Neurons, including their synapses, are generally ensheathed by fine processes of astrocytes, but this glial coverage can be altered under different physiological conditions that modify neuronal activity. Changes in synaptic connectivity accompany astrocytic transformations so that an increased number of synapses are associated with reduced astrocytic coverage of postsynaptic elements, whereas synaptic numbers are reduced on reestablishment of glial coverage. A system that exemplifies activity-dependent structural synaptic plasticity in the adult brain is the hypothalamo-neurohypophysial system, and in particular, its oxytocin component. Under strong, prolonged activation (parturition, lactation, chronic dehydration), extensive portions of somatic and dendritic surfaces of magnocellular oxytocin neurons are freed of intervening astrocytic processes and become directly juxtaposed. Concurrently, they are contacted by an increased number of inhibitory and excitatory synapses. Once stimulation is over, astrocytic processes again cover oxytocinergic surfaces and synaptic numbers return to baseline levels. Such observations indicate that glial ensheathment of neurons is of consequence to neuronal function, not only directly, for example by modifying synaptic transmission, but indirectly as well, by preparing neuronal surfaces for synapse turnover.
neurons, glia, and synaptic inputs are not static elements because they undergo dynamic transformations. This can take place under normal physiological conditions and highlights the adult nervous system’s remarkable capacity to undergo restructuring to meet particular functional requirements. A change in the morphology of neurons, and especially of afferent inputs controlling their activity, can have important consequences on their respective functions. These modifications are often accompanied by remodeling of adjacent glia, which has further impact on neuronal activity by modifying the immediate extracellular microenvironment, thereby influencing synaptic and volume transmission.
The hypothalamo-neurohypophysial system (HNS), and in particular, its oxytocinergic component, is a familiar model for this kind of morphological neuronal and glial plasticity (57, 76). Oxytocin (OT) neurons accumulate in well-defined regions in the hypothalamus, the supraoptic (SON), and paraventricular (PVN) nuclei, and release OT mainly from axon terminals in the neurohypophysis to act as a neurohormone in vital functions like parturition, lactation, and osmotic regulation. Critical for this systemic secretion is the concomitant exocytotic release of the peptide from somata and dendrites of OT neurons themselves in the SON and PVN. This central release facilitates the synchronous high-frequency activity of OT neurons, which is the basis of the intermittent, bolus, systemic release of the neurohormone required for activation and upregulation of OT receptors in the uterus and mammary gland (66).
Under stimulation, the overall morphology of OT neurons is significantly modified since they progressively hypertrophy, their dendrites shorten and display few branches, whereas their axons enlarge and ramify. These changes occur together with modifications in the morphology of their accompanying glia, represented by astrocytes in the hypothalamus and astrocyte-like pituicytes in the neurohypophysis. In addition, there is a significant remodeling of their afferent inputs, which results in increased numbers of inhibitory and excitatory synapses controlling their activity. The system reverts to its prestimulated morphology when central and peripheral OT release return to low baseline levels. This plasticity has been described in detail in several recent reviews (49, 57, 76, 79, 82), and we will here limit our discussion to those features relevant to synaptic remodeling. Moreover, while such plasticity involves the whole system in all the magnocellular nuclei, we will focus on the SON, where, because of its cellular homogeneity and ready accessibility, most work has been carried out.
SYNAPSE TURNOVER IN THE MAGNOCELLULAR NUCLEI
The term “synaptic plasticity” generally connotes long-lasting functional modifications within preexisting synapses, like long-term potentiation or depression. Nevertheless, it has been obvious for at least the last two decades that synaptic transmission can be significantly modified by structural synaptic plasticity as well. The OT system is exemplary in this domain and was one of the first neuronal systems in which we obtained clear evidence of activity-dependent formation and elimination of synapses in the adult brain (86).
In the magnocellular nuclei, the most obvious manifestation of synapse turnover is a changing number of synapses originating from boutons making synaptic contacts onto more than one postsynaptic element in the same plane of section (shared synapses) (86). Visible in the magnocellular nuclei under all conditions, their incidence significantly increases during conditions of enhanced or sustained OT release (parturition, lactation, chronic dehydration, restrain stress), an increase that affects those boutons synapsing onto OT profiles (84). When we applied stereology to evaluate synaptic densities, we found that synaptic remodeling in the OT system is even more extensive than reported originally, because it affects terminals making single, as well as shared synaptic contacts (25, 29). Thus in the SON of rats that have lactated for several days, in spite of neuronal hypertrophy (9, 24, 25, 29, 78), the overall synaptic density is not significantly less than that evaluated in virgin animals (about 40 × 106 synapses/mm3 SON) (25, 29, 81). A clear increase in synaptic densities becomes obvious when they are evaluated in relation to the neuropil (25, 29).
In vivo observations indicated that synapse formation occurs within 24 h of the onset of stimulation (52, 89). From our more recent in vitro work on acute hypothalamic slices containing the SON, we have found that new synapses appear very rapidly indeed. Within 2 h after activation of the system, there is a significantly increased incidence of synaptically coupled OT neurons (43), whereas significantly increased synaptic densities can be detected even after 1 h (Trailin A and Theodosis DT, unpublished observations). These findings thus add further evidence to the contention that mature neurons can form synapses reliably and quickly (see also, Refs. 27 and 58). There is synapse elimination in the OT system as well since the number of synapses return to unstimulated values on arrest of stimulation (52, 83), a phenomenon that also takes place within hours (43).
REMODELING OF ASTROCYTES ACCOMPANIES SYNAPSE TURNOVER
In the SON, the majority of glia are astrocytes, and like most other central astrocytes, they are quite heterogeneous, both morphologically (4, 6, 36) and physiologically (36). A large number have a radial morphology, with cell bodies lined up along the base of the nucleus and thick, long, processes oriented ventrodorsally from which arise finer processes. They express both glial fibrillary acidic protein (GFAP) and vimentin (6), are not electrotonically coupled, and possess GABA-A receptors (36). Interestingly, although they have the glutamate transporters, GLT-1 and GLAST (Piet R and Theodosis DT, unpublished observations), they do not appear to respond to glutamate (36). The other major population is stellate (protoplasmic), and they also emit thin processes (6). They are coupled by prominent gap junctions, contain GFAP, and possess glutamate transporters. OT neurons often occur in tightly packed clusters, but they remain separated by fine processes from both types of astrocytes. This kind of astrocytic coverage is widely seen in adult neuronal tissues (20, 30, 32, 93). It characterizes the OT and vasopressin (VP) systems under basal conditions of neurosecretion. In contrast, under stimulated conditions, when the neurons are contacted by an increased number of synapses, glial coverage is significantly reduced, and the surfaces of OT somata and dendrites are directly and extensively juxtaposed. When OT secretion returns to baseline levels (for example, after weaning the young), astrocytic processes reappear between neuronal profiles. Ultrastructural analysis of immunoidentified profiles established that these glial changes, like the synaptic changes, are specific to the OT system (84). While VP neurons display some juxtapositions, their incidence and extent are low and, more importantly, do not vary with changing conditions of neurosecretion (12, 78, 85).
As noted earlier, magnocellular neurons hypertrophy progressively with stimulation, and reduced glial ensheathment of oxytocinergic elements may merely represent failure of astrocytes to establish their initial coverage. However, it is difficult to see how cell hypertrophy can explain reduction of astrocytic coverage of dendrites, which often appear as directly juxtaposed bundles (of 3 to 8 dendrites). Moreover, our recent in vitro experiments (43) revealed that these glial transformations occur very rapidly, within 2 h, when the neurons have not significantly hypertrophied. Finally, as noted earlier, VP neurons do not display increased juxtapositions even under strong, persistent stimuli like lactation or chronic dehydration, yet they significantly hypertrophy under these conditions (12, 24, 29). A more likely explanation is that neuronal juxtapositions are the result of active retraction and elongation of astrocytic processes over neuronal surfaces whose morphology is constantly changing. Application of live imaging techniques to the SON should illustrate this more clearly. In the mouse brain stem, for example, a high degree of motility of astrocytic processes over neurons and synapses was visualized even under normal conditions (32). Application of such techniques to the SON should also make it possible to visualize the synapse turnover that is taking place concurrently on OT surfaces free of astrocytic processes.
NEUROCHEMICAL IDENTITY OF SYNAPSES UNDERGOING REMODELING
Like most other central neurons, OT and VP cells receive an abundant GABAergic input (19, 25, 29, 38, 81). Our ultrastructural analyses have estimated that under basal conditions, about 35% of all synapses in the SON are GABAergic, with twice as many on dendrites than on somata (25, 29). Moreover, they have clearly demonstrated that GABAergic inputs undergo remodeling so that in the SON of lactating rats, axosomatic and axodendritic GABAergic synapses represent close to 50% of all synapses. By analyzing ultrathin sections in which both pre- and postsynaptic partners were identified immunocytochemically, we confirmed that the increased inputs were on OT neurons (25, 29, 81). This synaptic remodeling can be reproduced in vitro in acute hypothalamic slices, a preparation that permits more direct pharmacological manipulations and therefore, a finer analysis of the modalities of such plasticity. These ongoing studies, using electron microscopy and patch clamp electrophysiology, have already made it clear that synapse formation is very rapid, occurring within a couple of hours and results in the formation of inhibitory synapses that are functional (78a). In vivo observations indicated that the origin of GABAergic inputs to SON neurons is local, from interneurons in close proximity to the magnocellular nuclei (38, 67, 81), as well as from long distance GABA afferents (37). In slices, synaptic densities of putative GABAergic synapses are closely similar to those evaluated in the SON in vivo, which strongly suggests that most GABA afferents to the SON derive from a local interneuron network.
The other major afferent input to OT and VP neurons uses glutamate as the neurotransmitter (24, 25, 38, 47, 90). Under basal conditions, glutamatergic synapses represent about 22% of all synapses in the rat SON; they are twice as numerous on dendrites than on somata (24, 25). Glutamate inputs undergo morphological remodeling under the same conditions that induce changes in GABAergic inputs, so that at lactation, they represent close to 25% of all synapses in the nucleus. The increase is represented only by glutamatergic synapses impinging on OT somata and dendrites (24, 25). Electrophysiological (7, 39) and morphological (14, 38) observations have indicated that many of these glutamatergic afferents derive from local interneurons in adjacent hypothalamic areas.
Finally, an excitatory, noradrenergic input, which represents ∼10% of all synapses on magnocellular neurons (48) undergoes activity-dependent remodeling, as well. Thus in virgin rats, noradrenalin afferents innervate equally OT and VP neurons, whereas in lactating rats, there is a significant increase in the noradrenergic (NA) innervation of OT neurons (48). These afferents have an extrahypothalamic source, namely neurons in the A1 and A2 areas of the ventrolateral medulla and nucleus tractus solitarii, respectively (15).
MECHANISMS OF SYNAPSE TURNOVER
In comparison with developing or regenerating systems, one would expect bouton sprouting and degeneration in neuronal centers where synaptic numbers constantly vary. However, we have no clear evidence of growth cones or degenerating boutons in the magnocellular nuclei. It is possible then that new synapses are formed by enlargement and splitting of active zones in already existing boutons. Postsynaptic densities would follow suit if only because there is a tight association of pre- and postsynaptic membranes, mediated by molecules that span or are anchored to the synaptic cleft (94). Shared synapses may be extensions of preexistent terminals, which induced de novo postsynaptic density formation in a neighboring neuronal element. This scenario can be reversed, especially in view of recent studies in the developing Drosophila neuromuscular junction showing formation of postsynaptic densities before the appearance of presynaptic specializations (64). Adhesion between pre- and postsynaptic sites would help to coordinate growth of the synaptic contact. Further maturation of synapses would involve acquisition of the full complement of pre- and postsynaptic components required for stable synaptic transmission, processes that depend on gene regulation and new protein synthesis. This occurs in the SON, because in our in vitro model, synaptic remodeling can be inhibited if the incubation medium includes the protein synthesis inhibitor anisomycin (Langle SL, Trailin A, and Theodosis DT, unpublished observations).
It is obvious that intervening astrocytic processes must participate in the regulation of synapse formation, stability, and elimination. Thus they cannot be present between pre- and postsynaptic partners if synapse formation is to occur while their reinsertion could contribute to undoing of synaptic contacts (see also Refs. 1, 28, and 32). After synapse elimination, pre- and postsynaptic protein components could simply be redistributed in their respective compartments, without involving degeneration of boutons. These possibilities await experimental evidence. What is certain is that synapse formation in the SON, as elsewhere in the central nervous system, must be accompanied by a coordinated development of pre- and postsynaptic molecular and structural specializations, which requires exchange of information from pre- and postsynaptic partners, as well as adjacent astrocytes. Astrocytic membranes are equipped with numerous signaling or cell recognition molecules [transmitter receptors, transporters, adhesion molecules (93)], so that astrocytes are well suited for all of these purposes.
Finally, one should also consider that synaptic remodeling includes a phase of stabilization for which neuronal activity appears critical. Whereas the establishment of synaptic contacts per se does not depend on the secretion of a particular neurotransmitter (91), the specification of synaptic inputs requires neuronal activity (10). It is now clear that membrane depolarization, elevation of intracellular calcium, and action potential generation can lead to posttranslational modifications of synaptic proteins and regulation of gene activity (59). However, the molecular links between these events and synaptic plasticity are not precisely defined. The most likely possibility is that correlated synaptic activity provides critical information about the appropriateness of synapse connections, influencing synapse stability, and elimination (40). This criterion is certainly met by OT neurons, which fire in a highly coordinated manner during parturition and lactation. Nevertheless, what is unclear is how such neuronal activity is translated into long-term structural changes.
MOLECULAR FACTORS PERMISSIVE FOR SYNAPTIC PLASTICITY
In view of the evidence presented above, there can be no doubt that the morphology of the OT system, including its synaptic inputs, is repeatedly modified throughout life. This kind of plasticity connotes dynamic cell-cell and cell-matrix interactions that will bring into play cell surface molecules, extracellular matrix, and cytoskeletal proteins, as well as soluble trophic factors, many of which have been identified in developing systems undergoing synaptogenesis and glial transformations (23, 44). It is not too surprising then that the adult HNS continues to express many of these juvenile molecules. The list includes members of the immunoglobulin superfamily of cell adhesion molecules, such as the neural cell adhesion molecule (NCAM), as well as complex extracellular matrix glycoproteins and several cytoskeletal proteins (76, 82). We will here focus our attention on NCAM whose expression is particularly striking in all portions of the HNS and which we are now sure participates in HNS plasticity.
NCAM, which occurs in several isoforms differing in their proteic and carbohydrate moieties, is thought to intervene in most cell interactions via modulation of cell adhesivity and intracellular signaling (see, for example, Ref. 23). Regulation of NCAM-mediated interactions can be achieved by transcriptional control of the type and amount of NCAM expressed on the cell surface. Further regulatory possibilities are provided by the posttranslational addition of the unique carbohydrate polymer, α-2–8-linked polysialic acid (PSA), on its extracellular domain (68). The resulting high Mr isoforms (about 220 kDa), known as PSA-NCAM, contain more than 30% PSA. PSA-NCAM is abundant in embryonic tissues, whereas most adult tissues contain NCAM with little PSA. Nevertheless, PSA-NCAM continues to be expressed in the adult HNS (5, 41, 55, 87), as it does in other systems endowed with the capacity for morphological and/or physiological plasticity (5, 71).
Neuronal expression of PSA-NCAM is well documented (68) but in the HNS, it is particularly abundant on astrocytic surfaces, especially in the SON and PVN (77). It is also abundant on the surface of all neurosecretory axons and terminals (41, 50, 62, 77). PSA-NCAM is constitutively expressed in HNS neurons and glia (62), which explains why its expression is not markedly affected by changing conditions of neurosecretion (5, 77, 87). Nevertheless, this does not mean that its expression is of no consequence to plasticity. On the contrary, it is a prerequisite for all its dynamic phases. Thus specific enzymatic removal of PSA from NCAM in the SON in situ inhibited the neuronal, glial, and synaptic remodeling associated with lactation and chronic dehydration; there was no effect if such pertubation was performed once remodeling of the nuclei had taken place (77). Likewise, PSA removal from cell surfaces in the neurohypophysis prevented stimulation-related induction and reversal of axon and glial changes but had no effect once remodeling had occurred (50). We do not know how PSA intervenes to permit morphological changes in the HNS, but a possible mechanism is that large quantities of PSA on cell surfaces attenuates adhesion via physical impedance or charge repulsion (69). Cells could then detach from their neighbors or from the extracellular matrix and be able to undergo changes in their conformation. In addition, PSA-NCAM may intervene by transducing intracellular signals resulting in synapse formation (21).
In the HNS, there is concomitant expression of less sialylated isoforms of NCAM, like the trans-membrane NCAM-140 and NCAM-180 (5, 87). Their role in HNS function and plasticity remains unknown. Since they are generally associated with stability of cell contacts (23), they could intervene to maintain the anatomy of the system once it has undergone remodeling. In view of its action in the hippocampus (70), NCAM-180 may also participate directly in the synaptic plasticity of the OT system, via complex intracellular signaling mechanisms.
We consider a molecule like NCAM permissive for plasticity since its expression is not tightly linked to the activity of HNS neurons and yet, as seen from our enzyme pertubation experiments (50, 77), its expression appears obligatory for plasticity. Nonetheless, other molecular mechanisms can substitute for NCAM, because remodeling does occur in lactating and salt-loaded mice genetically deficient in NCAM (88). Similar phenomena have been described in other neuronal systems in which NCAM is highly expressed and which continue to undergo plasticity in NCAM −/− mutants (22). In a sense, the specific identity of molecules permissive for remodeling may not be important, provided that they activate the same intracellular mechanism. On the other hand, the response of HNS cells would be invariant and independent of the identity of the factor, provided the proper stimulus intervenes. As will be discussed next, one such stimulus is OT itself. In other adult neuronal systems capable of morphological plasticity and expressing PSA-NCAM (5, 21, 33, 54, 72), other inductive factors must intervene.
OT INDUCES MORPHOLOGICAL PLASTICITY
If molecules like PSA-NCAM act as priming molecules to provide a permissive environment for synaptic remodeling, then there must be molecules whose expression is particular to the HNS that act as specific stimuli to induce structural transformations, triggering cascades of intracellular events that result in neuron-glial transformations and synapse formation. Such molecules include those that would be expressed at the onset of enhanced activity to bring on the neuronal and glial rearrangements as well as molecules, not necessarily identical to the former, which signal the system to revert to its original condition.
In the OT system, the physiological conditions during which remodeling is most striking are parturition and lactation. As noted earlier, these conditions are characterized not only by more OT released from the neurohypophysis but in increased secretion of the peptide within the hypothalamic nuclei themselves. Intranuclear OT is due primarily to somatodendritic exocytotic release, which can be brought on by action potential firing and which is modulated by calcium release from internal stores (42). Central OT facilitates the activity of the OT system by various means, including enhancing activity of excitatory glutamate inputs (34, 39) and inhibiting, via a presynaptic action, GABA release (18). In fine, it facilitates firing of its own neurons and therefore, neurohypophyseal release into the peripheral circulation. It is therefore not too surprising that intracerebroventricular infusion of the peptide, presumably mimicking this central release, induced neuronal-glial and synaptic changes in the SON similar to those detected under physiological stimulation (80). This particular action of central OT is receptor mediated because it can be mimicked by a close analog, 4-Th-OT, whereas peptides like VP or cholecystokinin have no effect. Nevertheless, OT does not act alone but in synergy with sexual steroids (51).
Ongoing experiments using acute hypothalamic slices show that this effect of OT can be reproduced in the SON in vitro. As in vivo, an OT receptor mechanism appears responsible, because synaptic and glial morphological changes were mimicked by application of the specific OT agonist 4-Thr-7-Gly-OT and were inhibited when OT was applied together with the specific OT receptor antagonist, dOVT (43). Experiments with our in vitro preparation also highlight the participation of sexual steroids since inclusion of 17-β estradiol potentiated the effects of OT. This is in accordance with numerous observations showing that estrogens enhance OT secretion by modulating the electrical activity of OT neurons (35), increasing OT gene expression (13) and central and peripheral release (95, 96). However, some caution is necessary in interpreting our results, because reports of gonadal steroid effects on magnocellular neurons are conflicting (74). Indeed, the site of action of the steroid in OT neurons has not been identified, and we still do not know whether its effect is direct or indirect. Long-term genomic effects are mediated by nuclear estrogen receptors, ERα or ERβ (46, 73). Rat OT neurons lack the former (73, 74) and <10% express the latter (75). In view of the rapidity of the effects of the steroid on the anatomy of the SON (43), it is more probable that estrogens act via a plasmalemmal mechanism (65), mobilizing intracellular calcium and inducing signaling cascades that result in activation of gene expression and ultimately synapse formation. Such a mechanism is not that unlikely when one considers the extremely rapid effects of the steroid on the electrical activity of OT neurons resulting in enhanced OT secretion (35). It is noteworthy that estrogens affect OT gene regulation if treatment is accompanied by a progesterone withdrawal protocol (2), endocrine conditions similar to those occurring at the end of gestation (8).
Although OT, with or without steroids, appears essential for remodeling in the magnocellular nuclei, our experiments fail to tell us the level at which it acts or whether it acts indirectly or directly, via its receptor on OT neurons (3) and/or associated astrocytes (31). There are other candidates for such signaling, and OT may act in a secondary fashion by facilitating their expression.
OTHER SIGNAL CANDIDATES FOR PLASTICITY
From our recent work on acute hypothalamic slices it has become obvious that OT can be replaced by other molecules to induce morphological changes. Thus exposure to a combination of metabotropic and ionotropic glutamate receptor antagonists prevented OT/estrogen-mediated increases in the proportion of juxtaposed and synaptically coupled profiles (43). Likewise, such a treatment inhibits the appearance of new GABA synapses (Trailin A and Theodosis DT, unpublished observations). This strongly suggests that glutamate itself may be a more immediate signaling agent than OT. The excitatory amino acid is an excellent candidate in this respect, because it can serve as a bidirectional signal to transfer information between activated neurons and adjacent astrocytes. Glutamate is taken up by astrocytes once released from synapses and can be released by astrocytes in response to physiological increases in intracellular Ca2+ (93). Even small increases in intracellular Ca2+ in astrocytes are sufficient to induce large glutamate-activated currents in neighboring neurons (61). Such a mechanism, alone or in combination with synaptically released glutamate, may serve both to activate the electrical activity of OT neurons and to signal initiation of morphological transformations. In line with this reasoning are data showing that OT cannot induce morphological changes in slices if it is used together with the chelator BAPTA-AM (43).
GABA, on the other hand, appears to inhibit synaptic and astrocytic transformations in the SON. Our preliminary observations indicate that application of the GABA-A receptor antagonist bicuculline to slices, even in the presence of the specific OT receptor antagonist dOVT, can reproduce the effects of OT and results in increases of juxtaposed and synaptically coupled neurons similar to those obtained with OT alone (43). Either GABA acts directly or the bicuculline effect reveals the intervention of other signaling or inducing agents.
As seen from the observations presented in this review, both inhibitory and excitatory synaptic inputs on OT neurons increase during enhanced neurosecretion. This could reflect a compensatory mechanism of the cells to adjust to their hypertrophy induced by sustained stimulation, a mechanism which would provide a degree of regulation equivalent to that in unstimulated animals. Data showing that the number of synaptic contacts per unit length of postsynaptic membrane in the SON of lactating rats remains stable, as does the size of postsynaptic densities (24, 25), would support such a premise. In addition, electrophysiological data obtained in lactating rats indicated that in spite of increased numbers of GABA release sites per hypertrophied OT neuron, synaptic current densities were no different from those recorded in virgin animals (9).
Nevertheless, an increased number of synapses may play an additional, more direct role in facilitating the particular electrical activity that characterizes OT neurons at parturition and lactation (63). That synaptic remodeling involves only those afferents impinging on OT neurons and that it occurs at every parturition and lactation is in favor of our contention that this kind of synaptic change directly influences such activity. A mechanism for increased inhibition could be through synchronization of GABA release from several terminals arising from one single inhibitory neuron and converging on several cells. This strong inhibition may be necessary to prevent the neurons from being stimulated by factors other than those relevant to lactation and would be a means by which the response of the OT system to stimuli other than suckling is significantly attenuated (26, 45). This kind of filtering action of GABA may also contribute to maintain OT neurons at a level of depolarization that would facilitate their synchronous activation just before milk ejection (53, 92).
On the other hand, an enhanced glutamatergic input probably serves to induce and further potentiate this activity. It is now obvious from many observations that glutamate plays a key role in the control of OT neuron firing. For instance, intracellular recordings from OT neurons in organotypic slice cultures showed that each burst of high-frequency action potentials is the result of a volley of postsynaptic potentials arising from glutamatergic synapses (39) originating from interneurons in adjacent hypothalamic areas (38). Moreover, simultaneous recordings from two OT cells revealed a high degree of synchronization due to synchronization of afferent input originating from these glutamate neurons (34). These in vitro observations are supported by in vivo data showing that the activity of OT neurons is under strong control by glutamate circuits from glutamate neurons within the hypothalamus but outside the SON and PVN (7, 16). What is still not clear are the possible effects that would be exerted by the additional glutamate input provided by newly formed glutamate synapses (24, 25).
To the potential facilitation of excitatory drive from newly formed glutamatergic synapses, one must include one derived from the additional noradrenergic synapses that appear at lactation (48). Noradrenergic afferents provide an excitatory input to OT neurons (17). In addition, there appears to be a central synergistic action of noradrenaline with glutamate (16, 60), an action which would further facilitate excitation of OT firing.
Nevertheless, some caution is necessary when attributing too strong a participation from additional synapses in the control of bursting activity of OT neurons at lactation. As noted earlier, enzymatic removal of PSA in all magnocellular nuclei at late gestation inhibits synaptic and glial remodeling (77), yet parturition and lactation, bursting firing, and reflex milk ejections occur normally (11). One possible explanation is that OT bursting activity, resulting from the activity of the intrahypothalamic burst generator, of which OT and glutamate neurons are a part (34, 39), is already controlled by strong high-frequency discharges and will appear unaffected by any additional glutamate and noradrenalin inputs. Within the pulse generator, glutamate neurons driving OT cells display an intense activity at milk ejection, which would bypass any subtle modulation of synaptic transmission (56), including that provided by a few additional excitatory synapses. It is also possible, however, that the additional synapses, whether inhibitory or excitatory, derive from pools of neurons not involved directly in the control of the pulsatile, coordinated firing of OT neurons but which intervene in fine tuning of their activity outside periods of high-frequency firing.
This work was supported, in part, by short-term fellowships from Institut National de la Santé et de la Recherche Médicale (Poste Vert) and the Ministry of Research (to A. Trailin).
- Copyright © 2006 the American Physiological Society