Hippocampal neurons isolated from rats subjected to the valproic acid model mimic in vivo synaptic pattern: evidence of neuronal priming during early development in autism spectrum disorders

Wistar rats (Facultad de Ciencias Exactas y Naturales, UBA) were housed in an air-conditioned room (20 ± 2 °C) and maintained on a 12-h light/dark cycle with food and water ad libitum. Experiments were carried out in accordance with the Guide for the Care and Use of Laboratory Animals provided by the NIH, USA. The experimental protocols were approved by the Ethics Committee for the Care and Use of Laboratory Animals of the School of Pharmacy and Biochemistry at the University of Buenos Aires (Approval No. 180613-1 and 2320). Special care was taken to minimize the number of animals used and their suffering.

VPA injection during pregnancy was performed as previously described [36, 38] with modifications. Embryonic day (E) 0 was considered the day when spermatozoa were found in vaginal smears. On E10.5, dams received a single intraperitoneal injection of saline solution or VPA (450 mg/kg, i.p.) (Fig. 1a). Sodium valproate was dissolved in saline solution (250 mg/ml). Each female was housed individually from E18 to raise their own litter until weaning at postnatal day (PND) 23. Only male pups were used for in vivo and in vitro experiments, those prenatally exposed to VPA were named VPA animals and those exposed to saline, control animals. At PND1, 3 control and 3 VPA animals were used to perform each primary culture of hippocampal neurons (Fig. 1b). Besides, starting from different litters, 4 control and 4 VPA animals were used at PND3 to obtain each mixed glial primary hippocampal culture (Fig. 1b). Pups used for neuronal and glial cultures were born from different dams and all their male siblings were used to validate the behavioral phenotype and perform biochemical analysis at PND3 and 35 (Fig. 1b). Litter size was kept at 8–12 pups (evenly distributed between male and female) per dam by culling at PND3 if necessary [38]. At PND23, males were separately housed in groups of four. The reabsorption rate and model efficacy were according to that previously described [38].

Immunofluorescence images from tissue slices and fixed cultures were captured by an epifluorescence Olympus IX81 microscope (40× magnification objective) equipped with a CCD model DP71 digital camera (Olympus). GAD-67 and phase contrast images were obtained with an Olympus IX83 microscope (40× magnification objective) equipped with a CMOS model C13440-22CU ORCA Flash 4.0 V2 digital camera from Hamamatsu Photonics. Images were analyzed using ImageJ (NIH) software.

In tissue sections, unless otherwise stated, two adjacent non-overlapping images were taken to sample the CA3 hippocampal region from each hemisphere. SYN, NCAM and PSA-NCAM immunolabeling was measured as relative immunoreactive area, quantified as positive immunoreactive area normalized to the total area corresponding to the CA3 in the field of view. Using the ImageJ (NIH) software as previously described by Codagnone et al. [38], images were transformed into gray scale and the positive immunoreactive area was defined as that which exceeded the established gray value threshold selected to differentiate the immunolabelled structures from the background. Once established for each marker, threshold values were kept constant between experimental groups. To analyze Iba1 immunolabeling, three parameters were studied: relative immunoreactive area to total area corresponding to CA3, Iba1 (+) cell number and microglia morphology by classifying Iba1 (+) cells into ramified and unramified categories. Ramified microglia were considered when a cell presented a small soma and more than two long and thin processes, while unramified showed a larger soma and few short and thick processes. Iba1 (+) cell number was estimated based on the optical fractionator method described by West et al. [66] and the area of the CA3 was calculated according to Sosa-Díaz et al. [67]. Every 20th section (665 µm apart) corresponding to plates 29–35 (from bregma – 2.80 to – 4.30) of the atlas of Paxinos and Watson [54] was processed. To measure the CA3 area in each of the three serial sections employed, both hemi-hippocampi were photographed at low magnification (10× objective, NA 0.3) and the CA3 region delimited according to the rat atlas of Paxinos and Watson [54] using the ImageJ (NIH) software. The CA3 areas measured in the serial sections were averaged for total area estimation per animal. Cells were counted manually using a 40× magnification objective (NA 0.6) whose xy movements corresponded to two adjacent non-overlapping fields of view (equivalent to 276 × 270 × 10 µm³ counting frame and a 552 × 540 µm² xy step) along the CA3 subfield. Counting spots in each section corresponded to 50% of the CA3 area. Only Iba1 (+) cells that came into focus within the optical plane were included in the measurement. The volume of the CA3 sampled region was calculated for each animal as mean CA3 area x brain thickness sampled with three serial sections. Finally, Iba1 (+) cell number/CA3 volume ratio was calculated for each animal and averaged per group.

In neuronal culture, for synaptic immunostaining patterns (SYN, PSD-95, NR1, vGLUT1, GAD-67), a single threshold was set for every condition to capture puncta. The cluster of synaptic proteins labeled with antibodies are termed synaptic puncta [68]. For each marker, a single threshold was established to clearly distinguish individual puncta from the background and to minimize the probability of including merged structures in the quantification. The size range selected to define SYN, PSD-95, NR1 and vGLUT1 puncta was 0.15–0.6 µm²; for GABAergic marker was 0.4–1.2 µm² [53, 57, 69, 70]. Synaptic puncta per neuron was quantified as total puncta number, which relates to the number of synapses, individual puncta area as a size parameter, and total area occupied by puncta, which can be influenced by the two former parameters. Actin and PSA-NCAM were measured as total immunoreactive area per neuron. Dendritic tree was determined by subtracting the area of the soma to the total MAP-2 area. Synapses relative to dendritic tree were quantified as SYN puncta number relative to 30 µm length of secondary dendrites labeled with MAP-2 to standardize the procedure across experimental conditions. FM4-64 loading analysis was quantified as number, size, and total area of synaptic puncta. Unloading analysis consisted in measuring the drop in fluorescence intensity of 20 individual puncta from a unique neuron (3 neurons from each group were studied).

Microglia morphology in culture was assessed by measuring mean circularity per field using ImageJ (NIH) software. The circularity, calculated by the software as 4π (area/perimeter²), can range from 1 to 0, being 1 a perfectly circular cell and 0 an elongated or ramified microglia. This parameter is an index of microglia reactivity, a more circular shape is associated to a more reactive cell in vitro [59, 71]. Besides, cell area was calculated as the mean Iba1 (+) cell area per field, by setting a threshold using ImageJ (NIH) software.

Final figures were created with Photoshop CS6. In case of bright and contrast adjustments, they were equally applied to all groups.

InfoStat (Facultad de Ciencias Agropecuarias, Universidad Nacional de Córdoba, Argentina) software was used to perform the statistical analysis. When comparing two normal, independent variables, a Student’s t test was applied; otherwise, the analysis was performed using a non-parametric Mann–Whitney U test. In the case of comparing more than two normal variables, once variance homogeneity was proved, two-way or one-way analysis of variance (ANOVA) followed by Tukey test was used, if not, a non-parametric Kruskal Wallis test was performed. Each analysis and level of significance between groups is described in the figure legend. All (*) represent a parametric test, while (#) represents a non-parametric one. Statistical significance was set at p < 0.05.

Alterations in synapse markers have been previously described in the hippocampus of VPA animals at postnatal ages concomitant with exploratory and social deficits [38, 72, 73]. In fact, the structural synaptic marker synaptophysin (SYN) is reduced in the CA3 hippocampal region of VPA animals in the juvenile period [38]. To address when this synapse change is established, we evaluated whether SYN labeling diminution was recognized soon after birth in the neonatal period or later during the juvenile postnatal period. To this aim, we first addressed early postnatal maturation and, in the juvenile period, autistic-like behavior in each animal. At early postnatal period, VPA rats showed significantly lower body weight at PND14 and scored lower than controls in the swimming test at PND12; both parameters did not reach statistical significance at PND7 and 8, respectively, indicating that growth and maturation deficits in the VPA group became evident later than PND7 (Fig. 2a). Since exploratory and social deficits emerge in VPA rats at the juvenile period, exploratory activity and social interaction were evaluated at PND30-35. Juvenile VPA animals showed an exploratory deficit measured as fewer hole-pokings compared to controls and exhibited social interaction impairment shown as fewer pinnings (Fig. 2a).

Therefore, we evaluated SYN immunolabeling in the hippocampus at PND3, before early behavioral deficits were evidenced and before synaptogenic and synaptic pruning peaks had occurred [51, 74‐76]. The same analysis was performed at PND35 once the VPA phenotype was fully demonstrated. The study was done in the CA3 since synapses in this subregion belong to the intrahippocampal circuit, thus excluding the contribution of non-hippocampal neurons [77]. At PND3, the CA3 of VPA animals showed similar SYN immunostaining to control rats (Fig. 2b). SYN immunolabeling displayed the typical synapse pattern depicted as protein clusters (Fig. 2c). Quantification revealed the absence of statistical differences in SYN relative immunoreactive area at PND3 (Fig. 2d). In contrast, at PND35 the CA3 hippocampal region of VPA animals showed a significant reduction in SYN immunolabeling (Fig. 2b, d). As this SYN reduction suggested a decrease in synapse number, we assessed this parameter by morphological evaluation of asymmetric synapses in the stratum radiatum of the CA3 by electron microscopy. This stratum is mostly comprised by excitatory synapses (asymmetric) from intrahippocampal pyramidal neurons [77]. VPA rats showed fewer asymmetric synapses in this hippocampal region compared to control animals at PND35 (Fig. 2e). Thus, results indicate that the reduction in the number of excitatory synapses in the hippocampus of VPA animal was established postnatally after the neonatal period, most probably after the synaptogenic and synaptic pruning peaks, and concomitantly with the development of the VPA phenotype.

We previously reported that male VPA rats exhibit an increase in NCAM along with a decrease in PSA-NCAM expression in the CA3 hippocampal region at PND35 [38]. Since NCAM participates in vesicle recruitment and synapse stability [17] and PSA-NCAM has a key role in dendritic outgrowth and synapse formation [18, 19], we evaluated whether synaptic changes in the hippocampus of VPA animals were accompanied by alterations in NCAM/PSA-NCAM balance. At PND3, PSA-NCAM and NCAM levels did not differ between control and VPA pups (Fig. 2f–i). Remarkably, at PND35 whereas PSA-NCAM decreased, NCAM increased in the CA3 hippocampal region of VPA rats (Fig. 2f–i). These results indicate that the increase in NCAM/PSA-NCAM balance was concomitant with synapse number reduction and that it was established postnatally after the neonatal period.

We studied the neuronal contribution to the hippocampal synapse alterations observed in vivo by performing an in vitro assessment of neuronal differentiation in the absence of glial cells. Before the active period of synapse formation in vitro (DIV7), cultured hippocampal neurons isolated from VPA pups (PND1) showed similar SYN puncta number, individual and total puncta area when compared with neurons isolated from control animals (Fig. 3a–d). It is accepted that puncta number and individual puncta size indicate synapse number and pre-/post-synapse size, respectively [53, 57, 69, 70]. During the active period of synapse formation in vitro (DIV7-14) [64, 65], hippocampal neurons from both experimental groups increased SYN puncta number throughout differentiation (Fig. 3a, b). However, neurons from VPA animals showed fewer SYN puncta number at DIV10 and 14 when compared with neurons from control animals (Fig. 3a, b). Thus, even though neurons from control and VPA animals showed similar SYN puncta number at DIV7 and synaptogenesis progressed in both groups, neurons from VPA animals were unable to match the control group throughout the synaptogenic period in vitro (Fig. 3a, b). Quantification of individual puncta size did not reveal any robust change in synaptic size throughout differentiation (Fig. 3c). Besides, changes in total SYN puncta area confirmed the results obtained with SYN puncta number from DIV7 to DIV14 in neurons from both control and VPA animals (Fig. 3d). Thus, hippocampal neurons isolated from VPA animals showed in vitro a time-dependent reduction in SYN puncta number in the absence of glia. Also, preserved total SYN expression in hippocampal neurons from VPA animals (Fig. 3e) ruled out decreased synapse number due to a reduction in the level of this structural synaptic protein and suggests that synapse formation and/or stability could be compromised in hippocampal neurons from VPA animals.

Since synapse number is deeply influenced by neuronal arborization [78], we studied neurite differentiation in cultured neurons immunostained for actin and MAP-2. Neurons from VPA animals depicted similar actin arborization throughout differentiation when compared with neurons from control animals (Fig. 3f). However, slight differences were observed when evaluating the progression of differentiation, since neurons from control animals showed a significant actin increase between DIV10 and DIV14 while neurons from VPA animals did so between DIV7 and DIV10 (Fig. 3f). The dendritic tree was evaluated by MAP-2 immunostaining at DIV7 and DIV14 when no changes and a deep decrease in synaptic puncta were observed, respectively. While at DIV7, neurons from VPA animals showed a conserved dendritic tree, at DIV14, a significant reduction was observed compared to the control group (Fig. 3g, h). Although the latter might account for the decrease in SYN puncta number at DIV14, quantification of SYN puncta number relative to dendrite length revealed that the decrease in synapse number was independent of the reduction in the dendritic tree in neurons from VPA animals (Fig. 3i). These results also point out synapse formation and/or stability as the processes involved in the reduction of synapse number in hippocampal neurons from VPA animals. Overall, these results show that, in the absence of glia, hippocampal neurons from VPA animals formed fewer synapses after active synaptogenesis and thus, mimicked in vitro the time-dependent synaptic pattern described in vivo.

Considering that hippocampal neurons isolated from VPA animals mimicked in vitro the time-dependent changes in synapse number previously seen in vivo, we tested whether NCAM/PSA-NCAM balance was affected after active synapse formation. At DIV14, cultured neurons isolated from VPA animals showed an increase in NCAM/PSA-NCAM balance concomitant with the reduction in synapse number, depicted as a decrease in PSA-NCAM immunolabeling (Fig. 3j, k) and an increase in NCAM expression (Fig. 3l). Thus, hippocampal neurons from VPA animals also mimicked in vitro the increase in NCAM/PSA-NCAM balance shown in vivo.

Considering the decrease in SYN puncta number observed in hippocampal neurons isolated from VPA animals when studied in culture at DIV14, we evaluated glutamatergic and GABAergic contribution to such synapse number reduction. GABAergic synapses were found unaltered in neurons from VPA animals immunostained for GAD-67 (Fig. 4a) and quantified as GAD-67 puncta number (Fig. 4e), individual puncta size (Fig. 4f) and total puncta area (Fig. 4g). To study glutamatergic synapses, vGLUT1, PSD-95 and NR1 immunolabeling were assessed. Neurons from VPA animals showed a decrease in number and size of glutamatergic pre-synapses evaluated by vGLUT1 immunolabeling (Fig. 4b, e–g). This result was in line with fewer PSD-95 puncta, confirming a diminution of glutamatergic synapses (Fig. 4c, e–g). Moreover, the constitutive subunit of NMDA receptor NR1 labeling exhibited not only a decreased puncta number but also smaller NMDA receptor clusters (Fig. 4d, e–g). Similar to what was described for SYN reduction, the diminution in NR1 puncta number was independent of the dendritic tree modification (Fig. 4h). These results indicate that hippocampal neurons isolated from VPA animals also mimicked the reduction in the number of excitatory synapses seen in the hippocampus of juvenile VPA animals.

We performed the labeling of pre-synaptic boutons with the FM4-64 dye (Fig. 4i) to determine whether synapses formed in vitro by neurons from VPA animals were functional. Cultured neurons from VPA animals exhibited fewer and smaller functional synapses as seen by reduced number and size of FM4-64 labeled puncta (Fig. 4j, k). However, dye unloading after a depolarizing stimulus remained unaltered compared to neurons obtained from control rats (Fig. 4l, m) indicating that even though there were fewer synapses and smaller vesicular pools in neurons isolated from VPA animals, they exhibited conserved unloading vesicle kinetics.

It has been reported that a brief exposure to a low glutamate concentration induces structural remodeling characterized by dendritic retraction and synapse loss in the absence of neuronal death. PSA-NCAM and NCAM display crucial roles in this glutamate-induced synapse remodeling, which is NMDA receptor-dependent [57]. Thus, considering the increased NCAM/PSA-NCAM balance seen in hippocampal neurons from VPA animals at DIV14, this strategy emerges as an excellent tool to study synapse remodeling as a functional approach. After glutamate exposure, both neurons from control and VPA animals reduced their dendritic tree (Fig. 5a, c) but only neurons from control animals diminished their SYN puncta number and increased their individual puncta size (Fig. 5b, d–f). Furthermore, the remodeling was proved to be NMDA receptor-dependent in both experimental groups, since pre-incubation with MK-801, an NMDA antagonist, but not CNQX, an AMPA antagonist, prevented the alterations induced by glutamate in both groups (Fig. 5g–j). These results indicate that even when neurons from both VPA and control animals responded to glutamate stimulus with dendritic retraction, only neurons from control animals exhibited synaptic changes, thus suggesting that hippocampal neurons isolated from VPA animals exhibit higher resistance to glutamate-induced synapse remodeling. These results suggest that distinctive synaptic features of hippocampal synapses from VPA animals such as NCAM/PSA-NCAM balance could underlie the resistance to structural plasticity.

Our results show that a reduction in synapse number was established in the hippocampus after the neonatal period (PND3). Microglia are involved in synapse formation and synaptic pruning in early life [30]. These glial cells have been proposed to play a role in ASD neuropathology [79‐81]. Some studies from our and other laboratories have evaluated microglia reactivity in the hippocampus of VPA animals, but results are still not conclusive [38, 39]. In order to elucidate whether hippocampal microglia may contribute to synapse number reduction, we studied morphological parameters of microgliosis in the CA3 hippocampal region of VPA animals at PND3 and 35, when changes in synapse number are absent or already proven, respectively. At PND3, qualitative analysis of Iba1 immunostaining showed similar microglia pattern, distribution, and density in the hippocampus of control and VPA animals (Fig. 6a). At PND35, both microglia positive area (Fig. 6b) and microglia cell number (Fig. 6c) did not differ between experimental groups. Ramified and amoeboid microglia profiles are associated with resting and reactive microglia, respectively [82, 83]. Evaluation of ramified and unramified microglia at PND35 showed similar proportion of each population both in control and VPA animals (Fig. 6d). Thus, results indicate the absence of microgliosis in the CA3 hippocampal region of VPA rats before and after the establishment of synapse number reduction.

To further study hippocampal microglia from VPA animals before the establishment of the reduction in synapse number, we isolated microglia from the hippocampus at PND3 and grew them in culture in the absence of neurons. This approach allowed us to evaluate intrinsic properties of microglia by ruling out adaptive changes induced by microenvironmental cues during the postnatal period [84‐86]. In basal conditions, hippocampal microglia isolated from VPA animals showed similar size and complexity (Fig. 6e) and shape profile (Fig. 6f–l) when compared with microglia isolated from control animals. In this case, internal complexity, circularity and cellular area are parameters of microglia reactivity [59, 62, 71, 83] (Fig. 6f). We then evaluated microglia reactivity when exposed to different stimuli. In response to LPS, hippocampal microglia both from control and VPA animals increased circularity (Fig. 6g–i). When exposed to synaptic terminals, hippocampal microglia isolated from both VPA and control animals responded by decreasing their cellular area (Fig. 6k, l). These results indicate that hippocampal microglia isolated from neonatal VPA animals and cultured in the absence of neurons showed preserved basal and stimuli-mediated reactivity. Since microglia reactivity profile is associated to cell functionality [82], our results suggest that hippocampal microglia from VPA rats are not intrinsically compromised.