| |
Functional organization of the lateral geniculate nucleus
For those who still consider the lateral geniculate nucleus as a simple relay, the complexity of its organization will no doubt come as a surprise. The nucleus is organized into a number of layers with a detailed topographic map of visual space that cuts across layers, and its circuitry involves several cell types, many distinctive groups of afferents, and complex synaptic relationships.
Maps
There is a precise map of the contralateral visual hemifield in the lateral geniculate nucleus of all species so far studied, and in this the visual system is like other sensory systems that are mapped in their thalamic relays. The map in the lateral geniculate nucleus is laid out in fairly simple Cartesian coordinates in all species (see Fig. 35.1, where the visual field and its retinal and geniculate representations are shown as tapered arrows). Each layer maps the contralateral hemifield, either through one or the other eye, and all of these maps are aligned across the various layers that characterize the lateral geniculate nucleus of most species (see below and Fig. 35.1). Thus, a point in visual space is represented by a line, called a line of projection, that runs perpendicularly through all the layers. The precise alignment of these maps, which matches inputs from the nasal retina of one eye across layers with inputs from the temporal retina of the other eye, is seen in all species and is somewhat surprising when one thinks about the developmental mechanisms needed to produce such a match, which forms before the eyes open and before the two visual images can be matched. We shall see that there are nonretinal afferent axons that innervate the lateral geniculate nucleus and distribute terminals along the lines of projection. In this way, these afferents can have a well-localized action on just one part of the visual input, even though this comes from different eyes and distributes to distinct sets of layers (see the section “Afferents to the A Layers”).
Figure 35.1..
Schematic view of the representation of the retina and visual field in the layers of the lateral geniculate nucleus of a cat. The visual field is represented by a straight arrow, and the projection of part of this arrow onto each retina is shown. Small white areas of the visual field and corresponding parts of the retina are labeled “1” and “2.” The representation of this part of the visual field in the lateral geniculate nucleus is shown as a corresponding white column going through all of the geniculate layers “like a toothpick through a club sandwich” (Walls, 1953). Each such column is bounded by the lines of projection, which also pass through all of the laminae. A, A1, and C, the major geniculate layers; L LGN and R LGN, left and right lateral geniculate nuclei; LE and RE, left eye and right eye; asterisk, central point of fixation.
Layering
The lateral geniculate nuclei of all mammalian species so far studied show some form of layering, although there is considerable difference among species as to what the layers represent functionally. For all mammals, each layer receives input from only one eye, but the distribution of distinct functional types of retinal afferents to the layers differs greatly from one species to another. Chapters 30 and 31 describe the various classes of retinal ganglion cell that give rise to several parallel retinogeniculate pathways, and these frequently relate to specific geniculate layers. Figure 35.2 shows the layering of the lateral geniculate nucleus in the macaque monkey2 and the cat, the two best-studied species. The figure illustrates the variation in layering seen across species and introduces the several parallel pathways that are relayed through the lateral geniculate nucleus to the cortex.
Figure 35.2..
Comparison of layering in lateral geniculate nucleus of cat and macaque monkey. See text for details. For simplicity, the medial interlaminar nucleus, which is part of the lateral geniculate nucleus medial to the main laminated portion, present in the cat but not in most other species, is not shown.
Both species have three main retinal ganglion cell classes that project to the lateral geniculate nucleus. For the macaque monkey (Casagrande, 1994; Casagrande and Kaas, 1994; Hendry and Reid, 2000), these are the P (for parvocellular, meaning small-celled), M (for magnocellular; large-celled), and K (for koniocellular; tiny or dust-like cells) cells, comparable respectively to the X, Y, and W cells in the cat (reviewed in Casagrande and Norton, 1991; Lennie, 1980; Sherman, 1985). The terminology for the macaque monkey relates to the geniculate layers to which the cell classes project. P and M cells project to parvocellular and magnocellular layers, respectively. In the macaque monkey, K cells project to the ventral regions of all layers, where very small cells lie scattered, and the projections overlap with those of M and P cells. However, in Galago, a prosimian primate, each of the homologous retinal cell types projects to a separate set of layers (not shown)—koniocellular, parvocellular, and magnocellular—and it was in this species that the koniocellular pathway was first clearly recognized (Conley et al., 1985; Itoh et al., 1982; Norton et al., 1988). In the cat, X and Y cells have overlapping projections to the A layers, Y cells also project to layer C, and W cells project to layers C1 and C2; there is no retinal input to layer C3, which is therefore shown in black in Figure 35.2 (Hickey and Guillery, 1974). Strictly speaking, layer C3 should perhaps not be included in the lateral geniculate nucleus, which can be defined as the thalamic relay of retinal inputs. What is common to cat and macaque monkey is that each layer is innervated by only one eye. What is different is the partial segregation of parallel pathways through each layer. In the macaque monkey, the P and M pathways use separate layers, the parvocellular for P and the magnocellular for M; the K pathway overlaps with each and is thus represented in all layers. In the cat, the W pathway uses separate layers (C1 and C2), and the Y pathway has exclusive use of layer C, but the X and Y pathways are mingled in the A layers.
Despite the overlap within layers of many of the parallel pathways, there is no functional overlap at the cellular level; within the A layers of the cat, retinal X and Y axons innervate their own classes of relay cell. A similar pattern exists for the macaque monkey, with each of the K, P, and M retinal axons targeting separate classes of geniculate relay cell, whose axons, in turn, have distinct distributions in the visual cortex (see Chapter 31).
New World primates have a slightly different layering arrangement from that shown for the macaque monkey. So far as we know, the human lateral geniculate nucleus is closely comparable to that of the macaque monkey, although there are commonly more than six layers (Hickey and Guillery, 1979), and we have no direct evidence concerning the distribution of distinct functional types to the different layers.
Thus, layering in the lateral geniculate nucleus separates left eye from right eye afferents and also, to a more limited extent, relates to the separation of functionally distinct parallel pathways. The binocular separation is constant across species, but the functional separation into distinct layers is highly variable, so that the number of layers, their sequence from superficial to deep, and their total number show great variability across species. Most of the information we have for cell and circuit properties of the lateral geniculate nucleus specifically and the thalamus more generally comes from the A layers in the cat, and most details presented below are from these layers. However, this focus on the A layers should not obscure two important facts. One is that in terms of the general arrangements of synaptic circuitry, there is a common thalamic plan that applies to the pulvinar region and to most other parts of the thalamus; the other is that structural details vary significantly among different layers and species. That is, there are details of functional organization that remain unknown but almost certainly will ultimately have to be added to our account of the A layers. For some details of geniculate relays beyond the A layers in the cat and in other species, see Jones (1985), Casagrande and Norton (1991), and Casagrande and Kaas (1994).
Cell Types within the A Layers
There are three basic cell types in the A layers of the cat's lateral geniculate nucleus (see Fig. 35.3). These include the two relay cell types, X and Y, and interneurons. The relay cells use glutamate as a neurotransmitter, whereas interneurons use GABA.
Figure 35.3..
Reconstruction of an X cell, Y cell, and interneuron from A layers of the cat's lateral geniculate nucleus. The larger scales are for the insets for the X cell and interneuron.
Relay cells
X and Y cells represent geniculate relays of two parallel and independent geniculocortical pathways, each innervated by its own retinal X or Y axons, which are excitatory. Retinal Y axons are thicker and conduct more rapidly than do X axons (reviewed in Sherman, 1985). Also, within the A layers, the terminal arbors of retinal Y axons are much larger than those of X axons and give rise to many more synaptic terminals (Bowling and Michael, 1984; Sur et al., 1987). As a result of this and because the postsynaptic relay cell types do not differ much in the number of synapses that they receive from the retina, each retinal Y axon innervates many more relay cells than does an X axon. It has been estimated that the X:Y ratio, which is roughly 10:1 in the retina (Leventhal, 1982; Wässle et al., 1975; Wässle et al., 1981), becomes 1:1 or 2:1 for geniculate relay cells (Sherman, 1985).
These geniculate relay cells differ from one another with respect to their functional and morphological properties. Both cell types have the center/surround organization for their receptive fields typical of retinal and geniculate cells generally. However, Y cells have larger receptive fields and respond better to higher temporal and lower spatial frequencies, whereas X cells have smaller receptive fields and respond better to lower temporal and higher spatial frequencies. Further, Y cells exhibit subtly more nonlinear summation (for further details, see Hochstein and Shapley, 1976a, 1976b; Sherman, 1985). All these receptive field differences are already present in the retinal afferents, and for this reason they will not be considered further.
Morphologically, there are some differences between these relay cell types (Friedlander et al., 1981; Guillery, 1966; LeVay and Ferster, 1977; Wilson et al., 1984). At the light microscopic level (see Fig. 35.3), Y cells have larger cell bodies and smooth dendrites that have cruciate branches in a relatively spherical arbor, with peripheral segments of dendrites often crossing from one layer to another. X cells have arbors that tend to be bipolar and oriented perpendicular to the layers' borders. Their dendrites also have numerous clusters of grape-like appendages located mostly near primary branch points (Fig. 35.3). The functional significance of these morphological differences remains to be defined, although the clustered appendages of the X cells represent the postsynaptic site of retinal inputs, where complex synaptic relationships are formed that are characteristic of X but not of Y cells (the triadic arrangements described below). These and other differences in the microcircuitry of these cells types are described below.
Interneurons
Interneurons have the smallest cell bodies in the A layers and long, sinuous dendrites, and the dendritic arbors are always oriented perpendicular to the layers, often spanning an entire layer (see Fig. 35.3). The dendrites have the appearance of terminal axonal arbors and for that reason have been described as axoniform in appearance (Guillery, 1966). Terminals of these dendritic arbors are presynaptic to local dendrites, containing synaptic vesicles and thus resembling axon terminals, and they are also postsynaptic to other axons, generally those coming from either the retina or the brainstem (see also below and Erişir et al., 1997; Famiglietti and Peters, 1972; Hamos et al., 1985; Ralston, 1971). In addition, most, if not all, interneurons have conventional axons that terminate in the general vicinity of the dendritic arbor. The receptive fields of the few identified interneurons that have been studied are like those of X relay cells and unlike those of Y cells, which suggests that the retinal inputs responsible for their firing are X rather than Y axons (Friedlander et al., 1981; Sherman and Friedlander, 1988).
Afferents to the A Layers
The major sources of inputs to the A layers, besides the retina, include the thalamic reticular nucleus, layer 6 of cortex, and the parabrachial region3 of the midbrain. These are summarized in Figure 35.4. Other afferent sources not shown in Figure 35.4 include the nucleus of the optic tract (midbrain), the dorsal raphé nucleus (midbrain and pons), and the tuberomamillary nucleus (hypothalamus) (reviewed in Sherman and Guillery, 1996, 2001).
Figure 35.4..
Neuronal circuitry related to A layers of the cat's lateral geniculate nucleus. Shown are the various inputs, the neurotransmitters associated with them, and the type of receptor, ionotropic or metabotropic, that each activates. Driver versus modulator inputs are also shown (see text for details).
Retinal afferents
Retinal afferents to the A layers are glutamatergic (see Fig. 35.4). They are relatively thick axons and have a distinct terminal structure involving richly branched, dense terminal arbors with boutons densely distributed mostly in flowery terminal clusters (not illustrated; see Guillery, 1966). In contrast, most nonretinal inputs described below are thinner, and have an equally distinct structure with smaller terminals en passant or on short side branches (see Fig. 35.5, right). The retinal axons innervate both relay cells and interneurons in the A layers. Each retinal axon has an arbor strictly limited to one of the A layers, although Y axons and occasional X axons branch to innervate the C layers and/or the medial interlaminar nucleus4 as well. As would be expected from the earlier description of the lines of projection, the terminal arbors of the retinogeniculate axons are also organized so that they occupy relatively narrow columns that are bounded by lines of projection, either within a single layer or across more than one layer. Y arbors are larger and contain more boutons than do X arbors. A more subtle difference between them is that the boutons in Y arbors are fairly evenly distributed, while those in X arbors tend to be found in clusters with gaps between them (Bowling and Michael, 1984; Sur et al., 1987). All retinal X and Y axons innervating the lateral geniculate nucleus branch to innervate the midbrain as well (a point that is discussed further below; see also Guillery and Sherman, 2002a,b), but they do not innervate the thalamic reticular nucleus.
Figure 35.5..
Tracings of partial terminal arbors of three corticothalamic axons in the pulvinar region labeled by biotinylated dextran amine. The axon on the left exhibits driver morphology from cortical layer 5, and the two axons on the right exhibit modulator morphology from layer 6. (From Sherman and Guillery, 2001.)
Afferents from the thalamic reticular nucleus
The thalamic reticular nucleus is a thin shell of GABAergic neurons that surrounds the entire thalamus laterally, extending somewhat dorsally and ventrally. It derives from the ventral thalamus, together with the ventral lateral geniculate nucleus (see footnote 1), and is divided into sectors, each related to thalamic relay nuclei concerned with a particular modality or function (e.g., auditory, somatosensory, and motor), and as in the main part of the thalamus, sensory surfaces are mapped in the sectors (Crabtree, 1992, 1996, 1998; Crabtree and Killackey, 1989; Montero et al., 1977; reviewed in Guillery et al., 1998). There are strong reciprocal connections between relay cells and reticular cells linking corresponding parts of the reticular and geniculate maps (Pinault and Deschênes, 1998; Pinault et al., 1995a, 1995b; Uhlrich et al., 1991), and the cortical afferents from layer 6 (next section) are mapped along the same coordinates (Murphy and Sillito, 1996). That is, the portion of the thalamic reticular nucleus innervating the lateral geniculate nucleus5 is mapped in retinotopic coordinates. In addition, the visual sector of the reticular nucleus is linked reciprocally to the pulvinar region (Conley and Diamond, 1990; Pinault et al., 1995a). There is some evidence that the visual sector of the reticular nucleus can be split into two parts, an inner part linked to the lateral geniculate nucleus and an outer part linked to the pulvinar region. The retinotopic mapping in the pulvinar sector of the reticular nucleus is less accurate than that for the geniculate sector, and there may be no map at all in the former.
The cells of the thalamic reticular nucleus, which lie just dorsal to layer A, have moderate to large cell bodies and dendrites oriented mostly parallel to layer A (Uhlrich et al., 1991). Their axons descend into the A layers, generally along the lines of projection, with terminal arbors that are moderately branched and contain numerous boutons, mostly en passant. These terminals innervate geniculate relay cells, but they provide only a very sparse innervation to interneurons (Cucchiaro et al., 1991; Wang et al., 2001). Thus, the thalamic reticular nucleus provides a potent inhibitory GABAergic input to relay cells (Fig. 34.4). Their receptive fields tend to be larger than those of relay cells and are often binocular (So and Shapley, 1981).
Afferents from layer 6 of the cortex
Cortical afferents from layer 6, which are glutamatergic, have thin axons with most boutons located at the end of short side branches (Fig. 34.5; see Murphy and Sillito, 1996). They are topographically organized, with each axon having terminal arbors bounded by lines of projection and passing across more than one layer. These axons enter the A layers after traveling through the thalamic reticular nucleus, where they also give off branches to innervate cells there; this projection, too, is topographic.
Afferents from the parabrachial region
Most of the input from the brainstem to the A layers derives from the parabrachial region (Bickford et al., 1993; de Lima and Singer, 1987). These axons are cholinergic but also appear to colocalize nitric oxide (Bickford et al., 1993; Erişir et al., 1997). Light microscopically, they resemble the cortical afferents more than the retinal afferents, but their terminal arbors are rather diffuse and most appear to terminate in a nontopographic fashion. These axons contact both relay cells and interneurons in the A layers and also branch to innervate cells in the thalamic reticular nucleus.
Other afferents
Some other afferents to the A layers not shown in Figure 35.4 have been described, but they are small in number, not well documented, and will be mentioned only briefly here (for further details, see Sherman and Guillery, 1996, 2001). Lying among the cholinergic cells of the parabrachial region, there are also some noradrenergic cells that innervate the A layers. There is limited serotonergic input from the dorsal raphé nucleus in the midbrain and pons. GABAergic cells of the nucleus of the optic tract in the midbrain also provide limited input. Finally, the tuberomamillary nucleus of the hypothalamus provides a small histaminergic input (Uhlrich et al., 1993).
Postsynaptic receptors
In addition to showing the inputs and their transmitters onto relay cells, Figure 35.4 shows the associated postsynaptic receptors. Note that both ionotropic and metabotropic receptors are postsynaptic in relay cells. There are a number of differences between these two receptor types, but only a few concern us here (for details, see Brown et al., 1997; Conn and Pin, 1997; Mott and Lewis, 1994; Nicoll et al., 1990; Pin and Duvoisin, 1995; Recasens and Vignes, 1995).
Ionotropic receptors include AMPA receptors for glutamate, GABAA, and nicotinic receptors for acetylcholine. These are complex proteins found in the postsynaptic membrane, and when the transmitter contacts the receptor, it leads to a rapid conformational change that opens an ionic channel, leading to transmembrane flow of ions and a postsynaptic potential. Activation of ionotropic receptors leads to fast responses, typically with a latency for postsynaptic potentials of less than 1 msec and a duration of a few tens of milliseconds. Metabotropic receptors include various glutamate receptors, GABAB, and various muscarinic receptors for acetylcholine. These are not directly linked to ion channels. Instead, when the transmitter contacts the receptor protein in the membrane, a series of complex biochemical reactions takes place that ultimately leads to the opening or closing of an ion channel, among other postsynaptic events. For thalamic cells, this is primarily a K+ channel that, when opened, produces an inhibitory postsynaptic potential (IPSP) as K+ flows out of the cell and, when closed, produces an excitatory postsynaptic potential (EPSP) as leakage of K+ is reduced. However, these postsynaptic responses are slow: there is usually a latency of 10 msec or longer and a duration of hundreds of milliseconds or more. Also, in general, metabotropic receptors require higher firing rates from inputs to be activated. This is thought to be related to the observation that electron micrographs show these receptors to be located slightly farther from the synaptic site than are ionotropic receptors, so that more transmitter must be released to reach them.
The more sustained responses associated with metabotropic receptors have a number of important implications. One has to do with the fact that retinal inputs activate only ionotropic receptors in relay cells. This means that retinal EPSPs are relatively fast and brief. This has the virtue that firing in the retinal afferents can reach relatively high levels before temporal summation of the EPSPs occurs, and thus each retinal action potential has a unique postsynaptic response associated with it. If, instead, metabotropic glutamate receptors were activated, the sustained EPSPs would mean that relatively low rates of firing in the afferent would produce temporal summation postsynaptically. This, in turn, means that higher-frequency information would be lost in retinogeniculate transmission or, more formally, that the retinogeniculate synapse would operate like a low-pass temporal device, filtering out higher frequencies. Thus, the fact that retinal inputs activate only ionotropic receptors serves to maximize the transfer of higher temporal frequencies. Note that the population of nonretinal inputs as a whole activates both metabotropic and ionotropic receptors (reviewed in Sherman and Guillery, 1996, 2001). However, it is not clear whether any individual nonretinal axon can activate both ionotropic and metabotropic receptors. Nonetheless, the activation of metabotropic receptors means that these inputs can create sustained changes in baseline membrane potential, which, among other things, means that these inputs can have sustained effects on the overall responsiveness of relay cells. Other consequences of these sustained postsynaptic responses are considered below.
Synaptic structures
Over 95% of all synaptic terminals in the A layers can be placed into one of four categories (reviewed in Sherman and Guillery, 1996, 2001): (1) RL (Round vesicle, Large profile) terminals, which are the retinal terminals, are the largest terminals in the A layers. They form asymmetric6 contacts consistent with their identity as excitatory inputs. (2) RS terminals (Round vesicle, Small profile) are smaller than RL terminals but also form asymmetric contacts. The vast majority of these come from either layer 6 of cortex or the parabrachial region. (3) F1 terminals (Flattened vesicles) form symmetric contacts consistent with their origin from axons of reticular cells or interneurons. (4) F2 terminals represent the dendritic outputs of interneurons; they also have flattened vesicles and form symmetric contacts. Unlike all of the other terminals, which are strictly presynaptic, these are both presynaptic and postsynaptic, with inputs from either retinal or parabrachial terminals.
Triadic synaptic arrangements involving F2 terminals are common in the A layers (Fig. 35.6). In some triads, an RL terminal contacts both an F2 terminal and the dendrite of a relay cell, and the F2 terminal contacts the same relay cell dendrite. A slightly different kind of triad can be formed by a parabrachial terminal contacting an F2 terminal and a different parabrachial terminal from the same axon contacting a relay cell dendrite, again with the F2 terminal contacting the same relay cell dendrite (Fig. 35.6). Nearly all F2 terminals are involved in one or the other form of triad. Curiously, these triads are quite common for relay X cells and rare for Y cells, the latter thus receiving very few inputs from F2 terminals.7 Triads are typically found in complex synaptic zones that lack astrocytic processes but are surrounded by sheets of astrocytic cytoplasm; these are called glomeruli. It is not at all clear how the triads function.
Figure 35.6..
Schematic view of triadic circuits in a glomerulus of the lateral geniculate nucleus in the cat. The arrows indicate presynaptic to postsynaptic directions. The question marks postsynaptic to the dendritic terminals of interneurons indicate that it is not clear whether or not metabotropic (GABAB) receptors exist there.
Distribution of inputs to relay cells
The dendritic arbors of relay cells can be divided into two distinct sectors with little or no overlap (Erişir et al., 1997; Wilson et al., 1984): a proximal region (up to about 100 µm from the cell body or generally close to the first branch point) and a distal region (farther than about 100 µm from the cell body). Retinal terminals contact the former region, whereas cortical terminals contact the latter. F2 and parabrachial terminals also contact relay cells in the proximal zone. Axonal inputs from interneurons mostly contact the proximal zone, whereas those from reticular axons mostly contact the distal zone.
A small minority of synaptic inputs onto geniculate relay cells derive from retina. In the A layers of the cat's lateral geniculate nucleus, for instance, only about 5–10% of the synaptic input to relay cells comes from the retinal axons: roughly 30% comes from local GABAergic cells (interneurons plus reticular cells), 30% from the cortical input, and 30% from the parabrachial region (Van Horn et al., 2000). In the parvocellular C layers, even fewer synapses—2% to 4%—onto relay cells derive from the retina (Raczkowski et al., 1988). If one had only the anatomical data, and for many other thalamic relays that is all we have, one might well conclude that the lateral geniculate nucleus relays parabrachial input to cortex and that the retinal input plays only a minor, undetermined role For the lateral geniculate nucleus we know that it is the retinal input that is relayed to cortex, so we accept that the small number of retinal afferents serve as the crucial drivers of geniculate function. However, for thalamic nuclei that we do not understand as well as the lateral geniculate nucleus, the point is important, as we will see when we discuss corticocortical communication.
Drivers and modulators
It follows that not all inputs to the thalamus are equal in their action on the relay cells. We have distinguished two different types of input (Sherman and Guillery, 1998, 2001): drivers and modulators. The drivers are the information-bearing input that is to be relayed to cortex, and this is the retinal input for the lateral geniculate nucleus. All other inputs are modulators. Examples of modulation are provided below, and details of how drivers might generally be distinguished from modulators is provided elsewhere (Sherman and Guillery, 1998, 2001). One difference is seen in Figure 35.4: the driver (retinal) input activates only ionotropic receptors, whereas the modulators activate metabotropic and often ionotropic receptors. Note that, of the main extrinsic inputs to the lateral geniculate nucleus, the retinal input is a driver, but the layer 6 cortical and parabrachial inputs are both modulators. The corticothalamic input must be seen as modulatory because, among other reasons, its elimination (by cooling, ablation, etc.) has only subtle effects on the receptive fields of geniculate relay cells, not altering their basic center/surround organization (Cudeiro and Sillito, 1996; Geisert et al., 1981; Jones and Sillito, 1991; Kalil and Chase, 1970; McClurkin and Marrocco, 1984; McClurkin et al., 1994). This is in contrast to corticothalamic afferents from layer 5, which go to higher-order thalamic relays but not to the lateral geniculate nucleus, and which must be regarded as drivers because their elimination essentially abolishes the characteristic receptive fields in such target thalamic relays as the posterior medial nucleus or pulvinar region (Bender, 1983; Chalupa, 1991; Diamond et al., 1992; see also the section “The Pulvinar Region as a Visual Relay”).
Intrinsic Properties of Thalamic Cells in the A Layers
The relay nature of the lateral geniculate nucleus depends on the mechanisms by which retinal inputs evoke firing in geniculate relay cells, and these mechanisms are also present in thalamic relays more generally. There are three factors that largely control this retinogeniculate transmission, and they are considered below. First are the intrinsic membrane properties of relay cells, including their passive and active membrane properties, because these determine the effect of retinal EPSPs at the cell body or region of action potential generation. Second is the geniculate circuitry that, by affecting many of the intrinsic membrane properties, also controls the effect of retinal EPSPs on relay cell firing. Third, the nature of the postsynaptic receptors largely determines the postsynaptic response of relay (and other) cells to their active inputs; this feature is considered in the section “Control of Response Mode.”
Generally, all thalamic cells show a wide range of intrinsic membrane properties that are found generally in neurons of the brain (reviewed in Sherman and Guillery, 1996, 2001). These include passive cable properties, voltage-sensitive and -insensitive conductances, and conductances sensitive to other factors, such as Ca2+ concentration. The conductances underlie transmembrane currents, including a leak K+ current (IK[leak]) that helps control the resting membrane potential, various voltage- and Ca2+-gated K+ currents (IA, IK[Ca2+], etc.), and a voltage-gated cation current (Ih). Since these are properties found widely in the brain, they will not be considered further here, but additional details of these as they apply to thalamic neurons can be found in Sherman and Guillery (1996, 2001) Two features that are of particular interest in thalamic neurons are considered below. One is the apparent cable properties of interneurons, which suggests that synaptic inputs onto their dendritic terminals affect them locally but have little effect on the cell body and axonal output, permitting these cells to provide numerous input/output routes independently and simultaneously (see below). The other is the presence in all cells of a voltage-gated Ca2+ conductance based on T-type (for transient) Ca2+ channels that, when activated, leads to a current (IT) large enough to produce an all-or-none Ca2+ spike (reviewed in Sherman and Guillery, 1996, 2001). This spike alters the response properties of the thalamic cells in a functionally highly significant manner.
Passive membrane or cable properties
Determining how current spreads through cells with complex geometries of their dendritic trees poses a formidable computational problem. Modeling the cell and its membranes as a cable provides a useful means of approximating current flow. Details of cable modeling for thalamic neurons can be found in Bloomfield et al. (1987) and Bloomfield and Sherman (1989) and will be briefly summarized here. Relay cells and cells of the thalamic reticular nucleus have relatively thick dendrites that branch in a way that suggests efficient current flow through the dendrites. The result of cable modeling for these cells indicates that they are electrotonically compact, meaning that EPSPs and IPSPs generated even on peripheral dendrites will produce significant voltage changes at the cell body and axon hillock (Bloomfield et al., 1987; Bloomfield and Sherman, 1989). Thus, all synaptic inputs to these cells can be considered influential in affecting the cell's firing. Since the dendrites of relay cells are purely postsynaptic structures, the only synaptic output of these cells is via their axons, so that any synaptic input to the distal dendrites that could not produce a voltage change at the spike-initiating region would be ineffective for spike initiation.
In contrast to relay cells, interneurons are built differently in two ways. First, as noted above, in addition to having a conventional axonal output, these cells have presynaptic terminals on peripheral dendrites, which provide these cells with numerous dendritic outputs. Second, these presynaptic terminals are attached to the stem dendrites by long, thin stalks; overall, the thin dendrites and the nature of the dendritic branching suggest poor current flow through the dendritic trees. Cable modeling suggests that EPSPs and IPSPs generated on the peripheral dendrites, where the retinal and brain stem axons provide significant input to the dendritic presynaptic boutons, will have little influence on the cell body and axon hillock (Bloomfield and Sherman, 1989). Thus, the interneuron appears capable of multiplexing: the axonal output is controlled by synapses located on proximal dendrites, whereas the several separate dendritic outputs are controlled locally and independently of each other by local synaptic inputs. However, this concept of the functioning of the interneuron remains a hypothesis that requires more direct evidence than currently exists (Cox and Sherman, 2000).
Properties of IT in relay cells
The voltage-dependent low-threshold Ca2+ spikes that are based on T channels are ubiquitous to thalamic relay cells: they have been found in every relay cell of every thalamic nucleus of every mammalian species so far studied (reviewed in Sherman and Guillery, 1996). Figure 35.7 shows the voltage dependence of the T channels and that of K+ channels also involved in the generation of the low-threshold spikes. The T channels have two voltage-sensitive gates, an activation gate and an inactivation gate, and both must be open for Ca2+ to flow into the cell and thus depolarize it. At the normally hyperpolarized resting membrane potentials (Fig. 35.7(1)), the activation gate is closed but the inactivation gate is open, so the channel is deinactivated. The single gate of the K+ channel is closed at this membrane potential. If the cell is now sufficiently depolarized (e.g., by an EPSP), the activation gate pops open, and Ca2+ flows into the cell, providing the uspwing of the low-threshold spike (Fig. 35.7(2)). However, depolarization causes the inactivation gate to close (Fig. 35.7(3)), but this takes time, on the order of 100 msec or so.8 The single gate of the K+ channel, because of its voltage dependency, also opens, and the combined inactivation of the T channel and activation of the K+ channel serve to repolarize the cell (Fig. 35.7(4)). Although not shown, in addition to voltage-dependent K+ channels, Ca2+-dependent ones are also activated by the Ca2+ entry and assist the repolarization process. While the membrane is repolarized to its initial potential, the T channel remains inactivated, because it takes roughly 100 msec of this hyperpolarization to deinactivate these channels, after which time the initial conditions are reestablished (Fig. 35.7(1)). To reiterate: when the cell is sufficiently hyperpolarized for more than about 100 msec, the T channel is deinactivated; if deinactivated, a suitable depolarization can then activate the channel, but continued depolarization for more than about 100 msec will inactivate it; the inactivation can then be removed by suitable hyperpolarization for more than about 100 msec.
Figure 35.7..
Highly schematized view of the actions of voltage-dependent T (Ca2+) and K+ channels underlying a low-threshold Ca2+ spike. The four numbered panels show the sequence of channel events, and the central graph shows the effects on membrane potential. The T channel has two voltage-dependent gates: an activation gate that closes at hyperpolarized levels and opens with depolarization, and an inactivation gate that shows the opposite voltage dependency. The K+ channel shown is really a conglomeration of several such channels that have only a single gate that opens during depolarization; thus, these channels do not inactivate. 1, At a relatively hyperpolarized resting membrane potential (∼70 mV), the activation gate of the T channel is closed but the inactivation gate is open, so the T channel is deinactivated. The single gate for the K+ channel is closed. 2, With sufficient depolarization to reach its threshold, the activation gate of the T channel opens, allowing Ca2+ to flow into the cell. This depolarizes the cell, providing the upswing of the low-threshold spike. 3, The inactivation gate of the T channel closes after roughly 100 msec (“roughly,” because closing of the channel is a complex function of time and voltage), and the K+ channel also opens. These combined actions lead to the repolarization of the cell. While the inactivation gate of the T channel is closed, the channel is said to be inactivated. There are probably several different kinds of K+ channels involved with different time constants, but in general, they open more slowly than does the activation gate of the T channel. Also, not shown, K+ channels dependent on Ca2+ entry are probably involved. 4, Even though the initial resting potential is reached, the T channel remains inactivated, because it takes roughly 100 msec (“roughly” having the same meaning as above) of hyperpolarization to deinactivate it; it also takes a bit of time for the various K+ channels to close. Note that the behavior of the T channel is qualitatively exactly like that of the Na+ channel involved with the action potential, but with several quantitative differences: the T channel is slower to inactivate and deinactivate, and it operates in a more hyperpolarized regime.
This voltage dependency of the T channels provides the relay cell with two different firing modes: if the cell is relatively depolarized, the T channels are inactivated and do not participate in the cell's responses; here the cell is said to be in tonic firing mode. If the cell is relatively hyperpolarized, the T channels are deinactivated and thus primed for action. They can become activated and, on the basis of mechanisms considered below, can affect how the cell fires; here the cell is said to be in burst firing mode.
The voltage-dependent properties of the T channels are qualitatively identical to those of the Na+ channels underlying the action potential, but there are several important quantitative differences: (1) Opening or closing of the inactivation gate is roughly two orders of magnitude faster for the Na+ channel. (2) The T channels are found in the cell body and membranes, but not in the axon; thus, the low-threshold spike can be propagated through the dendrites and cell body but not along the axon to cortex. That is, the T channels can affect the signal reaching cortex by the effect of the low-threshold spike on the generation of action potentials (see below). (3) The T channels operate in a somewhat more hyperpolarized regime, and their activation at more hyperpolarized levels is the reason the resultant Ca2+ spike is called low threshold. (4) The T channels operate in a slightly more hyperpolarized regime than the Na+ channels.
Figure 35.8 shows recordings from relay cells of the cat's lateral geniculate nucleus, illustrating some of the functional consequences of IT. When the membrane is more depolarized than roughly −60 to −65 mV for more than ∼100 msec, IT becomes inactivated (Fig. 35.8A), and activation by a depolarizing pulse evokes a steady stream of action potentials lasting for the duration of the stimulus: this is tonic firing. When the membrane is more hyperpolarized than about −65 to −70 mV for more than ∼100 msec (Fig. 35.8B), IT becomes deinactivated. Now the very same depolarizing pulse activates IT, leading to a low-threshold spike, which, in turn, leads to a burst of a few action potentials: this is burst firing. Note that the same excitatory stimulus produces two very different signals relayed to cortex, and the difference depends on the initial membrane potential of the relay cell, because this determines the inactivation state of IT. The stimulus in this example is a current pulse, but the same would apply to a sufficiently large EPSP.
Figure 35.8..
Properties of IT and the low-threshold spike; examples from intracellular in vitro recordings of geniculate relay cells of the cat. A, B, Voltage dependency of the low-threshold spike. Responses are shown to the same depolarizing current pulse delivered intracellularly but from two different initial holding potentials. With relative depolarization of the cell (A), IT inactivates, and the cell response is a stream of unitary action potentials lasting for the duration of a suprathreshold stimulus. This is the tonic mode of firing. With relative hyperpolarization of the cell (B), IT deinactivates, and the current pulse activates a low-threshold spike with four action potentials riding its crest. This is the burst mode of firing. C, Voltage dependency of low-threshold spike amplitude and associated burst response. Examples for two cells are shown. The number of action potentials were recorded first in the experiment, and then the procedure was repeated after tetrodotoxin (TTX) application to eliminate action potentials and isolate the low-threshold spike for measurement. The more hyperpolarized the cell before activation (Initial Membrane Potential), the more action potentials (AP) in the burst (open circles) and the larger the low-threshold spike (filled squares and curve). D, All-or-none nature of low-threshold spikes in another geniculate cell, measured in the presence of TTX. After initial hyperpolarization of the cell, current pulses were injected in 10 pA incremental steps. Smaller (subthreshold) pulses produced resistive-capacitative responses, but all larger (suprathreshold) pulses evoked low-threshold spikes that are all of the same amplitude, regardless of how far the depolarizing pulse exceeded activation threshold. (C redrawn from Sherman, 2001; Zhan et al., 2000; D redrawn from Zhan et al., 1999.)
The size of the activated low-threshold spike depends on the extent to which the cell is hyperpolarized before being activated, because the more the cell is hyperpolarized, the more T channels are deinactivated and thus available to contribute to the low-threshold spike (Fig. 35.8C). As would be expected, the larger the low-threshold spike, the larger the number of action potentials evoked in the associated burst (Fig. 35.8C). However, in addition, the low-threshold spike is activated in an all-or-none manner, because at any one initial membrane potential, suprathreshold activating inputs activate low-threshold spikes of essentially the same amplitude, regardless of how far above threshold the activating input is (Fig. 35.8D). One implication of this for the difference in input/output relationships between burst and tonic firing is shown in Figure 35.9. This relationship is fairly linear for tonic firing, because there is a direct link between the input depolarization and activation of action potentials, leading to the monotonic relationship as shown. However, the relationship is indirect for burst firing, since it is the low-threshold spike that controls firing, and, as Figure 35.8D shows, larger activating inputs do not produce larger low-threshold spikes.
Figure 35.9..
Input output relationship for a geniculate relay cell recorded intracellularly in vitro. The input variable is the amplitude of the depolarizing current pulse, and the output is the firing frequency of the cell. To compare burst and tonic firing, the firing frequency was determined by the first six action potentials of the response, since this cell usually exhibited six action potentials per burst in this experiment. The initial holding potentials are shown; −47 mV and −59 mV reflect tonic mode (open squares and curves), whereas −77 mV and −83 mV reflect burst mode (filled curves). (Redrawn from Zhan et al., 1999.)
Properties of IT in interneurons and reticular cells
IT is present in both interneurons and reticular cells, but its action is subtly different in these cells. Its activation in interneurons is a matter of some controversy. Several studies suggest that IT is rarely activated in these cells, because it is masked by IA (Pape et al., 1994). IA is a K+ conductance with a voltage dependence similar to that of IT: it is inactivated at depolarized levels and deinactivated at hyperpolarized levels, from which it can be activated (for details, see McCormick, 1991). However, unlike IT, IA creates hyperpolarization, since the current is composed of K+ ions leaving the cell. Some evidence suggests that, in interneurons, the relative voltage dependency of these two currents is such that IA is typically activated before IT, thereby preventing activation of IT (Pape et al., 1994). In relay cells, the relative voltage dependency is different, so that IT is activated first (Pape et al., 1994). However, other evidence suggests that if the input resistance of the interneuron is high enough, low-threshold spiking is readily produced (Zhu et al., 1999). At present, it is difficult to reconcile these views. Nonetheless, it should be appreciated that tonic firing and burst firing in the interneuron, to the extent that they exist, describe axonal outputs of interneurons, and their dendritic outputs may follow quite different and independent patterns.
Cells of the thalamic reticular nucleus also have burst firing based on voltage-dependent T channels, but the temporal properties are slightly different from those in relay cells, leading to longer low-threshold spikes and more prolonged bursts in reticular cells (for details, see Destexhe et al., 1996).
Significance of Burst and Tonic Firing
The first studies of thalamic bursting in vivo emphasized the presence of rhythmic bursting in thalamic relay cells during slow-wave sleep and certain pathological conditions, such as epilepsy, in which this bursting is synchronized across large cell populations. In these conditions, such bursting also interferes with normal relay functions, and it was thus considered not to be a relay mode of firing (Fanselow et al., 2001; Steriade and Llinás, 1988; Steriade and McCarley, 1990; Steriade et al., 1990, 1993).
Because such rhythmic bursting was not seen during waking behavior, this led to the notion that bursting occurs only during sleep or pathological state and that tonic firing is the normal relay mode during waking behavior. However, it is now clear that bursting also occurs during normal waking behavior (Edeline et al., 2000; Fanselow et al., 2001; Lenz et al., 1998; Magnin et al., 2000; Ramcharan et al., 2000; Swadlow and Gusev, 2001), but because the bursting then is not rhythmic it is harder to detect, and this perhaps explains why it was not emphasized in earlier studies. It may also be that the rhythmic bursting seen during slow-wave sleep provides a positive signal to cortex that nothing is being relayed despite the possible presence of sensory stimuli, and this is less ambiguous than no activity, which could mean either no relay or no stimulus. In contrast, when the bursting is arrhythmic, this arrhythmicity can represent responses evoked by sensory stimuli.
Most relay cells during waking behavior fire more often in tonic mode, and the amount of bursting seems to go down to small values as the animal becomes more alert (Ramcharan et al., 2000; Swadlow and Gusev, 2001). Nonetheless, both modes effectively relay retinal information to cortex (Ramcharan et al., 2000; Reinagel et al., 1999; Sherman, 2001), and thus the presence in an awake animal of both modes, tonic and nonrhythmic bursting, raises the obvious question: what is the significance of these two modes? They represent very different ways in which the relay cell responds to the same input, indicating that the same message is relayed to cortex in one of two different ways. Thus, when messages arrive at the relay cell, the level of its membrane potential, which determines the inactivation state of IT, can strongly influence the nature of the information that is transmitted to cortex. Receptive field analysis from relay cells of the cat's lateral geniculate nucleus indicates that both response modes convey comparable levels of information in the relay to cortex (Reinagel et al., 1999), although it is also clear that the nature of that information differs between modes (Sherman, 1996, 2001). There may be many differences related to these two modes, but two that have received considerable attention are linearity of the relay process and detectability of the message that is relayed to cortex.
Linearity
From the cellular properties described above (e.g., Fig. 35.9), it is clear that tonic firing represents a more linear relay mode. This is also seen in the responses of geniculate relay cells to visual stimuli. A clear example is shown in Figure 35.10, which shows the responses to a drifting sinusoidal grating of a relay cell recorded in vivo in an anesthetized cat. When the cell is in tonic mode, the response to the grating has a sinusoidal profile (Fig. 35.10A, lower). This means that the response level closely matches the changes in contrast, indicating a very linear relay of this input to cortex. However, when the same stimulus is applied to the same cell, but now in burst mode, the response no longer seems sinusoidal (Fig. 35.10B, lower), indicating considerable nonlinear distortion in the relay. Thus, tonic mode is better at preserving linearity in the relay of information to cortex (Sherman, 1996, 2001).
Figure 35.10..
Tonic and burst responses of relay cells from the cat's lateral geniculate nucleus to visual stimulation. A, B, Average response histograms of responses of one cell to four cycles of drifting sinusoidal grating (lower) and during spontaneous activity (upper). The contrast changes resulting from the drifting grating are shown below the histograms. The cell was recorded intracellularly in vivo, and current injected through the recording electrode was used to bias membrane potential to more depolarized (−65 mV), producing tonic firing (A), or more hyperpolarized (−75 mV), producing burst firing (B).
Detectability
The upper histograms of Figure 35.10 show further that spontaneous activity is lower during tonic than burst firing. Higher spontaneous activity is actually useful for maintaining linearity of response, because it helps prevent rectification of the response to inhibitory phases of visual stimulation, and rectification is a nonlinearity. Perhaps more interesting is the notion that spontaneous activity represents firing without a visual stimulus and can thus be considered a noisy background against which the signal—the response to the visual stimulus—must be detected. In this regard, the signal-to-noise ratio appears to be higher during burst firing, and indeed, the use of a method from signal detection theory involving the calculation of receiver operating characteristic curves (Green and Swets, 1966; Macmillan and Creelman, 1991) shows that stimulus detectability is improved during burst firing compared to tonic firing (Sherman, 1996, 2001).
Bursting as a “wake-up call.”
The above differences in firing modes as regards linearity and detectability suggest the following hypothesis (Sherman, 1996, 2001). Tonic firing is better for faithful, detailed reconstruction of the stimulus, because the nonlinear distortions created during burst firing would limit the extent to which cortex receives an accurate copy of the messages that are being passed through the thalamus. However, burst firing would be better for initial stimulus detectability. For instance, when an animal is drowsy, it might be advantageous to have geniculate relay cells in burst mode to maximize detection of a new visual stimulus; after detection, the relay can be switched to tonic firing for more faithful stimulus analysis (for details of this hypothesis, see Sherman, 1996, 2001). Indeed, there is evidence from the somatosensory system of awake, behaving rabbits that thalamic relay cells in burst mode are much more likely to evoke an action potential in their cortical target cells than when these relay cells are firing in tonic mode (Swadlow and Gusev, 2001). This finding that bursts punch through the thalamocortical synapse much more effectively than tonic firing is consistent with, but of course falls short of proving, the notion that bursts serve as a wake-up call.
Control of response mode
For this hypothesis to be plausible, there must be ways in which thalamic circuitry can be employed to control the firing mode. In fact, the functional circuitry shown schematically in Figure 35.4 provides this requirement. As noted above, to switch the inactivation state of IT requires a change in membrane voltage that must be sustained for roughly ≥100 msec. Sustained depolarization is needed to inactivate IT and sustained hyperpolarization for deinactivation. Activation of ionotropic receptors with their fast postsynaptic potentials seems poorly suited to this task, because without extensive temporal summation, the changes in membrane polarization would be too transient to affect the inactivation state of IT significantly. Metabotropic receptors are much better suited for this task, since their activation would produce sufficiently sustained postsynaptic potentials. Thus, activation of metabotropic glutamate receptors from cortex or muscarinic receptors from the parabrachial region produces a sufficiently long EPSP to inactivate IT and switch the firing mode from burst to tonic. In contrast, activation of GABAB receptors, mainly from activation of reticular inputs but also possibly from interneuronal inputs, produces a sufficiently long IPSP to deinactivate IT and switch the firing mode from tonic to burst. Indeed, evidence for such switching from activation of these inputs exists from both in vivo and in vitro studies (reviewed in Sherman and Guillery, 1996, 2001).
Ultimately, it is the cortical and parabrachial inputs that control firing mode via their direct inputs to relay cells, which promote tonic firing, and their indirect inputs, via reticular (and possibly interneuron) inputs, which promote burst firing. At the cellular level, both inputs seem to have the same effect. However, the corticogeniculate pathway is highly topographic and purely visual, so that this pathway would presumably control firing mode for groups of geniculate cells based on specific visual parameters such as different locations within the visual field. The diffuse nature of the parabrachial input suggests that its effects are more widespread in the lateral geniculate nucleus and might relate more globally to overall levels of attention (e.g., more bursting exists during states of drowsiness; see Ramcharan et al., 2000; Swadlow and Gusev, 2001) or to which sensory system is being used.
This is not to say that the only purpose of cortical and brainstem inputs is to control firing mode. For example, the corticogeniculate input seems quite heterogeneous, and several other functions have been proposed for it (e.g., McClurkin and Marrocco, 1984; McClurkin et al., 1994; Schmielau and Singer, 1977; Sillito et al., 1993, 1994). The point here is that these multiple functions are not mutually exclusive.
| |
