Abstract
A growing number of molecules have been identified as regulators of inhibitory synapse development, but whether dysregulated expression of these molecules contribute to brain disorders is poorly understood. Here we show that Down syndrome cell adhesion molecule (Dscam) regulates the inhibition of neocortical pyramidal neurons (PyNs) in a level-dependent fashion. Loss of Dscam impairs inhibitory neuron development and function. In the Ts65Dn mouse model for Down syndrome, where Dscam is overexpressed, GABAergic innervation of cortical PyNs by chandelier and basket cells is increased. Genetic normalization of Dscam expression rescues the excessive GABAergic innervation and the increased inhibition of PyNs. These findings demonstrate excessive GABAergic innervation and inhibition in the neocortex of Down syndrome mouse model and identify Dscam overexpression as the cause. They also implicate dysregulated Dscam levels as a potential pathogenic driver in related neurological disorders.
Introduction
GABAergic neurons mediate synaptic inhibition in the mammalian brain (1, 2), and altered GABAergic inhibition has emerged as a nexus of circuit dysfunction in a number of neurodevelopmental disorders (3, 4). A growing number of molecules have been identified in GABAergic synapse formation and synaptic transmission (4–9). Intriguingly, copy number variations of these genes are commonly found in different neurological diseases (7), suggesting that dysregulated expression of these genes might be a pathogenic driver. However, whether or not this is true remains largely unknown.
Down syndrome cell adhesion molecule (Dscam) is an evolutionarily conserved type I transmembrane protein (10). In human, Dscam gene is localized in the Down syndrome critical region on chromosome 21 (hCh21) (10). We previously showed in Drosophila that Dscam protein levels determine the sizes of presynaptic terminals without requiring the ectodomain diversity of the Drosophila Dscam gene (11). Moreover, others reported that overexpression of Dscam impairs synaptic targeting and transmission in Drosophila (12, 13). These findings suggest that dysregulated Dscam levels might contribute to neuronal defects in brain disorders. In fact, altered Dscam expression levels have been reported in multiple brain disorders, including Down syndrome (DS) (14), autism spectrum disorders (ASD) (15–17), intractable epilepsy (18), bipolar disorder (19), and possibly Fragile X syndrome (11, 12, 20, 21). Although recent findings suggest a conserved role of Dscam in promoting presynaptic growth in vertebrates (22, 23), whether dysregulated Dscam expression results in neuronal defects in brain disorders remains unknown.
In this work, we sought to determine the effects of altered expression levels of Dscam in mouse models of DS, in which the Dscam gene is triplicated. Previous studies have shown that enhanced GABAergic inhibition impairs cognition in Ts65Dn mice (24–28), the most widely used DS animal model (29, 30). A series of studies have demonstrated excessive GABAergic inhibition in the hippocampus of Ts65Dn mice (27, 31–39). However, whether GABAergic signaling is altered in other brain regions and, if it is, what genes on hCh21 cause it, are poorly understood. Here, by combining recently developed genetic tools for sparse labeling and whole-cell patch-clamp, we found excessive GABAergic innervation and inhibition of PyNs by chandelier (ChC) and basket cells in the neocortex of Ts65Dn mice. Genetic normalization of Dscam expression levels rescued the presynaptic overgrowth and excessive synaptic transmission of GABAergic neurons. Consistently, loss of Dscam impaired ChC presynaptic terminals and reduced GABAergic inhibition of neocortical PyNs. In addition, we found that ChC axon terminal growth and synaptogenesis are coupled and positively correlated with neocortical Dscam levels in both wide-type and Ts65Dn mice. These findings uncover novel molecular, cellular and synaptic mechanisms for DS brain disorders and highlight the critical role of Dscam levels in regulating neocortical inhibition. Dysregulated Dscam expression levels may be a common contributor to GABAergic dysfunctions in associated neurological diseases.
Materials and Methods
Mice breeding and tamoxifen administration
Age-matched littermates were used for all experiments. Ts65Dn (Ts(17(16))65Dn) was purchased from Jackson Laboratory (Stock No: 005252) and maintained by crossing with C57BL/C3H F1 hybrid males. For labeling single ChCs in Ts65Dn mice, Ts65Dn females were crossed with Nkx2.1-CreER+/−, Ai14−/−, Dscam+/− male mice of mixed C57BL/C3H background to obtain euploid, Ts65Dn and Ts65Dn: Dscam+/+/− littermates with Nkx2.1-CreER+/−, Ai14+/− transgenes. Tamoxifen (Sigma, T5648-1G), dissolved in corn oil, was delivered to P0 pups by intraperitoneal injection at the dosage of 80 mg/kg body weight. Pseudodam was prepared in advance to lactate the pups as the pups were often discarded by Ts65Dn dam after the drug delivery. For labeling single ChCs in Dscam2j mice, female Dscam+/− mice of C3H background were crossed with Nkx2.1-CreER+/−, Ai14−/−, Dscam+/− male mice of mixed C57BL/C3H background to obtain Dscam+/− and Dscam- pups with Nkx2.1-CreER+/−, Ai14+/− transgenes. Tamoxifen was delivered to P0 pups as above. Strong pups were sacrificed at P0 to keep the litter size 4-5 to increase the surviving rate of Dscam- pups. Nkx2.1-CreER+/−, Ai14−/−, Dscam+/− male mice were removed from the cage before female laboring as the male adults is likely to attack and sacrifice Dscam-pups.
For electrophysiology recordings, Ts65Dn females were crossed with C3H Dscam+/− male mice to obtain euploid, Ts65Dn and Ts65Dn: Dscam+/+/− littermates. C3H Dscam+/− mice were interbred to obtain Dscam+/− and Dscam- pups for recordings.
PCR genotyping was performed on purified tail tips at least 2 times to confirm the genotype according to the protocol by Jackson Laboratory. All work involving mice was approved by the University of Michigan Institutional Animal Care and Use Committee (IACUC) and in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals (NIH).
Tissue processing
P28 mice were euthanized by CO2 and immediately followed by intracardial infusion of 4% paraformaldehyde (PFA) with 4% sucrose in 1x PBS. The brains were then removed and post-fixed in 4 % PFA overnight at 4 °C, followed by incubation in 30% sucrose (wt/vol) for at least 1 day. The brains were embedded in the OCT compound (Fisher HealthCare), frozen in −20 °C overnight and then sectioned to 100 µm-thick slices with a Leica CM3050S cryostat. Sectioned brain slices were kept in 1x PBS containing 0.05% sodium azide at 4 °C until immunostaining.
Immunohistochemistry
For ChC and AIS staining, coronal brain sections (100 μm) were blocked with 8% BSA in PBST (1x PBS + 0.1% Triton x-100) containing 0.05 % sodium azide at 37°C for 1 hour, and then incubated with the following primary antibodies in the blocking solution at 37°C overnight: anti-mCherry (AB0081; 1:300; SICGEN) for neuronal morphology and anti-phospho-Iϰβ-α (14D4 rabbit monoclonal antibody; 1:500; Cell Signaling) for labeling the AIS. After washing 3 time (1 hour each time) at 37°C in PBST, brain slices were incubated with the following secondary antibodies in the blocking solution at 37°C overnight: donkey anti-goat-rhodamine RX (RRX) (705-297-003; 1:300; Jackson ImmunoResearch), donkey anti-rabbit-AlexaFluor 488 (711-545-152; 1:300; Jackson ImmunoResearch). After washing 3x for a total of 1 hour at 37°C in PBST, the slices were mounted in sRIMS (40).
The procedure for staining of GABAergic neurons and PyNs was the same as above, except that antibody incubations were done at room temperature and different primary antibodies were used. Primary antibodies used were anti-parvalbumin (PVG213; 1:1,000, Swant Inc), anti-CAMKIIα (6G9; 1:5,000, LSBio). Secondary antibodies used were donkey anti-goat-RRX (705-297-003; 1:500; Jackson ImmunoResearch) and donkey anti-mouse-AlexaFluor 488 (715-545-150; 1:500; Jackson ImmunoResearch).
For staining perisomatic GABAergic synapses, antigen retrieval was conducted in 10 μM sodium citrate for 20 min at 95°C (41). After a brief rinse in PBS, the slices were blocked in blocking buffer (1x PBS, 0.3% Triton x-100, 3% normal donkey serum, 0.05 % sodium azide) for 1 hour at room temperature (RT), and then incubated with the following primary antibodies in the blocking solution at 4°C overnight: mouse anti-Bassoon (SAP7F407; 1:1,000; ENZO), G.P. anti-VGAT (131 004; 1:1,000; Synaptic System), rabbit anti-GRASP1 (1:2,000) (42). After washing 3 time (20 mins each time) at RT in PBST, brain slices were incubated with the following secondary antibodies in the blocking solution at RT 3 hours: donkey anti-mouse-AlexaFluor 488 (715-545-150; 1:300; Jackson ImmunoResearch), donkey anti-G.P.-RRX (706-295-148; 1:300; Jackson ImmunoResearch), donkey anti-rabbit-AlexaFluor 647 (711-605-152; 1:300; Jackson ImmunoResearch),. The slices were mounted for imaging after washing 3x for a total of 1 hour at RT in PBST.
The procedure for c-fos staining was as described previously (43). Briefly, brain sections were washed three times (10-min each) with 1x PBS, and then treated sequentially with the following solutions: 2% H2O2, 0.3% NaOH/1x PBS for 20 min, 1x PBS for 5 min twice, 0.3% glycine/1x PBS for 10 min, 1x PBS for 5 min twice, 0.03% SDS/1x PBS for 10 min, and 1x PBS for 5 min twice. The brain sections were then blocked in 3% normal donkey serum/0.1% Triton X-100/1x PBS for 1 hr, and subsequently incubated overnight in blocking solution that contained anti-c-Fos antibody (#2250; 1:2000 cell signaling). The sections were washed in 1x PBS for three 20-min washes, and incubated with biotinylated secondary antibody (711-066-152; 1:200; Jackson Immunoresearch) in blocking solution for 2 hours. After washing in 1x PBS for three 10-min washes, the sections were processed for ABC amplification (PK-6100; 1:500 in PBS dilution; Vector Laboratories) for 1 hour, washed in 1x PBS for 10 min three times, incubated in 0.05% diaminobenzidine (DAB)/0.015% H2O2/1x PBS for 3 min, and then washed in 1x PBS for 10 min three times.
Image acquisition and quantitative analysis
All images were acquired from layer II/III of the ACC in the mouse neocortex by using a Leica SP5 confocal microscope equipped with a resonant scanner, except the samples for perisomatic GABAergic synapses (see below). For ChCs imaging, different fluorescence channels were imaged sequentially with the pinhole set at airy 1. A 63x objective lens with a numerical aperture of 1.4 was used. Confocal image stacks were collected with 100 continuous optical sections at 0.3-μm z-steps. The cell body was positioned around the middle of the 30-μm in depth. Before quantification, the image stacks were maximally projected along the z-axis. A region of 120 μm (length) × 80 μm (width) with the cell body in the top middle was set for quantification. Cartridges and boutons that colocalized with AIS were quantified by the NeuroLucida software (MBF Bioscience). Cartridge/bouton number was defined as the number of cartridges/boutons within this region. Cartridge length was defined as the distance from the first bouton that colocalized with AIS to the last one colocalized with AIS in that cartridge. Bouton size was defined as the length of the bouton in parallel to the AIS. We quantified the sizes of boutons in the 10 cartridges nearest to the cell body. Interbouton distance was defined as the distance between two neighboring boutons.
Samples for perisomatic GABAergic synapses were imaged on a Leica SP8 confocal microscope with a 63x objective lens (NA 1.4). For each PyN, a single confocal image was taken at the z-position where PyN cell body occupied the most area. Images were deconvoluted with the Huygens software (Scientific Volume I maging). Perisomatic VGAT+ and Bassoon+ puncta were quantified i n d efined Py N so ma regions, as shown in Figure 6B, by the NeuroLucida software (MBF Bioscience).
For c-fos and parvabumin imaging, a 20x objective lens with a NA of 1.4 was used. The z-steps were 1 μm. Five optical sections were maximally projected before quantification. Three fields were imaged for each mouse. Cells were counted in a defined area (252.4 μm × 200 μm) for each field manually with the assistance of the NeuroLucida software. To eliminate experimenter’s bias, these experiments were carried out in double-blind fashion. Two double-blind methods were used. First, the images acquired by the primary experimenter (H.L.) was coded and randomized by the second lab member (R.L. or M.V.). After the primary experimenter quantified the data, the data were decoded for statistical analysis. ChCs from 2 littermates were quantified in this method. Second, the mouse brains from the primary experimenter (H.L.) were coded and randomized by the second lab member (R.L. or M.V.). The second lab member sent the encoded mouse brains to the third lab member (T.Y.) for sectioning. After the primary experimenter immunostained and quantified the encoded brain sections, the data were decoded for a final statistical analysis. ChCs from 3 littermates and the c-fos, parvabumin and basket cell synapse quantifications were performed in this method.
Western blotting
Mouse neocortices were removed immediately after PFA perfusion. Similar volumes of tissues were homogenized in SDS sample buffer and loaded for electrophoresis. The proteins were transferred to PVDF membranes and blocked with 5% milk (wt/vol) for 1 h at room temperature. The blots were incubated overnight at 4 °C with goat anti Dscam (AF3666, 1:500, R&D systems) and mouse anti-tubulin (12G10, 1:5,000, DSHB). After washing, blots were incubated for 2 h with HRP-conjugated secondary antibodies and developed by chemiluminescence (Catalog# 32106, Pierce). Quantification was performed using ImageJ.
Electrophysiology recordings and analysis
Electrophysiological recordings of spontaneous and miniature inhibitory postsynaptic currents were performed as described previously (44). Brains were obtained from euploid, Ts65Dn, and Ts65Dn/Dscam+/+/− mice at around P28. The animals were decapitated under isoflurane and USP anesthesia, and the brain was quickly removed from the skull and placed in 4°C slicing solution containing (in mM) 62.5 NaCl, 2.5 KCl, 1.25 KH2PO4, 26 NaHCO3, 5 MgCl2, 0.5CaCl2, 20 glucose and 100 sucrose (pH maintained at 7.4 by saturation with 95% O2 + 5%CO2). Coronal brain slices (300-350 μm thick) containing layers II/III of the ACC neocortex were cut with a microtome (VF-300, Compresstome). The slices were then transferred to a holding chamber and maintained at room temperature in artificial cerebrospinal fluid (ACSF) containing (in mM) 125 NaCl, 2.9 KCl, 1.25 KH2PO4, 26 NaHCO3, 1 MgCl2, 2 CaCl2 and 20 glucose, pH 7.4 (with 95% O2 and 5% CO2 bubbling through the solution) for at least 1 hour prior to recording. After equilibration, individual slices were transferred to the recording chamber continuously perfused with ACSF (1-2 mL/min). Recording micropipettes were pulled from borosilicate glass capillaries (1.5 mm O.D. Harvard Apparatus) for a final resistance of 3-6 MΩ and filled with a solution containing (in mM) 135 CsCl, 4 NaCl, mM GTP, Mg-ATP, 0.5 CaCl2, 5 EGTA and 10 HEPES. The signals were recorded with an Axoclamp 700B amplifier (Axon Instruments, Union City, CA). Voltage clamp recordings were obtained from neurons in layers II/III of the ACC region; the cells in these brain regions were identified using a Nikon Eclipse FN-1 microscope with a 40x water-immersion objective and a DAGE MTI IR-1000 video camera. Neurons were visualized using IR-DIC to evaluate their orientation and morphology. Whole-cell patch-clamp recordings for cells with a high cell resistance (greater than 8 GΩ before break-in). sIPSC and mIPSC recordings in voltage-clamp configuration were acquired at 2 kHz fixing the voltage at −70 mV. The IPSCs were recorded in the presence of the NMDA receptor antagonist DL-2-amino-5-phosphonopentanoic acid (AP-5) at 50 mM and the AMPA/kainite antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) at 10 μM. For measurement of mIPSCs, 1 μM tetrodotoxin (TTX) was added to the perfusion solution to block synaptic responses dependent on the AP. Access resistance was monitored throughout the experiment and experiments were canceled if changes greater than 20% occurred. Peak events were identified automatically using Minianalysis (Synaptosoft Inc.) and visually monitored to exclude erroneous noise. The frequency, amplitude and distribution of events were analyzed. One euploid neuron (1 out of 20) with mIPSC frequency 8.9 Hz was defined as an outlier by the Grubbs’ test and removed from quantification. Mean values were compared using the Student’s t-test. Data are presented as mean ± SEM.
Statistical analysis
Data are presented as mean ± SEM. Comparisons of mean differences between groups were performed by one way ANOVA followed by Student’s t-test. P < 0.05 was considered to be statistically significant. For all quantification, *: p < 0.05; **: p < 0.01; ***: p < 0.001; ns: not significant (p > 0.05).
Results
Loss of Dscam impairs ChC presynaptic development in mice
To determine whether Dscam regulates GABAergic neuron development in the neocortex, we focused the analysis on ChCs for two reasons. First, ChCs are thought to be the most potent inhibitory neurons in the neocortex (45). Each ChC innervates roughly two hundred PyNs at their axon initial segments (AIS) (46), where action potentials are generated (47, 48). Second, the morphology of ChCs is relative stereotypical and reliably quantifiable (49–52). We specifically label single ChCs in the neocortex b y crossing Nkx2.1-CreER mouse line with the tdTomato reporter line Ai14 (53)(Figure 1A). A single ChC extends only a few dendritic branches but several hundreds of presynaptic terminals, called axonal cartridges, each of which innervates one or two PyNs (54). The axonal cartridges and presynaptic boutons of single ChCs were quantified i n a threedimensional volume defined based on the position of ChC cell body (Figure S1). AIS-colocalized cartridges and presynaptic boutons were quantified, as off-target varicosities lack presynaptic markers and symmetric synaptic densities (55). The AIS was immunolabeled by an anti-phospho-IϰB antibody (pIϰB) (50, 51, 53, 56), which strongly correlates with the AIS labeled by ankyrin-G antibodies in the mouse neocortex (57)(Figure S2). Quantification w as p erformed in a double-blinded fashion to avoid experimenters’ bias. The ChCs in the cortical layer II and III of the anterior cingulate cortex (ACC) were analyzed as the morphology of ChCs in this region is most stereotypical. Mice were analyzed on postnatal day 28 (P28), a time after ChC development is complete (54). Similar to its role in presynaptic development in Drosophila, loss of Dscam significantly impeded the development of ChC presynaptic terminals (Figures 1B-E). Compared to heterozygous littermates, the number and length of individual cartridges were significantly reduced in Dscam-- mice by 8% and 12%, respectively so that the total cartridge length for each ChC was reduced by 23% (Figures 1C-E).
Moreover, the numbers and sizes of presynaptic boutons were significantly reduced by 20% and 16%, respectively (Figures 1F-H). Interestingly, no difference was observed in the interbouton distance between these two groups (Figure S3A), indicating that Dscam does not regulate bouton density. These results demonstrate that Dscam is required for ChC presynaptic development in mice.
Loss of Dscam impaired GABAergic inhibition of PyNs To determine whether defective GABAergic synaptogenesis caused by loss of Dscam impairs GABAergic synaptic transmission, whole-cell patch-clamp was applied to record PyNs in layers II/III of the ACC from acute neocortical brain slices. Consistent with impaired axonal growth and synaptogenesis in ChCs, we found that the average frequency of miniature inhibitory postsynaptic currents (mIPSCs) was 35% less in Dscam- mice than that in heterozygous littermates (Figures 2A-B). In addition, the average amplitude of mIPSCs was 42% less, ¬suggesting that postsynaptic responses were impaired by loss of Dscam (Figures 2A and C). Similar changes were observed in the frequency and amplitude of spontaneous IPSC (sIPSCs) (Figures 2D-F). Consistent with the reduced GABAergic transmission, we observed hyperactivity in neocortical PyNs, as indicated by c-fos expression levels (58). There was 5-fold more c-fos-positive PyNs in Dscam- than in Dscam+/− mice (Figures 2G-H). Taken together, these data suggest that Dscam is required for the normal level of inhibition of PyNs in the neocortex.
ChCs exhibit excessive presynaptic terminals and boutons in Ts65Dn mice Next, we determined whether the presynaptic development of ChCs is altered in the Ts65Dn mouse model of DS, where Dscam is present in 3 copies (29) and overexpressed (Figure S4). Compared to euploid littermates, the number and length of individual axonal cartridges were significantly increased in Ts65Dn mice by 28% and 11%, respectively (Figures 3A-C).
The total cartridge length, which is the sum of individual cartridges in the quantified volume, was increased by 37% (Figure 3D). In addition, ChC showed a significant increase in both the number and size of synaptic boutons in Ts65Dn mice. The bouton number for each ChC is increased by 36% (Figures 3E-F). The average size of presynaptic boutons was enlarged by 21% (Figures 3E and G). The increased ChC presynaptic bouton number and size are consistent with previous reports of increased number of the puncta positive of vesicular GABA transporter proteins (VGAT) and enlarged inhibitory synapses in the neocortex of Ts65Dn mice (59, 60). The average interbouton distance between neighboring boutons showed a significant, though subtle, decrease in the trisomy mice (Figure S3B), suggesting a slight increase in bouton density. Altogether, by labeling and quantifying single interneurons in the ACC, we found excessive presynaptic terminals and boutons in Ts65Dn mice, which is opposite of what we observed in Dscam- mice.
Normalization of Dscam levels rescues ChC presynaptic overgrowth in Ts65Dn mice Dscam is overexpressed in both the brains of DS patients and those of Ts65Dn mouse model of DS (14)(Figure S4). Previous studies in Drosophila have demonstrated that increased levels of Dscam lead to excessive growth of presynaptic terminals (11, 61). To determine whether the overexpressed Dscam in Ts65Dn mice causes presynaptic overgrowth in ChCs, we normalized Dscam gene dosage in Ts65Dn mice by crossing female Ts65Dn mice with male Dscam+/− mice to obtain the Ts65Dn:Dscam+/+/− genotype (Figure 4A). This genetic scheme also yielded regular Ts65Dn littermates (Ts65Dn:Dscam+/+/+). In Ts65Dn mice with 2 copies of functional Dscam genes, the average levels of Dscam proteins were statistically indistinguishable from the euploid mice (Figure 4B). Normalizing Dscam levels rescued ChC presynaptic overgrowth in Ts65Dn mice. Compared with Ts65Dn littermates, the ChC cartridge number, individual cartridge length and total cartridge length were reversed to levels indistinguishable from euploid in Ts65Dn:Dscam+/+/− mice (Figures 4C-F). In addition, the increased bouton number and bouton size were mostly rescued by normalizing Dscam expression (Figures 4G-I), while no change was observed in the interbouton distance (Figure S3B). These results demonstrate that ChC presynaptic terminal overgrowth is mainly caused by Dscam overexpression in Ts65Dn mice.
ChC axon terminal growth and synaptogenesis are positively coupled Presynaptic terminal growth and synaptogenesis are concurrent processes for forming proper neuronal connections (62). Little is known about whether and how these two events are orchestrated in mammalian GABAergic interneurons. The sparse labeling of ChCs offers an opportunity to address these questions at the single-cell resolution. We examined the relationship between cartridge length and three morphological aspects of synaptic boutons, namely bouton number, size, and density (as reflected by the interbouton distance), in single ChCs. We found a strong correlation between the cartridge length and the bouton number of each ChC in both wild-type (euploid) (R2 = 0.82, p < 10-7) and Ts65Dn mice (R2 = 0.79, p < 10-15) (Figure 5A). Although Dscam positively regulated both cartridge length and bouton number (Fig. 1e, g and Fig. 4f, h), loss of Dscam did not impair their coupling (R2 = 0.89, p < 10-8) (Figure 5B). There was also a significant correlation between cartridge length and bouton size in both wild-type and Ts65Dn mice (Figure 5C), though it was weaker than that between cartridge length and bouton number. Loss of Dscam seemed to mildly impair the coupling between cartridge length and bouton size (Figure 5D). There was no correlation between cartridge length and bouton density in any of the genotypes tested (Figures 5E and F). Taken together, these results suggest that presynaptic terminal growth and synaptogenesis are positively coupled in ChC development.
ChC presynaptic cartridge length and synaptogenesis are proportional to neocortical Dscam levels In Drosophila, Dscam levels determine the size of presynaptic arbors (11). Western blotting results suggested that Dscam levels in the neocortex exhibited individual variations in euploid, Ts65Dn, and Ts65Dn:Dscam+/+/− mice (Figure 4B and S S5A). We plotted the neocortical Dscam level of each mouse, as assayed by western blotting, against the average cartridge length, bouton number, size or density in each mouse. We observed a strong correlation between Dscam levels and cartridge lengths, bouton number and bouton sizes in mice (Figures S5B-D). In contrast, no correlation was found between Dscam levels and interbouton distance (Figure S5E), again supporting that bouton density is not regulated by Dscam (Figures S3A-B). The dosage-dependence highlights the importance of Dscam expression levels in regulating ChC presynaptic development.
Dscam overexpression increases the number of GABAergic synapses on PyN somas in Ts65Dn mice Dscam is expressed in different subtypes of GABAergic neurons (63). Does its overexpression in Ts65Dn mice cause developmental defects in other types of GABAergic neurons? Basket cells are the most common cortical interneurons that preferentially innervate the somas of PyNs (1, 2). Because basket cell morphology is highly heterogeneous (64), we analyzed the GABAergic synapses formed by basket cells on PyN soma (i.e., perisomatic synapses), rather than quantifying the morphology of single basket cells as we did on ChCs. To visualize perisomatic GABAergic synapses, brain sections were triply labeled with anti-Bassoon (for presynaptic active zone) (65), anti-VGAT (for presynaptic GABAergic boutons) (66), and anti-GRASP1 (for PyN soma and proximal dendrites) (42). Most Bassoon+ puncta around a PyN soma were surrounded by or overlapped with VGAT signals (Figure S6A). We found that the average number of GABAergic synapses around each PyN soma—as indicated by Bassoon+ puncta that were either apposed to or overlapped with VGAT+ signals—was 20% more in Ts65dn neocortices than euploid controls (Figures 6A-B). Strikingly, normalizing Dscam expression level completely rescued the enhanced synaptogenesis (Figures 6B-D). These data suggest that the overexpressed Dscam in Ts65Dn mice also increases the number of GABAergic synapses on PyN somas.
Normalizing Dscam levels rescues the excessive GABAergic synaptic transmission in the Ts65Dn neocortex Previous studies in Ts65Dn neocortices have demonstrated enhanced GABAergic neurogenesis (31), increased numbers of GABAergic synapses (60), and enlargement of these synapses (59). However, whether GABAergic synaptic transmission in the neocortex is increased in these mice remains to be determined. Our finding that the overexpressed Dscam in Ts65Dn mice increased GABAergic synaptogenesis in the neocortex prompted us to test this possibility. We examined GABAergic synaptic transmission in acute neocortical brain slices, using the whole-cell patch-clamp technique to record from PyNs in layers II/III of the ACC. We found that mIPSC frequency was increased by 63% in Ts65Dn mice compared to euploid littermates (Figures 6E-F). These results suggest an increase in presynaptic neurotransmitter release and align well with our findings of enhanced synaptogenesis in ChCs and basket cells (Figures 3E-G and 6A-D). By contrast, we found no difference in mIPSC amplitudes between euploid littermates and Ts65Dn mice (Figures 6E, G), ¬suggesting that postsynaptic responses may not be affected by the trisomy. Consistent with the role of Dscam in the excessive synaptogenesis in Ts65Dn mice (Figures 4G-I), normalizing Dscam expression prevented the increase in mIPSC frequency in these mice (Figure 6E-F), suggesting that the overexpressed Dscam causes excessive GABAergic synaptic transmission in the Ts65Dn neocortex. Similar changes in the frequency, but not amplitude, were observed in spontaneous inhibitory postsynaptic currents (sIPSCs) among euploid, Ts65Dn, and Ts65Dn:Dscam+/+/− mice (Figure 6H-J). Notably, normalizing Dscam levels did not rescue the increased parvalbumin-expressing (PV+) neuron density in the ACC region of Ts65Dn mice (Figures S7A-B). Consistently, the number of PV+ neurons was unaffected by loss of Dscam (Figures S7C-D). Thus, the overexpressed Dscam in the Ts65Dn neocortex causes excessive GABAergic synaptic transmission by increasing the number of GABAergic synapses without affecting the number of PV+ GABAergic neurons.
Discussion
In this study, we provide evidence for the important role of Dscam expression levels in the pathogenesis of brain disorders in DS. We show that Dscam overexpression leads to prominent presynaptic overgrowth in ChCs and basket cells. The hyperinnervation and excessive GABAergic inhibition of PyNs in Ts65Dn were rescued by normalizing the Dscam levels. The converse phenotypes were observed in Dscam loss-of-function mutant mice. The sensitivity of GABAergic synapse development and function to Dscam expression levels suggest that dysregulated Dscam expression may underlie GABAergic dysfunction in neurological disorders that exhibit abnormal Dscam expression, including DS, ASDs, intractable epilepsy, bipolar disorders and, possibly, Fragile X syndrome.
Dscam expression levels determine presynaptic terminal sizes in mammalian neurons
The stereotypical morphology of ChCs axon arbors and presynaptic terminals are advantageous for quantitative assessment (50, 51). Moreover, recent advances in genetic labeling of ChCs have allowed sparse labeling of single ChCs for quantifying presynaptic terminals at single-cell resolution (53, 67). By taking advantage of this system, we show in this study that Dscam overexpression causes presynaptic overgrowth of ChCs in the neocortex. Notably, ChC cartridge length and synaptogenesis are proportional to neocortical Dscam expression levels, suggesting the sensitivity of ChC development to Dscam levels. Thus, this work supports a conserved role of Dscam in regulating presynaptic terminal growth in Drosophila and mice. An important question is whether Dscam functions cell-autonomously to promote presynaptic growth in GABAergic neurons. Previous studies in Drosophila sensory neurons and gain-of-function studies in mouse retinal ganglion cells suggest a cell-autonomous role of Dscam in promoting axonal growth (11, 22). Genetic deletion of Dscam in single ChCs is challenging because the Cre activity in Nkx2.1-CreER mouse line is weak. Since administration of tamoxifen cannot guarantee the deletion of target genes in all or vast majority of cells in floxed mice, immunostaining or in-situ hybridization is required to confirm the deletion. However, available anti-Dscam antibodies are not conducive for immunostaining in neocortices (14)(data not shown). Moreover, the experimental procedures for in situ hybridization are not compatible with morphological studies of ChCs by immunostaining. Future studies with stronger Cre in ChCs will determine whether Dscam functions cell-autonomously.
The coupling between presynaptic terminal growth and synaptogenesis in ChCs
Among the different types of GABAergic neurons, ChCs are thought to be the most powerful inhibitory neurons in the neocortex (45). These neurons form unique axo-axonic GABAergic synapses that selectively innervate PyNs at their AIS, where action potentials are generated (47, 48); each ChC innervates roughly two hundred PyNs (46). Impaired ChC presynaptic growth is present in individuals with epilepsy and schizophrenia, both of which are thought to be caused, in part, by disrupted GABAergic signaling (45, 68–71). Recent studies demonstrate that deleting Erbb4, a schizophrenia-associated gene, in ChCs causes a schizophrenia-like phenotype in mice, indicating a causal relationship between ChC defects and schizophrenia (52, 72). In addition to ErbB4, Neuregulin 1, DOCK7, Fgf13 and L1CAM have also been found to regulate the synapse formation between ChCs and PyNs (49–51, 73). In the present study, detailed investigation of the cartridge growth and synaptogenesis at single-cell resolution uncovered several interesting features of ChC development. First, presynaptic cartridge growth is strongly associated with bouton number and, to a lesser extent, bouton size (Figure 5A-D). This observation supports the synaptotropic model proposing that synaptogenesis stabilizes axonal arbor growth in neurodevelopment (74). Second, the factors that regulate cartridge growth and synaptogenesis are not necessarily the factors coupling these two processes. Although Dscam regulates both cartridge growth and bouton number, the coupling of these two processes remains intact in mice that are deficient of Dscam function, suggesting that Dscam is not the coupling factor. Identifying and distinguishing coupling factors from regulators is an important step toward a mechanistic understanding of ChC development.
Dscam overexpression leads to excessive neocortical inhibition in DS mouse model
Overproliferation of GABAergic neurons caused by triplication of Olig1 and Olig2 contributes to excessive GABAergic signaling in the hippocampus (31, 32, 35). However, several lines of evidence suggest that heterogeneous etiology may exist in DS brain disorders such that different brain regions exhibit distinct molecular, cellular, and physiological defects. For example, mIPSC frequency is increased in the dentate gyrus, but not the CA1 region, of the hippocampus in adult Ts65Dn mice (31, 37). In contrast to the large body of literature on GABAergic signaling in the hippocampus of DS animal models, very little is known about whether GABAergic signaling is altered in the neocortex. Previous studies showed that the sizes of synaptic boutons and inhibitory synapses are enlarged in the neocortex of Ts65Dn mice, suggesting possible alterations in GABAergic synaptic functions in the neocortex (59, 75, 76). In the present study, we demonstrate that GABAergic inhibition of neocortical PyNs is excessive in Ts65Dn mice and that Dscam overexpression plays a key role in this process. Our results support the notion that alterations in GABAergic signaling in DS brains are region-specific. Normalizing Dscam expression reverses both increased mIPSC and sIPSC rates in neocortical PyNs (Figures 6E-J), while normalizing Olig1 and Olig2 expression specifically corrects s IPSC frequency in hippocampal CA1 (31). In addition, the effects of Dscam unlikely originate from mitotic proliferation for several reasons. Firstly, Dscam is expressed in differentiating neurons but not mitotic progenitors (10, 77–79). Secondly, our results show that normalizing Dscam levels does not rescue the increased density of parvalbumin-expressing (PV+) neurons in the ACC region of Ts65Dn mice (Figures S7A-B). Consistently, the number of cortical PV+ neurons was not affected in Dscam knock out mice (Figures S7C-D). Taken together, our findings suggest that the overexpressed Dscam in Ts65Dn mice causes presynaptic overgrowth and enhances neocortical GABAergic inhibition. They also suggest that systemic screening of hCh21 genes in Drosophila might be an effective approach to identify candidate genes that drive the pathogenesis of DS diseases.
Insights into other brain disorders
Altered Dscam expression levels has been associated with many brain disorders, including DS, ASD, intractable epilepsy and bipolar disorder (14–19), and possibly Fragile X syndrome (11, 12, 20, 21). The present work has established a causal relationship between dysregulated Dscam levels and the developmental and functional defects of GABAergic neurons. The dose-dependent function of Dscam suggests that dysregulated Dscam levels may be a common pathogenic driver of GABAergic dysfunctions in related neurological diseases. For example, genetic analyses have revealed multiple disruptive single-nucleotide variants (SNVs) and copy number variants (CNVs) in Dscam gene in idiopathic ASD individuals, which raises the possibility of altered Dscam expression levels (15–17, 80). Given the regulation of GABAergic signaling by Dscam levels discovered in the present study and the established role of impaired GABAergic signaling in ASD (81), one may hypothesize that reduced Dscam expression causes impaired GABAergic signaling in ASD. In support of this, some ASD individuals with CNV of deleted enhancer region in Dscam show nonfebrile seizures (17), a symptom also found in mice with deficient Dscam function (82). It is thus important to examine Dscam levels in post-mortem samples of autistic individuals and determine whether altered Dscam expression causes GABAergic dysfunctions in ASD mouse models.
AUTHOR CONTRIBUTIONS
H.L. and B.Y. conceived the project and designed the experiments. H.L. performed mouse breeding, drug delivery, western blotting, immunostaining, and data analysis. R.C.F, J.M.H, and G.P. performed electrophysiology recordings and analysis. T.Y., R.L. and M.V. assisted on mouse breeding and immunostaining. J.Z.H. and P.F. provided Nkx2.1-CreER and Dscam2j mice, respectively. B.Y. supervised the project. H.L., R.C.F., L.L.I., K.Y.K, P.M.J., and B.Y. wrote the paper.
COMPETING FINANCIAL INTERESTS
The authors declare no competing interests.
ACKNOWLEDGEMENTS
We thank Roman Giger, Jun Wu, Andrew Nelson, Dawen Cai, Jonathan Flak, and Martin Myers for sharing reagents or technical support, Miao He, Yongjie Hou, Pedro Lowenstein, Ken Inoki, Yukiko Yamashita, and Dawen Cai for helpful discussions. This work was supported by grants from NIH (R21 NS094091 and R01 MH112669 to BY, and R37NS076752 to LLI) and Protein Folding Disease Initiative of the University of Michigan to B.Y., the Heinz C. Prechter Bipolar Research program and Richard Tam Foundation, a University of Michigan Rackham Merit Fellowship and MI-BRAIN Predoctoral Fellowship to JMH, University of Michigan Depression Center (PMJ) and the One Mind Bipolar Disorder Translational Research Award to P.M.J.
Footnotes
↵9 Lead contact
Bibliography
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.
- 6.
- 7.↵
- 8.
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.
- 26.
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.
- 34.
- 35.↵
- 36.
- 37.↵
- 38.
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.
- 70.
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.
- 79.↵
- 80.↵
- 81.↵
- 82.↵