Project Topics
Get Complete Project Material File(s) Now! »
Inhibitory neurons in brain circuits: functions of inhibition
Inhibitory neurons can mediate several types of inhibition within local cir-cuits. When GABAergic neurons are excited by an external source and in turn inhibit glutamatergic neurons, this inhibition is called « feedforward in-hibition » (1.3,A). In contrast, when excitatory neurons drive their own inhi-bition, one refers to this inhibition as « feedback inhibition » (1.3,B). In other words, from the point of view of a principal neuron population, if inhibition is self-generated the inhibition is termed « feedback », however it is called « feedforward » when inhibitory neurons are externally driven. Feedforward and feedback inhibition can be mediated by separate groups of inhibitory neurons. For instance, in the anterior piriform cortex (APC), afferent inputs stimulate layer 1a horizontal and NG cells, which in turn mediate feedfor-ward inhibition mainly onto SL cells. SL cells then drive SP cells, that recruit fast-spiking multipolar (fMP) cells of layer 3 and induce feedback inhibition onto both SL and SP cells (Suzuki and Bekkers, 2012). Interestingly, at the network level, feedforward inhibition onto SP cell dendrites dominates for a weak stimulation, while a stronger stimulation produced mainly somatic feedback inhibition (Stokes and Isaacson, 2010). This shift from dendritic to somatic inhibition can be explained by short-term plasticity mechanism at afferent synapse to GABAergic neurons: while synapse between affer-ent axons to layer 1a GABAergic neuron depresses, the synapse onto layer 3 inhibitory neurons facilitates. In the neocortex and in the hippocampus however, a train of excitatory stimuli first elicit somatic inhibition medi-ated via PV+ cells, and progressively shift distally, where SOM+ neuron synapses are formed (Tremblay et al., 2016). This can similarly be explain by the depressing and facilitating nature of inputs to PV and SOM cells, respectively.
Feedback inhibition can be further divided in two classes of microcircuits. If an external drive excite a subpopulation of excitatory neurons, which in turn drive their own inhibition, this form of inhibition is termed « recurrent » (Figure1.3,B’), however the inhibition is called « lateral » when the first ex-cited population recruit inhibition onto another subset of principal neurons, within the same functional circuit (Figure 1.3,B”). In the APC, afferent in-puts are first recruiting SL cells, which in turn drives fMP cells and therefore trigger recurrent inhibition onto SL cells and lateral inhibition onto SP cells. Lastly, inhibition can also come from an external source, termed « direct inhibition » (Figure 1.3,C) and, finally, inhibition of inhibitory neurons results in « disinhibition » onto principal cells (Figure 1.3,D).
The fundamental question that remains open is which type of GABAer-gic neuron contributes to specific brain circuit functions, behavior and the alteration of which function can lead to disease. Inhibitory activity is likely to depend on the network activity pattern in addition to neuronal charac-teristics. In this introduction, I will now focus on the synaptic physiology of GABA, the main functions supported by local inhibitory neurons, and finally I will review the literature and some new functions supported by long-range projecting inhibitory neurons.
Inhibitory neuron activity in behavior
Recent development in genetically engineered mice and calcium activity sen-sors has allowed investigation of specific neuronal types in behaving mice. Since anesthesia greatly impacts the function, I will mainly focus on data from the awake literature in this section and the following ones (Haider et al., 2012; Kato et al., 2012; Wachowiak et al., 2013). The innovative use of head-restrained mice has greatly eased recordings in awake animals, sometimes engaged in goal-directed behavior or during learning.
Activity of inhibitory neurons is greatly influenced by behavior. For instance, activity of PV+ neurons in the hippocampus substantially varies with oscillation regimes. In the absence of oscillatory activity, PV firing is low (6.5Hz), and it increases during theta oscillations (21Hz). During sharp wave ripples, PV firing augment by an order of magnitude (> 120Hz, Hu et al. (2014)). In the sensory cortex, Carl Petersen’s group found that barrel cortex SOM+ cells hyperpolarized during whisker defltection (Gentet et al., 2012) and PV+ cells fired at lower rates in hit versus miss trials in an associative learning task (Sachidhanandam et al., 2016). In the medial prefrontal cortex (involved in complex cognitive tasks), SOM+ neurons are suppressed when mice entered the reward zone in a reward foraging task, while PV+ activity increased as animals left the reward zone (Kvitsiani et al., 2013).
« Balanced » inhibition and excitation
It follows from the strong interconnections between inhibitory and excita-tory neurons that neuronal network dynamics can only be maintained if the excitatory drive is counterbalanced by inhibition. Through feedforward and/or feedback inhibition, excitatory afferents are somehow scaled, or « bal-anced » by inhibition (Isaacson and Scanziani, 2011; Roux et al., 2014). These changes in excitation and inhibition strength, which are temporally close, have been observed in multiple cortical regions, such as the auditory (Wehr and Zador, 2003), somatosensory (Wilent and Contreras, 2004), visual (Xue et al., 2014), and olfactory cortex (Poo and Isaacson, 2009). Similarly, in the medial prefrontal cortex, disruption of the excitatory/inhibitory balance by tonic depolarization of excitation disrupted social exploration behavior. Interestingly enough, this phenotype was rescued by selective activation of PV+ neurons. In addition, individual layer 2/3 pyramidal cells in the visual cortex receive inhibition that scales with the excitation they receive (Xue et al., 2014). When excitatory drive was genetically manipulated, inhibition from PV-, but not SOM-, expressing cells varied with excitation (Xue et al., 2014), suggesting that PV-containing neurons are the inhibitory neurons re-sponsible for balancing inhibition in mouse visual cortex. In addition to scaling individual cell’s excitatory input, the group of Scanziani showed that in hippocampal gamma oscillations, inhibition can rapidly follow excitation, in a cycle-to-cycle basis, inducing modulation of gamma oscillation over a wide band of frequencies (Atallah and Scanziani, 2009). Excitation and inhi-bition do not balance strictly in space and time, but rather, a ratio between excitatory and inhibitory conductances seems to be overall maintained. In-hibition and excitation are spatially distributed along the dendrites, somas and axon’s initial segments of neurons and are temporally shifted upon ap-propriate stimulation, such that they do not cancel each other out.
Inhibition narrows opportunity window for input inte-gration
Temporally precise disruption of excitation and inhibition occurs in the pres-ence of external or internal stimulation. Afferent inputs or firing of local principal neurons recruit feedforward or feedback inhibition, respectively and therefore excitation on principal cells precedes inhibition with a monosynap-tic delay (Pouille and Scanziani, 2001; Stokes and Isaacson, 2010). In circuit engaging PV cells, this synaptic delay can be as brief as 1 ms (Pouille and Scanziani, 2001). In the olfactory cortex, feedforward inhibition does not recruit PV cells, rather it recruits horizontal and NG cells (Stokes and Isaac-son, 2010; Suzuki and Bekkers, 2010, 2012). Although feedforward inhibition occurs at the dendrites of principal neurons in this example, afferent input stimulation also elicit short-latency inhibition of principal cells (< 2ms) (Stokes and Isaacson, 2010).
Inhibition shapes tuning properties of excitatory neu-rons
A basic property of cortical neurons is that specific features of the envi-ronment differentially drive the spike output of individual cells. For ex-ample, different bar orientations differentially drive different neuron in the visual cortex, different odors differentially drive neurons in the olfactory cortex, etc. Preferential tuning of a cell corresponds to maximum activ-ity (as assessed by cell firing or membrane depolarization; see Figure 1.4,A, black traces). Inhibition plays a clear role in selective tuning. For instance, GABAAR pharmacoligical blockade broadens tuning of principal cells in a variety of cortices (Isaacson and Scanziani, 2011), and notably in the APC (Poo and Isaacson, 2009). Inhibitory neurons are more broadly tuned than excitatory neurons in auditory, visual and olfactory cortex (Isaacson and Scanziani, 2011; Poo and Isaacson, 2009). Consistently, inhibitory current in principal cells were found to be more broadly tuned than their excitatory counterpart (Poo and Isaacson, 2009). Therefore, non-prefered stimuli elicit an excitation/inhibition ratio in principal cells in favor of inhibition, while preferred excitation evokes a ratio permitting further spike output. Stimuli eliciting the best overall responses (« preferred » stimuli) are stimuli eliciting the biggest excitatory/inhibitory conductance ratio. Interestingly, tuning of inhibitory neurons (and notably PV neurons) is correlated with the ex-tent of the dendritic arborization. The more developed the arborization, the broader the tuning (Tremblay et al., 2016). In addition, timing of excitation compared to inhibition can also generate selective tuning. In response to a sensory stimuli in auditory (Wehr and Zador, 2003), somatosensory (Wilent and Contreras, 2005) and visual cortex (Liu et al., 2010) inhibition will follow excitation with a few ms delay (with a feedforward inhibition mechanism, as seen before). This lag was greater for preferred stimulus compared to non-preferred stimuli (Wilent and Contreras, 2005), allowing more time for synaptic integration.
In the olfactory cortex, GABABRs were found to preferentially depress in-tracortical inputs, with minimal impact on afferent fiber synapses (Poo and Isaacson, 2011). Odor stimulation strongly drives intracortical activity, and GABABRs agonist baclofen was shown to have stronger effect on broadly-tuned neurons (Poo and Isaacson, 2011), suggesting a role for GABABRs in shaping olfactory cortical neuron tuning properties. Thus, in sensory sys-tems, the spatio-temporal between excitation and inhibition depend on stim-ulus features, such as odor identity in the olfactory cortex (Poo and Isaacson, 2009). Inhibition can sharpen tuning of cortical neurons without being itself tuned to the opposite direction. Because neighboring cortical neurons have different excitation/inhibition ratios (Xue et al., 2014), neighboring neurons are not tuned to the same stimuli, which shapes the population response to a stimulus (e.g, olfactory cortex neurons respond sparsely to odor stimulation (Stettler and Axel, 2009)).
Inhibition controls the gain of neurons
As a stimulus increases (either in intensity or frequency), a given neuron typically responds with an increase in action potentials emitted, until it eventually reaches an asymptotic firing rate. The relationship between in-put and cell output is the transfer function of a neuron, also called gain of the neuron (Figure 1.4,A). Therefore, the gain of a neuron determines its ability to respond over a range of input dynamics, with weak inputs eliciting a minimal response, increasing inputs inducing higher responses, with a cer-tain slope in the input/output response, and finally strong inputs saturate cell output. The minimal to maximal response that a given neuron is ca-pable of is known as it’s dynamic range. Through a mechanism called gain control, a neuron can undergo changes in its threshold, slope or saturation level, thus revealing that a neuron’s gain is not fixed. This is thought to be important in order for a sensor to have an optimal response to inputs that can vary strength. Below, I first review the different types of gain control that shape neuronal activity and then finish by discussing studies that report gain control in brain circuits.
Often, neuronal gain takes the form of a sigmoidal curve (with a log scale used to describe input strength). Gain control can affect the sigmoidal curve in at least three ways: it can shift the curve on the x axis, y axis, or change the slope without necessarily affecting the threshold or the saturation. 1) A shift on the x axis results in a modification to the threshold of the cell to B. Spatial transformations. Each neuron is represented in a line and in the corresponding row. Correlation between the activity of the neuron x and neuron y is represented in the pixel row x and the column y. Red indicates high correlations and blue low correlations. Spatial decorrelation separates the activity of correlated neurons while spatial correlation renders the activity of uncorrelated neurons more similar.
Functional role GABABR-mediated inhibition
GABABRs are receptors with slow kinetics, and are classically thought to be activated in context of synchronized and repeated GABA release, which will lead to accumulation of GABA and spilling-over outside the synaptic cleft to bind extrasynaptic GABABRs (Scanziani, 2000). However, emerg-ing evidence suggests a more spatially and temporally constraint mode of GABABR activation, both pre- and postynaptically. In this section, I re-view a few of the currently accumulating examples arguing in favor of a spatially or temporally restricted activation of GABABRs.
Function according to GABABR subunit composition. Subunit com-position is associated with different subcellular locations. Elegant stud-ies by the group of Bettler (Vigot et al., 2006) and Larkum (Pérez-Garci et al., 2006) demonstrated the subunit composition-dependent localization of GABABRs and investigated their differential functions. Vigot and col-leagues (2006) generated two specific knock-in mice, lacking either GABAB1a or 1b subunit specifically to show, within the hippocampus, that GABAB1a was localized at CA3 axon terminals while GABAB1b where confined on postsynaptic CA1 neurons. In mice lacking GABAB1a, object recognition was disrupted, while no change was observed in mice lacking the 1b subunit (Vigot et al., 2006). In the neocortex, layer 5 pyramidal cells are unusual in having both axonal and dendritic sites for action potential generation (Larkum et al., 2001, see « 1.1 »). Using the mice generated by Vigot et al. (2006), Perez-Garci and coworkers (2006; 2013) found that GABA release activates GABAB1b to generate a long-lasting inhibition of layer 5 pyra-midal cells. Furthermore, GABAA-mediated IPSPs on pyramidal neurons were blocked by GABAB1a-containing GABABR activation, therefore show-ing that GABAB(1A,2)Rs act as presynaptic autoreceptors on inhibitory neu-rons (Pérez-Garci et al., 2006). In addition, GABABRs containing either the 1a or 1b isoform of the R1 subunit have been associated with different functions during sleep network dynamics. During sleep, cortical networks alternate between synchronized periods of depolarization and burst firing (Up states), and periods of hyperpolarization (Down states). Up states are involved in memory consolidation and Down states may play a role in regu-lating neuronal homeostasis. In mice lacking GABAB1a, layer 1a stimulation terminates Up states but blocking the remaining GABAB1b failed to prolong Up states as found with wild-type mice. In contrast, in GABAB1b-lacking mice, layer 1a stimulation did not terminate Up states but blocking of the remaining GABAB1a did prolong Up states (Craig et al., 2013). Thus, these studies collectively show that pre- and postsynaptic GABABRs vary in sub-unit composition and play different roles in network dynamics.
GABABRs mediate confined shunting inhibition. In the example above, Perez-Garci and coworkers (2006; 2013) show that postsynaptic GABABRs produce inhibition on layer 5 pyramidal cells. Interestingly, they further demonstrate that GABABRs induce a long-lasting inhibition of Ca2+ spikes, but not back-propagating Na+ spikes, by directly inhibiting Ca2+ channels (Pérez-Garci et al., 2013). A similar study dissecting a circuit med-itating GABABR-dependent inhibition onto these layer 5 pyramidal cells came to similar conclusions Palmer et al. (2012). In this study, hindpaw stimulation induced a disynaptic inhibition of ipsilateral layer 5 pyramidal neurons. Disynaptic inhibition is mediated by interhemispheric connections from contralateral principal cells relaying onto NG cells (Palmer et al., 2012).
The resulting feedforward inhibition could be induced by GABABR activa-tion (Palmer et al., 2012). Importantly, hyperpolarization evoked on these neuron somas was minimal, rather GABABR-dependent inhibition was me-diated by blockade of Ca2+ channels (Palmer et al., 2012). Taken together, these results show that GABABRs can mediate shunting inhibition to prin-cipal cell, which is restricted to the dendritic tree.
Restricted spatial recruitment of pre- and postsynaptic GABABRs. GABABRs are not necessarily activated by non-specific volume transmission. Indeed, in the somatosensory cortex, NG cell stimulation activate postsynap-tic GABABRs on layer 4 fast-spiking GABAergic neurons but not presynap-tic GABABRs at thalamic axons (Chittajallu et al., 2013). Similarly, Booker et al. (2013) showed that postsynaptic GABABRs inhibit perisomatic-, but not dendritic tuft-, targeting PV neurons in hippocampus (Booker et al., 2013). Furthermore, in substantia nigra dopaminergic neurons, subcellular enrichment in postsynaptic GABABRs associated with different inputs lead to qualitatively different recruitment of GABABRs in responses to striatal, pallidal or nigral GABAergic afferent stimulation in vivo (Brazhnik et al., 2008). Taken together, these results suggest that differential recruitment of GABABRs, can shift the balance between incoming inputs, and between somatic and dendritic inhibition onto principal cells, thus leading to a shift in information encoded by these principal neurons.
Table of contents :
I Introduction
1 GABAergic signaling in the brain
1.1 Inhibition in the brain
1.1.1 GABAergic receptors
1.1.2 Mechanisms of inhibition
1.1.3 Activation of GABAergic receptors
1.1.4 Inhibitory neuron diversity
1.2 Inhibitory neurons in brain circuits: functions of inhibition
1.2.1 Inhibition in microcircuits
1.2.2 Inhibitory neuron activity in behavior
1.2.3 « Balanced » inhibition and excitation
1.2.4 Inhibitin narrows opportunity window for input integration
1.2.5 Inhibition shapes tuning properties of excitatory neurons
1.2.6 Inhibition controls the gain of neurons
1.2.7 Inhibition of inhibition
1.2.8 Cell assembly recruitment
1.2.9 Oscillations
1.2.10 Inhibition in plasticity and learning
1.3 Functional role GABABR-mediated inhibition
1.4 Long-range GABAergic projections in the brain
1.5 Conclusion on the role of inhibition in brain circuits
2 Introduction to the olfactory system
2.1 Signal transduction: from the nose to the brain
2.2 Synaptic organization in the Olfactory Bulb
2.2.1 Microcircuits in the Glomerular Layer
2.2.2 Olfactory bulb output neurons: routing the information to the olfactory cortex
2.2.3 Reciprocal connection with Granule cells
2.2.4 Additional microcircuits
2.3 From the Olfactory Bulb to the Olfactory Cortex
3 Top-down to the Olfactory Bulb
3.1 Glutamatergic feedback from the olfactory cortex
3.1.1 Projections originating from the anterior olfactory nucleus
3.1.2 Projections originating from the anterior piriform cortex
3.1.3 Topography of cortical feedback projections
3.1.4 Conclusion on glutamatergic cortical feedback
3.2 Top-down inputs from neuromodulatory regions
3.2.1 Serotoninergic neuromodulation
3.2.2 Inputs from the locus coeruleus
3.2.3 Cholinergic inputs from the basal forebrain
3.2.4 Conclusion on neuromodulatory top-down
3.3 GABAergic top-down
3.4 Conclusion on top-down inputs to the olfactory bulb
4 Olfactory Coding in the Olfactory Bulb
4.1 Large scale temporal coding in the olfactory bulb
4.2 Transformation in the Glomerular Layer
4.3 Transformation by Granule cells
4.4 Transformation by other Olfactory Bulb neurons
4.5 Odor responses in M/T cells
4.6 Olfactory code in the Olfactory Bulb
4.7 Cortical influences on olfactory coding and behavior
II Results
1 Cortical top-down inputs to the olfactory bulb are regulated by GABABRs (Article 1)
1.1 Results from the article
1.2 Supplementary results
2 GABAergic cortico-bulbar projections to the olfactory bulb alter olfactory perception (Article 2, in preparation)
2.1 Results from the article in preparation
2.2 Preliminary results
III Discussion
0.1 Extended discussion on Article 1
0.2 Discussion on Article 2 and future directions
0.3 General discussion
GABAergic Signaling in Cortical Feedback to the OB
Bibliography
