Gliomas are primary tumors of the central nervous system. The most malignant form, named glioblastoma, has a dismal prognosis. Even with the best currently available treatment, the median survival time from diagnosis is 15 months [1
Besides the typical radio- and chemo-resistance, the high infiltrative phenotype of glioma cells in the brain parenchyma is a crucial feature of gliomas that makes surgical resection ineffective and relapses very common [2
]. In this context, a better understanding of the mechanisms underlying cell infiltration is essential to defeat this tumor and many studies contribute in dissecting the molecular pathways underlying this process [3
]. One of the analyzed mechanisms, typical of metastatic epithelial tumors, is the epithelial to mesenchymal transition (EMT) [5
]. In this process, the activation of signaling pathways such as TGF-β or Wnt/β-catenin induces a switch from E-Cadherin (Cdh1) to N-cadherin (Cdh2) that loosens the adherent junctions and promotes cell migration [3
]. In gliomas, the role of EMT is still debated. Many studies show the upregulation in gliomas of genes typically involved in EMT [4
], but the role of cadherins is still under analysis. On the contrary of other epithelial tissues, neural cells normally express Cdh2 instead of Cdh1 [11
]. Moreover, while some data indicate that Cdh2 expression correlates with tumor grade or invasiveness [14
] other studies reported opposite results [17
]. Camand and collaborators explained this contradiction with a difference between Cdh2 mRNA and protein expression level, with the former highly expressed and the latter scarcely expressed in glioma samples. The same authors further demonstrated that Cdh2 downregulation increases cell migration of both normal neural cells and glioma cells [17
] showing an anti-migratory role for Cdh2 in neural tissue.
In line with the previous data we recently demonstrated, for the first time, the existence of an alternative cadherin switch during gliomagenesis taking place between Cdh2 and Cdh4 [20
], a classical cadherin is expressed in the neural tissue both during development and adult life [13
]. This event, that takes place at the adherent junctions of malignant cells, allows glioma cells overcoming the mechanisms of cell–cell inhibition of proliferation (CIP) and migration (CIM), and makes them proliferate in an uncontrolled manner and infiltrate the brain parenchyma. The results showed that Cdh4 is necessary for glioma invasion and for the maintenance of the tumorigenic potential of these glioma cells. All these data clearly show that a cadherin switch plays a crucial role in regulating proliferation and cell migration in glioma in a similar way to what has been demonstrated for epithelial tumors, although involving slightly different molecular players. Moreover, Cdh4 expression seems important for tumorigenesis in different tumor types. However, the role of this molecule has not been fully clarified yet, since in some tumors it promotes malignancy while in other it acts as a tumor suppressor [27
In the present study, to elucidate the role of Cdh4 on glioma cells migration, we extended our analysis in patient-derived glioblastoma initiating cells (GIC). Our data suggest that Cdh4 is a useful prognostic marker for glioma and, although down-regulation is not sufficient to fully inhibit glioma invasion as was in the murine model, we showed a partial rescue of CIM mechanisms in vitro and a reduced migratory activity in vivo. Moreover, Cdh4 silencing reduces the tumorigenic potential of GIC cells in vivo.
Cdh4 is typically expressed in the nervous system during both development and adulthood [13
]. Many studies showed its central role in regulating neuronal progenitors migration [38
] especially during radial migration in the cortical plate along radial glia cells [26
]. This cells/neurites guidance activity seems at least in part due to the combinatorial cadherins expression that during development generates a molecular code allowing cell sorting and fasciculation based on preferential cadherin homophilic dimerization [21
]. The Cdh4 role in cancer is still less clear since in some studies it seems to have oncogenic properties [29
] maintaining stemness features and increasing tumor cell migration and malignancy, while in other cases it acts as an oncosuppressor since its dowregulation increases malignancy in different tumor types [27
]. These contrasting results can be due to different functions performed by Cdh4 in different tumor types. It has been shown for other cadherins a context-depending activity, since both intracellular and extracellular matrix (ECM) molecules can highly influence cadherin adhesive properties modifying the strength of both cell-cell and cell-matrix junctions [44
]. A particular context can for instance promote a more pronounced migratory ability for Cdh1 expressing cells without the need of a cadherin switch [48
In a previous work on a glioma model in mouse, we demonstrated that Cdh4 is necessary to sustain cell infiltration and proliferation in vitro and in vivo by overcoming both CIP and CIM mechanisms. This characteristic is typical during the EMT process often occurring in tumor malignancy where, thanks to a switch between Cdh1 and Cdh2, tumor cells loosen cell-cell adhesivity [50
]. We demonstrated for the first time that during gliomagenesis an alternative cadherin switch occurs between Cdh2 and Cdh4 [20
In this work we focused on the role of Cdh4 in human glioblastoma. Altogether our data demonstrated that, in such a context, Cdh4 could be a useful prognostic marker. Its expression is heterogeneous within glioma patients and correlates with a shorter survival time in glioma patients and with the acquisition, by glioma cells, of high malignant features as the loss of cell-cell contact inhibition mechanisms. In line with our previous results on murine cells, we noticed a switch on the cell membranes between Cdh2 and Cdh4. In GBM-23, where Cdh4 is not expressed (Figure 2
a–c), Cdh2 is localized at the cell-cell junctions often forming visible septa (Figure 2
g) while, there is a delocalization in Cdh4 expressing tumors (Figure 2
h–j). Unlike in mouse however, Cdh4 downregulation does not revert this phenotype. Several reasons can explain this result as the presence of other molecules able to compensate Cdh4 activity. Alternatively, post-translational modifications on Cdh2 (as phosphorylation) could prevent its recycling on the membrane [51
]. Additionally, there is the possibility that the residual Cdh4 levels upon miRcdh4 transduction are still sufficient to induce Cdh2 delocalization.
Our results showed that, although Cdh4 silenced glioma cells can proliferate and infiltrate the brain parenchyma, these abilities result highly impaired. In an in vivo competition assay, Cdh4 downregulated cells migrate less than control cells and six months after orthotopic transplantation they represent less than 10% of the initial mixed cell population. We wonder whether the observed differences in the migratory behavior could have alternative explanations and, in particular, could depend on the known difference in cell proliferation. Although proliferation cannot explain, by itself, differences in migration, and faster proliferative activity might lead to a more abundant cell population and the different population size might, in turn, translate in an artefactual difference in migration. In fact, for purely statistical reasons, a larger cell population will have maximal distance from the injection site larger than the maximal distance reached by cells from smaller population, even if they have the same infiltrative abilities. In our experiments, however, Cdh4 cells (which show lower proliferation rate) were injected in a larger amount and, more importantly, they were still in larger amount at the time of the analysis (18–21 days post injection).
The Cdh4-dependent inhibition of CIP and CIM mechanisms allowing cells to proliferate and migrate notwithstanding the presence of other cells, can be explained by a Cdh-4 dependent loosening of the adherent junctions, as in the classical EMT, where a cadherin with a reduced adhesive strength (Cdh2) substitutes a cadherin with a higher adhesive strength [55
]. Further data on Cdh4 adhesion strength will help to clarify this point. An alternative explanation could be that Cdh4 expressing tumor cells can form mainly heterophilic interactions with parenchymal cells which expresses only Cdh2. Cdh4-Cdh2heterophilic intercellular dimers (trans-heterodimers) are known to be weaker than respective homodimers [26
]. Through its cytoplasmic tail, Cdh4 could also interferes with specific intracellular signaling pathways regulating cell proliferation and migration. Two of the main pathways dysregulated during the EMT process in malignant tumors are the Wnt and the Hippo pathways [3
]. The lack of β-catenin in the nucleus of the analyzed control and Cdh4-silenced GIC cells suggests however that the canonical Wnt pathway is not activated in this context. A data analysis on microarray assay performed on GICs with different Cdh4 expression level does not highlight a differential expression level in genes involved in the Hippo pathway as NF2
(Dr. Antonio Daga, Ospedale Policlinico San Martino, Italy personal communication). Even if this data can not completely rule out a role of this signaling pathway in gliomagenesis, it is unlikely that Cdh4 can influence its activity. Further experiments showing the level of Hippo pathway activation by analyzing, for example, the phosphorylation level and the localization of the downstream effector YAP1 are necessary to completely clarify this issue.
Other two intracellular signaling molecules, key regulators of cell proliferation and migration, are ERK and AKT and often they result activated in cancer [61
]. In our previous work on murine glioma cells we found a correlation between Cdh4 expression and ERK phosphorylation. In line with these results, we found in Cdh4-dowregulated GIC cells a reduction of the activated phosphorylated form of ERK respect to control GIC cells. This ERK inactivation can, at least in part, be responsible for the migratory and proliferative impairment of these cells.
4. Materials and Methods
Six microRNA against Cdh4 mRNA (sequences are available on request) were designed using the BLOCK.iT RNAi Designer software of Life Technologies (version, Carlsbad, CA, USA). Each microRNA was separately cloned using the BLOCK.iT method into the pCDB-GW retroviral vector together with an EmGFP cassette as described previously [65
]. The control vector is a pCAG:DsRed retroviral vector (kindly provided by Magdalena Goetz., Helmholtz Center, Munich, Germany).
Replication incompetent retroviral particles were obtained as previously described [66
4.2. Glioma Initiating Cells Culture
Human GIC cultures derived by glioma patients (kindly provided by Antonio Daga, Ospedale Policlinico San Martino, Genova, Italy) were maintained on Matrigel coated flasks (1:200; BD Biosciences, San Jose, California, USA) in 50% Neurobasal, 50% DMEM/F12 media (Life Technologies, Paisley, UK), 1X B27 Supplement (Life Technologies), 10 ng/mL bFGF, 20 ng/mL EGF (Peprotech, London, UK), 2 mM glutamine (Life Technologies) and 2 µg/mL heparin (Sigma-Aldrich, Milano, Italy).
Control and Cdh4 microRNA transduced cultures were sorted thanks to DsRed or GFP expression on a FACSAria II (BD Biosciences) as previously described [67
4.3. Animal Procedures
The in vivo experiments were approved by the Animal Ethics Committee (OPBA) of “Ospedale Policlinico San Martino” and by the Italian Ministry of Health (Authorization N° 859/2016-PR). All the experiments were performed according to the Italian law D. lgs 26/2014 and the European Directive 2010/63/EU of the European Parliament. In all the experiments were used the NOD.CB17-Prkdcscid/J strains (hereinafter referred as NOD/SCID) from ENVIGO (Huntingdon, UK).
Intracranial cell transplantation was performed as previously described [68
]. Briefly, deeply anesthetized animals were mounted on a stereotaxic table and the skull was pierced with a 22G needle. By using a Hamilton Syringe, 4 × 105
cells were injected at the following coordinates respect to the Bregma: 1 mm anterior; 1.5 mm lateral and 2.5 mm under the skull surface. At different time points, described in the Results section, brains were dissected and fixed over-night in 4% PFA at 4 °C and then cryoprotected in 20% sucrose over-night at 4 °C. Cryostat sections were stained with 1 µg/mL Hoechst-33342 (Sigma-Aldrich) for nuclei visualization.
Cultured cells were usually fixed in 4% PFA for 15 min and washed three times in PBS. For Cdh2 staining, cells were fixed in methanol/acetone 1:1 at −20 °C for 10 min and washed as previously described. For the immunostaining, we used the following primary antibodies: Rabbit anti-Cdh2 (AbCam, Cambridge, UK); rat anti-Cdh4 (MRCD5, Hybridoma Bank, Iowa City, Iowa, USA); mouse anti-GM-130 (BD Biosciences); rabbit anti-β-catenin (Thermo Scientific, Waltham, Massachusetts, USA). We then used the following fluorescence secondary antibody from Jackson Immunoresearch Laboratories: Dylight 549- and Cy2-conjugated goat anti-mouse IgG, Dylight 549- and Dylight 488-conjugated goat anti-rabbit IgG, Cy3- and Alexa Fluor 488-conjugated goat anti-rat IgG. Nuclei were stained with 1 µg/mL Hoechst-33342 (Sigma-Aldrich).
4.5. Quantitative PCR and Western Blot
For quantitative PCR, cultured GICs were harvested in RLT lysis Buffer (Qiagen, Milano, Italy) and RNA was extracted following manufacturer instructions. cDNA was synthetized from 500 ng of RNA using the iScript Reverse Transcription Supermix (Bio-Rad) following the manufacturer’s instructions. One hundredth of the cDNA solution was used for q-PCR reaction using Luna Universal qPCR Master mix (AbCam). Cdh4 expression was normalized to the Rpl41 housekeeping gene. The sequences of the primers are available on request.
For Western blot, cultured GICs were harvested in lysis buffer containing 50 mM HEPES (pH 7.5), 5 mM EDTA, 150 mM NaCl, 1% Triton X-100 detergent, and protease inhibitors (Complete, Roche Applied Science) and separated using Mini-Protean TGX gels (Bio-Rad, Segrate, Milano, Italy). Protein expression was normalized to α-tubulin. The following primary antibodies were used: Rat anti-Cdh4 (MRCD5, Hybridoma Bank, Iowa City, Iowa, USA), mouse monoclonal anti-α-tubulin (Sigma-Aldrich), rabbit anti-ERK1/2 (Cell Signaling, Danvers, Massachusetts, USA), rabbit anti-phosphoERK1/2 (Cell Signaling), rabbit anti-AKT (Cell Signaling), rabbit anti-phospho-AKT (Cell Signaling), rabbit anti p27 (Cell Signaling). Anti-mouse HRP-conjugated (Sigma-Aldrich), anti-rabbit HRP-conjugated (Sigma-Aldrich) and anti-rat HRP-conjugated (GE Healthcare, Little Chalfont, UK) secondary antibodies were used. Proteins were then revealed with the ECL Star substrate (Euroclone, Milano, Italy) and visualized with the MINI HD9 chemiluminescence system (Uvitec, Cambridge, UK). Quantification of the protein expression level was performed with the software ImageJ [69
Data are reported as mean and standard errors of the mean.
4.6. Data Analysis
The software FIJI [70
] was used for the analysis of the subcellular localization of β-catenin. Nuclear areas where identified by thresholding Hoechst signal. Perinuclear areas where defined as the 3 μm band surrounding the nuclear areas. Cytoplasmic/membrane-bound localization was inferred by subtracting nuclear and perinuclear from total cell area.
For the analysis of cell polarity, we used FIJI to obtain for each micrograph, the coordinates of cell nuclei, Golgi apparatuses and the rim of the scratch (details of the procedure are available on request). We then used a self-produced script in R [71
], available on request, to measure, for each cell, the angle between the line joining the center of Golgi apparatus with the center of the nearest nucleus and the line perpendicular to the rim of the scratch. Cells with an angle between −60° and +60° were considered as polarized towards the scratch. Cells whose Golgi apparatus was not visible were not included in the analysis. Statistical analysis was performed by a Chi-squared test with two classes: Polarized cells (in null hypothesis expected to be one third of the total) and not-polarized cells (in null hypothesis expected to be two-third of the total).
For the in vivo analysis of the migratory ability of GICs transplanted cells, we used an Imager.M2 fluorescence microscope (version, ZEISS, Milano, Italy) controlled by the Slide Explorer plugin of μManager software (version v1.4, Ron Vale San Francisco California, USA) [72
] to take 10× micrographs of brain sections around the injection sites. We then extracted, for each section, the coordinates of the needle track and of cell nuclei. With a self-produced script in R [71
], available on request, we measured the distance of each GIC transplanted cell, detectable thanks to fluorescence reporter expression, from the rim of the injection site. Normalization between different brain sections were obtained by creating for each Section 10 bins for cell distance. We then measured the proportion of Cdh4 miRNA injected cells respect to control injected cells per each distance bin. To measure whether the imbalances in the cell frequencies depended on the cell distance from the injection site we used Pearson’s correlation test.