Neuronal Growth and Differentiation Factors and Synaptic Plasticity

Paul H. Patterson

P. H. Patterson: Biology Division, California Institute of Technology, Pasadena, California 91125.

There is a growing realization that mechanisms and molecules that regulate the development of circuits in the embryonic nervous system can also influence the flow of synaptic information in maturity. This is done both by modifying the efficacy of transmission at established connections and by regulating the rearrangement of such connections. It is now well known that neuronal survival and growth in the embryo is controlled in part by proteins and steroids that act as trophic (Greek for nourishment) factors. These factors can govern the number of neurons that innervate a target cell in a retrograde fashion, through secretion of the factors by the target cell. Trophic factors can also act anterogradely, allowing a neuron to influence its targets, as well as more globally, either through the circulation or via glial or immune cells. In addition to controlling life, death, and growth, these and other families of factors can act instructively, to direct neurons to adopt one or another phenotype. For example, a noradrenergic neuron can be converted to the cholinergic phenotype by such a differentiation factor. In fact, this switch in phenotype is known to occur during the normal development of a subpopulation of sympathetic neurons when they contact their particular target cell type, the sweat gland. Dramatic changes in transmitter and associated neuropeptide systems like these are due to the ability of instructive neuronal differentiation factors to direct gene expression (see Molecular and Cellular Mechanisms of Brain Develoopment, The Development of Brain and Behavior, and The Neurobiology of Treatment-Resistant Mood Disorders).

It is less well appreciated that these phenomena of plasticity in growth and gene expression persist into maturity. There is evidence implicating growth and differentiation factors in synaptic plasticity and particular behaviors. The implications of this plasticity are threefold: (i) The adult system is metastable, representing a balance between growth and withdrawal, gene induction, and repression; (ii) neuronal trophic and differentiation factors are required for maintenance of the mature system; and (iii) these proteins and hormones could therefore be used for therapeutic intervention. Thus, growth and differentiation factors are potent and highly specific tools, not only for the rescue of dying neurons, but also for changing the balance of transmitter and neuropeptide systems, thereby modifying behavior (see The Development of Brain and Behavior).

The classic trophic factor is, of course, nerve growth factor (NGF). Its best known activity is the rescue of particular sets of embryonic neurons from death. In many parts of the nervous system, 50% of the neurons produced during embryogenesis die during subsequent development, and the target tissues that neurons innervate can play a major role in controlling how many neurons survive (125). This survival effect of targets is mediated in part by trophic factors such as NGF. The family of NGF-related survival factors is now composed of at least four members [the neurotrophins: NGF, brain-derived neurotrophic factor (BDNF), neurotrophins 3 and 4/5 (NT-3, NT-4/5); see Table 1 (135). The neurotrophins are not only required for survival, but they enhance growth of neuronal cell bodies and processes in a dose-dependent manner.

The neurotrophins act on partially overlapping populations of neurons in both the central nervous system (CNS) and peripheral nervous system (PNS) (61). Responsivity to the various family members is determined by which neurotrophin receptor is expressed on a neuron's surface. The high-affinity receptors for this family belong to the trk family of transmembrane protein kinases (21). The basal forebrain cholinergic neurons, well known because of their potential role in learning and memory and their loss in Alzheimer's disease, are the most intensively characterized neurotrophin-sensitive cells in the CNS. Of particular clinical interest are studies demonstrating that administration of NGF to rats with lesioned CNS cholinergic neurons enhances the survival of these neurons as well as performance in spatial memory tasks (29). Similar results were obtained with unlesioned but impaired, aged rats (38). Moreover, BDNF mRNA is decreased in the hippocampus of Alzheimer brains, relative to other mRNAs assayed (96). BDNF is also implicated in motor neuron disease and its potential treatment. This protein prevents the death of motor neurons after nerve section (60, 121, 150), and its mRNA is expressed at the appropriate stages in the embryonic spinal cord and in the limb bud for it to act on spinal neurons during development (48). NT-3 and NT-4/5 also share this expression pattern and can rescue motoneurons in culture (48). In addition, two proteins belonging to a different cytokine family, the neuropoietic factors cholinergic differentiation factor/leukemia inhibitory factor (CDF/LIF) and ciliary neurotrophic factor (CNTF) (discussed below), also promote motor neuron survival (73, 88). Moreover, CNTF prevents the degeneration of motor neurons in a mutant mouse model of progressive motor neuronopathy (122). For these reasons, CNTF is currently being tested in clinical trials with amyotrophic lateral sclerosis (ALS) patients.

Other trophic factors active on CNS neurons include insulin and insulin-like growth factors [IGFs (56, 65); also being tested on ALS patients], fibroblast growth factors [FGFs (51, 139); FGF-5 is also a good candidate as the motor neuron trophic factor (51)], platelet-derived growth factors (23, 52 ), and members of the transforming growth factor superfamily [TGFs (78, 98)]. Of particular interest is the recent discovery of a trophic factor for midbrain dopamine neurons, called glial-cell-line-derived neurotrophic factor [GDNF (68)]. A member of the TGFb superfamily, GDNF is more selective in its action on midbrain cultures than other factors that enhance survival of dopaminergic cells such as IGF-I and II, epidermal growth factor (EGF), aFGF and bFGF, and BDNF. It will be of interest to see the results of tests of the utility of GDNF in Parkinson's disease models.

In addition to neuronal survival and growth, target tissues can also control the phenotype of the neurons that innervate them. Phenotypic traits regulated qualitatively by targets in vivo include (a) the neuron's transmitter and neuropeptide profile (62) and (b) the type of synapses the neuron receives on its dendrites. That is, targets encountered by a neuron's axons in the periphery can influence the connections made on that neuron's dendrites in the CNS (reviewed in ref. 39). A classic example of the presumptive ability of a postsynaptic cell to control the phenotype of its presynaptic input is the type of synapses formed by the various branches of single auditory nerve fibers. When these axons contact neurons in one region of the cochlear nucleus, they form the very large end bulbs of Held; when collateral branches of the same axons encounter neurons in other regions of the nucleus, they form small boutons (102). Qualitative control of this order can be regulated by the neuronal differentiation factors, and such effects can be distinguished from the classical NGF survival and growth activities. Differentiation factors characteristically alter neuronal gene expression and phenotype without changing neuronal survival or growth (90).

The most intensively studied family of instructive differentiation factors is termed the neuropoietic cytokines, named for their effects on both the nervous and hematopoietic systems. Unlike the neurotrophins, this group of proteins does not share an extensive degree of amino acid sequence identity. The neuropoietic family members are linked by (a) the many biological activities they have in common (36, 59, 93), (b) the protein structure they are predicted to share in common with growth hormone (14, 111), and (c) their promiscuous use of common receptor subunits (28, 41).

Five of the neuropoietic cytokines [CDF/LIF, CNTF, OSM (oncostatin M), GPA (growth-promoting activity), and SGF (sweat gland factor; not yet cloned)] evoke nearly identical changes in gene expression when added to cultured sympathetic neurons, while the somewhat more distantly related family members, interleukins 6 and 11 (IL-6 and IL-11), evoke a subset of these changes (36, 104, 116). There is a striking overlap in the activities of CDF/LIF, IL-6, and IL-11 on other types of cells such as hepatic and myeloid cells, however (59). The molecular basis for this apparent functional redundancy is the sharing of common subunits in the receptor complexes for these cytokines (28, 41). Not only can some of these cytokines displace others at high-affinity ligand binding sites, but the receptor complexes can all employ the same transducing subunit (129). Presumably, a shared signal transducer ensures an identical set of effects on gene expression in the target cells. While there are a number of inconsistencies in this current picture (41), the sharing of receptor subunits is a major chapter in the story of this family.

Initially surprising, redundancy in ligand activity and promiscuity in the use of receptor subunits has also become a theme of cytokine action in the hematopoietic system (85). Moreover, it occurs in the neurotrophin family (21). One possibility is that other, more selective and nonoverlapping receptors may be discovered in the future. This would allow the members of the cytokine family to evoke unique as well as overlapping effects, leading to interesting combinatorial possibilities. Another parallel between the generation of cell diversity in the hematopoietic and nervous systems lies in the pyramidal structures of their lineages. It appears that multipotential stem cells generate progenitors committed to particular pathways, each of which may yield multiple phenotypes (8). Moreover, proliferation and differentiation at each of these steps can be influenced by cytokines that are shared between the hematopoietic and nervous systems (82, 93).

Not all intercellular signals that direct neuronal gene expression belong to the neuropoietic family. Activin A, a member of the TGF superfamily, induces a different, but partially overlapping, set of genes when compared to the neuropoietic cytokines in the cultured sympathetic neurons (36). Activin A can also mimic a target-derived factor that induces expression of the neuropeptide somatostatin (SOM) in cultured ciliary ganglion neurons (25). Moreover, activin A mRNA is found in cells cultured from this target (the choroid; see ref. 24).

What is the role of these differentiation factors in the nervous system? In the adult, evidence is emerging that these proteins can serve interesting functions in the response to injury (see below). In development, attention has focused on a possible role for the neuropoietic cytokines in the switch in phenotype that sympathetic neurons undergo when they contact the sweat glands. Landis (62) demonstrated that noradrenergic sympathetic neurons switch their transmitter phenotype to cholinergic when their axons contact sweat glands, even if this target tissue is placed in ectopic locations. The change in gene expression that occurs in this switch is very similar to that evoked in cultured sympathetic neurons when either CDF/LIF, CNTF, GPA, or OSM are added. Studies on the cholinergic differentiation factor extracted from rat sweat glands suggest that it is a unique protein, resembling CNTF (103, 110). There is, in fact, good evidence that CDF/LIF can convert noradrenergic sympathetic neurons to the cholinergic phenotype in vivo. A transgenic mouse line was created in which an insulin transcriptional promoter was used to ectopically express CDF/LIF in pancreatic islet cells. The result is an induction of cholinergic properties in the sympathetic innervation of the pancreas (R. Palmiter, unpublished data). Recent experiments have further demonstrated that labeled CDF/LIF can be taken up by the endings of sympathetic neurons and retrogradely transported back to the neuronal cell body (49).

It is now clear that the same protein can act instructively as a differentiation factor or permissively as a growth factor, depending on the responsive cell population. CDF/LIF, for instance, can act as a survival factor, enhancing the growth of embryonic sensory and motor neurons (44, 73, 80). CNTF and GPA can act as trophic factors for ciliary neurons, and GPA is expressed in chick eye during the period of naturally occurring cell death for ciliary neurons that innervate eye muscles (66). CNTF can also prevent axotomy-induced death in the CNS (22). By the same token, the trophic factor BDNF can selectively induce the expression of the neuropeptides SOM and neuropeptide Y (NPY) in cultures of rat cortical neurons, without affecting neuronal survival in this population (83). Similarly, NGF can selectively induce the expression of particular neuropeptides in sensory neurons (70).

The discussion to this point has emphasized the role of target tissues in the control of neuronal survival and gene expression. It is clear that anterograde effects, from neurons to their targets, can be another major mechanism in development and in maturity. Anterograde influences can be mediated by small molecules or by proteins. In the former class are the neurotransmitters and neuropeptides released by the presynaptic terminal. These signaling agents evoke changes in ion fluxes and intracellular messengers in the postsynaptic cell that have long been known to up-regulate transmitter biosynthesis in many types of neurons. This transsynaptic effect is an effective way for the stimulated postsynaptic neuron to replenish its transmitter stores that are transiently depleted by its newly elevated rate of activity and consequent release rates. When stimulation is prolonged, transsynaptic induction of mRNAs for the transmitter biosynthetic enzymes and neuropeptide precursors is elicited, thereby chronically elevating transmitter and neuropeptide production (10, 42, 153). Evoked activity can also influence the choice of which transmitter is to be produced. For instance, depolarization of cultured sympathetic and sensory neurons blocks their responses to certain differentiation factors (CDF/LIF but not CNTF; see ref. 105). There are also numerous examples of regulation of neuropeptide as well as neuronal trophic factor expression by activity (see refs. 16, 32, and 55). It is also important to note that there is evidence that neurotransmitters themselves can regulate cell proliferation and differentiation (63).

More novel is the notion that proteins may also act as anterograde trophic or differentiation signals. While not known to be a neuronal differentiation or trophic factor, the protein agrin is a good example of signal that can act in an anterograde fashion. Motor neurons in the spinal cord can produce agrin, transport it anterogradely to their synaptic endings, and release it (108, 113). Agrin then binds to the muscle cell and evokes clustering of several different muscle surface proteins under the synaptic endings (114).

Neurons also anterogradely transport known cytokines. In addition to acting as a trophic factor for many types of neurons in culture, bFGF has been shown to be anterogradely transported by retinal ganglion neurons to their target sites in the lateral geniculate body and the superior colliculus (37). FGF is also known to be synthesized and released by retinal cells in vivo (45). There is also evidence that the neurotransmitter vesicles of adrenal chromaffin cells contain bFGF (97, 144). Moreover, preliminary reports indicate that neurotrophic activity is released from these cells when they are depolarized (138). Chromaffin cells also contain several TGF-bs and a CNTF-like trophic factor (138, 140). Because chromaffin cells resemble neurons in many respects, and neurons themselves produce neurotrophic factors (see refs. 34, 119 and 145), it seems highly likely that these factors are used in the nervous system in both the antero- and retrograde directions. There is ample evidence that anterograde influences can regulate neuronal survival and gene expression during development (42, 87).

A key feature of cytokines acting in retrograde and anterograde pathways is that these transsynaptic actions make use of the exceedingly complex circuitry that underlies nervous system function. This is important for two reasons. First, because the circuitry is designed for discrete, cell-to-cell interactions, cytokines can regulate neuronal survival and gene expression with the same degree of precision that is inherent in the wiring. This allows for unique differentiation decisions by individual neurons within layers or large groups of cells. Such small, minority populations are, in fact, a common feature of many parts of the nervous system. Second, using circuitry to control gene expression will help ensure that the phenotypes of neurons linked in a given pathway are functionally appropriate. Postsynaptic receptors must match transmitters released presynaptically, for example. If the genes for these proteins are regulated in part by interactions between the synaptic partners, phenotype need not be completely preprogrammed earlier in development. Moreover, phenotypic decisions can be reversed at very late stages, because axons encounter distinct synaptic partners in postnatal life (62) (see Interactions Between the Nervous System and the Immune System: Implications for Psychopharmacology, and The Treatment of Tardive Dyskinesias).

Growth and differentiation factors are known to act in the normal, undamaged, adult nervous system, and they play a role in the response to injury as well. There is considerable evidence in the older literature that the adult system, while seemingly very stable, is actually in a state of dynamic equilibrium (99). For example, it has been known for many decades that denervation of skeletal muscle causes a series of changes in the myotubes that can be viewed as a return to an embryonic, preinnervation state. When nerves subsequently reinnervate the myotubes, a sequence of changes that were seen in development unfolds once again, producing a mature muscle. Thus, the nerve controls the state of differentiation of the muscle. Much of the influence of the nerve is mediated through synaptic transmission; that is, it is based on activity. Indeed, many of the changes observed in denervation can be prevented by electrically stimulating the muscle after cutting the nerve.

Similar phenomena are observed in the nerve fibers when imbalances are introduced experimentally. If only part of the muscle is denervated, the remaining, intact axons are observed to sprout and innervate the denervated myotubes. An activity-based mechanism can be invoked here as well, because sprouting can also be induced by blocking nerve–muscle transmission or paralyzing the muscle, and electrical stimulation of the muscle prevents sprouting induced by partial denervation. These and other experiments illustrate the state of dynamic balance that the mature nerve–muscle system represents. Similar phenomena have been repeatedly observed in neuron–neuron synapses. In fact, contemporary imaging methods are sufficiently sensitive that they have demonstrated that even in intact, undisturbed neuron–neuron and neuron–muscle synapses, small axonal sprouts are constantly arising, parts of postsynaptic gutters are being vacated, and new contacts are being formed (100, 146). Moreover, sprouting from intact synapses can be evoked by administration of trophic factors such as IGF-2 and CNTF (18, 44).

These and other observations suggest that neurotrophic and differentiation factors may participate in the regulation of synaptic circuitry in the intact adult system. Consistent with this hypothesis is the evidence that neuronal activity can control the transcription of the genes for these factors. For example, the balance between the activity of the glutaminergic and g-aminobutyric acid systems can regulate the level of BDNF and NGF mRNA in the adult rat hippocampus (151). In the visual system, physiological variations in sensory stimulation can elicit dramatic changes in neurotrophin expression. One hour of exposure to light after a period in the dark can nearly double the levels of BDNF mRNA in adult rat visual cortex (20). Because light-evoked activity can control the growth and sprouting of axons in the visual cortex, it was also of interest to test the effects of neurotrophin administration on this phenomenon. In fact, NGF can prevent the shift in ocular dominance normally observed in monocularly deprived rats and cats (72).

The neuropoietic cytokines may also be involved in such regulatory events. For instance, CDF/LIF mRNA levels are highest in the visual cortex and hippocampus, reaching maximal values in adulthood (77, 91, 149). The same conclusions hold for the CDF/LIF receptor (11). Moreover, mice in which the CDF/LIF gene has been disrupted by homologous recombination display severe alterations in the hippocampus and visual cortex (94). It is not yet known, however, whether these alterations occur during development or in maturity. It is now clear that the adult sensory cortex is capable of enormous plasticity, making very large changes in sensory maps exceedingly quickly (35). It will be of great interest to see what role the trophic and differentiation factors play in these remarkable changes in wiring. Another very promising area is that of gene expression in the adult CNS. There is evidence that continual presence of differentiation factors such as CDF/LIF is required for maintenance of neuropeptide expression in peripheral sensory neurons, for instance (82).

Another key group of trophic and differentiation factors active in the CNS are the steroid hormones. In addition to their organizational effects on the embryonic brain (17), gonadal steroids can direct neuropeptide expression in the adult CNS independently of effects on neuronal survival and growth. Estrogen regulates the expression of cholecystokinin (CCK) and substance P (SP) mRNAs differentially in a sexually dimorphic pathway in the amygdala (123). The selectivity of this control is particularly striking because these two neuropeptides are co-expressed in the same neurons. The hormonal influence is exerted during normal physiological events as indicated by the fact that the number of CCK-expressing neurons varies over the estrous cycle (89). The variation in CCK content makes it likely that the character of the synaptic transmission between this subset of estrogen-sensitive neurons in the amygdala and their target cells in the preoptic area is altered during the estrous cycle. Such alterations have been termed "chemical switching" of transmission by these cells (128). Another striking example of this phenomenon is the differential regulation of galanin (GAL) and luteinizing-hormone-releasing hormone (LHRH) in neurons that express both neuropeptides simultaneously. In female rats, such neurons in the medial preoptic area and their axons in the median eminence contain higher levels of GAL during proestrus than during estrus, while the number of neurons expressing LHRH is unaffected by the hormonal state of the organism (76).

These examples of hormonal regulation are especially interesting because the presence of a co-expressed neuropeptide that does not change with hormonal fluctuations serves as a good control for true trophic effects. Thus, the steroid effects described here are "activational," altering gene expression in the mature brain rather than (or in addition to) guiding the morphological organization of the system during development. Such activational alterations in gene expression have been observed in several other areas of the brain and PNS that subserve reproductive behaviors (30, 46, 115). Also worth noting is that the changes in neuropeptide expression represent only part of this story. Estrogen may also regulate neurotransmitter and steroid receptor levels (5, 9). Finally, estradiol can alter neuronal circuitry on a very rapid time scale. In the 24-hr period between proestrus and estrus, for example, synaptic density in the CA1 region of the hippocampus declines about a third (147), a result consistent with changes in synaptic density evoked by experimental manipulations in hormone levels.

Glucocorticoids instruct neuropeptide and neurotransmitter expression in several systems. The corticotropin-releasing factor (CRF)-containing neurons of the paraventricular nucleus of the hypothalamus express at least eight different transmitters/peptides simultaneously. Glucocorticoid exerts a selective, negative feedback on the expression of CRF and vasopressin (VP), without affecting levels of enkephalin and neurotensin (see ref. 128). CRF mRNA levels follow the diurnal surge in corticosterone, and adrenalectomy results in higher CRF and a massive increase in VP (54, 143). Independent control of the many neuropeptides in these neurons is likely to reflect the fact that these neurons are thought to form various synapses with different functions as their axons traverse the hypothalamus, median eminence, and anterior pituitary (128). The three physiological conditions of chronically low, medium, and high circulating corticosterone would yield paraventricular neurons of three distinct chemical states with discrete functional consequences. This steroid, in its neuronal differentiation factor role, thereby alters synaptic function in an anatomically stable circuit that is the final common pathway for mediating the pituitary–adrenal response to stress, on a minute-to-minute time scale (see ref. 54). In addition, there is evidence that VP/CRF co-expression is enhanced in response to behavioral stress paradigms in the absence of adrenal glands (J. Barrett, A.-J. Silverman, and D. Kelly, personal communication).

Glucocorticoids may also act indirectly to alter neuronal gene expression. For example, corticosterone (as well as testosterone) can regulate NGF expression in neurons and astrocytes (12, 69). This hormone can also inhibit the production of CDF/LIF by heart cells and non-neuronal cells of sympathetic ganglia (40, 47, 75). This inhibition can thereby affect the phenotype of neurons cultured with such non-neuronal cells.

Another aspect of cytokine action in the adult nervous system is the response to injury. Wounds in the CNS or severing a peripheral nerve results in the up-regulation of many growth and differentiation factors, including CDF/LIF, NGF, TGF-b1, and glial maturation factor b (see ref. 95). Tumor necrosis factor a (TNF-a), IL-1a, FGF, and EGF have also been implicated in the response to nerve injury (13, 26, 33, 112, 132). Although the interplay of these signaling agents with neurons, glia, and immune cells is not well understood, there is evidence that damaged neurons may participate in feedback loops involving neuronal differentiation factors in the injury response. For example, transecting postganglionic nerves, or culturing sympathetic ganglia for 24 hr, causes a dramatic increase in the neuropeptides vasoactive intestinal polypeptide (VIP), SP, and GAL (53, 106, 107, 154). This is of interest in the context of the neuropoietic cytokines because these agents can induce the same peptides in sympathetic neurons. Moreover, cutting these nerves or isolating the ganglia in culture causes an enormous rise in CDF/LIF mRNA (11). Tying these observations together is the recent finding that ganglia from CDF/LIF-deficient mice (produced by disrupting this gene by homologous recombination) do not display this striking increase in VIP and GAL expression upon explanation to culture or axotomy (107). Thus, CDF/LIF mediates a major part of the neuropeptide induction that occurs in response to injury.

Why is the neuropeptide phenotype altered when sympathetic neurons are damaged? One possibility is that the neuropeptides play a trophic role. Tissue injury and inflammation, for instance, alter neuronal gene expression through enhanced nociceptor (pain receptors) activity, and the induced neuropeptides and excitatory amino acids are thought to be involved in the axonal sprouting and plasticity associated with these injuries and with nerve damage (31, 127). A novel possibility in the case of sympathetic neurons is that it may be important to change the neuropeptides the neurons are producing. Release of CDF/LIF could induce the particular neuropeptides that are known to attract immune cells and activate them (VIP and SP) (reviewed in ref. 95). There is evidence that sympathetic axons participate in the inflammation associated with arthritis (67). A role for CDF/LIF in the nervous system injury response fits nicely with its proposed functions in the hematopoietic and hepatic responses to infection (59) (see Corticotropin-Releasing Factor: Physiology, Pharmacology, and Role in Central Nervous System, Galanin: A Neuropeptide with Important Central Nervous System Actions, and Neuroendocrine Interactions).

A major, and perhaps unexpected, feature of the studies discussed thus far is that dynamic alterations in transmitter/neuropeptide expression can occur in postnatal, fully functional neurons. This can occur as a response to normal fluctuations in neuronal activity or hormone levels, or in response to injury or other insults to the system. Because these alterations in neuronal gene expression can entail qualitative changes in the mode of synaptic transmission (e.g., chemical switching), it is worth asking how many higher functions/behaviors may be influenced by instructive differentiation factors in the normal organism. As discussed in the prior section, the behaviors associated with the estrous cycle are clearly higher functions that can be included in this context. A second example is the circadian rhythms that drive many different behaviors.

In mammals, daily rhythms such as the sleep–wake cycle are controlled by a circadian clock located in the suprachiasmatic nucleus (SCN) of the anterior hypothalamus. Ablation of the SCN disrupts the daily rhythm of locomotor activity (125), and implantation of an SCN drives the circadian rhythm of the host animal according to the clock of the donor (64, 101). The SCN contains two major neuronal populations, one containing VIP and the other containing arginine-VP (142). Neuropeptide mRNAs, neuropeptide levels, and neuropeptide release vary precisely with the circadian cycle (6, 86, 137). The importance of the SCN peptides in controlling the rhythm is shown by the ability of injected VIP and peptide histidine isoleucine to shift the cycle (7).

What drives the cyclic changes in SCN neuropeptide mRNAs? Neuronal activity from visual afferents presumably regulates VIP expression, because its mRNA levels are influenced by light (86). In contrast, the diurnal change in VP mRNA and peptide is not influenced by visual inputs or endocrine products (19). Moreover, the diurnal cycle of VP release is retained by SCN explants in vitro, and it is also observed in dissociated neurons in long-term cell cultures (79). These observations suggest that the SCN can serve as a circadian pacemaker independently of exogenous neural and endocrine regulation. Possible mechanisms include cell autonomous oscillations and oscillations in the interactions among cells within the SCN, including potential variations in paracrine/autocrine cytokine release or action. There is evidence to support both of these mechanisms (79, 130).

There is also evidence that oscillations in paracrine/autocrine factors could play a role. Levels of the neurotrophic cytokine S100b exhibit circadian variation in the SCN and visual cortex (141). In addition, there is correlative evidence that S100b may be involved in the development of serotonin neurons in the CNS (4, 81, 136). Thus, it is possible that the diurnal variations in S100b in the SCN could contribute to the variations in neuropeptide expression there. Although there is presently only circumstantial evidence implicating cytokines in circadian rhythms, these systems are rich areas for investigation.

Alterations in synaptic transmission are also believed to be central to the complex events comprising learning and memory. Long-term potentiation (LTP) of synaptic efficacy is one of the most intensively studied mechanisms in this regard (see ref. 126). Is there a role for cytokines in LTP and similar types of synaptic plasticity? It is clear that expression of neurotrophins such as NGF and BDNF can be regulated by neuronal activity in a time scale of hours (55, 71, 151). As discussed above, cytokines can be transported anterogradely to synaptic endings and can be released by depolarization, and they can act retrogradely as well. In fact, tetanic stimulation of intact neocortex releases heat-labile factors that enhance neurite outgrowth from PC12 cells (a sympathetic neuron-like line) as well as induce LTP in slices of hippocampus (117). The LTP-inducing factors do not appear to influence resting membrane potential or input resistance. The LTP-inducing activities in the cortical superfusate are heterogeneous in size, ranging from <3 to >50 kD (148). The high-molecular-weight fraction induces LTP soon after its application, whereas the smaller fractions require 50 min to induce potentiation. These results suggest that diverse molecules and mechanisms are involved. It is also worthwhile examining known factors; a stimulus paradigm used to induce LTP in hippocampal slices produces enhanced levels of BDNF and NT-3 mRNAs (92).

The complementary type of experiment—testing known cytokines and their antagonists in LTP paradigms—is also informative. In hippocampal slices, EGF and FGF enhance potentiation of the population spike amplitude and field excitatory postsynaptic potential slope in the CA1 region after tetanic, but not low-frequency, stimulation (133, 134). These effects have been reproduced in several laboratories, and they are significant at 6–60 ng/ml of cytokine (1). The action of the two cytokines is distinguishable: Enhancement of LTP by EGF is more obvious in the earlier phase, whereas the effect of FGF is more significant in the later phase (1, 134). Such time periods correspond to the induction versus the maintenance phases of LTP. Cytokine administration is also effective in the living animal. EGF and FGF, but not NGF, increase the magnitude and probability of LTP induction in the dentate gyrus of the intact hippocampus (57). Moreover, differences have been observed between the actions of aFGF and bFGF on LTP induction in fasted versus nonfasted rats (50). The mechanism of LTP facilitation by EGF is being pursued in dissociated hippocampal neurons. Using the fura-2 assay, EGF and bFGF are found to significantly enhance the intracellular calcium increase induced by N-methyl-D-aspartate (NMDA; a glutamate analogue) (2, 3). These results suggest that both proteins selectively enhance NMDA-receptor-mediated responses in hippocampal neurons, an effect that could contribute to the facilitation of LTP by these cytokines.

In contrast to the effects of EGF and FGF, IL-1b induces synaptic inhibition in rat hippocampal pyramidal neurons (152). This result could be due to the inhibitory function of SOM, which can be induced by IL-1b in the cortex (118). Other cytokines are active in such assays; interferon and IL-2 can suppress previously established LTP, as well as inhibit its induction in the hippocampus (15, 27). TNF-a can affect LTP an hour after its addition to hippocampal slices (131).

The relatively rapid effects of the cytokines suggest that they could influence intercellular signaling directly, rather than through the alteration of gene expression. Many cytokine receptors have tyrosine kinase domains (120), and electrophysiological and genetic evidence implicate protein kinases in LTP (see ref. 126). Thus, it is entirely possible that cytokines could act directly as neuromodulators or neurotransmitters. It is also possible, however, that cytokine-driven changes in gene expression may be involved in learning and memory. For example, long-term changes in synaptic efficacy involve RNA and protein synthesis, and there is evidence that some aspects of LTP may also require new protein synthesis (124). In fact, cytokines can induce major changes in neuropeptide expression rather quickly; 20-fold increases in SP and VIP mRNAs were observed in sympathetic ganglia within 24 hr (109). Thus, the instructive actions of cytokines could be involved in relatively short-term synaptic plasticity, as well as in the consolidation phase of memory or its conversion from short- to long-term storage. An intriguing correlation in this respect is the recent finding that one of the genes induced in the hippocampal dentate gyrus by the glutamate analogue kainate is MyD118, a gene that is also induced by CDF/LIF and IL-6 in myeloid cells (84). If nothing else, this finding suggests common signaling pathways for cytokine action and activity involved in long-term plasticity.

Another point at which cytokines could influence learning is in the morphological changes that can accompany learning and related behavioral paradigms. It is now clear that anatomical connections are continually remodeled in adult as well as developing nervous systems, and some of these changes can be linked directly to learning (see refs. 43 and 95). A particularly intriguing example in this context are the structural changes that accompany long-term synaptic facilitation in Aplysia. In this case, growth of presynaptic processes requires the presence of the postsynaptic neuron, suggesting the action of a retrogradely acting trophic factor (74). Once more, the suggestion is that same developmental mechanisms employed to set up the wiring system can be used in the mature brain to modify these connections (58).

It is clear that cytokines can act as neuronal differentiation factors, and that these agents are probably utilized in both the retrograde and anterograde directions. There are newly emerging families of cytokines and receptors that, along with other superfamilies, are likely to be involved in the following: (i) the response to injury, as well as the feedback between the immune and nervous systems; (ii) the ongoing daily and monthly biological rhythms of the organism; (iii) the changes in synaptic plasticity involved in learning and memory; and (iv) synaptic transmission directly. The ability to regulate neuropeptide expression, both quantitatively and qualitatively, adds another dimension to the plasticity and capacity for change that is becoming clear from experimental manipulations of cortex. This regulation can dramatically alter synaptic function, both rapidly (in short-term physiological assays) or over the course of a day or month. These changes in synaptic function appear to contribute in key ways to complex animal behaviors, such as estrus, the stress response, and possibly circadian rhythms. While evidence is accumulating that cytokines may be involved in experimental paradigms of learning and memory, it is not yet clear if this involvement is in the context of neurotransmitter/neuropeptide regulation, through physical rearrangement of synaptic contacts, or through direct action on membrane tyrosine kinases and signal cascades. Another frontier is the potential use of cytokines in the treatment of pathological conditions in the nervous system. Clinical trials are underway to test the efficacy of these agents in the neurotrophic context where neurons are dying, but they could also prove useful in the manipulation of transmitter/neuropeptide imbalances in mental disorders. Moreover, genetic testing for predisposition to particular conditions could open the way for early, prophylactic intervention with these factors.

I thank Floreen Rooks-Les Pierre for help in preparing the manuscript, and I also thank Herman Govan and Ming-ji Fann for their helpful comments on the text.

published 2000