DRR Drives Brain Cancer Invasion by Regulating Cytoskeletal-Focal Adhesion Dynamics

Oncogene (2010) 29, 4636–4647 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc ORIGINAL ARTICLE DRR drives brain cancer invasion by regulating cytoskeletal-focal adhesion dynamics

PU Le, A Angers-Loustau, RMW de Oliveira, A Ajlan, CL Brassard, A Dudley, H Brent, V Siu, G Trinh, G Mo¨lenkamp, J Wang, M Seyed Sadr, B Bedell, RF Del Maestro and K Petrecca

Department of Neurology and Neurosurgery, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada

Malignant glioma invasion is a primary cause of brain whereas malignant gliomas are highly invasive (Kleihues cancer treatment failure, yet the molecular mechanisms and Cavenee, 2000). underlying its regulation remain elusive. We developed One approach to convert this devastating cancer into a novel functional-screening strategy and identified down- a more manageable one is to inhibit malignant glial cell regulated in renal cell carcinoma (DRR) as a regulator of (MGC) invasion; maintaining MGCs in a local environ- invasion. We show that DRR drives invasion in vitro and ment leaves further treatment options open. The current in vivo. We found that while DRR is not expressed understanding of cell invasion is a composite derived in normal glial cells, it is highly expressed in the invasive from studies of different cell types and environments component of gliomas. Exploring underlying mechanisms, (Friedl and Wolf, 2003; Ridley et al., 2003). It includes we show that DRR associates with and organizes the actin the extension of a cellular process, attachment through and microtubular cytoskeletons and that these associa- focal adhesion (FA) formation, degradation of the tions are essential for focal adhesion (FA) disassembly and extracellular matrix to create space to accommodate cell invasion. These findings identify DRR as a new the moving cell, translocation of the cell body forward, cytoskeletal crosslinker that regulates FA dynamics and and release of cell rear FAs (Friedl and Wolf, 2003). cell movement. This multistep process requires the coordinated action Oncogene (2010) 29, 4636–4647; doi:10.1038/onc.2010.216; of cell surface receptors, signaling pathways, cytoskele- published online 14 June 2010 tal elements, FA components, and extracellular matrix degrading enzymes (Burridge and Chrzanowska-Wod- Keywords: brain cancer; cytoskeleton; cell invasion nicka, 1996; Lauuffenburger and Horwitz, 1996; Ridley et al., 2003). Within this scheme, recent studies have pointed to the importance of actin/microtubule (MT) Introduction dynamics in both cell front membrane protrusion and cell rear retraction (Palazzo and Gundersen, 2002; Gliomas, the most common primary brain cancers, are Rodriguez et al., 2003). among the most devastating of human malignancies Cell rear retraction requires regulated FA disassembly (Louis et al., 2002). Benign gliomas, known as pilocytic (Friedl and Wolf, 2003). Earlier studies have shown that astrocytomas, are seen in children. They are very well the actin/MT system has an important function in the treated by complete surgical resection, with patients process (Kaverina et al., 1998, 1999; Krylyshkina et al., typically maintaining a full life expectancy (Fernandez 2002, 2003; Ezratty et al., 2005); however, the molecular et al., 2003). Malignant gliomas, known as astrocyto- mechanisms that underlie FA disassembly remain mas, oligodendrogliomas, or glioblastomas, are adult unclear. A recent study has shown that conditional loss neoplasms. They can be further divided into low grade of the spectraplakin, ACF7, leads to abnormalities in and high grade. Low-grade gliomas are highly invasive, wound repair and defects in epidermal cell migration. but have low proliferation rates, often invading multiple ACF7 achieves these functions by coordinating actin/ lobes before clinical presentation. Over time, low-grade MT dynamics to regulate FA turnover (Wu et al., 2008). gliomas incur genetic changes that convert them to a Although there are many similarities between cell higher grade (Louis et al., 2002). The distinguishing movement in normal physiologic conditions and in feature between benign and malignant gliomas is brain cancer, MGCs are thought to use additional or alternate invasion; benign gliomas do not invade normal brain, mechanisms (Beadle et al., 2008). Instead of using a movement paradigm similar to epithelial or mesenchy- Correspondence: Dr K Petrecca, Department of Neurology and mal cells, MGCs invade the dense substance of the brain Neurosurgery, Montreal Neurological Institute and Hospital, McGill using a mode of cell movement that is more similar to University, 3801 University Avenue, Suite 109c, Montreal, Quebec neural progenitor cell movement. To uncover molecular Canada H3A 2B4. E-mail: [email protected] mechanisms that specifically regulate MGC invasion, we Received 21 September 2009; revised 20 April 2010; accepted 25 April developed a novel forward genetic-screening method 2010; published online 14 June 2010 and identified downregulated in renal cell carcinoma DRR regulates glioma invasion PU Le et al 4637 (DRR) as a novel powerful promoter of MGC invasion. with their WT counterparts, which express endogenous DRR was originally cloned from the short arm of DRR, all DRRÀ cell lines exhibited a significant chromosome 3 from patients with renal cell carcinoma reduction in their invasiveness (Supplementary Figures (Wang et al., 2000). Although a definitive function for S4C–E). Interestingly, DRR expression also leads to a this protein has not been reported, a function as a tumor profound change in cell morphology as DRR þ cells are suppressor has been suggested as heterologous expres- elongated and spindle shaped, whereas DRRÀ cells are sion reduced cell division (Wang et al., 2000). round (Figures 1h, i, and k). Experiments in 2D We show here that DRR drives MGC invasion in migration assays also reveal differences in the morphol- both in vitro and in vivo invasion assays. Importantly, we ogy of cells as they migrate. DRR þ cells migrate with have clinically validated the importance of DRR; long thin protrusions, whereas WT and DRRÀ cells although DRR is not expressed in normal human brain migrate with a uniform broad lamella (Supplementary glia, it is highly expressed in the invasive component of Figure S2). An elongated spindle cell shape has been malignant gliomas. Thus, DRR expression in malignant shown to be the preferred mode of MGC movement gliomas correlates with invasion. We delved into the through brain (Beadle et al., 2008). molecular mechanism through which DRR functions We next examined DRR’s function as an invasion and uncovered that DRR associates with and organizes promoter in a mouse model. DRR þ and DRRÀ tumors both the actin and MT cytoskeletons. By uncoupling were implanted into the subcallosal/caudate region of this association, we show that DRR interaction with mice and invasion was assessed (Figures 1l and m). actin and MTs is essential for FA disassembly and cell DRRÀ tumors grow as a well-circumscribed mass invasion. Our findings suggest that DRR has an without invasion into the adjacent parenchyma, and important function in glioma biology by augmenting these cells have a round morphology. Conversely, FA dynamics, thus driving MGC invasion. Together, DRR þ tumors are highly invasive. These invasive cells, these findings identify DRR as a new and important which are distinguished by their large, hyperchromatic cytoskeletal crosslinker that regulates FA dynamics and and elongated nuclei, have an elongated shape, separate directional cell movement. from the tumor mass, invade parenchyma, and, importantly, move toward and into the corpus callosum. Invasion into white matter tracts such as the corpus callosum is a preferred invasion paradigm used by Results human malignant glial tumors (Pedersen et al., 1995). Furthermore, DRR þ tumors were smaller than DRRÀ Functional-screening assay identifies DRR as a promoter tumors (Supplementary Figure S1B and C), suggesting a of invasion decrease in cell proliferation as earlier described (Wang MGC invasiveness can be assayed using a 3D invasion et al., 2000). We assessed the function of DRR in cell model (Del Duca et al., 2004). Using this model as division and also found that cell division is inversely a starting point, we developed a novel functional- correlated with DRR expression (Figure 1n). The notion screening assay by retrovirally transducing MGCs, the that MGC invasion and proliferation are temporally U251 glioma cell line, to express an entire brain cDNA exclusive events has been described (Giese et al., 1996). library. We reasoned that if we could make a cell Our data show that we have developed a robust- heterologously express a gene that promotes invasion, it screening strategy to identify molecules that drive cell would be distinguishable from other cells as a hyperinvasion. Using this assay, we have identified DRR as a invasive cell. Tumor spheroids were generated from candidate regulator of cell movement. We have vali- these transduced MGCs and their invasiveness was dated this finding in both in vitro and in vivo invasion assessed in the 3D invasion model. Distinguishable assays thus confirming that DRR is an important hyperinvasive cells were then captured and expanded in molecule in cell invasion. culture and the originally transduced gene was identified (Figure 1a). DRR was identified as a strong promoter of invasion using this forward genetic approach. DRR is expressed in neurons and human gliomas To test whether or not DRR acts as an effector of but not in normal glia MGC invasion, we generated composite tumor spher- To clinically validate our in vitro and in vivo findings, we oids made up of both DRR overexpressing MGCs set out to determine the expression pattern of DRR in (DRR þ ; Supplementary Figure S1) and wild-type (WT) normal human brain and malignant gliomas. We found MGCs and studied the invasion parameters of each cell that in normal human brain DRR is strongly expressed line (Figures 1b–e). Although MGCs endogenously in neurons, but not in astrocytes or in oligodendrocytes express DRR, DRR þ cells invade 240% farther than (Figures 2a–f; see Supplementary Figure S3 for high- WT cells. In contrast, reducing DRR expression in magnification images). Colabeling of cultured embryo- MGCs, using RNA interference (DRRÀ, Supplementary nic rat neurons and glia with neuronal and glial markers Figure S1) causes a significant decrease in invasion also shows that DRR is expressed in neurons, but not in (Figures 1f, g, and j). To test whether reducing DRR glial cells (Figures 2g–l). DRR antibody controls expression decreases invasion in other glial cell lines, we including antigenic peptide competition, immunolabeling developed U343-DRRÀ, C6-DRRÀ, U87MG-DRRÀ with preimmune serum, and single secondary antibody stable cell lines and tested their invasiveness. Compared immunolabeling were negative. Analysis of DRR

Oncogene DRR regulates glioma invasion PU Le et al 4638 expression in eight malignant gliomas of each grade uniformly express DRR, whereas the central prolifera- indicates that DRR is highly but not uniformly tive tumor region showed variability in DRR expres- expressed in all malignant glial tumors (Figures 2m). sion. The central tumor in 5/8 grade 4 gliomas showed Grade 2 and 3 gliomas, which are highly invasive tumors little to no DRR expression, whereas the central tumor with low proliferation rates, uniformly express DRR. In was DRR positive in 3/8 tumors. Taken together, these contrast, grade 4 gliomas, which are both highly data show that although DRR is not expressed in invasive and highly proliferative, express DRR in a normal glial cells, it has a robust and differential suggestive pattern. The invasive peripheral tumor cells expression pattern in malignant gliomas.

120 2.5 * 100 2 80 1.5 60 Infection with retroviral brain cDNA library 1

Cell Number 40

glial cells 0.5

20 Fold increase in Invasion

Generation of spheroids and 0 implantion into 3D matrix 0 + 0 10 80 WT DRR <70 >280 70-14140-2 210-2 Invasion Distance (µm)

700 * 100 600 90 80 Isolation and expansion of hyperinvasive cells 500 70 400 60 50 300 40 Cell Number 200 30

Percent Round Cells 20 100 Identification of 10 inserted gene 0 0 - + - DRR WT DRR WT DRR 1600 1400 1200 1000 800 600 WT 400 Cell Number (x1000) 200 0 24 48 72 96 Time (hrs) Figure 1 Validation of DRR as a regulator of invasion. (a) Outline of functional genetic-screening assay. (b) Mixed tumor spheroid containing WT glial cells (cytotracker red label) and DRR overexpressing cells (DRR þ , transparent) showing hyperinvasion of DRR þ cells. Solid circle demarcates invasion front of WT cells, dashed circle demarcates invasion front of DRR þ cells. (c) Control mixed tumor spheroid showing equal invasion of WT cytotracker red-labeled cells and WT-unlabeled cells showing that cytotracker red labeling does not influence invasion. (d) Quantitative analysis of invasion. (e) Quantification of maximal invasion of WT (red bars) and DRR þ (empty bars) cells. Data are mean±s.e.m. (n ¼ 14 for each cell line). Asterisk, Po0.001. (f) Tumor spheroid generated from DRRÀ cells. Circle demarcates invasion front. (g) Tumor spheroid generated from WT cells. Circle demarcates invasion front. (h) High-magnification image of inset in (f) shows that DRRÀ cells have a round cell shape. (i) High-magnification image of inset in (g) shows that WT cells have an elongated cell shape. (j) Quantification of cell invasion comparing DRRÀ cells and WT cells. Cells invading greater than 400 mm were counted. Data are mean±s.e.m. (n ¼ 8 for each cell line). Asterisk, Po0.001. (k) Quantification of the effect of DRR expression on cell shape showing that DRR expression promotes an elongated cell shape. (l) DRR þ cells implanted into mouse brain showing elongated cell shape and invasion into corpus callosum (cc). Arrows indicate MGCs that have invaded the corpus callosum. Arrowheads delineate tumor border. Arrow in inset indicates tumor implantation site. Bar ¼ 100 mm. (m) DRRÀ cells implanted into mouse brain showing round cell shape and no evidence of invasion toward the corpus callosum. Arrowheads delineate tumor border. Arrow in inset indicates tumor implantation site. Bar ¼ 100 mm. (n) Quantification of cell proliferation in DRR þ ,WT, and DRRÀ cells.

Figure 2 DRR is expressed in neurons and human gliomas, but not in normal glia. DRR immunolabeling of normal human brain at low (a, b) and high (c, d) magnification shows that DRR is found within the cortex, but not in white matter (wm). Expression of the glial marker GFAP does not overlap with DRR (a–d). DRR is not expressed in the aneuronal molecular layer (ml) of the cortex (c). High-magnification imaging shows that DRR is highly expressed in neurons (e) but not in white matter (f). Rat brain cultures similarly show that DRR expression overlaps with the neuronal marker MAP2 in neurons (g–i) but not with the glial marker GFAP in glia (j–l). (m) DRR expression in eight malignant gliomas of each grade was assessed. Both grade 2 and grade 3 gliomas (left panels, top and bottom) uniformly express high levels of DRR. In contrast, only the invasive peripheral tumor (PT) portions of grade 4 gliomas uniformly express DRR (right panel, bottom). The central tumor (CT) portion exhibits variable DRR expression, negative in five and positive in three tumors (right panel, top and middle). (h, e): hematoxylin and eosin, Ki-67: marker of cell division revealing high levels of proliferation in the central tumor region.

Oncogene DRR regulates glioma invasion PU Le et al 4639 DRR associates with the cytoskeleton Endogenous or heterologously expressed DRR predo- To uncover how DRR functions to drive cell movement, minantly localizes along actin stress ﬁbers, FAs, and we began by localizing DRR at the subcellular level. membrane rufﬂes (Figure 3a). In agreement with an

Oncogene DRR regulates glioma invasion PU Le et al 4640

Oncogene DRR regulates glioma invasion PU Le et al 4641 earlier report (Wang et al., 2000), DRR can also be DRR-associated protein, and determined that this found in the nucleus (Supplementary Figure S5). These association could be disrupted by mutation of the results suggest that DRR could be promoting invasion amino-terminal HRE region (DRRDHRE). through a direct influence on the cytoskeletal apparatus or through a regulatory function in the nucleus. DRR association with the cytoskeleton is required To address this issue, we sought to uncouple DRR for cell movement from the actin cytoskeleton by identifying minimal The ability to disrupt the DRR-actin and DRR-LC2 domains required for actin association. To do so, we associations provides two methods to determine whether generated sequential N- and C-terminal truncated DRR association with the cytoskeleton is required to constructs of human DRR and assayed for localization using fluorescent tags. Two minimal regions capable of drive cell movement. We thus generated stable cell lines expressing DRRDPEPE or DRRDHRE and tested their actin association were identified, amino acids 62–100 invasiveness in a 3D invasion assay. We found that loss and 108–120 (Supplementary Figure S7). We then B speculated that cytoskeletal association would be a of the DRR-actin association leads to a threefold reduction in invasion compared with WT cells, suggest- property of DRR that would be conserved across ing that this mutant form of DRR is acting as a species and, therefore, domains required for actin functional dominant negative (Figure 4). A similar association would also be conserved across species. finding is also seen when the DRR–LC2 interaction is We identified and mutated the amino acids within the disrupted (Figure 4). Together, these data show that 62–100 and 108–120 regions that were conserved across DRR association with actin and LC2 is required to drive human, mouse, rat, and zebrafish DRR (Supplementary cell invasion. Figure S6). We found that the combined mutation of the conserved PE motifs to alanines in both segments (DRRDPEPE) leads to a significant perturbation of the DRR regulates FA dynamics actin cytoskeleton and abolishes actin association with The process of cell movement requires regulated FA the remaining stress fibers (Figure 3a). dynamics (Lauuffenburger and Horwitz, 1996; Friedl Along the same line, we also wanted to learn more and Wolf, 2003). To determine whether DRR expression about DRR’s mechanism of action by identifying affects FAs, we expressed a GFP–paxillin fusion protein molecules that affect DRR association with the cytos- and studied the effect of DRR expression on FA keleton. We used yeast two-hybrid screening of normal dynamics using confocal videomicroscopy (Figure 5). brain libraries to identify DRR-binding partners. The In non-polarized DRR þ cells, we found that the total light chain (LC2) subunit of MAP1A was identified as a time taken for FAs to form and disassemble is candidate DRR-binding protein. Indeed, we found that 40.05±3.00 min (Figures 5a–c). The rate constant for DRR and LC2 colocalize along actin stress fibers and GFP–paxillin incorporation into FAs was membrane ruffles, and can be coimmunoprecipitated (6.2±0.9) Â 10À3 per min and the rate constant for when heterologously expressed (Figures 3b and c), GFP–paxillin disassembly was (8.6±0.7) Â 10À3 per suggesting their association. Interestingly, mutation of min. Conversely, FAs were not dynamic in WT control the DRR-actin-binding sites seems to increase DRR cells. We were unable to detect FAs that formed or association with LC2 (Figure 3b). We also found, in disassembled within the 170 min imaging interval some cells, that the non-actin-binding form of DRR (Figure 5d). These data strongly support a mechanism (DRRDPEPE) can localize to MTs (data not shown). We whereby DRR drives cell invasion by enhancing FA then developed non-LC2-binding forms of DRR using dynamics. truncation and amino-acid mutagenesis analysis. A It has been established that FA disassembly requires minimal N-terminal HRE sequence was found to be polymerized MTs (Kaverina et al., 1998, 1999; required for LC2 binding (Figure 3b). When this region Krylyshkina et al., 2002, 2003; Ezratty et al., 2005). As was mutated, DRRDHRE, there is a significant perturba- we have shown that DRR interacts with the LC2 tion of the actin cytoskeleton and the localization subunit of MAP1A and that DRR overexpression leads pattern of DRR changes. There is an increase in DRR to less stable FAs, we wanted to examine directly expression in the nucleus and diffuse cytoplasmic whether DRR promotes FA disassembly. localization (Figure 3a). We also observed slight actin One way to examine MT control of FA disassembly is association. In summary, these data show that DRR to disassemble MTs using the microtubular depolymer- localizes to the actin cytoskeleton, FAs, and nucleus. We izing agent nocodazole. After nocodazole application, defined minimal regions required for actin association FAs increase in size as there are no MTs available for (DRRDPEPE), identified the LC2 subunit of MAP1A as a disassembly. On nocodazole washout, MTs polymerize

Figure 3 DRR associates with the actin cytoskeleton and interacts with LC2. (a) Transfected DRR localizes along actin stress fibers and focal adhesions. Arrows indicate expression at FA sites. Actin is labeled with phalloidin. The non-actin-binding DRRDPEPE, and the non-LC2-binding DRRDHRE, is diffusely expressed in the cytoplasm. They do not localize to actin or FAs. DRRDHRE can also be found in the nucleus. (b) Co-localization of FLAG-DRR and MYC-LC2 along actin stress fibers, lamellipodia, and membrane ruffles. (c) Co-immunoprecipitation of heterologously expressed FLAG-DRR and MYC-LC2 from glial cells. MYC-LC2 co-immuno- precipitates with FLAG-DRR and FLAG-DRRDPEPE but not when the conserved N-terminal HRE sequence, DRRDHRE,ismutated.

Oncogene DRR regulates glioma invasion PU Le et al 4642

Figure 4 DRR association with actin and LC2 is required for cell invasion. (a) 3D invasion assays of WT, DRR þ , DRRDPEPE, and DRRDHRE in a 3D collagen matrix. (b) Closer view of the spheroid margins showing cell invasion. Asterisk indicates the spheroid edge in DRRDPEPE cells. (c) Quantitative analysis of cell invasion after 24, 48, and 72 h.

and FAs disassemble (Ezratty et al., 2005). When this 1999). We tested this hypothesis and found that DRR experiment was performed using DRR þ and DRRÀ cells regulates both the MT and actin cytoskeletons. DRR plated on the extracellular matrix component fibronec- expression leads to a highly organized MT system that tin, two striking findings were observed. First, on strongly parallels the localization pattern of the actin nocodazole application and MT depolymerization, cytoskeleton (Figure 7a). In contrast, DRR deficiency DRR þ cells develop more and larger FAs compared leads to an irregular, poorly organized MT cytoskeleton with non-treated cells (Figure 6a). In contrast, we did that does not parallel the actin cytoskeleton (Figure 7b). not observe differences in FA number and size in DRR deficiency also leads to a profound change in the nocodazole treated versus non-treated DRRÀ cells actin cytoskeleton with loss of stress fiber formation and (Figure 6a; Supplementary Figure S8) or WT cells (data the promotion of a cortical actin system (Figures 7a and not shown), suggesting that DRR deficiency leads to b). The promotion of a stress fiber actin system allows large and mature FAs. Second, when MTs repolymerize for actomyosin contraction and thus cell rear retraction in cells overexpressing DRR, FAs begin to disassemble (Verkhovsky et al., 1995). Importantly, we also found within 5 min, whereas FAs in DRRÀ cells (or WT cells, that DRR expression is required for MTs to reach FAs. data not shown) only begin to disassemble after 15 min MTs in DRR-deficient cells do not approach FAs and do not completely disassemble (Figure 6a). When (Figure 7b). In contrast, DRR expression leads to a the same experiment was performed with the non-actin- close association between MTs and FAs (Figure 7a). binding form of DRR (DRRDPEPE) or the non-LC2 form Together, these data strongly suggest that DRR is a of DRR (DRRDHRE), the FA disassembly kinetics were novel regulator of FA dynamics by controlling both the similar to DRR-deficiency conditions (Figure 6b, data actin and MT cytoskeletons (Figure 7c). not shown for DRRDHRE). The findings that DRR expression promotes FA disassembly whereas DRR deficiency leads to stable mature FAs point to DRR as a novel regulator of FA dynamics. Discussion

MGC invasion into normal brain is a deﬁning feature of DRR organizes the actin and microtubular cytoskeletons malignant gliomas and is directly related to the The results from the FA disassembly assay suggest that aggressive nature of these cancers. Many molecules DRR association with both the actin cytoskeleton and have been implicated in malignant glioma invasion the LC2 subunit of MAP1A are required for FA (Salhia et al., 2006; Johnston et al., 2007; Wang et al., disassembly (Figure 6). One mechanism through which 2008), but the molecular mechanisms that regulate DRR could achieve this result is to alter MT dynamics MGC invasion remain elusive. This study shows a novel by placing MTs in the vicinity of FAs (Kaverina et al., functional-screening assay to identify promoters of

Oncogene DRR regulates glioma invasion PU Le et al 4643

Figure 5 DRR promotes focal adhesion dynamics. DRR þ and WT cells were transfected with GFP–paxillin and imaged using confocal videomicroscopy for 170 at 1 min intervals. (a) DRR þ cells transfected with GFP–paxillin. Representative cell showing dynamic membrane protrusions and FA assembly and disassembly. Arrows indicate areas of robust FA assembly and disassembly. Boxes, (b), and (c), represent high-magniﬁcation areas shown in panels (b) and (c). (d) WT cell transfected with GFP–paxillin. Representative cell showing a lack of membrane protrusions and stable FAs. No FAs were identiﬁed that assembled or disassembled over the imaging interval.

Figure 6 DRR promotes focal adhesion disassembly. (a) DRRÀ, DRR þ and (b) DRRDPEPE were starved for 24 h and left untreated or treated for 4 h with 10 mM nocodazole. The MT depolymerizer was then washed out for the indicated time. DRR expression promotes FA disassembly, whereas DRR deﬁciency leads to more stable FAs.

Oncogene DRR regulates glioma invasion PU Le et al 4644

Figure 7 DRR organizes the actin and microtubular cytoskeletons. (a)DRRþ and (b)DRRÀ cellsweregrownonFN(10mg/ml) for 48 h before ﬁxation. Cells were then labeled for MTs (green), actin (red), and vinculin (blue).The insets represent higher magniﬁcation of the indicated outlined boxes. Arrows indicate that MTs are targeted to FAs in DRR þ cells, whereas MTs do not reach FAs in DRRÀ cells. Bars ¼ 20 mm. (c)Aworkingmodel summarizing the function of DRR in cytoskeletal organization and invasion. We propose that with LC2, DRR acts as an actin-MT crosslinker. DRR targets MTs to FAs promoting their disassembly, cell rear retraction, and cell invasion.

Oncogene DRR regulates glioma invasion PU Le et al 4645 invasion. Using this assay, we identified DRR as an links the actin and MT cytoskeletons by directly or invasion promoter. We show here that DRR promotes indirectly interacting with actin and indirectly interact- MGC invasion in 3D in vitro and in mouse models of ing with MTs by associating with the LC2 subunit invasion. Importantly, characterization of DRR expres- of MAP1A (Figure 7c). In support, we show that sion in normal human brain and gliomas reveals that in DRR localizes to the actin cytoskeleton and FAs and normal brain DRR is abundantly expressed in neurons, interacts with the LC2 subunit MAP1A. We show that but not in glia. In contrast, DRR is uniformly and DRR expression organizes both the actin and MT highly expressed in the invasive regions of both low- and cytoskeletons so that MTs approach FAs and promote high-grade gliomas, whereas its expression in the central their disassembly. DRR deficiency, or the disruption of proliferative region of high-grade gliomas is variable. this complex by abolishing DRR-actin or DRR-LC2 Together, these findings implicate DRR as an important association, leads to a loss of coordination between regulator of glioma invasion and suggest that DRR may actin and MTs, as well as the inability of MTs to be useful as a biomarker to delineate invasive regions reach FAs. and grade malignant gliomas. These findings translate into important consequences A recent study has linked DRR and malignant for cancer cell invasion. We propose that the de novo gliomas, reporting that DRR expression is reduced in expression of DRR observed in invasive MGCs leads to high-grade gliomas compared with low-grade gliomas, more rapid FA disassembly and thus invasion. The whereas expression in normal brain was not described inability of MGCs to separate from the tumor mass (van den Boom et al., 2006). We believe that our work when the DRR-actin association is abolished has supports and expands on this report and could be clinical implications. Indeed, a CD151-specific metasta- explained by the anti-proliferative function of DRR sis blocking monoclonal antibody has been shown to shown here and reported elsewhere (Wang et al., 2000). inhibit metastasis by preventing cell rear retraction and We interpret our results in the following way: low-grade thus cell detachment and migration from the primary gliomas express DRR and thus exhibit an invasive tumor mass (Zijlstra et al., 2008). Future studies will phenotype. Over time, a subpopulation of these cancer be needed to identify potential anti-invasive chemo- cells develops additional genetic alterations, which therapeutic targets within the DRR-associated macro- include loss of DRR expression, converting low-grade molecular complex. gliomas to a higher grade. The loss of DRR expression Finally, we observed that DRR expression converts leads to high cell proliferation, as is seen in the central cells from a round to an elongated spindle tumor region of high-grade gliomas. In this scheme, shape, reminiscent of that recently shown to be the DRR not only drives cell invasion, it can also regulate preferred mode of MGC invasion and similar to the cell proliferation. This implicates DRR as a central mode of cell movement used by neural progenitor player in glioma biology, possibly acting as a molecular cells (Beadle et al., 2008). Significant progress has been switch that converts invasive low-grade gliomas to their made toward understanding the origins of brain cancers high-grade proliferative counterparts. More work will with recent data pointing toward neural progenitor be required to address this possibility. Indeed, it has cells (Dirks, 2008). Our findings that DRR, in normal been shown that invasion and division are temporally brain, is expressed in neurons but not glia adds further exclusive events in MGCs (Giese et al., 1996). intrigue to this notion. In the normal physiologic Cell invasion requires a cycle of events that uses both condition, is DRR an effector of neuronal migration? the actin cytoskeleton and MTs along with the Is there an oncogenic event that triggers its de novo continuous formation and disassembly of FAs. The expression in brain cancer, or, is it expressed in the early function of MTs in FA turnover has been well neural progenitor cells that may be the origins of these established (Kaverina et al., 1998, 1999; Krylyshkina cancers? Future work is necessary to answer these et al., 2002, 2003; Ezratty et al., 2005). More recently, questions. FAs have been postulated to be the site of actin and MT crosstalk (Rodriguez et al., 2003). Accumulating evidence suggests that MTs grow toward FAs in a process that is Materials and methods coordinated by F-actin. Much less is known about the constituents of FAs in normal glial cells and gliomas, and Functional-screening assay the molecular mechanisms that regulate FA turnover in A normal human adult brain cDNA library (Clontech, normal glial cells and gliomas have not been described. Mountain View, CA, USA) was subcloned into the pLib retroviral vector (Clontech) and used to transfect the PT67 However, a recent study reported that the Rho effector packaging cell line using Lipofectamine PLUS reagent mDia1 (mammalian homologue of the Drosophila gene (Clontech). The secreted replication deficient retrovirus was Diaphanous 1) has a function in FA turnover by localizing collected from the supernatent 24–72 h post-transfection and c-SRC to FAs in the C6 glioma cell line, possibly through used to consecutively transduce, over a 72 h time course, the actin association (Yamana et al., 2006). WT-U251 glial cell line (Figure 1a). Our study provides evidence in support of the view that MTs facilitate FA disassembly and identifies DRR Human glioma analysis as a new player in this normal physiologic process. We Human glioma samples were obtained from the Brain Tumor show here that DRR is a novel actin/MT crosslinker Research Centre Tissue Bank at the Montreal Neurological that regulates FA disassembly. We propose that DRR Institute and Hospital (Montreal, Quebec, Canada). Written

Oncogene DRR regulates glioma invasion PU Le et al 4646 consent was obtained from all patients and the project was drug was washed out along a variable time course (5, 15, 30, approved by the Institutional Ethics Board at the Montreal and 60 min) using serum-free media. Fluorescently labeled cells Neurological Institute and Hospital. were visualized with a Zeiss 510 confocal microscope (63 Â objective). The number and surface area of vinculin- Cells stained FAs were quantified using Image J software. At least U251 human oligodendroglioma cell line (WT), human glial 10 fields from three independent experiments were quantified. tumor cell line (U343MG), rat astrocytoma cell line (C6), DRRÀ, and DRR þ cells were cultured in DMEM high-glucose Confocal videomicroscopy supplemented with 10% FBS and a penicillin–streptomycin WT or DRR þ cells were seeded on 35 mm glass bottom culture antibiotic mixture. Human glioblastoma cell line (U87MG) dishes (MatTek Corporation) before being transfected with (Cavanee lab, University of California at San Diego) was GFP–paxillin; 24–48 h post-transfection, the images were grown in DMEM high-glucose supplemented with 10% captured every 1 min for 170 min using a Zeiss 510 confocal inactivated FBS and a penicillin–streptomycin antibiotic microscope (63 Â objective). Five DRR þ and five control mixture. (WT) cells were analyzed, a total of 17 FAs were analyzed for DRR þ cells. The apparent rate constants for the incorporation Cell proliferation, migration, and invasion assays of GFP–paxillin into FAs and its disassembly from FAs was See Supplemental Information for details. quantified using the technique described in Webb et al., 2004. Measurements were obtained from five cells, 5–10 FAs/cell. In control cells, no FAs were identified that assembled or Mouse intracerebral tumor implantation disassembled within the 170 min imaging interval. Data is All animal experimentation was approved by the Institution’s presented as mean±standard error. Animal Care Committee and conformed to the guidelines of the Canadian Council of Animal Care. Six-week-old CD1 nu/nu athymic mice (Charles River, Canada) were anesthetized by an intraperitoneal injection containing Ketamine, Xylazine, Conflict of interest and Acepromazine. The authors declare no conflict of interest. Yeast two-hybrid screening Yeast two-hybrid screens were performed using the Match- maker Two-Hybrid System 3 (Clontech). Full-length DRR Acknowledgements was used as the bait to screen a human brain cDNA library (Clontech). We are grateful to Carmen Sabau for her technical expertise. This work was supported by Goals for Lily, the Alex Pavanel Immunocytochemistry Family, the Franco Di Giovanni Funds for Brain Tumor Cells were grown on glass coverslips or on fibronectin (10 mg/ Research, the Montreal English School Board, the B-Strong ml)-coated coverslips, fixed with 4% PFA and permeabilized Foundation to RFDM, and the Montreal Neurological with 0.5% Triton X-100 before being immunolabeled. The FA Institute to KP. RMWO was supported by the Canadian disassembly assay was performed as described earlier (Ezratty Institute of Health Research doctoral fellowship. AAL was et al., 2005). Briefly, cells were incubated in serum-free media supported by the National Cancer Institute of Canada Terry for 24 h before being treated with nocodazole (10 mM; 4 h). The Fox Studentship.

References

Beadle C, Assanah MC, Monzo P, Vallee R, Rosenfeld SS, Canoll P. Johnston AL, Lun X, Rahn JJ, Liacini A, Wang L, Hamilton MG (2008). The role of myosin II in glioma invasion of the brain. Mol et al. (2007). The p75 neurotrophin receptor is a central regulator of Biol Cell 19: 3357–3368. glioma invasion. PLoS Biol 5: e212. Burridge K, Chrzanowska-Wodnicka M. (1996). Focal adhesions, Kaverina I, Rottner K, Small JV. (1998). Targeting, capture, and contractility, and signaling. Annu Rev Cell Dev Biol 12: 463–519. stabilization of microtubules at early focal adhesions. J Cell Biol Del Duca D, Werbowetski T, Del Maestro RF. (2004). Spheroid 142: 181–190. preparation from hanging drops: characterization of a model of Kaverina I, Krylyshkina O, Small JV. (1999). Microtubule targeting of brain tumor invasion. J Neurooncol 67: 295–303. substrate contacts promotes their relaxation and dissociation. J Cell Dirks PB. (2008). Brain tumor stem cells: bringing order to the chaos Biol 146: 1033–1044. of brain cancer. J Clin Oncol 26: 2916–2924. Kleihues P, Cavenee WK. (2000). World Health Organization Classiﬁca- Ezratty EJ, Partridge MA, Gundersen GG. (2005). Microtubule- tion of Tumours of the Nervous System. WHO/IARC: Lyon. induced focal adhesion disassembly is mediated by dynamin and Krylyshkina O, Kaverina I, Kranewitter W, Steffen W, Alonso MC, focal adhesion kinase. Nat Cell Biol 7: 581–590. Cross RA et al. (2002). Modulation of substrate adhesion Fernandez C, Figarella-Branger D, Girard N, Bouvier-Labit C, dynamics via microtubule targeting requires kinesin-1. J Cell Biol Gouvernet J, Paz Paredes A et al. (2003). Pilocytic astrocytomas 156: 349–359. in children: prognostic factors—a retrospective study of 80 cases. Krylyshkina O, Anderson KI, Kaverina I, Upmann I, Manstein DJ, Neurosurgery 53: 544–553. Small JV et al. (2003). Nanometer targeting of microtubules to focal Friedl P, Wolf K. (2003). Tumour-cell invasion and migration: adhesions. J Cell Biol 161: 853–859. diversity and escape mechanisms. Nat Rev Cancer 3: 362–374. Lauuffenburger DA, Horwitz AF. (1996). Cell migration: a physically Giese A, Loo MA, Tran N, Haskett D, Coons SW, Berens ME. (1996). integrated process. Cell 84: 359–369. Dichotomy of astrocytoma migration and proliferation. Int Louis DN, Pomeroy SL, Cairncross JG. (2002). Focus on central J Cancer 67: 275–282. nervous system neoplasia. Cancer Cell 1: 125–128.

Oncogene DRR regulates glioma invasion PU Le et al 4647 Palazzo AF, Gundersen GG. (2002). Microtubule-actin cross-talk at Wang L, Darling J, Zhang JS, Liu W, Qian J, Bostwick D et al. focal adhesions. Sci STKE 2002: PE31. (2000). Loss of expression of the DRR 1 gene at chromosomal Pedersen PH, Edvardsen K, Garcia-Cabrera I, Mahesparan R, segment 3p21.1 in renal cell carcinoma. Genes Chromosomes Cancer Thorsen J, Mathisen B et al. (1995). Migratory patterns of lac-z trans- 27: 1–10. fected human glioma cells in the rat brain. Int J Cancer 62: 767–771. Wang L, Rahn JJ, Lun X, Sun B, Kelly JJ, Weiss S et al. (2008). Ridley AJ, Schwartz MA, Burridge K, Firtel RA, Ginsberg MH, Gamma-secretase represents a therapeutic target for the treatment Borisy G et al. (2003). Cell migration: integrating signals from front of invasive glioma mediated by the p75 neurotrophin receptor. to back. Science 302: 1704–1709. PLoS Biol 6: e289. Rodriguez OC, Schaefer AW, Mandato CA, Forscher P, Bement WM, Webb DJ, Donais K, Whitmore LA, Thomas SM, Turner CE, Waterman-Storer CM. (2003). Conserved microtubule-actin interac- Parsons JT et al. (2004). FAK-Src signalling through paxillin, tions in cell movement and morphogenesis. Nat Cell Biol 7: 599–609. ERK and MLCK regulates adhesion disassembly. Nat Cell Biol 6: Salhia B, Tran NL, Symons M, Winkles JA, Rutka JT, Berens ME. 154–161. (2006). Molecular pathways triggering glioma cell invasion. Expert Wu X, Kodama A, Fuchs E. (2008). ACF7 regulates cytoskeletal-focal Rev Mol Diagn 6: 613–626. adhesion dynamics and migration and has ATPase activity. Cell van den Boom J, Wolter M, Blascke B, Knobbe CB, Reifenberger G. 135: 137–148. (2006). Identification of novel genes associated with astrocytoma Yamana N, Arakawa Y, Nishino T, Kurokawa K, Tanji M, Itoh RE progression using suppression subtractive hybridization and real- et al. (2006). The Rho-mDia1 pathway regulates cell polarity and time reverse transcription-polymerase chain reaction. Int J Cancer focal adhesion turnover in migrating cells through mobilizing Apc 119: 2330–2338. and c-Src. Mol Cell Biol 26: 6844–6858. Verkhovsky AB, Svitkina TM, Borisy GG. (1995). Myosin II filament Zijlstra A, Lewis J, Degryse B, Stuhlmann H, Quigley JP. (2008). The assemblies in the active lamella of fibroblasts: their morphogenesis inhibition of tumor cell intravasation and subsequent metastasis via and role in the formation of actin filament bundles. J Cell Biol 131: regulation of in vivo tumor cell motility by the tetraspanin CD151. 989–1002. Cancer Cell 13: 221–234.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene