Short-Term Facilitation of Long-Range Corticocortical Synapses Revealed by Selective Optical Stimulation

Animals

All procedures were carried out in accordance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals and approved by the Michigan State University Institutional Animal Care and Use Committee (IACUC). We used the following mouse lines in this study: Crl: CD1 (ICR) (Charles River: 022), Ai14 (Jackson Labs: 007908) (Madisen et al. 2010), PV-Cre (Jackson Labs: 008069) (Hippenmeyer et al. 2005), SOM-IRES-Cre (Jackson Labs: 013044) (Taniguchi et al. 2011), VIP-IRES-Cre (Jackson Labs: 010908) (Taniguchi et al. 2011), Scnn1a-Tg3-Cre (Jackson Labs: 009613) (Madisen et al. 2010), Syt7 knockout (Jackson Labs: 004950) (Chakrabarti et al. 2003), Ntsr1-Cre (MMRRC: 017266-UCD) (Gong et al. 2007), Rbp4-Cre (MMRRC: 031125-UCD) (Gong et al. 2007), and 5HT3a-EGFP (MMRRC: 000273-UNC). The Ntsr1-Cre, Rbp4-Cre, Scnn1a-Tg3-Cre, Syt7 knockout, and 5HT3a-EGFP mouse lines had ICR genetic backgrounds. All mice, except for ICR, Syt7 knockout, and 5HT3a-EGFP mice, were bred by crossing homozygous or heterozygous Cre mice with homozygous Ai14 reporter mice, resulting in experimental mice that were heterozygous for the indicated genes. Animals were group-housed with same-sex littermates in a dedicated animal care facility maintained on a 12:12 h light–dark cycle. Food and water were available ad libitum. We used both male and female mice in this study.

Stereotactic Virus Injections

For all functional experiments characterizing the properties of CC synapses, we used an adeno-associated virus (AAV2) that encoded genes for hChR2 (H134R)-EYFP fusion proteins (rAAV2/hSyn-hChR2[H134R]-eYFP-WPREpA, AV4384, University of North Carolina Viral Vector Core). For the AAV serotype experiments, we also used the following AAV vectors obtained from Addgene: pAAV1-hSyn-hChR2(H134R)-EYFP, pAAV5-hSyn-hChR2(H134R)-EYFP, and pAAV9-hSyn-hChR2(H134R)-EYFP (Addgene plasmid # 26973). For the fusion protein experiment, we used AAV2/hSyn-hChR2(H134R)-mCherry (AV4385, University of North Carolina Viral Vector Core). Retrograde-Cre experiments were carried out similarly, with injections of AAVretro-Ef1a-mCherry-IRES-Cre (Addgene plasmid # 55632) into vS1 followed by an AAV injection in vM1 that drove Cre-dependent expression of ChR2 (pAAV1-hSyn-hChR2(H134R)-EYFP, Addgene plasmid # 26973). We performed dual-optogenetic experiments with injections of rAAV2/hSyn-hChR2[H134R]-eYFP-WPREpA (AV4384, University of North Carolina Viral Vector Core) into vS1 followed by injecting AAV2/hsyn-Flex-ChrimsonR-tdT (AV6555, University of North Carolina Viral Vector Core) into vM1 1-week after, in a PV-Cre mouse. We achieved selective optical control of local excitatory neurons and axons/terminals by injecting a Cre-dependent AAV2 in Cre-driver mice (Scnn1a-Tg3-Cre and Rbp4-Cre) (rAAV2/EF1a-DIO-hChR2(H134R)-eYFP, AV4378, University of North Carolina Viral Vector Core).

For all surgeries, the virus was injected unilaterally into vS1 or vM1 of mice in vivo, as previously described (Crandall et al. 2015; Crandall et al. 2017). Injections were normally performed on mice ~3 weeks old (Mean injection age: 22.2 ± 0.7 days, range 17–42 days). Briefly, mice were anesthetized with a Ketamine–Dexdomitor mixture diluted in sterile saline (KetaVed, 70 mg/kg; Dexdomitor, 0.25 mg/kg; intraperitoneally). Once deeply anesthetized, mice were placed into a digital stereotaxic frame with an integrated warming base that maintained core body temperature (Stoelting). A thin layer of ophthalmic ointment was applied to the eyes to prevent drying (Patterson Veterinary Artificial Tears). Next, an incision was made over the skull by scalpel or fine surgical scissors, the scalp and periosteum overlying the skull deflected, and a small craniotomy made over the target site. A small volume of the virus was then pressure-ejected via a glass micropipette attached to a Picospritzer pressure system (0.1–0.2 μL per injection site over 5–10 min; titer = 3.1–3.5 × 10¹² viral genomes/mL). In some experiments, we used a virus-retrobead or saline-retrobead mixture (red RetroBeads, Lumafluor, Cat# R180). When comparing different AAV serotypes, we kept titer levels consistent by diluting each viral preparation to the same titer (3.1 × 10¹² viral genomes/mL) just before intracranial injection. Following injection, the pipette was held in place for an additional 5–10 min before being slowly advanced or withdrawn from the brain. After surgery, the scalp was closed with a surgical adhesive (GLUture), and animals were given Antisedan (2.5 mg/kg) to reverse the effects of Dexdomitor. Mice were allowed to recover on a heating pad for 1 h before returning to their home cage. Most experiments were performed ~3 weeks after the virus injections to allow sufficient expression (Mean expression time: 20.4 ± 0.3, range 11–25 days). Coordinates from bregma for vS1 were 3.4 mm lateral, 0.8 mm posterior, 0.40, and 1.0 mm depth. Coordinates for vM1 were 1.25 mm lateral, 0.9 and 1.3 mm anterior, 0.40, and 1.0 mm depth. Coordinates for VPm were 1.8 mm lateral, 0.75 mm posterior, and 3.05 mm depth.

In Vitro Slice Preparation

After allowing ~3 weeks for ChR2-EYFP expression (Mean experimental age: 44.0 ± 3.3 days, Median age: 41 days, range 30–259 days), acute coronal brain slices (300 μm thick) containing vM1 and vS1 were prepared for in vitro recording, as described previously (Crandall et al. 2010; Crandall et al. 2017). In four mice, acute thalamocortical brain slices (300 μm thick) containing VPm and vS1 were prepared, as described previously (Agmon and Connors 1991; Crandall et al. 2017). Briefly, animals were anesthetized with isoflurane before being decapitated. Brains were removed and placed in a cold (~4 °C) oxygenated (95% O₂, 5% CO₂) slicing solution containing (in mM) 3 KCl, 1.25 NaH₂PO₄, 10 MgSO₄, 0.5 CaCl₂, 26 NaHCO₃, 10 glucose, and 234 sucrose. Brain slices were cut using a vibrating tissue slicer (Leica, VT1200S) and then transferred to a holding chamber with warm (32 °C) oxygenated artificial cerebrospinal fluid (ACSF) containing (in mM): 126 NaCl, 3 KCl, 1.25 NaH₂PO₄, 2 MgSO₄, 2 CaCl₂, 26 NaHCO₃, and 10 glucose. Slices were maintained at 32 °C for 20 min and then at room temperature for a minimum of 40 min before recording. Slices containing the injection site were always collected for imaging to confirm injection site accuracy and assess tissue health. We only considered mice in which there were no signs of tissue damage or off-target injections.

In Vitro Electrophysiological Recordings and Data Acquisition

Individual brain slices (300 μm) were transferred to a submersion-type recording chamber and bathed continually (2.5–3.0 mL/min) with warm (32 ± 1 °C) oxygenated ACSF containing (in mM): 126 NaCl, 3 KCl, 1.25 NaH₂PO₄, 1 MgSO₄, 1.2 CaCl₂, 26 NaHCO₃, and 10 glucose. Neurons were visualized using infrared differential interference contrast optics (IR-DIC) with a Zeiss Axio Examiner.A1 microscope equipped with a 40× water immersion objective (Zeiss, W Plan-Apo 40×/1.0 NA) and video camera (Olympus, XM10-IR). Whole-cell recordings were obtained using patch pipettes with tip resistances of 4–6 MΩ when filled with a potassium-based internal solution containing (in mM): 130 K-gluconate, 4 KCl, 2 NaCl, 10 HEPES, 0.2 EGTA, 4 ATP-Mg, 0.3 GTP-Tris, and 14 phosphocreatine-K (pH 7.25, 290 mOsm). Voltages were corrected for a − 14 mV liquid junction potential. The average reversal potential for GABA_A receptor-mediated inhibitory responses was −91 mV when measured in excitatory cortical cells (n = 3 cells). The calculated reversal potential was −95 mV. Neurobiotin (0.25%, w/v; Vector Laboratories) was added to the internal solution to inject into neurons during whole-cell recordings for subsequent identification in a subset of experiments (Supplementary Fig. 3).

Electrophysiological data were acquired and digitized at 20 kHz using Molecular Devices hardware and software (MultiClamp 700B amplifier, Digidata 1550B4, pClamp 11). Signals were low-pass filtered at 10 kHz (current-clamp) or 3 kHz (voltage-clamp) prior to digitizing. During recordings, the pipette capacitances were neutralized, and series resistances (typically 10–25 MΩ) were compensated online (100% for current-clamp and 70–80% for voltage-clamp). Series resistances were continually monitored and adjusted during experiments to ensure sufficient compensation. The local field potential (LFP) was monitored with a patch-style pipette (~0.6 MΩ) filled with 3 M NaCl, and signals were band-pass filtered between 0.1 and 4 kHz. The LFP pipette was placed in L1 for recordings in vS1 and L2/3 for recordings in vM1. Cell-attached recordings were obtained using patch-style pipettes filled with a potassium-based internal solution (see above), and signals were high-pass filtered (100 Hz). All pharmacological agents were bath-applied at least 10 min before subsequent experimental tests. Pharmacological agents included Tetrodotoxin citrate (TTX, Tocris, Cat# 1069), 4-Aminopyridine (4-AP, Sigma, Cat# A78403), DL-AP5 (Sigma, Cat# A5282), DNQX (Sigma, Cat# D0540), Picrotoxin (Sigma, Cat# P1675), and CGP 55845 hydrochloride (Tocris, Cat# 1248).

Photostimulation

ChR2 was optically excited using a high-power white light-emitting diode (LED) (Mightex, LCS-5500-03-22) driven by an LED controller (Mightex, BLS-1000-2). For the two-color photostimulation experiments (Supplementary Fig. 4), ChR2 and Chrimson were excited using a high-power blue (455 nm; LCS-0455-03-22) and red (625 nm; LCS-0625-03-22) LED, respectively, combined by a multiwavelength beam combiner (Mightex, LCS-BC25-0495). LED on/off times were fast (<50 μs) and of constant amplitude and duration when verified with a fast photodiode (Thorlabs, DET36A). The light was collimated and reflected through a single-edge dichroic beam-splitter (Semrock, FF660-FDi02) and a high-magnification water immersion objective (Zeiss, W Plan-Apo 40×/1.0 NA), resulting in an estimated spot diameter of ~1500 μm and maximum white LED power at the focal plane of ~32 mW (~18 mW/mm²). Stimuli were delivered as 0.5 ms flashes and were directed at ChR2-expressing axons/terminals by centering the light over the recorded cell. The LED intensity was typically adjusted to evoke a 200 pA EPSC in the recorded neuron when held in voltage-clamp at −94 mV, near the reversal potential for inhibition (see Supplementary Table 1 for the LED intensities needed for each cell type). This intensity was always subthreshold for the recorded cell (~3 mV; Supplementary Table 1). For the overterminal stimulation experiments (Fig. 6), ChR2 was excited using a high-power white LED (Mightex, BLS-GCS-6500-65-A0510) attached to a Digital Mirror Device (DMD)-based pattern illuminator (Mightex, Polygon 400). The pattern illuminator was used to briefly illuminate a small spot (0.5-ms flashes; 150 μm diameter) centered over the soma (Max LED power across all mirrors ~44.3; Max LED power when illuminating a 150 μm diameter spot ~4.9 mW or 276 mW/mm²). For these experiments, the LED light was reflected through the same water immersion objective as above.

Live Slice Imaging

Before recording, all live brain slices (300 μm) were imaged using a Zeiss Axio Examiner.A1 microscope equipped with a 2.5× objective (Zeiss, EC Plan-Neofluar) and Olympus XM10IR camera and cellSens software. Brightfield and fluorescence images were obtained to confirm the accuracy of the injection, tissue health at the injection site, and the overall expression in axons/terminals in the target region. During imaging, live slices were kept in a submersion recording chamber and continually bathed with oxygenated ACSF (at room temperature).

Immunohistochemistry and Tissue Processing for Neurobiotin

All tissues for immunohistochemistry were prepared from acute coronal brain slices, as described previously (Crandall et al. 2017) (Mean age of genotype pairs: 97.8 ± 9.3 days, range 73–116 days). Briefly, acute brain slices (300 μm) containing vS1 or vM1 were fixed by immersion overnight at 4 °C in a vial containing 4% paraformaldehyde in 0.1 M phosphate buffer (PB) solution. Slices were then transferred to a 30% sucrose in 0.1 M PB solution until re-sectioned (4 °C; 2–3 days). Tissue was resectioned at 60 μm using a freezing microtome (Leica, SM2010 R). Next, sections were immunostained for Syt7, as described previously (Jackman et al. 2016; Turecek et al. 2017). Briefly, sections were washed two times in 0.1 M phosphate buffer followed by three times in 0.1 M phosphate buffer with 0.9% NaCl, pH 7.4 (PBS) (5 min per wash). After washing, sections were preincubated for 1 h at room temperature with a blocking solution (0.1% Tween, 0.25% Triton X-100, 10% normal goat serum in PBS) and then incubated with primary antibody for 24 h with rotation at 4 °C. Following primary incubation, sections were washed five times in PBS (5 min per wash), pre-incubated for 45 min in blocking solution (same as above), incubated with secondary antibody for 2 h at room temperature, and then washed three times in PBS (10 min per wash) and two times in PB (5 min per wash). Sections were mounted and cover-slipped using Vectashield with DAPI (Vector Laboratories H-1500). The primary antibody was rabbit polyclonal anti-synaptotagmin 7 (1:200, Synaptic Systems, Cat# 105-173), and the secondary antibody goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody Alexa Fluor 568 (1:500, Thermo Fisher Scientific, Cat# A-11011). Antibodies were stored and prepared as recommended by the manufacturer and did not undergo repeated freeze/thaw cycles. For experiments comparing genotypes (WT and Syt7 KO), all tissue was prepared, stained, and processed in parallel. For each staining run, no primary and no secondary controls were conducted. Images were taken using a Zeiss Axio Imager.D2 fluorescence microscope equipped with appropriate filter sets and a high-resolution monochrome digital camera (Zeiss, AxioCam) and Zen software. For each genotype pair, images were collected from identical anatomical locations, using the same microscope settings and processed identically in Fiji: ImageJ.

Slices containing cells injected with neurobiotin were transferred to 4% paraformaldehyde in 0.1 M PB solution overnight at 4 °C (18–24 h). The next day, slices were changed to 30% sucrose in 0.1 M PB until resectioned (4 °C; 2–3 days). Individual brain slices were resectioned at 100–150 μm using a freezing microtome (Leica SM2010 R). Floating sections were washed twice in 0.1 M PB followed by three washes in 0.1 M PBS, pH 7.4 (5 min per wash). After washing, sections were incubated for 1 h at room temperature in a blocking solution containing 0.1% Tween, 0.25% Triton X-100, and 10% normal goat serum in PBS. Sections were then incubated using Streptavidin/Biotin Blocking Kit (Vector Labs, SP-2002), 30 min in streptavidin solution, followed by 30 min in biotin solution with a brief rinse of PBS after each. Sections were then incubated with Alexa Fluor 568-conjugated streptavidin (Thermo-Fisher Scientific, S11226, 1:5000, 2 μg/μL) solution in blocking solution for 3 h with rotation at room temperature. Following incubation, sections were washed three times in PBS, then two times in 0.1 M PB solution (10 min per wash), mounted, and cover-slipped using Vectashield Vibrance with DAPI (Vector Laboratories, H1800). Confocal image stacks of labeled neurons were taken using an Olympus FluoView 1000 filter-based Laser Scanning Confocal Microscope with an Olympus 40× Oil UPLFLN O (NA 1.3) objective and updated Version 4.2 software (laser excitation 543 nm).

Electrophysiological Data Analysis

Analysis of electrophysiological data was performed in Molecular Devices Clampfit 11 and Microsoft Excel. Synaptic responses to optical stimulation were measured from postsynaptic neurons recorded in whole-cell voltage-clamp. The amplitude of an evoked EPSC was measured relative to a baseline before the stimulus (0.5–10 ms depending on the frequency of the stimulation). The EPSC peak was measured over the 10 ms immediately after stimulus onset or over the 4.5 ms immediately after stimulus onset for all layer comparisons. Reported values were based on average responses to 3–20 stimuli (typically 10). Values for the fiber volley and opsin potential were based on average responses to ~200 stimuli, and their peaks were measured over the 10 ms immediately after stimulus onset. The field postsynaptic potential (fPSP) slope was measured over a 0.5-ms region immediately after the peak negativity. Intrinsic physiological properties were measured using previously described methods (Crandall et al. 2017). Regular-spiking (RS) neurons were identified using established physiological criteria, including their spike-frequency adaptation, growing afterhyperpolarization (AHP) during spike trains, and relatively long spike half-widths (McCormick et al. 1985). In a subset of experiments, we confirmed that RS cells were excitatory (pyramidal) neurons (Supplementary Fig. 3F–H).

Layer Comparisons Using Cre-Driver Mouse Lines

Previous work has shown that the Cre-driver mouse lines used in this study are selective for excitatory cells of distinct layers (L4: Scnn1a-Tg3-Cre; L5: Rbp4-Cre; L6: Ntsr1-Cre) (Olsen et al. 2012; Kim et al. 2014; Adesnik 2018). Live slices prepared from Cre mice crossed with a tdTomato (tdT) reporter (Ai14) confirmed that Cre-expressing cells were mostly confined to their respective layers and that the subpial distances for the layers were in agreement with previous reports (Supplementary Table 3). Layers were identified by drawing horizontal lines where the density of tdTomato-expressing cells decreased (Frandolig et al. 2019). Depth measurements for cells are reported as absolute distance or normalized distance from the top of their respective layer. Whole-cell recordings further confirmed that all tdTomato positive cells had intrinsic physiological properties consistent with excitatory cortical cells (McCormick et al. 1985; Crandall et al. 2017). No Cre expression was observed in vM1 of the Scnn1-Cre mouse, consistent with the idea that the agranular cortex lacks spiny stellate cells (Harris and Shepherd 2015). A Cre-driver mouse line was not used for L2/3 recordings. However, all L2/3 cells were confirmed post hoc to be located within the superficial layers of the cortex (average recorded depth from pia: vS1 L2/3 cells, 190.5 ± 3.9 μm, range: 120–310 μm; vM1 L2/3 cells: 238.0 ± 5.4 μm, range: 135–361 μm). To control for variability in ChR2 expression in different slices and mice, we sequentially recorded from an identified L2/3 RS cell and an excitatory neuron (tdTomato-positive and tdTomato-negative) in a separate layer of the same column. For synaptic strength comparisons, brief flashes of the same intensity were delivered over the recorded soma.

Experimental Design and Statistical Analysis

No formal method for randomization was used in this study. Experiments and data analyses were performed blind to the conditions of the Syt7 experiments. For all other studies, the experimenters were not blind to experimental groups. No statistical methods were used to predetermine sample size, but our sample sizes are similar to those reported in previous studies (Crandall et al. 2015; Crandall et al. 2017). Statistical comparisons were performed in OriginPro 2019. The Shapiro–Wilk test was first applied to determine whether the data had been drawn from a normally distributed population, in which case parametric tests were used. If the assumption of normality was not valid, nonparametric tests were used. Significance was assessed using the appropriate parametric (Paired t-test or two-sample t-test) or nonparametric test (Wilcoxon paired signed-rank test or Mann–Whitney U test) as indicated in Results. A two-way analysis of variance (ANOVA) was used to compare short-term synaptic dynamics across 20 Hz trains. A one-way ANOVA was used for multiple comparisons. All tests were two-tailed. Data are represented as mean ± SEM, and statistical significance was defined as p < 0.05.

Optical Control of CC Synaptic Transmission

Here we applied optogenetic control strategies to examine the connections between vS1 and vM1 (Crandall et al. 2015; Crandall et al. 2017). We identified vM1 and vS1 using viral-mediated anterograde tracing and the presence of large L4 barrels, respectively (Supplementary Fig. 1). vM1 was defined as a narrow region in the posteromedial part of the frontal cortex that reciprocally connected with vS1, consistent with previous studies (Hoffer et al. 2003; Ferezou et al. 2007; Mao et al. 2011). Next, we injected an adeno-associated virus (AAV2)-carrying genes for a channelrhodopsin-2/enhanced yellow fluorescent protein construct (ChR2-EYFP) in vS1 or vM1 of mice in vivo. Three weeks after vS1 injections, there was robust ChR2-EFYP expression in vS1 axons/terminals within vM1 (Fig. 1A and Supplementary Fig. 1A–D). vS1 terminal arbors formed a narrow band ascending from the white matter within vM1 and concentrated in layer 1 (L1) and L2/3, consistent with a previous report (Mao et al. 2011). Similarly, three weeks after vM1 injections, there was intense ChR2-EYFP expression in vM1 axons/terminals within vS1 (Fig. 1B and Supplementary Fig. 1E). vM1 terminal arbors covered a large expanse of vS1 and concentrated in L1 and L5/6, precisely where they are known to terminate (Veinante and Deschenes 2003). Brief LED flashes directly over cortical cells located in the injection site evoked large inward currents with very short onset latencies (<0.35 ms) that were resistant to blocking fast glutamatergic and GABAergic transmission, confirming they expressed ChR2 (Supplementary Fig. 3A).

Previous studies suggest that ChR2 desensitization impairs axons’ ability to follow during repetitive stimulation (Olsen et al. 2012; Hass and Glickfeld 2016), resulting in artificial synaptic depression. To determine the efficacy of ChR2 for activating CC pathways, we stimulated vM1 and vS1 axons/terminals with 20 Hz optical trains (10 pulses, 0.5 ms flashes, 1500 μm diameter spot) and measured effects on the extracellular, pharmacologically isolated fiber volley (Hass and Glickfeld 2016) (Supplementary Fig. 2). Fiber volley amplitude directly relates to the number of axons recruited by stimulation (Andersen et al. 1977). We found that optical stimulation evoked remarkably consistent fiber volley waveforms with modest amplitude decay (8.2 ± 3.8%) and minimal latency shift within trains (0.35 ± 0.05 ms; Supplementary Fig. 2E–G). Thus, 20 Hz optical stimulation of CC pathways is reliable under our conditions and can be used to study these synaptic connections.

Long-Range CC Synapses Express Short-Term Facilitation

We first examined the pathway linking vS1 to vM1. We used voltage-clamp recordings to measure optically evoked excitatory postsynaptic currents (EPSCs) from excitatory regular-spiking (RS) neurons in L2/3 of vM1 (Fig. 1C). Here the RS cells were identified using previously described physiological criteria (see Materials and Methods). We used brief, low-intensity LED stimuli to avoid mixed mono- and polysynaptic excitatory activity, keeping the initial synaptic current ~200 pA when measured near the inhibitory reversal potential (−94 mV; see Materials and Methods). This intensity evoked subthreshold potentials for all cells tested (~3 mV) (Supplementary Table 1).

For all L2/3 RS cells recorded near ChR2-expressing vS1 arbors within vM1, brief LED flashes evoked fast EPSCs with onset latencies averaging 2.34 ± 0.04 ms (Supplementary Fig. 3A). These onsets are consistent with the timing of neurotransmission at fast synapses (Sabatini and Regehr 1999) and indicate that cells did not express ChR2. We next used paired-pulse stimulation to assess short-term plasticity (Zucker and Regehr 2002). Pairs of closely spaced optical flashes resulted in facilitation for interstimulus intervals of 50–200 ms and a peak enhancement of ~1.5-fold (Fig. 1C). However, we observed intense short-term depression for interstimulus intervals of 10–20 ms, consistent with reports that ChR2 does not reliably evoke spikes above 25 Hz (Lin et al. 2009). Since presynaptic neurons often discharge several action potentials in vivo, we also assessed synaptic dynamics using trains of flashes. Consistent with paired-pulse stimulation, all responses underwent facilitation during 5–20 Hz optical trains (Fig. 1D and Supplementary Fig. 3D).

We next examined the reciprocal pathway linking vM1 to vS1 (Fig. 1E). Again we recorded from L2/3 RS cells, which receive direct input from vM1 on their apical dendrites in L1 (Petreanu et al. 2009). For most cells located near ChR2-expressing vM1 axons/terminals in L1, brief photostimulation evoked clear synaptic responses with short onset latencies (1.79 ± 0.03 ms; Supplementary Fig. 3A). When we used paired-pulse stimulation, cells responded with an increase in synaptic strength for short interstimulus intervals of 50–500 ms and a peak enhancement of ~2-fold (Fig. 1E). Furthermore, all responses evoked by 2–20 Hz optical trains exhibited facilitation (Fig. 1F and Supplementary Fig. 3E).

The short-onset latencies of the CC responses suggest they are likely mediated by fast ionotropic glutamate receptors. Indeed, the combined application of the selective NMDA and AMPA kainate-type glutamate receptor antagonists APV and DNQX (50 and 20 μM, respectively) eliminated the responses in all cells tested (6/6 cells from four mice) (Supplementary Fig. 3B). Responses were also abolished after blocking sodium channels with tetrodotoxin (TTX: 1 μM) but returned with the subsequent addition of the potassium channel blocker 4-aminopyridine (4AP: 1 mM; 5/5 cells from two mice) (Supplementary Fig. 3C). The latter finding confirms that the light-evoked responses were both action potential-dependent and monosynaptic (Petreanu et al. 2009; Cruikshank et al. 2010).

To rule out the possibility that the observed facilitation reflects the superposition of long-range and local inputs, we compared the responses evoked under control conditions with those recorded on trials that we reduced the likelihood of postsynaptic firing (Supplementary Fig. 4). We suppressed the responsiveness of local excitatory cells either optogenetically by activating cortical parvalbumin (PV)-expressing interneurons that conditionally expressed the red-shifted opsin ChrimsonR (Klapoetke et al. 2014) or pharmacologically by blocking nonlinearities in the synaptic conductances associated with NMDA receptors (APV: 50 μM) (Larkum et al. 2009). We found that silencing local excitatory cells during the activation of long-range projections did not affect the observed facilitation (Supplementary Fig. 4), indicating that local inputs do not confound the CC responses.

To determine if our light stimulation method, which includes overterminal illumination, significantly blunted short-term facilitation, we compared responses evoked by our approach and overaxon stimulation, which produces more physiological responses (Jackman et al. 2014). We found that the two stimulation methods produced similar dynamics during 20 Hz trains (Supplementary Fig. 5), suggesting that our stimulation method offers a reasonable strategy for assessing the short-term plasticity of CC synapses.

Altogether, these results established that reciprocal CC connections between vS1 and vM1 give rise to facilitating synapses capable of fast sustained excitatory transmission during brief periods of elevated presynaptic activity, contrasting with several previous reports (Lee et al. 2013; Kinnischtzke et al. 2014; Petrof et al. 2015; Naskar et al. 2021).

The Temporal Dynamics of Long-Range and Local Intracortical Excitation

The finding that long-range CC responses underwent short-term facilitation is somewhat surprising and intriguing given a key feature of most local excitatory-to-excitatory connections within the cortex is short-term depression (Gil et al. 1999; Reyes and Sakmann 1999; Feldmeyer et al. 2002). To better understand how different excitatory-to-excitatory connections within the neocortex are differentially affected by activity, we compared evoked responses at three general classes of excitatory synapses. At long-range CC synapses, 20 Hz optical stimulation resulted in paired-pulse facilitation (vS1-vM1 L2/3: 1.42 ± 0.08; vM1-vS1 L2/3: 1.80 ± 0.06; Fig. 2A). This facilitation contrasts sharply with long-range subcortical inputs to the cortex from the thalamus, such as core sensory thalamocortical synapses (VPm-vS1 L4/L6) that exhibited robust paired-pulse depression when tested under identical conditions (0.71 ± 0.02; Fig. 2A) (Gabernet et al. 2005; Cruikshank et al. 2010). Conversely, local connections between pairs of excitatory cells in vS1 or vM1 were more stable at 20 Hz, depressing only slightly across the tested population (0.98 ± 0.14; Fig. 2B). Optical stimulation of local ChR2-expressing cortical cells (0.94 ± 0.05) or axons/terminals without parent somata evoked similar paired-pulse ratios (0.95 ± 0.05; Fig. 2B). These data demonstrate that the synaptic dynamics of three common types of glutamatergic synapses in the neocortex are dramatically different (P < 0.0001, two-way ANOVA, stim. 2–10; Fig. 2C), suggesting that they might serve dynamically distinct roles in neocortical operations.

Remarkably, local and long-range CC input to excitatory cells displayed different synaptic properties, suggesting that local and long-distance synaptic terminals of the same pyramidal neurons could exhibit different release properties. To test this hypothesis, we next injected AAVretro-Cre into vS1 to gain retrograde access to vS1-projecting(P) neurons in vM1 and then a Cre-dependent AAV-DIO-ChR2-EYFP into the ipsilateral vM1 to optically control the activity of these cells (Fig. 2D) (Tervo et al. 2016). After allowing for expression, we observed ChR2-EYFP expression in vS1P neurons in vM1. We next recorded nonexpressing cells in vM1 to study the dynamics of vS1P inputs onto local excitatory neurons. In contrast to our finding for long-range vM1-vS1 synapses, we found that local vS1P-evoked EPSCs depressed (P < 0.0001, two-way ANOVA, stim. 2–10; Fig. 2E). These findings imply that the presynaptic properties of CC projection neurons can differ according to whether the target cortical neuron is in their local circuit or a distant cortical area.

Short-Term Plasticity of CC Input as a Function of Cortical Layer

Given the stark laminar differences in the spatial arrangement of vS1 and vM1 projections (Fig. 1), and previous details about long-range CC connectivity (Petreanu et al. 2009; Mao et al. 2011; Kinnischtzke et al. 2014), we wondered if the short-term dynamics of these synapses depended on cortical layer. To investigate this, we compared the responses in excitatory cells located across the vertical depth of the cortex using layer-specific Cre-recombinase driver mouse lines crossed with a tdTomato (tdT) reporter (see Materials and Methods and Supplementary Table 2). To control for variability in ChR2 expression in different slices and mice, we sequentially recorded from an identified L2/3 RS cell and an excitatory neuron (tdT-positive and negative) in a separate layer of the same column.

In vM1, excitatory vS1 currents were strongest in L2/3 and increased 50–70% during 20 Hz trains, whereas currents in L5/6 cells were significantly weaker, initially facilitated (10–15%), and then depressed with repetitive stimulation (Fig. 3A–C). In contrast, in vS1, excitatory vM1 currents were strongest in L6, and responses approximately doubled during 20 Hz trains in L2/3 and L6, whereas short-term plasticity was highly heterogeneous across L5 connections (Fig. 3D–F). Together, these results indicate that many long-range CC synapses undergo short-term facilitation, with absolute strength and dynamics differing with layer.

Properties of Excitatory CC Synapses onto Different L2/3 Interneuron Subtypes

Long-range CC connections in vS1 and other cortical areas also target GABAergic inhibitory interneurons (Dong et al. 2004; Lee et al. 2013; Yang et al. 2013; Kinnischtzke et al. 2014; Zhang et al. 2014; Naskar et al. 2021). To compare CC responses among different subtypes of L2/3 interneurons, we measured optically evoked CC responses from an identified interneuron and a neighboring RS excitatory cell. To target different L2/3 interneuron subtypes, we crossed three well-established Cre-driver mouse lines with tdT reporter mice (Ai14) to label PV-, somatostatin (SOM)-, or vasoactive intestinal peptide (VIP)-expressing interneurons (see Materials and Methods).

Figure 4A,E shows optically evoked CC responses from an interneuron and neighboring RS cell. We found that EPSCs recorded from PV cells had the largest amplitudes and showed slightly less facilitation at 20 Hz than RS cells for both pathways (EPSC peak normalized to RS: vM1 PV cells, 13.3 ± 5.8, median = 4.9; vS1 PV cells, 6.1 ± 1.4, median = 4.8; Fig. 4B,F). EPSCs from SOM cells had much smaller amplitudes for both CC pathways but displayed dramatic short-term facilitation at 20 Hz (EPSC peak normalized to RS: vM1 SOM cells, 0.17 ± 0.03, median = 0.18; vS1 SOM cells, 0.66 ± 0.11, median = 0.55; Fig. 4C,G). In contrast to those of PV and SOM cells, the EPSCs of VIP cells were significantly weaker in the vS1-vM1 pathway, whereas the EPSC amplitudes were statistically similar to RS cells in the vM1-vS1 pathway (EPSC peak normalized to RS: vM1 VIP cells, 0.14 ± 0.06, median = 0.10; vS1 VIP cells, 0.89 ± 0.28, median = 0.64; Fig. 4D,H). Furthermore, the excitatory currents in VIP cells initially facilitated (10–45%) and then depressed with repetitive stimulation for both pathways (Fig. 4D,H). Despite similar EPSC amplitudes, vM1 stimuli always evoked action potentials in the VIP cell (8/8 pairs) but rarely in the RS cell (1/8 pairs) (Supplementary Fig. 6). Altogether, these data suggest that long-range CC synapse strength and dynamics depend on the postsynaptic interneuron subtype and pathway.

Experimental Strategies That Affect CC Responses

Our results establish that facilitation is the predominant communication mode for conveying information between vS1 and vM1. To understand why our data differ from those previously reported, we first compared the effect of AAV serotype on optically evoked short-term synaptic plasticity (Jackman et al. 2014). Here, we injected mice with an identical vector to our previous work using AAV2 but with a different viral serotype (AAV1, AAV5, and AAV9; same titer). All three vectors produced robust ChR2-EYFP expression, similar to levels observed using AAV2 (Supplementary Fig. 7A). When we drove ChR2 expression by AAV5, responses evoked by optical trains exhibited less facilitation for the vS1-vM1 synapse but not for the vM1-vS1 synapse (Fig. 5A,B). However, when using AAV1 to express ChR2, facilitation was weaker for the vM1-vS1 synapse but not for the vS1-vM1 synapse (Fig. 5A,B). Responses were not different from those obtained using AAV2-ChR2 when using AAV9-ChR2. When we examined the effects of different fluorescent fusion proteins on optically evoked responses, we also found that currents evoked by ChR2-mCherry exhibited significantly less facilitation than ChR2-EYFP (Fig. 5C), perhaps due to the overall poor surface expression observed with ChR2-mCherry (Supplementary Fig. 7B). These results indicate that using AAV1 and AAV5 produces only modest facilitation in a pathway-dependent manner and that mCherry tagging of ChR2 caused adverse effects.

We also encountered several cells intermingled with ChR2-EYFP expressing axons/terminals that were light-sensitive when using AAV1, AAV5, and AAV9 (Supplementary Fig. 7C–E). However, we rarely encountered ChR2-expressing cells when using AAV2. These observations suggest that retrograde spread within the cortex is more frequent with some AAV serotypes.

Given direct excitation of ChR2-expressing terminals with light can cause artificial depression (Zhang and Oertner 2007; Jackman et al. 2014), we next wondered if the power density of overterminal illumination affects the magnitude of facilitation at CC synapses. Our approach was to stimulate using brief wide-field illumination (1500 μm diameter spot) and low light power density to minimize complications that may arise due to direct stimulation of presynaptic terminals. To assess the impact of a more focused overterminal illumination, we compared optical responses evoked by the same total light power but with light-spots of different diameters (1500 or 150 μm) (Fig. 6A). We found that restricting the excitation light to the distal ends of ChR2-expressing axons caused the amplitude of facilitation to decrease significantly during repetitive stimulation (Fig. 6B,C). We also found that pulse duration disrupted paired-pulse facilitation in a pathway-dependent manner (Supplementary Fig. 8A,B).

We next examined the Ca²⁺ dependence of optically evoked EPSCs and facilitation. It is well established that Ca²⁺ ions play a crucial role in short-term plasticity (Zucker 1999), and thus far, we have conducted our studies in the presence of 1.2 mM Ca²⁺, which is around physiological conditions (Somjen 2002). However, many studies are performed in the presence of 2.0 mM external Ca²⁺. We found that increasing Ca²⁺ caused a significant increase in EPSC amplitude and reduced paired-pulse facilitation for both synapses (Supplementary Fig. 8C,D). To test further the impact of Ca²⁺, we measured how EPSC trains evoked with repetitive optical stimulation changed with external Ca²⁺. Increasing the concentration of Ca²⁺ to 2.0 mM significantly reduced the magnitude of facilitation observed during 20 Hz trains (Figs 5E and 6D).

Our findings indicate that both light stimulation technique and external Ca²⁺ levels can strongly affect the magnitude of facilitation at CC synapses, suggesting that combining these two experimental approaches could convert facilitating responses to depression. To test this hypothesis, we compared CC responses evoked by our standard approach to those triggered by focused stimulation of the distal ends of ChR2-expressing axons while in the presence of 2.0 mM Ca²⁺ (Fig. 6F). Under control conditions, the responses underwent strong short-term facilitation, whereas subsequent changes to the light and external Ca²⁺ levels led to synaptic currents that depressed (Fig. 6G,H).

Overall, these results reveal how different AAV serotypes, fluorescent fusion proteins, methods of light stimulation, and Ca²⁺ concentrations can influence the short-term synaptic plasticity of CC synapses and perhaps the reliability of synaptically driven circuit activity.

Synaptotagmin 7 Mediates Facilitation at Long-Range CC Synapses

Recent work has shown that the slow presynaptic Ca²⁺ sensor, synaptotagmin 7 (Syt7), plays critical roles in synaptic transmission, especially in short-term facilitation at many central synapses (Jackman et al. 2016). Therefore, we tested whether Syt7 could mediate facilitation at CC synapses. Immunohistochemistry revealed a Syt7 pattern in the hippocampus and thalamus similar to a previous report (Supplementary Fig. 9A) (Jackman et al. 2016). In the neocortex, we observed that Syt7 was present in vM1 and vS1 of wild-type (WT) mice but was absent in Syt7 knockout (KO) animals (Fig. 7A,B). In WT mice, Syt7 labeling was prominent in L1, where CC synapses are dense and are known to contact the dendrites of pyramidal cells (Cauller et al. 1998; Petreanu et al. 2009; Mao et al. 2011), suggesting a possible expression of Syt7 in long-range CC terminals.

We next examined the functional role of Syt7 in facilitation at both CC synapses by performing similar optogenetic experiments in Syt7 KO mice and age-matched WT littermates. In WT mice, pairs of closely spaced optical stimuli and 5–20 Hz trains of flashes resulted in synaptic facilitation (Fig. 7C–F). In Syt7 KO mice, facilitation was eliminated (Fig. 7C–F and Supplementary Fig. 9C). The loss of facilitation in Syt7 KO mice cannot be accounted for by the inability of ChR2 to drive presynaptic axons because optical stimulation evoked fiber volleys comparable to that of WT animals (Supplementary Fig. 9D,E). Nor can it be explained by differences in intrinsic properties between WT and KO cortical cells (Supplementary Table 3). Altogether, these data indicate that Syt7 mediates short-term facilitation at CC synapses linking vS1 and vM1.

Role of Short-Term Dynamics in the Direct Modulation of L2/3 Excitability

Having determined the short-term dynamics of CC synapses, we next examined how these inputs might control the excitability of L2/3 pyramidal neurons. The short-term plasticity of excitatory CC synapses suggests that they may influence L2/3 processing by directly increasing the overall excitation and spike probability of pyramidal neurons with repetitive activity. To test this idea, we recorded in current-clamp from L2/3 RS neurons and performed optical stimulation with 20 Hz trains (Fig. 8A,B). During repetitive stimulation, the membrane potential of L2/3 vM1 and vS1 cells depolarized progressively, reaching its peak toward the ends of the trains. The overall magnitude of facilitation observed with the excitatory postsynaptic potentials (EPSPs) was consistent with the dynamics of the previously recorded EPSCs (Fig. 8A,B). To determine the impact of CC input on L2/3 excitability, we next monitored the spiking behavior in response to 20 Hz trains while holding the cell at a depolarized membrane potential (−74 mV) (Fig. 8C). Under these conditions, trains of optical stimuli reliably drove spiking in L2/3 RS cells. For both vM1 and vS1 neurons, the probability of spike discharge was lowest after the first stimulus and gradually increased following subsequent stimuli so that the maximal spiking probability occurred at the end of the train (Fig. 8D). Together, these data indicate that direct CC effects on L2/3 excitability are regulated dynamically, mainly generating enhancement during brief periods of sustained activity.

In general, the short-term synaptic dynamics of the CC evoked excitatory responses appeared to parallel the CC modulation of L2/3 excitability. To test the causal relationship between short-term facilitation and the changes in excitability, we performed similar current-clamp experiments in Syt7 KO and WT animals. In Syt7 KO mice, the facilitating CC-evoked EPSPs were abolished (Fig. 8E,F). This loss of facilitating CC-evoked excitatory responses eliminated the late enhancement in L2/3 excitability (Fig. 8G,H). These data indicate that the short-term dynamics of inputs to L2/3 cells can account for the CC-triggered late enhancement of their excitability.

Long-Range Projections from vS1 to vS2 Have Similar Synaptic Properties

To determine whether other CC projections might share the same short-term plasticity described for the connections between vS1 and vM1, we examined the vibrissal secondary somatosensory cortex (vS2) and its inputs from vS1. We found that the short-term plasticity of vS1 synapses in L2/3 of vS2 was very similar to that of the vS1-vM1 CC system (Fig. 9). These data suggest that short-term facilitation might be a general feature of long-range CC connectivity.