CNQX

TARP mediation of accelerated and more regular locus coeruleus network bursting in neonatal rat brain slices

Bijal Rawal, Vladimir Rancic and Klaus Ballanyi
Department of Physiology, Faculty of Medicine & Dentistry, 7-50 MSB, University of Alberta, Edmonton, T6G2H7 Alberta, Canada

Transmembrane AMPA receptor (AMPAR) regulatory proteins (TARP) increase neuronal excitability. However, it is unknown how TARP affect rhythmic neural network activity. Here we studied TARP effects on local field potential (LFP) bursting, membrane potential and cytosolic Ca2+ (Cai) in locus coeruleus neurons of newborn rat brain slices. LFP bursting was not affected by the unselective competitive ionotropic glutamate receptor antagonist kynurenic acid (2.5mM). TARP-AMPAR complex activation with 25µM CNQX accelerated LFP rhythm 2.2-fold and decreased its irregularity score from 63 to 9. Neuronal spiking was correspondingly 2.3-fold accelerated in association with a 2-5 mV depolarization and a modest Cai rise whereas Cai was unchanged in neighboring astrocytes. After blocking rhythmic activities with tetrodotoxin (1µM), CNQX caused a 5-8 mV depolarization and also the Cai rise persisted. In tetrodotoxin, both responses were abolished by the non-competitive AMPAR antagonist GYKI 53655 (25µM) which also reversed stimulatory CNQX effects in control solution. The CNQX-evoked Cai rise was blocked by the L-type voltage-activated Ca2+ channel inhibitor nifedipine (100µM). The findings show that ionotropic glutamate receptor-independent neonatal locus coeruleus network bursting is accelerated and becomes more regular by activating a TARP-AMPAR complex. The associated depolarization-evoked L-type Ca2+ channel-mediated neuronal Cai rise may be pivotal to regulate locus coeruleus activity in cooperation with SK-type K+ channels. In summary, this is the first demonstration of TARP-mediated stimulation of neural network bursting. We hypothesize that TARP-AMPAR stimulation of rhythmic locus coeruleus output serves to fine-tune its control of multiple brain functions thus comprising a target for drug discovery.

1. Introduction
In mammalian neurons, fast excitatory neurotransmission is mostly mediated by an α-amino-3- hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR) comprising one type of ionotropic glutamate receptors (iGluR) (Traynelis et al., 2010). Molecularly diverse AMPAR subtypes are coupled to a member of the family of auxiliary transmembrane AMPAR regulatory proteins (TARP) (Greger et al., 2017; Jackson and Nicoll, 2011; Maher et al., 2017). As reviewed in the latter reports, TARP activation can modulate the excitability of certain neuron types and competitive AMPAR antagonists, like the quinoxalinediones 6-cyano-7-nitroquinoxaline-2,3- dione (CNQX) and 6,7-dinitroquinoxaline-2,3(1H,4H)-dione (DNQX), are pivotal tools for studying this. Specifically, they act as partial agonists at the TARP-AMPAR complex and consequently depolarize and enhance firing, e.g. in principal neurons and (inhibitory) interneurons in the hippocampus (Hashimoto, et al., 2004; Maccaferri and Dingledine, 2002; McBain et al., 1992), thalamus (Lee et al., 2010), cerebellum (Brickley et al., 2001; Menuz et al., 2007) or spinal cord (Sullivan et al., 2017).
Regarding functional TARP roles, most of the above studies on acute brain slices showed that quinoxalinedione-evoked firing augments spontaneous (inhibitory) postsynaptic currents. Such currents can be analyzed in control mice compared to animals with a ‘knock-out’ of TARP-coding genes like in a report pointing out the requirement of ‘Type-I TARP -2’ for inflammation- associated spinal AMPAR plasticity (Sullivan et al., 2017). Moreover, ‘stargazer’ mice that lack -2, and were therefore instrumental in TARP discovery, show neurological deficits like ataxia, dyskinesia and seizures (Jackson and Nicoll, 2011; Maher et al., 2017).
While TARP effects at the cellular level are established, their role in active neural networks is currently unknown. This is due to the fact that most rhythmically active brain circuits operate via iGluR and quinoxalinediones thus block their activity like in locomotor (Hägglund et al., 2010) or respiratory (Ballanyi and Ruangkittisakul, 2009) networks, entorhinal cortex (Garaschuk et al., 2000) or hippocampus (Sipilä and Kaila, 2008). The locus coeruleus (LC) in the brainstem might be an important model for studying TARP-AMPAR complex roles for two reasons. Firstly, it innervates most brain structures and thus modulates many behaviors including arousal, sleep-wake cycle, memory, anxiety or opioid (withdrawal) effects (Berridge and Waterhouse, 2003; Foote et al., 1983). Secondly, in newborn rats gap junction-coupled LC neurons generate presumably synchronous Na+ action potentials via a mechanism that may not depend on iGluR as such ‘spiking’ is not blocked by quinoxalinediones (Alvarez-Maubecin et al., 2000) similar to persistence of tonic spiking in adult rats (Alvarez et al., 2002). Nevertheless, at both developmental stages LC neurons have functional NMDA- and AMPA/Kainate-type iGluR as specific agonists increase their spike rate (Kogan and Aghajanian, 1995; Olpe et al., 1989; Zamalloa et al., 2009).
Local field potential (LFP) recording is a potent tool to analyze neural network functions (Ballanyi and Ruangkittisakul, 2009; Buzsáki et al., 2012; Einevoll et al., 2013; Totah et al., 2018). We recently reported that spiking of LC neurons at ~1 Hz in newborn rat slices generates a rhythmic ~0.2 s-lasting crescendo-decrescendo-shaped LFP (Rancic et al., 2018). It was the aim of the present study to investigate by combining LFP monitoring with either ‘whole-cell’ membrane potential (Vm) recording or imaging of the free cytosolic Ca2+ concentration (Cai) in LC neurons and neighboring astrocytes whether CNQX has a TARP-mediated modulatory action on rhythmic population activity in this spontaneously active neonatal neural network.

2. Materials and methods
2.1 Preparation and solutions
The experiments were performed on horizontal brain slices from 0-5 days-old CD-001 (SD) rats of unknown sex (Charles River Laboratory Inc., Wilmington, MA, USA). All procedures were approved by the University of Alberta Animal Care and Use Committee and in compliance with guidelines of the Canadian Council for Animal Care and in accordance with the Society for Neuroscience’s ‘Policies on the Use of Animals and Humans in Neuroscience Research’.
Procedures for generating LC-containing brain slices are described elsewhere in detail (Kantor et al., 2012). In brief, rats were anesthetized with 2-3 % isoflurane to a level that caused disappearance of the paw withdrawal reflex. They were then decerebrated and the neuraxis was isolated at 18-20 °C in superfusate containing (in mM): 120 NaCl, 3 KCl, 1.2 CaCl2, 2 MgSO4, 26 NaHCO3, 1.25 NaH2PO4, and 10 D-glucose (pH adjusted to 7.4 by gassing with carbogen, i.e. a mixture of 95 % O2 plus 5 % CO2). The brain was glued on its ventral surface to a metal cutting plate which was then inserted into the vise of a vibratome (Leica VT1000S; Leica Microsystems, Richmond Hill, ON, Canada). In carbogenated superfusate, serial horizontal brain slices were cut at room temperature, initially at 400-600 µm steps, until the apex of the 4th ventricle appeared. Slice thickness was then reduced to 100 µm. Once the LC started to appear as a dark oval area close to the lateral border of the 4th ventricle, a single 400 µm thick slice was cut. This slice typically contained >50% of the dorsoventral aspect of the round LC soma area which extends in newborn rodents by 300-400 µm in the horizontal plane (Ishimatsu and Williams, 1996).
As the LC is bilaterally-organized, one hemisected slice was immediately used for recording whereas the contralateral slice half could be stored without apparent changes in LC properties at 30 oC for up to 5 h in a glass beaker filled with continuously carbogenated superfusate. For recording, a slice was mechanically fixed with a platinum ‘harp’ in an acrylic chamber (volume ~1 ml) with glass bottom (Warner Instruments, Hamden, CT, USA). The LC and individual cells were visualized with a x20 objective (XLUMPlanF1, numerical aperture 1.0) of an MPE microscope (Olympus, Markham, ON, Canada) or an IR-DIC video camera (OLY-150, Olympus). A peristaltic pump (Sci-Q 403U/VM, Watson-Marlow, Wilmington, MA, USA) was used to apply at a rate of 5 ml/min carbogenated superfusate which was removed from the chamber with vacuum applied to a hypodermic needle. Superfusate temperature in the chamber was kept at 28 °C using a heat control system (Thermo-Haake DC10-V15/B, Sigma Aldrich, Canada).

2.2 Pharmacology
The following bath-applied agents were used from stock solutions: CNQX (25 mM in H2O; Sigma Aldrich), kynurenic acid (100 mM in DMSO; Sigma Aldrich), GYKI 53655 (GYKI; 25 mM in H2O; Tocris, Canada), nifedipine (100 mM in DMSO; Sigma Aldrich), tetrodotoxin (TTX; 1 mM in 1 N NaOH; Alamone, Israel). CNQX was studied at 25 µM as this dose is similar to those used in previous TARP studies (for references, see Discussion). Also, as tested in 4 slices, 10 µM CNQX did not change LFP rate from its control value of 56.5 ± 13.3 bursts/min (P= 1.0) whereas 50 µM accelerated rhythm to 106.6 ± 7.1 bursts/min (P< 0.001), very similar to the effect of 25 µM (see Results). To facilitate recovery during Vm or Cai recording, CNQX was mostly applied for 1-2 min (which was sufficient for the effects to reach steady-state) while GYKI (25 µM) or TTX (1 µM) were preincubated for 2-3 min (the latter until rhythmic discharges were blocked). 2.3 Electrophysiological recording Patch pipettes were pulled from borosilicate glass capillaries (GC-150TF-10; 1.5 mm outer Ø, 1.17 mm inner Ø, Harvard Apparatus) to an outer tip Ø of ~2 μm using a vertical puller (PC-10, Narishige International Inc., Amityville, NY, USA). For suction electrode LFP recording, patch pipettes were broken and subsequently beveled manually with sand paper (Ultra Fine 600 Grit, Norton-Saint Gobain, Worcester, Wa, USA) at an angle of 45º to an oval-shaped tip with an inner opening of 40-60 μm. After filling with superfusate, the d.c. resistance of the suction electrodes was ~200 k (Rancic et al., 2018). Following insertion into a patch electrode holder (ESP-M15P and MHH-25, Warner Instruments), 15-30 mmHg negative pressure was applied to the electrode with a syringe (BD Diagnostics, Franklin Lakes, NJ, USA) and controled via a differential pressure sensor (Honeywell, SCX05DN, Fort Worth, TX, USA). A MP-285 micromanipulator (Sutter Instrument Company, Novato, USA) was used to position the suction electrode opening at a flat angle of ~30º on the slice surface followed by application of <5 mmHg negative pressure. The ‘raw’ suction electrode signal was amplified (x10k) and band-pass-filtered (0.3-3 kHz) using a Model-1700 differential amplifier (AM-Systems, Sequim, WA, USA). In parallel, the electrode signal was further processed for another recording channel by integration with a time constant of 50 ms using a MA-821/RSP unit (CWE, www.cwe-inc.com). For whole-cell Vm recording using an EPC-10 amplifier and ‘Pulse’ software (version 10, HEKA, Lambrecht, Germany), fire-polished patch pipettes were filled with (in mM): 140 K-gluconate, 1 NaCl, 0.5 CaCl2, 1 MgCl2, 1 Na2-ATP, 1 K4-BAPTA and 10 Hepes; pH adjusted to 7.4 with 1 N KOH; d.c. resistance in superfusate was 5-8 MΩ. When dimpling of the soma membrane occurred while visually targeting a neuron using the MP-285 micromanipulator, 20 mmHg positive pressure was released and negative pressure was applied for gigaseal formation (>1 GΩ). Whole-cell recording was established by abrupt suction (~100 mmHg). Only cells were analyzed in which spike amplitude was >70 mV while Vm was stable for a 5 min control period during which series resistance ranging from 10 to 50 MΩ (n= 7) was compensated and input resistance (238 ± 82 M, n= 7) was measured by injection of -60 pA d.c. current for 500 ms at an interval of 10-15 s. In ~30% of neurons, up to 100 pA d.c. current was injected to minimize random spike failures. Due to ongoing ‘subthreshold’ Vm oscillations (Christie et al., 1989), LC neurons do not have a ‘resting’ Vm which was thus defined as the value at 50 % of the time interval between single spikes evoked at the oscillation peak. Electrophysiological signals were sampled at 20 kHz (4 kHz for exclusive LFP recording) into a digital recorder (Powerlab 8/35 + LabChart 7 software, ADInstruments, Colorado Springs, CO, USA) connected to a personal computer.

2.4 Multiphoton population Cai imaging
Population Cai imaging enables visualization of brain activity (Yang and Yuste, 2017). We have used this approach in neonatal rat brain slices to visualize synchronous neuronal depolarizations in the inspiratory center (Ballanyi & Ruangkittisakul, 2009; Ruangkittisakul et al., 2009) and hippocampal CA3 area (Ruangkittisakul et al., 2015). Also in the LC, it is principally possible, using fast ‘line-scanning’ or a small region of interest (ROI), to visualize in single neurons Cai rises associated with single Na+ spikes at rates of ~1 Hz (Sanchez-Padilla et al., 2014). However, we aimed here at analyzing the activity of multiple LC cells in the same confocal plane. As this required xy-scanning at a rate of 1.1 frames/s, fast rhythmic Cai rises could not be resolved here.
For population Cai imaging, a broken patch pipette (outer diameter, 5-10 μm) was filled with the membrane-permeant green fluorescent Ca2+ dye Fluo-4-AM (5 mM in 20 % DMSO + pluronic acid, further diluted to 0.5 mM in superfusate, Invitrogen, Carlsbad, CA, USA) and then pressure- injected (30-50 mmHg) for 10 min at 40-50 µm depth into the LC. Within 10 min, an area of 150- 300 μm in diameter was stained. Fluorescence signals were typically measured 50-60 μm below the slice surface using the above specified Olympus x20 objective and MPE system emitting Ti:Sa laser light at 810 nm. Cai baseline changes were visualized in the somata of 15-25 cells in a single xy-image plane at 2x digital zoom with a sampling rate of 1.1 frames/s.

2.5 Immunohistochemistry
The vast majority of LC cells are neurons whereas the rest comprises astrocytes and some oligodendrocytes (Alvarez-Maubecin et al., 2000). Here, we showed with double-staining using the neuronal marker tyroxine hydroxylase (TH) and the astrocyte marker S100 that ~90 % of LC cells are neurons with a soma diameter >20 µm vs. ~10 % of notably smaller astrocytes. For their identification, 3 brains from 0 days-old pups were isolated and then chemically fixed in a solution comprising 4% paraformaldehyde (Sigma-Aldrich) in phosphate buffer (i.e. a 1:2 mixture of 0.1 M NaH2PO4 + 0.1 M Na2HPO4 in H2O, pH 7.2). After at least 24 h, 60 μm thick horizontal sections containing the LC were sliced and incubated overnight at 4 ºC in monoclonal rabbit anti-TH (1:1000, Life Technologies, Camarillo, CA, USA) and mouse anti-S100 (1:1000, Sigma Aldrich) antibodies in TBS (i.e. 0.9 % w/v NaCl plus 0.6 % w/v tris base, sigma, pH 7.4) also containing 0.3 % Triton X-100 (Sigma-Aldrich). Subsequently, the sections were incubated at room temperature in secondary goat anti-rabbit Alexa 594 (1:500, life technologies) and goat anti-mouse Alexa 488 (Jackson Immunoresearch Laboratories Inc, West Grove, PA, USA) for 90 min. From each brain, cells in the section with the largest LC extension were analyzed using the MPE system.

2.6 Data Analysis
Data were analyzed from slices in which LFP bursts had a crescendo-decrescendo shape as reported previously (Rancic et al., 2018). This was the case in >70% of slices. Burst rates and amplitudes were quantified during 1 min recording time periods in control or at steady-state of drug effects. LFP burst duration was defined with Clampfit software (Molecular Devices Corporation, Chicago, Il, USA) as the time interval from when the averaged signal increased above and decreased below a threshold set at 10 % of peak amplitude, respectively. The extent of LC network synchronicity was determined with Clampfit software by comparing over a time period of 10 s the cross-correlation between whole-cell-recorded single neuron spiking and the integrated LFP burst. The peaks in cross-correlograms refer to the cross-correlation function estimate (CFE) values for the extent of synchronicity. The ‘lag time’ quantifies the shift in the peak between these events and thus gives a measure of spike jitter. The regularity of LC network bursting was determined by quantifying the irregularity score which is an established parameter to analyze rhythmic neural network bursting, e.g. of the inspiratory center (Garcia et al., 2017). The formula for its determination is: Irregularity score = 100 * [(Pn-Pn-1)/Pn-1]/N, where N is the number of bursts, Pn is the period of nth burst and Pn-1 is the period of the preceeding burst. Note that the score has no unit. The lower the value, the more regular is the network rhythm.
Cai imaging analysis was done using Olympus Fluoview software (FV10-ASW, version 03.01.01.09). Relative changes in Cai baseline were referred to a percentage change in fluorescence intensity. Quantified was the ratio of the difference of the final value (average of the peak response) and the initial value (average baseline values) over the initial value multiplied by 100. Cai imaging revealed in ~30 % of recordings a modest linear drift of baseline fluorescence intensity which was compensated during offline analysis.
Values are given as means ± SD and n-values correspond to measurements in 1 slice per animal. Significance values, i.e., non significant (ns): P>0.05, *P< 0.05, ** P< 0.01, ***P< 0.001) were assessed by paired two-tailed t test (electrophysiological data) and one-way ANOVA with Dunnett’s post-test (Cai imaging data) using Prism software (GraphPad Software Inc., La Jolla, CA, USA). To facilitate reading of the Results section text, quantified data shown with statistical details in the scatter plots of Figs. 1, 2, 5 and 6 or Tab. 1 are only represented as means ± SD. 3. Results Initial analysis of iGluR blocker effects on LFP recording-based network bursting was followed by neuronal Vm analysis and Cai imaging in groups of Fluo-4-loaded neurons and astrocytes. 3.1 Stimulatory CNQX effect on iGluR-independent network bursting As a precondition for testing TARP roles in LC network function, it was studied if LFP rhythm persists during iGluR blockade (see Introduction). This was indeed the case as the non-selective competitive iGluR antagonist kynurenic acid had no effect on the LFP in 4 slices (Fig. 1). As exemplified in Fig. 2A for 1 of 12 slices, 5 min bath-application of 25 µM CNQX caused within 1-2 min a stable effect on the LFP that washed out within 5-15 min. While CNQX did not change LFP amplitude (101.6 ± 2.6 % of control, P= 0.07) (Fig. 2B), it increased its rate 2.24 ± 0.5 -fold (from 53.4 ± 10.6 to 116 ± 19.1 bursts/min) (Fig. 2C) and decreased its single burst duration from 268 ± 56.1 to 205 ± 29.4 ms (Fig. 2D). In 3 of the 12 slices, the accelerated rhythm had an almost sinusoidal oscillatory pattern lacking an inactivity phase between bursts and could include burst amplitude fluctuations by <15% (Fig. 2A, see also Figs. 3A, 4A, 6A). As exemplified in Fig. 3A, recording of neuronal Vm in 7 of the 12 slices revealed a very similar 2.29 ± 0.43 -fold acceleration of spike rate (from 50.4 ± 9.0 to 113 ± 9.2 bursts/min) (Fig. 2E, F) and a concomitant depolarization by 2-5 mV (3.1 ± 0.9 mV) from a resting Vm of -45.8 ± 3.2 mV (Tab. 1). These results show that neonatal LC network bursting does not depend on iGluR and indicate that a TARP-AMPAR complex is functional. LFP and Vm were next simultaneously recorded after abolishing rhythmic activities with TTX to (i) better quantify the CNQX depolarization, (ii) minimize a potentially short-circuiting Vm effect of SK-type K+ channel activation due to spike- related Ca2+ influx (Kulik et al., 2002) and (iii) exclude that enhanced spiking causes endogenous neurotransmitter/modulator release (Schwarz and Luo, 2015; Singewald and Phillipu, 1998). As tested in 5 of the above 7 neurons, TTX blocked in 0.5-3 min the LFP, Na+ spikes and subthreshold oscillations with no effect on resting Vm (which stabilized at –46.6 ± 3.9 mV) (Fig. 3B) (Tab. 1). Subsequent CNQX application depolarized Vm by 5-8 mV (6.6 ± 1.3 mV) with no change in input resistance (238 ± 82 M in control vs. 234 ± 75 M in CNQX, P= 0.42) (Fig. 3C) (Tab. 1). 3.2 Reversal of stimulatory CNQX effect by non-competitive iGluR antagonist In cerebellar neurons, the CNQX-evoked TARP-mediated depolarization was blocked by the non- competitive AMPAR antagonist GYKI (Menuz et al., 2007). Also here, CNQX failed, after preincubating slices in GYKI + TTX, to change Vm in 6 neurons (-44.8 ± 5.1 mV vs. -44.7 ± 5.2 mV) (Fig. 4A) (Tab. 1). In 3 different neurons, CNQX was continuously applied in control solution. When the accelerating effect on spiking and LFP rate was stable after 1-2 min, addition of GYKI to the CNQX-containing solution reversed the CNQX effects within ~1 min (Fig. 4B). 3.3 Uniform L-type Ca2+ channel-mediated CNQX-evoked Cai rise in LC neurons The finding that CNQX depolarized all neurons to a similar (modest) extent indicates that the entire nucleus may show a uniform response. To test this, population Cai imaging was done to determine (i) whether the depolarization and the associated accelerated spiking raises Cai, (ii) how many LC neurons in the same optical plan show such a response and (iii) whether CNQX also affects Cai in neighboring astrocytes that show in control solution depolarizations which are synchronous with subthreshold neuronal Vm oscillations (Alvarez-Maubecin et al., 2000). Fig. 5A illustrates that TH-immunostained LC neurons with a soma diameter >20 µm comprise ~90 % of LC cells and notably smaller S100-immunostained astrocytes the remaining ~10 %. Fig. 5B shows Fluo-4-loaded neurons and astrocytes in one image plane during the steady-state effect of CNQX. Fig. 5C exemplifies for 4 of these neurons that CNQX caused a very similar modest Cai increase whereas 4 astrocytes did not respond. Corresponding findings were obtained in a total of 174 neurons and 44 astrocytes of 8 slices (Fig. 5D). Two of the 4 astrocytes in Fig. 5C, and 23 % in total, showed spontaneous Cai ‘spikes’ unrelated to CNQX application. Similar modest Cai rises were seen during CNQX in control solution (30.14 ± 0.08 % ) and in TTX (27.83 ± 0.07 %) which itself lowered Cai baseline by 28.13 ± 0.03 % (Fig. 5C, D). After recovery from CNQX in TTX and preincubation of 4 of the above 8 slices in TTX plus either the L-type voltage- activated Ca2+ channel blocker nifedipine or GYKI, CNQX did not change neuronal Cai (Fig. 5D).

3.4 CNQX effect on network synchronicity and regularity of bursting
The above finding that CNQX shortened the LFP burst suggests that CNQX enhances LC network synchronization. This was studied by correlating cellular spiking with LFP events at steady-state, typically 90 s after start of application as exemplifed in Fig. 6A for the neuron shown in Fig. 3. The correlograms in Fig. 6B did not reveal a CNQX-induced change of the CFE value for this neuron whereas the lag time was shortened. For all 7 neurons in which CNQX effects have been analyzed above, the scatter plots did not indicate a change in the CFE value (0.33 ± 0.04 in control vs. 0.33 ± 0.04 in CNQX) while there was a trend for a decrease in lag time (87.8 ± 26.1 ms in control vs. 59.3 ± 23.5 ms in CNQX) (Fig. 6C). But, as analyzed in the 12 slices analyzed above, CNQX made the LFP rhythm more regular as evident from a decrease of the irregularity score from 63.1 ± 19.7 to 8.8 ± 4.2 % (Fig. 6D).

4. Discussion
This study reports for the first time TARP effects on spontaneous neural network activity, specifically in the LC of newborn rat brain slices. This network was chosen because it comprises a small nucleus in which coordinated discharge of a single spike in each neuron causes a robust rhythmic LFP that can be analyzed pharmacologically (Rancic et al., 2018). LFP recording in combination with either whole-cell Vm recording or Cai imaging unraveled, in summary, (i) non- synchronous, yet phase-locked, spiking does not rely on iGluR, (ii) CNQX causes in all neurons a modest depolarization and Cai rise, (iii) CNQX-evoked depolarization more than doubles cellular and network discharge rates with a concomitant decrease in jitter and increased regularity of network bursting. Underlying mechanisms and the potential physiological relevance are discussed.

4.1 LFP analysis of CNQX stimulation of iGluR-independent network bursting
Spontaneous neonatal LC neuron spiking is presumably synchronous (Christie, 1997). But, one of the reports establishing this view stated that subthreshold Vm oscillations are synchronous whereas Na+ spikes at their peak not necessarily (Christie et al., 1989). In our recent study (Rancic et al., 2018), we showed that spiking is not synchronous, but rather jittered, yet phase-locked, to the LFP burst. We also found in that report that increasing LC excitability with raised extracellular K+ transforms the LFP pattern from ~0.2 s-lasting crescendo-decrescendo-shaped bursts to ~3 s- lasting multi-peak events. Here, we show firstly that LFP properties are not changed by the broad spectrum competitive iGluR antagonist kynurenic acid (Traynelis et al., 2010). It was already reported that single LC neuron spiking in newborn and adult rat brain slices persists in presence of diverse iGluR antagonists (Alvarez-Maubecin et al., 2000; Alvarez et al., 2002; Olpe et al., 1989). The lack a blocking effect by these antagonists is not due to absence of functional iGluR in LC neurons. In fact, both NMDA-type and AMPA/Kainate-type iGluR agonists evoke inward currents that depolarize these neurons to accelerate their spontaneous spiking (Kogan and Aghajanian, 1995; Olpe et al., 1989; Williams et al., 1991; Zamalloa et al., 2009). Our results point out, in addition, that blocking all iGluR subtypes with kynurenic acid does not affect LC network coupling. This contrasts with findings in other spontaneously active (neonatal) rodent neural networks showing that iGluR antagonists block rhythm generation. Specifically, these networks are the locomotor central pattern generator (Hägglund et al., 2010), the breathing center (Ballanyi and Ruangkittisakul, 2009), the inferior olive (Devor and Yarom, 2002; Placantonakis and Welsh, 2001), the entorrhinal cortex (Garaschuk et al., 2000) and the hippocampal CA3 area (Sipilä and Kaila, 2008). A study on LC neurons in newborn rat brainstem-spinal cords anecdotally mentioned that CNQX caused a small depolarization and accelerated their tonic spiking (Oyamada et al., 1998). Also here, CNQX accelerated neuronal spiking (and LFP rate as well). Already these findings prove that pharmacological LFP analysis is a powerful tool to analyze this neural network.

4.2 TARP mediation of a uniform excitatory CNQX effect
Like in LC neurons of newborn rat brainstem-spinal cords (Oyamada et al., 1998), CNQX caused here a modest (~5 mV) depolarization that nevertheless more than doubled LFP and spike rates. The lack of a CNQX-evoked decrease in input resistance indicates that (synaptic) ion channels causing the depolarization might be located primarily on the pericoerulear dendrites and space- clamp might thus not be sufficient to reveal their activation with somatic injection of current pulses (Ishimatsu and Williams, 1996). Accordingly, the depolarization and concomitant Cai rise might be more pronounced in distal dendrites. The inhibitory TTX effect on both Na+ spikes and subthreshold Vm oscillations were similar to those in the latter report (Oyamada et al., 1998).
Contrary, others noted that Vm oscillations are TTX-resistant in (most) LC neurons of slices from juvenile rats (Christie et al., 1989) or mice (Sanchez-Padilla et al., 2014). The fact that TTX did not attenuate the CNQX depolarization indicates that accelerated spiking in CNQX does not cause neurotransmitter/modulator release which might potentially contribute to this response (Schwarz and Luo, 2015; Singewald and Phillipu, 1998). The observation that CNQX raised Cai in all neurons within the same optical plane indicates that the agent causes a depolarization of similar amplitude in the entire network. However, this does not mean that all LC neurons respond directly to the drug as they are coupled by gap junctions which act as a ‘low-pass filter’ (Christie, 1997; Christie et al., 1989; Ishimatsu and Williams, 1996).
LC neurons do not comprise a homogeneous class of brain cells. For example, ventrally-located adult rat LC neurons with shorter spikes and smaller afterhyperpolarizations than LC core neurons act as a ‘pontospinal-projecting module’ (Li et al., 2016). Moreover, dorsomedially-located small and densely packed GABAergic neurons in juvenile mice show faster spiking with enhanced adaptation (Jin et al., 2016). Finally, topographically distinct LC modules exist also regarding both afferent synaptic inputs and efferent projections (Schwarz and Luo, 2015) and their in vivo activity (Totah et al., 2018). Consequently, future studies may determine whether a TARP-AMPAR complex is functional in all (neonatal) LC neurons or rather only in a cluster that transmits TARP- enhanced activity via the gap junction-coupling to neighboring cells. In any case, a sustained activity burst from afferent glutamatergic neurons, particularly in the nucleus paragigantocellularis, the lateral habenula or prefrontal cortex (Aston-Jones et al., 1986; Herkenham and Nauta, 1979; Jodo and Aston-Jones, 1997; Singewald and Phillipu, 1998) may depolarize these neurons more effectively than those lacking that TARP-AMPAR complex. But, the outcome may be similar as the additional excitation plus the resulting Cai rise can potentially spread via the gap junctions throughout the nucleus. Contrary to the uniform CNQX-evoked neuronal Cai rises, there was no effect on neighboring astrocytes. This is somehow surprising as LC neurons and astrocytes are gap junction-coupled as evidenced by their synchronous spontaneous Vm oscillations and dye diffusion between them (Alvarez-Maubecin et al., 2000).
Our finding of an excitatory neuronal CNQX action is in line with results from previous slice studies. Early reports showed that both excitatory and inhibitory neurons are depolarized by CNQX and (mostly) DNQX, but not NBQX (Brickley et al., 2001; Hashimoto, et al., 2004; Maccaferri and Dingledine, 2002; McBain et al., 1992). The authors noted that the effect indicates a novel type of excitatory action of these quinoxalinedione-type iGluR antagonists. In more recent work on neurons in slices or after dissociation, it was hypothesized (Lee et al., 2010) or proven (Menuz et al., 2007; Rigby et al., 2015; Sullivan et al., 2017) that CNQX activates a TARP- AMPAR complex. Since their discovery a decade ago, TARP-AMPAR complex structures and functions have been analyzed thoroughly, often in expression systems or genetically-engineered (e.g. ‘stargazer’) mice as reviewed comprehensively (Greger et al., 2017; Jackson and Nicoll, 2011; Maher et al., 2017). This work also established that TARP-AMPAR complex activation by the ‘partial agonist’ CNQX is prevented by non-competitive antagonists drugs (including ‘GYKI’- type agents) which, by themself, exert no partial agonist action on this structure (Brickley et al., 2001; Menuz et al., 2007). Accordingly, we found that preincubation with GYKI in TTX blocked both the CNQX-evoked depolarization plus Cai rise and reversed its stimulatory action in control solution. Based on the above arguments, our results strongly suggest that the excitatory CNQX action on the LC network is due to activation of a TARP-AMPAR complex.

4.3 Functional TARP role
Regarding possible functions of the AMPAR-TARP complex, augmentation of a glutamate- mediated Cai rise might regulate the excitability of LC neurons which seem to be very sensitive to this pivotal second-messenger and vice versa. This assumption is based on our finding that nifedipine (in TTX) abolished the likely modest somatic CNQX-evoked Cai rise indicating that L- type voltage-activated Ca2+ channels are activated by the ~5 mV depolarization from a resting Vm of about -50 mV. Likely, these channels are also responsible for the TTX-sensitive spike-related tonic increase in Cai baseline also seen, for example, in dorsal vagal neurons that spike spontaneously at a similar rate (Kulik et al., 2002). As one explanation for the latter findings, these, and possibly also other voltage-activated Ca2+ channel subtypes, may interact closely with SK- type Ca2+-activated K+ channels in the neonatal rat LC. In fact, in adult rats in vivo (Aghajanian et al., 1983) and slices (Andrade and Aghajanian, 1984), the number of current-evoked spikes correlates with the duration of a prounced hyperpolarization that is attenuated by increased cellular Ca2+ buffering. Moreover, the rate of spontaneous spiking in LC neurons of adult mouse slices is regulated by cooperation of L- and T-type Ca2+ channels with SK2 channels (Matschke et al., 2015, 2018). If this is the case also in the neonatal rat LC, even the modest TARP-dependent depolarization, that might though be more pronounced in the dendrites (see above), may serve to fine-tune network activity while an excessive increase in such activity may be a crucial factor in neurodegenerative diseases. In that regard, findings from combined whole-cell-recording and Cai imaging in LC neurons of juvenile mouse slices indicate that activity-related L-type channel- mediated Ca2+ entry and resulting mitochondrial stress may contribute to the etiology of Parkinson’s or Alzheimer’s disease (Sanchez-Padilla et al., 2014).
Regarding physiological TARP roles, the neural network in the inferior olive shares properties with that in the LC, including subthreshold Vm oscillations that are synchronized by gap junction- coupling (Devor and Yarom, 2002; Placantonakis and Welsh, 2001). A more recent modeling study on the inferior olive network concluded that ’subthreshold oscillations of the individual neurons and the electrical gap junctions make this system a powerful encoder and generator of spatiotemporal patterns with different, but coordinated oscillatory rhythms’ (Latorre et al., 2013). Similarly, recent findings from our lab (Rancic et al., 2018) and other groups, both in vivo and in vitro, indicate that the LC transforms its activity pattern under the influence of (glutamatergic) inputs or neuromodulators like noradrenaline, cocaine or opioids (Chandley and Ordway, 2012; Safaai et al., 2015; Totah et al., 2018; Zhu and Zhou, 2005). In addition to our previous observation that opioids and high K+ transform the LFP pattern (Rancic et al., 2018), we found here that TARP makes the rhythm more regular while, at the same time, more than doubling its output population burst rate. These dynamic properties may serve to fine-tune the LC control of multiple brain circuits and thus of behaviors including arousal, sleep-wake cycle, breathing, memory, pain sensation, anxiety and opioid (withdrawal) effects (Berridge and Waterhouse, 2003; Foote et al., 1983). Consequently, regarding various diseases in these systems, TARP within the LC might be a potent target to improve drug efficacy while mitigating adverse effects (Jackson and Nicoll, 2011; Maher et al., 2017).