Serotonergic Suppression of Mouse Prefrontal Circuits Implicated in Task Attention

Visual Abstract


Introduction
The medial prefrontal cortex (mPFC) is critical for "topdown" executive control of attention (Miller and Cohen, 2001;Knudsen, 2007), and disruption of its signaling impairs normal performance on attention tasks (Muir et al., 1996;Miner et al., 1997;Granon et al., 1998) Here, we investigated how mPFC L6 pyramidal neurons are modulated by 5-HT, how L6 excitation affects the two major classes of L5 interneurons, and how 5-HT modulates these internal mPFC circuits. Our results reveal a strong 5-HT-elicited inhibition of L6 pyramidal neurons mediated by 5-HT 1A receptors and a lesser, statedependent, and somewhat unexpected, contribution by 5-HT 2A receptors. We find that L6 pyramidal neurons robustly activate fast-spiking (FS) as well as non-fastspiking (nFS) interneurons in L5. Finally, we show that these intracortical circuits in mPFC are strongly suppressed by 5-HT.

Experimental animals
We used BAC transgenic Swiss Webster mice with expression of eGFP driven by the synaptotagmin 6 promoter (Syt6-EGFP EL71, MMRRC; RRID:MMRRC 010557-UCD) made by the GENSAT Project (Gong et al., 2003). L6 pyramidal neurons in mPFC strongly express eGFP and facilitate visual targeting of these neurons for recording (Tian et al., 2014). Syt6 mice were kept heterozygous, and there were no significant differences in their 5-HT responses compared to their wild-type littermate controls or wild-type C57BL/6 mice (F (2,64) ϭ 0.58, p ϭ 0.56, one-way ANOVA). To investigate the downstream synaptic connection of prefrontal L6 pyramidal neurons, we crossed GENSAT epiphycan BAC transgenic mice expressing Cre-recombinase (Epyc-Cre KR363, a gift from Dr. Nathaniel Heintz at Rockefeller University; RRID: MMRRC_036145-UCD) with Ai:32 mice (Jackson Laboratories, Bar Harbor, ME; RRID:IMSR_JAX:024109) to achieve eGFP-channelrhodpsin-2 expression in prefrontal L6 neurons (Epyc-ChR2). Wild-type littermates of the Epyc-ChR2 were used as controls to ensure that the UV light did not have effects in brain slices from mice lacking channelrhodopsin-2. Translating ribosome affinity purification (TRAP) and quantitative reverse-transcription (qRT)-PCR were used to confirm that Syt6 and Epyc-Cre neurons indeed represent an overlapping population of L6 glutamatergic neurons. All experimental animal procedures were performed in accordance with the University of Toronto and Rockefeller University Institutional Animal Care and Use Committee regulations.

Electrophysiology
Coronal brain slices (400 m) for electrophysiological recordings were obtained from adult male mice (postnatal 60 to 170 days; mean Ϯ SEM; 101 Ϯ 4 days; n ϭ 41 mice). Brains were rapidly excised and chilled in 4°C oxygenated sucrose artificial cerebrospinal fluid (ACSF; 254 mM sucrose, 10 mM D-glucose, 24 mM NaHCO 3 , 2 mM CaCl 2 , 2 mM MgSO 4 , 3 mM KCl, 1.25 mM NaH 2 PO 4 ; pH 7.4). Coronal slices (400 m thick, 2.34 -0.74 mm from bregma) were cut on a Dosaka Linear Slicer (SciMedia, Costa Mesa, CA) and recovered in 30°C oxygenated ACSF (128 mM NaCl, 10 mM D-glucose, 26 mM NaHCO 3 , 2 mM CaCl 2 , 2 mM MgSO 4 , 3 mM KCl, 1.25 mM NaH 2 PO 4 ; pH 7.4) for at least 2 h. Recovered slices were transferred to a perfusion chamber on the stage of a BX50W1 microscope (Olympus, Tokyo, Japan). ACSF was bubbled (95% O 2 , 5% CO 2 at room temperature) and perfused the chamber at a rate of 3-4 ml/min. In addition to recording from L6 pyramidal neurons based on neuronal morphology and anatomical landmarks in wild-type mice, L6 in Syt6 mice was landmarked with fluorescently identified eGFP-positive neurons (X-cite Series 120; Lumen Dynamics, Mississauga, Canada; Tian et al., 2014). Recording electrodes (2-4 M⍀) containing 120 mM potassium gluconate, 5 mM KCl, 2 mM MgCl 2 , 4 mM K 2 -ATP, 0.4 mM Na 2 -GTP, 10 mM Na 2phosphocreatine, and 10 mM HEPES buffer (adjusted to pH 7.3 with KOH) were used to patch L6 pyramidal neurons. Interneurons in L5 were identified visually based on their unique morphology in infrared differential interference contrast (small, circular somata) in contrast to L5 pyramidal neurons (oriented, triangular shaped somata, relatively thick apical dendrites toward pia). A subset of patched interneurons was filled with Alexa Fluor 594 (20 M) or Texas red dextran (0.15%) in the patch solution for morphological confirmation of these criteria. Interneurons were further subclassified as FS or nFS based on their electrophysiological spike pattern and maximal spike frequency. Multiphoton images were acquired with a Ti: sapphire laser (Mai Tai, Spectra-Physics, Fremont, CA) using an Olympus Fluoview FV1000 microscope and an Olympus XLPlan N 25ϫ water-immersion objective. Neuronal membrane potential and holding current were recorded with an EPC10 (HEKA Electronik, Lambrecht/ Pfalz, Germany) and corrected for the liquid junction potential (14 mV). All data were acquired at 20 kHz and low-pass filtered at 3 kHz with pClamp software (Molecular Devices, Palo Alto, CA). Threshold potentials for action potentials were detected using a derivative threshold of at least 20 mV/ms, and action potential amplitude was calculated as the change in membrane potential from threshold to the peak of the action potential. Intrinsic properties of L6 pyramidal neurons, as well as L5 FS and nFS interneurons, are summarized in Table 1.
To examine the effects of 5-HT on L6 pyramidal neurons near rest and during spiking, we performed wholecell patch-clamp recording in voltage clamp at -75 mV and in current clamp with current injections to elicit either constant spiking (2-3 Hz) at baseline or an initial membrane potential of -75 mV before depolarizing current injections (1 s, 25-pA steps, 15-s intervals) were used to assess input-output relationships. For the latter experiment, the frequency of action potential firing was measured for each depolarizing current step and plotted against the magnitude of the injected current step.

Pharmacology
Acute responses to 5-HT were probed by bath application of 5-HT (serotonin creatinine sulfate, Sigma-Aldrich, St. Louis, MO; 10 M; 30 s) in ACSF. To examine the effect of 5-HT on the excitability of L6 pyramidal neurons, 5-HT (10 M) was bath applied until a steadystate response was reached, and remained in bath throughout the duration of the input-output test protocols (ϳ2 min total application). Selective antagonists and agonists were from Tocris (Bristol, UK), except where mentioned. Antagonists for 5-HT 1A receptors (30 nM WAY100635, 10 M NAN-190) and 5-HT 2A receptors (30 nM MDL100907; 2 M ketanserin; 300 nM to 3 M ritanserin) were applied in bath for 10 min before further experiments with 5-HT. There were no significant differences between effects of 300 nM and 3 M ritanserin, and results were grouped for analysis. TCB-2 was used as a specific agonist of 5-HT 2A receptors (300 nM to 1 M).

Optogenetic stimulation
Channelrhodopsin-expressing neurons in Epyc-ChR2 mice were stimulated by blue LED light (473 nm) delivered by optic fiber (Thorlabs, Newton, NJ) mounted on a mechanical micromanipulator (Narishige International, East Meadow, NY). Light stimulation was directed directly to L6 of mPFC by targeted positioning of the optic fiber. Twenty light pulses (2-5 ms each) were delivered at 20 Hz to stimulate L6 neurons. This stimulation profile was sufficient to elicit robust activation of L6 pyramidal neurons expressing channelrhodopsin. In control experiments with brain slices from littermate mice lacking channelrhodopsin, light stimulation did not elicit a response in either L6 pyramidal neurons or L5 interneurons. Responses to light stimulation in L6 pyramidal neurons and L5 interneurons were measured in current-clamp from a baseline membrane potential of -75 mV held by continuous injection of depolarizing current. Response latency in L6 pyramidal neurons expressing channelrhodopsin was calculated from the time of light-on to the onset of the corresponding membrane potential change. Time-to-spike for L6 neurons from light-on was also calculated using the peak of the first resulting action potential. In L5 interneurons, the latency to response from L6 activation was calculated in voltage clamp as the time taken from light-on to the onset of the postsynaptic current, then corrected by the timeto-spike in L6 pyramidal neurons. Pairwise analysis of the effects of 5-HT on the excitation of L5 interneurons by optogenetic activation of L6 were performed using light stimulus that was able to elicit at least four action potentials in patched L5 interneurons. Light stimulus intensity to elicit a baseline of at least four action potentials did not differ between FS and nFS interneurons (t 14 ϭ 1.4, p ϭ 0.18, unpaired t test).

Statistical analysis
All recordings were analyzed using Clampfit software (Molecular Devices). Statistical analyses were performed with GraphPad Prism (GraphPad Software, La Jolla, CA). Analyses performed were one-sample t test, unpaired Student's t test, paired Student's t test, one-way ANOVA, two-way ANOVA, and two-way repeated-measures ANOVA. All tests were two-sided. Dunnett's multiple comparison tests were performed post hoc to compare changes in action potential firing in L6 neurons elicited by 5-HT. Sidak's multiple comparison tests were used to compare differences in spike frequency at individual injected current steps in the presence of 5-HT. All data are presented as mean Ϯ SEM.

Translating ribosome affinity purification and quantitative RT-PCR
Adult (8 -12 weeks old) Epyc-Cre mice under ketamine/ xylazine (100/10 mg/kg) anesthesia received single bilateral stereotaxic injections of 0.25 l AAV-FLEX-EGFPL10a virus (3.75 ϫ 10 12 genome copies/ml) into the mPFC (1.54 AP from bregma, 0.4 ML, -1.80 DV from dura). Animals were sacrificed in a controlled CO 2 chamber 3 weeks after surgery, brains were rapidly dissected in ice-cold Hanks balanced salt solution containing 2.5 mM HEPES-KOH (pH 7.4), 35 mM glucose, 4 mM NaHCO 3 , and 100 g/ml cycloheximide. The cortex was isolated from the rest of the brain, and each hemisphere was split along the coronal plane at the level of the genu of the corpus callosum (ϳ1.6 mm AP from bregma). The rostral portion was saved as the "PFC" and used for TRAP. Tissue from three mice (male and female) was pooled for each sample, and three biological replicates were collected. Polysome immunoprecipitations (IPs) were carried out as previously described ( . IPs were carried out overnight at 4°C. Bound RNA was purified using the Absolutely RNA Nanoprep kit (Agilent, Santa Clara, CA). RNA was also purified from the pre-IP supernatant to serve as whole-PFC "input" samples. RNA quantity was measured with a Nanodrop 1000 spectrophotometer, and quality was assayed on an Agilent 2100 Bioanalyzer. Only samples with RNA integrity values Ͼ7.0 were used for qRT-PCR analysis. cDNA was synthesized from 15 ng of IP or input total RNA using the Ovation qPCR System (NuGEN Technologies, Carlos, CA) following the manufacturer's protocol. qRT-PCR was performed on an Applied Biosystems StepOnePlus Fast Real-Time PCR System using commercially available Taqman assays (Table 2) and following standard cycling conditions (50°C for 2 min, 95°C for 10 min, then 40 cycles of 95°C for 15 s and 60°C for 1 min). Ten nanograms of cDNA was used for each qRT-PCR reaction, and technical triplicates were run for each of the biological triplicates from TRAP IP and input samples. The mean C T for technical replicates was used for quantification. Data were normalized to Gapdh by the comparative C T (2 -⌬⌬CT ) method (Livak and Schmittgen, 2001). Data are presented as mean Ϯ SEM of biological triplicates. Statistical significance was calculated between the normalized expression values (2 -⌬CT ) from the IP and input biological replicates for each gene by Student's t test in Microsoft Excel.

Serotonin robustly inhibits L6 pyramidal neurons of mPFC
Here, we investigated the electrophysiological consequences of 5-HT on pyramidal neurons in L6 of mPFC.
To investigate the functional effects of 5-HT on L6 pyramidal neurons during excitation, we used currentclamp and bath-applied 5-HT in the presence of injected positive depolarizing current sufficient to elicit action potential firing (2-3 Hz). Under these conditions, 5-HT hyperpolarized L6 neurons (-16.3 Ϯ 1.6 mV, n ϭ 17) and fully and significantly inhibited action potential firing in every recorded neuron (t 7 ϭ 13.5, p Ͻ 0.0001, n ϭ 8, paired t test; Fig. 1E, H, I). This suppression was repeatable in the same neuron after washout and was not affected by the presence of synaptic blockers (Fig. 1E, H,  I). Antagonism of 5-HT 1A receptors by WAY100635 significantly reduced the 5-HT-mediated hyperpolarization (-7.8 Ϯ 0.7 mV, t 18 ϭ 5.0, p Ͻ 0.0001, n ϭ 19, unpaired t test; Fig. 1F-I). Unexpectedly, however, 5-HT still robustly and significantly inhibited action potential firing in every neuron (t 13 ϭ 12, p Ͻ 0.0001, n ϭ 14, paired t test; Fig.  1F-I). This strong and significant suppression of L6 spiking by 5-HT was also observed in the presence of synaptic blockers (t 3 ϭ 4.8, p ϭ 0.02, n ϭ 4, paired t test). These data show a robust and repeatable 5-HT inhibition of L6 neurons by 5-HT with a component mediated by 5-HT 1A receptors. However, the continued suppression of action potential firing by 5-HT after blockade of 5-HT 1A receptors suggests the involvement of an additional 5-HT-mediated mechanism for inhibition of mPFC L6 pyramidal neurons.    5.4, p Ͻ 0.0001, n ϭ 21). I, Action potential firing was significantly affected by 5-HT (F (4,103) ϭ 37, p Ͻ 0.0001, one-way ANOVA). Post hoc analyses reveal that baseline firing was strongly suppressed by 5-HT (q ϭ 9, p Ͻ 0.0001, n ϭ 17, Dunnett's multiple comparison test) and remained suppressed by 5-HT in synaptic blockers (q ϭ 6.1, p Ͻ 0.0001, n ϭ 6). The suppression was not blocked by WAY100635 (q ϭ 8.8, p Ͻ 0.0001, n ϭ 16) and returned to baseline levels after washout of 5-HT (q ϭ 2.3, p Ͼ 0.05, n ϭ 36). To probe further the power of 5-HT 2A receptors to inhibit L6 pyramidal neurons in mPFC, we applied a potent 5-HT 2A agonist, TCB-2 (300 nM to 1 M). Here, we observed a strong inhibition of L6 neuronal excitability, with a significant right-shift of the input-output relationship (inhibitory effect of TCB-2: F (1,80) ϭ 24, p Ͻ 0.0001, n ϭ 11, repeated-measures two-way ANOVA; Fig. 2D). Pretreatment with MDL100907 abolished the inhibitory effect of TCB-2 (F (1,16) ϭ 1.2, p ϭ 0.3, n ϭ 3, repeated-measures two-way ANOVA). Taken together, our results suggest that 5-HT inhibition of mPFC L6 pyramidal neurons is mediated by a combination of 5-HT 1A and 5-HT 2A receptors acting in concert. However, substantial future work will be needed to elucidate the mechanisms by which these receptors individually and together work to suppress the excitability of L6 pyramidal neurons. Figure 3A    late tagged polysomes. Bound mRNAs were then purified and analyzed by qRT-PCR. These data are plotted in Figure 3C. There was a significant enrichment for the excitatory neuron marker, Slc17a7 (VGluT1), in the Epyc TRAP IP compared with whole PFC input. Two genes known to be expressed in L6 corticothalamic cells, Ntsr1  (Ferland et al., 2003), were also significantly enriched in IP samples. In contrast, genes that label inhibitory interneurons (Gad1), astrocytes (Aldh1l1), or oligodendrocytes (Cnp) were significantly depleted from the IPs. Taken together, these data suggest that Epyc-Cre labels a population of L6 corticothalamic pyramidal cells. Importantly, the qPCR also revealed that Syt6 was highly enriched in the Epyc cells, demonstrating that the Syt6-eGFP and Epyc-Cre mice label an overlapping population of neurons. In contrast, levels of the housekeeping gene Gapdh were found to be similar between the TRAP IPs and whole PFC input. Similar results were obtained for the 5-HT receptors, Htr1a and Htr2a, suggesting these genes are expressed but not enriched in the Epyc cells, which was not surprising given

Optogenetic activation of L6 pyramidal neurons is sensitive to serotonin
In electrophysiological experiments from Epyc-ChR2, we found that L6 pyramidal neurons, but not nonpyramidal neurons, were strongly depolarized upon light stimulation (473 nm, train of 2-ms-duration pulses at 20 Hz for 1 s), which was targeted to L6 mPFC with optic fiber (Fig.  4A, B). In contrast, prefrontal L6 neurons of littermate controls lacking channelrhodopsin did not respond to light stimulation. To verify that L6 pyramidal neurons were directly activated by light stimulation, we measured the kinetics of their light-evoked excitation. L6 pyramidal neurons rapidly responded to light (Ͻ1-ms latency to onset of excitation), consistent with direct activation through expressed channelrhodopsin (Ernst et al., 2008). This response rises to threshold, giving an action potential peak at 4.7 Ϯ 0.7 ms (time to L6 spike; n ϭ 5).

Optogenetic activation of L6 drives excitation of L5 interneurons
Because fast-spiking GABAergic neurons in mPFC are important to normal performance in attention tasks (Kim et al., 2016), we patched mPFC L5 interneurons in Epyc-ChR2 mice as potential downstream projection targets of L6 pyramidal neurons. We anticipated that light-mediated activation of L6 pyramidal neurons by targeted optic fiber would elicit postsynaptic responses in patched L5 interneurons. GABAergic interneurons were visually identified by their morphology and intrinsic properties, and their spiking patterns in response to depolarizing current steps were documented. This experimental protocol yielded two distinct populations of interneurons: FS cells with characteristic action potential firing Ͼ40 Hz and nFS cells that displayed lowthreshold firing characteristics. The intrinsic properties of these neurons are illustrated in Table 1 (Fig. 4C). Activation of both FS and nFS interneurons by optogenetic stimulation of L6 was substantially and significantly reduced by TTX (F (1,80) ϭ 19, p Ͻ 0.0001, twoway ANOVA). Together with the need for light activation over L6, it appears that channelrhodopsin is predominantly localized in the L6 pyramidal cell bodies and not in axon terminals impinging on the L5 interneurons.
Light stimulation over L6 elicited action potential firing in 100% of FS cells (Fig. 4D) and 70% of nFS cells. The firing pattern elicited in these two types of interneurons was different, with a greater number of spikes seen at the start of L6 stimulation in FS neurons and a more evenly distributed firing pattern observed in the nFS neurons (Fig.  5). The minimal L6 light to elicit a suprathreshold excitatory response did not differ significantly between FS and nFS L5 interneurons (t 14 ϭ 0.2, p ϭ 0.8, unpaired t test), despite a significant difference in input resistance (t 14 ϭ 4.3, p ϭ 0.0006, unpaired t test; Table 1). In response to maximal L6 light stimulation of L6, FS interneurons fired more action potentials than nFS interneurons (t 14 ϭ 4.4, p ϭ 0.0007, unpaired t test).

Serotonin suppresses L6 activation of L5 interneurons
Because optogenetic stimulation in L6 resulted in robust and highly stable excitation of interneurons in L5 that did not decrease over time at baseline conditions (t 11 ϭ 0.8, p ϭ 0.4, paired t test; Fig. 6A), it was straightforward to test the effect of 5-HT on this local circuit. We found that 5-HT strongly and significantly suppressed the number of action potentials elicited in L5 interneurons by optogenetic activation of L6 (FS cells: t 8 ϭ 3.8, p ϭ 0.005,  n ϭ 9, paired t test; nFS cells: t 6 ϭ 5.7, p ϭ 0.001, n ϭ 7,  paired t test; Fig. 6B). Of note, this suppression appeared to arise from 5-HT effects in L6, since interneurons in L5 showed minimal direct responses to 5-HT at -75 mV (2.7   : t 7 ϭ 1.9, p ϭ 0.1, unpaired t test). 5-HT 2A antagonists. Taken together, these results suggest that 5-HT 1A and 5-HT 2A receptors mediate a strong inhibitory drive in L6 that can suppress its local activation of cortical targets in L5, which others have shown to be critical to attention (Kim et al., 2016).

Prefrontal L6 pyramidal neurons excite a diverse group of interneurons in L5
We found that L6

Serotonergic inhibition of this L6-to-L5 intracortical circuit
We found that 5-HT, via stimulation of 5-HT 1A and 5-HT 2A receptors, inhibited L6 pyramidal neurons and their activation of L5 interneurons. These two receptors show substantial colocalization in L6 pyramidal neurons in mPFC of mouse (Table 3 in  Complex and carefully controlled future work will be necessary to identify the mechanisms underlying the 5-HT 2A receptor-mediated inhibition of L6 pyramidal neuron excitability. In investigating FS and nFS interneurons in L5, we found that the majority of these cells did not respond strongly to 5-HT. A minority showed electrophysiological responses (FS, 4/13; nFS, 2/16), predominantly inward currents (less than -20 pA) that were insufficient to elicit spiking. Taken together, our data support the hypothesis that the combined activation of both 5-HT 1A and 5-HT 2A receptors can inhibit neuronal excitability in L6 neurons of prefrontal cortex. However, further pharmacological work is required to examine the specific downstream mechanisms underlying this inhibition of L6 pyramidal neurons. Furthermore, additional investigations into the consequences of serotonergic inhibition of L6 on local network dynamics will provide more insight into the nature of these important associative circuits and how they control attention.

Serotonin, prefrontal attention circuitry, and attention deficits in psychiatric illness
Prefrontal attention circuitry is complex, and attentional performance can be perturbed by extremes of mPFC activity in either direction ( Our study is the first to demonstrate the strong inhibitory modulation exerted on L6 by 5-HT and the resultant decrease in its ability to stimulate interneuron activity in L5. Taken together, this evidence shows that L6 of mPFC is a candidate locus of action for the modulatory effects of 5-HT on attention. Based on recent work in nonhuman primates (Watson et al., 2015), it is tempting to speculate that 5-HT levels in deep mPFC may modulate the balance between social vigilance and attentional task performance, a phenomenon that is impaired in several neuropsychiatric illnesses.