|Home | About | Journals | Submit | Contact Us | Français|
We studied the expression and distribution of the microtubule-severing enzyme spastin in 3 human glioblastoma cell lines (U87MG, U138MG, and T98G) and in clinical tissue samples representative of all grades of diffuse astrocytic gliomas (n= 45). In adult human brains, spastin was distributed predominantly in neurons and neuropil puncta, and to a lesser extent, in glia. Compared to normal mature brain tissues, spastin expression and cellular distribution were increased in neoplastic glial phenotypes, especially in glioblastoma (p < 0.05 vs. low-grade diffuse astrocytomas). Overlapping punctate and diffuse patterns of localization were identified in tumor cells in tissues and in interphase and mitotic cells of glioblastoma cell lines. There was enrichment of spastin in the leading edges of cells in T98G glioblastoma cell cultures and in neoplastic cell populations in tumor specimens. Real-time PCR and immunoblotting experiments revealed greater levels of spastin mRNA and protein expression in the glioblastoma cell lines vs. normal human astrocytes. Functional experiments indicated that spastin depletion resulted in reduced cell motility and higher cell proliferation of T98G cells. To our knowledge, this is the first report of spastin involvement in cell motility. Collectively, our results indicate that spastin expression in glioblastomas might be linked to tumor cell motility, migration, and invasion.
Gliomas are the most frequent group of central nervous system (CNS) neoplasms, accounting for more than 70% of all brain tumors (1). Diffuse gliomas are particularly ominous forms as they are highly invasive within the brain and are difficult to treat. Glioblastoma (GBM) is the most common and malignant, as well as the deadliest, glioma in adults (1).
One of the most effective strategies for treating malignant tumors has focused on disrupting the integrity of the microtubule system and preventing mitotic division (2–4). Microtubules are dynamic polymers that undergo rapid bouts of assembly and disassembly during interphase, and then reorganize into a bipolar spindle during mitosis (5). Targeting microtubules represents a strategy for attenuating the invasive properties of tumor cells, as these cytoskeletal structures are implicated in tumor cell motility, the formation of long dynamic protrusions (6), and the elongation of invadopodia (7). Unfortunately, despite the potential usefulness of a new generation of tubulin-binding agents from the bench to the bedside, the emergence of drug-resistant tumor cells remains an overriding problem that contributes to treatment failures (8). As a result, the efficacy of drugs such as the taxanes, which aggressively stabilize microtubules in tumor cells, is hindered by the development of resistance ascribed, in part, to cytoskeletal alterations; these include the aberrant overexpression of class III β-tubulin isotype (4, 9–11). Moreover, many of these drugs do not readily cross the blood-brain barrier (12), while they also exert deleterious effects on host neural tissues, as exemplified in paclitaxel-induced peripheral neuropathy (13). Hence, there is great urgency in expanding the body of knowledge about the regulation of microtubules in cancer cells so that more sophisticated strategies can be developed.
A number of different proteins impact the dynamic properties of the microtubules, such as stathmin, which promotes microtubule disassembly, and the classic microtubule-associated proteins, which promote their assembly and stabilization (14). Molecular motor proteins are also needed to organize microtubules into higher level structures such as the mitotic spindle; motor proteins s such as kinesin-5 have recently been proposed as potential targets for anti-cancer drugs (15, 16). Another category of microtubule-related proteins is endowed with the specialized property of severing or breaking the lattice of microtubules, but these proteins have received little attention in the context of cancer to date. These so-called microtubule-severing proteins are enzymes capable of severing long microtubules generated at the centrosome by breaking the microtubules into short pieces that are able to translocate through the axonal and glial cytoplasm (5, 17).
The microtubule-severing protein spastin, encoded by the SPG4 gene, is a member of the ATPases associated with various cellular activities (AAA) family of proteins. Spastin has a 220 amino acid AAA domain at its carboxy terminus, as well as other functional domains (18). A microtubule interacting and endosomal trafficking (MIT) domain is located at the amino terminus; the microtubule-binding domain comprises amino acids 270 to 328 (19). Two functional nuclear localization signals responsible for targeting to the nucleus and 2 nuclear export signals have also been identified, but their functional significance is not understood (18). Biochemical and cell biological experiments demonstrated that both microtubule-binding domain and AAA domains are essential for the microtubule-severing activity of spastin (19).
Various spastin isoforms with heterogeneous patterns of distribution among different tissues have been demonstrated (20). These isoforms mainly result from a combination of translation at 2 different start codons (18, 21) and the inclusion/exclusion of exons 4, 8 and 15 within these transcripts (20). Distribution studies in rat tissue showed that the longest spastin isoform (hereafter referred to as M1) is selectively expressed in the adult spinal cord, whereas the shorter isoform (referred to as M87) has a more ubiquitous tissue distribution (22). To date, the main interest in spastin has derived from the fact that mutations in this gene account for 40% of cases of hereditary spastic paraplegias (23).
In the normal human and mouse CNS, spastin is expressed constitutively in neurons (24, 25), but induced spastin expression has been reported in reactive astrocytes of the hippocampus in pilocarpine-induced status epilepticus (25). To our knowledge, the expression of spastin in brain tumors in general and in gliomas in particular is hitherto unknown. Given the role of microtubules on the invasive properties of tumor cells and the high propensity of glioblastoma cells for brain invasion, we hypothesized that the microtubule-severing ATPase spastin might be aberrantly expressed and/or regulated in these cells. To test this hypothesis, we evaluated the expression levels and the intracellular distribution of spastin in human glioblastoma cell lines, and in surgically excised tumor samples representative of all grades of diffuse astrocytic gliomas. In addition, functional experiments evaluated effects of spastin depletion on cell motility and proliferation in glioblastoma cells.
All cell lines were obtained from American Type Culture Collection (Manassas, VA). Human glioblastoma cell lines T98G, U87MG and U138MG, human neuroblastoma cell lines SH-SY5Y and SK-N-SH, as well as human embryonic kidney (HEK) cells were maintained in Dulbecco’s modified Eagle’s media (Invitrogen, Prague, Czech Republic) containing 10% fetal bovine serum (PAA Laboratories, Cölbe, Germany), 4 mmol/L L-glutamine and antibiotics. SH-SY5Y cells were cultivated for 4 days in the presence of 1 uM all trans-retinoic acid (Sigma-Aldrich, Prague, Czech Republic) to induce neurite formation (26). Proliferating nonimmortalized, nontransformed human fetal astrocytes, isolated from the cerebral hemisphere of an 18-g.w. human male fetus (Clonetics Astrocyte Cell Systems), were purchased from Cambrex Bio Science (Walkersville, MD) and cultivated as previously described (27).
A rabbit polyclonal antibody (Sp/AAA) (22), and 2 mouse monoclonal antibodies, Sp3G11/1 (Millipore, Temecula, CA; catalog number MAB5634) and Sp6C6 (Sigma-Aldrich [catalog number S7074] and Santa Cruz Biotechnology, Santa Cruz, CA [catalog number sc-81624]), were used for immunohistochemistry, immunofluorescence and immunoblotting experiments. The Sp/AAA antibody was raised against the recombinant fragment coding mouse spastin polypeptide (amino acids [a.a.] 337–446) (22); the Sp3G11/1 (IgG2a) was prepared using full-length recombinant human spastin for immunization. For the preparation of antibody Sp6C6 (IgG2a), recombinant human M87 spastin lacking the 32 amino acids encoded by exon 4 (a.a.197–228) was used as immunogen (21). Mouse monoclonal antibody TU-01 to α-tubulin (IgG1) (28, 29) and affinity-purified polyclonal rabbit antibody to αβ-tubulin dimer (30) were described previously. A rabbit antibody to actin and a mouse monoclonal antibody to Golgi matrix protein GM-130 were purchased from Sigma-Aldrich and BD Biosciences (San Jose, CA), respectively. A rabbit antibody to kinesin-like protein KIF11 (Eg5) was from Cytoskeleton Inc. (Denver, CO). 5Cy3-conjugated anti-mouse and anti-rabbit antibodies and DY 488-conjugated anti-mouse and anti-rabbit antibodies for multiple staining were obtained from Jackson Immunoresearch Laboratories (West Grove, PA). Horseradish peroxidase (HRP)-conjugated anti-mouse and anti-rabbit antibodies were obtained from Promega (Madison, WI).
T98G cells in 24-well plates were transfected with siRNAs (20 nM final concentration) using Lipofectamine RNAi MAX (Invitrogen) according to the manufacturer’s instructions. Two Silencer Select siRNAs (Applied Biosystems/Ambion, Prague, Czech Republic) that target regions present in all human spastin isoforms (5′-CAACCTTGCTAACCTTATA-3′; siRNA#1 and 5′-GGAAGUCCATTGACCCAAA-3′; siRNA#2) were used. dTdT overhangs were added to the 3′ of the oligomers with exception of siRNA#1 antisense strand where dCdT overhang was used. The maximal depletion was reached 48 to 72 hours after siRNA transfection. Negative control siRNA was from Ambion (Silencer Negative Control #1 siRNA).
Total RNA from glioblastoma cell lines and normal human astrocytes (NHA) was isolated by the RNeasy Mini kit (QIAGEN, Valencia, CA). Aliquots of 1 μg of total RNA in 20-μl reaction mixture were converted to cDNA using the ImProm-II RT kit (Promega) with random hexamers in a 20-μl reaction volume. These cDNA reaction mixtures were diluted 5 times in DEPC-treated water to prevent inhibition of Taq polymerase in a subsequent PCR reaction. One μl of diluted cDNA product was used for each PCR reaction. Amplifications were performed in 10 μl PCR reaction mixtures containing QuantiTect SYBR Green PCR Master Mix (QIAGEN) and 0.5 μM of each human gene-specific primers for spastin (SPAST, NCBI Refseq ID: NM_014946 and NM_199436; primers anneal to both M1 and M87 transcript isoforms) or β-actin (ACTB, NCBI Refseq ID: NM_001101). Primer sequences were as follows: spastin forward 5′-TCAGGCTGGTCTTGAACTCC-3′, reverse 5′-ATGCATCTTCTGGCTGGG-3′; and β-actin forward 5′-TCCTTCCTGGGCATGGAGT-3′, reverse 5′-AAAGCCATGCCAATCTCATC-3′. Oligonucleotides were from East Port (Prague, Czech Republic). Real-time quantitative PCRs were carried out on Master-cycler realplex (Eppendorf, Wesseling-Berzdorf, Germany) as described (27). Experiments were performed twice with triplicate samples (cDNA isolated from 3 separate cultures of tested cells). The expression of analyzed genes was normalized to the expression of β-actin. Levels of β-actin did not differ significantly between cell lines and NHA. The identity of spastin PCR product was verified by cDNA sequencing. Statistical analysis was performed with the Student unpaired t-test.
Full-length human spastin constructs M1 (Swiss-Prot identifier Q9UBP0-1, 616 a.a.) and M87 (Swiss-Prot identifier Q9UBP0-3, 530 a.a) were subcloned in the lentiviral plasmid vector FCbAGW-IRES-GFP (D. Huang and G. Morfini, unpublished data). Lentiviral constructs were co-transfected into 293T HEK cells, along with VSVg and ΔR8.9 envelope and packaging plasmids by calcium phosphate transfection. Viral supernatants were harvested 24 to 36 hours after transfection, concentrated by ultracentrifugation, resuspended in sterile PBS, and stored at −80°C. HEK cells were then infected with functional lentivirus at a multiplicity of infection of 100. HEK cells lysates were obtained 72 hours post-infection. Lysates derived from HEK cells overexpressing full-length M1 and M87 spastin isoforms facilitated the identification of specific spastin isoforms in immunoblots.
Whole-cell extracts were prepared by washing the cells in cold PBS, solubilizing them in hot SDS-sample buffer (31) and boiling for 5 minutes. Proteins were separated by SDS-PAGE on 7.5% Tris gels and transferred to nitrocellulose membranes. Cultured T98G cells, NHA and cortical neurons were lysed in 1% SDS, and protein concentration measured using Pierce BCA protein assay kit (Cat. #23225). Lysates were run on NuPAGE Novex 4–12% Bis-Tris gels (to increase M1/M87 separation) and transferred to PDF membranes. Details on the immunoblotting procedure have been described elsewhere (32). Monoclonal antibodies to spastin (Sp6C6, Sp3G11/1) and polyclonal antibody to actin were diluted 1:1,000 and 1:3,000, respectively. The polyclonal anti-spastin antibody (Sp/AAA) was diluted from 1:2,000 to 1:20,000. The polyclonal antibody to Eg5 was diluted 1: 5,000. Monoclonal antibody to α-tubulin (TU-01) was used as undiluted hybridoma culture supernatant. Primary antibodies were detected after incubation of the blots with HRP-conjugated secondary antibody diluted 1:10,000. HRP signal was detected with chemiluminescence reagents (Pierce, Rockford, IL) in accordance with the manufacturer’s instructions.
Immunofluorescence microscopy on fixed cells was performed as described previously (33). Briefly, cells grown on coverslips were rinsed with microtubule stabilizing buffer (MSB; 1M Mes adjusted to pH 6.9 with KOH, 2 mM EGTA, 2 mM MgCl2) supplemented with 4% polyethylene glycol 6000, fixed for 20 minutes in 3% formaldehyde in MSB and thereafter extracted for 4 minutes in 0.5% Triton X-100 in MSB. Anti-spastin antibodies Sp/AAA, Sp3G11/1 and Sp6C6 were diluted :1:1,000, 1:100 and 1:100 respectively. Antibodies to GM130 and tubulin were diluted 1:100 and 1:20, respectively. Cy3-conjugated anti-mouse and anti-rabbit antibodies were diluted 1:500. DY 488-conjugated anti-mouse and anti-rabbit antibodies were diluted 1:100. FITC-conjugated phalloidin (Sigma-Aldrich) for detection of microfilaments was used at concentration 0.2 μg/ml. 4,6-diamidino-2-phenylindole was used to label cell nuclei. The preparations were mounted in MOWIOL 4–88 (Calbiochem, San Diego, CA) and examined with an Olympus A70 Provis microscope. Conjugates alone did not give any detectable immunoreactivity.
T98G cells were transfected with either spastin-specific siRNA or negative control siRNA as described above. Seventy-two hours after transfection cells were detached, counted and diluted in culture media. 2 × 104 transfected cells were plated in a single well of a 6-well tissue culture dish and allowed to adhere for 24 hours. Cells were thereafter incubated in medium containing Hoechst 33342 (Sigma-Aldrich) at final concentration 0.5 μg/ml and imaged by 10X/N.A. 0.30 dry objective on an Olympus IX-81 microscope equipped with Cell^R system, motorized stage and objectives, and a temperature-controlled chamber at 37 C and 5% CO2. Images were taken in bright-field channel to visualize whole cells and in fluorescence channel (Ex: 350/50, Em: 460/10) to visualize nuclei. The total imaging time was 10 hours with 10-minute intervals. For each experiment, time-lapse sequences of 10 different fields were acquired for negative control or spastin-depleted cells (n:3). Time-lapse sequences were adjusted and nuclei movement analyzed using an object-tracking plug-in written in house; 8-bit grayscale images were used for nuclei detection. The images were initially smoothed by Gaussian filter (with σ equal 2 pixels) in order to increase signal to noise ratio. Then the morphological opening by octagon (16 pixels wide) grayscale opening with cone (with slope 4 gray levels per pixel) and Gaussian smoothing (with σ equal 2 pixels) were applied to simplify the images and to filter out small artifacts. Maxima of the image intensity were then detected by morphological reconstruction and the coordinates of nuclei were calculated as centers of mass of the maxima dilated by octagon 20 pixels wide (34). The corresponding nuclei in subsequent images were detected by pairing the mutually closest particles with distance less than 100 pixels. The particle trajectories were constructed from corresponding particles connecting the mutually closest ends and origins of trajectories with time interval less than 6 frames and with distance less than 100 pixels. The speed of particles was calculated as ratio of the particle trajectory length and of the trajectory duration. The algorithms were implemented as plug-in modules of the Ellipse program, version 2.07 (ViDiTo Systems, Ko ice, Slovakia). Statistical analysis was performed with the Student two-tailed unpaired t-test using Microsoft Excel.
Cell proliferation was assessed by manual cell counting of T98G cells transfected with either spastin-specific siRNA or negative control siRNA. Twenty-four hours after transfection, cells were detached, diluted in culture media and counted. 2×105 transfected cells were plated on a 6-cm-diameter Petri dish. Cells were counted at various time intervals from 2 to 6 days after transfection. Samples were counted in doublets in a total of 3 independent experiments.
Formaldehyde-fixed, paraffin-embedded biopsy/resection samples from adults (n = 37) and children (n = 8) representative of all World Health Organization (WHO) grades of astrocytic gliomas (WHO grades I-IV) (n = 45) were collected retrospectively from the Gade Institute, Department of Pathology, University of Bergen, Haukeland Hospital (Bergen, Norway), Department of Pathology, University of Patras Hospital (Rion, Patras, Greece), and St. Christopher’s Hospital for Children (Philadelphia, PA) under the approval by an Institutional Review Board (IRB) Exempt Review (Drexel University IRB Protocol No.16660). A number of tumor specimens utilized in this study were also used in previous studies (35–37).
Histological classification was according to the recommendations of the 2007 WHO classification of tumors of the central nervous system (38). Tumor specimens from adult patients included low-grade diffuse astrocytomas (WHO grade II; n = 7), anaplastic astrocytomas (WHO grade III; n = 5) and GBM (WHO grade IV; n = 25). Pediatric glioma specimens included diffuse gliomas (WHO grade II) of the cerebral hemispheric white matter (n = 2) and brainstem (n = 2), and anaplastic gliomas of the thalamus and brainstem (WHO grade III; n = 4). The histologic features of tumor samples were evaluated independently by 2 neuropathologists (S.J.M. and C.D.K.) and a pediatric pathologist (J.P.D.), who were blinded to the original histopathological diagnosis for each specimen.
Five-μm-thick microtome sections from archived formalin-fixed, paraffin-embedded tissue blocks were placed on electromagnetically charged slides and stained with hematoxylin and eosin for morphological evaluation. Adjacent, serially cut, and sequentially numbered sections were processed for immunohistochemistry.
Control tissues included surgical and autopsy tissue samples from cases devoid of tumor (n = 10). Non-neoplastic cerebral and cerebellar autopsy tissue samples (n = 10) were examined from 3 infants (ages 2, 4 and 8 months), 2 children (6 and 7 years old), and 5 adults (21, 49, 54, 59 and 63 years old) who had died of non-neurological conditions and the postmortem CNS examination of whom was devoid of pathological findings. Some of these tissues were utilized in previous studies (27). The use of existing autopsy tissue specimens was approved by an Institutional Review Board Exempt Review protocol (Drexel University IRB Protocol No. 17148). No patient identifiers were used.
Immunohistochemistry was performed according to the avidin biotin complex (ABC) peroxidase method using Rabbit and Mouse IgG ABC Elite detection kits (Vector Labs, Burlingame, CA), as previously described (35–37). Five-μm-thick paraffin-embedded tissue sections were subjected to non-enzymatic antigen retrieval using 0.01 M sodium citrate buffer (pH 6.0) (for immunostaining with Sp6C6 and Sp/AAA antibodies) and Tris-EDTA (pH = 9.0) (for immunostaining with Sp3G11/1 monoclonal antibody) at 95–100°C for 30 minutes. Slides were then allowed to cool to room temperature (RT) and endogenous peroxide activity quenched with 3% H2O2 (v/v) in methanol for 30 minutes.
Tissues were incubated at RT in corresponding blocking solution (5% normal horse serum or 5% normal goat serum in 0.1% phosphate buffered saline-bovine serum albumin (PBS-BSA) for mouse monoclonal antibodies or rabbit polyclonal antibodies, respectively) for 2 hours. Tissues were then incubated overnight at RT in primary antibodies diluted in blocking solution diluted as follows: Sp/AAA 1:2,000, Sp3G11/1 1:20 and Sp6C6 1:50 to 1:100. Pre-diluted secondary antibodies from the ABC Elite detection kits were applied at RT for 1 hour, followed by a 1-hour incubation with avidin-biotin-complex solution under the same condition. Negative controls included omission of primary antibody, as well as substitution with nonspecific mouse IgG2a or nonspecific rabbit antibody (Becton Dickinson, Franklin Lakes, NJ). Experiments using isotype-matched control antibodies did not show any non-specific binding of secondary antibodies.
Immunodetection visualizations were performed using 3′3-diaminobenzidine solution (DAB Substrate Kit, Vector Labs) and were allowed to develop on tissues for 3 to 7 minutes. Slides were counterstained for 4 minutes in Harris hematoxylin (Fisher, Norristown, PA), dehydrated through graded alcohol-water sequence, cleared in xylene, and mounted with cover slides (Fisher) using Permount (Fisher).
Manual cell counting of spastin-labeled tumor cells in immunohistochemical preparations was performed by 3 observers independently (C.D.K., P.S.L. and K.K.). Cell counting and statistical analysis were carried out on preparations immunostained with Sp3G11/1 and Sp/AAA only in the adult group of diffuse astrocytic gliomas for which clear spastin immunoreactivity was detected (n = 27). Preparations stained with Sp6C6 were evaluated qualitatively/semi-quantitatively. Between 568 and 928 tumor cells were evaluated per case in 20 non-overlapping high-power (40x) fields and a labeling index was determined for each case. The labeling index was expressed as the percentage (%) of spastin-labeled cells out of the total number of tumor cells counted in each case and for each antibody, as previously described (35). The level of interobserver agreement was quantitated using generalized kappa and pairwise kappa statistics (39, 40). The immunohistochemical preparations were reviewed at a multiheaded microscope in order to achieve consensus. Interobserver agreement was substantial (κ = 0.65). For purposes of labeling index recording, the consensus opinion was considered as conclusive. The median labeling index (MLI) and the interquartile range (IQR) -delimited by the 25th and 75th percentiles- were determined for the set of cases in each histological grade using one-way ANOVA (Jandel software, Sigmastat). The statistical significance of differences in labeling indices between WHO histological grades was examined with non-parametric statistical techniques using Kruskal-Wallis analysis of variance tests. A p value of less than 0.05 was considered as statistically significant. Because of the small number of pediatric gliomas with spastin immunoreactivity included in this study, only qualitative assessment was performed in these cases.
Immunofluorescence staining of various cell lines was first performed using the Sp3G11/1 anti-spastin monoclonal antibody. In NHA, a faint granular staining was observed throughout the cytoplasm (Fig. 1A). In contrast, there was more prominent and widespread spastin localization in the form of confluent punctate and diffuse cytoplasmic aggregates in human glioblastoma cell lines (T98G, U138MG, U87MG) (Fig. 1B–D). Comparable staining patterns were observed for T98G cells stained with the anti-spastin monoclonal antibody Sp6C6 (Fig. 1E) or the polyclonal antibody Sp/AAA (Fig. 1F). In control human neuroblastoma cell lines (SH-SY5Y, SK-N-SH) both prominent cytoplasmic diffuse and granular staining were observed (Fig. 1G, H).
When SH-SY5Y cells are grown for 4 days in the presence of 1 uM all trans-retinoic acid, prominent neurites are generated (26). Spastin labeling was detectable along these neurites and in the cell periphery (Fig. 2A–C). Interestingly, vesicle-like immunoreactivity for spastin was also observed at the leading edges of migrating T98G cells, a cellular compartment where few microtubules are present (Fig. 2D–I). The association of spastin with leading edges was confirmed by double-label fluorescence in T98G cells with an antibody to spastin and FITC-conjugated phalloidin as a marker of microfilaments. The distribution of spastin and microfilaments is shown in Figure, Supplemental Digital Content 1, http://links.lww.com/NEN/A262. Spatial dissociation of spastin and GM-130 was noted in glioblastoma cells (Fig. 3A–C). Interestingly, strong spastin staining was observed in close proximity to mitotic spindle microtubules, as well as the cytoplasm of round glioblastoma cells undergoing mitosis (Fig. 3D–F). All 3 anti-spastin antibodies used in this study rendered similar patterns of immunoreactivity in T98G cells (data not shown).
Differences in the staining intensity with anti-spastin antibodies between NHA (Fig. 1A) and glioblastomas (Fig. 1B–D) were confirmed by immunoblotting experiments. Under our SDS-PAGE Tris-Glycine conditions, antibodies Sp6C6 and Sp3G11/1 recognized a major band of ~54 kDa when probed against T98G human glioblastoma cell lysates (Fig. 4A). The antibody Sp/AAA stained only very faintly blotted proteins in the 70–50 kDa region (not shown). To verify that the band recognized by Sp6C6 and Sp3G11/1 antibodies indeed corresponds to endogenous spastin, T98G cells were treated with either control or spastin-specific siRNAs, and lysates derived from these cells probed by immunoblotting. Immunoreactivity of the ~54-kDa band was substantially diminished by spastin-specific siRNA#1 and siRNA#2, but not by control siRNA, confirming its identity as endogenous spastin (Fig. 4B). Immunoblotting experiments showed higher levels of spastin expression in glioblastoma cell lines compared to NHA cells (Fig. 4C), consistent with results from immunocytochemical experiments (Fig. 1). Immunoblots of HEK cell lysates overexpressing full-length versions of human M1 and M87 spastin isoforms showed that all anti-spastin antibodies used in this study similarly recognized both M1 and M87 spastin isoforms (Fig. 5A). Moreover, immunoblots of T98G, NHA, and cortical cell lysates indicate that the main spastin isoform predominantly expressed in these cells corresponds to M87 (Fig. 5B). A weaker immunoreactive band running at a slightly lower molecular weight than M87 was also observed, likely corresponding to M87 lacking exon 4 (Fig. 5B) (20).
Consistent with results from immunoblotting experiments, quantitative real-time PCR experiments showed increased expression of spastin transcripts in human glioblastoma (U138MG, U87MG, T98G) and neuroblastoma (SK-N-SH) cell lines compared to NHA (Fig. 6). Collectively, these results indicated increased expression of spastin in human glioblastoma cell lines.
To determine whether spastin affects cell motility, this protein was depleted in T98G cells using spastin-specific siRNA and cell motility was monitored using time-lapse imaging for 10 hours. When compared to cells transfected with control siRNA, spastin-depleted cells displayed lower migration velocities (Fig. 7A). Velocities in control and spastin-depleted cells were 21.7 ± 13.48, and 16.5 ± 8.37 μm/hour (mean ± SD; n = 503 in control cells, n = 615 in spastin-depleted cells), respectively (p < 1 × 10−14) (Fig. 7A). When compared with controls, spastin-depleted cells had also shorter mean trajectory length. Mean track length during 10-hour time-lapse imaging in control and spastin-depleted cells were 176 ± 114, and 139 ± 80 μm(mean ± SD; n = 503 in control cells, n = 615 in spastin-depleted cells), respectively (p < 1 × 10−9) (Figure, Supplemental Digital Content 2, http://links.lww.com/NEN/A264). Examples of cell trajectories in control and spastin-depleted cells are shown in Figure 7B. Time-lapse imaging of control and spastin-depleted T98G cells revealed that some cells within the spastin-depleted cell population had distinctively long protrusions. Video, Supplemental Digital Content 3, http://links.lww.com/NEN/A265 demonstrates motility in control and spastin-depleted cells. One long protrusion was typically observed per moving cell, as documented by still images from time-lapse experiments (Fig. 8A, B). These protrusions, characteristic of spastin-depleted T98G cells, were packed by dense arrays of microtubules (Fig. 8D). Protrusions and changes in the organization of microtubules were not detected in control cells (Fig. 8C).
To assess the effect of spastin depletion on cell division, growth curves were determined in control and spastin-depleted cells (days 1–6 after siRNA transfection) together with immunoblot analysis of spastin expression. The number of viable cells was clearly higher in spastin-depleted cells; statistically significant differences were observed on day 3 after transfection (Fig. 9A); effective spastin depletion at this time point was confirmed by immunoblotting (Fig. 9B). Higher proliferation of spastin-depleted cells was also confirmed by demonstrating a higher amount of kinesin-like protein KIF11 (Eg5), which associates with mitotic spindles and represents a marker of mitotic cells (Figure, Supplemental Digital Content 4, http://links.lww.com/NEN/A266). Collectively, these results document substantial changes in cell motility and proliferation in spastin-depleted glioblastoma T98G cells.
In the normal brain, antibodies Sp/AAA, Sp3G11/1 and Sp6C6 produced similar, yet not identical, immunoreactivity profiles. In the cerebellum, Sp/AAA rendered widespread staining in Purkinje cell perikarya and apical dendrites in the molecular layer (Fig. 10A), Golgi II neurons (Fig. 10B) and to a lesser extent, in glial cells of the molecular layer (Fig. 10C). In contrast, antibodies Sp3G11/1 and Sp6C6 did not stain Purkinje cell bodies but instead rendered discrete punctate staining in the neuropil of the Purkinje cell and molecular layers (Fig. 10D) and in the cerebellar white matter (Fig. 10E, arrows), suggesting localization in axo-dendritic or axo-axonal synapses. Scant cytoplasmic staining was detected in glial cells with Sp3G11/1 (Fig. 10F), much as observed with the Sp6C6 antibody.
Compared to the normal brain, spastin immunoreactivity profiles were significantly increased in diffuse astrocytic gliomas according to an ascending gradient of malignancy. Immunolabeling with all 3 anti-spastin antibodies employed in this study was detected in 27/37 adult diffuse astrocytic gliomas. Ten out of 37 tumor specimens (9 GBM and 1 grade III anaplastic astrocytoma) showed no spastin immunoreactivity. Grade for grade, spastin immunoreactivity was significantly increased in GBM (p < 0.05 vs. grade II astrocytoma). Amongst the 27 tumor specimens exhibiting positive labeling by immunohistochemistry, the spastin median labeling index (MLI) for GBM was 17.2%; (IQR: 6.4%-38.4%) and 15.4% (IQR: 5.5%-36.7%) using antibodies Sp3G11/1 and Sp/AAA respectively, as compared to the low-grade (grade II) diffuse astrocytomas (MLI: 2.1%; IQR: 1.4%-5.6% and 1.7%; IQR: 0.7%-3.8%) also with antibodies Sp3G11/1 and Sp/AAA, respectively (p < 0.05). A similar trend was noted in pediatric tumors, but the number of available cases was too small for statistical analysis.
Spastin immunoreactivity profiles varied widely among different tumor specimens of the same histological grade and among different areas within individual samples, consistent with marked intratumoral staining heterogeneity. In diffuse low-grade astrocytomas (Fig. 11A, B, D, E) and anaplastic astrocytomas (Fig. 11C, F, I, J), spastin staining was detected in randomly dispersed tumor cells (Fig. 11B, D), including in areas of gray matter infiltration (Fig. 11A, B, G, H). Aside from robust somato-dendritic spastin labeling in entrapped neurons of the cerebral cortex and deep gray nuclei (Fig. 11A–C, G, H), overlapping diffuse and punctate localizations (Fig. 11E, F) (some with a distinctive tendency for the cell periphery) were noted in neoplastic glial cells (Fig. 11F, I, J). As a consistent trend (but not invariably), distinctive punctate staining (with a predilection for the cell periphery) was present in large pleomorphic/multinucleated (“ganglioid”) tumor cells with ample cytoplasm (Fig. 11F, I, J). Immunostaining of such tumor cells was almost indistinguishable from the punctate pattern of spastin localization in entrapped indigenous neurons (Fig. 11G, H), except for the presence of supernumerary and/or atypical nuclei in neoplastic glial cells (Fig. 11F, I, J).
Robust, mostly diffuse and multi-punctate spastin immunoreactivity was detected in tumor cells clustered around tumor blood vessels in GBM (Fig. 12A–C), but was also variously distributed throughout the tumor parenchyma (Fig. 12D–K). In contrast to spastin-expressing neoplastic cells there was a paucity of spastin labeling in hypertrophic endothelial cells in foci of angiogenesis (Fig. 12A–C). There was a diffuse and dense cytoplasmic staining pattern in the preponderant nondescript tumor cells (Fig. 12E, F). Randomly scattered tumor cells with an irregular multipolar astroglial-like morphology exhibited a variant robust fibrillary/filamentous staining (Fig. 12G). Strong, diffuse and/or micro-punctate, juxtanuclear staining was also noted in small (“anaplastic”) GBM cells and in the intervening tumor-infiltrated brain parenchyma in areas of pseudopalisading necrosis (Fig. 12H–J). Mitotic figures were for the most part spastin-negative in these sections (Fig. 11C), although weak spastin labeling was noted in a small number of mitoses (Fig. 12L). There was an overall trend for increased spastin expression in overtly astroglial morphologic phenotypes as compared to neoplastic cells with spindle cell/sarcomatoid features (not shown).
Immunoreactivity with Sp3G11/1 and Sp/AAA was qualitatively and semi-quantitatively more robust than that of Sp6C6. The labeling pattern with Sp3G11/1 was also more micro-punctate as compared to that of Sp/AAA on paraffin sections. However, no statistically significant differences were detected in tumor labeling indices obtained using Sp3G11/1 and Sp/AAA antibodies.
Collectively, the immunohistochemical findings on the tumor samples suggest that increased spastin expression represents a distinct feature of neoplastic glial phenotypes, especially in GBM.
This study demonstrates that in the context of human gliomas, the microtubule-severing ATPase spastin is expressed at higher levels in neoplastic glial phenotypes, especially in GBM, than in normal brain tissue. In the latter, spastin localization is predominantly neuronal and, only to a lesser extent, glia-associated. Increased spastin protein and transcripts levels were confirmed by immunofluorescence, immunoblotting and quantitative RT-PCR experiments in 3 human glioblastoma cell lines compared to cultured NHA. Notably, all 3 antibodies used in this study yielded similar immunofluorescence profiles on cultured glioblastoma cell lines. The lack of colocalization of spastin and GM-130 in these cells suggests that spastin does not undergo trafficking along the early secretory pathway. By immunohistochemistry, the anti-spastin antibodies Sp3G11/1 and Sp/AAA gave statistically significant increases in spastin labeling indices in primary, surgically excised, GBM specimens vs. grade II diffuse astrocytic gliomas. Collectively, these results indicate that increased expression of spastin is linked to an overall tendency toward high-grade glioma malignancy.
Spastin expression has been documented in brain, spinal cord, heart, kidney, liver, lung, pancreas, placenta and skeletal muscle (23). Our results in the normal human CNS confirm previous studies reporting a predominantly neuronal trend for spastin localization (24, 25). Our immunohistochemical studies also revealed different patterns of spastin immunoreactivity for different spastin antibodies in neuronal cells in situ. Specifically, the antibody Sp/AAA gave strong somato-dendritic and axonal Purkinje cell immunoreactivity. In contrast, antibodies Sp3G11/1 and Sp6C6 produce no Purkinje cell staining but instead rendered discrete puncta-like localizations in the neuropil, for example, at the boundary of Purkinje and molecular layers in the vicinity of the basket fibers and in the white matter. Because spastin localization has also been described in synapses (41), we speculate that these spastin-immunopositive puncta may represent axo-dendritic and/or axo-axonal synaptic sites. Also, the Sp/AAA antibody stained glial cell populations in the cerebellar molecular layer and white matter, whereas the Sp3G11/1 and Sp6C6 showed no significant glial cell immunoreactivity besides weak/scanty labeling in occasional astrocytes. Consistent with these observations, deletion experiments mapped the Sp3G11/1 and Sp6C6 epitopes to a region within amino acids 87 and 274, far away from the Sp/AAA epitope(s) located between amino acids 337 to 465 (M. A. Burns and G. Morfini, unpublished data). Thus, the differences in Sp3G11/1-Sp6C6 vs. Sp/AAA immunoreactivity most likely result from differential epitope availability in situ.
The human spastin gene yields 2 major isoforms (M1 and M87) that result from 2 different translational start sites (18, 21). The exclusion of specific exons (exons 4, 9 and 15) further increases the heterogeneity of spastin isoforms in different tissues (20). In normal human brain, 2 major spastin transcripts have been identified by RT-PCR, the most prominent one corresponding to full-length M87 and a less abundant one corresponding to M87 lacking exon 4 (20). Previous immunoblotting studies established selective expression of M87, but not M1, in the adult rat hippocampus and cerebral cortex (22). These observations, coupled with immunoblotting data herein, strongly suggest that NHA and glioblastoma cells mainly express M87 spastin. A weaker spastin-immunoreactive band running at a slightly lower molecular weight than full-length M87 was also observed. Based on prior studies, this band likely represents M87 lacking exon 4 (20). However, the expression of M87 spastin isoform variants was not directly evaluated in the present study.
Overlapping punctate and diffuse patterns of spastin localization were identified in primary tumors and in interphase cells of glioblastoma cell lines, independent of Golgi staining, as well as in mitoses. Intriguingly, recruitment of spastin to the leading edges of growing tumor cells, where few microtubules are present, was detected both in T98G cells and in tumor cells from clinical samples. Aside from actin-driven protrusions, including α-actinin-enriched ruffled membranes and lamelipodia (42), a similar recruitment to the leading edges of glioblastoma cells has been observed with the intermediate filament protein synemin, which contributes to the migratory properties of astrocytoma cells by influencing the dynamics of the actin cytoskeleton (43). We observed a substantial decrease in cell motility in spastin-depleted vs. control cells. These results were also supported by a radial cell migration assay (44) using T98G cells attached on laminin (E. Dráberová, unpublished observations). In some cells, spastin depletion resulted in generation of long protrusions but immunofluorescence microscopy in T98G cells did not reveal overt changes in microtubule organization in the close vicinity of plasma membranes. On the other hand, depletion of katanin, another microtubule-severing protein, in addition to affecting cell migration, reportedly results also in the generation of dense arrays of microtubules running parallel to the inner face of the plasma membrane (cell cortex) (45). Both proteins might, therefore, regulate different microtubule functions in moving cells. To our knowledge, there are no previous reports indicating involvement of spastin in cell motility.
Gene expression profiling experiments in gliomas have shown upregulation of genes related to motility, and functional studies demonstrated that increased cell motility may be the main contributor to the invasive phenotype of diffuse gliomas (46). Similarly, the commitment of glioma cells to migrate and invade appears to be inversely related to proliferative activity (47). Intriguingly, depletion of spastin in T98G cells in the present study also resulted in inhibition of cell migration and stimulation of cellular proliferation. These data suggest that spastin levels may correlate with changes in proliferation and migratory rates, thus making spastin a novel determinant in the emergence of divergent proliferating or invasive cell populations in gliomas. Accordingly, the inverse and dichotomous role(s) of spastin in cell motility vs. proliferation in glioblastoma cells are consistent with the concept that migration and proliferation constitute antagonistic cellular behaviors within a glioma cell population (46). Further studies are needed to elucidate the role of spastin in the growth and invasion of gliomas, as this microtubule-severing enzyme may be subject to differential regulation by divergent signaling effectors within the brain tumor microenvironment. The permissiveness of the cellular environment may promote migration and decrease proliferation and pro-apoptotic disposition in tumor cells; collectively, these effects may contribute to tumor resistance to chemotherapy and radiation therapy (46).
Abnormalities in microtubule dynamics and organization underlie a common mechanism for genetic instability in cancer cells and for altered tumor cell architecture; however, the molecular mechanisms underlying such changes are poorly understood (48). In normal CNS development, spatial control of microtubule severing is important for mobilizing large numbers of microtubules at growth-related sites, such as centrosomes/microtubule organizing centers (49), sites of branch formation (50), and growth cones (51). Previous studies have shown that spastin is enriched in cell regions containing dynamic microtubules including the spindle pole, the central spindle and the midbody, as well as the distal axon and the branching points (52).
Microtubule severing may increase polymer mass by generating shorter microtubules that can serve as seeds for nucleating new microtubules (53, 54). In previous studies, we and others, demonstrated overexpression and aberrant patterns of non-centrosomal γ-tubulin compartmentalization in diffuse astrocytic gliomas and glioblastomas (36, 37, 55), as well as in breast cancer cell lines (56), suggesting ectopic microtubule nucleation in cancer cells (55, 56). Interestingly, spastin interacts with centrosomal proteins and co-fractionates with γ-tubulin (52), a form of tubulin essential for centrosomal (57, 58) and non-centrosomal (59, 60) microtubule nucleation. Future studies are warranted to determine whether γ-tubulin overexpression may affect abnormal microtubule severing activity and aberrant microtubule nucleation in cancerous glial cells.
Spastin labeling was present in both human cell lines and tumor cells of surgically excised GBM, but more robust and widespread labeling was consistently encountered the former. A striking feature in GBM specimens was markedly heterogeneous immunoreactivity profiles associated with large areas of non-staining of the tumors. This cellular distribution pattern contrasts with that of perivascular and/or randomly dispersed aggregates of tumor cells that exhibited unequivocal spastin staining. This again raises the question of differential epitope availability. Moreover, hindered detection of spastin by immunohistochemistry in formalin-fixed tissues may be due to epitope loss as a consequence of proteolysis in deeper portions of surgically excised or biopsy tissue samples because of slower tissue penetration by the fixative. Alternatively, it may be related to epitope masking due to protein conformational changes by formaldehyde-induced cross linkages or other factors related to histological processing and/or embedding.
To our knowledge, this is the first study demonstrating increased levels of spastin expression in human glioblastoma cell lines vs. NHA. Moreover, greater spastin immunoreactivity was found in clinical samples of GBM compared to low-grade diffuse astrocytic gliomas or nascent indigenous astrocytes, albeit with markedly heterogeneous cellular distribution. These results indicate that spastin overexpression in GBM is linked to a trend toward high-grade malignancy and aggressive growth potential. While aberrant upregulation of spastin in a subset of tumor cells may not necessarily have functional significance, the observation of spastin enrichment in the leading edges of glioblastoma cells in vitro and in situ, coupled with results from functional experiments, call for future investigations into the possible role of spastin in tumor cell motility and invasion. Because increased spastin expression may promote cellular migration in the face of decreasing proliferation, targeting spastin may offer a promising therapeutic strategy directed against glioma cell invasion with a potentially added benefit on tumor cell response(s) to conventional cytotoxic and/or tubulin-targeted treatments.
Supported in part by Grants # P302/10/1701 (to ED) 204/09/1777 (to PD) and 204/09/H084 (to SV) from the Czech Science Foundation, grant KAN200520701 (to VS) from Grant Agency ASCR, grant LC545 from Ministry of Education, Youth and Sports of the Czech Republic and by the Institutional research support (AVOZ 50520514) (to PD); Institutional research support (AVOZ 50110509) (to JJ); NIH R01 NS066942A and ALS Therapy Alliance research grant (to GM); NIH RO1 NS28785, NSF 0841245, and State of Pennsylvania Tobacco Settlement Funds (to PWB); and grant #203 from the St. Christopher’s Foundation for Children (to CDK).