GABAergic Synapses and Neurotransmission

About 20% of synapses (most of the remaining being excitatory glutamatergic), they use GABA (gamma-aminobutyric acid). If only excitatory synaptic transmission took place, activity would become explosive, with all cells continuously active, so there is a need for inhibition to balance the excitement. GABA binding to postsynaptic receptors tends to reduce the probability of action potential firing by causing an inhibitory (hyperpolarizing) postsynaptic potential.

There are many types of GABAergic neurons, and their synapses can be highly specific for particular compartments, including the distal dendrites (mainly on the dendritic shaft rather than on spines), the soma and the axon initial segment. The axon initial segment in particular is targeted almost exclusively by GABAergic synapses, whereas glutamatergic synapses mostly land on dendritic spines and sometimes on the soma

GABA is synthesized from glutamate in a single enzymatic step by glutamic acid decarboxylase (GAD). GAD removes a carboxyl group (-COOH) from glutamate, releasing it as CO2. Glutamate is an amino acid present in the cytosol of all cells, so GAD is the only additional component needed to make GABA.

GAD is present in the cytosol of presynaptic boutons, and some GAD appears to be bound directly to the synaptic vesicle membrane. As a result, GABA is produced locally, exactly where it is packaged into vesicles and released.

The vesicular GABA transporter (VGAT) in the synaptic vesicle membrane packages GABA into vesicles. V-type ATPases in the vesicle membrane use ATP to pump protons into the vesicle, creating a pH gradient (acidic inside) and an electrical gradient (positive inside) across the vesicle membrane. VGAT uses the energy of these gradients to accumulate GABA, exchanging protons leaving the vesicle for GABA entering it. The GABA concentration inside the vesicle reaches ≈100 mM.

The process is the same as at glutamatergic synapses:
1. The action potential invades the nerve terminal.
2. The depolarization drives calcium influx.
3. The standard exocytotic release machinery (synaptotagmin, SNARE complex) drives vesicle fusion, releasing GABA into the synaptic cleft.
4. GABA diffuses a few nanometers across the cleft (20-50 nm) and binds receptors in the postsynaptic specialization, with fast neurotransmission being mediated by ligand-gated ion channels.

Fast inhibition is mediated by the GABA_A receptor, a ligand-gated ion channel located in the postsynaptic specialization. GABA binds directly to the GABA_A receptor and increases its open probability within ≈1 ms. The GABA_A receptor is permeable to anions, and chloride (Cl−) is its major permeant ion. At resting membrane potential, Cl- ions enter the cell because of their concentration gradient (higher conc outside the cell), which hyperpolarizes the cell.

Opening GABA_A receptors produces an outward current. Negatively charged Cl− ions enter the cell, and because current direction is defined by the flow of positive charge, an influx of anions counts as an outward current. The added negative charge inside the postsynaptic cell hyperpolarizes it.

Intracellular Cl− is ≈5 mM and extracellular Cl− is ≈120 mM = strong inward concentration gradient. Chloride reversal potential = 61.5 mV × log10([Cl−]inside / [Cl−]outside) = 61.5 mV × log10(5/120) ≈ −85 mV. The ratio is inside over outside because chloride carries a negative charge (or you can do outside/inside and just divide everything by the -1 valence).

The chloride reversal potential (≈ −85 mV) is more negative than the resting membrane potential and well below the action potential threshold of ≈ −45 mV. Opening GABA_A receptors therefore pulls the postsynaptic membrane potential toward ≈ −85 mV, which prevents action potential firing. The inhibitory action of GABA thus depends on the low intracellular and high extracellular concentration of Cl-.
===
"opening a chloride conductance" is shorthand for opening chloride channels which increases the membrane's chloride conductance

1. Passive distribution: leak chloride channels are open at rest in all cell membranes, and the negative resting membrane potential of ≈ −70 mV (set largely by potassium leak conductances) repels the negatively charged Cl− ions, driving them out of the cell.
2. Active extrusion by the KCC2 transporter, which moves Cl− out of the cell together with K+, pumping Cl− against its concentration gradient and lowering intracellular Cl− further than passive distribution can.

If leak channels were the only pathway for Cl−, Cl− would stop leaving the cell once the chloride reversal potential reached the resting potential (≈ −70 mV). At that point, the concentration gradient pushing Cl− back in balances the electrical force pushing it out. So reaching a chloride reversal potential of −85 mV, more negative than the resting potential of most neurons, requires KCC2 to actively pump out additional Cl− against its concentration gradient.

KCC2 is a neuron-specific transporter with 12 transmembrane segments that extrudes Cl− from the cell. It cotransports K+ and Cl− with an electroneutral one-to-one stoichiometry. Because K+ concentration is high inside the cell and low outside, K+ leaving through KCC2 releases energy, which is coupled to moving Cl− out against its concentration gradient. Like other transporters, KCC2 has multiple binding sites and a slow transport rate compared with ion channels.
====
Analogy: think of water flowing downhill, it releases energy that can be used to turn a wheel. So K+ going down its gradient releases energy which is then used for Cl- to go up its gradient.

1. At postnatal day 0 (corresponding to a gestational stage in humans, since rodents are born less developed), KCC2 expression is extremely low. Neurons therefore cannot extrude chloride, and intracellular Cl− is ≈30 mM, while extracellular Cl− is still ≈120 mM.
2. This gives a chloride reversal potential of ≈ −40 mV, which is positive to the resting membrane potential. So when GABA_A receptors open, Cl− leaves the cell, depolarizing the neuron and possibly contributing to exciting it.
3. During the first postnatal days, KCC2 expression increases dramatically, lowering intracellular Cl− and shifting the chloride reversal potential to ≈ −80 mV. By one week after birth, GABA clearly has an inhibitory function.

KCC2 expression differs between dendritic and axonal membranes. Chloride concentrations, and therefore the reversal potential of GABA responses, may consequently differ across neuronal compartments, adding complexity to how GABAergic inhibition works.

At glutamatergic synapses, glutamate is packaged into vesicles by the vesicular glutamate transporter (VGLUT). GABAergic synapses need one additional step: GAD first synthesizes GABA from glutamate, and then the vesicular GABA transporter (VGAT) fills the vesicles. The action potential and neurotransmitter release machinery are similar at both synapse types.

At both synapse types, fast transmission is mediated by ligand-gated ion channels: the neurotransmitter binds directly to the channel and opens a conductance. At glutamatergic synapses, AMPA and NMDA receptors are cation channels with a reversal potential of ≈0 mV, so they are depolarizing and excitatory. At GABAergic synapses, GABA_A receptors are chloride channels with a reversal potential of ≈ −80 mV, so they are hyperpolarizing and inhibitory.

The mechanism is the same at both synapse types. Receptors are present at low concentration extrasynaptically (i.e. outside the postsynaptic density), where they diffuse within the plasma membrane. When a receptor encounters the postsynaptic density, it binds to scaffolding proteins that hold it in apposition to the presynaptic release site so that the receptor is right where it needs to be. The scaffolding protein differs: PSD95 anchors glutamate receptors at glutamatergic synapses, whereas gephyrin anchors GABA_A receptors at GABAergic synapses.

Glutamatergic synapses are often located on dendritic spines, which as you recall are very dynamic (they can grow towards specific locations, appear, and disappear), which may allow them to contribute to rewiring of the brain during learning and memory.

GABAergic synapses are typically located on the dendritic shaft, the soma, or the axon initial segment, with a small fraction on some dendritic spines. Spines always have a glutamatergic input, but a small fraction of spines also receive a GABAergic input --> these synapses may regulate the glutamatergic synapse on the same spine, possibly by switching its function on or off.

The other type is the GABA_B receptor, a metabotropic, 7-transmembrane, G-protein-coupled receptor. Postsynaptically, GABA_B receptors are located at extrasynaptic sites, some distance from the synapse, so released GABA must diffuse to reach them. GABA_B receptors are also present on the presynaptic boutons of most synapses, whether those synapses release glutamate, GABA or another neurotransmitter.

1. Postsynaptic inhibition: GABA_B receptors on the postsynaptic membrane act through G proteins to open GIRK channels (G-protein-coupled inwardly rectifying potassium channels). This produces a slow inhibitory postsynaptic potential (IPSP) that starts ≈50 ms later than the GABA_A-mediated IPSP, and lasts longer as well (several hundred milliseconds). Postsynaptic GABA_B receptors also inhibit voltage-gated calcium conductances, reducing postsynaptic excitability.

2. Presynaptic inhibition: GABA_B receptors on presynaptic boutons act through G proteins to inhibit the calcium channels, thereby reducing neurotransmitter release.

GABA_A-mediated IPSPs are fast: the chloride channels open within ≈1 ms of GABA release, and the IPSP lasts tens of milliseconds. GABA_B-mediated IPSPs start ≈50 ms later than GABA_A-mediated IPSPs and last several hundreds of milliseconds. The delay arises because GABA must diffuse from the synapse to the extrasynaptic GABA_B receptors, and because several G-protein signaling steps are needed before the potassium channels open.

1. Fast postsynaptic inhibition: Postsynaptic GABA_A receptors open a chloride conductance causing an influx of Cl-
2. Slow postsynaptic inhibition: Postsynaptic GABA_B receptors activate GIRK potassium channels causing an efflux of K+, and can also inhibit voltage-gated Ca2+ channels to reduce postsynaptic excitability.
3. Presynaptic inhibition: Presynaptic GABA_B receptors inhibit voltage-gated Ca2+ channels, reducing neurotransmitter release.

Glycine is the main inhibitory neurotransmitter in the spinal cord and brainstem, where GABA has a less prominent role. There, glycine plays an almost identical role to that of GABA in the rest of the brain, and it works in almost the same way as GABA at GABA_A receptors.

Glycine is packaged into synaptic vesicles by the same vesicular GABA transporter (VGAT). Postsynaptic glycine receptors are ligand-gated chloride channels, very similar to GABA_A receptors, that allow chloride to enter the cell when glycine binds.

Without GABA in the pipette, the GABA_A channel stays closed. With GABA in the pipette, the channel's open probability increases, producing single-channel currents with a conductance of ≈10 to 30 pS.

Synaptically released GABA arrives as brief pulses that may reach ≈1 mM. To mimic this, a membrane patch containing GABA_A receptors is voltage-clamped in front of a laminar flow electrode, which delivers two adjacent streams of solution: one without GABA and one with 1 mM GABA. The electrode is mounted on a piezo device that rapidly shifts the streams across the patch, exposing it to GABA for only 1 ms. The current flowing through the GABA_A receptors is measured.

The 1 ms GABA pulse almost immediately activates a chloride conductance, producing an outward current. Although GABA is present for only 1 ms, the current lasts ≈10 to 20 ms, which is the duration of the GABA_A conductance evoked by a single brief pulse. In the equivalent experiment, a 1 ms pulse of glutamate activates an AMPA conductance lasting only ≈2 ms, so the GABA_A conductance is 5 to 10 times slower.
(so when plotting current in the voltage clamped cell you have a positive peak that is wider vs for AMPA it is a negative peak that is narrower)

The IPSC is inward at very negative potentials of around −100 mV, absent at the reversal potential of ≈ −75 mV (so there is almost no IPSC seen at resting membrane potentials), and outward at potentials positive to the reversal potential, such as −50 mV. The current-voltage relationship of GABA_A receptors is linear and symmetrical around the reversal potential, whose exact value depends on local chloride concentrations. The IPSC lasts ≈10 ms, similar to the current evoked by a 1 ms GABA pulse on isolated receptors.

Neuronal membrane potentials typically lie between ≈ −75 and −40 mV. GABA_A currents reverse at ≈ −75 mV, so in this range they are outward and inhibitory. AMPA currents reverse near 0 mV, so in this range they are inward and excitatory. Both GABA_A and AMPA receptors have linear I-V relationships just with different x-intercepts (~-75 mV for GABA_A and 0 mV for AMPA)

1. Time course: the GABA_A conductance and the IPSC it produces both last ≈10 ms, and the IPSP lasts slightly longer than both. This is because the membrane capacitance introduces a membrane time constant: the charge that the synaptic current moves onto the membrane capacitance takes time to discharge.
2. At the resting membrane potential (≈ −70 mV): almost no IPSP is seen. The resting potential is close to the chloride reversal potential (≈ −75 mV), so there is almost no driving force on Cl− and little current flows through the GABA_A channels.
3. At a depolarized membrane potential (for example ≈ −50 mV, reached by injecting 100 pA of current): a clear hyperpolarizing IPSP lasting some tens of milliseconds appears, because Cl− now has a large driving force into the cell.
==
Thus in order to record an IPSP, the postsynaptic cell needs to be depolarized

The GABA_A conductance mediates shunting inhibition by lowering the input resistance of the soma. Open GABA_A channels add extra conducting pathways across the membrane, like additional leak channels, which increases the leak conductance of the cell, equivalent to reducing the input resistance. Part of any excitatory current arriving at the cell then escapes through the open GABA_A channels, so the same current produces a smaller and shorter depolarization.
======
So the two ways in which GABA_A conductance can mediate inhibition are via the hyperpolarizing inhibition that occurs when the membrane potential is depolarized relative to the chloride reversal potential, so chloride hyperpolarizes the cell, or via shunting inhibition.

Glutamate activates AMPA and NMDA receptors on the dendrites, and the resulting current flows toward the soma. There, it either charges the membrane capacitance or flows out through the leak conductances to the extracellular space, which is at 0 mV by definition. Opening the GABA_A conductance adds a parallel conducting pathway, decreasing the somatic input resistance (R, the inverse of the total leak conductance). As a result, the EPSPs recorded at the soma become smaller and shorter.

V(t) = IR × (1 − e^(−t/τ)), where V is the change from the starting membrane potential (before the current was injected), I is the injected current, R is the input resistance, and τ = RC is the membrane time constant. When the current is switched off, the voltage change decays back to zero with the same time constant τ as during charging. Lowering R has two effects, which apply equally to EPSPs:

1. Smaller amplitude: the steady-state voltage change (=IR) is reduced.
2. Shorter duration: τ (=RC) decreases, so the voltage change rises and decays faster.

The GABA_A conductance is still activated: GABA opens the GABA_A channels regardless of membrane potential. No net current flows, however, because the driving force on Cl− (membrane potential minus chloride reversal potential) is zero, and current = conductance × driving force. There is therefore no IPSP, but the open conductance still lowers the input resistance, which is the basis of shunting inhibition.

GABAergic synapses are found on many subcellular compartments, and each location probably performs a slightly different function:

1. Distal dendrites: strong inhibition of nearby excitatory synapses on the same branch, with less effect on synapses on other branches. This allows spatially specific inhibition of individual dendritic arborizations. GABAergic synapses are also found throughout the basal and proximal dendrites.

2. Soma, where glutamatergic synapses are usually absent: by lowering the somatic input resistance, GABAergic synapses may have a divisive effect, proportionally scaling down by the same factor all the EPSPs produced by excitatory currents flowing in from the dendrites.

3. Axon initial segment, where action potentials are initiated in most cells and there are no excitatory synapses: inhibition here may act as an editing function, deciding whether or not the cell outputs an action potential (and acting independently of the synaptic computations in the rest of the cell--> it doesn't change dendritic computation).

4. Presynaptic specializations: GABA_A synapses directly on presynaptic terminals occur only in the spinal cord.

1. AMPA receptors: CNQX and NBQX
2. NMDA receptors: APV and MK-801
3. GABA_A receptors: picrotoxin, bicuculline and gabazine

These antagonists have highly specific actions, which makes them extremely useful in experiments. Therapeutically they have not proven useful, presumably because these receptors are of general importance for all types of brain function.

Benzodiazepines enhance GABAergic inhibition and are used clinically as:
1. Anxiolytics, to treat panic disorder and anxiety
2. Anticonvulsants, to treat epileptic seizures
3. Sedatives, to treat insomnia
The best-known benzodiazepine is diazepam (Valium). Many others have since been developed with higher specificities and better therapeutic outcomes. Unlike receptor antagonists, benzodiazepines have proven clinically useful even though they act on many types of GABA_A receptors, and their specific mechanism of action is well understood at the molecular level.

The GABA_A receptor is a pentamer: five subunits, each one an individual protein, assemble to form the ligand-gated receptor with the Cl− conducting pore in the middle. The subunits are encoded by many genes, divided into families: α β, γ, ρ, δ, π, ε and θ. Most GABA_A receptors are composed of two α subunits, two β subunits and one γ subunit, although the γ position can be occupied by a subunit from one of the other families.

Each subunit has a large extracellular N-terminal domain followed by four transmembrane α-helices (M1-M4). The M2 helices of the five subunits together line the pore that hosts the chloride channel. Both the GABA and benzodiazepine binding sites lie in the N-terminal domains, on the outside of the cell. The two GABA binding sites are at the interfaces between the α and β subunits (since most receptors have two α and two β subunits), whereas the benzodiazepine binding site is at the interface between the α and γ subunits (which needs to be a γ2 or γ3 subunit).

Benzodiazepines do not activate GABA_A receptors by themselves. Instead, when they bind (at their own binding site separate from the GABA binding site), they increase the receptor's affinity for GABA so that GABA binds more tightly. As a result, IPSCs keep the same amplitude but last longer (see image). This increases the total charge transfer and so enhances fast GABAergic inhibition. Because benzodiazepines only enhance GABA's effect, they act only when and where GABA is normally released.

The receptor must contain a γ2 or γ3 subunit, which forms the benzodiazepine binding site together with an α subunit. It must also contain an α1, α2, α3 or α5 subunit. Benzodiazepines have no effect on receptors containing an α4 or α6 subunit instead.

The α subunits have highly similar amino acid sequences, but a single substitution distinguishes benzodiazepine-sensitive from insensitive subunits. Benzodiazepine-sensitive subunits (α1, α2, α3, α5) contain a histidine at this position, whereas insensitive subunits (α4, α6) contain an arginine. In α1 and α2 this is the 101st amino acid counting from the N-terminus, so it is called histidine 101. In the other α subunits, the same residue sits at the same place within a shared stretch of sequence, but its number differs slightly because those subunits have a few more or fewer amino acids earlier in the chain. This position lies in the extracellular N-terminal domain, which contains the benzodiazepine binding site.

Express the α1 subunit together with a β and a γ subunit in a heterologous system (cells that do not naturally make these proteins), whose cells therefore lack GABA_A receptors of their own. Compare two versions of α1: the wild-type subunit, and a mutant in which histidine 101 is replaced by arginine, the residue found in the benzodiazepine-insensitive α4 and α6 subunits. Then measure the Cl− currents evoked by GABA, with and without 1 µM diazepam administered.

1. Wild-type α1: GABA evokes a Cl− current, and in the presence of diazepam the GABA-evoked currents are much larger.
2. Mutant α1: a functional receptor still forms, but diazepam causes no potentiation of GABA-evoked currents. Without diazepam, the mutant receptor's GABA-evoked currents are the same as those of the wild-type receptor.

Histidine 101 is therefore essential for diazepam to potentiate the GABA_A receptor, and a single amino acid appears to distinguish benzodiazepine-sensitive from benzodiazepine-insensitive α subunits.

The same point mutation (histidine 101 to arginine aka H101R) is made in the mouse genome in either the α1 or the α2 subunit, making only GABA_A receptors containing that subunit insensitive to diazepam. Because the mutation does not change GABA-evoked currents in the absence of diazepam, the point-mutant mice behave like wild-type mice at baseline (running around and exploring their environment normally). Giving diazepam and comparing the behaviour of wild-type mice with each point-mutant line reveals which subunit mediates each effect.

1. Wild-type mice: sedation. The mouse moves much less, and may go to sleep or stay still in a corner.

2. α1 point-mutant mice: no sedative effect. The mouse keeps running around and exploring.

3. α2 point-mutant mice: sedation still occurs normally.

Sedation by benzodiazepines is therefore specifically mediated by GABA_A receptors containing α1 subunits, not by those containing α2 subunits.

Mice are nocturnal and anxious in large, open, well-lit spaces, so under control conditions they mostly stay in the dark area, only briefly poking into the light area before returning.

1. Wild-type mice: a low dose of diazepam, much lower than the sedative dose, reduces anxiety, so the mouse explores the light area much more.

2. α1 point-mutant mice: the anxiolytic effect still occurs.

3. α2 point-mutant mice: there is no anxiolytic effect.

Anxiolysis (reduction in anxiety) by benzodiazepines is therefore mediated by GABA_A receptors containing α2 subunits, at least in mice.

It is unclear to what extent the mouse results (sedation via α1, anxiolysis via α2) carry over to humans. Nonetheless, they have inspired the pharmaceutical industry to develop agents acting specifically at α1- or α2-containing GABA_A receptors to cause different effects. The other GABA_A receptor subunits may also mediate different therapeutic actions worth exploring.

In the neocortex, GABAergic neurons are predominantly local: their axons extend only a few hundred micrometers. In other brain regions, GABAergic projection neurons send long-range axons out of their own area to inhibit other brain nuclei.

(Most long-range projections in the brain are formed by excitatory glutamatergic neurons, but long-range GABAergic projections also play an important role in communication between brain areas).

1. Neurons of the reticular nucleus (nucleus reticularis) of the thalamus (aka NRT): provide feedback inhibition to the thalamus, presumably gating and filtering the sensory information relayed to the neocortex.

2. Striatal projection neurons of the basal ganglia: thought to be involved in goal-directed motor control.

3. Purkinje cells: form the sole output of the cerebellar cortex and are involved in fine-tuning motor output.

Except for olfaction, all sensory information reaches the neocortex via the thalamus, whose excitatory principal neurons project to the neocortex. Whisker deflection opens mechano-gated ion channels in sensory neurons at the whisker follicle, then sequentially activates:

1. Primary sensory neurons of the trigeminal nerve, with cell bodies in the trigeminal ganglion, which release glutamate in the brainstem
2. Neurons in the principal trigeminal nucleus of the brainstem, which release glutamate in the thalamus
3. Neurons in the ventral posterior medial (VPM) nucleus of the thalamus, which release glutamate in the barrel cortex
4. Neurons in the primary somatosensory barrel cortex

Both VPM and barrel cortex contain an anatomically identifiable somatotopic map of the whiskers, in which each whisker has its own region, laid out in the same arrangement as the whiskers on the snout. In VPM, each whisker is represented by a "barreloid", and in barrel cortex by a "barrel". Deflecting the C2 whisker makes neurons in the C2 barreloid of VPM fire an action potential, driven by glutamatergic input from the brainstem. This action potential travels along the thalamocortical axon to the C2 barrel of primary somatosensory cortex, where sensory percepts are thought to be generated.

On their way to the cortex, VPM axons send off collaterals that release glutamate onto the NRT, which contains exclusively GABAergic neurons. NRT neurons therefore receive EPSPs and, if sufficiently excited, fire action potentials. NRT axons project back to the thalamus, and some innervate exactly the region of VPM that excited them. This produces fast, GABA_A-mediated hyperpolarizing IPSPs in those VPM neurons, which is feedback inhibition (The full sequence is: whisker-evoked EPSP in VPM → VPM spike → EPSP in NRT → NRT spike → IPSP in VPM).

1. Timing: feedback inhibition restricts VPM firing to a very short time window after a brief whisker deflection, so VPM action potentials are timed with millisecond precision. This happens because the first VPM action potentials excite the NRT, which sends an IPSP back to VPM that prevents any later firing. The precise timing presumably tells the somatosensory cortex exactly when the whisker moved.
2. Selection: only VPM neurons whose EPSP is large and fast reach threshold before the IPSP arrives. A VPM neuron whose EPSP rises more slowly is cut short by the IPSP and never fires. As a result, only the most strongly excited VPM neurons relay the whisker signal to the cortex.

NRT neurons also produce surround inhibition: they inhibit VPM neurons representing nearby whiskers and even neurons in other thalamic nuclei. This creates a competition for the flow of sensory information to the cortex, partly regulated by the NRT. Overall, fast feedback inhibition from the NRT sharpens the spatiotemporal receptive fields of VPM neurons: spatially, surround inhibition makes each VPM neuron respond mainly to its own whisker, and temporally, feedback inhibition makes it fire only in a brief window after the deflection.

The basal ganglia are an interconnected set of brain nuclei, including the striatum and the substantia nigra, whose major function is thought to be controlling goal-directed behavior. The striatum lies immediately below the cortex and above the thalamus, and thalamocortical axons pass through it on their way to the cortex.

About 90% of striatal neurons are GABAergic projection neurons. They receive excitatory glutamatergic input from the cortex and from higher-order thalamic nuclei (though not from primary sensory nuclei such as VPM). There are at least two types of striatal projection neurons (D1 and D2 MSNs), which differ in gene expression and in the targets of their long-range axons.

Direct pathway neurons express dopamine type 1 (D1) receptors and project to the substantia nigra pars reticulata (SNr) in the midbrain. The SNr contains tonically active GABAergic projection neurons: at rest they fire at a high rate and inhibit brainstem motor areas, preventing the animal from moving. By inhibiting the SNr, direct pathway neurons disinhibit the brainstem motor nuclei and promote movement initiation. The direct pathway is therefore a "go" pathway that drives goal-directed action.

Indirect pathway neurons express dopamine type 2 (D2) receptors and project to the external part of the globus pallidus, which in turn inhibits the substantia nigra pars reticulata (SNr). Activating indirect pathway neurons inhibits the globus pallidus, which disinhibits the SNr. The SNr then inhibits brainstem motor areas, preventing movement. With two inhibitory steps before the SNr, compared with one in the direct pathway, the indirect pathway is a "no-go" pathway.

The dopaminergic input comes from the substantia nigra pars compacta (SNc), which contains the midbrain dopaminergic neurons. The majority of SNc axons innervate the striatum.

Dopamine is involved in reward signalling and in motivating goal-directed behaviour and is released by volume transmission: the boutons along dopaminergic axons have no tight postsynaptic apposition and release dopamine into the extracellular space. The dopamine diffuses and acts on D1 receptors on direct pathway neurons and on D2 receptors on indirect pathway neurons (both metabotropic receptors)

Dopamine is released in response to reward and is thought to act as a reinforcement signal that drives synaptic plasticity. The striatum receives glutamatergic input from the cortex and thalamus, and these inputs synapse onto both D1 and D2 striatal projection neurons. Dopamine changes the plasticity of these glutamatergic synapses in opposite ways in the two cell types:
1. Through D1 receptors, dopamine appears to promote potentiation of glutamatergic synapses onto direct pathway neurons.
2. Through D2 receptors, it may have the opposite effect on glutamatergic synapses onto indirect pathway neurons.

After an unexpected reward, dopamine would therefore strengthen the recently active glutamatergic synapses onto the direct pathway, perhaps reinforcing the sensorimotor processing that led to the reward (making the animal more likely to repeat what it just did, because the same situation will drive the go pathway more strongly next time).

The degeneration of dopaminergic neurons especially in the SNc. Its main symptoms include rigidity, tremor, and bradykinesia (slowness/difficulty initiating movements). The missing dopamine may cause a deficit in the D1 direct ("go") pathway, leading to a lack of self-motivated, goal-directed movements.

Like the main brain, the cerebellum ("little brain" in Latin) has a cortex that projects to deeper brain structures. In the main brain, the output neurons of the neocortex are glutamatergic pyramidal neurons, which excite deeper brain areas. In the cerebellum, the output neurons of the cerebellar cortex are GABAergic Purkinje cells, which form its sole output and inhibit the deep cerebellar nuclei. The two cortices thus have a similar layout, but the neocortex sends excitatory output while the cerebellar cortex sends inhibitory output.

The cerebellum is thought to be involved in sensorimotor learning and in higher cognitive functions. It receives sensory information from both descending and ascending pathways and uses it to fine-tune motor output. Purkinje cells appear to act as integrators that coordinate sensorimotor actions.

Purkinje cells have elaborate, spiny dendritic arborizations lying in a flat plane, the largest and most complex dendritic arbors in the mammalian brain. Their axon descends to the deep cerebellar nuclei, where it releases GABA.

(Santiago Ramón y Cajal drew them over 100 years ago using Golgi staining)

Inhibitory interneurons are GABAergic neurons with local axons stretching only a few hundreds of micrometers, which regulate the activity of the local microcircuit they are embedded in (e.g. by inhibiting nearby pyramidal neurons or other interneurons). Most neocortical GABAergic neurons are interneurons, without long-range axonal projections. They make up ≈15% of neocortical neurons, while ≈85% are excitatory glutamatergic pyramidal neurons. Despite their small number, they have a huge diversity in morphology, electrophysiological properties, synaptic connectivity and, molecular expression patterns (which are used to classify them).

1. Parvalbumin (PV)-expressing neurons
2. Somatostatin (Sst)-expressing neurons
3. Vasoactive intestinal peptide (VIP)-expressing neurons
4. Neurogliaform cells

The first three groups are defined by which of these marker genes they express. Their expression is largely non-overlapping, so they form distinct classes. Neurogliaform cells currently lack a good molecular marker. Together, the four groups account for most neocortical GABAergic neurons.

PV neurons are fast-spiking: when strongly depolarized by injected current, they fire at up to several hundred Hz, and each action potential has a half-width (its duration at half its amplitude) of ≈300 µs (0.3 ms). Excitatory pyramidal neurons reach maximum firing rates of only a few tens of Hz, about an order of magnitude lower, and each action potential has a half-width longer than 1 ms. PV neurons make up almost half of all neocortical GABAergic neurons, making them the largest group.

=======
This is the example we saw in previous weeks where we saw that some inhibitory neurons (e.g. PV neurons) can fire 10x faster than excitatory due to faster repolarization phases.

1. Basket cells, the major type: their axons innervate excitatory pyramidal neurons close to the soma, often surrounding the soma in a basket shape, and also the proximal dendrites. Their action potentials cause fast, GABA_A-mediated hyperpolarizing IPSPs.

2. Axo-axonic cells, also called chandelier cells: their axons specifically innervate the axon initial segment of excitatory pyramidal neurons, where action potentials are initiated. They may leave integration in the dendrites and soma largely unaffected and instead "edit" whether the pyramidal neuron can output an action potential (i.e. inhibition at the AIS can block the pyramidal neuron from firing an AP without affecting how its dendrites and soma integrate synaptic inputs)

Several nearby neurons in a brain slice are recorded simultaneously in current clamp, each with its own electrode. Current is injected into one cell to make it fire an action potential, while the membrane potential of every other recorded cell is monitored. An EPSP or IPSP appearing in a cell right after the action potential shows that the firing cell connects to it. This is repeated for each cell in turn, revealing which cells connect to which and in which direction. Each cell is then identified in two steps: fluorescence imaging shows whether it is GABAergic, and its firing pattern during current injection shows whether a GABAergic cell is a fast-spiking PV neuron.

1. Fluorescence: in his example, the slice came from a transgenic mouse expressing green fluorescent protein in all GABAergic neurons, and each recording electrode filled its cell with red dye. The recorded pyramidal neurons therefore appeared red, while the recorded GABAergic neuron appeared yellow (red plus green).

2. Firing pattern (measured with current clamp): when current was injected, the pyramidal neurons fired broad action potentials at low rates, whereas the GABAergic neuron was fast-spiking, identifying it as a PV neuron.

1. Action potential in cell 1: an EPSP in cell 3, and no response in cell 2.
2. Action potential in cell 2: an EPSP in cell 3 and no response in cell 1.
3. Action potential in cell 3: IPSPs in both cell 1 and cell 2, because the PV neuron and the pyramidal neurons are connected in both directions.

To see the IPSPs, cells 1 and 2 must be depolarized to −50 mV. At their resting membrane potential, which is close to the chloride reversal potential of ≈ −70 mV, GABA_A activation opens a conductance but causes no visible hyperpolarization.

=====
This circuit is a typical example, because the conneciton probability of pyramidal neurons with each other is only about 10% while the connection probability of pyramidal<-->PV neurons is about 50%.

Nearby excitatory neurons are connected to each other with ≈10% probability, whereas excitatory neurons and PV neurons are connected with ≈50% probability in each direction. An excitatory neuron therefore excites PV neurons far more often than it excites other excitatory neurons, and those PV neurons inhibit excitatory neurons in return.

The result is strong inhibition of both the active excitatory neurons themselves (feedback inhibition) and their neighbours (lateral inhibition). Any increase in excitation is therefore rapidly suppressed. In layer 2/3 of primary somatosensory cortex, the circuit is so strongly over-damped that polysynaptic explosive excitation is very difficult to evoke.

Sst neurons inhibit excitatory pyramidal neurons on their distal dendrites. In particular, they target the tuft dendrites, the top branches of the apical dendrite, which extend into cortical layer 1. Layer 1 is the most superficial layer of the neocortex. It contains GABAergic neurons but very few excitatory neurons, and consists mainly of synapses.

Layer 1 receives inputs thought to carry top-down signals for high-level integrative processing. For example, in mouse primary somatosensory cortex, signals from motor cortex arrive in layer 1, where they reach the tuft dendrites of pyramidal neurons. By inhibiting these tuft dendrites, Sst neurons may gate the entry of motor information into the primary somatosensory cortex.

1. PV neurons inhibit proximally: they synapse onto the soma, axon initial segment and proximal dendrites of pyramidal neurons, close to where action potentials are generated. They provide fast feedback inhibition that acts from the very start of activity, because they fire at very high rates and are densely connected with excitatory neurons.

2. Sst neurons inhibit distally: they synapse onto the distal dendrites of pyramidal neurons, where top-down inputs arrive in layer 1. They provide a slower form of inhibition that builds up during sustained activity, because the EPSPs they receive from pyramidal neurons grow during high-frequency firing. Their inhibition may therefore gate the inputs arriving on the distal dendrites.

A single action potential evokes a relatively small EPSP of variable amplitude. During a 50 Hz train, each action potential evokes a larger EPSP than the one before, so the later EPSPs become much larger than the first (see image). This is presynaptic facilitation: repeated firing increases the probability of neurotransmitter release from the presynaptic terminal. So, Sst neurons may integrate bursts of pyramidal action potentials over longer periods, causing them to become strongly depolarized in response to trains of action potentials (this is unique to Sst cells; no other neocortical neurons show such prominent facilitation).

Onto PV neurons, the same pyramidal axon evokes EPSPs that stay constant in amplitude during a train or even depress (shrink), whereas onto Sst neurons the EPSPs facilitate (no other neocortical cells besides Sst neurons show such prominent facilitation). Synapses from a single axon can therefore have very different properties depending on the postsynaptic cell type.

VIP neurons inhibit other GABAergic neurons rather than excitatory neurons. They primarily inhibit Sst neurons, and to a lesser extent PV neurons. By reducing the inhibition that these neurons exert on pyramidal neurons, VIP neurons increase excitation in the microcircuit through disinhibition. The function of the VIP peptide itself in the neocortex is unknown, but it serves as a useful marker for this cell type.

Neurogliaform cells are small, compact GABAergic neurons with an extremely dense axonal arborization close to the soma. The boutons along their axons release GABA without an obvious postsynaptic partner, so GABA is released into the extracellular space rather than at a classical synapse (volume transmission). Neurogliaform cells may therefore regulate extracellular GABA concentrations, inhibiting at least in part through extrasynaptic metabotropic GABA_B receptors to cause slow IPSPs.
Recall the approximate values for:
-Concentrations: vesicle GABA, synaptic GABA pulse, adult and neonatal intracellular and extracellular Cl−
-Voltages: chloride reversal potentials (adult, neonatal), resting potential, action potential threshold, AMPA reversal, depolarization needed to see IPSPs
-GABA_A: opening time, conductance duration vs AMPA, IPSC and IPSP durations, single-channel conductance
-GABA_B: slow IPSP delay and duration
-Development: when GABA becomes inhibitory
-Firing: PV vs pyramidal maximum rates and action potential half-widths
-Proportions: GABAergic synapses, GABAergic neurons, PV fraction, connection probabilities

Concentrations: vesicle GABA ≈100 mM; synaptic GABA ≈1 mM for ≈1 ms; adult Cl− ≈5 mM inside, ≈120 mM outside; neonatal Cl− ≈30 mM inside
Voltages: adult chloride reversal ≈ −85 mV by Nernst (also given as ≈ −80, −75 and −70 mV); neonatal ≈ −40 mV; resting ≈ −70 mV; threshold ≈ −45 mV; AMPA ≈0 mV; depolarize to ≈ −50 mV to see IPSPs
GABA_A: opens within ≈1 ms; conductance lasts ≈10 to 20 ms vs AMPA ≈2 ms (5 to 10 times slower); IPSC ≈10 ms; IPSP some tens of ms; single channel ≈10 to 30 pS
GABA_B: slow IPSP starts ≈50 ms after the GABA_A IPSP and lasts several hundred ms
Development: GABA inhibitory by ≈1 week after birth in rodents
Firing: PV up to several hundred Hz, half-width ≈300 µs; pyramidal up to a few tens of Hz, half-width >1 ms
Proportions: GABAergic synapses ≈20%; GABAergic neurons ≈15% of neocortical neurons; PV ≈ half of neocortical GABAergic neurons; excitatory to excitatory ≈10%, excitatory to and from PV ≈50% each direction
Length: interneuron axons extend some hundreds of µm