GABAA receptor subunit deregulation in the hippocampus of human foetuses with Down syndrome
The function, regulation and cellular distribution of GABAA receptor subunits have been extensively documented in the adult rodent brain and are linked to numerous neurological disorders. However, there is a surprising lack of knowledge on the cellular (sub-) distribution of GABAA receptor subunits and of their expressional regulation in developing healthy and diseased foetal human brains. To propose a role for GABAA receptor subunits in neurodevelopmental disorders, we studied the developing hippocampus of normal and Down syndrome foetuses. Among the α1–3 and γ2 subunits probed, we find significantly altered expression profiles of the α1, α3 and γ2 subunits in developing Down syndrome hippocampi, with the α3 subunit being most affected. α3 subunits were selectively down-regulated in all hippocampal subfields and developmental periods tested in Down syndrome foetuses, presenting a developmental mismatch by their adult-like distribution in early foetal development. We hypothesized that increased levels of the amyloid precursor protein (APP), and particularly its neurotoxic β-amyloid (1–42) fragment, could disrupt α3 gene expression, likely by facilitating premature neuronal differentiation. Indeed, we find increased APP content in the hippocampi of the Down foetuses. In a corresponding cellular model, soluble β-amyloid (1–42) administered to cultured SH-SY5Y neuroblastoma cells, augmented by retinoic acid-induced differentiation towards a neuronal phenotype, displayed a reduction in α3 subunit levels. In sum, this study charts a comprehensive regional and subcellular map of key GABAA receptor subunits in identified neuronal populations in the hippocampus of healthy and Down syndrome foetuses and associates increased β-amyloid load with discordant down-regulation of α3 subunits.
KeywordsInterneuron Differentiation Pyramidal cell Cell migration Neurodegeneration
γ-Aminobutyric acid (GABA) is the main inhibitory neurotransmitter in the adult brain, which signals through GABAA and GABAB receptors (GABAA/B-Rs) (Johnston et al. 1978; Wilkin et al. 1981). GABAA-Rs are ligand-gated ion channels organized by pentameric assembly from a panoply of nineteen subunits (Sieghart 1995). Although theoretically, many arrangements are possible to form a pentameric receptor, only a limited number of combinations seem to confer functionality in vivo (Olsen and Sieghart 2008). By virtue of their distinct subunit composition, GABAA-Rs show substantial diversity in their biophysical and pharmacological properties, as well as distribution throughout the brain (Belelli et al. 2009; Eyre et al. 2012; Hortnagl et al. 2013; Pirker et al. 2000; Ramerstorfer et al. 2011; Sieghart 1995; Sieghart et al. 2012; Varagic et al. 2013a, b). Their regional diversity in the adult brain allows GABAA-Rs to drive region- and cell-type-specific inhibition, underlying, e.g., sensory and motor processing, sleep-wakefulness, emotional control, learning, memory and cognition (Fritschy and Panzanelli 2014).
A pivotal role for GABAA-Rs during embryonic brain development has been proposed because of their gradual enrichment from early foetal development (Cobas et al. 1991; Fiszman et al. 1993; Stojanovic et al. 2016) to gate the GABA-mediated control of cell proliferation (Martinez-Cue et al. 2013), migration (Behar et al. 2000; Heck et al. 2007) and differentiation (Cuzon et al. 2006), particularly for cortical interneurons and pyramidal cells. Since impaired GABA signaling in development and adulthood is strongly correlated with pathological states associated with excess excitation, including epilepsy (Pavlov et al. 2011), anxiety and depression (Kalueff and Nutt 2007), a possible pathological contribution of altered GABAA-R-mediated signalling has been posited in neurodevelopmental disorders such as Rett syndrome (Blue et al. 1999; Yamashita et al. 1998), fragile X syndrome (D’Hulst et al. 2006), Prader-Willi/Angelman syndrome (Braat and Kooy 2015) and Down syndrome (Braudeau et al. 2011; Martinez-Cue et al. 2014; Potier et al. 2014).
Down syndrome (trisomy 21) is the most common congenital cause of mental retardation with an incidence of approximately 1 in 750 births (Gardiner et al. 2010; Lott and Dierssen 2010; Parker et al. 2010). Histological studies showed that disrupted proliferation and migration of neurons and glial cells in human foetal brains with Down syndrome is associated with reduced hippocampal and cortical volume, delamination and delayed myelination (Abraham et al. 2012; Contestabile et al. 2007; Golden and Hyman 1994; Guidi et al. 2008; Kanaumi et al. 2013; Larsen et al. 2008). Recent studies in mouse models of Down syndrome showed that excess inhibition through GABAA-Rs could underlie morphological deficits and the ensuing cognitive decline (Braudeau et al. 2011; Fernandez et al. 2007; Martinez-Cue et al. 2013). As such, RO4938581, a GABAA-R α5 subunit-specific negative allosteric modulator, rescued these learning and memory deficits (Martinez-Cue et al. 2013, 2014). However, if the developmentally altered GABAA-R subunit expression profiles could underlie phenotypic deficits of specific neuronal contingents in Down syndrome remain unknown.
Here we sought to investigate the distribution of α1, α2, α3 and γ2 GABAA-R subunits at the network, cellular and subcellular levels in the developing human hippocampal formation from healthy subjects and age-matched cases with Down syndrome. These particular subunits were chosen, since α2 and α3 are predominantly expressed during in-utero development in rodents with their substitution for α1 subunits timed for around birth (Fritschy et al. 1994). Together with their widespread pairing with the γ2 subunit (Hortnagl et al. 2013), we probed possible alterations to the spatiotemporal hippocampal distribution in human Down syndrome foetuses and dissected cell-type-specific enrichment in mouse hippocampi. We find that the spatiotemporal expression of the α3 subunit is especially disrupted in Down syndrome, seen as a premature down-regulation in the CA1–CA3 subfields and dentate gyrus, mimicking an adult-like phenotype at early developmental stages. Since β-amyloid deposits are correlated with the cognitive decline of the ageing brain (Rodrigue et al. 2009), and Down syndrome patients have elevated β-amyloid(1–42) levels in plasma (Obeid et al. 2016) and present early β-amyloid plaque formation (Motte and Williams 1989), we hypothesized that soluble β-amyloid(1–42), produced from an increased pool of amyloid precursor protein (APP) encoded on chromosome 21 (Korenberg et al. 1989) and linked to cellular growth responses (Freude et al. 2011; Wang et al. 2009), is responsible for the adult-like phenotype observed in late-gestational foetuses. Indeed, in our cell culture models, including human SH-SY5Y neuroblastoma cells, we find that soluble β-amyloid(1–42) promotes the progression of differentiating neuroblasts towards a neuron-like phenotype with coincident down-regulation of α3 subunits. This suggests that β-amyloid(1–42) might partake in imposing premature maturation events in the brains of Down syndrome subjects.
Human tissue collection and cataloguing
We collected 28 cases with diagnosed Down syndrome (by karyotyping) and 24 age-matched cases of normal brain development and adulthood at the Brain Bank of the Institute of Neurology, Medical University of Vienna, Austria (Nr. 1316/2012). Foetal brain tissue was obtained from spontaneous or medically induced abortions with their parameters published previously (Milenkovic et al. 2017). Only brains of foetuses [15–34 gestational weeks (GW)], infants (up to 6 months of the early postnatal period) and adults (15–62 years) whose cause of death was unrelated to other genetic disorders, head injury, neurological diseases or other known diseases (e.g. infections) were included. Exclusion criteria were other chromosome aberrations, major CNS malformations, as well as brains with post-mortem autolysis, severe hypoxic/ischemic encephalopathy, intraventricular haemorrhages, severe hydrocephalus, meningitis or ventriculitis (Supplementary Table 1). Biological periods were defined according to paediatric guidelines (Kanaumi et al. 2013): period 1 included early pregnancy period [14–16 gestational weeks (GW)], period 2 spanned the middle pregnancy period (16–28 GW), period 3 corresponded to the late pregnancy period (28 GW onwards, newborns and babies) and adults (15–62 years). Adult cases were used for photographic illustrations, but excluded from our statistical evaluation. Tissue was obtained and used according to the Declaration of Helsinki and compatible institutional guidelines (Kanaumi et al. 2013).
Foetal tissue processing for immunohistochemistry
Foetal tissues were kept at 4° for 24–48 h due to local regulations before immersion fixation in formalin (10%). Tissues were kept in formalin on average for 34–37 days before embedding in paraffin. Formalin-fixed, paraffin-embedded tissue blocks containing the hippocampus and temporal and insular cortices were cut in the coronal plane at 3-µm thickness and mounted on gelatin pre-coated glass slides (Star Frost). Anatomical regions were delineated according to the atlas of human central nervous system development (Bayer and Altman 2005) on hematoxylin–eosin-stained specimens. The immunohistochemical staining methods used here were described previously (Alpar et al. 2014; Kanaumi et al. 2013). In brief, following deparaffinization and rehydration, sections were preincubated either in low pH EnVision™ FLEX antigen retrieval solution at 98 °C for 20 min (BD24 and α3) or with high pH (α2 and γ2), and subsequently manually stained overnight with antibody concentrations listed in Supplementary Table 3. DAKO’s EnVision detection kit was used to visualize the horseradish peroxidase/3,3′-diaminobenzidine (DAB) reaction with H2O2 substrate (0.01%; DAKO). Sections were counterstained with hematoxylin–eosin, dehydrated in ascending concentrations of ethanol, cleared with xylene and covered with Consil-Mount (Thermo Scientific). Sections were inspected on a Nikon Eclipse E400 microscope.
For mouse tissues, sections were blocked with 5% normal donkey serum (NDS; Jackson), 2% bovine serum albumin (BSA; Sigma) and 0.3% Triton X-100 (Sigma) in PBS. Next, sections were incubated with select combinations of primary antibodies (Supplementary Table 3) in 2% NDS, 0.1% BSA and 0.3% Triton X-100 in PBS at 4 °C for 72 h. Secondary antibodies were applied at a concentration of 1:300 at 22–24 °C for 2 h with 2% BSA in PBS. Sections were routinely counterstained with Hoechst 33,342 (Sigma), a nuclear marker, before being coverslipped with Entellan (in toluene; Merck).
Tissue collection from live animals conformed to the 2010/63/EU directive and was approved by the Austrian Ministry of Science and Research (66.009/0145-WF/II/3b/2014). Particular care was taken to minimize the number and suffering of experimental subjects. Adult mice [C57BL/6J and GAD67 gfp/+ (Tamamaki et al. 2003)] were sedated with isoflurane (5%) and subsequently humanely killed by cervical dislocation for fresh foetal tissue collection or transcardially perfused with 4% paraformaldehyde (PFA) in 0.1 M PB (pH 7.4) for immunohistochemistry (n = 2–3 per experiment). The brains were dissected, post-fixed in 4% PFA overnight and subsequently cryoprotected in 30% sucrose for at least 48 h before being cryosectioned (Leica CM 1850 UV) at 50 µm thickness (free-floating sections) in phosphate-buffered saline (0.01 M PBS; pH 7.4) and processed for immunohistochemistry (Alpar et al. 2014). Neonatal mice [C57BL/6J and CCKBAC/DsRed::GAD67 gfp/+ (Calvigioni et al. 2017)] were decapitated with their brains’ immersion fixed in 4% PFA overnight. Brains were cryoprotected in 30% sucrose and freeze–thaw cryosectioned at 20 µm thickness onto electrically charged glass slides (SuperFrost Plus).
Primary neuronal and SH-SY5Y cultures
Foetal hippocampal neurons were obtained (Alpar et al. 2014) by harvesting embryos from C57BL/6NRj pregnant mice on embryonic day (E)18.5. Tissues were dissociated in 0.1% trypsin (Gibco) and DNAse Type 1 (Sigma) in DMEM (Invitrogen) for 5 min at 37 °C. Dissociated cells were washed with 0.4% BSA in DMEM followed by repeated washes in Neurobasal medium containing 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco), 1 mM Glutamax (Gibco) and 1× B27 supplement (Gibco). Primary neurons were grown on poly-d-lysine (PDL; Sigma)-coated glass coverslips at 37 °C for 2, 4 and 7 days in vitro (DIV), and fixed with 4% PFA in PBS on ice for 15 min before being processed for immunohistochemistry.
SH-SY5Y human neuroblastoma cells were maintained in DMEM/F12 (1:1) containing 10% foetal bovine serum (FBS; Gibco), 1 mM sodium pyruvate (Gibco), 1 mM Glutamax and penicillin–streptomycin (as above). For differentiation, SH-SY5Y cells were seeded on PDL-coated glass coverslips in full growth medium for > 3 h. Subsequently, their differentiation was initiated with a growth medium containing 1% FBS and 10 µM all-trans retinoic acid (Sigma) for 7 DIV (Cheung et al. 2009). Differentiated cells were treated once (starting on day 7 and for 48 h) with synthetic β-amyloid (1–42), from 25 nM to 5 µM, custom-synthesized at the Department of Medical Chemistry, Szeged, Hungary (Bozso et al. 2010). Purified peptides were freshly dissolved in distilled water at room temperature, sonicated and stored at − 20 °C until use to prevent peptide aggregation (Bozso et al. 2010) at a stock concentration of 0.5 mM. β-Amyloid(1–42) was then freshly used at concentrations ranging from 25 nM to 5 µM (Lee et al. 2013). Cell density, to mark cell survival, was monitored on an EVOS XL Core microscope (Thermo Fisher), and quantified (20× magnification, 3–5 fields per condition) using Imaris ×64 (Bitplane, 8.3.0). Thereafter, SH-SY5Y cells were either fixed in 4% PFA in PBS and processed for immunocytochemistry or lysed for western blotting or qPCR.
Gene expression analysis
mRNA extraction was performed from fresh frozen tissues and cultured cells using a SPLIT RNA extraction kit (Lexogen). One µg mRNA was converted to cDNA using a High Capacity cDNA Reverse Transcription kit (Thermo Fisher) on a T100 thermal cycler (Bio-Rad) and PCR amplified, as applicable, by mouse or human specific primers (Supplementary Table 2). PCR products were resolved on a 1.5% agarose gel and imaged on a ChemiDoc XRS+ system (Bio-Rad).
Brain tissues and cultured cells were collected in lysis buffer containing (in mM): 25 HEPES (Sigma), 1 EDTA (Sigma), 6 MgCl2 (Sigma), 1 DTT (Sigma) and 1x protease inhibitor cocktail (EDTA Free; Roche) and disrupted by ultrasonication (5 pulses, 50% intensity; Bandelin Electronic). Protein concentrations were measured on a Nanodrop 2000 spectrophotometer (Thermo Fisher) and diluted, if necessary, to 1 µg/µL. Samples were resolved on a 13% SDS-containing gel cartridge (20 µg protein load) in an Amersham WB system (GE Healthcare). Primary antibody concentrations are referred to in Supplementary Table 3. Secondary anti-rabbit and anti-mouse antibodies were used at a concentration of 1:2500. Protein prevalence was analysed with the Amersham WB software package with total protein labelled by carbocyanine-5 in bulk as loading control. Fresh–frozen hippocampal tissue from adult subjects for antibody validation was provided by the Brain Bank of the Institute of Neurology, Medical University of Vienna, Austria, and processed similarly.
Antibody generation, characterization and specificity
The GABAA receptor subunit-specific antibodies used in this study were generated earlier in the laboratory of Dr. W. Sieghart (Hortnagl et al. 2013; Milenkovic et al. 2013; Sieghart 1995; Sieghart and Sperk 2002; Stojanovic et al. 2016). They were characterized extensively for specificity on knock-out brain lysates with western blot (Milenkovic et al. 2013), on knock-out brain sections with immunohistochemistry (Pirker et al. 2003; Zimprich et al. 1991) and displayed similar staining patterns as reported by others using alternative antibody sources (Waldvogel et al. 1999, 2004, 2008). Polyclonal antibodies against GABAA receptor subunits α2, α3 and γ2 were raised in rabbits using maltose-binding protein (MBP) fusion proteins produced in Escherichia coli as previously described (Kasugai et al. 2010; Pirker et al. 2000; Sperk et al. 1997). Antisera were purified by affinity chromatography using the respective GST-subunit fusion proteins (Mossier et al. 1994). All three polyclonal antibodies were raised against sequences identical for rats and humans (C-terminal epitope of α2 and N-terminal epitope of α3) and showed strong homology between rat and human sequences (loop epitope of γ2). The commercially available monoclonal anti-α1 antibody (BD24; Millipore, 1:100) showed a similar cellular hippocampal staining pattern as previously reported (Waldvogel et al. 1999, 2004).
For chromogenic immunostainings, images of entire glass slides were captured with a high-resolution digital slide scanner (NanoZoomer 2.0-HT: C9600-13, Hamamatsu Photonics). Single images were exported from digitalized slides using the compatible viewer software (NDP.view, NanoZoomer Digital Pathology Image) containing hippocampal structures: the dentate gyrus (DG), cornus ammonis 1–4 (CA1–4) subfields and subiculum. Individual layers within subfields (molecular layer (ML), granule cell layer (GL) and polymorph layer (PL) of DG; deep polymorph layer (dPL) of CA4; ventricular zone (VZ), intermediate zone (IZ), stratum oriens (SR), pyramidal cell layer (PyL), stratum radiatum (SR) and stratum lucidum (SL) of CA3; VZ, IZ, SO, PyL and SR for CA2 and CA1; VZ, IZ, PyL and ML for subiculum) were densitometrically analysed in detail. Cases of early development, where the CA1 subfield could not be separated from the subiculum, were used only to describe immunoreactivity distribution but were excluded from our statistical analysis. In addition, a white matter structure (internal capsule (IC)) was imaged to normalize signal density values. For each receptor subunit, three images at 40× magnification were exported from the CA1 and subiculum, while other smaller subregions in CA2, CA3 and the internal and external limbs of the DG were cropped out using a free-hand tool at 10× magnification. Quantitative analysis of immunoreactivity was performed in ImageJ (NIH, 1.50 g). After colour deconvolution to spectrally unmix the hematoxylin pigment (Colour Deconvolution Plugin), images were converted to 8 bit. The threshold defining immunopositivity was set uniformly for all images. Coverage of immunopositive structures was calculated for all exported images and expressed as percentage of the total surface area. Thus, potential technical bias due to different fixation and variable enzymatic DAB reactions were minimized. For immunofluorescence, images were captured with a Zeiss LSM880 laser-scanning microscope and linearly adjusted using the ZEN2010 software (Zeiss, Jena, Germany).
Data were tested for outliers (boxplots), normal distribution (Lilliefors test) and homogeneity of variance (Levene’s test). ANOVA univariate general linear model (GLM) was applied to ranked values of each subunit (α1, α2, α3, and γ2) in all structures to determine differences between Down syndrome and control cases within individual age periods. Subsequently, post hoc analysis with Tukey’s range test (if equal variances and equal sample sizes were assumed) or Games–Howell test (if equal variances and equal sample size could not be assumed) was performed. A p value of < 0.05 was considered statistically significant. Values for α3 were not normalized, as this subunit is predominantly expressed in the white matter during foetal development (see Fig. 4), thus excluding the collection of tissue “background”. Statistical analyses were performed by using IBM SPSS version 21.0.
In this study, we investigated the distribution of developmentally prevalent γ2, α1, α2 and α3 GABAA receptor subunits in the hippocampal CA regions (CA1–4), subiculum and dentate gyrus during development in human foetuses, infants and adults diagnosed with Down syndrome and age-matched controls (Supplementary Table 1). Since the migratory transition of neurons from the CA1 to the subiculum occurs gradually in period 1 (for a description of developmental periods see above), these two regions were designated as CA1/subiculum for that time (Arnold and Trojanowski 1996). To prevent stochastic staining variability (‘edge effects’), we divided the dentate gyrus into inner (DGi; continuation of CA3) and outer limbs (DGo; flanking the CA4).
To verify the cellular identify of the observed γ2 staining patterns, we tested γ2 expression in the developing mouse hippocampus by means of multiple fluorescence immunohistochemistry. In neonates (n = 2), γ2 immunoreactivity was mainly localized to the pyramidal layer in the CA1 and co-localized with Brn-1, a transcription factor transiently expressed in developing pyramidal cells (Alvarez-Bolado et al. 1995) (Fig. 1f, f1). In the adult dentate gyrus (n = 2), γ2 immunoreactivity was mainly seen as neuropil labelling in the molecular layer and perisomatically in the granule cell layer. Intermingled interneurons, genetically co-labelled for GAD67 (Gad1) by in-frame GFP knock-in (Tamamaki et al. 2003), exhibited somatic γ2 immunoreactivity (n = 2; Fig. 1g, g1).
To further examine γ2’s subcellular localization, we cultured primary mouse neurons to identify its spatiotemporal distribution pattern during the developmental period encompassing neurite outgrowth, growth cone motility and synaptogenesis. Up to 4DIV, γ2 immunoreactivity was found indiscriminately on the perikarya, neurites and growth cones of pyramidal-like cells (Supplementary Fig. 1A, A1). In more mature cultures (7DIV) with pre-formed neuronal networks, γ2-positive pearl–lace-like punctae were found on neuronal perikarya and processes, and were opposed by parallel-running processes, likely axons, positive for vesicle associated membrane protein 2 (VAMP2), which participates in the docking and fusion of synaptic vesicles (Washbourne et al. 1995), as well as the vesicular GABA transporter (VGAT) (McIntire et al. 1997) (Supplementary Fig. 1D–D1). Occasional VGAT+ interneurons were found expressing somatodendritic γ2 subunits (Supplementary Fig. 1D2) in a punctate fashion.
The γ2 expression pattern in early developmental periods of Down syndrome cases, periods 1 (n = 4) and 2 (n = 17), was comparable to that observed in age-matched controls (Fig. 1a, b vs. a1, b1). However, the amount of γ2 immunoreactivity was generally reduced in the hippocampus. A significant loss of γ2 immunoreactivity was found in the subiculum (F (1,37) = 6.627, p = 0.014), the CA1 (F (1,38) = 15.008, p = 0.000), and both the internal limb (F (1,30) = 7.181, p = 0.012) and the external limb (F (1,30) = 5.337, p = 0.028) of the dentate gyrus (Fig. 1h, h3; Supplementary Fig. 3E, E1) of Down syndrome cases (n = 6 control vs. n = 5 Down syndrome) around birth (period 3). In turn, γ2 immunoreactivity in the CA2 and CA3 subfields was comparable between diseased and heathy cases (Fig. 1f1, f2).
Immunohistochemistry for the α1 subunit in neonatal mouse hippocampus showed that pyramidal cells positive for α1 in the pyramidal layer were mostly negative for Brn-1 (Fig. 2f, f1). These cells layered above Brn-1 + cells and are either maturing (down-regulated Brn-1) or a subpopulation positive for other transcription factors, such as Brn-2 (Alvarez-Bolado et al. 1995). Interneurons defined by the expression of GAD67 in the CA1 and dentate gyrus of the mouse hippocampus were positive for the α1 subunit, being comparable to humans (Fig. 1g, g1).
The cellular distribution of α1 subunits was determined in mouse primary neuronal cultures. Labelling for α1 was not detected in DIV4 pyramidal-like neurons except minimal immunoreactivity on the proximal segment of their neurites (Supplementary Fig. 1B, B1). α1 labelling was not seen in growth cones either (Supplementary Fig. 1B1). At 7DIV, densely packed post-synaptic α1 immunoreactivity puncta apposed VAMP2 and VGAT immunoreactivities along somas and dendrite-like processes of pyramidal-shaped neurons (Supplementary Fig. 1e, e1). Interneurons (with VGAT+ axons) were occasionally found decorated by α1 puncta along their somatodendritic axis (Supplementary Fig. 1e2). This pattern is similar to that of the γ2 subunit, reminiscent of extrasynaptic receptors involved in tonic GABA signalling (Milenkovic et al. 2013).
Although the layer-specific and region-specific distribution of α1 subunits in the hippocampus of subjects with Down syndrome was comparable to those of normal foetuses, we noted an overall reduction in α1 immunoreactivity (Fig. 2a–d vs. a1–d1). This reduction was mainly due to the loss of neuropil immunoreactivity, while neuronal perikarya retained labelling with an intensity equalling control cases (Fig. 2e1). Interestingly, morphologically identified interneurons seemed to be the least affected by this reduction in period 3 (Fig. 2e1) with statistically significant decreases observed only in the neuropil of the CA1 subfield (F (1,40) = 5.494, p = 0.024) (Fig. 2f). Indeed, the majority of Down syndrome cases in period 3 showed a significant decrease in α1 immunoreactivity in the CA1 (Fig. 2c1) which persisted until adulthood (Fig. 2d1). In other CA and DG regions, α1 immunoreactivity was in general weaker (Supplementary Fig. 3F, F1), but not statistically significant in any of the investigated periods (Fig. 2h–h3).
During successive stages of human development, immunoreactivity on processes became gradually reduced (Fig. 3e1), leaving a strong neuropil labelling, as well as numerous cells (including pyramidal-like and granule-like cells), in the adult CA regions and dentate gyrus (Fig. 3e2, e3). Notably, α2 immunoreactivity seemed to be more pronounced in the inner 1/3 of the molecular layer of the adult DG (Supplementary Fig. 3B, B1). In the adult mouse, α2 immunoreactivity was similarly decreased in the pyramidal layer, leaving a perisomatic staining pattern on pyramidal cells, but was not found expressed by GFP-containing interneurons on a GAD67gfp/+ background (Fig. 3g).
Similar to the patterns found in vivo, α2 subunits were seen on perikarya, processes and growth cones of cultured mouse pyramidal-like neurons at 4DIV (Supplementary Fig. 1C, C1). By 7DIV in maturing cultures, α2 immunoreactivity was mainly restricted to postsynapse-like structures on pyramidal-like neurons apposing VGAT and VAMP2 boutons (Supplementary Fig. 2E–E2) with minimal IMMUNOREACTIVITY remaining on the somatodendritic compartment relative to earlier time points.
The overall distribution of α2 subunits in Down syndrome was similar compared to the control cases (Fig. 3a–d vs. a1–d1). The majority of hippocampal subregions (the dentate gyrus, CA2, CA3, CA4) presented neither significant differences in immunoreactivity nor cellular distribution (Fig. 3h–h3). The initial higher neuropil immunoreactivity in Down cases reached levels comparable to those in control cases in period 2 (Fig. 3b, b1, f). For all other areas (Supplementary Fig. 3G, G1), the immunoreactivity measured was generally weaker in Down syndrome albeit not reaching statistical significance.
Throughout development in Down syndrome, we noted substantially lower levels of both α3 immunoreactivity intensity and its more restricted layer distribution (Fig. 4a–d vs. a1–d1) relative to controls. In period 1, α3 immunoreactivity was significantly lower in the CA1 of Down syndrome cases (F (1,39) = 4.582, p = 0.039) (Fig. 4h). In period 2, levels of immunoreactivity were still significantly lower than in control cases in the subiculum (F (1,39) = 19.118, p < 0.001), CA1 (F (1,39) = 12.121, p = 0.001), CA2 (F (1,34) = 6.478, p = 0.016) and CA3 (F (1,32) = 4.985, p = 0.03) (Fig. 4h–h3; Supplementary Fig. 3H). Furthermore, the adjacent intermediate and marginal zones in the CA regions exhibited a reduction in expression patterns, too. In period 3, we observed a significant and subtotal loss of α3 immunoreactivity, leaving barely any immunoreactivity detectable throughout the hippocampus in Down syndrome cases (Fig. 4c1); the subiculum (F (1,39) = 19.919, p < 0.001), CA1 (F (1,39) = 31.205, p < 0.001), CA2 (F (1,34) = 43.841, p < 0.001), CA3 (F (1,32) = 24. 975, p < 0.001), the internal limb of the dentate gyrus (F (1,34) = 20.741, p < 0.001) and external limb of the dentate gyrus (F (1,34) = 18.879, p < 0.001) (Supplementary Fig. 4D–D1, H–H3). Noteworthy, α3 immunoreactivity on interneuron-like cells seemed more robust than pyramidal cells, which were devoid of α3 immunoreactivity around periods 1 and 2 (Fig. 4b, b1, e2, e3). In contrast, α3 expression in interneurons was reduced at much later developmental stages (Fig. 4e1), while lower α3 levels on interneurons were observed in adult Down syndrome cases (Fig. 4d, d1).
α3 Subunits in neurochemically defined hippocampal neurons
In addition, α3 subunits were found on bipolar cells [negative for the neuronal migration marker doublecortin (DCX) (Gleeson et al. 1998)], surrounded by a meshwork of VGAT+ puncta (Fig. 5c–e1). To validate if these bipolar cells are either GAD67+ and/or cholecystokinin (CCK)-containing interneurons (Klausberger and Somogyi 2008), we utilized a novel CCKBAC/DsRed::GAD67 gfp/+ dual-colour reporter mouse (Calvigioni et al. 2017). Almost all α3+ cell bodies co-localized with GAD67, with a subpopulation being also positive for CCK (Fig. 5f–g1), reinforcing that morphologically differentiating interneurons can express α3. Comparable to its human developmental distribution (Fig. 3d), α3 immunoreactivity was reduced in the adult mouse hippocampus: residual punctate labelling in the pyramidal layer and mainly GFP+ interneurons were noted (Fig. 5h–j1). These data suggest that interneurons retain their α3 expression, whereas pyramidal cells down-regulate this GABAA subunit during hippocampal development.
Cellular exposure to β-amyloid leads to down-regulation of α3 subunits
To seek a more mechanistic view on the down-regulation of α3 subunit expression in Down syndrome, we selected amyloid precursor protein 1 (APP) as a candidate for upstream regulation due to its presence on chromosome 21 and for it being implicated in driving neuronal differentiation and synaptogenesis (Korenberg et al. 1989; Wang et al. 2009). In addition, inhibition of γ-secretase, a multiprotein complex with enzymatic activity to cleave APP and generate full-length β-amyloid (1–42), restores neurogenesis and synaptogenesis in a mouse model of Down syndrome (Giacomini et al. 2015).
Differentiation of SH-SY5Y cells was initiated by application of retinoic acid (Cheung et al. 2009) and verified by their up-regulation of β-III-tubulin (TUJ1), a cytoskeletal marker of immature neurons (Fig. 7c–d1; n = 3 coverslips). Coincidently, α3 subunits appeared in processes of SH-SY5Y cells undergoing retinoic acid-induced differentiation (Fig. 7d, d1). After application of fresh β-amyloid(1–42) at concentrations (100–250 nM) that did not affect the survival of cells undergoing retinoic acid-induced differentiation in our culture paradigm (Supplementary Fig. 4B–C2), TUJ1 levels significantly increased relative to untreated yet differentiated controls (Fig. 7e; n = 3 each). This response was anti-parallel with α3 subunit levels (that decreased; Fig. 7e). When applying μM concentrations of β-amyloid(1–42), we noted cytotoxicity (Harkany et al. 2000a, b) (Supplementary Fig. 4B–C2), indicating that there is a narrow developmental window in which β-amyloid(1–42) might exert morphogenic, perhaps even differentiation-promoting effects on developing neurons, if a tandem of TUJ1 and α3 are seen as read-outs. In sum, our data suggests that β-amyloid, a pathogenic outcome of increased APP levels, in Down syndrome could contribute to the remodelling of GABA synapses by altering their subunit composition (Fig. 7f).
In this study, we present the first comprehensive description of the expression pattern of α1, α2, α3 and γ2 GABAA receptor subunits in the human hippocampus of Down syndrome cases and age-matched controls along a broad developmental trajectory. Although the existence of functional GABAA receptors in primate and human brains was demonstrated by visualization of benzodiazepine binding sites (Reichelt et al. 1991; Shaw et al. 1991), there is a surprising gap of knowledge at the protein level due to the lack of appropriate tools. Therefore, we applied novel knock-out-tested antibodies against GABAAR subunits to address this issue.
We find differential maturation of α1 and α2 subunits in the hippocampus. Similar to data from animal models, α1 seems to have late expressional onset in the human brain (Fritschy et al. 1994). In contrast, an abundance of α3 subunits throughout the developmental hippocampus was observed with significantly diminished levels in the adult. GABAARs have been implicated in the migration of post-mitotic neurons, including both pyramidal cells (Behar et al. 2000) and interneurons (Cuzon et al. 2006), postulating a central role for this subunit to control migration and differentiation. Our finding that α3 subunits are robustly expressed during development on both interneurons and pyramidal cells in vivo, as well as on actin-rich motile growth domains in vitro, reinforces a non-synaptic trophic and/or guidance role for GABA signalling through α3 subunits in the developing hippocampus. In addition, the α3-subunit seems to be exchanged perinatally to other subunits (i.e. α1 and α2) in the human foetal brain, which might account for the change in electrophysiological properties that GABAAR signalling undergoes during normal brain development (Owens et al. 1999). While, the γ2 subunit showed widespread expression in the hippocampus, we observed a striking similarity in expression pattern between α3 and γ2 subunits in the CA1 and subiculum, suggesting their probable functional co-occurrence. The temporal co-expression of these two subunits suggests that these possible complexes are destined for synaptic terminals, since γ2 is classified as the post-synaptic subunit and is involved in the trafficking of the complex towards the synapse (Vithlani et al. 2011). Noteworthy, we did not detect significant amounts of α3 in the DG, suggesting different assembly partners for γ2 in this region.
In the hippocampus of foetal Down syndrome subjects, our data indicates complex expression profiles coincidently affecting α1, α3 and γ2 subunits. For instance, α1 and γ2 subunits showed selective depletion in the CA1 and subiculum, whereas in other hippocampal areas these changes were not as obvious. Interestingly, the distribution of α2 subunits in the hippocampus bore a striking resemblance to the γ2 subunit, both of them exhibiting a subtotal depletion in the CA1 and subiculum in period 3. In contrast, the α3 subunit demonstrated the most robust changes in Down syndrome cases. A strong reduction was observed throughout all investigated time periods with an overall lower expression and an almost complete loss from period 3 onwards. Most adult animal studies focused on pharmacological manipulation of α5 subunit containing GABAARs (Braudeau et al. 2011; Fernandez et al. 2007; Martinez-Cue et al. 2013), due to their involvement in cognition (Redrobe et al. 2012; Rudolph and Mohler 2014; Wang et al. 2012). We were unable to reliably probe the α5 subunit with the antibodies available to us in developmental tissues, revealing only a weak punctate staining that we deemed unspecific due to a lack of regional and temporal patterns (data not shown). Therefore, we only focussed on GABAAR subunits whose consistent staining pattern we deemed specific.
Given the fact that GABAARs are thought to provide a stop signal for migrating cells in the cortex (Behar et al. 2000), a disturbance in GABAAR subunits would suggest alterations to neuronal migration in Down syndrome. Indeed, a change in neurogenesis and reduced neuronal number in the cortex have been described in Down syndrome before (Becker et al. 1991; Golden and Hyman 1994; Ross et al. 1984; Wisniewski et al. 1984). Since we localized α3 subunits to motile growth domains on both pyramidal cells and interneurons, our data suggest that a premature down-regulation of α3 subunits might contribute to those observed changes in neuronal migration, especially if this reduction leads to altered GABAAR-mediated currents or second messenger signalling systems (Luscher et al. 2011). Thus, a premature redistribution of subunits could eventually lead to improper neuronal placement and differentiation, and disturb prenatal giant depolarizing potentials which are important for the maturation of synapses (Ben-Ari et al. 2007) overall, resulting in cognitive deficits.
A concurrent comorbidity in Down syndrome is the early development of β-amyloid plaques in the adult brain and Alzheimer’s-like neurological abnormalities (Masters et al. 1985; Motte and Williams 1989). Due to the position of APP on chromosome 21, APP is thought to elevate levels of full-length β-amyloid, leading to early plaque deposits. In addition, inhibitors of γ-secretase, restored neurogenesis and synaptogenesis in a mouse model of Down syndrome (Giacomini et al. 2015). However, β/γ secretase cleavage of APP can generate bioactive peptides other than β-amyloid(1–42) itself, including small C-terminal fragments that interact with signalling proteins (van der Kant and Goldstein 2015) and its intracellular domain (AICD50) retained after γ secretase cleavage which can act as a transcription factor (Konietzko 2012). Recent data demonstrate that AICD50 overexpression affects the levels of Sox2, a key transcription factor in the regulation of stem cell maintenance and lineage commitment (Sarlak et al. 2016). Since the physiological role of these cleavage products are still poorly understood, we focussed on β-amyloid as it has been correlated with cellular differentiation (Freude et al. 2011; Wang et al. 2009) and the peculiarity of accelerated ageing in Down syndrome (Obeid et al. 2016). We thus posit an increase of APP, and elevated levels of soluble β-amyloid (1–42) as a factor in promoting early brain maturation.
In our developmental Down syndrome cohort, we were not able to find β-amyloid plaques, suggesting that any pathological effect of β-amyloid is not due to its aggregation in foetal brains. Instead, we observed the first signs of plaque formation in tissues of a 15-year-old patient from a separate cohort, with more substantial deposits in patients > 30 years old. As β-amyloid oligomers are more toxic than un-aggregated peptide (Zhao et al. 2012), an increase in the soluble un-aggregated form could promote cellular morphological changes without compromising cell survival. In our foetal Down syndrome cases, we find elevated quantities of APP, and experimental challenges with external β-amyloid (1–42) delivery triggered the down-regulation of α3 subunits similarly to what is observed in human foetal brain. This observation is in line with our previous results, showing that developmental physiologic tau phosphorylation is disturbed in individuals with Down syndrome (Milenkovic et al. 2017), most probably by shifting the phosphorylation pattern towards an adult pattern. In sum, we attribute both APP and β-amyloid fragments as key features in the premature and erroneous differentiation of brain circuitry by deregulating α3 subunits, leading to miswired networks that suppress cognitive performance.
Open access funding provided by Medical University of Vienna. The authors thank Dr. W. Sieghart for providing antibodies and Dr. M. Ernst for valuable discussions. This study was supported by the EU FP7 Project DEVELAGE (#278486; G.G.K.), the Hungarian Brain Research Program (# KTIA_13_NAP-A-III/7; L.F., Z.B.), the Swedish Medical Research Council (T.H.), Swedish Brain Foundation (‘Hjärnfonden’, T.H.) and the European Research Council (ERC-2015-AdG-695136, T.H.) and intramural funds of the Medical University of Vienna (E.K., T.H.).
Compliance with ethical standards
Conflict of interest
The authors declare no conflict of interest.
- Bayer SA, Altman J (2005) Atlas of human central nervous system development. CRC Press, Boca RatonGoogle Scholar
- Calvigioni D et al (2017) Functional differentiation of cholecystokinin-containing interneurons destined for the cerebral cortex. Cereb Cortex 47:2453–2468Google Scholar
- Milenkovic I et al (2017) The physiological phosphorylation of tau is critically changed in fetal brains of individuals with Down syndrome. Neuropathol Appl Neurobiol. https://doi.org/10.1111/nan.12406
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.