Cerebellar Influences on Descending Spinal Motor Systems
- 3 Downloads
The cerebellar nuclei, and to some extent the vestibular nuclei, mediate the ultimate result of cerebellar processing to the rest of the brain. Cerebellar output is directed to the diencephalon and to a score of brainstem regions. This chapter reviews the cerebellar nuclear projections to the brainstem areas that give rise to descending connections that can influence motor programming at spinal cord level, i.e., the reticulospinal, vestibulospinal, rubrospinal, tectospinal, and interstitiospinal pathways. In addition, cerebellar projections to other areas will be briefly considered. Although cerebellar output is structured by the modular internal organization of cerebellar circuitry and related olivocerebellar connections, it is concluded that the modular output, i.e., output of individual cerebellar nuclei or parts thereof, still reaches many areas in the brainstem and diencephalon. In addition, multiple modules may converge their outputs indirectly to the same muscles. This suggests that multiple modules may each take part in different aspects of control of the same muscle or muscle group. Conversely, individual modules, due to the distributed nature of their outputs, may simultaneously affect several descending motor systems with the same ensuing goal. Tools for detailed anatomical and physiological studies are now available and are expected to enhance our knowledge on the effects of the cerebellum on descending motor pathways.
KeywordsCerebellar nuclei Reticulospinal tracts Rubrospinal tract Vestibulospinal tracts Tectospinal tract
List of Abbreviations
- 4/5 Cb
Cerebellar lobule 4/5
Anterior interposed nucleus
Anterior pretectal nucleus
Basal interstitial nucleus
Basilar pontine nuclei
Nucleus of Darkschewitsch
External cuneate nucleus
Fields of Forel
Gigantocellular reticular nucleus
Interstitial cell groups
Inferior cerebellar peduncle
Interstitial nucleus of Cajal
Lateral cerebellar nucleus
Lateral geniculate nucleus
Lateral reticular nucleus
Lateral vestibular nucleus
Medial cerebellar nucleus
Middle cerebellar peduncle
Medullary reticular nucleus
Medial geniculate nucleus
Medial longitudinal fascicle
Motor trigeminal nucleus
Medial vestibular nucleus
Nucleus reticularis tegmenti pontis
Parvicellular reticular formation
Parafascicular thalamic nucleus
Posterior interposed nucleus
Caudal pontine reticular formation
Oral pontine reticular formation
Magnocellular red nucleus
Parvicellular red nucleus
Superior cerebellar peduncle
Spinal trigeminal nucleus
Spinal vestibular nucleus
Superior vestibular nucleus
Ventral group thalamic nuclei
The cerebellum, which is dominantly involved in the coordination, adaptation, and learning of motor behavior and, most likely, also participates in visceral, affective, and cognitive functions, exerts its influence on these functions by way of the cerebellar nuclei. In addition, selected sets of Purkinje cells provide a direct input to specific regions of the vestibular nuclei. The specific connections of virtually all regions of the cerebellar nuclei with those parts of the thalamus that are connected to the primary and premotor cortices were already recognized and stressed in the early literature (Allen and Tsukahara 1974). However, the cerebellar nuclei also have a prominent direct impact on a number of premotor centers with connections to the spinal cord or lower brainstem. Here, a brief overview of the cerebellar nuclei will be provided followed by a review of their connections to these descending motor systems. Finally, these connections will be briefly assessed in relation to other regions that receive cerebellar output and their joint impact on cerebellar functioning and dysfunctioning.
The Cerebellar Nuclei
The MCN, in cat, macaque, and human often referred to as the fastigial nucleus as it is positioned directly adjacent to the fastigium (apex, roof) of the fourth ventricle, is the most medially located nucleus. Its rostral half is found directly dorsomedial to the rostral half of the fourth ventricle. Caudally, it is found dorsal to the lateral part of the nodulus. In the rat, as in mice, the MCN is characterized by a prominent outgrowth that reaches up into the white matter, consists of generally large cells, and is known as the dorsolateral protuberance (DLP; Korneliussen 1968). The rostral tip of the MCN is in ventrolateral direction continuous with the medial aspect of the superior vestibular nucleus. Caudally, the MCN is separated from both parts of the interposed nuclei by passing corticovestibular fibers. Located within these fibers are small groups of nuclear cells which are collectively known as the interstitial cell groups (ICG; Buisseret-Delmas et al. 1998). The ICG, based on their cortical and efferent connections, seem to be mostly associated with the posterior interposed nucleus (PIN; Buisseret-Delmas et al. 1993; Pijpers et al. 2005). Although the ICG are not mentioned in the description of the cerebellar nuclei of other mammals, they can also be recognized in the cat. In man, the fastigial nucleus also displays a dorsally extending protuberance; however, it is not known to what extent this is homologous to the DLP of rodents.
The PIN is found directly lateral to the caudal half of the MCN where it rests on the roof of the fourth ventricle. More rostrally and laterally, its borders with the anterior interposed and lateral cerebellar nuclei, respectively, are not always clear, especially in transverse sections. The rostromedial extension of the PIN continues as the interstitial cell groups (Buisseret-Delmas et al. 1993, 1998). In cat, the PIN carries a medial and lateral dorsal-ward protrusion. The human equivalent of the PIN is designated globose nucleus (Voogd and Ruigrok 2012).
The LCN, also referred to as dentate nucleus in cat and primates, is rostrally separated from the AIN by traversing fibers from the inferior cerebellar peduncle. In the rat, the LCN consists of the dorsolateral magnocellular part and a ventromedial parvicellular part. In the rhesus monkey, these areas actually form a dentated area, with the medial sheath consisting of mostly small neurons, whereas the lateral sheath is made up of mostly large cells. Only in the human, a multiple dentated nucleus is present, which is considerably larger than the other cerebellar nuclei. Within its dentated form, a dorsal microgyric part and a ventral macrogyric part are recognized (Voogd and Ruigrok 2012). Throughout the neuropil of all cerebellar nuclei, small cells are located that have been demonstrated to supply GABAergic input to the inferior olive (De Zeeuw et al. 1989; Fredette and Mugnaini 1991; Teune et al. 1995). In primates, including man, a loosely packed group of neurons, the basal interstitial nucleus (BIN), is found ventral to the LCN that is continuous in a rostrolateral direction toward the floccular peduncle and which maintains reciprocal connections with the vestibulocerebellum (Langer 1985; Ding et al. 2016). In the cat and rat, neurons in a similar position are found but form an even less homogeneous nuclear contour (Ruigrok 2003). The BIN has not been indicated in the reconstructions of Fig. 1.
Medial and Lateral Descending Motor Systems from the Brainstem
Hans Kuypers divided the pathways descending to the spinal cord into three groups (Kuypers 1981, 1985). The first group consists of the corticospinal tract, which is subdivided into the crossed lateral and uncrossed medial descending tracts. A second group of descending tracts originates from the brainstem and is also divided into a medial and a lateral descending system based on the funicular course of the fibers and the termination pattern of the participating fibers within either the medial or lateral part of the intermediate region (i.e., layers V–VIII) of the spinal cord. A third group originates from the locus coeruleus and subcoeruleus as well as from the raphe nuclei and terminates diffusely throughout the gray matter of the spinal cord. This review will focus on the cerebellar influence of the second group of descending connections, i.e., the reticulospinal, vestibulospinal, tectospinal, rubrospinal, and interstitiospinal tracts.
The medial and medioventral pontomedullary reticular formation are the origin of several long descending tract systems (Torvik and Brodal 1957). Delineation and identification of various subregions has proven to be difficult and may be different in different species. For example, in the rat, over 25 reticular regions have been shown to contain neurons with descending fibers (Newman 1985a, b). Many of these areas have recently been shown to contain neurons that are presynaptic to motoneurons or to spinal interneurons (Arber 2017; Esposito et al. 2014). Here, we will describe the cerebellar influence on the reticulospinal tracts in a general sense.
It is well established that reticulospinal connections may follow several routes to the spinal cord (Nyberg-Hansen 1965; Peterson et al. 1975). Reticulospinal neurons located in the pontine and rostral medullary reticular formation mostly, but certainly not exclusively, follow a course by way of the ventral funiculus where they are found adjacent to the fibers of the medial longitudinal fascicle (mlf). They terminate throughout the length of the spinal cord predominantly in the medial and central parts of the intermediate zone (i.e., in laminae VII–VIII). Additional reticulospinal tracts are found in the ipsilateral and contralateral ventrolateral funiculus. Both tracts are followed by the axons of reticulospinal neurons from more caudal regions. Projections from these tracts are found throughout all laminae of the spinal cord. Individual reticulospinal axons can collateralize over considerable distances, thereby affecting different regions of the body (Peterson et al. 1975). To further exemplify the heterogeneous nature of the reticulospinal tracts, it has been shown that the individual fibers may be excitatory (glutamatergic), inhibitory (containing GABA, glycine, or both), or modulatory (e.g., serotonin) (Holstege and Kuypers 1987). At least part of the reticulospinal axons can have monosynaptic excitatory effects on motoneurons located throughout the length of the spinal cord (Shapovalov and Gurevitch 1970; Shapovalov 1972) as was also recently elegantly verified in mice using selective, monosynaptic, and conditional viral tracing techniques (Capelli et al. 2017; Esposito et al. 2014).
From the above, it follows that the actions of the reticulospinal system are extremely diverse (for review, see Arshavsky et al. 1986). Indeed, reticulospinal systems have been suggested to be involved in maintaining and controlling ongoing motor activity such as locomotion, in the gating of somatosensory information to segmental as well as supraspinal levels and in the control of autonomic activity including pain modulatory systems (Fields 2004; Saper 2004; Arber 2017).
Overview of brainstem and diencephalic regions of the rat where labeled varicosities were found after injection with an anterograde tracer in the part of the cerebellar nuclei indicated and, at least partly, depicted in Figs. 2 and 3. Large dots denote dense labeling, intermediate dots indicate fair labeling, and small dots indicate sparse labeling. Question mark indicates that the area was not available for analysis; i indicates that the labeling was found ipsilateral to the injection side. (Based on Teune et al. 2000)
Medial Cerebellar Nucleus
Projections from especially the caudal aspect of the MCN have been shown to terminate in the dorsal aspect of the contralateral medial pontomedullary reticular formation by way of the uncinate tract in rat (Fig. 2d), cat, and monkey (Asanuma et al. 1983; Batton et al. 1977; Teune et al. 2000; Voogd 1964). In the monkey, the caudal region of the fastigial nucleus has been implicated in oculomotor functions subserving saccades (Noda 1991; Gonzalo-Ruiz et al. 1988). This area not only projects to the contralateral medial pontine reticular formation, known to be involved in saccade control (Enderle 2002), but also to the contralateral dorsomedial medullary reticular formation (Noda et al. 1990). This latter area would overlap the region from where neck muscles can be directly activated or inhibited (Peterson et al. 1975). Ipsilateral reticular projections that originate from the MCN are conspicuously less dense. They enter the brainstem by way of the direct fastigiobulbar tract located directly medial to the superior cerebellar peduncle (Teune et al. 2000; Voogd 1964; Ruigrok et al. 1990). The dorsolateral protuberance (DLP) of the MCN, which is prominent in rodents, has been shown to more specifically reach intermediate and lateral (i.e., parvicellular) parts of the contralateral reticular formation (Fig. 2c; Rubertone et al. 1990; Teune et al. 2000). The rostral MCN seems to preferentially target the vestibular nuclei (see below) and only sends moderate projections to the intermediate (mediolateral) levels of the reticular formation and to the lateral paragigantocellular nucleus (Fig. 2b). Bagnall et al. (2009) provided evidence that the ipsilateral connections of the MCN are glycinergic whereas glutamatergic MCN neurons project contralaterally. Glutamatergic but not glycinergic projections from the medial cerebellar nucleus control reticulospinal neurons involved in controlling manipulatory actions of the forelimbs (Esposito et al. 2014).
By definition, the Purkinje cells that innervate the MCN belong to the A zone (Voogd 1964; Voogd and Glickstein 1998), which has been subdivided into a number of subzones, each of which relate to a specific region of the MCN (Apps and Hawkes 2009; Voogd and Ruigrok 2004). Sugihara and Shinoda (2007) have suggested that three regions may be recognized in the MCN dealing with spinal (DLP and rostral MCN), eye movement (centrodorsocaudal MCN), and head orientation (centrocaudal MCN) aspects of motor control. The A zone has also been implicated in the control of emotional behavior, such as fear learning, affective state, cardiovascular control, and freezing behavior (Apps and Strata 2015; Bradley et al. 1991; Koutsikou et al. 2014), but it is not known to what extent this involvement is mediated by way of MCN projections to the reticular formation.
The Posterior Interposed Nucleus and the Interstitial Cell Groups
The ICG of the rat, which are intercalated between the MCN and both interposed nuclei, are the target of the Purkinje cells of the X and CX zone (Buisseret-Delmas et al. 1993, 1998; Voogd and Ruigrok 2004). They have been shown to have projections to the contralateral pontomedullary reticular formation, where they mostly terminate in the gigantocellular reticular nucleus (Fig. 2e; Teune et al. 2000). In the rat, apart from the nucleo-olivary projections, virtually no pontomedullary projections have been described that originate from the main body of the PIN.
The Anterior Interposed Nucleus and the Dorsolateral Hump
The dorsolateral hump (DLH) is sometimes also incorporated as part of the LCN (Angaut and Cicirata 1990), because it maintains reciprocal connections with the dorsomedial group of the principal olive (Ruigrok and Voogd 2000; Ruigrok et al. 2015). Nevertheless, it was originally described as part of the AIN (Goodman et al. 1963; Korneliussen 1968), and the DLH and lateral part of the AIN are the origin of the ipsilateral descending tract of the cerebellum (Bentivoglio and Kuypers 1982; Bentivoglio and Molinari 1985; Mehler 1967). They terminate mostly in the parvicellular regions of the ipsilateral pontomedullary reticular formation but also invade the deeper layers of the spinal trigeminal nucleus (Fig. 2f; Teune et al. 2000). Stimulation of the DLH induces movement of the lips, neck, and forelimbs (Cicirata et al. 1992; Angaut and Cicirata 1990). As such it has been suggested to be specifically involved in food manipulation. Indeed, the lateral and intermediate medullary reticular formation of the rat have been demonstrated to contain pre-oromotor neurons (Travers et al. 2000) and are involved with mastication and swallowing (Luo et al. 2001). More rostral regions of the lateral reticular region have been shown to be involved in vocalization in the squirrel monkey (Hannig and Jurgens 2006). The ventral part of the medullary reticular formation, on the other hand, has been shown to be critically involved in grasping (Esposito et al. 2014). Transneuronal transport of viral tracers suggests that the DLH may also be involved with eye blink as well as hind limb muscles (Morcuende et al. 2002; Ruigrok et al. 2008). The DLH receives its Purkinje fiber input from the D0 zone, which is intercalated between the D1 and D2 zones of lobules V–VII (Pijpers et al. 2005; Sugihara and Shinoda 2004).
The Lateral Cerebellar Nucleus
The connections of the LCN with the pontomedullary reticular formation are well documented for rat, cat, and monkey (Chan-Palay 1977; Teune et al. 2000; Tolbert et al. 1980). Particularly dense projections are noted in the contralateral gigantocellular reticular nuclei (Fig. 2g). The projection originates mostly from the magnocellular dorsal aspects of the nucleus suggesting that region receives its Purkinje cell projections mostly from the D2 zone. Projections also reach the ipsilateral reticular formation and have been shown to activate monosynaptically reticulospinal neurons to the lumbar cord in cat (Bantli and Bloedel 1975; Tolbert et al. 1980) and to the cervical cord in mice (Esposito et al. 2014). Especially, the recent work in mice suggests that the disynaptic dentate-reticulo-spinal may be used to control skilled motor behavior.
Classically, the vestibular nuclear complex is divided into a medial (MV), spinal (SpV), lateral (LV), and superior vestibular nucleus (SuV). In addition, several subgroups have been recognized in various animals (i.e., groups F, L, X, Y, and Z) (Brodal 1974; Highstein and Holstein 2006). Vestibulospinal tracts are divided into a medial vestibulospinal tract, which descends by way of the mlf, originates mainly from the MV, and reaches to cervical levels where its fibers terminate bilaterally, and into a lateral vestibulospinal tract, which runs lateral to the mlf, originates from the LV, and descends and terminates mostly ipsilaterally throughout the length of the spinal cord (Brodal 1974; Holstege and Kuypers 1982; Nyberg-Hansen 1964b; Highstein and Holstein 2006; Kasumacic et al. 2015; Lambert et al. 2016; Voogd 2016). Apart from their termination in the ventromedial part of the intermediate zone of the spinal cord, projections to the dorsal horn have been described (Bankoul and Neuhuber 1992). Moreover, the medial vestibulospinal tract may carry both inhibitory and excitatory fibers, whereas the LV neurons that contribute to the lateral vestibulospinal tract are excitatory. In addition to the vestibulospinal tracts, projections from especially the MV and SV are directed to the oculomotor centers where they can have excitatory or inhibitory influences. Some neurons, mostly located in the rostral, magnocellular part of the MV, collateralize to both the oculomotor centers as well as the cervical cord (Highstein and Holstein 2006; Ruigrok et al. 1995; Voogd and Barmack 2006). Finally, vestibular connections to autonomic brainstem systems are known that may regulate cardiovascular responses (Balaban and Porter 1998; Highstein and Holstein 2006).
The vestibular nuclei are intimately connected with the cerebellum (Voogd 2016; Voogd et al. 1996). Not only is a large part of its output directed to the cerebellar cortex and nuclei, but it also receives a main input from the cerebellum (e.g., see Fig. 2b). The cerebellar input to the vestibular complex is special because it is the only brainstem system that receives afferents from the cerebellar nuclei as well as directly from the cerebellar cortex.
Cerebellar Corticovestibular Projections
Direct projections from the cerebellar cortex to the vestibular nuclei arise from the vermis (lateral A and B zones), from the caudal vermis (ventral uvula and nodulus), and from the flocculus and adjacent ventral paraflocculus (Voogd et al. 1996). The projections from the Purkinje cells of the B zone, which is mostly located in lobules I–VIa but also has a small component in lobule VIII (Ruigrok et al. 2008; Sugihara and Shinoda 2004), to the neurons of the LV constitute the most direct influence of the cerebellar cortex on a descending pathway (Ito and Yoshida 1966; Voogd 2016). The B-zone-lateral vestibulospinal connection is mostly involved in the control of extensor or anti-gravity musculature (Arshavsky et al. 1986). Purkinje cell projections from the nodulus and ventral uvula can be found throughout most of the vestibular complex with the exception of the LV (Bernard 1987; Wylie et al. 1994). This region is implicated in the control of head movement. Floccular projections to the vestibular nuclei seem to terminate in a more orderly fashion. Purkinje cells that show modulation in their firing frequency upon visual stimulation around a vertical axis project predominantly to the magnocellular part of the MV, whereas Purkinje cells modulated by visual stimulation around a horizontal axis project to SV and the Y-group (De Zeeuw et al. 1994; Schonewille et al. 2006; Tan et al. 1995; Balaban et al. 2000). Floccular function is predominantly attributed to the control of the vestibulo-ocular reflex and its coordination with pursuit movements (Ilg and Thier 2008; Voogd and Barmack 2006). The Purkinje cells of the flocculus target a specific population of the MV, termed floccular target neurons, which have physiological characteristics resembling cerebellar projections neurons (Sekirnjak et al. 2003).
Cerebellar Nucleo-Vestibular Connections
More indirect control of vestibular function is also induced by the Purkinje cells of the A zone by way of the MCN. Apart from the reticular connections mentioned above, prominent, usually bilateral, projections to virtually all components of the vestibular nuclei are found (Teune et al. 2000), which mostly, but not exclusively, originate from the rostral part of the MCN (Fig. 2, Table 1). In the mouse ipsilateral MCN projections, like the projections to the reticular formation, have been shown to be glycinergic, whereas contralateral vestibular projections are glutamatergic (Bagnall et al. 2009). It is not known if cortical and MCN projections converge upon the same vestibular regions or neurons. Sparse vestibular projections from the other cerebellar nuclei arise from ICG, AIN, DLH, and LCN and are mostly ipsilateral (Delfini et al. 2000; Teune et al. 2000). They suggest that cerebellar processes are widely integrated with vestibular functions.
In most mammals the red nucleus (RN) can be recognized as a prominent round structure located centrally in the rostral part of the mesencephalic tegmentum. Usually, a distinction is made between a caudal magnocellular part (RNm) and a rostral parvicellular part (RNp). It should be recognized that this distinction to some extent relates to the difference in projections of both regions. The magnocellular part is the origin of the crossed rubrospinal tract. This part can be identified in the mesencephalon of most limb-carrying terrestrial vertebrates (ten Donkelaar 1988). In primates and carnivores, the parvicellular part is mostly associated with the uncrossed projection to the inferior olive by way of the central tegmental tract. Miller and Gibson (2009) have proposed the additional term of parvicellular contralateral red nucleus (RNpc) to identify small- and medium-sized neurons that also project to the contralateral spinal cord or the contralateral brainstem (Pong et al. 2008; Esposito et al. 2014). Thus, the axons of RNm and RNpc neurons cross at the level of the red nucleus, project mostly contralaterally, and basically only differ from each other with respect to the distance they have to travel to their respective main projection region, i.e., caudal cord versus rostral cord or brainstem. On the other hand, RNp neurons, by definition, project to the ipsilateral olive by way of the central tegmental tract. As such, the rodent red nucleus, as defined by the atlas of Paxinos and Watson (1998), essentially only consists of RNm and RNpc neurons (Ruigrok 2004; Rutherford et al. 1984). In many other species, such as cat, macaque, and human, a “real” RNp has been identified (for review see Onodera and Hicks 2009). In addition, in these animals, as well as in the rodent, immediately rostral, medial, and dorsal to the RN, several areas are located that basically surround the retroflex fascicle and carry names such as the nucleus of Bechterew, nucleus of Darkschewitsch, interstitial and rostral interstitial nucleus of the medial longitudinal fascicle, subparafascicular nucleus, and prerubral field. Homologies between various names that are presently in use are difficult to establish (e.g., see Onodera and Hicks 2009; Ruigrok 2004), but these nuclei mostly have in common the fact that they contain small neurons, which project massively to the inferior olive. For this reason they have been collectively named as nucleus parafascicularis prerubralis in the rat (Carlton et al. 1982; Ruigrok 2004). Onodera and Hicks (2009) have provided evidence that the projection from this mesodiencephalic area to the inferior olive is topographically organized and that some species differences may exist in whether these projections course by way of either the medial or central tegmental tracts.
Rubrospinal fibers terminate mostly within the intermediate zone of the spinal cord; however terminals are also found at cervical levels in lamina IX of monkey, cat, and rat (Miller and Gibson 2009) where they may sparsely terminate on motoneurons that innervate distal muscles (Al-Izki et al. 2008; Kuchler et al. 2002). Rubrospinal fibers also collateralize to the lateral reticular nucleus and to the anterior interposed nucleus (Beitzel et al. 2017; Huisman et al. 1983; Rajakumar et al. 1992; Shokunbi et al. 1986).
The function of the rubrospinal tract is still under debate. Selective lesion of this system produces only mild deficits in motor control. In cat and monkey, this affects grasping behavior and causes dragging of the dorsum of the paw during locomotion (Horn et al. 2002; Miller and Gibson 2009). In rat, mild changes in locomotion pattern have been described (Muir and Whishaw 2000; Whishaw et al. 1998). RNm stimulation mostly seems to affect distal extensor muscles in cat (Horn et al. 2002), whereas most RNm units display up-modulation during the swing phase in stepping decerebrate cats (Arshavsky et al. 1986, 1988).
Many authors have mentioned that the rubrospinal tract diminishes with the expansion of the corticospinal tract. In man, it is questioned if the rubrospinal tract, which may consist of only several hundreds of fibers, descends beyond the cervical level (for review see Onodera and Hicks 2009). The decrease in size of the rubrospinal tract is countered by an increase in the importance of the RNp and adjacent regions to the inferior olive. It has been speculated that the relative increase in the projection from the mesodiencephalic junction nuclei to the inferior olive or to specific parts thereof mimics the increased control of cerebellar and cortical structures over specific motor functions and, in man, may have enabled the advent of bipedalism and speech (Onodera and Hicks 1999, 2009; Hicks and Onodera 2012).
Cerebellar Projections to the Red Nucleus
The medial aspect of the PIN projects to the medial margin of the RNm in rat (Fig. 3a, e; Daniel et al. 1987) and cat (Robinson et al. 1987). This projection is less dense compared to the input derived from the AIN. RNm projections originating from the MCN and LCN are sparse or non-existent, the latter mostly targeting regions immediately lateral and rostral to the RNm (Fig. 3). The PIN and LCN projections to the primate and cat RNp and the rodent area of the nucleus parafascicularis prerubralis function in a cerebellar nucleo-midbrain-olivary loop (De Zeeuw and Ruigrok 1994; Onodera and Hicks 2009; Ruigrok and Voogd 1995; Guillain et al. 1933).
The superior colliculus is mostly associated with visual input and is involved in directing gaze to objects of interest. This is performed by saccades, head movements, or a combination of both. The control of the superior colliculus over the paramedian pontine reticular formation from where saccades are initiated is well established in most mammals (Enderle 2002). However, in addition, arm or even whole body movements may be evoked by stimulation of the superior colliculus. These body movements are triggered by tectoreticular and tectospinal pathways (Harting 1977; Nyberg-Hansen 1964a; Rose and Abrahams 1978). The tectospinal component seems to be small in most mammals with the largest component to be found in carnivores (Meredith et al. 2001; Murray and Coulter 1982; Nudo and Masterton 1989). Most tectospinal neurons do not project beyond the cervical segments from where the neck muscles are innervated (Nudo and Masterton 1989). Most axons arise from the caudolateral quadrant and, after decussating in the dorsal tegmental decussation, descend contralaterally close to the mlf where they form the predorsal bundle. In cat, a small ipsilateral projecting group of tectospinal fibers has been described (Olivier et al. 1994). Although tectospinal cells may terminate directly on cervical motoneurons (Olivier et al. 1995), conclusive anatomical proof has not yet been obtained (Muto et al. 1996; Shinoda et al. 2006). Therefore, the most direct impact of superior colliculus on cervical neck motoneurons seems to involve at least a segmental or reticular relay (Kakei et al. 1994; Shinoda et al. 2006).
Cerebellar Nucleo-Tectal Connections
The cerebellar nuclei have a profound and often neglected impact on the mesencephalic tectum. Cerebellar projections to this region mostly terminate in the intermediate and deep layers of the contralateral superior colliculus (Fig. 3, Table 1). In various animals they have consistently been shown to originate predominantly from two areas of the cerebellar nuclei. The lateral aspect of the PIN and adjacent area of the LCN supply the bulk of cerebellotectal projections (Hirai et al. 1982; Kurimoto et al. 1995; May and Hall 1986; May et al. 1990; Uchida et al. 1983). Both regions mediate information from the paraflocculus, which is known to receive mostly extrastriate visual information by way of the dorsolateral basal pons in rat, cat, and monkey (Gayer and Faull 1988; Glickstein et al. 1994; Robinson et al. 1984; Kralj-Hans et al. 2007). In the macaque, neurons in the lateral PIN and LCN have been shown to connect disynaptically to visually related regions in the posterior parietal cortex (Prevosto et al. 2010).
Projections to the superior colliculus also arise from the caudal, oculomotor, half of the MCN (Fig. 3a, b). This part mostly receives cortical input from the oculomotor vermis (caudal lobule VI and lobule VII) and has been suggested to be specifically involved with the control and adaptation of saccades (Thier et al. 2002; Noda and Fujikado 1987; Fujikado and Noda 1987; Takagi et al. 1998). Interestingly, the MCN of the rabbit, and hardly that of the gray squirrel, do not project to the superior colliculus (Uchida et al. 1983; May and Hall 1986), a finding that seems to correlate well with the observation that the rabbit superior colliculus does not seem to be essential for generating either saccades or optokinetic nystagmus (Collewijn 1975). The lateral aspect of the AIN has also been reported to provide input to the superior colliculus in the rat and rabbit (Kurimoto et al. 1995; Uchida et al. 1983). In the cat and monkey, the MCN has been shown to terminate bilaterally, generally located somewhat more superficial compared to the PIN projections which terminate strictly contralateral and deeper in the intermediate layer (Kawamura et al. 1982; May et al. 1990). Both types of projections seem to make use of a different type of synapse as judged from their respective degeneration characteristics (Warton et al. 1983) and have been suggested to play different roles in the control of saccades. A clear topographic pattern, however, has not been described in a variety of mammals (Hirai et al. 1982; Kawamura et al. 1982; Kurimoto et al. 1995; May and Hall 1986; May et al. 1990; Noda et al. 1990; Uchida et al. 1983; Teune et al. 2000). The ventrolateral aspect of the caudal superior colliculus reportedly receives the densest cerebellar input (Fig. 3). However, cerebellar terminals are also found in tectal areas that contain large tectospinal neurons. Presently, no specific anatomical information seems to be available that positively identifies the collicular targets of the cerebellar output (Warton et al. 1983). Although GABAergic projections to the pretectum have been reported to arise from the lateral PIN and ventral LCN in the cat (Nakamura et al. 2006), no evidence for cerebellar inhibitory synapses in the superior colliculus has been found (Warton et al. 1983).
The interstitiospinal tract originates from scattered large neurons located dorsomedially to the red nucleus and lateral to the oculomotor nuclei and periaqueductal gray within and surrounding the mlf. This poorly defined region is known as the interstitial nucleus of Cajal (INC). Fibers descend ipsilaterally by way of the mlf to the spinal cord where they terminate in laminae VII and VIII of the spinal gray. The interstitiospinal tract has excitatory monosynaptic contacts with neck musculature but also provides di- and polysynaptic mostly excitatory activation of back, forelimb, and hind limb muscles (Fukushima et al. 1978; Holstege and Cowie 1989). The INC, apart from acting as origin of the interstitiospinal tract, is also involved in the control of eye movements. Neurons participating in this function appear to form a different population from the cells that give rise to the interstitiospinal tract (Bianchi and Gioia 1995; Zuk et al. 1983). Projections from the INC also reach the oculomotor nuclei, the pontine reticular formation, and the vestibular nuclei. As such it plays an important role in coordinating eye and head movements. The area directly rostral the INC, known as the rostral interstitial nucleus of the mlf and containing many parvalbumin-containing neurons, provides input to the INC and is specifically involved in the control of vertical eye movements (Buttner-Ennever 2006; Fukushima-Kudo et al. 1987; Fukushima 1991; Horn and Buttner-Ennever 1998; Tolbert et al. 1978b).
Cerebellar Projections to the Interstitial Nucleus of Cajal
Most prominent cerebellar projections to the INC arise from the oculomotor part of the MCN in rat (Fig. 3b), cat, and monkey (Noda et al. 1990; Teune et al. 2000; Sugimoto et al. 1982). However, it also receives an afferent contribution from other parts of the cerebellar nuclei (Table 1) (Chan-Palay 1977; Teune et al. 2000).
Cerebellar Projections to Other Areas
Apart from the cerebellar projections to the five classic descending premotor tracts, all cerebellar nuclei also provide input to various regions of thalamus (Aumann et al. 1994; Chan-Palay 1977; Teune et al. 2000), which will not be reviewed here (but see Table 1). Also, many cerebellar nuclear neurons have been shown to provide input to the cerebellar cortex (Provini et al. 1998; Batini et al. 1992; Buisseret-Delmas and Angaut 1988; Tolbert et al. 1978b; Gao et al. 2016; Houck and Person 2015; Ankri et al. 2015), some of which may be inhibitory (Uusisaari and Knopfel 2010; Batini et al. 1989; Ankri et al. 2015).
In addition, the cerebellar nuclei provide input to a number of precerebellar nuclei in the brainstem. Cerebellar nuclear inputs to the reticulotegmental nucleus of the pons and the basal pontine nuclei are well-known and have been shown to participate in potential functionally important reverberating circuitry involving cerebello-bulbo-cerebellar loops (Mock et al. 2006; Tsukahara et al. 1983). However, projections to nuclei that give rise to the descending motor tracts may also provide feedback to the cerebellum, i.e., the vestibular nuclei and reticular formation form an important source of mossy fiber input to the cerebellum. In addition, many rubrospinal fibers have been demonstrated to collateralize specifically to the AIN (Huisman et al. 1983; Beitzel et al. 2017). A more prominent connection to a precerebellar nucleus, be it of a very different category, is formed by the cerebellar nuclear projections to the inferior olive. This connection has been shown to precisely match the collateral projection of the cerebellar climbing fibers to the cerebellar nuclei (Ruigrok and Voogd 1990, 2000) and to consist exclusively of GABAergic fibers (De Zeeuw et al. 1989; Fredette and Mugnaini 1991) that originate from a population of small neurons intermingled with the larger and generally excitatory projection neurons and to possess their own physiological characteristics (Fredette and Mugnaini 1991; Uusisaari et al. 2007; Najac and Raman 2015). The population of small GABAergic neurons in the cerebellar nuclei most likely exclusively targets the inferior olive (Teune et al. 1995; De Zeeuw and Ruigrok 1994).
Another major target of cerebellar nuclear projections is formed by nuclei that themselves supply a major input to the inferior olive. Especially the nuclei in the mesodiencephalic junction participate in this cerebellar nucleo-mesodiencephalo-olivo-cerebellar loop (De Zeeuw and Ruigrok 1994; Ruigrok and Voogd 1995). In man, the main circuit is formed by the projections from the dentate nucleus to the parvicellular RN, which, by way of the central tegmental tract, targets the principal olive. Although the function of this circuit is far from clear (Ruigrok 1997; Ruigrok and Voogd 1995; Hoebeek et al. 2010; Kennedy 1990), this so-called triangle of Guillain-Mollaret (Guillain et al. 1933) has received attention as lesions of this circuit may be instrumental to inducing myoclonus, or rhythmic tremor, of the palatal muscles (sometimes also involving pharyngeal, laryngeal, diaphragm, or extraocular muscles) but also of conspicuous enlargement of particular parts of the inferior olive (Gautier and Blackwood 1961; Boesten and Voogd 1985; Ruigrok et al. 1990; Shaikh et al. 2010).
Finally, cerebellar nuclear connections have been described that do not fall within any of the categories mentioned above (Table 1). Examples include the cerebellar projections to hypothalamic areas, zona incerta, parvicellular reticular formation, parabrachial nuclei, and periaqueductal gray (Teune et al. 2000; Zhu and Wang 2008; Haines et al. 1997). These connections suggest a wide impact of cerebellar processing involving various autonomic as well as pain functions.
Divergence of Cerebellar Projections
Convergence of Cerebellar Projections
The modular characteristics of the cerebellar circuitry, where strips of Purkinje cells converge upon an entity of the cerebellar nuclei and that is matched by the organization of the olivocerebellar climbing fiber system, have been suggested to lie at the basis of the functional working blocks of the cerebellum (Apps and Garwicz 2005; Ruigrok 2011). However, from the account sketched above, it will be obvious that as yet it is by no means clear how individual cerebellar modules interact with brainstem structures to result in functionally meaningful signals. For example, individual modules, by way of their collateralizing output, are capable of influencing numerous brainstem structures simultaneously, which may include several nuclei that have descending connections to the spinal cord. On the other hand, as shown with viral tracing techniques, multiple modules may participate in the control of the same muscle. This particular organization suggests that each cerebellar module may serve a specific function in the control of muscles. The interaction of multiple strips of Purkinje cells with several targets areas has most clearly been demonstrated in the relatively simple system of floccular control on reflexive eye movements (van der Steen et al. 1994; Voogd and Wylie 2004; Ruigrok et al. 1992; Sugihara et al. 2004; Wylie et al. 2017). However, studies that attempted to study the contribution of individual modules to skeletomotor function also suggest that a single module may affect a particular type of muscle control (Horn et al. 2010; Cerminara and Apps 2010; Pijpers et al. 2008). For example, impairment of the C1 hind limb module has been shown to affect the phase-locked modulation of reflexes during locomotion (Pijpers et al. 2008). Inactivation or lesion of various olivary regions, resulting in inactivation or severe functional impairment of related cerebellar nuclear areas, also results in highly specific deficits in motor control (Horn et al. 2010; Cerminara and Apps 2010). Hence, different cerebellar modules seem to be able to participate in the control of potentially the same muscles that are being used in different functional contexts.
From this account it will be obvious that the impacts of cerebellar malfunctioning are diverse and not infrequently difficult to fully understand. Due to the distributed connectivity, specific damage of the vestibulocerebellar, vermal, paravermal, and hemispheral components of the cerebellum results in various combinations of specific cerebellar disorders (Thach and Bastian 2004). Conversely, due to the nature of ongoing spinocerebellar degenerative disease, this may differentially influence a multitude of cerebellar modules. Likewise, cerebellar tumors or lesions will almost invariably involve several adjacent modules in an incomplete way (i.e., the modules are organized in longitudinal cortical arrays which can be discontinuous and may be found in both the anterior and posterior lobes (Pijpers et al. 2006; Ruigrok 2011). The resulting clinical syndrome, therefore, is hard to predict or explain.
Nevertheless, and despite the recent recognition that the cerebellum may be involved in a large array of non-motor brain functions, cerebellar disorders can lead to very clear motor deficits. Obviously, due to the double decussation of most cerebellofugal projections (i.e., the decussation of the superior cerebellar peduncle in the caudal mesencephalon, followed by the decussation of corticospinal and rubrospinal tracts), unilateral cerebellar deficits usually involve affected movements of the ipsilateral side of the body.
Cerebellar-based handicaps in motor control commonly involve three symptoms: muscle weakness (hypotonia), ataxia, and tremor (Thach and Bastian 2004).
Sudden removal of cerebellar nuclear output in cases of, e.g., trauma, will diminish the tonic excitatory drive to premotor regions. This is likely to result in a general reduction of activation of the related motoneuron pools, which will be reflected in reduced muscle tone (hypotonia). Similarly, however, cerebellar lesions that do not involve the nuclei are likely to diminish the inhibitory drive of the cortical Purkinje cells on the cerebellar nuclei, thereby increasing nuclear output resulting in hypertonic responses. Hypotonia and hypertonia may (partly) disappear after a period of weeks or months, probably by adjustment of the tonic drive to premotor regions by other, unaffected, cerebellar or extracerebellar brain regions.
Although ataxia is typical for cerebellar disorders, it is not easily characterized. It usually reflects the inability of a patient to execute voluntary movements in a “normal” way. Movements, or parts thereof, are not initiated at the correct moment, are not terminated at the appropriate time, and are not corrected adequately. This results in improperly conducted and dysmetric movements. To some extent hypo- and hypertonia of muscles or muscle groups form the foundation of these motor deficits. Because antagonistic and correcting activity is inadequately timed, oscillations will easily occur around the intended target or result in pendular reflexes. Due to dysmetria and failure to time movements, cerebellar patients also encounter problems with the execution of fast rhythmic movements (dysdiadochokinesis). When these disorders concern the mouth, pharynx, or larynx musculature, it results in cerebellar dysarthria that typically involves changes in rhythm and amplitude in combination with a careless or “slurred” pronunciation.
This form of tremor is also very characteristic in cerebellar disorders and may be related to dysfunction of properly timed components of movements, resulting in a disturbance in the beat and rhythm of movements with tremor as a consequence. Increased pendular reflexes are clearly related to this phenomenon. Cerebellar nystagmus may be seen as a form of intention tremor reflecting dysfunction of the vestibulocerebellum.
Conclusions and Future Directions
The cerebellum is usually appreciated as a structure with a uniform internal structure that will perform a particular type of information processing. Within this internal structure, based on the organization of corticonuclear and olivocerebellar connections, a number of parallel, longitudinally organized, modules can be recognized which form functional entities. The organization of the input to these modules and in particular the organization of their output channels, therefore, will determine the type of information processed within a module and which structures will be informed of its result. The above account shows that the output of a particular module is directed to many regions in brainstem and diencephalon and may affect multiple centers with descending connections to the spinal cord. In addition, several modules may ultimately affect activity patterns of the same muscle or muscle group. In order to fully understand the cerebellar involvement in motor control and learning (and in other functions), it will be imperative to understand to what extent individual modules or micromodules control different aspects of controlling muscles. In addition, a precise description of the often di- and polysynaptic pathways involved will be necessary. The advent of new neuroanatomical and physiological techniques such as selective (modular) lesioning, conditional transneuronal tracing, single-axon reconstructions, and optogenetics is expected to greatly aid in fulfilling these requirements for further understanding cerebellar function and dysfunction.
This work was supported by the Dutch Ministry of Health, Welfare and Sports (T.R.).
- Ankri L, Husson Z, Pietrajtis K, Proville R, Lena C, Yarom Y, Dieudonne S, Uusisaari MY (2015) A novel inhibitory nucleo-cortical circuit controls cerebellar Golgi cell activity. Elife 4. https://doi.org/10.7554/eLife.06262
- Apps R, Garwicz M (2005) Anatomical and physiological foundations of cerebellar information processing. Nat Rev 6(4):297–311Google Scholar
- Apps R, Hawkes R (2009) Cerebellar cortical organization: a one-map hypothesis. Nat Rev 10(9):670–681Google Scholar
- Arshavsky YI, Gelfand IM, Orlovsky GN (1986) Cerebellum and rhythmical movements. Studies of brain function, vol 13. Springer, Berlin/HeidelbergGoogle Scholar
- Aumann TD, Rawson JA, Finkelstein DI, Horne MK (1994) Projections from the lateral and interposed cerebellar nuclei to the thalamus of the rat: a light and electron microscopic study using single and double anterograde labelling. J Comp Neurol 349(2):165–181. https://doi.org/10.1002/cne.903490202CrossRefPubMedGoogle Scholar
- Bagnall MW, Zingg B, Sakatos A, Moghadam SH, Zeilhofer HU, du Lac S (2009) Glycinergic projection neurons of the cerebellum. J Neurosci 29(32):10104–10110Google Scholar
- Batton RR, Jayaraman D, Ruggiero D, Carpenter MB (1977) Fastigial afferent projections in the monkey: an autoradiographic study. J Comp Neurol 174:281–306Google Scholar
- Beitzel CS, Houck BD, Lewis SM, Person AL (2017) Rubrocerebellar feedback loop isolates the interposed nucleus as an independent processor of corollary discharge information in mice. J Neurosci 37(42):10085–10096. https://doi.org/10.1523/JNEUROSCI.1093-17.2017CrossRefPubMedPubMedCentralGoogle Scholar
- Bentivoglio M, Molinari M (1985) Crossed divergent axon collaterals from cerebellar nuclei to thalamus and lateral medulla oblongata in the rat. Brain Res 362:180–184Google Scholar
- Brodal A (1974) Anatomy of the vestibular nuclei and their connections. In: Kornhuber HH (ed) Handbook of sensory physiology, vol VI, I: Vestibular system. Springer, New York, pp 239–352Google Scholar
- Cerminara NL, Apps R (2010) Behavioural significance of cerebellar modules. Cerebellum. https://doi.org/10.1007/s12311-010-0209-2
- Chan-Palay V (1977) Cerebellar dentate nucleus: organization, cytology and transmitters. Springer, BerlinGoogle Scholar
- Ding SL, Royall JJ, Sunkin SM, Ng L, Facer BA, Lesnar P, Guillozet-Bongaarts A, McMurray B, Szafer A, Dolbeare TA, Stevens A, Tirrell L, Benner T, Caldejon S, Dalley RA, Dee N, Lau C, Nyhus J, Reding M, Riley ZL, Sandman D, Shen E, van der Kouwe A, Varjabedian A, Write M, Zollei L, Dang C, Knowles JA, Koch C, Phillips JW, Sestan N, Wohnoutka P, Zielke HR, Hohmann JG, Jones AR, Bernard A, Hawrylycz MJ, Hof PR, Fischl B, Lein ES (2016) Comprehensive cellular-resolution atlas of the adult human brain. J Comp Neurol 524(16): 3127–3481. https://doi.org/10.1002/cne.24080CrossRefPubMedPubMedCentralGoogle Scholar
- Fields H (2004) State-dependent opioid control of pain. Nat Rev 5(7):565–575Google Scholar
- Fukushima K, van der Hoeff-van Halen R, Peterson BW (1978) Direct excitation of neck motoneurons by interstitiospinal fibers. Exp Brain Res 33(3–4):565–581Google Scholar
- Graf W, Gerrits N, Yatim-Dhiba N, Ugolini G (2002) Mapping the oculomotor system: the power of transneuronal labelling with rabies virus. Eur J Neurosci 15(9):1557–1562Google Scholar
- Guillain GC, Mollaret P, Bertrand IG (1933) Sur la lesion responsable du syndrome myoclonique de tronc cerebral. Rev Neurol (Paris) 3:666–673Google Scholar
- Holstege G, Kuypers HG (1982) The anatomy of brain stem pathways to the spinal cord in cat. A labeled amino acid tracing study. Prog Brain Res 57:145–175Google Scholar
- Horn AK, Buttner-Ennever JA (1998) Premotor neurons for vertical eye movements in the rostral mesencephalon of monkey and human: histologic identification by parvalbumin immunostaining. J Comp Neurol 392(4):413–427Google Scholar
- Kasumacic N, Lambert FM, Coulon P, Bras H, Vinay L, Perreault MC, Glover JC (2015) Segmental organization of vestibulospinal inputs to spinal interneurons mediating crossed activation of thoracolumbar motoneurons in the neonatal mouse. J Neurosci 35(21):8158–8169. https://doi.org/10.1523/JNEUROSCI.5188-14.2015CrossRefPubMedPubMedCentralGoogle Scholar
- Kuypers HGJM (1981) Anatomy of the descending pathways. In: Brookhart JM, Mountcastle VB, VBrooks VB, Geiger SR (eds) Handbook of physiology. Volume II, Motor control Part 1. American Physiological Society, Bethesda, pp 597–666Google Scholar
- Kuypers HGJM (1985) The anatomical and functional organization of the output of the motor system. In: Swash E (ed) Science basis of clinical neurology. Churchill Livingstone, Edinburgh, pp 1–17Google Scholar
- Langer TP (1985) Basal interstitial nucleus of the cerebellum: cerebellar nucleus related to the flocculus. J Comp Neurol 235:38–47Google Scholar
- Lu X, Miyachi S, Ito Y, Nambu A, Takada M (2007) Topographic distribution of output neurons in cerebellar nuclei and cortex to somatotopic map of primary motor cortex. Eur J Neurosci 25(8):2374–2382Google Scholar
- Mehler WR (1967) Double descending pathways originating from the superior cerebellar peduncle: an example of neural species differences. Anat Rec 157:374Google Scholar
- Miller LE, Gibson AR (2009) Red nucleus. In: Squire LR (ed) Encyclopedia of neuroscience, vol 8. Elsevier, Amsterdam, pp 55–62Google Scholar
- Onodera S, Hicks TP (1999) Evolution of the motor system: why the elephant’s trunk works like a human’s hand. Neuroscientist 5(4):217–226Google Scholar
- Paxinos G, Watson C (1998) The rat brain in stereotaxic coordinates, 4th edn. Academic, San DiegoGoogle Scholar
- Paxinos G, Huang X-F, Toga AW (2000) The rhesus monkey in stereotaxic coordinates. Academic, San DiegoGoogle Scholar
- Prevosto V, Graf W, Ugolini G (2010) Cerebellar inputs to intraparietal cortex areas LIP and MIP: functional frameworks for adaptive control of eye movements, reaching, and arm/eye/head movement coordination. Cereb Cortex 20(1):214–228Google Scholar
- Ruigrok TJH (2004) Precerebellar nuclei and red nucleus. In: Paxinos G (ed) The rat nervous system, 3rd edn. Elsevier Academic, San Diego, pp 167–204Google Scholar
- Ruigrok TJ, Voogd J (2000) Organization of projections from the inferior olive to the cerebellar nuclei in the rat. J Comp Neurol 426(2):209–228Google Scholar
- Ruigrok TJH, de Zeeuw CI, Voogd J (1990) Hypertrophy of inferior olivary neurons: a degenerative, regenerative or plasticity phenomenon. Eur J Morphol 28:224–239Google Scholar
- Ruigrok TJ, Pijpers A, Goedknegt-Sabel E, Coulon P (2008) Multiple cerebellar zones are involved in the control of individual muscles: a retrograde transneuronal tracing study with rabies virus in the rat. Eur J Neurosci 28(1):181–200. https://doi.org/10.1111/j.1460-9568.2008.06294.xCrossRefPubMedGoogle Scholar
- Ruigrok TJH, Sillitoe RV, Voogd J (2015) Cerebellum and cerebellar connections. In: Paxinos G (ed) The rat nervous system, 4th edn. Elsevier, Amsterdam, pp 133–205Google Scholar
- Saper CB (2004) Central autonomic system. In: Paxinos G (ed) The rat nervous system, 3rd edn. Elsevier Academic, San Diego, pp 761–796Google Scholar
- Voogd J (1964) The cerebellum of the cat: structure and fiber connections. Van Gorcum, AssenGoogle Scholar
- Voogd J (2004) Cerebellum. In: Paxinos G (ed) The rat nervous system, 3rd edn. Elsevier Academic, San Diego, pp 205–242Google Scholar
- Voogd J, Glickstein M (1998) The anatomy of the cerebellum. Trends Neurosci 2:305–371Google Scholar
- Voogd J, Ruigrok TJH (2012) Cerebellum and precerebellar nuclei. In: Mai JK, Paxinos G (eds) The human nervous system, 3rd edn. Elsevier, Amsterdam, pp 471–545Google Scholar