Activity-dependent modulation of hippocampal synaptic plasticity via PirB and endocannabinoids

The threshold for Hebbian synaptic plasticity in the CNS is modulated by prior synaptic activity. At adult CA3-CA1 synapses, endocannabinoids play a role in this process, but how activity engages and maintains this retrograde signaling system is not well understood. Here we show that conditional deletion of Paired Immunoglobulin-like receptor B (PirB) from pyramidal neurons in adult mouse hippocampus results in deficient LTD at CA3-CA1 synapses over a range of stimulation frequencies, accompanied by an increase in LTP. This finding can be fully explained by the disengagement of retrograde endocannabinoid signaling selectively at excitatory synapses. In the absence of PirB, the NMDAR-dependent regulation of endocannabinoid signaling is lost, while CB1R-dependent and group I mGluR-dependent regulation are intact. Moreover, mEPSC frequency in mutant CA1 pyramidal cells is elevated, consistent with a higher density of excitatory synapses and altered synapse pruning. Mice lacking PirB also perform better than WT in learning and memory tasks. These observations suggest that PirB is an integral part of an NMDA receptor-mediated synaptic mechanism that maintains bidirectional Hebbian plasticity and learning via activity-dependent endocannabinoid signaling.

somatic deletion of floxed alleles in offspring in addition to brain-specific excision. To ensure that only mice with brain-specific deletion of PirB were used in experiments, genotyping was designed to differentiate fully deleted PirB vs. floxed PirB vs. WT PirB alleles in peripheral somatic tissue (ear punch) for each mouse. Only Cre+ mice with floxed PirB alleles or WT PirB alleles, but not fully excised PirB alleles in the peripheral somatic tissue, were used for experiments. This protocol was employed in addition to recommended breeding strategies for each line designed to minimize the effect of ectopic excision in gametes.
Ai14TdTomato mice were crossed to CamKIIa-Cre+ or Grik4-Cre+ mice for one generation, and the pattern of TdTomato expression was assessed in Cre+ mice of the F1 generation. Previous studies have shown that excision of PirB from PirBfl/fl by Cre recombinase under the control of the Ubc promoter occurs within one week, accompanied by a complete loss of PirB protein after about 3 weeks from the onset of Cre recombinase expression (e.g. Figure 1, (4)). Experiments reported here with conditional postnatal PirB deletion were all performed in mice older than P90, weeks to months after Cre expression, at which point PirB protein should be fully removed from targeted cell types.

Behavior: Delayed match-to-place (DMP) dry maze
A DMP dry maze test was used to assess spatial working/episodic-like learning and memory. The dry maze is thought to measure similar learning abilities to the original DMP water maze (5,6). It was conducted using a modified Barnes (dry) maze apparatus (7) consisting of a 122-cm diameter circular platform with 40 escape holes, each with a diameter of 5 cm placed along three rings of varying distances from the center of the platform (Fig. 1a). The outer ring has 16 holes at 50 cm from the center, the middle ring has 16 holes at 35 cm from the center, and the inner ring has eight holes at 20 cm from the center. An escape box was attached to one of these holes and all holes were left uncovered. Bright overhead lighting (750 lux) and air turbulence created by fans were used to create aversive conditions that would encourage mice to seek out the target hole. Visual cues were placed on all four sides of the maze. Mice were given a series of four trials with 2 minute inter-trial intervals (ITI); the maximum duration of each trial was 90 sec. For each trial, mice were placed in different locations at the edge of the maze and held under a dark cover to prevent directional bias. After 10 sec, the cover was removed and the trial started. The distance from the release point to the escape box was the same within a day. The trial ended if a mouse entered the escape box before the end of the 90 sec. Mice unable to find the escape box were led to it by the experimenter and allowed to enter. As soon as the mouse entered the escape box, the noise was turned off and the mouse remained within the box for 10 sec before being returned to its home cage. Mice were habituated to the escape box for 1 minute on the first day of testing. During the initial two days of testing the escape box was marked by an elevated ping-pong ball (Visual Training protocol) to help mice learn the task, and also enable the experimenter to detect potential gross sensorimotor or visual deficits. Subsequently the ping-pong ball was removed, and mice were tested for five additional days (Hidden Escape Hole protocol). The location of the target escape hole was moved each day during this testing period, while all other parameters remained unchanged. The maze was cleaned with 70% alcohol between each trial. All data was recorded using Ethovision (Noldus Information Technology, Wageningen, The Netherlands). Parameters measured were escape latency, distance moved, and velocity. Mice with low mobility were excluded from the analysis prior to un-blinding.

Hippocampal physiology
Adult hippocampal slice preparation.
Acute hippocampal slices were made from mice older than P90, when hippocampal circuitry is fully mature in terms of ionic conductances (8,9), synaptic plasticity mechanisms (10,11), and the eCB retrograde signaling system (12,13). Slices were prepared using a "protective recovery" method (14). This approach minimizes excitotoxicity and hypoxia. Mice were overdosed via intraperitoneal injection of Ketamine HCl (KetaVed, VEDCO; 0.4 mg/g body weight), xylazine HCl (AnaSed, AKORN; 0.04 mg/g body weight), and acepromazine maleate (AceproJect, Henry Schein; 6x10 -4 mg/g body weight) cocktail. Under deep anaesthesia, mice were perfused transcardially with ice-cold carbogenated ACSF (~3.5 ml/min), then decapitated, and the brains gently but quickly extracted from the scull and immersed in ice-cold N-methyl-d-glucamine (NMDG)-ACSF (see below) for 1 min. To ensure proper slicing angle for transverse hippocampal slices, a 60 degree wedge centered at the midline was cut out from the rostral end of the forebrain. Hemispheres were separated, and each was glued cut side down onto a mounting disk. The disk carrying the hemispheres was immersed in a chamber containing ice-cold carbogenated NMDG-ACSF; a block of 5% agar was glued behind the hemispheres to provide additional support during cutting. For LTP, stimulation intensity used to generate field excitatory postsynaptic potentials (fEPSPs) during both baseline and induction stimulation was set at 30% of the constant current needed to elicit a population spike. LTD was induced with a single low-frequency stimulation (LFS) epoch consisting of 900 pulses delivered at 1Hz; in this case stimulation intensity was set at 50% of spiking threshold both during baseline and induction. These stimulation and induction paradigms are used extensively (15,16). For LTP and LTD measurements in Pyr-KO/Pyr-WT and CA3-Pyr-KO/CA3-Pyr-WT mice, baseline stimulation consisted of a pair of pulses separated by 80 ms. Paired-pulses were delivered at 0.033 Hz; slopes of resulting fEPSPs were used to calculate paired-pulse ratio (PPR) before and after plasticity induction: PPR= fEPSP2 slope/fEPSP1 slope. For chemical LTD, bath application of WIN-55,212-2 (5 µm in ACSF) for 15 min, DHPG (100 µm in ACSF) for 15 min, and NMDA (20 µm in ACSF) for 3 minutes was used. After either LTP or LTD induction, fEPSPs were recorded for 60-90 minutes. Throughout the experiment, a linear fit was used to quantify fEPSP slopes between 20-60% of peak amplitude, ensuring minimal contamination from the fiber volley.
To isolate excitatory synaptic inputs, all LTP and LTD measurements were done in the presence of 100 µm picrotoxina GABAA receptor blocker -in the bath. Picrotoxin eliminates both synaptic and tonic GABA transmission (17), as well as heterosynaptic effects driven by endocannabinoid-dependent inhibitory-LTD (I-LTD) on the threshold for LTP at neighboring CA3-CA1 synapses (18). These experimental conditions allowed us to assess the effect of PirB deletion on homosynaptic plasticity specifically at excitatory CA3-CA1 synapses. To prevent epileptic activity, the CA3 region was cut off prior to each experiment.
Picrotoxin also increases excitability of hippocampal slices via blockade of tonic inhibition, resulting in lower spiking threshold; consequently, the stimulation needed for 30% or 50% of the spiking threshold was readjusted after addition of picrotoxin. In a small number of recordings, slice health deteriorated and slices became hyper-excitable, as evident by the appearance of a population spike; these experiments were discarded.
Flow rate of ACSF through the recording chamber was 2 ml/min. Signals were acquired with a DP304 Differential amplifier (Warner Instruments, Hamden, CT), DigiData 1322A digitizer, using Clampex software (Molecular Devices LLC, Sunnyvale, CA), and analyzed with Clampfit module. Intracellular CA1 recordings.
The whole-cell patch-clamp technique was used to record evoked synaptic currents, mEPSCs, and action potentials from CA1 cells. Healthy CA1 cells were visualized using IR-DIC illumination in an Olympus Evoked currents and spikes were analyzed with Clampfit module. Analysis of mEPSCs was done using MiniAnalysis software, ver. 6.0.7 (Synaptosoft). Prior to analysis, mEPSCs were filtered with a 2 kHz, lowpass Elliptical filter. Synaptosoft parameters were set to: threshold -8 pA, search for a local maximum -20,000 µs; time before a peak for baseline -5,000 µs; period to search a decay time -5,000; fraction of peak to find a decay time -0.5; period to average a baseline -2,000 µs; area threshold -10; number of points to average for peak -3; direction of peak, negative. Immunostaining.

List of Drugs and Vendors
For CB1R and VGlut1 immunostaining, fixed brain tissue (per above) was sectioned at 10 µm through the (Adobe Systems, San Jose, CA). Immunofluorescent images in Fig.4 and Fig. S5. were generated as follows: original Leica microscopy files were imported into Fiji (19) as separate red (VGlut1) and green (CB1R) channel via Bio-Formats Importer plugin (20); each channel was processed to subtract background (Process => Substract Background), and remove laser speckle (Process => Noise => Despeckle); these preprocessed VGlut1 and CB1R channels were then merged (Image => Color => Merge Channels) to visualize co-localization between the two signals as yellow pixels.

Western Blots
For CB1R and NR1 Western blots, cortices were isolated from P90 or older Pyr-WT and Pyr-KO mice (five mice per genotype), and whole-cell lysates were prepared by douncing ten times in a dounce homogenizer and NR2B were normalized to GAPDH levels for each sample.

Supplementary Figure S1. Average savings during DMTP is larger in PirB-/-vs. WT mice.
A) Bar-graphs of average savings in distance between trial 1 and trial 4 for the three non-cued days.   following LTD induction, resulting in LTP instead of LTD (see Fig. 2B, 2E)