Differential expression of DLG1 as a common trait in different human diseases: an encouraging issue in molecular pathology

Epithelial cancer progression

A particular interest in DLG1 is due to its altered expression during the malignant progression of epithelial tumors (Figure 1A). Initial evidence was gathered from the analysis of HPV-induced cervical carcinogenesis. Cell culture experiments had shown that DLG1 was targeted for proteasomal degradation by the transforming oncoprotein E6 from high-risk HPV (Gardiol et al., 1999, 2002). Such findings prompted an investigation of DLG1 expression in the evolution of cervical lesions. In this sense, immunohistochemistry assays revealed a loss of DLG1 expression in invasive stages of the disease (Watson et al., 2002; Cavatorta et al., 2004). The fact that HPV E6 oncogene were highly upregulated in such stages suggested DLG1 proteasomal degradation, which was then supported by experiments showing that cervical cancer cell lines restored DLG1 levels and localization at the cell borders upon E6 ablation or proteasome inhibition (Massimi et al., 2004). However, striking changes in DLG1 expression were also observed in the early onset of disease, namely the preceding low-grade intraepithelial lesions (LSIL). In certain LSIL cases, reduced localization at the cell borders was evidently accompanied by increased expression in the cytoplasmic region (Cavatorta et al., 2004, 2017). Altogether, these initial analyses served to associate loss of DLG1 expression with cervical invasive stages, supporting the hypothesis of tumor suppressor functions of DLG1. Moreover, they also raised the possibility of a role of DLG1 mislocalization during the onset of cervical pathology. In addition, these analyses led to an interest to study the scenario in epithelial tumors unrelated to HPV infection. In this sense, biopsies from different colon adenocarcinomas were analyzed for DLG1 expression and, interestingly, some changes were detected. Semi-differentiated colon adenocarcinomas exhibited a reduction of DLG1 expression at the cell borders with a diffuse and increased cytoplasmic staining (Cavatorta et al., 2004; Gardiol et al., 2006). Furthermore, DLG1 protein was nearly undetectable in poorly differentiated stages. DLG1 expression was also analyzed in lung, larynx, breast and, more recently, in hepatocellular cancers. In all cases, a clear downregulation of its levels was found in late stages of malignancy (Fuja et al., 2004; Byeon et al., 2011; Szymanowska-Narloch et al., 2013; Wu et al., 2016). Moreover, in some of these reports, DLG1 upregulation and loss from cell borders were also observed, as well as mislocalization in intermediate lesions.

Figure 1:

Alterations in DLG1 expression are associated to different human pathologies.

(A) Loss of DLG1 normal cell localization at the cell borders and changes in its abundance are observed during epithelial tumor development at diverse anatomical sites. (B) Different neurological diseases are linked to diverse genetic mutations in the DLG1 gene and probably to changes in DNA methylation status. (C) Redistribution of DLG1 in T-lymphocytes with loss of its expression at the immunological synapses can deregulate signaling pathways, driving to inflammatory illnesses.

Taken together, the results described suggest a common pattern for DLG1 expression during epithelial tumor evolution. Loss of DLG1 expression could indeed be considered a late stage of epithelial malignancy. On the other hand, the loss of DLG1 from its regular site of cell-to-cell interactions, accompanied by eventual overexpression, may take place in less severe lesions. The intriguing question of possible effects of such misexpression on the pathology outcome (better or worse prognosis) have been addressed in some exciting recent works. In the case of cervical cancer, our group performed a follow-up analysis associating the DLG1 staining pattern at LSIL stage with progression of malignancy or regression to a normalization of the epithelium (as frequently occurs at this stage). Interestingly, all LSIL cases with a membrane-associated DLG1 staining, resembling that of a normal tissue, regressed. However, LSILs with DLG1 misdistribution and overexpression progressed to high-grade intraepithelial lesions. In another study which analyzed colorectal tumor samples, increased and cytoplasmic expression of DLG1 was closely associated with malignancy (Zhu et al., 2017). Interestingly, such patients suffered a more aggressive form of the disease being reflected on enhanced tumor evasion and low overall survival. Finally, research carried out in endometrial cancer tissues showed that patients with DLG1 cytoplasmic staining were linked to nodal metastasis and poorer overall survival than those with a strong or weak membrane-associated expression (Sugihara et al., 2016). Therefore, these data clearly suggest that DLG1 misexpression is highly associated with tumor progression or more malignant forms of the disease. Considering DLG1 scaffolding properties, it is possible that its contribution or implication in cancer relates to its inability to interact with the correct partners in a spatial/temporal manner. As DLG1 redistribution to the cytoplasm is normally achieved during the S phase of the cell cycle, this pattern of expression could be associated with a proliferative state of epithelial cells. In the context of a tumor cell, such misexpressions are likely to synergize with abnormal expression of wild-type or mutated oncogenes in order to fully contribute to cell transformation. In line with this, more studies are necessary to understand the precise mechanism by which DLG1 deregulation contributes to malignancy. Nevertheless, it is important to note that the DLG1 staining pattern could serve to predict evolution of disease.

Neurological disorders

DLG1 contributes to the maintenance of neuronal homeostasis. DLG1 participates in the trafficking of glutamate and aspartate receptors and α-subunits of the potassium channels Kv1.4 and Kv4.2 to the cell membrane, all of which are crucial for basal synaptic transmission and synaptic plasticity (Gardoni et al., 2007; Fourie et al., 2014). Accordingly, there is mounting evidence that associates DLG1 deficiency or DLG1 mutant variants with the development of a number of neurological disorders (Figure 1B) (Soler et al., 2018). Perhaps the best defined is schizophrenia (SZ), which relies on altered glutamate neurotransmission (Sato et al., 2008; Uezato et al., 2012). Studies analyzing schizophrenic post-mortem brains reported a marked loss of DLG1 expression in the prefrontal cortex region. Microdeletions in the 3q29 locus, which contains the DLG1 gene, were found in such patients (Toyooka et al., 2002; Mulle et al., 2010; Quintero-Rivera et al., 2010; Kirov et al., 2012; Glassford et al., 2016). More recently, additional causes that may contribute to SZ susceptibility in regard to DLG1 expression have been reported. A study employing DNA methylation chip arrays demonstrated changes in the methylation pattern of the DLG1 gene and the upstream region, whereas other reports showed SZ-associated non-synonymous single nucleotide variants (SNVs) within the DLG1 coding sequence (Fromer et al., 2014; Purcell et al., 2014; Xing et al., 2016; McKinney et al., 2017). Remarkably, similar genetic variations in the DLG1 gene were demonstrated in further neurological pathologies. SNV and microdeletion were observed in autism-spectrum disorders while single nucleotide polymorphisms (SNP, which represent SNV that are at least as 1% frequent in population) were detected in cases of frontotemporal dementia. This last disorder affects the frontal-temporal lobes of the brain and is linked to progressive deterioration of behavior and cognition (Rabinovici and Miller, 2010). Interestingly, an abnormal hypomethylation state of DLG1 promoter region was also found in this pathology (Taskesen et al., 2017).

Parkinson’s disease (PD) is a degenerative disorder of the central nervous system, and the etiological causes are poorly understood. In animal models, a reduced expression and altered cellular distribution of the rat DLG1 homologue, SAP97, was observed in Parkinson’s drug induced experiments (Nash et al., 2005). Furthermore, microarray studies analyzing gene expression in whole blood identified DLG1 as a candidate gene for PD (Sun et al., 2014). These reports suggest a potential involvement of DLG1 misexpression in the pathobiology of this neurological disorder.

Finally, interstitial microduplication of 3q29 has recently been described in patients with widely variable phenotypes. A case report study described a 1.607Mb duplication at 3q29 locus, involving several genes including DLG1, in an individual with intellectual disability, epilepsy and cerebral palsy. The association of this genetic change with DLG1 expression level and with these severe clinical manifestations requires further investigation to understand whether or not DLG1 dysfunctions could play a role in these syndromes (Fernández-Jaén et al., 2014).

Inflammatory and autoimmune diseases

Regulatory T cells (Tregs) are central players in maintaining immune tolerance to self-antigens, preventing autoimmune diseases (Kitagawa and Sakaguchi, 2017). Treg immunosuppressive functions rely on TCR engagement and activation of downstream signaling pathways (Li and Rudensky, 2016). Evidence suggests that DLG1 accumulation at the immunological synapse (IS) must be particularly important for Tregs activation as these cells were shown to recruit a four-fold higher level of DLG1 than other T-cell types (Zanin-Zhorov et al., 2012, 2017). Importantly, DLG1 was demonstrated to modulate both nuclear factor of activated T cells (NFAT) and phosphatidylinositol 3-kinase (PI3K)/protein kinase B (Akt) signaling pathways whose precise regulation is necessary for proper Treg functions. In this regard, DLG1 interacts to p38 MAP kinase and stimulates its activity which is required for NFAT activation and immunosuppressive cytokine production. In addition, DLG1 confers stability to PTEN phosphatase, which in turn downregulates the PI3K activity that normally impairs Treg functions (Zanin-Zhorov et al., 2012, 2017).

Current evidence indicates an association of autoimmune and inflammatory diseases with a reduced capacity of Tregs to recruit a sufficient amount of DLG1 at IS. One example is rheumatoid arthritis (RA), an autoimmune disease where diminished activity of Tregs has been observed. In RA patients Tregs are defective in suppressing CD8+ T-cell proliferation and, surprisingly, they were shown to accumulate lower levels of DLG1 at IS than Tregs from healthy individuals (Figure 1C) (Zanin-Zhorov et al., 2012). A similar scenario might occur in patients with the neuro-inflammatory disease known as HTLV-1-associated myelopathy/tropical spastic paraparesis (HAM/TSP). This pathology is linked to HTLV-1 infection and is characterized by a high infiltration of HTLV-1 infected Tregs into the central nervous system. Interestingly, such cells not only present low suppressive activity but also change to a pro-inflammatory phenotype (Araya et al., 2014). This phenomenon was explained in part by the activity of the HTLV-1 Tax protein, a key player for HTLV-1 replication, which is expressed at high levels in HAM/TSP (Araya et al., 2011; Enose-Akahata et al., 2017). HTLV-1 Tax has been reported to downregulate foxp3 expression (a transcription factor crucial for Tregs suppressive activities); however, it may use further mechanisms to interfere with Tregs functions. In this regard, we and others have demonstrated that Tax can directly interact with DLG1 and promote its hyperphosphorylation leading to DLG1 mislocalization from the cell borders to the MTOC/Golgi region (Figure 1C) (Hirata et al., 2004; Marziali et al., 2017) . Thus, this interaction might interfere with DLG1 accumulation at IS upon TCR engagement, negatively affecting suppressive responses of infected Tregs. Interestingly, HAM/TSP individuals have a tendency to develop RA and studies using transgenic mouse models have confirmed a role of HTLV-1 Tax in the onset of such disease (Habu et al., 1999; Yakova et al., 2005).

Transcriptional level

The first attempt to identify elements regulating DLG1 expression was carried out by cloning the minimal region with promoter activity upstream from the transcription start site initially reported by Lue and collaborators (Lue et al., 1994; Cavatorta et al., 2008). A detailed analysis of such regions revealed the presence of several consensus binding sites for transcription factors like nuclear factor kappa B, AP-1, AP-2, SP-1, Snai1 (Cavatorta et al., 2008). More recently, a study has added the human zinc finger 191 (ZNF191) factor to the list (Wu et al., 2016). Although their functions have been well studied, the influence on DLG1 promoter has only been ascertained for Snai1 and ZNF191 (Figure 2A).

Snai1 is a transcription factor member of the Snail family which participates actively in the epithelial-mesenchymal transition (EMT) during embryonic development, driving epithelial cell migration. Reactivation of EMT is one of the most striking features of epithelial tumor cells in advanced stages of malignancy. This can be triggered by the tumor environment through secretion of cytokines like epithermal growth factor, transforming growth factor-beta, Notch, Wnt and tumor necrosis factor-alpha. In response to such signals, epithelial tumor cells upregulate the EMT promoting factor Snai1 (Wang et al., 2013). Interestingly, Snai1 was shown to downregulate DLG1 expression by interacting with E-box-like elements present in DLG1 promoter and repressing gene transcription (Cavatorta et al., 2008). Therefore, the observed decrease in DLG1 levels frequently reported in advanced stages of epithelial tumors may be, in part, a consequence of EMT program reactivation triggered by Snai1.

DLG1 expression was recently found to also be regulated by the transcription factor ZNF191, which plays a role in mammal embryonic development as well as in adult organs (Li et al., 2009). In hepatocytes, ZNF191 inhibits migration by downregulating the Hippo-Yes-associated protein (YAP) signaling pathway. More specifically, ZNF191 promotes the cytoplasmic retention of the signaling activator YAP through phosphorylation. Strikingly, ZNF191-mediated YAP phosphorylation was shown to depend on DLG1 expression, suggesting a functional link between ZNF191 and DLG1 during HIPO-YAP signaling regulation. This relationship was clarified having found that ZNF191 interacts with TCAT repetitive motifs on DLG1 promoter and activates its transcription (Wu et al., 2016). Importantly, ZNF191 was found to be downregulated in invasive hepatocellular carcinoma. Hence, this fact could strongly contribute to understanding the loss of DLG1 expression found in such stages of malignancy (Wu et al., 2016).

Finally, the promoter region of DLG1 presents sequences enriched with CpG dinucleotides. Such elements allow transcription from different start sites giving rise to pools of mRNA for DLG1 with high variability in the 5′ extremes, however, the usage of particular sites under different biological contexts has not yet been studied (Cavatorta et al., 2011). In addition, CpG dinucleotides allow epigenetic control by hypo or hyper-DNA methylation. Nevertheless, to date no studies have deeply analyzed how the epigenetic status of DLG1 promoter relates to DLG1 protein levels. This knowledge, as well as the identification of epigenetic regulators, could be particularly important for the neurological disorders explained before.

Post-transcriptional level

Genes that must be precisely regulated, such as oncogenes and genes encoding transcription factors, receptors, transduction pathway regulators and tumor suppressors, often give rise to transcripts with variable 5′ untranslated regions (5′UTR) (Hughes, 2006; Willimott and Wagner, 2010; Ishii and Sakuma, 2011). DLG1 could be placed in such category as a 5′UTR alternative splicing event was identified in that region (Figure 2B). Such a mechanism was detected in different cell types, highlighting its conserved nature. Of note, the spliced exon contains two elements that regularly have a negative influence on translation efficiency: a short upstream open reading frame (uORF) and stable stem loops. uORF are genetic elements reported to decrease the translation rate of main ORFs by reducing the availability of ribosomes. On the same hand, stem loops within 5′UTR provoke ribosome stalling and may also create binding sites for cellular factors that affect translation efficiency (Smith, 2008; Hinnebusch et al., 2016). Importantly, the negative impact of both elements on DLG1 expression was confirmed having found a higher translation efficiency for the spliced variant (Cavatorta et al., 2011). These findings opened the possibility of 5′UTR splicing as a mechanism to regulate DLG1 levels in biological contexts in which DLG1 plays an important role. In this regard, an upregulation of 5′UTR spliced version relative to the unspliced form was observed during adherens-junction formation of colon epithelial cells, where an increase in DLG1 levels has been reported (Marziali et al., 2015). However, a less significant role of this mechanism was found in DLG1 decreased levels observed during cell cycle progression from G1 to the S phase (Marziali et al., 2015). Therefore, these data revealed that the extent of 5′UTR splicing as a mechanism to control DLG1 protein levels is highly dependent on specific cellular behaviors. Thus, it is tempting to speculate that interferences in this 5´UTR splicing mechanism could have potential to contribute to oncogenic processes by increasing or decreasing DLG1 levels depending on the tumoral context.

Post-translational level

DLG1 plays an important role in the organization of signaling pathways recruiting several kinases at specific cell regions and supporting the availability of substrates near the kinase active sites. In turn, phosphorylation is a major post-translational modification that controls DLG1 activities, interacting partners and cellular localization. Many different phosphorylation events by diverse kinases were described for DLG1 with different outcomes in DLG1 functions (Figure 2C).

DLG1 is important for cell regulation and it was shown to be phosphorylated in cells exposed to different stress signals. Under hyperosmotic stress eukaryotic cells may shrink and cells may develop responses to restore cell integrity, volume and homeostasis. For this, several signalling pathways are activated in order to reinforce and modulate the cytoskeletal network. Particularly, the p38 mitogen-activated protein kinases (MAPKs) are stimulated and specifically, SAPK3/p38gamma activation is rapid and strong compared to other p38s (Sabio et al., 2004). It was shown that DLG1 is a substrate of this kinase and this phosphorylation prompts the dissociation of DLG1 from the guanylate kinase-associated protein (GKAP). GKAP interacts with intermediate filaments and is able to associate with DLG1 and recruit it to the cytoskeleton. Phosphorylated DLG1 by osmotic shock may suffer conformational changes that prevent its interaction with GKAP and induce its release from the cytoskeletal fraction, probably for regulating intercellular complexes and polarity under changes in the extracellular environment (Sabio et al., 2005). In line with this study, Massimi and colleagues demonstrated that the osmotic shock induces the phosphorylation of DLG1 in parallel with its accumulation in the cells’ borders, at sites of cell-to-cell junctions, suggesting that the release of DLG1 from the cytoskeleton could contribute to the maintenance of cell contacts under stress conditions (Massimi et al., 2006). This idea was also supported by the fact, that under osmotic shock, DLG1 moves to the insoluble fraction of the protein cell extracts, most likely because of its participation in molecular complexes with membrane-associated proteins. In this study, the phosphorylation sites were mapped within the first 185 amino acids, and it was shown that the Jun N-terminal kinase (JNK) was responsible for DLG1 relocalization. Moreover, the redistribution of DLG1 to the cell junctions renders this protein more susceptible to proteasome degradation in the presence of HVP E6 oncoprotein; hence, phosphorylation may not only control DLG1 cell distribution but also its stability (Massimi et al., 2006).

It was shown that phosphorylation can also control particular DLG1 functions in specific cell types. As described before, in T cells DLG1 localizes at the IS and controls TCR signal transduction by interacting with cytoskeletal and signaling regulators through its several interacting protein domains (Xavier et al., 2004). In CD8+ T cells, DLG1 directs TCR stimulation to the activation of pro-inflammatory cytokine gene expression and cytotoxic granule release, processes necessary for destroying infected or transformed cells which require the activation of different signal networks. In these cells, DLG1AB, an alternative spliced isoform of DLG1, is tyrosine phosphorylated (Y222) by LCK kinase after TCR activation (Silva et al., 2015). This specific modification renders DLG1 able to modulate p38 activation and the subsequent cytokine expression, highlighting the importance of this posttranslational modification as a key control for the DLG1 activities involved in lymphocyte biology. Moreover, the blocking of DLG1AB phosphorylation was proposed as a novel therapeutic tool to specifically block proinflammatory cytokine production (Crocetti et al., 2014)

DLG1 is able to interact with some tumor suppressors, such as adenomatous polyposis coli and PTEN, and it was shown to regulate cell proliferation in many contexts. Hereafter, the regulation of DLG1 during the cell cycle should be important in order to strictly control cell growth.

The first report about DLG1 in association with kinases during cell replication showed that the PDZ binding kinase (PBK) interacts with DLG1 PDZ domains and this interaction should be important for the association of this kinase with the central spindle and the promotion of cytokinesis. Both PBK and DLG1 were shown to be phosphorylated at mitosis in HeLa cells, and the mitotic phosphorylation of PBK by cyclin B/cdc2 was required for its activity. These initial findings suggested a link for DLG1 to signal transduction pathways regulating the cell cycle or cellular proliferation (Gaudet et al., 2000). Afterwards, an analysis of DLG1 expression during the cell cycle identified that DLG1 localized at the cell junctions in G1, in the cytoplasm in S phase, relocating to the mitotic spindle in M phase and finally accumulating in the midbody during cytokinesis, demonstrating a cell cycle-dependent distribution of DLG1 that may impact on the protein activities. It was also shown that both cyclin dependent kinase 1 and 2 (CDK1 and CDK2) phosphorylate DLG1 on Ser158 and Ser442. These phosphorylation events modulate DLG1 nuclear localization, and particularly, protein stability, together with the redistribution of DLG1 to the mitotic spindle. These findings demonstrated the importance of cell cycle phosphorylation in the regulation of DLG1 localization and abundance during the different phases with potential consequences on its oncosuppressor functions (Narayan et al., 2009).

Other authors have also reported the phosphorylation of DLG1 at multiple sites during mitosis, like Ser102, Ser122 and Ser158. However, using specific phospho-antibodies the authors showed that, while DLG1 is accumulated at the midbody, phospo-DLG1 is absent from such structures, suggesting that phosphorylation may induce the dissociation of DLG1, likely by changing the protein binding pattern. Moreover, these authors also demonstrated that DLG1 phosphorylation in M phase is regulated by the ERK pathway which is involved in cell cycle control, potentially linking DLG1 to the regulation of entry and progression of mitosis (Inesta-Vaquera et al., 2010).

These data remark on the presence of several sites in DLG1 that could be phosphorylated by diverse kinases, either independently or simultaneously, depending on the activated signal pathways, cell type and extracellular microenvironment.

As described already, DLG1 seems to have a dual role in oncogenesis with both oncogenic or tumor suppressor activities, according to the biological context. It is likely to hypothesize that the phosphorylation status could be a way of controlling such activities as the activation of many kinases depends on the stimulation of signal translation pathways involved in the regulation of tumor development or prevention. In line with this, protein kinase C α isoform (PKCα) interacts with DLG1 in a PDZ-dependent manner and is able to phosphorylate DLG1 in the Thr-656 site. This phosphorylation correlates with increased invasiveness in non-small cell lung cancer lines and this event was considered a marker of PKC-mediated invasion, showing the requirement of DLG1 for PKCα to promote cellular migration. PKCα and DLG1 colocalize at the leading edge of the migrating cells and this depends on PKCα activity. These findings emphasize the involvement of DLG1 phosphorylation in both DLG1 cell distribution and functions, in this particular case, contributing to PKCα mediated tumor progression and metastasis (O’Neill et al., 2011).