Postnatal development of the cerebellar cortex
In human, the cerebellum continues to develop through childhood and adolescence, reaching its full structural growth by the 15th to 20th years of life (Diamond, 2000). In rodents, the cerebellum is also remarkably immature at birth (Woodward et al., 1971), and undergoes an intense period of maturation that lasts for about 3 weeks (Figure 1.6), and that prolongs up to 7-8 weeks (McKay and Turner, 2005).
In early postnatal days, all Purkinje cells are innervated by multiple CFs (Crepel et al., 1976; Mariani and Changeux, 1981; Crepel, 1982). These multiple CFs initially form synapses on the perisomatic processes of Purkinje cell in newborn mice (Chedotal and Sotelo, 1993). Progressively, CFs forsake the soma to invade the Purkinje cell’s proximal dendrites where they make strong synaptic contacts (Palay and Chan-Palay, 1974). Supernumerary CFs are eliminated eventually with the progress of postnatal development, and mono-innervation is attained by the end of the third postnatal week in mice (Kano et al., 1995). There have been several decades of investigations to understand the mechanisms of this selective activity-dependent regression. However, this question remains uncompletely resolved. The current view on the subject makes the developmental long term plasticities key players of these processes.
In the mouse, morphological data demonstrated that the synapses between PFs and Purkinje cell are established at around P7 (Zhao et al., 1998). Granule cells are generated by the vigorous proliferation of their progenitors in the external germinal layer during the first two postnatal weeks. This leads to a huge number of cells. Post-mitotic granule cells then bilaterally extend their axons, the PFs, and their cell bodies migrate downward in the developing molecular layer. By the third postnatal week, they finally settle in the internal granular layer underneath the Purkinje cell layer. During these first three weeks, cells in the pia mater play a role in granule cell proliferation and migration, whereas Bergman glia extends their processes into the molecular layer, guiding migrating granule cells (Altman, 1975). Internal granule cells further differentiate, forming synapse complexes in glomeruli that gather excitatory afferent mossy fibers and inhibitory Golgi cell axons. Some differentiating granule cells undergo an apoptotic cell death, which is thought to ensure the fine-tuning of the proper cell numbers and connectivity. Simultaneously, Purkinje cells undergo a massive outgrowth of dendrites and form elaborate arborizations containing numerous synapses with extending PFs. Supernumerary CFs are pruned away by competing with PFs. In addition, the competition for Purkinje cell territories is mutually regulated between PFs and CFs by their respective activity.
Cerebellum and motor control
How is information processed in the stereotyped cerebellar microcircuit? Independently of learning, the cerebellum plays a role in the adequate execution of movement. There are several trends of theories about cerebellar functions, sometimes controversial. Many of them are based on the notion that cerebellum contains “i nternal models” of the motor apparatus. These internal models encode the representation of dynamic properties of body part that enables the central nervous system to predict the consequences of motor commands and to determine those required to perform specific tasks. The internal model mimics the behaviour of the sensorimotor system in the external environment, and helps the brain, by prediction, to perform the movement precisely, without the need to refer to feedback from the moving body part. These internal models allow to predict the more adequate set of actions in a given context but the changes in the context, or the learning of new movements, requires them to be plastic. For this reason, in addition to this classical view of its acute correction of motor programs, the cerebellum has long been proposed to be the place of motor learning.
Cerebellum and motor learning
It is thought that the cerebellum might store motor memory in the form of internal models (Imamizu et al., 2000; Kawato et al., 2003).The unusual architecture of the cerebellar cortex first inspired theoretical models of its function, later confirmed by experimentation. Brindley (1969) proposed that we initially generate movements « consciously, » under higher cerebral control. As the movement is practiced, the cerebellum learns to link this movement to the context in which it is executed. Marr and Albus (in 1969-1971), soon followed by Ito (1972), proposed that the linkage between the contextual input and the appropriate motor output is established and stored in the cerebellar neuronal microcircuit, through the PF-Purkinje cell synapses. This is the so called Marr-Albus-Ito model of learning in the cerebellum (Figure 1.7). The PF-Purkinje cell connection is modified during the period of learning by the activity of the CF which conveys error signals and induces a long term change in synaptic strength. This change is input specific and is named the long-term depression (LTD) of PF–Purkinje-cell synapses (for reviews, see Ito, 2001, 2006). When the linkage is complete, the occurrence of the context (represented by a certain input to the cerebellum) will trigger through the cerebellar microcircuit the appropriate motor response (output). This explains how we become able to move skillfully after repeated practice. The « learned » movement is distinguished from the « unlearned » conscious movement by being automatic, rapid, and stereotyped. It is worth mentioning here that the first experimental evidence of LTD in the early 1980s came from the observation of changes in the rate of discharge of rabbit Purkinje cells in vivo (Ito et al., 1982; Ekerot and Kano, 1985). These changes were attributed to synaptic plasticity. Even though this LTD was confirmed by the study of synaptic currents later on, one can not exclude that a part of this plasticity could rely on intrinsic plasticity.
Delta2 glutamate receptors (GluRdelta2)
Remarkably, Purkinje cells also express delta2 glutamate receptors (GluRdelta2), which are orphan receptors expressed almost exclusively by these cells. GluRdelta2 are located specifically at PF–Purkinje cell synapses, except d uring a short period of the early development in which they are also found at CF synapses (Araki et al., 1993; Takayama et al., 1996; Zhao et al., 1998). GluRdelta2 have been cloned by sequence homology with AMPA-Rs and NMDA-Rs (Yamazaki et al., 1992a; Araki et al., 1993; Lomeli et al., 1993). Interestingly, the structural comparison with other ionotropic glutamate receptors has recently demonstrated that the amino acid composition of the ligand-binding cavity of GluRdelta2 is most similar to that of the NR1 subunit of the NMDA-R (Naur et al., 2007). Like NR1 subunit, GluRdelta2 can bind Glycine or D-serine. However, neither its agonist (if one), nor its physiological role, has been clearly identified.
Last but not least, NMDA receptors, which are the focus of this thesis, are also expressed in postnatal and adult Purkinje cells. They will be extensively treated in the last part of this review.
NMDA-R: general properties
NMDA-Rs are ionotropic glutamate receptors that play a major role in many cerebral processes like development, neuroplasticity and neuronal death. They display unique features among ligand-gated ionotropic receptors. To be activated, they not only must bind both glutamate and the co-agonist glycine -or D-serine- (Johnson and Ascher, 1987), but they also require a coincident membrane depolarization, in order to relieve a Mg2+ block of the receptor ion channel (Mayer et al., 1984). This remarkable property of both voltage-dependent and ligand-gated channel confers on the NMDA-R the capacity to act as a molecular coincidence detector of simultaneous pre- and postsynaptic excitation. Since NMDA-Rs display a high permeability to Ca2+ ions, their activation causes a large influx of Ca2+ into cells (MacDermott et al., 1986) that initiates signal transduction cascades, triggering for instance LTD or LTP of synaptic currents.
NMDA-Rs are heteromeric complexes of different subunits classified, to date, in three families: NR1 (eight splice variants), NR2 (NR2A, -2B, -2C, and -2D), and the recently characterized NR3 subunits family (NR3A and NR3B). In the mouse, some NMDA-R subunits are differently denominated: NR-Zeta, NR-epsilon1, -epsilon2, -epsilon3, -epsilon4, NR-chi1, -chi2 which correspond to NR1, NR2A, -2B, -2C, -2D, NR3-A and -3B respectively. To simplify the understanding of the present thesis report, I have chosen the first denomination, whatever the species.
NMDA-Rs are tetramers of two mandatory glycine-binding NR1 subunits and two glutamate-binding NR2 subunits that can be identical or different. The type of NR2 subunit is critical in determining some of the key biophysical and pharmacological properties of the receptor, like agonist affinity, magnesium sensitivity, deactivation kinetics, modulation by polyamines and channel conductance (for review, see Dingledine et al., 1999). The Glycine-binding NR3 subunit could act as a dominant-negative subunit in the NMDA-R complex, notably by reducing calcium permeability (Nishi et al., 2001; Perez-Otano et al., 2001), but the role of this subunit remains to be clearly determined. Each type of NMDA-receptor subunit and isoform exhibit a different developmental and regional pattern of expression in the brain, resulting in an important source of functional diversity among NMDA-Rs. In Purkinje cells, let us see what kinds of NR1 isoforms and/or NR2 subunits are successively expressed throughout postnatal life and adulthood.
Purkinje cells express NR1 subunits throughout the postnatal and adult life
While NR2 subunit expression has long been controversial, there was early compelling evidence for the NR1 expression by Purkinje cells, in young as well as in mature animals. Indeed, most of in situ hybridization (Moriyoshi et al., 1991; Monyer et al., 1992; Akazawa et al., 1994; Laurie and Seeburg, 1994; Monyer et al., 1994; Watanabe et al., 1994; Nakagawa et al., 1996) and immunohistochemical studies (Petralia et al., 1994a; Garyfallou et al., 1996; Hafidi and Hillman, 1997; Thompson et al., 2000) showed that the NR1 subunit is expressed by Purkinje cell as early as E13 in the mouse (Watanabe et al., 1994) and throughout adulthood.
NR1 occurs as eight distinct splice variants that influence NMDA-Rs properties and that are regionally and developmentally regulated (Sugihara et al., 1992; Laurie and Seeburg, 1994) for review see (Zukin and Bennett, 1995). NR1 isoforms result from the alternative splicing of exon 5 (the N1 amino-terminal cassette), exon 21 (the C1 carboxy-terminal cassette), and 22 (when this C2 carboxy-terminal cassette is deleted, the C2’ cassette replaces it); see Figure 1.10. In the terminology of Hollmann et al. (1993), the eight NR1 splice variants are denominated NR1-1a, -1b, -2a, -2b, -3a, -3b, -4a and -4b. The number indicates the variant at the carboxy-terminal end (1 = no deletion; 2, 3, and 4 = deletions of C1, C2, C1+C2 respectively). In the a-subtype, N1 is absent, whereas the b-subtype displays N1 (Hollmann et al., 1993).
The N1 cassette directly controls proton inhibition (Figure 1.10A) and voltage-independent inhibition by Zn2+ ions of NMDA-Rs. Alternative splicing of the NR1 C-terminal affects the trafficking, cell surface expression, and synaptic targeting of NMDA-Rs in an activity-dependent manner (Ehlers et al., 1995; Okabe et al., 1999; Mu et al., 2003).
· All NR1 isoforms also contain a C0 cassette which, together with C1, mediates protein–protein interactions for instance with calm odulin (Ehlers et al., 1996), (Baude et al.)-actinin (Wyszynski et al., 1997), neurofilaments (Ehlers et al., 1998), and a protein YOTIAO (Lin et al., 1998) as well as signalling to the nucleus (Bradley et al., 2006) and export from the endoplasmic reticulum (ER) (Wenthold et al., 2003).
· The C1 cassette has been shown to contain an endoplasmic reticulum retention signal (Scott et al., 2001) that can be masked by the C2’ cassette (Standley et al., 2000), and/or by the association with NR2 subunit (Hawkins et al., 2004; Yang et al., 2007). Thus, NR1-1 subunits (C1/C2) alone are likely to be retained in the ER (see further).
· The C2 cassette is necessary for maintenance of dendritic spines in hippocampal pyramidal neurons (Alvarez et al., 2007). Indeed, either NR1-1 -containing (C0/C1/C2) or NR1–2 -containing (C0/C2) receptors a re able to preserve some levels of spine density.
· The C2′ cassette contains a terminal PDZ-binding motif that mediates interactions with PSD-95, PSD-93, synapse-associated protein-102 (SAP-102), and SAP-97 (for review, see O’Brien et al., 1998). mRNA splicing that regulates C2/C2′ expression is activity-regulated such that activity blockade leads to enhanced expression of C2′-containing NR1 subunits and promotes surface expression of NMDA-Rs (Mu et al., 2003). In contrast, NR1–3 -containing receptors (C0 /C1/C2′) are insufficient to maintain spine density and do not enhance the spine recovery by C2-containing receptors. Thus, although the C2′ domain is required for interactions with PSD-95 and may be important for anchoring to the PSD, it seems not to be necessary for spine maintenance.
Table of contents :
Table of illustrations
List of abbreviations
1.1.1 Anatomical architecture and main connections
1.1.2 Cellular organization of the cerebellum
1.1.3 Postnatal development of the cerebellar cortex
1.2 Cerebellar functions
1.3 Actors of the glutamatergic transmission in Purkinje cells
1.3.1 Glutamate receptors of Purkinje cells
1.3.2 Glutamate Transporters
1.3.3 Excitatory responses evoked in Purkinje cells by its glutamatergic afferents
1.4 NMDA-receptors in Purkinje cells
1.4.1 NMDA-R: general properties
1.4.2 Purkinje cells express NR1 subunits throughout the postnatal and adult life
1.4.3 NMDA-Rs of Purkinje cells in neonatal rodents (first postnatal week)
1.4.4 Second postnatal week: a gap in the expression of NMDA-Rs
1.4.5 NMDA-Rs of the adult Purkinje cells: a controversy
1.5 Problem and work hypothesis
2 Materials and methods
2.2 Slice preparation
2.4 Solutions and pharmacology
2.5 Data analysis and inclusion criteria
3.1 Publication 1: Neuroprotective effect of mifepristone involves neuron depolarization
3.1.2 Summary of the results
3.2 Publication 2: NMDA receptor contribution to the climbing fiber response in the adult mouse Purkinje cell
3.2.2 Summary of experimental results
3.3 Manuscript to submit: NMDA receptor participation to climbing fiber EPSC in adult mouse Purkinje cells lacking the delta2- glutamate receptor
3.3.1 Introduction and summary
3.3.3 Discussion in the context of the present thesis
3.4 Results in progress: NMDA receptor involvement in bidirectional plasticity of parallel fiber to Purkinje cells synapses
3.4.2 Preliminary results
4 Summary and conclusions