Ephrina5 Acts As a Tumor Suppressor in Glioma by Negative Regulation of Epidermal Growth Factor Receptor

Home , CYR61, Ephrin

ORIGINAL ARTICLE EphrinA5 acts as a tumor suppressor in glioma by negative regulation of epidermal growth factor receptor

J-J Li1,2, D-P Liu1, G-T Liu3 and D Xie1

1Laboratory of Molecular Oncology, Institute for Nutritional Sciences, Shanghai Institutes of Biological Sciences, Shanghai, People’s Republic of China; 2Graduate School of the Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, People’s Republic of China and 3Cedars-Sinai Medical Center, Los Angeles, CA, USA

Eph receptors, the largest subfamily of receptor tyrosine among the receptor tyrosine kinases (RTKs) in that kinases, and their ephrin ligands play important roles in their endogenous ligands, the ephrins, are bound to nervous system development. Recently, they have been the surface of neighboring cells. Ephrins are divided implicated in tumorigenesis of different cancers. In this into two major classes, differing by their modes of study, we showed that the expression of ephrinA5 was attachment to plasma membrane (Gale et al., 1996). Class dramatically downregulated in primary gliomas compared A ephrins are tethered to the plasma membrane by virtue with normal tissues. Forced expression of ephrinA5 of a glycosylphosphatidylinositol anchor, whereas class B reduced tumorigenicity of human glioma U373 cells. ephrins are transmembrane proteins. All of the ephrins Epidermal growth factor receptor (EGFR), which fre- interact with specific Eph receptors, although promiscuity quently acts as an oncoprotein in glioma, was greatly exists as some ephrins bind to more than one Eph decreased in ephrinA5-transfected glioma cells, and the receptor. This receptor–ligand system is involved in a two molecules exhibited a mutually exclusive expression variety of biological events (Poliakov et al., 2004; pattern in primary glioma samples. We found that Huberman et al., 2005; Kuijper et al., 2007). Although ephrinA5 enhanced c-Cbl binding to EGFR, thus pro- a majority of investigators have studied the role of Eph/ moted ubiquitylation and degradation of the receptor. ephrin in development, emerging evidence suggests strong Either ephrinA5-Fc or EphA2-Fc treatment simulating involvement in tumorigenesis, including metastasis, inva- bidirectional signaling of Eph/ephrin system resulted in sion, and angiogenesis (Surawska et al., 2004; Brantley- EGFR decrease. This study discovered that ephrinA5 Sieders et al., 2006; Liu et al., 2007). acted as a tumor suppressor in glioma, and its negative Gliomas are the most common primary brain tumors regulation of EGFR contributed to the suppressive effects. with limited therapeutic options, high recurrence rate In addition to identifying a novel mechanism underlying and mortality. The epidermal growth factor receptor tumor suppressor activity of ephrinA5, we also showed (EGFR) and its ligands figure prominently in glioma cross-talk between different receptor tyrosine kinase pathogenesis, and the malignant nature of glioma is families in glioma. These findings may improve therapeu- partially due to the aberrance of EGFR, including gene tic strategies for glioma. amplification, mutation, abnormal overexpression, and Oncogene (2009) 28, 1759–1768; doi:10.1038/onc.2009.15; improper activation of the receptor and its downstream published online 9 March 2009 signaling (Nicholas et al., 2006). Considering the importance of RTKs in brain tumors, investigation into Keywords: ephrinA5; EGFR; c-Cbl; glioma RTK networks provides a better understanding of the biology of glioma. However, studies concerning the cross-talk between Eph/ephrin and ErbB families are rare. In this study, we showed that ephrinA5, a member belonging to ephrinA subclass, could negatively regulate Introduction EGFR, which contributed to its suppressive effect in glioma. A mutually exclusive expression pattern be- Eph/ephrin family members are expressed in adult tween EGFR and ephrinA5 was disclosed in both stable human tissues and in the developing nervous system cell lines and clinical tissue samples. We further revealed (Hafner et al., 2004). The Eph receptors are unique that ephrinA5 acted as a negative regulator of EGFR through promoting c-Cbl-mediated degradation. Im- Correspondence: Dr D Xie, Laboratory of Molecular Oncology, portantly, both ephrinA5-Fc and EphA2-Fc fusion Institute for Nutritional Sciences, Shanghai Institutes of Biological protein treatment led to decreased level of EGFR. Sciences, Chinese Academy of Sciences, 294 Tai-Yuan Road, Shanghai Therefore, the bidirectional signaling, which is char- 200031, People’s Republic of China. acteristic of the Eph/ephrin system, is involved in E-mail: [email protected] Received 7 July 2008; revised 17 November 2008; accepted 22 December ephrinA5-induced EGFR reduction. Thus, we have 2008; published online 9 March 2009 identified a novel mechanism of EGFR regulation in EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1760 glioma. Our findings may provide the assets for proliferation in MTT (3-(4,5-dimethylthiazol-2-Yl)-2,5- developing improved therapeutic strategies. diphenyltetrazolium bromide) assay (data not shown). In contrast, U373/ephrinA5 cells developed significantly fewer and smaller colonies in soft agar compared with Results U373/v cells (Figure 2a), indicating its inhibitory effect on anchorage-independent clonogenic growth. EphrinA5 was expressed at a low level in primary We further examined whether ephrinA5 would affect glioblastomas the tumorigenicity of glioma cells in vivo. U373/v and We examined the expression pattern of ephrinA5 in U373/ephrinA5 cells were subcutaneously injected into both quick-frozen brain samples (23 glioblastoma 8-week-old nude mice, and the resulting tumors were multiforme and 5 normal) and primary cells (derived measured at the end of 4 weeks. U373/ephrinA5 cells from 23 glioblastoma multiforme and 5 normal) by real- barely developed any observable tumors, whereas the time PCR. Four out of five normal tissues exhibited tumors derived from U373/v cells were easily visible and significantly higher ephrinA5 expression level compared measurable. A significant difference in tumor volume with tumor tissues (Figure 1a). The similar expression between the two groups was revealed by statistical pattern of ephrinA5 was also observed in primary cells analysis (Po0.05, Figure 2b). These results proved the (Figure 1b). We performed immunohistochemistry in suppressive effect of ephrinA5 in glioma. both normal and tumor tissues. As shown in Figure 1c, ephrinA5 protein was easily detected in normal tissues, EphrinA5 downregulated either the basal or the exhibiting pronounced membrane staining. In contrast, phosphorylation level of EGFR and impaired EGFR it was almost undetectable in tumor tissues. Down- signaling in glioma cells regulation of ephrinA5 at both mRNA and protein level To investigate the underlying mechanism for reduced in glioma suggested its role as a tumor suppressor. tumorigenicity of glioma cells by ephrinA5, we examined the expression of several important molecules Forced expression of ephrinA5 in glioma cells reduced involved in tumorigenesis and development of glioma. their clonogenic growth in soft agar and tumorigenicity We found that the expression level of EGFR was in nude mice significantly decreased in U373/ephrinA5 cells To examine the potential suppressive effects of ephrinA5 (Figure 2c). The significant decrease of EGFR was also in glioma, we established stable cell lines using human observed in both U343 and U373 cells infected with a glioma U373 cells. Two identified clones exhibiting virus vector overexpressing ephrinA5 (Figure 2d). On different levels of ephrinA5 were selected for further the basis of this data, we investigated the expression studies (U373/ephrinA5L and U373/ephrinA5H, pattern of ephrinA5 and EGFR in glioma tissue Figure 2c). EphrinA5 showed little effect on cell samples. As shown in Figure 3a, the expression of

Figure 1 Quantitative expression and Immunohistochemistry of ephrinA5 in human clinical samples. (a) Real-time PCR examination of ephrinA5 expression in normal brain and primary glioma tissue samples. (b) Quantitative expression of ephrinA5 in primary cells derived from normal brain and glioma tissues. T, tumor; N, normal. (c) Immunohistochemistry displays clear ephrinA5 signals (brown) in normal brain tissue samples (N); however, it is almost undetectable in glioma tissues (T).

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1761

Figure 2 Forced expression of ephrinA5 in U373 cells, soft agar assay and tumorigenicity assay. (a) Soft agar assay: A total of 1.0 Â 103 cells per well were seeded in soft agar for 20 days, and colonies were enumerated. Each experiment was performed in triplicate, and results represented the mean±s.d. of three experiments. (b) Tumorigenicity assay. Either 1.0 Â 106 U373/ephrinA5 cells or U373/v cells were subcutaneously injected on opposite ﬂanks of nude mice (three nude mice per group), respectively. After 4 weeks, resulting tumors were removed and the volumes were measured. Statistical signiﬁcance was determined with a Student’s t test (bars, s.d.; *Po0.05 versus U373/v). (c) Expression of ephrinA5 and several other genes in U373/v and U373/ephrinA5 cells were examined by western blot. (d) Expression of both basal and phosphorylation level of epidermal growth factor receptor (EGFR) was examined in U343 and U373 cells, which was infected with either control (U343 con, U373 con) or virus vector overexpressing ephrinA5 (U343/ ephrinA5, U373/ephrinA5). The same examination was also conducted in U373/v and U373/ephrinA5H cells.

Figure 3 Mutually exclusive pattern of epidermal growth factor receptor (EGFR) and ephrinA5 in glioma tissue samples, and reduced response to epidermal growth factor (EGF) stimulation in U373/ephrinA5 cells. (a) Expressions of EGFR and ephrinA5 in 18 glioma samples examined by western blot. Blot densities were quantified by densitometry using Scion Image beta 4.03. Statistical analysis was carried out with Pearson correlation test. (b) Impaired response to EGF stimulation in U373/ephrinA5 cells. Cells were plated in a 24-well plate at a density of 80% confluence. After being starved overnight, cells were stimulated with 20 ng/ml EGF. Protein samples were collected at the indicated time points, and the level of ERK1/2 and phospho-ERK1/2 (p-ERK1/2) (pY202/pY204) was examined by western blot. ephrinA5 and EGFR was strikingly different from each (Pearson correlation coefficient ¼À0.559, P ¼ 0.016). other in 18 clinical samples, and the two molecules These results suggested potential negative regulation of exhibited a mutually exclusive expression pattern. The EGFR exerted by ephrinA5. negative correlation between ephrinA5 and EGFR We further examined whether EGFR signaling was expression was revealed by Pearson correlation analysis altered in ephrinA5-overexpressing glioma cells. The

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1762 tyrosine phosphorylation level of EGFR, which indicates its activation status, was downregulated in both virus-infected cells and stable clone cells (Figure 2d). Furthermore, we observed a mild decrease of both total and phosphorylated stat3 (pY705) in U373/ephrinA5 cells (Figure 2c). Ligand-bound EGFR can recruit and activate stat3 by phosphorylation, thus the decreased phosphorylation level of both stat3 (pY705) and EGFR itself suggested the impaired EGFR signaling. To test this hypothesis, we examined the activation of Ras–Raf– MEK—ERK (extracellular signal-regulated kinase) pathway upon epidermal growth factor (EGF) stimulation, which is a key pathway downstream of EGFR. As shown in Figure 3b, ERK phosphorylation triggered by EGF stimulation was obvious in both U373/v and U373/ephrinA5 cells at 10 min. Whereas the level of phospho-ERK was comparable between U373/v and U373/ephrinA5L cells, it was much lower in U373/ ephrinA5H cell at this time point. Furthermore, the level of ERK (pT202/pY204) declined more rapidly in both U373/ephrinA5 cells compared with U373/v cells during the treatment. At 120 min, phospho-ERK was quite weak in U373/ephrinA5 cells, whereas it was still readily detectable in control cell. Thus, the response to EGF stimulation was actually impaired in U373/ephrinA5 cells, consistent with the decreased EGFR level. The close correlation between the response to EGF stimulation and the expression level of EGFR has been strongly Figure 4 EphrinA5 downregualted epidermal growth factor proved in EGFR-knockdown U373 cells (Supplemen- receptor (EGFR) by enhancing c-Cbl association with EGFR tary Figure 1b and c). and the resulting EGFR ubiquitylation. (a) HEK-293T (human embryonic kidney) cells were cotransfected with ephrinA5 expression construct at different doses and wild-type EGFR (EGFR WT) or mutant EGFR (Y1045F). Protein samples were collected 24 h EphrinA5-facilitated EGFR degradation by promoting after transfection and analysed with indicated antibodies. (b) HEK- c-Cbl binding and ubiquitylation 293T cells co-expressing EGFR and empty vector/ephrinA5 were Ubiquitylation of EGFR and subsequent degradation is extracted and analysed either directly or following IP (immuno- a well-known mechanism regulating its level. Thus, we precipitation) with the indicated antibodies. Ub, ubiquitin. examined the effect of ephrinA5 on EGFR stability. We used human embryonic kidney (HEK-293T) cells instead of U373 cells because of the relatively low level association of c-Cbl with EGFR, thus enhancing of EGFR and much lower transfection efficiency in the ubiquitylation and degradation of this receptor. latter. When ectopically expressed in HEK-293T cells, ephrinA5 resulted in decreased level of EGFR, reminis- EphrinA5-Fc treatment-induced EGFR decrease cent of what we observed in either stable cell lines or and sustained EGF-mediated EGFR degradation clinical samples (Figure 4a). Furthermore, the suppres- Bidirectional signaling in ‘Eph/ephrin’ system is respon- sion of EGFR by ephrinA5 was in a dose-dependent sible for a variety of biological events. The signaling manner (Figure 4a). Interestingly, one EGFR mutant pathway downstream of the transmembrane Eph (Y1045F) showed resistance to downregulation by RTKs is referred to as Eph forward signaling, whereas ephrinA5 (Figure 4a). The 1045 tyrosine of EGFR that downstream of ephrins is referred to as ephrin provides a docking site for c-Cbl, which sorts the reverse signaling. We first investigated whether the receptors for degradation by tagging them with ubiqui- forward signaling was involved in ephrinA5-mediated tin (Oksvold et al., 2003). We assumed that ephrinA5 EGFR degradation. Several EphA receptors are ex- might influence c-Cbl binding to EGFR. To test this pressed in U373 parental cells, such as EphA2 hypothesis, EGFR was immunoprecipitated in the and EphA4, which decrease in U373/ephrinA5 cells absence or presence of ephrinA5 in HEK-293T cells, (Supplementary Figure 1a). We employed ephrinA5-Fc and then the ubiquitylation and c-Cbl association was fusion protein to mimic forward signaling activation examined. As expected, ubiquitylation of EGFR was (Supplementary Figure 2a). Treating U373 cells with significantly enhanced when ephrinA5 was present ephrinA5-Fc resulted in decreased EGFR level (Figure 4b). Consistent with the enhanced ubiquityla- (Figure 5a). However, the extent of EGFR decrease tion, more c-Cbl bound to EGFR when ephrinA5 was induced by ephrinA5-Fc was milder than that in co-expressed (Figure 4b). This result strongly supported HEK-293T cells (Figure 5b), possibly because of the our hypothesis that ephrinA5 could promote the relatively lower basal and phosphorylation level of

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1763

Figure 5 EphrinA5-Fc-induced epidermal growth factor receptor (EGFR) decrease and sustained epidermal growth factor (EGF)- triggered EGFR degradation. (a) U373 cells were plated in a 24-well plate at the density of 80% confluence. After being starved overnight, cells were incubated with control-Fc or ephrinA5-Fc at a final concentration of 2 mg/ml for 15 min on ice, then the cells were transferred to 37 1C. Proteins were extracted at different time points indicated for further western blot analysis. (b) The same treatment described in (a) was conducted in 293T cells overexpressing EGFR and EphA2. (c) Co-stimulation with EGF and Fc recombinant proteins on U373 cells. U373 cells were plated on cover slides. Serum-starved cells were stimulated with a combination of EGF (50 ng/ ml) and Fc recombinant proteins (2 mg/ml). The samples were fixed at different time points and localization of EGFR and ephrinA5-Fc were examined by immunofluorescence (Green, ephrinA5-Fc; Red, EGFR). (d) In the treatment described in (c) membrane EGFR (120 min) was isolated by phase separation method using Triton X-114, and subjected to immunoprecitation analysis. IP, immunoprecipitation. endogeneous EGFR in U373 cells. The time-course cells, whereas it remained undetectable in ephrinA5-Fc- assay showed that decrease of EGFR became more treated cells. This difference was further proved by pronounced at later time points, such as 120 min, the phase separation of membrane EGFR (120 min) suggesting the accumulative effect of ephrinA5-Fc- (Figure 5d). The above results suggested that the induced EGFR reduction. forward signaling activated by ephrinA5 could prolong To further investigate the effect of ephrinA5-Fc on EGF-induced EGFR degradation and prevent mem- ligand-activated EGFR, we treated U373 cells with a brane EGFR recovery. Thus, the Eph forward signaling combination of EGF and either control or ephrinA5-Fc, should contribute to the negative regulation of EGFR and examined the degradation of EGFR by immuno- by ephrinA5. fluorescence. As shown in Figure 5c, there was no difference between the control-Fc and ephrinA5-Fc treatment at the beginning (1 min). Endocytic vesicles EphA2-Fc treatment also reduced EGFR in either bearing EGFR signals were observed in differently HEK-293T cells or U373/ephrinA5 cells treated cells. The green signals representing ephrinA5- Both EGFR and ephrinA5 were reported to reside in a Fc exhibited a typical membrane distribution, suggest- specific microdomain on the membrane in which various ing its ligation to the cognate receptors. However, at downstream signals could be transduced. Although 30 min, no detectable EGFR signals appeared on the ephrinA5 is a glycosylphosphatidylinositol-linked mem- membrane of ephrinA5-Fc-treated cells, whereas membrane protein, it is also able to elicit intracellular brane EGFR staining in control cells was visible despite signaling transduction, such as activation of Src family being weak. Figure 5c also showed that by 30 min, kinases (Davy et al., 1999). We supposed that the ephrin ephrinA5-Fc aggregated around the nuclei, indicating its reverse signaling also participated in the downregulation lysosomal degradation. The difference in membrane of EGFR, considering the presence of endogeneous EGFR level extended to as long as 120 min. At this time EphA receptors for ephrinA5. Firstly, we examined the point, both of the cells showed perinuclear EGFR localization of EGFR and ephrinA5 in U373/ephrinA5 signals, indicating degraded receptors. However, mem- cells. As expected, the two molecules showed nearly brane EGFR recovered, at least partially, in control identical colocalization on the membrane.

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1764

Figure 6 (a) Cellular localization of epidermal growth factor receptor (EGFR) and ephrinA5 were examined by immunoﬂuorescence in U373/ephirnA5 cells (Green, EGFR; Red, ephrinA5). (b) HEK-293T (human embryonic kidney) cells either co-expressing EGFR and ephrinA5 or expressing ephrinA5 alone were treated with control Fc or EphA2-Fc (2 mg/ml). Proteins were collected at indicated time points and analysed by western blot. (c) Either U373/ephrinA5 cells or U373 cells were treated with either control Fc or EphA2-Fc (2 mg/ml). Proteins were collected at the indicated time points for western blot examination.

Owing to the rather low ephrinA5 expression level Discussion in all the available glioma cell lines (data not shown), we instead employed HEK-293T cells to examine the A majority of earlier studies focused on the role of effect of ephrin reverse signaling on EGFR. HEK-293T ephrinA5 in developmental biology, especially the cells co-overexpressing EGFR and ephrinA5 nervous system (Bolz et al., 2004; Otal et al., 2006). were treated with EphA2-Fc to activate the reverse Recently, the role of ephrinA5 in cancer has gained signaling. As shown in Figure 6b, EphA2-Fc caused attention, and the expression pattern of ephrinA5 has reduced level of EGFR compared with control Fc, been examined in several types of cancers, such as and was associated with decreased ephrinA5 and ovarian cancer (Herath et al., 2006) and osteosarcoma increased Fyn (pY416) (Supplementary Figure 2b), (Varelias et al., 2002). In this study, we showed that indicating the activation of ephrinA5 reverse signaling. ephrinA5 played as a suppressor in glioma through its We performed the same treatment on cells expressing negative regulation on EGFR, which is frequently EGFR alone to preclude the possibility that decreased aberrant in glioma. level of EGFR was due to other endogenous ligands for Dysregulation of EGFR signaling contributes a large EphA2 in HEK-293T cells other than ephrinA5. EGFR part to the development and malignant nature of high- expression showed little change upon EphA2-Fc treat- grade gliomas (Halatsch et al., 2006). Thus, regulation of ment in the absence of ephrinA5, proving that the this receptor has been largely investigated to ﬁnd suppressive effect on EGFR was mainly mediated indications for clinical therapy. Ubiquitylation of EGFR through ephrinA5. and subsequent degradation controls the intensity of The same treatment was also conducted on U373/ downstream signals after receptor activation. c-Cbl, a ephrinA5 cells, using U373 parental cells as control, ubiquitin E3 ligase, is recruited to EGFR upon ligand and similar results were observed (Figure 6c). However, stimulation, and then initiates the ubiquitylation of the the change of EGFR was less dramatic than that in activated receptor (Joazeiro et al., 1999; Levkowitz et al., HEK-293T cells, possibly due to the relative lower 1999). The ubiquitylated EGFR is internalized into early expression level of both EGFR and ephrinA5 in U373/ endosomes, from where it is either recycled to plasma ephrinA5 cells. These results suggested that the ephrin membrane, or delivered to multivesicular bodies and late reverse signaling as well as the forward signaling endosomes for degradation (Waterman and Yarden, participated in the downregulation of EGFR by 2001; Ravid et al., 2004). In our study, we reported for ephrinA5. the ﬁrst time that ephrinA5 could enhance c-Cbl

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1765 association with EGFR and subsequent ubiquitylation of above, it is possible that activation of Src family kinases the receptor, and thereby promote EGFR degradation. through ephrinA5 reverse signaling, combined with the Owing to various mitogens contained in the cell culture stimulation of EphA2 forward signaling, can facilitate medium, increased EGFR-associated c-Cbl and enhanced the activation and recruitment of c-Cbl to EGFR receptor degradation could be detected even without synergistically, and thus enhance the subsequent EGFR additional EGF stimulation. Owing to the rather low degradation. Therefore, the negative regulation of EGFR level, it was impossible to perform the immuno- EGFR by ephrinA5 is an integrated event combining precipitation assay in U373/ephrinA5 cells. However, this multiplex signalings and effects. mechanism should contribute to decreased EGFR level in In addition to protein stability, EGFR can also be U373/ephrinA5 cells. regulated at the transcriptional level. It was reported EphrinA5 is a glycosylphosphatidylinositol-linked that expression of the human EGFR gene could be protein tethered to the plasma membrane. This raised regulated by ligand-activated thyroid hormone receptor the question, how it can influence the interactions (Xu et al., 1993) and estrogen (Yarden et al., 2001). We between c-Cbl and EGFR in the cytoplasmic region to also examined the mRNA level of EGFR in transfected negatively regulate EGFR. The forward and reverse U373 cells, and found it decreased by more than 50% in bidirectional signal transduction distinguishes Eph/ the U373/ephrinA5 cells compared with U373/v cells ephrin family from other RTK families. In this study, (Supplementary Figure 3a). EGFR promoter reporter ephrinA5-Fc and EphA2-Fc treatments were used to assay showed that ephrinA5 could suppress the promo- mimic the activation of the forward and reverse signals, ter activity in a dose-dependent manner (Supplementary respectively. Several EphA receptors were expressed in Figure 3b). However, this suppression was modest, U373 cells, and could be activated after ephrinA5-Fc which could not fully interpret the dramatic decrease in stimulation, accompanied with elicitation of the forward the level of EGFR in stable cell lines. Taken together, signal (Supplementary Figure 1a, 2a and c). c-Cbl has ephrinA5 can negatively regulate EGFR at multilevels. been reported to participate in EphA receptor degrada- Oncogenic dysregulation of RTK signals usually tion induced by ligand stimulation (Walker-Daniels involves the inactivation of negatively regulating tumor et al., 2002; Wang et al., 2002; Sharfe et al., 2003). suppressor gene products, which is the case of ephrinA5 Larsen et al., 2007 showed that other than transcrip- in our study. Our results suggest that glioma cells tional target of EGFR, EphA2 colocalized with possibly silence ephrinA5 expression to avoid its activated EGFR. Thus, recruitment of c-Cbl by suppressive effect on EGFR. The downregulation of activated EphA receptor upon ligand binding poten- EGFR by ephrinA5 also provides new insights into the tially facilitates its access to EGFR, and promotes intricate interaction between the signaling pathways of subsequent ubiquitylation of the receptor, which could ErbB and Eph RTK families. Aberrant expression and interpret the potential role of Eph forward signaling. activation of EGFR represents one of the major Glycosylphosphatidylinositol-anchored ephrinA5 can molecular alterations resulting in malignancy in glioma. induce compartmentalized signaling within a membrane Although a variety of agents targeting EGFR have been microdomain when bound to its cognate Eph receptor. developed, their antitumor effects are often not as This signaling seemed to require the activity of the Fyn optimal as predicted from preclinical studies. Their protein tyrosine kinase (Davy et al., 1999). We showed modest activity is possibly due to the complex modula- that ephrinA5 colocalized with EGFR in certain regions tion of strongly enhanced EGFR signaling pathway in on the plasma membrane. As phosphorylation by Fyn glioma. Further understanding of the negative regula- has been reported to regulate the E3 ligase activity of tion on EGFR by ephrinA5 will expand our knowledge c-Cbl (Grossmann et al., 2004; Li et al., 2007), we of the molecular pathogenesis of glioma. Moreover, our hypothesize that the ephrinA5 reverse signaling trig- work indicates that therapeutic intervention using gered by EphA2-Fc would activate Fyn, which could existing anti-EGFR agents combined with ephrinA5- phosphorylate c-Cbl, and facilitate c-Cbl-mediated related signal activation will potentially improve the EGFR degradation. drug efficacy. In transiently transfected HEK-293T cells, parental U373 cells and U373/ephrinA5 cells, EGFR was Materials and methods attenuated in response to ephrinA5-Fc or EphA2-Fc treatment. However, the reduction of EGFR was less Tissue samples and cell culture significant in U373 or U373/ephrinA5 cells when Forty-six brain tumor samples, all of which are glioblastoma compared with HEK-293T cells. This might be due to multiforme, and five normal brain samples were obtained from the much higher expression level of extrogenous EGFR Cedars-Sinai Medical Center, UCLA School of Medicine at and ephrinA5 in transfected HEK-293T cells, which the time of surgery after the pathology specimens had been possibly enhanced the intensity of signals, and led to obtained. Part of each tissue sample was quick frozen in liquid 1 more pronounced phenotype. As shown in Supplemen- nitrogen and stored at À70 C until they were processed, whereas the remaining tissue was used for primary cell culture. tary Figure 2c, in U373/ephrinA5 cells, despite reduced Our work was approved by the Institutional Review Board of membrane expression of EphA2, its localization the Institute for Nutritional Sciences, Chinese Academy of exhibited obvious overlap with ephrinA5 in the same Sciences. HEK-293T cells, HeLa cells, U343 and U373 glioma cell. Their colocalization indicated potential cooperation cells, were maintained in Dulbecco’s modified Eagle’s on the suppression of EGFR expression. As mentioned medium (GIBCO, GrandIsland, NY, USA) supplemented

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1766 with 10% fetal bovine serum (Hyclone, Logan, UT, USA) at from Dr De Camilli P; CD19-Fc and EphA2-Fc expression 37 1C in a humidiﬁed atmosphere containing 5% CO2. vectors from Dr Hans-Christian Aasheim (Finne et al., 2004).

Antibodies and reagents Luciferase reporter assay Anti-ephrinA5 antibody and recombinant human EGF were HeLa cells were plated at subconfluent density in 24-well purchased from R&D Systems (Minneapolis, MN, USA). dishes and cotransfected with a total of 0.5 mg plasmids, Anti-EGFR, anti-Stat3, anti-b-actin, anti-a-tubulin and anti- including 0.1 mg reporter plasmid, different amounts of empty Ubiquitin antibodies were obtained from Santa Cruz Biotech- or expression vector and 0.05 mg Renilla luciferase pRL-TK nology Inc. (Santa Cruz, CA, USA). Anti-EphA4 and PY20 plasmid. Cell lysates were prepared 24 h after transfection, and antibodies were from BD Transduction Laboratories (San reporter activity was measured using the Dual-Luciferase Diego, CA, USA). Anti-Fyn and anti-EphA2 antibodies were Reporter Assay System (Promega, Madison, WI, USA). from Chemicon (Temecula, CA, USA) and Upstate (Lake Transfections were performed in triplicate, and repeated thrice Placid, NY, USA), respectively. Anti-ERK1/2, anti-phospho- to ensure reproducibility. ERK1/2 (pT202/pY204), anti-phospho-Stat3 (pY705), anti- phospho-Src family (Tyr416) and horseradish peroxidase- Immunofluorescence microscopy conjugated secondary antibodies were purchased from Cell Cells (U373, U373/v or U373/ephrinA5) were digested with Signaling Technology Inc. (Danvers, MA, USA). The second- trypsin and plated on cover slides. On the next day, the cells ary antibodies for immunofluorescence were from Molecular were washed thrice with phosphate buffered saline (PBS) at Probe Inc. (Eugene, OR, USA). room temperature, and fixed in 4% formaldehyde on ice for 30 min. Cells were blocked with 2% bovine serum albumin at RNA preparation and Real–Time PCR room temperature for 1 h. Primary antibody incubations were Preparation and construction of RNA and cDNA preparation performed overnight at 4 1C (ephrinA5, 1:100; EGFR, 1:100; were done as described earlier (Xie et al., 2004). The primer EphA2, 1:200). After extensive washing with PBS, secondary pairs used for real-time PCR are as follows: ephrinA5: F: antibody (Alexa Fluor 488-conjugated donkey anti-rabbit, 50-GCAATCCCAGATAATGGAAGAA-30,R:50-TCGGCT mouse, goat IgG and Alexa Fluor 555-conjugated goat anti- GACTCATGTACGGT-30; EGFR: F: 50-CGGGACATAGT mouse, rabbit and donkey anti-goat IgG, 1:1000) incubations CAGCAGTG-30,R:50-GCTGGGCACAGATGATTTTG-30; were performed at room temperature for 1 h. Samples were b-actin: F: 50-GATCATTGCTCCTCCTGAGC-30,R:50- washed again with PBS and then mounted in Mowiol 4–88 ACTCCTGCTTGCTGATCCAC-30. Amplification reactions (Sigma-Aldrich Inc., St Louis, MO, USA). Fluorescence was were performed as described earlier (Deng et al., 2007). monitored by inverted confocal laser microscopy (Carl Zeiss, NY, USA). Plasmids construction and cell transfection Protein extraction with Triton X-114 To generate the ephrinA5 expression vector, the open reading frame of human ephrinA5 cDNA was cloned into the U373 cells treated with Fc fragments were lysed with Triton eukaryotic expression vector pcDNA3.1 (Invitrogen, Carls- X-114 extraction buffer (10 mM Tris (pH 7.4), 150 mM NaCl, bad, CA, USA) and was fused to a C-terminal myc tag. The 1% Triton X-114). The proteins were subjected to phase expression construct and empty pcDNA3.1 vector were separation as described earlier (Bordier, 1981). Briefly after transfected into U373 cells, respectively, using Lipofectamine removal of the particulate matter by centrifugation, the sample 1 2000 Reagent (Invitrogen). The transfected cells were selected was layered over a 6% (w/v) sucrose cushion, warmed to 30 C in the presence of 800 mg/ml G418 and resistant clones were to separate the phases and centrifuged at 300 g at room further confirmed by western blot analysis. A modified temperature. The detergent phase (an oily droplet at the bottom lentivirus-based FG12 vector with a cytomegalovirus promoter of the tube) was diluted and analysed by Immunoprecipitation. (Sun et al., 2008) was also used to overexpress human ephrinA5. HEK-293T cells were transfected using the calcium Immunohistochemistry phosphate precipitation method, and HeLa cells were trans- Glioblastoma multiforme and normal brain tissues were fixed in fected using FuGENE6 Transfection Reagent (Roche Applied formalin,embeddedinparaffin,and10mm-thick consecutive Sciences, Penzberg, Germany). EphrinA5-Fc fusion protein sections were cut and mounted on glass slides. The sections were was generated as described earlier (Munthe et al., 2000), and fixed in ice-cold acetone for 30 min, washed thrice in 0.01 mol/l the primer pair for the construct was: F: 50-CATAAGCTTC PBS (8 mmol/l Na2HPO4, 2 mmol/l NaH2PO4 and 150 mmol/l CACCATGTTGCACGTGGAGATGTT-30,R:50-ATCGGAT NaCl) for 5 min each, blocked for 1 h in 0.01 mol/l PBS CCTGACTCATGTACGGTGTC-30. supplemented with 0.3% Triton X-100 and 5% normal goat To knockdown the endogenous EGFR, three small inter- serum, followed by addition of ephrinA5 (1:100) antibody at fering RNA sequences against different sites of hEGFR 4 1C overnight. After brief washes in 0.01 mol/l PBS, sections mRNA were designed using the online software ‘siRNA were exposed for 2 h to 0.01 mol/l PBS containing horseradish Target Finder’ (Ambion Inc., Austin, TX, USA): (1) peroxidase-conjugated rabbit anti-goat IgG (1:500), followed by 0 GGCACGAGTAACAAGCTCA; (2) GAGAAAGAATAC development with 0.003% H2O2 and 0.03% 3,3 -diaminobenzi- CATGCAG; (3) GTGCTGGATGATAGACGCA. Oligonu- dine in 0.05 mol/l Tris–HCl (pH 7.6). Immunohistochemistry for cleotide with a scrambled sequence was used as a negative each sample was performed at least three separate times, and all control. RNA interference constructs were cloned and sections were counterstained with hematoxylin. Negative con- produced as described earlier (Qin et al., 2003). trols consisted of substitution of the primary antibody with normal goat serum at the same dilution. The following constructs were kindly provided: EGFR promoter reporter plasmid (pER-Luc) from Dr Western blot analysis Alfred C. Johnson (Liu et al., 2000); Expression constructs for Cultured cells were washed twice with PBS and lysed in hEGFR and hEGFR (Y1045F) mutant were from Dr Yosef radioimmunoprecipitation assay buffer for 15 min on ice. Yarden (Levkowitz et al., 1999); pcDNA3.1-HA-Ubiquitin Glioma tissues were also extracted with radioimmunoprecipi-

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1767 tation assay buffer. Cell lysates were clarified by centrifugation 1.0 Â 103 cells per well in a volume of 400 ml per well. After 20 at 10 000 g for 15 min, and protein concentration was days of incubation, the colonies were counted and measured. determined by the Bradford Reagent (Rio-Rad Laboratories, All of the experiments were done at least thrice using triplicate Hercules, CA, USA). Lysates were separated on either 10% or plates per experimental point. 8% sodium dodecyl sulfate polyacrylamide gel electrophoresis, proteins were then transferred to Immobilon membrane Tumorigenicity assay (Millipore, Bedford, MA, USA) and immunoblotted with Six 8-week-old nude mice were separated into two groups. indicated antibodies. All immunoblots were visualized by The experimental group received 1.0 Â 106 U373/ephrinA5 enhanced chemiluminescene (Pierce, Rockford, IL, USA). The cells subcutaneously injected on opposite flanks; the control band intensity was quantified by arithmetic analysis using the group received subcutaneously lateral blank injection of software Scion Image beta 4.03. The ratios of EGFR/tubulin 1.0 Â 106 U373/v cells. Tumors were harvested 4 weeks after expression between different lanes were marked in the form of injection and individually measured. Tumor volumes (mm3) relative values under the blots. Statistical analysis was carried were calculated using the standard formula: length Â out with Pearson correlation test. width Â height Â 0.5326. Statistical analysis was performed using the Student’s t test. Immunoprecipitation Cells were washed with ice-cold PBS and lysed in Tris-buffered saline (pH 7.4), containing 50 mM Tris, 150 mM NaCl, 1% Acknowledgements NP-40, 1 mM EDTA, 1 mM Na3VO4,10mM NaF, 2.5 mg/ml aprotinin and leupeptin, 1 mM b-glycerophosphate and AEBSF We thank Dr Alfred C Johnson for the EGFR promoter (4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride), and reporter plasmid, Dr Yosef Yarden for the hEGFR and 10 mM iodoacetate. Lysates were incubated on ice for 15 min hEGFR (Y1045F) plasmids, Dr De Camilli P for pcDNA3.1- before cellular debris and nuclei were removed by centrifugation HA-Ubiquitin plasmid and Dr Hans-Christian Aasheim for at 5000 g for 5 min. Cell lysates were incubated with anti-EGFR CD19-Fc and EphA2-Fc constructs. antibody overnight at 4 1C. Protein A–Sepharose (Amersham We also would like to thank Dr H Phillip Koeffler (Cedars- Biosciences, Piscataway, NJ, USA) beads in a 50:50 mixture in Sinai Medical Center) for helpful discussions, suggestions and 50 mM Tris buffer, pH 7.0, were added, and further incubated critical reading of this paper. for another 4 h at 4 1C. The immunoprecipitates were washed This work was supported by Ministry of Science and thrice in Tris-buffered saline and boiled for 5 min in 100 ml Technology 863 Program 2007AA02Z474, 973 Program Laemmli buffer containing 0.02% blue bromophenol and 2% b- 2007CB914704, and 2008ZX10207, National Natural Science mercaptoethanol. Funds for Distinguished Young Scholar 30725010, National Natural Science Foundation of China 90813023 and 30528003. Soft agar assay Chinese Academy of Sciences Grant KSCX2-YW-R-152 and For clonogenic assay, cells were plated into 24-well flat- KSCX-YW-R-73, and Science and Technology Commission of bottomed dishes using a two-layer soft agar system with Shanghai Municipality 08140902300 to D Xie.

References

Bolz J, Uziel D, Muhlfriedel S, Gullmar A, Peuckert C, Zarbalis K determine the phosphorylation sites within c-Cbl. FEBS Lett 577: et al. (2004). Multiple roles of ephrins during the formation 555–562. of thalamocortical projections: maps and more. J Neurobiol 59: Hafner C, Schmitz G, Meyer S, Bataille F, Hau P, Langmann T et al. 82–94. (2004). Differential gene expression of Eph receptors and ephrins in Bordier C. (1981). Phase separation of integral membrane proteins in benign human tissues and cancers. Clin Chem 50: 490–499. Triton X-114 solution. J Biol Chem 256: 1604–1607. Halatsch ME, Schmidt U, Behnke-Mursch J, Unterberg A, Wirtz CR. Brantley-Sieders DM, Fang WB, Hwang Y, Hicks D, Chen J. (2006). (2006). Epidermal growth factor receptor inhibition for the Ephrin-A1 facilitates mammary tumor metastasis through an treatment of glioblastoma multiforme and other malignant brain angiogenesis-dependent mechanism mediated by EphA receptor tumours. Cancer Treat Rev 32: 74–89. and vascular endothelial growth factor in mice. Cancer Res 66: Herath NI, Spanevello MD, Sabesan S, Newton T, Cummings M, 10315–10324. Duffy S et al. (2006). Over-expression of Eph and ephrin genes in Davy A, Gale NW, Murray EW, Klinghoffer RA, Soriano P, advanced ovarian cancer: ephrin gene expression correlates with Feuerstein C et al. (1999). Compartmentalized signaling by GPI- shortened survival. BMC Cancer 6: 144. anchored ephrin-A5 requires the Fyn tyrosine kinase to regulate Huberman AD, Murray KD, Warland DK, Feldheim DA, Chapman cellular adhesion. Genes Dev 13: 3125–3135. B. (2005). Ephrin-As mediate targeting of eye-specific projections to Deng YZ, Chen PP, Wang Y, Yin D, Koeffler HP, Li B et al. (2007). the lateral geniculate nucleus. Nat Neurosci 8: 1013–1021. Connective tissue growth factor is overexpressed in esophageal Joazeiro CA, Wing SS, Huang H, Leverson JD, Hunter T, Liu YC. squamous cell carcinoma and promotes tumorigenicity through (1999). The tyrosine kinase negative regulator c-Cbl as a RING- beta-catenin-T-cell factor/Lef signaling. J Biol Chem 282: type, E2-dependent ubiquitin-protein ligase. Science 286: 309–312. 36571–36581. Kuijper S, Turner CJ, Adams RH. (2007). Regulation of angiogenesis Finne EF, Munthe E, Aasheim HC. (2004). A new ephrin-A1 isoform by Eph-ephrin interactions. Trends Cardiovasc Med 17: 145–151. (ephrin-A1b) with altered receptor binding properties abrogates the Larsen AB, Pedersen MW, Stockhausen MT, Grandal MV, van Deurs cleavage of ephrin-A1a. Biochem J 379: 39–46. B, Poulsen HS. (2007). Activation of the EGFR gene target EphA2 Gale NW, Holland SJ, Valenzuela DM, Flenniken A, Pan L, Ryan TE inhibits epidermal growth factor-induced cancer cell motility. Mol et al. (1996). Eph receptors and ligands comprise two major Cancer Res 5: 283–293. specificity subclasses and are reciprocally compartmentalized during Levkowitz G, Waterman H, Ettenberg SA, Katz M, Tsygankov AY, embryogenesis. Neuron 17: 9–19. Alroy I et al. (1999). Ubiquitin ligase activity and tyrosine Grossmann AH, Kolibaba KS, Willis SG, Corbin AS, Langdon WS, phosphorylation underlie suppression of growth factor signaling Deininger MW et al. (2004). Catalytic domains of tyrosine kinases by c-Cbl/Sli-1. Mol Cell 4: 1029–1040.

Oncogene EphrinA5 suppresses tumorigenesis in glioma J-J Li et al 1768 Li Z, Dong T, Proschel C, Noble M. (2007). Chemically diverse Ravid T, Heidinger JM, Gee P, Khan EM, Goldkorn T. (2004). toxicants converge on Fyn and c-Cbl to disrupt precursor cell c-Cbl-mediated ubiquitinylation is required for epidermal growth factor function. PLoS Biol 5: e35. receptor exit from the early endosomes. JBiolChem279: 37153–37162. Liu DP, Wang Y, Koeffler HP, Xie D.. (2007). Ephrin-A1 is a negative Sharfe N, Freywald A, Toro A, Roifman CM. (2003). Ephrin-A1 regulator in glioma through down-regulation of EphA2 and FAK. induces c-Cbl phosphorylation and EphA receptor down-regulation Int J Oncol 30: 865–871. in T cells. J Immunol 170: 6024–6032. Liu XW, Katagiri Y, Jiang H, Gong LJ, Guo LY, Shibutani M et al. Sun ZJ, Wang Y, Cai Z, Chen PP, Tong XJ, Xie D. (2008). (2000). Cloning and characterization of the promoter region of the Involvement of Cyr61 in growth, migration, and metastasis of rat epidermal growth factor receptor gene and its transcriptional prostate cancer cells. Br J Cancer 99: 1656–1667. regulation by nerve growth factor in PC12 cells. J Biol Chem 275: Surawska H, Ma PC, Salgia R. (2004). The role of ephrins and Eph 7280–7288. receptors in cancer. Cytokine Growth Factor Rev 15: 419–433. Munthe E, Rian E, Holien T, Rasmussen A, Levy FO, Aasheim H. Varelias A, Koblar SA, Cowled PA, Carter CD, Clayer M. (2002). (2000). Ephrin-B2 is a candidate ligand for the Eph receptor, EphB6. Human osteosarcoma expresses specific ephrin profiles: implications FEBS Lett 466: 169–174. for tumorigenicity and prognosis. Cancer 95: 862–869. Nicholas MK, Lukas RV, Jafri NF, Faoro L, Salgia R. (2006). Walker-Daniels J, Riese II DJ, Kinch MS. (2002). c-Cbl-dependent Epidermal growth factor receptor - mediated signal transduction EphA2 protein degradation is induced by ligand binding. Mol in the development and therapy of gliomas. Clin Cancer Res 12: Cancer Res 1: 79–87. 7261–7270. Wang Y, Ota S, Kataoka H, Kanamori M, Li Z, Band H et al. (2002). Oksvold MP, Thien CB, Widerberg J, Chantry A, Huitfeldt HS, Negative regulation of EphA2 receptor by Cbl. Biochem Biophys Res Langdon WY. (2003). Serine mutations that abrogate ligand- Commun 296: 214–220. induced ubiquitination and internalization of the EGF receptor Waterman H, Yarden Y. (2001). Molecular mechanisms underlying do not affect c-Cbl association with the receptor. Oncogene 22: endocytosis and sorting of ErbB receptor tyrosine kinases. FEBS 8509–8518. Lett 490: 142–152. Otal R, Burgaya F, Frisen J, Soriano E, Martinez A. (2006). Ephrin- Xie D, Yin D, Tong X, O’Kelly J, Mori A, Miller C et al. (2004). Cyr61 A5 modulates the topographic mapping and connectivity of is overexpressed in gliomas and involved in integrin-linked kinase- commissural axons in murine hippocampus. Neuroscience 141: mediated Akt and beta-catenin-TCF/Lef signaling pathways. Cancer 109–121. Res 64: 1987–1996. Poliakov A, Cotrina M, Wilkinson DG. (2004). Diverse roles of eph Xu J, Thompson KL, Shephard LB, Hudson LG, Gill GN. (1993). receptors and ephrins in the regulation of cell migration and tissue T3 receptor suppression of Sp1-dependent transcription from assembly. Dev Cell 7: 465–480. the epidermal growth factor receptor promoter via overlapping Qin XF, An DS, Chen IS, Baltimore D. (2003). Inhibiting HIV-1 DNA-binding sites. J Biol Chem 268: 16065–16073. infection in human T cells by lentiviral-mediated delivery of Yarden RI, Wilson MA, Chrysogelos SA. (2001). Estrogen suppres- small interfering RNA against CCR5. Proc Natl Acad Sci USA sion of EGFR expression in breast cancer cells: a possible 100: 183–188. mechanism to modulate growth. J Cell Biochem Suppl 36: 232–246.

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

Oncogene