Oxytocin signals via Gi and Gq to drive persistent CA2 pyramidal cell firing and strengthen CA3-CA1 neurotransmission

The oxytocin receptor (OXTR) is concentrated in specific brain regions, exemplified by hippocampal subregion CA2, that support social information processing. Oxytocinergic modulation of CA2 directly affects social behavior, yet how oxytocin regulates activity in CA2 remains incompletely understood. We found that OXTR stimulation acts via closure of M-current potassium channels in all OXT-sensitive CA2 neurons. M-current inhibition was persistent in CA2 pyramidal cells, whose prolonged burst firing required functional coupling of the OXTR to both Gαq and Gαi proteins. Other neuromodulators acted via distinct patterns of G-protein signaling to induce CA2 pyramidal neuron burst firing, underscoring its likely importance. CA2 burst firing impacted hippocampal subregion CA1 where stratum oriens-resident CA1 interneurons were targeted more strongly than CA1 pyramidal cells. Oxytocinergic modulation of interneurons, via CA2 pyramidal cell input and directly, triggered a long-lasting enhancement of CA3-CA1 transmission. Thus, transient activation of oxytocinergic inputs may initiate long-lasting recording of social information.


Introduction Figure 1. Excitatory and inhibitory hippocampal neurons show cell-type specific responses to
OXTR stimulation. PV+ interneurons in CA1 and CA2 (a) and CA2 pyramidal cells (b) are depolarized by OXTR stimulation (TGOT, 400 nM, bath application). PV+ interneurons were targeted for recording using a transgenic mouse line (PV-ires-Cre X Ai9). Inset shows an example recorded cell filled with biocytin (streptavidin labeling in purple) and PV-expressing cells in green). The pyramidal cell layer, stratum pyramidale (SP), is outlined in white to distinguish it from the stratum oriens (SO) layer. Arrows point to axonal arborizations. Scale bar = 50 µm. Response latencies were comparable between cell-types; p = 0.39, unpaired t-test (c), while burst durations sharply differed, p = 1.7e-4, Kolmogorov-Smirnov test (d). Raster plots for CA2 neurons are shown in (e), which are collapsed to reflect instantaneous firing frequency in (f). Numbers in parentheses indicate group sizes for data in c -f. Example 2D morphological reconstructions of PV+ neurons are shown in (g). The soma and dendrites are shown in black, while the axon is in red. The SP is demarcated in gray. SR refers to the stratum radiatum. PYR group data from 11 cells / 10 mice. PV group data from 12 cells / 5 mice. 1 supplemental figure.

151
In the prefrontal cortex, a specific class of somatostatin-expressing (SST+) interneurons 152 express the OXTR and are implicated in social-sexual behavior (Nakajima et al., 2014). 153 To test for the recruitment of hippocampal SST+ interneurons by OXTR stimulation, we 154 recorded from fluorescently labeled SST-expressing CA2 interneurons in a transgenic 155 mouse line (SST-Cre x Ai9). The TGOT responses were generally small in this group 156 (2.5 ± 0.8 mV, n=9), though 3 of 9 SST+ cells did display burst firing ( Fig. 1 -Supp. 1). 157 The magnitude of TGOT-induced depolarization was variable even within a defined 158 SST+ subclass: anatomically confirmed OLM (oriens-lacunosum moleculare) 159 interneurons ( Fig. 1 -Supp. 1). 160 In contrast to the PV+ interneuron response, CA2 pyramidal cell responses were highly 161 variable and often long outlasted the stimulus (Fig. 1). We next considered what 162 mechanisms might underlie the cell type-specific persistence of this response to OXTR 163 stimulation. Because TGOT responses in CA2 pyramidal cells are long-lasting even in 164 the presence of excitatory synaptic blockers (Tirko et al., 2018), we focused on 165 intracellular, not synaptic, signaling mechanisms. We first asked whether or not bursting 166 activity was perpetuated via a "latch" mechanism, whereby once the cell started spiking, 167 it entered a self-perpetuating bursting state. To test this, we forced CA2 pyramidal cells In a complementary series of experiments, we asked whether the persistent response 176 was dependent on a continually depolarized membrane potential. After inducing burst 177 firing and depolarization by TGOT application, we held the cell at a hyperpolarized 178 membrane potential (hyperpolarizing the cell by ~8 mV, to near baseline potential, for 179 200 s; Fig. 2b). To our surprise, this sustained hyperpolarization was unable to trigger a 180 return to the cell's baseline state (pre vs post: -58.3 ± 1.2 mV vs. -59 ± 1.3 mV, p = 0.32, 181 paired t-test). After cessation of the hyperpolarizing current, cells went right back to a 182 depolarized membrane potential and burst firing (Fig. 2b). Thus, depolarizing the cell 183 was not enough to induce continual burst firing (Fig. 2a) and hyperpolarizing the cell 184 during TGOT-induced bursting was not sufficient to stop previously triggered activity. 185

186
These results prompted us to consider possible biochemical underpinnings of the 187 persistent activity. In response to OXTR stimulation, CA2 PYR cell input resistance 188 Figure 2. Sustained M-current inhibition is responsible for persistent OXTR responses. When forced to fire bursts of action potentials by a ramping current injection at the soma (example cell in a1), the membrane potential of CA2 pyramidal cells is unchanged (a2; p = 0.44, paired t-test; n = 6 cells / 5 mice). Imposition of a hyperpolarizing current for 200 seconds did not return the TGOT-excited cell to its baseline membrane potential (b; n = 5 cells / 4 mice). Input resistance and membrane potential remain elevated following TGOT application in CA2 pyramidal cells (c). Application of the KCNQ channel opener, retigabine (100 µM, retig.), repolarized the TGOT-excited CA2 PYRs and returned cellular input resistance to baseline values (d). Group data summarized in (e). Control group data comes from 11 cells / 8 mice. Retigabine group data from 5 cells / 3 mice. All error bars reflect the s.e.m. Schematic of known signaling machinery downstream of the OXTR in CA2 cells (modified from (Tirko et al., 2018), (f)). One supplemental figure. increased and remained elevated for the duration of the voltage response (Fig. 2c, Fig.  189 2 -Supp. 1). Previously, we pharmacologically demonstrated that OXTR stimulation 190 increased input resistance due to closure of the M-channel (Tirko et al., 2018). This 191 observation led us to suspect that sustained channel inhibition might be responsible for 192 the long-lasting depolarization and burst firing, even though swifter recovery from 193 inhibition has been found in other neurons, like sympathetic ganglion cells (Suh & Hille, 194 2002). To test explicitly for sustained IM inhibition, we applied the M-current opener, 195 retigabine (100 µM), a few minutes after TGOT removal to determine if this could 196 reverse TGOT-induced excitation (Fig. 2d). Input resistance (Rin), an indirect assay of 197 M-current conductance, was sampled every 10 s with a small hyperpolarizing 198 current step (magnified trace, Fig. 2c, d) (Suh & Hille, 2002;Zhang et al., 2003). We have 208 previously reported that pre-treatment with U73122, commonly employed as a PLC 209 inhibitor, prevents TGOT-mediated excitation (Tirko et al., 2018). We next tested 210 whether U73122 treatment was also able to reverse OXTR-driven depolarization. Unlike 211 retigabine, U73122 was unable to return CA2 PYRs to their resting Vm or input 212 resistance (mean change in Vm following U73122 treatment: 0.31±0.9 mV, n = 11; p = 213 0.37). These data are consistent with sustained PLC activation being dispensable for 214 sustained M-current inhibition. unpaired t-test) to those elicited by TGOT (Fig. 3 -Supp. 1). The similarity between 233 AVP and TGOT responses suggested that signaling through Gaq proteins alone, via the 234 AVP1bR, was capable of producing a long-lasting depolarization. While AVP is capable 235 of signaling through both the OXTR and the AVP1bR (Song & Albers, 2018), application 236 of AVP is capable of producing sustained depolarization in OXTR KO animals ( Fig. 3 -237 Supp. 2). 238 239 Both TGOT and AVP elicit bursting firing in CA2 pyramidal cells, but only TGOT caused 240 highly variable responses. We hypothesized that some of this variability might be due to 241 the activation of multiple G-proteins downstream of the OXTR. To first test whether the 242 OXTR signaled via Gaq, whose activation directly stimulates PLC, we measured the 243 TGOT response in slices pre-treated with the specific Gaq -family inhibitor FR900359 244 In an independent series of experiments, we pre-treated mice with the specific Gai 255 inhibitor pertussis toxin (PTx; intraventricular injection, 24-72 hours before preparing 256 slices). The long pre-treatment with PTx did not alter the electrical properties of CA2 257 PYRs or the frequency of synaptic input (Fig. 3 -Supp. 3). PTx pre-treatment did, 258 however, blunt the depolarization usually caused by TGOT (control v. drug-treated: 259 5.6±0.8 mV (TGOT) v. 1.7±0.5 mV (TGOT+PTx) p = 0.02; unpaired t-test), but not that 260 caused by CCh (control v. drug-treated: 9.6±2.6 (CCh) v. 8.4±0.8 (CCh+PTx) mV; p = 261 0.64; unpaired t-test; Fig. 3d-f). Like FR, PTx pre-treatment also inhibited the TGOT-262 induced increase in input resistance (Fig. 3 -Supp. 4). This PTx sensitivity was specific 263 to pyramidal cells; CA2 PV+ interneurons treated with PTx still responded to TGOT 264 (mean depolarization: 8.8 ± 1.8 mV, n = 3). Furthermore, the PTx sensitivity did not 265 extend to CCh responsiveness (Fig. 3e), indicating that involvement of Gai is not a 266 general prerequisite for persistent bursting. It is equally surprising to find a GPCR whose neuronal signaling requires activation of 273 both Gaq and Gai, although there is precedent for such joint dependence in immune 274 cells (Shi et al., 2007).  Pre-treatment with FR900359 blocks the response to OXTR (a) and acetylcholine receptor (b) activation. Group data summarized in (c). Pre-treatment with pertussis toxin blocked TGOT-induced depolarization in CA2 PYRs (d), but not the response to carbachol, CCh, (e). Group data summarized in (f) as the mean change in membrane potential after drug treatment. Results of one-way ANOVAs (p = 0.001 and p = 0.004 for TGOT and CCh comparisons, respectively) prompted us to make pairwise comparisons between groups, which are reported in text with Tukey-Kramer correction for multiple comparisons. Group sizes are as follows: TGOT only (18 cells / 12 mice); TGOT+FR (9 cells / 7 mice); CCh only (5 cells / 2 mice); CCh+FR (5 cells / 2 mice); TGOT+PTx (9 cells / 5 mice); CCh+PTx (6 cells / 3 mice). 4 supplemental figures.
a Cre-dependent virus encoding ChETA-YFP into the CA2 regions of Amigo2-Cre mice 283 and quantified YFP signal density (Fig. 4a). Consistent with reports from other groups 284 TGOT-stimulated CA2 PYR drive onto excitatory CA1 cells was independent of distance 299 from CA2 and CA1, and whether or not the CA1 PYR was in the deep or superficial 300 pyramidal layer (data not shown). To understand why we did not observe significant 301 synaptic excitation onto CA1 PYRs during TGOT presentation, we revisited CA2 PYR 302 cell 303 304 305 anatomy. As CA2 PYR axons most strongly innervate regions rich in interneurons ( Fig.  306 4a, b), we sought to determine the relative strength of CA2 PYR cell synapses onto 307 excitatory and inhibitory cells in the CA1 SO. In these experiments, we made serial 308 recordings from neighboring CA1 PYR and SO interneuron "pairs", while optogenetically 309 stimulating CA2 PYR cell fibers and keeping the intensity of light stimulation the same.  properties and axonal projection anatomy (Fig. 4 -Supp. 1). In this analysis, we 336 identified multiple interneuron subclasses in our data set. In general, strongly targeted 337 interneurons, (EPSP >5 mV) were characterized by significantly lower input resistance 338 We next asked how strong targeting of CA1 interneurons by CA2 PYRs might locally 349 regulate evoked activity in CA1. First, we considered how acute stimulation of CA2 350 PYRs influences spike transmission between CA3 and CA1 PYRs, evoked via 351 stimulation of Schaffer Collaterals (SC). To do so, we optogenetically mimicked CA2 352 burst firing by delivering light pulses at 20 Hz for 1 s to ChETA-bearing CA2 fibers in 353 CA1 while simultaneously stimulating the SC (Fig. 5 -Supp. 1). Baseline spike 354 probability in response to SC stimulation was established over 20 trials, before 355 interleaving every other stimulus with delivery of the blue light. This burst-like 356 stimulation of CA2 PYR fibers had no effect on CA3-CA1 spike transmission or EPSP 357 amplitude (Fig. 5 -Supp 1), prompting us to ask if optogenetic release of oxytocin, stability, we continuously monitored input resistance, which remained stable throughout 369 the recording period (Fig. 5 -Supp. 2). The enhancement of evoked CA3-CA1 370 transmission was not simply due to direct synaptic modulation by oxytocin, insofar as 371 TGOT did not affect the amplitude or dynamics of SC-evoked synaptic currents in CA1 372 pyramidal cells (Fig. 5 -Supp. 3). Consistent with a role for interneurons in this 373 phenomenon, we observed a trend for the compound IPSP to enlarge upon 374 oxytocinergic stimulation (Change in net IPSP: 1.07±0.5 mV; p = 0.07, one-sample t-375 test) that was not observed following OTA pre-treatment (0.59±0.5 mV; p = 0.36, one-376 sample t-test; Fig. 5c). Also consistent with activation of interneurons, the SC-evoked 377 EPSP narrowed following optogenetic stimulation (change in EPSP width: -0.99±0.4 378 mV; p = 0.04, one-sample t-test), while OTA pre-treated slices actually showed a 379 broadening of the PSP waveform (0.77±0.19 mV; p = 0.03, one-sample t-test; Fig. 5d). 380 of the SC fiber volley was unchanged (Fig. 5 -Supp. 4). In sum, CA1 pyramidal cell 387 input resistance (Fig. 5 -Supp. 2), evoked CA3-CA1 EPSC amplitude (Fig. 5 -Supp.  388 3) and CA3 axonal excitability (Fig. 5 -Supp. 3) were all unchanged in response to 389 oxytocin release, lending support for an underlying circuit mechanism for the 390 enhancement and narrowing of the evoked EPSP. Liang, 2019). It is important to note that a minority of CA1 pyramidal cells, across 462 different data sets, did show an increase in excitatory input upon TGOT application to 463 the slice (Fig. 4; Fig. 5 -Supp. 3). These cells may represent a functionally distinct 464 CA1 pyramidal subclass that is uniquely targeted by CA2. CA3-CA1 PSP is refined and enhanced by oxytocin. The observations that 1) CA1 488 pyramidal cell membrane potential and input resistance are unchanged upon TGOT 489 exposure or endogenous oxytocin release (Fig. 5 -Supp. 2) and 2) CA3 pyramidal cell 490 fiber volley (Fig. 5 -Supp. 2) is unaffected by endogenous oxytocin release weigh 491 against modulation of CA1 or CA3 cell excitability accounting for the potentiation. 492 Similarly, because TGOT did not modulate the amplitude or synaptic dynamics of SC-493 evoked EPSCs, we do not think that oxytocin directly affects CA3 presynaptic release 494 (Fig. 5 -Supp. 3). We propose that a circuit mechanism, perhaps involving interneuron 495 modulation, may underlie the observed increase in EPSP amplitude. Modulation of 496 inhibitory output is implied by the narrowing of the compound PSP and may have been 497 predicted by increased excitability of CA1 interneurons (caused directly by their OXTR 498 stimulation, or indirectly by excitatory synaptic drive coming from CA2 pyramidal 499 neurons). Of these two possibilities, we regard CA2-CA1 interneuron drive to be the 500 weightier contributor to the increased IPSP, which was observed 10+ minutes after light 501 stimulus, well after the direct effect of oxytocin has subsided in interneurons. 502

503
We have previously reported that TGOT application acutely reduces evoked inhibition in 504 the CA1 region of juvenile rats via activity-dependent depression of inhibitory synaptic 505 output (Owen et al., 2013). Here we observed a similar decrease in feed-forward 506 inhibition immediately following TGOT exposure in CA1 pyramidal cells of adult mice 507 ( Fig. 5 -Supp. 3). It remains open whether these acute responses are linked via 508 disinhibition to the induction of sustained excitatory synaptic enhancement (Fig. 5a,b), 509 or to the later strengthening of disynaptic inhibition (Fig. 5c), which may be secondarily 510 further complemented by experiments that discriminate between plasticity at CA2®CA1 519 PYR synapses per se (Tirko et al., 2018) and a heterosynaptic influence on plasticity on 520 other synapses targeting CA1 PYR (Fig. 5). The indirect promotion of synaptic 521 potentiation could provide modulatory enhancement of the storage of incoming 522 information via more classical pathways (e.g. CA3®CA1), switching on recording in 523 dorsal CA1 to support object recognition and in ventral CA1 to promote social 524 memorization, as functionally separated with optogenetics (Raam et al., 2017).  differed from those weakly targeted in the amount of "sag" produced by a 649 hyperpolarizing pulse (relative to the steady state response; c) and input resistance (d).

650
Example responses to hyperpolarizing current injection are shown in the insets.

Stereotaxic injections 704
For all stereotaxic surgeries, mice (aged 4 -10 weeks) were anesthetized with 705 isofluorane (2%-5%) and secured in a stereotaxic apparatus (Kopf). Glass pipettes 706 (Drummond Scientific) were formed using a P-2000 puller (Sutter Instrument) and were 707 characterized by a long taper and 10-20 µm diameter tips. Pipettes were back-filled with 708 mineral oil (Fisher Scientific) before being loaded with virus or toxin (Nanoject II, 709 Drummond Scientific) and positioned at the stereotaxic coordinates indicated below. A 710 small drill hole was made in the skull to allow for pipette insertion. To optogenetically 711 excite cells, we injected pAAV5-EF1a-DIO-ChETA-eYFP-WPRE-HGHpa (Addgene Throughout the surgery, body temperature, breathing and heart rate were monitored. 715 Saline was administered subcutaneously (s.c) to maintain hydration and the animal was MultiClamp 700B amplifier (Axon Instruments, Union City, CA). Signals were filtered at 745 10 kHz using a Bessel filter, digitized at 20kHz with a Digidata 1322A analog-digital 746 interface (Axon Instruments) and analyzed using custom MATLAB scripts (MathWorks). 747 Cellular input resistance was monitored, every 10 seconds, throughout most recordings 748 by regularly giving a small hyperpolarizing step. Negative input resistance values, and 749 those that were more than 2.5 times away from the baseline value were omitted.