Excitation of medium spiny neurons by ‘inhibitory’ ultrapotent chemogenetics via shifts in chloride reversal potential

Ultrapotent chemogenetics, including the chloride-permeable inhibitory PSAM4-GlyR receptor, were recently proposed as a powerful strategy to selectively control neuronal activity in awake, behaving animals. We aimed to validate the inhibitory function of PSAM4-GlyR in dopamine D1 receptor-expressing medium spiny neurons (D1-MSNs) in the ventral striatum. Activation of PSAM4-GlyR with the uPSEM792 ligand enhanced rather than suppressed the activity of D1-MSNs in vivo as indicated by increased c-fos expression in D1-MSNs and in vitro as indicated by cell-attached recordings from D1-MSNs in mouse brain slices. Whole-cell recordings showed that activation of PSAM4-GlyR depolarized D1-MSNs, attenuated GABAergic inhibition, and shifted the reversal potential of PSAM4-GlyR current to more depolarized potentials, perpetuating the depolarizing effect of receptor activation. These data show that ‘inhibitory’ PSAM4-GlyR chemogenetics may activate certain cell types and highlight the pitfalls of utilizing chloride conductances to inhibit neurons.


Introduction
Novel research tools like opto-and chemogenetics have been instrumental in dissecting brain circuits and understanding their relevance to normal and maladaptive behaviors. However, like any new technology, these tools have inherent limitations that could confound data interpretation.
In control neurons, uPSEM 792 (50 nM) did not change the number of evoked action potentials (n = 4-7, ANOVA mixed-effects analysis; no significant group effect: F 1, 6 = 0.83, p=0.40, and no interaction: F 2.1, 8.73 = 1.30, p=0.32; Figure 3B and F) or firing frequency (no significant group effect: F 1, 6 = 0.29, p=0.61, and no interaction: F 2.97, 12.55 = 2.78, p=0.08; Figure 3G). In PSAM 4 -GlyR + neurons, all neurons treated with 10 nM uPSEM 792 continued to fire action potentials with somatic current injections (n = 14, Figure 3C). In 50 nM uPSEM 792 , 10 of 16 neurons continued to In PSAM 4 -GlyR + neurons, 10 or 50 nM uPSEM 792 produced an inward current. Dashed line is baseline whole-cell current for ease of visualization. (B) Plot of the magnitude of the inward current produced by 10, 50, or 100 nM uPSEM 792 in PSAM 4 -GlyR + neurons. Line and error bars represent means ± SEM. (C) To measure membrane resistance (R m ), a 10 mV voltage step (À88 to À78 mV) was made in aCSF and during the uPSEM 792 -induced inward current, and the instantaneous change in current following the capacitive transient was measured (DI). Representative traces are shown below the voltage step command. Dashed line is 0 pA. (D) In PSAM 4 -GlyR + neurons, 50 nM uPSEM 792 significantly decreased R m (paired t-test, p<0.0001) while 10 nM showed a trend toward lower R m (paired t-test, p=0.08). * indicates statistical significance, ns denotes not significant. The online version of this article includes the following source data for figure 2: Source data 1. of PSAM4-GlyR depolarizes D1-MSNs (source data).

PSAM 4 -GlyR induces action potential firing in D1-MSNs in cell-attached recordings
Whole-cell recording conditions affect intracellular chloride levels, which could confound data interpretation. Therefore, we sought to determine the effect of PSAM 4 -GlyR activation on D1-MSN firing using loose cell-attached recording configuration, which does not perturb intracellular chloride concentrations. After obtaining a steady baseline, uPSEM 792 (100 nM) was applied for 5 min, followed by high potassium-containing CSF (~40 mM) to depolarize the neurons. High potassium results in rapid depolarization and firing by shifting the reversal potential of potassium to more positive values. Activation of PSAM 4 -GlyR in D1-MSNs did not suppress high-potassium-induced firing of action potentials ( Figure 4A and B). In fact, in one of five neurons, activation of PSAM 4 -GlyR induced firing prior to high potassium application (not shown). Further, the frequency of high-potassium-induced firing was not different in uPSEM-treated control vs. PSAM 4 -GlyR + neurons, suggesting the absence of a silencing effect by PSAM 4 -GlyR activation.
The whole-cell recording data in Figure 3A suggest that PSAM 4 -GlyR activation results in~20 mV depolarization of D1-MSNs, which is not sufficient to reach firing threshold (~À43 mV [Gertler et al., 2008]) from the resting membrane potential (~À83 mV). Therefore, to determine if the membrane depolarization induced by PSAM 4 -GlyR activation is capable of triggering action potential firing when the membrane potential is closer to firing threshold, the cell-attached experiment was repeated after depolarizing the neurons to subthreshold potentials by increasing the extracellular potassium concentration from 4.5 mM to~11.5 mM (10-13 mM). Under these conditions, cell-attached recordings show that activation of PSAM 4 -GlyR triggered firing of D1 MSNs ( Figure 4C and D). In control cells, uPSEM 792 did not result in firing, unlike high potassium application ( Figure 4C and D). A mixed-effects two-way ANOVA showed a significant interaction effect between PSAM 4 -GlyR expression (+ vs. À) and recording condition (aCSF vs. uPSEM vs. high potassium) on the firing frequency (F 2, 13 = 12.40, p=0.001). Post-hoc Sidak's multiple comparisons tests showed a significant increase in firing frequency with application of uPSEM 792 (compared with aCSF) in PSAM 4 -GlyR + but not control neurons. In contrast, high potassium after uPSEM 792 caused a significant increase in firing frequency in control neurons (compared to aCSF and uPSEM 792 ), as opposed to reduced firing frequency in PSAM 4 -GlyR + neurons, likely because of depolarization block, as is also suggested by the decrement in the amplitude of action currents with uPSEM 792 -induced firing ( Figure 4C).
Taken together, the whole-cell and cell-attached recording results demonstrate that (1) activation of PSAM 4 -GlyR significantly depolarized the membrane potential, (2) the magnitude of the decrease in membrane resistance produced by opening of PSAM 4 -GlyR channels was not sufficient to inhibit neuronal firing via electrical shunting, and (3) PSAM 4 -GlyR activation could trigger action potential firing when neurons are depolarized to subthreshold potentials.

Instability of chloride equilibrium underlies PSAM 4 -GlyR-mediated depolarization
In order to gain a mechanistic understanding of PSAM 4 -GlyR-mediated depolarization, we examined tail currents generated by stepping back to a holding potential of À88 mV from preceding voltage steps. In principle, relaxation of outward currents during voltage steps could reflect fewer available channels (e.g., channel desensitization, inactivation, or voltage-dependent block), which would be expected to result in smaller tail currents. Alternatively, the relaxation of outward currents could be a manifestation of reduced driving force of chloride influx through PSAM 4 -GlyR channels due to the accumulation of intracellular chloride over the course of the voltage steps, which would be expected to result in larger tail currents. Indeed, many reports show that sustained activation of chloride conductances including GABA A receptors (Huguenard and Alger, 1986;Staley et al., 1995;Thompson and Gähwiler, 1989), glycine receptors (Karlsson et al., 2011), and Halorhodopsin (Raimondo et al., 2012;Alfonsa et al., 2015) can lead to chloride influx that overwhelms homeostatic mechanisms to pump chloride out, thereby elevating the intracellular chloride concentration. An increase in intracellular chloride manifests as a rightward shift of the chloride reversal potential, commensurate with our observation of a rightward shift of the PSAM 4 -GlyR reversal potential from À61 mV to À44 mV. Further, the magnitude of the tail current increased substantially with preceding depolarizing steps ( Figure 5F and G; RM two-way ANOVA, significant PSAM 4 -GlyR activation effect: F 1, 30 = 18.05, p=0.0002; significant PSAM 4 -GlyR activation Â voltage step interaction: F 9, 270 = 9.55, p<0.0001; Dunnett's multiple comparisons test shows that tail currents from every preceding voltage step were significantly different from current at À88 mV without a preceding voltage step). These data reveal that the decline in outward current was not due to fewer available channels but was consistent with a change in the driving force and shift in the reversal potential during the depolarizing voltage steps. In addition, the magnitude of the shift in reversal potential was positively correlated with the conductance of PSAM 4 -GlyR, measured at V hold À88 mV (Pearson's correlation, r = 0.57, p=0.0008, n = 21; Figure 5H). The tail current analysis also suggested some intrinsic voltage-sensitivity of PSAM 4 -GlyR. The magnitude of tail currents decreased with preceding hyperpolarizing steps (e.g., À98, -108, or À118 mV) and increased with preceding depolarizing steps (e.g., À78 or À68 mV; Figure 5F and G), suggesting that hyperpolarization reduced the number of available channels, or that depolarization increased the number of available channels. Taken together, the results suggest that activation of PSAM 4 -GlyR in D1-MSNs is largely depolarizing at sub-and peri-threshold membrane potentials likely due to two mechanisms: increased channel availability with depolarization due to intrinsic voltage-sensitive mechanisms, and more prominently, a rightward shift of PSAM 4 -GlyR reversal potential due to accumulation of intracellular chloride at depolarized potentials.

Discussion
Strategies to manipulate neuronal activity in defined cell populations have become instrumental in mapping neural circuits and correlating neuronal and circuit activity with behavior. Engineered receptors to promote excitation (e.g., Channelrhodopsin-2, G q -DREADDs) have been largely successful, albeit with careful consideration of ligand-related off-target effects (Gomez et al., 2017;Manvich et al., 2018). In contrast, silencing of neuronal firing has been uniquely challenging, especially when relying on chloride conductances (Wiegert et al., 2017). The utility of a chloride conductance like PSAM 4 -GlyR to silence firing relies on hyperpolarization of the membrane at potentials more positive than the chloride reversal potential and/or the efficacy of the channels to shunt membrane depolarization (Doyon et al., 2016). The present study shows that PSAM 4 -GlyR in D1-MSNs is predominantly depolarizing, fails to inhibit neuronal firing, triggers action potential firing when neurons are depolarized to subthreshold potentials, and has a limited capacity to hyperpolarize these neurons even at more depolarized potentials. When activated near the resting membrane potential (À83 mV), PSAM 4 -GlyR passes inward current and depolarizes D1-MSNs. When activated at potentials more depolarized than the chloride reversal potential, PSAM 4 -GlyR passes transient or small outward currents, likely due to influx and accumulation of intracellular chloride that overwhelms endogenous mechanisms to restore the chloride gradient (Huguenard and Alger, 1986;Staley et al., 1995;Thompson and Gähwiler, 1989;Karlsson et al., 2011;Raimondo et al., 2012;Alfonsa et al., 2015). Bath application or systemic injection of high potency exogenous agonists of chloride conductances (e.g., uPSEM 792 ) is a very different phenomenon than physiological activation of native chloride conductances (e.g., GABA A and glycine receptors) by endogenous agonists. The release and clearance of endogenous neurotransmitters are usually tightly controlled and occur over the scale of milliseconds, allowing the neurons to better maintain a stable intra-and extracellular chloride gradient. This is unlike exogenous agonists that result in sustained activation over seconds or minutes and cause massive chloride influx and a shift in chloride reversal potential. In addition, the expression of endogenous chloride conductances is also regulated, unlike viral overexpression of exogenous conductances (e.g., PSAM 4 -GlyR) which only augments the magnitude of chloride influx, hence the unpredictable nature of neuronal responses to the activation of exogenous chloride conductances with exogenous agonists.
More commonly, chloride conductances inhibit action potential firing through electrical shunting. Opening channels decrease membrane resistance which reduces (shunts) depolarization in response to inward current. While there was a significant decrease in membrane resistance associated with opening PSAM 4 -GlyR in D1-MSNs (~30% reduction), it was not sufficient to prevent suprathreshold depolarization by current injection. The majority of neurons were equally capable of firing action potentials to current injection. Despite using high viral titer in our experiments, the decrease in membrane resistance we observed was lower than previously reported in cortical layer 2/3 neurons where PSAM 4 -GlyR activation suppressed firing (Magnus et al., 2019). While increasing PSAM 4 -GlyR conductance by further increasing expression levels could theoretically improve shunting efficacy, the data in this study show a strong correlation between PSAM 4 -GlyR conductance and the rightward shift in PSAM 4 -GlyR reversal potential, suggesting that increasing expression levels will only promote further depolarization with PSAM 4 -GlyR activation. The lack of shunting efficacy in D1-MSNs is likely due to their low membrane resistance (~78 MW in brain slices as measured above) relative to other neurons in the brain (e.g., pyramidal neurons in cortex and hippocampus). It is important to note that membrane resistance is inversely proportional to temperature, age, and number/ activity of synaptic inputs (Kroon et al., 2019;Waters and Helmchen, 2006;Fernandez et al., 2018), and therefore, the shunting efficacy of PSAM 4 -GlyRs during behavioral experiments in vivo is likely to be lower than in brain slices.
In the awake behaving state in vivo, MSNs show brief and irregular spontaneous membrane potential fluctuations between depolarizations (À50 to À60 mV) that facilitate action potential firing and hyperpolarizations (À80 to À90 mV), with the overall membrane potential following a Gaussian distribution (mean~À69 mV) (Mahon et al., 2006). The cell-attached data show that activation of PSAM 4 -GlyR triggers action potential firing when the neurons are depolarized to subthreshold potentials by increasing extracellular potassium. A caveat of these data is that increasing extracellular potassium may indirectly impair chloride extrusion via the potassium-chloride cotransporter, KCC2, which may exacerbate instability of the chloride gradient and make the cells more excitable. Such KCC2 perturbance is expected to affect both control and PSAM 4 -GlyR cells. However, the increased excitability observed in our experiments was limited to the PSAM 4 -GlyR cells supporting the conclusion of PSAM 4 -GlyR driven increase in excitability. The current-clamp data showed that activation of PSAM 4 -GlyR depolarized D1-MSNs by~20 mV and was sufficient to produce depolarization block in a proportion of neurons. This implies that in vivo, PSAM 4 -GlyR activation will effectively increase the probability and duration of spontaneous depolarizations, and increase the probability of D1-MSN firing especially during these depolarizations. The results also show that PSAM 4 -GlyR activation results in 'apparent cross-desensitization' of GABA A synaptic currents, producing a rightward shift in the GABA A reversal potential and a profound reduction of inhibitory GABA A synaptic currents onto D1-MSNs, likely attributed to electrical shunting since the reduction was observed at all potentials. Given the critical role of local GABAergic inhibition in the striatum in regulating neuronal activity (Burke et al., 2017;Koó s and Tepper, 1999), PSAM 4 -GlyR triggered reduction of GABAergic inhibition will further increase the likelihood of D1-MSNs activation rather than inhibition. Thus, increased c-fos expression in D1-MSNs following in vivo activation of PSAM 4 -GlyR is likely caused by PSAM 4 -GlyR-induced depolarization and firing in addition to loss of GABAergic inhibition, and reflects increased neuronal activity of D1-MSNs in vivo.
Finally, PSAM 4 -GlyR-induced depolarization may allow robust calcium influx via voltage-gated calcium channels and NMDA receptors relieved from voltage-dependent magnesium pore-block (Mayer and Westbrook, 1987;Mayer et al., 1984). Calcium influx activates calcium/calmodulindependent protein kinases (CaMKs), calcium response elements in genes, and various other signaling cascades (Pasek et al., 2015;Clapham, 2007), which are known to influence synaptic plasticity and behavior (Lisman et al., 2002;Wayman et al., 2008). Therefore, it is possible that when used to silence neurons, PSAM 4 -GlyR activation could confound the interpretation of experimental results due to depolarization-induced calcium influx, independent of PSAM 4 -GlyR's effect on action potential firing like in the case of depolarization-block.
In summary, activation of PSAM 4 -GlyR expressed in D1-MSNs in the ventral striatum enhanced neuronal activity through direct depolarization and did not suppress action potential firing via membrane shunting. The results of our study show that the PSAM 4 -GlyR approach to silence neurons may not be suitable for all cell types and highlight the need to validate the inhibition of neuronal firing by PSAM 4 -GlyR in the cell type of interest prior to behavioral studies. More broadly, these data demonstrate that achieving neuronal silencing with chloride conductances continues to be challenging and may result in unexpected neuronal activation.

Subjects
All experimental procedures were conducted in accordance with the guidelines of the National Institutes of Health Guide for the Care and use of Laboratory Animals, and approved by the Animal Care and Use Committee of the National Institute on Drug Abuse. We used adult female and male prodynorphin-Cre mice (pdyn-Cre; 8-12 weeks old, breeding facility at the Intramural Research Program, National Institute on Drug Abuse) for electrophysiology and c-fos experiments. We crossed pdyn-Cre and Ai9 Rosa-tdTomato mice to validate selectivity of viral transduction ( Figure 1A). Mice were group housed four per cage and maintained under a 12 hr light cycle at 21 ± 2˚C. Food and water were freely available.

Brain slice preparation and electrophysiological recordings
After >4 weeks of viral incubation, mice were anesthetized with Euthasol (i.p., Virbac AH, Inc, TX) and then decapitated. Brains were rapidly removed and placed in room temperature (25˚C) modified Krebs' buffer containing (in mM): 125 NaCl, 4.5 KCl, 1.0 MgCl 2 , 1.2 CaCl 2 , 1.25 NaH 2 PO 4 , 11 D-glucose, and 23.8 NaHCO 3 , bubbled with 95/5% O 2 /CO 2 with 5 mM MK-801 to increase slice viability. Using a vibrating microtome (Leica Biosystems, IL). Coronal brain slices (220 mm) containing the ventral striatum were collected and incubated at 32˚C for 4 min, then transferred to room temperature until use. All recordings were from neurons in the ventro-medial shell of the nucleus accumbens.
Assuming permeability to Cl À only, the calculated reversal potential of PSAM 4 -GlyR was À62.3 mV. Assuming that PSAM 4 -GlyR has similar Cl À /HCO 3 À permeability to GlyR (P HCO3 À:P Cl À=0.14) (Jun et al., 2016), the calculated reversal potential of PSAM 4 -GlyR in our recording conditions was À60.7 mV ([Cl À ] out 133.9 mM, [Cl À ] in 12.8 mM, [HCO 3 À ] out 23.8 mM, and assuming [HCO 3 À ] in 8 mM). Series resistance was monitored throughout the recordings and not compensated. Reported voltages were corrected for a liquid junction potential of À8 mV between the internal and external solutions. To measure neuronal excitability, current was injected in 50 pA increments (2 s). In determining the current-voltage (I-V) relationship using voltage steps, the external solution of some recordings included 1 mM tetrodotoxin (TTX) to eliminate sodium channel-dependent spiking. There were no differences in the shape of the I-V with or without TTX, so data were combined. Reversal potentials were determined by linear regression considering each replicate an individual point. Recordings in which current did not cross 0 pA were omitted from analysis (n = 2). GABA A receptormediated synaptic currents were evoked with electrical stimulation delivered by a bipolar stimulating electrode placed in the brain slice, in the presence of ionotropic glutamate receptor antagonists (NBQX/DNQX) to isolate fast GABAergic synaptic transmission. Peak amplitude of GABA A receptormediated synaptic currents was measured 3-4 ms from the apparent peak to remove any potential contribution from a stimulation-induced 'escaped' action potential.

Immunohistochemistry and confocal microscopy
For immunohistochemistry, confocal microscopy, and image analysis, the experimenter was blinded to the animal treatment group. Mice microinjected with PSAM 4 -GlyR vs. control AAV into the medial ventral striatum received an i.p. injection of uPSEM 792 (3 mg/kg) or saline followed 30 min later by an i.p. injection of fentanyl (0.2 mg/kg) or saline ( Figure 1B). After 90 min, at the expected peak of c-fos protein expression (Barros et al., 2015), mice were euthanized with Euthasol (i.p.) and transcardially perfused with 1Â PBS followed by ice-cold 4% paraformaldehyde (pH 7.4, Sigma Aldrich, MO). The brains were collected and fixed overnight in 4% paraformaldehyde at 4˚C, and then moved to PBS. Coronal brains slices (50 mm) containing the viral injection sites were collected using a vibratome (Leica Biosystems, IL). Free-floating brain slices were washed in PBS (Â3), permeabilized and blocked in a solution containing PBS, 0.3% Triton-X, and 5% normal donkey serum for 2 hr. Slices were then incubated at 4˚overnight in a solution containing PBS, 0.03% Triton-X, 5% normal donkey serum, and 1:4000 rabbit anti-c-fos primary antibody (Cell Signaling, MA). This was followed by a wash in PBS (Â3) and incubation in 1:500 donkey anti-rabbit secondary antibody, conjugated to Alexa-Fluor-647 (Jackson ImmunoResearch, PA) for 1 hr and 45 min. Slices were then washed in PBS (Â3) and mounted with DAPI Fluoromount mounting medium (Thermo Fisher Scientific, MA). Confocal images were collected using an Olympus Fluoview FV1000 confocal microscope with 20Â (0.75 NA) or 40Â (0.95 NA) objective lens and processed using ImageJ. Z-stack images of the viral injection site in the right hemisphere (two to three brain sections) were collected. Expression of c-fos and eGFP/eYFP was manually quantified by counting fluorescent cells within the image frame. The number of non-transduced cells was obtained by subtracting the number of transduced cells from the number of cells stained with DAPI. Results from all brain sections were averaged for each mouse.