Modulation of Dendritic Spines by Protein Phosphatase-1

Abstract

Protein phosphatase-1 (PP-1), a highly conserved multifunctional serine/threonine phosphatase, is enriched in dendritic spines where it plays a major role in modulating excitatory synaptic activity. In addition to established functions in spine maturation and development, multi-subunit holoenzyme forms of PP-1 modulate higher-order cognitive functions such learning and memory. Mechanisms involved in regulating PP-1 activity and localization in spines include interactions with neurabin and spinophilin, structurally related synaptic scaffolding proteins associated with the actin cytoskeleton. Since PP-1 is a critical element in synaptic development, signaling, and plasticity, alterations in PP-1 signaling in dendritic spines are implicated in various neurological and psychiatric disorders. The effects of PP-1 depend on its isoform-specific association with regulatory proteins and activation of downstream signaling pathways. Here we review the role of PP-1 and its binding proteins neurabin and spinophilin in both developing and established dendritic spines, as well as some of the disorders that result from its dysregulation.

Introduction

A precisely regulated balance in the activities of protein kinases and phosphatases governs reversible phosphorylation states of key proteins essential for cellular function. Protein phosphatases, similar to protein kinases, include both protein serine-threonine phosphatases and protein tyrosine phosphatases based on the amino acid residues that they dephosphorylate. However, unlike the hundreds of kinases encoded by the human genome, protein phosphatases are relatively fewer in number. Their selectivity is conferred largely through formation of supramolecular complexes containing a variety of regulatory and targeting subunits that direct phosphatase catalytic subunits to specific subcellular locations and substrates.

Protein phosphatase-1 (PP-1) is a highly conserved serine/threonine phosphatase with multifunctional roles in diverse cellular processes. These multiple roles are mediated through interactions of the PP-1 catalytic subunit (PP-1c) with various regulatory subunits that regulate its activity and substrate specificity (Cohen, 2002). In addition, the PP-1c activity is modulated by the phosphorylation state of a number of regulatory or targeting subunits including inhibitor-1, inhibitor-2, dopamine- and cyclic AMP-regulated phosphoprotein-32 (DARPP-32), and PP-1 nuclear targeting subunit (PNUTS), which facilitate complex integrations of multiple cellular signaling pathways (Watanabe, Huang, Horiuchi, da Cruze Silva, Hsieh-Wilson, Allen et al., 2001; Kim, Watanabe, Allen, Kim, Lee, Greengard et al., 2003). PP-1 was initially identified based on its role in controlling glycogen metabolism and protein synthesis in skeletal muscle (Strålfors, Hiraga, & Cohen, 1985; Shenolikar & Nairn, 1991), but multiple functions for PP-1 have since been discovered in neurons including roles in neurite outgrowth (Oliver, Terry-Lorenzo, Elliott, Bloomer, Li, Brautigan et al., 2002), synapse formation (Malchiodi-Albedi, Petrucci, Picconi, losi, & Falchi, 1997), ion channel function (Wang, Orser, Brautigan, & MacDonald, 1994; Surmeier, Bargas, Hemmings, Nairn, & Greengard, 1995) and neurotransmission (Yan, Hsieh-Wilson, Feng, Tomizawa, Allen, Fienberg et al., 1999; Colbran, 2004). Identification of PP-1 in dendritic spines highlighted its role in higher-order functions such as synaptic plasticity (Mulkey, Endo, Shenolikar, & Malenka, 1994; Morishita, Connor, Xia, Quinlan, Shenolikar, & Malenka, 2001), learning and memory, and behavior (Genoux, Haditsch, Knobloch, Michalon, Storm, & Mansuy, 2002; Yu, Huang, Chang, & Gean, 2016). This review focuses on the roles of PP-1 in dendritic spines including the roles of its binding proteins neurabin and spinophilin in modulating both normal and pathological functions.

Expression and targeting of PP-1 in dendritic spines

In mammals, PP-1c is encoded by three genes giving rise to four isoforms: PP-1α, PP-1β, PP-1γ1, PP-1γ2, with the latter two originating by differential splicing of the same gene (Sasaki, Shima, Kitagawa, Irino, Sugimura, & Nagao, 1990). The catalytic subunits for all four isoforms are ~80% identical in amino acid sequence and possess distinct tissue distributions and subcellular localizations (Cohen, 2002). Brain immunoreactivity for isoforms PP-1α and PP-1γ1 is heterogeneously distributed, with intense labelling in the prefrontal cortex, striatum, and hippocampal formation (Ouimet, da Cruz e Silva, & Greengard, 1995; Muly, Greengard, & Goldman-Rakic, 2001). Relatively more widespread brain distribution is found for PP-1β (Strack, Kini, Ebner, Wadzinski, & Colbran, 1999), with PP-1γ2 showing greater expression in non-neuronal cells (Shima, Haneji, Hatano, Kasugai, Sugimara, & Nagao, 1993) compared to neurons. Evidence from immunofluorescence and immuno-electron microscopy also reveals differential subcellular localization of brain PP-1 isoforms. PP-1β is found in discrete areas of the soma and dendrites co-localized with microtubules, while PP-1α and PP-1γ1 are highly enriched in presynaptic terminals and dendritic spines where they largely associate with the actin cytoskeleton (Fig. 1; Ouimet et al., 1995; Muly et al., 2001; Strack et al., 1999).

Figure 1:

Schematic diagram showing subcellular localizations of protein phosphatase-1 isoforms. PP-1β is found in the soma and dendrites co-localized with microtubules (left). PP-1α and PP-1γ1 are enriched in presynaptic terminals and postsynaptic dendritic spines associated with the actin cytoskeleton (inset).

Dendritic spines are actin-rich, postsynaptic structures that compartmentalize biochemical and cell biological processes critical for excitatory synaptic transmission and plasticity. These dynamic structures concentrate second messenger pathways, signaling enzymes, cytoskeletal elements, and transmembrane and scaffolding molecules appropriate for integration of regulated signaling. PP-1 complexes purified from rat cytoskeletal extracts include neurabin and spinophilin as scaffolding proteins that target PP-1α and PP-1γ1 to dendritic spines (Allen, Ouimet, & Greengard, 1997; MacMillan, Bass, Cheng, Howard, Tamura, Strack et al., 1999; Terry-Lorenzo, Carmody, Voltz, Connor, Li, Smith et al., 2002). This facilitates efficient regulation of synaptic phosphoproteins involved in spine formation, signaling, and synaptic plasticity.

Neurabin and spinophilin, also known as neurabin 2, are structurally similar proteins containing shared regions with 43% to 86% homology (Fig. 2; Satoh, Nakanishi, Obaishi, Wada, Takahashi, Satoh et al., 1998): N-terminal actin-binding domains, PP-1 binding domains, PDZ domains (transmembrane receptor binding), and C-terminal coiled-coil domains (GTPase regulatory protein binding) (Ryan, Alldritt, Svenningsson, Allen, Wu, Nairn et al., 2005; Kelker, Dancheck, Ju, Kessler, Hudak, Nairn et al., 2007). In addition, neurabin, but not spinophilin, contains a sterile α motif (SAM) domain in its C-terminus that may mediate additional protein interactions (Ju, Ragusa, Hudak, Nairn, & Peti, 2007). Both proteins are highly concentrated in the postsynaptic density [PSD] (Muly, Allen, Mazloom, Aranbayeva, Greenfield, & Greengard, 2004) and promote F-actin crosslinking in a phosphorylation-dependent manner (Hsieh-Wilson, Allen, Watanabe, Nairn, & Greengard, 1999; Grossman, Futter, Snyder, Allen, Nairn, Greengard et al., 2004), and thereby play an important role in the formation and function of dendritic spines (Nakanishi, Obaishi, Satoh, Wada, Mandai, Satoh et al., 1997). Deletion and genetic knockout studies reveal a role for neurabin in regulating F-actin polymerization involved in spine morphogenesis and glutamatergic synapse formation (Zito, Knott, Shepherd, Shenolikar, & Svoboda, 2004; Terry-Lorenzo, Roadcap, Otsuka, Blanpied, Zamorano, Garner et al., 2005), while spinophilin modulates spine morphology and density, and glutamatergic transmission (Feng, Yan, Ferreira, Tomizawa, Liauw, Zhuo et al., 2000; Evans, Robinson, Shi, & Webb, 2015).

Figure 2:

Schematic diagram showing domain structures of spinophilin (top) and neurabin-1 (bottom) with corresponding amino acid residue numbers (Kelker et al., 2007).

Regulation of protein phosphatase-1 by neurabin and spinophilin

PP-1 interactions with neurabin, spinophilin, and the actin cytoskeleton within dendritic spines are ideally positioned to regulate dephosphorylation of substrates such as NMDA and AMPA glutamate receptors (Snyder, Fienberg, Huganir, & Greengard, 1998; Yan et al., 1999), calcium and potassium ion channels (Endo, Critz, Byrne, & Shenolikar, 1995; Surmeier et al., 1995), and calmodulin-dependent protein kinase II (CaMKII) and protein kinase A (PKA) signaling (Blitzer, Connor, Brown, Wong, Shenolikar, Iyengar et al., 1998; Allen, 2004). Interestingly, neurabin, spinophilin, and many of their interacting partners are regulated by reversible phosphorylation, further modulating assembly and/or activity (McAvoy, Allen, Obaishi, Nakanishi, Takai, Greengard et al., 1999; Sarrouilhe, di Tommaso, Metaye, & Ladeveze, 2006). These multi-protein complexes regulate rapid signaling cascades involved in PP-1-mediated synaptic transmission and plasticity.

Activity

Neurabin is a F-actin binding protein that copurifies with PP-1α and PP-1γ from rodent brain extracts, interacts with PP-1α in a yeast two-hybrid screen, and potently inhibits the catalytic activity of PP-1α (McAvoy et al., 1999). Neurabin migrates at ~180 kDa on SDS-PAGE, but based on its amino acid composition is ~123 kDa (Nakanishi et al., 1997). At least two domains in neurabin have been identified to bind and target PP-1 isoforms to F-actin in vitro and in intact cells. The canonical PP-1 binding motif, particularly Phe-460, can interact with both PP-1α and PP-1γ1 isoforms, but auxiliary residues (473-479) C-terminal to the PP-1 binding domain further stabilize targeting of PP-1γ1 (Carmody, Baucum II, Bass, & Colbran, 2010). Regulation of neurabin interactions with PP-1 is further modulated by phosphorylation. For example, phosphorylation of Ser-461 in neurabin by PKA, a site located C-terminal to the PP-1 binding domain, attenuates PP-1c interactions with neurabin, reducing its inhibition of PP-1c activity and revealing a novel mechanism whereby neuromodulators of the PKA pathway can regulate downstream activity of PP-1 (McAvoy et al., 1999; Oliver et al., 2002).

Spinophilin, named for its enrichment in dendritic spines, also has a lower molecular weight (89 kDa) compared to its apparent mass (134 kDa) estimated by SDS-PAGE (Allen et al., 1997). It has similar structural interactions with PP-1α and PP-1γ1 to those of neurabin, including the RVxF PP-1c-binding motif as well as residues outside this domain (478-494) that determine binding of PP-1c and inhibition of its activity, respectively (Hsieh-Wilson et al., 1999). Mutation of Phe-451 or deletion of the conserved motif abolishes binding of PP-1 as observed by co-precipitation and competition binding assays (Hsieh-Wilson et al., 1999). Phosphorylation of residues in the actin-binding domain by PKA reduces the affinity of spinophilin for F-actin and shifts its subcellular localization away from the PSD, with no major effect on its PP-1c binding affinity (Hsieh-Wilson, Benfenati, Snyder, Allen, Nairn, & Greengard, 2003). These phosphorylation sites (Ser-94 and Ser-177) are not conserved in neurabin, and suggest further differential regulation through phosphorylation. In addition to PKA, spinophilin interaction with F-actin is regulated through phosphorylation by CaMKII, cyclin-dependent kinase 5 (cdk5), extracellular signal-regulated kinase (ERK), and protein kinase C (PKC) (Grossman et al., 2004; Futter, Uematsu, Bullock, Kim, Hemmings, Nishi et al., 2005).

Spine formation and maturation

Spine density and morphology are dynamic during development as well as in the adult brain, which demands active remodeling of the cytoskeleton (Harris & Kater, 1994; Rochefort & Konnerth, 2012) and suggests key roles for neurabin, spinophilin, and PP-1 in spine modulation. Early studies using heterologous expression in non-neuronal cells identified a role for neurabin in neurite outgrowth (Nakanishi et al., 1997) and for PP-1-dependent regulation of cell morphology (Oliver et al., 2002). Subsequent studies in cultured hippocampal neurons showed that the isolated N-terminal actin-binding domains of both neurabin and spinophilin can induce filopodia extension, delaying spine formation and maturation (Zito et al., 2004; Terry-Lorenzo et al., 2005). This effect is antagonized not only by mutation of the actin-binding domains, but also by association of neurabin with PP-1. The neurabin/PP1 complex remodels F-actin to transform filopodia into spines, increases surface expression of GluA1 AMPA receptor subunits, and enhances AMPA receptor currents underlying functional maturation of excitatory synapses (Zito et al., 2004; Terry-Lorenzo et al., 2005). Potential postsynaptic substrates that may be dephosphorylated by neurabin/PP-1 to regulate the spine actin cytoskeleton include actin-depolymerizing factor (ADF)/cofilin (Meberg, Ono, Minamide, Takahashi, & Bamburg, 1998), and CaMKII (Lisman & Zhabotinsky, 2001).

In addition to actin remodeling, PP-1 modulates the function of various ion channels through dephosphorylation, including glutamate receptors (Wang, Salter, & MacDonald, 1991; Greengard, Allen, & Nairn, 1999) involved in spine formation and function (Moser, Trommald, & Andersen, 1994; McKinney, Capogna, Durr, Gahwiler, & Thompson, 1999). Spinophilin regulates excitatory synaptic transmission by targeting PP-1 to AMPA and NMDA glutamate receptors via the PDZ domain (Kelker et al., 2007) and promotes channel down-regulation through dephosphorylation of GluA1 at Ser-845 (Snyder, Allen, Fienberg, Valle, Huganir, Nairn et al., 2000) and NR1 NMDA receptor subunits at Ser-890 and Ser-897 (Snyder et al., 1998; Yan et al., 1999). Deletion of the spinophilin gene reduces this ability of PP-1 to regulate glutamatergic synaptic transmission by uncoupling PP-1 from glutamate receptor channels and producing persistent AMPA currents and reducing long-term depression (LTD; Feng et al., 2000). Moreover, mice lacking neurabin or spinophilin have reduced PP-1 expression in striatum and hippocampus, further affecting phosphatase activity (Allen, Zachariou, Svenningson, Lepore, Centonze, Costa et al., 2006; Salek, Edler, McBride, & Baucum II, 2019).

Structural spine changes from alterations in synaptic transmission depend on the brain region: in the caudatoputamen, spinophilin knockout increases or has no effect on spine density depending on age (Feng et al., 2000). Alternatively, in hippocampal neurons, removal of spinophilin decreases dendritic spine density (Evans et al., 2015). As multifunctional scaffolding proteins, association of neurabin or spinophilin with different binding partners may also influence phenotypic differences in spine modulation. Neurabin and spinophilin are localized to spines via a conserved N-terminal F-actin binding domain (Grossman, Hsieh-Wilson, Allen, Nairn, & Greengard, 2002) and also share a PP-1 binding domain to target PP-1 to dendritic spines (Allen et al., 1997; Hsieh-Wilson et al., 1999). In addition to F-actin and PP-1 binding domains, both neurabin and spinophlin contain a PDZ and C-terminal coiled coil domains that have been implicated in regulating dendritic spine morphology. An interaction between neurabin and the Rho guanine nucleotide exchange factor (GEF) Lfc was identified in a yeast two-hybrid analysis showing a novel mechanism for F-actin reorganization, and in turn, regulation of spine morphology (Ryan et al., 2005). Upon neuronal stimulation, Lfc accumulates in spines through neurabin/spinophilin interaction, locally activating Rho to stabilize F-actin and cause spine shrinkage (Ryan et al., 2005). In contrast, Asef2, another GTPase activating GEF, is recruited to synaptic sites by spinophilin, promoting formation of dendritic spines and synapses in hippocampal neurons by a Rac-dependent mechanism (Evans et al., 2015).

Synaptic plasticity

Hippocampal-dependent plasticity

Synaptic plasticity is the strengthening or weakening of synapses in response to activity patterns; it involves complex modulation by distinct signaling proteins that alter cellular activity. The two main types of synaptic plasticity are either a persistent decrease [LTD] or increase [long-term potentiation (LTP)] in synaptic efficiency (Citri & Malenka, 2008), involving in part endocytosis or increased surface expression of synaptic AMPA receptors, respectively (Luscher, Xia, Beattie, Carroll, von Zastrow, Malenka et al., 1999; Man, Ju, Ahmadian, & Wang, 2000; Carroll, Lissin, von Zastrow, Nicoll, & Malenka, 2001). LTD, like LTP, is predominantly mediated by activation of synaptic NMDA receptors (NMDAR-LTD; Mulkey & Malenka, 1992) or by mGluRs (mGluR-LTD; Oliet, Malenka, & Nicoll, 1997; Bellone, Lüscher, & Mameli, 2008) at hippocampal synapses.

In addition to synaptic stimulation, pharmacological agents can induce NMDAR-or mGluR-LTD. The low-frequency pattern of synaptic activity necessary to induce NMDAR-LTD is ineffective in activating specific protein kinases, but does activate PP-1 via dephosphorylation of I-1 by calcineurin (Mulkey, Herron, & Malenka, 1993). Induction of LTD via NMDARs recruits PP-1 to the activated synapses (Fig. 3; Morishita et al., 2001), where internalized AMPARs are stabilized through dephosphorylation of Ser-845 in GluA1 (Fig. 3; Ehlers, 2000) or Ser-880 in GluA2 (Matsuda, Mikawa, & Hirai, 1999), with additional involvement of protein tyrosine phosphatases (Collingridge, Isaac, & Wang, 2004). AMPAR endocytosis is further permitted by dephosphorylation of Ser-295 of PSD-95 by PP-1, causing dispersal of PSD-95 from synaptic sites and subsequent internalization of AMPAR (Kim, Futai, Jo, Hayashi, Cho, & Sheng, 2007). Initially presumed to involve phosphatases, various kinases have also been implicated in NMDAR-LTD, including PP-1 activation of GSK-3β by dephosphorylation of GSK3β itself and its upstream inhibitor, Akt (Szatmari, Habas, Yang, Zheng, Hagg, & Hetman, 2005). Targeting of PP-1 to its substrates necessary for LTD requires the scaffolding protein spinophilin, as peptides that prevent the PP-1-spinophilin interaction and examination of glutamatergic responses in spinophilin knockout mice reveal potentiation of glutamate receptor responses (Yan et al., 1999) and a failure to elicit NMDAR-LTD (Morishita et al., 2001; Allen et al., 2006).

Figure 3:

Role of PP-1 activity in the bidirectional control of hippocampal NMDA-mediated synaptic plasticity. A. Low frequency stimulation-long-term depression (LFS-LTD, orange lines): LFS activates a calcineurin cascade that leads to inhibitor-1 (I-1) dephosphorylation and disinhibition of PP-1 activity. PP-1 then dephosphorylates Ser-845 on GluA1 leading to internalization of dephosphorylated AMPA receptors. Additional dephosphorylation of Ser-295 on PSD-95 and activation of GSK-3β (dephosphorylation of Ser-9) further permits AMPAR endocytosis and expression of LTD. B. High frequency stimulation-long-term potentiation (HFS-LTP, green lines): HFS leads to phosphorylation of I-1 by PKA and subsequent inhibition of PP-1. Activation of PKA and of CaMKII leads to GluA1 phosphorylation at Ser-831 and Ser-845, leading to increased conductance, membrane insertion of AMPA receptors and expression of LTP. CaM, calmodulin; CaMKII, calcium/calmodulin-dependent protein kinase II; cAMP, cyclic adenosine monophosphate; GSK-3β, glycogen synthase kinase-3β; I-1, inhibitor-1; PKA, protein kinase A; PP-1, protein phosphatase-1; PSD-95, post-synaptic density-95.

Spinophilin also interacts with Group I mGluRs in the hippocampus to regulate receptor trafficking and signaling involved in mGluR-mediated synaptic plasticity. Binding of spinophilin to the C-terminal PDZ domain of mGluRs and subsequent prevention of endocytosis is necessary for mGluR activity, as chemically (DHPG)-induced LTD mediated by mGluR5 (Palmer, Irving, Seabrook, Jane, & Collingridge, 1997) is impaired in spinophilin knock-out mice (Di Sebastiano, Fahim, Dunn, Walther, Ribeiro, Cregan, et al., 2016). However, acute peptide disruption of PP-1 interaction with spinophilin did not inhibit mGluR-LTD (Morishita et al., 2001). Tyrosine dephosphorylation of GluA2 also plays a role in internalization of AMPARs during DHPG-LTD (Gladding, Collett, Jia, Bashir, Collingridge, & Molnar, 2009; Collingridge, Peineau, Howland, & Wang, 2010). The specific roles of various regulatory subunits of PP-1 in synaptic plasticity thus require further study.

Unlike LTD, inhibition of PP-1 activity occurs in NMDAR-LTP (Fig. 3; Mulkey et al., 1993; Mulkey et al., 1994). Large depolarizations lead to activation of kinases such as CaMKII following influx of Ca²⁺ (Pettit, Perlman, & Malinow, 1994). However, CaMKII also activates enzymes such as adenylyl cyclase and calcineurin with opposing actions that determine the phosphorylation state of inhibitor-1(I-1), a potent inhibitor of PP-1 activity and of LTP. Calcineurin, by dephosphorylating and inactivating I-1, allows PP-1 to disrupt CaMKII signaling and thus LTP (Mulkey et al., 1994). Alternatively, adenylyl cyclase antagonizes effects of calcineurin by producing cAMP and activating PKA, which inhibits PP-1 by phosphorylating I-1 (Blitzer, Wong, Nouranifar, Iyengar, & Landau, 1995). Thus, regulation of these opposing enzyme activities (Blitzer et al., 1995, 1998) is necessary for persistent strengthening of synapses underlying learning and memory (Bliss & Collingridge, 1993).

In addition to spinophilin, the neurabin-PP-1 complex also modulates synaptic transmission in the hippocampus, further solidifying the role of PP-1 in synaptic plasticity. For example, in rat CA1 pyramidal neurons, mutant neurabins that are unable to bind to either PP-1 or F-actin enhance LTP and inhibit LTD (Hu, Huang, Roadcap, Shenolikar, & Xia, 2006), while wild-type neurabin inhibits LTP and elicits more robust LTD (Hu et al., 2006). However, genetic deletion of neurabin reduces LTP and hippocampus-dependent fear memory suggesting that developmental modulation by neurabin of synaptic plasticity may differ from acute knockdown (Wu, Ren, Wang, Kim, Cao, & Zhuo, 2008).

Striatal-dependent plasticity

Principal activity in the striatum is largely controlled by glutamatergic synapses onto medium spiny neurons (Calabresi, Pisani, Mercuri, & Bernardi, 1996). High-frequency stimulation (HFS) or negative spike timing of this pathway leads to LTD of excitatory synaptic transmission (Calabresi, Maj, Pisani, Mercuri, & Bernardi, 1992; Lovinger, Tyler, & Merritt, 1993). Unlike plasticity mechanisms in the hippocampus, striatal LTD does not require glutamate receptor trafficking and is functionally expressed as a reduction in the probability of glutamate release (Choi & Lovinger, 1997) mediated by retrograde endocannabinoid (eCB) signaling (Gerdeman, Ronesi, & Lovinger, 2002). Postsynaptic activation of mGluRs and L-type calcium channels lead to production and release of endocannabinoids that cross the synaptic cleft and activate presynaptic CB1 receptors (Sung, Choi, & Lovinger, 2001; Adermark & Lovinger, 2007).

Dopamine plays a prominent neuromodulatory role in regulating fast synaptic transmission in the striatum associated with motor control and psychological function. Successful induction of eCB-LTD requires activation of dopamine D1 or D2 receptors, depending on striatal pathway (Calabresi et al., 1992; Shen, Flajolet, Greengard, & Surmeier, 2008). D1 and D2 receptors exhibit opposing effects on adenylyl cyclase to regulate intracellular levels of cAMP and in turn, PKA activity (Stoof & Kebabian, 1981). Phosphorylation by PKA in response to D1 receptor stimulation converts DARPP-32 into a potent inhibitor of PP-1, removing antagonism of downstream PKA signaling involved in ion channel activity, neurotransmitter release, and locomotor activity (Fienberg, Hiroi, Mermelstein, Song, Snyder, Nishi et al., 1998; Feng et al., 2000). LTD and LTP are absent in DARPP-32 knock-out mice and are restored by pharmacological inhibition of PP-1, suggesting a critical involvement of the DARPP-2/PP-1 cascade in corticostriatal synaptic plasticity (Calabresi, Gubellini, Centonze, Picconi, Bernardi, Chegui, et al., 2000). PP-1 mediated regulation in neuromodulatory signal transduction is further revealed in mice lacking spinophilin; genetic deletion of spinophilin reduces striatal PP-1 levels and prevents LTD in a dopamine-dependent manner (Allen et al., 2006). Spinophilin serves to promote PP-1 activity toward substrates critical for synaptic depression, but specific targets are unclear. Although AMPAR trafficking is not associated with eCB-LTD, regulation of glutamate receptors by phosphorylation may still play a role as AMPAR-mediated membrane depolarization is needed to open voltage-gated calcium channels crucial for LTD induction (Wang, Kai, Day, Ronesi, Yin, Ding, et al., 2006). In contrast, striatal LTP does require NMDA receptor activation (Calabresi, Pisani, Mercuri, & Bernadi, 1992), and dopaminergic neurotransmission regulates a variety of substrates in a DARPP-32/PP-1-dependent manner (Calabresi et al., 2000) including NMDA receptor function (Calabresi et al., 1992; Flores-Hernandez, Cepeda, Hernandez-Echeagaray, Calvert, Jokel, Fienberg, et al., 2002). The absence of LTP in mice lacking DARPP-32 or D1 receptors suggests inhibition of dephosphorylation by the D1/cAMP/PKA/DARPP-32/PP-1 pathway along with activation of kinases leading to increased phosphorylation and enhancement of NMDA currents (Calabresi et al., 2000; Flores-Hernandez et al, 2002; Centonze, Grande, Saulle, Martin, Gubellin, Pavon et al., 2003). Collectively, these findings confirm the role of dopamine transmission in both forms of striatal plasticity, but demonstrate involvement of distinct signaling cascades. Moreover, despite similar PP-1 interactions, spinophilin and neurabin are required for distinct, but converging aspects of dopaminergic neurotransmission. Deletion of spinophilin or neurabin results specifically in loss of LTD or LTP induction, respectively, but are both rescued by exogenous application of dopamine receptor agonists (Allen et al., 2006).

Behavior

Learning and memory

Potentiation of LTD by PP-1 emphasizes its physiological importance as a suppressor of learning and memory, as well as a key factor in age-related cognitive decline. Repetition and practice in learning are required for formation of accurate and long-lasting memories. PP-1 activation limits acquisition and favors memory decline. Genetic deletion of PP-1 in mutant mice expressing a constitutively active form of inhibitor-1 decreases the time and number of training episodes for optimal performance during a spatial memory task (Genoux et al., 2002). Biochemically, this effect correlates with increased phosphorylation of both CaMKII (Thr-286) and GluA1 (Ser-845), two known PP-1 targets (Genoux et al., 2002). Furthermore, inhibition of PP-1 prolongs memory when induced after learning, suggesting that PP-1 also promotes forgetting, particularly in aging animals (Genoux et al., 2002).

Drug addiction

Synaptic plasticity mechanisms, such as LTP, are physiological correlates of learning and memory. Memory consolidation is a dynamic process, and consolidated memory can become destabilized under various conditions or treatments (Nader, Schafe, & Le Doux, 2000; Nader, 2003). The NMDA receptor subunits NR2B and NR2A have been reported to be necessary for memory destabilization and re-stabilization, respectively (Milton, Merlo, Ratano, Gregory, Dumbreck, & Everitt, 2013). Other studies have identified roles for AMPA receptor subunit GluA2 and PP-1 in memory stabilization (Yu et al, 2016). Drug addiction is a process of strengthened learning, and re-instatement of drug motivation is reinforced when encountered with drug-related environmental cues (Thanos, Bermeo, Wang, & Volknow, 2009; Tang & Dani, 2009). Destabilization of drug memory is mechanistically consistent with a form of LTD; memory retrieval induces activation of NR2B-containing NMDA receptors resulting in Ca²⁺ influx and activation of PP-1, leading to dephosphorylation of GluA1 and subsequent endocytosis of synaptic AMPA receptors (Yu et al, 2016). In contrast, cellular adaptations that accompany drug sensitization and addiction include up-regulation of AMPA receptor surface expression in the nucleus accumbens (Berke & Hyman, 2000; Boudreau & Wolf, 2005). Corticostriatal LTP is a correlate of reward-related learning (Reynolds, Hyland, & Wickens, 2001), and enhancement of the rewarding effects of cocaine seen in spinophilin knockout mice may be related to preferential expression of LTP over LTD at cortico-striatal synapses (Allen et al., 2006). Progressive increases in locomotor activity is a hallmark effect of consecutive drug administrations associated with increased dopaminergic neurotransmission (Kalivas & Duffy, 1990; Henry & White, 1991) and LTP induction (Borgland, Malenka, & Bonci, 2004). These drug-induced changes in behavioral sensitization paradigms are also influenced by scaffolding proteins. Standard immunocytochemistry and cell-type specific proteomic screening show increased striatal levels of spinophilin following psychostimulant administration (Boikess & Marshall, 2008; Watkins, True, Mosley, & Baucum II, 2018), enhancing its interactions with various proteins involved in synaptic signaling (Watkins et al., 2018) and concomitantly leading to more dendritic spines (Li, Kolb, & Robinson, 2003; Lee, Kim, Kim, Helmin, Nairn, & Greengard, 2006). These alterations indicate greater targeting of PP-1 to its many substrates and greater regulation of its activity. Behaviorally, deletion of spinophilin has no effect on the acute locomotor responses to cocaine, but attenuates cocaine-induced locomotor sensitization compared to wild-type controls (Areal, Hamilton, Martins-Silva, Pires, & Ferguson, 2019). Similar findings were also reported in spinophilin-deficient mice administered amphetamine (Morris, Watkins, Salek, Edler, & Baucum II, 2018). Importantly, these results may be protocol-dependent as a shorter observation time and higher cocaine dosage in mice lacking spinophilin or neurabin did lead to acute hyperactivity with preserved locomotor sensitization (Allen et al., 2006).

Protein phosphatase-1 and neurodegenerative disorders

Disorders such as Alzheimer’s Disease (AD), Parkinson’s Disease (PD), and schizophrenia are correlated to dysfunction in learning and memory, and are characterized by functional and structural changes in dendritic spines. This implicates dysregulation of PP-1 as a candidate mechanism in the pathophysiology of various neurodegenerative and psychiatric disorders. Although few human diseases are reported to be caused by defects in PP-1 per se, neurodegenerative diseases are correlated with dysregulation of protein phosphorylation and/or decreased availability and function of PP-1 regulatory/targeting subunits.

Alzheimer’s disease is a progressive, degenerative disorder characterized by the presence of extracellular amyloid-beta (Aβ) plaques and intracellular neurofibrillary tangles (NFT; Terry, Gonatas, & Weiss, 1964; Braak & Braak, 1991). Alterations in spine number and size, caused by oligomeric forms of the Aβ peptide, precede these hallmark changes and are directly associated with early memory loss (Masliah, Terry, DeTeresa, & Hansen, 1989; Selkoe, 2002) and severity of disease (Terry, Masliah, Salmon, Butters, DeTeresa, Hill et al., 1991). Aβ is proteolytically cleaved from the amyloid precursor protein (APP), a process regulated by PKC (Rebelo, Vieira, Esselmann, Wiltfang, da Cruz e Silva, & da Cruz e Silva, 2007) and PP-1 (da Cruz e Silva, da Cruz e Silva, Zaia, & Greengard, 1995). The activity of different isoforms of PP-1 is specifically and potently inhibited by Aβ both, an effect that is further potentiated by Aβ aggregation (Vintém, Henriques, da Cruz e Silva, & da Cruz e Silva, 2009). Thus, concomitant activation of PKC and inhibition of PP-1 may contribute to hyperphosphorylation of key proteins associated with AD pathology such as APP and tau (Bennecib, Gong, Grundke-Iqbal, & Iqbal, 2000). Moreover, LTP impairment induced by Aβ oligomers can be reversed by PP-1 inhibition in vitro (Knobloch, Farinelli, Konietzko, Nitsch, & Mansuy, 2007). Taken together, increased Aβ production can lead to abnormal PP-1 activity and contribute to AD pathology by mediating reduced synaptic contacts, LTD (Morishita et al., 2001) and age-related memory defects (Genoux et al., 2002).

PP-1 regulatory proteins are also affected by AD. Human AD patients show a marked reduction in spinophilin-labeled hippocampal dendritic spines, which correlates directly with cognitive dysfunction as assessed by both a global neuropsychological measure (MMSE) and a dementia severity scale (CDR; Akram, Christoffel, Rocher, Bouras, Kovari, Perl et al., 2008). Moreover, lower spinophilin-reactive dendritic spine numbers also correlate with greater Aβ plaque burden (Mi, Abrahamson, Ryu, Fish, Sweet, Mufson et al., 2017).

In both human and animal models of PD, there is a specific loss of dopamine neurons that project from the substantia nigra to the striatum leading to progressive motor deficits (Calabresi, Mercuri, Sancesario, & Bernardi, 1993; Stephens, Mueller, Shering, Hood, Taggart, Arbuthnott et al., 2005). These neurons synapse with dendritic spines in the striatal medium spiny neurons (MSNs) predominantly onto the necks of spines (Freund, Powell, & Smith, 1984) whereas excitatory synaptic input from the cortex and thalamus synapse primarily onto spine heads (Bolam, Hanley, Booth, & Bevan, 2000). Loss of dopamine signaling leads to a significant reduction in MSN spine density (Ingham, Hood, van Maldegem, Weenink, & Arbuthnott, 1993; Villalba, Lee, & Smith, 2009), glutamatergic synapses (Day, Wang, Ding, An, Ingham, Shering et al., 2006) and synaptic plasticity, particularly LTD (Calabresi et al., 1992; Calabresi et al., 1993). Dysregulation of kinase and phosphatase activities are observed in animal models of PD including reduction of PP-1γ1 activity in striatal MSNs (Brown, Baucum, Bass, & Colbran, 2008). Decreased activity of PP-1 occurs concurrently with increased association between PP-1 and spinophilin (Brown et al., 2008), consistent with spinophilin’s ability to inhibit PP-1 activity towards specific substrates (Ragusa, Dancheck, Critton, Nairn, Page, & Peti, 2010), which influences spine density (Feng et al., 2000) and synaptic plasticity (Allen et al., 2006). Moreover, proteomics-based assays of dopamine depletion in a PD mouse model observed overall reduction in spinophilin association with various substrates including synaptic signaling, cytoskeletal and scaffolding proteins (Hiday, Edler, Salek, Morris, Thang, Rentz et al., 2017).Decreased spinophilin association with possible PP-1 targets may further lead to reduced PP-1 activity. Attenuated PP-1 activity leads to less dephosphorylation and unopposed kinase activity, and hyperphosphorylation of glutamate subunits (NR2A, NR2B) by CaMKII induces motor disability and maladaptive striatal plasticity in a PD mouse model (Picconi, Gardoni, Centonze, Mauceri, Cenci, Bernardi et al., 2004; Brown, Deutch, & Colbran, 2005). Spinophilin-dependent PP-1 targeting and activity may underlie these motor effects as spinophilin-deficient mice have deficits in motor behaviors and motor learning (Edler, Salek, Watkins, Kaur, Morris, Yamamoto et al., 2018).

Protein phosphatase-1 and neuropsychiatric disorders

Neuropsychiatric conditions such as schizophrenia obsessive compulsive disorder (OCD), and mental retardation alter synaptic and dendritic spine function, making PP-1 a likely factor in their pathophysiology. Schizophrenia is a neurodevelopmental disorder described by impairments in perception, affect, and cognition, reflecting altered connectivity within and between multiple brain regions. Structural imaging studies have shown smaller whole brain volumes (Levitt, Bobrow, Lucia, & Srinivasan, 2010) and reductions in grey matter underlying smaller cortical and hippocampal volumes (Glahn, Laird, Ellison-Wright, Thelen, Robinson, Lancaster et al., 2008). Electron microscopy of postmortem brain shows synapse-and region-specific effects on spine density; consistent with the imaging studies, spine density was reduced in multiple neocortical areas (Garey, Ong, Patel, Kanani, Davis, Mortimer et al., 1998), while hippocampal (Rosoklija, Toomayan, Ellis, Keilp, Mann, Latov et al., 2000; Kolomeets, Orlovskaya, Rachmanova, & Uranova, 2005) and striatal (Roberts, Conley, Kung, Peretti, & Chute, 1996; Roberts, Roche, & Conley, 2005) spines show variable effects based on sub-regions or type of synapses. Changes in spinophilin expression may contribute to these differences as genetic deletion of spinophilin increases the number of spines in the caudatoputamen of young mice, while its acute knockdown reduces spine number in hippocampal cultures (Evans et al., 2015).

Additional evidence that disparate spinophilin regulation in schizophrenia may be brain region-specific is observed at the transcript level. Measurements of spinophilin mRNA show increased or unchanged levels in dorsolateral prefrontal cortex (Weickert, Straub, McClintock, Matsumoto, Hashimoto, Hyde et al., 2004; Barackskay, Haroutunian, & Meador-Woodruff, 2006), whereas spinophilin transcript levels are increased in several sub-regions of the hippocampus (Law, Weickert, Hyde, Kleinman, & Harrison, 2005). PP-1 and spinophilin are also implicated in OCD, a disease that is characterized by persistent intrusive thoughts, repetitive actions, and excessive anxiety. Loss of the scaffolding protein SAP90/PSD95-associated protein 3 (SAPAP3) in rodent models leads to OCD-like behaviors such as excessive grooming due to increased AMPAR internalization via mGluR5 signaling (Welch, Lu, Rodriguiz, Trotta, Peca, Ding et al., 2007; Chen, Wan, Ade, Ting, Feng, & Calakos, 2011; Wan, Feng, & Calakos, 2011). Spinophilin is associated with both SAPAP3 and mGluR5 in mouse brain (Morris et al., 2018). Association of spinophilin with mGluR5 is enhanced by SAPAP3, while mGluR5 increases association of spinophilin and SAPAP3 (Morris et al., 2018). Normally, spinophilin and SAPAP3 regulate anchoring and trafficking of mGluR5 leading to silencing or potentiation of AMPA signaling, respectively (Wan et al., 2011; di Sebastiano et al., 2016). However, aberrant regulation of the spinophilin/mGluR5/SAPAP3 complex may underlie OCD pathology as loss of spinophilin enhances mGluR endocytosis (di Sebastiano et al., 2016), decreases reduction of AMPA signaling (Allen et al., 2006), and attenuates behaviors associated with OCD (Morris et al., 2018). Moreover, a role for spinophilin/PP-1 interaction is implicated in OCD as overexpression of the catalytic subunit of PP-1α abolishes the association of spinophilin with SAPAP3 and mGluR5 (Morris et al., 2018).

Several mental retardation disorders are caused by mutations in the gene encoding ATRX, a chromatin remodeling protein. A milder form of this disorder has been recapitulated in ATRX^ΔE2 mice, mimicking an exon 2 mutation seen in patients that causes learning and memory dysfunction. Mental retardation is associated with changes in dendrite arborization and dendrite spine morphology (Kaufmann & Moser, 2000). ATRX^ΔE2 mice have longer and thinner spines in the medial prefrontal cortex compared to wild-type mice, which is mediated by abnormal elevation of CaMKII activity due to reduced expression and activity of PP-1 (Shioda, Beppu, Fukuda, Li, Kitajima, & Fukunaga, 2011). CaMKII localized in the rat forebrain PSD fraction is predominantly dephosphorylated by PP-1 (Strack, Barban, Wadzinski, & Colbran, 1997), and inhibition of PP-1 in cultured hippocampal or cortical neurons enhances autophosphorylation of CaMKII and spine elongation (Jourdain, Fukunaga, & Muller, 2003) by spinophilin targeting. Disrupting the spinophilin/PP-1 complex in vitro enhances spine elongation and filopodia in cultured HEK293 cells (Oliver et al., 2002) and dissociated rat hippocampal neurons (Terry-Lorenzo, et al., 2005). Similarly, decreased PP-1 activity correlates with increased phosphorylation of hippocampal CaMKII in the PSD with disruption of synaptic plasticity, learning, and memory in a mouse model of Angelman syndrome (Weeber, Jiang, Elgersma, Varga, Carrasquillo, Brown et al., 2003).

Synaptic dysfunction is implicated in affective disorders such as anxiety and depression, and both neurabin and spinophilin play distinct age-dependent roles in their regulation. Loss of neurabin by genetic deletion is anxiolytic in young adult mice compared to wild-type controls (Kim, Wang, Li, Chen, Mercaldo, Descalzi et al., 2011; Wu, Cottingham, Chen, Wang, Che, Liu et al., 2017). In contrast, there is reduced anxiety-like behavior in middle-aged spinophilin knock-out mice, whereas the anxiolytic effect of neurabin loss is absent in older animals. However, attenuation of anxiety-related behaviors in neurabin or spinophilin knock-out mice was not consistent across different types of behavioral testing, suggesting selective sensitivity to particular anxiogenic conditions (Kim et al., 2011; Wu et al., 2017). Opposing age-dependent regulation of depression-like behavior were also seen in both knock-out strains as loss of spinophilin increased behaviors of depression in young adult mice (Wu et al., 2017), while conflicting findings were reported in mice lacking neurabin (Kim et al., 2011; Wu et al., 2017). In contrast, loss of spinophilin provided protection from depression-like behavior in aged mice (Wu et al., 2017). These distinct phenotypes are of interest as overall protein expression levels did not change in these knock-out mouse models (Wu et al., 2017), suggesting an interplay of region-specific protein expression, age, and differences in synaptic plasticity. Selective inhibition of neurabin in the anterior cingulate cortex using siRNA reproduced anxiety-like behavior without affecting depression-like behavior (Kim et al., 2011); similar studies will be necessary to exclude developmental abnormalities or compensation.

Together these data describe mechanisms modulating the interactions between PP-1 and its regulatory proteins involved in neuropsychiatric disease pathology. Understanding how PP-1 regulates the phosphorylation status of these substrates in vivo and in specific cell types promises to provide insights into disease mechanisms, therapeutic treatments, and prevention.

Conclusion

Aberrant PP-1 regulation has been linked to several pathological states, making pharmacological manipulation of phosphatase activity or levels an attractive therapeutic strategy. Drugs that target specific phosphatase complexes may provide useful approaches with target specificity. For example, deletion of the C terminus of the PP-1 catalytic subunit prevents PP-1 inhibition by endogenous inhibitors (Conner, Kleeman, Barik, Honkanen, & Shenolikar, 1999), identifying this region as an appealing target for the design of novel, selective inhibitors that prevent substrate access. Disorders that are treated by protein kinase inhibitors may also be responsive to phosphatase activators. Synthetic small molecule peptides can activate PP-1 in the absence or presence of endogenous inhibitors, indicating that binding of these peptides to the regulatory domain of PP-1 can directly influence phosphatase activity as well as its interaction with regulatory proteins (Tappan & Chamberlin, 2008).

The ability to target a specific phosphatase complex directly as opposed to overall PP-1 activity is an appealing approach to prevent unwanted toxic side effects of broad catalytic subunit inhibitors. Methods to accomplish this include disruption of interactions between the phosphatase (catalytic or regulatory subunit) and its substrates and disruption of interactions between catalytic and regulatory subunits (McConnell & Wadzinski, 2009). There are several examples of molecules that affect multiprotein complexes (Vassilev, Vu, Graves, Carvajal, Podlaski, Filipovic et al., 2004), including phosphatase complexes (McConnell & Wadzinski, 2009), to impair phosphatase-substrate interaction. Targeted dissociation of the catalytic subunit from the associated regulatory subunit can also influence the function of PP-1 holoenzymes. For example, introduction of synthetic peptides that disrupt PP-1 complexes has been shown to influence synaptic transmission (Yan et al., 1999; Morishita et al., 2001), apoptosis (Boyce, Bryant, Jousse, Long, Harding, Scheuener et al., 2005), and cancer progression (Chen, Weng, Tseng, Lin, & Chen, 2005). Thus, synthetic or natural compounds that regulate PP-1 activity have a broad range of potential therapeutic uses.