SK Channel Modulates Synaptic Plasticity by Tuning CaMKIIα/β Dynamics

N-Methyl-D-Aspartate Receptor 1 (NMDAR)-linked Ca++ current represents a significant percentage of post-synaptic transient that modulates synaptic strength and is pertinent to dendritic spine plasticity. In the hippocampus, Ca++ transient produced by glutamatergic ionotropic neurotransmission facilitates Ca++-Calmodulin-dependent kinase 2 (CaMKII) Thr286 phosphorylation and promote long-term potentiation (LTP) expression. At CA1 post-synaptic densities, Ca++ transients equally activate small conductance (SK2) channel which regulates excitability by suppressing Ca++ movement. Here, we demonstrate that upstream attenuation of GluN1 function in the hippocampus led to a decrease in Thr286 CaMKIIα phosphorylation, and increased SK2 expression. Consistent with the loss of GluN1 function, potentiation of SK channel in wild type hippocampus reduced CaMKIIα expression and abrogate synaptic localization of T286 pCaMKIIα. Our results demonstrate that positive modulation of SK channel at hippocampal synapses likely refine GluN1-linked plasticity by tuning dendritic localization of CaMKIIα.

Activation of NMDAR, and accompanying calcium release, are not isolated events in LTP. Modulation of synaptic potentiation events that underlie LTP involves the movement of monovalent K + ions through the SK2, and constitutes an ion channel gating mechanism (Xia et al., 1998;Stackman et al., 2002;Ngo-Anh et al., 2005;Hammond et al., 2006;Lin et al., 2008). Together, LTP embodies the movement of ions in an organized process of oscillation between Ca ++ and K + currents. The efflux of K + ions through SK2 decreases synaptic strength by attenuating neuronal firing while creating a state of refractoriness. Positive modulation of SK2 has also been shown to alter frequency-dependent LTP and neural encoding in the hippocampus (Maylie et al., 2004;Lin et al., 2008). Previous studies suggest that SK2 modulate LTP by producing afterhyperpolarization current (I AHP ) that determines the interspike interval for evoked potentials (Hammond et al., 2006;Lin et al., 2008;Maingret et al., 2008). Additionally, by suppressing transient Ca ++ current associated with NMDAR, SK2 refine synaptic strength through an activity-linked feedback loop (Ngo-Anh et al., 2005;Kuiper et al., 2012;García-Negredo et al., 2014;. Elsewhere, we reported that the pharmacologically-induced NMDAR hypofunction suppressed CaMKII expression, and increased SK2 expression in the hippocampus of WT mice (Ogundele and Lee, 2018). This outcome suggests that T286 phosphorylation of CaMKIIα and expression of SK2 may be inversely related in NMDAR hypofunction. Here, we tested the hypothesis that SK2 regulation of synaptic excitability may impact CaMKII synaptic localization. We assessed the structural and physiological properties of hippocampal synapses after genetic ablation of GluN1 sub-unit of NMDAR. In addition to dendritic spine perturbations, an upregulated SK2 expression was accompanied by a loss T286/T287 phosphorylation of CaMKII in the hippocampus. Likewise, positive modulation of SK channel in a wild type (WT) hippocampus (normal GluN1) dysregulates T286/T287 phosphorylation, and synaptic localization of CaMKII. Together, our results suggest that SK2 regulate hippocampal excitability-in part-by modulating dendritic localization of CaMKII.

Animals
Adult male and female animals were acquired from the Jackson's Lab (Bar Harbor, ME, USA). The animals were bred to generate F 1 offspring which were later inbred for two generations. All experimental animals were housed under standard laboratory conditions of 12 h alternating light and dark cycle. Food and water were provided ad libitum. All animal handling procedures were approved by the Institutional Animal Care and Use Committee of the Louisiana State University School of Veterinary Medicine. The animals used for this study weighed ∼24 gm.

Adeno-Associated Virus (AAV) Injection
Mice were anesthetized by intraperitoneal Ketamine/Xylazine (100 mg/Kg:10 mg/Kg) injection and assessed for pain sensation by toe pinching. The head of the mouse was gently fixed in a multi-rail stereotaxic apparatus (Köpf Instruments, Tujunga, CA, USA) for a craniotomy procedure. The Adeno-associated virus (AAV) cocktail was injected into the anterior hippocampus at coordinates (AP: −1.94 mm, ML: 1.0 mm, DV: 1.5 mm) relative to the Bregma. Approximately 800 nl of AAV cocktail was delivered at the rate of 60 nl per min (3 min interval) using a manual Hamilton's syringe holder (World Precision Instruments Inc., Sarasota, FL, USA). After the last injection, the syringe was allowed to stay in place for 15 min before it was gradually withdrawn.

Adult Hippocampal SK Channel Positive Modulation (48 h)
An ICV cannula guide was positioned in the CA1 of a GluN1 flx/flx mouse by stereotaxic surgery (AP: −1.94 mm, ML: 1.0 mm), and affixed with a dental cement. The cannula guide was covered with a dummy cannula. Seven days after the implant, the dummy cannula was replaced with an ICV cannula for a single dose (10 µM CyPPA) drug delivery [SK2/3(+)]. The drug solution was delivered at the rate of 10 µL/min using a manual Hamilton's syringe holder (World Precision Instruments Inc., Sarasota, FL, USA).

Specimen Preparation
Deeply anesthetized mice were euthanized in an isoflurane chamber. The animals were transcardially perfused with 10 mM PBS (pH 7.4) and the whole brain was harvested. The brain was rapidly placed in cold artificial cerebrospinal fluid (ACSF) maintained on ice, and saturated with 95% O 2 /5%CO 2 . A clean razor blade was used to cut the brain along the sagittal plane. The hippocampus was microdissected and extracted from the right and left hemispheres.

Immunofluorescence
After transcardial perfusion with PBS, the animal was perfused with 4% phosphate-buffered paraformaldehyde (PB-PFA). The whole brain was removed and fixed overnight in 4% PB-PFA. Subsequently, the whole brain was transferred into freshly prepared 4% PB-PFA containing 30% sucrose for cryopreservation for at least 72 h. Free-floating cryostat sections (40 µm) were obtained and preserved in 48-well plates containing 10 mM PBS at 4 • C. The sections were washed three times (5 min each) in 10 mM PBS (pH 7.4) on a slow orbital shaker (35 rpm). Blocking was performed in 5% Normal Goat serum (Vector Labs #S-1000), prepared in 10 mM PBS + 0.03% Triton-X 100, for 1 h at room temperature. The sections were incubated overnight at 4 • C in Rabbit anti SK2 Antibody (ThermoFisher Scientific #PA5-41071) and Rabbit anti α-actinin Antibody (ThermoFisher #701914). The primary antibodies were diluted in blocking solution (10 mM PBS + 0.03% Triton-X 100 and 5% normal goat serum). After primary antibody incubation, the sections were washed twice in 10 mM PBS (5 min each), and labeled with a secondary antibody-Goat anti-Rabbit Alexa 568 (ThermoFisher Scientific #A-11036)-diluted in the blocking solution. Secondary antibody incubation was done for 1 h at room temperature, with gentle shaking. Immunolabeled sections were washed and mounted on gelatin-coated slides using ProLong TM Diamond Antifade Mountant (ThermoFisher Scientific #P36970).

Microscopy and Quantification
Fluorescence images were acquired in a Nikon-NiU fluorescence upright microscope that is configured for 3D imaging. Z-stacks were obtained across a depth of 15 µm and rendered as 3D (eGFP) or converted into 2D images through the ''extended depth of focus'' option in Nikon Element Advanced Research software. Normalized fluorescence intensity for immunolabeled proteins in the hippocampus was determined in optical slices for serial section images (n = 5 per mouse brain). Cell counting and fluorescence intensity quantification were done using Nikon Element AR and ImageJ software. Average count and intensity were determined per field for n = 12 fields of view.

Spine Morphology
Fluorescence Vibrotome sliced 100 µm thick sections containing the hippocampus were prepared in oxygenated aCSF and rapidly fixed in 4% PFA. DIL Neurotracer paste (ThermoFisher scientific #N22880) was applied to the hippocampus with the tip of a pulled capillary glass tube. After DIL treatment, the sections were incubated in a PBS humidified chamber overnight at room temperature. Subsequently, the sections were washed in PBS and blocked with 5% normal goat serum prepared in PBST. After blocking, sections were incubated overnight in Rabbit anti-Mouse Synaptophysin antibody (ThermoFisher Scientific #MA5-14532) prepared in the blocking serum solution. The morphology of dendritic spines with synaptophysin expression was imaged in an Olympus FluoView10i Confocal laser scanner. Subsequent quantification of dendritic spines was done in ImageJ.

Transmission Electron Microscopy (TEM)
Microdissected hippocampal tissue was fixed in primary EM fixative composed of 1.6% paraformaldehyde, 2.5% glutaraldehyde, and 0.03% CaCl 2 in 0.05M cacodylate buffer (pH 7.4). The hippocampus was trimmed into ∼1 mm sections using a sharp razor and transferred into a fresh fixative for 2 h at room temperature. In subsequent steps, the samples were washed in 0.1 M cacodylate buffer supplemented with 5% sucrose and fixed in 2% osmium tetroxide for 1 h at RT. The sections were washed in water, then in-block stained with 2% uranyl acetate prepared in 0.2 M sodium acetate buffer (pH 3.5), for 2 h. Stained sections were dehydrated in ascending grades of alcohol and propylene oxide. The processed sections were embedded in Epon-Araldite mixture and polymerized for 24 h at 60 • C. Tissue blocks were sectioned using a Leica Ultratome (Leica EM UC7). Thin (80 nm) sections were recovered and stained with lead citrate for 5 min. Transmission electron photomicrographs were obtained in a JEOL 1400 Transmission Electron Microscopy (TEM) microscope, equipped with a GATAN digital camera. All reagents for electron microscopy were from EMS (Hatfield, PA, USA). A typical synapse was identified by the presence of synaptic vesicles, postsynaptic mitochondria, and post-synaptic density. The length of dendritic spines was determined in TEM photomicrographs. As such, we estimated the distance between the post-synaptic density, and the stalk of the dendritic spine.

Acute Extracellular Recording
In vivo recording was carried out in the hippocampus of anesthetized mice. Animals were deeply anesthetized with Urethane (0.2 mg/Kg i.p.), then the head was affixed on a stereotaxic frame. For combined recording and photostimulation, CaMKII-Cre::ChR2 mice were anesthetized with ketamine/xylazine. Mice were tested for toe pinch response to ensure the absence of sensation before the commencement of the procedure. A craniotomy was done in order to expose the dura. Drops of ACSF was applied to this area to prevent dryness. Under a digital dissection microscope, the dura over the exposed brain area was carefully excised using a bent needle tip. An acute neural probe, with a 10 mm long and 50 µm thick shank was used for this procedure (Neuronexus, Ann Arbor, MI, USA). The probe shank carried four electrodes arranged as a tetrode, with an inter-electrode distance of 25 µm. The electrodes were connected to a pre-amplifier head stage (Intantech, Los Angeles, CA, USA), tethered to a recording controller and amplifier system (Intantech, Los Angeles, CA, USA). The electrode was gently lowered into the brain tissue using an ultrafine (µm range) hydraulic micromanipulator (Narishige, Japan) to reach the CA1 dendritic field (radiatum/molecular layer) at stereotaxic coordinates (AP: 1.94 mm, ML: 1.0 mm, DV: 1.5 mm) relative to the Bregma. Stainless steel ground wires soldered onto the head stage-electrode adapter (Neuronexus; A4 to Omnetics CM32 adapter) were tied to a ground screw that was fixed in the occipital bone. For recording procedure involving optogenetics, the recording electrode shank, optic fiber and Hamilton's syringe were stereotactically positioned in the CA1. TTL (Prizmatix, Southfield, MI, USA) driven 470 nm LED source (Thorlabs, Newton, NJ, USA) was triggered to generate 50 ms pulses over a 789 s duration (at 1 Hz). In three separate groups, we applied: (1) photostimulation (470 nm); (2) photostimulation with 100 nM Apamin (SK2 blocker; Kasumu et al., 2012) infusion; and (3) photostimulation combined with 10 µM CyPPA (SK2 positive modulation) infusion.
The stereotaxic apparatus, micromanipulator, electrode and subject mouse were kept in a Faraday cage and connected to the amplifier ground. At the onset of each recording procedure, the impedance of the electrodes was determined at 1 KHz. For all recording sessions in this study, impedance measurement for the silicon tetrodes ranged between 0.6 and 3.1 . Single unit activity was recorded by setting the cut-off frequency as 250 Hz and 7.5 KHz, respectively for lower and upper limits, sampled at 20 KHz/s. Neural activity was monitored for 20 min to ensure the stability of the animal's vitals before the commencement of recordings. Continuously recorded spike trains from the CA1 dendritic field was processed in an Offline Spike Sorting software (OFSS; Version 4; Plexon Inc., Dallas, TX, USA). Further analysis of the sorted spikes was done in Neuroexplorer Version 5 (Nex Technologies, Colorado Springs, CO, USA).

Neural Spike Processing and Analysis
Neural spikes were extracted from the continuous data through threshold crossing in the OFSS. The extracted spikes were sorted into clusters using a combination of Valley seeking and K-means clustering methods. Spikes were assigned to single unit clusters through a 3-dimensional space principal component analysis (PCA) projection. Where necessary, unsorted spikes were assigned to clustered units, or invalidated if outlying. Sorted neural spike waveforms, clustered units, and up-sampled continuous data were exported into the Neuroexplorer software for analysis of spike properties.

Social Interaction Test
Mice were habituated in the testing area for 24 h before the commencement of the behavioral test (Kaidanovich-Beilin et al., 2011). At each phase of the test, the compartment was wiped with 70% isopropyl alcohol to prevent odor-specific cues and bias in the subsequent steps. Two smaller holding compartments in the testing area were designated ''E (Empty)'' during the habituation trial (5 min). For the sociability test, a stranger mouse (S1 0 ; stranger 1) was introduced into one of the compartments while the subject animal was re-introduced into the chamber (5 min). After an inter-trial time of 30 min, we introduced a second stranger mouse (S2), along with the first stranger (now S1 n ) into two separate holdings (social novelty test). The contact with (E or S1 0 ), and (S1 n , S2) were estimated to determine sociability [S1 0 /(E+S1 0 )] and social novelty [S2/(S2+S1 n )] indices respectively. While ''0'' represents the first encounter with S1 during sociability test, ''n'' represents the second encounter with S1 in a novel position during the social novelty test. Therefore, different time measurements were done at S1 0 and S1 n . The time spent in contact with E, S2, S1 0, and S1 n in the sociability and social novelty tests were estimated blindly by an independent investigator using a software-Ethovision (Noldus, Leesburg, VA, USA).

Dendritic Spine Morphology and Synaptic Structure
The general outline of dendritic spines was demonstrated using a combination of DIL Neurotracer, and synaptophysin immunofluorescence labeling (Figures  2A,B). Low magnification images demonstrate a significant increase in the distribution of DIL positive cytoskeletal aggregates in the CA1 of the ∆GluN1 hippocampus when compared with the WT and GluN1 flx/flx . At higher magnification, this translates into a significant loss of dendritic spines in the ∆GluN1 hippocampus ( Figure 2B; p < 0.01). To verify this outcome, dendritic spine morphology was further examined by ultrastructural imaging (TEM) of hippocampal synapses. As illustrated in Figures 2C,D, transmission electron photomicrographs revealed a prominent loss of cytoskeletal filament assembly (yellow arrowheads) in ∆GluN1 hippocampal synapse when compared with WT and GluN1 flx/flx . In addition to the loss of dendritic spines and synaptic cytoskeletal perturbation, TEM images of the ∆GluN1 hippocampus also revealed a prominent decrease in the thickness of post-synaptic densities (Figures 2E,F; p < 0.001).

Morphology of Dendrites
The expression of membrane-bound eGFP conveyed by the AAV vectors was used to assess the morphology of the dendrites. Given that the expression of eGFP in the neuropil outlines dendrite branching, we used 3D-fluorescence imaging to demonstrate and measure the length of ∆GluN1 CA1 dendrites (µm). However, in order to assess the length of dendritic spines, we used ultrastructural synapse images acquired in TEM. Figure 3A illustrates the general outline of CA1 pyramidal neurons where DL and Ds.L represent the dendrite length and dendritic spine length, respectively. ∆GluN1 CA1 pyramidal neurons exhibits a significant decrease in dendrite length when compared with the GluN1 flx/flx (p < 0.001; Figures 3B,C). In assessing dendritic spine length (TEM), the distance between the dendrite shaft and the tip of the spine was determined (Figure 3A; Ds.L). The synapse (sy) in Figure 3D illustrates the tip of the dendritic spine where postsynaptic densities are present. Yellow arrowheads connote the dendritic spine shaft and synaptic expansion. In addition to a decrease in dendrite length (DL), the ∆GluN1 hippocampal neurons were also characterized by a decrease in dendritic spine length (Ds.L: p < 0.001;

Burst Activity in the CA1 Neural Network
The relationship between the observed synaptic perturbations and neural encoding in the ∆GluN1 hippocampus was assessed by in vivo extracellular neural recording. Spontaneously evoked neural spikes in the CA1 dendritic field was analyzed to determine the impact of GluN1 loss of function on multisynaptic burst encoding in vivo. Continuously recorded spike trains for n = 4 mice per group ( Figure 4A) were sorted offline to isolate single units (GluN1 flx/flx : n = 77 and ∆GluN1: n = 54) across six sessions. Spike train properties were assessed using the interspike interval (ISI) parameters which represent neural refractoriness over a time window. The firing property of each cell was determined with the ISI histogram (ISIH). Based on the ISIH properties, we grouped the neuron units as bursty, tonic, or irregular firing ( Figure 4B). For the GluN1 flx/flx hippocampus, about 48% of the neurons exhibited distinct burst firing patterns while 38% are irregular firing neurons (p < 0.01). Conversely, in the ∆GluN1 hippocampus, 20% of the neurons showed characteristic burst firing pattern while ∼60% were irregular firing neurons (p < 0.01; Figures 4B,C). Consistent with the suppression of burst firing, ∆GluN1 CA1 neurons exhibit a prolonged ISI peak duration (p < 0.001; Figure 4D), and an increased mean ISI duration (p < 0.05; Figure 4E).
In addition to a decreased burst firing, structural changes in ∆GluN1 hippocampal synapses was accompanied by an increased firing irregularity. Here, we determined the regularity of firing as the coefficient of variation (CV 2 ) of the ISI in a fixed time window. A score of CV 2 < 1 or CV 2 = 1 depicts a regular firing pattern, while CV 2 > 1 represents irregular firing between two fixed time points (τ n to τ n+1 ). Regularity of firing based on the CV 2 illustrates a time window computation of the ISI and is not affected by slight changes in firing frequency (spikes/s). Consistent with the percentage of irregular firing neurons, an average score of CV 2 = 1.2 was recorded for the ∆GluN1 neurons when compared with the GluN1 flx/flx which showed significant firing regularity (CV 2 = 0.5; Figure 4F). A prolonged ISI and increased irregularity of firing in the ∆GluN1 CA1 spike train was also associated with a decrease in burst count per unit time when compared with the GluN1 flx/flx spike train (p < 0.001; Figure 4G).

∆GluN1-Linked Synaptic Perturbations Led to a Reduced Neuronal Firing Rate
In addition to ISIH characterization of spike train properties (bursty, tonic and irregular), we further examined the neuron population dynamics based on the percentage of fast or slow spiking neurons determined by autocorrelogram plots (Figures 5A,B). This was further verified by the peri-event raster plots which demonstrates number of events (ticks) per trial ( Figure 5C) when the activity of all neurons (per session) was compared with a neuron with the highest raster count. By combining the autocorrelogram and peri-event raster plots, the neurons were further grouped based on their firing rate r(t) over a fixed time window (τ ms). Together with ISIH characterization of neuron firing pattern, it was noted that the fast-spiking neurons were also bursty in their firing properties. On the other hand, irregular firing neurons exhibited a skewed autocorrelogram plot that connotes slow or desynchronized firing. As illustrated by the pie charts, the ∆GluN1 spike train exhibited a significant increase in the percentage of slow and irregularly firing neurons when compared with the GluN1 flx/flx spike train (p < 0.001; Figure 5D). Consequently, the mean firing rate r(t) for the ∆GluN1 CA1 network reduced significantly vs. the GluN1 flx/flx (p < 0.01; Figure 5E). In order to ascertain the impact of the highlighted synaptic changes on neural function, ∆GluN1 mice were assessed for sociability and social novelty. Heat maps in Figure 5F illustrates a significant decrease in sociability and social novelty performance index for ∆GluN1 mice when compared with the GluN1 flx/flx . Together with the observed synaptic perturbations the ∆GluN1 mice also exhibited a decline in cognitive function (p < 0.001; Figure 5G).

Cellular Basis of Firing Rate and ISI Dynamics in ∆GluN1 CA1
Synaptic potentiation in the hippocampus is mediated by several factors. However, it has been established that T286 phosphorylation of CaMKIIα and activation of SK2 can respectively enhance and attenuate synaptic activity ( Figure 6A). Here, we showed that an increased ISI duration for the ∆GluN1 spike train was linked with a significant upregulation of hippocampal SK2 expression when compared with the controls (WT and GluN1 flx/flx ). Double labeling immunofluorescence in (Figures 6B,C) demonstrates a significant increase in SK2 expression for eGFP positive neurons in the ∆GluN1 CA1 (p < 0.01). Since SK2 activity determines the ISI, it is logical to speculate that the increased peak and mean of ISI duration (Figures 4D,E) were-in part-as a result of upregulated CA1 SK2 expression. T286 pCaMKIIα synaptic localization and substrate targeting is required for LTP expression that is pertinent to long-term synaptic plasticity. As such, we determined the firing r(t) and burst discharge rate as indirect measures of synaptic T286 pCaMKIIα translocation efficiency in the CA1 neural network in vivo. In the ∆GluN1 hippocampus, the expression of CaMKIIα did not change significantly when compared with the control level (WT and GluN1 flx/flx ; Figures 6D-F). However, reduced firing ( Figure 5E) and burst rates (Figure 4G) in the ∆GluN1 CA1 dendritic network was accompanied by a prominent decrease in T286 pCaMKIIα and T287 pCaMKIIβ (p < 0.01; Figures 6D-G; also Supplementary Figure S3). Together, ∆GluN1-induced dendritic spine perturbations and loss of neural plasticity involved the loss CaMKII activity and increased SK2 expression ( Figure 6H).

Activity-Coupled Regulation of SK2 and GluN1 Expression
While assessing synaptic function in the ∆GluN1 hippocampus, a notable change in spike train properties was an increase in the mean ISI duration (Figure 4E), and a reduced firing rate ( Figure 5E). A possibility is that ∆GluN1 caused a decrease in transient Ca ++ current that is necessary for SK2 activation and may cause SK2 overexpression through an activity coupled feedback mechanism. To test this hypothesis, we pharmacologically activated SK2 and determined the expression of GluN1 and SK2 by western blotting. Previously, we demonstrated that ablation of GluN1 (upstream) led to an increase in SK2 expression ( Figure 6B). Here, we noted that positive modulation of SK2 caused a moderate increase in GluN1 expression (p < 0.05; Figures 7A,B), and significantly downregulate hippocampal SK2 expression (p < 0.01; Figures 7C,D; Supplementary Figure S4). These outcomes suggest that the expression of GluN1 and SK2 is inversely linked to their activity. Consequently, in the ∆GluN1 hippocampus, reduced Ca ++ current required for the activation of SK2 may drive an increase in SK2 expression to compensate for the initial reduction in its activation. Likewise, repetitive activation of SK2-downstream of NMDAR-attenuates synaptic potentiation, and lead to an increase in GluN1 expression through a similar feedback mechanism. Figure 7E illustrates the site for stereotaxic implant of the cannula guide and drug delivery through the cannula.

SK2-NMDAR Cross-Talk in the Regulation of CaMKIIα Activity
Given that the expression and activity of GluN1 and SK2 are tightly linked, an important question yet to be addressed is whether downstream positive modulation of SK channel can impact CaMKII activity and if such interaction might be dependent on NMDAR. In the ∆GluN1 hippocampus, T286/T287 phosphorylation of CaMKII decreased significantly. Although GluN1 expression was upregulated after a positive modulation of SK channel, the expression and T286 phosphorylation of CaMKIIα was significantly suppressed in the SK2/3(+) hippocampus (Figures 8A-C; p < 0.001; Supplementary Figure S5). Likewise, SK channel potentiation caused a significant decrease in T287 pCaMKIIβ ( Figure 8D; p < 0.001). Ultimately, SK channel potentiation led to a decrease in percentage T286/T287 phosphorylated CaMKII in the hippocampus (Figure 8E; p < 0.001). These results are indicative of the role of SK2 in the tuning synaptic plasticity by CaMKII modulation.

Compartmental Localization of CaMKII in the Hippocampus
We processed microdissected whole hippocampus to isolate synaptosomal extracts containing post-synaptic densities (PSD), and cytosolic extracts containing dendritic shaft. We validated the tissue extracts by western blotting to detect PSD proteins-PSD-95 and GluN1-in the synaptosomal extract. Likewise, the shaft-cytosol extract was validated by immunoblot detection of ErK1/2 and pErK1/2. The validation protocol was set up as a reciprocal detection experiment where synaptic proteins are also assessed in shaft-cytosol extracts, and vice versa. For all samples, PSD-95 (S2.1,S2.2) and GluN1 (S2.3,S2.4) were enriched in the synaptosomal extract, and not in the shaft-cytosol domain. Similarly, ErK1/2 (S2.5,S2.6) and pErK1/2 (S2.7,S2.8) were enriched in the shaft-cytosol domain only. In order to determine compartmental localization of CaMKIIα, we performed immunoblot detection of CaMKIIα, and T286α/T287β phosphorylated isoforms in the synaptosome and shaft-cytosol extracts. Our results revealed that CaMKIIα is predominantly localized in the synaptic compartment when compared with the shaft-cytosol (Figures 10A,B; p < 0.001). Likewise, T286 pCaMKIIα and T287 pCaMKIIβ were also enriched in the synaptic compartment when compared with the shaft-cytosol extract (Figures 10C,D; p < 0.001).
In subsequent experiments, we incubated the microdissected hippocampus in glucose-rich oxygenated Ca ++ -ACSF to determine how an ex vivo treatment might affect the compartmental localization of CaMKII (Figures 10A-D). After 1 h of incubation in ACSF, CaMKIIα and T286α/T287β were predominantly localized in the shaft-cytosol domain ( Figure 10E). However, when 10 µM L-Glutamate is included the ACSF, CaMKIIα was predominantly localized in the synaptic compartment, and was accompanied by a near total T286α/T287β synaptic translocation (Figures 10E,F; p < 0.001). These results suggest that L-Glutamate-driven synaptic potentiation is required for spine recruitment of CaMKIIα and T286α/T287β (Figure 10G). Figure 8 demonstrates a significant loss of neural CaMKII, T286α and T287β after SK channel potentiation in vivo (48 h). In order to determine the short-term (1 h) effect of SK channel on synaptic localization of CaMKII, we paired L-Glutamate treatment with SK channel potentiation in acute preparations ( Figure 11A). L-Glutamate treatment significantly increased synaptic localization of CaMKIIα (p = 0.0031; Figure 11B). Normalized synaptic expression of CaMKIIα also increased when L-Glutamate was paired with CaMKII inhibitor (A2RIP; p = 0.0024) or SK channel potentiator (p = 0.0010; Figure 11B). Similar to L-Glutamate treatment, a combination of L-Glutamate with A2RIP or CyPPA did not impact the shaftcytosol localization of CaMKIIα ( Figure 11C). L-Glutamate treatment increased synaptic localization of CaMKIIβ when compared with ACSF only (p = 0.0216; Figure 11D). However, when combined with CyPPA, the synaptic translocation of CaMKIIβ was significantly upregulated (p = 0.0015). L-Glutamate treatment caused a significant decrease in shaftcytosol localization of CaMKIIβ when compared with ACSF (p = 0.0453). This was attenuated by SK channel potentiation and CaMKII inhibition. For these treatment combinations, there was no significance vs. ACSF (baseline; A2RIP and CyPPA; Figure 11E).

SK Channel Potentiation Reduced the Efficiency of CaMKII Translocation
In order to ascertain the effect of SK channel potentiation (CyPPA) on CaMKII translocation efficiency, we normalized synaptosomal pCaMKII (T286, T287, T305/306) with shaftcytosolic expression level. While L-Glutamate significantly increased synaptosomal translocation of T305/306 pCaMKII (p = 0.0158), pairing L-Glutamate with either CaMKII inhibition or SK channel potentiation suppressed synaptic translocation of T305/306 ( Figure 11F). Accordingly, when L-Glutamate is paired with A2RIP or CyPPA, there is no significant change in synaptic T305/306 expression vs. ACSF ( Figure 11F). L-Glutamate treatment increased the synaptic localization of T286 pCaMKIIα when compared with ACSF (p = 0.0016). Similar to T305/306, pairing L-Glutamate with SK channel potentiation reduced synaptic T286 pCaMKIIα localization ( Figure 11G). Here, the significance recorded for L-Glutamate-vs. ACSF only-was abolished when L-Glutamate is paired with CyPPA (ns). L-Glutamate treatment also increased the synaptic localization of T287 pCaMKIIβ when compared with ACSF only (p = 0.0170; Figure 11H). As shown in Figure 11D, CyPPA treatment induced a more prominent increase in synaptic CaMKIIβ expression (p = 0.0015) when compared with L-Glutamate (p = 0.0216); vs. ACSF. To this effect, pairing L-Glutamate with CyPPA also caused a prominent increase in synaptic T287 pCaMKIIβ (p < 0.0001) when compared with L-Glutamate (p = 0.0170); vs. ACSF ( Figure 11H). Together, our results suggest that SK channel potentiation suppressed synaptic translocation of T286 pCaMKIIα while increasing T287β pCaMKII phosphorylation in acute experiments.

SK Channel Potentiation Modulates CA1 Burst Encoding
The physiological implication of a suppressed T286/T287 phosphorylation and imbalanced translocation to the synapse is that LTP may be abrogated. In addition, long-term synaptic plasticity modulated by CaMKII may become significantly dysregulated. In a separate experiment, we examined the effect of SK channel positive modulation (10 µM CyPPA) or inhibition (100 nM Apamin) on CA1 dendritic neural network burst encoding in vivo. In anesthetized mice, SK channel agonist (CyPPA) was infused, and spontaneously evoked spikes were recorded with a tetrode. Subsequently, Apamin (SK2 blocker) was infused to attenuate the activity of SK2 in the CA1 dendritic field. As illustrated in Figure 12A, positive modulation of SK channel function significantly reduced spontaneous CA1 firing while Apamin treatment rescued CA1 firing and burst activity (Figures 12A,B). Based on ISI characterization of firing pattern (Figures 4B,C), there was a significant increase in the percentage of irregular firing neurons and a decrease in bursty neurons for SK2/3(+) CA1 spike train (p < 0.001; Figure 12C). Subsequent Apamin treatment increased the percentage of cells with characteristic bursty FIGURE 9 | SK channel potentiation impacts hippocampal α-actinin expression. (A-B) Fluorescence images illustrating α-actinin expression in the hippocampus (so, stratum oriens; pyr, pyramidal layer; sr, stratum radiatum; mol, molecular layer; gr, granule cell layer, and hil, hilus). Scale bar = 0.4 mm (A), 100 µm (B). (C) Bar graph representation of α-actinin fluorescence intensity in the hippocampus (One-Way ANOVA). Schematic illustration of SK2 modulation of CaMKII spine anchorage: (D) synaptic pCaMKII is structurally linked with to the GluN2B sub-unit of NMDAR and is anchored to the dendritic spine cytoskeleton (F-actin) by α-actinin. In NMDAR-mediated synaptic potentiation, synaptic accumulation of pCaMKIIα/β is facilitated by α-actinin. (E) SK channel potentiation induced a loss CaMKII and α-actinin expression (C; * * p < 0.01).
firing pattern, while also reducing the percentage of irregular firing neurons (Figure 12C; p < 0.001). Further evaluation of the ISI revealed that SK2 blockade by Apamin reduced the mean ISI duration (p < 0.01; Figure 12D) and increased the regularity of firing (Figure 12E) of CA1 neurons. This was seen as a decrease in CV 2 score for the ISI (p < 0.01) following Apamin infusion.

SK Channel Modulation of Firing Rate
Here, we measured CA1 network firing rate (Hz) when pyramidal cell dendritic field photostimulation is paired with pharmacological modulation of SK2 in vivo. Figures 13A,B illustrates the expression of ChR2 in the hippocampus and dendritic spine. Dotted lines represent the electrode/optic fiber track in the hippocampus (Figure 13B). The experimental set up for placement of the recording electrode, optic fiber, and needle for drug infusion is schematically illustrated in Figure 13C. Blue light pulse (470 nm) was delivered by a square wave (TTL) generator at 1 Hz (50 ms) to potentiate the spines. The effect of the photostimulation regime is demonstrated by the 470 nm light ON phase (Figures 13D-F). When the ISI histogram was used for the characterization of the firing pattern, ChR2 photostimulation increased burst firing as indicated by the percentage of bursty neurons ( Figure 13G; p < 0.001). Interestingly, when SK2 inhibition (Apamin) is paired with photostimulation, the threshold of bursting activity increased drastically with ∼100% of neurons showing burst activity. However, when photostimulation is paired with SK channel potentiation, a significant decrease in burst activity was recorded (p < 0.001). This is further evident in perievent histogram plots that depict response strength and synchrony of CA1 synaptic units. A paired photostimulation and SK2 inhibition regime increased the firing rate and synchrony when compared with photostimulation only. Conversely, SK channel activation by CyPPA scrambled the synchrony and response strength of the CA1 units (Figures 13H-K).
The significance of this outcome is further evident in the analysis of the ISI histogram plot. For paired photostimulation and SK channel inhibition, the ISI histogram shows strong burst activity that is characterized by a decreased ISI, and rapid onset for ISI decay (red arrow head; Figures 14A,B). On the other hand, pairing SK channel potentiation with photostimulation prolonged the ISI duration, and delayed the onset of ISI decay (>400 ms; Figure 14C). This outcome is further supported by analysis of autocorrelogram which revealed a prominent increase in firing strength as a result photostimulation ( Figure 14D; red arrow heads); compared with spontaneously firing bursty neuron (inset). Pairing photostimulation and SK2 inhibition further enhanced burst strength of CA1 network as illustrated by a further depression of the edge of autocorrelogram plot ( Figure 14E). However, when photostimulation is coupled with SK channel potentiation, synaptic response strength and synchrony decreased significantly as shown by a decrease in autocorrelogram peak strength ( Figure 14F). This is consistent with an increased ISI duration (ms; Figure 14C), and reduced synchrony of firing shown by the perievent raster plot (Figure 13K). Suppression of SK2 activity by Apamin during photostimulation increased the firing rate and the number of spikes in a burst when compared with photostimulation only (Figures 14G,H, p < 0.0001). As expected, pairing SK channel potentiation with photostimulation abolished the firing rate increase recorded in photostimulation only (Figures 14G,H). Interestingly, our results revealed that the mean burst duration was not altered irrespective of the paired SK2 modulation event. Based on these outcomes, we deduced that the dendritic SK channel modulates burst encoding by directing the firing rate and number of spikes in a burst. While these parameters are linked to the probability of burst firing, we noted that SK channel modulation did not alter the duration of burst events in the CA1 neural network.

DISCUSSION
NMDAR is a glutamate type ionotropic receptor that is relatively abundant in the hippocampus and mediates LTP (Hatton and Paoletti, 2005;Coultrap and Bayer, 2012;Lisman et al., 2012;Babiec et al., 2017). A change in the expression of NMDAR or loss of function of its sub-units (GluN1 and GluN2A and 2B) have been identified in the etiology and progression of neuropsychiatric disorders (Hansen et al., 2014;Huang and Gibb, 2014;Wesseling et al., 2014;Bustos et al., 2017). Neurocognitive disorders such as autism, schizophrenia, and depression exhibit various forms of NMDAR hypofunction (Duffney et al., 2013;Snyder and Gao, 2013;Cohen et al., 2015;Zhou et al., 2016;Gulchina et al., 2017;Nakazawa et al., 2017;Ogundele and Lee, 2018;Rebollo et al., 2018). Although the causes of developmental neuropsychiatric defects are wide and varied, recent evidence suggests that loss of NMDAR function is a determinant of the synaptic and behavioral defects.
In LTP, synaptic potentiation by NMDAR-linked ionotropic neurotransmission is regulated through various mechanisms. Notably, positive modulation of synaptic strength which is pertinent to memory encoding is facilitated by CaMKIIα T286 phosphorylation (Gustin et al., 2011;Coultrap and Bayer, 2012;Lisman et al., 2012;Coultrap et al., 2014). On the other hand, negative modulation of synaptic strength could be driven by small conductance (SK2) ion channel (Ngo-Anh et al., 2005;Hammond et al., 2006;Maingret et al., 2008). The NMDARmediated Ca ++ current constitutes 75% of post-synaptic transient that is central to synaptic potentiation and LTP. The mechanism involves activation (T286 autophosphorylation) of a closed CaMKII structure (T305/T306) to promotes kinase activity and substrate targeting (Lisman et al., 2002(Lisman et al., , 2012. In the T286 autophosphorylated state, CaMKII holds an increased (10-folds) affinity for Ca ++ -calmodulin binding, and promotes high-frequency firing that leads to LTP expression (Gustin et al., 2011;Coultrap and Bayer, 2012;Hell, 2014;Ma et al., 2015). On the other hand, activation of SK2 by the Ca ++ surge suppresses neuronal firing by spike frequency adaptation that is pertinent to intrinsic excitability regulation (Ngo-Anh et al., 2005;Hammond et al., 2006;Lin et al., 2008;. Mechanistically, the activity of SK2 constitutes the after-hyperpolarization phase, that depicts the interspike interval, during which a synaptic unit is least expected to fire (Stackman et al., 2002(Stackman et al., , 2008Ngo-Anh, 2006;Kim and Hoffman, 2008;Lin et al., 2008;Maingret et al., 2008;Prescott and Sejnowski, 2008;Toporikova and Chacron, 2009;Kuiper et al., 2012;Trimmer, 2015). Based on these concepts, NMDAR hypofunction may also imply-to an extent-the suppression of CaMKIIα T286 phosphorylation and (or) a positive modulation of SK2 function. Consistent with this idea, genetic ablation of NMDAR or pharmacological potentiation of SK2 suppressed T286 phosphorylation of CaMKIIα and reduced the firing rate of CA1 neurons in vivo.
The frequency of NMDAR Ca ++ transients represents a significant aspect of synaptic potentiation that impacts dendritic spine plasticity in the hippocampus. As such, the structure of dendritic spines, distribution of dendritic spines, and CA1 network firing pattern are directly related to LTP expression and structural plasticity (Jauregui et al., 2017;Nakahata and Yasuda, 2018). While an increased CaMKII T286 phosphorylation is characteristic of high-frequency neuronal firing, an increased SK2 function attenuates neuronal firing. Given that NMDAR activation produces the transient Ca ++ current that activates CaMKII and SK2 for synaptic potentiation and depression respectively, a significant aspect of this interaction that is yet to be considered is whether SK2 refine synaptic plasticity by modulating the spine dynamics of CaMKII. Taken together, our results demonstrate that genetic ablation of GluN1 in the hippocampus significantly perturbed dendritic spine morphology, and attenuate CA1 burst firing in vivo. This is attributable-in part-to the activity-coupled upregulation of SK2 expression, and a suppression of T286 CaMKIIα phosphorylation. Interestingly, in the presence of normal GluN1 function, positive modulation of SK channel equally reduced hippocampal synaptic T286 pCaMKIIα localization.

Cytoskeletal Anchorage of CaMKII
Our results demonstrate that SK channel potentiation in vivo caused prominent loss of CaMKII, and suppressed T286/287 phosphorylation after 48 h. However, in acute treatment (1 h) performed with ex vivo slices, we found that SK channel potentiation disrupts T286 pCaMKIIα and T287 pCaMKIIβ synaptic homeostasis. It follows that the CaMKII heteromeric dodecamer (9α/3β) is anchored to dendritic spine cytoskeleton by an F-actin binding protein-αactinin (Shen et al., 1998;Robison et al., 2005;Jalan-Sakrikar et al., 2012;Khan et al., 2016). However, it has been previously established that CaMKIIβ holds a stronger affinity, and is primarily involved in anchoring the holoenzyme α-actinin/Factin (Coultrap and Bayer, 2012). T287β phosphorylation allows the holoenzyme to detach from the cytoskeletal anchor and facilitates substrate targeting (kinase activity) of the T286 pCaMKIIα component. Here, our results demonstrate that SK channel potentiation suppresses synaptic localization of T286 pCaMKIIα while promoting T287β phosphorylation (Figures 11G,H). The physiological implication is that the synaptic substrate activity (T286 pCaMKIIα) and cytoskeletal anchorage (CaMKIIβ) are significantly reduced following SK channel potentiation. The latter is further evident by the prominent loss of α-actinin in the hippocampus after SK channel potentiation (Figure 9).

SK2 in CA1 Burst Encoding
Spontaneously evoked synaptic potentials are necessary for hippocampal neural plasticity and are representative of the state of CA1 neural network (Winnubst et al., 2015). Here, neurons were characterized using the shape of the ISIH which is mostly dependent on SK2 function and represents a distinct firing signature based on repetitive patterns of refractoriness (Maylie et al., 2004;Ngo-Anh et al., 2005;Hammond et al., 2006). Additionally, we assessed burst and firing rate events that are dependent on the threshold of T286 pCaMKIIα synaptic localization (Coultrap and Bayer, 2012;Hell, 2014;Penny and Gold, 2018). Together with the structural perturbations of dendritic spine, our results suggest that loss of GluN1 function may also connote the suppression of CaMKII T286 phosphorylation and upregulation of SK2 function ( Figure 6H). Here, we showed that the suppression of CaMKII function in the ∆GluN1 hippocampus was associated with a reduced burst firing (Figure 4). Likewise, upregulated SK2 expression may underlie an increase in the ISI duration within the CA1 neural network (Figures 4, 12).
Spontaneously evoked neural spikes recorded from the CA1 dendritic field were characterized by mean a firing rate of 30 spikes/s (Figure 5). As a result of GluN1 loss of function (∆GluN1), there was a significant decrease in firing rate. This is further supported by an increased percentage of irregularly firing neurons, and a decrease in the count of bursty neuron units. To this effect, loss of GluN1 function in the hippocampus was associated with an increased irregularity of firing which was determined by the coefficient of variation (CV 2 ) of the ISI. Determining the regularity of firing based on CV 2 is solely dependent on the ISI and is not affected by a slight change in firing frequency. Since SK2 function impacts the ISI, the CV 2 over a time window may be representative of the effect of SK2 on firing regularity. To support this outcome, in normal GluN1 function (Figure 12), potentiation of SK2 reduced the regularity of firing while suppressing the firing rate.
In addition to an increased ISI, positive modulation of SK2 also suppressed burst and firing rates in the CA1 neural network (Figures 12A,B). In a typical synaptic potentiation event, high-frequency firing is dependent on NMDAR Ca ++ transient, and CaMKIIα T286 phosphorylation at postsynaptic densities. Here, we noted that SK channel potentiation reduced the fringing rate-number of spikes per unit time (n/t)-by suppressing synaptic localization of T286 pCaMKIIα (Figures 8, 11). Although GluN1 ablation led to a loss of T286 pCaMKIIα (Figures 6D,E), the associated upregulation of SK2 likely enhanced the suppression of T286 pCaMKIIα synaptic localization.

SK2 Regulation of Synaptic Plasticity Involves CaMKII Modulation
Our results suggest that GluN1 regulation of synaptic plasticity is dependent on a balanced activity of SK2 and CaMKII in hippocampal dendritic spines. Activation (T286 phosphorylation) of CaMKIIα promotes synaptic potentiation, and is a determinant of neuronal firing rate (n/t), and burst activity (Giese et al., 1998;Coultrap and Bayer, 2012;Coultrap et al., 2014). A coupled activation of SK2 by Ca ++ transient determines the pattern of neural refractoriness and duration between successive spikes (ISI). Since SK2 modulate the ISI duration, it is also directly involved in the regulation of burst rates (Stackman et al., 2002(Stackman et al., , 2008Lin et al., 2008;Maingret et al., 2008). As a result, there is a possibility of an overlap in CaMKII activity and SK2 function in LTP. Consistent with the previously described changes in dendritic spine morphology, we noted that an increase in ISI inversely correlated with both regularity of firing ( Figure 4F) and the firing rate in the ∆GluN1 CA1 network ( Figure 5E). Likewise, when SK2 was activated in the WT hippocampus (normal GluN1 function), both the firing rate and regularity of firing reduced significantly (Figure 12); similar to ∆GluN1. In addition to a decreased firing rate, positive modulation of SK channel function led to an abrupt suppression of CaMKIIα and its T286 phosphorylation in the hippocampus (Figure 8). To this effect, a subsequent infusion of SK2 blocker-Apamin-increased neuronal firing rate and percentage bursty neurons when compared with the SK channel potentiation.

SK2 Modulation of Firing Rate and Neural Burst Discharge
Synaptic translocation of T286 pCaMKIIα is dependent on the stimulation frequency. Accordingly, high-frequency stimulation events facilitate a rapid synaptic T286pCaMKIIα recruitment (LTP) compared to low-frequency stimulation (Håvik et al., 2003;Coultrap and Bayer, 2012;Coultrap et al., 2014;. Since positive modulation of SK channel activity attenuate hippocampal T286 pCaMKIIα localization (Figure 8) and burst discharge rate (Figure 12), we ask whether SK modulation of CA1 firing properties is dependent on the stimulation state. Photostimulation of genetically encoded ChR2 in the hippocampus of mice increased burst activity and synaptic strength when the ISI histogram and autocorrelogram plots were evaluated (Figures 13H, 14). Interestingly, Apamin-induced SK2 blockage further enhanced burst firing and the firing rate of CA1 neurons when compared with photostimulation only. In the absence of an external stimulus in vivo, spontaneously evoked CA1 firing rate and burst rate were significantly attenuated during SK channel potentiation (Figure 12). Likewise, when we paired photostimulation with SK channel potentiation, it significantly abolished the burst and firing rate increase associated with the photostimulation event (Figures 14G-I).

SUMMARY
Together, our results suggest that GluN1 modulation of spine plasticity and neuronal firing involves the co-regulation of SK2 and CaMKII. Additionally, SK2 can tune long-term synaptic plasticity by refining spine-specific localization and activity of CaMKII.

ETHICS STATEMENT
All animal handling procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Louisiana State University (LSU) School of Veterinary Medicine (SVM).