Primary cilia are sensory non-motile microtubule-based organelles (Lee and Gleeson, 2010) protruding from the surface of GCPs in the EGL at early post-natal stages (Del Cerro and Snider, 1972), whose requirement for Shh-induced expansion and cerebellar development has been proved using mutants of genes involved in the ciliary formation and maintenance (Chizhikov et al., 2007; Spassky et al., 2008). Among them, the genetic ablation of primary cilia by removing Kif3a (which encodes the microtubule plus end-directed kinesin motor 3A protein), blocked MB formation driven by a constitutively active Smoothened protein (Han et al., 2009). Therefore, Kif3a down-regulation blocks MB Shh-type formation in a primary cilia-dependent manner; moreover, its activity is not required for GCPs differentiation (Chizhikov et al., 2007). In our model, we observe that Kif3a is up-regulated in Ptch1 heterozygous mice, irrespective of the presence or absence of Tis21, which is therefore not involved in the Kif3a-dependent phenotype (Figure 2). This is consistent with the finding that Kif3a is required for the proliferation of the GCPs (Chizhikov et al., 2007) and with our observation that Tis21 in cerebellum regulates the migration of the GCPs but not their proliferation, while the opposite occurs for Ptch1.

Nevertheless, in our model, several genes encoding for the coiled-coil domain containing proteins are deregulated in Set A, and thus are dependent on Tis21, i.e., Ccdc157 and Ccdc171 (Table 1). One-fourth of the deregulated genes in Set A corresponds to coiled-coils proteins (data not shown), whose highly versatile protein folding motif is related to different biological processes, from subcellular infrastructure maintenance to trafficking control (Burkhard et al., 2001; Rose et al., 2005) and cilia-related (McClintock et al., 2008; Munro, 2011). A coiled-coil containing protein is also Rab11 family-interacting protein 4 encoded by Rab11fip4 (Muto et al., 2006, 2007), whose role in our model will be discussed more in detail in other paragraphs together with the functional product of Rab11fip2, and their wide implication in Shh signaling at primary cilium as a protein involved in microtubule-based vesicle trafficking. Another protein, Nesprin-2 encoded by Syne2, is known to mediate centrosome migration and is essential for early ciliogenesis and formation of the primary cilia by the interaction with the coiled-coil domain of Meckelin protein (Dawe et al., 2009). Notably, Ccdc157, Ccdc171, Rab11fip2, and Rab11fip4 are significantly down-regulated in Set A, while Syne2 is up-regulated. Also a novel repressor of hedgehog signaling, whose gene Rgs5 is up-regulated in set A, has been proven to be present with Smo in primary cilia (Mahoney et al., 2013). This would suggest that Tis21-dependent tumorigenesis in a (proliferation-independent) way involves ciliogenesis. This latter may be also enhanced by Syne2 after Tis21 ablation.

Evidences of direct involvement of Shh signaling on the increase of Ca²⁺ levels (Ca²⁺ spikes) have been shown at the primary cilium of chicken embryonic spinal neurons. In this system has been observed that Shh (a recombinant N-terminal molecule) may recruit second messengers (i.e., calcium—Ca²⁺—and inositol triphosphate) by a non-canonical pathway, through the activation of the Smoothened protein, which translocates to the cilium and becomes activated by phosphorylation at its C-terminal from a G-protein-coupled receptor kinase 2 (Riobo et al., 2006; Belgacem and Borodinsky, 2011; Brennan et al., 2012). Belgacem and Borodinsky (2011) proposed a model in which the primary cilium acts as a subcellular compartment for Shh signaling allowing the spatiotemporal integration of the second messengers through a Smoothened-dependent recruitment of G_i proteins and Phospholipase C that in turn increases inositol triphosphate levels. The opening of Inositol triphosphate receptors-operated stores and the following activation of Transient receptor potential cation channel 1 (Trpc1) results in an increased Ca²⁺ spike activity. This model could fit with our data (Figure 3) in which Plc-gamma2, Ip3r3, Trpc1, Trpc2, and Trpc3 are up-regulated in Ptch1 heterozygous mice (but independently from Tis21). Notably, this is the first time that such regulation is observed directly in a mouse model of Shh-type MB, as it had been previously only suggested (Briscoe and Therond, 2013). Furthermore, the authors suggested that the Smoothened-dependent Ca²⁺ spike activity is necessary for Shh-induced differentiation of spinal postmitotic neuron. Moreover, the role of second messenger signaling in the regulation of cerebellar granule cell migration has been studied in different mouse models (Komuro et al., 2015), which highlighted the direct evidence of the role of Ca²⁺ signaling in granule cell turning and modulation of their migration rate. The revision of these studies, performed by Komuro et al. (2015), suggested the role of Ca²⁺ as potential therapeutic target for some deficits in granule cell migration since its downstream effectors control the assembly and disassembly of cytoskeletal elements.

In the last years, the discovery of the role of the primary cilium in Shh signaling captured the attention of the scientific community, leading to test a large number of molecules that modulate SMO cilial translocation acting on different therapeutic potential targets in different types of cancer among which MB (Amakye et al., 2013). Loss of cilia in cancer has been suggested to be responsible for an insensitivity of cancer cells to environmental repressive signals, based in part on derangement of cell cycle checkpoints governed by cilia and centrosomes (Plotnikova et al., 2008). The importance of the role of cilia in Shh-driven medulloblastoma allografts derived from Ptch1^+/−P53^−/− mice has been shown using a Shh antagonist, i.e., arsenic trioxide (a therapeutic agent for acute promyelocytic leukemia), which inhibits the growth of tumor through the prevention of Shh ciliary accumulation and the reduction of the stability of the Gli2 transcriptional effector (Kim et al., 2010).

Other genes deregulated in Set A are involved in endocytic trafficking clathrin-dependent (Figure 5), and a certain number is related to cytoskeletal remodeling and primary cilium that could be very interesting for their implications for target therapy.

The clathrin-dependent endocytic mechanism is a receptor-mediated endocytosis type, which involves clathrin-coated vesicles, early endosomes, microtubule-based vesicle trafficking, lysosomes and recycling transport vesicles in its pathway (Le Roy and Wrana, 2005). Evidences of a deregulated clathrin-mediated endocytosis pattern have been reported for MB group 4 during early human neurogenesis (Hooper et al., 2014), and detected in our Set B (data not shown) as well as in Set A, where a large number of deregulated genes belongs to this signaling: Rab11fip4 (Horgan and McCaffrey, 2009), Ehbp1 (Guilherme et al., 2004), Rab11fip2 (Fan et al., 2004), Zfyve20 (de Renzis et al., 2002; Naslavsky et al., 2004), Cxcl12 (Fan et al., 2004; Teicher and Fricker, 2010; Zhu et al., 2012), Sgsm2 (Yang et al., 2007), Ckap5 (Foraker et al., 2012), Vps35 (Foraker et al., 2012), and Rab18 (Foraker et al., 2012).

Rab11fip4 is known to interact with Rab11 (a Rab GTPase involved in the regulation of intracellular membrane trafficking), which in turn interacts with Rab-coupling protein, a protein involved in the formation of lamellipodia, which facilitates invasive cancer cell migration (Kelly et al., 2012). Ehbp1 gene product is involved in actin reorganization coupled to endocytosis clathrin-mediated (Guilherme et al., 2004). Zfyve20 is known to regulate structure, sorting, endocytic/recycling pathway of the early endosomal compartment by the interaction with Rab4 and Rab5 (de Renzis et al., 2002; Naslavsky et al., 2004). Vps35, among other genes involved in intracellular membrane trafficking, is deregulated in Set A. It encodes for the vacuolar protein sorting-associated protein 35, a component of the trimeric cargo-selective retromer complex, required for the efficient endosome-to-Golgi recycling of many membrane proteins among which Wntless. It is also known to regulate endosomal tubule dynamics (Harbour et al., 2010; Berwick and Harvey, 2014). The oncogene Rab18 is required for normal ER structure and the regulation of endocytic traffic (Lütcke et al., 1994; Gerondopoulos et al., 2014). This Rab GTPasi is also known as a MB antigen (Behrends et al., 2003). Interestingly, Rab18 has been localized at lipid droplets (LDs) level (Martin et al., 2005), and its overexpression causes close apposition of LDs to membrane cisternae connected to the rough ER (LD-associated membrane or LAM) (Ozeki et al., 2005). Indeed, de novo lipid synthesis has been found in certain highly proliferative and aggressive tumors such as MB Shh-type, suggesting that the Shh-E2F1-FASN axis regulating de novo lipid synthesis in cancers could be used as therapeutic target in hedgehog-dependent tumors (Bhatia et al., 2011). Remarkably, Rab11fip2 encodes for a Rab11 family-interacting protein 2, which directly interacts with the actin-based myosin Vb motor protein regulating Cxcr2 recycling and receptor-mediated chemotaxis; hence, Rab11fip2 links Rab11 to molecular motor proteins and cell migration (Jones et al., 2006; Horgan and McCaffrey, 2009). Moreover, the passage of the internalized Cxcr2 through the Rab11-recycling system appears to have a key role in the physiological response to a chemokine that follows to the formation of clathrin-coated vesicles; in fact, Rab11 family-interacting protein 2 can form a complex with AP-2 that is a major clathrin adaptor complex in cells (Cullis et al., 2002; Fan et al., 2004; Le Roy and Wrana, 2005). Furthermore, the Rab11 involvement into the regulation of vesicular trafficking during primary ciliogenesis has been already put in light (Knödler et al., 2010; Hsiao et al., 2012). A consequence of the above findings, indicating an interaction between Cxcr2 and the clathrin pathway, is that, since Cxcl3 binds the Cxcr2 receptor (Zlotnik et al., 2006), we can infer that this chemokine receptor-mediated chemotaxis mechanism is clathrin-dependent and linked through Rab11fip2 to the primary cilium, in which the Shh signaling takes part. Other evidence of an involvement of Rab11fips and Shh signaling derives from Rab11fip4 (see the retina development section). This protein seems to be involved in the regulation of the membrane trafficking system through interaction with other small GTPases, among which could be Ras-related protein Rab-23, and in the negative regulation of Shh signaling at the primary cilium (Muto et al., 2007; Hsiao et al., 2012).

All this points to an important link between Shh signaling, operating through the primary cilium, and GCPs impaired cell migration, through a clathrin-Cxcl3-Cxcr2-mediated chemotaxis and microtubule-based endocytic vesicle recycling trafficking. Furthermore, the findings of a study about the role of endosomes around mother centriole appendages, and their Rab11-dependent recycling activity that requires centrosome-associated endosome proteins (Hehnly et al., 2012), seem to be in line with our data (see drug target section). Interestingly, they showed that (i) the appendages of the mother centriole and recycling endosomes are in intimate contact, as first evidence for a novel centrosome-anchored molecular pathway and regulation of endosome recycling; (ii) there is a structural association between the endosome and the centrosome with new and unexpected implications for recycling endosome functions, such us that one related to cilia formation; (iii) it is also possible that Rab11, and other endosome-associated molecules bound to the centrosome, may play dual roles in endosome and centrosome function (Hehnly et al., 2012).

In mice, retinal development occurs between E11.5 and P8, as uncommitted neuroblasts leave the cell cycle and commit to retinal cell fates (Mu et al., 2001). Thanks to mice models, it is known that aberrant proliferation during the development of the neural tube, of cerebellum and retina, leads to embryonal and early postnatal tumors (Dyer, 2004). The potent mitogen Shh positively controls the proliferation of their neuronal precursor cells (Martí and Bovolenta, 2002). In particular, Shh signaling plays a pivotal role in regulating the proliferation of retinal progenitor cells (RPCs) and the differentiation of retinal ganglion cells (RGCs) during vertebrate retinal development, acting in a cell-specific manner; namely, in mouse Shh is required as positive regulator of RPCs proliferation and as negative regulator of RGCs production, by inhibiting cell-cycle exit (Wang et al., 2005; Wallace, 2008).

A further molecular target which may be responsible for the regulation of retinal cell proliferation and hence for cancer cell proliferation was suggested to be Rb; in fact, the levels of Rb protein appear critical for the development of retinal tumors (Sicinski et al., 1995). The rationale for this is that Rb, when active, inhibits the cycle at the G1 checkpoint, prior to cell differentiation, whereas its inactivation, exerted by phosphorylation from cyclinD1/CDK4, is known to start the cell cycle progression. Thus, high levels of Rb may be more difficult to inactivate and vice-versa, thus critically linking the Rb-dependent developmental regulation of proliferation during neurogenesis to cancer cell proliferation for certain types of tissues (Sicinski et al., 1995). Indeed, in our model we observe that in GCPs heterozygous for Ptch1 (set B: Ptch1^+/−/Tis21^+/+ vs. Ptch1^+/+/Tis21^+/+) cyclin D1 expression does not change while Rb does increase. This suggests that in Shh-driven neoplastic GCPs the increase of Rb protein is compensatory.

In the case of MB, a parallel between developmental neurobiology and oncology was suggested for the proliferating progenitor cells of the retina and cerebellar granule neurons, where the failure to exit the cell cycle leads to aberrant cell proliferation during development in mice (Romer and Curran, 2004). Notably, MB-types Cluster D and E (also referred to as groups 4 and 3, respectively) have been found to be marked by a deregulated expression of retinal photoreceptor genes suggesting a distinct origin (i.e., non-cerebellar) from stem cells during the embryonic development, with respect to MB Shh-type in human (Kool et al., 2008).

In this context, concerning the contribution of Tis21 deletion to the MB development, in Set A we noticed a great number of deregulated genes that have been previously described as involved in retinal development. This comparison could be useful to suggest some common mechanisms related to the progenitor cells cell-cycle exit failure. A consistent subset of Set A genes has been previously described as being involved in cellular expression patterns of mouse early retinal development; these gene were previously recognized by analyzing the outer retinal neuroblastic layer, which in early developmental stages consists almost entirely of mitotic progenitor cells: Vdac1, Taf7, Emd, 9130401M01Rik, Taok2, Hist3h2ba, Tomm22, Vps35, H19, Slc6a6, Pafah1b1, Akap2, Raly, Rps12, Nlk, Pa2g4 and Srpk2 (Blackshaw et al., 2004). These genes are all up-regulated in Set A except for H19. Furthermore, a number of other genes that were identified in other studies as being involved in retinal development, are down-regulated in set A: Col4a6 (Bai et al., 2009), Rab11fip4 (Muto et al., 2007), Bsn (Dick et al., 2003), Efna4 (Marcus et al., 1996; Poliakov et al., 2004; Triplett and Feldheim, 2012), Egflam (whose product is also known as Pikachurin) (Omori et al., 2012); conversely, other retinal genes up-regulated in set A are: Dgkq (Pilz et al., 1995), Cdc27 (Leung et al., 2008), Syne2 (Yu et al., 2011), Slc6a6 (Vinnakota et al., 1997; Warskulat et al., 2007), Ripk3 (Trichonas et al., 2010).

The genes listed above belong to different functional clusters, and some of them will be discussed more in detail in their paragraph of pertinence. Interestingly, the mouse Rab11fip4, whose product regulates the Rab GTPases and is predominantly expressed in the developing neural tissues, among which retina, acts as regulator of RPCs cell-cycle exit and their subsequent differentiation (Muto et al., 2007). Rab11fip4 seems to be involved in the regulation of membrane trafficking system through interaction with other small GTPases and in the negative regulation of Shh signaling (Muto et al., 2007). Moreover, the Syne2 gene product is known to mediate nuclear migration during mammalian retinal development connecting the nucleus with dynein/dynactin and kinesin proteins (Yu et al., 2011).

This comparison is in line with the evidence that progenitors from the developing cerebral cortex, cerebellum and retina share a common expression program, suggesting a common evolutionary origin of the different progenitors cells (Livesey et al., 2004), and implying a possible common differentiation program. Our data suggest that this developmental process, possibly at the steps of migration and differentiation, is regulated by Tis21.

The migration of the GCPs is known to be induced by responses to local environmental cues. However, the alterations of migratory behavior may also depend on intrinsic programs (Komuro et al., 2013). In our study, we observed in Set A the differential expression of a great number of genes belonging to migration, cell adhesion and differentiation mechanisms. Among the genes implicated in migration seven are down-regulated after ablation of Tis21 in Shh-activated background, namely, Cxcl3, Jmy, Efna4, Timp1, Frk, Prrx1, and Rab11fip2, and three are up-regulated: Cxcl12, Pdgfd, and Pafah1b.

Cxcl3 is a chemokine known to have chemotactic activity for neutrophils (Wuyts et al., 1999). Chemotaxis but also angiogenesis seems to be activated by the interaction with its receptor (Addison et al., 2000; Zlotnik et al., 2006). In our previously published data, Cxcl3 has been identified as a gene whose transcription is regulated positively by Tis21 (which directly associates to the Cxcl3 promoter), suggesting that the evident role played in vivo by Tis21 as inducer of the migration of the GCPs out of the EGL might occur through its functional product; this implies Cxcl3 as a new pharmacological target for medulloblastoma therapy, also considering that MB lesions were reduced using this chemokine as treatment (Farioli-Vecchioli et al., 2012a,b). Recently, the anti-cancer role of this chemokine has also been recognized for non-small lung cancer, mediated by the activity of interleukin-27 (Airoldi et al., 2015). Jmy, a junction mediating and regulatory protein, is known to induce cell motility by promoting actin assembly and regulating cadherins in the cytoplasm; in DNA damage conditions it undergoes nuclear accumulation, where acts as p53-cofactor promoting apoptosis (Shikama et al., 1999; Coutts et al., 2007; Zuchero et al., 2009). Together, these findings suggest that the ability of Jmy to regulate actin and cadherin, by coordinating cell motility with the p53 response, could underlie a common pathway (Coutts et al., 2009). Furthermore, actin assembly has been found to regulate the nuclear import of Jmy in response to DNA damage (Zuchero et al., 2012). Thus, an abnormal activity and/or localization of Jmy could contribute to tumor invasion, thus making this gene as a potential tumor therapeutic target (Wang, 2010). In theory, the down-regulation of Jmy may contribute to enhance the rate of MB in our mouse model, possibly by controlling the migration of the GCPs as shown for Cxcl3. Regarding Efna4, this gene encodes for the EphrinA4 protein that has been linked with the migration of neuronal cells during development (Poliakov et al., 2004). A recent study showed that EphrinA4 and the Ephrin type-B receptor 4 are almost exclusively expressed in Shh MBs, while other Ephrins are expressed in non–Shh MBs; furthermore, a strategy of overexpression and silencing, applied specifically to EphrinB1, showed its ability to control the migration of MB cell lines (McKinney et al., 2015). Further evidence of the migration-promoting activity of EphrinA4 has been observed in human glioma cells, where it acts by accelerating a canonical FGFR signaling pathway (Fukai et al., 2008). Timp1 protein product is a metalloproteinase inhibitor known to be involved in cell adhesion and migration of human neural stem cells (hNSCs). In fact, through its cell surface receptor CD63, TIMP-1 activates β1 integrin, FAK (focal adhesion kinase) and the PI3K (phosphoinositide 3-kinase) signal transduction pathway, resulting in the migration of hNSCs (Lee et al., 2014). Due to its chemoattractant properties, TIMP-1 has been identified in the same study as a therapeutic target for human glioma. The Frk gene product is a Src kinase known as Tyrosine-protein kinase FRK, which controls the migration and invasion of human glioma cells by regulating JNK/c-Jun signaling (Zhou et al., 2012). Moreover, the Tyrosine-protein kinase FRK acts as a tumor suppressor in breast cancer by regulating the stability of PTEN, as the loss of Rak (i.e., Frk) induced tumorigenicity in immortalized normal mammary epithelial cells (Yim et al., 2009). In mouse brain, Pten is known to be expressed starting at approximately postnatal day 0 (Lachyankar et al., 2000) and has also been correlated with the regulation of neuronal precursor cell migration (Li et al., 2002). In Set B Pten is up-regulated. The paired-related homeobox transcription factor 1, which is encoded by Prrx1, is an epithelial-mesenchymal transition (EMT) inducer in embryos, where this process is required for the formation of tissues for which cells originate far from their final destination (Ocaña et al., 2012). EMT is modified and exploited by cancer cells for metastatic dissemination and also in cancer cells. In particular, the loss of Prrx1 has been related to the ability of cancer cells to acquire tumor-initiating abilities concomitantly with stem cells properties (Ocaña et al., 2012). Furthermore, paired-related homeobox transcription factor 1 has been found to promote tenascin-C–dependent fibroblast migration when its expression was induced by Focal adhesion kinase 1 (McKean et al., 2003). Fak is up-regulated in both Set B and Set D. Very interesting is the down-regulation of Rab11fip2, whose role in the endocytic recycling pathway has been linked to cell migration (Jones et al., 2006) as previously discussed (Section Receptor-Mediated Endocytosis Mechanisms, Microtubule-Based Vesicle Recycling and Intracellular Membrane Trafficking).

Among the genes up-regulated in Set A related to migration there is Cxcl12, which encodes for a deeply studied chemokine involved in different mechanisms in cancer development and metastatic invasion (Duda et al., 2011; Hattermann and Mentlein, 2013), but also described as involved in the migration of neuronal cells through both its receptor, C-X-C chemokine receptor type 4 and Atypical chemokine receptor 3 (Tiveron and Cremer, 2008; Memi et al., 2013; Yang et al., 2013). Cxcl12 appears to exert an action opposite to Cxcl3, as it promotes the localization of the GCPs to the EGL by chemoattraction, being released from meninges (Klein et al., 2001; Zhu et al., 2002). Thus, the upregulation of Cxcl12, consequent to the ablation of Tis21, synergizes with the downregulation of Cxcl3 in preventing the migration of the GCPs from the EGL. Notably, the chemoattraction of cerebellar granule cells by Cxcl12, whose receptor C-X-C chemokine receptor type 4 is coupled to a G protein, is selectively inhibited by the soluble EphB receptor; this inhibition is blocked by a truncated PDZ-RGS3 lacking the RGS domain, which activates the G-proteins. Therefore, this points to the existence of a pathway connecting B ephrins and Cxcl12 to the regulation of G protein–coupled chemoattraction, and leads to a model for regulation of migration in cerebellar development (Lu et al., 2001). In this regard, in our model (Set A) we have detected not only a down-regulation of Efna4, which is a cell surface GPI-bound ligand for Eph receptors, but also the up-regulation of a regulator of heterotrimeric G protein signaling, i.e., Rgs5. Therefore, in our model the decrease of Cxcl3 and Efna4 and the increase of Cxcl12 after ablation of Tis21 in Ptch1 heterozygous mice would synergize in impairing the migration of GCPs from the EGL. As mentioned above, the increase of Cxcl12 in the Tis21-null EGL GCPs would prevent their migration by chemoattraction (Zhu et al., 2002). Moreover, the PDGFR pathway members have been recently studied in correlations with MB Shh-driven and MB cell migration, and an upregulation of PDGFRA (receptor for PDGF-A), PDGF-D (ligand for PDGFRB) and Cxcl12 has been observed in Shh-type MB (vs. non Shh MBs). In particular, the activation of the PDGFR pathway has been shown to activate the Cxcl12 receptor (i.e., CXCR4, via inhibition of GRK6) and thus cell migration (Yuan et al., 2013). A relevant caveat of that study is that the direction of cell migration, relative to the EGL, was not assessed, having the analysis been conducted only in vitro (Yuan et al., 2013). In our model, we noticed an up-regulation of Pdgfd and Cxcl12 genes in Set A, while both Pdgfra and Pdgfrb and also Cxcl12 are up-regulated in Set B and D (see Figure 3). This indicates that the heterozygosity of Ptch1 is in itself a condition inducing the expression of both the Cxcr4 and Pdgfr pathways, and that the additional ablation of Tis21 further enhances their activation. Finally, Pafah1b1, up-regulated in Set A, encodes for Lis1, a microtubule regulator that is required for correct neuronal migration during development (Hippenmeyer et al., 2010; Escamez et al., 2012).

The transcriptional reprogramming of the epigenetic patterns is among the causes of tumorigenesis. Nowadays, the epigenetic regulation of transcription and genome organization in MB Shh-type pathogenesis is extensively studied (Batora et al., 2014; Hovestadt et al., 2014; Shi et al., 2014). We have previously provided a first functional genomic analysis of epigenetically regulated genes or interacting proteins for our mouse model Tis21 KO, identified in background either Ptch1 heterozygous or wild-type (Farioli-Vecchioli et al., 2012b). We also have recently shown that the Tis21 protein binds to histone deacetylases (HDAC1/4/9) in GCPs, where they are required for the Tis21-dependent inhibition of cyclin D1 expression (Micheli et al., 2016). In the present analysis, a great number of genes encoding proteins involved in epigenetic regulation appear to be deregulated in Set A (three down-regulated and 12 up-regulated). They mostly belong to the category of the Histone modification rather than to the Chromatin remodeling one (Table 2), which categories are described in (Arrowsmith et al., 2012; Plass et al., 2013). Our data are in line with those previously published highlighting a great number of deregulated or mutated histone modifiers involved in medulloblastoma (Northcott et al., 2012) and their importance as oncogenes or as tumor suppressors (Cohen et al., 2011). Furthermore, we highlight also the deregulation of five Histone modifier Regulators, up-regulated in Set A, namely: Emd (Berk et al., 2013), Anp32a (Seo et al., 2001; Fan et al., 2006; Kular et al., 2009), Taf7 (Gegonne et al., 2001, 2013; Kloet et al., 2012), Pag2g4 (Zhang et al., 2003), Ipo7 (Jäkel et al., 1999; Mühlhäusser et al., 2001). A detailed description of Set A genes involved in epigenetic modulation is presented in Supplementary Data, at section “Epigenetic modulation.”

The increasing recognition of the heterogeneity of molecular basis underlying cancer allows the identification and development of molecularly targeted agents and their transfer to the patients (Rask-Andersen et al., 2011; Saletta et al., 2014). Here we provide a drug target identification through the genomic analysis of deregulated MB Shh-type Tis21 knockout-dependent genes in Set A and, where possible, the identification of potentially druggable targets (Figure 3, Tables 3–5), performed using the methods described in the appropriate section. This analysis has given as a result 20 primary, 31 secondary, 53 druggable targets and 18 gene targets among Set A elements (Tables 3–6). Their distribution within the functional classes is showed in Figure 5. and the putative drug targets identified in SetA, here discussed, are shown in Figure 3.

Among the Set A genes showing a change of expression influenced exclusively by the ablation of Tis21, there is Pdgfd that has been discussed in developmental (see Supplemtary Data) and migration processes as up-regulated gene. Since the overactivity of PDGF signaling can drive tumorigenesis (Pietras et al., 2003), and since PDGF-D in particular has been found to be a potent transforming and angiogenic growth factor (Li et al., 2003) highly expressed in Shh-type medulloblastoma (Yuan et al., 2013), we propose targeting PDGF-D as therapeutic strategy for medulloblastoma Shh-type, as already studied in other tumors (Heldin, 2013). Another very interesting drug target could be the phosphatidylinositol 3-kinase-related kinase Smg1, which here has been discussed in developmental and nonsense-mediated decay processes (Supplementary Material) as down-regulated in Set A (up-regulated in Set D). This protein seems to act antagonistically with mTOR signaling (González-Estévez et al., 2012; Du et al., 2014), and this is in functional synergy with the up-regulation in Set A of Deptor, which negatively regulates mTOR signaling (Beauchamp and Platanias, 2013). Also Deptor is one of the druggable target identified in our study. Taken together, these findings seem to support the importance of mTOR pathway and its upstream PDGF signaling in the pathogenesis of medulloblastoma (Mohan et al., 2012).

Other two druggable targets are regulators of cell cycle and developmental processes: Sik2 and Lats2. The functional product of Sik2 is localized at the centrosome where its absence leads to a delay of G1/S transition (Ahmed et al., 2010), while Lats2 encodes for a centrosomal protein (a serin threonin kinase) whose loss leads to centrosome fragmentation (Yabuta et al., 2007), also involved in the regulation of cell cycle G1/S checkpoint, for its relevance to MB tumorigenesis due to its influence in the reduction of expression of cyclin-D1 and N-CoR (Park et al., 2012; Lit et al., 2013). The deregulation of centrosome and cilia biogenesis have been already described in different human diseases, in particular, in cancer where a derangement of cell cycle checkpoints is governed by cilia and centrosomes (Plotnikova et al., 2008; Nigg and Raff, 2009; Bettencourt-Dias et al., 2011). In addition to that, Sik2 has been characterized as negative regulator of Hippo signaling in Drosophila (Wehr et al., 2013). In our data, other two regulators of Hippo signaling appear to be down-regulated after ablation of Tis21, Lats2, and Fat4 that we discuss for their role in the developmental process (Supplementary Data); both act also as tumor suppressors. This evidence supports the involvement of Hippo signaling (Roussel and Hatten, 2011) and centrosome assembly in the pathogenesis of MB.

Another putative drug target belonging to developmental processes, Rgs5, encodes for an endogenous repressor of Shh signaling and has been proposed in a recent study as potential therapeutic target in Hh-mediated diseases. In fact, it was shown that (i) Rgs5 inhibits the Shh-mediated signaling by activating the GTP-bound Gαi downstream of Smo and (ii) a physical complex between Rgs5 with Smo is present in primary cilia (Mahoney et al., 2013).

The apoptosis is not the only form of cellular death in which the deregulated genes in Set A are involved. In fact, the Ripk3 functional product, a receptor-interacting protein kinase 3, has been reported to contribute to both apoptotic and necroptotic cell death, depending on target availability (Cook et al., 2014; Vanden Berghe et al., 2014). Since many anti-cancer drugs are inducers of apoptosis, the induction of RIP3-dependent necrosis is an attractive strategy to circumvent apoptosis resistance of cancer cells (Moriwaki and Chan, 2013) that is currently under investigation (Moriwaki et al., 2015). As Ripk3 expression is induced after knockout of Tis21 in Shh-activated background, we may hypothesize that Ripk3 plays in our model a tumor suppressor role.

The therapeutic advantage of targeting the ubiquitin-proteasome system has already being successfully investigated with proteasomal inhibitors in Shh-type MB with in vivo preclinical studies (Ohshima-Hosoyama et al., 2011) and in a preliminary study with personalized targeted therapy for pediatric brain tumors among which MB (Wolff et al., 2012). However, the targeting of specific enzymes regulating the ubiquitylation process, e.g., SKP2, a SCF ubiquitin ligase, up-regulated in Set D (Figure 3), has been recently proposed as a more specific approach than the previous one (Hede et al., 2014). Two genes belonging to the ubiquitin-dependent degradation processes are up-regulated in Set A and have been identified in this study as putative drug target: Ups36 and Cdc27. Usp36, encoding for a deubiquitinating enzyme (Quesada et al., 2004), has been detected as overexpresed in human ovarian cancer compared to normal ovaries (Li et al., 2008). An Usp36 homolog, recently detected in Drosophila stem cells, has been linked to the deubiquitylation of histone H2B and to the silencing of key differentiation genes, including genes target of Notch (Buszczak et al., 2009). Whereas the anaphase-promoting complex component Cdc27 is an E3 ubiquitin ligase able to control cell cycle progression at the G1 to S transition, and also to induce the non-proteolytic disassembly of the spindle checkpoint during mitosis (Jin et al., 2008; Manchado et al., 2010; Pawar et al., 2010). For these reasons, Cdc27 functional product has been identified as a molecular target of the curcumin-induced cell cycle arrest and apoptosis in Shh-type MB (Lee and Langhans, 2012). The therapeutic properties of this natural substance were already shown in Shh-driven MB models, highlighting its ability to inhibit the Shh signaling, to reduce the level of β-catenin and to inhibit HDAC4 (Elamin et al., 2010; Lee et al., 2011). Thus, the up-regulation of Cdc27 after ablation of Tis21 may be a tumorigenic factor.

Of interest as drug target among the genes belonging to the migration processes is Timp1. Its signal trasduction pathway results in the migration of hNSCs, showing chemoattractant properties and being identified in the same study as a therapeutic target for human glioma (Lee et al., 2014). Here we suggest the same prospective use of Timp1 in MB Shh-type.

The Cdc42bpb produces a Rho GTPase activated serine/threonine kinase, which, by regulating the (non-muscle) cytoskeletal actomyosin, influences cell shape and promotes motility and migration (Tan et al., 2008, 2011), thus acting also as a key regulator of tumor cell invasion. Cdc42bpb is therefore an interesting drug target (Heikkila et al., 2011; Unbekandt et al., 2014). It's down-regulation in our Set A data is in line with the GCPs migration failure, consistently with the deregulation of others migration regulators discussed above.

Genes involved in the RNA processing are Smg1 and Upf3b, respectively down- and up-regulated in Set A. Interestingly, both genes encode for proteins involved in alternative splicing (AS) coupled with Nonsense-Mediated Decay (NMD) mechanism by which NMD controls transcript abundance by regulating AS; in fact, the SURF complex, which includes the SMG1–UPF1–eRF1–eRF3 proteins, forms a bridge between the ribosome and the downstream exon-junction complex (EJC) associated with UPF3b and UPF2 (Chamieh et al., 2008; McIlwain et al., 2010). Moreover, Smg1 has been showed to be involved in the predisposition to tumor formation and inflammation in Smg1 heterozygous mice; this mouse model presents elevated basal tissue and serum cytokine levels—indicating low-level inflammation—and can progress to chronic inflammation or enhanced cancer development. Therefore, this is a model of inflammation-enhanced cancer development (Roberts et al., 2013), also suggesting that Smg1 is a tumor suppressor (Du et al., 2014). To our knowledge, the possibility to target inflammation through Smg1 has never been applied to medulloblastoma until now. Furthermore, in our data Il1b, for example, is up-regulated in Set D and B (See Figure 2); also Cxcl12 is up-regulated in Set D, B, and also in Set A, as mentioned previously, while Cxcl3 is down-regulated in Set A. It is also worth noting the existence of an anticoagulant inhibitor of CXCL12/CXCR4 axis, the Tinzaparin, as well as other heparinoids that have been studied in brest cancer models for their heparinoid-mediated inhibition of chemotaxis activity (Harvey et al., 2007; Mellor et al., 2007). As Cxcl12 favors the localization of the GCPs to the EGL by chemoattraction, an inhibitor of Cxcl12 might achieve an opposite effect, similar to the promigratory effect exerted by Cxcl3.