Abstract
Normal neural circuits and functions depend on proper neuronal differentiation, migration, synaptic plasticity, and maintenance. Abnormalities in these processes underlie various neurodevelopmental, neuropsychiatric, and neurodegenerative disorders. Neural development and maintenance are regulated by many proteins. Among them are Par3, Par6 (partitioning defective 3 and 6), and aPKC (atypical protein kinase C) families of evolutionarily conserved polarity proteins. These proteins perform versatile functions by forming tripartite or other combinations of protein complexes, which hereafter are collectively referred to as “Par complexes.” In this review, we summarize the major findings on their biophysical and biochemical properties in cell polarization and signaling pathways. We next summarize their expression and localization in the nervous system as well as their versatile functions in various aspects of neurodevelopment, including neuroepithelial polarity, neurogenesis, neuronal migration, neurite differentiation, synaptic plasticity, and memory. These versatile functions rely on the fundamental roles of Par complexes in cell polarity in distinct cellular contexts. We also discuss how cell polarization may correlate with subcellular polarization in neurons. Finally, we review the involvement of Par complexes in neuropsychiatric and neurodegenerative disorders, such as schizophrenia and Alzheimer's disease. While emerging evidence indicates that Par complexes are essential for proper neural development and maintenance, many questions on their in vivo functions have yet to be answered. Thus, Par3, Par6, and aPKC continue to be important research topics to advance neuroscience.
Introduction
Normal neural functions rely on proper neurogenesis, neuronal differentiation and maintenance, and interneuronal wiring. Defects in these processes can lead to neuropsychiatric and neurodegenerative disorders, such as intellectual disability, autism spectrum disorders, bipolar disorder, schizophrenia, Alzheimer's disease (AD), and Parkinson's disease (Penzes et al., 2011; Batool et al., 2019). These disorders can be caused by a variety of environmental and genetic factors as well as by their complex interplay (Dick et al., 2010; Hannan, 2013; Banerjee et al., 2014; Takumi and Tamada, 2018; Willsey et al., 2018; Zarrei et al., 2019; Hollander et al., 2020). To date, a complete understanding of the etiologies of these disorders is lacking.
Proper neurodevelopment is regulated by many proteins, including Par3, Par6 (partitioning defective 3 and 6), and aPKC (atypical protein kinase C) polarity proteins (Hakanen et al., 2019). Par3, Par6, and aPKC proteins have many homologs in vertebrates. The names and chromosome location of these homologous genes in Caenorhabditis elegans, Drosophila, zebrafish, mice, and humans are listed in Table 1. Here, we use Par3, Par6, and aPKC to generically and collectively refer to the homologs identified in various species, but when necessary, the specific name of a protein/gene of a particular species is clarified.
Par3, Par6, and aPKC homologsa
Par3, Par6, and aPKC polarity proteins perform their biological functions by forming protein complexes in tripartite or other combinations (Kemphues, 2000; Ohno, 2001; Goldstein and Macara, 2007). Thus, we use a generic plural term, Par complexes, to refer to these protein complexes collectively. To better understand the roles of Par complexes in both normal neurodevelopment and pathologic conditions, we here review the findings on Par complexes in the following three aspects: (1) The basic biophysical binding and biochemical modification properties of Par3, Par6, and aPKC proteins in a variety of cell types and species. Although some of these properties were discovered in non-neuronal cell types, they underlie the basic biology of Par complexes and are likely conserved to a certain extent in neurons. Thus, such information forms the basis to understand their regulation and functions in neurodevelopment. (2) The expression and localization of Par complexes in the nervous system and their roles in neuroepithelial polarity, neurogenesis, neuronal migration, neurite differentiation, synaptic plasticity, and memory. (3) The evidence that human PARD3, PARD3β, PARD6, PARD6β, PARD6γ, and PRKCι, PRKCζ, and PKMζ are involved in neuropsychiatric and neurodegenerative disorders. On each aspect, we also highlight a few outstanding questions and challenges.
In sum, accumulating evidence shows that, at the nexus of various signaling pathways, Par complexes play essential roles in neurodevelopment, and that malfunction of Par complexes is associated with neurodevelopmental, neuropsychiatric, and neurodegenerative disorders. With many important questions yet to be answered, Par3, Par6, and aPKC are exciting subjects for neuroscience research.
Biophysical and Biochemical Properties of Par Complexes
It was in C. elegans that Par-3, Par-6, and PKC-3 were first discovered to form complexes and to play important roles in cell polarization by partitioning cytoplasmic and membrane-bound components to distinct subcellular regions (Kemphues et al., 1988; Watts et al., 1996; Tabuse et al., 1998; Kemphues, 2000). In insects and vertebrates, Par3, Par6, and aPKC are conserved, but with multiple homologs and isoforms (Knust and Bossinger, 2002; Goldstein and Macara, 2007). In this section, we emphasize vertebrate homologs when we summarize the findings on the physical binding and biochemical interactions among Par3, Par6, and aPKC proteins (Table 2).
The binding partners of Par3, Par6, and aPKC homologsa
The physical binding and biochemical interactions among Par3, Par6, and aPKC for complex formation
Two Par3 homologs exist in mammals (Pard3 and Pard3β in mice, and PARD3 and PARD3β in humans) and four in zebrafish (Pard3αa, Pard3αb, Pard3βa, and Pard3βb) (Table 1). Par3 has the following domains aligned from the N terminus to the C terminus: one PB1 (Phox and Bem1) domain, three PDZ (postsynaptic density 95, discs large, and zonula occludins-1) domains (PDZ1, PDZ2, and PDZ3), one aPKC-binding region, and one coiled-coil region (Fig. 1). The PB1 domain can mediate oligomerization to form helical filaments via front-to-back electrostatic interactions (Feng et al., 2007; Zhang et al., 2013); this oligomerization plays a role in localizing Par complexes to the tight junctions in the epithelium (Mizuno et al., 2003). PDZ1 domain can bind to the PDZ domain of Par6 (Lin et al., 2000). The aPKC-binding region can bind to aPKC's kinase domain (Izumi et al., 1998). The binding of Par3 to aPKC can inhibit aPKC's kinase activity, thus simultaneously regulating the anchoring and activity of aPKC at specific subcellular regions (Lin et al., 2000; Soriano et al., 2016). In turn, aPKC can phosphorylate Par3's aPKC binding region to destabilize the Par3–aPKC binding and make aPKC available to phosphorylate other substrates, creating a negative regulatory loop (Nagai-Tamai et al., 2002). Unlike Pard3, Pard3β, which is not prominently expressed in neurons (Huo and Macara, 2014), does not bind to aPKC, but Pard3β can still localize to the tight junctions in MDCK cells, suggesting that the apical localization of Pard3β is independent of aPKC. In addition, some Pard3β isoforms bind to Par6 with varying affinities (Gao et al., 2002; Kohjima et al., 2002). In sum, with multiple protein–protein binding domains, Par3 plays an important scaffolding role in Par complexes.
Domain structures of Par3, Par6, and aPKC. Left, N terminals of the proteins. Right, C terminals. Double-headed arrows indicate the binding interactions between domains. PB1, Phox and Bem1 domain; PDZ, PSD95-Dlg1-ZO1 domain; aPKC BR, aPKC binding domain; IQ, IQ calmodulin-binding motif; CRIB, Cdc42/Rac interactive binding; PS, pseudosubstrate peptide; C1, cysteine-rich zinc-finger domain; kinase, kinase domain; coiled-coil, coiled-coil domain.
Vertebrates have multiple Par6 homologs (Pard6α Pard6β, and Pard6γ in mice; PARD6α, PARD6β, and PARD6γ in humans; and Pard6α, Pard6β, and Pard6γa, Pard6γb in zebrafish) (Table 1). Par6 homologs contain the following domains that are aligned from the N terminus to the C terminus: a PB1 domain, an IQ calmodulin-binding motif, a CRIB-like (Cdc42/Rac interactive binding) motif, and a PDZ domain (Fig. 1). In addition to the interaction with Par3 mentioned above, the PB1 domain of Par6 binds to the PB1 domain of aPKC (Joberty et al., 2000; Noda et al., 2001; Hirano et al., 2005).
Two aPKC genes exist in vertebrates (prkcι/λ, and prkcζ) (Table 1). The prkcι/λ genes (originally named aPKCι for the human homolog and aPKCλ for the mouse homolog) encode predominantly one protein product, generically referred to as aPKCλ in this review, whereas prkcζ encodes two protein products by using alternative transcription start sites: the full-length aPKCζ and the constitutively active C-terminal form, PKMζ, which lacks an N-terminal regulatory region (Hernandez et al., 2003) (Fig. 1). PKMζ can also be generated from proteolytic cleavage of aPKCζ, but this mechanism is mainly used in invertebrates (Frutos et al., 1999; Smith et al., 2000; Bougie et al., 2012). Full-length aPKCλ and aPKCζ contain the following domains, aligned from the N termini to the C termini: a PB1 domain, a pseudosubstrate sequence motif, a lipid-binding cysteine-rich zinc-finger domain (C1), and a kinase domain (Suzuki et al., 2003) (Fig. 1). The PB1, pseudosubstrate sequence, and C1 motif/domains constitute the N-terminal regulatory region of aPKC. The pseudosubstrate and C1 domain bind to aPKC's kinase domain to inhibit its activity (Lopez-Garcia et al., 2011). aPKCλ and aPKCζ belong to the atypical subfamily of the PKC family of serine/threonine kinases, which also encompass the conventional (cPKC) and novel (nPKC) subfamilies (Nishizuka, 1988, 1995; Steinberg, 2008). For membrane targeting and activation, cPKC and nPKC require phorbol 12-myristate 13-acetate/diacylglycerol, and cPKC's membrane targeting also requires anionic phospholipids in a calcium-dependent manner (Ohno and Nishizuka, 2002); by contrast, aPKC require phosphatidylinositol (PI)−3,4,5-trisphosphate (PIP3), which displaces the pseudosubstrate sequence and C1 to release their inhibition of the kinase domain and promotes the association with Par3 (Ono et al., 1989; Nakanishi et al., 1993; Hirai and Chida, 2003; Ivey et al., 2014). This displacement of the pseudosubstrate motif can also be achieved by binding to Par6 via their PB1 domains (Suzuki et al., 2003; Graybill et al., 2012). Because of the lack of the N-terminal regulatory region-mediated intramolecular autoinhibition, PKMζ is believed to be constitutively active (Hernandez et al., 2003).
The formation of the Par complexes via physical binding partially explains their mutual dependence for subcellular localization in some cellular contexts. However, Par3, Par6, and aPKC do not always coexist in a tripartite complex; their physical binding interactions are not static but rather fluid, and the involved domains can recruit different proteins to suit various cellular contexts. For example, Par3 can perform some functions in the absence of Par6 and aPKC (Chen and Macara, 2005; Harris and Peifer, 2005; Zhang and Macara, 2006); Par3 binds to dynein light intermediate chain 2 in the absence of Par6 and aPKC (Schmoranzer et al., 2009); and aPKC and Par6 can complex with Lgl independent of Par3 to regulate epithelial polarity (Plant et al., 2003; Yamanaka et al., 2003; Yamanaka et al., 2006). This flexibility in complex formation greatly expands their functional versatility.
Another factor that contributes to the versatility in the complex formation and functionality of Par3, Par6, and aPKC is the multiplicity of their homologs and isoforms, particularly in vertebrates. For example, vertebrate Pard3 is expressed in at least 20 isoforms through alternative splicing and transcription start sites, with 100, 150, and 180 kDa being the major isoforms in mammals (Lin et al., 2000; Macara et al., 2009) and 150 and 180 kDa as the major isoforms in zebrafish (Wei et al., 2004). Like Pard3, Pard3β is also expressed as various splicing variants (Fang and Xu, 2001; Gao et al., 2002). These isoforms have different domain compositions, implying the formation of different complexes for different functions. For example, the 100 kDa Pard3 isoform cannot directly associate with aPKC because it lacks the aPKC-binding domain (Lin et al., 2000).
Of note, the current understanding of Par complex formation and functions has been largely based on limited homologs or isoforms. Because many homologs and isoforms are expressed in tissue-specific manners, the binding interactions among Par3, Par6, and aPKC can be very complex and may vary drastically in different cellular contexts. When more homologs and isoforms are investigated, new biochemical and biophysical interactions and functions are expected. Because of the complexity of the interactions among Par3, Par6, and aPKC, it is necessary to use compatible homologs to study their protein–protein interactions. For example, one might need to be cautious to generalize a protein–protein interaction detected between a mouse protein and a human protein in vitro because such an interaction never exists in nature.
Interactions with other molecules in various signaling pathways
Par3, Par6, and aPKC not only physically bind to each other to form Par complexes, but they also interact with many other proteins or lipids through physical binding or biochemical modifications. In this section, we review some prominent upstream and downstream interactions in various signaling pathways as examples (Table 2), which shall not be treated as a comprehensive list.
Interactions with Rho GTPases
Par6's CRIB-like motif together with its adjacent PDZ domain can bind to many small GTPases, such as Cdc42 (cell division control protein 42 homolog), Rac1, RhoA, and TC10 (Joberty et al., 2000; Lin et al., 2000; Noda et al., 2001; Pichaud et al., 2019). Cdc42, Rac1, RhoA, and TC10 belong to the Rho GTPase family, one of the five families of the Ras GTPase superfamily (Adams et al., 1990; Johnson and Pringle, 1990; Ridley and Hall, 1992; Nobes and Hall, 1995; Etienne-Manneville, 2004; Rojas et al., 2012). Rho GTPases act as small molecular switches by converting between an active GTP-bound state and an inactive GDP-bound state (Hall, 2012). This conversion is regulated by guanine nucleotide exchange factors (GEFs), which activate GTPases by promoting the exchange of GDP for GTP, by GTPase-activating proteins (GAPs), which inactivate GTPases by activating their intrinsic GTPase activity, and by guanine nucleotide dissociation inhibitors, which keep these GTPases inactive by maintaining them in the GDP-bound state (Diekmann et al., 1991; Hart et al., 1991; Hoffman et al., 2000; Lin et al., 2003; Hall, 2012). By interacting with their effectors, GTPases regulate many cellular processes (Harris and Tepass, 2008; Pichaud et al., 2019).
The binding of GTP-bound Cdc42 to Par6 enhances the affinity of Par6's PDZ domain to its ligands, such as Par3's PDZ1 domain, through allosteric regulation (Garrard et al., 2003; Peterson et al., 2004; Whitney et al., 2011). In addition, GTP-bound Cdc42 causes conformational changes in Par6 to activate aPKC (Yamanaka et al., 2001; Garrard et al., 2003). Because Cdc42 distributes in a polarized fashion, Cdc42 can also facilitate the apical localization and activation of Par complexes (Etienne-Manneville and Hall, 2001). Par complexes are not just effectors of Rho GTPases. Pard3 can also positively regulate Cdc42 and Rac1 by activating two Rac GEFs, Tiam1 (T-lymphoma invasion and metastasis 1) and STEF (also called Tiam2), thus establishing a positive feedback loop (Chen and Macara, 2005; Nishimura et al., 2005; Zhang and Macara, 2006). This partially explains the observation that the loss of Cdc42, Par-3, Par-6, and PKC-3 causes similar polarity defects in diverse C. elegans cell types (Welchman et al., 2007). Furthermore, Par6 can also inhibit RhoA GTPase in dendritic spine formation in cultured neurons by activating p190 RhoGAP (Zhang and Macara, 2008). In addition, the RhoA effector Rho-kinase can phosphorylate Pard3 to disrupt Par complex formation in cell migration (Nakayama et al., 2008). Thus, Rho GTPase signaling can act both upstream and downstream of Par complexes.
Regulation by phosphorylation states
In a variety of cellular contexts, the activities of Par complexes are regulated by many kinases and phosphatases. For example, kinases Ndr1 and Ndr2, whose activity is stimulated by Rassf5, can phosphorylate mouse Pard3 at Ser383 to inhibit its interaction with retrograde dynein and thus to restrict Pard3 to axon tips for axonal specification (Yang et al., 2014). Human PARD3 can bind to, and be phosphorylated by, mitotic kinase Aurora-A at Ser962 to regulate axon specification (Khazaei and Puschel, 2009). Rat Pard3 can also be phosphorylated by Extracellular Signal-Regulated Kinase 2 (ERK2) at Ser1116 of its C-terminal coiled-coil domain, to which ERK2 binds; this phosphorylation promotes Pard3's unloading at the growth cone during axonal transportation by inhibiting Pard3's binding to KIF3A (kinesin-2) (Funahashi et al., 2013). In addition, Aurora also phosphorylates Par-6 in Drosophila neuroblasts for their polarized localization (Wirtz-Peitz et al., 2008). Par complexes can also be regulated via dephosphorylation. For example, mouse Pard3 can be dephosphorylated by serine/threonine phosphatase PP4 via interacting with PP4's regulator Smek1 through Pard3's coiled-coil domain; this modification removes Pard3's suppression on the differentiation of neural stem cells (Lyu et al., 2013).
Regulation by planar cell polarity (PCP)
Par3, Par6, and aPKC have traditionally been considered as proteins important for epithelial apicobasal polarity (Knust and Bossinger, 2002), which is defined as the polarization of cells along the axis from their apical end, which faces the luminal or external surfaces, to the basal end, which attaches to the basement membrane. Par complexes establish and maintain apicobasal polarity by regulating apical junctional complexes, such as adherens junctions and tight junctions. Interestingly, recent studies suggest that they are also linked to PCP (cell polarization within the two-dimensional plane of tissues, orthogonal to the apicobasal axis of individual cells). In cultured hippocampal neurons, aPKCζ can be activated to promote axon differentiation by binding to PCP component Dishevelled (Dvl), which is activated by noncanonical Wnt signaling ligand Wnt5a and accumulated at the tip of the growing axon but not that of other neurites (Zhang et al., 2007). This Dvl-mediated aPKCζ activation for axon differentiation can be inhibited by nucleoporin Nup358 via the physical binding of Nup358's N terminus to the kinase domain of aPKCζ and the PDZ domain of Dvl (Vyas et al., 2013). Also, Pard3 itself can localize in a planar polarized fashion in mouse inner ear sensory epithelia and regulates the planar orientation of stereociliary hair bundles by regulating the core PCP pathway (Landin Malt et al., 2019). Because apicobasal polarity and PCP together regulate the three-dimensional organization of tissues, Par complexes play a pivotal role in tissue morphogenesis by linking apicobasal and planar cell polarities.
Regulation by phosphoinositide (PIP) signaling
PIP signaling regulates a vast number of cellular processes (Di Paolo and De Camilli, 2006; Fruman et al., 2017), including the apical localization and activation of Par complexes. Drosophila Bazooka can associate with the cell membrane by directly binding to PIPs using its C terminal region (Krahn et al., 2010; McKinley et al., 2012); in addition, PDZ2 domain of rat Pard3 binds to PIPs (with assistance from nonspecific electrostatic interactions and the insertion of hydrophobic side chains into lipid membrane) (Wu et al., 2007). Also, the PDZ1-3 region of mouse Pard3 can associate with the cell membrane by binding to phosphatidic acid (Yu and Harris, 2012). In turn, the binding of Bazooka or Pard3 to PIPs indirectly modulates the makeup of PIPs because Par3's PDZ3 can recruit protein and lipid phosphatase PTEN2, which converts PI(3,4,5)P3 to PI(4,5)P2 at the apical cell–cell junctional regions (von Stein et al., 2005; Pinal et al., 2006; Wu et al., 2007; Feng et al., 2008). This localized PTEN-mediated dephosphorylation, together with phosphorylation by PI4P5 kinase Skittles (SKTL) and PI(4,5)P3 kinase DP110 to produce PI(4,5)P2 and PI(3,4,5)P3, respectively (Claret et al., 2014), regulates the regional balance between PI(3,4,5)P3 and PI(4,5)P2. A proper level of PI(3,4,5)P3 modulates PDK1 (PI(3,4,5)P3-dependent protein kinase 1), which activates rat aPKCζ by localizing to the cell membrane through the binding of its pleckstrin homology domain to PI(3,4,5)P3 (Chou et al., 1998; Le Good et al., 1998; Dainichi et al., 2016). Incidentally, PI(3,4,5)P3 can recruit GEFs via their pleckstrin homology domains to activate Cdc42 and Rac1, which in turn activate aPKC by binding to Par6 (Zheng, 2001; Di Paolo and De Camilli, 2006). Thus, the phosphorylation states of PIPs modulate the localization and activities of Par complexes, making Par complexes integration sites of PIP signaling.
Regulation by TGFβ signaling
Human PARD6γ can bind to Type I TGFβ receptor, TβRI, via its PB1 domain. Upon stimulation by TGFβ, Type II TGFβ receptor is recruited to the Par6-TβRI complex to phosphorylate Par6. During TGFβ-mediated epithelial-to-mesenchymal transition, this phosphorylation promotes PARD6γ's binding to E3 ubiquitin ligase Smurf1, which mediates ubiquitination of RhoA at tight junctions and consequently disrupts tight junction (Ozdamar et al., 2005). Epithelial-to-mesenchymal transition is based on drastic cell shape changes, which also occur during tissue morphogenesis, such as neurulation. Indeed, TGFβ regulates neural tube closure by interacting with Par complexes in the neuroepithelium to modulate cell shape changes (Amarnath and Agarwala, 2017). These findings suggest that, in a variety of cellular contexts, Par complexes are regulated by TGFβ signaling, which has been implicated in normal neurodevelopment and neuropsychiatric disorders (Galvez-Contreras et al., 2016; Hiew et al., 2021; Mitra et al., 2022).
Regulation of basolateral Lgl-Dlg-Scrib complex
As apicobasal polarity proteins, one way Par complexes contribute to epithelial apicobasal polarity is by directly limiting basolateral proteins from localizing at the apical side. Mouse Pard6γ-aPKCλ-Cdc42 complex can bind to and phosphorylate basolateral protein Lethal giant larvae (Lgl), and the binding occurs between Pard6γ's PDZ domain and Lgl's N terminus (Plant et al., 2003). These interactions were also verified in human cells (Yamanaka et al., 2003). This binding brings Lgl into close contact with aPKCλ for phosphorylation at the apical side of the cell, thus restricting phosphorylated Lgl to the basolateral region (Betschinger et al., 2003), where it negatively regulates the apical membrane domain (Hutterer et al., 2004; Yamanaka et al., 2006). As a result of its basolateral localization, Lgl also restricts cell fate determinant Miranda to the basal side of Drosophila neuroblasts (Betschinger et al., 2003), thus playing a critical role in asymmetric cell division. In this regulatory process, mitotic kinase Aurora-A phosphorylates and activates DmPar-6, which then activates aPKC to phosphorylate Lgl to release Lgl's prevention of Bazooka from binding to the DmPar-6-aPKC complex; this alters the substrate specificity of aPKC and consequently phosphorylates cell fate determinant Numb to promote its basal localization in asymmetric cell division (Wirtz-Peitz et al., 2008). Lgl is only one example of a vast number of polarity proteins that can be phosphorated by aPKC. For more information on aPKC-mediated protein phosphorylation, interested readers are referred to a more detailed review (Hong, 2018).
Interaction with Crumbs (Crb) protein complex
Par complexes also interact with the evolutionarily conserved Crb apical protein complex (Crb, Pals1, Lin7c, and PATJ). In the neuroepithelium and many other epithelia, Par3, aPKC, and Par6 colocalize with Crb complex in a narrow region apical to the tight junctions, as we and others have revealed (Wei and Malicki, 2002; Sotillos et al., 2004; Wei et al., 2006; Yang et al., 2009; Zou et al., 2013; Guo et al., 2018; Tan et al., 2020). Mammalian Pard6β's PDZ domain can bind to the N terminus of Pals1 (proteins associated with Lin7) (Hurd et al., 2003; Wang et al., 2004), which is the mammalian homolog of zebrafish Nok and fly Stardust (Hong et al., 2001; Wei and Malicki, 2002). The Pard6β-Pals1 interaction competes with Pals1-PATJ (PALS1-associated tight junction protein) interaction (Wang et al., 2004). In addition, DmPar-6's N terminus can also bind to the PDZ3 domain of Dlt (Discs lost), the fly homolog of mammalian PATJ, to regulate the localization of Crb complex for Drosophila embryonic epithelial polarity and photoreceptor morphogenesis, respectively (Nam and Choi, 2003; Hutterer et al., 2004). DmPar-6 can also bind to the C-terminus of Crb via its PDZ domain, thus mutually reinforcing each other's apical accumulation (Nunes de Almeida et al., 2019). Similar binding also exists between human CRB3 and PARD6α (Lemmers et al., 2004). Drosophila aPKC can phosphorylate Crb and directly bind to Crb and PATJ for apical localization (Sotillos et al., 2004). These physical binding interactions between Par and Crb complexes suggest that these apical polarity proteins may form super complexes and collectively regulate cell polarization.
Interactions with Hippo signaling
Par complexes also interact with Hippo signaling, which mediates diverse cellular processes, such as cell proliferation, differentiation, and survival, by regulating the nuclear localization of the Yki/YAP/TAZ transcription cofactors and consequently gene expression (Zheng and Pan, 2019; Schroeder and Halder, 2012; Enderle and McNeill, 2013; Yu and Guan, 2013). Hippo signaling is based on a kinase cascade, in which the upstream Hpo/Mst kinases phosphorylate and activate Wts/Lats kinases, which then phosphorylate Yki/YAP/TAZ to inhibit their nuclear localization and activity. Proper expression and activation of aPKC are important for Hippo signaling. For example, constitutively active aPKCζ can transform MDCK cells and murine mammary epithelial cells into tumor-like cells by interacting and blocking the membrane localization of Mst1/2, thus suppressing Hippo signaling and increasing nuclear localization and activity of unphosphorylated YAP1 (Archibald et al., 2015). In addition, aPKCζ can phosphorylate and inhibit Kibra, an upstream activator of Mst/Hippo (Buther et al., 2004; Yoshihama et al., 2009). On the other hand, Kibra can also inhibit aPKCζ by directly binding to its kinase domain (Yoshihama et al., 2011). Thus, Par complexes and Hippo signaling mutually inhibit each other.
Inhibition of glycogen synthase kinase 3β (GSK3β)
GSK3β regulates many neurodevelopmental processes, such as neurogenesis, neuronal migration, neuronal polarization, and axon growth and guidance (for more information on GSK3β, readers are referred to a more detailed review, Hur and Zhou, 2010). GSK3β activity can be inhibited by Par complexes through two indirect mechanisms. First, rat Pard3 binds to and activates PI3K (Itoh et al., 2010), which phosphorylates and activates PKB/Akt to phosphorylate and inactivate GSK3β (Cross et al., 1995; Frame and Cohen, 2001). Second, mouse or rat aPKCζ can recruit Dvl, which binds to and locally inhibits GSK3β's function by disrupting the complex of GSK3β, axin, adenomatous polyposis coli, and β-catenin (Etienne-Manneville and Hall, 2003; Schlessinger et al., 2007). Interestingly, localization of mammalian Pard3 at axon tips also requires local inhibition of GSK3β (Shi et al., 2004), suggesting a feedback regulation between Par3 and GSK3β. Thus, Par complexes can regulate neurodevelopment via GSK3β signaling.
Regulation of phospholipase C
When cotransfected in HEK293 cells, human PARD3 and PARD6 were found to bind to G-protein-activated phospholipase C-β (PLC-β) via PARD3's PDZ1 and PARD6's PDZ domains; this association can stimulate PLC-β to hydrolyze PI(4,5)P2 into second messengers diacylglycerol and inositol 1,4,5-triphosphate (Cai et al., 2005). Given that PLC-β signaling plays important roles in neurodevelopment and that PLC-β is involved in neuropsychiatric and neurodegenerative disorders (Cockcroft, 2006; Kang et al., 2016; Yang et al., 2016), it would be interesting to further evaluate how Par complexes regulate neurodevelopment by interacting with PLC-β signaling in neurons in vivo.
In sum, by interacting with many other proteins and lipids through physical binding and biochemical modifications, Par complexes integrate various signaling pathways.
Par Complexes in Neurodevelopment
Par complexes are essential for the development of neurons and glial cells from undifferentiated neuroepithelial cells. The current understanding of the roles of Par complexes in the nervous system has been largely limited to early development by studying embryonic development and neurons cultured in vitro. However, the functions of Par complexes in the adult nervous system are much less understood because embryonic lethality caused by systemic loss-of-function approaches has made it difficult to study their roles in the adult nervous system in vivo. For example, the loss of zebrafish Pard3αb (Wei et al., 2004; Guo et al., 2018), Pard6γb (Munson et al., 2008; Guo et al., 2018), and aPKCλ (Horne-Badovinac et al., 2001) usually resulted in embryonic lethality. One exception is the zebrafish pard3αbfh305 mutation, which was supposed to be a null nonsense mutation at the N-terminus. However, the protein expression level in zebrafish pard3αbfh305 was not examined to rule out the possibility of an alternative translation or transcription that bypasses the pard3αbfh305 mutation and produces shorter and yet functional products (Blasky et al., 2014). Loss of Pard3 in mice also resulted in embryonic lethality (Hirose et al., 2006). In this section, we summarize major findings regarding the functions of Par complexes in neuroepithelial polarization, neurogenesis, neuronal migration, neurite development, synaptic plasticity, memory, and myelination. We also discuss how cell polarization in various cellular contexts can be correlated.
Par complexes in neuroepithelial polarization and integrity
The CNS develops from the neural tube, which is made of a single layer of pseudostratified neuroepithelial cells. These cells contact both apical luminal surface and basal basement membrane, displaying apicobasal and planar cell polarities. Although neuroepithelium is much thicker than simple epithelia, these epithelia are similar in the overall organization of their apical adhesion complexes, with Crb-mediated adhesions, tight junctions, and adherens junctions aligned sequentially in the apical-to-basal direction (Zou et al., 2012; Guo et al., 2018) (Fig. 2A). This similarity suggests that conserved apical adhesion mechanisms are used to build various monolayer epithelia. Par complexes establish and maintain epithelial polarity and integrity by stabilizing and organizing apical adhesions through physical and biochemical interactions with many proteins in adhesion complexes, for example, Pard6β with Pals1 (Hurd et al., 2003; Wang et al., 2004), Pard3 with tight junction-associated protein JAM (Ebnet et al., 2001), and Pard3 with nectins (Takekuni et al., 2003).
The functions of Par complexes and various manifestations of cell polarity during neurodevelopment. A, Par complexes are essential for establishing and maintaining neuroepithelial polarity and integrity by regulating three apical adhesion mechanisms: Crb-mediated adhesion (Crb), tight junction (TJ), and adherens junction (AJ). B, Par complexes regulate both symmetric proliferative and asymmetric differentiative divisions in neurogenesis by controlling the biased distribution of cell fate determinants and the orientation and position of the mitotic spindle. C, Localizing to two opposite ends of neurons, Par complexes play important roles in axon specification, axon growth, and pathfinding, and dendritic spine morphogenesis. D, During synaptic plasticity, Par complexes recruit synaptic molecules and scaffold synaptic subcellular structures at presynaptic and/or postsynaptic terminals. Par complexes are also expected to regulate synaptogenesis. The presence and localization of Par complexes at synapses may vary depending on cell types. Red represents the apical ends of cells. Blue represents the basal ends of cells. Green represents the localizations of Par complexes.
In the neuroepithelium, the essential roles of Par complexes in maintaining epithelial polarity and integrity have been demonstrated by many in vivo studies. For example, the loss of Pard3 and Pard6γb caused abnormal junctional association between the left and right neuroepithelial cells during zebrafish neurulation, preventing the proper formation of the neural tube lumen (Wei et al., 2004; Munson et al., 2008; Guo et al., 2018), and the loss of aPKCλ disrupted neuroepithelial polarity of the retina and brain in zebrafish and mice (Horne-Badovinac et al., 2001; Imai et al., 2006). During neuroepithelial development, Par complex components are expressed at proper levels in a strict spatiotemporal fashion (Yang et al., 2009; Guo et al., 2018). Overexpression of Pard3, for example, disturbed the proper localization of other polarity proteins in the chicken neuroepithelium, causing neuroepithelial cells to detach from the apical surface and form rosettes in the basal regions (Afonso and Henrique, 2006). In short, Par complexes, along with other polarity proteins, establish and maintain the proper apicobasal polarity and integrity of the neuroepithelium from which the sophisticated CNS develops.
Par complexes in neurogenesis
Neuroepithelial cells and the later radial glial cells are neural stem cells. These cells divide either symmetrically, parallel to the plane of the neuroepithelium, or asymmetrically, along the apicobasal axis. Symmetric divisions expand the progenitor population and are thus called proliferative divisions, whereas asymmetric divisions produce an apical daughter cell that retains neural stem cell properties and a basal daughter cell that differentiates into a neuron and are thus called differentiative divisions (Fig. 2B). Whether a neuroepithelial cell divides symmetrically or asymmetrically depends on the orientation and position of the mitotic spindle, which determines not only the orientation and position of cytokinesis but also the distribution of fate determinants relative to the plane of cytokinesis. These processes are all regulated by Par complexes.
As fundamental polarity proteins, Par complexes play an important role in distributing cell fate determinants in a polarized fashion. For example, in Drosophila neuroblasts, aPKC-mediated phosphorylation prevents cell-fate determinants (the adaptor protein Miranda, the Notch signaling repressor Numb, and the transcription factor Prospero) from localizing apically (Rolls et al., 2003; Prehoda, 2009). Par complexes also regulate spindle orientation by apically localizing the NuMA-LGN-Gαi complex, which orients the mitotic spindle by anchoring and pulling astral microtubules of the spindle to the cell cortex via dynein (Kiyomitsu and Boerner, 2021; Lechler and Mapelli, 2021). The apical localization of NuMA-LGN-Gαi requires LGN binding to the adaptor protein Inscuteable, whose apical localization, in turn, depends on Par3 as revealed in various systems (Kuchinke et al., 1998; Wodarz et al., 1999; Schaefer et al., 2000; Yu et al., 2000; Petronczki and Knoblich, 2001; Du and Macara, 2004). aPKC can also negatively regulate the NuMA-LGN-Gαi complex. For example, in the first cell division of C. elegans embryonic development, anterior polarized PKC-3 phosphorylates NuMA homolog Lin-5 and attenuates cortical pulling force in the anterior region, leading to the posterior positioning of the spindle (Galli et al., 2011). In addition, in 3D MDCK cell cultures, aPKCλ can phosphorylate LGN to release it from the cell cortex by disrupting its attachment to membrane-associating GαI, thus promoting symmetric cell division for proper lumen formation (Hao et al., 2010).
During vertebrate neurogenesis, Par complexes are involved in both proliferative and differentiative cell divisions. In proliferative divisions during early cortical neurogenesis in mice, the protein levels of Pard3, Pard6, and aPKCλ at the apical surface are higher than they are during late cortical neurogenesis (Costa et al., 2008). Furthermore, when Pard3 was suppressed at E12 or E13 by in utero injection of anti-Pard3 shRNA expression constructs, smaller clone sizes of brain cells were induced, suggesting that high levels of Par complexes promote the expansion of the neural tissue through proliferative cell divisions (Costa et al., 2008; Bultje et al., 2009). However, conditional KO of Pard3 in developing mouse cortex at E9.5 resulted in overproduction of radial glial progenitor cells, causing megalencephaly with ribbon heterotopia and epilepsy, suggesting that Pard3 regulates neurogenesis in sophisticated spatiotemporal manners (Liu et al., 2018). In asymmetric divisions of mouse neuroepithelial cells, Par complexes remain apically localized, suggesting a conserved role in the asymmetric inheritance of cell-fate determinants (Manabe et al., 2002). This notion was supported by studies of asymmetric division of mouse radial glial progenitor cells, in which one daughter cell was enriched with Pard3 and Notch and took a radial glial progenitor cell fate and another daughter cell inherited less Pard3 and Notch and took a neuronal fate (Bultje et al., 2009).
While Par complexes are involved in both modes of cell divisions in the vertebrate neuroepithelium, the molecular mechanisms that underlie the spatiotemporal regulation of these two division modes are not well understood. Nevertheless, emerging evidence supports that Par complexes regulate vertebrate neurogenesis via their evolutionarily conserved roles in polarizing the distribution of fate determinants and in controlling spindle positioning and orientation.
Par complexes in neuronal migration and cellular pattern formation
Neurons are often born far from their final destinations and need to migrate and establish neural structures and circuits at a distant location. Neuronal migration involves the extension of the leading process, forward translocation of the centrosome and the nucleus, and retraction of the trailing end; these processes require dynamic regulation of cytoskeletons and unfold while the neuron is still attached to supporting substrates, such as a radial glial cell, suggesting the need of intercellular interactions for proper neuronal migration (Feng and Walsh, 2001; Nadarajah et al., 2001; Ayala et al., 2007; Minegishi and Inagaki, 2020). In glial-guided migration of purified cerebellar granule neurons in culture, Pard6α and aPKCζ are enriched in the centrosome and underlie the integrity and movement of the centrosome; Pard6α and aPKCζ also participate in the assembly of the perinuclear microtubule cage that encompasses the cell nucleus for translocation (Solecki et al., 2004). The forward migration of the centrosome and the soma requires acto-myosin-mediated contractility in the leading process, and this contractility is activated by the phosphorylation of the myosin light chain in a Pard6a-dependent manner (Solecki et al., 2004, 2009). In addition, Pard3 recruits tight junction adhesion molecule C to the surface contacts between migrating cerebellar granule neurons and neighboring neurons or glia for proper radial migration (Ebnet et al., 2001; Famulski et al., 2010). Furthermore, Pard6γb, aPKCλ, and aPKCζ regulate the proper migration of facial branchiomotor neurons in a cell nonautonomous fashion in zebrafish, suggesting that neuronal migration mediated by intracellular Par complexes is dependent on intercellular interactions (Grant and Moens, 2010), attesting to the importance of studying the functions of Par complexes in vivo. Because of these roles in migration, the disruption of Par complex-mediated neuronal migration resulting from Pard3 deficiency is expected to contribute to abnormal tissue architectures, such as cortical heterotopia (Liu et al., 2018) and retinal lamination defects (Wei et al., 2004).
Mechanisms of the subcellular localization of Par complexes in neurons
Unlike in neuroepithelial cells, where Par complexes localize apical to the adherens junctions and tight junctions (Guo et al., 2018), the localization of Par complexes in differentiated neurons is quite complicated because of the extensive neuronal arborization (Fig. 2C). In this section, we review findings regarding the mechanisms of expression and localization of Par complex components at tissue/cell and subcellular levels in neurons. This provides background information for understanding the functions of Par complexes in neurons, reviewed in later sections.
Homologs of Par3, Par6, and aPKC are differentially transcribed across various tissue and cell types in vertebrates. Northern blotting has revealed that human PARD6α is expressed in the nervous system, muscles, and kidneys, whereas PARD6β and PARD6γ are strongly expressed in the kidney but weakly in other tissues (Noda et al., 2001). However, in zebrafish, Pard6γb is expressed in the neural tube and is required for neurulation (Munson et al., 2008), suggesting possible differences in tissue specificity among species. Similarly, the human PARD3β homolog (also known as Par3-L) is expressed strongly in the kidney but weakly in the other tissues, including the nervous system; this differs from the broader expression of PARD3 (Gao et al., 2002). As for aPKC, Western blotting and ISH revealed that aPKCλ and PKMζ were expressed in many regions of the rat and mouse brain, including the neocortex, cerebellum, and CA1 regions; in contrast, the full-length aPKCζ was not detectable in the forebrain by Western blotting, although it was found in the lateral olfactory tract by ISH, and was also expressed in the kidney and the hindbrain, including the cerebellum (Hernandez et al., 2003; Oster et al., 2004). These different expression patterns suggest that the nervous system and even different regions of the nervous system require distinct isoforms or homologs of Par complex components for their normal functions.
The subcellular localization of Par3, Par6, and aPKC is achieved via several mechanisms, including translocation, local translation, and physical binding to other molecules (Table 2). Pard3 can be transported to the axon tip anterogradely along microtubules by binding to the plus-end-directed kinesin KIF3A using its C-terminal coiled-coil domain (Nishimura et al., 2004). This Pard3/KIF3A complex is also associated with the plus-end binding protein adenomatous polyposis coli, which further facilitates Pard3's enrichment at the axon tip (Shi et al., 2004). The complex formation with KIF3A is inhibited by phosphorylation of rat Pard3 by ERK2 (Funahashi et al., 2013). Along with Pard3, aPKCλ is translocated via binding to Pard3 and KIF3A (Nishimura et al., 2004).
Par3 and PKMζ can be locally translated and enriched in synapses via localized accumulation and translation of mRNA, which allows for quick responses to stimuli. This mechanism is particularly useful for differentiated neurons because of the long distance between the cell body and the synapse (Sutton and Schuman, 2005; Willis et al., 2007). For example, in cultured hippocampal neurons, PKMζ mRNA is transported rapidly to and stored at dendritic postsynaptic compartments for local translation; this mRNA transport requires cis-regulatory elements located at the junction between the 5′ UTR and open reading frame and in the 3′ UTR (Muslimov et al., 2004); these cis elements are recognized by specific RNA binding proteins that bind to microtubules for transport (Tiedge et al., 1999). These RNA binding proteins also suppress translation (Hernandez et al., 2003) until relieved by neuronal activity (Kelly et al., 2007a,b) or by local stimuli, such as the hypothalamic neuropeptide oxytocin (Lin et al., 2012). This localized de novo synthesis of PKMζ is believed to play an important role in synaptic long-term potentiation (LTP) and memory maintenance (Sacktor, 2011). Similarly, in Aplysia, aPKC mRNA is stored presynaptically for local translation during long-term facilitation; this local translation is promoted by CPEB (cytoplasmic polyadenylating element-binding protein), which is normally inhibited by miRNA-22 until relieved by the neuromodulator serotonin (Fiumara et al., 2015).
The localization of Par complex components may also be regulated by chemical modifications, such as phosphorylation. For example, mouse Pard3 can be phosphorylated by Ndr kinases at Ser383. This phosphorylation blocks Par3's binding to dynein for retrograde transport, thus facilitating its axonal accumulation (Yang et al., 2014).
Par complex components can be retained in the synapses via physical association with other local proteins or lipids. Pard3/Tiam1 complex can be recruited to hippocampal synapses via physical binding to brain-specific angiogenesis inhibitor 1, a synaptic adhesion GPCR required for both spinogenesis and synaptogenesis (Duman et al., 2013). In addition, Bazooka's enrichment at axon tips can be facilitated by binding to PIP3 using its C terminal region and PDZ2 domain (Wu et al., 2007; Krahn et al., 2010; McKinley et al., 2012). This mechanism is consistent with the finding that the axonal localization of mouse Pard3 requires PI 3-kinase (Shi et al., 2004), a key enzyme for synthesizing PIP3, which is concentrated at axon tips (Menager et al., 2004).
In addition to the above-mentioned localization mechanisms, liquid-liquid phase separation (LLPS) may play a role in the localization of Par complexes. In the past decade, growing evidence has suggested that many intracellular nonmembrane-bound structures behave as liquid condensates of molecules, or droplets as they are often referred to, which can fuse, drip, and quickly exchange content with the surrounding, less concentrated liquid phase of cytoplasm or nucleoplasm (Hyman and Simons, 2012; Lee et al., 2013; Hayashi et al., 2021). In Drosophila neuroblasts, Bazooka condenses as punctate foci that quickly exchange contents with their surrounding cytoplasmic liquid phase, fuse into larger foci, or separate into smaller ones, suggesting that they possess an LLPS property (Liu et al., 2020). This liquid phase separation is facilitated by binding interactions among Par3, Par6, and aPKC, particularly the oligomerization of Par3 mediated by its N-terminal PB1 domain (Liu et al., 2020). In addition, dynamic and polarized condensation of Drosophila Bazooka, DmPar-6, and aPKC was reconstructed in S2 cells on overexpression of Bazooka, suggesting that protein concentration is a critical driving factor for Par complex liquid condensation (Kono et al., 2019).
Although LLPS suits the need for dynamic Par complexes in the ever-changing cellular environment, it is unclear how the stochastic dissolution and condensation process of LLPS reconcile with the polarized localization of Par complexes, which occurs reproducibly and robustly in many cellular contexts. In the subcellular environment, other factors, such as local protein–protein binding, protein-lipid binding, biochemical modifications, and translation-mediated local protein enrichment, may collectively dictate the nucleation of Par complexes in specific subcellular regions in a predictable fashion. While challenging, how these factors collectively orchestrate the localization of Par complexes during neuronal development is an important question to address in the future.
Another outstanding question is how the localization of Par complexes in differentiating or differentiated neurons is regulated relative to other polarity proteins in space and time. Previously, we showed that during neurulation, groups of apical polarity proteins localize to the apical side of the neuroepithelium in three steps: pioneer, intermediate, and terminal. Whereas N-cadherin and ZO-1 belong to the pioneer protein group, Pard3αb, Pard6γb, and aPKCλ, along with Crb complex, belong to the intermediate protein group, and Na+/K+ ATPase belongs to the terminal group (Yang et al., 2009; Guo et al., 2018). In morphologically complex neurons, the proper function of Par complexes and other polarity proteins may also require a certain spatiotemporal order. For example, in neurons, both N-cadherin and Na+/K+ ATPase are expressed, but it is unknown how their localization is spatiotemporally coordinated with that of Par complexes.
Par complexes in axon specification and development
Neuronal polarization is manifested by the long extension of an axon and numerous dendrites for transmitting and receiving neuronal signals, respectively. Par complexes play a critical role in defining which neurite develops into the axon (Fig. 2C). When neurites start to differentiate in cultured hippocampal neurons, enrichment of Pard3, Pard6, and aPKCζ at the tip of one presumptive axon may be essential for axon specification and development because suppression of Pard3, Pard6, and aPKCζ blocked neuronal polarization (Shi et al., 2003). The regulation of axon specification by Par complexes requires mediation by Wnt signaling as suggested by two lines of evidence: First, on the loss of the downstream effector of Wnt signaling Dvl, which directly associates with aPKC to promote its axon accumulation, axon specification was disrupted. Second, Wnt5a promoted axon specification (Zhang et al., 2007). In addition, Par complex-mediated axon specification is regulated by TGFβ signaling, as evidenced by the fact that TGFβ could stimulate axon specification through Type II TGFβ receptor by phosphorylating Par6 (Yi et al., 2010). Interestingly, aPKCλ and PKMζ isoforms localize to presumptive axon- and non–axon-forming neurites, respectively (Parker et al., 2013). Moreover, overexpression of PKMζ inhibited axon formation, whereas overexpression of aPKCλ resulted in supernumerary axons, suggesting that aPKCλ promotes axon formation and PKMζ inhibits axon specification. Instead, PKMζ may promote dendrite formation, possibly via antagonistic competition for Par3 (Parker et al., 2013).
Schematic illustration of the upstream regulators and downstream effectors of Par complexes during neurodevelopment. Double arrows indicate the binding interactions among the constitutes of Par complexes.
Once an axon is specified, it needs to extend and find its target neurons. Par complex also plays an important role in axon growth and pathfinding. To control axon growth, Par complexes regulate microtubules, the major cytoskeletal components of axons. Pard3 can directly stabilize microtubules by promoting their bundling using its N-terminal-mediated oligomerization and the physical binding between microtubules and Pard3's PDZ3 domain (Chen et al., 2013). Par complexes can also enhance microtubule assembly by suppressing microtubule affinity-regulating kinase 2 (Chen et al., 2006), which otherwise phosphorylates microtubule-associated proteins (MAPs), such as tau, and destabilizes microtubules. In addition, because MAPs regulate not only microtubule dynamics but also axonal transport and synaptic structure and function (Tapia-Rojas et al., 2019), Par complexes can exert a broad impact on neuronal development by regulating microtubule affinity-regulating kinase 2.
The promotion of axon growth and pathfinding by Par3 is influenced by extracellular signals. For example, in primary cultures of rat embryonic DRG neurons, nerve growth factor, and netrin-1 can stimulate local translation of Pard3 from stored mRNA to stimulate axon growth (Hengst et al., 2009). In contrast, in the context of CNS injury where axon growth is lacking and oligodendrocytes, Pard3, Pard6, and aPKCζ inhibit axon growth, and this inhibition is activated by chondroitin sulfate proteoglycans secreted by oligodendrocytes in the glial scar (Lee et al., 2013). Thus, different extracellular signals regulate the activity of Par complexes differently to either promote axon elongation in normal development or inhibit axon growth during CNS injury.
However, the requirement of Par complexes for axon morphogenesis is not universal. In Drosophila, not all neuronal lineages require Bazooka and DmPar-6 for axon morphogenesis and pathfinding (Rolls and Doe, 2004; Spindler and Hartenstein, 2011). Of note, Drosophila neurons' unipolar branching pattern (i.e., a single neurite branching into both dendritic and axonal terminals) is distinct from those of vertebrate neurons. This morphologic difference may necessitate different regulatory mechanisms (Spindler and Hartenstein, 2011).
Par complexes in dendritic spinogenesis
While Par complexes are enriched at axon tips during the early neuronal development, at the late stages of neuronal differentiation, in addition to localizing in axon tips, they also localize to dendritic spines to regulate spinogenesis and maintain excitatory synapses (Fig. 2C). In hippocampal neurons, spine-localized Pard3 binds to Rac GEF TIAM1 and spatially restricts it to dendritic spines, thus locally modulating the Rac–GTP-dependent actin cytoskeleton for proper spinogenesis (Zhang and Macara, 2006). As discussed earlier, this dendritic localization of Pard3 is regulated by synaptic adhesion molecule GPCR brain-specific angiogenesis inhibitor 1 (Duman et al., 2013). In addition, the coiled-coil domain of Pard3 can bind to scaffolding protein FRMD4A (FERM domain containing 4A) and form a complex with guanine nucleotide exchange factor cytohesin-1 to activate GTPase ARF6 (ADP-ribosylation factor 6) (Ikenouchi and Umeda, 2010), which modulates actin cytoskeleton dynamics in dendritic branching (Jaworski, 2007). The interaction between rat Pard3 and FRMD4A might be important for normal neurodevelopment because the loss of FRMD4A was associated with congenital microcephaly, intellectual disability, and dysmorphism (Fine et al., 2015). Similarly, in Drosophila neuromuscular junctions, Bazooka is required for postsynaptic F-actin enrichment (Ramachandran et al., 2009). Interestingly, independent of Pard3, the Pard6/aPKCζ complex also regulates the actin cytoskeleton in spinogenesis by inhibiting RhoA through p190A RhoGAP (Zhang and Macara, 2008). PKMζ also regulates the length and morphology of spines as revealed by overexpression of PKMζ in primary cultures of rat cortical neurons, which reduced spine length, although it did not affect spine density and dendritic branching (Ron et al., 2012). Similarly, in vivo overexpression and suppression of PKMζ in the optic tectum of Xenopus tadpoles revealed that PKMζ stabilized the growth of dendritic filopodia and restricted dendritic arborization, suggesting that PKMζ helps with synapse maturation during embryonic brain circuit development (Liu et al., 2009).
Par complexes in synaptic plasticity and memory
Neurotransmission at synapses is the central function of neurons. Thus, proper synaptic development, maintenance, and plasticity underlie cognition (Fromer et al., 2014; Volk et al., 2015). How specific synaptic connections are established and modulated in learning and memory remains incompletely understood, however (Sudhof, 2018; Sanes and Zipursky, 2020). Given the fact that Par complexes regulate spinogenesis and axon specification and growth, it is expected that Par complexes also regulate synaptogenesis, but research about this important topic has been scarce. Nevertheless, evidence suggests that Par complexes play important roles in synaptic plasticity at both presynaptic and postsynaptic sites and in memory formation and maintenance (Fig. 2D). In the following section, we review the findings in this regard.
In developed neurons, synapse-localized Par complexes regulate the dynamics of the synaptic cytoskeleton. For example, in the presynaptic compartment of Drosophila neuromuscular junctions, aPKC colocalizes with DmPar-6 and Bazooka and promotes the association of Futsch, a MAP1B-related protein, with microtubules, thus stabilizing microtubules. In contrast, at F-actin-rich postsynaptic sites, aPKC regulates the segregation of actin-rich regions from microtubule-rich regions without colocalizing with DmPar-6 and Bazooka; yet DmPar-6 and Bazooka regulate the formation of synaptic boutons by promoting the F-actin-rich regions (Ruiz-Canada et al., 2004). While not colocalizing with aPKC, the postsynaptic targeting of Bazooka to the F-actin-rich postsynaptic site requires its phosphorylation by aPKC; however, the subsequent retention of Bazooka in the F-actin-rich region requires its dephosphorylation by phosphatase PTEN (Ramachandran et al., 2009).
The synaptic localization of Par complex components is important for synaptic plasticity because they facilitate the recruitment of synaptic membrane proteins by physical binding and biochemical modifications. PKMζ enriched at dendritic spines can phosphorylate and activate ZDHHC8, an enzyme that palmitoylates PSD-95 to facilitate its attachment to membranes and targeting to synapses (Yoshii et al., 2011). In addition, aPKC facilitates glutamate receptor localization in the postsynaptic compartment of Drosophila neuromuscular junctions (Ruiz-Canada et al., 2004). During LTP in rats, aPKCλ can phosphorylate the GluA1 subunit of AMPAR (Parker et al., 2013; Ren et al., 2013) to facilitate its incorporation into active synapses (Boehm et al., 2006; Lin et al., 2009). Furthermore, PARD3's PDZ3 domain can bind to the intracellular PDZ domain of CNTNAP2 (contactin-associated protein-like 2), a Type I transmembrane cell adhesion molecule, to promote its punctate localization in interneurons (Gao et al., 2020); this localization is necessary for CNTNAP2 to cluster potassium channels at the juxtaparanode (Poliak et al., 2003).
LTP of synaptic strength may directly involve PKMζ (Kandel, 2001; Sacktor, 2011). According to the synaptic tagging and capture theory (Frey and Morris, 1997), several studies suggested that PKMζ is one of the plasticity-related proteins, which are synthesized locally and captured by synaptic “tags” in newly potentiated synapses at the early phase of LTP and promote these synapses to develop from the early into the late phase of LTP (Osten et al., 1996; Sajikumar et al., 2005; Serrano et al., 2005; Kelly et al., 2007a; Sajikumar and Korte, 2011). PKMζ was suggested to be necessary and sufficient to maintain late phase of LTP in hippocampal slices (Ling et al., 2002), and that overexpression of WT PKMζ, but not dominant-negative PKMζ, enhanced long-term memory (Shema et al., 2007, 2009, 2011; Serrano et al., 2008; Kwapis et al., 2009; von Kraus et al., 2010). Overexpression of PKMζ also promoted memory in Drosophila and rats (Drier et al., 2002; Shema et al., 2011). In contrast, inhibition of PKMζ by ZIP (ζ inhibitory peptide, a myristoylated pseudosubstrate peptide derived from the autoinhibitory region of PKC) reversed LTP and erased memory (Pastalkova et al., 2006; Madronal et al., 2010; Gao et al., 2018), and suppression of PKMζ by chelerythrine induced LTD (Hrabetova and Sacktor, 1996). LTD may require differential regulation of PKC isoforms because upregulation of cPKC and downregulation of PKMζ were both associated with LTD (Hrabetova and Sacktor, 2001).
Of note, methodological issues have raised doubts about the above-mentioned findings implicating PKMζ in LTP and LTD. First, the specificity of ZIP is a major concern about the specific role of PKMζ in vivo because ZIP also blocks cPKCs (Karaman et al., 2008; Wu-Zhang et al., 2012). Supporting these doubts, ZIP was reported to broadly disrupt neural activity (LeBlancq et al., 2016). In addition, whereas genetic loss of PKMζ did not cause learning and memory defects, ZIP treatment of PKMζ-deficient mice showed such defects (Lee et al., 2013; Volk et al., 2013). Finally, the ZIP-mediated loss of LTP and memory could be nonspecific because ZIP was excitotoxic to cultured hippocampal neurons (Sadeh et al., 2015).
Despite the above concerns regarding the effects of ZIP, a role for PKMζ in late LTP and long-term memory cannot be excluded. For example, PKMζ may be redundant with aPKCλ, which is more stable than newly synthesized PKMζ in synapses after chemical LTP induction (Palida et al., 2015). Supporting this notion, KO of PKMζ in mice can be compensated for by aPKCλ, which can also be inhibited by ZIP (Tsokas et al., 2016; Farah et al., 2017). Likewise, the loss of aPKCλ can be compensated for by PKMζ (Sheng et al., 2017). The potential redundancy between PKMζ and aPKCλ in mammals is consistent with the observation that the loss of aPKC compromised synaptic bouton formation in Drosophila, which has only one aPKC (Ruiz-Canada et al., 2004).
An involvement of PKMζ in LTP and memory is also consistent with the immuno-EM observation that age-related loss of memory is correlated with a decrease in the dendritic synaptic expression of PKMζ and AMPA receptors in rhesus monkeys (Hara et al., 2012). In addition, overexpression of PKMζ-GFP in postsynaptic compartments was correlated with the enrichment of postsynaptic scaffold protein PSD-95 at excitatory synapses (Shao et al., 2012) and with the shortening of dendritic spines and the increase in the amplitude of excitatory postsynaptic currents (Ron et al., 2012).
All these findings suggest that both PKMζ and aPKCλ play important roles in memory formation, possibly in a partially redundant fashion to ensure robust LTP. This redundancy is supported by mathematical modeling (Jalil et al., 2015). The full-length aPKC might be more involved in short-term memory because it can be regulated by external signals, whereas PKMζ may predominantly regulate long-term memory because of its constitutive activity (Hernandez et al., 2003) that is not easily influenced by external signals. This notion is also supported by the differential effects of RNAi-mediated knockdown of aPKCλ and PKMζ in the mouse dorsal hippocampus on short-term and long-term memory, respectively (Wang et al., 2016).
Par complexes in glial development
Par complexes regulate not only neuronal development but also the development of glial cells, although research on this topic is not as abundant as that on neurons. It has been found, for example, that Pard3 is required for myelination. In cocultures of Schwann cells and DRG neurons of rats, Pard3 recruits p75 neurotrophin receptor (NTR) to the axon-glial junction through physical binding between Pard3's PDZ1 domain and p75 NTR in initiating myelination in a BDNF-dependent manner (Chan et al., 2006). Nectin-like cell adhesion protein 4 (Necl-4), which is expressed in Schwann cells, also interacts with Pard3, and this interaction is required for the proper localization of Pard3 at the interface between Schwann cells and axons (Meng et al., 2019). However, it is unknown whether Pard3 regulates Necl-4 localization by binding to Necl-1 (Meng et al., 2019). Finally, Pard3 is required for Reelin-induced Schwann cell migration by directly interacting with Reelin receptor ApoER2 (Pasten et al., 2015). Further research is expected to reveal additional functions of Par complexes in glial development.
A question about versatile manifestations of cell polarity in the nervous system
From morphologically simple neuroepithelial cells to complex neurons and glia of various shapes, cells of the nervous system are polarized, but with different manifestations, as exemplified by the subcellular localizations of Par complexes in the nervous system. This morphologic diversity raises questions about cell polarity and subcellular polarity: Can subcellular structures be polarized independent of cell polarization? If the presynaptic and postsynaptic compartments can be considered as the two opposing ends of polarized neurons, how do Par complexes localize to both ends? Can they still be considered polarity proteins? Can the seemingly opposite roles of Par complexes in axon versus dendrite specification be reconciled by differential localization of distinct homologs and/or isoforms on a spatiotemporal scale, such as aPKCλ in axons versus PKMζ in dendrites? Alternatively, spatially segregated subcellular structures, such as axonal tips and dendritic spines, should develop their own polarity without being influenced by polarity proteins at distal subcellular structures. In this way, the same apical polarity apparatus can be repurposed at opposite ends of the same neurons. It would be interesting to determine if and how these two mechanisms, or any other mechanisms, coordinate to regulate neuronal polarization.
Involvement of Par Complex Genes in Human Neuropsychiatric and Neurodegenerative Disorders
Defects in neuronal migration, neuronal polarization, and synaptogenesis are major characteristics of neurodevelopmental disorders. In addition, synaptic defects often precede massive cell loss in neurodegenerative and cognitive conditions, such as Parkinson's disease and AD (Marcello et al., 2012; Picconi et al., 2012; Lepeta et al., 2016; Lanctot et al., 2017; Soukup et al., 2018). Thus, both neurodevelopmental and neurodegenerative disorders can undermine neural circuits and functions and manifest symptoms, such as intellectual disability, bipolar disorder, autism spectrum disorder, attention-deficit/hyperactivity disorder, and schizophrenia (Valnegri et al., 2012; Yin et al., 2012; Batool et al., 2019). Although emerging evidence has linked some genetic risk factors with neurodevelopmental, neurodegenerative, and neuropsychiatric disorders (Volk et al., 2015; Taylor et al., 2020; Rizalar et al., 2021), a complete understanding of the genetic causes of these disorders has yet to be achieved.
Given the important roles of Par complexes in neuroepithelial polarization, neurogenesis, neuronal migration, neurite development, synaptic plasticity, and memory (Fig. 2), it is reasonable to speculate that malfunction of Par complexes may contribute to neurodevelopmental, neurodegenerative, and neuropsychiatric disorders. For example, a PARD3 genetic defect may contribute to neuropsychiatric disorders by affecting its role in subcellular targeting of CNTNAP2 (Gao et al., 2020), malfunction of which is a risk factor for many neuropsychiatric disorders, including intellectual disability, autism spectrum disorder, schizophrenia, epilepsy, and cortical dysplasia focal epilepsy syndrome (Anney et al., 2012; Zweier, 2012; Rodenas-Cuadrado et al., 2016). It is also possible that hyperactivity of PRKCζ and PARD6α may promote neurodegeneration because increased expression of PRKCζ and PARD6α promoted the secretion of tau in cell culture (Yan et al., 2016). And tau secretion is a possible mechanism to propagate misfolded tau from cell to cell in a prion-like fashion, thus spreading the fibrillization of tau in healthy neighbor neurons, causing tauopathy (Frost et al., 2009; Sanders et al., 2014).
Supporting the above speculation, some genetic variants of Par complex genes have been associated with neuropsychiatric and neurodegenerative disorders in humans (Table 3). In this section, we review these variants and discuss their implications.
Genetic variants of human par complex genes and their associated clinical phenotypesa
Neuropsychiatric disorders
PARD3 and PARD3β
Genetic variants of both PARD3 and PARD3β are found to be associated with neuropsychiatric disorders. In a study of 204 schizophrenia patients and 351 normal controls in a Korean population, synonymous single-nucleotide variant (SNV) rs3781128 and two intronic SNVs rs1936429 and rs671228 of PARD3 were found to be associated with schizophrenia (Kim et al., 2012). In another study of 2458 cases of schizophrenia, 1108 cases of autism spectrum disorder, and 2095 healthy controls in a Japanese population, two small deletion copy number variants (CNVs), RCV000754238 and RCV000754239, affecting only PARD3β but not its neighboring gene, were associated with schizophrenia (Kushima et al., 2018). In another study of 11,730 patients with neurodevelopmental disorders and 2867 controls, two likely gene-disruptive variants in PARD3β were associated with autism spectrum disorder (Stessman et al., 2017). In a different study, intronic SNV rs4675502 of PARD3β was found to be associated with autism spectrum disorder (Anney et al., 2010, 2012). In another study, missense SNV rs80119103, which causes a Pro1145His mutation in PARD3β at the C-terminus that is not conserved with PARD3, was found in 1 of 8 Saudi children with attention-deficit/hyperactivity disorder (Bogari et al., 2020).
PARD6β and PARD6γ
Two of the three human PARD6 genes have been implicated in neuropsychiatric disorders. A study of 188 bipolar disorder patients and 376 healthy controls in a Bulgarian population suggested that SNV rs6122972, located in the 3′ UTR of PARD6β, is associated with bipolar disorder (Yosifova et al., 2011). This SNV may affect mRNA stability, thus affecting the PARD6β protein level. In another study of 909 subjects with schizophrenia and 917 controls from the South African Xhosa population, missense mutation Gly165Cys of PARD6γ was identified in one case, as a heterozygote. This mutation altered a highly conserved amino acid residue in the PDZ domain and was not present in any public database (Gulsuner et al., 2020).
PRKCι and PRKCζ
Abnormalities in both PRKCι and PRKCζ genes have also been associated with neuropsychiatric disorders. A comparison between SNVs and gene expression levels in the human PFC revealed that intronic SNV rs2140825-based reduction in PRKCι expression was associated with bipolar disorder (Wellcome Trust Case Control, 2007; Iwamoto et al., 2011). Another study revealed that intronic SNV rs1392366 of PRKCι was associated with bipolar disorder comorbid with alcohol use disorder (Dalvie et al., 2016). The expression of PRKCι as well as many other genes was promoted by topiramate, a drug that has been used to treat bipolar disorder and methamphetamine dependence (Li et al., 2014). In addition, higher PRKCι mRNA expression in the PFC was found in suicide victims (Choi et al., 2011). Moreover, a CpG island in the promoter region of PRKCι was shown to be entrained by light through methylation, suggesting that PRKCι protein expression levels are subject to seasonal and geographical influences, providing a molecular explanation for mood swings caused by these environmental factors (Aslibekyan et al., 2014).
In addition to PRKCι, PRKCζ and PKMζ have also been suggested as risk factors for neuropsychiatric disorders. In a study of 738 patients, intronic SNV rs3753242 in PRKCζ was found to be associated with schizophrenia (McClay et al., 2011). In another study of 600 bipolar disorder patients and 605 normal controls, intronic SNVs rs3128396, rs2503706, and rs3128309 in PRKCζ were found to be associated with bipolar disorder (Kandaswamy et al., 2012). In the same study of the South African Xhosa population mentioned above, a Val557Met missense mutation of PRKCζ was identified in another heterozygous schizophrenia patient, affecting a highly conserved amino acid residue in the kinase domain (Gulsuner et al., 2020). In addition, SNV rs3128296, located in the promoter region of PKMζ, is associated with bipolar disorder (Kandaswamy et al., 2012).
In contrast to the above evidence of the association of genetic variants of human Par complex genes with neuropsychiatric conditions, variants of these genes were not identified as risk factors for autism spectrum disorder in four other studies of a total of 965 autistic families (Iossifov et al., 2012; Neale et al., 2012; O'Roak et al., 2012; Sanders et al., 2012), suggesting that genetic variations in Par complexes may not be common causes of autism spectrum disorder.
Neurodegenerative disorders
Evidence also suggests that Par complex genes are implicated in neurodegenerative disorders. For example, in a study of 86 AD patients and 404 controls in Sweden's Uppsala population, synonymous SNV rs14920 in PARD6α was reported to be associated with AD (Giedraitis et al., 2009). In addition, PKMζ was found enriched in phospho-tau-rich neurofibrillary tangles, one of the hallmark intracellular pathologic features of AD (Crary et al., 2006), implying that improper aggregation of PKMζ may be related to neuronal degeneration. Similarly, PRKCι was also found enriched in tau-positive neurofibrillary inclusions and actin-rich Hirano bodies in AD patients as well as in α-synuclein-rich Lewy bodies in idiopathic Parkinson's disease and dementia with Lewy bodies (Shao et al., 2006). However, the mechanism of such PKMζ and PRKCι aggregation is unknown; it is unclear whether any genetic alteration in PKMζ and PRKCι caused their intracellular aggregation, which then resulted in neurodegeneration.
Effects of partial versus complete loss of Par complex functions
Most of the above variants are synonymous SNVs, intronic SNVs, or CNVs, except for a few missense mutations. Synonymous SNVs may reduce the efficiency of protein translation because of changes in codon usage; intronic SNVs may affect cis-regulatory elements that either affect transcription levels and specificity or alter splicing isoform selection; CNVs may cause gene dosage imbalance, whereas missense mutations in heterozygous patients may cause either a haploinsufficient or a dominant-negative effect (Kearney et al., 2011; Girirajan et al., 2012, 2013; Kirov, 2015; Shaikh, 2017; Takumi and Tamada, 2018). The common feature of these variants is that they may only partially compromise gene functions because any severe disruption of Par complex functions would likely have resulted in embryonic lethality because of their pivotal roles in many cellular processes, as suggested by animal models (Horne-Badovinac et al., 2001; Wei et al., 2004; Hirose et al., 2006; Munson et al., 2008; Guo et al., 2018). Such embryonic lethality would obscure the manifestation of any postnatal neuropsychiatric disorders. In addition, deleterious mutations would not be preserved in the gene pool because of purifying selection. For example, several PARD3 variants were associated with fatal neural tube defects in humans: in one study, 11 of 138 human fetuses with craniorachischisis and anencephaly had defects in PARD3 (Chen et al., 2017); in another study of 224 fetuses with anencephaly, 6 SNVs in PARD3 were associated with anencephaly (Gao et al., 2012).
In addition to germline mutations, somatic mutations of Par complex genes may also be a cause of neuropsychiatric and neurodegenerative disorders. Because somatic mutations can occur in local regions in a mosaic fashion, pathogenic mutations may completely abolish the functions of the concerned genes while still allowing the affected people to survive and manifest neuropsychiatric or neurodegenerative disorders (D'Gama and Walsh, 2018). Of course, it is challenging to identify patients with somatic mutations, and any manifested neuropsychiatric symptoms in such patients could vary drastically from case to case, depending on the timing of mutations and the scope of the affected areas. As far as we know, no such cases have been identified for Par complex genes.
Together, partial but not complete and systemic loss of the functions of Par complex genes likely contributes to neuropsychiatric and neurodegenerative disorders. However, current human genetics research on these genes has been at the level of finding and confirming the association between genetic variations and the disorders. Such associations need to be validated in animal models to rule out the possibility that any clinical symptoms were caused by defects in other genes. In addition, animal models will help with understanding how the genetic variants promote these disorders at the molecular and cellular levels. To study these genes in animal models, thoughtful genetic approaches may be necessary because of potential embryonic lethality and functional redundancy as mentioned above. For example, conditional KO of aPKCλ in differentiated mouse brain neurons did not result in loss of neurons (Yamanaka et al., 2013); this could suggest either that aPKCλ is not required for neuronal maintenance or that its role is compensated for by aPKCζ, as aPKCζ compensated for the loss of aPKCλ in LTP and memory (Sheng et al., 2017). The ever-expanding databases of human genetic variants (Sherry et al., 2001; Landrum et al., 2016; Azzariti et al., 2018), such as ClinVar, dbVar, dbSNP, and gnomAD, will help to reveal the genetic and molecular basis for various neuropsychiatric and neurodegenerative disorders.
In conclusion, at the nexus of various signaling pathways, the evolutionarily conserved Par3, Par6, and aPKC polarity proteins play essential roles in neural development and maintenance by regulating neuroepithelial polarity, neurogenesis, neuronal migration, neuronal differentiation, neuronal polarization, synaptic plasticity, and memory (Fig. 3). As a result, dysfunction of these proteins can contribute to neurodevelopmental, neuropsychiatric, and neurodegenerative symptoms. While it is practical and important to study the functions of these genes with in vitro approaches, one must realize the limitations of in vitro methodology because neurons interact more naturally with their neighboring cells in vivo and they may behave quite differently in cell culture. Future study of these genes in various types of neurons and glial cells with in vivo approaches will be challenging and yet rewarding.
Footnotes
L.Z. was supported by the Department of Psychology at Dalian Medical University. X.W. was supported by the Department of Ophthalmology at the University of Pittsburgh; the Eye and Ear Foundation of Pittsburgh; and an unrestricted grant from Research to Prevent Blindness to the Department of Ophthalmology at the University of Pittsburgh. We thank the reviewers and editors for the constructive feedback and editing; and Dr. Mary-Claire King for providing detailed information on some schizophrenia cases and suggestions for the manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Lili Zhang at zhangll{at}dmu.edu.cn or Xiangyun Wei at weix{at}upmc.edu