Mapping multiple synaptic connections to a single neuron
In order to stimulate different sites in the ACC region, we replaced a typical whole-cell patch-clamp recording chamber with the MED64 probe (Fig. 1B). A brain slice was transferred into the chamber and placed so that the ACC was located above the electrodes on the MED64 probe (Fig. 1C). A single neuron located above an electrode was then located and patched (Fig. 1D). Neurons were patched typically in either layer II or III of the ACC and responded to stimulations elicited in surrounding channels up to 600 µM away. After collecting general mapping data, future neurons were only stimulated from directly neighbouring channels. In total, data from 24 neurons, 190 channels, and 23 mice were collected for whole cell experiments using MED64 probe stimulation. 16 neurons were pyramidal while 8 were interneurons further separated as fast-spiking interneurons (FS INs) (n = 5) and regular-spiking interneurons (RS INs) (n = 3). In order to compare observations made from single unit whole-cell recordings to evoked field excitatory postsynaptic potentials (fEPSPs) in multi-channel recordings, data from 7 neurons, 56 channels, and 7 mice were collected using a traditional 64 multi-channel setup. Based on whole-cell experiments, most of the responses from pyramidal neurons were monosynaptic, while others may be polysynaptic (n = 99 monosynaptic connections, n = 42 polysynaptic connections, total n = 141 connections). Thus, approximately 70% of the responses recorded in the patched pyramidal neuron by stimulating different channels were monosynaptic. However, stimulation of other channels using the same stimulation values and protocols led to undetectable responses which was different for each neuron. Typically, a range of 5 to 20 channels would trigger a response in the patched neuron while all other channels would result in an undetectable response.
Excitatory responses recorded from a single pyramidal cell
Pyramidal neurons located in layers II/III were primarily recorded in order to investigate neuronal connectivity in the ACC. These layers were chosen as our previous studies have characterized neurons in these areas, as well as shown them to participate in ACC related functions such as pain transmission and anxiety [17, 18]. Electrical stimulation was delivered by one of the electrodes on the 64-array located beneath the ACC. The following experiment was performed using the selective antagonist of GABAA receptor picrotoxin (PTX) (100µM). A neuron located in layers II/III were patched and neighbouring channels were stimulated in order to check which ones elicited a response in the patched neuron (Fig. 2A). The neuron was identified as a pyramidal neuron of the ACC by injecting depolarizing currents which induced repetitive action potentials with a pattern different from that of a FS- or RS-IN (Fig. 2B). In order to test whether the input responses from neighbouring channels were monosynaptic in nature, 5 shocks at 5 Hz and 20 shocks at 20 Hz were delivered. These synaptic responses followed the repetitive stimuli without failure in the presence of picrotoxin (100 µM) which suggests that these inputs are monosynaptic in nature (Fig. 2B). Different inputs from different channels led to different magnitudes of response in the patched cell as well different latencies from the time of stimulation to the time of response (Fig. 2C). In order to control the number of channels which would elicit a response across multiple slices, the smallest stimulation required to trigger a response in the patched cell was used. Recorded responses could be roughly categorized into three general types and depicted the large variation and heterogeneity between evoked EPSCs: Type I – Rapid onset/slow inactivation; Type II – Slow onset/slow inactivation; Type III – Rapid onset/rapid inactivation (Fig. 2D).
Nerve conduction velocity in the ACC and signal distance
To examine the different conduction velocities of ACC fibers from the surrounding regions to the targeted neuron, we utilized the MED64 to measure the conduction latency between the stimulation site and the recording site. We then calculated the conduction velocity according to the travelling distance based off of the dimensions of the MED64 probe. As shown in Fig. 3A, the arrays were placed on the ACC area based on the brain atlas as well as previous work [19, 20], and the surrounding stimulation channels (Deep layer channels: channel 29, 37 and 45; Superficial layer channels: channel 30, 31, 39, 46, 47) and the recording site (channel 38, located at superficial layer) were marked. Channel stimulation and recording were performed according to typical protocols [21, 22]. The sample traces were recorded at the target site, which show the responses induced when stimulating different ACC layers (Fig. 3B). All the analyzed channels (n = 56 channels/7 slices/7 mice) were divided into two categories based on different distances between the recording site and target site: the long-distance group (283 µm) and the short distance group (200 µm). We found that the conduction velocities in long distance group (0.20 ± 0.04 m/s, n = 19 channels) are significantly faster than those in short distance group (0.14 ± 0.03 m/s, n = 23 channels; t40 = 5.079, p < 0.001, paired t test, Fig. 3C). In other words, EPSC latency did not increase proportionately to distance from the patch-clamped neuron. However, because calculations were done using the shortest distance from stimulus electrode to patch-clamped neuron without taking into consideration actual length of connections and high variation in latency measurements, it is difficult to confidently conclude whether this is a shared trait among long distance connections. There was no significant difference in latency between long and short-distance groups (long distance group, 1.50 ± 0.08 ms; short distance group, 1.46 ± 0.07 ms; t40 = 0.3035, p = 0.7631; data not shown). The overall slow conduction velocity of both groups suggests type C unmyelinated nerve fibers, which are known to play a role in pain signaling and descending pain modulation [23–25].
Conduction velocity between dorsal and ventral ACC
Previous studies have demonstrated that the nerve fibers in the dorsal ACC (ACCd) are more densely packed and longer than those of the ventral ACC (ACCv) [26]. To test the possibility that distinct myelination levels at specific ACC regions may influence conduction velocity, we compared the conduction velocity from both ACC dorsal and ventral sides with the same distance (283 µm). The results show that there is no significant difference of conduction speed between these two groups (ACCd group, 0.19 ± 0.02 m/s, n = 8 channels; ACCv group, 0.20 ± 0.01 m/s, n = 11 channels; t17 = 0.7364, p = 0.4716, paired t test, Fig. 3D). In addition, we also compared the conduction velocity of projections from both deep layers and superficial layers (with the same distance (283 µm) from stimulation site to response site), and the transmission speed between these two groups did not exhibit significant difference (Deep layer group, 0.20 ± 0.01 m/s, n = 11 channels; Superficial layer group, 0.19 ± 0.01 m/s, n = 8 channels; t17 = 0.7306, p = 0.4750, Fig. 3E).
Synaptic transmission response kinetics in interneurons and pyramidal cells in the ACC
We recorded responses from 44 channels across eight interneurons located in layers II/III of the ACC. In terms of channel connectivity, interneurons typically respond to less channels when compared to most, but not all, pyramidal neurons. We were unable to identify any other differences in channel connectivity (Fig. 4A). Two different types of interneurons were identified based on the shape of action potentials: 5 fast-spiking interneurons and 3 regular-spiking interneurons (Fig. 4B). Similar to pyramidal neurons, multiple inputs to a single patched interneuron are not homogenous which is consistent with evidence that connections within neuronal networks are highly heterogeneous, both in structure and activity [27–29] (Fig. 4C). To see if there was a significant difference in response kinetics between neuron types, EPSCs from pyramidal neurons, fast-spiking, and regular-spiking interneurons were analyzed. Both differences in rise and decay time were found to be significantly significant between each pair across all three groups (Rise time; PN: 3.35 ± 0.24 ms; FS IN: 1.81 ± 0.16 ms; RS IN: 6.40 ± 1.32 ms, and Decay time; PN: 17.43 ± 0.93 ms; FS IN: 40.34 ± 6.60 ms; RS IN: 23.78 ± 3.90 ms, ***p < 0.001, one-way ANOVA, Fig. 4D).
For PNs, we found that neurons located farther away from the target site on average had a higher conduction velocity than neurons closer to the target site which was found to be statistically significant (long distance group, 0.13 ± 0.01 m/s, p < 0.05, one-way ANOVA, n = 45 channels/6 mice) (Fig. 5A left). However, this was found to be not statistically significant when the mean conduction velocity was measured according to layer and not distance (short distance group, 0.10 ± 0.01 m/s, p = 0.05167, one-way ANOVA, Fig. 5A right). We also examined the rise and decay of the response observed in the patch-clamped neuron when stimulating a connected channel. Rise and decay time were calculated using Clampfit 8.0. Differences in the rise time of responses induced from deep and superficial layers were found to be statistically insignificant (deep layer group: 2.01 ms ± 0.35; superficial layer group: 2.30 ms ± 0.42, p = 0.3479, one-way ANOVA), as well as in the decay time (deep layer group, 12.37 ms ± 3.10, superficial layer group, 13.70 ms ± 3.07, p = 0.3819, one-way ANOVA, n = 20 channels/3 mice, Fig. 5B left). Differences in response kinetics when categorizing data by long and short distance also showed no statistical significance (long distance rise time, 2.20 ms ± 0.35, short distance rise time, 1.978 ± 0.41, p = 0.6851, one-way ANOVA and long-distance decay time, 12.19 ± 2.97, short-distance decay time, 13.88 ms ± 3.19, p = 0.7024, one-way ANOVA, Fig. 5B right). These data indicate that distance alone does not determine the speed of transmission to the patch-clamped neuron which is consistent with evidence that the conductance velocity of a nerve fiber is dependent on multiple factors such as the level of axon myelination, internode distance, and axon diameter [30, 31]. Comparing averaged values for rise and decay time between distance and layer indicate no difference on the population level, although individual neurons show large variation.
Glutamate-mediated synaptic transmission
We investigated which receptors may be involved in the transmission of synaptic responses within these neuronal networks. Baseline recordings in PNs were performed in the presence of picrotoxin (100 µM) and D-AP5 (50 µM) (Fig. 6A top). This was followed by GYKI 53655 (50 µM) perfusion resulting in a decrease in recorded EPSC amplitude. Because GYKI is an AMPA receptor antagonist, residual EPSCs recorded were considered to be mediated by KA receptors [32] (Fig. 6A middle). Next, perfusion of CNQX (20 µM), a competitive antagonist of both AMPA and KA receptors, was able to block residual currents in all channels (Fig. 6A bottom). A single channel was chosen from 3 mapping recordings to observe the effect of both GYKI 53655 and CNQX over a time course. The average EPSC amplitude of AMPA/KA EPSCs from these recordings were: 133.3 ± 12.3 pA; KA EPSCs: 8.7 ± 0.5 pA, n = 3 neurons/3 mice (Fig. 6B). Relative to the baseline EPSCs consisting of both AMPA/KA mediated EPSCs, KA EPSCs was 11.5 ± 2.9% which was blocked by CNQX (total averaged AMPA/KA EPSCs: 75.2 ± 16.3 pA; KA EPSCs: 7.6 ± 2.0 pA, n = 41 channels, 6 neurons/3mice) (Fig. 6C). The same procedure was also used to measure baseline response followed by drug perfusion in FS INs and RS INs. In FS-INs, GYKI 53655 resulted in a 94.3 ± 1.4% decrease from 105.5 ± 17.3 pA to 6.1 ± 0.6 pA (**p < 0.01). In RS-INs, GYKI 53655 resulted in a 94.2 ± 1.6 % decrease from 60.3 ± 10.6 pA to 3.5 ± 0.4 pA (***p < 0.001) (Fig. 6D). The data shows that the degree of EPSC peak amplitude decrease after application of GYKI 53655 is not the same across all channels suggesting variability of receptor composition located on different synapses.
In order to better visualize the spread of rise and decay time, data were split between monosynaptic and polysynaptic connections and reanalysed (Fig. 6E-G) before the recorded values for individual neurons were mapped on a distribution plot, color-graded to show peak EPSC amplitude. The values for rise time ranged from approximately 0.7 to 10 ms and 1.5 to 58 ms for decay time, representing a large variation in response kinetics observed from the patch-clamped neuron. Rise time values below approximately 3.2 ms and decay time values after 10 ms were associated with higher peak EPSC amplitudes (Fig. 7A). To see if there were any differences in the spread of data when separating monosynaptic and polysynaptic responses, we replotted the data and found that monosynaptic responses have rise and decay time values within a much narrower range compared to that of polysynaptic responses which are spread over a larger area (Fig. 7B). In a subset of experiments, baseline EPSCs were recorded in the presence of PTX and D-AP5, an NMDA receptor antagonist, followed by bath application of GYKI 53655, a non-competitive AMPA receptor antagonist, which attenuated most of the recorded responses. Remaining current was completely blocked by CNQX, suggesting that the receptors responsible for the remaining current were KA receptors excluding GluK3 subunits, which is also blocked by GYKI 53655. This allowed us to plot the ratio between AMPA and KA receptor-mediated responses showing that AMPA receptor-mediated responses composed majority of the response in most of the neurons and that a higher AMPA receptor component was associated with higher peak EPSC amplitude (Fig. 7C left). In a similar experiment, we recorded baseline EPSCs in the presence of PTX and glycine, a co-agonist required for NMDA receptor activation, followed by a change in membrane holding potential to record combined AMPA and NMDA receptor-mediated responses. Application of D-AP5 was able to isolate the AMPA receptor component allowing us to plot the ratio between AMPA and NMDA receptor-mediated responses indicating an average 3:2 ratio (Fig. 7C right). The same approach was used for both FS INs and RS INs which showed that the AMPA receptor-mediated component was much higher than the KA receptor component in the same group as well as what was observed in PNs (Fig. 7D, E). These data suggest possible differences in expression of receptor type and number between neurons types as well as across synapses of the same neuron. Differences in measured ratios of different receptor types observed in the same patch-clamped neuron is likely explained by the high level of molecular diversity of glutamatergic and GABAergic synapses located throughout the brain [33, 34].