The short-term plasticity of VIP interneurons in motor cortex

Abstract

Short-term plasticity is an important feature in the brain for shaping neural dynamics and for information processing. Short-term plasticity is known to depend on many factors including brain region, cortical layer, and cell type. Here we focus on vasoactive-intestinal peptide (VIP) interneurons (INs). VIP INs play a key disinhibitory role in cortical circuits by inhibiting other IN types, including Martinotti cells (MCs) and basket cells (BCs). Despite this prominent role, short-term plasticity at synapses to and from VIP INs is not well described. In this study, we therefore characterized the short-term plasticity at inputs and outputs of genetically targeted VIP INs in mouse motor cortex. To explore inhibitory to inhibitory (I → I) short-term plasticity at layer 2/3 (L2/3) VIP IN outputs onto L5 MCs and BCs, we relied on a combination of whole-cell recording, 2-photon microscopy, and optogenetics, which revealed that VIP IN→MC/BC synapses were consistently short-term depressing. To explore excitatory (E) → I short-term plasticity at inputs to VIP INs, we used extracellular stimulation. Surprisingly, unlike VIP IN outputs, E → VIP IN synapses exhibited heterogeneous short-term dynamics, which we attributed to the target VIP IN cell rather than the input. Computational modeling furthermore linked the diversity in short-term dynamics at VIP IN inputs to a wide variability in probability of release. Taken together, our findings highlight how short-term plasticity at VIP IN inputs and outputs is specific to synapse type. We propose that the broad diversity in short-term plasticity of VIP IN inputs forms a basis to code for a broad range of contrasting signal dynamics.

Keywords: VIP; inhibitory interneurons; motor cortex; plasticity; short-term plasticity.

Figure 1

Short-term depression at VIP IN outputs onto MCs and BCs. (A) Schematic illustrating the experimental paradigm. MCs and BCs were targeted for whole-cell recording in L5 of the mouse motor cortex. L2/3 VIP INs were activated with blue laser light which resulted in IPSPs in connected MCs and BCs. (B) Sample traces illustrating recorded IPSPs in a MC and BC following stimulation delivered with blue laser light. Average traces are in blue (MC) and red (BC) while individual responses are in light-blue (MC) and pink (BC). (C) Overlay of average IPSP traces (from B) aligned with respect to onset of blue laser light pulse (light blue square). The initial phase of the IPSP is shown at a higher resolution as an inset. (D) VIP IN→MC and VIP IN→BC connections differed in kinetics. Mean ± SEM was calculated for VIP IN→MC (n = 60 connections, N = 53 animals) and VIP IN→BC (n = 41 connections, N = 35 animals) groups. (Left) Compared to VIP IN→MC synapses, VIP IN→BC had faster 20–80% rise time (VIP IN→MC: 6.0 ms ± 0.2 ms vs. VIP IN→BC: 4.0 ms ± 0.2 ms, unequal variances t-test p < 0.001) and (middle) shorter latency (VIP IN→MC: 6.6 ms ± 0.6 ms vs. VIP IN→BC: 2.6 ms ± 0.2 ms, unequal variances t-test p < 0.001). (Right) IPSP amplitude was larger at VIP IN→MC compared to VIP IN→BC synapses (VIP IN→MC: −1.4 mV ± 0.1 mV vs. VIP IN→BC: −0.86 mV ± 0.2 mV, t-test p < 0.01). Rise time, latency, and amplitude were analyzed based on averages from individual connections. (E) Connection probability (left) was similar for VIP IN→MC connections (97/222, 44%) and VIP IN→BC connections (61/136, 45%). Path strength (right), however, was almost twice as strong for VIP IN→MC compared to VIP IN→BC connections (VIP IN→MC: 62 vs. VIP IN→BC: 38). (F) PPR revealed that VIP IN→MC synapses and VIP IN→BC synapses were short-term depressing when activating L2/3 VIP INs with blue laser light (ChR2: VIP IN→MC: 0.49 ± 0.02, n = 58 connections, N = 51 animals; VIP IN→BC: 0.53 ± 0.03, n = 40 connections, N = 34 animals) and with extracellular stimulation (stim: VIP IN→MC: 0.54 ± 0.07, n = 19 connections, N = 3 cells; VIP IN→BC: 0.62 ± 0.06, n = 23 connections, N = 5 cells). LMM statistics revealed that PPR did not differ between methods for VIP IN→MC synapses (p = 0.69) or for VIP IN→BC synapses (p = 0.31). (G) PPR revealed that VIP IN→MC synapses exhibited short-term depression at all tested frequencies (PPR at 2 Hz: 0.82 ± 0.06, 5 Hz: 0.78 ± 0.06, 10 Hz: 0.64 ± 0.04, 20 Hz: 0.59 ± 0.07, 30 Hz: 0.42 ± 0.05, 40 Hz: 0.46 ± 0.07, 50 Hz: 0.44 ± 0.08). Top right inset: Sample trace of VIP IN→MC IPSPs due to 5 blue light pulses (blue bars) delivered at 20 Hz (n = 9 cells, N = 8 animals). X-axis scale bar: 50 ms; y-axis scale bar: 1 mV. (H) PPR revealed that VIP IN→BC synapses exhibited short-term depression at all tested frequencies (PPR at 2 Hz: 0.76 ± 0.1, 5 Hz: 0.48 ± 0.02, 10 Hz: 0.44 ± 0.1, 20 Hz: 0.45 ± 0.07, 30 Hz: 0.30 ± 0.09, 40 Hz: 0.36 ± 0.1, 50 Hz: 0.39 ± 0.2). Top right inset: Sample trace of VIP IN→BC IPSPs due to 5 blue light pulses (blue bars) delivered at 20 Hz (n = 4 cells, N = 4 animals). X-axis scale bar: 50 ms; y-axis scale bar: 0.2 mV.

Figure 2

VIP IN inputs exhibited heterogeneous short-term dynamics. (A) Schematic illustrating the experimental paradigm. VIP INs were targeted for whole-cell recording in L2/3 of the motor cortex. EPSPs from local excitatory inputs onto VIP INs were generated using extracellular stimulation and were recorded in the patched VIP INs. (B) Example traces from a patched VIP IN illustrating EPSPs in response to 5 pulses of extracellular stimulation delivered at 2, 5, 10, 20, 30, 40, and 50 Hz. Gray traces represent individual responses and black trace represents the average response. (C) On average, E → VIP IN synapses had an EPSP amplitude of 4.1 mV ± 0.4 mV, a 20–80% rise-time of 2.7 ms ± 0.2 ms, and a latency of 1.4 ms ± 0.1 ms (n = 25 cells, N = 18 animals). (D) PPR at all tested frequencies revealed that E → VIP IN synapses had diverse short-term dynamics (PPR at 2 Hz: 1.1 ± 0.04, 5 Hz: 1.2 ± 0.1, 10 Hz: 1.4 ± 0.1, 20 Hz: 1.5 ± 0.1, 30 Hz: 1.4 ± 0.1, 40 Hz: 1.4 ± 0.2, 50 Hz: 1.3 ± 0.2). Top left inset: Sample traces of short-term depressing (left) and short-term facilitating (right) E → VIP IN connections. Scale bars: 50 ms, 3 mV. (E) LMM statistics revealed that PPR at E → VIP IN synapses was different compared to PPR at VIP IN→MC and VIP IN→BC synapses (p < 0.001). In this figure panel, PPR data was pooled across all tested frequencies for each synapse type.

Figure 3

E → VIP IN short-term dynamics linked with target cell rather than with synaptic input (A) In this experimental paradigm, we targeted VIP INs for whole-cell recording and moved the extracellular stimulation pipette to several different positions in the slice to activate multiple presynaptic inputs onto individual VIP INs. (B) Compared to the population average, cells 5 and 7 (red) short-term facilitated (one-tailed t-test of log PPR), an outcome that data shuffling demonstrated was unlikely (18 such outcomes across 100,000 shuffled trials, implying p < 0.001). In conclusion, short-term dynamics associated with the postsynaptic cell rather than with the synaptic inputs. Thick vertical lines: mean.

Figure 4

Wide variability in probability of release at E → VIP IN synapses. (A) Sample experiment illustrating the tuning of a 2-parameter (red) vs. 3-parameter (purple) vs. 4-parameter (blue) model to our fixed frequency short-term plasticity data (black). Due to the presence of short-term facilitation, the 2-parameter model — which only has short-term depression — performs poorly. (B) R² goodness of fit (2-P: 0.34 ± 0.08; 3-P: 0.81 ± 0.02; 4-P: 0.85 ± 0.02), RMS error (2-P: 0.89 mV ± 0.1 mV; 3-P: 0.47 mV ± 0.05 mV; 4-P: 0.41 mV ± 0.04 mV), and KS test (p-values: 2-P: 0.25 ± 0.06; 3-P: 0.52 ± 0.05; 4-P: 0.67 ± 0.05) revealed that the 4-parameter and 3-parameter models fit the data better than the 2-parameter model (n = 25 cells, N = 18 animals). (C) We validated our models using pseudo-random Poisson spike trains. Here, black traces represent a sample Poisson spike train and the purple circles represent the EPSP predictions using the 3-parameter model. The model fit is indicated by the fractional error, calculated as the difference between observed and predicted EPSPs, divided by observed EPSPs. (D) R² goodness of fit (2-P: −0.46 ± 0.3; 3-P: 0.40 ± 0. 2; 4-P: −0.48 ± 0.4;), RMS error (2-P: 0.40 mV ± 0.04 mV; 3-P: 0.25 mV ± 0.02 mV; 4-P: 0.36 mV ± 0.02 mV), and KS test (p-values: 2-P: 0.093 ± 0.09; 3-P: 0.71 ± 0.08; 4-P: 0.65 ± 0.07) revealed that the 3-parameter model best predicted the EPSPs for the Poisson spike trains (n = 9 Poisson, N = 4 cells). (E) The 3-parameter TM model revealed that the depression time constant, D, and the facilitation time constant, F, distributed tightly, whereas there was a relatively broad range of values of the release probability, U.

Figure 5

Short-term plasticity at VIP IN inputs and outputs. We found that E → VIP IN synapses were heterogeneous, showing instances of facilitation as well as of depression. However, VIP IN→MC and VIP IN→BC synapses were consistently short-term depressing.