p27^Kip1 expression inhibits glioblastoma growth, invasion, and tumor-induced neoangiogenesis

p27^Kip1 (hereafter p27) belongs to the Cip/Kip family of cyclin-dependent kinase inhibitors that binds and inhibits, although with a different threshold, all the cyclin/cyclin-dependent kinase complexes, thus often referred to as an universal cyclin-dependent kinase inhibitor (1). It has been extensively shown that for the proper control of cell proliferation p27 has to bind and regulate the cyclin-dependent kinase nuclear activity, because its displacement in the cytoplasm results in loss of cell cycle inhibition (2–4). Recently, several research groups reached the conclusion that diverse biological functions of p27 are correlated to its cytoplasmic localization. Among these studies, the role of p27 on cell motility has become an interesting matter of scientific literature (5–9). We recently reported that cytoplasmic p27 interacts with and inhibits the microtubule-destabilizing protein stathmin, thus inhibiting sarcoma cell motility through extracellular matrix (ECM) substrates (5).

Cell cycle deregulation is a hallmark of cancer; in particular, reduced expression of p27 has been extensively observed in human cancers and often its low levels are associated to a worse prognosis (10, 11). Moreover, lack of p27 has also been linked to increased invasion and metastasis formation in several types of human tumors (12–17). Few studies have addressed systematically the effect of p27 cytoplasmic localization in tumor progression. The largest one done in breast carcinomas showed that patients with high cytoplasmic expression of p27 had a better survival with respect to patients with low expression, although the best prognosis was found in patients with high nuclear p27 staining (18). In malignant glioblastoma, low p27 expression levels have been often associated with poor prognosis and increased proliferation (19–22). Smaller studies did not confirm this correlation (23, 24), a discrepancy that could be due to the heterogeneity of gliomas (25) or to the different cutoff used to evaluate p27-positive cells within the tumors. Malignant glioblastoma represents an ideal candidate for local gene therapy approaches, because it prevalently invades the surrounding parenchyma rarely forming distant metastasis and because it is almost invariably resistant to current therapeutic approaches. Being p27 an attractive target for new therapeutic opportunity in cancers (3, 10), we explored the effects of p27 expression using a tetracycline-inducible adenoviral system in vitro and in vivo gene therapy experiments on glioblastoma cells.

A total of eight brain tumors were collected and diagnosed at University La Sapienza Rome according to the WHO criteria. For immunohistochemical staining, four glioblastoma multiformes (grade 4) and four anaplastic astrocytomas (grade 3) were evaluated using an anti-human p27 antibody from DAKO.

Human embryonic kidney cells [HEK 293, ATCC CRL-1573, and 293 H (Life Technologies)] were used to generate recombinant adenovirus. Human malignant glioma cell lines U87MG, U138MG, T98G, U251MG, SF268, SBN75, U373, SF539, SF295, and SBN19 were kindly provided by Prof. A. Fusco (Università Federico II di Napoli) and maintained in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum.

Human umbilical vein endothelial cells (HUVEC) were isolated from human umbilical cords as described previously (26). Cells were grown on 1% gelatin-coated plates in M199 supplemented with 20% fetal bovine serum, 1% penicillin/streptomycin, and bovine brain extract (0.5%).

Total protein and nuclear/cytoplasmic extracts were prepared as reported previously (5). For immunoblotting, protein extracts were separated in 4% to 20% SDS-PAGE (Criterion Precast Gel; Bio-Rad) and transferred to nitrocellulose membranes (Amersham). Primary antibodies were from Transduction Laboratories (p27, CDK1, and CDK2), Santa Cruz Biotechnology (p27 C19, p27 N20, p27, CDK1, cyclin B1, vinculin, Akt, and c-Abl), Sigma (stathmin and α-tubulin), Cell Signaling (pS473Akt), and Upstate Biotechnology (histone H1).

Membranes were incubated with primary antibodies according to experimental design and then with horseradish peroxidase–conjugated (ECL Kit; Amersham) or Alexa-conjugated secondary antibodies (Odyssey Infrared Detection System; Li-Cor). Immunoprecipitation experiments were done using 0.5/1.0 mg total lysate and protein A/G-Sepharose (Amersham). True Blot reagent (e-Bioscience) was used as secondary anti-mouse or anti-rabbit antibody.

Human glioblastoma xenografts were established by s.c. injection of 10⁷ U87MG cells into female athymic nude mice (Harlan; 8 weeks old). When tumors were ∼40 to 60 mm³ (∼15 days from injection), the animals (10 mice per group of treatment) were randomly divided into groups according to experimental design and intratumoral injections were done with 2 × 10⁹ total infectious units (AdTRE/AdTet-ON, 80:40) and repeated five times every 2 days. Tumor size was measured with a caliper three times weekly and volume was calculated as 0.5 × length × width². Unless differently indicated, animals were sacrificed after 15 days of treatment and tumor analysis was done.

In the tumor prevention models, 7 × 10⁶ U87MG cells were transduced with 80:40 AdTRE/AdTet-ON and then harvested 48 h later. Cells were then washed, resuspended in DMEM without red phenol, and injected s.c. into nude mice (8 mice per group of treatment). Animals were sacrificed after 25 days for tumor analysis. For in vivo induction of p27 expression, 2 days before cell injection and for all the course of the experiment, drinking water was supplemented with 1 mg/mL doxycycline and 2.5% (w/v) sucrose and changed every 2 days.

Explanted tumors were formalin-fixed and included in paraffin. At least 10 sections per tumor stained with H&E were analyzed in blind by an expert pathologist to evaluate the presence of local invasion.

For microvessel density analysis, explanted tumors were included in OCT and frozen in liquid N₂. Sections were incubated with anti-mouse CD31 (PECAM) monoclonal antibody (1:20; BD PharMingen) and visualized via three-step staining procedure in combination of biotinylated polyclonal anti-rat Ig mouse-adsorbed (Vector Laboratories; 5 μg/mL), and streptavidin-horseradish peroxidase (Sigma; 1:400) using 3,3-diaminobenzidine as substrate. The slides were then counterstained with hematoxylin. The number and the length of CD31⁺ vessels were calculated in blind using a Leica microscope coupled with the Leica IM software. Ten randomly selected fields per section and at least four sections per tumor were analyzed. For each treatment group, four different tumors were analyzed. Only vessels >10 μm in length were considered.

Adhesion and migration assays were done essentially as described previously (5, 27). Briefly, for migration, bottoms of HTS Fluoroblok (Becton Dickinson) were coated overnight at 4°C with 20 μg/mL fibronectin, 10 μg/mL vitronectin, 10 μg/mL collagen I, or 10 μg/mL collagen VI and then saturated 2 h at room temperature with PBS-1% bovine serum albumin. Cells were labeled with DiI (Molecular Probes) for 20 min at 37°C before being seeded in the Fluoroblok upper chamber and then incubated at 37°C for the indicated times. Migrating cells were evaluated by reading at different time points the lower and upper sides of the membrane with the Spectrafluor reader (Tecan). Each experiment was done at least three times in duplicate. In some experiments, migration was blocked at the indicated time point, by fixing the Fluoroblok membranes in 4% paraformaldehyde and the Fluoroblok membrane mounted on a slide, to allow the count of migrated cells.

For three-dimensional Matrigel evasion assay, cells (3 × 10⁵/mL) were transduced as indicated and 48 h later were included in Matrigel (6 mg/mL; Becton Dickinson) or collagen I (1 mg/mL; Becton Dickinson) drops and incubated for the indicated times in complete medium. Cell motility was observed by transmission microscopy using a Nikon TS100/F microscope. Images were collected using a digital camera (Nikon).

For three-dimensional Matrigel invasion assay, glioblastoma cells were transduced as indicated and 48 h later seeded in the Fluoroblok upper chamber coated with 80 μg/mL Matrigel and then incubated at 37°C for 2 days. Fluoroblok membranes were then fixed in 4% paraformaldehyde, cells were stained with propidium iodide as indicate above, and the Fluoroblok membranes were mounted on a slide to allow the count of migrated cells. At least 10 randomly selected fields per membrane were counted in two different experiments done in duplicate.

HUVEC were transduced using a 60:30 recombinant AdTRE/AdTet-ON in the presence of 1.0 μg/mL doxycycline 48 h before seeding on Matrigel (BD PharMingen; 12.9 mg/mL)–coated LabTek (55,000 cells per chamber) and incubated for 2 h with normal growth medium or U87 conditioned medium. In some experiments, HUVEC were also transduced with Ad/stathmin small interfering RNA (siRNA) at multiplicity of infection 500 and 24 h later with AdTRE p27^wt as described above. Nontransduced HUVEC were used as control. Results are expressed as number of tube-like structures per field (magnification, ×200).

The others methods used in this work are fully described in the Supplementary Material.⁶

To study the role of p27 in human glioblastoma, p27 expression and localization was evaluated in 10 different glioblastoma cell lines by Western blot. Except for T98G cells, p27 is generally expressed at low levels in the analyzed glioblastoma cell lines (Fig. 1A, left). Differential extraction of cytoplasmic and nuclear proteins showed that in the majority of these cell lines p27 presented an enriched nuclear localization, although some protein was also present in the cytoplasm (Supplementary Fig. S1B).⁶ A comparison of only the cytoplasmic fractions revealed that in U251 and SBN75 cell lines p27 was almost completely absent, whereas two cell lines, T98G and U138MG, expressed high levels of cytoplasmic p27 (Fig. 1A, right). The analysis of p27 expression and localization by immunohistochemistry on a small panel of primary human brain tumors confirmed the results obtained on cell lines showing that high-grade brain tumors express low levels of p27 protein. In fact, all glioblastoma cases (n = 4) showed a lower expression of p27 compared with anaplastic astrocytoma (n = 4; Supplementary Fig. S1A).⁶ In both types of tumors, p27 staining was predominantly nuclear.

To assess if the different cellular localization of p27 could be associated to differences in the growth rate and/or ECM-driven cell motility, proliferation and migration assays were done on four representative cell lines: T98G and U138MG cells displaying strong cytoplasmic p27 expression and U87MG and SF268 displaying low cytoplasmic levels. Results showed that no direct correlation could be established between cytoplasmic levels of p27 and proliferation rates (Fig. 1B), whereas T98G and U138MG cells, which had strong p27 cytoplasmic expression, exhibited a decreased migration rate through collagen IV (Fig. 1C), vitronectin, collagen I (Supplementary Fig. S1C),⁶ and fibronectin (Fig. 2; data not shown) compared with SF268 and U87MG cells that expressed low cytoplasmic p27 levels.

We showed previously that a p27 deletion mutant lacking the last 28 amino acids (p27^1-170) retains the growth inhibition ability while failing to inhibit cell motility through ECM substrates (5). Based on this notion and to determine the effects of p27 overexpression on growth and motility of glioblastoma cells, different recombinant adenoviruses (AdTRE) were prepared expressing p27^wt, p27^1-170, and p27^T187A (see Supplementary Material).⁶ This last protein carries a point mutation resulting in the substitution of threonine 187 with alanine, which in turn impairs p27 degradation via the ubiquitin-dependent proteasome pathway (28, 29), thus representing a better control of the p27^1-170 mutant than the wild-type protein, because it also lacks the T187. Standardization of this expression system for different glioblastoma cells is described in Supplementary Fig. S2.⁶ It consisted in determining the most suitable cotransduction ratio and doxycycline concentration to achieve a good inducible protein expression in at least 80% of transduced cells. As a control, AdTRE β-Gal (able to express the β-galactosidase protein on induction with doxycycline) was used. In all the cell lines tested, similar expression of the three p27 proteins was obtained (Supplementary Fig. S2; data not shown).⁶ Biological characterization of p27 overexpression on cell growth was done on U87MG and U251MG. The three adenoviruses were able to inhibit in vitro proliferation of both cell lines with a similar profile as shown by cell growth curve (Supplementary Fig. S3A)⁶ and fluorescence-activated cell sorting analysis of DNA content (Supplementary Fig. S3B).⁶ In long-term growth assay (that is, colony assay), all the adenoviruses were able to significantly impair cell growth, although AdTRE p27^1-170 was the most effective one (Supplementary Fig. S3C).⁶

To determine the effects of the different p27 proteins on glioblastoma cell motility, U87MG and U251MG cells were transduced with the AdTRE p27/AdTet-ON and then tested in migration experiments on FN (Fig. 2A). Overexpression of p27^wt and p27^T187A induced a significant impairment of U87MG cell migration (65% and 67.5% of inhibition, respectively), whereas p27^1-170 was much less effective (30% of migration inhibition; P ≤ 0.01 versus p27^wt and p27^T187A, Student's t test; Fig. 2A). Conversely, overexpression of p27 did not alter the migration pattern of U251MG cells (Fig. 2B). In agreement with the above results, p27^wt and p27^T187A inhibited U87MG cell motility also in other migration assays, evaluating cell movements under three-dimensional conditions (Fig. 2C and D). Again, p27^1-170 failed to properly inhibit cell migration in both types of assays, confirming that the COOH-terminal portion of p27 is required to fully inhibit ECM-driven cell motility (5). Also in the three-dimensional experiments, U251 cells transduced with the three AdTRE p27 migrated at a similar rate (data not shown). To understand the molecular basis of this different biological behavior, we first investigated the localization of endogenous and overexpressed p27 proteins in U87MG and U251MG cells adhered to fibronectin for 1 h. U87MG cells showed a predominant cytoplasmic localization in either nontransduced or transduced cells as evaluated by Western blot and immunofluorescence analysis (Fig. 3A and B). In contrast, in U251MG cells, p27 mainly localized in the nuclear compartment (Fig. 3). Western blot analysis revealed that, albeit very small, a fraction of overexpressed protein is also present in the cytoplasm of U251MG cells, but it is mainly cleaved at its COOH terminus as revealed using antibodies recognizing the COOH terminus (C19) or NH₂ terminus of p27 protein (N20; Fig. 3A, right). The caspase inhibitor Z-VAD -FMK almost completely prevented p27 cytoplasmic degradation (Fig. 3C), suggesting that the cytoplasmic cleavage of p27 in U251MG cells is dependent on caspase activity.

Doxycycline-regulated p27 expression impairs glioblastoma growth and invasion in vivo. To verify whether in vivo the various AdTRE p27 were able to inhibit glioblastoma cell growth, U87MG cells, known to be able to grow when implanted s.c. in nude mice, were used. On the contrary, U251MG cells could not be used because they are unable to form tumors in the same type of assay (30). U87MG xenografts were established using pretransduced cells and their growth was followed in animals fed with doxycycline. As controls, untreated and AdTRE β-Gal-transduced cells were used. From all controls, we obtained similar results (Fig. 4A; data not shown) and Western blot analysis showed that also in vivo the transgene expression was regulated by doxycycline (Fig. 4B, right; data not shown). In this experiment, AdTRE p27^1-170 transduction was the most effective in growth inhibition (60%), whereas p27^wt and p27^T187A overexpression gave 43% and 54% of tumor growth inhibition, respectively, in line with the in vitro experiments (Supplementary Fig. S3C).⁶ These data showed that p27 overexpression is able to reduce glioblastoma cell growth in vivo.

Next, we asked whether the various AdTRE p27 could be used to treat glioblastoma in vivo. To this aim, U87MG xenografts tumors were established in nude mice by s.c. injection. Tumor treatments, consisting of five intratumoral injections of the various AdTRE p27, started as the tumors reached the volume of ∼50 mm³ (Fig. 4B, left). In a pilot experiment (n = 3 mice per group of treatment), mice transduced with AdTRE p27^wt and AdTRE p27^1-170 were fed or not with doxycycline and the tumor growth was evaluated every 3 days up to 24 days after injection unless noted. Induction of p27 expression by doxycycline treatment (Fig. 4B, right) reduced tumor growth and increased the survival of mice of ∼10 days (Fig. 4B, left), showing that growth inhibition was specifically due to p27 expression and not to adenoviral transduction. In a larger experiment (n = 8-10 mice per group of treatment), all AdTRE p27 tested significantly inhibited U87MG growth (Fig. 4C, left) compared with untreated and AdTRE β-Gal-transduced tumors. When tumor growth inhibition was evaluated using the tumor weight as a variable, a significantly stronger activity was observed for AdTRE p27^wt and p27^T187A compared with AdTRE p27^1-170 (60% and 70% versus 45% of tumor growth inhibition, respectively; Fig. 4C, left).

Histologic analysis of explanted tumors was carried out to search for signs of local invasion. Samples were judged as positive when at least two foci per section of invasive growth were found. Interestingly, we found that only one of seven different AdTRE p27^wt-treated tumors exhibited signs of local invasion (12.5%; Fig. 4C, right), whereas untreated and AdTRE p27^1-170-treated tumors showed a local invasion in 57% and 60% of cases, respectively.

p27 inhibition of cell invasion was linked to its ability to bind the microtubule-destabilizing protein stathmin (5). Accordingly, in tumors treated with AdTRE p27^wt, but not with AdTRE p27^1-170, stathmin coprecipitated with the transduced protein (Fig. 4D), whereas the two proteins similarly bound to cyclin/cyclin-dependent kinase complexes (Supplementary Fig. S3D-F; data not shown).⁶ Overall, these data showed that intratumoral gene therapy approach based on p27 overexpression is able to significantly inhibit glioblastoma cell growth and invasion and that the COOH-terminal portion of p27 is necessary for the proper p27 activity in vivo.

The data collected using the intratumoral treatment were in apparent contrast with the tumor growth inhibition observed when U87MG cells were transduced with the various AdTRE p27 before tumor cell injection in nude mice. In this type of experiment, in fact, AdTRE p27^1-170 showed the strongest tumor growth inhibition, in accord with in vitro experiments. Thus, we speculated that intratumoral injections of AdTRE p27 could function not only on cancer cells but also by altering the tumor microenvironment. This hypothesis was also supported by the fact that AdTRE p27^wt and p27^T187A exerted a higher growth-inhibitory activity when injected in preformed tumors with respect to when pretransduced in U87MG cells (compare Fig. 4A and C).

Previous works showed that inducible overexpression of p27^wt and p27^T187A in endothelial cells blocked DNA replication, inhibited cellular migration and tubulogenesis in vitro, and impaired angiogenesis in a mouse model of hind limb ischemia (7). Taking into account these results and our in vivo observations (Fig. 4), we investigated whether p27 overexpression had any effect on U87MG tumor neoangiogenesis. To this aim, we analyzed microvessel density in explanted tumors derived from the intratumoral or the pretransduced treatments using the anti-CD31 (PECAM) antibody. When AdTRE p27^wt and p27^T187A were administered intratumorally, they significantly reduced the number (Fig. 5A) and the length (Fig. 5C) of CD31⁺ vessels with respect to control, AdTRE β-Gal, or AdTRE p27^1-170-treated tumors. Conversely, when the same analysis was carried on tumors formed by pretransduced cells, no significant difference was observed among the treatments, and tumors formed by untreated or pretransduced U87MG cells presented similar vascularization (Fig. 5B and C). A semiquantitative proteomic analysis of U87MG conditioned medium derived from nontransduced and AdTRE p27^wt-transduced cells on 174 known cytokine showed no significant differences in the amount of proangiogenic factors produced by U87MG cells (Supplementary Table S1),⁶ supporting the idea that the different tumor vascularization was due to a direct effect of p27 on mouse endothelial cells.

To dissect the role of p27 on tumor-induced angiogenesis, HUVEC were used as a model system. We first evaluated the effects of U87MG conditioned medium on HUVEC growth and motility, comparing it with the HUVEC standard growth medium. U87MG conditioned medium did not significantly stimulate HUVEC growth (data not shown) but profoundly increased their ability to form tube-like structures when cells were seeded on top of a Matrigel matrix (Supplementary Fig. S4A).⁶

Next, AdTRE p27 overexpression in HUVEC was standardized (Supplementary Fig. S4B-E).⁶ Data showed that similar levels of protein expression were achieved using the three AdTRE p27 (Supplementary Fig. S3D)⁶ and that, as expected, overexpression of p27^wt, p27^1-170, and p27^T187A resulted in a similar proliferation inhibition (Supplementary Fig. S4E).⁶ Then, the effect of p27^wt, p27^1-170, and p27^T187A on HUVEC motility stimulated by U87MG conditioned medium was evaluated. p27^wt and p27^T187A significantly impaired tube-like structures formation by HUVEC (4.8 ± 2.1 and 5.5 ± 3 structures per field, respectively, with respect to 18 ± 8 of nontransduced cells; P ≤ 0.01), whereas no significant difference was induced by p27^1-170 with respect to control cells (11 ± 4 versus 18 ± 8, respectively; P = 0.2; Fig. 6A). We described previously that the negative effects of p27 on sarcoma cell migration requires two fundamental modifications: the cytoplasmic localization of p27 in response to cell-ECM interaction and the consequent p27 binding to the microtubule-destabilizing protein stathmin (5). We thus verified whether in endothelial cells the same mechanism was functioning. Nucleocytoplasmic translocation of p27 was clearly observed in HUVEC after cell adhesion to fibronectin (Fig. 6B). Accordingly, an increased adhesion-dependent association between p27 and stathmin was observed by coimmunoprecipitation (Fig. 6C). These data support the hypothesis that p27 impairs HUVEC motility by inhibiting stathmin activity. To prove this point, we down-regulated stathmin in HUVEC using a specific siRNA. Inhibition of stathmin expression resulted in a strong inhibition of tube-like structures formation (Fig. 6D); more importantly, the overexpression of p27^wt protein in stathmin silenced cells did not further inhibit HUVEC motility on Matrigel (Fig. 6D). This result suggests that stathmin expression is necessary for proper endothelial cell motility and that its association with p27 could be an important event in the regulation of endothelial cell motility both in vitro and in vivo. Accordingly, endogenous p27/stathmin colocalization was readily observed in cells forming tube-like structures on a three-dimensional Matrigel matrix as shown by the immunofluorescence analysis (Supplementary Fig. S4F).⁶

Many previous works suggested that p27 could represent a proper candidate for targeted gene therapy in several types of cancer including glioblastoma (31–37), and our data support and expand our knowledge of the tumor-suppressive role of p27 in vivo at least in mice.

Our in vitro work showed that p27 is able to inhibit both cell growth and motility in glioblastoma cells. The latter effect depends on the COOH-terminal portion of the protein and on its cytoplasmic displacement as supported by several evidences presented here. (a) The analysis of p27 expression and localization coupled with motility and proliferation experiments in different glioblastoma cell lines (Fig. 1; Supplementary Fig. S1)⁶ support the concept that p27 inhibits cell growth when located into the cell nucleus, whereas the control of cell motility is to ascribe to its cytoplasmic pool. (b) In U87MG cells, cytoplasmic p27 expression inhibits motility through its COOH-terminal portion (Figs. 2 and 3). (c) In U251MG cells, p27 overexpression did not interfere with cell motility probably due to the fact that in these cells p27 is retained in the nucleus and is cleaved at its COOH terminus by caspase activity (Fig. 3). Previous work showed that caspase are able to cleave p27 and p21 (38), although its significance for p27 was not assessed. Our data showed that in U251MG cells cleavage of p27 occurs at high rate in the cytoplasm, suggesting that several mechanisms control p27 expression, localization, and activity in glioblastoma cells.

Here, we provide the first demonstration that p27/stathmin interaction plays a pivotal role in the control of tumor cell invasion in in vivo models, confirming our previous results using in vitro experiments (5). Accordingly, recent data suggest that the knockout of p27 in human mammary carcinoma cells results in increased cell growth and motility (39), supporting again its tumor-suppressive role, at least in human cancer cells. However, a protumorigenic role for p27 has been recently shown in mice (40); however, whether the “oncogenic” activity of p27 depend on its cytoplasmic or nuclear part is still to be determined and whether it is dependent on its role in the control of cell motility is similarly unexplored (41).

The use of different mutants allowed us to show that p27 may also play important roles in the control of tumor microenvironment modification, specifically inhibiting tumor-induced neoangiogenesis. At least part of the effects exerted by p27 seems to be mediated by its interaction with stathmin in the cytoplasm of tumor and endothelial cells. This was shown either by the use of a deletion mutant unable to bind stathmin or by the use of a siRNA approach on HUVEC. It is increasingly clear that modulation of microtubule dynamics plays an important role in the regulation of several aspects of cell physiology, including cell motility (42). Thus, it is not surprising that interfering with stathmin activity can result in the alteration of several aspects of the cell physiology. Accordingly, Atweh et al. recently confirmed a prominent role for stathmin in the control of HUVEC motility (43), thus independently confirming our finding.

Previous work on U87MG cells injected in nude mice showed that these cells locally invaded the surrounding tissues and suggested that stable p27 silencing could impair this effect (44). We therefore tried to inhibit p27 expression in U87MG cells using our adenoviral siRNA system. None of the tested siRNA (included the one used by Wu et al.) was able to significantly reduce p27 levels in these cells (Supplementary Fig. S5),⁶ suggesting that other mechanisms, linked with cell clones selection, could be responsible for the reported effects in that study. Other studies suggested that the effects of p27 on cell motility could be due to the inhibition of the RhoA-ROCK pathway (6, 44). It is increasingly clear that RhoA hyperactivation inhibits two-dimensional motility, e.g., when wound healing or two-dimensional random motility assays are used, whereas it increases cell migration and invasion when the cells are tested in a three-dimensional context and in vivo (45–48). Moreover, several observations suggest that altering the microtubule network results in altered RhoA activity and/or localization, which in turn could contribute to altered cell motility (reviewed in ref. 49). Thus, it is conceivable that p27 could indirectly contribute to regulating RhoA activity through its interaction with stathmin and the consequent modulation of microtubule dynamics. Indeed, several experimental evidences produced in our laboratory support this hypothesis.⁷

In summary, we report here that, at least in glioblastoma, p27-based gene therapy could represent a promising approach and provide evidence that its tumor-suppressive activity is due not only to block of cell proliferation but also to tumor microenvironment modification, opening new way to look at gene therapy approach in human cancer.

No potential conflicts of interest were disclosed.

We thank Sara D'Andrea for excellent technical assistance.