Mouse CaMKIγ cDNA was subcloned into pEGFPC1 (BD Clontech) and pcDNA3 (Invitrogen), and mutants were generated by stepwise substitution of each codon by site-directed mutagenesis. GODZ cDNA (a kind gift from Dr. Masayoshi Mishina, University of Tokyo, Japan) (Uemura et al., 2002) was subcloned into pcDNA3-HA (Takemoto-Kimura et al., 2003). pEGFP-V12Rac1 was generated by site-directed mutagenesis from pEGFP-N17Rac1 (a kind gift from Dr. Shuh Narumiya, Kyoto University, Japan), and pcDNA3-Flag-PHnTSS STEF (N-terminal PH domain and Tiam-STEF-SIF homologous domain as a dominant negative for STEF/Tiam2) was generated as described previously (Matsuo et al., 2002).

We used commercial rat and mouse antibodies: anti-GFP (Nacalai), anti-HA (Roche Diagnostics, Cell Signaling), anti-Flotillin-1, anti-Caveolin-1, anti-Rac1, anti-GM130 (BD), horseradish peroxidase-conjugated IgG (Amersham Biosciences), and Alexa 488-, Alexa 555-, and Alexa 594-conjugated IgG (Molecular Probes).

COS-7 cells were maintained in Dulbecco's modified Eagle medium (DMEM), containing 10% heat-inactivated fetal calf serum (FCS). PC12 cells were grown in DMEM containing 10% heat-inactivated horse serum (HS) and 5% heat-inactivated FCS using a PRIMARIA culture dish (BD Falcon).

Mouse hippocampal neurons were prepared from P0 ICR mice, as described previously (Bito et al., 1996). Dissociated cells were plated onto 12-mm Matrigel-coated coverslips, transfected with each expression vector using Lipofectamine 2000 (Invitrogen) at 7 days in vitro, and fixed 2–3 days later for immunostaining. Neurons were treated with either 60 μM 2-bromopalmitate (Sigma) overnight or 10 mM methyl-β-cyclodextrin (Sigma) for 20 min before fixation (Figures 1C, D).

Cells were fixed in 4% paraformaldehyde/4% sucrose/phosphate-buffered saline (PBS)(–) for 20 min and washed with 0.1 M glycine/PBS(–). Immunostaining was carried out as described (Bito et al., 1996; Takemoto-Kimura et al., 2003, 2007; Ageta-Ishihara et al., 2009). Fluorescence imaging was performed using scanning laser confocal microscopes (LSM 510 and LSM900 with Airyscan 2, Zeiss) with 63× and 40× objective lenses (Plan-Apochromat 63× /NA 1.4, oil; Plan-Neofluar 40× /NA 1.3, oil, Zeiss) and wide-field microscopes (BX-51, Olympus) with 20× objective lenses (UplanFI 20× /NA 0.5, air, Olympus) and a color CCD camera (DP-70, Olympus). Projected images of confocal sections are shown, but occasionally, for clear separation of membrane- and/or Golgi-fluorescent signals, single confocal sections are shown (Figures 1B–E). Fluorescence intensity was measured using Fiji (ImageJ). The localization of CaMKIγ to the Golgi was calculated as follows: (%) = (the summation of the GFP-CaMKIγ signal in GM130 (a marker of the Golgi-positive area)/(the summation of the GFP-CaMKIγ signal in the soma area) × 100).

For co-immunoprecipitation experiments, COS-7 cells were transfected with wild-type and mutant GFP-CaMKIγ vectors and wild-type HA-CaMKIγ using the Fugene6 reagent (Roche Diagnostics). Cell lysates were prepared 36 h after transfection, and immunoprecipitation was performed as follows. Cells were washed twice in PBS(–), lysed with lysis buffer [50 mM Tris-HCl pH 7.5, 100 mM NaCl, 2 mM MgCl₂, 10% Glycerol, 1% Triton X-100, and protease inhibitors (Roche Diagnostics)], and mixed with protein A Sepharose (Amersham Biosciences) and GFP antibody (Molecular Probes). Each sample was mixed with an equal amount of 4× sample buffer (0.2 M Tris-HCl pH 6.8, 8% SDS, 40% glycerol, 40% 2-mercaptoethanol, and 0.02% bromophenol blue) and boiled for 3 min and was subjected to Western blot analysis.

Detergent-insoluble membrane fractions were obtained according to a previously described method (Morishima-Kawashima and Ihara, 1998; Takemoto-Kimura et al., 2007). Briefly, COS-7 cells cultured in three 10-cm dishes were transfected with the expression vectors, harvested 2 days later, and homogenized in 2.5 ml of MES-buffered saline (MBS; pH 6.5), containing 1% Triton X-100 and protease inhibitors (Roche Diagnostics). These obtained lysates were adjusted to a final 40% sucrose concentration in MBS, and 4 ml of each lysate was placed at the bottom of an ultracentrifuge tube and overlaid with a discontinuous sucrose gradient consisting of 4 ml of 5%/4 ml of 35% in MBS. Ultracentrifugation was carried out at 39,000 rpm for 20 h at 4°C in an SW 41 rotor (Beckman). A total of 12 fractions (1 ml each) and an extract of the resultant pellet were collected along with 15 μl of each fraction.

Western blot analysis was carried out as described (Takemoto-Kimura et al., 2007; Ageta-Ishihara et al., 2009). Films were digitally scanned, and rectangular windows of an identical area were defined around each band of interest. After background subtraction, the average pixel intensity of each region of interest was calculated using ImageJ 1.36 (NIH) and Fiji (ImageJ).

COS-7 cells were transfected with GFP-CaMKIγ or GFP-CaMKIγ-C474S and Myc-CaMKKactive or an empty vector using Lipofectamine 2000. After transfection for 48 h, an immunoprecipitate kinase assay was performed using MBP as a substrate (Takemoto-Kimura et al., 2003). CaMKIγ immunopurification was evaluated by silver staining using a PlusOne Silver Staining Kit (Amersham). CaMKK expression was functionally confirmed by the similar intensity of the phosphorylated CaMKIγ band under extended exposure.

Total RNAs were extracted from the P12 rat brain and PC12 cells using Trizol (Life Technologies). After DNase treatment, cDNAs were synthesized from 2 μg of total RNA using the Omniscript RT Kit (Qiagen). PCR reactions were performed using the following primers:

PC12 cells, previously used for studies of small GTPases (kind gift from Dr. Shuh Narumiya, Kyoto University, Japan), were grown onto poly-Lysine-coated glass coverslips or culture dishes (BD Biosciences).

For a neuritogenic activity assay, wild-type or mutant HA-CaMKIγ vectors were co-transfected with pEGFPC1 as a morphological marker using Optifect reagent (Invitrogen). Alternatively, in experiments employing co-transfection of another signaling molecule, such as GFP-N17Rac1 or Flag-PHnTSS STEF, cell contours were reliably identified using DIC images. Transfected PC12 cells were treated with a high K⁺ medium containing 55 mM KCl at either 4 h or 24 h after transfection, kept for another 24 h, fixed, and immunostained for morphological analyses. We defined neuritogenic activity by counting the percentages of cells bearing at least one neurite with a length that exceeded the larger diameter of the cell body (neurite-positive cells) within many randomly chosen fields of view. We excluded from our analyses all cells in which either one of the co-transfected molecules was not detected. An observer blind to the transfection type counted the number of neurite-positive cells. In each experiment, an average percentage of the neurite-positive cells from two coverslips was obtained for each construct (N > 500 transfected cells per coverslip). The shown results are means ± SEM from three to five independent experiments.

For Rac1 pulldown assays, PC12 cells were transfected with wild-type or mutant HA-CaMKIγ vectors using the Lipofectamine 2000. After 24 h, the cells were stimulated by a high K⁺ medium containing 55 mM KCl for 1 min. Then, the cells were washed once in ice-cold PBS(–) and lysed with lysis buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 1% NP-40, 1 mM DTT, 5% glycerol, and EDTA-free protease inhibitors). The lysates were centrifuged at 16,000 g for 15 min at 4°C, and the supernatant was collected. The protein concentrations were determined using a BCA Protein Assay Reagent Kit (Pierce). Activated Rac1 was detected by a pulldown assay using an EZ-detect Rac1 activation kit (Pierce), according to the manufacturer's protocol. Briefly, SwellGel-immobilized glutathione disks were placed into spin columns together with 20 μg of GST-human Pak1-PBD and cell lysates containing at least 1 mg of proteins. The reaction mixtures were incubated for 1 h at 4°C with gentle rocking. The resin was washed three times in the lysis buffer, followed by the addition of 50 μl of 2× SDS sample buffer and boiling for 5 min. The eluted fractions were collected and subjected to SDS–PAGE and Western blot analysis. To quantify the intensity of a Rac1-immunoreactive band, ECL reaction time and film exposure times were adjusted to maintain the signals within a linear dynamic range of detection.

For comparisons between two groups and more than three groups, Student's t-test and one-way analysis of variance (followed by a post hoc Tukey-Kramer test) were used, respectively [Prism 4.0 and 9.0 (GraphPad Software) and JMP 5.1.2 (SAS Institute)]. The results were represented as the mean ± standard error of means (SEM).

We reasoned that the addition of a palmitoyl moiety to prenylated wild-type CaMKIγ may facilitate the redistribution of CaMKIγ into specific membrane microdomains such as lipid rafts (Anderson and Jacobson, 2002). To substantiate this, we expressed CaMKIγ in COS7 cells with or without GODZ/DHHC3, a Golgi-enriched PAT that does not localize to lipid rafts. Under similar conditions, palmitate incorporation into CaMKIγ was increased by 4.3-fold (Takemoto-Kimura et al., 2007). We thus measured the amount of CaMKIγ in the Triton X-100-insoluble fractions, considered here as lipid rafts. We found that increased palmitoylation by GODZ co-expression significantly augmented the amount of CaMKIγ co-fractionated in the lipid raft fractions, identified using specific markers such as caveolin and flotillin (Figure 1A, top panels). In contrast, a palmitoylation-deficient 4CS (C417S, C419S, C420S, and C423S) mutant (Takemoto-Kimura et al., 2007) was completely absent from the raft fractions irrespective of the absence or presence of GODZ (Figure 1A, bottom panels).

We next investigated the effect of GODZ activity on CaMKIγ localization in cultured hippocampal neurons. The expression of GFP-CaMKIγ revealed a significant Golgi localization (identified using GM130 as a cis-Golgi marker) over a background of relatively diffuse cytoplasmic fluorescence (Figure 1B, –GODZ). Upon additional expression of HA-GODZ, however, an increased amount of GFP-CaMKIγ now appeared localized to the Golgi apparatus and in somatodendritic puncta near the plasma membranes, with a marked diminishment of any diffuse signals (Figure 1B, +GODZ). The Golgi localization of CaMKIγ induced by the expression of GODZ was inhibited by overnight treatment with 2-bromopalmitate (2-Br-Pal), a competitive inhibitor of PAT activity (Figure 1C). Furthermore, cholesterol depletion by methyl-beta-cyclodextrin (MeCD) also diminished the accumulation of CaMKIγ in the Golgi compartment (Figure 1D). These manipulations of palmitoylation and lipid raft states indicated that Golgi accumulation of CaMKIγ depends on both palmitoylation and intact cholesterol content on the Golgi membranes. Interestingly, throughout these manipulations, GODZ completely co-localized with a cis-Golgi marker, GM130, suggesting that GODZ localization to the Golgi is stable and may not depend on either palmitoylation or membrane cholesterol. Consistently, GODZ was not biochemically co-fractioned in the raft fractions either (Figure 1A).

Fluorescent localization of a prenylation-deficient mutant, C474S, which also lacks palmitoylation, shows a completely diffuse and cytosolic distribution devoid of any Golgi apparatus accumulation in accordance with our previous results (Figure 1E; Takemoto-Kimura et al., 2003, 2007). In contrast, the palmitoylation-deficient mutant 4CS, which is prenylated to a similar extent as the WT CaMKIγ (Takemoto-Kimura et al., 2007), showed undistinguishable localization in Golgi as compared with the wild type (Figure 1E). Together, we conclude that the prenylation of CaMKIγ is sufficient to enable its membrane insertion, permitting full access to the Golgi. Further association with a Golgi PAT, GODZ, may then facilitate CaMKIγ's subsequent palmitoylation that leads to its final association with lipid raft-like puncta.

What is the biochemical significance of sequential lipid modifications—prenylation and palmitoylation—and consequential stepwise lipid membrane targeting? We tested whether these lipidation states facilitated the kinase's self-association or its kinase activity itself. Indeed, many signaling molecules localized to lipid rafts form multimers by lipidation (Zacharias et al., 2002; Huang and El-Husseini, 2005). Consistent with this, raft-enriched, HA-tagged CaMKIγ-WT was co-immunoprecipitated with GFP-CaMKIγ-WT. In contrast, non-palmitoylated CaMKIγ mutants, 4CS or non-lipidated C474S, were unable to assemble with GFP-CaMKIγ-WT (Figure 2A). Because the lipid-modified cysteine residues were located close to the regulatory domain of CaMKIγ, we next asked whether the lipidation states might affect its kinase activity. We evaluated the in vitro Ca²⁺/CaM-dependent kinase activity for WT and the lipidation-deficient mutant, C474S. In contrast to a previous finding that a kinase-deficient CaMKIγ had received a lesser amount of palmitoylation (Takemoto-Kimura et al., 2007), we here observed less difference with regard to CaMKK- and Ca²⁺-dependent kinase activity toward a substrate, MBP, between WT and the C474S mutant (Figure 2B). Together, these findings indicate that while the active conformation of CaMKIγ may be structurally beneficial to become a GODZ palmitoylation substrate on the Golgi membranes, the lipidation states may primarily modulate kinase self-association in the lipid rafts and not the kinase activity per se.

Our experiments so far pointed to the possibility that CaMKIγ is redistributed via palmitoylation to specific membrane signalosomes, such as lipid rafts (Anderson and Jacobson, 2002). We previously found that CaMKIγ palmitoylation is necessary for BDNF- and Ca²⁺-regulated dendritogenesis (Takemoto-Kimura et al., 2007). However, we then realized that further clean mechanistic dissection through functional rescue experiments was hampered in primary neurons because BNDF release is strongly stimulated downstream of BDNF-stimulated Ca²⁺ signaling in neurons. Thus, the direct examination of the significance of CaMKIγ palmitoylation on lipid raft signaling will be inevitably contaminated by TrkB-mediated lipid raft signaling triggered upon stimulation (Suzuki et al., 2004). To overcome this, we sought a tangible reconstitution mammalian cell model expressing a large number of neuronal PATs, including DHHC3/GODZ, a known PAT for CaMKIγ (Greaves and Chamberlain, 2011). Messenger RNAs for four neuronal PATs (DHHC3/GODZ, DHHC7, DHHC15, and DHHC17) were previously detected in PC12 cells (Greaves et al., 2008). Furthermore, PC12 cells fulfilled three advantageous conditions: first, numerous reports have previously shown activation of a lipid raft-associated signal complex in PC12 cells (e.g., Scott-Solomon and Kuruvilla, 2020). Second, depolarization by treatment with a 50 mM KCl solution known to induce Ca²⁺ influx via voltage-dependent Ca²⁺ channels (VDCC) (Banno et al., 2008) significantly increased the number of cells with neurites longer than their main cell body in PC12 cells (Figure 3A), enabling a rapid unbiased assay of activity-dependent neurite growth that was independent of any activity-dependent growth factor release. Finally, CaMKK was expressed in PC12 cells, and neurite extension triggered by depolarization was attenuated by inhibition of CaMKK with a specific inhibitor, STO-609 (Figure 3B). Using RT-PCR, we examined which CaMKK isoforms and which CaMKI or IV subtypes were expressed in PC12 cells. We found the expression of CaMKKα, CaMKKβ, CaMKIα, and CaMKIβ, but were unable to detect the amounts of CaMKIγ, δ, and CaMKIV mRNAs in PC12 cells (Figure 3C). These results suggested that Ca²⁺-CaMKK-dependent pathways may regulate activity-dependent neurite extension in PC12 cells through CaMKIγ-independent mechanisms. These experiments suggested that PC12 cells may offer a privileged condition to straightforwardly design and perform functional rescue experiments using distinct lapidated states of CaMKIγ.

We took advantage of the PC12 cells, as these had no detectable levels of CaMKIγ expression. Overexpression of CaMKIγ-WT itself had no stimulatory effect on neurite extension (Figure 4A, left bars). While a 24-h exposure to high K⁺ solution significantly triggered neuritogenesis, the expression of CaMKIγ-WT in combination with a 24 h exposure to high K⁺ doubled the amount of neurite outgrowth compared to non-transfected cells, suggesting that Ca²⁺ signaling via heterologously expressed CaMKIγ can contribute to neurite formation, presumably downstream of endogenous CaMKK (Figure 4A, right bars). In contrast, this high K⁺-induced neurite outgrowth was significantly attenuated in cells expressing a palmitoylation-deficient mutant (4CS) or a prenylation-deficient mutant (C474S; Figure 4A), both of which also lacked palmitoylation. This confirmed that a sizable portion of neurite formation was elicited via reconstituted CaMKIγ signaling to require lipidation that might facilitate lipid raft recruitment and not via a simple augmentation of CaMKIγ kinase activity (Figure 2B).

A robust increase in neuritogenesis in PC12 cells was likely to involve dynamic remodeling of the actin cytoskeleton, and we first tested whether Rac1 small GTPase activity was modulated by high K⁺ stimulation in the CaMKIγ-WT-expressing cells. Indeed, a significant rise in GTP-bound Rac1 was shown (Figure 4B). In sharp contrast, neither the C474S nor the 4CS mutants of CaMKIγ were able to stimulate Rac1 activity under high K⁺ conditions, suggesting that the CaMKIγ-mediated increase in Rac1 was mediated by palmitoylation and required its lipid raft recruitment (Figure 4C). Is Rac1 activation specifically downstream of palmitoylated CaMKIγ, or is it also required when the lipidation state of CaMKIγ is perturbed during high K⁺-triggered neuritogenesis? A dominant negative Rac1 mutant, N17Rac1, that inhibits RacGEF activity, completely abolished all forms of high K⁺-dependent neuritogenesis, suggesting that all forms of activity-dependent neuritogenesis, irrespective of the direct activation mechanism, required intact RacGEF activity (Figure 4D). A contrasting result was found, however, when we tested the overexpression of PHnTSS, a dominant interfering fragment of a lipid raft-associated RacGEF, STEF/Tiam2, which is known to be involved in neurite outgrowth (Hoshino et al., 1999; Matsuo et al., 2002; Kawauchi et al., 2003). While we found less effect of the blocking peptide overexpression on PC12 cells in the absence of CaMKIγ, a significant block of neurite formation was found in PC12 cells expressing palmitoylated wild-type forms of CaMKIγ. Furthermore, the STEF-blocking peptide overexpression showed no significant effect in lipidation-deficient mutant CaMKIγ-expressing cells (Figure 4E). Thus, a palmitoylated CaMKIγ may recruit lipid raft-compartmentalized RacGEF activity and functionally couple Ca²⁺/CaM to polarized actin cytoskeletal remodeling during neuritogenesis (Figure 5).