Recordings from neuron–HEK cell cocultures reveal the determinants of miniature excitatory postsynaptic currents

Synaptic transmission proceeds through a sequence of steps, beginning with presynaptic Ca²⁺ entry, followed by Ca²⁺ sensor activation, fusion pore opening, neurotransmitter diffusion into and across the synaptic cleft, postsynaptic receptor activation, local depolarization, and finally electrotonic spread of dendritic voltage changes to the cell soma (Sabatini and Regehr, 1999; Lisman et al., 2007). These steps are generally quite rapid, with sub-millisecond kinetics, and they all can impact the nature of a synaptic response. This makes it quite challenging to test the contributions of specific molecules and extract information about individual steps. One of these steps, neurotransmitter flux through a fusion pore, tracks the progress of membrane fusion, and modeling studies indicate that the pore must expand rapidly to produce transmitter concentrations high enough to activate synaptic receptors (Khanin et al., 1994; Clements, 1996; Stiles et al., 1996). In endocrine secretion, fusion pore flux can be measured directly with amperometry recording (Lindau and Alvarez de Toledo, 2003; Chang et al., 2017), but the complicated sequential nature of synaptic transmission makes synaptic fusion pores more difficult to study.

One can simplify this problem by focusing on unitary, quantal events produced by the spontaneous exocytosis of single vesicles. These miniature excitatory postsynaptic currents (mEPSCs) bypass the upstream steps involving Ca²⁺ entry and triggering. However, a number of downstream steps remain, and mEPSCs introduce a new difficulty because of their high variance (Finch et al., 1990; Bekkers, 1994). Factors contributing to this variance may include receptor density and subunit composition, vesicle size, intravesicular transmitter concentration, pre- and postsynaptic apparatus registration, auxiliary proteins, and electrotonic distance (Bekkers et al., 1990; Lomeli et al., 1994; Geiger et al., 1995; Wu et al., 2007). Electrotonic distance is especially likely to obscure fusion pore effects because it slows the mEPSC onset, which depends critically on fusion pore size. Thus, it is difficult to use mEPSCs to investigate the function of exocytotic proteins in synapses, and only a few studies have successfully demonstrated that presynaptic molecular manipulations can alter quantal release (Pawlu et al., 2004; Guzman et al., 2010; Bao et al., 2018; Chiang et al., 2018).

To resolve the various contributions to quantal synaptic release, we have investigated mEPSCs in synapses formed by neurons on heterologous human embryonic kidney 293 (HEK 293) cells. Neuronal–HEK cell cocultures had been developed to assay specific postsynaptic receptor subunits in a synaptic context (Fu et al., 2003; Dixon et al., 2015) and to evaluate cell adhesion molecules in synaptogenesis (Sara et al., 2005; Dong et al., 2007). Synapses will form on HEK cells provided that they express the appropriate postsynaptic proteins. Expression of a synaptic receptor enables a nonneuronal cell to respond to a neurotransmitter such as glutamate. Neuroligins, which complex with neurexins from nerve terminals, recruit excitatory and inhibitory synaptic inputs to nonneuronal cells (Scheiffele et al., 2000; Dean et al., 2003; Dean and Dresbach, 2006). Neuroligins also recruit glutamate receptors into postsynaptic clusters through interactions with PSD-95 and stargazin (TARP γ2; Chen et al., 2000; Schnell et al., 2002; Dean and Dresbach, 2006). HEK cells expressing neuroligin 1, PSD-95, and NMDA receptors (NR1-1a/NR2A) or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (GluA4) recruit synaptic inputs from cerebellar neurons to produce robust mEPSC activity as well as synaptic release evoked by Mg²⁺ (Fu et al., 2003). HEK cells expressing either neuroligin 1 or SynCAM (synaptic cell adhesion molecule) with GluA2 recruit synaptic inputs from hippocampal neurons, producing mEPSCs and synaptic responses evoked by high K⁺ (Biederer et al., 2002; Sara et al., 2005). HEK cells expressing neuroligin-2 and γ-aminobutyric acid A (GABA_A) receptors recruit inhibitory synaptic inputs from hypothalamic neurons to produce miniature inhibitory postsynaptic currents and synaptic responses to action potentials (Dong et al., 2007). Neuroligin 1 and PSD-95 colocalize with the presynaptic protein synapsin in cerebellar–HEK cell cocultures (Fu et al., 2003), and nerve terminals cluster around glutamate receptors expressed in HEK cells in heterologous synapses (Biederer et al., 2002). In addition to permitting precise molecular control of the postsynaptic apparatus, the compact shape of HEK cells eliminates electrotonic distortion. This should mitigate significant sources of variance.

Here, we investigated synapses between mouse hippocampal neurons and HEK cells expressing the GluA2 AMPA-type glutamate receptor subunit, neuroligin 1, and PSD-95. In addition to these molecules employed by Fu et al. (2003), we also included stargazin, which improves receptor targeting (Chen et al., 2000). Detailed comparisons highlighted various contributions to the shapes of mEPSCs. Fusion pore perturbation with synaptobrevin harboring mutations in its transmembrane domain recapitulated changes seen in mEPSCs produced by synapses between neurons (Chiang et al., 2018), thus supporting a role for SNARE protein transmembrane domains in synaptic fusion pores. These studies of mEPSCs in neuron–HEK cell cocultures permit a dissection of the roles of diverse molecular determinants of quantal synaptic release.

DNA constructs encoding neuroligin 1 (harboring a hemagglutinin tag), stargazin-GFP (TARP γ2 fused with GFP), PSD-95, and GFP-GluA2 (GFP fused with the AMPA receptor subunit GluA2) separated by internal ribosome entry site elements were provided by Dr. Lu Chen (Stanford University, Stanford, CA). To provide a separate fluorescence channel, mCherry was subcloned into the GluA2 construct in place of enhanced GFP. PCR was used with a forward primer containing the signaling peptide sequence. The P2A sequence was introduced to replace the internal ribosome entry site using PCR and Gibson assembly. Neuroligin 1–P2A-stargazin-GFP and PSD-95–P2A–mCherry–GluA2 were then subcloned into a retroviral vector driven by a cytomegalovirus promoter containing hygromycin B and puromycin resistance gene cassettes, respectively, for antibiotic selection.

Hippocampal neurons were cultured from embryonic E17.5–18.5 WT C57BL/6 mice. The hippocampus was dissected and dissociated with 0.05% (wt/vol) trypsin-EDTA. Cells were plated at a density of 80–120 × 10³ per well onto 12-mm #1 coverslips coated with 0.2 mg/ml poly-D-lysine (borate buffer, pH 8.5, overnight) in 24-well plates. Neurons were incubated at 37°C with 5% CO₂/air in Neurobasal plus culture medium supplemented with GlutaMAX-I, B27 plus, and ampicillin/streptomycin. AraC (1.5 µM) was added at 2–3 d in vitro (DIV) to reduce glial cell growth. Half the medium was replaced every 3–4 d for the first 2 wk and every 2 d after 16–17 DIV.

HEK 293T cells (from American Type Culture Collection; product #ATCC CRL-3216) were passaged twice a week up to 32 times. Cells in six-well plates were transfected with the above-described neuroligin 1–P2A-stargazin-GFP and PSD-95–P2A–mCherry–GluA2 constructs using calcium phosphate or X-tremeGENE HP (Merck Millipore). Cells were incubated with serum-free opti-MEM (Gibco) during transfection. Medium was switched back to Dulbecco’s modified Eagle’s medium with 10% FBS after 4–6 h. After another 18 h, the cells were digested with 0.05% (wt/vol) trypsin-EDTA and resuspended with the same Neurobasal culture medium used for hippocampal neurons. Cells were added directly to the coverslips (15 × 10³ cells/well) with neurons at either 13–14 DIV or 17–18 DIV. GFP and mCherry fluorescence were used to find HEK cells expressing both constructs for patch clamp recordings.

Whole-cell patch clamp recording was performed between 1 and 3 d after plating HEK cells for coculture. We recorded from both neurons and double-fluorescent HEK cells in the same cultures. An Axopatch 200B (Molecular Devices) amplifier controlled by pClamp 10 (Molecular Devices) was used to acquire one to four 1-min gap-free recordings. Coverslips with cocultured cells were transferred to 35-mm dishes containing extracellular solution consisting of (in mM) 140 NaCl, 5 KCl, 5 CaCl₂, 1 MgCl₂, 10 HEPES, and 10 glucose (pH 7.4 at room temperature). Note that 5 mM of calcium increases mEPSC frequency and facilitates their study (Scanziani et al., 1992; Bekkers and Stevens, 1995; Chiang et al., 2018). NMDA receptors, γ-aminobutyric acid receptors, and action potentials were blocked with 50 µM amino phosphonovaleric acid, 10 µM SR95531, and 0.5 µM tetrodotoxin, respectively. Aniracetam was added to the extracellular solution for the indicated experiments. Patch pipettes (3–7 MΩ when filled) were pulled with Narishige PC-10 puller. After coating with Sylgard, the pipettes were filled with intracellular solution consisting of (in mM) 120 Cs-gluconate, 8 CsCl, 2 NaCl, 10 EGTA, 5 phosphocreatine, 5 HEPES, 2 Mg-ATP, and 0.3 Na-GTP. Membrane potential was held at −65 mV throughout recordings. Series resistance was compensated >75%, and cells with series resistance >20 MΩ before compensation were discarded.

A postsynaptic neuron presents a variety of diverse sites for presynaptic input, including dendritic shafts, spines, and somata. Dendritic inputs can range from proximal to distal locations, and distal synapses elicit slower voltage changes at the soma compared with proximal synapses (Rall, 1967; Rall et al., 1967; Jack and Redman, 1971; Rall, 1989). Thus, with synapses between neurons, variations in this distance can impact electrotonic conduction, making mEPSCs more variable. HEK cells lack these processes and present uniform sites for synaptic inputs. The resulting close proximity between synaptic input and recording electrode eliminates electrotonic distortion. To induce synapse formation on HEK cells, we transfected them with four postsynaptic proteins described in Materials and methods and then cocultured them with hippocampal neurons.

Patch clamp recordings from HEK cells demonstrated functional synaptic inputs from neurons in coculture. Cells were selected based on visual emission in both the red and green fluorescence channels to detect expression of the GluA2-mCherry-PSD-95 construct and the neuroligin 1–stargazin-GFP construct, respectively. Recordings from cells with fluorescence in only one channel never produced mEPSCs. In HEK cells cocultured with 14-DIV neurons, recordings 1–4 d later revealed very little mEPSC activity (Fig. 1 A). Only 4 out of 58 cells produced mEPSCs (7%), and the frequency was very low (Fig. 1 B; cell numbers given at base of bars). This indicated that neurons at this stage either fail to innervate HEK cells or their synapses are not functional. Increasing the plating density of neurons from 80–120K to 120–150K (per coverslip) failed to increase the mEPSC frequency. By contrast, in cocultures with 17-DIV neurons, patch clamp recordings from HEK cells the next day (at 18 DIV) revealed robust mEPSC activity at a frequency of 4.3 ± 0.77 Hz (Fig. 1, A and B) in 24 out of 39 cells (61%; Fig. 1 B; cells with zero mEPSCs not included in frequency average). Recordings from neurons in these cocultures revealed mEPSCs at both 14 and 18 DIV (Fig. 1 A), with a similar frequency of ∼1.3 Hz in both. This frequency was approximately threefold lower than in HEK cells from 18-DIV cocultures (Fig. 1 B). Because mEPSCs were much more robust in HEK cells with 18-DIV neurons, we used these cocultures for further study.

Individual mEPSCs recorded from both neurons and HEK cells were generally well fitted by a function with an exponential rise and decay (Eq. 1 in Materials and methods; Fig. 2 A). The averages computed from the pooled events revealed striking differences, with mEPSCs in HEK cells being larger and faster than those in neurons (Fig. 2 B). To explore these differences in detail, the parameters obtained from fits to Eq. 1 were used to calculate the peak amplitude, area, decay time, 10–90% rise time, and maximum rise rate for each event. Fig. 2 C presents the means for these quantities, with standard errors (based on cell number, stated in bars for amplitude). The amplitude and area showed modest ∼25% increases between neurons at 14 and 18 DIV, but the kinetic parameters were all indistinguishable between the two age groups (Fig. 2 C). mEPSCs in HEK cells had significantly larger amplitudes, but the areas were similar. As apparent from the comparison of plotted mEPSC averages (Fig. 2 B), mean mEPSC rises and decays were more rapid in HEK cells than in neurons (Fig. 2 C). The similar areas, larger amplitudes, and more rapid kinetics in HEK cell mEPSCs are consistent with the elimination of electrotonic conduction. The similar areas suggest that a synaptic input gates the entry of similar amounts of charge at synapses in both neurons and HEK cells. The different kinetic parameters could reflect either more dendritic filtering in neurons or differences in receptor function; dendritic filtering will be examined below.

To focus on mEPSC variability, we plotted the coefficient of variation (CV; Fig. 2 D). All mEPSC parameters had greater variation in neurons at 18 DIV than at 14 DIV (Fig. 2 D), as expected for the greater morphological complexity of older neurons (the plating densities were the same). Although HEK cells present more uniform locations for synapse formation, the amplitude and decay time had significantly higher CVs in HEK cells than in neurons (Fig. 2 D). The amplitude CV was 0.50 ± 0.02 for neurons versus 0.62 ± 0.02 for HEK cells (two-sample t test, t = 3.27, P = 0.0019), while the CV for decay was 0.38 ± 0.01 for neurons and 0.50 ± 0.04 for HEK cells (two-sample t test, t = 3.21, P = 0.0023). This indicates that neuronal synapses have other sources of variation in addition to electrotonic distance and molecular composition. Since the only excitatory receptor subunit in the HEK cells is GluA2, the variability in their mEPSCs cannot be attributed to heterogeneity of subunit composition. Nonuniform receptor density is a likely source of variance, and this could arise from variable expression levels of any of the transfected proteins. They can all influence receptor density, as well as receptor recruitment to synaptic contacts. Furthermore, stargazin modulates glutamate receptors (Priel et al., 2005; Tomita et al., 2005; Zhang et al., 2006; Carrillo et al., 2020), and its expression levels could influence mEPSC amplitude and shape.

The CV for area and 10–90% rise times did not differ significantly between neurons and HEK cells (Fig. 2 D). Interestingly, the mEPSC rise rate was not only faster in HEK cells than in neurons but also less variable. The rise rate does not depend on receptor density, so this lower variation may reflect uniformly proximal electrotonic locations of the synaptic inputs in HEK cells. Since the maximum rise rate depends in part on fusion pore size, the reduced variance suggests that mEPSCs in HEK cells may provide a better way to assess synaptic fusion pore function. Since maximum rise rate reflects the earliest and most rapid phase of mEPSC onset, we focused on this quantity and omitted 10–90% rise times from further analysis.

To examine variations in greater detail, we plotted cumulative distributions of mEPSC amplitude (Fig. 3 A), area (Fig. 3 B), decay time (Fig. 3 C), and rise rate (Fig. 3 D). The shifts in distributions of amplitude, decay, and rise rate between neurons and HEK cells are consistent with the differences in mean (Fig. 2 C) and indicate that these differences reflect overall changes across the population rather than changes in the proportions of distinct subsets of mEPSCs. The distributions of area were very similar (Fig. 3 B), and we examined mEPSC area more closely in an attempt to evaluate the vesicle populations for synapses on neurons and HEK cells. Assuming a constant intravesicular glutamate concentration that does not vary between vesicles of different sizes and further assuming a proportional relation between the number of glutamate receptors activated and the number of molecules of glutamate released should result in a scaling relation between vesicle volume and mEPSC area.

Since volume is proportional to diameter cubed, we examined the distribution of the cube root of mEPSC area. In both HEK cells (Fig. 3 E) and neurons (Fig. 3 F), these distributions had a similar appearance, with main peaks at similar values and shoulders on the right. These distributions were well fitted by a sum of two Gaussians with similar parameters (see legend of Fig. 3). Interestingly, the second peak was located at a value of (area)^1/3 that was 39% higher than the first peak for HEK cells and 32% higher than the first peak for neurons. The second peak has 67% of the events in the first peak in the HEK cell distribution and 51% of the events in the first peak in the neuron distribution. Taking the cube of the ratios of the positions of the peaks suggests that the larger population of vesicles has 2.69 times more glutamate molecules in neuron–HEK cell synapses and 2.30 times more glutamate molecules in neuron–neuron synapses. Thus, the larger population of vesicles may result from the fusion of two smaller vesicles. The similar distributions of (area)^1/3 in Fig. 3, E and F support the interpretation that similar populations of vesicles are released at both neuron–neuron and neuron–HEK cell synapses. Thus, in terms of vesicle size, hippocampal nerve terminals function similarly in their targeting of neuronal and nonneuronal cells; the postsynaptic target does not impact the size of the vesicles in the innervating terminal. Thus, the differences between mEPSCs in the two cell types are likely to reflect postsynaptic components.

To evaluate the contributions of AMPA receptors to mEPSC shape, we tested aniracetam, a drug that binds to AMPA receptors and inhibits deactivation and desensitization (Ito et al., 1990; Isaacson and Nicoll, 1991; Vyklicky et al., 1991; Brenowitz and Trussell, 2001). In neuronal synapses, aniracetam increased mEPSC amplitudes by reducing ambient receptor inactivation (Lawrence et al., 2003). In mEPSCs recorded from HEK cells, we found that aniracetam (1 mM) left the amplitude and total area of mEPSCs unchanged but reduced the rates of mEPSC rise and decay (Fig. 5). These changes in kinetics were not seen in mEPSCs in cultured hippocampal neurons (Chiang et al., 2018). The differences in aniracetam actions are likely to reflect differences in pharmacological sensitivity of receptors formed by different glutamate receptor subunits (Johansen et al., 1995). Aniracetam actions are very sensitive to the ambient levels of glutamate and to the residence time of glutamate in the synaptic cleft, so the difference between the actions on HEK cell mEPSCs versus neuronal mEPSCs could also indicate that their synapses differ in one or both of these respects.

To test the impact of fusion pores on mEPSCs in HEK cells, we transfected the hippocampal neurons with synaptobrevin 2 (Syb2) mutants harboring tryptophan in the transmembrane domain. These mutants had been tested previously for effects on mEPSCs in synapses between neurons (Chiang et al., 2018). For these experiments, we prepared neurons from Syb2/cellubrevin double-knockout mice, which have a very low background frequency of mEPSCs. This enabled us to view mEPSCs that depend on an expressed form of Syb2. Neurons were transfected by lentiviral infection at 14 DIV with WT and mutant Syb2. HEK cells were added 3 d later, and patch clamp recordings were made at 18 DIV. These cocultures were thus similar to those in our other 18-DIV experiments. mEPSCs recorded from HEK cells cocultured with neurons from double-knockout mice expressing WT Syb2 were indistinguishable from mEPSCs seen in HEK cells cocultured with neurons from WT mice (compare means for WT Syb2 in Fig. 6 with means for HEK cells in Fig. 2 C).

Expressing Syb2 mutants with tryptophan at position 101, 102, and both 101 and 103 (2W) had no significant effect compared with WT Syb2 on the amplitudes and areas of mEPSCs in cocultured HEK cells (Fig. 6, A and B). However, the 101 and 2W mutations both produced significant reductions in the mEPSC rise rate of ∼25% (Fig. 6 C; 101W: t = 3.57, P = 0.0021; 2W: t = 3.13, P = 0.0057; two-sample t test). A statistically significant increase in the decay time of 45% was also observed with the 101 and 2W mutants (Fig. 6 D; two-sample t test, t = 3.41, P = 0.0030 and t = 4.82, P = 1.4 × 10⁻⁴, respectively). These mutants were previously shown to produce similar changes in mEPSCs of neuronal synapses (Chiang et al., 2018). The residue at position 102 is not a fusion pore liner (Chang et al., 2015; Chiang et al., 2018), and as reported previously with neuron–neuron synapses, this mutant did not change mEPSCs significantly in neuron–HEK cell synapses in any respect (Fig. 6). In neurons, the 2W double mutation was previously shown not to alter the mEPSC decay time (Chiang et al., 2018), but in HEK cells it did. Thus, HEK cells offer improved resolution of the impact of fusion pore perturbations on mEPSC shape, revealing clear and significant effects with data from half the number of cells.

Turning to the average mEPSC of the 2W domain mutant, we found that adjusting only the fusion pore parameter, τ_fp, to 0.38 ms successfully reproduced both the reduced amplitude and the slower rise time but worsened the fit to the decaying phase (Fig. 7 D). Despite the poor fit to the decay, it is nevertheless significant that adjusting a single parameter accounted for the changes in both the rise rate and amplitude. This suggests that the 2W mutation reduces glutamate flux through the synaptic vesicle fusion pore by 58%.

mEPSCs recorded from neurons and cocultured HEK cells exhibited quantitative differences in their properties. The comparison provided insight into how synaptic transmission is influenced by a variety of molecular and cellular controls. Differences between mEPSCs, some of which were subtle, revealed the distinct roles of vesicles, receptors, dendrites, and fusion pores in shaping synaptic transmission. The elimination of electrotonic distortion in the postsynaptic HEK cells of these cocultures provides an experimental system with significant advantages for biophysical studies of mEPSCs, providing improved resolution over cultured neurons. The defined molecular composition of the postsynaptic apparatus represents another advantage. It will be important to extend the study of homomeric GluA2 receptors presented here to naturally occurring heteromeric glutamate receptors. This will depend on how readily the palette of expressed molecules can be expanded beyond the four proteins expressed in the present work. The developmental state of the synapses in this system remains unclear. Although these synapses release spontaneously, a test of evoked release will be necessary for a full comparison with native synapses between neurons. The structural organization of the postsynaptic apparatus in heterologous cells has not been investigated, and comparisons with native synapses will be enhanced by super-resolution imaging studies of synapses on heterologous cells.

This study showed that mouse hippocampal neurons formed functional synapses with HEK cells expressing GluA2, neuroligin 1, PSD-95, and stargazin. These four proteins thus support robust postsynaptic function, but their sufficiency cannot be judged because HEK cells express additional proteins that could contribute (Inada et al., 2016). We found that 18-DIV neurons formed synapses on HEK cells, but 14-DIV neurons did not, even with a higher plating density. Neurons at 14 DIV formed synapses with one another and produced mEPSCs at the same frequency as 18-DIV neurons. Synaptogenesis between 14-DIV neurons may involve signaling molecules not recognized by the neuroligin 1 we expressed in HEK cells. It is also possible that the release at synapses on HEK cells arises from more mature synapses with a larger reserve pool of vesicles (Mozhayeva et al., 2002; Andreae et al., 2012). We observed an mEPSC frequency, amplitude, and decay time in HEK cells similar to those reported by Fu et al. (2003), who used HEK cells expressing the proteins we used except for their omission of stargazin. The mEPSC frequency in our HEK cells was approximately threefold higher than in both 14- and 18-DIV neurons (Fig. 1 B). HEK cells, despite their smaller surface area and lack of dendrites, could conceivably present more sites receptive to innervation, but individual synaptic contacts on HEK cells could also release more vigorously. Previous work on synapses in HEK cells expressing the adhesion molecule SynCAM and the AMPA receptor GluA2 cocultured with hippocampal neurons reported mEPSCs in 30% of the heterologous cells (Biederer et al., 2002) with slightly smaller amplitudes and approximately eightfold lower frequency (Sara et al., 2005). The higher fraction of HEK cells with mEPSCs, higher mEPSC frequency, and larger amplitude we observed with neuroligin 1, PSD-95, GluA2, and stargazin (Fig. 1 B and Fig. 2 C) could reflect either the presence of PSD-95 and stargazin in our study, or a greater efficacy of neurexin–neuroligin signaling (Nam and Chen, 2005; Varoqueaux et al., 2006), or the younger neurons and different species (rat) used by Biederer et al. (2002) and Sara et al. (2005).

Our results support the view that synapses on heterologous cells recapitulate many aspects of native synaptic function (Fu et al., 2003; Sara et al., 2005). mEPSCs in HEK cells had greater amplitudes and more rapid rises and decays. Speed is generally viewed as an important adaptive property in synaptic transmission (Jackson, 1989). From this perspective, the present results suggest that the performance of synapses on HEK cells exceeds that of synapses on neurons. Rapid synaptic transmission depends on narrow synaptic clefts and close alignment between pre- and postsynaptic elements. The present results indicate that neuron–HEK cell synapses recapitulate these structural features. If detailed structural studies of the molecular organization of synapses between neurons (Biederer et al., 2017) can be complemented with comparable studies in heterologous synapses, we may be in a position to attempt more detailed interpretations of the quantitative differences in functional dynamics.

With receptors in excess, mEPSC area should reflect the quantity of glutamate released per vesicle (Liu et al., 1999; Hanse and Gustafsson, 2001; Franks et al., 2003), which depends on vesicle size (Bekkers et al., 1990; Zhang et al., 1998; Bruns et al., 2000; Karunanithi et al., 2002) and intravesicular concentration (Song et al., 1997; Wojcik et al., 2004; Wilson et al., 2005; Wu et al., 2007). The similar areas (Fig. 2 C) and distributions of (area)^1/3 (Fig. 3, E and F) suggest that vesicles release the same amount of glutamate from synapses on neurons and on HEK cells. A higher receptor density in HEK cells would then reduce the radial distance within the synaptic cleft that glutamate diffuses before binding to receptors, thus making the rise more rapid and synchronizing channel opening to give a higher amplitude (Fig. 2 C). A higher subsynaptic receptor density could simply reflect stronger GluA2 expression in HEK cells, or better recruitment due to higher expression of PSD-95 and/or stargazin (Bats et al., 2007; MacGillavry et al., 2013), or endogenous N-cadherin (Inada et al., 2016).

The CVs of amplitude and decay time were significantly higher for HEK cell mEPSCs than 18-DIV neuronal mEPSCs, but the CV of rise rate was significantly lower (Fig. 2 D). Since glutamate receptors in neurons show considerable variation in subunit composition (Wenthold et al., 1996; Lu et al., 2009), the comparison presented here suggests that this variation has relatively little impact on the shape and variance of mEPSCs in neurons. The greater CV for amplitude and decay time in HEK cells is consistent with heterogeneity of receptor density. Receptor density will impact mEPSC amplitude and decay directly, and packing synaptic inputs at a higher density will reduce the distance between contacts to enhance transmitter spillover and further increase variations (Nielsen et al., 2004; Sargent et al., 2005). By contrast, the CV of mEPSC rise rate was lower in HEK cells (Fig. 2 D), and this may reflect the uniform absence of dendritic filtering. mEPSC onset in HEK cells may depend on less variable factors such as fusion pore flux.

Q→R edited flop is the most abundant GluA2 form in the brain (Wright and Vissel, 2012). We used this subunit to generate postsynaptic glutamate receptors, and with only this subunit expressed in HEK cells, their synapses will have a uniform population of homomeric GluA2 receptors. The majority of AMPA receptors in adult hippocampal neurons are GluA1/GluA2 and GluA2/GluA3 heteromers (Wenthold et al., 1996; Lu et al., 2009). Although GluA1 is more abundant in hippocampal neurons (Craig et al., 1993), homomeric GluA2 receptors target synaptic contacts (Shi et al., 2001). Ablating GluA1 and GluA3 produces a large reduction in evoked synaptic AMPA receptor–mediated synaptic currents, but the remaining homomeric GluA2 receptors produce mEPSC with amplitudes only ∼15% lower than controls, with decays that are ∼30% longer (Lu et al., 2009). By contrast, decays were approximately twofold faster in HEK cells than in neurons (Fig. 2 C), suggesting an influence of factors other than GluA subunits. Aniracetam increased the area and amplitude of mEPSCs in neuronal synapses (Chiang et al., 2018), and the different actions seen here in HEK cell mEPSCs (Fig. 5) are likely due to receptor differences (Johansen et al., 1995).

Our kinetic model was drawn from experiments with the same receptor used here, Q→R GluA2 flop, and the same cells, HEK293 (Krampfl et al., 2002). Although a single exponential fitted the decays of individual events (Fig. 2 A), averaged mEPSCs reduced the noise level to reveal a bi-exponential decay that diverged from the simulated mEPSC (Fig. 7). A likely source for these disparities is the accessory protein stargazin, which was not employed by Krampfl et al. (2002). Stargazin modulates AMPA receptors, including those formed from GluA2 (Priel et al., 2005; Tomita et al., 2005; Carrillo et al., 2020), and has been proposed to promote bi-exponential decays in excitatory synaptic currents (Zhang et al., 2006). Thus, cocultures provide a system for testing how molecular variations in the postsynaptic apparatus control the mEPSC time course.

The impact of dendrites on synaptic responses is well established (Rall, 1967; Jack and Redman, 1971; Rall, 1989). Dendritic filtering has been shown to affect mEPSCs and contribute to their variance, but considerable variance remains after correction for cable effects (Ulrich and Lüscher, 1993; Bekkers and Stevens, 1996). The comparison of mEPSCs in neurons and morphologically compact HEK cells (Fig. 1 A) provides a new and direct test of the role of electrotonic conduction. The slower rises and decays in neuronal mEPSCs (Fig. 2 C) qualitatively agree with expectations of cable filtering. However, the higher CVs of amplitude and decay time in HEK cells than in neurons (Fig. 2 D) support the idea of other significant contributions to variance (Ulrich and Lüscher, 1993; Bekkers and Stevens, 1996). Simulations of the effect of dendritic filtering over a range of dendritic locations (Fig. 4) accounted for the slower rate of rise and suggested that inputs are relatively proximal (X = 0.1). However, the simulations failed to provide a full account for the slower decay. Our modeling of cable effects indicated that the mEPSCs recorded in neurons could not be generated by electrotonic distortion of synaptic inputs of the form recorded in HEK cells (Fig. 4, B and C). These results suggest that electrotonic conduction slows the onset of neuronal mEPSCs, while receptors, auxiliary subunits, and electrotonic conduction work together to shape the decay.

Flux through the fusion pore shapes the mEPSC rise and is thought to set up a transient steady state in glutamate concentration balanced by radial diffusion away from the release site (Khanin et al., 1994; Clements, 1996; Stiles et al., 1996; Chiang et al., 2018). The role of fusion pore flux was tested in neuron–HEK cell synapses with Syb2 transmembrane domain mutations. As found previously in neuron–neuron synapses (Chiang et al., 2018), incorporating a bulky tryptophan side chain into position 101 or both 101 and 103 reduced the mEPSC rise rate and slowed the decay (Fig. 6). These results are consistent with the role of Syb2 in synaptic (Chiang et al., 2018) and endocrine (Chang et al., 2015) fusion pores. A model of the WT HEK cell mEPSC recapitulated the concomitant reduction in amplitude and rise rate reasonably well by adjusting only one parameter, the fusion pore flux (Fig. 7). These results thus support a model of the initial fusion pore lined by SNARE protein transmembrane domains (Chang et al., 2017). The HEK cell coculture yielded clearer results than pure neuronal cultures, suggesting that eliminating dendritic filtering can improve such studies.

A detailed comparison of mEPSCs in neurons and heterologous HEK cells permitted an evaluation of a number of important factors that contribute to the time course of synaptic transmission. This study strengthened the view that synapses on heterologous cells recapitulate native synaptic function with high fidelity but also revealed subtle differences. Receptor properties, fusion pores, and dendritic filtering impact the shapes of mEPSCs in distinct ways. Neuron–HEK cell cocultures offer a system for investigation of mEPSCs mediated by a postsynaptic apparatus with a defined molecular composition. The high resolution afforded by this system will be useful in further studies to probe the molecular mechanisms of the distinct steps that occur sequentially during synaptic transmission.

Jeanne M. Nerbonne served as editor.

We thank Dr. Lu Chen for providing constructs encoding neuroligin 1, stargazin-GFP, PSD-95, and GFP-GluA2. We thank Mazdak Bradburry for advice on cocultures, and Che-Wei Chang and Vasanthi Jayaraman for helpful discussions and comments.

This study was supported by National Institutes of Health grant NS044057.

The authors declare no competing financial interests.

Author contributions: C.-W. Chiang designed the experiments, generated DNA constructs, prepared the cultures and viral vectors, performed the patch clamp recordings, analyzed data, and wrote the paper. W.-C. Shu assisted in interpretations, analysis, and writing. J. Wan and B.A. Weaver generated DNA constructs. M.B. Jackson designed experiments, analyzed data, and wrote the paper.