Defining inhibitory neurone function in respiratory circuits: opportunities with optogenetics?

Abstract

Pharmacological and mathematical modelling studies support the view that synaptic inhibition in mammalian brainstem respiratory circuits is essential for generating normal and stable breathing movements. GABAergic and glycinergic neurones are known components of these circuits but their precise functional roles have not been established, especially within key microcircuits of the respiratory pre-Bötzinger (pre-BötC) and Bötzinger (BötC) complexes involved in phasic control of respiratory pump and airway muscles. Here, we review briefly current concepts of relevant complexities of inhibitory synapses and the importance of synaptic inhibition in the operation of these microcircuits. We highlight results and limitations of classical pharmacological studies that have suggested critical functions of synaptic inhibition. We then explore the potential opportunities for optogenetic strategies that represent a promising new approach for interrogating function of inhibitory circuits, including a hypothetical wish list for optogenetic approaches to allow expedient application of this technology. We conclude that recent technical advances in optogenetics should provide a means to understand the role of functionally select and regionally confined subsets of inhibitory neurones in key respiratory circuits such as those in the pre-BötC and BötC.

Abbreviations

BötC

Bötzinger complex

CPG

central pattern generator

GAD

glutamic acid decarboxylase

GlyT

glycine transporter

KCC2

K⁺–Cl⁻ cotransporter 2

PI

post-inspiratory

PSD

postsynaptic density

pre-BötC

pre-Bötzinger complex

RTN

retrotrapezoid nucleus

VGAT

vesicular GABA transporter

VIAAT

vesicular inhibitory amino acid transporter

VRC

ventral respiratory column

VRG

ventral respiratory group

Introduction

Understanding the structural–functional organization of neural circuits in the mammalian brainstem generating respiratory movements represents a major research challenge in physiology and neuroscience. These circuits consist of spatially distributed populations of inter-acting excitatory and inhibitory neurones that subserve distinct functional roles in respiratory neural and motor output pattern generation. Based on classical neurophysiological and neuropharmacological approaches, a number of models of respiratory central pattern generator (CPG) networks have been proposed that incorporate interactions of excitatory and inhibitory circuits. Inhibitory circuits are thought to be particularly important for temporally coordinating inspiratory and expiratory phases of the respiratory cycle as well as dynamically shaping activity patterns within phases necessary for normal motor behaviour (Lindsey et al. 2012; Richter & Smith, 2014). However, many of the circuit interactions and functional roles of different populations of the active neurones as portrayed in these models have been assumed and others remain unknown. The problem of defining the functional roles of inhibitory neurones has been of longstanding interest. Application of optogenetic approaches, which enable targeted optical control of neural activity, may provide for the first time an opportunity to selectively interrogate the function of inhibitory (as well as excitatory) microcircuits of the respiratory network. This review discusses virtues and limitations of both classical pharmacological approaches, which have provided important insights, and optogenetics-based strategies for functional analyses of key inhibitory microcircuits thought to be involved in respiratory pattern generation. Before considering the roles of synaptic inhibition for respiratory pattern generation, it is necessary to consider some peculiarities of most fast inhibitory synapses that have relevance to the brainstem respiratory network.

Overview of the neurobiology of fast inhibitory synapses

GABA and glycine are the main inhibitory neurotransmitters in the brainstem and often co-exist in inhibitory presynaptic terminals (Fig. 1), yet they can serve very different roles (Legendre, 2001; Luscher & Keller, 2004). But how is this possible? Both neurotransmitters are incorporated into synaptic vesicles by the same vesicular inhibitory amino acid transporter (VIAAT) also known as vesicular GABA transporter (VGAT; McIntire et al. 1997). This means that the relative concentrations of GABA and glycine in synaptic vesicles is directly dependent on the relative intracellular concentration of each amino acid (Sagne et al. 1997). Glycine is biosynthesized in the body from the amino acid serine. Its concentration in the cytoplasm of inhibitory neurones also depends on the activity of the plasma membrane bound glycine transporter-2 (GlyT₂), whereas GlyT₁ is concentrated in glial cells (Fig. 1) and functions to control extracellular levels of glycine in the synaptic cleft (Zafra & Gimenez, 2008). Intracellular GABA is synthetized from glutamate by two cytosolic glutamic acid decarboxylase enzymes (GAD65 and GAD67) and also originates from reuptake by plasma membrane GABA transporters (GAT; Zhou & Danbolt, 2013). Cytosolic levels of GABA mainly depend on the availability of the precursor glutamate and can be affected by its extracellular concentration (Mathews & Diamond, 2003). Differential availability of GABA and glycine in the inhibitory presynaptic terminal is one of the mechanisms that can confer selectivity to the fast-inhibitory transmission in different central circuits.

Figure 1. Inhibitory amino acid co-transmission

The main precursors for γ-aminobutyric acid (GABA) synthesis in neurones are glucose, pyruvate and glutamine, which are converted to glutamate. The concentration of GABA in the presynaptic terminal depends on the activity of the synthetic enzyme glutamic acid decarboxylase (GAD, converts glutamate to GABA), and re-uptake by the GABA transporter 1 (GAT₁). The major precursor of glycine in the brain is serine, and there is also re-uptake of glycine into the presynaptic terminal by the glycine transporter 2 (GlyT₂), which is preferentially expressed in neurones. Both glycine and GABA are incorporated into synaptic vesicles by the vesicular inhibitory amino acid transporter (VIAAT), also known as vesicular GABA transporter (VGAT). GABA and glycine bind to the GABA_A and glycine receptor (GlyR), respectively, opening chloride (Cl⁻) channels in the postsynaptic terminal. The activity of the K⁺–Cl^– cotransporter 2 (KCC2) maintains Cl⁻ homeostasis regulating the Cl^– equlibrium potential that determines the resulting change in membrane potential (IPSP, inhibitory postsynaptic potential), usually hyperpolarization in mature neurones. Clustering of GlyR and GABA_A receptors at the postsynaptic membrane is determined by interactions with gephyrin scaffolds. The concentration of GABA and glycine in the synaptic cleft is controlled by both neuronal (GAT₁ and GlyT₂, respectively) and astrocyte re-uptake (GAT₃ and GlyT₁, respectively).

Fast postsynaptic inhibition is mediated by GABA_A and glycinergic receptors (Alexander et al. 2011). A mechanism that allows for selectivity of an inhibitory synapse type is the differential expression of these receptors. Both are ligand-gated chloride channels of the Cys-loop family (Alexander et al. 2011). The high time resolution and strength of postsynaptic inhibition required for respiratory pattern generation depends on the amount of transmitter released but also on the configuration of the postsynaptic density (PSD), i.e. a high concentration of ligand-gated chloride channels. Not only expression, but clustering of receptors within the PSD is also key for synaptic strength and functional plasticity (Luscher & Keller, 2004; Tyagarajan & Fritschy, 2014). The shared vesicular packaging system for GABA and glycine means that the same presynaptic neurone can provide both GABAergic synaptic transmission at one synapse and glycinergic transmission for another, depending on the type and availability of the receptor expressed postsynaptically. While the postsynaptic localization is particularly important for phasic activity-related activation of glycine and GABA_A receptors, the latter can also have extra-synaptic localization, mediating via low levels of GABA in the extracellular space so-called ‘tonic’ inhibition (Farrant & Nusser, 2005; Ferando & Mody, 2014). This regulates mass neuronal excitability that may also play a role in brainstem respiratory circuits to ensure stability of rhythmic pattern generation (Abdala et al. 2010a).

Another twist to fast synaptic inhibition is that in many CNS regions GABA_A receptor-mediated currents and, in some brainstem regions, glycine receptor-mediated currents (Ehrlich et al. 1999) are depolarizing and thus excitatory in immature neurones postnatally. This is due to high intracellular chloride (Cl⁻) concentrations associated with low expression of the K⁺–Cl⁻ cotransporter 2 (KCC2), in contrast to mature neurones where transcriptional induction of KCC2 results in a switch in the Cl⁻ equilibrium potential during postnatal maturation, accompanied by hyperpolarizing inhibitory currents. However, in respiratory circuits, this switch has already occurred at birth (Ren & Greer, 2006), at least in rodents.

Importance of inhibitory circuits within the medullary respiratory network

A central hypothesis in the respiratory neurobiology field is that the key populations of excitatory and inhibitory interneurones, which constitute the neural machinery for generating the inspiratory and expiratory phases of the respiratory cycle in the respiratory CPG, are bilaterally distributed in three anatomically defined and functionally distinct compartments in the ventrolateral medulla – the Bötzinger complex (BötC), pre-Bötzinger complex (pre-BötC) (see Fig. 2), and rostral ventral respiratory group (rVRG). These regions are components of a more extensive rostro-caudally distributed ‘ventral respiratory column’ (VRC) on each side of the brainstem (Feldman et al. 2013; Smith et al. 2013). Electrophysiological recording studies have shown that major populations of inspiratory neurones are concentrated in the pre-BötC and rVRG, while populations of expiratory neurones are in the BötC and the region termed the caudal VRG (cVRG), adjacent and caudal to the rVRG. Sub-populations of the active neurones in these regions are inhibitory neurones (below), and phasic synaptic inhibition in these regions is critical for generating the normal respiratory pattern (see Richter & Smith, 2014, for a recent review), which consists of three neural activity ‘phases’: inspiration (I, phase 1), post-inspiration (PI, phase 2), and the later second stage (called E2) of expiration constituting phase 3 of the respiratory cycle (see Fig. 3). Interacting excitatory and inhibitory circuits especially within the pre-BötC and BötC are postulated to be core components of the circuitry required for stable inspiratory–expiratory pattern generation (Lindsey et al. 2012; Smith et al. 2013; Richter & Smith, 2014). Inhibitory neurone populations within these two regions are also thought to be a major source of synaptic inhibition throughout the brainstem respiratory network. Circuit diagrams depicting some of the postulated interactions of inhibitory and excitatory neural populations including within and between the pre-BötC and BötC have been proposed (e.g. see Smith et al. 2007, 2013; Lindsey et al. 2012; Richter & Smith, 2014). The functional roles of synaptic inhibition within the medullary respiratory network are discussed next.

Figure 2. Glycinergic neurones in the pre-Bötzinger complex (pre-BötC) and Bötzinger complex (BötC) regions of transgenic mouse medulla oblongata

This parasagittal view (single optical plane confocal image) of one side of the ventral medulla (VS, ventral medullary surface) portrays the dense concentrations of glycinergic neurones (red) in these regions as labelled by expression of tdTomato fluorescent protein, obtained by crossing a GlyT₂-Cre driver mouse strain with a Cre-dependent tdTomato reporter strain (J. C. Smith, H. Koizumi, N. Koshiya & R. Zhang, unpublished observations). Subsets of these glycinergic neurones are components of the respiratory network inhibitory connectome. The concentrations of inhibitory neurones reflect their potential role in circuit function of these classically defined respiratory regions. Motoneurones (green) in compact and semi-compact divisions of nucleus ambiguus (NAc and NAsc, respectively) and facial motor nucleus (VII), serving as anatomical landmarks, are labelled with an antibody against choline acetyltransferase (ChAT).

Figure 3. Schematic diagram of a three-phase respiratory cycle and its neuro-mechanical components

Top schematic plots and neurograms represent lung volume, sub-glottal pressure (SGP), phrenic (PN), recurrent laryngeal (RLN) and internal intercostal (int IN) nerve activities during the three phases of a respiratory cycle, i.e. inspiration (I), post-inspiration (PI) and late expiration (E2). Note that RLN conveys outputs from both abductor and adductor motoneurones, which fire respectively during inspiration (to dilate the glottis during inhalation) and post-inspiration (to narrow the glottis during exhalation). Bottom overlay plots represent hypothetical time courses of glycine- (Gly, dotted line) and GABA-mediated (continuous line) inhibition to inspiratory (I) neurones during the respiratory cycle. These time courses reflect the activity of medullary post-inspiratory (PI) inhibitory neurones, thought to be predominantly glycinergic, and GABAergic augmenting expiratory (E-AUG) inhibitory neurones active during E2 phase, both of which inhibit inspiratory neurones during expiration. Phasic inhibition of inspiratory neurones is minimal during inspiration, when active inspiratory neurones inhibit expiratory neurones, but rises abruptly during PI to orchestrate the inspiratory–expiratory phase transition and initiate exhalation.

Inhibitory neurones in the pre-BötC

The pre-BötC, locus of the excitatory circuitry that generates rhythmic inspiratory activity (Smith et al. 1991; Rekling & Feldman, 1998; Koshiya & Smith, 1999), incorporates a rich network of inhibitory inspiratory neurones and synapses (Koizumi et al. 2013). Figure 2 shows the high concentrations of glycinergic inhibitory neurones in the pre-BötC and BötC regions many of which form local respiratory circuits. Although the function of pre-BötC inhibitory neurones has not been established, since many of these neurones fire during inspiration, it has been proposed that these neurones are connected to expiratory neurones in the BötC and elsewhere to silence expiratory neurones during inspiration (see Richter & Smith, 2014). Nearly half of pre-BötC inhibitory inspiratory neurones express both GAD67 and GlyT₂ mRNA, indicating potential for co-transmission also confirmed by immunohistochemistry (Fig. 4). Although the functional role of co-transmission in the respiratory network is unexplored, this has been documented in the cerebellum and brainstem auditory system. Cerebellar Golgi cells, which co-release GABA and glycine, exert inhibition of granule cells purely via GABA_A receptors, whereas their inhibition of unipolar brush cells is mediated by glycinergic currents (Dugue et al. 2005). In auditory brainstem neurones, GABA and glycine are co-released and GABA fine tunes the duration of glycinergic inhibitory postsynaptic currents (Lu et al. 2008; Apostolides & Trussell, 2014).

Figure 4. Co-expression of glycine in GABAergic neurones in the pre-Bötzinger complex (pre-BötC) region of GAD67-GFP transgenic mouse

A, GAD67-GFP positive neurones distributed within the pre-BötC (coronal plane, ×10 objective). Motoneurones in the semi-compact division of nucleus ambiguus (NAsc) are labelled with antibody against choline acetyltransferase (ChAT) (white). Ba–b, single optical plane confocal images of GAD67-GFP positive (Ba) and glycine (Bb) immunolabelled neurones (red) in the pre-BötC area marked by dashed box in A. Merged image (Bc) shows the heterogeneous population of pre-BötC inhibitory neurones, including GABAergic neurones co-expressing glycine (yellow, white arrowheads) or neurones without glycine (green), and a subpopulation of glycinergic only neurones (red, GAD67-GFP negative). Modified from Koizumi et al. (2013) with permission.

Inhibitory neurones in the BötC

In the BötC, where augmenting (E-AUG) and decrementing expiratory neurones are concentrated, there seems to be a clearer specialization of inhibitory transmission (Ezure et al. 2003). Decrementing (called E-DEC/post-inspiratory or PI) neurones, hypothesized to inhibit inspiratory neurones including in the pre-BötC to mediate inspiration–expiration phase transition (Smith et al. 2007; Shevtsova et al. 2014), preferentially express GlyT₂ mRNA (Ezure et al. 2003). In contrast, some cells of this type are excitatory when located in the pre-BötC of rat and cat (Schwarzacher et al. 1995) as well as the rVRG and cVRG (Ezure et al. 2003). These neurones receive glycinergic-mediated synaptic inhibition during inspiration which when antagonized with strychnine reverts to inspiratory modulated depolarization (Dutschmann & Paton, 2002a). This suggests that PI neurones receive phasic inhibition and excitation (Richter & Smith, 2014) simultaneously from inspiratory neurone populations.

E-AUG neurones provide extensive phasic inhibition to medullary inspiratory and post-inspiratory neurones and appeared to preferentially use GABA based on single-cell pharmacological blockade studies (Champagnat et al. 1982; Haji et al. 1992; Schmid et al. 1996). However, other studies found them to be glycinergic (Schreihofer et al. 1999; Ezure et al. 2003; Fortuna et al. 2008), which may indicate glycine–GABA co-expression or distinct sub-classes of expiratory neurones. In sum, at least according to pharmacological studies on inspiratory neurones, glycine seems to dominate fast inhibition in the early phase of expiration whereas GABAergic inhibition may predominate in the late 2/3 of expiration (see Fig. 3).

Role of inhibitory synaptic transmission for respiratory phase transitions and pattern formation: pharmacological perturbations

Although the role of inhibition for respiratory rhythm generation is still unresolved and the debate about contributions of inhibitory vs. excitatory circuit mechanisms for inspiratory rhythm generation per se in the pre-BötC is ongoing (see Ramirez et al. 2012; Feldman et al. 2013; Janczewski et al. 2013; Smith et al. 2013; Richter & Smith, 2014), there is a general agreement that inhibition is critical for phase transitions to coordinate the orderly generation of inspiratory and both expiratory phases. According to models (below), this may involve a complex set of inhibitory–excitatory circuit interactions, as well as reciprocal interactions between different inhibitory neurone populations. It is clear that within the VRC phasically active inspiratory neurones receive inhibitory synaptic inputs during expiration, whereas expiratory neurones are inhibited during the inspiratory phase. This inhibition is ultimately essential for coordination of spinal and cranial respiratory motor outputs including physiological respiratory modulation of the upper airway (Dutschmann & Paton, 2002a,b,c). Not surprisingly an antagonist of glycinergic receptors, strychnine, severely disrupts normal phase transitions and the three-phase respiratory pattern as well as altering the rhythm in several species when applied either systemically or locally into the pre-BötC (Schmid et al. 1991; Paton & Richter, 1995; Pierrefiche et al. 1998; Busselberg et al. 2001; Dutschmann & Paton, 2002a; Bongianni et al. 2010). Specifically, blockade of glycinergic synaptic transmission abolished the phrenic nerve ramping pattern and caused PI neurones to fire during inspiration, which resulted in paradoxical glottal closure during inhalation in rats (Dutschmann & Paton, 2002a,b). In rabbits, local pharmacological blockade of GABA_A receptors in the BötC severely depressed the respiratory rhythm and induced apnoea (Bongianni et al. 2010). The combined blockade of GABA_A and glycine receptors in both pre-BötC and BötC also inhibited the Hering–Breuer (lung inflation) reflex in rats, affecting inspiratory–expiratory phase transition (Janczewski et al. 2013). Thus, inhibition is essential for generating stable respiratory activity patterns and breathing movements. Next, we consider potential pitfalls with classic pharmacological approaches that have driven more targeted genetic-based strategies.

Caveats with current pharmacological approaches for understanding synaptic inhibition within respiratory circuits

Interpreting functional outputs in terms of inhibitory circuit interactions based on cellular level data and pharmacology alone is not simple. Expiratory neurones also inhibit each other as well as inspiratory neurones (Rybak et al. 2007; Richter & Smith, 2014) and pharmacological blockers are blunt tools meaning they have limited spatial and neuronal type selectivity. Studies using antagonists of fast inhibitory receptors address the postsynaptic inhibition without differentiating between the source(s) of presynaptic drive, whether that is from phasic or tonically active neurones located in the same or different regions of the brainstem, or indeed respiratory related or not. This classical approach also cannot readily address the functional nuances of fast-synaptic co-transmission, receptor clustering and postsynaptic localization within intact circuits. Finally, bicucculline's quaternary salts, widely used antagonists of GABA_A receptors for their solubility in water, are in fact inverse agonists (Ueno et al. 1997) and can also affect cholinergic neurotransmission and other ion channels (Debarbieux et al. 1998; Seutin & Johnson, 1999), off-target actions that can limit their value for assessing GABAergic circuit interactions. Gabazine is a more selective allosteric inhibitor with preferential antagonism of phasically activated synaptic GABA_A receptors (Yeung et al. 2003), although there may be low affinity glycine receptor antagonism.

Nevertheless, several mathematical models of the respiratory network have been constructed to help translate cellular level data to network level output and disturbances with pharmacological perturbation. One model in particular (Rybak et al. 2007), and later variants (Molkov et al. 2011; Shevtsova et al. 2011, 2014), based on circuit diagrams for inhibitory neurone population connectivity, have reproduced experimental results with perturbations of network activity following disruptions of synaptic inhibition pharmacologically (Shevtsova et al. 2011, 2014) or via reduction of extracellular chloride to attenuate inhibitory currents (Smith et al. 2007). These models provide a proof of principle for the essential role of fast inhibitory synaptic transmission for stable breathing.

The mathematical models have allowed exploration of synaptic inhibition, circuit connectivity and functional roles of different types of respiratory neurones in generating and shaping circuit activity and motor output. Based on this hypothetical work, we now require tools with the potential to more precisely modulate the excitability of selected neuronal populations of known transmitter content to better understand their contribution to respiratory activity generation. The following section discusses a potential important role for optogenetics.

Targeting inhibitory neurone populations optogenetically: opportunities and challenges

Optogenetics involves directed cell-specific expression of photo-sensitive ion channel opsins such as channelrho-dopsins, halorhodopsins, and archaerhodopsins in genetically defined cells (see Fenno et al. (2011) for a comprehensive review, and Mattis et al. (2012) for principles for applying optogenetic tools). When incorporated into neurones, either by transgenic and/or viral vector-based transduction strategies, these light-activated opsins enable photo-stimulation or photo-inhibition of neuronal activity (Zhang et al. 2007a). In non-respiratory rhythmic networks such as rodent spinal locomotor pattern generating circuits, optogenetic approaches have enabled dissection of circuit organization revealing the critical involvement of excitatory interneurones for rhythmogenesis (Hagglund et al. 2013). To date there have been only a few published studies applying optogenetic strategies to probe respiratory/respiratory-related circuits (Abbott et al. 2009, 2011; Gourine et al. 2010; Pagliardini et al. 2011), none of which have focused on the problem of defining roles of inhibitory neurone populations in respiratory pattern generation. In the brainstem respiratory network where timing of neuronal activity is so critical for rhythmic pattern generation, optogenetic tools would be ideal as they will allow high-resolution temporal control of neural activity to probe functional roles of key populations of inhibitory neurones. Below we consider some of the challenges for applying optogenetic approaches in respiratory circuits.

Optogenetic targeting of different inhibitory (and other) neurone types depends on selectivity of molecular markers for cell-type specific activity modulation. Applying optogentics to dissect inhibitory neurone population function in respiratory circuits is challenging, primarily because in the VRC a precise correlation between neuronal function and molecular phenotype has not been established. This is in contrast to some excitatory respiratory neurone populations in which expression of certain homeobox genes are very useful for targeting well-defined functional phenotypes, e.g. Phox2b in the retrotrapezoid nucleus (RTN), a site for central chemo-sensitivity (Hwang et al. 2001; Abbott et al. 2009, 2011; Marina et al. 2010) and Dbx1 in the pre-BötC (Lu et al. 1996; Bouvier et al. 2010). Because chemo-sensitive glutamatergic RTN neurones have a defined phenotype (Phox2b), channelrhodopsin-based optogen-etics has been very successful in demonstrating their role in inspiratory and expiratory motor output generation (Abbott et al. 2009). Peptidergic markers have also been used to delineate sub-populations of excitatory respiratory neurones (Leibstein et al. 1985), but peptidergic expression profiles of inhibitory neurones have not been delineated. Somatostatin has showed value for identifying a sub-population of pre-BötC neurones (>90% glutamatergic) that are essential for breathing in vivo (Stornetta et al. 2003; Tan et al. 2008), although a recent study suggested a role in vocalization (Tupal et al. 2014). Galanin is expressed in about 50% of the glutamatergic chemosensitive neurones in RTN (Lazarenko et al. 2009; Bochorishvili et al. 2012), but is also present in propriobulbar and bulbospinal rVRG neurones (Lazarenko et al. 2009; Bochorishvili et al. 2012; Spirovski et al. 2012; Tan et al. 2012). In terms of inhibitory cell-specific markers, parvalbumin, a small calcium-binding albumin protein often found in cortical GABAergic interneurones (Kawaguchi & Kubota, 1997), could be useful for differentiating a subset of GABAergic BötC neurones (Alheid et al. 2002) but others reported it may not be a valid marker of GABAergic cells in the brainstem (Dehkordi et al. 2007).

Although advances have been achieved in delineating subsets of excitatory respiratory neurones, targeting of inhibitory neurones is presently restricted to GlyT₂ and GAD promoters (Taniguchi et al. 2011). Listed below are issues that will impact on the interpretation of data from studies targeting inhibitory neurones in the VRC:

(1) Besides the presently lacking correlation between neurochemical/molecular and functional phenotypes, some presumed inhibitory neurone types, as defined electrophysiologically, are not restricted to a nucleus or defined region within the VRC. For example, E-DEC neurones are distributed throughout the VRC (Schwarzacher et al. 1995; Ezure et al. 2003).

(2) The synthesis of inhibitory amino acids is not exclusively limited to respiratory neurones in the VRC. In the pre-BötC and BötC regions inhibitory neurone populations controlling sympathetic activity co-exist (Schreihofer & Guyenet, 2002).

(3) Glycinergic neurones in the pre-BötC mutually inhibit other local glycinergic neurones (Winter et al. 2009), and mutual inhibitory interactions between BötC inhibitory neurones that are also connected to excitatory neurones are likely. These circuit arrangements could result in complex competing effects for example during optical stimulation experiments, i.e. photo-stimulation of inhibitory neurones could cause concomitant disinhibition/excitation.

(4) Glycine is an obligatory co-agonist of glutamate and is necessary for activation of NMDA receptors (Alexander et al. 2011); therefore tonic optical inhibition of glycinergic neurones could invariably have a competing negative effect on glutamatergic actions at the postsynaptic cell.

(5) The inhibitory and excitatory synaptic drives governing the phasic firing patterns of respiratory neurones are extremely strong (Rybak et al. 2007; Richter & Smith, 2014). This synaptic control is an essential requirement for respiratory central pattern generation, but it also means that the task of overcoming these powerful waves of synaptic drive via exogenously expressed ion channels is challenging. Stated differently, the powerfully modulated oscillating membrane potential may occlude any effect of transient optical excitation or inhibition. In particular, out-of-phase photo-activation or in-phase photo-inhibition of these cells in an operating respiratory network is likely to require very high levels of opsin expression which may generate its own problems like disturbances to ion homeostasis and cell morpho-physiology (Raimondo et al. 2012; Miyashita et al. 2013).

(6) Activity silencing strategies using inward Cl⁻ pumps (i.e. with halorhodopsin, NpHR; Zhang et al. 2007b) would be disadvantageous for the study of respiratory circuits as intracellular Cl⁻ ion accumulation may change the reversal potential of Cl⁻-mediated synaptic inhibition (Raimondo et al. 2012), and some Cl⁻ pumps (e.g. NpHR) present slow channel kinetics with long post-stimulus inactivation periods (Chow et al. 2010). Conversely, proton pumps like archaerhodopsin-3 (Arch) can produce stronger hyperpolarizing currents, recover rapidly from activation and do not affect Cl⁻ or in a major way challenge pH homeostasis (Chow et al. 2010).

(7) Hyperpolarization-activated currents and rebound firing after photo-silencing and excessively large membrane hyperpolarizations should be considered (Madisen et al. 2012).

(8) Photo-stimulation intensity and frequency are critical for properly controlled photo-coupling to avoid spike failure or induced plateau potentials from over-excitation. Most of these problems have been solved with newly engineered opsins with improved channel kinetic properties (Gunaydin et al. 2010; Mattis et al. 2012).

Hypothetical wish list for optogenetics as applied to the inhibitory connectome of the medullary respiratory network

A list of ideal qualities for an optogenetic strategy to allow a functional dissection of the role of inhibitory neurones in the respiratory network in vivo comprises: (1) high spatio-temporal resolution opsins allowing highly precise activation (and inactivation) for respiratory phase-gated optical control; (2) high levels of opsin transduction/expression efficiency in genetically identified and spatially confined neurone subsets; (3) inducible acute expression and applicability to multiple species; (4) the ability to target respiratory cell groups not only based on genetic profile but also functional phenotype. Recent advances are now in place that help fulfil some of the items we have listed.

(1) New engineered opsins provide improved temporal resolution

The latest generation opsin mutants have a stronger light-coupling (Chow et al. 2010; Gunaydin et al. 2010) and avoid disturbing the reversal potential of chloride with sustained activation (Raimondo et al. 2012; Ting & Feng, 2013). Indeed, new chloride-conducting channelrhodopsins (ChloC and iC1C2) can now mimic hyperpolarization induced physiologically (Berndt et al. 2014; Wietek et al. 2014), thus are less likely to artificially induce rebound. Chronos, an ultra-light sensitive blue channelrhodopsin has faster kinetics than previous opsins (Klapoetke et al. 2014). The new red-activated channelrhodopsin ReaChR, with higher photocurrents and improved kinetics (Lin et al. 2013; Azimipour et al. 2014), as well as the red-light drivable channelrhodopsin CrimsonR with enhanced channel kinetics (Klapoetke et al. 2014) also provide better tissue penetration of light for in vivo experiments. When used with Chronos, for example, red-shifted opsins enable two-colour, independent optical stimulation of two different neural populations (Klapoetke et al. 2014) to investigate synaptic interactions.

(2) Transgenic Cre recombinase driver mice provide coverage of target cell type

Crossing of mouse Cre recombinase (Cre; Abremski & Hoess, 1984) driver lines targeting inhibitory neurones (Taniguchi et al. 2011) with Cre-dependent transgenics expressing opsins (Madisen et al. 2012; Ting & Feng, 2013) yields very high levels of expression. It also guarantees fuller coverage of the targeted cell type, unlike viral targeting vectors that can often offer less than 60% transduction efficiency in the injection site (Abbott et al. 2009, 2011; Marina et al. 2010). For a comprehensive resource of available Cre-driver mouse lines refer to Taniguchi et al. (2011) and Ting & Feng (2013). Proven-effective GlyT₂-Cre mice for controlling Cre-dependent expression of reporters (see Fig. 2) and therefore likely opsins to target glycinergic neurones in respiratory regions are available. VIAAT-Cre driver mouse lines possibly covering GABAergic, glycinergic and GABA–glycine co-expressing neurones are available (Hagglund et al. 2013) that should be evaluated for targeting respiratory regions. But the Cre-dependent transgenic approach has a major limitation for functional studies including a lack of spatial selectivity as well as optical activation or inhibition of fibres of passage and local synapses expressing the channel opsin in the region targeted with light (Ting & Feng, 2013). This can make interpretation difficult as the optical signal affects neuronal somata, dendrites, axons and fibre terminals. Additionally, in transgenic animals long term expression during pre- and postnatal development of certain opsins could have a deleterious effect on neuronal cell morphology and physiology, potentially causing alterations in the exact mechanisms under study (Miyashita et al. 2013).

(3) Viral vectors using Cre recombinase-mediated amplification produce high expression with spatial resolution

An approach that provides better spatial resolution without risk of developmental effects is Cre-dependent opsin expression via a viral vector in Cre-driver transgenics, although currently limited (for inhibitory neurones at least) to mice and potentially by viral transduction efficiency. The discovery of an active region of the GlyT₂ promoter that is small enough to be packaged into lentiviral vectors while retaining its selectivity (Abdala et al. 2010b,c) allows inducible, targeted expression of opsins in multiple species. This approach is providing direct experimental evidence for predicted inhibitory connectivity from BötC neurones to parafacial neurones proposed to be important for controlling late expiratory activity (Molkov et al. 2010, 2011; Moraes et al. 2014). More recently, we tested a Cre-dependent dual vector approach, in which a viral vector containing the inhibitory neurone cell-specific promoter (GAD67 or GlyT₂) driving expression of Cre recombinase is co-applied with a Cre-dependent viral construct driving expression of an opsin of interest (A. P. Abdala, B. Liu, S. Kasparov & J. F. R. Paton, unpublished data). This method provided higher levels of expression and refined spatial resolution. However, Cre-based strategies are not without their own caveats in these applications as transgene expression of opsins is usually driven by a strong ubiquitous promoter such as elongation factor 1 alpha (EF1α; Uetsuki et al. 1989; Mizushima & Nagata, 1990) or cytomegalovirus early enhancer fused with chicken β-actin (CAG; Niwa et al. 1991; Chung et al. 2002). While this works to amplify expression levels, it can introduce ‘noise’, i.e. non-specific expression from the cell-specific promoter that is driving Cre. This is somewhat offset by the fact that the Cre-mediated recombination reaction appears to occur stoichiometrically (Abremski & Hoess, 1984; Ringrose et al. 1998), hence optimum ratios of Cre to substrate would still be required to produce robust transgene expression. Conversely, with a cell-selective single-promoter approach the impact of leaky expression may be virtually negligible since opsin expression would only reach the threshold for photo-activation in cells with the highest levels of activity of the gene of interest. This means that the choice between promoter-based or Cre-mediated approaches should take into account not only levels of expression desired, but also construct ‘leakiness’ that needs to be evaluated for each individual neuronal population targeted.

(4) Activity- and multiple-feature dependent approaches could help simultaneous functional–neurochemical targeting of respiratory neurones

We believe major progress can ultimately be reached when targeting respiratory cell groups based on neurochemical profile simultaneously with functional/electrophysiological phe-notype is achievable. Recent developments in genetic molecular tools offer new possibilities. For example, approaches that allow for controlled expression of opsins, and other genes of interest, in groups of neurones that are active in a given time window may prove useful (Reijmers et al. 2007; Ramirez et al. 2013). The use of engineered introns to control the expression of genes of interest conditional upon multiple neurochemical phenotypes, effectively allowing the application of Boolean logical operations in one single construct (Fenno et al. 2014) make it possible to combine more than one transmitter, peptidergic marker, and/or transcription factor to potentially restrict opsin expression to a single functional cell type in a given VRC compartment. Furthermore, approaches combining optical control and neuronal activity recording to identify and manipulate specific types of inhibitory interneurones based on their physiological activity patterns in relation to network dynamics and behaviour, as proposed for cortical GABAergic neurones (Roux et al. 2014), could provide new insights on functional roles.

Outlook

It becomes apparent that different optogenetic tools for the functional dissection of the inhibitory component of the respiratory network offer a broad range of choices that have limitations and potential advantages. These tools now require experimental testing in the VRC. To date optogenetic approaches have been successfully applied to study functional roles of some populations of excitatory respiratory neurones. But the challenging problem of targeting inhibitory respiratory neurones based on both genetic and ultimately functional profiles for optogenetic control has not yet been adequately addressed. Effective application of the existing technology to interrogate the roles of inhibitory neurones should provide revealing information that addresses key, yet currently unanswered questions in the field. These include defining the functional role(s) of inhibitory neurone populations in the pre-BötC and BötC, and their mutual interactions, in respiratory rhythm control and activity pattern generation. When applied in conjunction with other complementary techniques for mapping circuit connectivity, the optogenetics toolbox should play a major role in formulating a structural–functional map of the inhibitory/excitatory connectome of respiratory circuits. It is true that some complexities of inhibitory synaptic transmission, including co-transmission and postsynaptic receptor identity, which are relevant functional components, are not readily addressable by current optogenetic approaches and await technological advancement. Nevertheless, optogenetics provides potentially powerful tools for interrogating functional roles of different respiratory microcircuits, and their rigorous application to this system is a current frontier.

Biography

• 

  Dr Abdala is a Research Fellow at the University of Bristol. She obtained honours BSc in Biomedical Research, an MSc in Pharmacology and a PhD in Physiology from the Federal University of Sao Paulo in Brazil. She trained with Julian Paton as a post-doctoral fellow at the University of Bristol in the UK, and then obtained a Research Fellowship from the International Rett Syndrome Foundation. Her research interests span a range of topics centering on the pathophysiology of autonomic function. Notably, she has used cutting edge technology for interrogating neural circuits involved in generating the respiratory rhythm. Recently, Dr Abdala has applied the basis of her research to reveal new therapeutic targets for diseases involving the autonomic control of breathing, sympathetic activity and blood pressure.

Additional information

Competing interests

The authors have no competing interests to disclose.

Author contributions

A.P.A, J.F.R.P. and J.C.S. drafted the manuscript and revised it critically for important intellectual content; A.P.A., J.F.R.P. and J.C.S. approved the final version to be published and are accountable for all aspects of the work.

Funding

A.P.A. and J.F.R.P. are funded by the National Institutes of Health (NIH - R01 NS069220). A.P.A. is funded by the International Rett Syndrome Foundation (IRSF). J.C.S. is supported by the Intramural Research Program of the NIH, NINDS.

Acknowledgements

The authors thank Ruli Zhang, NINDS, NIH, for confocal imaging used for Fig. 2 in the manuscript and Hide Koizumi for his permission to use Fig. 4.