Septin 11 Is Present in GABAergic Synapses and Plays a Functional Role in the Cytoarchitecture of Neurons and GABAergic Synaptic Connectivity

Abstract

Mass spectrometry and immunoblot analysis of a rat brain fraction enriched in type-II postsynaptic densities and postsynaptic GABAergic markers showed enrichment in the protein septin 11. Septin 11 is expressed throughout the brain, being particularly high in the spiny branchlets of the Purkinje cells in the molecular layer of cerebellum and in the olfactory bulb. Immunofluorescence of cultured hippocampal neurons showed that 54 ± 4% of the GABAergic synapses and 25 ± 2% of the glutamatergic synapses had colocalizing septin 11 clusters. Similar colocalization numbers were found in the molecular layer of cerebellar sections. In cultured hippocampal neurons, septin 11 clusters were frequently present at the base of dendritic protrusions and at the bifurcation points of the dendritic branches. Electron microscopy immunocytochemistry of the rat brain cerebellum revealed the accumulation of septin 11 at the neck of dendritic spines, at the bifurcation of dendritic branches, and at some GABAergic synapses. Knocking down septin 11 in cultured hippocampal neurons with septin 11 small hairpin RNAs showed (i) reduced dendritic arborization; (ii) decreased density and increased length of dendritic protrusions; and (iii) decreased GABAergic synaptic contacts that these neurons receive. The results indicate that septin 11 plays important roles in the cytoarchitecture of neurons, including dendritic arborization and dendritic spines, and that septin 11 also plays a role in GABAergic synaptic connectivity.

We have recently developed a method for the preparation of a brain fraction enriched in GABAergic postsynaptic complex (1). This fraction, insoluble in Triton X-100, was enriched in Gray's type-II postsynaptic densities (type-II PSDs)² and in the postsynaptic GABAergic markers GABA_A receptors (GABA_ARs) and gephyrin. Here we report that septin 11 is a major component of the type-II PSD fraction.

Septins are a family of proteins with GTPase activity that form heterooligomeric filaments and ringlike structures that act as diffusion barriers and scaffolds. Septins are involved in cytokinesis, positioning of the mitotic spindle, cellular morphology, vesicle trafficking, apoptosis, neurodegeneration, and neoplasia (2–5). In mammals, 14 septin genes have been identified. Each septin gene is expressed in several spliced forms. Although most septins are highly expressed in the brain (6), only recently is their role in neuronal function (7–9) and in neuropathology (10–14) is beginning to be addressed for some septins.

Septin 11 is expressed in various tissues, including the brain (15), but little is known about the role of septin 11 in the brain. Septins 3, 5, 6, and 7 are localized in the presynaptic terminals, frequently associated with synaptic vesicles (6, 16, 17). In neurons, septin 11 forms heterooligomeric complexes with septin 7 and septin 5 (9, 18). Nevertheless, the regional and developmental distribution of septin 11 in the brain and in hippocampal cultures is not identical to that of septin 7 or septin 5 (8). These results and other heterooligomerization studies show that septin 11 is not always associated with septin 7 and septin 5 (7, 15, 19). Thus, septin 11 is expected to have functional properties both similar to and different from those of septin 7 and other septins that heterooligomerize with septin 11. In the present paper, we show that septin 11 is associated with the GABAergic synapses, particularly with the postsynapse, and concentrates at the neck of dendritic spines in the intact brain. Others have recently shown that another septin (septin 7) accumulates at the base of dendritic protrusions of cultured neurons (8, 9). However, it is not known whether septins also accumulate at the base of the dendritic spines in the brain. To the best of our knowledge, this is the first time that (i) a septin has been shown to be associated with GABAergic synapses and (ii) a septin has been shown to concentrate at the neck of dendritic spines and dendritic branching points in the intact brain.

EXPERIMENTAL PROCEDURES

Animals

All of the animal protocols have been approved by the Institutional Animal Care and Use Committees of the University of Connecticut and followed the National Institutes of Health guidelines.

Antibodies

A novel rabbit antiserum (Rb anti-septin 11-N) was raised to the N-terminal synthetic peptide amino acids 1–14 of rat septin 11 (MAVAVGRPSNEELR; Fig. 1) that was covalently coupled, via an added C-terminal cysteine, to keyhole limpet hemocyanin (Pierce). The antibody was affinity-purified on immobilized synthetic peptide, and the specificity was tested by enzyme-linked immunosorbent assay, immunoblotting, immunofluorescence of transfected HEK 293 cells, brain immunocytochemistry, and displacement by the antigenic peptide. Another Rb anti-septin 11 antibody to amino acids 366–429 of the porcine septin 11 (Rb anti-septin 11–366) was affinity-purified on the immobilized fusion protein as described elsewhere (15, 19). The mouse anti-septin 7 ascites fluid was a gift from Dr. Randall Walikonis (University of Connecticut) (20). The guinea pig anti-rat γ2 (to amino acids 1–15) antibody was raised in our laboratory and affinity-purified on immobilized peptide. The specificity of the guinea pig anti-rat γ2 antibody has been determined previously (21–28). The Rb anti-gephyrin antiserum used in immunoblotting was a gift from Dr. Ben Bahr (University of Connecticut) (29), and the mouse monoclonal antibody to gephyrin (clone 7a) used in rat brain immunocytochemistry was from Synaptic Systems (Gottingen, Germany). The sheep anti-GAD antiserum was a gift from Dr. Kopin (National Institutes of Health). For other primary and secondary antibodies, see the supplemental material.

FIGURE 1.

Septin 11 is enriched in a GABAergic type-II PSD fraction. A, comparative analysis by SDS-PAGE of the main protein components revealed by Coomassie Blue stain in one-Triton PSD and the fractions enriched in glutamatergic type-I PSDs and GABAergic type-II PSDs. The arrow shows the most prominent protein band (∼50 kDa) enriched in type-II PSDs. LC-MS/MS shows that septin 11 is the main component of this protein band. The same amount of total protein (4 μg) was added in each lane. B, immunoblots of the one-Triton PSD, type-I PSD fraction, and type-II PSD fraction with Rb anti-septin 11-N or Rb anti-septin 11–366 show that septin 11 (filled arrows) is enriched in type-II PSDs over type-I PSDs. Septin 11 is also present in one-Triton PSDs, which contain both type-I and type-II PSDs. Immunoblots with anti-gephyrin and anti-PSD-95 antibodies show that gephyrin (93 kDa; empty arrow) concentrates in the type-II PSD fraction, whereas PSD-95 (95 kDa; arrowhead) concentrates in type-I PSD fraction. The same amount of total protein (2 μg) in each PSD fraction was transferred to each strip. C, amino acid alignment of the N and C termini of rat septin 11 splice variants. All of the splice variants have identical amino acid sequence except for the 6 or 7 amino acids of the C terminus. The Rb anti-septin 11-N antibody was raised to amino acids 1–14 of the N terminus, which is common to all of the splice variants. The numbers at the top represent the amino acid positions.

cDNA Cloning of Rat Septin 11

See the supplemental material.

Generation of the Small Hairpin RNAs (shRNAs)

Two shRNAs (sh1 and sh2) targeting the coding sequence of the rat septin 11 were subcloned into mU6pro vector, respectively. The antisense strand of each shRNA perfectly matched the target mRNA, whereas the sense strand included a mismatch near the middle of the sequence (supplemental Fig. 3A). A DNA oligonucleotide encoding both arms of the shRNA was annealed with the corresponding reverse DNA strand, and the double-stranded DNA was ligated between the BbsI and XbaI sites of mU6pro vector (30, 31). The control shRNAs were generated by introducing three point mutations in the sense and antisense arms of the corresponding shRNA (supplemental Fig. 3A).

FIGURE 3.

Septin 11 concentrates in many GABAergic synapses and also in dendritic branching points and at the base of dendritic protrusions. A–D, cultured hippocampal neurons (21 DIV) were triple-labeled with Rb anti-septin 11-N (A), guinea pig anti-γ2 (B), and sheep anti-GAD (C). The arrows show septin 11 clusters (red) colocalizing with γ2-GABA_AR clusters (green) and GAD-containing boutons (blue). E–H, neurons were triple-labeled with Rb anti-septin 11–366 (E), sheep anti-GAD (F), and guinea pig anti-vGlut1 (G). The white arrowheads show septin 11 clusters (red) colocalizing with GAD-containing boutons (blue), and the black arrowhead indicates a septin 11 cluster colocalizing with vGlut1 (green). I–N, neurons were triple-labeled with guinea pig anti-γ2 (I), Rb anti-septin 11-N (J), and mouse anti-septin 7 (K). The black arrows in I, K, M, and N show septin 11 clusters (green) colocalizing with septin 7 (blue) and γ2 (red) clusters. The black arrowheads in I and M show septin 11 clusters (green) colocalizing with γ2 (red) in the absence of septin 7 clusters. The white arrows in L show septin 11 clusters (green) colocalized with septin 7 clusters (red). In L only, septin 7 is shown in red to better appreciate the colocalization with septin 11 (green) in the overlay. O–S, confocal image of the molecular layer of the rat cerebellum from brain sections triple-labeled with Rb anti-septin 11-N (P), mouse monoclonal antibody anti-gephyrin (Q), and guinea pig anti-vGAT (R). The arrows indicate septin 11 clusters (red) colocalizing with gephyrin (green) and vGAT (blue). T–Z′, neurons were labeled with Rb anti-septin 11-N (T and X), phalloidin-TRITC (U and Y), and monoclonal antibody MAP2 (V), and color overlays are shown in W, Z, and Z′. The arrow indicates septin 11 accumulating at dendritic branching points. The arrowheads point to septin 11 (green) localized at the base of dendritic protrusions. The scale bar in A represents 5 μm for A–N. The scale bar in O represents 30 μm for O and 18 μm for P–S. The scale bar in T represents 5.9 μm for T–W, 3.4 μm for X–Z, and 1.4 μm for Z′. over, panels with color overlays.

Subcellular Fractionation of Rat Brain and Preparation and Fractionation of the “One-Triton PSD” Fraction

Three-month-old Sprague-Dawley male rats were used in these and other fractionation experiments. The preparation of the crude synaptosomal P2 fraction was done according to Carlin et al. (32). The one-Triton PSD fraction (20) was prepared from purified synaptosomes, as described elsewhere (1). For the preparation of type-I and type-II PSD fractions, the one-Triton PSD pellet was suspended in 0.32 m sucrose and loaded onto a continuous sucrose gradient 0.32–2.0 m in 1 mm NaHCO₃, centrifuged at 201,800 × g for 16 h, and fractionated in 1-ml fractions, as described elsewhere (1). Sucrose fraction number 6 (ρ = 1.10 g/ml) is enriched in GABAergic type-II PSDs, whereas sucrose number 10 (ρ = 1.20 g/ml) is enriched in glutamatergic type-I PSDs (1). All steps were carried out at 4 °C, and all fractions were suspended in 50 mm Tris-HCl, pH 7.4, aliquoted, and stored at −70 °C. Immunoblots for were done according to De Blas and Cherwinski (33), and for supplemental Fig. 1 and supplemental Fig. 3, the chemiluminiscent protein bands were visualized with a Chemi Doc (Bio-Rad), and the intensities of the bands were quantified by Quantity One (Bio-Rad).

Identification of Proteins in the Type-II PSD Fraction by Mass Spectrometry

The type-II PSD and type-I PSD fractions were loaded on SDS-PAGE in parallel and stained with Coomassie Blue. The main protein band, enriched in the type-II PSD over the type-I PSD fraction, was collected, digested with trypsin, and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) at the Yale University Cancer Center Mass Spectrometry Resource and W. M. Keck Foundation Biotechnology Resource Laboratory (New Haven, CT). A Waters Q-Tof mass spectrometer (Waters, Milford, MA) was used. The MS/MS fragmentation spectra were analyzed by Mascot Distiller, and proteins were identified by searching a Mascot data base.

Cell Cultures, Immunofluorescence, and Transfection

Hippocampal cultures were prepared according to Goslin et al. (34), as described elsewhere (22, 23, 35). HEK 293 cells or rat neuroblastoma B103 cells were cultured on poly-l-lysine-coated 18-mm coverslips, as described elsewhere (25, 36). The immunofluorescence procedure has been described elsewhere (24, 37). Cultured high density hippocampal neurons (10 DIV), HEK 293 cells, or rat neuroblastoma B103 cells were transfected with 2 μg total of various plasmids for protein or shRNA expression using the CalPhos mammalian transfection kit (BD Biosciences), following the instructions of the manufacturer. Four days later, after transfection (14 DIV for neurons), cells were subjected to immunofluorescence, as described above. Quantification of the effect of shRNAs on the expression of septin 11 was done after nucleofection of rat neuroblastoma B103 cells using the Cell Line Nucleofector Kit L (Amaxa GmbH, Koln, Germany), following the instructions provided by the manufacturer.

Image Acquisition and Analysis

Fluorescence images of neuronal cultures were collected using a ×60 panfluor objective on a Nikon Eclipse T300 microscope with a Sensys KAF 1401E CCD camera, driven by IPLab version 4.0 (Scanalytics, Rockville, MD) acquisition software. Brain section immunofluorescence images were acquired on a Leica TCS SP2 laser confocal microscope using a HCXPL Apo ×100 oil CS objective lens and a pinhole set at 1 Airy unit. For qualitative analysis, images were processed with Photoshop version 7.0 (Adobe, San Jose, CA), as described elsewhere (1, 24, 26, 27).

Quantification of Clusters, Dendritic Protrusions, and Sholl Analysis

For determining the density of γ2 clusters, septin 11 clusters, and dendritic protrusion and the length of the latter, 30 dendrites were randomly selected from 15–20 cells from 3–4 separate experiments. The maximum intensities of the fluorophore channel were normalized, and the low intensity and diffuse nonclustered background fluorescence signal seen in the dendrites was subtracted. The density of clusters or dendritic protrusions was calculated as the number of clusters or dendritic protrusions/100 μm². Cluster colocalization in two different fluorescence channels was determined by overlaying the images. A cluster in a fluorescence channel was considered to colocalize with a cluster (or dendritic protrusion) in the other channel when >66% of the surface of one of the clusters overlapped with the other cluster (or dendritic protrusion). The length of the dendritic protrusions was determined after manual tracing followed by measurement by IPLab version 4.0 (Scanalytics). The total number of clusters, spines, or puncta counted per transfection condition (mean ± S.E.) was 461 ± 15 for septin 11, 535 ± 63 for γ2-GABA_AR, 623 ± 43 for spine density, 989 ± 52 for spine length, and 821 ± 54 for PSD-95 and 845 ± 48 for vGut1. For quantifying the GABAergic innervation, 25 pairs of transfected neurons and their corresponding nontransfected sister neighbor neurons from 3–4 experiments were selected, and the number of GAD⁺ boutons was counted. The total number of GAD boutons counted per transfection condition was 732 ± 71 in transfected cells and 867 ± 74 in nontransfected cells. For Sholl analysis (38), the images of 30 transfected neurons from three experiments were taken with a ×40 objective, and the dendrites of the transfected neuron were visualized by EGFP or mCherry fluorescence. Concentric circles with 10-μm differences in radius were drawn around the cell body, and the number of dendrites crossing each circle was manually counted.

Light Microscopy Immunocytochemistry of Rat Brain Sections

This procedure has been described elsewhere (27, 39–42). See supplemental material.

Preembedding Electron Microscopy (EM) Immunocytochemistry

The procedure has been described elsewhere (25, 36). See supplemental material.

Postembedding Electron Microscopy Immunogold

For EM immunogold, the Lowicryl-embedded block was a kind gift of Drs. Peter Somogyi (Oxford University, Oxford, United Kingdom) and Zoltan Nusser (Institute of Experimental Medicine, Budapest, Hungary). The preparation of the block has been described elsewhere (43). The postembedding immunogold labeling was done in our laboratory, as described elsewhere (44, 45). See the supplemental material. All sections were visualized with a Tecnai Biotwin 12KV transmission electron microscope (FEI Corp., Eindhoven, Holland).

RESULTS

Septin 11 Is Enriched in a GABAergic Type-II PSD Brain Fraction

There is limited knowledge about the molecular components and organization of the GABAergic type-II PSDs compared with that of the glutamatergic type-I PSDs, in part because until recently (1), there was no method for the isolation from the rat brain of subcellular fractions enriched in type-II PSDs. We have recently shown that the classical one-Triton PSD fraction (made of the insoluble material after Triton X-100 extraction of the brain synaptosomal fraction) (20, 32, 46) contains PSDs from both GABAergic and glutamatergic synapses, which can be separated by centrifugation in a continuous sucrose gradient (1). The GABAergic type-II PSDs, which are enriched in GABA_ARs and gephyrin, have lower density than the glutamatergic type-I PSDs, which are enriched in the postsynaptic marker PSD-95 protein (1). In an effort to identify other proteins present in the GABAergic type-II PSDs, we did a comparative analysis by SDS-PAGE of the major protein components in the fractions enriched in type-II and type-I PSDs. The type-II PSD fraction showed a prominent protein band (∼50 kDa) that was not obviously present in the type-I PSD fraction (Fig. 1A, arrow). The LC-MS/MS analysis of the tryptic peptides of this protein band identified septin 11 as the major component with 14 peptide matches. Immunoblotting with the Rb anti-septin 11-N and Rb anti-septin 11–366 confirmed the enrichment in the 50-kDa septin 11 in the type-II PSD fraction over the type-I PSD fraction (Fig. 1B, filled arrows). Septin 11 was also present in the one-Triton PSD, from which type-I and type-II PSD fractions were derived. The type-II PSD fraction was also enriched in gephyrin (93 kDa) over the type-I PSD fraction (4.1-fold), whereas the type-I PSD was enriched in PSD-95 (95 kDa) over the type-II PSD fraction (6.2-fold), as shown in Fig. 1B (empty arrow and filled arrowhead, respectively). We have shown elsewhere that the type-II PSD fraction is also enriched in GABA_ARs, containing the γ2 subunit (1). Thus, although the type-II PSD fraction is enriched in GABAergic postsynaptic markers, it contains some type-I PSDs. In turn, the type-I PSD preparation contains some type-II PSDs, as we have discussed elsewhere (1). Nevertheless, septin 11 is significantly enriched in the type-II PSD fraction over the type-I PSD fraction (7-fold, as shown in Fig. 1B (arrows)).

Cloning of Splice Variants of Septin 11 cDNAs from a Rat Brain Library

Four mRNA splice variants of septin 11 (transcripts I–IV) were cloned from an adult rat brain Marathon-ready cDNA library. The four variants differ in (i) the presence or absence of nucleotides encoding the 6 or 7 amino acids of the C terminus (Fig. 1C) and (ii) the 3′-UTRs. Sequence comparison of these isoforms to the rat genome revealed that the septin 11 gene is localized on chromosome 14, region p22. The four splice variants share nine exons (exons 1–9), which include bases 1–1274 of the coding region. Additional exons encode a 6- or 7-amino acid tail, added to the C terminus, and the 3′-UTR, which varies depending on the splice variant. Thus, transcripts I and II also share the 10th exon encoding the same C terminus heptapeptide and part of the 3′-UTR, but they differ in the 11th exon, thus having the 3′-UTR partially different from each other. Transcript III has the 10th and 11th exons both different from that of the other splice variants. The 10th and 11th exons of this variant encode a C terminus hexapeptide (instead of the heptapeptide encoded in transcripts I and II) and a different 3′-UTR. Transcript IV has no equivalent of the 10th exon that is present in transcripts I–III, and therefore transcript IV has the 6 or 7 amino acids of the C terminus missing (Fig. 1C). Thus, the septin 11 protein encoded by this transcript is 6 amino acids shorter than that of transcript III and 7 amino acid shorter than that of transcripts I and II. A different splice variant of rat septin 11 has also been reported in GenBank^TM (NP_001100678), which encodes the same septin protein as transcripts I and II, including the 7-amino acid C terminus (Fig. 1C). Nevertheless transcript NP_001100678 and transcripts I and II have different 3′-UTRs encoded by three different 11th exons. Therefore, different mRNA transcripts can either encode the same septin 11 amino acid sequence or septin 11 with different C termini. We have deposited the sequences of transcripts I–IV in GenBank^TM (EU711414; EU711415, EU711416 and EU711417, respectively).

Distribution of the Expression of Septin 11 in Rat Brain

A rabbit antibody (Rb anti-septin 11-N) to amino acids 1–14 of the N terminus of rat septin 11, common to all of the rat splice variants, was generated. Immunoblots of rat forebrain showed that the affinity-purified antibody specifically recognized a 50-kDa protein and that the immunoreactivity was blocked by the antigenic peptide (Fig. 2H, arrow). Light microscopy immunocytochemistry showed that septin 11 is highly expressed in cerebellum, olfactory bulb, hippocampus, cerebral cortex, thalamus, and corpus striatum (Fig. 2A). Higher magnification of the hippocampus (Fig. 2, B and C) revealed stronger immunoreactivity around the pyramidal cells of the stratum pyramidale (SP) and in the stratum lucidum (SL) of the CA2-CA3 regions (Fig. 2C), in regions known to be enriched in synapses. High immunoreactivity was found in the various layers of the olfactory bulb (Fig. 2D), particularly in the external plexiform layer (EP). Relatively strong immunoreactivity was also found in various layers of cerebral cortex (Fig. 2E). In the cerebellum, septin 11 concentrates in the molecular layer of cerebellum (Fig. 2F), particularly in the spiny branchlets of the Purkinje cells (Fig. 2G). Beaded accumulations of immunoreactivity were also observed along the main Purkinje cell dendrites (Fig. 2G, arrows), which is consistent with the notion that septin 11 accumulates at the branching points of the Purkinje cell dendrites.

FIGURE 2.

Immunocytochemical distribution of septin 11 in the rat brain. A, parasagittal section of the rat brain immunolabeled with Rb anti-septin 11-N. Very high immunoreactivity is observed in the molecular layer of the cerebellum (CB) and the olfactory bulb (OB). Less, but still high, immunoreactivity was observed in the hippocampus (HP), cerebral cortex (CC), and thalamus (TH). B–G, higher magnification of several brain regions, including the hippocampus and dentate gyrus (B), CA3 region of the hippocampus (C), olfactory bulb (D), cerebral cortex (E), cerebellum (F), and the molecular layer of the cerebellum (G). SO, stratum oriens; SP, stratum pyramidale; SL, stratum lucidum; SR, stratum radiatum; GR, granule cell layer; GL, glomerular layer; EP, external plexiform layer; ML, molecular layer; PK, Purkinje cell layer. Scale bar, 2.4 mm (A), 250 μm (B), 100 μm (D and F), 50 μm (C and E), and 20 μm (G). H, immunoblot of the P2 brain fraction with the affinity-purified antibody to septin 11 N-terminal (lane 1). The immunoreactivity of the 50-kDa protein (arrow) was blocked with 20 μg/ml antigenic peptide (lane 2).

Expression of Septin 11 during Rat Development

See the supplemental material.

Septin 11 Concentrates in Many GABAergic Synapses but Is Also Present in Some Glutamatergic Synapses

We have shown above that septin 11 is enriched in a GABAergic type-II PSD fraction over a glutamatergic type-I PSD fraction (Fig. 1), suggesting that septin 11 concentrates on the postsynaptic complex of GABAergic synapses. Thus, we have tested whether septin 11 is localized in GABAergic synapses in the cultured hippocampal neurons and in the intact brain. Triple label immunofluorescence of 21 DIV cultured hippocampal neurons showed that in dendrites, septin 11 forms clusters of 0.20 ± 0.01 μm² average size (n = 486, ranging from 0.09 to 0.95 μm²). The cluster immunofluorescence was blocked by the antigenic peptide, and the number of immunofluorescent clusters was highly reduced by knocking down septin 11 with shRNAs (see below). There was also a “background” fluorescence, much of which corresponded to nonclustered septin 11, since it was highly reduced by the antigenic peptide or septin 11 shRNAs. Some septin 11 clusters were associated with GABAergic synapses, as shown by both the colocalization of septin 11 clusters with postsynaptic γ2-GABA_AR clusters and colocalization/apposition with presynaptic GAD-containing terminals (arrows in Fig. 3, A–D). Similar results were obtained with the two different anti-septin 11 antibodies, Rb anti-septin 11-N (Fig. 3A) and Rb anti-septin 11–366 (Fig. 3E). Triple label fluorescence with GAD and vGlut1 showed that although many GAD⁺ puncta were colocalized with or apposed to septin 11 clusters (white arrowheads in Fig. 3, E–H), some but fewer vGlut1⁺ puncta were also colocalized with or apposed to septin 11 clusters (black arrowhead in Fig. 3, E–H). These septin 11 clusters were associated with glutamatergic synapses, since they colocalized with PSD-95 and vGlut1 (not shown).

Quantification showed that 54 ± 4% (S.E.) of the GABAergic synapses (with both presynaptic GAD⁺ puncta and postsynaptic γ2-GABA_AR⁺ clusters) and 25 ± 2% of glutamatergic synapses (with both vGlut1⁺ puncta and PSD-95⁺ clusters) had colocalizing septin 11 clusters. Thus, a considerably higher proportion of GABAergic synapses over glutamatergic synapses had colocalizing septin 11. The location of these large clusters in dendrites, the better matching of the shape and size of these clusters with that of the postsynaptic markers, and the higher association of these clusters with GABAergic over glutamatergic synapses are consistent with the enrichment in septin 11 in the GABAergic type-II PSD fraction over the glutamatergic type-I PSD fraction.

Neurons triple-labeled with septin 11, septin 7, and γ2 showed that there are considerably more septin 11 clusters (20.8 ± 2.3 clusters/100 μm²) than septin 7 clusters (13.9 ± 1.4 clusters/100 μm², p = 0.01), as shown in Fig. 3, J and K. The majority (92.7 ± 2.0%) of septin 7 clusters (Fig. 3L, red color for septin 7 only in this panel for better visualization of septin 7 in the overlay) colocalized with septin 11 clusters (Fig. 3L, green, white arrows), whereas many septin 11 clusters did not colocalize with septin 7 clusters (Fig. 3, J–L). Although colocalization of septins 11 and 7 does not necessarily result from heterooligomerization of the two septins, these results support the notion that some but not all septin 11 heterooligomerizes with septin 7. Some of the γ2 clusters colocalized with both septin 7 clusters (black arrows in Fig. 3, I, K, and N) and septin 11 clusters (black arrows in Fig. 3, I and M). Other γ2 clusters colocalized only with septin 11 clusters and not with septin 7 clusters (black arrowheads in Fig. 3, I and M). However, we found no γ2 clusters that colocalized with septin 7 but not with septin 11 clusters.

Confocal microscopy analysis of triple label immunofluorescence of the rat brain cerebellum with antibodies to septin 11, the GABAergic postsynaptic protein gephyrin, and the GABAergic presynaptic marker vesicular GABA transporter (vGAT) showed that a significant number of GABAergic synapses in the molecular layer of the cerebellum were associated with septin 11 clusters (Fig. 3, O–S). Often, the colocalizing septin 11 corresponded to individual septin 11 clusters rather than to a continuous dendritic profile. Quantification showed that 53.5 ± 3.4% of gephyrin clusters colocalized with septin 11 clusters (Fig. 3, P and Q, white arrows). Random colocalization of gephyrin and septin 11, measured in the molecular layer of the intact cerebellum after flipping horizontally the gephyrin image, was 8.0 ± 1.6%. The majority of the gephyrin clusters represented GABAergic synapses, since 88.5 ± 1.9% of gephyrin clusters also colocalized with vGAT (white arrows in Fig. 3, Q–S). In contrast, only 19.6 ± 2.5% of PSD-95 clusters had colocalizing septin 11 clusters. In the CA1 region of the intact hippocampus (strata oriens, pyramidale, and lacunosum), 40.7 ± 3.3% of gephyrin clusters colocalized with septin 11 clusters, whereas only 7.7 ± 1.1% of PSD-95 clusters colocalized with septin 11 clusters. Random colocalization of gephyrin and septin 11, calculated as described above, was 4.6 ± 0.5%. Thus, in hippocampal cultures and brain tissue, a considerably higher proportion of GABAergic synapses over glutamatergic synapses have associated septin 11 clusters.

In Cultured Hippocampal Neurons, Septin 11 in Dendrites Concentrates at the Dendritic Branching Points and at the Base of Dendritic Protrusions

Triple label immunofluorescence of 21 DIV neurons with anti-septin 11-N (green), phalloidin-TRITC (to visualize dendritic protrusions and spines, which are rich in F-actin; red), and anti-MAP2 (microtubule-associated protein 2) (to visualize dendrites; blue) revealed that large and abundant septin 11 clusters concentrated at dendritic branching points (arrow in Fig. 3, T–W). Septin 11 clusters were also frequently localized at the base of the dendritic protrusions (arrowheads in Fig. 3, T–Z′). Thus, 79 ± 3% of the dendritic protrusions had associated septin 11 clusters. Neurons triple-labeled with phalloidin-TRITC (red), septin 11 (green), and septin 7 (blue) showed that septin 7 frequently colocalized with septin 11, at the base of the dendritic protrusions (arrowheads in supplemental Fig. 2, A–D). Thus, it is likely that septin 11 and septin 7 are part of the heterooligomeric septin complexes at the base of many of the dendritic protrusions in these hippocampal cultures. Septin 11 was also found in the axons of the cultured hippocampal neurons (21 DIV), as shown by double label immunofluorescence with anti-Tau1 and anti-septin 11-N. However, the fluorescence intensity for septin 11 in the axons was very low and did not show the large septin 11 clusters observed in the dendrites (not shown, because visualization of septin 11 in the axon of mature cultured hippocampal neurons required high enhancement of the signal compared with dendrites). These results in culture are consistent with the EM data of intact brain (see below).

The localization of septin 11 in 3 DIV hippocampal cultures is shown in the supplemental material.

EM Immunocytochemistry

We have shown above by light microscopy immunocytochemistry that septin 11 is highly expressed in the Purkinje cell dendrites (Fig. 2G). Preembedding EM immunocytochemistry of the molecular layer of the rat cerebellum shows that septin 11 is present in Purkinje cell dendrites (PKD in Fig. 4, A, B, and E), concentrating at the neck of dendritic spines of various lengths and shapes (Fig. 4, A–D, F, and G, filled arrows). These dendritic spines received type-I glutamatergic synaptic contacts (Fig. 4, A–D, F, and G, arrowheads, pointing to the type-I PSD). Nevertheless, no septin 11 immunolabeling was found at these glutamatergic synapses, either pre- or postsynaptically. Except for the neck area, the immunoreactivity was absent from the spines. Septin 11 also concentrated at the branching point of the Purkinje cell dendrites (Fig. 4E, empty arrow). Septin 11 immunoreactivity was found also concentrating postsynaptically at type-II GABAergic synapses (Fig. 4H), which could be identified by the flattened synaptic vesicles that are present in the presynaptic terminal and a thin postsynaptic density. The septin 11 immunoreactivity associated with GABAergic synapses was not just the result of a general labeling of the dendritic shaft, where GABAergic synapses occur, since the intensity of the immunolabeling at the EM level of the GABAergic synapses was considerably stronger than in the dendritic cytoplasm, in agreement with the immunofluorescence data in the hippocampal cultures, which showed that the distribution of septin 11 immunoreactivity in dendrites was not uniform. In these cultures, septin 11 forms clusters in dendrites, and a significant number of GABAergic synapses have associated septin 11 clusters. Septin 11 immunolabeling was absent from axons and presynaptic terminals of GABAergic or glutamatergic synapses (Fig. 4, A–H), indicating that the level of septin 11 in the presynaptic terminals is considerably lower than in the dendrites, in agreement with the immunofluorescence results.

FIGURE 4.

EM immunocytochemistry of rat brain cerebellum. A–H, preembedding EM immunoperoxidase labeling with anti-septin 11-N antibody of the molecular layer of the cerebellum. Immunolabeling is observed in the Purkinje cell dendrites (PKD). Intense immunolabeling is observed at the neck of the dendritic spines (filled arrows in A–D and F–G). The dendritic spines receive type-I synaptic contacts (arrowheads in A–G). Intense immunolabeling is present at the branching points of the Purkinje cell dendrites (empty arrow in E) and postsynaptically to type-II GABAergic synapses (H). I–K, postembedding EM immunogold labeling with Rb anti-septin 11-N. Gold particles are associated with the postsynaptic side of some type-II synapses (I) and with the neck (arrows) of the dendritic spines (Sp). The arrowheads point to the type-I PSD of the synaptic contacts on the spines. In H and I, the presynaptic terminal containing flattened presynaptic vesicles is located in the upper half of the panel. The scale bar represents 500 nm in all panels except in I, where it represents 155 nm.

Considerably weaker labeling with the anti-septin 11 antibodies was obtained with the EM postembedding immunogold technique, probably due to the denaturation of the epitope following tissue fixation and embedding. Nevertheless, an accumulation of gold particles corresponding to septin 11 was observed at some type-II synapses (Fig. 4I) and at the neck of dendritic spines (Sp in Fig. 4, J and K, arrows). The arrowheads in Fig. 4, J and K, point to the type-I PSDs present in these spines.

The EM data in the intact cerebellum agrees with the immunofluorescence data of hippocampal cultures, showing that septin 11 clusters concentrate (i) at the base of dendritic spines/protrusions, (ii) at the dendritic branching points, and (iii) in some GABAergic synapses.

Knocking Down Septin 11 by shRNA Leads to Reduced Dendritic Arborization

We designed two shRNAs (sh1 and sh2), specifically targeting all known splice variants of rat septin 11 mRNA, and the corresponding control shRNAs, each containing three point mutations (sh1 3m and sh2 3m; supplemental Fig. 3A). Immunoblots of homogenates from rat neuroblastoma B103 cells nucleofected with the septin sh1 or sh2 showed reduced expression of septin 11 protein (to 19.6 ± 5.9% for sh1 (p < 0.001) and to 42.1 ± 3.2% for sh2 (p < 0.001)) when compared with cells transfected with the empty vector mU6pro (supplemental Fig. 3C). In contrast, cells transfected with sh1 3m (84.8 ± 12.5%, p = 0.21) and cells transfected with sh2 3m (94.0 ± 6.3%, p = 0.30) had no significant effect on the septin 11 expression when compared with cells nucleofected with the mU6pro vector. These experiments show that sh1 and sh2 effectively knock down the expression of rat septin 11.

When cultured hippocampal neurons were transfected with sh1, the density of septin 11 clusters (7.6 ± 0.7 clusters/100 μm²) was decreased compared with that of the sister nontransfected neurons (14.6 ± 0.9 clusters/100 μm², p < 0.001) or neurons transfected with sh1 3m (13.8 ± 1.1 clusters/100 μm², p < 0.001), as shown in supplemental Fig. 3, D and E. Similarly, hippocampal neurons transfected with sh2 showed a significant reduction in the density of septin 11 clusters (7.7 ± 0.8 clusters/100 μm²) compared with that of the sister nontransfected neurons (p < 0.001) or neurons transfected with sh2 3m (14.2 ± 1.2 clusters/100 μm², p < 0.001). In rescue experiments of neurons cotransfected with sh1 or sh2 and the corresponding rescue septin 11 mRNA, the normal density of septin 11 clusters was restored (15.1 ± 1.2 clusters/100 μm² (p = 0.72) for sh1 plus sh1 rescue mRNA and 15.2 ± 1.8 clusters/100 μm² (p = 0.76) for sh2 plus sh2 rescue mRNA), when compared with nontransfected neurons. The rescue septin 11 mRNAs (derived from transcript II) for sh1 or sh2 each carried five silent mutations in the corresponding mRNA sequence targeted by sh1 or sh2, respectively (supplemental Fig. 3B). We chose this transcript for the rescue experiment, because its translation gives the same septin 11 amino acid sequence as transcripts I and NP_001100678 (Fig. 1C). The neurons transfected with EGFP only (15.2 ± 1.0 clusters/100 μm², p = 0.64) have no significant difference in the septin 11 clusters when compared with nontransfected neurons. Comparison between groups using one-way ANOVA Tukey test showed that the neurons transfected with the mutated sh1 3m or sh2 3m or with EGFP or in the sh1 and sh2 rescue experiments or the nontransfected cells had no significant difference (p > 0.05) in septin 11 cluster density.

We believe that the effects of sh1 and sh2 on septin 11 cluster density and the other effects described below are specific. (i) The shRNA target sequences are specific for septin 11 as a data base search shows; (ii) similar effects are obtained with two different septin 11 shRNAs; (iii) the effects were abolished by introducing three point mutations in the shRNAs; and (iv) the phenotypes were rescued by septin 11 mRNA carrying five silent mutations in the target sequence of the corresponding shRNA.

Knocking down septin 11 in cultured hippocampal neurons with sh1 or sh2 led to decreased dendritic arborization (supplemental Fig. 4A) compared with controls (supplemental Fig. 4, B–D). The neurons transfected with sh1 or sh2 (and EGFP) had thinner and shorter dendrites and less branching than the neurons transfected only with EGFP or with sh1 3m or sh2 3m (and EGFP). Supplemental Fig. 4, A–D, shows representative examples for sh1 and various controls, including the rescue control. Quantification by Sholl analysis (38) shows that there was significantly fewer crossings of the dendritic branches with the concentric circles (Fig. 5A) in the neurons transfected with sh1 and EGFP compared with neurons transfected with sh1 3m and EGFP or with EGFP only (reduced to 51.9 ± 2.5% of EGFP, p < 0.001 in a paired t test) or sh2 and EGFP compared with sh1 3m and EGFP or EGFP only (reduced to 50.3 ± 2.8% of EGFP, p < 0.001 in a paired t test). In contrast, there is no significant difference in branching between the neurons transfected with EGFP only and the neurons cotransfected with the mutated sh1 3m or sh2 3m and EGFP (p = 0.96 and p = 0.74, respectively). Cotransfection of neurons with sh1 or sh2 and the corresponding rescue septin 11 mRNA led to a significant recovery of the neuronal branching (p < 0.001, paired t test in both conditions compared with neurons transfected with the sh1 or sh2, respectively), increasing from 51.9 ± 2.5 to 75.6 ± 2.4% for sh1 plus sh1 rescue mRNA and from 50.3 ± 2.8 to 72.4 ± 1.6% for sh2 plus sh2 rescue mRNA. Although the rescue experiments restored the density of septin 11 clusters as shown above, it did not fully restore the dendritic branching. These results suggest that additional septin 11 splice variants are needed for the full rescue of dendritic branching.

FIGURE 5.

Knocking down septin 11 by shRNA leads to decreased dendritic arborization, decreased density, and increased length of dendritic protrusions and decreased number of GABAergic contacts that these cells receive. A, Sholl analysis of dendritic branching of the hippocampal neurons transfected with EGFP only; with EGFP plus sh1, sh1 3m, sh2, or sh2 3m; or with EGFP plus sh1 or sh2 plus the corresponding septin 11 mRNA carrying silent mutations used for rescue. B, quantification of the density and length of the dendritic protrusions in the hippocampal neurons transfected with the aforementioned plasmids. C, triple-labeled immunofluorescence of hippocampal neurons transfected with sh1 and EGFP compared with neurons transfected with sh1 3m and EGFP. The antibodies used were guinea pig anti- γ2 (red) and sheep anti-GAD (blue). The arrowheads indicate the γ2-GABA_AR clusters apposed/colocalized with GAD bouton on the transfected (green) neurons. D, quantification of the density of the γ2-GABA_AR clusters on the neurons transfected with the aforementioned plasmids. E, quantification of the percentage of the GABAergic synaptic contacts (GAD⁺ and γ2⁺ GABA_AR) in transfected neurons compared with those of sister nontransfected neighbor neurons (100%) in the same culture. Scale bar, 5 μm in C and 2.5 μm in the selected dendrites of the right-hand panels in C. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Knocking Down Septin 11 by shRNAs Leads to Longer and Fewer Dendritic Protrusions

The septin 11 knockdown by sh1 and sh2 also had a significant effect on the density and length of dendritic protrusions (Fig. 5B) (illustrated in supplemental Fig. 4, E–H, arrowheads). Compared with neurons transfected with sh1 3m and EGFP (16.6 ± 1.0 protrusions/100 μm²) or sh2 3m and EGFP (16.3 ± 1.1 protrusions/100 μm²), the neurons transfected with sh1 or sh2 (and EGFP) had decreased density of dendritic protrusions (8.8 ± 0.7 protrusions/100 μm² for sh1 and 8.8 ± 0.8 protrusions/100 μm² for sh2, p < 0.001 in both cases). In contrast, neurons transfected with sh1 3m or sh2 3m (and EGFP) had no effect on dendritic protrusion density compared with the neurons transfected with EGFP only (16.5 ± 0.9 protrusions/100 μm²) (p = 0.97 for sh1 3m and p = 0.87 for sh2 3m). Moreover, the mutated septin 11 mRNA rescued the dendritic protrusion density (14.8 ± 0.8 protrusions/100 μm² for sh1 plus sh1 rescue mRNA and 14.7 ± 0.9 protrusions/100 μm² for sh2 plus sh2 rescue mRNA, p < 0.001 for both rescues compared with sh1 and sh2, respectively). Comparison between groups using one-way ANOVA Tukey test showed that the neurons transfected with the mutated sh1 3m or sh2 3m or with EGFP or in the sh1 and sh2 rescue experiments had no significant difference (p > 0.05) in dendritic protrusion density.

Compared with neurons transfected with sh1 3m and EGFP (1.00 ± 0.03 μm) or sh2 3m and EGFP (1.09 ± 0.03 μm), the neurons transfected with sh1 or sh2 (and EGFP) had an increased dendritic protrusion length (1.57 ± 0.08 μm for sh1 and 1.81 ± 0.08 μm for sh2, p < 0.001 in both cases), as shown in Fig. 5B and supplemental Fig. 4, E–H. Neurons transfected with sh1 3m and EGFP had no significant effect on dendritic protrusion length compared with the neurons transfected with EGFP only (0.95 ± 0.05 μm, p = 0.29). Comparison between groups using one-way ANOVA Tukey test showed that the cells transfected with mutated sh1 3m or sh2 3m or with EGFP only had no significant difference (p > 0.05) in the length of dendritic protrusions. Moreover, the mutated septin 11 mRNA led to a significant rescue of the phenotype, restoring the length of the dendritic protrusions (1.17 ± 0.05 μm for sh1 plus sh1 rescue mRNA and 1.19 ± 0.04 μm for sh2 plus sh2 rescue mRNA, p < 0.001 in both cases) compared with that of the neurons transfected with sh1 and sh2, respectively. Comparison between groups using one-way ANOVA Tukey test showed that the length of the dendritic protrusions in the rescue experiment was not significantly different from that of the cells transfected with the controls sh1 3m or sh2 3m. However, there was a small but significant difference in the length of the protrusions of the cells transfected only with EGFP (p < 0.05). Knocking down septin 11 had no significant effect on the size of the head of the dendritic protrusions (not shown).

Knocking Down Septin 11 by shRNAs Leads to a Decreased Number of GABAergic Contacts

Since a significant number of GABAergic synapses have colocalizing septin 11 clusters (Fig. 3), we tested whether knocking down septin 11 affects the density of γ2-GABA_AR clusters and the GABAergic synaptic innervations. Fig. 5, C and D, shows that compared with neurons transfected with sh1 3m and EGFP (12.7 ± 0.8 γ2 clusters/100 μm²) or sh2 3m and EGFP (12.2 ± 1.1 γ2 clusters/100 μm²), neurons transfected with sh1 or sh2 and EGFP showed a small but significant decrease in γ2 cluster density (10.2 ± 0.7 clusters/100 μm² (p = 0.02) for sh1, 9.7 ± 0.6 clusters/100 μm² (p = 0.04) for sh2). The neurons cotransfected with sh1 3m and EGFP or sh2 3m and EGFP had no significant change in the density of GABA_AR γ2 clusters (p = 0.62 and p = 0.95, respectively) when compared with the neurons transfected only with EGFP (12.1 ± 0.9 γ2 clusters/100 μm²). Neurons cotransfected with mutated full-length septin 11 with shRNAs rescued the density of GABA_AR γ2 clusters compared with neurons transfected with sh1 or sh2 and EGFP (13.3 ± 1.2 clusters/100 μm² (p = 0.027) for sh1 plus sh1 rescue mRNA and 12.2 ± 1.0 clusters/100 μm² (p = 0.018) for sh2 plus sh2 rescue mRNA, respectively). Comparison between groups using one-way ANOVA Tukey test showed that the neurons transfected with the mutated sh1 3m or sh2 3m or with EGFP or in the sh1 and sh2 rescue experiments had no significant difference (p > 0.05) in γ2-GABA_AR cluster density. In these low density cultures, there is a limited GABAergic innervation of the pyramidal neurons. Under these conditions, many of the γ2-GABA_AR clusters are nonsynaptic (23, 47). Although 54 ± 4% of the GABAergic synapses (GAD⁺ and γ2⁺) had associated septin 11 clusters, only 18 ± 2% of all of the γ2-GABA_AR clusters, mostly the synaptic ones, had colocalizing septin 11 clusters. The results also show that septin 11 is not involved in the nonsynaptic clustering of γ2-GABA_ARs.

We have also quantified the GABAergic innervation that these neurons received by counting the number of presynaptic GAD⁺ boutons contacting these neurons that showed apposed postsynaptic γ2-GABA_AR clusters (Fig. 5, C and E). We analyzed pairs of transfected neurons and sister nontransfected neighbors in the same culture. Because the GABAergic innervation in these hippocampal cultures is limited, some neurons receive GABAergic innervation while others do not. Therefore, the measurement of “density” of GABAergic synapses per unit of dendritic length shows very large S.D. values. More practical comparative values were derived from measuring the total number of GABAergic synapses (GAD⁺ and γ2⁺) occurring in the whole neuron. We have previously shown that these GAD⁺ and γ2⁺ contacts correspond to synapses with actively recycling synaptic vesicles (23). The neurons cotransfected with sh1 and EGFP or sh2 and EGFP received 61.6 ± 13.4 or 59.2 ± 2.3%, respectively, of the GABAergic contacts that the corresponding nontransfected sister neighbors received (100%). These values are significantly lower than the innervation of neurons transfected with sh1 3m and EGFP (95.9 ± 7.4% of the sister nontransfected neighbors, p = 0.036) or of neurons transfected with sh2 3m and EGFP (92.1 ± 7.0% of the sister nontransfected neighbors, p = 0.002) (Fig. 5E). The GABAergic innervation of the neurons transfected with sh1 3m and EGFP or sh2 3m and EGFP was not significantly different from that of the nontransfected neurons, when compared with neurons transfected only with EGFP (103.1 ± 9.3%, p = 0.48 and p = 0.34, respectively). In the rescue experiments, where neurons were cotransfected with mutated septin 11 mRNA and the corresponding shRNA, the GABAergic innervation was restored, being 84.3 ± 11.0% for the sh1 plus sh1 rescue mRNA and 85.7 ± 6.9% for the sh2 plus sh2 rescue mRNA, not significantly different from the control neurons transfected with sh1 3m or sh2 3m or EGFP only, as shown by one-way ANOVA Tukey test (p > 0.05).

Neurons transfected with sh1 and EGFP had no significant effect on the density of PSD-95 (17.1 ± 1.2 clusters/100 μm²) or vGlut1 (17.4 ± 1.1 puncta/100 μm²) when compared with neurons transfected with sh1 3m and EGFP (16.6 ± 1.0 clusters/100 μm², p = 0.75 for PSD-95 and 16.6 ± 1.0 puncta/100 μm² p = 0.58 for vGlut1). Neurons transfected only with EGFP had 16.1 ± 1.2 clusters/100 μm² for PSD-95 and 16.1 ± 1.1 puncta/100 μm² for vGlut1. Comparison between groups using one-way ANOVA Tukey test showed that there is no significant difference (p > 0.05) in the PSD-95 cluster density or vGlut1 puncta density in the various conditions. Thus, the decreased density in dendritic protrusions observed after knocking down septin 11 with shRNAs reported above does not affect the density of postsynaptic PSD-95 clusters. There are two possible explanations: (i) that the PSD-95 clusters, formerly present (or destined to be) in dendritic spines, translocated to (or never left) the dendritic shaft after the retraction (or lack of formation) of the corresponding dendritic spines in the neurons transfected with septin shRNAs or (ii) because in nontransfected 14DIV neurons only 13.2 ± 1.6% of the PSD-95 clusters were associated with dendritic protrusions, being the majority of PSD-95 clusters localized in the dendritic shaft, a reduction in dendritic protrusion density induced by the shRNAs could lead to a statistically nonsignificant change in the total density of PSD-95 clusters.

Overexpression of EGFP- or mCherry-Septin 11 Fusion Proteins in Cultured Hippocampal Neurons Has a Dominant Negative Effect on Neuronal Cytoarchitecture and on GABAergic Synaptic Contacts

We have also tested whether overexpression of septin 11 had effects on the neuronal morphology and GABAergic synapses. The overexpression of nontagged septin 11 had no effect on the cytoarchitecture of neurons (see below). However, neurons overexpressing EGFP-septin 11 or Septin 11-EGFP (supplemental Fig. 4, J and K) formed aggregates in dendrites and cell bodies (arrowheads, insets, green). The aggregates were absent in neurons transfected with EGFP. Overexpression with mCherry-septin 11 also led to the formation of aggregates in the cell body and dendrites (arrowheads in supplemental Fig. 4N, red). No aggregates were observed when neurons were transfected with mCherry. These experiments were done with EGFP or mCherry fusion proteins of septin 11 transcript II. Similar results were obtained by using fusion protein constructs of other septin 11 transcripts (not shown). Septin11-EGFP, EGFP-septin 11, and mCherry-septin 11 also formed aggregates in transfected HEK 293 cells (not shown). We used these aggregation artifacts to our advantage, since the EGFP- or mCherry-septin 11 fusion proteins had a dominant negative effect in the transfected neurons by recruiting the endogenous septin 11 to these aggregates, as shown by the considerably reduced number of endogenous septin 11 clusters (EGFP⁻) that were recognized by the anti-septin 11-N antibody compared with that of nontransfected neurons. Moreover, the overexpression of these fusion proteins in neurons also led to decreased dendritic branching (supplemental Fig. 4, J, K and N), as when septin 11 was knocked down by septin 11 shRNAs. Thus, the similar effects observed (see also below) with the two different methods of disrupting septin 11 clusters strengthen the findings obtained with the knockdown experiments.

The Sholl analysis (Fig. 6A) showed that neurons transfected with EGFP-septin 11 or mCherry-septin 11 had reduced dendritic branching compared with that of neurons transfected with EGFP or mCherry, although the reduction in dendritic branching was more extensive in EGFP-septin 11 than in mCherry-septin 11. The branching of neurons transfected with EGFP-septin 11 was 40.9 ± 2.3% that of the neurons transfected with EGFP, and the branching of neurons transfected with mCherry-septin 11 was 64.0 ± 2.1% that of the neurons transfected with mCherry (p < 0.001 in both cases, paired t test).

FIGURE 6.

Overexpression of EGFP and mCherry septin 11 fusion proteins have a dominant negative effect, leading to decreased dendritic arborization, decreased density of dendritic protrusions, and decreased number of GABAergic synaptic contacts. A, Sholl analysis of the hippocampal neurons transfected with EGFP, EGFP-septin 11, nontagged septin 11 and EGFP, mCherry, mCherry-septin 11, or nontagged septin 11 and mCherry. B, quantification of the effect on the density and length of the dendritic protrusions in neurons transfected with the aforementioned plasmid constructs. C, triple-labeled immunofluorescence of hippocampal neurons transfected with EGFP compared with neurons transfected with EGFP-septin 11. The antibodies used were guinea pig anti-γ2 (red) and sheep anti-GAD (blue). The arrowheads indicate the γ2-GABA_AR clusters apposed/colocalized with GAD bouton on the transfected (green) neurons. The arrow points to a large EGFP-septin aggregate on the dendrite. D, quantification of the density of the γ2 GABA_AR clusters of the neurons transfected with the aforementioned plasmid constructs. E, quantification of the percentage of the GABAergic synaptic contacts (GAD⁺ and γ2⁺ GABA_AR) in transfected neurons compared with those of sister nontransfected neighbor neurons (100%) in the same culture. Scale bar, 5 μm in C and 2.5 μm in the selected dendrites of C. *, p < 0.05; **, p < 0.01; ***, p ≤ 0.001.

Fig. 6B shows that neurons transfected with EGFP-septin 11 (12.1 ± 1.0/100 μm²) or mCherry-septin 11 (11.2 ± 1.0/100 μm²) also had significantly fewer dendritic protrusions than neurons transfected with EGFP (16.5 ± 0.9/100 μm², p = 0.001) or with mCherry (17.9 ± 1.3/100 μm², p < 0.001), respectively. Fig. 6B also shows that neurons transfected with mCherry-septin 11 (1.31 ± 0.04 μm) or EGFP-septin 11 (1.14 ± 0.06 μm) had significantly longer dendritic protrusions than neurons transfected with mCherry (0.97 ± 0.03 μm, p < 0.001) or EGFP (0.95 ± 0.05 μm, p = 0.02). An illustrative example (supplemental Fig. 4, O and P) shows the decreased density and increased length of the dendritic protrusions (arrowheads) in neurons transfected with mCherry-septin 11 compared with neurons transfected with mCherry.

Neurons transfected with EGFP-septin 11 had decreased density of γ2-GABA_AR clusters (9.0 ± 0.7 clusters/100 μm², p = 0.007) compared with neurons transfected only with EGFP (12.1 ± 0.9 clusters/100 μm²). Neurons transfected with mCherry-septin 11 showed a nonsignificant effect on the density of γ2-GABA_AR clusters (10.1 ± 0.8 clusters/100 μm², p = 0.074) compared with neurons transfected with mCherry (12.3 ± 0.9 clusters/100 μm²). Neurons transfected with EGFP-septin 11 or mCherry-septin 11 had reduced GABAergic innervation (74.8 ± 5.6% (p = 0.04) and 66.9 ± 7.4% (p = 0.029), respectively) compared with the neurons transfected with EGFP (103.1 ± 9.3%) or mCherry (94.8 ± 7.5%). The percentage values refer to the number of GABAergic contacts of transfected cells compared with that of sister nontransfected neighbors. The EGFP-septin 11 had no effect on PSD-95 cluster density (16.5 ± 1.4 clusters/100 μm², p = 0.84) or vGlut1 puncta density (17.6 ± 1.4 puncta/100 μm², p = 0.41) compared with EGFP (16.1 ± 1.2 clusters/100 μm² for PSD-95 and 16.1 ± 1.1 puncta/100 μm² for vGlut1). Thus, the overexpression of EGFP- or mCherry-tagged septin 11 led to dominant negative effects on dendritic branching, dendritic protrusions, and GABAergic synapses, similar to the effects obtained by knocking down septin 11 with shRNAs described above.

Overexpression of Nontagged Septin 11 Does Not Lead to Change in Dendritic Arborization, Dendritic Protrusions, or GABAergic Innervation of the Transfected Neurons

We also tested whether overexpression of nontagged septin 11 (transcript II) leads to phenotypic change opposite to that of knocking down septin 11 or that of the dominant negative EGFP- or mCherry-septin 11 fusion proteins. Neurons cotransfected with nontagged septin 11 (and EGFP or mCherry), contrary to the EGFP- or mCherry-septin 11 fusion proteins, showed no large aggregates of septin 11 in the neuronal cell bodies or dendrites (supplemental Fig. 4, I and L). The neurons transfected with nontagged septin 11 showed increased general immunofluorescence, as shown with the Rb anti-septin 11-N antibody. However, there was no significant difference in the dendritic branching between the neurons transfected with EGFP or mCherry and neurons cotransfected with nontagged septin 11 (p = 0.60 and p = 0.62, respectively, as shown in Fig. 6A).

Neurons cotransfected with nontagged septin 11 and EGFP (18.3 ± 1.3/100 μm²) or mCherry (17.1 ± 1.3/100 μm²) had no significant effect on dendritic protrusion density when compared with the neurons transfected with EGFP only (p = 0.24) or mCherry (p = 0.65), respectively (supplemental Fig. 4, P and Q, and Fig. 6B). Neurons cotransfected with nontagged septin 11 and EGFP (0.93 ± 0.03 μm) or mCherry (0.97 ± 0.03 μm) had no significant effect on protrusion length compared with the neurons transfected with EGFP (p = 0.84) or mCherry (p = 0.15), respectively (supplemental Fig. 4, P and Q, and Fig. 6B).

We have also studied the effect of overexpression of nontagged septin 11 on γ2-GABA_AR clusters and GABAergic innervation. There was no significant difference in the density of γ2 clusters between neurons cotransfected with nontagged septin 11 and EGFP (12.6 ± 0.8 clusters/100 μm², p = 0.70) and neurons transfected with EGFP (12.1 ± 0.9 clusters/100 μm²) or between neurons cotransfected with nontagged septin 11 and mCherry (11.4 ± 0.9 clusters/100 μm², p = 0.49) and neurons transfected with mCherry (12.3 ± 0.9 clusters/100 μm²). The neurons transfected with nontagged septin 11 with EGFP or with mCherry received similar GABAergic innervation (105.0 ± 12.3% (p = 0.51) and 116.9 ± 11.4% (p = 0.16), respectively) when compared with the neurons transfected only with EGFP (103.1 ± 9.3%) or mCherry (94.8 ± 7.5%). The 100% value corresponds to the number of GABAergic contacts of sister nontransfected neighbors from the same culture.

Thus, overexpression of septin 11 in neurons does not have any significant effect on the dendritic branching, the density and length of dendritic protrusions, the γ2GABA_AR cluster density, or the GABAergic innervation that these neurons receive.

DISCUSSION

Septin 11 came to our attention when LC-MS/MS revealed it as the main component of a protein band that was enriched in a GABAergic type-II PSD fraction over a glutamatergic type-I PSD fraction from rat brain. Immunoblots with two anti-septin 11 antibodies confirmed the enrichment of the (50-kDa) immunoreactive septin 11 protein in type-II PSDs over type-I PSDs. We corroborated the presence of septin 11 in many GABAergic synapses by immunofluorescence of hippocampal cultures and brain slices and by EM immunocytochemistry. To the best of our knowledge, this is the first time that a septin has been shown to be associated with GABAergic synapses. In contrast, there was little colocalization of septin 11 with glutamatergic synapses, as shown by immunofluorescence. Moreover, there was no pre- or postsynaptic immunolabeling of glutamatergic synapses on the dendritic spines of the molecular layer of the cerebellum, as shown by EM immunocytochemistry.

Septin 11 immunofluorescence in cultured hippocampal neurons reveals both diffuse immunofluorescence and the presence of large clusters at dendrites. Both are blocked by the antigenic peptide and by the septin shRNAs. We hypothesize that the diffuse fluorescence corresponds to septin 11 monomers or heterohexamers, whereas the clusters associated with the base of dendritic spines, dendritic branching points, and GABAergic synapses correspond to aggregates of septin 11 organized into filaments, rings, and/or gauzes, which are preferential forms of septin organization (5, 48).

In mammals, septins polymerize with other septins forming heterohexamers containing two subunits of each septin (49, 50). The heterohexamers form filaments that show lateral polarity, which allows them to be organized in sheets of bundled filaments forming gauzes and rings. During yeast cytokinesis, the septin filaments adopt different organization, transitioning from hourglass to ring filaments to two separate rings (51). Some of these structures become diffusion barriers and scaffolds (51, 52). The septin C-terminus has a coiled-coiled structure and is oriented toward one side of the filament interacting with non-septin partners. Thus, the variability in the C terminus of the septin 11 splice variants that we are reporting would allow the interaction of septin 11-containing filaments with an increased number of molecular partners. The other side of the septin filament interacts with the membrane, F-actin, and microtubules. It has been shown that in nonneuronal cells, septin 11 colocalizes with microtubules and actin stress fibers, modulating microtubule dynamics and actin organization (15, 19, 53–55).

The concentration of septin 11 at the neck of the dendritic spines places septin 11 and other septins that heterooligomerize with septin 11 in a privileged position for acting as scaffolds providing structural support for stabilizing dendritic spines and the glutamatergic synapses that these spines receive. This notion is supported by our RNA interference experiments with septin 11 shRNAs, which led to decreased number and longer dendritic protrusions. It is worth noting the discrepancy in the reported effect of knocking down another septin (septin 7) with RNA interference on dendritic protrusions. Thus, Xie et al. (9) reported a decreased number of and longer dendritic protrusions, whereas Tada et al. (8) reported an increased number of and longer dendritic protrusions.

The accumulation of septin 11 at the neck of the dendritic spines could also be involved in the assembly of a diffusion barrier between the dendritic spine and the dendritic shaft, reminiscent of the accumulation of septins at the bud neck in yeast (51, 56, 57). The association of septin 11 clusters with many GABAergic synapses in the dendritic shaft might also act as a diffusion barrier and/or a scaffold in GABAergic synapses. Septin 11 might also regulate the exocytosis of postsynaptic GABAergic proteins, since it has been proposed that some septins guide vesicles to points of exocytosis (58). Thus, septin 5 binds to the Sec6/8 exocyst complex (59). Septin 5 also binds to syntaxin, inhibiting exocytosis, concentrating at the tip of the neurites, where it associates with vesicles regulating vesicle targeting and fusion (59–64).

The accumulation of septin 11 at dendritic branching points is likely to function as a scaffold for cytoskeletal proteins involved in the formation and stabilization of the dendritic arborization. This would be consistent with the high expression of septin 11 in Purkinje cell dendrites, which show a very complex dendritic arborization. This notion is supported by the knocking down of septin 11 with shRNAs, which leads to reduced dendritic branching. Others have shown that knocking down another septin (septin 7) also leads to reduced dendritic branching (8, 9). The levels of septin 11 at the axon and presynaptic terminals in mature neurons are relatively low compared with those of the dendrites, as shown by both immunofluorescence of cultured hippocampal neurons and EM immunocytochemistry. Nevertheless, in developing axons of 3 DIV hippocampal cultures, septin 11 concentrates at the axonal branching points and at the base of the growth cones. Thus, septin 11 might also establish diffusion barriers at the neck of the growth cones and promote the branching of developing axons.

Others have recently shown that septin 7 accumulates at the branching points in dendrites and at the base of dendritic protrusions, but these studies were done in cultured neurons (8, 9). Dendritic protrusions in cultured neurons are not necessarily equivalent to dendritic spines in the brain. We are now showing by EM immunocytochemistry of brain tissue that septin 11 concentrates at both the neck of morphologically identified dendritic spines and at the dendritic branching points. Septin 11 can form heterooligomers with septin 7 and septin 5 (9, 18). Since septin 11, septin 7, and septin 5 concentrate and colocalize at the base of dendritic protrusions in cultured neurons (8, 9) (supplemental Fig. 2, A–D), it is very likely that these three septins form heterooligomers at the neck of dendritic spines.

Not all septin 11 heterooligomerizes with septins 7 and 5. Thus, septin 11 also heterooligomerizes with septins 7 and 9, and the regional and developmental distribution of septin 11 in the brain and hippocampal cultures is not identical to that of septin 7 or septin 5 (7, 8, 15, 19). Although we find that septin 11 and septin 7 highly colocalize at the base of dendritic protrusions, many GABAergic synapses that show associated septin 11 clusters do not show associated septin 7 clusters, although some GABAergic synapses showed colocalization with both septin 11 and septin 7 clusters. We do not know yet if these results reflect variability in the heterooligomerization of septin 11 with septin 7 and other septins in GABAergic synapses.

In an effort to understand the functional roles that septin 11 plays, we transfected cultured hippocampal neurons with septin 11 shRNAs or overexpressed EGFP-septin 11 or mCherry-septin 11 fusion proteins. The shRNAs act by knocking down septin 11 protein expression. The overexpressed EGFP-septin 11 or mCherry-septin 11 operates through a different mechanism, by acting as dominant negatives, forming septin 11 fusion protein aggregates that trap the endogenous septin 11 into these aggregates, thus interfering with the normal delivery and assembly of the endogenous septin 11 to the appropriate cell compartment(s). The septin 11 shRNA knockdown and the EGFP-septin 11 or mCherry-septin 11 overexpression experiments led to similar outcomes: decreased dendritic arborization, decreased density and increased length of dendritic protrusions, and decreased GABAergic synaptic contacts that the transfected neurons received. However, it had little effect on the density of nonsynaptic γ2-GABA_AR clusters, indicating that the nonsynaptic clustering of γ2-GABA_ARs is largely independent of septin 11.

The decrease in GABAergic innervation and the reduction in synaptic GABA_AR clusters observed after knocking down septin 11 suggest that septin 11 could be involved in the stability of GABAergic synapses and play a role in GABAergic connectivity. An alternative explanation is that the decreased GABAergic innervation of the neurons transfected with septin 11 shRNAs is an indirect effect resulting from the reduced dendritic arborization of these neurons (Fig. 5A). The latter would decrease the probability of a GABAergic axon contacting the pyramidal neuron. We favor the first explanation, because the decreased density of synaptic γ2-GABA_AR clusters after knocking down septin 11 with shRNAs was not accompanied by parallel changes in the density of PSD-95 clusters or vGlut1 puncta. Moreover, we and others have shown that in these hippocampal cultures, interfering with the postsynaptic clustering of GABA_ARs or gephyrin in the postsynaptic cells leads to decreased GABAergic innervation of these cells (27, 31, 65). Therefore, if septin 11 clusters are involved in the assembly and/or maintenance of the postsynaptic GABAergic complex, their disruption would lead to decreased GABAergic innervation.

Overexpression of nontagged septin 11, which does not lead to the formation of abnormal septin 11 aggregates, does not have a significant effect on dendritic arborization, the density and length of dendritic protrusions, or the number of GABAergic synapses. These results contrast with the reported effects of overexpressing septin 7, which leads to increased dendritic branching and increased density of dendritic protrusions (8). These results together suggest that septin 7 is limiting in the heterooligomerization with septin 11 and agree with our finding that in cultured hippocampal neurons, there is a higher density of septin 11 over septin 7 clusters.

It has been reported that a septin 4 null mouse has a phenotype, including male infertility and anatomical abnormalities of the cerebellum (13, 66). However, the knock-out septin 3, septin 5, or septin 6 mice showed no apparent effect on brain phenotype (7, 66, 67), indicating that there is developmental compensation and/or functional redundancy of various septins. No knock-out mice for septin 11 or septin 7 have been reported, and we do not know yet whether these mutants would have any of the neuronal phenotypes resulting from the RNA interference-mediated knockdown of septin 11 (this work) or septin 7 (8, 9).