Dissociation of Golgi-associated DHHC-type Zinc Finger Protein (GODZ)- and Sertoli Cell Gene with a Zinc Finger Domain-β (SERZ-β)-mediated Palmitoylation by Loss of Function Analyses in Knock-out Mice*

The γ2 subunit of GABA type A receptors (GABAARs) is thought to be subject to palmitoylation by both Golgi-associated DHHC-type zinc finger protein (GODZ; also known as DHHC3) and its paralog Sertoli cell gene with a zinc finger domain-β (SERZ-β; DHHC7) based on overexpression of enzymes and substrates in heterologous cells. Here we have further investigated the substrate specificity of these enzymes by characterization of GODZ and SERZ-β knock-out (KO) mice as well as double KO (DKO) neurons. Palmitoylation of γ2 and a second substrate, growth-associated protein of 43 kDa, that is independently implicated in trafficking of GABAARs was significantly reduced in brain of GODZ KO versus wild-type (WT) mice but unaltered in SERZ-β KO mice. Accumulation of GABAARs at synapses, GABAergic innervation, and synaptic function were reduced in GODZ KO and DKO neurons to a similar extent, indicating that SERZ-β does not contribute to palmitoylation or trafficking of GABAARs even in the absence of GODZ. Notably, these effects were seen only when mutant neurons were grown in competition with WT neurons, thereby mimicking conditions of shRNA-transfected neurons previously used to characterize GODZ. However, GABA-evoked whole-cell currents of DKO neurons and the GABAAR cell surface expression in DKO neurons and GODZ or SERZ-β KO brain slices were unaltered, indicating that GODZ-mediated palmitoylation selectively controls the pool of receptors at synapses. The different substrate specificities of GODZ and SERZ-β in vivo were correlated with their differential localization to cis- versus trans-Golgi compartment, a mechanism that was compromised by overexpression of GODZ.

S-Palmitoylation is an important posttranslational modification that involves the addition of the 16-carbon fatty acid chain palmitate via a thioester bond to Cys residues (1). This modification in turn can alter a protein's conformational state, membrane association, and complex formation as well as its susceptibility to other posttranslational modifications (2)(3)(4). Global analyses of rat synaptosomal fractions led to the discovery of nearly 300 palmitoylated synaptosomal protein candidates that illustrate the particular importance of palmitoylation in regulating the function of neuronal synapses (5).
In mammalian cells, the palmitoylation reaction is catalyzed principally by a super gene family of 23 palmitoyl acyltransferases containing a DHHC motif in a cysteine-rich domain (DHHC-CRD) 3 that are both essential for enzyme function (6 -10). The mechanisms that determine substrate specificity of PATs remain poorly understood, although some specificity is observed upon overexpression of substrates and PATs in heterologous cells. However, increasing evidence suggests that the substrate specificity of DHHC-type PATs in vivo is much more stringent than in vitro. For example, the postsynaptic density (PSD) protein of 95-kDa (PSD-95) can be palmitoylated in heterologous cells by at least five members of the DHHC family of PATs (11), whereas in vivo only DHHC8 and DHHC2 have so far been shown to be required for normal palmitoylation of PSD-95 (12,13). In part, the substrate specificity of DHHC enzymes may be determined by their segregation to different subcellular compartments, which has also implications as to how palmitoylation affects the trafficking or function of substrate proteins (14).
Here we have extended our characterization of GODZ (also known as DHHC3) and its most closely related paralog, Sertoli cell gene with a zinc finger domain-␤ (SERZ-␤; DHHC7), which are known as founding members of the mammalian DHHC family of proteins (15,16). We have previously provided evidence that in cultured neurons GODZ is important as a PAT of ␥2 subunit-containing ␥-aminobutyric acid (GABA) type A receptors (GABA A Rs), the principal receptors that mediate GABAergic synaptic inhibition (17). Moreover, GODZ shRNAor dominant negative GODZ-mediated knockdown of GODZ suggested that GODZ functions selectively at inhibitory but not excitatory synapses (18). Nevertheless, GODZ has independently been proposed to also regulate the trafficking of ionotropic glutamate receptors as both AMPA and NMDA receptors can be palmitoylated upon overexpression of GODZ in heterologous cells and neurons (19,20).
Based on mRNA expression analyses, GODZ and SERZ-␤ show similar tissue-specific expression in vivo and in brain are selectively expressed in neurons (18). When examined by overexpression in heterologous cells, both enzymes exhibit broad substrate specificity, and indeed virtually any palmitoylated protein can be palmitoylated by both GODZ and SERZ-␤ under these conditions (11,18,21). However, the extent by which any of these putative substrate proteins are palmitoylated by GODZ or SERZ-␤ in vivo has not been addressed.
To begin to address the role GODZ-mediated palmitoylation plays by loss of function analyses in vivo, here we have generated GODZ and SERZ-␤ knock-out (KO) mice. Analyses of GODZ KO and GODZ/SERZ-␤ double KO (DKO) primary cultured neurons confirmed that GODZ is important for normal accumulation of GABA A Rs at synapses and receptor-dependent presynaptic innervation. However, this phenotype was only observed when mutant neurons were grown and analyzed under conditions where they were forced to compete with wildtype (WT) neurons for GABAergic innervation, i.e. conditions we previously showed to be highly sensitive to reduced expression of the ␥2 subunit of GABA A Rs (22). By contrast, the cell surface expression and synaptic localization of GABA A Rs and the density of GABAergic synapses were unaffected in pure DKO cultures. Nevertheless, palmitoylation of ␥2 was markedly reduced in brain of GODZ KO compared with WT mice along with reduced palmitoylation of growth-associated protein of 43 kDa (GAP-43) (also known as neuromodulin) as a second substrate that was recently implicated in the formation of GABAergic synapses (23). Consistent with all these findings, electrophysiological analyses of DKO neurons revealed functional defects that were principally limited to inhibitory synapses. Notably, palmitoylation of PSD-95 and AMPARs was unaltered in brain of GODZ and SERZ-␤ KO mice. Moreover, palmitoylation of ␥2 and GAP-43 was unaltered in SERZ-␤ KO mice. The data indicate substantially greater substrate specificity of both PATs in vivo versus in vitro. Super-resolution microscopy showed that GODZ is highly restricted to the cis-Golgi, whereas SERZ-␤ was enriched in the trans-Golgi. Such differences in the subcellular localization of PATs might contribute to the increased substrate specificity of these enzymes in vivo versus after overexpression of substrate and enzyme in heterologous cells.

Generation and Verification of GODZ and SERZ-␤ Knockout Mice-To enable investigation of the role of GODZ-and SERZ-␤-mediated palmitoylation by loss of function analyses
in vivo, we generated KO mice in which the third and fourth protein-encoding exons of each of these two genes were deleted by homologous recombination in embryonic stem cells (Fig. 1, a and b, and "Experimental Procedures"). These two exons together encode the DHHC-CRD that is essential for enzyme function of GODZ and other DHHC family PATs (9,10). For both GODZ and SERZ-␤, deletion of exons 3 and 4 is predicted to cause a shift in the translational reading frame of the downstream exons with artificial stop sites and hence to result in enzymatically inactive, truncated, and likely unstable proteins that contain the protein sequence of only the first two of a total of six protein-coding exons. Successful deletion of exons 3 and 4 was verified by Southern blotting of genomic DNA using DNA probes that flanked the targeting vector (not shown) and by reverse transcription (RT)-PCR analyses of total brain RNA (Fig. 1c). For each gene, RT-PCR of RNA isolated from WT and heterozygous mice using primer pairs that annealed to exons 1 and 4 (GP1/2 and SP1/2, respectively; Fig. 1, a and b) revealed the expected RT-PCR amplification products in WT mice that were absent from corresponding KO mice as expected (Fig. 1c, upper micrographs, primer pairs GP1/2 and SP1/2). Moreover, PCR analyses of reverse transcribed RNA using primers that annealed to exons 1 and 6 of each gene (GP1/3 and SP1/3, respectively; Fig. 1, a and b) revealed the presence of the predicted cDNAs of WT mice and the predicted truncated products corresponding to cDNAs that had exons 3 and 4 deleted (Fig. 1c, lower micrographs, primer pairs GP1/3 and SP1/3). RNA isolated from heterozygous GODZ and SERZ-␤ mice gave rise to both the WT and truncated RT-PCR products as expected (Fig. 1c).
Next we confirmed the absence of GODZ in KO mice by Western blotting (Fig. 1d, left panels). Immunoblotting of whole brain lysates of WT and SERZ-␤ KO mice probed for GODZ detected a ϳ31-kDa protein corresponding in size to previously described epitope-tagged recombinant GODZ, which is slightly smaller than predicted based on its 299-amino acid sequence (17). Importantly, this protein was absent in brain lysates of GODZ KO mice as expected (Fig. 1d, left panels). Immunoblots for heat shock protein of 90 kDa (HSP90) analyzed in parallel were used as loading controls (Fig. 1d, lower  panels). Notably, the GODZ antiserum used for Western blotting was directed against a C-terminal 31-amino acid peptide of GODZ that is absent in the putative truncated protein of GODZ KO mice because of the deletion-induced, altered translational reading frame.
Parallel analyses of the same brain lysates with a SERZ-␤ antiserum detected a protein with an M r of ϳ30,000, which again is slightly less than the predicted molecular weight for this 308-amino acid protein. As predicted, this protein was present in extracts of WT and GODZ KO mice and absent in SERZ-␤ KO mice (Fig. 1d, right panels). The SERZ-␤ antiserum used for these experiments is directed against a peptide containing amino acids 40 -193 of human SERZ-␤ that includes a 37-amino acid exon that is absent in mouse SERZ-␤, whereas the rest of the immunogen is 98% conserved between human and mouse. In addition to the expected ϳ30-kDa SERZ-␤ protein species, this antiserum detected an additional, more slowly migrating SERZ-␤-immunoreactive band (ϳ35 kDa) that was also absent in SERZ-␤ KO mice and therefore may be SERZ-␤-related (Fig. 1d, right panels). It could represent a detergent-resistant complex of SERZ-␤ with another, as of yet unknown protein.
Next we verified the absence of GODZ immunoreactivity from cultured GODZ KO neurons and brain sections using a GODZ antibody directed against the C-terminal 50-amino acid peptide of mouse GODZ. Immunostaining for the microtubule-associated protein MAP2 was used to illustrate the outlines of neurons and dendrites. Immunofluorescent staining of primary cultured cortical neurons at 18 days in vitro (DIV 18) confirmed the Golgi-specific staining of GODZ described previously (17). Moreover, this staining was entirely absent in cultures prepared from GODZ KO embryos (Fig. 1e). In brain sections of WT mice, GODZ immunoreactivity was similarly delimited to Golgi-like structures of neurons as expected, and showing from top to bottom (i) the genomic loci including the protein-coding exons of GODZ and SERZ-␤, respectively (exons are shown as black boxes except for the DHHC domain-containing exons that are essential for enzymatic activity and shown in red), (ii) targeting vectors, (iii) mutated alleles resulting from homologous recombination, (v) the floxed alleles after Cre-mediated deletion of the PGK-neo cassette (gray shaded box), and the mutated alleles lacking the fourth and fifth protein-coding exons. The yellow triangles indicate the positions of the loxP sites. The locations of the primers used for RT-PCR analyses (GP1-3 for GODZ and SP1-3 for SERZ-␤; shown in 5Ј-3Ј direction) are indicated as black arrowheads. Exons are shown larger than to scale. For simplicity, the exons coding for non-translated region of the mRNAs are not shown. Restriction sites are indicated as follows: B, BamHI; Bg, BglII; RI, EcoRI; N, NcoI. c, RT-PCR analyses of brain RNA from GODZ (left) and SERZ-␤ (right) KO mice with the genotypes analyzed indicated at the top of panels, the primer pairs (GP1-3 and SP1-3) used indicated on the left, and the size of amplified cDNA fragments representing WT and KO alleles for each primer pair indicated on the right. d, immunoblots of brain lysates of GODZ and SERZ-␤ KO mice probed with antibodies raised against GODZ and SERZ-␤ and genotypes as indicated. Note the absence of GODZ-and SERZ-␤-specific bands of ϳ31 and ϳ30 kDa in lysates of GODZ KO and SERZ-␤ KO mice, respectively. The SERZ-␤ antiserum in addition to the ϳ30-kDa bands recognizes a ϳ36-kDa SERZ-␤-related protein species, both absent in SERZ-␤ KO mice. e, representative images of cultured cortical neurons derived from WT and homozygous GODZ KO mice immunostained for GODZ (green; top) and MAP2 (red; middle) with merged images shown at the bottom. f, immunostaining of cortical brain sections of a WT (left column) and GODZ KO mouse (right column) for GODZ (green; top) and gephyrin (red; middle) with merged images shown in the bottom row. Note the absence of the GODZ immunoreactivity in cultured neurons and brain sections of GODZ KO mice. Scale bars, 10 m. this staining was absent in sections of GODZ KO mice (Fig. 1f). Immunostaining for the inhibitory synapse marker gephyrin was used to visually delineate the neuropil from neuronal somata of brain sections. Notably, analogous immunostainings for SERZ-␤ were not feasible as the SERZ-␤ antiserum used for Western blotting revealed only unspecific staining under these (non-denaturing) conditions (not shown). Nevertheless, all our analyses were consistent with faithful and specific inactivation of the GODZ and SERZ-␤ genomic loci in the respective KO mice.
The Phenotype of Double KO Mice Suggests Functional Redundancy of GODZ and SERZ-␤ in Vivo-Both GODZ KO and SERZ-␤ KO mice were found to be viable and fertile without overt behavioral or physical phenotypes. Nevertheless, male GODZ KO mice showed a modest reduction in body weight (89.2 Ϯ 2.5% of WT at 1 month of age, p Ͻ 0.001; 90.4 Ϯ 1.7% of WT at 2 months of age, p Ͻ 0.01; n ϭ 14 -18 for all groups, t tests). By contrast, the body weights of female GODZ KO mice and male and female SERZ-␤ KO mice were not measurably affected (not shown). Previous in vitro experiments using overexpression of recombinant proteins had indicated largely overlapping substrate specificity of GODZ and SERZ-␤ (Ref. 18; and for review, see Ref. 6). Therefore, to test for possible functional redundancy of GODZ and SERZ-␤, we interbred GODZ and SERZ-␤ KO mice to generate GODZ/SERZ-␤ DKO mice. Unexpectedly, most DKO mice showed a perinatally lethal phenotype, although early on during the course of these studies a few DKO mice survived to adulthood. By 1 year of age these surviving DKO mice showed drastically reduced body and brain weights when compared with littermate controls (total body weight of DKO mice, 53.6 Ϯ 1.7% of GODZ ϩ/Ϫ SERZ-␤ ϩ/Ϫ and GODZ ϩ/ϩ SERZ-␤ Ϫ/Ϫ controls; DKO brain weight, 74.7 Ϯ 6.3% of controls; n ϭ 3 mice/group). No viable postnatal DKO mice were recovered beyond three generations of brother-sister matings of GODZ ϩ/Ϫ SERZ-␤ ϩ/Ϫ mice. As predicted, these findings suggested at least partial functional redundancy of GODZ and SERZ-␤ and, by extension, significant overlap in substrate specificity of these two enzymes. In addition, they dictated that further analyses of the neuronal mutant phenotypes were limited to single KO mice or to cultured neurons derived from KO or DKO embryos.
GODZ Is Required for Normal Trafficking of GABA A Rs and for GABAergic Innervation-Given the defects in GABA A R clustering and synapse formation seen previously in shRNAand dominant negative GODZ-transfected neurons (18), we predicted that such deficits should be replicated in neuron cultures prepared from GODZ KO neurons. Surprisingly, cortical neurons prepared from GODZ KO embryos (18 DIV) showed normal GABA A R clustering as evidenced by unaltered density of punctate immunostaining for the ␥2 subunit and the presynaptic marker glutamic acid decarboxylase (GAD) in GODZ KO compared with WT cortical cultures (Fig. 2, a- The normal density of GABA A Rs clusters and inhibitory synapses in KO cultures at first appeared to conflict with our previous results obtained by shRNA-or dominant negative GODZ-mediated knockdown of GODZ that suggested a role for GODZ in clustering of GABA A Rs and inhibitory synapse formation (18). However, we recently showed that ␥2-GABA A Rs are critically important for GABAergic synapse formation selectively under conditions where GABA A R-deficient neurons are forced to compete with WT neurons for presynaptic innervation by WT neurons (22), i.e. conditions that mimic cultures that were sparsely transfected with shRNA constructs. Thus, to assess whether GODZ was required for inhibitory synapse formation under competitive conditions, we cocultured GODZ KO neurons with WT neurons (1:9 ratio). WT neurons were prepared from a GFP transgenic embryo to facilitate distinction of mutant and WT neurons. Interestingly, under these conditions, GODZ KO neurons showed drastically reduced density of puncta for the ␥2 subunit and GAD compared with WT cultures (density of puncta in KO neurons mixed with WT neurons as a percentage of pure WT cultures: ␥2, 57.5 Ϯ 4.9%; GAD, 57.9 Ϯ 4.0%; p Ͻ 0.001, n ϭ 10 -12 cells for both comparisons, ANOVA, Tukey tests) (for absolute values, see Fig. 3, m and n) (Fig. 3

, a-d versus i-l, m, and n).
Residual GABA A R clusters in mutant neurons of mixed cultures remained invariably colocalized with presynaptic GAD (percent colocalization: GODZ KO/WT co-cultures, 89.2 Ϯ 4.3%; WT pure cultures, 91.9 Ϯ 3.0%; p, n.s.; n ϭ 12-18, ANOVA) (Fig. 3o). Notably, our analyses of mutant neurons focused on pyramidal shaped neurons that lacked somatic GAD immunoreactivity and GFP fluorescence. Therefore, the GABAergic terminals contacting mutant neurons in these mixed genotype cultures originated almost exclusively from WT neurons, and the synaptic phenotypes of mutant pyramidal cells can be attributed specifically to postsynaptic defects in palmitoylation. The data indicate that GODZ KO neurons replicate the phenotype previously described for GODZ shRNAand GODZ dominant negative construct-transfected cultures.
Unaltered clustering of the ␥2 subunit in (pure) GODZ KO cultures was unexpected and appeared inconsistent with a simple role of palmitoylation in forward trafficking of GABA A Rs to the plasma membrane, raising the possibility of functional redundancy of GODZ with SERZ-␤. To further assess whether the phenotype of GODZ KO neurons was compromised by functional redundancy between GODZ and SERZ-␤, we repeated the above experiments with neuron cultures prepared from DKO (GODZ Ϫ/Ϫ SERZ-␤ Ϫ/Ϫ ) embryos. Similar to results obtained with GODZ KO cultures, cortical cultures (18 DIV) prepared from pure DKO embryos showed normal density and colocalization of punctate immunoreactivity for the ␥2 subunit and GAD (density as percentage of WT: ␥2, 97.4 Ϯ 3.1%; GAD, 96.2 Ϯ 3.6%; p, n.s.; n ϭ 29 -46 cells; colocalization of ␥2 and GAD: DKO, 87.2 Ϯ 1.5%; WT, 87.3 Ϯ 1.9%; n ϭ 30 -46; p, n.s.; ANOVA) ( Fig. 3, a-e versus e-h and m-o). Moreover, DKO neurons co-cultured with GFP-tagged WT neurons showed defects in pre-and postsynaptic clustering comparable with GODZ KO/WT co-cultures described above (density of puncta in DKO/WT co-cultures as a percentage of pure WT cultures: ␥2, 49.7 Ϯ 1.9%, n ϭ 30 -59; GAD, 55.3 Ϯ 2.2%, n ϭ 29 -59; p Ͻ 0.001, ANOVA, Tukey tests) (Fig. 3, e-h versus i-l, m, and n). However, unlike in the case of GODZ KO/WT co-cultures, neurons of DKO/WT co-cultures showed significantly reduced colocalization of punctate ␥2 and GAD immunoreactivities (colocalization in DKO/WT co-cultures as a percentage of pure WT neurons, 82.6 Ϯ 2.9; p Ͻ 0.001, n ϭ 30 -59) (Fig. 3o). The size of ␥2 puncta in DKO neurons of DKO/WT co-cultures remained unaltered (DKO neurons co-cultured, 96.8 Ϯ 5.6% of WT; p, n.s.; n ϭ 30 -59, ANOVA). The data suggest that the reduced density of puncta observed in GODZ KO and GODZ/ SERZ-␤ DKO neurons that was observed selectively in co-cultures was not simply a consequence of reduced forward trafficking of GABA A Rs to the cell surface. Curiously, quantitative comparison of the ␥2 clustering defect observed in co-cultured GODZ KO neurons (density of ␥2 puncta in GODZ KO, 57.5% of WT) with the defect observed in co-cultured DKO neurons (density of ␥2 puncta in DKO, 49.7% of WT) revealed no significant difference (p, n.s; n ϭ 30 -59 (DKO) and n ϭ 12-18 (GODZ KO), two-way ANOVA) and indicated that SERZ-␤ did not contribute to palmitoylation and trafficking of ␥2-GABA A Rs even in the absence of GODZ (i.e. in GODZ KO neurons). Hence, although the lethal phenotype of DKO mice suggests functional redundancy of GODZ and SERZ-␤ with respect to at least some of their substrates, no such overlap in function is apparent with respect to GABA A R clustering.
GODZ Is Required for Normal Function of Inhibitory Synapses-To corroborate the above findings of altered inhibitory synapses with functional assays, we performed patch clamp recordings of DKO neurons grown in competition with GFP-transgenic WT neurons, analogous to conditions used for immunostaining above. Whole-cell currents evoked by 20-s 100 M GABA pulses recorded from DKO neurons (3741 Ϯ 651 pA) were indistinguishable from whole-cell currents recorded from WT neurons in the same culture (4418 Ϯ 770 pA; n ϭ 12, p Ͼ 0.05) (Fig. 4a), suggesting normal cell surface expression of GABA A Rs. Similarly, whole-cell glutamate-evoked currents GFP-transgenic WT cultures and GODZ KO cortical neurons were cultured as pure cultures, or the mutant neurons were co-cultured 1:9 with an excess of GFP-expressing WT neurons. GFP autofluorescence was used to distinguish WT from mutant neurons. a-l, representative micrographs of a GFP-WT neuron (a-d; note the GFP autofluorescence in a (green)), a GODZ KO neuron (e-h), and a GODZ KO neuron co-cultured with WT neurons (i-l) immunostained for the ␥2 subunit (b, f, and j; red) and GAD (c, g, and k; blue) with corresponding merged images shown in d, h, and l. Boxed dendritic segments are shown enlarged in the right-most column of micrographs with arrows pointing to colocalized pre-and postsynaptic structures of axons that have grown along and seem to adhere to dendrites and arrowheads pointing to axons (blue) that have grown across dendrites with minimal contact. m-p, summary quantifications of the density of ␥2 (m) and GAD (n) immunoreactive puncta, their colocalization (o), and the size of ␥2 puncta (p). Note the drastic reduction in the density of both pre-and postsynaptic structures selectively in GODZ KO neurons that were grown in the presence of an excess of WT neurons, whereas the density of puncta in pure KO cultures remained unaffected. In mutant neurons grown in competition with WT neurons, the defect in synaptic localization of GABA A Rs resulted in loss of innervation by WT neurons. Bar graphs with error bars indicate means Ϯ S.E. ***, p Ͻ 0.001, ANOVA, Tukey tests. Scale bar, 5 m.
(500 M; 20 s) were unaffected (DKO, 931 Ϯ 210 pA; WT, 1022 Ϯ 234 pA; p Ͼ 0.05, n ϭ 12, t test) (Fig. 4b). However, consistent with the reduced density of punctate GABA A R and GAD immunostaining, the frequency and amplitude of mIPSCs recorded from DKO neurons were significantly reduced compared with WT neurons (frequency mean values: DKO, 0.15 Ϯ 0.08 Hz; WT, 0.24 Ϯ 0.09 Hz; amplitude mean values: DKO, 19.9 Ϯ 1.9 pA; WT, 22.5 Ϯ 2.2 pA; n ϭ 12 for both comparisons, p Ͻ 0.05, Kolmogorov-Smirnov test) (Fig. 4, d and e) (Fig. 4, g and h). Frequency changes of mEPSCs that are accompanied by unaltered mEPSC amplitudes can generally be attributed to altered neurotransmitter release probabilities of presynaptic terminals, which in our co-cultures primarily were derived largely from WT neurons. Moreover, we recently showed that modest defects in GABAergic synapses of GABA A R ␥2 ϩ/Ϫ neurons result in a significant homeostatic-like reduction in the density of gluta-matergic synapses (24), indicating that the change in mEPSC frequency recorded from DKO neurons was indirect and not indicative of altered palmitoylation of proteins functioning at glutamatergic synapses.
GODZ Is Essential for Palmitoylation of the GABA A R ␥2 Subunit and GAP-43 in Vivo-Given the normal immunostaining of the ␥2 subunit in pure GODZ KO and DKO cultures, it was critically important to ascertain whether GODZ KO mice suffered from measurable defects in palmitoylation in vivo. Candidate protein substrates of GODZ/SERZ-␤ based on PAT assays of transfected cells, in addition to the ␥2 subunit, include GAP-43 (11). GAP-43 was of additional interest because it was recently identified as a gephyrin-interacting protein that regulates the degradation of gephyrin and, indirectly through altered gephyrin clustering, modulates the cell surface expression of GABA A Rs in developing neurons. Co-transfection of HEK293T cells with GAP-43 and GODZ resulted in significantly increased GAP-43 palmitoylation compared with transfection of GAP-43 alone (GODZ plus GAP-43, 620.4 Ϯ 105.7% of GAP-43 alone) (Fig. 5a). Moreover, GODZ failed to palmitoylate a mutant version of GAP-43, GAP-43 C3S,C4S , that had its FIGURE 3. Deficits in GABA A R postsynaptic clustering and inhibitory synapse density in DKO cortical neurons co-cultured with WT neurons. DKO cortical neurons were cultured on their own or co-cultured 1:9 with an excess of GFP-expressing WT neurons analogously to Fig. 2. a-l, representative micrographs of a GFP-WT neuron (a-d; green), a GFP-lacking DKO neuron (e-h), and a GFP-lacking DKO neuron co-cultured with WT neurons (i-l) immunostained for the ␥2 subunit (b, f, and j; red) and GAD (c, g, and k; blue) with corresponding merged images shown in d, h, and l. Boxed dendritic segments of each image are shown enlarged in the right-most column with arrows pointing to colocalized pre-and postsynaptic structures of axons that have grown along dendrites and seem to adhere to dendrites and with arrowheads pointing to axons (blue) that have grown across dendrites with minimal contact, seemingly ignoring the dendrites. m-p, summary quantifications of the density of immunoreactive puncta for ␥2 (m) and GAD (n), their colocalization (o), and the average size of ␥2 puncta per neuron (p). Note the reduced density of pre-and postsynaptic puncta selectively in DKO neurons co-cultured with WT neurons, whereas the size of ␥2 puncta was unaltered. Bar graphs with error bars indicate means Ϯ S.E. ***, p Ͻ 0.001, ANOVA, Tukey tests. Scale bar, 5 m.
To assess whether the ␥2 subunit and GAP-43 were dependent on GODZ for normal palmitoylation in vivo, we quantitated the steady state palmitoylation levels of these proteins by acyl-biotin exchange (ABE) assays of whole brain extracts of GODZ KO and WT mice. These assays involve blocking of the free cysteine residues of proteins with N-ethylmaleimide followed by hydrolyzing the thioester linkages between cysteine residues and palmitate with hydroxylamine and conjugating the newly freed cysteine residues to a cysteine-reactive biotin compound that then allows for pulldown of the previously palmitoylated proteins with streptavidin beads. Hydroxylamine-lacking reactions were routinely analyzed in parallel to qualitatively monitor unspecific labeling and pulldown of proteins independently of palmitate substitution. The amount of palmitoylated protein in the pulldown assay (hydroxylamine lanes) was quantitated by Western blotting using antibodies directed against the proteins of interest and normalizing to total amounts of the same proteins in aliquots of the same extracts on the same gel/Western blot. Based on such assays, steady state palmitoylation of both the ␥2 subunit and GAP-43 was markedly decreased in GODZ KO versus WT brain (␥2: KO, 67.2 Ϯ 4.7% of WT, p Ͻ 0.001, n ϭ 7-8 mice; GAP-43: KO, 66.2 Ϯ 12.2% of WT, p Ͻ 0.05, n ϭ 7 mice, t tests) (Fig. 5b). Therefore, both the ␥2 subunit and GAP-43 represent genuine substrates of GODZ in vivo. By contrast, palmitoylation of PSD-95 was unaffected (GODZ KO, 108.6 Ϯ 10.1% of WT, p ϭ 0.589, n ϭ 4, t test) (Fig. 5b), although this protein can be readily palmitoylated by GODZ when enzyme and substrate are overexpressed in HEK293T cells (11). Similarly, palmitoylation of the AMPAR subunits GluA2/3 was unaltered in GODZ KO versus WT brain (KO, 104.4% Ϯ 18.1% of WT, n ϭ 5) even though these receptors can be palmitoylated by overexpression of recombinant GODZ (19). The total expression of ␥2, GAP-43, and GluA2/3 (normalized to ␤-tubulin) was unaltered in KO mice (KO: ␥2, 105.4 Ϯ 15.4% of WT; GAP-43, 159.4 Ϯ 32.9% of WT; GluA2/3, 82.4 Ϯ 3.2% of WT; n ϭ 5; p, n.s. for all three comparisons; t tests) (Fig. 5, e and f), indicating that differences apparent in ABE assays did not involve alterations in protein expression. Notably, there was a non-significant trend for increased expression of GAP-43 in KO versus WT brain (p ϭ 0.16) that is consistent with reduced GABAergic inhibitory tone and evidence that GAP-43 expression is subject to neural activity-mediated up-regulation. Interestingly, neither palmitoylation of ␥2 nor of GAP-43 or GluA2/3 was measurably reduced in brain extracts of SERZ-␤ KO mice (KO: ␥2, 109.2 Ϯ 20.4% of WT; GAP-43, 110.6 Ϯ 11.2% of WT, GluA2/3, 96.7 Ϯ 17.6% of WT; p, n.s; n ϭ 5 for all three comparisons; t tests) (Fig. 5, g and h). Therefore, the in vivo substrates of GODZ and SERZ-␤ represent only a subset of the large number of proteins that can be palmitoylated by overexpression of the enzymes and substrates in heterologous cells.
GODZ and SERZ-␤ Are Dispensable for Cell Surface Accumulation of GABA A Rs-To independently assess whether GODZ and SERZ-␤ are required for accumulation of GABA A Rs or AMPARs at the neuronal cell surface, we measured the expression of these proteins by cell surface biotinylation, starting with analyses of hippocampal brain slices. The cell surface values determined for the ␥2 and GluA2/3 subunits in hippocampal brain slices of WT and GODZ KO were normalized to the amounts of ␤-tubulin determined in aliquots of total extracts from the same biotinylated brain samples set aside before affinity purification. We found no significant changes in the cell surface expression of ␥2 and GluA2/3 subunits in GODZ KO mice (␥2, 104.8 Ϯ 24.33% of WT, p ϭ 0.89; GluA2/3, 138.6 Ϯ 31.0% of WT, p ϭ 0.29; n ϭ 5 mice per genotype for both comparisons; t tests) (Fig. 6, a-d).
To further assess the purity of cell surface-biotinylated protein fractions and to validate the negative results of the above experiments, we retroactively estimated an enrichment factor for GluA2/3 in cell surface-biotinylated versus total protein fractions using either ␤-tubulin or ␤-actin as a cytoplasmic reference protein (Fig. 6, a and d). Although ␤-tubulin was only partially reduced in cell surface fractions versus total extracts, ␤-actin was highly depleted and barely detectable in some cell surface fractions (Fig. 6, a and d). Thus, ␤-tubulin but not ␤-actin remains partially associated with cell surface-biotinylated, affinity-purified membrane fractions. To estimate an enrichment factor for cell surface proteins in cell surface-biotinylated versus total protein fractions, we first normalized the GluA2/3 expression level for both the cell surface and total fractions to the respective ␤-actin level. The average enrichment factor for cell surface-biotinylated protein fractions, defined as normalized surface GluA/total GluA, was 192 Ϯ 82-fold (n ϭ 12 brains). Analogous determination of a cell surface enrichment factor using normalization to ␤-tubulin instead of ␤-actin revealed an enrichment factor of 55 Ϯ 26 (n ϭ 12), confirming that ␤-tubulin remained partially associated with purified cell surface fractions.

SERZ-␤ Does Not Contribute to GABA A R Trafficking and Function Even in the Absence of GODZ-
To test for possible functional redundancy of GODZ and SERZ-␤, we further examined primary cultured neurons prepared from DKO embryos. The cell surface expression levels of ␥2 and GluA2/3 subunits were not measurably altered in DKO versus WT cultures (␥2, 109 Ϯ 11.7% of WT; p, n.s.; n ϭ 5; GluA2/3, 107.3 Ϯ 9.0% of WT; p, n.s.; n ϭ 4) (Fig. 6, h-j). The average cell surface enrichment factor in this experiment using normalization to ␤-tubulin was 7.8 Ϯ 2.5-fold (n ϭ 8 cultures). Unaltered cell surface expression of ␥2 and GluA2/3 subunits in DKO is consistent with normal densities and sizes of postsynaptic ␥2 puncta in pure DKO cultures (Fig. 3, m and p) and unaltered GABA-and glutamate-evoked whole-cell currents in DKO neurons of mixed genotype cultures (Fig. 4, a and b). Taken together, the results indicate that neither GODZ nor SERZ-␤ is essential for normal cell surface expression of GABA A Rs nor do these PATs measurably regulate exocytosis of AMPARs or show any appreciable redundancy with respect to palmitoylation of these putative substrates. Although GODZ is essential for normal palmitoylation of the ␥2 subunit and for accumulation of GABA A Rs at synapses and GABA A R-dependent presynaptic innervation, the detailed mechanism by which palmitoylation controls the trafficking or accumulation of these receptors at synapses remains ill-defined (see "Discussion").
GODZ and SERZ-␤ Are Differentially Restricted to cis-and trans-Golgi-The collective evidence so far has pointed to a selective role for GODZ in palmitoylation of the ␥2 subunit and with HA-GODZ. Input lanes represent aliquots of assays removed before pulldown of biotinylated samples and used for normalization. Hydroxylaminecontaining (HAϩ) lanes represent palmitoylated substrate, and hydroxylamine-lacking (HAϪ) lanes represent negative controls (see "Experimental Procedures"). b, summary quantification of data in a showing GODZ-induced palmitoylation in a GAP-43 Cys-3/4-dependent manner. c and d, representative ABE assays of brain extracts from GODZ KO and WT mice analyzed on the same gel (c) and summary quantification of substrate palmitoylation normalized to input and WT (d). Note the marked reductions in steady state palmitoylation of the ␥2 subunit and GAP-43 and unaltered palmitoylation of PSD-95 and GluA2/3 subunits in brain of GODZ KO versus WT mice. e and f, representative Western blots comparing total brain extracts of two GODZ KO and two WT mice probed for the ␥2 subunit, GAP-43, or GluA2/3 subunits (e). The three proteins were analyzed in separate aliquots of the same extracts. Parallel blots of the same extracts were probed for ␤-tubulin (␤-tub) and used to normalize ␥2, GAP-43, and GluA2/3 protein amounts for summary quantification (f). Note that the expression of ␥2, GAP-43, and GluA2/3 was unchanged except for a weak trend for increased expression of GAP-43 in KO versus WT brain. g and h, ABE assays of WT and SERZ-␤ KO brain extracts analyzed on the same gel and probed for ␥2, GAP-43, or GluA2/3 (g). Summary statistics (h) indicate that palmitoylation of all three proteins was unchanged in SERZ-␤ KO versus WT brain. Bar graphs with error bars indicate means Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, t tests.

GAP-43 in vivo.
Furthermore, although GODZ and SERZ-␤ show indistinguishable substrate specificity in vitro and neuron-specific expression in brain in vivo (18), SERZ-␤ was dispensable for normal palmitoylation of the ␥2 subunit and GAP-43 in vivo. We wondered, therefore, whether functional differences between GODZ and SERZ-␤ might reflect subtle differences in the precise localization of these two enzymes in the Golgi complex. To address these questions, we first assessed the suborganellar localization of GODZ in the Golgi using transfection of HEK293T cells with either the cis/medial-Golgi marker N3 (also known as Golgi phosphoprotein of 130 kDa) fused to GFP (GFP-N3) or the trans-Golgi marker galactosyltransferase fused to YFP (YFP-GalT) (26,27). Immunofluorescent images of cells stained for endogenous GODZ and the two Golgi markers were captured by super-resolution structure-illuminated microscopy (SIM) followed by intensity correlation analysis of their colocalization in the Golgi area (Fig. 7).
Although GODZ appeared perfectly colocalized with the cis-Golgi marker GFP-N3, GODZ immunoreactivity was largely excluded from areas labeled by the trans-Golgi marker YFP-GalT (Fig. 7, a and b). The mean intensity correlation quotient (ICQ) for colocalization of GFP-N3 with GODZ (0.22 Ϯ 0.017) was significantly greater than that for GalT-YFP and GODZ (0.078 Ϯ 0.019, p Ͻ 0.001, n ϭ 7-9, t test) (Fig. 7g), indicating that GODZ was 2.8-fold more restricted to the cis-than the trans-Golgi complex. To address whether Golgi localization was affected by overexpression of GODZ (and to replicate experimental conditions used for analyses of SERZ-␤ below), we repeated the above experiment using transfection of epitope-tagged GODZ (Fig. 7, c and d). HA-GODZ remained significantly more colocalized with GFP-N3 (ICQ ϭ 0.302 Ϯ 0.021) than with GalT-YFP (ICQ ϭ 0.157 Ϯ 0.017, p Ͻ 0.001, n ϭ 11-17, t test) (Fig. 7h). However, the degree of colocalization of transfected HA-GODZ with the cis-Golgi marker FIGURE 6. Unaltered cell surface expression of GABA A Rs and AMPARs in brain slices of KO mice and DKO neuron cultures. a-f, cell surface-biotinylated protein fractions of hippocampal brain slices of GODZ KO (a-c) and SERZ-␤ KO mice (d-f) were analyzed for expression of the ␥2 and GluA2/3 subunits of GABA A Rs and AMPARs, respectively, and compared with respective fractions of WT littermates. Cell surface values were normalized to the amount of ␤-tubulin present in aliquots of total extracts set aside before pulldown of biotinylated cell surface proteins. In addition, the ratio of GluA2/3 to ␤-actin expression was compared in total versus cell surface fractions to estimate a cell surface enrichment factor for GluA2/3 subunits (see text for further details). Note the depletion of cell surface fractions of ␤-actin and the drastically increased ratio of GluA2/3/␤-actin in cell surface versus total extracts of both experiments (a and d). By contrast, significant amounts of ␤-tubulin remained associated with affinity-purified cell surface fractions. Summary statistics of cell surface fractions normalized to ␤-tubulin revealed no significant differences in cell surface expression of ␥2 (b and e) and GluA2/3 subunits (c and f) in GODZ KO (b and c) and SERZ-␤ KO mice (e and f) (p, n.s. for all four comparisons; t tests). g-i, analyses of cell surface-biotinylated protein fractions of cortical cultures prepared from WT and DKO embryos and probed for ␥2 and GluA2/3 subunits. Duplicate aliquots of cell surface fractions and total extracts from pairs of WT and DKO cultures prepared and processed in parallel were compared on the same gel. The cell surface expression levels of ␥2 (h) and GluA2/3 subunits (i) (normalized to ␤-tubulin (␤-tub)) were not measurably altered in DKO versus WT cultures (p, n.s; t tests). Bar graphs with error bars indicate means Ϯ S.E.

Discussion
Here we have presented an initial characterization of GODZ and SERZ-␤ KO mice lacking expression of two of the most prominent members of the DHHC family of PATs that show  and f1; red). Colocalization of the Golgi markers with GODZ or SERZ-␤ is evident in the merged images depicted in the right-most column (a2-f2; yellow). g-i, summary quantification of colocalization by intensity correlation analyses of immunofluorescence in the Golgi area for the indicated Golgi markers with endogenous GODZ (g), transfected HA-GODZ (h), and transfected HA-SERZ-␤ (i). Note that both endogenous GODZ and transfected HA-GODZ were strongly colocalized with the cis-Golgi marker GFP-N3 but not with the trans-Golgi marker GalT-YFP. By contrast, HA-SERZ-␤ showed greater colocalization with GalT-YFP than with GFP-N3. Bar graphs with error bars indicate means Ϯ S.E. **, p Ͻ 0.01; ***, p Ͻ 0.001, t tests. Scale bar, 2.5 m.
broad tissue distribution and that have garnered a reputation of largely indistinguishable and broad substrate specificity. Our studies confirm that GODZ is an important PAT for the ␥2 subunit of GABA A Rs and GAP-43, two proteins that contribute to inhibitory synapse function by entirely different mechanisms. The ␥2 subunit is known to be required for postsynaptic clustering of GABA A Rs and for GABAergic inhibitory synapse formation (28 -30). This latter phenotype is specifically evident under competitive conditions, such as under artificial conditions where ␥2 ϩ/Ϫ neurons are grown in competition with WT neurons (22) or following sparse in utero electroporation of embryonic cortex with a ␥2 shRNA vector in vivo (30), and it may be physiologically relevant in the context of adult-born hippocampal neurons that must compete for innervation by axons of mature neurons (22). However, similar defects were not observed upon global deletion of the ␥2 subunit in homogeneous ␥2 KO cultures (28). Thus, reduced clustering of GABA A Rs and defects in GABAergic innervation of GODZ KO neurons that were selectively observed in neurons grown in competition with WT neurons replicate the phenotype of ␥2-deficient neurons grown in competitive conditions. The defect in inhibitory synapses observed in GODZ KO neurons is entirely consistent with moderately reduced expression of ␥2-containing GABA A Rs.
A putative role for GODZ-mediated palmitoylation of GAP-43 in the formation of GABAergic synapses, however, is more abstruse. There is general agreement that palmitoylation of GAP-43 is essential for its translocation from the Golgi to the plasma membrane, whereas its association with the plasma membrane can persist even after the protein has been depalmitoylated (for a review, see Ref. 31). GAP-43 is known to traffic through the cis-Golgi (32), which is consistent with restricted localization of GODZ specifically to this Golgi compartment. However, GAP-43 is subject to dynamic palmitoylation and depalmitoylation also at the plasma membrane as has been shown for GAP-43 localized to axonal growth cones (33). Given the restricted localization of GODZ to the cis-Golgi, this dynamic change in GAP-43 palmitoylation at the plasma membrane is likely mediated by PAT family members other than GODZ. More importantly, GAP-43 has recently been shown to associate with cytoplasmic gephyrin in a mechanism that is regulated by the phosphorylation state of a PKC site of GAP-43 (23). PKC-mediated phosphorylation of GAP-43 regulates the degree of cytoplasmic gephyrin aggregation and its sequestration into aggresomes and degradation, which in turn is thought to regulate the residual pool of gephyrin available for gephyrindependent clustering and stabilization of GABA A Rs at inhibitory synapses (23). Although a direct mechanistic link between GODZ-mediated palmitoylation of GAP-43 and PKC-mediated phosphorylation of GAP-43 remains elusive, it is conceivable that defects in trafficking of GAP-43 through the early secretory pathway of GODZ KO neurons affect GAP-43 phosphorylation and hence contribute to impaired gephyrin-and GABA A R-dependent formation of GABAergic synapses. However, much more elaborate experimentation would be needed to dissociate ␥2-dependent from GAP-43-dependent defects in inhibitory synapse formation.
At the outset of our studies, we predicted that the mutant phenotype of GODZ KO mice might become compromised by functional redundancy between GODZ and SERZ-␤, which prompted us to generate KO mice also for SERZ-␤. These two PATs are not only closely related in their structure, they also exhibit largely indistinguishable substrate specificity when analyzed by overexpression of enzymes and putative substrates in transfected heterologous cells (6,18) along with similar broad mRNA tissue distribution in peripheral tissues and similar neuron-specific expression in brain (11,18). The absence of overt behavioral and physical phenotypes in single KO mice together with the severely runted or lethal phenotype of DKO mice supported our prediction that GODZ and SERZ-␤ show overlapping substrate specificity. However, we were surprised to find that the synaptic defects of DKO neurons were not more severe than those of GODZ KO neurons and that palmitoylation of ␥2 and GAP-43 was perfectly normal in SERZ-␤ KO mice. We also noted that other paralogs of GODZ might contribute to GABA A R and GAP-43 palmitoylation in vivo even though these enzymes were ineffective in palmitoylating these proteins in transfected HEK293T cells (11,18). In particular, we found that the mRNA coding for the somewhat more distantly related GODZ/SERZ-␤ paralog DHHC15, as quantitated by RT-PCR, was significantly up-regulated in brain tissue of GODZ KO and DKO mice compared with WT littermate controls (DHHC15 mRNA in percentage of WT: GODZ KO, 129.3 Ϯ 10.9% of WT, p Ͻ 0.05; DKO, 207.1 Ϯ 40.1% of WT, p Ͻ 0.05; n ϭ 4, t tests). Thus, DHHC15 could account for residual palmitoylation of ␥2 and GAP-43 observed in GODZ KO mice even if it does not contribute to palmitoylation of these substrates in WT mice. By contrast, mRNA levels for yet another GODZ/SERZ-␤ paralog, DHHC21, were unchanged in DKO versus WT brain (not shown).
GODZ has previously been implicated as a PAT of PSD-95 (11) and AMPA receptor subunits GluA1-4 (19). However, we found that steady state palmitoylation of PSD-95 was unaffected in GODZ KO mice, which is consistent with convincing evidence that PSD-95 palmitoylation is mainly mediated by DHHC2 (13), DHHC5 (34), and DHHC8 (35). In the case of AMPA receptors, overexpression of GODZ was previously reported to limit the receptor cell surface expression in cultured neurons (19). If this effect of overexpressed GODZ was due to increased AMPAR subunit palmitoylation one would predict that GODZ KO mice would show increased cell surface expression of AMPA receptors. However, we found no measurable change in GluA2/3 palmitoylation and cell surface expression in GODZ KO and SERZ-␤ KO mice. Moreover, we found no change in the cell surface expression of GluA2/3 subunits, glutamate-evoked whole-cell currents, or mEPSC amplitudes in DKO cultured neurons, indicating that GODZ and SERZ-␤ do not regulate constitutive trafficking of AMPARs to the plasma membrane. Instead, reduced AMPAR cell surface expression seen following overexpressed GODZ is likely due to artificial mass-driven interaction with and trapping by GODZ in the Golgi compartment as we described previously for the ␥2 subunit and shorter polypeptide substrates coexpressed with GODZ or SERZ-␤ in transfected cells (18). When overexpressed in heterologous cells, GODZ, SERZ-␤, and their puta-tive substrates form stable complexes in the Golgi region, seemingly reflecting trapping in a very limited membrane compartment where these substrates do not normally accumulate. Such complexes appear useful to map protein-protein interaction domains, but they interfere with the normal trafficking of substrates in a non-physiological manner (18). It is also noteworthy that GODZ DKO neurons grown in competition with WT neurons showed a slight reduction in the frequency of mEPSCs in DKO neurons that is reminiscent of GABAergic deficit-induced homeostatic down-regulation of glutamatergic synapses observed in ␥2 ϩ/Ϫ cultures and mice (24). Collectively, our studies indicate that the substrate specificity of GODZ is much narrower than previously thought, and they call for caution when interpreting DHHC enzyme-substrate relationships based on overexpression of these proteins in cultured cells.
The finding of unaltered GABA A R and GAP-43 palmitoylation in SERZ-␤ KO mice was unexpected and raised the question of whether SERZ-␤ and GODZ differed in their precise subcellular localization. Analyses by immunoelectron microscopy had previously suggested an asymmetric distribution of GODZ across the Golgi stack, but whether GODZ was restricted to the cis-or trans-face of the complex had remained elusive (18). The super-resolution imaging used here unambiguously identified GODZ as a cis-Golgi PAT, whereas SERZ-␤ was enriched in the trans-Golgi compartment. Moreover, we showed that overexpression of GODZ compromised its restricted localization to the cis-Golgi. It stands to reason then that the different substrate specificity of GODZ and SERZ-␤ is at least in part due to their differential localization in the secretory pathway.
None of the neuronal substrates examined in this study could be identified as genuine in vivo substrate of SERZ-␤. However, we recently identified SERZ-␤ as a principal PAT for the glucose transporter GluT4 in adipocytes and muscle cells. 4 SERZ-␤ KO mice were shown to suffer from hyperglycemia and glucose intolerance, consistent with prominent defects in insulin-induced palmitoylation of GluT4 and translocation of GluT4 to the plasma membrane. GODZ and SERZ-␤ KO mice will be instrumental in evaluating the role these enzymes play in a range of physiological processes both in the CNS and in peripheral tissues.

Experimental Procedures
Animals-KO mice were generated by homologous recombination of the genomic loci in BK4 embryonic stem (ES) cells (37) using targeting vectors that contained a loxP site between the second and third protein-coding exons and a PGK-neo cassette flanked by a loxP site inserted between the fourth and fifth protein-coding exons of the genes and coding GODZ and SERZ-␤ (Fig. 1, a and b). Exons 3 and 4 encode the DHHC-CRD of each gene, which is known to be essential for PAT activity of DHHC family PATs (38). Deletion of exons 3 and 4 in each case is predicted to result in a shift of the reading frame downstream of the deleted exons. Thus, deletion of exons 3 and 4 of each gene is predicted to result in truncated proteins lacking the DHHC-CRD domains that are essential for PAT activity as well as all other sequences downstream of exon 2, most likely resulting in unstable proteins. The targeting vector for GODZ was derived from the bacterial artificial chromosome (BAC) clone RPCI-23-338G21 (BACPAC Resources Center, Children's Hospital Oakland Research Institute, Oakland, CA) and contained a 2999-bp NcoI-NheI left arm fragment with a loxP site at a unique EcoRI site (1098 bp from the NcoI site) followed by a loxP site-flanked PGK-neo cassette (inserted in the same transcriptional orientation as the target locus) and a 2951-bp NheI-EcoRI right arm fragment. Properly targeted ES cell clones (BK4 cell line (37); cells obtained from the Pennsylvania State University transgenic core) were screened and verified by Southern blotting and PCR, expanded and injected into blastocysts, and transferred to foster mothers following standard procedures of the Pennsylvania State University transgenic core facility. High grade chimeric offspring mice were then crossed with B6.FVB-Tg(EIIa-cre) C5379Lmgd/J mice (stock number 003724, The Jackson Laboratory, Bar Harbor, ME) to delete the loxP site-flanked exons and PGK-neo cassette. Offspring mice with germ line transmission and proper Cre-mediated deletion of the PGK-neo cassette (fGODZ locus; Fig. 1a) or both the PGK-neo cassette and exons 4 and 5 (GODZ KO allele; Fig. 1a) were identified by PCR analyses of genomic DNA of tail biopsies; the Cre transgene was subsequently removed by outcrossing. The recombined loci were confirmed by Southern blotting analyses of genomic DNA using probes that mapped to regions of the GODZ locus upstream and downstream of the targeting vector. SERZ-␤ KO mice were produced analogously using BAC RP23-87F1 as a source for generating the SERZ-␤ targeting vector (BACPAC Resources Center, Children's Hospital Oakland Research Institute). The "left arm" of the targeting vector consisted of a 2941-bp BglII fragment that was modified by insertion of a loxP site at a unique BamHI site (1307 bp downstream of the 5Ј BglII site). This fragment was ligated to a PGK-neo cassette flanked by loxP sites with the PGK-neo cassette in the same transcriptional orientation as the SERZ-␤ locus followed by the "right arm" of the targeting vector consisting of a 4299-bp BglII-EcoRI genomic fragment. Properly targeted embryonic stem cells were identified and verified by genomic Southern blotting. High grade chimeric mice derived from these cells were crossed with Cre transgenic mice to delete the PGK-neo cassette as described above for GODZ KO mice, and the final floxed SERZ-␤ locus was verified in mice using Southern blotting and partial sequencing of PCR-amplified genomic fragments.
Husbandry and Genotyping of Mice-All mice were maintained on a standard 12:12 h light-dark cycle with food and water available ad libitum. The mice were maintained as heterozygotes on a 129-C57BL/6 mixed genetic background using brother-sister matings. KO mice used for experimentation were produced by mating of heterozygous mice, thereby allowing for experimental comparison with WT and heterozygous controls as littermates. All animal experiments were approved by the Institutional Animal Care and Use Committee of the Pennsylvania State University or by the Institutional Animal Care and Use Committee of the Tufts Medical Center and performed in accordance with relevant guidelines and regulations of the National Institutes of Health. Routine genotyping of GODZ KO mice by PCR was done using primers GZ1 (5Ј-GAGGC TTCAG AATAG TCTCT TAC-3Ј) and GZ4 (5Ј-GCTCC CCAAC TCTTA CTTGA ATG-3Ј) for the mutant locus and primers GZ3 (5Ј-TGCCA GCCCA GCCTC ATTTT ATT-3Ј) and GZ4 (5Ј-GCTCC CCAAC TCTTA CTTGA ATG-3Ј) for the GODZ wild-type locus. SERZ-␤ KO mice were genotyped using the primers SZ1 (5Ј-TGAGC CAGGA TGGAT TTCAG ACA-3Ј) and SZ6 (5ЈTGCCC TCGGA CGCAG GAGAT GAA-3Ј) for the mutant locus and SZ4 (5Ј-TCCCC TGATG TATGC GAATG TCC-3Ј) and SZ5 (5Ј-AACAG GTGCC TTTTG AATGT CAG-3Ј). B5/EGFP mice (stock number 003516, The Jackson Laboratory) were genotyped as described (39).
Cell Surface Biotinylation-Cell surface biotinylation of cultured neurons was done as described previously (40). Briefly, two to three 6-cm culture dishes per genotype of DIV14 cortical neurons were incubated with (EZ-Link TM sulfo-NHS-SSbiotin, Thermo Fisher Scientific). The unreacted biotin was quenched by washing once with cold 50 mM glycine followed three times with 0.1% bovine serum albumin (BSA; number 7030, Sigma) in phosphate-buffered saline (PBS). The cells were lysed in ice-cold lysis buffer (150 mM NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, pH 8.0, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1 g/ml each leupeptin, pepstatin A, antipain, and aprotinin), the protein was extracted for 15 min on ice, and the extracts were cleared by centrifugation at 10,000 ϫ g for 10 min at 4°C. Protein concentrations were determined using the Bradford (36) assay. Two hundred micrograms of protein were adsorbed to NeutrAvidin TM beads (Thermo Fisher Scientific) for 2 h at 4°C, the beads were washed with lysis buffer (three times), and the bound protein were eluted with SDS-containing loading buffer. Duplicate aliquots of cell surface fractions and total extracts from matched pairs of WT and DKO cultures were analyzed by Western blotting analyses in five sequential experiments.
ABE Assay-Extracts from brain tissue of 4-month-old mice or from transfected HEK293T cells were analyzed by the acylbiotin exchange assay with minor modifications (41). The cells and tissues were lysed in LB buffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 5 mM EDTA, and 10 mM N-ethylmaleimide supplemented with protease inhibitors (Complete protease inhibitor mixture). Triton X-100 (1.7%) was added, and the tissue was homogenized using a tip sonicator (three 10-s pulses at 2.5 setting) followed by incubation on ice for 1 min. Proteins were diluted in LB buffer with 50 mM N-ethylmaleimide and incubated at 4°C overnight with gentle rocking. Proteins were precipitated with chloroform/methanol, resolubilized in 4SB (4% SDS, 5 mM EDTA, and 50 mM Tris-HCl, pH 7.4) repeatedly. After the third precipitation, the samples were subjected to with 1% Triton X-100 for 1 h followed by blocking for 1 h in 5% normal goat serum. The sections were incubated in rabbit anti-GODZ (directed against amino acids 250 -299 of mouse GODZ; 1:500; Ab31837) and mouse anti-gephyrin mAb7a (1:2000; number 147 011, Synaptic Systems) overnight at 4°C and developed using Alexa Fluor 488-conjugated anti-rabbit and Cy3-conjugated anti-mouse secondary antibodies. Fluorescent images were captured using an Olympus Fluoview 1000 confocal microscope equipped with a 40ϫ 0.13 NA UPlanFL objective. Images were contrast-enhanced and assembled into figures using Adobe Photoshop.
Structured Illumination Microscopy and Intensity Correlation Analyses of Golgi Localization of GODZ/SERZ-␤-HEK293T cells grown on poly-lysine-coated glass coverslips were transfected and immunostained essentially as described above using the plasmid GFP-N3 (43) or GalT-YFP (44) alone or together with HA-GODZ or HA-SERZ-␤ (11) (1 g per plasmid), fixed, and immunostained using rabbit anti-GODZ (1:500; Ab31837) or rabbit anti-HA (1:1000; H6908, Sigma-Aldrich) and mouse anti-GFP (1:1000; A11120, Molecular Probes). Images were captured using a Nikon N-SIM microscope with a CFI Apochromat total internal reflection fluorescence 100ϫ oil objective (NA 1.49) with Nikon Elements software utilizing the Center for Biologic Imagining (Pittsburgh, PA). Colocalization of transfected proteins was analyzed using ImageJ (National Institutes of Health, Bethesda, MD). The images were background-subtracted using the Subtract Background from region of interest plug-in followed by quantitation of colocalization with the Intensity Correlation analysis plug-in (48).
Electrophysiology-Cortical neurons at DIV9 -11 were used for electrophysiological analyses. Whole-cell recordings were performed using Multiclamp 700A (Molecular Devices Corp., Sunnyvale, CA). The recording chamber was continuously perfused with a bath solution containing 128 mM NaCl, 30 mM D-glucose, 25 mM HEPES, 5 mM KCl, 2 mM CaCl 2 , and 1 mM MgCl 2 , pH 7.3 (adjusted using NaOH). Patch pipettes were filled with 147 mM KCl, 5 mM disodium phosphocreatine, 2 mM EGTA, 10 mM HEPES, 2 mM MgATP, and 0.3 mM Na 2 GTP, pH 7.3 (adjusted with KOH). Data were acquired using PCLAMP 8 software, sampled at 10 kHz, and filtered at 1-2 kHz. To assess whole-cell currents, pulses (20 s) of GABA (100 M) or glutamate (500 M) were applied through a glass pipette with the pipette tip close to the cell soma, and evoked currents were recorded with the membrane potential clamped at Ϫ70 mV. Current peak amplitudes were measured using Clampfit 9 software (Molecular Devices Corp.) to obtain GABA A or AMPA receptor-mediated whole-cell currents (pA). mIPSCs and mEP-SCs were recorded in the presence of 0.5 M tetrodotoxin and either 10 M 6-cyano-7-nitoquinoxaline-2,3-dione or 20 M bicuculline, respectively.
Statistical Analysis-Statistical comparisons of multiple means of biochemical and imaging data were performed using Prism software (GraphPad, La Jolla, CA). Multiple comparisons were done by one-way ANOVA followed by Tukey post hoc analysis. Normality of data distribution of imaging data was assessed by D'Agostino-Pearson test. Simple comparison of two group means was done by two-sample two-tailed t tests.