Background
Synaptogenesis is a developmental process in which neurons form specialized sites of contact that mediate intercellular communication via release of pre-synaptic neurotransmitters that bind to and modulate the activity of postsynaptic neurotransmitter receptors. Synapses can be either excitatory (synaptic neurotransmission results in an excitatory postsynaptic potential) or inhibitory (synaptic neurotransmission triggers an inhibitory postsynaptic potential). The balance between excitatory and inhibitory neurotransmission is critical for proper development and function of the central nervous system [
1]. Abnormalities in inhibitory synaptic function have long been implicated in the pathogenesis of epilepsy and other seizure disorders [
2], and recent hypotheses of the pathogenesis of at least some neurodevelopmental disorders (i.e., autism spectrum disorders) implicate imbalance of inhibitory and excitatory neurotransmission as a causal factor [
3]. Thus, developing high-throughput approaches for quantifying excitatory
versus inhibitory synaptogenesis is becoming increasingly important for mechanistic, toxicologic and drug screening studies of neurodevelopmental disorders.
Excitatory and inhibitory synapses are distinguished by the type of neurotransmitter that is released from the presynaptic terminal and by the profile of pre- and postsynaptic proteins within the synapse. For example, in the mature central nervous system (CNS), glutamatergic synapses are excitatory and are characterized by the release of glutamate from the presynaptic terminal, the presence of vesicular glutamate transporter 1 (vGLUT1) in the presynaptic vesicle pool and the presence of postsynaptic density 95 (PSD95) in the postsynaptic density [
4,
5]. In contrast, mature GABAergic synapses are inhibitory and are characterized by the release of γ-aminobutyric acid (GABA) from the presynaptic terminal, the presence of vesicular GABA transporter (vGAT) in the presynaptic vesicle pool and the presence of gephyrin in the postsynaptic density [
6]. Excitatory and inhibitory synapses have distinctly different roles in controlling nervous system function and may be differentially susceptible to events that modulate nervous system development, such as chemical exposure or pharmacologic intervention. Therefore, independent measurements of these two types of synapses are important for understanding how perturbations in neurodevelopmental processes affect the formation of a mature synaptic network.
A number of methods have been used to measure synaptogenesis
in vitro, including quantification of synaptic protein levels using antibody-based methods (e.g., ELISA, western blotting and immunocytochemistry) and functional assessment using electrophysiology [
7-
9]. However, the relationship between these measurements of synaptogenesis at different biological levels is unclear. For example, does an increase in expression of excitatory presynaptic proteins necessarily correlate with an increase in excitatory synaptic function? In addition, the majority of published studies of synaptogenesis have relied on low-throughput approaches that assess a small number of single cells (e.g., imaging of immunostained cells or electrode recordings) or single cultures (e.g., ELISA or western blotting). Recently, high-content imaging (HCI) has been used to rapidly quantify the development of synapses
in vitro at the cellular level based on immunocytochemical localization of synaptic proteins and to detect chemical-induced neurotoxicity [
10,
11]. Significant advantages of HCI include not only the increase in throughput relative to more conventional approaches, but also that it provides automated standardized acquisition of a very large number of images, which increases statistical power and removes the selection bias inherent with the more conventional single cell methods of assessing synaptogenesis. At the functional level, microelectrode arrays (MEAs) have been developed to rapidly assess the development of neuronal activity and network formation
in vitro [
12].
In the present study, we examined the ontogeny of synaptogenesis in two widely used in vitro models of neurodevelopment: primary cultures of rat cortical and hippocampal neurons. Synapse formation was measured over 28 DIV at differing levels of biological complexity: 1) at the molecular level using ELISA to quantify the level of synaptophysin protein; 2) at the cellular level using HCI to quantify the immunoreactivity of excitatory and inhibitory synaptic biomarkers; and 3) at the functional level using MEA recordings. Our data demonstrate quantitative and qualitative similarities and differences in measures of synaptogenesis across methods and cell types.
Discussion
Primary cultures of rodent neurons from various brain regions, including the cortex and hippocampus, have been powerful tools for elucidating the cellular and molecular mechanisms that control synapse formation and stabilization. However, the reported rates of synaptogenesis
in vitro vary considerably between studies depending upon the biomarkers, experimental models and experimental methods used to identify and quantify synapses [
15]. In the present study, we compared the ontogeny of synapse development in primary rat cortical and hippocampal cell cultures using three complementary methods: 1) quantification of the levels of synaptophysin protein by ELISA; 2) quantification of excitatory and inhibitory synapse number using HCI; and 3) quantification of synaptic network activity using MEAs. Each method demonstrated a general increase in synapses over time in both culture models. However, assessment at the cellular level using HCI demonstrated distinct differences between neuronal cell types with respect to the temporal profile of synapse development and the subcellular distribution of excitatory and inhibitory synapses. The functional significance of these differences was confirmed by the assessment of network activity using MEAs.
One of the challenges encountered in this study was the need to use different cell plating densities between cell types and experimental platforms. The basal rate of neuronal cell loss was greater in cortical cultures than hippocampal cultures, necessitating the use of higher cortical cell densities relative to hippocampal cell densities for the synaptophysin ELISA and high content imaging studies. Loss of cells over time has been observed previously in neuronal cultures [
16], although to our knowledge, differences in basal neuronal attrition rates in culture have not been systematically examined. In addition, more densely plated cultures were required to produce reliable measurements of synapse connectivity on the MEA platform as compared to cultures used for either ELISA or HCI experiments. These observations highlight the need for cell type and platform specific optimization of seeding densities to ensure to collection of high quality data.
The need to use different plating densities between cell types and across platforms complicates comparisons of the temporal profile of synaptogenesis. However, it must be noted that within each cell type, the same plating density was used for cultures prepared for ELISA and HCI experiments, thereby allowing direct comparisons of data obtained for each cell type across these two platforms. While it may be difficult to compare the timing of synapse formation between cell types, the general spatiotemporal patterns of excitatory and inhibitory synapse development are likely unaffected by differences in plating density. Moreover, hippocampal and cortical cultures were plated at equivalent densities for the MEA experiments, enabling comparisons between cell types with respect to ontogenetic profiles of synaptic activity. Even in light of these potential confounders, parallel use of complimentary methods for quantifying synaptogenesis will likely yield a more comprehensive understanding of responses to experimental manipulations or stressors (i.e., chemical exposure).
Synaptophysin is an integral membrane protein of synaptic vesicles whose expression at both the transcript and protein level increases as neurons mature [
14,
17]. Thus, synaptophysin has been used as a biomarker of presynaptic terminals
in vivo, and as a biomarker of synaptogenesis in cultured hippocampal neurons [
18,
19]. Using both synaptophysin immunocytochemistry and ELISA as a general marker for synaptogenesis, we observed that the amount of synaptophysin increased in both hippocampal and cortical cultures with increasing time in culture. The fold-increase in synaptogenesis we observed using this approach is similar to that reported in other studies that used ELISA or Western blotting to quantify synaptophysin levels in neurons cultured under similar conditions [
14,
20]. Similarly, we had previously reported increased expression levels of synapsin, another synaptic vesicle protein, in cultured cortical neurons with increasing time in culture [
10]. Qualitative comparisons of synaptophysin immunoreactivity between culture types suggested that cortical cultures express significantly more synaptophysin than hippocampal neurons at any given time in culture. This is consistent with a prior report that synaptophysin levels are higher in cortical
versus hippocampal neurons cultured from embryonic mice [
21]. However, analysis of synaptophysin expression level by ELISA (in which equal amounts of protein were loaded per sample) revealed no cell type-specific differences in synaptophysin expression level at any given time point. These data demonstrate that the amount of synaptophysin protein per amount of total protein did not differ between the culture models and suggests that the amount of synaptophysin protein produced on a per cell basis is not different between hippocampal and cortical neurons in culture.
Our findings are consistent with the literature that the expression of synaptophysin correlates well with neuronal maturation; however, because not all synaptic vesicles are localized to synapses, the immunocytochemical localization of synaptophysin does not necessarily indicate a true synapse [
22]. Furthermore, quantification of synaptophysin expression level by either immunocytochemistry or ELISA provides no information regarding the type of synapses formed or their function [
23], as we confirmed in hippocampal and cortical cultures immunostained for synaptophysin, vGLUT1 and vGAT. To address this issue, we applied high-throughput, HCI technology to quantify vGLUT1 and vGAT immunopositive puncta. To increase the likelihood that the vGLUT1 and vGAT immunopositive puncta included in our analysis represent the pre-synaptic half of a bipartite synapse, we quantified puncta that were immunopositive for these presynaptic proteins that co-localized with MAP2, a biomarker of postsynaptic structures, specifically dendrites and neuronal cell bodies.
The development of the dendritic arbor is intimately tied to synapse formation. Synaptic connections increase in parallel to dendritic development, and abnormalities in dendritic length are associated with changes in synapse number and function [
24-
29]. In our studies, dendrite length increased with increasing time in culture. On a per cell basis, the amount of dendritic growth was greater in hippocampal than in cortical neurons at each time point. This may reflect the fact that cortical cells were plated at a higher initial seeding density than the hippocampal cells, thus reducing the distance required for cortical dendrites to grow before contacting a neighboring neuron. Alternatively, inherent differences in the rate of
in vitro dendritic growth between the two cell types may contribute to this effect.
The numbers of vGLUT1 and vGAT immunopositive puncta also increased in hippocampal and cortical cell cultures with increasing time in culture, which is consistent with previous experiments performed in our lab [
10] and with reports in the literature [
30,
31]. However, the ontogenetic profile of these parameters differed depending on neuronal cell type in that significant increases in the numbers of vGLUT1 and vGAT immunopositive puncta increased much earlier in hippocampal neurons relative to cortical neurons. This is consistent with
in vivo observations that rates of synaptogenesis differ between and within brain regions even at similar stages of brain development [
32,
33]. Another possible explanation for the differences observed in our culture models is that cortical cell cultures have a broader distribution of neuronal types; whereas, in hippocampal cell cultures, pyramidal neurons are the predominant neuronal cell type [
34-
36]. This could likewise explain the more prominent numbers of vGLUT1 puncta observed in the hippocampal neurons as vGLUT1 has been shown to localize preferentially in the stratum pyramidale [
37]. However, the effect of cell density on the time course of synaptogenesis between the two culture types cannot be disregarded. While cell density does not appear to be a contributor in the determination of the composition of neuron cell types
in vitro, lower cell densities result in faster development of synapses and higher synapse-to-neuron ratios [
38]. This could explain the slower onset of synaptogenesis in the cortical cultures despite their higher cell density. Furthermore, the influence of our culture reagents and conditions must be taken into account, as they can have a robust and differential effect on neurite outgrowth in different neuronal cell populations [
39].
In addition to cell type-specific differences in the ontogeny of synaptogenesis, there was a significant difference in the number and ratio of excitatory and inhibitory synapses in hippocampal
versus cortical cell cultures. In general, there were significantly more excitatory and inhibitory synapses formed in hippocampal cell cultures relative to cortical cell cultures beginning on DIV 14 through DIV 28. With regards to the ratio of excitatory to inhibitory synapses, previous studies of hippocampal neurons have reported that the number of excitatory synapses generally is greater than the number of inhibitory synapses in mature neuronal cell populations [
40,
41]. Our data are consistent with these prior studies with respect to the total number of excitatory
versus inhibitory synapses and the ratio of excitatory to inhibitory synapses that form on MAP2 immunopositive dendrites. The one notable exception to this generalization was the ratio of excitatory to inhibitory synapses formed on neuronal cell bodies in hippocampal cell cultures. The number of vGAT immunopositive puncta formed on neuronal cell bodies was significantly higher in hippocampal
versus cortical neurons, and the ratio between these two synaptic types was such that vGAT immunopositive puncta outnumbered vGLUT1 immunopositive puncta at DIV 14 through 28. Inhibitory synapse input predominates in pyramidal cell somata and it has been speculated this is to match the inhibitory efficacy in dendrites, due to the relatively larger diameter of the pyramidal cell somata [
14,
15]. Thus, the present findings are consistent with previous literature and can perhaps again be attributed to the differing cell compositions of the cultures [
42].
The relative numbers of excitatory and inhibitory synapses are a critical determinant of network activity, which is the functional readout of synaptogenesis. To determine whether the data we obtained from high content imaging of excitatory
versus inhibitory synapses is predictive of network activity, we also measured synaptic function by recording activity in hippocampal and cortical cultures plated onto MEAs. MEAs enable simultaneous, noninvasive extracellular recording over long periods of time in a relatively high-throughput format compared to traditional electrophysiological techniques. Cultured neurons can be followed from the time of isolation until the development of spontaneous firing, providing a unique opportunity to record the ontogeny of neuronal network activity. These measurements have a wide variety of applications, from basic research to drug discovery and toxicology screening [
13,
43,
44]. While there are numerous studies investigating neuronal networks of specific neuronal cell populations [
45,
46], studies comparing the development of neuronal network activity between different neuronal cell types are lacking. In our experiments, the mean firing rate and mean bursting rate, both classical descriptors of macroscopic network activity state, were followed over time [
47,
48]. In both culture models, network activity was evident as measurable spike and burst activity by DIV 7, and this significantly increased by DIV 14, followed by a decrease at subsequent time points. This pattern of an increase in firing activity followed by a transient reduction is consistent with previous literature, although these shifts occurred earlier in our studies [
9,
49]. A possible explanation for the earlier peak and fall of firing activity is the comparatively high density of neurons utilized in our study (~235,000 cells/cm
2). Culture density has consistently been shown to have significant influence on the development, localization, and function of synapses, with higher density cultures exhibiting earlier developmental onset of network activity [
38,
50,
51]. Thus the ontogeny of network activity in our studies is comparable to studies that utilize similar higher-density cultures [
52] .
Interestingly, the ontogeny of network activity was similar between the two culture models; however, at DIV 14, cortical cultures displayed significantly higher firing rates and burst activity than hippocampal neurons. This difference may reflect the higher vGLUT1/vGAT ratio observed in the cell body compartment of cortical neurons relative to hippocampal neurons. Inhibitory GABAergic input modulates firing behavior such that decreased GABAergic neurotransmission shifts firing to a bursting pattern [
53,
54], and a lower ratio of excitatory to inhibitory input can depress bursting behavior in cultured neurons [
55]. The time at which the cell type-dependent differences in bursting behavior manifest may reflect differences in the rate of maturation of
in vitro excitatory connections [
56,
57]. However, a differential distribution of neuronal cell types or glial cells between the two culture models cannot be discounted as contributing to the differences in bursting behavior we observed in hippocampal
versus cortical cultures, as all of these factors can influence neuronal network activity [
58].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
JH participated in the conception and design of the study, conducted the high content imaging studies, performed the statistical analyses and drafted multiple sections of the manuscript. HC participated in the design of the study, set up the primary neuronal cell cultures, performed the MAP2, vGLUT1 and vGAD immunostaining, collected MEA recordings, analyzed MEA data and drafted significant sections of the manuscript. KS participated in the design of the study, performed the ELISA studies and helped edit the manuscript. DY participated in the conception and design of the study and performed initial pilot studies to optimize culture and immunostaining conditions for high content imaging. WM and PL participated in the conception and design of the study, helped with the acquisition of data and interpretation of the data. PL critically revised the draft manuscript. All authors read and approved the final manuscript for submission.