in GBMs compared to nontumor tissue following normalization for the vinculin loading control. These results suggest that loss of PTPμ protein expression may be an important event in glioma progression.
Because of the significant loss of PTPμ in GBM tumor samples and the dispersive nature of this tumor, we hypothesized that PTPμ suppresses cell migration. To determine whether PTPμ expression affects cell migration directly, shRNA was used to knock down PTPμ in U-87 MG cells, and the migration of these cells was analyzed in vitro using a scratch wound paradigm. U-87 MG cells were infected with lentivirus containing control shRNA or two PTPμ shRNAs targeting different mRNA sequences. Lysates of these cells at 3 days postinfection were analyzed for PTPμ protein levels by immunoblot and showed a 4.5-fold reduction of PTPμ protein expression as determined by densitometry (Fig. 3A). RT-PCR analysis of other PTPμ-like subfamily members (PTPρ, PTPκ, and PCP-2) indicated that only PTPμ mRNA is targeted for degradation by PTPμ shRNA (Fig. 3B). In addition, these data indicate that PTPκ and PCP-2 are expressed in U-87 MG cells (Fig. 3B).
Confluent monolayers of U-87 MG cells expressing either control or PTPμ shRNA were scratched with a pipette tip to induce a wound. The wound edges were monitored over time by microscopy. Control shRNA-infected U-87 MG cells failed to migrate inward to fully close the wound, even after 48 h. In contrast, PTPμ shRNA-infected cells migrated more rapidly and began to close the wound at 24 h (Fig. 4A). The distance migrated by the leading edge of the wound was quantified and shows a statistically significant increase in the distance the U-87 MG cells expressing PTPμ shRNA migrated in comparison to control cells (Fig. 4B). Furthermore, the morphology of cells with PTPμ knocked down appeared to differ from control U-87 MG cells. Loss of PTPμ induced broad lamellipodial structures at the leading edge of the wound, consistent with that of a migratory cell (Fig. 4C). To ensure the morphological change induced by expression of PTPμ shRNA was not due to an off-target effect, another PTPμ shRNA was used to repeat the experiment. Both shRNA constructs induced the same lamellipodial structures (Fig. 5A).
The formation of lamellipodia is dependent upon the activity of the rho family GTPase Rac1.33 Therefore, we hypothesized that the induction of lamellipodia by the loss of PTPμ may be Rac1 dependent. To test this, a specific inhibitor of Rac1 activity24,25 was added to the medium of U-87 MG cells expressing either control or PTPμ shRNA just prior to a scratch assay. The addition of the Rac1 inhibitor to the control U-87 MG cells had no effect on morphology (Fig. 5B). However, the Rac1 inhibitor prevented the induction of lamellipodia in the cells with reduced PTPμ expression (PTPμ shRNA 1), returning the cells to the morphology of the control cells (Fig. 5B). Additionally, the Rac1 inhibitor significantly suppressed the migration induced by the loss of PTPμ (Fig. 5C). Taken together, these data suggest that PTPμ suppresses migration of glioma cells in vitro by inhibition of Rac1 activity.
The results of the scratch wound assay suggest that PTPμ alters the migration of glial tumor cells, but the possibility remained that this effect was due to a change in proliferation. To rule out this hypothesis, the cell cycle profiles of U-87 MG cells expressing control or PTPμ shRNA were analyzed by flow cytometry. U-87 MG cells were infected with control or PTPμ shRNA and serum arrested with basal media at 24 h postinfection. Cells either remained serum arrested for 48 h or were stimulated after 24 h of serum arrest with complete medium containing 10% serum for an additional 24 h. As shown in Fig. 6, there were no significant differences in cell cycle distribution observed between either serum-arrested or serum-stimulated U-87 MG cells infected with control or PTPμ shRNA. Approximately 80% of the cells were diploid in all conditions, and the remaining aneuploid population had a cell cycle distribution similar to that of the diploid cells (data not shown). Therefore, PTPμ knockdown did not alter U-87 MG cell proliferation.
To determine whether the absence of PTPμ in shRNA-infected U-87 MG cells influenced their migratory behavior in the complex environment of the brain, PTPμ expression was knocked down in U-87 MG glioma cells for assay in a brain slice model. A modified version of the brain slice assay for dispersal was used that more closely approximates in vivo migratory conditions20 (Fig. 7A). This assay allows for evaluation of dispersal and proliferation of glioma cells through the complex matrix of the adult brain, an environment that simulates glioma cell dispersal in vivo. Previous studies have characterized the behavior of primary rat astrocytes using this assay in neonatal rat brains.20
U-87 MG cells were infected with either control or PTPμ shRNA containing a GFP reporter. After 3 days, these cells were injected into the cortex of ex vivo adult rat brain slices. Cell migration was measured in brain slices after 48 h. By following their migration over time using GFP fluorescence, cell movement through a three-dimensional matrix environment was evaluated. The assay was quantitated by measuring the average area of fluorescent cells that migrated away from the injection site in a given slice at the 48-h time point. Control shRNA-infected U-87 MG cells were not dispersive and remained as a tight clump of cells that did not migrate into the adult brain tissue of the slice (Fig. 7B, left). Knockdown of PTPμ by shRNA, however, induced a significant dispersal of cells away from the injection site (Fig. 7B, right). Migration of these cells occurred in several focal planes of the brain tissue, suggesting dispersal throughout the three-dimensional architecture of the brain slice. These results suggest that loss of PTPμ protein expression correlates with increased migration and dispersal of the glioma cells.
We observed a change in the morphology and dispersive phenotype of U-87 MG cells upon the loss of PTPμ in vitro and in a three-dimensional assay. Therefore, we hypothesized that these changes would be recapitulated in vivo using an intracranial mouse xenograft model. U-87 MG cells expressing control or PTPμ shRNA were intracranially implanted into nude mice to allow tumors to form. After 2 weeks, whole brains were sectioned to analyze the morphology of the two tumor groups. Both hematoxylin and eosin staining and GFP labeling of the tumor cells indicate a marked difference in the morphology of the tumors in each group. U-87 MG cells expressing control shRNA formed tightly packed tumors of cells that appear to retain cell–cell adhesion (Fig. 8A,B). In contrast, the tumors in which PTPμ expression was decreased by shRNA were composed of a loose mass of cells that were heterogeneous in size and shape (Fig. 8A,B). These changes are consistent with the pleomorphism and heterogeneity seen in infiltrative high-grade glioma tumors and support our data that the loss of PTPμ is an important molecular change in GBMs. Taken together, our results suggest that PTPμ may play an important role in contact-dependent signaling to negatively regulate migration of glial cells and that the loss of PTPμ protein expression may be advantageous to GBM formation during the migration and dispersal process.
GBM remains a significant treatment challenge due to the extensive dispersal of cells throughout the brain parenchyma. A prerequisite for dispersal is the loss of contact inhibition of movement. RPTPs are thought to be important mediators of the adhesion-dependent signals of contact inhibition of movement. Therefore, RPTPs such as PTPμ are attractive candidates as regulators of glioma cell dispersal. Primary rodent astrocytes express PTPμ, and the PTPμ expression levels increased as they became confluent at high density. We hypothesized that the lack of PTPμ expression in dispersive GBMs is a critical step in the loss of contact inhibition of movement, facilitating the migration of those cells throughout the neuropil. Consistent with this notion, substantial levels of PTPμ were detected in nontumorigenic human brain and low-grade astrocytomas, but PTPμ was absent in high-grade GBMs and migratory tumor cell lines. If PTPμ is a critical regulator of tumor cell dispersal, then its interacting molecules and downstream signaling components represent unique candidates for targeting brain tumor cells. This is especially important since in many cases, a population of GBM tumor cells migrates and disperses throughout surrounding brain tissue and cannot be detected or removed during surgical resection.8 Migration and dispersal of gliomas in vivo have been proposed to occur along myelinating axons and extracellular matrix found in the basement membrane of blood vessels and the pial surface.4,7 We have developed a novel strategy to assess the migration and dispersal of brain tumor cells in the three-dimensional environment of the brain parenchyma. Using an adult brain slice assay, we have demonstrated that U-87 MG cells do not migrate through the brain parenchyma but can be induced to disperse extensively by knocking down PTPμ protein levels. Interestingly, another RPTP, RPTP ζ/β, is also altered in gliomas and plays a role in regulating cell growth34 and migration.35,36 Together, these data suggest that changes in the expression and function of both tyrosine kinases and phosphatases are important events in the progression of gliomas.
The molecular mechanisms of glioma cell migration and invasion are complex and likely to be multifaceted.8 Although the precise signaling events of PTPμ-dependent regulation of glioma cell migration are not currently known, there are a number of insights from other systems that suggest downstream signaling from PTPμ is important for cell migration. PTPμ is known to regulate neuronal migration and axon guidance in the chick,12 where it has been extensively studied. For example, PTPμ-mediated neurite outgrowth of chick retinal ganglion cells requires both the receptor for activated C kinase (RACK1)37 and protein kinase C delta (PKCδ).38,39 Similar associations may apply in gliomas since PKCδ has been implicated in the regulation of proliferation, migration, and apoptosis of human glioma cells.40,41 PTPμ has also been shown to be required for both E- and N-cadherin–mediated neurite outgrowth,42,43 which correlates with our data suggesting N-cadherin expression levels are decreased in GBM. This change accompanies the loss of PTPμ protein expression. PTPμ and N-cadherin are known binding partners.43,44 It is interesting to hypothesize that alterations in the PTPμ/N-cadherin signaling complex contribute to the dysregulation of cell migration.
PTPμ also interacts with the scaffolding protein IQGAP1,45 whose associated proteins Rac1 and Cdc42 are important in regulating cell motility. The rho GTPases, including Cdc42 and Rac1, are particularly interesting as potential downstream mediators of PTPμ function in gliomas, since they have been shown to regulate migration and invasion in other systems.46–48 In fact, Cdc42 activity is required for PTPμ-dependent neurite outgrowth and axon guidance.45,49,50 IQGAP1 function and association with Cdc42 are required for PTPμ-dependent neurite outgrowth.45 We have recently demonstrated that PTPμ-mediated repulsion of temporal retinal neurons during axon guidance is induced by PTPμ’s ability to inhibit Rac1 activity.50 Data from the present study suggest that the loss of PTPμ in glioma cells induces Rac1-dependent lamellipodia, which further supports our previous findings that PTPμ negatively regulates Rac1.50 Clearly, PTPμ is implicated in several signaling pathways that are altered in human tumors, including GBM. It is an attractive notion that the various PTPμ signaling pathways converge to regulate cell adhesion and migration and that the loss of PTPμ in glioma tumor cells promotes dispersal.
This research was supported by National Institutes of Health grant R01-NS051520 (S.M.B.-K., S.R., M.A.V., and R.H.M.). Support to R.H.M. was provided by the Case Comprehensive Cancer Center (P20-CA103736). Additional support was provided by Visual Sciences Research Center core grant P30-EY11373 from the National Eye Institute and Case Comprehensive Cancer Center core grant P30-CA043703 from the National Cancer Institute. A.E.S. was supported by National Cancer Institute grants K08-CA101954 and R01-CA116257, the Ivy Brain Tumor Foundation, and the Cancer Genome Atlas project. A.M.B. was supported in part by National Institutes of Health grants T32-GM007250 (Medical Scientist Training Program) and T32-CA059366. We thank Carol Luckey, Scott Becka, Theresa Gates, and Dr. Moonkyung Caprara for technical support and Dr. Mark Cohen for analysis of xenograft specimens. We thank Sara Lou and Scott Howell for help with figures and graphs, as well as members of the Brady-Kalnay lab for insightful discussions.