Current knowledge about the thalamus is the outcome of a long process of data acquisition and conceptual refinement that is still ongoing. Strides along this process have been marked, to a large extent, by the introduction of new research tools. Here, I summarize this sequence of events along two main lines: (a) the identification of the thalamus and its constituent nuclei and (b) the development of conceptual frameworks about thalamus function.
The word thalamus is a Latin transliteration of the ancient Greek ϑάλαμος, the name for a bridal bed and, by extension, for the recess or chamber in a mansion for such a bed. Although formal and dated, the word tálamo retains this meaning in modern Latin-derived languages, such as Spanish, Portuguese, and Italian. Its first recorded use for naming a brain structure is by the second-century ce Roman physician Claudius Galenus, who investigated brain anatomy by skillfully dissecting fresh animal brain specimens. When he mentioned a “thalamus of the ventricles,” however, Galenus was not referring to what we today call the thalamus but to a recess of the lateral ventricles that, he surmised, contacted the end of the optic tract. Because Galenus believed that a “mind fluid” (pneuma psychicon) contained within the ventricles should somehow flow into the optic nerve toward the eye to make seeing possible, he regarded the identification of this point of contact as a major discovery (Reference RoccaRocca 2003).
Following Galenus’s view, Mondino de Luzzi (thirteenth century), in his pioneering dissections of human brains, used a different term, “buttocks (anche) of the ventricles,” to name the two round, adjoining masses he observed at the base of the lateral ventricles (Reference Gailloud, Carota, Bogousslavsky and FaselGailloud et al. 2003). Similar names were used by anatomists in the following three centuries (Reference García-Cabezas, Pérez-Santos and CavadaGarcía-Cabezas et al. 2021). However, in their influential anatomical treatises, first Jean Reference RiolanRiolan (1626) and then Thomas Willis (1664; Figure 1A) labeled the two round masses thalami of the optic nerves. Moreover, they attributed this denomination to Galenus himself, hence consolidating its usage in subsequent literature. The association with the optic nerves persisted in the anatomical literature, although already by 1775, Giovanni Domenico Reference SantoriniSantorini (1775) had shown that the optic tract actually terminates in the lateral geniculate body, a gray mass he described as adjacent to but separate from the rest of the human thalamus.
The introduction in the early nineteenth century of ethyl alcohol as tissue fixative and hardener allowed the first macroscopic descriptions of the inner structure of the thalamus. Applying this method to human brains and aided with a hand magnifying glass, Karl Friedrich Reference BurdachBurdach (1822) identified the lamina medullaris interna within the gray mass of the thalamus and used this fiber tract as a reference to separate some of the main nuclear groups we recognize today, which he called superior (anterior), internal (medial), and external (ventral/lateral). He also named the pulvinar (Latin for “cushion”) because of its shape as seen from behind, protruding in the quadrigeminal cistern. He confirmed Santorini’s description of the medial and lateral geniculate bodies and noted that a stratum corneum (the external medullary lamina) wrapped them together with the rest of the thalamus.
The next advance was the combination by Reference GerlachGerlach (1858) of a fixative (chromic acid) with a dye (carmine-gelatin) to increase the contrast between tissue regions. The neurologist Julius Bernard Reference LuysLuys (1865) applied Gerlach’s method to identify the cell groups or cèntres (centers) that, he supposed, should exist within the thalamus to act as separate conduits for the sensory pathways to the cerebral cortex. Luys’s delineation of four “centers” and their associated sensory channels was not confirmed by subsequent studies, but his cèntre median label remains in use for a nucleus that is prominent in the thalamus of humans and other primates (numbered 10 and 10’ in Figure 1B).
In the second half of the nineteenth century, the introduction of improved light microscopes, along with sliding microtomes, tissue embedding in paraffin, mounting media, and new synthetic chemicals able to selectively stain specific tissue components, allowed an increasingly refined microscopical analysis of the brain. Employing potassium dichromate and gold chloride to selectively stain myelin, Theodor Reference Meynert and StrickerMeynert (1872) was able to identify and name in human and monkey brains most of the major myelinated fiber tracts entering or surrounding the thalamus, such as the optic tract, ansa peduncularis, fornix, brachium of the inferior colliculus, and habenulo-interpeduncular tract. He also identified the habenula and subdivided the external mass of Burdach into a lateral and a ventral portion (the modern lateral posterior and ventral posterior nuclei) because of the high prevalence of myelinated fiber bundles in the latter. Meynert’s student Auguste Reference ForelForel (1877) published the first complete collection of drawings of myelin-stained thalamus sections and reviewed and modified Meynert’s nomenclature. In doing so, Forel’s name became attached to some reticular regions (“fields”) of the ventral thalamus.
In 1894, as he was still a medical student in Munich, Franz Reference NisslNissl (1894) discovered that cresyl violet, an aniline dye derived from tar, could be applied on glass-mounted thin tissue sections to selectively stain acidic cell structures such as polyribosomes (known today as Nissl’s bodies), nuclear chromatin, and nucleoli in the neuronal cell somata and the glial and endothelial cells. His method was rapidly adopted by others and remains in wide use. The method allowed a delineation of brain cell masses based on population-level differences in their packing density, soma size, and/or staining intensity, as seen under low or medium light-microscope magnification. Using this so-called cytoarchitectonic approach, Nissl later published a complete description of the nuclei of the rabbit thalamus (Reference NisslNissl 1913).
In the following decades, thalamus research focused mainly on cytoarchitectonic and myelin architecture studies in different mammal species. As studies were conducted in different institutions by researchers writing in English, French, or German, they led to a proliferation of inconsistent delineations and terminologies. Finally, a comprehensive review on the structure and connections of the thalamus by Wilfrid Reference Le Gros ClarkLe Gros Clark (1932) opened a gradual process of consolidation that led to the adoption of a common nomenclature for about thirty nuclei in the thalamus of laboratory species such as rats, rabbits, cats, and dogs. Consensus took longer to emerge for the thalamus of nonhuman primates (Reference VogtVogt 1909; Reference FriedemannFriedemann 1912; Reference WalkerWalker 1938; Reference JonesJones 2007), mainly because of the complexities involved in subdividing the massive ventral and lateral posterior-pulvinar nuclei of primates.
Meanwhile, the nomenclature and delineation of the human thalamus followed an independent path for most of the twentieth century, as it remained mainly defined by neuropathological studies or stereotaxic atlases devised to guide selective “thalamotomies” intended as surgical treatments for depression, psychosis, pain, and movement disorders (Reference Déjerine and Dejerine-KlumpkeDèjerine 1901; Reference Vogt and VogtVogt & Vogt 1941; Reference Schaltenbrand and WahrenSchaltenbrand & Wahren 1977). After the decline of such methods in the 1970s and the introduction of modern brain-imaging techniques, a nomenclature consistent with that used in nonhuman primates was finally extended to the human thalamus (Reference Hirai and JonesHirai & Jones 1989; Reference Morel, Magnin and JeanmonodMorel et al. 2007, Reference Mai and MajtanikMai & Majtanik 2019). Based on multiarchitectonic data and on an extrapolation from connectivity data in nonhuman primates, this new nomenclature is now accepted as the international standard in the Terminologia Neuroanatomica (FIPAT 2017).
In the twentieth century, the application of increasingly sensitive connection labeling methods confirmed the heuristic value of the classic “cytoarchitectonic” nuclei delineations as a key principle of the functional organization of the thalamus. For example, studies of input pathways from the brainstem or retina revealed that the terminal distribution of these pathways largely matches nuclear boundaries. In addition, these studies redefined some boundaries and identified new subdivisions, such as the lateral geniculate domains receiving either retinal or tectal inputs (Reference Kaas, Guillery and AllmanKaas et al. 1972, Reference Kaas, Huerta, Weber and Harting1978) and the pulvinar subnuclei receiving tectal inputs (Reference Kaas and LyonKaas & Lyon 2007). Additional histochemical and immunolabeling methods developed in this period provided new references for nuclei delineation (Reference Graybiel and BersonGraybiel & Berson 1980; Reference Rausell and JonesRausell & Jones 1991; Reference Kuramoto, Furuta, Nakamura, Unzai, Hioki and KanekoKuramoto et al. 2009). From these studies, a consensus on nuclei layout and terminology based on cyto- and chemoarchitectonics as well as on connectivity criteria is now widely shared for the main animal model species (mouse: Reference Paxinos and FranklinPaxinos & Franklin 2019; Allen Mouse Common Coordinate Framework and Reference Atlas v.3 2017; rat: Reference Paxinos and WatsonPaxinos & Watson 2007; marmoset: Reference Paxinos, Watson, Petrides, Rosa and TokunoPaxinos et al. 2012; ferret: Reference Radtke-SchullerRadtke-Schuller 2018). Even in these best-studied species, however, some nuclei boundaries and/or subdivision denominations remain unsettled.
Studies of the thalamic output pathways showed that the cells in each thalamic nucleus selectively send their axons to one area or group of cortical areas and often also to other forebrain structures such as the striatum. Studies with single-cell-resolution methods have in recent years revealed that a relatively small number of high-level cell classes are found repeatedly throughout the thalamus. These cell classes are each characterized by a distinct axon wiring and transcriptomic profile (Reference Phillips, Schulmann, Hara, Winnubst, Liu, Valakh, Wang, Shields, Korff, Chandrashekar, Lemire, Mensh, Dudman, Nelson and HantmanPhillips et al. 2019; see Chapter 3 in this volume). The same cell class is often found in nuclei that are separate and receive different inputs. Interestingly, recent developmental cell lineage studies (Reference Shi, Xianyu, Han, Tang, Li, Zhong, Mao, Huang and ShiShi et al. 2017; Reference NakagawaNakagawa 2019) suggest that these common cell classes may reflect a deep level of thalamic cell diversity that is related to the clonal origins and migration of young neurons in the developing thalamus, well before the arrival of input connections or the emergence of cytoarchitectonically distinct nuclei.
As for other brain systems, the delineation of connections and electrophysiological properties is key for mechanistically defining the functioning of the thalamus and its contributions to behavior. A comprehensive framework for thalamic functions has been slow to emerge and is still being actively discussed. Historically, advances followed from three main research approaches: (a) observation of the effects of spontaneous or experimental lesions; (b) connection tracing; and (c) electrical recording, often in combination with the other two methods.
In the early nineteenth century, postmortem pathological reports became an increasingly common part of clinical practice. Some such reports linked, for the first time, lesions in the thalamus with premortem focal sensory and motor deficits (Reference BrightBright 1837; Reference JacksonJackson 1864). However, because of the nonselective nature of the lesions, usually vascular in origin, and the proximity to the thalamus of the internal capsule, the symptoms reported were quite variable. Likewise, experimental lesion studies conducted in frogs, birds, and mammals produced inconsistent behavioral effects (Reference FlourensFlourens 1824; Reference NothnagelNothnagel 1873). As a result, the idea of the thalamus as a sensory-related structure gained acceptance only slowly (Reference LuysLuys 1865; Reference JacksonJackson 1866, Reference FerrierFerrier 1886) against the alternative view that the thalamus was one more of the basal ganglia and primarily motor in function (Reference MagendieMagendie 1841; Reference VulpianVulpian 1866; Reference MeynertMeynert 1885).
As noted earlier, Reference LuysLuys (1865) was an early proponent of the thalamus as a gateway for sensory information to the cortex. As was common at the time, Luys assumed that brain cells formed continuous syncytia. Emboldened by the then-recent finding by Paul Reference BrocaBroca (1861) of a brain “center” for speech, Luys envisioned the thalamus as consisting of separate “centers” whose cells would “condense,” “store,” and “elevate” the sensory impressions they received from the “purely reflex” brainstem levels. Via the thalamic radiations, the centers would then convey this “energy” to particular regions of the cortex to excite or awaken (“erect”) them (Figure 2). Multisensory convergence could then occur in the upper layers of the cortex via tangential connections (Reference Parent and ParentParent & Parent 2011). Although prescient, Luys’s views were largely bold intuitions based on the case-report medical literature, as well as his knowledge of macroscopic brain anatomy and cellular theory, without much experimental support.
A few years later, Reference NothnagelNothnagel (1873) reported a persistent loss of visual responses in rabbits following thalamus lesions, along with somatic sensory-motor deficits. After the demonstration by Reference Fritsch and HitzigFritsch and Hitzig (1870) of the electrical excitability of the motor cortex, several researchers performed electric stimulation experiments in the thalamus, without conclusive motor effects. As part of one of these studies, David Reference FerrierFerrier (1886) performed a selective surgical ablation of the posterior thalamus in a monkey and observed a clear lack of response to cutaneous stimuli in the contralateral hemibody. This led him to conclude that the thalamus should be involved in somatic sensation, although his overall interpretation remained that it would receive the sensory information in a top-down manner from the sensory areas of the cortex. In Ferrier’s view, the thalamus was primarily involved in the guidance of voluntary movement through interactions with the striatum. A parallel line of early evidence for the implication of the thalamus in sensation came from clinical descriptions of numbness followed by persistent central pain (hyperalgesia and allodynia) associated with isolated infarctions of the posterolateral thalamus (Reference Déjerine and RoussyDéjerine & Roussy 1906). This thalamic pain syndrome (Déjerine-Roussy syndrome) is today recognized as an infrequent but well-defined neurological condition. Pain is believed to arise from multilevel plastic changes in the somatic sensory pathways, set in motion by the thalamic lesion (Reference Hong, Bai, Jeong, Choi, Chang, Kim, Ahn and JangHong et al. 2010).
Correct delineation of neural circuits is a prerequisite for mechanistically understanding their function. Not surprisingly, therefore, insights on thalamic functions have, to a large extent, followed from the application of new methods able to resolve the highly diverse neural connections of the thalamus with increasing precision.
As soon as Franz Nissl discovered cresyl violet staining, his mentor, Bernhard von Gudden, realized its potential for neural connection tracing. Hence, following a selective lesion in a part of the brain, Reference GuddenGudden (1870) was able to detect changes in cellular morphology and subsequent neuronal loss and gliosis in distant brain regions. Such changes were most evident when the lesions were performed in very young animals later examined as adults. Cell loss could be interpreted as evidence that the region had been selectively connected at a distance with the lesioned area (Figure 3A). Using this experimental approach by lesioning different cortical areas of cats, dogs, and rabbits and identifying which thalamus nuclei showed retrograde neuronal degeneration, Constantin Von Reference MonakowMonakow (1895) soon was able to delineate a first general topography of the thalamocortical connections.
Later, Wilfried Reference Le Gros ClarkLe Gros Clark (1932; Reference Le Gros Clark and BoggonLe Gros Clark & Boggon 1935; Reference Le Gros Clark and NorthfieldLe Gros Clark & Northfield 1937; Reference Le Gros Clark and PowellLe Gros Clark & Powell 1953) and A. Earl Reference WalkerWalker (1938) refined the method further and charted the projections of the thalamic nuclei to the cerebral cortex of monkeys. Other researchers extended the method to additional animal species, and by the 1960s, a general plan of the thalamocortical projection became established in the literature. However, the low sensitivity and indirect nature of the retrograde degeneration data fostered the image of a point-to-point, low-complexity wiring of the thalamocortical system (Reference WalkerWalker 1938, Figure 3B). Moreover, the degeneration results suggested that the thalamic axons reached some cortical areas of the cerebral hemisphere, but not others (Reference WalkerWalker 1938; Reference Akert, Warren and AkertAkert 1964; Reference LockeLocke 1967; Figure 3C). Likewise, the notion that some thalamic nuclei mainly innervate the striatum remained controversial for a long time (Reference Rose and WoolseyRose & Woolsey 1943). Besides, the observation that some thalamic nuclei degenerated massively following a lesion to a single cortical area, whereas others degenerated to a lesser extent or only when several areas were simultaneously lesioned, led to an influential distinction between thalamocortical projections that were “essential” to a given area (those that degenerated following a lesion limited to the area) versus those that were just “sustaining” to the area (which required the lesioning of several areas to produce detectable degeneration; Reference Rose and WoolseyRose & Woolsey 1949a, Reference Rose, Woolsey, Harlow and Woolsey1958). This distinction hinted at the possibility that some thalamocortical pathways might be more widely spread than others across the cortical mantle.
Improving on an impregnation method originally devised by Camillo Reference GolgiGolgi (1873), Santiago Ramón-y-Cajal visualized, for the first time, the basic neuronal circuitry of the thalamus (Reference Cajal1900, Reference Cajal1903; Figure 4A). The Golgi impregnation process stains at random, but entirely, a limited number of cells with a black silver precipitate, hence revealing dendrites, somata, and unmyelinated axons in exquisite detail. However, the myelinated part of long-range axons is mostly not impregnated; thus, Cajal’s discoveries centered on the thalamic gray matter (Figure 4A and B). These included the selective and orderly arborization within the thalamus of the ascending sensory and corticothalamic input fiber systems. He also noted the presence of projection neurons and local interneurons (Figure 4A) and even commented on the scarcity of the latter in rodents compared with carnivores. In addition, Cajal emphasized that the prethalamic reticular nucleus cell axons project back into the dorsal thalamus, but not toward the cortex (Figure 4C). Cajal also noted that axon-to-dendrite contacts (later called synapses) were never observed between the projection neurons.
Working with the Golgi method on embryonic and neonatal brains, whose white-matter tracts are still incompletely myelinated, Cajal and his collaborators were even able to visualize, in part, the thalamocortical and corticothalamic axons (Figure 4A). In a late review of his earlier Golgi work as a young student at Cajal’s laboratory, Rafael Reference Lorente de No and FultonLorente de No (1938) presumed that two types of axons innervating the cortex might come from the thalamus. The first type, which he called “specific” (“a” and “b” in Figure 4D), directly entered the cortex and arborized profusely in a single spot of the cortex middle layers. The second type, which he called “nonspecific” (“c” and “d” in Figure 4C), gave off a collateral branch that entered the cortex while the main trunk extended farther afield, toward other cortical regions. This collateral branch formed a simpler arborization that reached up to the pial surface. Because Lorente’s observations were limited to Golgi-stained material, the origins of these axons could not be established with certainty.
Reference Marchi and AlgeriVittorio Marchi and Giovanni Algieri (1885) introduced an empirical method using osmium salts in the presence of an oxidizing agent to selectively stain myelin sheaths in their process of post-lesion breakup but not in the normal tissue. This method could thus be applied on serial brain tissue sections to track myelinated fiber bundles degenerating after experimental brain lesions, from the lesion site to the vicinity of their target territory. However, because the terminal axon arborizations lack myelin ensheathing, axons could not be tracked to their end with this method. Nevertheless, “Marchi’s method” was instrumental to resolve disputes regarding the general sites of termination in the thalamus of the main subcortical input systems (Reference VogtVogt 1909; Reference EconomoEconomo 1911; Reference WalkerWalker 1938). Specifically, the method showed that the subcortical afferent systems always terminated in some nucleus of the thalamus without extending to the cerebral cortex (Reference MottMott 1892; Reference MonakowMonakow 1895). Along with Cajal’s Golgi observations, these studies definitively established the notion of the thalamus as a relay structure for sensory and motor subcortical inputs en route to the cortex.
In the 1950s, new methods based on metal impregnation were developed that allowed direct visualization of the degenerating axons themselves, including their terminal branches (Reference Nauta and GygaxNauta & Gygax 1954; Reference Fink and HeimerFink & Heimer 1967). This allowed a more accurate mapping of the distribution of afferent and efferent thalamic connections (Reference GuilleryGuillery 1967; Reference Montero and GuilleryMontero & Guillery 1968). In the same years, the introduction of electron microscopy for the study of brain tissue ultrastructure made possible the first descriptions of thalamic and cortical synapses (Reference SzentagothaiSzentagotai 1963; Reference Colonnier and GuilleryColonnier & Guillery 1964). Electron microscopy was soon shown to be useful for detecting synapses undergoing degenerative changes produced by a lesion in a distant brain region (Reference Gray and HamlynGray & Hamlyn 1962). This approach provided the first descriptions of synapses established by specific cortical and subcortical axon systems onto thalamic cells (Reference Colonnier and GuilleryColonnier & Guillery 1964; Reference Jones and PowellJones & Powell 1970), as well as the identification of the cells and dendritic domains postsynaptic to thalamic axons in the cortex (Reference Jones and PowellJones & Powell 1970; Reference Strick and SterlingStrick & Sterling 1974; Reference WhiteWhite 1978) or striatum (Reference Kemp and PowellKemp & Powell 1971b; Reference Chung, Hassler and WagnerChung et al. 1977).
In the early 1970s, entirely new kinds of connection-tracing methods that exploited the neuronal mechanisms for transmembrane uptake and axonal transport were developed. Each of these methods allowed the visualization of a particular exogenous molecule that, when deposited into the living nervous tissue, could be internalized by the local neuronal bodies and dendrites and actively transported to their distant axonal arborizations (anterograde transport) or be taken by the axons in the deposit region and transported to their parent cell bodies (retrograde transport), or both. Some of these methods remain in wide use. Detection of the transported substance was usually performed on thin tissue sections (Figure 5A–F). Depending on the substance injected, autoradiography (Reference Cowan, Gottlieb, Hendrickson, Price and WoolseyCowan et al. 1972), histochemistry (Reference LaVail and LaVailLaVail & LaVail 1972; Reference Gonatas, Harper, Mizutani and GonatasGonatas et al. 1979), epifluorescence (Reference Bentivoglio, Kuypers, Catsman-Berrevoets, Loewe and DannBentivoglio et al. 1980; Reference Kuypers, Bentivoglio, Catsman-Berrevoets and BharosKuypers et al. 1980; Reference Katz, Burkhalter and DreyerKatz et al. 1984; Reference Ju, Han and FanJu et al. 1989; Reference Nance and BurnsNance & Burns 1990), or immunolabeling with specific antibodies (Reference Gerfen and SawchenkoGerfen & Sawchenko 1984; Reference Veenman, Reiner and HonigVeenman et al. 1992; Reference Angelucci, Clascá and SurAngelucci et al. 1996) was required for detection (reviewed in Reference Lanciego and WouterloodLanciego & Wouterlood 2020). Besides, key insights on the embryonic development of thalamus connections were gained through the application of highly lipophilic fluorescent substances that, when deposited into fixed embryonic brain specimens, spread selectively and over relatively long distances along axons by simple passive diffusion (Figure 5G; Reference Godement, Vanselow, Thanos and BonhoefferGodement et al. 1987; Reference Molnár and BlakemoreMolnár & Blakemore 1995; Reference Clascá, Angelucci and SurClascá et al. 1995).
Because of their sensitivity and versatility, the axonal transport methods helped to resolve long-standing questions about the circuitry of the thalamus. First, they established that thalamic projection cells target the ipsilateral cerebral hemisphere and that, in contrast, the prethalamic reticular nucleus cells project instead to the nuclei of the dorsal thalamus (Reference JonesJones 1975). Likewise, these methods showed that all areas of the neocortex, as well as some hippocampal and olfactory areas, receive thalamic connections (Figure 5I) and that some thalamic nuclei innervate subcortical structures such as the striatum, accumbens, and amygdala (Reference Jones and LeavittJones & Leavitt 1974).
Second, axonal transport tracing studies clarified the extent and complexity of the corticothalamic projection. Corticothalamic connections were found to arise from two different cortical layers. One projection system arises from cortical layer 6 pyramidal neurons. All thalamic nuclei were found to receive massive numbers of these layer 6 connections, which may arise from the same cortical areas innervated by the cells of a given nucleus (“reciprocity principle”; Reference Diamond, Jones and PowellDiamond et al. 1969) but also from other functionally related cortical areas (“parity principle”; Reference Deschênes, Veinante and ZhangDeschênes et al. 1998), including even some in the contralateral cerebral hemisphere. Moreover, both the thalamocortical axons exiting a given thalamic nucleus and the layer 6 corticothalamic axons innervating that nucleus pierce the same region of the prethalamic reticular nucleus and leave branches in it; as a result, that reticular nucleus sector is functionally associated with that particular set of cortical and thalamic regions (Figures 5I and 10A; Reference JonesJones 1975; Reference Sherman and GuillerySherman & Guillery 1996).
A second corticothalamic projection system was found to originate in layer 5b pyramidal cells. These cells target the association nuclei but not the sensory relay nuclei of the thalamus (Reference Gilbert and KellyGilbert & Kelly 1975; Reference Catsman-Berrevoets and KuypersCatsman-Berrevoets & Kuypers 1978). Subsequent studies revealed that the differences between the two corticothalamic systems are profound. For example, layer 5 cells innervate the thalamus through branches of axons that also target the brainstem, whereas layer 6 cell axons target only the thalamus (Reference OjimaOjima 1994). In addition, each system establishes synapses onto different regions of the thalamic projection neuron dendrites (Reference GuilleryGuillery 1995). Moreover, the axon growth cones of each projection system invade the thalamus at markedly different ages during development (Reference Clascá, Angelucci and SurClascá et al. 1995).
Third, the axonal transport methods detected a marked variability in the convergence–divergence of the pathways originating in different nuclei. Cells from some nuclei could be retrogradely labeled from many cortical regions, indicating that their axons reach wide swaths of the cortical mantle, whereas cells in other nuclei were only labeled from a particular area or subarea, thus indicating that their axons make a more focal and non-overlapping targeting (Reference Herkenham, Jones and PetersHerkenham 1986; Reference Macchi, Bentivoglio, Minciacchi and MolinariMacchi et al. 1986; Figure 5H). Moreover, every cortical area was found to receive, as a rule, convergent inputs from more than one nucleus, in characteristic proportions (Reference Clascá, Llamas and Reinoso-SuárezClascá et al. 1997). Double-labeling strategies based on detecting the accumulation within the same cell body of two distinguishable retrograde tracers injected in distant points of the nervous tissue (Reference Rosina, Provini, Bentivoglio and KuypersRosina et al. 1980) were used to investigate the possibility that some cells branched their axons to target separate cortical areas. In most such experiments, double-labeled cells were observed in small numbers, and only in some thalamic nuclei, suggesting that the divergently branched thalamocortical axons were unusual (Reference Asanuma, Andersen and CowanAsanuma et al. 1985; Reference Minciacchi, Bentivoglio, Molinari, Kultas-Ilinsky, Ilinsky and MacchiMinciacchi et al. 1986; Reference Spreafico, Barbaresi, Weinberg and RustioniSpreafico et al. 1987; Reference Kishan, Lee and WinerKishan et al. 2008; Reference Cappe, Morel, Barone and RouillerCappe et al. 2009; Figure 5H).
Fourth, bulk-injected anterograde tracers in the thalamus revealed that the axons from different nuclei terminate in the cortex in specific laminar patterns (Reference HerkenhamHerkenham 1980, Reference Herkenham, Jones and Peters1986; Reference Killackey and EbnerKillackey & Ebner 1972, Reference Killackey and Ebner1973; Reference Burton and JonesBurton & Jones 1976), indicating that they may each target specific cell populations and/or dendritic domains within the cortical circuits. Based on a large collection of tracing experiments in the rat, Reference Herkenham, Jones and PetersHerkenham (1986) proposed a tripartite grouping of thalamic nuclei attending to these laminar patterns (Figure 6A and B).
Other studies examined this same issue by comparing the cells retrogradely labeled in the thalamus by tracer deposits limited to cortical layer 1 versus those involving the deeper layers. These confirmed that only some nuclei contain neurons that innervate layer 1 and often found differences in the soma size (Reference Carey, Fitzpatrick and DiamondCarey et al., 1979a, Reference Carey, Fitzpatrick and Diamond1979b; Reference Rausell and AvendañoRausell & Avendaño 1985) and/or calcium-binding protein content (Reference Hashikawa, Rausell, Molinari and JonesHashikawa et al. 1991; Reference Rausell, Bae, Viñuela, Huntley and JonesRausell et al. 1992; Reference Hendry and YoshiokaHendry & Yoshioka 1994) of the neurons innervating layer 1 versus those targeting only deeper cortical layers.
Drawing from the reports mentioned previously and from his own observations in macaque brains, Edward Jones (Reference Rausell and JonesRausell & Jones 1991; Reference JonesJones 1998, Reference Jones2007) proposed, as a fundamental principle of thalamus organization, the existence of two major systems: (a) a broad class of neurons spread as a “matrix” across the thalamus and characterized by their widely spread axons directed to the superficial layers of the cortex, and (b) another class of “core” cells present only in some thalamic nuclei characterized by focal axons centered in the cortex middle layers (Figure 7C). Jones emphasized that, at least in macaques, such two-cell populations each express specific calcium-binding proteins. Although later studies have shown the matrix-core hypothesis to be overly simplified, it provided an important conceptual framework for modeling the interactions between the thalamus and cortex, beyond those of simple relay (Reference Kuramoto, Furuta, Nakamura, Unzai, Hioki and KanekoKuramoto et al. 2009; Reference Bonjean, Baker, Bazhenov, Cash, Halgren and SejnowskiBonjean et al. 2012; Reference Cruikshank, Ahmed, Stevens, Patrick, Gonzalez, Elmaleh and ConnorsCruikshank et al. 2012; Reference Krishnan, Rosen, Chen, Muller, Sejnowski, Cash, Halgren and BazhenovKrishnan et al. 2018; Reference Müller, Munn, Hearne, Smith, Fulcher, Arnatkevičiūtė, Lurie, Cocchi and ShineMüller et al. 2020; see Section 2.3.2).
At the turn of this century, the technical possibilities for selective pathway tracing were multiplied by the introduction of a variety of recombinant viral vectors able to drive the expression of high levels of fluorescent proteins in neurons and their long-distance axons, either constitutively (Reference Chamberlin, Du, de Lacalle and SaperChamberlin et al. 1998; Reference Furuta, Tomioka, Taki, Nakamura, Tamamaki and KanekoFuruta et al. 2001), or in a Cre-recombinase–dependent fashion (Reference Atasoy, Aponte, Su and SternsonAtasoy et al. 2008). Other vectors were engineered to label pathways as they spread selectively across synapses (see reviews in Reference Xu, Holmes, Luo, Beier, Horwitz, Zhao, Zeng, Hui, Semler and Sandri-GoldinXu et al. 2020; Reference Saleeba, Dempsey, Le, Goodchild and McMullanSaleeba et al. 2019). And yet other vectors were created to drive the expression of both a fluorescent tag and a light-sensitive optogenetic probe in the transfected axons, hence making possible the simultaneous interrogation of thalamic axon wiring and synaptic effects with unprecedented resolution and flexibility (see, e.g., Reference Cruikshank, Ahmed, Stevens, Patrick, Gonzalez, Elmaleh and ConnorsCruikshank et al. 2012; Reference Zhou, Masterson, Damron, Guido and BickfordZhou et al. 2018; Reference Collins, Anastasiades, Marlin and CarterCollins et al. 2018).
2.2.3 Neurochemical Characterization of Thalamic Circuits
In parallel with the advent of axonal transport connection-tracing methods, a variety new histochemical, immunolabeling, and in situ messenger RNA (mRNA) hybridization methods became available in the 1980s and 1990s. These revealed the existence of diverse molecular mechanisms for signal transmission and allowed their mapping at the cellular and subcellular levels. As a result, we now know that the projection neurons in all dorsal thalamic nuclei use mostly glutamate as a neurotransmitter. Corticothalamic and most subcortical thalamus input pathways are glutamatergic as well. Interestingly, the glutamate transporters in the synaptic vesicles of the thalamocortical and subcortical afferent axon terminals are different from those employed in the corticothalamic terminals (Reference Fujiyama, Furuta and KanekoFujiyama et al. 2001). In contrast, the intrinsic interneurons and prethalamic reticular nucleus cells utilize the inhibitory gamma-amino butyric acid (GABA). Some subcortical pathways also signal to thalamic neurons through inhibitory synapses, using GABA or glycine (reviewed in Reference Halassa and AcsádyHalassa & Acsàdy 2016). Other studies detected the selective expression of proteins with specific calcium-binding dynamic profiles (calbindin28K, parvalbumin, calretinin) in different thalamic cell subpopulations (Reference Rausell and JonesRausell & Jones 1991; Reference Jones and HendryJones & Hendry 1989).
Immunolabeling studies also revealed a variety of “diffuse” brainstem input pathways to the thalamus, each identified by their use of a specific neurotransmitter, as well as by their origin in a different brainstem cell population. Specific antibodies against the neurotransmitter itself, its pathway enzymes, or its membrane transporters became key tools to map these systems in different species. The diffuse input pathways included the cholinergic axons from the peribrachial region and laterodorsal nucleus of the pons (Reference Saper and LoewySaper & Loewy 1980; Reference AhlsénAhlsén 1984; Reference Hallanger, Levey, Lee, Rye and WainerHallanger et al. 1987), the noradrenergic axons from the locus coeruleus (Reference Pickel, Segal and BloomPickel et al. 1974; Reference Leger, Sakai, Salvert, Touret and JouvetLeger et al. 1975; Reference Swanson and HartmanSwanson & Hartman 1975), and the serotonergic axons from the dorsal and median nucleus of the raphe (Reference Conrad, Leonard and PfaffConrad 1974; Reference Morrison and FooteMorrison & Foote 1986; Reference Rico and CavadaRico & Cavada 1998). In addition, a substantial dopaminergic system and several peptidergic systems were described in the primate thalamus (Reference Lechner, Leah and ZimmermannLechner et al. 1993; Reference García-Cabezas, Martínez-Sánchez, Sánchez-González, Garzón and CavadaGarcía-Cabezas et al. 2009). All components of the thalamus, including the dorsal thalamus and the prethalamic reticular nucleus, were found to be innervated by these pathways in a highly uneven fashion, as the axons branch profusely and cross nuclear boundaries. Studies combining immunolabeling and electron microscopy revealed that the distribution of these brainstem axons onto the thalamic cells and their presynaptic structure is specific (Reference Raczkowski and FitzpatrickRaczkowski & Fitzpatrick 1989; Reference Nothias, Onteniente, Roudier and PeschanksiNothias et al. 1988). The expression of receptors for each of the various neurotransmitters is highly heterogeneous across thalamic territories as well (Reference Pérez-Santos, Palomero-Gallagher, Zilles and CavadaPérez-Santos et al. 2021). Overall, these discoveries have revealed a rich combinatorial matrix of modulatory inputs superimposed onto that of the main, signal-carrying afferent pathways.
2.2.4 Single-Cell Labeling Studies
Although dendritic and local axon morphologies had been amenable to study since the introduction of the Golgi method, techniques able to reveal the entire axonal tree of individual long-range projection neurons did not become available until the late 1990s (Reference Deschênes, Bourassa and PinaultDeschênes et al. 1994; Reference PinaultPinault 1996). The first technique published required electroporation, under extracellular recording conditions, of a histochemically detectable marker molecule such as biocytin. This “juxtacellular labeling” strategy allowed the first visualizations of the complete morphology of individual thalamocortical, corticothalamic, and reticulo-thalamic neurons (Reference Deschênes, Veinante and ZhangDeschênes et al. 1998, Figure 8A–C). In the 2010s, new viral vectors (Reference Kuramoto, Furuta, Nakamura, Unzai, Hioki and KanekoKuramoto et al. 2009) or combinations of vectors (Reference Economo, Clack, Lavis, Gerfen, Svoboda, Myers and ChandrashekarEconomo et al. 2016) able to drive the expression of high levels of different fluorescent proteins in isolated cells expanded the array of tools available for visualizing complete thalamic projection neuron morphologies. To date, the method of single-cell tracing and reconstruction has been applied to a number of nuclei and cortical areas of the thalamus, mainly in rats and mice.
The most relevant insights gleaned to date from the single-cell studies regard the prevalence and specificity of branched axons. For example, many thalamic axons were found to innervate the cortex and striatum and/or amygdala simultaneously, dispelling the notion of strictly separate thalamocortical, thalamostriatal, and thalamoamygdaloid circuits. Within the cortex, branched axons were found to target separate areas, to arborize in different layers in each area, and to establish structurally diverse synapses that engage specific cell populations and receptor mechanisms (Reference Kuramoto, Furuta, Nakamura, Unzai, Hioki and KanekoKuramoto et al. 2009; Reference Rodriguez-Moreno, Porrero, Rollenhagen, Rubio-Teves, Casas-Torremocha, Alonso-Nanclares, Yakoubi, Santuy, Merchan-Pérez, DeFelipe, Lübke and ClascáRodriguez-Moreno et al. 2020). These findings upended the dual matrix-core model. Overall, the single-cell labeling data made it clear that the thalamic projection neurons are a diverse cell population that encompasses several major cell classes, each characterized by their axonal architecture as well as other differences in somatodendritic morphology and calcium-binding protein expression. The same basic cell types are found across different functional systems (see Chapter 3, in this book).
Likewise, branching in corticothalamic pathways was found to be highly specific. For example, the layer 5 projection to the thalamus is established exclusively via collateral branches from axons directed to brainstem motor centers; remarkably, these axons leave few or no branches in the prethalamic reticular nucleus. In contrast, corticothalamic layer 6 projections are established by axons that arborize only in the thalamus and reticular nucleus. The arborization patterns and presynaptic specializations of layer 5 and layer 6 inputs in the thalamus are strikingly different (Reference Bourassa, Pinault and DeschênesBourassa et al. 1995; Reference Deschênes, Veinante and ZhangDeschênes et al. 1998, Reference Kakei, Na and ShinodaKakei et al. 2001, Reference Guillery and ShermanGuillery & Sherman 2002a; see Section 2.3.3).
The technological effort around the Second World War led to the development of new systems capable of electrical recording, feedback, and amplification at the scales required for investigating the activity of neuronal populations or individual cells in the thalamus. Starting in the 1940s, these studies developed around two main themes: (a) the properties of thalamic neurons related to their role as relay stations for ascending information to the cerebral cortex and (b) the involvement of the thalamus in the synchronous activities of large numbers of forebrain neurons associated with different states of consciousness and sleep.
Relay function studies recorded changes evoked by sensory stimulation in the membrane potential of thalamic and cortical cells. The emphasis of this approach was mainly on the mapping of topographical representations of the skin surface, visual field, and auditory periphery in the principal relay nuclei and on the receptive fields and response properties of neurons in these nuclei. Prominent researchers in this field were Vernon Mountcastle and colleagues in the somatosensory system (Reference Rose and MountcastleRose & Mountcastle 1952; Reference Poggio and MountcastlePoggio & Mountcastle 1963), Reference Hubel and WieselDavid Hubel and Torsten Wiesel (1961) in the visual system, and Reference Rose, Woolsey, Harlow and WoolseyJerzy Rose and Clinton Woolsey (1958) in the auditory system.
Thalamic sensory nuclei were found to be ideal experimental models because of the ease of controlling input parameters during recording to gain insight into the specific functions performed in these circuits. An additional advantage is that neurons in sensory relay nuclei usually remain responsive under light anesthesia. Progress in electrophysiology created a positive-feedback loop for increasingly sophisticated structural and/or neurochemical investigations of the sensory nuclei and their connections. Eventually, the study of thalamic sensory pathways and their relay nuclei came to dominate thalamus research almost completely for several decades, even though these nuclei represent just one part of the thalamus and a relatively small one in humans (Reference GuilleryGuillery 1995).
The development of intracellular glass microelectrodes in the 1960s expanded the sensitivity and resolution of electrophysiological thalamic relay function studies. Such electrodes, applied in vivo or on brain slices, allowed the characterization of the functional properties of thalamic circuits at the single-cell level. In subsequent years, the microelectrodes were filled with marker substances that diffused, while recording, into the cells or terminal axons. This allowed detailed visualization of the recorded cells or axons under light and electron microscopy. This approach revealed, for example, the specific terminal morphologies of different retinofugal axon types in the dorsal lateral geniculate nucleus (Reference Bowling and MichaelBowling & Michael 1980, Reference Bowling and Michael1984), the synaptic relationships of interneurons in this nucleus (Reference Hamos, Van Horn, Raczkowski, Uhlrich and ShermanHamos et al. 1985, Reference Hamos, Van Horn, Raczkowski and Sherman1987), or the existence of three cells subclasses in the dorsal lateral geniculate nucleus, each carrying different signals from the retina and arborizing in specific layers of the primary visual cortex (Reference Friedlander, Lin, Stanford and ShermanFriedlander et al. 1981; Reference Stanford, Friedlander and ShermanStanford et al. 1981, Reference Stanford, Friedlander and Sherman1983; Reference Sur and ShermanSur & Sherman 1982).
Likewise, microelectrode studies established the existence of an extensive and functionally diverse set of corticothalamic connections. These were later shown to be excitatory but to elicit mainly inhibitory effects in the thalamic cells through a parallel activation of inhibitory neurons in the prethalamic reticular nucleus as well as of intrinsic interneurons in the relay nuclei (Reference SwadlowSwadlow 1994; Reference SteriadeSteriade 2001a, Reference Steriade2001b; Reference Sirota, Swadlow and BeloozerovaSirota et al. 2005; Reference BriggsBriggs 2020).
Intracellular recordings in thalamus slices by Rodolfo Llinás and coworkers (Reference Jahnsen and LlinásJahnsen & Llinás 1984a, Reference Jahnsen and Llinás1984b) made the crucial discovery that all the thalamic relay cells have in their membranes an unusual type of calcium channel (transient T-type Ca2+ channels). These channels mediate, in a voltage-dependent manner, an inward current (IT) that produces a transient depolarization of the cell (Figure 9A). Although T channels are common to neurons everywhere, what sets the thalamic projection neurons cell apart is that they exhibit a sufficiently dense distribution of T channels in their somata and dendrites that initiation of IT generally leads to an all-or-none Ca2+ spike propagated throughout the soma and dendritic arbor. This is seen only rarely in other neuronal types in the central nervous system (Reference ShermanSherman 2001a).
It was subsequently shown that as a result of having these channels, a thalamic projection neuron cell can respond to incoming excitatory stimuli in one of two very different modes, depending on its recent voltage history (Reference ShermanSherman 2001b; Figure 9B). In the “tonic” mode, when the cell is near the resting membrane potential (approximately −60 mV), T channels go along with the rest of membrane conductances, allowing the projection neuron to fire action potentials at frequencies linearly proportional to the amplitude of the stimuli it receives. Thus, while in tonic mode, thalamic projection neurons faithfully relay ascending sensory or motor information.
In contrast, the same cells enter “burst” mode at more hyperpolarized (–70 mV) membrane potentials. In this situation, the T channels open again, allowing a calcium spike that is sufficiently large (typically 25–40 mV) to trigger a transient, high-frequency burst of two to ten action potentials. Bursting is an all-or-none activity, and its temporal profile follows mainly from the T-channel dynamics. As a result, during bursting, signals are nonlinearly relayed, and information may be lost. In addition, burst firing provides a much higher signal-to-noise ratio in the incoming flow of signals to the cortex than tonic firing, implying that an evoked burst is more likely to be detected by the cortex and may be used as a “wake-up call” for novel, potentially relevant stimuli (Reference ShermanSherman 2001b; Reference Swadlow and GusevSwadlow & Gusev 2001).
In parallel with the studies of the thalamus as a relay, other electrophysiological studies examined a possible role of the thalamus in the generation of coordinated activity across widespread brain cell populations. This activity, detectable through macroelectrodes applied on the scalp or in the brain, had been shown to consist of large-scale voltage waves comprising different types of low-frequency and high-frequency oscillations that changed according to the state of consciousness (Reference BergerBerger 1929).
Interest in the thalamus as a potential source for some of these widespread oscillations arose from reports in the early 1940s that linked the thalamus with a type of slow (7- to 14-Hz) high-voltage waxing and waning wave burst that repeated spontaneously over wide zones of the cortex and thalamus during the entry or recovery from slow-wave sleep or under barbiturate anesthesia. These bursts, later known as spindles because of their appearance in the electroencephalogram (EEG), were first recorded in mesencephalon-transected cats (Reference BremerBremer 1937) or in barbiturate-anesthetized animals after sensory stimulation (Reference AdrianAdrian 1941). Intriguingly, spindles were observed to disappear from the cortex after transections of the underlying white matter (Reference BremerBremer 1937; Reference Morison, Finley and LothropMorison et al. 1943; Reference BurnsBurns 1950), yet they persisted in the thalamus even after complete decortication and extensive thalamic deafferentation (Reference Morison, Finley and LothropMorison et al. 1943; Reference Morison and BassettMorison & Basett 1945). These observations were interpreted as evidence that the thalamus could, in some circumstances, act as a pacemaker for cortical activity. Interest in this hypothesis was in part spurred by the assumption, later dismissed, that spindle activity could bear relation to the generation of the EEG alpha rhythm that characterizes relaxed wakefulness.
Another type of rhythmic activity associated with the thalamus was first reported in studies in barbiturate-anesthetized cats that recorded two different types of electrical activity “responses” on the cortical surface following the application of repeated low-frequency trains of electrical pulses in the thalamus (Reference Dempsey and MorisonDempsey & Morison 1942). The first or “primary” response, appearing at a short latency (around 3 ms) in primary sensory or motor areas after stimulation of their corresponding relay nuclei, consisted of a biphasic positive–negative that “augmented” in amplitude and cortical spread with pulse repetition. It is now known that this wave pattern reflects the distribution of some thalamocortical terminals in the middle layers of the cortex and their short-term synaptic plasticity mechanisms. The second (“recruiting”) type of response appeared at longer latencies (around 20–35 ms) over much wider territories as high-voltage surface-negative waves that waxed and waned recurrently. Remarkably, this recruiting response was reported to be best elicited by stimulation in the medial and intralaminar nuclei, often following a single stimulus, and appeared mainly in the same frontoparietal areas where the spontaneous spindle activity was usually observed (Reference Verzeano, Lindsey and MagounVerzeano et al. 1953).
The significance of spindles and recruiting responses remains unclear, and they may actually be epiphenomenal effects of thalamic cell membrane properties and synaptic wiring (see later discussion); however, they were profoundly influential in generating experiments that uncovered important aspects of the functional connectivity of the thalamus.
The pursuit of the mechanism of this recruiting response attracted much attention in the 1940s and 1950s. The surface-negative wave profile of the response was interpreted as evidence of its origin near the cortical surface (Reference JasperJasper 1949). Despite the long latencies, Reference Dempsey and MorisonDempsey and Morison (1942) favored the interpretation that it should involve a direct projection from a diffuse, thin-axon thalamocortical system preferentially targeting layer 1 (Figure 7B). Multidisciplinary approaches were unusual at the time, and these electrophysiology researchers apparently took at face value the opinion by Lorente de No that some of the “nonspecific” thalamocortical fibers he had observed in Golgi samples while at Cajal’s lab might originate in the intralaminar nuclei (Reference Lorente de No and FultonLorente de No 1938, Figure 4B).
Hence, the notion was born that the intralaminar nuclei would innervate layer 1 in a diffuse manner and be the substrate for the recruiting responses. Remarkably, functional studies of this hypothetical system eventually came to absorb a substantial effort of thalamus research well until the early 1960s. Other authors posited that the intralaminar nuclei of the thalamus would represent, together with the prethalamic reticular nucleus, a rostral extension of a multisynaptic “ascending reticular activating system” (a concept in vogue at the time; Reference Moruzzi and MagounMoruzzi & Magoun 1949) that would spread excitatory brainstem activity to the cortex for the regulation of consciousness and sleep.
Recruiting-response studies produced the first functional investigations of nonsensory thalamic nuclei in the medial and intralaminar regions of the thalamus, emphasizing their possible role in cortical arousal and EEG desynchronization (Reference JasperJasper 1949, Reference Jasper, Field, Magoun and Hall1960). Additionally, because these nuclei could not be selectively driven by peripheral sensory stimulation, this line of research required the development of new fine bipolar electrodes for deep-brain stimulation and lesioning. It also fostered the publication of the first stereotaxic brain atlases to guide the positioning of deep electrodes in experimental animals (Reference Jasper and Ajmone-MarsanJasper & Ajmone-Marsan 1954; Reference Reinoso-SuárezReinoso-Suárez 1961). As years passed, however, spindle activity was found to be evoked by microstimulation also in lateral nuclei such as the medial geniculate (Reference Galambos, Rose, Bromiley and HughesGalambos et al. 1952) or the ventrobasal complex (Reference Andersen and SearsAndersen & Sears 1964) and even in nonthalamic regions. Eventually, interest in the nonspecific / intralaminar system waned. In the following decades, axonal transport studies would reveal that the intralaminar nuclei do not target the superficial layers of the cortex and mostly innervate relatively restricted cortical territories (Reference Herkenham, Jones and PetersHerkenham 1986; Reference Groenewegen and BerendseGroenewegen & Berendse 1994; Reference Deschênes, Bourassa, Doan and ParentDeschênes et al. 1996).
In the 1990s and 2000s, the notion of a widely spread, upper-layer–directed thalamocortical system was revived within the “core-matrix” model of thalamocortical pathways of Edward Jones (Reference Llinás, Ribary, Contreras and PedroarenaLlinás et al. 1998; Reference JonesJones 2001, Reference Jones2007, Reference Jones2009 Figure 7C). However, Jones remarked that the projection neurons originating this system were not located in the intralaminar nuclei but spread instead across many other thalamic nuclei. Jones speculated that the “matrix” projection neurons would propagate synchronous oscillations across wide assemblies of cortical and thalamic cells. Taking into the model some then recently published single-cell tracing data on corticothalamic pathways (Reference Bourassa, Pinault and DeschênesBourassa et al. 1995; Reference Veinante, Lavallée and DeschênesVeinante et al. 2000; Figure 8B), he postulated that the double reciprocal pathways linking the thalamus and cortex, along with the widely projecting matrix thalamocortical cells, could be the substrate for the spread of coherent activity throughout the thalamocortical network (Reference JonesJones 2001, Reference Jones2009). However, rather than relating this effect to slow cortical oscillations as in the 1940s, Jones posited that the matrix thalamocortical system might facilitate the propagation of oscillations at faster frequencies (~40 Hz) across the network. These fast oscillations, known to originate in the cortex, had been associated a few years earlier with conscious sensory experiences in humans (Reference Desmedt and TombergDesmedt & Tomberg 1994; Reference Tononi and EdelmanTononi & Edelman 1998; Reference Llinás, Ribary, Contreras and PedroarenaLlinás et al. 1998). A link between the matrix cell pathway with reports of behavioral improvement in minimally conscious patients with brain injuries after thalamic stimulation was also proposed (Reference Schiff, Giacino, Kalmar, Victor, Baker, Gerber, Fritz, Eisenberg, Biondi, O’Connor, Kobylarz, Farris, Machado, McCagg, Plum, Fins and RezaiSchiff et al. 2007).
In addition, Jones noted that the complementary laminar termination upon cortical pyramidal cells of “matrix” and “core” axons laid the wiring for a coincidence detector (Figure 10B). Based on this model, it was also speculated that pathological alterations disrupting the low- and high-frequency oscillatory rhythms of the thalamocortical network could ultimately underlie some of the clinical manifestations of conditions as diverse as neurogenic pain, tinnitus, Parkinson’s disease, and depression (Reference Llinás, Ribary, Jeanmonod, Kronberg and MitraLlinás et al. 1999; Reference JonesJones 2009).
The notion that the thalamus was a pacemaker for some rhythmic EEG activities also triggered a search for the mechanisms that could generate such rhythmicity. First, Reference Dempsey and MorisonDempsey and Morison (1942) posited the existence in the medial and intralaminar nuclei of some cells with spontaneous rhythmic discharges that would be able to influence cells in other nuclei by means of hypothetical internuclear connections and/or corticothalamic feedback. Subsequent studies specifically aimed at testing these ideas combined microstimulation of the medial-intralaminar nuclei while intracellularly recording in the more lateral thalamic nuclei in awake, paralyzed animals (Reference Purpura and CohenPurpura & Cohen 1962; Reference Purpura and ShoferPurpura & Shofer 1963).
During these experiments, large inhibitory postsynaptic potentials (IPSPs) were observed in the thalamic projection neurons following the thalamic stimulus. These potentials were long (80–200 ms) and effectively suppressed the spontaneous discharge of the neurons. Shorter excitatory postsynaptic potentials (EPSPs) alternated with them in a stereotyped fashion. Moreover, the stereotyped sequence also occurred spontaneously during spindles. At the time, it was proposed that these sequences and the ensuing synchronization in the EEG were generated in reciprocally interconnected excitatory neurons in medial and lateral thalamic nuclei (Reference Purpura, Pappas and PurpuraPurpura 1972).
Evidence for connections between dorsal thalamic nuclei was not substantiated by subsequent anatomical investigations with axonal transport labeling methods, and the notion was abandoned. However, these studies brought the role of inhibition in the generation of rhythmic thalamocortical cell oscillations to the forefront of research. In this vein, Andersen and colleagues (Reference Andersen and SearsAndersen & Sears 1964; Reference Andersen and AnderssonAndersen & Andersson 1968) observed the same type of alternating excitation and inhibition sequences following peripheral nerve stimulation in the ventroposterior nucleus of barbiturate-anesthetized cats. Their interpretation was that the thalamic spindle sequence would start with the firing of a projection neuron produced spontaneously or by extrinsic inputs. The activity would then be propagated within the nucleus through the supposed collaterals of the projection cell axon inside the nucleus to inhibitory local circuit interneurons. These would fire repetitively and induce the long IPSPs in a population of surrounding projection cells. After the IPSPs, the projection cells would rebound, firing axon potentials that would again reach, via their supposed axon collaterals, a wider group of interneurons, which would in turn recruit a larger number of thalamic cells, and so on, in cyclic fashion. The sequence would wane after a number of cycles because differences in IPSP duration would decouple the oscillations in the cell population. According to this view, any nucleus could become a pacemaker and propagate its activity to the cerebral cortex.
In the following decades, however, the same observations were reinterpreted in the light of new functional, anatomical, and neurochemical discoveries. These showed that projection neuron axons do not leave collateral branches inside the thalamus. Instead, the long IPSPs actually depend on the collaterals of the projection axons exciting prethalamic reticular nucleus cells (Figure 10A). The reticular cells were shown to be GABAergic (Reference Houser, Vaughn, Barber and RobertsHouser et al. 1980) and to project back, in topographical fashion, to the thalamic nucleus that innervates them (Reference JonesJones 1975; Reference Steriade, Parent, Paré and SmithSteriade et al. 1987). Moreover, prethalamic reticular nucleus cells were shown to fire prolonged bursts of high-frequency action potentials during the spindles. Such bursts are enhanced by GABAergic connections between prethalamic reticular nucleus cells and the excitatory corticothalamic input to the nucleus.
In contrast, thalamic interneuron axons were found to remain confined to individual nuclei near their parent cell soma or to be missing altogether, as most interneuron synapses are dendro-dendritic (Reference JonesJones 2007; Reference CoxCox 2014). Because the thalamic projection cells do not connect directly to each other, local-scale connections within the thalamus are dominated by these resident inhibitory neurons (Reference ShermanSherman 2004; Reference Hirsch, Wang, Sommer and MartinezHirsch et al. 2015). Remarkably, interneuron numbers and distribution vary widely across mammalian phyla, being virtually absent from some or most nuclei of the thalamus in some species, such as small rodents, marsupials, and bats. This observation suggests that the local interneurons are an optional part of the thalamic networks, and they may be flexibly modified to develop specific evolutionary adaptations (Reference Arcelli, Frassoni, Regondi, Biasi and SpreaficoArcelli et al. 1997; Reference Letinic and RakicLetinic & Rakic 2001; Reference Rikhye, Wimmer and HalassaRikhye et al. 2018; Reference Jager, Moore, Calpin, Durmishi, Salgarella, Menage, Kita, Wang, Kim, Blackshaw, Schultz, Brickley, Shimogori and DeloguJager et al. 2021).
Studies in the 1980s and 1990s firmly established that the prethalamic reticular nucleus is the pacemaker for the thalamic-generated spindles and revealed the circuit and intrinsic cell properties of the reticular nucleus and thalamic cells from which it arises. The rebound discharges of thalamic cells were shown to depend on the expression, in thalamic cells of the T channels, the voltage-dependent conductances described in Section 2.3.2 (Reference Jahnsen and LlinásJahnsen & Llinás 1984a, Reference Jahnsen and Llinás1984b; Reference Roy, Clercq, Steriade and DeschênesRoy et al. 1984). During the spindling activity, low-threshold spikes and bursts occur in the projection cell as they recover from the strong inhibition imposed by the prethalamic reticular nucleus cells. Thalamic projection cells burst and then excite prethalamic reticular nucleus cells, and the cycle starts again (Reference Bal and McCormickBal & McCormick 1993). Importantly, the thalamic cells do not oscillate at spindle frequencies when deprived of inhibitory input from the prethalamic reticular nucleus (Reference Steriade, Parent, Paré and SmithSteriade et al. 1987). The population-level spindle waxes as more reticular nucleus and projection cells are recruited and wanes as the two populations become out of step. Beyond the spindle phenomenon, however, the most relevant effect of the prethalamic reticular nucleus-imposed inhibition on projection cells during normal information transmission probably occurs on those cells that are firing out of synchrony with the most activated projection cell population. In this way, activity in particular regions of the reticular nucleus can powerfully “focus” the flow of information through the thalamus (Reference CrickCrick 1984; Reference SteriadeSteriade 1999).
Overall, these investigations revealed that the dorsal thalamus and the prethalamic reticular nucleus form a tightly integrated functional network and that inhibition plays a major role in shaping thalamic cells’ responses. Changes in prethalamic reticular nucleus inhibitory activity under modulation from the thalamus, cortex, and brainstem cholinergic and monoaminergic systems thus may even be causal to changes in arousal states (Reference Dingledine and KellyDingledine & Kelly 1977; Reference BeierleinBeierlein 2014; Reference Lewis, Voigts, Flores, Schmitt, Wilson, Halassa and BrownLewis et al. 2015)
2.3.4 The Guillery and Sherman Synthesis
In the late 1990s, a comprehensive new synthesis of thalamus structure and function was proposed by Ray Guillery and Murray Sherman and subsequently elaborated over the course of a long collaboration (Reference GuilleryGuillery 1995; Reference Sherman and GuillerySherman & Guillery 1996, Reference Sherman and Guillery1998, Reference Sherman and Guillery2005, Reference Sherman and Guillery2012; Guillery & Sherman 2002). This multipronged model was based on both electron microscopic and microelectrode recording data from their own laboratories, mainly on sensory relay nuclei, as well as on insightful analysis of the literature.
On the anatomical side, a fundamental distinction was made between the “first-order relay” (FO) thalamic nuclei that relay to the cortex ascending signals from the subcortical sensory or motor systems versus the “higher-order relay” (HO) nuclei that relay back to the cortex signals received from the cortex itself, via collaterals of layer 5 axons primarily directed to the brainstem (Reference GuilleryGuillery 1995). It was also pointed out that because the layer 5 pathway originating in a given primary sensory cortical area is not directed to the FO nucleus that innervates it but to a nearby HO nucleus, a potential route for transthalamic communication between cortical areas is created (Reference GuilleryGuillery 1995). Direct evidence in support of this hypothesis was subsequently obtained on mouse brain slice preparations with preserved connections between the cortex and thalamus (Reference Theyel, Llano and ShermanTheyel et al. 2010).
A second tenet of the proposal was that the layer 5 input would represent the “primary source” of information for HO nuclei cells in a manner equivalent to what subcortical sensory inputs are for FO nuclei cells (Reference GuilleryGuillery 1995). This conclusion was based on the fact that both layer 5 axon branches and ascending sensory axons reaching the thalamus form clustered large terminal boutons on proximal projection cell dendrites (Reference Hoogland, Welker and Van der LoosHoogland et al. 1987) that show virtually identical synaptic ultrastructures (Reference Colonnier and GuilleryColonnier & Guillery 1964; Reference MathersMathers 1972a, Reference Mathers1972b). Perceptively, Reference Sherman and GuillerySherman & Guillery (1996) also remarked that, according to information in the literature, both the layer 5 corticothalamic and the peripheral sensory input systems to the thalamus are, as a rule, collateral branches of axons that simultaneously innervate lower (motor) centers. The functional implication is that the main content of information flowing through the thalamus must be about ongoing cortical or subcortical motor instructions (Figure 11B; Reference Guillery and ShermanGuillery & Sherman 2002b).
In striking contrast, the layer 6 cell corticothalamic axons were identified as a functionally different pathway that provides a highly convergent reciprocal feedback loop. Layer 6 axons terminate exclusively in the thalamus and the prethalamic reticular nucleus (Figures 8B and 11A), and they form disperse small terminal varicosities that establish synapses only on distal dendritic domains of the thalamic projection neurons (Reference RobsonRobson 1983; Reference GuilleryGuillery 1969).
The third element in the synthesis pertained to the dynamics of signal transmission of the various inputs reaching the thalamic projection neurons. Again, the activities evoked by the ascending subcortical inputs in FO nuclei neurons or layer 5 corticothalamic inputs in HO nuclei neurons were found to be identical. In contrast, the activities evoked by layer 6 corticothalamic inputs were markedly different. For example, in vitro studies showed that HO and layer 5 synapses produced fast-rising, large EPSPs that depressed with paired-pulse repetition, whereas the layer 6 corticothalamic synapses produced smaller, slower-rising currents that increased in amplitude with repetition and lasted longer (Figure 12C and D). Moreover, the application of selective channel blockers revealed that the ascending subcortical inputs to FO neurons and layer 5 corticothalamic inputs to HO neurons were both mediated only by ionotropic glutamate receptors, whereas the layer 6 corticothalamic inputs were mediated by ionotropic as well as metabotropic receptors.
An important generalization to other glutamatergic systems of the forebrain followed from these observations in the thalamus: that the glutamatergic synapses can be consistently separated into two large classes based on their functional impact on postsynaptic cells, subcellular structure, and electrophysiological parameters (Figure 12A–D). Reference Sherman and GuillerySherman and Guillery (1998, Reference Sherman and Guillery2011) coined the term driver for the glutamatergic synapses capable of reliably transmitting their information content (e.g., receptive field properties) to the postsynaptic cells, and they used the term modulator for all other synapses capable of altering the probability of certain aspects of that transmission. In the thalamus, the ascending glutamatergic inputs to FO nuclei and the layer 5 corticothalamic inputs to HO nuclei were both identified as “driver” synapses, whereas the layer 6 corticothalamic were identified as “modulator” synapses. As a result of the different functional impact of each, although the HO thalamic relay is massively dominated in terms of numbers by layer 6 corticothalamic synapses, it is the driver subcortical or cortical layer 5 synapses that determine the main content of the information relayed to the cortex. In subsequent studies, the driver/modulator analysis was extended to thalamocortical synapses (Reference Viaene, Petrof and ShermanViaene et al. 2011a, Reference Viaene, Petrof and Sherman2011b; Reference Rodriguez-Moreno, Porrero, Rollenhagen, Rubio-Teves, Casas-Torremocha, Alonso-Nanclares, Yakoubi, Santuy, Merchan-Pérez, DeFelipe, Lübke and ClascáRodriguez-Moreno et al. 2020) and even to some of the afferent subcortical inputs to the thalamus (Reference Lee and ShermanLee & Sherman 2010).
Over the past two decades, the conceptual framework proposed by Guillery and Sherman has been profoundly influential. Specifically, it has brought into focus the importance of advancing knowledge on the HO nuclei. As a result, inroads are being made in the understanding of the structural and functional diversity of HO thalamocortical circuits (Reference Rovó, Ulbert and AcsádyRovó et al. 2012; Reference Groh, Bokor, Mease, Plattner, Hangya, Stroh, Deschenes and AcsádyGroh et al. 2014; Reference Rodriguez-Moreno, Porrero, Rollenhagen, Rubio-Teves, Casas-Torremocha, Alonso-Nanclares, Yakoubi, Santuy, Merchan-Pérez, DeFelipe, Lübke and ClascáRodriguez-Moreno et al. 2020). In parallel, new in vivo experimental approaches are being specifically devised to investigate the roles that HO thalamic cell subnetworks may play in complex functions, such as the control of information transfer between cortical areas, spatial navigation, attention, and cognitive flexibility (Reference Purushothaman, Marion, Li and CasagrandePurusothaman et al. 2012; Reference Saalmann, Pinsk, Wang, Li and KastnerSaalman et al. 2012; Reference Saalmann and KastnerSaalman & Kastner 2015; Reference Schmitt, Wimmer, Nakajima, Happ, Mofakham and HalassaSchmitt et al. 2017). In the coming years, the combination of such approaches may provide important new clues about the functional diversity of thalamic circuits and their role within large-scale multiregional brain networks. Thalamus research history remains in the making.
This work was supported by funding from the European Union’s Horizon 2020 Research and Innovation Programme (Grant Agreement No. 945539 HBP SGA3). Additional support was provided by Spain’s Ministerio de Ciencia, Innovación y Universidades (PID2020-115780GB-I00) to the author.