Information is encoded and stored in response to changes in neural activity and Ca²⁺ signaling in circuits underlying memory formation in the brain. One molecule critical for these processes is CaMKII whose activity is acutely sensitive to frequency dependent changes in Ca²⁺ during long-term potentiation (LTP) during hippocampal memory formation (Lisman et al., 2002; Hell, 2014). In addition CaMKII is abundant in structures known to be required for memory and is localized specifically to the parts of the neuron important for memory formation. For instance, CaMKII is the main protein in the hippocampal post-synaptic density (PSD) (Kelly et al., 1984; Hell, 2014) and is similarly enriched in the mushroom body memory center of Drosophila (Takamatsu et al., 2003; Hodge et al., 2006). CaMKII is thought to act as a molecular memory switch during increased neuronal activity, when increased Ca²⁺ levels stimulate CaMKII autophosphorylation inducing the changes in synaptic strength that underlie learning. This occurs because the increased Ca²⁺/CaM binds to a subunit of the CaMKII dodecamer causing a conformational change exposing a T286 on mammalian CaMKII and T287 on Drosophila CaMKII that can be autophosphorylated (Figure 1A), resulting in a Ca²⁺ independent constitutively active kinase (Lisman and Zhabotinsky, 2001). CaMKII knockout mice or treatment with a CaMKII inhibitor resulted in mice with impaired LTP and memory (Silva et al., 1992a,b). Mice expressing either Ca²⁺ dependent CaMKII-T286A or CaMKII-T286D have abnormal LTP and memory (Mayford et al., 1996; Giese et al., 1998; Yasuda and Mayford, 2006). A second pair of autophosphorylation sites (TT305/6 equivalent to Drosophila TT306/7, Figure 1A); are exposed in the CaM binding domain when Ca²⁺/CaM dissociates from CaMKII, for instance during low synaptic activity, and are inhibitory as autophosphorylation prevents subsequent CaM binding blocking CaMKII function. CaMKII-TT305/6AA mice show enhanced LTP while CaMKII-TT305/6DD expression also disrupted LTP and memory (Elgersma et al., 2002; Pi et al., 2010). In Drosophila, there is no CaMKII null, which would be expected to be lethal (Park et al., 2002; Mehren and Griffith, 2004), however peptide inhibition of CaMKII led to synaptic defects and memory deficits in the courtship-conditioning assay (Griffith et al., 1993, 1994). Therefore, the control of CaMKII and its autophosphorylation is critical for synaptic plasticity and memory in Drosophila and mammals. But the mechanism of regulation of CaMKII autophosphorylation during memory formation is still unclear.

One molecule that in addition to CaM regulates CaMKII autophosphorylation is CASK (Figure 1B), a membrane-associated guanylate kinase (MAGUK) synaptic scaffolding protein that contains a CaMK-like and Lin-2/Lin-7 (L27) domain in addition to the canonical PDZ [post-synaptic density protein (PSD95), Drosophila disc large tumor suppressor (Dlg1), and Zonula occludens-1 protein (Zo-1)], SH3 (SRC Homology 3), and GUK (guanylate kinase) domains with the CaMK and GUK domains likely kinase dead in Drosophila (Hata et al., 1996; Lu et al., 2003). The CaMK domain of mammalian CASK has low levels of Ca²⁺/CaM independent kinase activity against neurexin that unlike other kinases is magnesium independent (Mukherjee et al., 2008; LaConte and Mukherjee, 2013). Drosophila CASK has two isoforms, a full-length CASK-ß isoform that contains the CaMK-like and L27 domains and PDZ, SH3, and GUK domains (Figure 1B). The other isoform, CASK-α, is short and contains only the common PDZ, SH3, and GUK domains and forms a molecule with structural homology to vertebrate MPP (Slawson et al., 2011). CASK-β associates with CaMKII at synapses and in the absence of Ca²⁺/CaM promotes TT306/7 phosphorylation (Figure 1A), inactivating the kinase (Lu et al., 2003). The function of CASK has also been studied in mice, and while CASK knock-outs are lethal due to a cleft palate phenotype, neurons cultured from these animals show abnormalities in glutamatergic synaptic release (Atasoy et al., 2007). However, the early lethality of these mice prevents the modeling of CASK function in behavior and disease.

Flies completely lacking CASK are viable, have decreased levels of synaptic CaMKII-TT306/7 autophosphorylation and display abnormal habituation behavior in a version of the courtship conditioning assay (Lu et al., 2003). Furthermore, CASK mutants increase T287 autophosphorylation thereby allowing CASK to regulate the CaMKII switch to Ca²⁺ independence (Hodge et al., 2006). CASK is expressed throughout the fly brain including the mushroom bodies (Martin and Ollo, 1996; Lu et al., 2003; Malik et al., 2013). Recently a CASK-β mutation that completely removes has been shown to cause a number of cognitive deficits in flies including disrupted sleep and place preference (Slawson et al., 2011; Donelson et al., 2012).

In order to determine the role of CASK and CaMKII autophosphorylation in learning and memory the Drosophila olfactory shock conditioning was used (Tully and Quinn, 1985; Malik et al., 2013; Malik and Hodge, in press). All CASK and CaMKII mutant genotypes learned (equivalent to immediate or 2 min memory) similar to wildtype (Figure 2E). When the flies were tested 3 h after training middle-term memory (MTM), the CASK-β mutant flies that lack just the long isoform of CASK had reduced MTM. This showed that the CaMK-like and L27 domains only present in this form of CASK (Figure 1B) were the key signaling domains required for memory formation. As CASK-β mutations that leave intact PDZ, SH3, and GUK containing CASK-α, wiped out memory to a similar extent as a deficiency that removed all forms of CASK (Malik et al., 2013). Previous work has shown that CASK-β regulates CaMKII autophosphorylation by its CaMK-like domain (Lu et al., 2003; Hodge et al., 2006; Gillespie and Hodge, 2013); therefore, it is likely that CASK functions in memory formation via its control of CaMKII autophosphorylation mediated by its CaMK-like domain. We then tested the effect of mushroom body specific reduction of CASK on learning (Figure 2A). We used a UAS-CASK-RNAi line which reduced the expression of CASK by ~50% (Gillespie and Hodge, 2013; Malik et al., 2013) to test if reduction of CASK in the mushroom body using the Gal4/UAS system affected memory. Expression of CASK-RNAi in either, all mushroom body neurons (OK107-Gal4, Figure 2B), or just α′/β′ neurons (c305a-Gal4, Figure 2C) similarly showed a drastic reduction in MTM (Figure 2E), while expression in the remaining α/β and γ neurons (MB247-Gal4, Figure 2D) had no effect. In order to distinguish the role of CASK in mushroom body development as opposed to an acute physiological role in signaling underlying memory we restricted the reduction of CASK to just the adult mushroom body using Gal80^ts (McGuire et al., 2003; Malik et al., 2013). Reduction of CASK specifically in the adult mushroom body α′/β′ as opposed to the adult α/β and γ neurons was sufficient to cause the reduction in MTM showing the effects were post-developmental.

Decreased levels of CASK are known to increase CaMKII-T287 autophosphorylation (Figure 1A; Lu et al., 2003; Hodge et al., 2006; Gillespie and Hodge, 2013). Consistent with this, we found that direct overexpression of CaMKII-T287D in the α′/β′ neurons caused a similar reduction in MTM as knocking-down CASK in the same neurons (Figure 2E), while CaMKII-T287D expression in the α/β and γ neurons had no effect. Expression of UAS-CASK-β just in the α′/β′ neurons of the CASK-β null flies fully rescued the MTM defect to a level indistinguishable from wildtype, confirming that CASK signaling in the mushroom body α′/β′ is necessary and sufficient for Drosophila MTM formation (Malik et al., 2013).

We also determined for the effect of CaMKII overexpression on memory, showing α′/β′ neuron expression completely removed MTM. In addition reduction of CASK just in neurons that express CaMKII (using a CaMKII-specific Gal4 promoter) was sufficient to remove MTM (Malik et al., 2013). Furthermore, increasing CASK in α′/β′ neurons also greatly reduced MTM. Such increases in CASK would be expected to block T287 autophosphorylation (Hodge et al., 2006), consistent with this idea we found α′/β′ neuron T287A overexpression gave a similar MTM phenotype (Figure 2E). The role of CaMKII-T287 autophosphorylation in the memory neurons is an acute physiological one as opposed to a developmental one, as changing CaMKII-T287 autophosphorylation just in the adult α′/β′ neurons was sufficient to remove memory (Malik et al., 2013). We found α′/β′ neuron overexpression of CaMKII-TT306/7AA removed MTM and overexpression of CASK completely rescued the memory deficit due to mushroom body overexpression of CaMKII-T306/7AA (Malik et al., 2013). Therefore, our data suggests that CASK regulates CaMKII autophosphorylation during memory in the mushroom body α′/β′ neurons. A role of mushroom body α′/β′ neurons in memory consolidation has previously been proposed (Krashes et al., 2007) however the molecular pathways involved remain largely unknown.

Previous work has shown mushroom body overexpression of CaMKII or CaMKII-T287D enhanced training but did not affect memory in the courtship conditioning assay, while CaMKII-T287A overexpression changed habituation and neuronal excitability, but resulted in no change in courtship conditioning memory (Mehren and Griffith, 2004, 2006). However, mushroom body expression of the CaMKII-RNAi transgene has been shown to decrease long-term memory (LTM) using the olfactory shock assay (Ashraf et al., 2006) and was associated with decreased mushroom body Ca²⁺ signaling (Akalal et al., 2010). The differences in effects of CaMKII on courtship and olfactory-shock learning phenotypes maybe due to differences in the circuitry employed in the two memory tasks along with the timing of memory measured in the two assays. Recently CaMKII has been shown to undergo CREB-dependent gene transcription and translation in mushroom body and dorsal anterior lateral (DAL) neurons during LTM (Chen et al., 2012). Consistent with these studies we also showed mushroom body expression of CaMKII-RNAi only affected LTM (Malik et al., 2013). In addition this is the only CASK or CaMKII transgene that gave a memory phenotype when expressed in the α/β or γ neuron, suggesting LTM is particularly sensitive and requires a certain baseline level of CaMKII activity in every type of mushroom body neuron in order to form LTM. This is in contrast to α/β or γ neuron expression of CaMKII-T287D, T287A, and TT306/7AA that had no effect on LTM (Figure 2E), possibly because the endogenous CaMKII in α/β or γ neuron maybe sufficient to support enough of the required autophosphorylation activity to generate LTM. This is in contrast to the critical role of α′/β′ neurons that required the correct level of CASK, CaMKII, and CaMKII autophosphorylation in order to form LTM (Figure 2E). Therefore, our data is consistent with the other studies showing α/β or γ neuron expression of CaMKII-RNAi disrupts LTM and decreased the peak GCaMP Ca²⁺ response, however it should be noted neither study tested a role of α′/β′ neurons (Ashraf et al., 2006; Akalal et al., 2010).

We also measured a reduction in peak GCaMP Ca²⁺ response in the α′/β′ neurons with CaMKII-RNAi (Figure 3); however this was also never tested for in the previous studies. We also found that the reciprocal CaMKII overexpression caused a large increase in peak Ca²⁺ response. Previous electrophysiological studies have shown neuronal expression of CASK-RNAi or CaMKII-T287D both decreased neural excitability in response to stimulation (Chen and Featherstone, 2011). Likewise we find expression of these transgenes caused a reduction in α′/β′ peak Ca²⁺ signaling. Therefore, the GCaMP data is consistent with the current model of CASK regulation of CaMKII autophosphorylation (Figure 1A; Lu et al., 2003; Hodge et al., 2006) and reveals the likely neurophysiological basis for the disruption of memory resulting from CASK and CaMKII transgene expression in the α′/β′ neurons.

Flies with the CASK-β null mutation or reduced CASK in the α′/β′ neurons also reduced LTM (Figure 2E). The LTM effects of CASK could be explained by its role in transcriptional activation of various plasticity molecules including NMDA receptors (Huang and Hsueh, 2009), as NMDA receptors having been shown to be required for Drosophila LTM (Wu et al., 2007). We also showed that the CaMKII molecular memory switch (pT287) is required for mushroom body α′/β′ LTM formation with phospho-mimic or block removing LTM (Figure 2E). Again this seems to be an evolutionarily conserved memory mechanism with T286 mutant mice also not being able to form LTM (Irvine et al., 2011).

Point mutations in human CASK have been associated with neurological and cognitive defects particularly involving the cerebellum, including severe learning difficulties resulting from mutations in the CaMK-like and SH3 domains (Najm et al., 2008; Piluso et al., 2009; Tarpey et al., 2009). Drosophila and human CASK (74% identical) and CaMKII (79% identical) are remarkably conserved at the protein level, suggesting that they might function in a similar way in both organisms (Cho et al., 1991; Hsueh, 2006). Therefore, we created UAS-human CASK flies and expressed the transgene in the α′/β′ neurons of flies that otherwise express no CASK-β. Whereas CASK-β nulls completely lacked MTM, expression of human CASK just in the α′/β′ neurons returned memory to wildtype levels indicating that Drosophila and human CASK show conserved neuronal function in memory formation (Malik et al., 2013). We also found that human CASK could regulate Drosophila CaMKII autophosphorylation at synapses (Gillespie and Hodge, 2013). Therefore, this work validates the use of Drosophila to study CASK and CaMKII in the healthy brain and in disease suggesting a number of directions where this research might lead (Table 1).