Dynorphin inhibits basal forebrain cholinergic neurons by pre‐ and postsynaptic mechanisms

Key points

• The basal forebrain is an important component of the ascending arousal system and may be a key site through which the orexin neurons promote arousal.

• It has long been known that orexin‐A and ‐B excite basal forebrain cholinergic neurons, but orexin‐producing neurons also make the inhibitory peptide dynorphin.

• Using whole‐cell recordings in brain slices, we found that dynorphin‐A directly inhibits basal forebrain cholinergic neurons via κ‐opioid receptors, and decreases afferent excitatory synaptic input to these neurons.

• While the effects of dynorphin‐A and orexin‐A desensitize over multiple applications, co‐application of dynorphin‐A and orexin‐A produces a sustained response that reverses depending on the membrane potential of basal forebrain cholinergic neurons. At −40 mV the net effect of the co‐application is inhibition by dynorphin‐A, whereas at −70 mV the excitatory response to orexin‐A prevails.

Abstract

The basal forebrain (BF) is an essential component of the ascending arousal systems and may be a key site through which the orexin (also known as hypocretin) neurons drive arousal and promote the maintenance of normal wakefulness. All orexin neurons also make dynorphin, and nearly all brain regions innervated by the orexin neurons express kappa opiate receptors, the main receptor for dynorphin. This is remarkable because orexin excites target neurons including BF neurons, but dynorphin has inhibitory effects. We identified the sources of dynorphin input to the magnocellular preoptic nucleus and substantia innominata (MCPO/SI) in mice and determined the effects of dynorphin‐A on MCPO/SI cholinergic neurons using patch‐clamp recordings in brain slices. We found that the orexin neurons are the main source of dynorphin input to the MCPO/SI region, and dynorphin‐A inhibits MCPO/SI cholinergic neurons through κ‐opioid receptors by (1) activation of a G protein‐coupled inwardly rectifying potassium current, (2) inhibition of a voltage‐gated Ca²⁺ current and (3) presynaptic depression of the glutamatergic input to these neurons. The responses both to dynorphin‐A and to orexin‐A desensitize, but co‐application of dynorphin‐A and orexin‐A produces a sustained response. In addition, the polarity of the response to the co‐application depends on the membrane potential of BF neurons; at −40 mV the net effect of the co‐application is inhibition by dynorphin‐A, whereas at −70 mV the excitatory response to orexin‐A prevails. This suggests that depending on their state of activation, BF cholinergic neurons can be excited or inhibited by signals from the orexin neurons.

Key points

• The basal forebrain is an important component of the ascending arousal system and may be a key site through which the orexin neurons promote arousal.

• It has long been known that orexin‐A and ‐B excite basal forebrain cholinergic neurons, but orexin‐producing neurons also make the inhibitory peptide dynorphin.

• Using whole‐cell recordings in brain slices, we found that dynorphin‐A directly inhibits basal forebrain cholinergic neurons via κ‐opioid receptors, and decreases afferent excitatory synaptic input to these neurons.

• While the effects of dynorphin‐A and orexin‐A desensitize over multiple applications, co‐application of dynorphin‐A and orexin‐A produces a sustained response that reverses depending on the membrane potential of basal forebrain cholinergic neurons. At −40 mV the net effect of the co‐application is inhibition by dynorphin‐A, whereas at −70 mV the excitatory response to orexin‐A prevails.

Abbreviations

ACSF

artificial cerebrospinal fluid

BF

basal forebrain

CTB

cholera toxin subunit B

Cy3‐p75^NTR‐IgG

indocarbocyanine‐coupled antibodies against p75 neurotrophin receptors

DynA

dynorphin‐A

GFP

green fluorescent protein

GIRK current or channels

G protein‐coupled inwardly rectifying potassium current or channels

I_DynA

dynorphin‐A mediated current

IR‐DIC

infrared differential interference contrast

K‐S test

Kolmogorov–Smirnov test

MCH

melanin‐concentrating hormone

MCPO

magnocellular preoptic nucleus

NMDG

N‐methyl‐d‐glucamine

Nor‐BNI

nor‐binaltorphimine

OxA

orexin‐A

Ox1R

orexin 1 receptor

Ox2R

orexin 2 receptor

s/mEPSC

spontaneous and miniature excitatory postsynaptic currents

s/mIPSC

spontaneous and miniature inhibitory postsynaptic currents

SI

substantia innominata

TPQ

tertiapin‐Q

Introduction

The basal forebrain (BF) is a key extrathalamic relay to the cerebral cortex and is essential for promoting wakefulness in conjunction with monoaminergic and cholinergic signals from more caudal regions (Szymusiak, 1995; Zaborszky & Duque, 2003; Jones, 2004). The BF contains a heterogeneous population of neurons including cholinergic, GABAergic, glutamatergic and peptidergic neurons as well as interneurons (Gritti et al. 1997; Hur & Zaborszky, 2005). BF cholinergic neurons are the major source of cholinergic innervation of the cortex (Saper, 1984). They fire in association with cortical activation during wakefulness and rapid eye movement sleep, and their firing rate positively correlates with gamma EEG activity across the sleep–wake cycle (Lee et al. 2005 a). Importantly, activation of BF cholinergic neurons is sufficient to promote cortical activation and facilitates state transitions to wakefulness (Cape et al. 2000; Han et al. 2014; Irmak & de Lecea, 2014).

A growing body of evidence suggests that the orexin neurons (also known as hypocretin neurons) are a key driver of BF activity (Fadel & Frederick‐Duus, 2008; Jones, 2008; Arrigoni et al. 2010; Saper et al. 2010). The orexin neurons are located in the perifornical area of the lateral hypothalamus, they heavily innervate the BF and they appear to synapse upon cholinergic neurons, suggesting direct innervation (Wu et al. 2004; Espana et al. 2005; Fadel et al. 2005). Optogenetic activation of the orexin neurons triggers awakenings from sleep, but this effect is not completely abolished in orexin knockout mice, indicating that additional neurotransmitters released by the orexin neurons may contribute (Adamantidis et al. 2007). The orexin neurons release glutamate, and some postsynaptic effects are driven through glutamate receptors (Sears et al. 2013; Schone et al. 2014). In addition, orexin neurons produce and co‐release the endogenous opiate dynorphin (Chou et al. 2001; Torrealba et al. 2003; Crocker et al. 2005; Muschamp et al. 2014). The BF and nearly all brain regions innervated by the orexin neurons express κ opiate receptors, the main receptor for dynorphin (DePaoli et al. 1994; Mansour et al. 1994; Marcus et al. 2001). This is puzzling because orexin‐A (OxA), orexin‐B and glutamate excite their target neurons, but dynorphin has inhibitory effects (Arrigoni et al. 2010). In the BF, OxA excites the cholinergic neurons (Eggermann et al. 2001; Wu et al. 2004; Arrigoni et al. 2010), but little is known about the response to dynorphin. We therefore mapped the source of dynorphin input to the magnocellular preoptic nucleus and substantia innominata (MCPO/SI), and we examined the electrophysiological effects of dynorphin‐A (DynA) upon the cholinergic neurons of the MCPO/SI region.

Methods

Animal care and ethical approval

For electrophysiological recordings, we used 126 8‐week‐old C57BL/6 mice (27–30 g) (The Jackson Laboratory, Bar Harbor, ME, USA). For retrograde tracing studies, we used three C57BL/6 wild type mice, seven pro‐dynorphin‐IRES‐Cre (Pdyn‐ires‐Cre); Z/EG mice (Pdyn‐ires‐Cre; Z/EG), and three Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice in which green fluorescent protein (GFP) is produced in dynorphin‐expressing neurons. These mice were produced by crossing Pdyn‐ires‐Cre mice (Krashes et al. 2014) with two Cre‐dependent GFP reporter mouse lines: Z/EG; JAX 004178 (Novak et al. 2000) and R26‐loxSTOPlox‐L10‐GFP (Krashes et al. 2014). We counted neurons in four Pdyn‐ires‐Cre; Z/EG mice that strongly expressed GFP in nearly all dynorphin neurons, and we excluded two other mice with globally low levels of GFP expression. We housed all mice in a pathogen‐free animal research facility maintained on a 12:12 h light–dark cycle (lights on at 07.00 h) at 22 °C ambient temperature and with ad libitum access to food and water. Care of the mice met the National Institutes of Health standards, as set forth in the Guide for the Care and Use of Laboratory Animals, and all protocols were approved by the Beth Israel Deaconess Medical Centre Institutional Animal Care and Use Committee.

Pre‐labelling of BF cholinergic neurons

We anaesthetized C57BL/6 mice with isoflurane and injected indocarbocyanine (Cy3)‐coupled antibodies raised against mouse p75 neurotrophin receptors (Cy3‐p75^NTR‐IgG) (Advanced Targeting Systems, San Diego, CA, USA) into the lateral ventricle. We injected the Cy3‐p75^NTR‐IgGs (300 nl, 0.4 mg ml⁻¹, intracerebroventricular) over 5 min using a silane‐coated glass micropipette (25–30 μm tip diameter). Stereotaxic coordinates for the left lateral ventricle were: AP = −0.34 mm, DV = −2 mm, ML = −1 mm. One to three days after surgery, we prepared brain slices from the injected mice for in vitro recordings. We found many Cy3‐p75^NTR‐IgG‐labelled neurons ipsilateral and contralateral to the injection, and we recorded neurons from either side of the BF.

Slice preparation and whole‐cell patch‐clamp recordings

To prepare brain slices, we anaesthetized mice with an intraperitoneal (ip) injection of ketamine 150 mg kg⁻¹ plus xylazine 15 mg kg⁻¹ and then transcardially perfused them with ice‐cold artificial cerebrospinal fluid (ACSF; N‐methyl‐d‐glucamine, NMDG‐based solution), followed by decapitation. Using a vibrating microtome (VT1000, Leica, Bannockburn, IL, USA), we cut coronal brain slices (250 μm thick) in ice‐cold ACSF (NMDG‐based). We incubated the slices containing the BF for 20 min at 34°C in the NMDG‐based ACSF, and then we transferred them into a holding chamber (at room temperature) in normal ACSF (Na‐based) until brought to the recording chamber. For recordings, we submerged and perfused the slices with normal ACSF (2 ml min⁻¹). We recorded only Cy3‐p75^NTR‐IgG‐labelled neurons in the MCPO/SI region. To guide our recordings, we used fluorescence and infrared differential interference contrast (IR‐DIC) video microscopy using a fixed stage upright microscope (BX51WI, Olympus America Inc.) equipped with a Nomarski water immersion lens (40×/0.8 W) and IR‐sensitive CCD camera (ORCA‐ER, Hamamatsu, Bridgewater, NJ, USA). Images were displayed on a computer screen in real time using AxioVision software (Carl Zeiss MicroImaging). We recorded neurons at room temperature in whole‐cell configuration using a Multiclamp 700B amplifier (Molecular Devices, Foster City, CA, USA), a Digidata 1322A interface, and Clampex 9.0 software (Molecular Devices). We monitored the series resistance at regular intervals, and we discarded the data if neurons showed an unstable resting membrane potential or if the series resistance changed by more than 25%.

Reagents and solutions

The composition of the NMDG‐based ACSF for slice preparation was (in mm): 100 NMDG‐Cl, 2.5 KCl, 20 Hepes, 1.24 NaH₂PO₄, 30 NaHCO₃, 25 glucose, 2 thiourea, 5 sodium ascorbate, 3 sodium pyruvate, 0.5 CaCl₂, 10 MgSO₄ (pH 7.3 with HCl when carbogenated with 95% O₂ and 5% CO₂). The composition of normal ACSF (Na‐based) for recordings was (in mm): 120 NaCl, 1.3 MgCl₂, 2.5 KCl, 1.3 MgSO₄, 10 glucose, 26 NaHCO₃, 1.24 NaH₂PO₄, 2 thiourea, 1 sodium ascorbate, 3 sodium pyruvate, 4 CaCl₂ (pH 7.4 when carbogenated with 95% O₂ and 5% CO₂, 310–320 mOsm). The composition of the pipette solution used for current‐clamp recordings and for most of the voltage‐clamp recordings unless otherwise specified was (in mm): 120 potassium gluconate, 10 KCl, 3 MgCl₂, 10 Hepes, 2.5 K‐ATP, 0.5 Na‐GTP (pH 7.2 adjusted with KOH; 280 mOsm). We recorded the Ca²⁺ currents in a modified ACSF in which Ca²⁺ is replaced by Ba²⁺ for which the Ca²⁺ channels have higher conductance and to reduce Ca²⁺‐mediated inactivation of the Ca²⁺ channels. The composition of the ACSF for Ca²⁺ current recording was (in mm): 105 NaCl, 40 TEA‐Cl, 2.5 KCl, 2 MgCl, 10 glucose, 10 Hepes, 5 BaCl₂ (pH 7.4 when carbogenated with 95% O₂ and 5% CO₂; 310–320 mOsm). The composition of the pipette solution used for Ca²⁺ current recording was (in mm): 120 CsCl, 5 EGTA, 1 MgCl₂, 10 Hepes, 4 Mg‐ATP and 0.5 Na‐GTP (pH adjusted to 7.3 with CsOH; 280 mOsm). We recorded spontaneous and miniature excitatory postsynaptic currents (s/mEPSCs) in ACSF, in 10 μm bicuculline, at a holding potential (V _h) = –70 mV, and with the potassium gluconate‐based solution in the recording pipette. We recorded spontaneous and miniature inhibitory postsynaptic currents (s/mIPSCs) in ACSF at V _h = −5 mV with a caesium methane sulfonate‐based pipette solution (Cl^– reversal potential = −64 mV). The composition of the pipette solution used for s/mIPSCs recording was (in mm): 125 caesium methane sulfonate, 11 KCl, 10 Hepes, 0.1 CaCl₂, 1 EGTA, 5 Mg‐ATP and 0.3 Na‐GTP (pH adjusted to 7.2 with CsOH; 280 mOsm). In the recording of s/mEPSCs and s/m/IPSCs we applied DynA for 10 min. In all the experiments, we added Lucifer yellow CH‐ammonium salt (0.1%) to the pipette solution to mark the recorded cells and to verify that recorded neurons were double‐labelled with Lucifer yellow and Cy3‐p75^NTR‐IgG. We purchased bicuculline methiodide, TTX, Tertiapin‐Q, nor‐binaltorphimine (nor‐BNI) and DynA, CTAP, SDM25N, BRL‐52537 from Tocris Bioscience (Ellisville, MO, USA). We purchased OxA from Tocris Bioscience and from American Peptides (Vista, CA, USA). We purchased all the other reagents from Sigma‐Aldrich (St Louis, MO, USA). We dissolved bicuculline methiodide in DMSO. The final concentration of DMSO in the ACSF was <0.1%. We dissolved all the other drugs in water.

Data analysis and statistics

We analysed all our data using Clampfit 9.2 (Molecular Devices) and IGOR Pro 6 (WaveMetrics, Lake Oswego, OR, USA). We measured action potential firing frequency in 1 min bin intervals. The control point is taken as the 1 min bin just prior to application of DynA and the response of DynA is measured during the 1 min bin beginning 2 min into DynA application. We analysed s/mEPSCs and s/mIPSCs off‐line using Mini Analysis 6 (Synaptosoft, Leonia, NJ, USA), and we measured the frequency and amplitude of s/mEPSCs and s/mIPSCs during 5 min in control, just before the application of DynA, during the last 5 min of 10 min DynA application, and during the last 5 min of a 15 min washout. We statistically compared s/mEPSC and s/mIPSC inter‐event interval and amplitude cumulative distributions using the non‐parametric Kolmogorov–Smirnov test (K‐S test). Statistical analyses were performed using StatView (SAS Institute, Cary, NC, USA) or Prism 6 (GraphPad Software, La Jolla, CA, USA). We normalized all our values by dividing the values of control and treatment samples by the mean of the control sample. Such normalization conserves the distribution and the relative variance of the samples, allowing the subsequent use of a t test. We used paired t‐tests or one‐way ANOVA (with repeated measures) followed by Fisher's PLSD tests for statistical analysis. An α < 0.05 was considered significant. Results are expressed as mean ± SEM and, unless otherwise specified, n refers to the number of cells.

Retrograde tracing and histology

We used four Pdyn‐ires‐Cre; Z/EG and three Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice as the Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice produces more intense soma labelling than the Pdyn‐ires‐Cre; Z/EG mice for the retrograde tracing studies. We also used thee C57BL/6 wild‐type mice for the double labelling of cell bodies and fibres for dynorphin and orexin peptides. We anaesthetized the Pdyn‐ires‐Cre; Z/EG and the Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice with ketamine (100 mg kg⁻¹, ip) and xylazine (10 mg kg⁻¹, ip) and stereotaxically unilaterally microinjected 6–9 nl of the retrograde tracer cholera toxin subunit B (CTB; List Biological, Campbell, CA, USA; 0.2% in saline) into the MCPO/SI (AP: −0.46 mm, DV: 5.3 mm, ML: −1.2 mm) and these animals were then killed 1 week later. We deeply anaesthetized all mice with ketamine (150 mg kg⁻¹, ip) and xylazine (15 mg kg⁻¹, ip) and transcardially perfused them with 50 ml of buffered 10% formalin (pH 7.0; Fisher Scientific, Fair Lawn, NJ, USA). We postfixed the brains for 12 h in formalin and cryoprotected them in a 20% sucrose solution overnight. We later coronally sectioned the brains at 30 μm into a 1:3 series on a freezing microtome.

For the retrograde tracing study, we first surveyed the entire brain using light microscopy. We incubated sections overnight in goat anti‐CTB primary antiserum (1:10,000; List Biological Laboratories; catalogue no. 703; lot no. 7032A7) and rabbit anti‐GFP (1:10,000; Invitrogen, Grand Island, NY, USA; catalogue no. A6455; lot no. 622086). We then incubated sections for 1 h in donkey anti‐goat biotinylated secondary antiserum (1:500; Jackson ImmunoResearch, West Grove, PA, USA; catalogue no. 705‐065‐147) followed by 1 h in avidin–biotin complex (Vectastain ABC Elite Kit; Vector Laboratories, Burlingame, CA, USA). We labelled CTB black using nickel‐DAB. We then stained GFP red by incubating for 1 h in donkey anti‐rabbit biotinylated secondary antiserum (1:500; Jackson ImmunoResearch; catalogue no. 711‐065‐152) followed by staining with alkaline phosphatase red (Vector Red Alkaline Phosphatase Substrate Kit; Vector Laboratories).

We next examined the areas of the brain containing CTB and/or GFP for co‐localization using double immunofluorescence staining. After overnight incubation in CTB and GFP primary antisera, we reacted sections for 1 h in donkey anti‐goat IgG conjugated to Alexa Fluor 555 (1:500; Invitrogen; catalogue no. A21432) and donkey anti‐rabbit IgG conjugated to Alexa Fluor 488 (1:500; Invitrogen; catalogue no. A21206) to label CTB‐containing cells red and GFP‐containing cells green. We imaged the tissue with an Olympus slide scanner VS120 and counted double‐labelled cells in each section with OlyVIA software. We then photographed double‐labelled sections using a Zeiss LSM 510 META confocal microscope in 1 μm optical sections. We analysed image z‐stacks using ImageJ (NIH, Bethesda, MD, USA). We calculated the percentage of double labelled neurons for CTB and GFP in the lateral hypothalamus based on cell counting in a 1:4 series (30 μm sections). The average number of double labelled neurons was 20 ± 1 in Pdyn‐ires‐Cre; Z/EG mice (n = 4 animals) and 19 ± 3 in Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice (n = 3 animals).

To examine if the retrograde labelling in the lateral hypothalamus was within orexin neurons, we examined native GFP fluorescence in sections additionally labelled for CTB and orexin using the goat anti‐CTB primary antiserum and a rabbit anti‐OxA primary antiserum (1:5000; Phoenix Pharmaceuticals, Burlingame, CA, USA; catalogue no. G‐003‐30; lot no. 01282‐3). We then incubated tissue for 1 h in donkey anti‐goat IgG conjugated to Alexa Fluor 555 and donkey anti‐rabbit IgG biotinylated secondary (1:500; Jackson ImmunoResearch) followed by 1 h in streptavidin conjugated to Cy5 (1:1000; Invitrogen; catalogue no. SA1011).

To double immunolabel DynA and OxA fibres in the MCPO/SI region, we incubated tissue from a wild‐type mouse overnight in rabbit anti‐DynA primary antisera (1:2000; Peninsula Laboratories; catalogue no. T‐4268; lot no. 061353) and goat anti‐OxA (1:5000; Santa Cruz; catalogue no. sc‐8070; lot no. C0512) and followed by 1 h in donkey anti‐goat IgG conjugated to Alexa Fluor 555 and donkey anti‐rabbit IgG conjugated to Alexa Fluor 488.

To immunolabel DynA cell bodies in the lateral hypothalamus, we treated two C57BL/6 wild‐type mice with colchicine. We sterotaxically injected colchicine (990–1000 μl; 20 μg μl⁻¹ in saline; Sigma, catalogue no. C9754) into the lateral ventricle (AP: 0.2 mm, DV: −2.0 mm, ML: 0.9 mm) and they were then killed 42 h later as described above. To double label neurons of the lateral hypothalamus for DynA and OxA immunoreactivities, we incubated hypothalamic sections overnight in rabbit anti‐DynA (1:2000; Peninsula Laboratories) and goat anti‐orexin (1:5000; Santa Cruz) primary antisera, followed by 1 h in donkey anti‐rabbit IgG conjugated to Alexa Fluor 555 (1:1000; Invitrogen; catalogue no. A31572) and donkey anti‐goat IgG conjugated to Alexa Fluor 488 (1:1000; Invitrogen; catalogue no. A11055).

Results

Source of dynorphin inputs to MCPO/SI neurons

To identify the dynorphin inputs to the BF, we microinjected the retrograde tracer CTB into the MCPO/SI of Pdyn‐ires‐Cre; Z/EG (n = 4) mice and Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP (n = 3) mice (Fig. 1 A and B). We first surveyed the entire brain using light microscopy to map the distribution of retrogradely labelled neurons and neurons expressing GFP. We then used double fluorescence immunolabelling to map retrogradely labelled neurons containing both GFP and CTB.

Figure 1. Sources of dynorphin inputs to the MCPO .

A, a typical microinjection of CTB (red) into the MCPO of a Pdyn‐ires‐Cre; Z/EG mouse. The caudate and putamen (CPu) contain many dynorphin neurons expressing GFP immunoreactivity (green), and GFP‐labelled fibres innervate the ventral pallidum (VP). Other abbreviations: ac, anterior commissure; 3V, third ventricle. B, map of the CTB injection sites in Pdyn‐ires‐Cre; Z/EG mice (n = 4; coloured lines) and in Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice (n = 3; dashed black lines) overlaid on a map of cholinergic neurons immunolabelled for choline acetyltranferase. C, the perifornical region of the lateral hypothalamus contains many GFP‐labelled neurons (green; Pdyn‐ires‐Cre; Z/EG mouse) indicative of dynorphin and many of these are retrogradely labelled with CTB (red). Arrows indicate three double‐labelled neurons (yellow). D, confocal photomicrographs showing orexin immunoreactivity (magenta), dynorphin neurons expressing GFP immunoreactivity (green; Pdyn‐ires‐Cre; Z/EG mouse) and CTB immunoreactivity (red). White arrowheads mark co‐localization of the dynorphin neurons (GFP‐positive) with orexin immunoreactivity and asterisks mark co‐localization of GFP, orexin and CTB immunoreactivities, indicating that orexin neurons expressing dynorphin project to the MCPO/SI region. E, immunolabeled OxA (in green) and DynA (in red; in a C57BL/6 wild‐type mouse) cell bodies in the lateral hypothalamus. F, immunolabeled OxA (in red) and DynA (in green; in a C57BL/6 wild‐type mouse) fibres in the MCPO region. Most of the dynorphin fibres are double labelled for orexin (yellow), indicating that the orexin input to the BF contains dynorphin. Scale bars = 500 μm in A and B, 50 μm in C, 25 μm in D, 50 μm in E, and 20 μm in F.

Only the lateral hypothalamus contained a substantial number of double‐labelled neurons (Fig. 1 C). Specifically, the orexin/dynorphin neuron field contained 52.6% of all retrogradely labelled neurons that also contained GFP in the Pdyn‐ires‐Cre; Z/EG (n = 4) mice and 39% of double labelled neurons in the Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP (n = 3) mice. In addition, in both GFP mouse lines, all of the neurons that contained CTB and GFP in the orexin/dynorphin neuron field also contained orexin (Fig. 1 D), indicating that the orexin/dynorphin neurons are the main source of dynorphin input to the MCPO/SI.

Within the lateral hypothalamus/orexin field of Pdyn‐ires‐Cre; Z/EG mice we found that 19.2 ± 0.2% of the GFP neurons also contained CTB. In addition, 40.0 ± 2.3% of the CTB‐containing neurons expressed GFP, indicating that non‐orexin neurons of this region also innervate the BF.

To test if these results were unique to the Pdyn‐ires‐Cre; Z/EG mice, we also counted the number of double labelled cells in the Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice. The Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP mice had more GFP soma labelled in the lateral hypothalamus and throughout the brain. Within the lateral hypothalamus/orexin field, 10.6 ± 0.7% of the GFP neurons also contained CTB. In this region, 30.2 ± 4.4% of CTB‐containing neurons expressed GFP (n = 3 mice).

Neurons double‐labelled for GFP and CTB were scattered in other brain regions, including the medial prefrontal cortex, olfactory tubercle, cortical amygdala areas, ventromedial hypothalamus and subparafasicular area of the thalamus. However, the number of double‐labelled cells in these regions was low, and none of these regions accounted for more than a few per cent of all double‐labelled cells (Table 1).

Table 1.

Sources of dynorphin input to the MCPO

			Pdyn‐ires‐Cre; R26‐loxSTOPlox‐
	Pdyn‐ires‐Cre; Z/EG mice		L10‐GFP mice
	(n = 4)		(n = 3)
Medial orbital cortex	2 ± 0.5	5.3%	6 ± 0.7	12.2%
Prelimbic cortex	0 ± 0.0	0%	2 ± 2.8	4.1%
Dorsal tenia tecta	0 ± 0.0	0%	2 ± 0.0	4.1%
Olfactory tubercle	3 ± 1.7	7.9%	3 ± 4.2	6.1%
Infralimbic cortex	0 ± 0.0	0%	3 ± 0.7	6.1%
Accumbens n. shell	3 ± 0.0	7.9%	1 ± 1.4	2.0%
Striatum	0 ± 0.0	0%	0 ± 0.0	0%
Lateral septum	2 ± 2.5	5.3%	1 ± 0.7	2.0%
IPAC	1 ± 0.6	2.6%	1 ± 0.7	2.0%
Cortical amygdaloid n.	2 ± 0.5	5.3%	4 ± 4.2	8.2%
Central amygdala n.	0 ± 0.0	0%	1 ± 1.4	2.0%
Lateral hypothalamus	20 ± 0.6	52.6%	19 ± 2.7	38.9%
Supraoptic n.	0 ± 0.0	0%	0 ± 0.0	0%
Ventromedial hypothalamic n.	4 ± 1.4	10.5%	4 ± 4.9	8.2%
Dorsomedial hypothalamic n.	0 ± 0.0	0%	0 ± 0.0	0%
Parafasicular thalamic n.	1 ± 0.6	2.6%	2 ± 0.7	4.1%
Parabrachial n.	0 ± 0.0	0%	0 ± 0.0	0%

Average number of neurons immunolabeled for GFP and CTB across the entire brain (± SD) in two GFP reporter mouse lines: Pdyn‐ires‐Cre; Z/EG and Pdyn‐ires‐Cre; R26‐loxSTOPlox‐L10‐GFP. IPAC, interstitial nucleus of the posterior limb of the anterior commissure.

We closely examined many other brain regions known to contain dynorphin‐producing neurons that might innervate the BF. The parabrachial nucleus and pedunculopontine nucleus contained neurons singly labelled for GFP or CTB, but we found no double‐labelled cells. Dynorphin‐containing neurons were abundant in the central nucleus of the amygdala, accumbens nucleus, caudate/putamen, dorsomedial hypothalamus and supraoptic nucleus, but these areas contained few retrogradely labelled neurons and almost none was double‐labelled (Table 1). To confirm that the dynorphin fibres in the BF contained orexin, we double labelled BF sections for OxA and DynA in a wild‐type mouse (Fig. 1 E and F). In the MCPO/SI we observed that most of the dynorphin fibres were double labelled for orexin, indicating that the orexin input to BF also contains and possibly releases dynorphin. There were a few dynorphin‐labelled fibres that did not contain orexin that probably originated in other dynorphin‐containing brain regions that innervate the MCPO/SI.

DynA directly inhibits MCPO/SI cholinergic neurons

We next studied the effect of DynA on MCPO/SI cholinergic neurons using in vitro brain slices. We selectively labelled these neurons in vivo using fluorescent antibodies against the neurotrophin receptor p75 receptor (Cy3‐p75^NTR‐IgG) as described in our previous studies (Hawryluk et al. 2012) (Fig. 2 A). In the BF, Cy3‐p75^NTR‐IgG is internalized selectively by cholinergic neurons and thus is a useful tool to label BF cholinergic neurons for in vitro electrophysiological recordings. Consistent with prior studies (Hawryluk et al. 2012), all the Cy3‐p75^NTR‐IgG‐labelled neurons recorded in the MCPO/SI region exhibited the electrophysiological fingerprint of BF cholinergic neurons, which included a distinctive delayed rebound firing on recovery from hyperpolarizing current pulses mediated by the A‐type current, the presence of inward rectification and lack of depolarizing sag in response to hyperpolarizing pulses (Fig. 2 B).

Figure 2. DynA inhibits MCPO/SI cholinergic neurons through κ opioid receptors .

A, in vivo labelling of MCPO/SI cholinergic neurons using the fluorescent Cy3‐p75^NTR‐IgG (left), and the same neurons visualized under IR‐DIC (right) (scale bar = 20 μm). B, in response to hyperpolarizing current pulses, MCPO/SI cholinergic neurons show no I _h‐mediated depolarizing sag, a voltage‐dependent inward rectification and delayed rebound firing in response to the repolarization to resting membrane potential after hyperpolarizing pulses (arrow). C, DynA (1 μm) inhibits the firing of MCPO/SI cholinergic neurons, and this effect is blocked by the κ opioid receptor antagonist nor‐BNI (1 μm). D, in voltage‐clamp recordings, DynA activates a steady state outward current that is abolished in the presence of nor‐BNI. E, the κ‐opioid receptor agonist BRL52537 (1 μm) mimics the effect of DynA. F, the μ‐ and δ‐opioid receptor antagonists CTAP (1 μm) + SDM25N (500 nm) do not block DynA responses. G, nor‐BNI (1 μm) applied alone has no effect on MCPO/SI cholinergic neurons. H, summary bar chart of the mean evoked current in response to DynA (n = 17), BRL52537 (n = 6), DynA in the presence of CTAP + SDM25N (DynA + CTAP + SDM25N; n = 5) or nor‐BNI (DynA + nor‐BNI; n = 8) and to nor‐BNI applied alone (n = 6). **P < 0.01 and *P< 0.05, paired t test, comparing holding currents before and during drug applications (DynA, BRL52537 and nor‐BNI). ^†† P < 0.01, one‐way ANOVA, F _4,37 = 11.91 and Fisher's PLSD, comparing all the groups vs. DynA. All voltage‐clamp recordings were conducted at a V _h = −70 mV and in TTX (1 μm).

DynA (1 μm) inhibited the great majority of the MCPO/SI cholinergic neurons (92% of 61 neurons tested). Only a small number of these neurons were unaffected (n = 5), and none was excited. In current‐clamp recordings, DynA produced membrane hyperpolarization (−8.27 ± 2.84 mV, n = 6; P = 0.015, paired t test, membrane potential in control vs. DynA) and depressed the firing of MCPO/SI cholinergic neurons (1.17 ± 0.24 Hz in control and 0.25 ± 0.11 Hz in DynA, n = 6, P = 0.005, paired t test, Fig. 2 C). These effects were mediated by κ‐opioid receptors as they were blocked by the specific κ‐opioid receptor antagonist nor‐binaltrophine (nor‐BNI, 1 μm). In the presence on nor‐BNI, DynA did not change the resting membrane potential or firing frequency of cholinergic MCPO/SI neurons (Fig. 2 C). Resting membrane potential was −46.06 ± 2.37 mV in nor‐BNI and −45.61 ± 2.27 mV in nor‐BNI + DynA (n = 6, P = 0.47, paired t test, and P = 0.02, unpaired t test when comparing membrane potential changes by DynA in control ACSF and by DynA in nor‐BNI). Firing frequency was 1.35 ± 0.15 Hz in nor‐BNI and 1.36 ± 0.18 Hz in nor‐BNI + DynA (n = 6, P = 0.85, paired t test, and P < 0.001, unpaired t test when comparing firing frequency changes by DynA in control ACSF and by DynA in nor‐BNI).

In the presence of TTX (1 μm), DynA hyperpolarized the membrane potential (−7.43 ± 1.27 mV, n = 6; P = 0.002, paired t test, resting membrane potential in control vs. DynA) and reduced the input resistance (control: 817.36 ± 104.85 MΩ; DynA: 661.72 ± 76.14 MΩ, n = 5; P = 0.016, paired t test, control vs. DynA), indicating a direct postsynaptic effect.

In voltage‐clamp recordings (V _h = −70 mV and in TTX, 1 μm), DynA evoked an outward current (Fig. 2 D and H; I _DynA 10.16 ± 1.07 pA, n = 17; P < 0.001, paired t test, control vs. DynA). This effect was mimicked by the κ‐opioid receptor agonist BRL52537 (1 μm; 8.00 ± 2.3 pA, n = 6; P = 0.018, paired t test, control vs. BRL52537; Fig. 2 E and H). In addition, the I _DynA was abolished by the κ‐opioid receptor antagonist nor‐BNI (1 μm; 0.13 ± 0.67 pA, n = 8; P = 0.842, paired t test, nor‐BNI vs. nor‐BNI + DynA; Fig. 2 D and H), but was unaffected by the μ‐ and δ‐opioid receptor antagonists CTAP (1 μm) and SDM25N (500 nm; 7.65 ± 2.6 pA, n = 5; P = 0.044, paired t test, CTPA + SDM25N vs. CTPA + SDM25N + DynA; no difference between I _{CTPA + SDM25N + DynA} and I _DynA, one‐way ANOVA and Fisher's PLSD, P = 0.25; Fig. 2 F and H), confirming that DynA directly inhibits MCPO/SI cholinergic neurons selectively through κ‐opioid receptors. Application of nor‐BNI alone did not change the holding currents (Fig. 2 G and H; n = 6; P = 0.96, paired t test, control vs. nor‐BNI), suggesting that inhibitory κ‐opioid tone is absent in MCPO/SI brain slices.

DynA inhibits MCPO/SI cholinergic neurons by activating a GIRK channel and inhibiting voltage‐dependent Ca²⁺ channels

To identify the ionic mechanisms that mediate the DynA inhibition of MCPO/SI cholinergic neurons, we performed a series of experiments using voltage ramp and voltage step protocols. DynA was previously shown to inhibit neuronal activity by activating a G protein‐coupled, inwardly rectifying K⁺ (GIRK) current and by inhibiting voltage‐gated Ca²⁺ channels (Li & van den Pol, 2006; Parsons & Hirasawa, 2011). To examine the involvement of GIRK channels, we used a voltage ramp protocol (from −140 to −20 mV, 0.3 mV ms⁻¹; Fig. 3 A) and tested whether I _DynA was: (1) affected by changes of extracellular K⁺ concentration, (2) blocked by BaCl₂ and (3) blocked by a specific GIRK current blocker, tertiapin‐Q (TPQ). The voltage ramp protocol showed that I _DynA inwardly rectified, reversed near the K⁺ equilibrium potential (−89.73 ± 3.51 mV, n = 5) and was blocked by BaCl₂ (100 μm; at −70 mV, n = 5; P = 0.73, paired t test, holding current in BaCl₂ vs. BaCl₂ + DynA; comparison of changes in holding currents by DynA in control ACSF vs. DynA in BaCl₂ are shown in Fig. 3 D), suggesting that inward rectifying K⁺ channels could mediate the effects of DynA (Fig. 3 B). The involvement of a K⁺ conductance was further confirmed by the rightward shift of the reversal potential of I _DynA in response to increased extracellular K⁺. Specifically, raising the external K⁺ concentration from 2.5 to 7.5 or 12.5 mm shifted the I _DynA reversal potential from −89.73 ± 3.51 mV (n = 5; in 2.5 mm K⁺) to −56.91 ± 2.16 mV (n = 7, in 7.5 mm K⁺) or to −46.93 ± 1.11 mV (n = 9, in 12.5 mm K⁺), respectively, as predicted by the Nernst equation for the K⁺ reversal potential, demonstrating that I _DynA was a K⁺ current (Fig. 3 B). Furthermore, I _DynA was blocked by pretreatment of the recording slices with the GIRK channel blocker TPQ (300 nm, n = 5; P = 0.29, paired t test, holding current in TPQ vs. TPQ + DynA, Fig. 3 C, comparison of changes in holding currents by DynA in control ACSF vs. DynA in TPQ are shown in Fig. 3 D).

Figure 3. DynA inhibits the MCPO/SI cholinergic neurons by activating GIRK channels and by inhibiting voltage dependent Ca²⁺ channels .

A, I–V relationship using voltage ramps (−140 to −20 mV in 416 ms) recorded before and during the application of DynA (1 μm). B, right shift of the reversal potential of the DynA‐mediated current (I _DynA = DynA – Con) in response to raised extracellular K⁺ concentrations (2.5, 7.5 and 12.5 mm) as predicted by the Nernst equation for the K⁺ reversal potential. I _DynA is absent in BaCl₂ (100 μm). C, the effect of DynA is also blocked by the GIRK channel blocker tertiapin‐Q (TPQ, 100 nm). D, summary bar chart of the mean I _DynA recorded in control ACSF (con; n = 17, data also represented in Fig. 2 H), in BaCl₂ (n = 5) and in TPQ (n = 5). **P < 0.01, paired t test, comparing holding currents before and during DynA applications. ^†† P < 0.01, one‐way ANOVA, F _2,24 = 18.46 and Fisher's PLSD, comparing I _DynA in BaCl₂ and in TPQ vs. I _DynA in control ACSF. E, DynA (1 μm) reduces the amplitude of voltage‐dependent Ca²⁺ currents (CdCl₂ 200 μm). Recordings of Ca²⁺ currents were done at a V _h = −80 mV (step to −20 mV) in a modified ACSF (Ca²⁺ is replaced by Ba²⁺), in TTX (1 μm) and with a CsCl‐based pipette solution.

We next tested the effects of DynA on voltage‐dependent Ca²⁺ currents. We recorded Ca²⁺ currents in modified ACSF (Ca²⁺ replaced by Ba²⁺), in the presence of TTX, using a caesium‐based pipette solution and using voltage depolarizing steps from −80 to −20 mV. DynA (1 μm) reduced the amplitude of voltage‐dependent Ca²⁺ currents to 79.48 ± 8.14% (Fig. 3 E), and this effect was reversible with washout (n = 5; repeated‐measures one‐way ANOVA and Fisher's PSLD, P = 0.017, control vs. DynA).

Together, these data demonstrate that DynA inhibits MCPO/SI cholinergic neurons by activating GIRK channels and by inhibiting voltage‐gated Ca²⁺ channels.

DynA inhibits glutamatergic input to MCPO/SI cholinergic neurons through presynaptic kappa opioid receptors

We next examined the effects of DynA on the glutamatergic input to MCPO/SI cholinergic neurons. DynA (1 μm) almost halved the frequency of the glutamatergic sEPSCs (Fig. 4 A), and this effect was reversed after 20 min washout (control: 4.32 ± 0.78 Hz; DynA: 2.6 ± 0.49 Hz; washout: 3.62 ± 0.84 Hz, n = 6; repeated‐measures one‐way ANOVA and Fisher's PLSD, P < 0.01, control vs. DynA and P = 0.011, washout vs. DynA). DynA had no effects on the amplitude of sEPSCs (Fig. 4 A; control: 8.47 ± 1.12 pA; DynA: 8.27 ± 1.10 pA; and washout: 7.73 ± 1.00 pA, n = 6; repeated‐measures one‐way ANOVA, F _2,10 = 2.37, P = 0.14).

Figure 4. DynA inhibits the glutamatergic input to MCPO/SI cholinergic neurons through presynaptic κ opioid receptors .

A, mean effect of DynA (1 μm) on sEPSC frequency (left graph; n = 6; **P < 0.01 and *P < 0.05, one‐way ANOVA, F _2,10 = 13.69 and Fisher's PLSD) and on sEPSC amplitude (right graph; n = 6; one‐way ANOVA, F _2,10 = 2.37, P = 0.144). B and C, DynA reduces mEPSC frequency without changing mEPSC amplitude. The effect of DynA on mEPSCs of a representative cholinergic neuron is shown in B (left; current traces) and the corresponding cumulative distribution plots of mEPSC inter‐event intervals (right graph; P < 0.01, control vs. DynA and P < 0.01 DynA vs. washout; K‐S tests; DynA: thicker line) and mEPSC amplitudes (C; P = 0.1, control vs. DynA and P = 0.06, DynA vs. washout, K‐S tests; DynA: thicker line). Inset: averaged mEPSCs (in control, and DynA; n = 200 synaptic events). D, DynA reduces mEPSC frequency (left; n = 7; **P < 0.01, one‐way ANOVA, F _2,12 = 14.35 and Fisher's PLSD) but does not affect mEPSC amplitude (n = 7; one‐way ANOVA, F _2,12 = 3.02, P = 0.92). E, DynA reduces mEPSC frequency for ∼30 min (n = 7). F and G, nor‐BNI (1 μm) blocks the effects of DynA on mEPSC frequency, indicating that DynA acts on presynaptic κ‐opioid receptors. The response of a representative cholinergic neuron to DynA in the presence of nor‐BNI is shown in F (current traces). G, the cumulative distribution plots of mEPSC inter‐event intervals (left; P = 0.6, nor‐BNI vs. nor‐BNI + DynA, K‐S test; nor‐BNI + DynA) and mEPSC amplitudes (right; P = 0.38, nor‐BNI vs. nor‐BNI + DynA, K‐S test; nor‐BNI + DynA: thicker line). Mean effect of DynA on mEPSC frequency (centre; n = 6, P = 0.72, paired t test) and mEPSC amplitude (insert right; n = 6, P = 0.72, paired t test) in the presence of nor‐BNI. H and I, tertiapin‐Q (TPQ; 300 nm) blocks the effects of DynA on mEPSC frequency, indicating that DynA acts on presynaptic terminals by the activation of GIRK channels. H (current traces), the response of a representative cholinergic neuron to DynA in the presence of TPQ. I, cumulative distribution plots of mEPSC inter‐event intervals (left; P = 0.70, TPQ vs. TPQ + DynA, K‐S test) and mEPSC amplitudes (right; P = 0.26, TPQ vs. TPQ + DynA, K‐S test; TPQ + DynA: thicker line). Mean effect of DynA on mEPSC frequency (centre; n = 5, P = 0.64, paired t test) and mEPSC amplitude (insert right; n = 5, P = 0.57, paired t test) in the presence of TPQ. sEPSCs and mEPSCs were recorded at a V _h = −70 mV and in bicuculline (10 μm), and mEPSCs were recorded in TTX (1 μm).

To determine whether DynA acted on presynaptic glutamatergic terminals, we tested its effects on glutamatergic mEPSCs. DynA decreased the frequency of glutamatergic mEPSCs by ∼40% (Fig. 4 B, D and E; control: 3.30 ± 0.94 Hz; DynA: 1.97 ± 0.6 Hz; and washout: 2.9 ± 0.8 Hz, n = 7; repeated‐measures one‐way ANOVA and Fisher's PLSD, P < 0.01, control vs. DynA and P < 0.01, DynA vs. washout), but DynA had no effect on the amplitude of mEPSCs (Fig. 4 C and D; control: 8.67 ± 0.55 pA, DynA: 8.62 ± 0.42 pA, and washout: 7.57 ± 0.35 pA, n = 7; repeated‐measures one‐way ANOVA, F _2,12 = 3.02, P = 0.08), consistent with a presynaptic effect. In addition, DynA did not affect the cumulative distribution plots of mEPSC amplitude of all the neurons tested (n = 7; P > 0.05, K‐S test). DynA did not affect the mEPSC rise time (mean mEPSC 10–90 rise time: control = 1.39 ± 0.06 ms and DynA = 1.47 ± 0.08 ms, n = 7; P = 0.19, paired t test) or decay time (mean mEPSC decay tau: control = 3.8 ± 0.3 ms and DynA = 4.120 ± 0.39 ms, n = 7; P = 0.24, paired t test), indicating that DynA did not target a select pool of mEPSCs.

Application of nor‐BNI (1 μm) completely blocked the effects of DynA on mEPSC frequency (Fig. 4 F and G; nor‐BNI: 5.9 ± 1.67 Hz and nor‐BNI + DynA: 5.86 ± 1.67 Hz, n = 5; P = 0.72, paired t test, and P = 0.008 unpaired t test when comparing changes in mEPSC frequency by DynA in control ACSF and by DynA in nor‐BNI), demonstrating that DynA reduces the glutamatergic input to the cholinergic MCPO/SI neurons through presynaptic κ opioid receptors. However, nor‐BNI applied alone did not change the mEPSC frequency (control: 2.78 ± 1.04 Hz; nor‐BNI: 2.72 ± 1.02 Hz, n = 6; P = 0.62, paired t test) or mEPSC amplitude (control: 9.72 ± 0.56 pA; nor‐BNI: 9.54 ± 0.59 pA, n = 6; P = 0.19, paired t test), indicating that the glutamatergic input was not under inhibitory tone by endogenous dynorphin.

Furthermore, application of TPQ (300 nM) blocked the effect of DynA on mEPSC frequency (Fig. 4 H and I; TPQ: 3.65 ± 0.50 Hz and TPQ + DynA: 3.57 ± 0.48 Hz, n = 5; P = 0.64, paired t test, and P = 0.017, unpaired t test when comparing changes in mEPSC frequency by DynA in control ACSF and by DynA in TPQ), demonstrating that DynA inhibits glutamate release through the activation of presynaptic GIRK channels.

DynA does not affect the GABAergic input to MCPO/SI cholinergic neurons

We also examined the effects of DynA on GABAergic input to MCPO/SI cholinergic neurons. We recorded GABAergic sIPSCs and mIPSCs at a holding potential of −5 mV using a caesium methane sulfonate‐based pipette solution. Both sIPSCs and mIPSCs were completely abolished by the GABA_A receptor antagonist bicuculline (10 μm; n = 5). Application of DynA (1 μm) had no effects on sIPSC frequency (control: 0.55 ± 0.08 Hz; DynA: 0.54 ± 0.08 Hz; and washout: 0.54 ± 0.08 Hz, n = 6; repeated‐measures ANOVA, F _2,10 = 0.09, P = 0.91) and sIPSC amplitude (control: 9.34 ± 0.84 pA; DynA: 9.33 ± 0.84 pA; and washout: 9.12 ± 0.74 pA, n = 6; repeated‐measures ANOVA, F _2,10 = 0.52, P = 0.61) (Fig. 5 A). In addition, DynA did not change mIPSC frequency (control: 0.82 ± 0.14 Hz; DynA: 0.89 ± 0.13 Hz; and washout: 0.9 ± 0.17 Hz, n = 6; repeated‐measures ANOVA, F _2,10 = 0.50, P = 0.62) or mIPSC amplitude (control: 9.57 ± 1 pA; DynA: 9.31 ± 0.84 pA; and washout: 9.59 ± 0.82 pA, n = 6; repeated‐measures ANOVA, F _2,10 = 0.35, P = 0.71) (Fig. 5 B). No neurons showed significant changes in the cumulative distribution plots of mIPSC frequency (n = 6; P > 0.05, K‐S test) or amplitude (n = 6; P > 0.05, K‐S test), or in mIPSC rise time (mean mIPSC 10–90 rise time: control = 4.66 ± 0.54 ms and DynA = 4.96 ± 0.52 ms, n = 6; P = 0.52, paired t test) and mIPSC decay time (mean mIPSC decay tau: control = 19.05 ± 2.09 ms and DynA = 20.83 ± 2.37 ms, n = 6; P = 0.13, paired t test).

Figure 5. DynA does not affect GABAergic input to MCPO/SI cholinergic neurons .

A, the effects of DynA on GABAergic sIPSCs. Current traces of a representative cholinergic neuron (left panel) and the corresponding cumulative distribution plots of sIPSC inter‐event intervals (right; P = 0.78, control vs. DynA and P = 0.7, DynA vs. washout, K‐S tests; DynA: thicker line) and sIPSC amplitudes (bottom left; P = 0.13, control vs. DynA and P = 0.38, DynA vs. washout, K‐S tests). Mean sIPSC frequency (n = 6; one‐way ANOVA, F _2,10 = 0.09, P = 0.91) and sIPSC amplitude (n = 6; one‐way ANOVA, F _2,10 = 0.52, P = 0.61) in control (c), Dyn‐A (d) and washout (w). B, the effects of DynA on GABAergic mIPSCs. Current traces of a representative cholinergic neuron (left) and the corresponding cumulative distribution plots of mIPSC inter‐event intervals (right; P = 0.45, control vs. DynA and P = 0.26, DynA vs. washout, K‐S tests) and mIPSC amplitude (bottom left; control vs. DynA: P = 0.7; DynA vs. washout: P = 0.38; K‐S tests; DynA: thicker line). Mean mIPSC frequency (n = 6; one‐way ANOVA, F _2,10 = 0.5, P = 0.62) and mIPSC amplitude (n = 6; one‐way ANOVA, F _2,10 = 0.35, P = 0.71) in control (c), Dyn‐A (d) and washout (w). Spontaneous IPSCs and mIPSCs were recorded in normal ACSF, at a V _h = −5 mV with a caesium methane sulfonate‐based pipette solution (Cl⁻ reversal potential = −63 mV) and mIPSCs were recorded in TTX (1 μm).

Responses to the co‐application of DynA and OxA depend on the membrane potential of MCPO/SI cholinergic neurons

Kappa opioid signalling can desensitize, with reduced responses to agonists after repeated activation (Liu‐Chen, 2004). To determine whether the response to DynA in MCPO/SI cholinergic neurons desensitizes, we tested the responses to three brief applications of DynA in voltage‐clamp at V _h = −40 mV and −70 mV in the presence of TTX (1 μm), kynurenic acid (1 mm) and bicuculline (10 μm) to block glutamatergic and GABA_A transmissions. We applied DynA (1 μm) for 1 min and then washed it out for 9 min and repeated this three times. At −40 mV, after the first application, I _DynA was 16.6 ± 2.6 pA (n = 8), but this current was smaller with subsequent applications (2nd application: I _DynA = 9.8 ± 2.9 pA and 3rd application: I _DynA = 8.2 ± 2.2 pA; repeated‐measures one‐way ANOVA P < 0.01; Fisher's PLSD, P < 0.01 between 1st and 2nd applications and P = 0.038 between 2nd and 3rd applications) (Fig. 6 A and C). At −70 mV the I _DynA was small and still desensitized with repeated applications of DynA (1st application: I _DynA = 5.1 ± 0.8 pA; 2nd application: I _DynA = 4.1 ± 1.2 pA, and 3rd application: I _DynA = 3.7 ± 1.2 pA, n = 5; repeated‐measures one‐way ANOVA, P = 0.049; Fisher's PLSD, P = 0.013 between the 1st and 3rd application) (Fig. 6 A and D).

Figure 6. The responses to the co‐application of DynA and OxA depend on the membrane potential of MCPO/SI cholinergic neurons .

A, the outward currents evoked by DynA (1 μm) in MCPO/SI cholinergic neurons recorded at a V _h = −40 mV (left traces) and at a V _h = −70 mV (right traces). The I _DynA is much smaller at V _h = −70 mV. At both holding potentials, I _DynA is reduced with repeated applications of DynA (1 min DynA application, followed by a 9 min washout, and repeated three times). B, the inward current evoked by OxA (300 nm) in MCPO/SI cholinergic neurons (V _h = −40 mV, left traces and V _h = −70 mV, right traces). The I _OxA is smaller at V _h = −40 mV and at both holding potentials the I _OxA is also reduced with repeated application of 300 nm OxA (1 min OxA application, followed by a 9 min washout, and repeated three times). C, mean amplitude of I _DynA (back bars) and I _OxA (grey bars) in response to repeated applications of DynA and OxA in MCPO/SI neurons recorded at V _h = −40 mV (n = 8, for DynA applications; n = 6, for OxA applications; repeated measures one‐way ANOVA: for DynA data F _2,7 = 45.8; P < 0.01 and for OxA data F _2,5 = 17.8; P < 0.01; Fisher's PLSD: **P < 0.01, comparing I _DynA or I _OxA during 2nd and 3rd applications with I _DynA or I _OxA during 1st application and ^† P < 0.05 comparing I _DynA or I _OxA during 2nd and 3rd applications). D, mean amplitude of I _DynA (back bars) and I _OxA (grey bars) in response to repeated applications of DynA and OxA at V _h = −70 mV (n = 5 for DynA applications and n = 6 for OxA applications; repeated measures one‐way ANOVA: for DynA data F _2,4 = 5.3; P = 0.049 and for OxA data F _2,5 = 23.5; P < 0.01; Fisher's PLSD: **P < 0.01 and *P < 0.05 comparing I _DynA or I _OxA during 2nd and 3rd applications with I _DynA or I _OxA during 1st application and ^†† P < 0.01 comparing I _DynA or I _OxA during 2nd and 3rd applications). E, the response of MCPO/SI cholinergic neurons to the co‐application of DynA (1 μm) + OxA (300 nm) depends on the membrane potential. Upon co‐application the inhibitory effect of DynA is the most prominent at V _h = −40 mV (outward current, left traces) whereas the excitatory effect of OxA prevails at V _h −70 mV (inward current, right traces). At both holding potentials the response to the co‐application of DynA + OxA is not significantly reduced with repeated applications (1 min DynA + OxA application, followed by a 9 min washout, and repeated three times). F, the mean amplitude of I _DynA + I _OxA in response to repeated co‐applications of DynA + OxA in MCPO/SI neurons recorded at V _h = −40 mV (left graph, n = 11; repeated measures one‐way ANOVA, F _2,10 = 1.42; P = 0.26) and at V _h = −70 mV (right graph, n = 8; repeated measures one‐way ANOVA, F _2,7 = 0.77; P = 0.45). All recordings were conducted in TTX (1 μm), kynurenic acid (1 mm) and bicuculline (10 μm).

We also determined whether the responses to OxA on MCPO/SI cholinergic neurons desensitize. We used the same protocol as for DynA applications (OxA 300 nm, applied three times for 1 min and washed for 9 min, in TTX, kynurenic acid and bicuculline). The responses to OxA desensitized with repeated applications of OxA at both holding potentials (at V _h = −40 mV, 1st application: I _OxA = −3.9 ± 0.9 pA; 2nd application: I _OxA = −1.5 ± 0.5 pA, and 3rd application: I _OxA = 0.01 ± 0.4 pA, n = 6; repeated‐measures one‐way ANOVA, P < 0.01; and at V _h = −70 mV, 1st application: I _OxA = −9.9 ± 2.2 pA; 2nd application: I _OxA = −6.5 ± 2.1 pA and 3rd application: I _OxA = −4.9 ± 2.1 pA, n = 6; repeated‐measures one‐way ANOVA, P < 0.01; Fig. 6B). At both holding potentials, there was a significant difference in amplitude between the first and second application of OxA (Fisher's PLSD, P < 0.01 at V _h = −40 mV and P = 0.01 at V _h = −70 mV), and also between the second and third applications (Fisher's PLSD, P = 0.04 at V _h = −40 mV and P < 0.01 at V _h = −70 mV) (Fig. 6 C and D).

We next examined the responses to multiple co‐applications of DynA and OxA on MCPO/SI cholinergic neurons. Using the same protocol as for DynA and OxA applications, we co‐applied DynA (1 μm) and OxA (300 nm) three times for 1 min and washed for 9 min (at V _h = −40 and −70 mV, in the presence of TTX, kynurenic acid and bicuculline). The responses to the co‐application of DynA and OxA were opposite at the two holding potentials. At −40 mV the effect of DynA dominated, whereas at −70 mV the response to OxA prevailed (Fig. 6 E). At −40 mV the current evoked by the co‐application (I _{DynA + OxA}) was significantly smaller than the current evoked by DynA alone (I _DynA; P < 0.01, unpaired t test comparing I _{DynA + OxA} vs. I _DynA, first application), whereas at −70 mV I _{DynA + OxA} was smaller but not significantly different from the current evoked by OxA alone (P = 0.341, unpaired t test comparing I _{DynA + OxA} vs. I _OxA, first application).

In addition, at both holding potentials, the net response of the co‐application of DynA and OxA is fairly stable over multiple applications (at V _h = −40 mV, 1st co‐application: I _{DynA +} I _OxA = 6.2 ± 1.5 pA; 2nd co‐application: I _{DynA +} I _OxA = 6 ± 1.1 pA; and 3rd co‐application: I _{DynA +} I _OxA = 4.0 ± 1.2 pA, n = 11; repeated‐measures one‐way ANOVA, P = 0.26 and at V _h = −70 mV, 1st co‐application: I _{DynA +} I _OxA = −6.6 ± 2.4 pA; 2nd co‐application: I _{DynA +} I _OxA = −7.8 ± 2.4 pA; and 3rd co‐application: I _{DynA +} I _OxA = −8.0 ± 3.2 pA, n = 8; repeated‐measures one‐way ANOVA, P = 0.78) (Fig. 6 F).

Discussion

We found that the orexin neurons are the main source of dynorphin to the MCPO/SI. We also found that DynA inhibits MCPO/SI cholinergic neurons through several mechanisms mediated by κ‐opioid receptors. These cholinergic neurons are inhibited postsynaptically by activation of a GIRK current and reductions in voltage‐gated Ca²⁺ currents and this response desensitizes over several minutes. In addition, the cholinergic neurons are inhibited presynaptically by reductions in glutamatergic EPSCs. Finally, we found that the response to the co‐application of DynA and OxA depends on the membrane potential of BF neurons. This finding suggests that, depending on their state of activation, BF cholinergic neurons can be excited or inhibited by orexin neuron signalling.

Dynorphin directly inhibits BF cholinergic neurons

Orexin‐immunoreactive appositions are common on SI cholinergic cell bodies and dendrites (Wu et al. 2004; Espana et al. 2005), but the orexin neurons may release additional neurotransmitters. In addition to OxA and orexin‐B, the orexin neurons express dynorphin and glutamate, as well as the secreted synaptic protein neuronal activity‐regulated pentraxin, and the endogenous opioid peptides nociceptin/orphanin (Chou et al. 2001; Reti et al. 2002; Torrealba et al. 2003; Maolood & Meister, 2010; Muschamp et al. 2014). Furthermore, the orexin and dynorphin peptides are colocalized in the same synaptic large dense core vesicles, supporting the conclusion that they are released together (Muschamp et al. 2014). BF neurons express receptors for orexin, glutamate and dynorphin (DePaoli et al. 1994; Mansour et al. 1994; Marcus et al. 2001). Orexin and glutamate excite BF cholinergic neurons (Eggermann et al. 2001; Wu et al. 2004; Arrigoni et al. 2010), yet we have found that DynA directly and indirectly inhibits MCPO/SI cholinergic neurons.

The postsynaptic effects of DynA on BF cholinergic neurons fit well with prior electrophysiological studies. DynA has high affinity for the κ‐opioid receptors, but it also binds with low affinity to μ‐ and δ‐opioid receptors (Raynor et al. 1994). We found that the effects of DynA were mimicked by a κ‐opioid receptor agonist and were abolished by a κ‐opioid receptor antagonist, but were unaffected by μ‐ and δ‐opioid receptor antagonists. This is consistent with an effect mediated purely through κ‐opioid receptors. The κ‐opioid receptors are coupled to the G_i/o G‐protein family, and they can activate GIRK channels through their βγ subunits (Yamada et al. 1998). In line with prior work on κ‐signalling in other brain regions (Spadoni et al. 2004; Li & van den Pol, 2006; Parsons & Hirasawa, 2011; Zhang & van den Pol, 2013), we found that DynA inhibits MCPO/SI cholinergic neurons by activating a GIRK channel. In addition, through βγ subunits of the G_i/o‐proteins, κ signalling can also inhibit the voltage‐activated Ca²⁺ channels (Herlitze et al. 1996). Accordingly, we found that DynA reduces I _Ca in the MCPO/SI cholinergic neurons. This effect might contribute to the inhibition of the firing of MCPO/SI neurons (Fig. 2) as well as somato‐dendritic integration of synaptic inputs.

In addition, we found that the response to DynA desensitized quickly over three brief applications, suggesting that inhibition of BF cholinergic neurons by DynA may fade over time. Opioid receptors are known to desensitize (Liu‐Chen, 2004), and attenuation of κ‐signalling in response to repeated applications of DynA occurs in the orexin, melanin‐concentrating hormone (MCH), dorsal raphe and locus coeruleus neurons (Li & van den Pol, 2006; Land et al. 2009; Reyes et al. 2010; Parsons & Hirasawa, 2011). Desensitization can occur through three main processes over a time scale ranging from seconds to days (Liu‐Chen, 2004). Phosphorylation of the κ‐opioid receptors can functionally uncouple them from G proteins and reduce κ signalling in seconds. Transient internalization of the κ receptor can occur within minutes after receptor activation. The receptors are then recycled to the cell surface in about 60 min. With exposure to dynorphin for hours to days, κ‐receptors are first internalized and then degraded, leading to long‐term reductions in the number of receptors. As we observed a relatively rapid desensitization in BF cholinergic neurons, phosphorylation and/or internalization of the κ‐receptors seems most likely, and this could be investigated in future studies.

These findings clearly demonstrate that DynA directly inhibits BF cholinergic neurons. This suggests that when co‐released with orexin and glutamate, DynA should antagonize the effects of glutamate and orexin. The net postsynaptic effects may ultimately depend on the levels of receptor expression and relative concentration of these neurotransmitters in the extracellular space. Currently, little is known about the co‐release and synaptic concentrations of orexin, glutamate and dynorphin, so we cannot realistically reproduce their combined effects using modelling or bath co‐applications. Experiments using optogenetic stimulation of orexin neurons and terminals have recently demonstrated co‐release of glutamate and orexin (Sears et al. 2013; Schone et al. 2014) and future studies should soon unravel the dynamics and synaptic effects of dynorphin as well.

It has been reported that in the lateral hypothalamus the response to DynA desensitizes, whereas the response to OxA does not (Li & van den Pol, 2006). We found instead that the responses of MCPO/SI cholinergic neurons to both peptides desensitize, suggesting that the effects of both DynA and OxA may be transient and that, over time, the glutamate signal may produce a more sustained response in BF cholinergic neurons.

DynA inhibits the excitatory input to BF cholinergic neurons by presynaptic kappa receptors

BF cholinergic neurons are innervated by a robust glutamatergic input which represents more than 50% of their total innervation (Hur et al. 2009). The output of BF cholinergic neurons depends on this excitatory, glutamatergic tone (Rasmusson et al. 1994; Fadel et al. 2001), and therefore changes in the strength of this input should have a large effect on the excitability of BF cholinergic neurons. We observed that DynA acts through presynaptic κ receptors and the activation of GIRK channels to reduce the glutamatergic synaptic drive to MCPO/SI cholinergic neurons, a finding that provides an additional mechanism through which DynA can reduce the activity of BF cholinergic neurons. Specifically, DynA decreases the frequency of the mEPSCs without affecting their amplitude, indicating a presynaptic effect with no change in postsynaptic efficacy. These results are consistent with the description of κ‐opioid signalling on presynaptic terminals and with DynA reducing glutamatergic transmission via κ‐opioid receptors in several other brain regions (Tallent, 2008).

Glutamatergic input to BF cholinergic neurons arises from multiple regions (Hur et al. 2009; Hawryluk et al. 2012), but it remains unclear which of these sources provide the excitatory glutamatergic input to BF cholinergic neurons that is presynaptically inhibited by DynA. Furthermore, opioid neuropeptides can also act as retrograde signals regulating their own release (Weisskopf et al. 1993; Tallent, 2008; Dicken et al. 2012). There is therefore the possibility that, in addition to modulating the glutamatergic input to cholinergic MCPO/SI neurons (heterosynaptic modulation), dynorphin released from orexin terminals could also provide feedback to inhibit the release of glutamate, orexin and dynorphin from orexin terminals via homosynaptic modulation.

Desensitization to dynorphin and orexin and the response to their co‐application

It has been previously reported that orexin and dynorphin can have effects that change over time. For example, the MCH neurons of the lateral hypothalamus are inhibited by dynorphin and excited by orexin, but while the response to dynorphin desensitizes, the response to orexin remains stable; therefore, when the two peptides are co‐applied, the MCH neurons are first inhibited by dynorphin, but over time the excitatory effect of orexin prevails (Li & van den Pol, 2006). In contrast, we found that in MCPO/SI cholinergic neurons the response to both DynA and OxA desensitized quickly. Our results are consistent with other pharmacological studies demonstrating that as with many other G protein‐coupled receptors, κ‐receptors and as well as orexin receptors (OxRs) internalize with agonist binding. This has been shown in in vitro (Milasta et al. 2005; Li & van den Pol, 2006; Duguay et al. 2011; Parsons & Hirasawa, 2011) and in whole‐animal studies (Land et al. 2009; Reyes et al. 2010; Heydendael et al. 2014). The majority of published studies on OxR desensitization have concentrated on the Ox1Rs, but based on more recent work, it appears that due to a small difference in the intracellular carboxy‐terminal domain, Ox2Rs might not internalize like the Ox1Rs (Duguay et al. 2011). Our results, however, indicate that OxRs on MCPO/SI cholinergic neurons undergo desensitization, despite Ox2Rs being the main OxR in BF cholinergic neurons (Trivedi et al. 1998; Eggermann et al. 2001; Marcus et al. 2001; Wu et al. 2004).

Our results also suggest cross‐talk between dynorphin and orexin intracellular pathways. For example, we found that at –40 mV the current evoked by co‐application of DynA and OxA (I _{DynA + OxA}) is significantly smaller than the current evoked by application of DynA alone, but at −40 mV, the I _OxA is too small to fully explain this difference. Orexin excites BF cholinergic neurons through the activation of an Na⁺/Ca²⁺ exchanger (Wu et al. 2004), but orexin can also act through the inhibition of the GIRK channels (Leonard & Kukkonen, 2014), thus directly competing with the DynA response. Potential cross‐talk and/or sharing of effectors between dynorphin and orexin signalling is also shown in the results of the multiple co‐applications. While responses to individually applied DynA and OxA both desensitize quickly over three repeated applications, co‐application of DynA and OxA produces a more sustained response suggesting the possibility of reciprocal interference in the dynorphin and orexin desensitization pathways. A similar condition has been reported for opioid receptors, in which occlusion of desensitization occurs with co‐activation of several non‐opioid G‐protein coupled receptors (Levitt et al. 2011).

We also found that co‐application of DynA and OxA can produce opposite effects depending on the membrane potential of BF cholinergic neurons. At −40 mV, co‐application of the two peptides inhibits MCPO/SI cholinergic neurons due to the stronger effect of dynorphin signalling over orexin. At −70 mV, the OxA excitatory response prevails, and the net effect of co‐application is an excitation of MCPO/SI cholinergic neurons. This suggests that the response of BF cholinergic neurons to co‐release of dynorphin and orexin directly depends on their state of activation.

Functional significance of dynorphin inhibition of BF cholinergic neurons

These results provide new perspectives on how orexin neuron signalling regulates wakefulness and sleep via the BF. The orexin neurons mainly fire during active wakefulness and sometimes fire several seconds before an awakening from sleep (Lee et al. 2005 b). In addition, optogenetic activation of the orexin neurons reliably triggers awakenings from sleep (Adamantidis et al. 2007). Orexins and co‐released glutamate excite BF cholinergic neurons (Eggermann et al. 2001; Wu et al. 2004; Arrigoni et al. 2010), and our new findings now show that co‐released orexin and dynorphin have net excitatory or inhibitory effects on BF cholinergic neurons, depending in their state of activation. During slow wave sleep, BF cholinergic neurons are quiescent and presumably hyperpolarized, and orexin neuron signalling probably has a net excitatory influence, mediated by orexin, which promotes sustained awakenings from sleep (Lee et al. 2005 a; Adamantidis et al. 2007; Branch et al. 2015). Conversely, during wakefulness, when BF cholinergic neurons fire at a maximal rate (Lee et al. 2005 a) and are probably depolarized, dynorphin counterbalances the response of orexin, preventing over‐excitation of the BF cholinergic neurons. These opposing roles of orexin and dynorphin may not be restricted to the BF cholinergic neurons but may also apply to other sleep/wake‐regulating neurons which co‐express orexin and κ−opioid receptors.

Additional information

Competing interests

The authors have no conflicts of interest, financial or otherwise, to declare.

Author contributions

LLF, TES and EA designed the experiments. LLF performed the in vivo labelling and whole‐cell recordings and LJA performed the retrograde tracing studies. MJK and BBL generated the Pdyn‐ires‐Cre mice. LLF, LJA, TES and EA did data analysis and interpretation. TES and EA wrote the manuscript with comments from all the authors. All authors have approved the final version of the manuscript.

Funding

This study was supported by NIH grants: RO1NS061863, the Administrative Supplement Utilizing Recovery Act Funds RO1NS061863, R21NS082854, P01HL095491 and R01NS091126.