Lyn Kinase Activity Is the Predominant Cellular Src Kinase Activity in Glioblastoma Tumor Cells

Src family members are non–receptor tyrosine kinases that are activated rapidly on the engagement of multiple cell surface receptors and modulate several cellular processes, including cell adhesion, migration, proliferation, and differentiation (reviewed in refs. 1–3). Despite the similar domain structure of the nine known Src family members, recent evidence suggests that individual Src family members may promote specific cellular functions. For example, Fyn has been shown to promote the differentiation of oligodendrocytes in the developing brain (4, 5) and hemidesmosome disassembly in A431 epidermoid squamous carcinoma cells (6), c-Src but not c-Yes promotes cytoskeletal reorganization in fibroblasts (7), and we have reported recently that Lyn but not Fyn is necessary for the migration promoted by the cooperation of integrin α_vβ₃ and the platelet-derived growth factor receptor β (PDGFrβ) in human glioblastoma cells (8).

The mechanism(s) accounting for the selective association of specific Src family members with particular functions is not entirely clear. Summy et al. (7) showed that the amino-terminal domains of c-Src (SH4-unique-SH3-SH2) dictate specificity in signaling that differentiates c-Src and c-Yes, suggesting that a Src family member–specific unique domain in cooperation with the SH4, SH3, and SH2 domains is responsible, at least in part, for Src family member–specific signaling. Other investigators have shown that certain Src family members can associate preferentially with specific cell surface receptors. For example, in platelets, integrin α_IIbβ₃ coimmunoprecipitates with c-Src rather than with Fyn and Lyn (9), whereas Lyn and Fyn associate with the Fc receptor γ chain (10).

The functional roles of the members of the Src family suggest that they may be involved in tumorigenesis and this possibility has been explored extensively using cell lines. Analyses of tumor biopsies have indicated that c-Src activity is significantly elevated in breast and colon cancer biopsy samples compared with normal tissue samples (11–16), and there has been one report of elevated levels of c-Src protein in breast cancer samples (17). Interestingly, Han et al. (16) reported differential activation of c-Src and c-Yes in primary colon cancer biopsy samples compared with metastatic colon cancer samples.

Based on our observation that Lyn but not Fyn is necessary for the PDGF-stimulated migration of glioblastoma cells when integrin α_vβ₃ is engaged (8) and the known up-regulation of the PDGFr and integrin α_vβ₃ in these tumors (18–23), we compared the levels of Lyn kinase activity, protein, and message in glioblastoma (grade IV) tumor biopsy samples with those in anaplastic astrocytoma (grade III) tumor samples, nonneoplastic brain, and normal autopsy brain samples. We found that Lyn kinase activity is significantly elevated in the glioblastoma biopsy samples. Furthermore, Lyn protein is more strongly expressed in the glioblastoma tumor cells compared with the tumor endothelial cells, suggesting that the elevated Lyn kinase activity detected in glioblastoma biopsy samples is poised to promote the highly invasive and/or proliferative phenotype of these tumors.

Tissue samples. Frozen samples of nonneoplastic brain, normal autopsy brain, anaplastic astrocytoma (grade III) tumor, and glioblastoma (grade IV) tumor biopsies were obtained from the Brain Tumor Bank at the University of Alabama at Birmingham Medical Center or from the Cooperative Human Tissue Network of the National Cancer Institute in accordance with the University Human Tissue Committee policies. Tumors were diagnosed and graded according to the WHO Classification of Brain Tumors (24). The nonneoplastic brain biopsy samples were from the temporal lobe (cortex and white matter) of patients with a seizure disorder (tissue removed to gain access to the seizure focus), from patients with head trauma where normal brain tissue had to be removed to control bleeding, or from patients undergoing repair of a vascular malformation and normal brain tissue had to be removed to gain access to the abnormal vessel. Histopathologic examination showed increased numbers of reactive astrocytes and decreased numbers of neurons in nonneoplastic brain tissue from seizure patients and edema in nonneoplastic brain from trauma patients. The normal autopsy brain samples were from frontal, parietal, temporal, or occipital lobes, included both gray and white matter, and were obtained within 24 hours of death.

Immunoblotting. Samples were lysed in 10 mmol/L Tris Base, pH 7.4, 150 mmol/L NaCl, 1% Deoxycholate, 1% Triton X-100, and 0.1% SDS (RIPA) lysis buffer with protease inhibitors (10 μg/mL aprotinin, 10 μg/mL leupeptin, 10 μg/mL N-α-tosyl-l-lysyl-chloromethyl-ketone, 100 μmol/L phenylmethylsulfonyl fluoride, and 100 μmol/L sodium vanadate) as described (25). Equivalent amounts of protein (20 μg) were resolved on a 12% disulfide-reduced SDS-PAGE gel, transferred to Immobilon-P membrane (Millipore Corp., Bedford, MA), blocked with 2% nonfat milk in TBST (3 hours, 22°C), and reacted with the primary antibody overnight at 4°C followed by a secondary antibody conjugated to horseradish peroxidase (HRP; Sigma Chemical Co., St. Louis, MO). Membrane-bound antibodies were detected by enhanced chemiluminescence (Amersham Biosciences, United Kingdom). Band intensities were determined by averaging the densitometry readings from three different autorad exposures. Src family member–specific rabbit anti-Lyn, anti-Fyn, anti-c-Src, anti-c-Yes, and anti-Lck IgG were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), and rabbit anti-c-Yes IgG was purchased from Upstate Biotechnology (Lake Placid, NY); each antibody was used from one lot and the optimal concentration was determined as described (8). Anti-Src [pY418] IgG was purchased from Upstate Biotechnology, monoclonal antibody (mAb) anti-actin from Sigma Chemical, and mAb anti–glyceraldehyde 3-phosphate dehydrogenase (G3PDH) from Research Diagnostics (Flanders, NJ). The following antibody concentrations were used for blotting: rabbit anti-c-Src, anti-Lyn, and anti-Fyn IgG (0.1 μg/mL), rabbit anti-c-Yes and anti-Lck IgG (0.2 μg/mL), rabbit anti-phosphospecific Src [pY418] IgG (0.01 μg/mL), and loading control mAbs anti-actin and anti-G3PDH (0.1 μg/mL).

Lyn and pan-Src kinase activity assay. The Lyn and pan-Src kinase activities were measured using a Src kinase activity assay kit (Upstate Biotechnology) as described (8). Tissue samples were lysed as described above, and following overnight immunoprecipitation at 4°C of 300 μg lysate with rabbit anti-Lyn IgG, mouse anti-pan-Src IgG, or rabbit IgG coupled to Gamma-bind-Sepharose, the immunoprecipitated samples were washed as follows: thrice with PTO buffer (PBS with 0.1% ovalbumin and 1% Tween 20), six times with RIPA lysis buffer, thrice with PT buffer (PBS with 1% Tween 20), and once with PBS. The kinase activity was estimated based on phosphorylation of a specific Src kinase substrate peptide and the activity was expressed as cpm. Nonspecific activity detected in the rabbit IgG immunoprecipitates was subtracted.

Real-time reverse transcription-PCR analysis. Total RNA from patient biopsies was extracted with Trizol (Invitrogen, Carlsbad, CA), purified on RNeasy columns (Qiagen, Valencia, CA), and quantitated with RiboGreen (Molecular Probes, Eugene, OR). The GeneAmp 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) was used for the detection of real-time PCR products amplified from reverse-transcribed total RNA (25 ng). Reverse transcription-PCR (RT-PCR) reactions were done in triplicate for each sample along with an internal S9 control. Levels of Lyn, Fyn, and c-Src message were normalized to S9 message and plotted relative to the level of message found in RNA from pooled normal brain (Ambion, Austin, TX). Primers and probes were purchased from Applied Biosystems and were as follows: Fyn forward primer 5′-CAGCAAGACAAGGTGCAAAGTT-3′, Fyn reverse primer 5′-TTGATTGTGAACCTCCGTAC-3′, Lyn forward primer 5′-GGCATACATCGAGCGGAAGA-3′, Lyn reverse primer 5′-GACTCGGAGACCAGAACATTAGC-3′, c-Src forward primer 5′-AGCACAGGACAGACAGGCTACA-3′, c-Src reverse primer 5′-CGTCTGGTGATCTTGCCAAAATA-3′, Fyn probe 5′-(6-FAM)-CATCAAGTGGACGGCCCCCG-TAMRA-3′, Lyn probe 5′-(6-FAM)-CTACATTCACCGGGACCTGCGAGC-TAMRA-3′, and c-Src probe 5′-(6-FAM)-CGTGGCGCCCTCCGACTCC-TAMRA-3′.

Immunohistochemistry. Frozen sections were reacted with 5.0 μg/mL rabbit anti-c-Src, anti-Lyn, or anti-Fyn IgG (rabbit IgG as a negative control) in 1% bovine serum albumin in PBS with 0.01% Tween 20 (22°C, 60 minutes), washed, and reacted with a multilink secondary antibody (22°C, 20 minutes) followed by a biotin-streptavidin label and 3,3′-diaminobenzidine substrate using the kit from BioGenex (San Ramon, CA) as described (18). The sections were then counterstained with hematoxylin and coverslipped. The intensity of staining was compared with the negative control of rabbit IgG and graded as follows: 3+, dark brown; 2+, medium brown; 1+, light brown; and wk+, very light brown. In all sections, >30% of the indicated cell type in the tissue section were stained with the indicated antibody.

Statistical analysis. The data were summarized using mean, SD, median, and range for each group. A Kolmogorov-Smirnov test was applied to assess whether the data were distributed normally. If the data were distributed normally, an ANOVA was applied to determine if there were differences among the mean levels. If the data were not distributed normally, a Kruskal-Wallis test was applied to test if there were differences among the median levels. If significance at a level of 0.05 was achieved based on ANOVA or Kruskal-Wallis tests, a two-sample t test or Wilcoxon rank sum test using normal approximation was applied to test if the difference in the mean or median, respectively, between tumor tissue and normal tissue or between two grades of tumor tissue was significant. As there were five comparisons, a Bonferroni's correction was used (i.e., the significance level for the pairwise comparison was 0.01). A Pearson correlation coefficient (r) was calculated between the normalized level of Lyn protein and Lyn activity as well as between the normalized level of Lyn activity and Lyn kinase activity for each group. Potential interactions between group and sex also were examined.

Elevated Lyn kinase activity in glioblastoma biopsy samples. We first determined the level of Lyn kinase activity in those tissue samples for which sufficient sample lysate was available. Lysate from each sample was immunoprecipitated with specific anti-Lyn IgG and mAb anti-pan Src IgG and with normal rabbit IgG as a negative control, washed, and subjected to the Src kinase activity assay (Table 1). The kinase activity assay data were normally distributed in the seven glioblastoma samples, eight grade III tumor samples, five normal autopsy brain samples, and six nonneoplastic brain samples. As the ANOVA applied to evaluate differences in the means between groups indicated significant differences (significance level of 0.05), a two-sample t test using normal approximation was applied. This indicated a significantly elevated mean Lyn kinase activity in the glioblastoma samples compared with the anaplastic astrocytoma (grade III) samples (P < 0.001), nonneoplastic brain biopsy samples (P < 0.001), and normal autopsy brain samples (P < 0.001; Table 1). No significant interaction for age or sex between groups was found. These data indicate that Lyn kinase activity is significantly elevated in glioblastoma tumor biopsy samples.

Elevated Lyn activity in glioblastoma biopsy samples. Lyn kinase activity also can be determined in smaller samples using a standard immunoblotting technique with phosphospecific anti-Src [pY418] IgG. Using this technique, we were able to determine the level of Lyn activity in a total of 14 glioblastoma tumor samples, 13 grade III tumor samples, 13 nonneoplastic brain biopsy samples, and 7 normal autopsy brain samples. The anti-Src [pY418] IgG antibody detects only the active forms of all Src family members and recognizes phosphorylation of the autophosphorylation site (1). All of the samples used for the Lyn and pan-Src kinase activity assays in Table 1 were included in this analysis. We detected proteins with apparent molecular weights of 53, 56, 59, and 60 kDa in most of the samples (Fig. 1A,-E, a). In addition, a protein with an apparent molecular weight of 62 kDa on the anti-Src [pY418] IgG blot was detected in only two anaplastic astrocytoma samples (Fig. 1A,, a, lanes 9 and 10) but in no glioblastoma samples. To differentiate Lyn and Lck, which are the two Src family members that migrate with an apparent molecular weight of 56 kDa on disulfide-reduced SDS-PAGE, we immunoblotted with anti-Lyn and anti-Lck specific IgG. This revealed that the Lyn protein frequently migrated as a doublet with the apparent molecular weight of 56 and 53 kDa in the biopsy samples (Fig. 1A,-E, b), with the 53-kDa band most likely representing an alternatively spliced form (26). Lck protein was not detected in any of these samples (a blot of six representative samples shown in Fig. 1F,, a, lanes 2-7) but was consistently detected in the positive control, a lysate of T lymphocytes (Fig. 1F , a, lane 1).

Therefore, to determine the level of Lyn activity relative to total cellular Src activity (% Lyn activity) in each sample, we normalized the sum of the bands representing the active Lyn protein (apparent molecular weights of 56 and 53 kDa) to actin and then divided this ratio by the sum of the densitometric readings for total cellular Src activity (sum of bands with M_r of 60,000, 59,000, 56,000, and 53,000; Fig. 1A,-E, a). To determine the relative activity of the non-Lyn cellular Src in each sample, we summed the densitometric readings of the bands with a M_r of 59,000 and 60,000, normalized this value to actin, and divided this ratio by the sum of the densitometric readings for total cellular Src activity (% non-Lyn activity; Fig. 2A). The analysis excluded the active Src with an apparent molecular weight of 62 kDa (consistent with c-Yes) detected in two grade III tumor samples. The data were normally distributed; thus, ANOVA was applied to test for significant differences in the mean percentage of Lyn activity or non-Lyn activity. The mean percentage of Lyn activity in the glioblastoma samples was significantly higher (2-fold) compared with the mean percentage in the nonneoplastic brain and the normal autopsy brain samples (P < 0.0001 and P = 0.0001, respectively; Fig. 2A). No significant difference between % Lyn activity in the glioblastoma samples compared with the anaplastic astrocytoma samples was detected. These data support the Lyn kinase assay data and indicate an elevated Lyn activity in glioblastoma tumor biopsy samples.

As the above analysis did not take into account the levels of Lyn protein, we then determined the normalized level of Lyn activity for each sample by summing the densitometric readings of the bands with M_r at 56,000 and 53,000 on the anti-Src [pY418] IgG blot and dividing the sum of these densitometric readings by the level of Lyn protein normalized to actin (Fig. 2B). The data for the normalized levels of Lyn activity were not normally distributed; thus, a Kruskal-Wallis test was applied to test for differences among the median levels, and significant differences were found with a significance level of 0.05 (Fig. 2B). A Wilcoxon rank sum test using normal approximation was then applied, and the median of the normalized levels of Lyn activity in the glioblastoma samples was found to be significantly higher than that in the nonneoplastic brain (P = 0.002) and marginally higher than that in the normal autopsy brain (P = 0.06) and the grade III tumor samples (P = 0.06). Two of the glioblastoma samples had very elevated normalized levels of Lyn activity; thus, we also compared the normalized level of Lyn activity in the glioblastoma group with the three other groups after excluding the two highest values in the glioblastoma group. The data in the glioblastoma group were then normally distributed, and as the data in the three other groups were also normally distributed, a two-sample, two-sided t test was applied. Compared with the glioblastoma group, the nonneoplastic brain group had a significantly lower normalized level of Lyn activity (P = 0.007) and the normal autopsy brain and the anaplastic astrocytoma group were not significantly different. As 13 of the 14 glioblastoma samples were from primary (de novo) glioblastomas, we eliminated the measurement for the one secondary glioblastoma and did the same statistical analysis. Again, the median of the normalized level of Lyn activity in the glioblastoma tumor samples was significantly elevated compared with the nonneoplastic brain samples (P = 0.004) and marginally elevated compared with the normal autopsy brain (P = 0.08) and the grade III tumor samples (P = 0.08). No significant interaction for age or sex between groups was found. We found no correlation between survival and the normalized level of Lyn activity in the 11 of 14 glioblastoma patients in which follow-up survival data were available using a proportional hazard regression model (P = 0.65). These data support the Lyn kinase activity assay data and indicate that, even when the level of Lyn activity is normalized to the level of Lyn protein, Lyn activity remains significantly elevated in glioblastoma tumors compared with the nonneoplastic brain.

To test for a correlation between Lyn kinase activity and the normalized level of Lyn activity, a Pearson correlation coefficient was calculated. A correlation between Lyn kinase activity and the normalized level of Lyn activity was found when comparing these values for all four sample groups together (r = 0.6928; P = 0.0001). In addition, a correlation was found between Lyn kinase activity and the normalized level of Lyn activity for each sample group individually: glioblastoma samples (r = 0.9250; P = 0.003), anaplastic astrocytoma samples (r = 0.8347; P = 0.01), nonneoplastic brain samples (r = 0.9155; P = 0.01), and normal autopsy brain samples (r = 0.9893; P = 0.01). The presence of a correlation between Lyn kinase activity and the normalized level of Lyn activity support our conclusion that Lyn activity is significantly elevated in glioblastoma tumor biopsy samples.

Expression of Fyn and c-Src proteins as well as Lyn in glioblastoma tumor samples: the normalized level of Lyn protein in glioblastoma tumors is elevated. We then determined the level of Lyn protein relative to other cellular Src family members. Immunoblotting for Fyn and c-Src (Fig. 1A,-E, c and d, respectively) indicated expression of Fyn and c-Src in most of the tissue samples. An additional Src family member, c-Yes, has been reported to be expressed in neurons in the chicken and the rat brain (27–29). Furthermore, compared with the cerebellum, c-Yes was reported to be expressed at a 3- to 6-fold lower level in other regions of the brain, including cortex and white matter (27). On immunoblotting a subset of tissue samples with anti-c-Yes IgG, we detected a protein with an apparent molecular weight of 62 kDa, consistent with c-Yes, after a long exposure in four of four normal autopsy brain samples, three of five nonneoplastic brain samples, three of five grade III tumors, and three of seven glioblastoma tumors (six representative samples shown in Fig. 1F,, b). These data suggest that c-Yes is expressed at low levels in the normal brain cortex and white matter and at low levels in a subset of these tumors. To compare the levels of Lyn, Fyn, and c-Src proteins between the different groups, we normalized the densitometric readings for Fyn and c-Src proteins to actin (Fig. 3B and C) as described for Lyn protein (Fig. 3A). As an additional control to test the quality of the lysates, the membranes were all reprobed with mAb anti-G3PDH (data not shown; Fig. 1B , f, and F, c), and Lyn, Fyn, and c-Src proteins were then normalized to G3PDH. We found the ratio of Lyn, Fyn, and c-Src proteins normalized to actin or G3PDH to be highly similar, supporting the use of actin as a valid loading control.

Statistical analyses of the normalized levels of Lyn, Fyn, and c-Src proteins for each group indicated the data for Lyn protein were not distributed normally. Therefore, a Kruskal-Wallis test was applied to test for differences in the median of the normalized levels of Lyn protein between the groups. Significant differences were found with a significance level of 0.05; thus, a Wilcoxon rank sum test using normal approximation was applied. We found the median of the normalized levels of Lyn protein to be significantly elevated in the glioblastoma tumor samples compared with the nonneoplastic brain (P = 0.01) and the normal autopsy brain (P = 0.0479) samples, but it was not significantly different from the anaplastic astrocytoma samples (P = 0.1092; Fig. 3A). In 3 of 13 nonneoplastic brains, the normalized level of Lyn protein was high (Fig. 3A). This could be due to higher levels of Lyn protein in specific areas of the normal brain, as Lyn protein in the normal adult mouse brain shows stronger expression in specific neuronal populations (30). The normalized levels of Fyn protein were normally distributed and no significant difference in the means of the normalized levels of Fyn protein was detected between the glioblastoma samples and the three other groups (Fig. 3B). The normalized levels of c-Src protein were not normally distributed; thus, a statistical analysis was done as described for the normalized levels of Lyn protein. No significant differences in the median of the normalized levels of c-Src protein were detected between the glioblastoma samples and the three other groups (Fig. 3C). No significant interaction for age or sex between groups was found.

To test for a correlation between the normalized levels of Lyn protein (Fig. 3A) and the normalized levels of Lyn activity (Fig. 2B), a Pearson correlation coefficient was calculated. A correlation between the normalized level of Lyn protein and the normalized level of Lyn activity was only found for the normal autopsy brain samples (r = 0.8639; P = 0.0122). These data suggest that the elevated normalized level of Lyn activity detected in the glioblastoma samples is not due solely to an elevated normalized level of Lyn protein.

Characterization of Lyn, Fyn, and c-Src message levels in glioblastoma tumors. As Lyn protein was significantly elevated in glioblastoma biopsy samples, we examined Lyn message levels by real-time RT-PCR analysis, normalized these data to S9 message and expressed the levels relative to the message level detected in pooled commercially available normal brain RNA. The levels of Fyn and c-Src messages were evaluated as a control. A different patient sample group was used due to the limited amount of tumor tissue and nonneoplastic brain biopsy samples available. In this analysis, the nonneoplastic brain samples were used as a normal control. Lyn message levels were not normally distributed; therefore, a Kruskal-Wallis test was applied to test for differences in the median levels between groups, and significant differences were seen with a significance level of 0.05. A Wilcoxon rank sum test using normal approximation was then applied, and the median level of Lyn message in the glioblastoma samples was found to be significantly elevated compared with the nonneoplastic brain (P = 0.0266) and the anaplastic astrocytoma tumor samples (P = 0.0028; Fig. 4A). In contrast, the Fyn and c-Src message levels were not significantly different in the glioblastoma samples compared with the nonneoplastic brain samples (Fig. 4B and C, respectively). These data suggest that the elevated normalized level of Lyn message detected in the glioblastoma tumor biopsy samples could potentially contribute to the elevated normalized level of Lyn protein we detected in the glioblastoma samples; however, as the Lyn protein and message levels were analyzed in different patient groups, we cannot directly compare them.

Cellular localization of Lyn, Fyn, and c-Src in glioblastoma tumor biopsy samples. To determine whether the elevated Lyn kinase activity was attributable to a particular cell type in the glioblastoma tumor samples, immunohistochemical analysis was done with Src family member–specific anti-Lyn, anti-Fyn, and anti-c-Src IgG on frozen biopsy samples using rabbit IgG as a negative control. We found Lyn protein to be strongly expressed (2+ to 3+ intensity) in the glioblastoma tumor cells in five of five biopsies (Fig. 5A and B) and a less intense expression in the tumor endothelial cells (wk+ to 1+ intensity) in four of five biopsies (Fig. 5B) or no detectable endothelial cell expression (one of five biopsies). Fyn protein was expressed in the glioblastoma tumor cells and in the tumor endothelial cells (1+ to 2+ intensity) in five of five biopsy samples (Fig. 5C). In contrast, the expression of c-Src protein was considerably less intense in the glioblastoma tumor cells (wk+ intensity) in five of five biopsies. c-Src was detected in the tumor endothelial cells (wk+ to 1+ intensity) in five of five biopsies (Fig. 5D). These data taken together with the Lyn kinase activity data and the immunoblotting data for Lyn activity suggest that the elevated Lyn kinase activity we have detected in glioblastoma tumor biopsy samples is likely due in large part to Lyn kinase activity localized to tumor cells.

In this report, we show that Lyn kinase activity is significantly elevated in glioblastoma (grade IV) tumor biopsies compared with anaplastic astrocytoma (grade III) tumor biopsies, nonneoplastic brain biopsies, and normal autopsy brain cortex and white matter. Our kinase activity assays for Lyn and pan-Src (all Src family members) show a 5-fold higher level of Lyn kinase activity in the glioblastoma tumor samples compared with the three other categories of tissues. This is comparable with the elevation in c-Src activity reported in breast and colon cancer tissue samples relative to normal tissue (11–13, 15). Notably, in the glioblastoma samples, Lyn kinase activity accounted for >90% of the pan-Src kinase activity, whereas in the nonneoplastic brain, normal autopsy brain, and the grade III tumor samples Lyn kinase activity accounted for ≈30% of the total pan-Src kinase activity. These findings suggest that other Src family members, like Fyn and c-Src kinases, contribute substantially to the cellular Src kinase activity of the normal and nonneoplastic brain samples as well as to the anaplastic astrocytoma (grade III) tumor samples but that they contribute substantially less to the cellular Src kinase activity of the glioblastoma tumor samples. The kinase activity assay results are supported by our immunoblotting data that also indicate a significantly elevated level of Lyn activity in the glioblastoma tumor samples compared with the nonneoplastic brain. Furthermore, a correlation was found between the normalized level of Lyn activity and Lyn kinase activity. As 13 of 14 glioblastoma tumor samples were obtained from patients with a primary or de novo glioblastoma tumor (a tumor that does not arise from a lower-grade tumor), these data also suggest that elevated Lyn activity is characteristic of this type of glioblastoma tumor.

Although we found that the levels of Lyn protein and message were significantly elevated in the glioblastomas, we did not find a correlation between the normalized level of Lyn protein and Lyn activity in the glioblastoma tumors. This is consistent with a complex regulation of cellular Src family member activity that is not solely dependent on the level of Src family member protein expressed. Cellular Src activity is regulated by multiple mechanisms, which include the action of growth factor receptors and integrin receptors that activate cellular Src, the activity of the Csk kinase that phosphorylates the negative regulatory peptide in the carboxyl terminus of cellular Src inhibiting cellular Src activity, phosphatases such as SHP-2 that dephosphorylate the tyrosine residue in the negative regulatory peptide promoting cellular Src activity, the cell surface inhibitory receptor molecule SIRP1α that recruits SHP-2 to the membrane, and adaptor molecules such as AFAP-110 that bind to the cellular Src SH3 domain and disrupt intramolecular autoinhibitory interactions, thereby promoting cellular Src activity (1–3, 31–35). Typically, overexpression of a wild-type Src family member primes the cell for signaling (1); thus, elevated expression of a Src family member could increase cellular Src activity. Our data suggest that the elevated Lyn kinase activity detected in the glioblastoma tumor samples is most likely attributable to more than one mechanism. Based on previous studies, we speculate that these mechanisms include up-regulation of growth factor receptors known to activate cellular Src, such as the PDGFr and the epidermal growth factor receptor (22, 23, 36), and up-regulation of integrin receptors, such as α_vβ₃ and α_vβ₅ (18–21). In addition, the elevated level of Lyn protein could position more Lyn molecules to become stimulated and to transmit signals that promote an invasive phenotype.

Of the five Src family members (Lyn, Fyn, c-Src, c-Yes, and Lck) that have been reported to be expressed in the brain (1, 27–29, 37), we found that Lyn, Fyn, and c-Src were consistently expressed in all of the categories of brain samples tested. We also detected low levels of c-Yes in the normal brain cortex and white matter and low levels of c-Yes in a subset of the tumor samples but only after a long exposure. The absence of c-Yes expression in some tumor samples is most likely associated with the loss of neurons that occurs in grade III and IV tumors (24), as c-Yes exhibits neuron-specific expression in the normal brain (27–29). This finding is consistent with our previous inability to detect c-Yes protein and only barely detectable levels of c-Yes message in two glioblastoma cell lines propagated in vitro (8). Although Lck was consistently and readily detectable in our positive controls, we did not detect it in any of the tissue samples, suggesting that there is a significantly lower level of Lck protein in the normal brain and in grade III and IV tumors compared with Lyn, Fyn, and c-Src proteins. Lck message has been reported to be expressed in the adult mouse brain, but it was noted that its detection by Northern blotting and in situ hybridization required a very long exposure (37) and the intensity of the Lck protein signal in mouse brain lysate, detected by immunoblotting of an unspecified amount of lysate with anti-Lck antisera, was reported to be ≈10-fold lower than that found in the thymus (37). Thus, although we could not detect Lck protein in our tissue samples, we cannot entirely rule out the possibility that Lck is expressed at very low levels in these samples.

The strong expression of Lyn protein in the glioblastoma cells, compared with its less intense expression in the endothelial cells, of the glioblastoma tumors together with its predominant activity among the members of the cellular Src family that are coexpressed in these tumors indicates that Lyn kinase activity is poised to promote the highly migratory/invasive phenotype of glioblastoma tumors.

Grant support: NIH, National Cancer Institute grant P50 CA97247 (L.B. Nabors, G.Y. Gillespie, and C.L. Gladson), and grant CA97110 (C.L. Gladson).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Mrs. Jo Self for assistance in preparing this manuscript.