Phosphorylation of a New Brain-specific Septin, G-septin, by cGMP-dependent Protein Kinase*

The septins are a family of GTPase enzymes, some of which are required for the cytokinesis stage of cell division and others of which are associated with exocytosis. We purified and cloned the cDNA for a 40-kDa protein from rat brain that is a substrate for type I cGMP-dependent protein kinase (PKG). The amino acid sequences of two tryptic peptides of P40 showed high homology to the septins. Molecular cloning revealed the 358-amino acid P40 to be a new member of the septin family. P40 was named G-septin, as it is phosphorylated in vitro by PKG, but relatively poorly by the related cAMP-dependent protein kinase and not by protein kinase C. Two splice variants of G-septin ( a and b ) were found with distinct N and C termini, but a common GTPase domain. G-septin lacks the C-terminal coiled-coil domain characteristic of all other mammalian septins and uniquely has two predicted phosphorylation site motifs for type I PKG. Photoaffinity labeling a -(244–259) (DLEDKTENDKIRQESM). The peptides were individually conjugated by coupling, through C-terminal cysteines added to the sequences during their synthesis, to diphtheria toxoid as a carrier, and conjugates were then mixed together to immunize the same sheep (Chiron Technologies, Victoria, Australia). The production and characterization of the two anti-NEDD5 antibodies 6 and 11 were reported previously (14).

Some septins play an essential role in cytokinesis. In budding yeast, cell division involves assembly of a ring of four septin proteins encircling the junction between mother and bud, which acts like a scaffold for other proteins to complete cytokinesis (1,2). The ring first appears at the cell cortex, where a bud will emerge and grow and to which the divided nuclei migrate and eventually separate. Cytokinesis does not necessarily occur by ring contraction, but at least one role of the septin ring is for the assembly of Myop and subsequently of actin at the ring (20). Mutation in any of the four yeast septin proteins Cdc3p, Cdc10p, Cdc11p, and Cdc12p results in cell cycle arrest and defective cytokinesis and gives rise to multinucleated cells (1,3,21). The septin ring is absent in these mutant strains (1). In contrast to yeast, cytokinesis in insect and animal cells is achieved by a contractile mechanism in which a cleavage furrow of actin filaments forms at the cell cortex during anaphase and rapidly constricts to bisect the cell. Although there are mechanistic differences in division between yeast and flies, septins are also required for cytokinesis in flies. A mutation in Drosophila PNUT causes a lethal phenotype, consistent with a cytokinesis defect (7). Drosophila septins also encircle the cleavage furrow before and during cytokinesis, and the actomyosin contractile ring shrinks as the membrane pinches inwards. Mammalian NEDD5 localizes near the contractile ring during cytokinesis in HeLa cells, and microinjection of anti-NEDD5 antibodies into dividing cells results in binucleated cells, suggesting that NEDD5 plays an essential role during cytokinesis (14).
Groups of different septins assemble as filaments in vitro, and this probably also occurs in vivo. This was first described in yeast, where mutation in any of the CDC3, CDC10, CDC11, and CDC12 genes results in loss of 10-nm filaments that normally localize at the septin ring (22)(23)(24). Heterotrimeric septin complexes comprising PNUT, SEP1, and SEP2 from Drosophila also form filaments and exhibit GTP binding and GTPase activities in vitro (9). Septin complexes also occur in mammalian cells and contain CDC10, NEDD5, H5, DIFF6, and KIAA0128 (25). Like the Drosophila septin complex, the mammalian complex assembles as 8-nm filaments assembling endto-end in 25-nm unit lengths (25). The mammalian septin filaments display thin, 10-nm long strands arising from their sides every 16 nm along the filament, a feature not detected in Drosophila (9). Septin GTPase activity may be required for mammalian septin filament formation in vivo since microinjection of GTP␥S 1 and transfection of GTP-binding mutants of mouse Nedd5 into HeLa cells disrupt formation of NEDD5containing filaments (14).
A number of septins are found in post-mitotic neurons. High levels of Drosophila SEP1 were found in central and peripheral nervous tissue (8), and PNUT was also observed in Drosophila neural tissue (22). However, the function of the septins in these cells is unknown. CDCrel-1 is brain-specific in rats; CDC10, H5, and NEDD5 are abundant in brain, whereas DIFF6 is not highly expressed there (26). In humans, four septins, NEDD5, H5, DIFF6, and CDC10, were found in post-mortem brain tissues, the first three of which were identified in neurofibrillary tangles in brains affected by Alzheimer's disease (27). Notably, CDC10 was not localized to the tangles. The three septins were also found in fine fibrillar deposits in neuronal soma, which are considered the earliest detectable stages of neurofibrillary tangle formation. These findings suggest that septins may have a function in the etiology of neuronal disease. Neuronal septins probably do not play roles in cytokinesis since neurons in adult brain generally do not divide. New roles for neuronal septins are emerging in synaptic transmission. Four septins have been identified in a protein complex that can interact with the SEC6-SEC8 complex in rat brain and that can assemble as filaments (25). Since the SEC6-SEC8 complex is a cluster of molecules essential for exocytosis, septins may have a role that links the secretory machinery to the actin-based cytoskeleton beneath the plasma membrane (25). The brainspecific human CDCrel-1 and the more widely expressed NEDD5 proteins have been shown to be essential for exocytosis in neurons (26). Both proteins immunoprecipitated with the N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) protein syntaxin, and transfection of wild-type or dominant-negative mutants of human CDCrel-1 inhibited or enhanced, respectively, exocytosis from cultured neuroendocrine cells.
No septin is presently known to be regulated by phosphorylation or by cellular signal transduction cascades. This study was initiated as part of a program to identify neuronal proteins phosphorylated by type I cGMP-dependent protein kinase (PKG-I) (28 -31). A major signaling pathway common to many cells is mediated by the second messenger, cGMP. Intracellular levels of cGMP are responsive to hormones of the atrial natriuretic peptide family and to a variety of other hormones that activate or induce the synthesis of nitric-oxide synthase, leading to the production of nitric oxide (29). Cyclic GMP plays a role in a variety of physiological responses, from learning and memory to apoptosis. Major targets for cGMP signaling are the two Ser/Thr protein kinases PKG-I and PKG-II (29). PKG has a relatively limited substrate specificity compared with the closely related PKA, and only a small number of substrates have been identified (29). A previous study has identified a group of proteins in rat brain that were selectively phosphorylated by PKG in comparison with PKA (28). Here, a 40-kDa protein (P40) that was relatively specifically phosphorylated by PKG was purified and identified as a new member of the septin family of proteins. We named this protein G-septin as it is a relatively specific PKG substrate.
Protein Purification-P40 was partially purified from rat brain by extension of approaches previously described (28). A fraction enriched in Ca 2ϩ -sensitive lipid-binding proteins was prepared from 40 rat brains. Proteins in this fraction were precipitated by (NH 4 ) 2 SO 4 to 80% saturation; dialyzed against buffer containing 20 mM Tris-HCl (pH 7.7), 1 mM dithiothreitol, 1 mM EDTA, and 0.1% Tween 80; and applied to a Q-Sepharose high performance anion-exchange column (10 ϫ 1.6 cm; Amersham Pharmacia Biotech) at 2 ml/min. After washing the column with the same buffer, bound proteins were eluted with a 0 -400 mM NaCl gradient, and 60 2-ml fractions were collected. Aliquots of every second or third fraction were phosphorylated (see below), and fractions containing P40 were pooled. The pool of P40 was diluted to adjust the NaCl concentration to Ͻ40 mM and the pH to 7.0 and applied to an S-Sepharose high performance cation-exchange column (10 ϫ 1.6 cm; Amersham Pharmacia Biotech) that was pre-equilibrated with 20 mM Tris-HCl (pH 7.0) containing 1 mM EDTA, 0.1% Tween 80, and 1 mM dithiothreitol at 1 ml/min. After extensive washing, bound proteins were eluted with a 0 -400 mM NaCl gradient, and 60 1-ml fractions were collected. Aliquots of the column fractions were phosphorylated, and fractions containing P40 were pooled, concentrated, and desalted with an Ultrafree-4 centrifugal filter device (Millipore Corp., Bedford, MA) and stored at Ϫ20°C in the S-Sepharose column buffer (pH 7.0). The catalytic subunit of PKA and PKC (a mixture of PKC␣, PKC␤, and PKC␥) were purified from bovine heart and rat brain as described previously (32). Bovine lung PKG-I was purified to homogeneity using DEAE-cellulose and cAMP affinity columns (33), with additional steps as described (32), and was stored at Ϫ70°C in 10% glycerol with 0.05% Tween 80. The recombinant NEDD5 protein was expressed and purified as described previously (14).
Protein Phosphorylation-Phosphorylation of brain fractions eluted from the ion-exchange columns was performed by incubation at 37°C for 5 min with 0.125 Ci/l [␥-32 P]ATP. The incubation mixture (in a final volume of 80 l) contained 30 mM Tris-HCl (pH 7.4), 1 mM EGTA, 1 mM MgSO 4 , 40 M ATP, and 0.05% Tween 80 in the absence or presence of the activator of the appropriate protein kinase, such as 10 M cGMP, as indicated in the figure legends. As a control, PKI (0.01 mg/ml) was routinely included in the conditions used to stimulate PKG to prevent potential cross-activation of endogenous PKA by cGMP. After prewarming tubes to 37°C for 5 min, incubations were initiated by addition of ice-cold brain samples (usually 10 l), followed immediately by the purified protein kinase (20 -40 ng/tube). To define the relative specificity of PKG substrates, the fractions were also phosphorylated by the catalytic subunit of PKA. Phosphorylation reactions were terminated after 5 min by addition of 50 l of a 3ϫ concentrated SDS sample buffer and rapid freezing on dry ice as described previously (34). Half of each sample was used for SDS-PAGE, and phosphoproteins were detected by gel electrophoresis and autoradiography as described (34) using 7.5-15% linear gradients and 20-cm gels (Bio-Rad Protean II system).
Protein Kinase Assay-Protein kinase activity was determined using 30 mM Tris-HCl (pH 7.4) 1 mM EGTA, 200 M ATP, 2 Ci of [␥-32 P]ATP, and 10 mM MgSO 4 in 40-l incubations. Incubations were for 5 min at 30°C, and the synthetic peptide phospholamban-(8 -21) (TRSAIR-RASTIEMP; at a final concentration of 0.1 mg/ml) was the substrate (35). Incubations were initiated by addition of 40 ng of PKG or 20 ng of the catalytic subunit of PKA. These amounts of the protein kinases were determined in preliminary experiments to phosphorylate phospholamban- (8 -21) to the same level since this substrate has the same V max for both protein kinases (35). Reactions were terminated by addition of phosphoric acid to 75 mM, and aliquots were spotted onto Whatman P-81 paper, washed three times for 10 min each in 75 mM phosphoric acid, dried, and counted by liquid scintillation techniques (36).
Amino Acid Sequencing-For amino acid sequencing, P40 was ex-cised from a dried, Coomassie Blue-stained polyacrylamide gel and digested with trypsin, and the resultant peptides were separated by HPLC. Five major peaks were sequenced with an Applied Biosystems Model 494-HT Procise Sequencer.
Cloning of P40 and Sequence Analysis-The BLAST program was utilized to search the expressed sequence tag data base (NCBI, dbest) using the peptide amino acid sequences obtained from the amino acid sequencing (37). Five potential clones, AA002534, W64898, and AA003408 (matching the second peptide) and AA036534, and W13793 (matching the first peptide), were obtained from the IMAGE Consortium (Genome System, St. Louis, MO), and their sequences were confirmed by manual DNA sequencing. A PCR product from clone AA003408, a 400-bp fragment encoding the second peptide, was used as a probe to screen a rat brain cDNA library (ZAP II library, Stratagene). The Gigaprime DNA labeling kit (Bresatec Ltd.) was used to label the probe, and the library was screened according to the manufacturer's instructions. Subcloned cDNA was characterized by manual DNA sequencing with a Thermo-Sequenase radiolabeling terminator cycle sequencing kit (Amersham Pharmacia Biotech). G-septin-␤ was initially subcloned, but does not have a stop codon at the 3Ј-end of the sequence. Therefore, G-septin-␤ was used as a probe to screen the cDNA library, and another six positive clones were isolated. One of these was identified as full-length G-septin-␣. All sequence, homology, and structural analyses were performed with the BLAST program at NCBI. An unrooted phylogenetic tree analysis was carried out using MultAlign (38) with the Windows software TreeView (39).
Antibodies and Immunoblot Analysis-Antibodies were raised in sheep against two different synthetic peptides from regions common to both G-septin isoforms: G-septin-␣-(97-112) (EKIPKTVEIKAIGHVI) and G-septin-␣-(244 -259) (DLEDKTENDKIRQESM). The peptides were individually conjugated by coupling, through C-terminal cysteines added to the sequences during their synthesis, to diphtheria toxoid as a carrier, and conjugates were then mixed together to immunize the same sheep (Chiron Technologies, Victoria, Australia). The production and characterization of the two anti-NEDD5 antibodies 6 and 11 were reported previously (14).
Immunoprecipitation and Western Blotting-To confirm that the partially purified P40 preparation contained G-septin, we immunoprecipitated G-septin from the column fraction containing P40 with the anti-G-septin antibodies. Antibodies were incubated with partially purified P40 at 4°C overnight and incubated with protein G-agarose beads (Roche Molecular Biochemicals) for 2 h. The protein G beads were sequentially washed with phosphate-buffered saline (pH 7.4) and 20 mM Tris-HCl (pH 7.7), resuspended in 20 mM Tris-HCl (pH 7.7), and aliquoted to 40-l lots, each containing 0.02-0.04 g of the 40-kDa protein. These samples were phosphorylated on the beads as described above. Phosphoproteins were heated at 95°C for 2 min, centrifuged in a microcentrifuge at top speed for 1 min, and then separated by gel electrophoresis and autoradiographed using 7.5-15% linear gradient gels (34). For Western blot analysis, samples were denatured in a reducing SDS stop buffer and resolved on 12% polyacrylamide minigels. The proteins were transferred onto a nitrocellulose transfer membrane (0.45 m; Schleicher & Schü ll, Dassel, Germany) as described previously (40). The membrane was blotted in phosphate-buffered saline (pH 7.4) with 5% skim milk overnight, washed in Tris-buffered saline (pH 7.4) containing 0.1% Tween 20 (TBST), and then incubated with the first antibody for 2 h and with the second antibody for 1 h in TBST containing 0.5% polyvinylpyrrolidone 40; and the membrane was washed with TBST. Immunoblotting was performed by chemiluminescent detection (Pierce SuperSignal).
Expression of His 6 -tagged G-septin-␣-Wild-type G-septin-␣ cDNA was cloned into the pTrcHis vector (Invitrogen) and transfected into Escherichia coli. Recombinant G-septin-␣ was expressed using the Xpress TM protein expression system (Invitrogen) and purified with a Ni 2ϩ -nitrilotriacetic acid resin column (QIAGEN Inc.) according to the manufacturers' instructions.
[␣-32 P]GTP Photoaffinity Labeling-The GTP binding assay was performed in the well of a 96-well microtiter plate containing 50 mM KCl, 20 mM Hepes (pH 7.6), 1 mM EGTA, 3 mM MgCl 2 , 1 mM dithiothreitol, and 500 nCi of [␣-32 P]GTP. Partially purified P40 (purified as described above) from brain (5-10 g of protein) and recombinant G-septin-␣ (2 g) were used for photoaffinity labeling. The tubes were incubated in the dark at 4°C for 3 h and then irradiated with a short wavelength ultraviolet lamp at 315 nm for 15 min at a distance of 8 cm. The specificity of the photolabeling was determined by comparing the labeling in the presence and absence of 1 mM unlabeled GTP. The reaction was stopped by addition of SDS sample buffer and heating at 95°C for 2 min. Samples were analyzed by SDS-PAGE and autoradiography.
Immunoprecipitation of 32 P-Labeled G-septin-␣-Crude synaptosomes (P2) were isolated (34); washed twice with Hepes-buffered Krebs solution containing 20 mM Hepes, 118 mM NaCl, 4.7 mM KCl, 1.18 mM MgSO 4 , 0.1 mM Ca 2ϩ , and 10 mM D-glucose (pH 7.4); and then labeled with 500 Ci/ml 32 P i for 1 h at 37°C (31,32). After washing with warmed medium to remove free 32 P i , aliquots of the synaptosomes were incubated without additions or with the membrane-permeable cyclic nucleotide analogue 8-p-chlorophenylthio-cGMP (8-pCPT-cGMP; 200 M) or 8-p-chlorophenylthio-cAMP (8-pCPT-cAMP; 200 M) for 15 min at 37°C. Synaptosomes were sedimented for 1 min in a microcentrifuge and lysed with 50 mM Tris, 0.5% Triton X-100, 250 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, and 20 g/ml leupeptin (pH 7.4). Gseptin was immunoprecipitated as described above with some modifications. The lysates were precleared for 15 min with protein G-coupled agarose beads that had been incubated with serum from non-immunized sheep. G-septin antiserum (2 l/tube) was purified by preincubation with protein G-agarose beads for 2 h and then washed with buffer containing 20 mM Tris (pH 7.4), 1 mM dithiothreitol, 1 mM EDTA, 1 mM EGTA, and 7 g/ml leupeptin. The protein G beads coated with anti-G-septin antibodies were incubated with the lysate at 4°C overnight. The beads were recovered, washed with phosphate-buffered saline (pH 7.3), resuspended in 80 l of SDS sample buffer, and subjected to SDS-PAGE and autoradiography.
Northern Blotting-Total RNAs from different rat tissues (brain, liver, pancreas, kidney, heart, spleen, lung, thyroid gland, testis, stomach, intestine, aorta, and skeletal muscle) were prepared using the Tri-Reagent TM kit (Sigma). Rat brain mRNA was isolated using the PolyATtract ® mRNA isolation kit (Promega). Northern blotting was performed using standard procedures (41) with some variations. In brief, 2 g of mRNA and 20 g of total RNA were resolved on a 2.2 M formaldehyde-containing 1% agarose gel and transferred onto a nylon transfer membrane (Bresatec Ltd.). The DNA probes used for hybridization were a 196-bp PCR fragment that is unique to the N-terminal coding region of G-septin-␣ and a synthetic primer containing 61 bp (corresponding to G 1043 -G 1103 at the 3Ј-end of the coding region of G-septin-␣). The probes were labeled with [␣-32 P]dCTP using the Gigaprime kit. Hybridization was performed at 50°C overnight. The membrane was then washed twice with 2ϫ concentrated SSC (0.3 M NaCl and 0.03 M sodium citrate (pH 7)) and 0.1% SDS and once with 0.2ϫ SSC and 0.1% SDS at 65°C and autoradiographed.

Identification of a 40-kDa PKG Substrate as a New
Septin-A rat brain extract enriched in both peripheral membrane proteins and Ca 2ϩ -sensitive lipid-binding proteins was subjected to Q-Sepharose chromatography. Bound proteins were eluted with a 0 -400 mM NaCl gradient, and aliquots of every second or third column fraction were phosphorylated in the presence of exogenous PKG or the catalytic subunit of PKA (Fig. 1A). Equal specific activities of PKG and PKA were used as determined by phosphorylation of a synthetic peptide containing amino acids 8 -21 of phospholamban. A 40-kDa protein(s), called P40, eluted from the column at 260 -310 mM NaCl (primarily in Q-Sepharose column fractions 38 -48) and was preferentially phosphorylated by PKG rather than PKA (Fig. 1A, lanes 4 -6). Note that the catalytic subunit of PKA autophosphorylated and migrated at approximately the same position as P40; therefore, phosphorylation in the presence of PKA (lanes 3, 6, and 9) might primarily represent autophosphorylation of exogenous PKA. Fractions containing P40 were pooled and subjected to S-Sepharose chromatography (Fig. 1B). Bound proteins were eluted with a 0 -400 mM NaCl gradient, and aliquots of every second fraction were phosphorylated with PKG or PKA as described above. P40 eluted from the column at 240 -280 mM NaCl (primarily in S-Sepharose column fractions 36 -42). Approximately 10 g of P40 was obtained at ϳ2% purity as determined by staining gels with Coomassie Brilliant Blue (Fig. 1C).
To identify P40, the band containing the protein(s) was excised from a Coomassie Blue-stained gel and digested in gel with trypsin, and the released peptides were separated by high performance liquid chromatography. Five peptides (Table I) were obtained for amino acid sequencing. The first two, FDFN-IMVVGQSGLGK and EVTHNIHYETYR, were the predominant peptides. They exhibited 58 -93% identity to various members of the septin family (Table I). The third peptide, (R/S)GSNPDIPVVDHA, was related to only one septin protein, yeast Cdc12p. The other two peptides, VVDLESPNTS and AEXIENTXX, produced weak sequence signals and did not match any proteins in the non-redundant data base. The two major peptides indicate that the major component of the P40 band is a new member of the septin family of proteins.
Cloning and Splice Variants of G-septin-Cloning of P40 revealed that it is a new member of the septin family of proteins. We named P40 G-septin as it is a PKG substrate in vitro.
To clone G-septin, the sequences of peptides 1 and 2 were used to search the expressed sequence tag data base. Five potential clones, AA036534 and W13793 (matching the first peptide) and AA002534, W64898, and AA003408 (matching the second peptide), were found, and their sequences were confirmed by manual DNA sequencing. The AA002534 and AA003408 clones encoded the second peptide completely. A 400-bp fragment encoding this region was obtained by PCR using clone AA002534 as a template and was used to screen a rat brain cDNA library. Ten positive clones were isolated; only one of them, G-septin-␤, had an open reading frame region ( Fig. 2A). The first in-frame ATG resides in an appropriate context for translation initiation (42) and is preceded by an in-frame stop codon 9 bases upstream. The open reading frame has 1041 bases, without a stop codon at the 3Ј-end, and encodes a polypeptide of 347 amino acids with a predicted isoelectric point of 6.14 and a molecular mass of 39.6 kDa. The sequences of three other positive clones from this screen were compared with the 3Ј-end of G-septin-␤. Each contained an extra 4 amino acids (DSNP) at the C terminus, but was without an initiation site at the 5Ј-end (suggesting that they are fragments of fulllength G-septin-␤); therefore, 4 amino acids (DSNP) are missing at the end of the C terminus of the G-septin-␤ sequence.
To obtain a full-length G-septin cDNA, G-septin-␤ was used as a probe to screen a rat brain library, and six positive clones were isolated. One of these clones, G-septin-␣, had an open reading frame region of 1074 bases and encodes a 358-amino acid protein with a predicted isoelectric point of 6.72 and a   (Fig. 2, A and B). G-septin-␣ and -␤ were identical over most of their sequences, but differed in the N-and C-terminal regions, suggesting the use of alternative exons in these locations. Only G-septin-␣ contained both peptides 1 and 2, whereas G-septin-␤ contained only the second half of peptide 1, which is located across the putative splice junction (Fig. 2B). Therefore, P40 is most likely to represent G-septin-␣ rather than G-septin-␤; however, it is also possible that P40 contains a mixture of both isoforms, which are very similar in size. Data base searching revealed that G-septin shares similarities with yeast, mammalian, and Drosophila septins. Like other septins, G-septin-␣ contains a set of consensus sequence motifs for GTP binding, at amino acids 68 -76, 125-128, and 207-210. These regions are highly conserved in the septin family of proteins, but are distinct from those found in other GTPase subfamilies (43). Most mammalian septins are characterized by a predicted coiled-coil domain at their C termini (4,7,8,19,44). However, G-septin does not contain this feature. A primary feature of the N-terminal splicing is the presence in G-septin-␣ of a predicted binding motif for SH3 domain-containing proteins (PAVP), which is absent from G-septin-␤ (Fig.  2). Among the septins, G-septin uniquely contains two predicted phosphorylation sites for PKG (45), Thr 55 and Ser 91 in G-septin-␣ ( Fig. 2A), with G-septin-␤ missing the former potential site.
The amino acid sequence of G-septin-␣ was aligned with that of all known mammalian septins. Only the GTPase domain was used (amino acids 40 -325 in G-septin-␣), as the short N and C termini of the septins have very low homology. This revealed an overall protein sequence identity of 21% for the family of mammalian septins (Fig. 3A). H5 and human CDCrel-1 were the most related septins at 84% identity, and G-septin and KIAA0991 showed 73% identity in this region of the sequence (Fig. 3B). KIAA0991 is a new human septin sequence that was deposited in the GenBank TM /EBI Data Bank while this manuscript was in preparation. An unrooted phylogenetic analysis indicated that the eight mammalian septins fall into three groups comprising 1) G-septin and KIAA0991; 2) KIAA0128; and 3) H5, CDCrel-1, DIFF6, CDC10, and NEDD5 (Fig. 3C). The amino acid sequence identity between any two septins from different groups was 42-46%.
P40 Is G-septin, a New PKG Substrate-To confirm that cloned G-septin represents the 40-kDa protein originally identified as being a PKG substrate, a series of phosphorylations, Western blotting, and immunoprecipitations were performed. Note again that exogenous PKA autophosphorylated at 40 kDa (clearly seen in isolation in Fig. 4A, lane 3) and partly contributed to the relatively low labeling in the region of P40 on these gels. First, we compared phosphorylation of P40 with that of a related septin (NEDD5) that is also found in high levels in the brain and has a slightly larger predicted mass. P40 was phos- phorylated by PKG to a greater level than by PKA and was not phosphorylated by PKC (Fig. 4A). Recombinant NEDD5, however, was not phosphorylated by PKG in vitro, but was weakly phosphorylated by PKA and strongly phosphorylated by PKC (Fig. 4B). This confirms that P40 is not NEDD5 and suggests that other septins may be substrates for different protein kinases.
In the second approach to confirm that P40 is G-septin, two synthetic peptides (EKIPKTVEIKAIGHVI from G-septin-␣-(97-112) and DLEDKTENDKIRQESM from G-septin-␣-(244 -259)) were prepared from sequences in the domain common to both G-septin isoforms. Neither peptide had any significant sequence homology to any other known septins or to other proteins in the non-redundant data base. They were used together to raise specific antibodies to G-septin in the same sheep. Immunoblot analysis was performed with partially purified P40 and recombinant mouse NEDD5. To test the specificity of the antibodies to G-septin, P40 and recombinant NEDD5 were resolved on 12% SDS-polyacrylamide gel and transferred to nitrocellulose. NEDD5 antiserum 11, raised against a C-terminal sequence (NEDD5-(303-361)), detected only recombinant NEDD5 protein, but not G-septin (Fig. 5A). However, as expected, NEDD5 antiserum 6 cross-reacted with a 40-kDa protein in rat brain extracts since this antiserum was raised against NEDD5-(74 -85) near the conserved GTP-binding region, which is similar in sequence to G-septin (data not shown). The specific G-septin antiserum detected only a single 40-kDa band in the partially purified P40 preparation and did not detect recombinant NEDD5 (Fig. 5B). The specificity of the polyclonal G-septin antibody was further examined in rat brain cytosol and peripheral membrane extracts (Fig. 5C). The antiserum detected only a single band of 40 kDa in peripheral membrane fractions that comigrated on the same gel with P40. This result confirms that G-septin is a peripheral membrane protein. The G-septin antiserum detected only one other protein of 20 kDa in total brain cytosol and peripheral membrane fractions, which was not detected with affinity-purified G-septin antibodies (data not shown).
In the third approach to confirm that the phosphorylated 40-kDa protein is G-septin, we immunoprecipitated G-septin from the partially purified P40 fraction and phosphorylated the immunoprecipitated proteins on the protein G-agarose beads with PKG (Fig. 5D). The immunoprecipitated G-septin was a PKG substrate in vitro (Fig. 5D, lane 5). This band comigrated with P40 that had been phosphorylated before immunoprecipitation (lanes 1-3). This result confirms that the G-septin antiserum recognizes the P40 PKG substrate and that they are the same protein. Finally, G-septin-␣ was directly confirmed to be a PKG substrate by experiments with recombinant His 6 -tagged G-septin-␣. Recombinant G-septin-␣ was recognized by anti-G-septin antibodies (Fig. 6A). It was phosphorylated by PKG slightly better than by PKA in vitro (Fig. 6B, lanes 2 and 3). Since the PKG used in this study was purified by release from a cAMP affinity column with saturating cGMP concentrations, this protein kinase is not fully cGMP-dependent when used for in vitro phosphorylation. However, phosphorylation of recombinant Gseptin-␣ was partly dependent on cGMP, as phosphorylation was reduced when PKG was used in the absence of cGMP and was elevated in the presence of cGMP (Fig. 6C, lanes 2 and 3). Therefore, recombinant G-septin-␣ is a PKG substrate in vitro.
G-septin Is a GTP-binding Protein-The predicted amino acid sequences of all known septins contain consensus elements for GTP binding. To test whether G-septin binds to guanine nucleotides, duplicates of partially purified P40 from rat brain or recombinant His 6 -tagged G-septin-␣ were incubated with [␣-32 P]GTP for 3 h in the dark and photocrosslinked with a short wavelength ultraviolet light. The specificity of the photolabeling was determined by competition with 1 mM unlabeled GTP (Fig. 7). GTP was covalently cross-linked to a 40-kDa protein in partially purified P40 (lanes 1 and 2) and also to recombinant G-septin-␣ (lanes 5 and 8). An unidentified 45-kDa protein was also detected in the P40 preparation ( lanes  1 and 2). Binding to P40 and recombinant G-septin-␣ was efficiently competed by excess unlabeled GTP. This confirms that G-septin-␣ is a guanine nucleotide-binding protein.
Phosphorylation of G-septin in Nerve Terminals-To determine whether G-septin phosphorylation might be of physiological relevance, its phosphorylation in intact cells was investigated using rat brain synaptosomes, which are isolated nerve terminals (32). Synaptosomes were labeled with 32 P i , washed, and stimulated with a membrane-permeable cGMP or cAMP analogue (8-pCPT-cGMP or 8-pCPT-cAMP). G-septin was im-munoprecipitated from the lysate with anti-G-septin antibodies precoupled to protein G-agarose beads. The immunoprecipitates were separated by SDS-PAGE, and protein phosphorylation was detected by autoradiography (Fig. 8). A single major phosphoprotein of 40 kDa was immunoprecipitated from synaptosomes (Fig. 8A, lanes 1 and 2), and its phosphorylation was slightly increased in response to stimulation with either the cGMP and cAMP analogue (lanes 3-8). Each sample contained a constant amount of G-septin as determined by immunoblotting of the same lysates immediately prior to immunoprecipitation (Fig. 8B). The in vivo phosphorylation results were quantified by densitometry (Fig. 8C). The increase in phosphorylation induced by either cyclic nucleotide analogue was small, but statistically significant (p ϭ 0.042 for cGMP and p ϭ 0.002 for cAMP), suggesting that G-septin-␣ is an in vivo phosphoprotein regulated by cyclic nucleotides.
Distribution of G-septin mRNA-Northern blot analysis was performed to determine the tissue distribution of mRNA encoding G-septin-␣. Total RNAs from multiple adult rat tissues, including brain, liver, lung, stomach, testis, intestine, heart, skeletal muscle, aorta, kidney, spleen, pancreas, and thyroid gland, were prepared. The blot was probed with a PCR product containing 196 bp from the 5Ј-coding sequence of G-septin-␣, and the blot was then stripped and rehybridized with a probe to actin to confirm the presence of equal loads of RNA in all lanes (data not shown). Two transcripts of 5 and 2.5 kb were detected in brain (Fig. 9A). No transcript was detected in the other tissues, except for low levels of three bands of 2.6, 3.4, and 6.0 kb in total RNA from testis. As a control for nonspecific binding, a synthetic probe was made from the plasmid vector, and this probe did not detect any bands (data not shown). The G-septin result was confirmed by Northern blotting using a synthetic primer containing 61 bp corresponding to the nucleotide sequence G 1043 -G 1103 at the 3Ј-end of the G-septin-␣ coding sequence. This detected the 5.0-kb transcript strongly and the 2.5-kb transcript weakly in total rat brain RNA (Fig.  9B), suggesting that the 5.0-kb band is G-septin-␣ and that it is exclusively expressed in rat brain. DISCUSSION We report here on the discovery and cloning of a novel neuronal septin called G-septin, which is a relatively specific substrate for PKG in vitro. It is the first septin shown to be phosphorylated, and its phosphorylation is regulated by cyclic nucleotides in nerve terminals in vivo. G-septin was found to be brain-specific and to occur as two alternatively spliced variants. Although sharing the major GTPase domain with other mammalian septins, G-septin lacks their C-terminal coiled-coil domain.
Several criteria were used to identify the PKG-phosphorylated rat brain 40-kDa protein (P40) as G-septin. Amino acid sequencing of two major peptides from P40 revealed a homology to members of the septin family, but no identity, and GTP binding studies demonstrated that a GTP-binding protein comigrated with P40 on gels. Cloning of the cDNA revealed the sequence of a new septin and further demonstrated that there are different splice variants of G-septin in rat brain. An antibody against two synthetic peptides of G-septin-␣ was raised in sheep and used for immunoblotting. The antiserum specifically reacted with the partially purified 40-kDa protein, and conversely, immunoprecipitated G-septin was phosphorylated by PKG. Finally, to complete the identification of P40 as G-septin, we showed that recombinant G-septin-␣ was also an in vitro substrate for PKG. These results demonstrate that P40 is Gseptin. The amino acid sequence of three other minor peptides derived from P40 did not match the final cloned protein. It is possible that these peptides represent an additional protein or proteins that comigrated with the partially purified G-septin.
G-septin is the first septin reported to be post-translationally modified. We discovered G-septin in the context of a screen for neuronal PKG substrates (28,29,31). It was also found in this study that NEDD5 was an in vitro substrate for PKC. These results suggest that other septins may be controlled through the cellular protein phosphorylation network. Immunoprecipitation experiments revealed that G-septin is also an in vivo phosphoprotein in nerve terminals. Phosphorylation in nerve  . 8. Phosphorylation of G-septin in nerve terminals. Rat brain synaptosomes were labeled with 32 P i , washed, and stimulated with 8-pCPT-cGMP or 8-pCPT-cAMP (200 M) for 15 min. Synaptosomes were lysed, and G-septin was immunoprecipitated with protein G-purified anti-G-septin antibodies. A, phosphorylation of G-septin in vivo was detected by SDS-PAGE on a 7.5-15% acrylamide gel, followed by autoradiography. Results are representative of five independent experiments in duplicate or triplicate, but each was performed with slightly different immunoprecipitation conditions. B, aliquots from the synaptosomal lysates used in A were resolved by SDS-PAGE on a 12% acrylamide minigel, and the proteins were transferred to nitrocellulose. The membrane was probed with the anti-G-septin antibody to show constant G-septin concentrations in each of the samples. C, phosphorylation from A was quantitated by densitometry, and results are expressed in arbitrary units as a percent of unstimulated synaptosomes Ϯ S.E. (n ϭ 4 for controls (Ctrl) and 3 for the cyclic nucleotide-stimulated samples). terminals was elevated by cGMP and cAMP analogues. This suggests that G-septin might be a substrate for both protein kinases in vivo or that there is cross-talk between cyclic nucleotides for the activation of each protein kinase. The small increase in cGMP-stimulated phosphorylation might relate to the widespread distribution of G-septin in brain and the relatively more limited PKG distribution. The relatively high basal phosphorylation might also indicate a potential role for other protein kinases in neurons. We found that NEDD5 antiserum 6, which cross-reacts with G-septin, also immunoprecipitates a 32 P-labeled 40-kDa protein from intact nerve terminals. The phosphorylation of this protein was increased after cell treatment with cGMP, but not with phorbol esters. 2 During the isolation of P40, G-septin was phosphorylated much better by PKG than by PKA, but recombinant His 6 -tagged G-septin was phosphorylated only slightly better by PKG. Thus, it is not yet clear whether G-septin has high selectivity for PKG or whether the presence of the additional amino acids in the recombinant protein may have affected kinase specificity. Inspection of the predicted amino acid sequence of G-septin revealed two potential phosphorylation site motifs at Thr 55 (RKKTMK) and Ser 91 (SRKASSWNR) in G-septin-␣, which match the motif necessary for efficient PKG phosphorylation determined in studies with peptide libraries on cellulose papers (45). Those studies revealed the minimal motif RKX(S/T) and highlighted a key role for additional basic amino acids on either side of RK for optimal V max . Our preliminary phosphoamino acid analysis studies suggested that the in vitro PKG phosphorylation site for P40 is exclusively on serine, which points to Ser 91 as the most likely phosphorylation site. This site is present in both G-septin splice variants, but it is not present in the septin KIAA0991, which is highly related to G-septin in this region, or in any other septin. We are currently determining the phosphorylation site by amino acid sequencing of PKG-phosphorylated G-septin.
PKG or cGMP signaling pathways have been linked to a variety of functions in the brain that have been reviewed (29).
In several more recent studies, PKG signaling has been shown to regulate neurotransmitter release from nerve terminals (46), to control proliferation of sensory neurons (47), to enhance neuronal cell survival (48), to mediate long-term depression in the dentate gyrus (49), to mediate long-lasting changes in synaptic transmission in ciliary ganglions (30), to mediate earlystage memory formation in rats (50), and to switch nerve growth cone repulsion to attraction (51). Other major roles include gene expression at the transcriptional level (52) and at the level of pre-mRNA splicing (31). The phosphorylation of G-septin by PKG suggests that septin function may also be modulated by PKG in vivo, when cells are appropriately activated. Thus, it is possible that the PKG/G-septin connection described in this report may underlie some of these known roles for PKG in the brain.
Like other septins, G-septin contains conserved GTPase sequence motifs, and our data directly demonstrated that Gseptin is also a GTP-binding protein. In the purification scheme, another GTP-binding protein of ϳ45 kDa was detected in the G-septin extract, which might represent another septin that associates with G-septin during the purification, as observed for other septins (14,25). Both GTP binding and GTPase activities have been demonstrated in Drosophila and mammalian septins. Septin filaments composed of PNUT, SEP1, and SEP2 in Drosophila exhibited GTP binding and GTPase activities, but these activities did not affect the polymerization state of the septin filament in vitro (9). Microinjection of GTP␥S and transfection of the GTP-binding mutant of mouse neuronal Nedd5 into HeLa cells disrupted NEDD5-containing filaments, suggesting that the GTPase activity is required for mammalian septin filament assembly in vivo (14). The potential role of the GTPase activity of G-septin remains to be determined. However, the close localization of both potential phosphorylation sites in G-septin to the first GTP-binding motif raises the possibility that phosphorylation could regulate guanine nucleotide binding or the GTPase activity of the enzyme.
Septin filaments act as intracellular scaffolds involved in the organization of cytoskeletal proteins, neuronal polarity, and vesicle trafficking. Assembly of septin filaments is regulated by signals from growth factors, the actin filaments, and GTP hydrolysis (14). Most of the septins appear to interact with each other in order to assemble as filaments, and although it is not yet clear if G-septin shares this property, four major groups of evidence support this possibility. First, the septins co-localize in groups. In yeast, Cdc3p, Cdc10p, Cdc11p, and Cdc12p colocalize to the mother neckbud (21,53); in Drosophila, PNUT, SEP1, and SEP2 co-localize at cleavage furrows (7); and in yeast and mammalian brain, certain septins robustly copurify (8,14,25). Second, yeast two-hybrid screens showed that yeast Sep7p/Shs1p associates with Cdc12p, but not with other septins (6). Third, the septin localizations are interdependent. Temperature-sensitive mutations in Cdc3p, Cdc10p, Cdc11p, or Cdc12p in yeast cause disappearances of all four proteins from the septin ring (22)(23)(24), and Drosophila PNUT mutants fail to localize SEP1 in neurons (8). Fourth, purified septin complexes from Drosophila (9) or rat brain (25) assemble together as filaments. It is therefore possible that G-septin may also interact with other septins. The lack of a coiled-coil domain does not preclude this potential function, as yeast Cdc10p also interacts with other yeast septins, yet has no coiled-coil domain (21)(22)(23)(24)53). However, an analysis of the rat brain septin filament that binds neuronal SEC6-SEC8 proteins did not identify G-septin as a component (25).
While this work was being prepared for publication, the sequence of a mouse protein called septin 3 (SEP3) was deposited in the GenBank TM /EBI Data Bank (accession number AAD02884). Alignment of G-septin-␣ and -␤ with SEP3 revealed the latter to represent a potential third splice variant. It has an alternative C terminus, only partially shared with G-septin-␤, and a 117-amino acid extension comprising multiple repeats of CV, CM, or CVY that shows no homology to any protein in the non-redundant data base. An internal splice site in SEP3 resulted in the replacement of a 15-amino acid sequence from G-septin-␣ ( 221 KQRVRKELEVNGIEF 235 ) with a 6-amino acid sequence ( 208 LPSEGI 213 ) in SEP3. The N termini of G-septin-␣ and SEP3 are identical, except that the first 13 amino acids (MSKGLPEARTDTA) are absent in SEP3. Our data show no evidence for the expression of this protein in rat brain. SEP3 sequence predicts that it is almost 8 kDa larger than G-septin-␣ and would be expected to migrate at a higher position on SDS-PAGE. No such band was detected in rat brain cytosol or peripheral membrane extracts with our G-septin antiserum, which was raised against amino acid sequences that are identical in SEP3; however, our Northern blot analysis supports the possibility of a transcript for SEP3. The primer used from the 5Ј-end of G-septin-␣ was 71% identical to the 5Ј-end of SEP3 and detected 5-and 2.5-kb transcripts, whereas the second primer from the 3Ј-end of G-septin-␣ was 28% identical to SEP3 and detected the 2.5-kb transcript rather weakly. Thus, the 5-kb transcript is G-septin-␣, whereas the 2.5-kb transcript might represent SEP3 or another splice variant. Neither transcript is likely to be G-septin-␤, which has no homology to either probe employed in this study.
G-septin is highly expressed in brain, and Northern blot analysis demonstrated that G-septin-␣ mRNA is brain-specific. No message was detected in the two endocrine tissues tested, thyroid and pancreas. We expect that G-septin may prove to be enriched in neurons, as our preliminary data with the G-septin antiserum indicated high protein levels in nerve terminals isolated from rat brain, 2 and the protein is phosphorylated in nerve terminals. CDCrel-1 is also brain-specific (26). CDC10, H5, and NEDD5 are abundant in adult brain, whereas DIFF6 is not highly expressed there (26). NEDD5 was originally described as a neuronal protein highly expressed in neuronal precursor cells and was down-regulated in adult brain (12), but is present in a variety of tissues. However, G-septin expression in rat brain increases from birth to adulthood, as judged by Western blotting, 2 suggesting that G-septin has a role in mature neurons. The roles of other septins in cytokinesis in yeast and mammalian cells, formation of filaments, and regulation of exocytosis from neurons have now been established. The localization of the new G-septin to brain, which is largely nonmitotic, and the specific phosphorylation of this protein by PKG suggest that G-septin may mediate some cGMP and PKG signaling pathways in the brain that modulate neuronal function.