Chemicogenetic recruitment of specific interneurons suppresses seizure activity

Enhancing the brain’s endogenous inhibitory mechanisms represents an important strategy for suppressing epileptic discharges. Indeed, drugs that boost synaptic inhibition can disrupt epileptic seizure activity, although these drugs generate complex effects due to the broad nature of their action. Recently developed chemicogenetic techniques provide the opportunity to pharmacologically enhance endogenous inhibitory mechanisms in a more selective manner. Here we use chemicogenetics to assess the anti-epileptic potential of enhancing the synaptic output from three major interneuron populations in the hippocampus: parvalbumin (PV), somatostatin (SST) and vasoactive intestinal peptide (VIP) expressing interneurons. Targeted pre- and post-synaptic whole cell recordings in an in vitro hippocampal mouse model revealed that all three interneuron types increase their firing rate and synaptic output following chemicogenetic activation. However, the interneuron populations exhibited different anti-epileptic effects. Recruiting VIP interneurons resulted in a mixture of pro-epileptic and anti-epileptic effects. In contrast, recruiting SST or PV interneurons produced robust suppression of epileptiform activity. PV interneurons exhibited the strongest effect per cell, eliciting at least a five-fold greater reduction in epileptiform activity than the other cell types. Consistent with this, we found that chemicogenetic recruitment of PV interneurons was effective in an in vivo mouse model of hippocampal seizures. Following efficient delivery of the chemicogenetic tool, pharmacological enhancement of the PV interneuron population suppressed a range of seizure-related behaviours and prevented generalized seizures. Our findings therefore support the idea that selective chemicogenetic enhancement of synaptic inhibitory pathways offers potential as an anti-epileptic strategy. Significance statement Drugs that enhance synaptic inhibition can be effective anticonvulsants but often cause complex effects due to their widespread action. Here we examined the anti-epileptic potential of recently developed chemicogenetic techniques, which offer a way to selectively enhance the synaptic output of distinct types of inhibitory neurons. A combination of in vitro and in vivo experimental models were used to investigate seizure activity in the mouse hippocampus. We find that chemicogenetically recruiting the parvalbumin-expressing population of inhibitory neurons produces the strongest anti-epileptic effect per cell, and that recruiting this cell population can suppress a range of epileptic behaviours in vivo. The data therefore support the idea that targeted chemicogenetic enhancement of synaptic inhibition offers promise for developing new treatments.

antibiotic and antimycotic solution (with 10,000 units penicillin, 10 mg streptomycin 177 and 25 μg amphotericin B per mL) for up to two feeding sessions after injection and 178 slices were allowed at least two weeks for expression before being used. 179

180
The organotypic hippocampal slices were transferred to a recording chamber, where 181 they were maintained at 28°C and continuously superfused with artificial 182 cerebrospinal fluid (aCSF) containing (in mM): NaCl (120), KCl (3), MgCl2 (0.5-to-1.5), CaCl2 (2-to-3), NaH2PO4 (1.2), NaHCO3 (23), D-Glucose (11) and ascorbic acid 184 (0.2). Osmolarity was adjusted to 290 mOsm and pH was adjusted to 7.36 with 185 NaOH. Oxygen and pH levels were stabilised by bubbling the aCSF with 95% O2 and 186 5% CO2. Neurons within the hippocampal formation were visualised with 10x and 187 60x water-immersion microscope objectives (Olympus BX51WI) and targeted for 188 single or dual-patch whole-cell recordings. Patch pipettes of 4-to-9 MΩ tip resistance 189 were pulled from filamental borosilicate glass capillaries with an outer diameter of 190 To examine the direct effects of activating excitatory DREADDs upon 207 interneuron excitability, current clamp recordings were conducted in aCSF 208 containing kynurenic acid (3 mM) and hM3Dq receptors were activated by bath 209 application of CNO (10-20 μM, Tocris, Bio-Techne). The spontaneous action 210 potential firing rate of each interneuron was compared for a 5-minute period before 211 and after CNO application, having allowed 3 minutes for the CNO to reach the 212 chamber. To measure the post-synaptic GABAergic currents induced by activating 213 hM3Dq receptors in a specific interneuron population, voltage-clamp recordings were 214 conducted by clamping CA1 and CA3 pyramidal neurons at the reversal potential for 215 glutamatergic current (EGLUT) in the presence of kynurenic acid. Once recordings had 216 stabilised, the amplitude of post-synaptic inhibitory conductances were compared 217 across two-minute periods recorded under baseline conditions, after bath application 218 of CNO and then after co-administration of CNO and tetrodotoxin (TTX, 1-2 μM). 219 baseline, to 1.9±1.1 Hz in the presence of CNO (Fig. 1H). 394

Quantification of seizure-like events in vitro
To examine the potential to chemicogenetically enhance the output of two 395 other major hippocampal interneuron subtypes, we used organotypic hippocampal 396 slices generated from SST-cre mice and VIP-cre mice, in order to target hM3Dq 397 receptors to SST and VIP interneurons, respectively (Fig. 2). Consistent with 398 previous evidence that SST interneurons target the dendritic compartments of 399 hippocampal principal neurons (Katona et al., 1999; Lovett-Barron et al., 2012), we 400 confirmed that the soma and processes of SST interneurons are located within 401 stratum oriens and lacunosum-moleculare, and tend to avoid the pyramidal cell 402 layers of the CA regions ( Fig. 2A,B). Immunohistochemistry confirmed that the SST 403 interneurons could be efficiently and specifically targeted with hM3Dq receptors in 404 SST-cre slices. Two-to-four weeks after viral transduction with AAV8-hSyn-DIO-405 hM3Dq-mCherry, the majority of expressing neurons were immunopositive for SST 406 (94.1±1.6 specificity) and the majority of all SST immunopositive neurons were 407 expressing hM3Dq-mCherry (89.6±1.5% efficiency; Fig. 2C). Finally, to assess 408 whether the hM3Dq receptors could be activated and increase the output of the SST 409 interneuron population, current clamp recordings were targeted to hM3Dq-mCherry 410 positive neurons. The addition of CNO was shown to cause a significant increase in 411 the firing rate of SST interneurons, from 1.1±0.5 Hz during baseline, to 3.7±1.3 Hz in 412 the presence of CNO (Fig. 2D). 413 We next assessed the potential to chemicogenetically target the VIP 414 interneuron population (Fig. 2E) (Fig. 2F). 421 Immunohistochemistry confirmed that the VIP interneurons could be efficiently and 422 specifically targeted with floxed constructs delivered by AAV. Two-to-four weeks 423 after transduction, the majority of expressing neurons were immunopositive for VIP 424 (81.2±2.4% specificity) and the majority of VIP immunopositive neurons expressed 425 the floxed construct (81.6±4.4% efficiency ; Fig. 2G). Finally, activating hM3Dq 426 receptors targeted to VIP interneurons by CNO increased the firing rate from 427 0.01±0.01 Hz during baseline, to 0.1±0.1 Hz in the presence of CNO (Fig. 2H). quantification of the total SLE activity and then a breakdown in terms of the effects 447 upon SLE frequency and individual SLE duration. In PV-targeted slices (Fig. 3A) (Fig. 3B). In SST-453 targeted slices (Fig. 3C) (Fig. 3D). In VIP-targeted slices (Fig. 3E) (Fig. 3F). Finally, to assess potential off-target effects of CNO, the drug was 465 bath-applied during the recording of spontaneous SLEs in control slices that were 466 not expressing excitatory DREADDs (Fig. 3G) (Fig. 3H). 471 Comparisons across the different interneuron populations confirmed subtype-472 specific effects. Total SLE activity was reduced by more than half following activation 473 of either PV interneurons (down 58.2±10.3%) or of SST interneurons (down 474 50.8±10.5%), and both of these reductions were significantly greater than the change 475 in total SLE activity induced by recruiting VIP interneurons (up 12.7±15.4%) (Fig.  476   3I). Each of the interneuron populations was able to decrease SLE frequency: PV 477 interneurons by 55.4±10.1%, SST interneurons by 36.7±13% and VIP interneurons by 478 26.3±10.4% (Fig. 3J). Meanwhile, only the VIP population increased individual SLE 479 length by 57.5±28.4%, as compared to the effect upon SLE length of activating SST (-480 16.1±10.6%) or PV interneurons (4.4±14.7%) (Fig. 3K) neurons were recorded with a caesium-based internal solution containing QX-314, and held at EGLUT to isolate inhibitory post-synaptic currents (Fig. 4A,C,E). For each 495 of the three interneuron populations (PV, SST and VIP), CNO-activation of hM3Dq 496 receptors resulted in a significant increase in post-synaptic inhibitory input (Fig.  497   4A,C,E). Furthermore, for each interneuron population, the CNO-induced increases 498 in post-synaptic inhibition were abolished by bath application of the voltage-gated 499 Na + blocker, TTX (1-2 μM; Fig. 4A,C,E). This confirmed that in each case, the post-500 synaptic inhibition was the result of CNO-mediated increases in action potential-501 evoked GABA release from the interneurons. Following recruitment of PV 502 interneurons, the total inhibitory post-synaptic input to pyramidal neurons increased 503 from 11±4.2 pA/ms to 34±6.1 pA/ms and was abolished by TTX (-0.5±0.1 pA/ms; 504 input elicited by CNO activation was found to be similar for an individual PV interneuron (1.0 ± 0.3 fold, relative to a PV interneuron) and an SST interneuron (1.1 521 ± 0.1 fold, relative to a PV interneuron), both of which were five times greater than 522 for an individual VIP interneuron (0.2 ± 0.1 fold, relative to a PV interneuron; Fig.  523   5B). Second, to relate this to anti-epileptic efficacy, we normalised the effects upon 524 total SLE activity by the number of hM3Dq-expressing cells. These data indicated that 525 individual PV interneurons had the greatest anti-epileptic effect (1.0 ± 0.2 fold, 526 relative to a PV interneuron), which was at least five times more than the anti-527 epileptic effect associated with an individual SST interneuron (0.2±0.04 fold, relative 528 to a PV interneuron) or VIP interneuron (-0.1±0.1 fold, relative to a PV interneuron; 529 to-14-month-old PV-cre mice (Fig. 6A). The virus was delivered at multiple depths 543 in the ventral and dorsal hippocampus, which resulted in extensive expression of the 544 hM3Dq receptor across the rostro-caudal axis (Fig. 6B,C). The majority of all virally-transduced neurons were immunopositive for PV (86.42 ± 1.43% specificity) and the 546 majority of PV immunopositive neurons expressed hM3Dq-mCherry (89.17±2.12% 547 efficiency; Fig. 6D). 548 To investigate the effect of activating PV interneurons on epileptiform activity 549 in vivo, acute seizures were triggered by intra-hippocampal infusion of 4-AP. From 550 2-months after viral-mediated delivery of the hM3Dq receptor, the PV-cre mice were 551 implanted with an intra-hippocampal infusion cannula. Animals underwent four 552 seizure experiments, each separated by three days. Animals were randomised to 553 receive an i.p. injection of either CNO or vehicle in their first experiment, and then 554 alternated between CNO and vehicle for subsequent experiments. Each i.p. injection 555 was delivered 15 minutes before the animal was placed in an arena and their freely 556 moving behaviour was monitored for a period of 80 minutes using high-speed, high-557 definition cameras (Fig. 7A). Following a 20-minute baseline period, 4-AP was 558 infused directly into the hippocampus according to a spaced delivery protocol (three 559 4-AP infusions, each separated by 12 minutes; see Materials and Methods). 560 For our initial analysis, we scored the time of the first generalised motor 561 convulsion episode as the 'generalised seizure threshold' (i.e. reaching level 5 on the 562 Racine scoring, at which point 4-AP infusions were terminated). The probability of 563 reaching the generalised seizure threshold in the control experiments (i.e. vehicle 564 injections) was highest between the second and third 4-AP infusions, with a 565 probability of 0.43 ± 0.13 (N=12 experiments in 7 animals, reaching seizure 566 threshold 12-24 minutes after the start of the 4-AP infusion protocol). In contrast, 567 this probability was dramatically reduced to 0.07 ± 0.07 in animals that had received 568 CNO injections to enhance the activity of their hippocampal PV interneurons (N=14 569 experiments in 7 animals). These effects were evident across the monitoring period, 570 as shown by cumulative probability plots for generalised seizure threshold (Fig. 7B).
To assess whether this reduction in generalized seizures was associated with non-572 specific effects upon behaviour, we monitored the animals' locomotor activity 573 throughout the experiment via a camera mounted above the behavioural arena. 574 These data revealed that the distribution of time spent throughout the arena was 575 indistinguishable between the vehicle and CNO-treated groups (Fig. 7C,D), 576 supporting the conclusion that the CNO-mediated reduction in generalised seizures 577 was not associated with a non-specific effect upon locomotor activity. 578 To provide a more detailed description of the epileptic seizure activity, each 579 animal's behaviour was scored blindly using the five-point Racine scale, at a 580 sampling frequency of 1 Hz across at least 70 minutes per experiment ( Fig. 8A; see 581 Materials and Methods). Racine categorization of epilepsy-related behaviour 582 included orofacial clonic activity (Racine 1), head nodding (Racine 2), limb clonic 583 activity (Racine 3), retreating/rearing with orofacial clonic activity (Racine 4) and 584 rearing and falling and/or jumping (Racine 5). These analyses revealed that CNO-585 mediated recruitment of hippocampal PV interneurons caused a reduction in the 586 frequency of epileptic behaviours across the Racine categories (Fig. 8B). To 587 characterise the temporal aspects of these effects, the Racine scoring scale was used 588 to generate an integrated measure of epileptic-related behaviour that could be 589 tracked over time (see Materials and Methods). This integrated measure revealed 590 that CNO-mediated recruitment of hippocampal PV interneurons reduced the 591 occurrence of all epileptic behaviours by 85% compared to controls (Fig. 8C). In 592 summary therefore, these results are consistent with the conclusion that 593 chemicogenetic recruitment of PV interneurons is able to effectively supress seizure 594 activity in vivo. 595

596
Here we demonstrate the potential to generate anti-epileptic effects through the 597 chemicogenetic enhancement of distinct populations of GABAergic interneurons. 598  (Freund and Buzsáki, 1996). SST interneurons meanwhile, mainly synapse on the dendrites of 647 pyramidal neurons (Klausberger and Somogyi, 2008), where they regulate dendritic 648 activation (Miles et al., 1996)  to structures that may be deep within the brain, or to cells that may be distributed 695 over large regions. 696 Chemicogenetic intervention strategies may mitigate these issues, such as the 697 potential to modulate cellular activity on larger spatial and temporal scales. Effects from chemicogenetics can be coordinated across large areas of tissue, due to the 699 systemic delivery of the activating drug (Roth, 2016). Furthermore, the fact that 700 DREADDs are G-protein coupled receptors and act through endogenous cellular 701 mechanisms, may avoid unwanted effects associated with artificial synchronisation individual seizure (Ellender et al., 2014). For these reasons, multiple strategies may 716 be required, perhaps using one strategy for pathologically affected cells within the 717 epileptic focus and another strategy for surrounding healthier circuits. 718 In conclusion, the current work supports the use of selective chemicogenetic 719 targeting of the inhibitory system as an approach to disrupt seizure activity. Such a 720 cell-specific pharmacological strategy has the attraction of being controllable and yet 721 avoiding the system-wide effects of drugs that enhance GABA-mediated inhibition in 722 a non-cell selective fashion. Future work will be required before such cell-targeted 723 strategies can be adopted in a translational context, but increasing support is being 724 provided for this general principle. 725