A New Story of the Three Magi: Scaffolding Proteins and lncRNA Suppressors of Cancer

Simple Summary MAGI1, 2, and 3 belong to a subgroup of the MAGUK family of scaffolding proteins, and are comprised of 6 PDZ domains, 2 WW domains, and 1 GUK domain. MAGIs associate with cell surface receptors, junctional complexes, and interact selectively with a wide range of effectors, including the PTEN tumor suppressor, the β-catenin, and YAP1 proto-oncogenes. The regulation of the PI3K/AKT, the Wnt, and the Hippo signaling pathways, on the one hand, the downmodulation of MAGIs in various types of cancers, and its physiopathological significance, on the other, make these scaffolding proteins considered to be tumor suppressors. Interestingly, MAGI1 and MAGI2 genetic loci generate a series of long non-coding RNAs (lncRNAs) that act as promoters or suppressors of tumors in a tissue-dependent manner by sponging some sets of miRNAs or by regulating epigenetic processes. This review details current knowledge of paths followed by the three MAGIs to control carcinogenesis. Abstract Scaffolding molecules exert a critical role in orchestrating cellular response through the spatiotemporal assembly of effector proteins as signalosomes. By increasing the efficiency and selectivity of intracellular signaling, these molecules can exert (anti/pro)oncogenic activities. As an archetype of scaffolding proteins with tumor suppressor property, the present review focuses on MAGI1, 2, and 3 (membrane-associated guanylate kinase inverted), a subgroup of the MAGUK protein family, that mediate networks involving receptors, junctional complexes, signaling molecules, and the cytoskeleton. MAGI1, 2, and 3 are comprised of 6 PDZ domains, 2 WW domains, and 1 GUK domain. These 9 protein binding modules allow selective interactions with a wide range of effectors, including the PTEN tumor suppressor, the β-catenin and YAP1 proto-oncogenes, and the regulation of the PI3K/AKT, the Wnt, and the Hippo signaling pathways. The frequent downmodulation of MAGIs in various human malignancies makes these scaffolding molecules and their ligands putative therapeutic targets. Interestingly, MAGI1 and MAGI2 genetic loci generate a series of long non-coding RNAs that act as a tumor promoter or suppressor in a tissue-dependent manner, by selectively sponging some miRNAs or by regulating epigenetic processes. Here, we discuss the different paths followed by the three MAGIs to control carcinogenesis.


Introduction
The cellular response to external stimuli e.g., hormones, growth factor, cell-cell contact, stress requires the spatio-temporal integration of the several signaling pathways to produce an adapted biological response in terms of metabolism, proliferation, differentiation, cell-cell interaction, migration, or cell death. These processes are orchestrated at specific sites by scaffolding proteins that serve as the backbone for supramolecular signaling complexes termed signalosomes [1][2][3][4]. Scaffolding proteins allow the local and subtle tuning of effectors in the specific subcellular domain, the stimulation of distinct pathways simultaneously, or the coordination of positive and negative signals to impede

MAGI Structure
Among the scaffolding molecules with PDZ domains, the membrane-associated guanylate kinases (MAGUKs) are a superfamily of proteins characterized by the inclusion of one or three PDZ domains at the N-terminus, followed by one SH3 domain and then the guanylate kinase (GUK) domain. This GUK domain is catalytically inactive due to the absence of the P-loop that binds ATP. Nevertheless, it allows protein interaction, mainly with cytoskeleton proteins and effectors of signal transduction. Many MAGUK proteins contain other modules, e.g., homologous of CaMKII, WW, or L27 domains [11]. PSD95, DLG1, and ZO1 that serve as an acronym for "PDZ" are members of this family of proteins.
The three membrane-associated guanylate kinase inverted (MAGI) proteins constitute a subgroup of MAGUK proteins. They are characterized by an inverted arrangement: the GuK domain is at the NH2 terminus, the SH3 domain is replaced by two WW domains, and follows five PDZ domains -numbered 1 to 5- [29,30]. A more recent analysis of the MAGI structure allowed to identify an additional PDZ domain above the GUK domain that was designed PDZ0 ( Figure 1A) [31,32]. Overall, the primary sequence of MAGI3 is more similar to MAGI2 than to MAGI1 [8,33,34]. MAGI1 and MAGI3 are ubiquitous, whereas MAGI2 is mainly expressed in the brain and the renal glomeruli (podocytes), followed by skeletal muscle, pancreas, ovary, and heart ( Figure 1). These scaffolding molecules are mainly located at junctional complexes, including tight and adherens junctions [21,22,29,32,[35][36][37][38]. Several splice variants of MAGIs have been evidenced and proved to be expressed in a tissue-dependent manner ( Figure 1B) [23,29,32,39] Two MAGI2 variants termed (MAGI2α and MAGI2β) depleted of PDZ0 and part of the GUK domain as the result of alternative translational initiation sites have been also identified in the rat brain [40]. Such variants in human tissues have also been predicted from genomic analysis and expressed sequence tags.
The premature polyadenylation of MAGI1 and MAGI3 transcripts have been evidenced in human small intestinal neuroendocrine tumors and breast cancers, respectively [41,42]. The use of a polyA site in the 2nd intron of MAGI1 upstream the sequence encoding the GuK domain leads to the expression of an mRNA lacking 21 exons in the 3-end, whereas the premature polyadenylation of MAGI3 produces a truncated protein depleted of PDZ2-PDZ5 [41,42]. WW: WW domain (also termed rsp5-domain or WWP repeating motif). MAGI1. The isoforms MAGI1A, B, and C result from alternative splicing downstream the sequence encoding PDZ5, the related frameshifts generating unique C termini. The C terminus of MAGI1C contains potential nuclear targeting signals. Two internal splice variants have been further evidenced, these include an in-frame insertion of 28 amino-acid residues upstream PDZ3 domain (α domain). The 2nd splicing involves a region termed β, encoding much of PDZ4 [23,32]. Four different arrangements of the β region were identified, each encoding an intact PDZ4 which might exhibit distinct ligand binding properties. Splicing out the β domain results in the depletion of a functional PDZ4 domain, and in an isoform encompassing 5 PDZ domains. A MAGI1C variant with splice out of β region and part of the inter-domain connecting PDZ3 and PDZ4 was identified in the human cervix and designed MAGI1D [39]. The structure of the MAGI1 splice variants displayed in panel A is based on the initial nomenclature and sequences from Laura et al. (GenBank AF401655, AF401654, and AF401656) and van Doorslaer et al. [32,39]. Other MAGI1 predicted variants are not represented. MAGI2. The MAGI2 isoform 1 differs from isoform 2 by in-frame insertion of 14 amino acid residues in the inter-domain linker connecting PDZ2 and PDZ3 (alternative splice schematized by the disc and the dashed lines). Some of the predicted MAGI2 isoforms arising from alternative translation initiation sites are also displayed. MAGI3. The MAGI3 isoform 2 results from an alternate splice site at the 5' end of the last coding exon, generating a shorter and distinct C-terminus. MAGI3pPA originates from the use of a cryptic intronic poly(A) signal. (B) Tissular distribution of MAGI transcripts. Relative abundance of the transcripts encoding MAGI1 isoforms (left panel) [32], and MAGI1, 2, and 3 (right panel) in human tissues, as evaluated by RT-PCR and northern blot [31,[43][44][45][46]. Note that MAGI1 is also designated: atrophin-1-interacting protein 3 (AIP-3), BAI1-associated protein 1 (BAP-1), membrane-associated guanylate kinase inverted 1 (MAGI-1), trinucleotide repeat-containing gene 19 protein, WW domain-containing protein 3 (WWP3); MAGI2 is also named: atrophin-1-interacting protein 1 (AIP-1), atrophin-1-interacting protein A, synaptic scaffolding molecule (S-SCAM), membrane-associated guanylate kinase inverted 2 (MAGI-2); and MAGI3: membrane-associated guanylate kinase inverted 3 (MAGI-3).

MAGI Molecular Partners
Many ligands of MAGIs have been identified, including KRAS, NET1, PTEN, βcatenin, NMDA receptors, neuroligin, BAI-1, megalin, RA-GEF-1, activin type II receptor (Table 1). These ligands may compete with each other for interacting with the same MAGI binding module ( Figure 2). Some of these interactions are shared by the three MAGIs (e.g., PTEN, β-catenin), others seem to be restricted to a subset of these scaffolding molecules (e.g., the nerve growth factor receptor interacts with MAGI1, but not with MAGI2 or MAGI3; Figure 2, Table 1). Thus, it seems difficult to delineate a specific biological activity for each MAGI, although some selective tissue expression has allowed establishing some peculiar physiological role. In this context, the constitutive knockout of MAGI2 in mice leads to podocyte loss, diffuse glomerular extracapillary epithelial cell proliferation, and renal failure [47].
The 9 binding modules of MAGIs allow the spatiotemporal organization of signalosomes and the subcellular targeting/confinement of effector proteins (e.g., PTEN or βcatenin at adherens junctions), as well as gathering effectors of the same pathway (e.g., the Wnt receptor frizzled 4/Vang-like protein 2) or with antagonistic activities (e.g., ERBB4 tyrosine kinase receptor/tyrosine phosphatase R-PTP-zeta; Rho guanine nucleotide exchange factors NET1/Rho GTPase activating protein ARHGAP6) for the fine-tuning of downstream signaling pathways ( Figure 2). RA-GEF-1 Guanine nucleotide exchange factor (GEF); activates the Rap and Ras family of small GTPases by exchanging bound GDP for free GTP in a cAMP-dependent manner MAGI1 recruits RA-GEF-1 at tight junctions in epithelial cells. In endothelial cells, MAGI1 is required for Rap1 activation and adherens junction formation. Translocation of MAGI1 to cell-cell contacts is ascribed to its interaction with β-catenin. MAGI2 co-expressed with RA-GEF-1 enhances activation of Rap1 in podocytes, signaling essential for normal podocyte function [99][100][101][102]  SYNPO Synaptopodin Actin-associated protein, involved in actin-based cell shape and motility + ND ND 2nd WW Role in actin cytoskeleton dynamics? [24] Posttranslational Modifications

CASP3
Caspase-3; CASP-3 Cysteine-aspartic acid protease. Effector caspase is involved in the execution phase of apoptosis. Activated by proteolytic cleavage. CASP-3 shares several substrates with CASP-7 but exerts distinct functions during apoptosis.
MAGI1 is cleaved by CASP-3 at Asp761, between PDZ2 and PDZ3. The N-term fragment translocates in the cytoplasm and the C-term accumulates in the nucleus.
MAGI1 cleavages induce disruption of cell-cell contacts and apoptosis. MAGI2 cleavage site was not determined. Cysteine-aspartic acid protease. Effector caspase is involved in the execution phase of apoptosis. Activated by proteolytic cleavage. CASP-7 shares several substrates with CASP-3 but exerts distinct functions during apoptosis.
MAGI1 is cleaved by CASP-7 at Asp761, generating 97 kDa and 68 kDa fragments. The N-term fragment translocates in the cytoplasm and the C-term accumulates in the nucleus. MAGI1 cleavages induce disruption of cell-cell contacts and apoptosis. [103]

MAGIs in Cancer 2.3.1. Genetic Alteration, Silencing
The frequent genetic alteration and downregulation of MAGIs at transcripts and protein levels in many types of cancers, including breast, prostate, liver, stomach, colon, and cervix offer a valuable insight into the implication of these scaffolding proteins in the control of carcinogenesis ( Table 2).
The promoter methylation and the silencing of MAGI1 have been reported in anaplastic thyroid cancer and acute lymphocytic leukemia [121,122]. Methylation of multiple CpG islands of MAGI1 promoter is associated with a worse overall survival of patients with lymphocytic leukemia [122]. MAGI-1 expression proved also to be frequently downregulated at mRNA and protein levels in human T-cell leukemia cell lines [123]. Methylation of the MAGI2 gene was evidenced in cervical, gastric, endometrial, ovarian, and breast cancers [124][125][126][127][128]. In the cervix carcinoma HeLa cell line, this hypermethylation was attributed to the downregulation of TET1 -involved in DNA demethylation via the oxidation of 5-methylcytosine-by the HOTAIR lncRNA [129].
Some genomic rearrangements of MAGI2 have also been evidenced in melanoma cell lines and prostate cancers leading to in-frame deleted transcript with an unknown biological significance, and to MAGI2 invalidation, respectively [130,131].
A rearrangement between MAGI3 and AKT3 genes caused by chromosome 1 inversion, has been identified in triple-negative breast cancer (lacking estrogen and progesterone receptors and ERBB2 expression). The resulting chimeric protein combines MAGI3 truncated upstream PDZ1, fused with AKT3 [132]. Nevertheless, the overall prevalence of this constitutively active protein kinase seems to be below 1% of tumors [133,134].
MAGI1 mutation in early gastric cancer and adjacent mucosa is associated with an increased risk of developing metachronous gastric cancer following curative endoscopic submucosal dissection [135]. Overall, missense mutations of MAGI genes are identified in 16-20% of human cancers, non-sense mutation being in the range of 1-2% (Cosmic V94, URL http://cancer.sanger.ac.uk/cosmic/, released 28 May 2021, accessed on 22 July 2021, Table 2).
Another level of MAGI dysregulation relates to the targeting of the corresponding transcripts by a series of microRNAs (Table 2). MAGI1 mRNAs are down-regulated by miR-486-5p in the human erythroid leukemia K562 cell line, leading to decreased RA-GEF-1 /Rap1A signaling pathway and inhibition of erythroid differentiation [136]. In renal cell carcinoma, a low MAGI1 expression level, as a result of c-Myb-induced miR-520h transactivation, is associated with a poor prognosis [137]. MAGI2 is targeted by a series of miRNA including miR-629-5p, miR-487a, miR101, and miR-134/487b/655 cluster [138][139][140][141]. In breast cancers, miR-487a was positively correlated with lymph nodes metastasis and negatively correlated with the expression of MAGI2. At the mechanistic level, miR-487a is transactivated by the TGF-β /NFκB signaling pathways and promotes epithelial-mesenchymal transition (EMT) and invasiveness via targeting MAGI2/ PTEN axis [139]. In these cancers, miR-101 impedes the same pathway to enable estrogen-independent growth [140]. MiR-27a-3p functions as an oncogenic miRNA in breast cancer cells themselves, but also redesigns the tumor microenvironment. Accordingly, the release of this oncomiR in exosomes from breast cancer cells downregulates MAGI2 in macrophages, leading to upregulation of PDL1 and immune escape of cancer cells [142]. In lung carcinoma cell lines, miR-134/487b/655 cluster regulates TGF-β-induced EMT and drug resistance to gefitinib by targeting MAGI2 [141].
As regards MAGI3, its downregulation by oncomiR-34c-3p enhances β-catenin/ c-Myc -induced glycolytic gene expression/ Warburg effect and promotes the growth of human hepatocellular carcinoma cell lines in vitro and in vivo [143].
MAGIs are also subjected to posttranslational modifications, including phosphorylation, SUMOylation, and cleavage by caspase 3 and 7 (Table 1), whose regulation, biological impact, and pathological implications remain are so far poorly documented.

MAGIs Expression and Prognosis
As mentioned above, the accumulation of MAGI transcripts and proteins is proven to be associated with prognosis in various human malignancies (Table 3). In renal, liver, and gastric cancers and in lymphocytic leukemia, MAGI1 downregulation is associated with a poor prognosis (Table 3) [122,137,145,146]. In gliomas, MAGI1 is downregulated at both transcript and protein levels, and this expression is negatively correlated with the WHO grade of the tumors [56]. In estrogen receptor-positive (ER+)/human epidermal growth factor receptor 2 negatives (HER2−) breast cancers, MAGI1 is considered a tumor suppressor [88,147]. In this subtype of tumors, MAGI1 expression is positively correlated with the accumulation of the transcript encoding estrogen receptor 1, the luminal fate transcription factors GATA3 and FOXA1 levels, and a lower risk of relapse, whereas low MAGI1 levels correlate with more aggressive phenotype and worse prognosis. It turns out that MAGI1 and estrogen receptors mutually regulate each other. On the one hand, MAGI1 silencing in the MCF7 breast cancer cell line decreases ESR1 mRNA and ERα protein levels, activates PI3K/Wnt signaling, promotes cell proliferation, and reduces apoptosis and epithelial differentiation. On the other hand, estradiol treatment of the ER+ MCF7 cell line upregulates MAGI1 mRNA levels [147]. In the experimental model of metastasis, MAGI1 downregulation triggers the metastatic phenotype of the mouse mammary 67NR (ER+) cells, whereas its ectopic expression impairs the metastatic properties of the 4T1 (ER−) cells. MAGI1 expression negatively correlates with the inflammatory status of breast tumors and is downregulated by the prostaglandin E2/ COX-2 axis. These observations are in line with another study demonstrating that Cox-2 inhibitor upregulates MAGI1 in human colon cancer cell lines [148].
As regards MAGI2, conflicting observations have been reported in prostate cancers (Table 3). Cao et al. found a decreased accumulation of MAGI2 transcripts in prostate cancer as compared to control tissue, whereas the immunostaining of MAGI2 was unaffected [149]. They also reported that high MAGI2 expression is associated with longer disease-free survival. Similarly, Mahdian et al. reported a decreased accumulation of MAGI2 transcripts in prostate cancer [150]. In contrast, other studies evidenced increased MAGI2 immunostaining in high-grade prostatic intraepithelial neoplasia (HGPIN) and tumor, as compared with control tissue [151][152][153][154][155], with a decreased expression during progression from HGPIN to adenocarcinoma of low to high Gleason score [151]. Nevertheless, it is noteworthy that MAGI2 mutations are found at a relatively high incidence (26%) in these cancers (Cosmic V94, URL http://cancer.sanger.ac.uk/cosmic/, released 28 May 2021, accessed on 22 July 2021). In multiple myeloma, MAGI2 is overexpressed and is significantly associated with unfavorable patient outcomes [156]. A super-enhancer (large clusters of enhancers) in the MAGI2 gene is a driver of oncogenic phenotype in multiple myeloma and is further accentuated by binding the oncogenic MAF transcription factor [156]. Nevertheless, further studies are required to assign this phenotype to MAGI2 itself or other genetic elements, e.g., MAGI2-AS3 that shares the same promoter. In contrast, as stated above, MAGI2 promoter is hypermethylated in many solid tumors, including cervical, gastric, endometrial, ovarian, and breast cancers. This hypermethylation and MAGI2 silencing has been associated with a poor prognosis in gastric and breast cancers [125,128].
Concerning MAGI3, high levels of transcripts predict an excellent prognosis in ovarian cancers [157]. In human gliomas, MAGI3 is downregulated at both mRNA and protein levels, and this expression is negatively associated with tumor grade and prognosis [55]. Overexpression of MAGI3 in the human U373 and LN229, and C6 rat glioma cell lines reduces cell proliferation, arrests the cell cycle, and inhibits migration [55].
The role of MAGIs in the control of carcinogenesis is further highlighted by human papillomaviruses (HPV) E6 viral oncoprotein-induced degradation. The high-risk types HPV16 and HPV18, are associated with cancer onset in humans, where the most common is cervical cancer. This pathogenesis is tightly linked to the combined action of E6 and E7 oncoproteins, which cooperate efficiently to immortalize human keratinocytes and whose sustained expression is required for maintaining the proliferative potential and to prolong the survival of tumor-derived cells. The E6 oncoprotein targets several proteins with the PDZ domain for degradation by the proteasome, among them MAGI scaffolding molecules through interaction with PDZ1 [109,158]. The use of E6 protein deficient in the PDZ binding motif allowed us to delineate the impact of the selective downregulation of protein with the PDZ domain in neoplastic progression. Single depletion of MAGI1, SCRIB, or PAR3 significantly but partially restored the tumorigenic ability of the E6 protein depleted of the PDZ binding motif [159]. On the other hand, the re-expression of a MAGI1 variant resistant to E6 targeting in HPV-positive cells increases recruitment of ZO-1 and PAR3 at cell-cell contact, represses cell proliferation, and triggers apoptosis [118]. Phylogenetic analysis of oncogenic and non-oncogenic papillomaviruses suggests that the ability of their E6 proteins to induce degradation of PDZ-proteins is an enabling phenotype towards oncogenicity [39]. Another level of regulation of E6 oncogenic properties depending on the HPV type relies on their differential phosphorylation by PKA or AKT, which, in turn, inhibits PDZ recognition [160].

Signaling Pathways and Effector Systems Involved in the Tumor Suppressor Activities of MAGIs
As shown in Table 1 and Figure 2, MAGI proteins interact with a wide variety of effector proteins involved in the control of cell proliferation, apoptosis, or cell motility. Although not all of these interactions have so far been linked to MAGI antitumor activity (e.g., MAPK, Notch, Rho, Smad), some of them are clearly established (Figure 3) [164][165][166].

PTEN/PI3K/AKT Signaling Pathway
Phosphatase and tensin homolog (PTEN) is a tumor suppressor that represents one of the most prevalent targets for genetic alteration in human cancer. PTEN controls a broad range of physiopathological processes related to cell proliferation, differentiation, DNA/chromosome integrity, apoptosis, and invasiveness. PTEN dephosphorylates not only proteins but also phosphoinositides generated by phosphatidylinositol 3-kinase, thus counteracting the Akt signaling pathway [1,2]. PTEN exerts also some biological functions independently of its catalytic activity. PTEN contains a PDZ binding motif in its C-terminus and interacts with several proteins with PDZ domains including MAGI proteins through PDZ2 [22,[96][97][98]. This interaction-impaired by PTEN C-terminus phosphorylation-, increases PTEN stability and enhances PTEN mediated downregulation of PI3K/AKT activity [22,31,56,96,123,139,140,[167][168][169]. We demonstrated that the inhibition of the PI3K/ AKT pathways by the PTEN/MAGI1 signalosome exerts a critical role in restraining the invasive phenotype [22,170]. In this regard, MAGIs protein levels and PTEN accumulation and activity proved to be correlated [56,152,171,172]. This is of high importance since Alimonti et al. have demonstrated that, even a subtle reduction in PTEN level is sufficient to promote cancer susceptibility [173]. In this connection, celecoxib, a Cox-2 inhibitor upregulates MAGI1 in the human colon cancer SW480 and HCT116 cell lines, curtailing primary tumor growth and spontaneous lung metastasis in an orthotopic model of colorectal cancer [142]. Interestingly, celecoxib was also reported to promote the membrane translocation of PTEN and PTEN activity through deacetylation, leading to AKT inactivation [174]. One interesting point that was not addressed in this study, concerns the impact of MAGI1 expression on celecoxib-induced PTEN targeting to the plasma membrane. In the breast cancer MDA-MB-231 cell line, it has been reported that stimulation of PAR2 by tissue factor, transiently reduces PTEN interaction with MAGI2, this release being associated with enhanced PTEN activity [175].
MEK1 interacts with the GuK and WW domains of MAGIs proteins and proved to be necessary for PTEN recruitment at the plasma membrane and deactivation of the AKT pathway. This complex formation is independent of MEK1 kinase activity but requires MEK1 phosphorylation at T292 by ERK [94]. These interconnections between the RAF/MEK1/ERK and PI3K/AKT pathways suggest that RAF or MEK inhibitors might promote AKT activation and emphasize the requirement of combined targeting of these two pathways to circumvent the emergence of resistance to chemotherapies.

Beta-Catenin/Wnt Signaling Pathways
Beta-catenin is a genuinely multifunctional protein, with two major cellular pools. It localizes (i) at the plasma membrane allowing the anchoring of cadherin cell-cell adhesion molecules to the actin cytoskeleton, and (ii) free in the cytosol or nucleus as an effector of the Wnt pathways that when disturbed is a major driver of cancer. Activation of the Wnt receptor frizzled allows accumulation and nuclear translocation of β-catenin, interaction with the TCF/LEF family of transcriptional factors, and the transactivation of target genes involved in cell proliferation, survival, and migration. In absence of Wnt ligands, β-catenin is maintained at a low level by the glycogen synthase kinase 3β (GSK3β). GSK3β forms a multimeric complex with APC, AXIN1, and β-catenin and phosphorylates the N-terminus of β-catenin, leading to its ubiquitination and degradation by the proteasome [176].
Beta-catenin does interact with MAGI PDZ5 at the plasma membrane [21,22,36,55]. This interaction strengthens junctional complexes (Vide infra) and prevents the nuclear translocation of β-catenin and the transactivation of Wnt target genes (Figure 3). In this context, MAGI3 knockdown in the PTEN defective U373 glioma cell line induces accumulation of the Wnt transcriptional targets cyclin D1 and Axin2 and enhances cell proliferation and migration [55]. Similarly, downregulation of MAGI1 in the PTEN defective U87MG and U251 glioblastoma cell lines enhances the accumulation of β-catenin, increases the expression of mesenchymal biomarkers, stimulates cell proliferation, and reduces cell apoptosis [56].
MAGIs might also control Wnt signaling through the PTEN/PI3K/AKT pathway. Accordingly, AKT phosphorylates β-catenin at Ser 552, which causes its dissociation from cell-cell contacts and accumulation in the cytoplasm and the nucleus, leading to β-catenin enhanced transcriptional activity and cancer cell invasiveness [177]. In contrast, the phosphorylation and inactivation of GSK3 by AKT do not seem to be involved in the control of the Wnt signaling. There may be different pools of GSK-3: one linked with Axin that is resistant to phosphorylation by Akt, and another pool that is regulated by AKT but did not appear to modulate β-catenin signaling [178,179].
Interestingly, in brain synapses, MAGI2 (S-SCAM) forms a complex with β-catenin and competes with GSK3 for AXIN binding and may protect β-catenin from degradation [89].
HIPPO Signaling Pathway YAP1 (Yes1 Associated Transcriptional Regulator) is the critical regulatory target in the Hippo signaling pathway. Hippo pathway regulation of YAP1 mediates cellular responses to mechanical tension, extracellular ligands, energy, osmotic and hypoxic stress, and inflammation and tissue injury [181]. In the nucleus, YAP1 interacts with transcription factors including TEA domain family members (TEADS) that trigger the expression of target genes implicated in cell transformation, proliferation, invasion, and metastasis [182]. The Hippo pathway involves a kinase cascade, including the Ser/Thr kinases STK3/ STK4, that activates LATS1/2 in complex with their regulatory protein MOB1, which in turn phosphorylate and inactivate YAP1 oncoprotein through cytoplasmic sequestration by 14-3-3 proteins [183].
A role of MAGI3 in the control of YAP1-dependent transformation was recently evidenced in breast cancer, through the interaction of the C-terminus of the transcriptional regulator with the PDZ5 of the scaffolding molecule [42]. In the human mammary MDA-MB-231 cancer cells, the premature polyadenylation of MAGI3 produces a truncated protein depleted of PDZ2-PDZ5 (MAGI3 pPA , Figure 1) that acts in a dominant-negative manner to prevent full-length MAGI3 from interacting and impairing YAP1 nuclear translocation ( Figure 3) [42,90]. Silencing of MAGI3 pPA transcripts and YAP1 activity markedly decreases the growth of MDA-MB-231 cancer cells xenografted in NOD/SCID mice, whereas ectopic expression of the variant promotes anchorage-independent growth of MCF10A-SV40 cells. This premature polyadenylation at a cryptic intronic poly(A) signal in MAGI3 identified in 7.5% of breast cancers -but not in tumor-adjacent control tissues-likely contributes to breast carcinogenesis. This variant is not related to mutations in the exon-intron boundaries but is associated with reduced N6-methyladenosine in MAGI3 large internal exon [90]. The premature polyadenylation of MAGI1 has also been reported, but its physiological significance remains unknown.
In line with MAGI3 control of the Hippo signaling pathway, an RNA-interference screen revealed that MAGI1 silencing decreases YAP1 phosphorylation, triggers its nuclear translocation, and the transactivation of a reporter gene [184]. The underlying mechanisms have not been investigated. Interestingly, using the proximity-dependent biotinylation identification technology (BioID) in HeLa and HEK293 cells, Couzens et al. reported the interaction of MAGI1 with LATS1, presumably via the WW domain [93]. MAGI2 and MAGI3 were also recently reported to interact with LATS1/2 as well as with other effectors of this signaling pathway via their WW domain tandems, including AMOT (angiomotin) and PTPN14 [34]. These tandems enhance the binding affinity and specificity for their targets [34]. Alternatively, MAGI proteins might indirectly control the YAP1 oncogenic pathway through the PTEN tumor suppressor. Accordingly, loss of PTEN phosphatase activity leads to LATS/MOB1 complex destabilization, decreased YAP1 phosphorylation and nuclear translocation, and promotes cell proliferation and migration of gastric cancer cells in vitro, and in vivo after xenografting in nude mice [185]. Furthermore, GSK3, which is inhibited by AKT, phosphorylates and triggers the degradation of the YAP1 related molecule TAZ [186]. Nevertheless, the regulation of YAP1 by MAGIs is seemingly complex. Accordingly, in endothelial cells, MAGI1 depletion markedly increases LATS1/2 at mRNA and protein levels leading to a reduction in YAP expression and endothelial permeability [106].

Junctional Complex Stabilizer, Invasion Suppressor Activity
Besides their critical role in the control of cell signaling, MAGI scaffolding proteins allow complex formation between a series of transmembrane proteins and actin-binding proteins at tight and adherens junctions (Figure 3).
Tight junctions exert a critical role in the maintenance of epithelial barrier integrity [4,10,12,38]. Increased paracellular permeability of intestinal epithelium allows increased translocation of commensal bacteria from the gut lumen and the development of inflammation, a driver of colorectal cancer [4]. In this concern, it is worth noting that genetic variants of MAG1, 2, and 3 have been associated with the development of inflammatory bowel diseases, including ulcerative colitis and Crohn's disease [187][188][189][190][191]. These pathologies are at high risk in the subsequent development of colorectal cancers [4,[192][193][194]. The role of MAGIs in the maintenance of tight junctions is further highlighted by the disruption of these junctions by viral oncoproteins that drive MAGIs to proteasomal degradation. Restoration of MAGI1 expression in the HPV-positive cervix cancer HeLa cell line induces cell growth arrest and apoptosis [118]. MAGI1 proves also to be a target of caspase-3 and -7 during the apoptotic process, MAGI1 cleavage contributes to cell junction disassembly, a key feature of apoptosis [103]. In endothelial cells, thrombin stimulates p90RSK that directly phosphorylates MAGI1 at Serine 741 and induces MAGI1 deSUMOylation at Lysine 931, the consecutive release of MAGI1 from adherens and tight junctions leading to endothelial permeability [106]. In this model, this effect was mimicked by MAGI1 down-regulation. The interaction of β-catenin with MAGI1 seems to be involved in the maintenance of tight junction integrity [20].
MAGIs act also as a suppressor of invasiveness through the stabilization of adherens junctions. We previously demonstrated that ectopic expression of MAGI1B enhanced the recruitment of PTEN to junctional complexes, promoted E-cadherin-dependent cell-cell aggregation, and reversed the Src-induced invasiveness of kidney MDCKts-Src cells [22]. The local down-regulation of phosphatidylinositol-3,4,5-trisphosphate pools and downstream effector systems at the site of cell-cell contacts proved to be focal points for restraining both disruptions of junctional complexes and induction of tumor cell invasion [22]. Similarly, in the human colon cancer SW480 and HCT116 cell lines, MAGI1 overexpression stabilizes E-cadherin localization at cell-cell junctions, enhances actin stress fiber, and focal adhesion formation, increases cell adhesion to matrix proteins, and suppressed anchorageindependent growth, migration, and invasiveness [148]. Conversely, MAGI1 silencing decreases E-cadherin and β-catenin localization at cell-cell junctions, disrupts actin stress fiber and focal adhesion, and promotes anchorage-independent growth, migration, and invasion in vitro. In the breast cancer MDA-MB-231 and MCF7 cell lines, down-regulation of miR-487a restores MAGI2 expression, suppresses epithelial-mesenchymal transition (EMT), and inhibits the invasive phenotype [139]. Similarly, restitution of MAGI1 expression in glioblastoma cell lines decreases proliferation, migration, and invasion through AKT, MMP2, MMP9, and the E-cadherin/ N-cadherin/vimentin pathway [22,195].
In luminal breast cancer MCF7 cell line, downregulation of MAGI1 is associated with an increased accumulation of AMOTL2 (Angiomotin Like 2) and of E-cadherin, and to enhanced cell proliferation, but not cell migration or invasiveness [88]. This process seems to be related to an increased cell stiffness caused by actin cytoskeletal tension, leading to the activation of the p38 stress signaling pathway via AMOTL2 and ROCK. Interestingly, the Hippo and Wnt pathways were not affected.
In endothelial cells, MAGI1 promotes focal adhesion maturation and reduces its turnover. This process involves increased endothelial cells adhesion to the extracellular matrix, via the transient increase in both RhoA and Rac1 activities, and reduces invasiveness and tubulogenesis in vitro, and suppresses angiogenesis in vivo [196].

Competitions between Scaffolding Molecules and Molecular Partners
The role of MAGIs in the assembly and the subcellular targeting of supramolecular complexes enables alternative regulation of cell signaling through the competition with scaffolding molecules orchestrating other signalosomes. In this context, MAGI3 competes with NHERF2 for the binding to the carboxyl-terminal PDZ binding motifs of both lysophosphatidic acid receptor 2 (LPAR2) and phospholipase C β-3 (PLC-β3) (Figures 2 and 3, Table 1). In the HCT-116 colon cancer cell line, LPA triggers invasiveness through NHERF2 interaction with Gαq, and the activation of PLC-β3 and the downstream effectors nuclear factor-κB (NFκB) and c-Jun N-terminal kinase. By contrast, MAGI3 attenuated PLC-β activity and inhibited the invasive phenotype via coupling LPAR2 to both Gαq and Gα12, and potentiated LPA-induced RhoA activation [75]. The physiopathological relevance of this regulation is highlighted by the overexpression of NHERF2 and the downregulation of MAGI3 in sporadic human colorectal cancers and intestinal tumors from APC Min/+ mice. This antagonistic activity of NHERF2 and MAGI3 in activating PLC-β3 was also evidenced for the P2Y1 purinergic receptor [75].
On the other hand, MAGI molecular partners might compete for binding to the same MAGI interaction modules. In this connection, we have demonstrated that TRIP6-a LIM protein related to Zyxin, LPP, Ajuba, and LIMD1-competes with β-catenin for interacting with the PDZ5 of MAGI1B, leading to junctional complexes destabilization and invasiveness ( Figure 2) [26]. This release of β-catenins from junctional complexes has been also associated with their nuclear translocation and the transactivation of Wnt target genes [92]. The implication of TRIP6 in tumor progression is illustrated by its overexpression in various cancers, including colorectal, hepatocellular, gastric, breast, and cervical cancers [26,92,[197][198][199][200]. Besides junctional complex destabilization, we demonstrated that PI3-kinase/AKT and NFκB pathways were implicated in TRIP6-induced invasiveness [26]. TRIP6 also interacts with the Hippo/YAP1 pathway. Under high mechanical tension (low cell density), TRIP6 is recruited to adherens junctions of the mammary MCF10A cells by vinculin, and competes with MOB1 for its interaction with LATS1/2, leading to YAP1 activation and cell proliferation [201]. Likewise, in cervical cancer cell lines, TRIP6 promotes proliferation and invasiveness through the YAP1 activation [202]. Whether defective MAGI/ PTEN or MAGI/ β-catenin axes are involved in these processes remains an interesting point to investigate.
Different levels of MAGI tumor suppressor activities are displayed in Figure 3.

MAGI ceRNAs in Carcinogenesis
Interestingly, it was recently evidenced that MAGI gene loci are involved in the control of carcinogenesis, independently of the transcription of the corresponding MAGI scaffolding molecules, their splice variants, and alternative polyA truncated isoforms, but through transcripts that act as competing endogenous (ceRNA) RNAs. CeRNA appears as a novel regulatory mechanism between non-coding miRNA and coding RNA. These long non-coding RNAs (lncRNAs) can function as miRNA sponges, thus inhibiting the degradation of miRNAs target transcripts.
Similarly, MAGI1-IT1 is overexpressed in omental and mesenteric metastases of epithelial ovarian cancers as compared to primary tumors or control tissue and is positively associated with the III-IV FIGO stage. The overexpression or the downregulation of MAGI1-IT1 did not affect the proliferation of the SKOV and ES-2 ovarian cancer cell lines, in vitro and subcutaneously xenografted in nude mice. In contrast, MAGI1-IT1 promoted SKOV and ES-2 cell migration and invasion in vitro and intraperitoneal dissemination after orthotopic xenograft in nude mice [206]. MAGI1-IT1 promotes EMT, cell migration, and invasion by competitively binding miR-200a leading to the upregulation of the transcription repressors ZEB1 and ZEB2 [206] (Table 4). Epithelial-mesenchymal transition, invasiveness. Ovarian cancer cell lines [206] miR-302d-3p IGF1 (insulin-like growth factor 1) Structurally and functionally related to insulin. Growth-promoting activity through binding to IGF1R tyrosine kinase, leading to activation of the PI3K-AKT/PKB and the Ras-MAPK pathways.
Overexpression MAGI1-IT1 in gastric cancer is associated with poor overall survival. Knockdown in gastric cancer cell lines decreases proliferation in vitro and in vivo [207] miR-512p-3p AKT (protein kinase B, PKB) Proto-oncogene. Ser/Thr kinase downstream of phosphatidylinositol 3-kinase (PI3K). Involved in cell proliferation, survival, and migration.
Increased AKT activity. Increased lung cancer cell lines proliferation [205]

Transmembrane protein involved in cell aggregation
Decreased proliferation and migration in vitro, enhanced apoptosis, decreased tumor growth, and experimental metastases in nude mice. Lung squamous carcinoma xxxSW900, SK-MES_1, and A549 and cell line.

Induces apoptosis, decreases cell proliferation and migration
Downregulation of miR15b-5p and miR-374a/b-5p by MAGI2-AS3 decrease the viability and migration of the epithelial ovarian cancer PEA1, KURAMOCHI, and SKOV3 cell lines. Downregulation of MAGI2-AS3 due to promoter hypermethylation. Decreased migration and invasion of breast cancer cell lines [209,216] miR374b-5p SMG1 (genitalia family member 1) Ser/Thr protein kinase involved in nonsense-mediated decay of mRNAs containing premature Stop codons. Can phosphorylate and trigger TP53 activation after cell exposure to genotoxic stress Decreased proliferation and migration. Hepatocarcinoma cell lines [217] RECK * (reversion inducing cysteine-rich protein with kazal motifs; ST15, suppressor of tumorigenicity 15 protein) Extracellular protein anchored to the plasma membrane. Decreases diffusible gelatinase/collagenase activity Downregulation of miR15b-5p and miR-374a/b-5p by MAGI2-AS3 decrease the viability and migration of the epithelial ovarian cancer PEA1, KURAMOCHI, and SKOV3 cell lines. Downregulation of MAGI2-AS3 due to promoter hypermethylation [209] HOXA5 (homeobox A5) Transcription factor upregulates the tumor suppressor p53 Downregulation of miR15b-5p and miR-374a/b-5p by MAGI2-AS3 decrease the viability and migration of the epithelial ovarian cancer PEA1, KURAMOCHI, and SKOV3 cell lines. Downregulation of MAGI2-AS3 due to promoter hypermethylation [209] miR-525-5p MDX1 (MAX dimerization protein) Competes with MAX (Myc associated factor X) for interaction with the c-Myc proto-oncogene. Inhibits transactivation by this transcriptional activator complex Decreases proliferation and migration of ovarian cancer cell lines [218]    Involved in the regulation of lysosomal trafficking Decreased apoptosis, increased proliferation, and migration of the HCT116 and RKO colon cancer cell lines [229] : Inhibition of the target; : Activation of the target; * Secondary targets regulated by different miRNAs exhausted by MAGI2-AS3 are in bold.
MAGI1-IT1 is also overexpressed in gastric cancers and associated with a poorer patient overall survival. Mechanistically, MAGI1-IT1 sequesters miR-302d-3p enabling accumulation of its target insulin-like growth factor 1 (IGF1). Knock-down of the lncRNA in the human gastric cancer AGS and MGC-803 cell lines decreases proliferation in vitro and tumor growth in nude mice [207].
A MAGI1 antisense RNA has been identified, but to our knowledge, no physiological role has been described ( Figure 4).
Three MAGI2 antisense RNAs have been identified (Figure 4). If no function has been so far attributed to MAGI2-AS1 and MAGI2-AS2, the lncRNA MAGI2-AS3 seems to act as a tumor suppressor in a tissue-dependent manner through different molecular processes, including sponging some sets of miRNAs and regulating epigenetic mechanisms (Table 4).
In non-small cell lung carcinomas (NSCLCs), MAGI2-AS3 is downregulated and associated with tumor size, TNM stage, distant metastasis, and shorter overall survival [210,211,214,230]. In the human NSCLC H1993 cell line, the inhibition of migration and invasiveness by MAGI2-AS3 has been attributed to quenching miR-25 and upregulation of RECK (reversion inducing cysteine-rich protein with kazal motifs). Reck is an extracellular protein that may serve as a negative regulator for matrix metalloproteinase-9 [212]. Concurrently, MAGI2-AS3 may control NSCLC cell proliferation via miR-155/ SOCS-1 axis [214]. In lung squamous carcinoma cell lines, the subtype of NSCLC, MAGI2-AS3 overexpression inhibits proliferation and migration in vitro, as well as tumor growth and experimental metastasis of human adenocarcinoma A549 cells inoculated in nude mice. In this study, MAGI2-AS3 proved to sponge miR-374a/b-5p and miR-23a-3p that deplete cell adhesion molecule 2 (CADM2) and PTEN transcripts, leading to upregulation of CADM2, PTEN, caspase 3, and Bax, and reduced level of Bcl-2 [211]. MAGI2-AS3 was also reported to suppress the proliferative and invasive abilities of the A549 and PC9 NSCLC cell lines by targeting miRNA-23a-3p leading to restored expression of the PTEN tumor suppressor [210]. Interestingly, the level of MAGI2-AS3 is decreased in tumor-educated blood platelets and plasma of patients with NSCLC and is correlated with TNM stage, lymph node metastasis, and distant metastasis, making this lncRNA a potential biomarker for the noninvasive diagnosis of lung cancer [230].
In breast cancer, MAGI2-AS3 is downregulated compared to adjacent control tissues, and this accumulation is negatively correlated with histological grade, TNM stage, ER expression, PR expression, and HER-2 expression [223]. Among the clinical group of breast cancer, MAGI2-AS3 is poorly expressed in tumors with lymph node involvement [231]. Furthermore, an elevated level of MAGI2-AS3 is associated with better relapse-free survival for patients with triple-negative breast cancer subtype (absence of estrogen and progesterone receptors, no amplification of human epidermal growth factor receptor HER2) associated with a poor prognosis [232]. Mechanistically, Yang et al. showed that overexpression of the lncRNA in MDA-MB-231 and MCF-7 breast cancer cells inhibited proliferation and promoted cell apoptosis, at least partly through increased expression of Fas and Fas ligand (FasL) [223]. On the other hand, Du et al. reported that MAGI2-AS3 impedes miR-374a that exhausts the transcripts of the tumor suppressor PTEN [216].
Interestingly, in hepatocellular carcinoma (HCC), MAGI2-AS3 targets miR-374b-5p, and inhibits the proliferation and migration of HepG2, Hep3B, and MHCC-97H HCC cell lines in vitro, and the growth of HepG2 xenografted in nude mice [217]. In this study, Yin et al. identified genitalia family member 1 (SMG1) -involved in nonsense-mediated mRNA decay-as positively regulated by MAGI2-AS3, and showed that SMG1 knockdown reverses the suppressive function of MAGI2-AS3 in HCC cells [217]. Whether, PTEN, a target of miR-374b-5p [233] is also downregulated in HCC remains an intriguing point to investigate. The ectopic expression of MAGI2-AS3 also induces the downmodulation of ROCK2 (Rho-associated coiled-coil containing protein kinase 2) at both RNA and protein levels. Although the underlying mechanisms have not been investigated, this downmodulation decreases the migration and invasiveness and induces apoptosis in the Hep3B and MHCC97-H hepatocarcinoma cell lines [224].
As observed in NSCLC and breast cancer, downregulation of MAGI2-AS3 in HCC is strongly related to tumor size, lymph node metastasis, TNM stage, and shorter overall survival [217,221]. Furthermore, plasma levels of MAGI2-AS3 are significantly lower in patients with hepatocarcinoma compared with healthy individuals and decrease in case of distant recurrence after tumor resection [224]. Thus, plasma levels of this lncRNA might improve the diagnosis and follow-up of HCC.
Likewise, MAGI2-AS3 proved also to be downregulated in gliomas and associated with shorter overall survival [234].
An in sillico analysis of expression profiles of lncRNA, miRNA, and mRNA, along with the clinical information of bladder carcinoma patients allowed us to identify MAGI2-AS3 as a key node encompassing 17 miRNA nodes and 46 mRNA nodes. Among the putative miRNA targets, miR-200 and miR-143 are considered tumor suppressors in human bladder cancer [226]. Nevertheless, several studies have demonstrated in this cancer that MAGI2-AS3 was downregulated and negatively associated with tumor stage [208,213,235]. In this regard, Wang et al. showed that low levels of MAGI2-AS3 were associated with a poor prognosis and that ectopic expression of MAGI2-AS3 inhibited proliferation, migration, and invasion of the human bladder cancer cells T24 and RT4 in vitro, and suppressed the growth of T24 cells in nude mice [208]. MAGI2-AS3 could serve as a competing ceRNA for miR-15b-5p that targets CCDC19 (CFAP45, cilia, and flagella associated protein 45), a tumor suppressor in bladder cancer. Rescue assays demonstrated that knockdown of CCDC19 restored the proliferation, migration, and invasion of bladder cancer cells suppressed by MAGI2-AS3 overexpression [208]. Tang et al. also demonstrated that MAGI2-AS3 upregulates tensin 1 and inhibits the migration and invasion properties of bladder cancer cell lines by quenching miR-31-5p [213].
In the context of epithelial ovarian cancer, specifically high-grade serous ovarian carcinoma, MAGI2-AS3 acts as a tumor suppressor by sponging miR-15b-5p, miR-374a-5p, and miR-374b-5p, and may regulate the expression of their target mRNAs PTEN, RECK, Metastasis suppressor protein 1 (MTSS1), Homeobox -A5 (HOXA5) [209]. MAGI2-AS3 might also inhibit the proliferation and invasiveness of ovarian cancer cells through downregulation of c-Myc signaling. MAGI2-AS3 is poorly expressed in ovarian cancer tissues compared with adjacent control tissue. Restoration of MAGI2-AS3 expression in the human ovarian cancer SUN8 and Caov3 cell lines depletes miR-525-5p which exhausts MXD1 (MAX dimerization protein 1). MDX1 antagonizes c-Myc-mediated transcriptional activity by competing for the binding partner MAX and by recruiting repressor complexes containing histone deacetylases [218].
A polymorphism in MAGI2-AS3 (rs7783388 GG genotype) seems to enhance colorectal cancer risk as a result of decreased binding affinity of glucocorticoid receptor to the MAGI2-AS3 promoter, leading to lower MAGI2-AS3 accumulation [236]. In this connection, functional experiments showed that MAGI2-AS3 overexpression suppressed proliferation (cell cycle arrest at G0/G1 phase), decreased migration and invasiveness, and promoted apoptosis of the human colon cancer SW480 and SW620 cell lines [236]. In this cancer, a MAGI2-AS3 LncRNA-miRNA-mRNA network of differentially expressed genes has been recently proposed [237].
Another mechanism for the tumor suppressor activity of MAGI2-AS3 independent of miRNA sponging has also been evidenced in HCC cells [221]. MAGI2-AS3 binds and recruits the lysine demethylase 1A (KDM1A) to the RACGAP1 promoter in an RNA-DNA manner. The demethylation of histone H3 at lysine 4 (H3K4me2) enriched in the RACGAP1 promoter region downmodulates RacGAP, inhibits proliferation, and promotes apoptosis of HEPG2 cells in vitro, and decreases tumor growth in nude mice [221]. Likewise, MAGI2-AS3 impairs leukemia stem cells' self-renewal through recruitment of TET2 (Tet methylcytosine dioxygenase 2, involved in cytosine demethylation) to the LGIR1 promoter. The enhanced expression of LRIG1 (leucine-rich repeats and immunoglobulin-like domains 1) might negatively regulate the EGFR, MET, and RET receptor tyrosine kinases [222]. The low expression of MAGI2-AS3 in these malignant cells favors the self-renewal and the maintenance of the clonal hierarchy of acute leukemia stem cells. In esophageal cancer, MAGI2-AS3 can down-regulate HOXB7 expression by recruiting the histone-lysine Nmethyltransferase EZH2 (enhancer of zeste homolog 2) to promote the tri-methylation of lysine 27 on histone H3 in the HOXB7 promoter region. The ensuing decreased transcription of HOXB7 inhibits proliferation and radio-resistance of esophageal cancer cell lines in vitro and in vivo [219]. Both MAGI2-AS3 and EZH2 are expressed at low levels in esophageal carcinomas. In clear cell renal cell carcinoma, the downregulation of MAGI2-AS3 is associated with a poor prognosis. MAGI2-AS3 forms a nucleoprotein complex with HEY1 (Hes related family BHLH transcription factor with YRPW motif 1) in the nucleus and impairs its transcriptional repressor activity on ACY1 (aminoacylase 1) promoter region [220]. Restoration of ACY1 expression by MAGI2-AS3 decreases viability, expression of EMT markers, migration, and vasculogenic mimicry in vitro, and reduces the growth and angiogenesis of the human clear cell renal cell carcinoma RLC-310 cell line xenografted in nude mice [220].
Interestingly in breast cancer cells, MAGI2-AS3 proves to act as a cis-acting regulatory element, decreasing MAGI2 promoter methylation through TET1 recruitment. Restoration of MAGI2-AS3 expression in the MCF-7 mammary cell line induces MAGI2 expression, leading to the inactivation of AKT and Wnt/β-catenin pathways and impaired cell proliferation and migration [128].
In contrast, MAGI2-AS3 proved to be overexpressed in gastric cancers and nasopharyngeal carcinomas [225,228,240]. MAGI2-AS3 overexpression enhances nasopharyngeal cells proliferation and migration, and their resistance to cisplatin. This process is related to miR-218-5p downregulation, leading to increased accumulation of GDPD5 (glycerophosphodiester phosphodiesterase domain containing 5) that cleaves the glycosylphosphatidylinositol anchoring proteins to the plasma membrane, and SEC61A1 (SEC61 translocon subunit α1) involved in the insertion of polypeptides into the endoplasmic reticulum [228]. In gastric cancer, MAGI2-AS3 is negatively associated with overall survival and diseasefree survival of cancer patients and promotes epithelial-mesenchymal transition (EMT), cell migration, and invasiveness by sponging miR-141/200a and upregulating the ZEB1 transcription factor [225]. In the metastatic lymph node from a gastric cancer HGC-27 cell line, MAGI2-AS3 overexpression might be caused by the transcriptional regulator bromodomain containing 4 (BRD4) and its interaction with histone H3 acetylated at lysine 27, enriched in the promoter region MAGI2-AS3 [225]. In the HCT116 and RKO human colon cancer cells, depletion of MAGI2-AS3 restrains proliferation and increases apoptosis [229]. This effect seems linked to the quenching of miR-3163 and the upregulation of TMEM106B.
In cervical cancer, conflicting results have been reported. Wang et al. showed that the expression of lncRNA MAGI2-AS3 was reduced in tumors compared to control tissues and that the promoter region of the lncRNA was hypermethylated [241]. In this study, the prognostic value of MAGI2-AS3 was not statistically significant. Similarly, Hou et al. evidenced a decreased accumulation of MAGI2-AS3 in cervix squamous cell carcinomas compared with control tissue [215]. They further demonstrated that MAGI2-AS3 inhibited SiHa and HeLa cell invasion and migration by sponging miR-223 leading to up-regulation of the tumor suppressor EPB41L3. In contrast, Liu et al. noticed that high levels of MAGI2-AS3 were associated with the poor survival of patients with cervical squamous cell carcinoma [242]. In the human cervical cancer C-33A cells, MAGI2-AS3 overexpression promotes cell proliferation by affecting the cell cycle through CDK6 upregulation [242].
These dual effects of MAGI2-AS3 as a tumor promoter or suppressor might be related to the cellular context, including tissue origin, the patterns of mRNA/ miRNA expression, the genetic defects, the pathways dysregulated but also the existence of different variants of this lncRNA, -due to different translation initiation sites, and alternative splicing-that might target distinct miRNAs. It appears therefore of major importance to perform the fine analysis of the pattern of expression of these variants in relation to physiopathology.

Conclusions, Perspectives
Scaffolding molecules exert a critical role in orchestrating cellular response by increasing the efficiency and selectivity of signal transduction. The emerging role of PDZ domain molecules in carcinogenesis might be related to some redundancy between these different protein families. In this concern, PTEN does interact with the three MAGIs, but also with other scaffolding molecules such as DLG1 and SCRIB. This further highlights the essential role of these scaffolds in the control of signal transduction. Nevertheless, we report here strong evidence concerning the downmodulation of MAGIs in various types of cancers and their physiopathological implications (Figures 3 and 5). Biochemical, biophysical, and mathematical modeling approaches have allowed one to identify molecular partners and effectors systems driven by scaffolding molecules, mainly through a global determinist approach. A challenge in our knowledge of these networks concerns the characterization of the dynamic of these signalosomes and their composition at the molecular level in their native environment. Molecular noise, stochasticity in the assembly of these multiprotein complexes could contribute to generating cellular signaling diversity. For instance, interdomain allostery, where binding to one PDZ domain changes the affinity at other domains, should modify the selectivity for other ligands. Using an experimental approach mimicking the anchoring of scaffold protein to the plasma membrane, Erlendsson et al. evidenced two-three higher orders of magnitude in intrinsic binding affinity than isolated PDZ interaction [243]. The limited amount of each molecular partner, their affinity, their diversity, their bioavailability in terms of subcellular localization or commitment in other scaffolds, their posttranslational modifications might contribute in a part of randomness in the assembly of these complexes. McCam et al. reported that the scaffolding protein PSD-95 does not stabilize specific complexes, but rather increases the frequency of shortlived binding events [244]. Thus, these complexes might only infrequently contain all the components necessary for signal transmission. Furthermore, these interactions seem to occur under non-equilibrium conditions [243]. Whether conceptualization of protein scaffolds as molecular machine and cell signaling as engineering wiring proved effective, these generic snapshots do not take into account the cellular context, either that real life is in multiple dimensions. In an editorial, Frank Gannon made a parallel between biology and physics before the advent of quantum physics [245]. A change in paradigm in biology might concern the dynamic integration of these molecular interactions, the self-organizing nature of the assembly, the abundance of partners and competitors, and the plasticity of these components using more probabilistic and stochastic approaches [246]. Much work remains to integrate all these different aspects to understand this networking leading to cellular outcomes and heterogeneity under physiological and pathological conditions. Nevertheless, targeting the multiple levels of regulation of MAGI scaffolding molecules enables unique opportunities to modulate simultaneously several pro/suppressive oncogenic signaling pathways controlling carcinogenesis ( Figure 5). Hypermethylation of CpG islands in promoters is a common feature of neoplastic progression. In this context, a series of DNA demethylating agents such as 5-azacitidine, 5-aza-2-deoxycytidine (decitabine), 4 -thio-2 -deoxycytidine (TdCyd), and hydralazine are under clinical trials (https://clinicaltrials.gov, accessed on 6 July 2021). Such compounds might reactivate MAGI expression silenced by promoter methylation. The recent success of the RNA vaccine demonstrates the technological feasibility of RNA delivery in humans that could be applied to selectively exhaust some oncomiRs which sponge transcripts encoding tumor suppressors, including MAGIs. Decreased methylation in MAGI3 large exon is correlated with the accumulation of the MAGI3pPA pro-oncogenic variant in breast cancers. Thus, targeting RNA m6A demethylases, such as ALKBH5 which is overexpressed in breast cancer and promotes stem cell maintenance [247][248][249][250] might restore MAGI3 full-length expression, hampering YAP1 signaling and promoting the PTEN tumor suppressor pathway. In this context, some inhibitors of ALKBH5 are under development and proved to efficiently exert antiproliferative activity in a cancer-cell-type-selective manner [251]. As far as MAGI1 is wild-type, its accumulation could be raised by Cox2 inhibitors leading to increased PTEN stability and activity. In this connection, the regulation of endothelial cell activity, and the tuning of vascular permeability and angiogenesis by MAGI1 might constitute another level to control carcinogenesis [166]. Growing and promising efforts are devoted to evaluating the druggability of PDZ domains [7,8]. Such an approach could benefit neurological, viral, and oncological disorders. In this connection, using a structure-based computational design framework that models peptide flexibility, Zheng et al. designed peptides that bind the PDZ2 of NHERF2 but do not interact with the PDZ5 of MAGI3 [252]. These peptides might selectively hamper NHERF2/ LPA2R interaction and oncogenic signaling, without affecting LPA2R/ MAGI3 biological activity. Alternatively, some ligands of PDZ domains could be neutralized by cell-penetrable nanobodies (transbodies) favoring selective interaction of competing molecules, e.g., TRIP6 vs β-catenin. These approaches open up new perspectives in precision medicine to assess individual tumor responses to these targeted therapeutic agents. Interestingly, besides encoding scaffolding molecules, MAGI1 and MAGI2 genomic loci also generate lncRNAs that modulate neoplastic progression and might serve as non-invasive biomarkers of cancer progression [224,230]. These observations clearly illustrate that, as revealed by the human genome sequencing, the complexity of an organism is not solely a matter of intrinsic gene number, but rather reflects the functional role of epigenetic processes, non-coding genetic sequences, RNA editing, alternative splice, posttranslational modifications, multiprotein complexes assembly, and dynamic self-organizing systems that go well beyond the set of instructions of the "genetic program".