y,

Keywords:InammationLearningMemoryNeural plasticityLong-term potentiation (LTP)Neurogenesis

it brall/ps

with immune functions (particularly microglia and astrocytes), peripheral immune cells (particularly T

he immal procays frs by si

tively regulate neuroplasticity and neurogenesis, promoting learn-ing, memory, and hippocampal long-term potentiation (LTP). Thesecond part of the review will focus on the detrimental effectsof inammatory conditions induced by infections and injury aswell as severe or chronic stress, demonstrating that under such

addition, immune-like processes are involved in tissue remodeling,which is a continuous process of dynamic alterations in a specictissue or a whole organ that facilitates morphological and func-tional adaptations to the ever changing environmental demands.For example, in the bones macrophage-like cells (osteoclasts)continuously regulate bone structure and function by tissueresorption and by secreting myriad cytokines and chemokines thatadapt the bone to internal and external pressures and demands(Teitelbaum, 2000). Interestingly, osteoclasts (and other bone cells)

Corresponding author. Fax: +972 2 5882947.

Brain, Behavior, and Immunity xxx (2010) xxxxxx

Contents lists availab

r,

.eE-mail addresses: razyirmiya@huji.ac.il, a.kavelaars@umcutrecht.nl (R. Yirmiya).mune-like processes involving neuroglial communication withinthe brain. Behavioral and neural plasticity are among the mostimportant aspects of brain functioning that are modulated by im-mune mechanisms. The aim of the present review is to present acomprehensive and integrative view of the complex dual role ofthe immune system in learning, memory, neural plasticity andneurogenesis. The rst part of the review will focus on the physio-logical benecial effects of the immune system under normal, qui-escent conditions. Under such conditions, immune mechanismsare activated by environmental/psychological stimuli and posi-

in learning, memory and neuroplasticity.

1. The role of the immune system in learning, memory, neuralplasticity and neurogenesis under quiescent conditions

The immune system is primarily involved in surveillance ofbodily tissues and protection from infectious agents and variousforms of injury. It is also activated by, and participates in processesthat prepare the tissues for potential danger of such challenges. InCytokinesMicrogliaAstrocytesIL-1

It is now rmly established that tulate brain functioning and behavioris exerted by communication pathwmune system to the brain as well a0889-1591/$ - see front matter 2010 Elsevier Inc. Adoi:10.1016/j.bbi.2010.10.015

Please cite this article in press as: Yirmiya, R., Go(2010), doi:10.1016/j.bbi.2010.10.015cells and macrophages), neurons, and neural precursor cells. These interactions involve the responsive-ness of non-neuronal cells to classical neurotransmitters (e.g., glutamate and monoamines) and hor-mones (e.g., glucocorticoids), as well as the secretion and responsiveness of neurons and glia to lowlevels of inammatory cytokines, such as interleukin (IL)-1, IL-6, and TNFa, as well as other mediators,such as prostaglandins and neurotrophins. In conditions under which the immune system is stronglyactivated by infection or injury, as well as by severe or chronic stressful conditions, glia and other brainimmune cells change their morphology and functioning and secrete high levels of pro-inammatorycytokines and prostaglandins. The production of these inammatory mediators disrupts the delicate bal-ance needed for the neurophysiological actions of immune processes and produces direct detrimentaleffects on memory, neural plasticity and neurogenesis. These effects are mediated by inammation-induced neuronal hyper-excitability and adrenocortical stimulation, followed by reduced production ofneurotrophins and other plasticity-related molecules, facilitating many forms of neuropathology associ-ated with normal aging as well as neurodegenerative and neuropsychiatric diseases.

2010 Elsevier Inc. All rights reserved.

une system can mod-esses. This modulationom the peripheral im-gnals produced by im-

conditions the delicate physiological balance between immuneand neural processes is disrupted, resulting in neuronal hyper-excitability, hormonal aberrations, reduced neurotrophic factorsproduction and suppressed neurogenesis, leading to impairmentsAvailable online xxxx cuits, promoting memory consolidation, hippocampal long-term potentiation (LTP) and neurogenesis.These benecial effects of the immune system are mediated by complex interactions among brain cellsNorman Cousins Lecture

Immune modulation of learning, memor

Raz Yirmiya , Inbal GoshenDepartment of Psychology, The Hebrew University of Jerusalem, Jerusalem 91905, Israel

a r t i c l e i n f o

Article history:Received 8 October 2010Accepted 16 October 2010

a b s t r a c t

Over the past two decadeslearning, memory and neuactivated by environmenta

Brain, Behavio

journal homepage: wwwll rights reserved.

shen, I. Immune modulation ofecame evident that the immune system plays a central role in modulatingplasticity. Under normal quiescent conditions, immune mechanisms areychological stimuli and positively regulate the remodeling of neural cir-neural plasticity and neurogenesis

le at ScienceDirect

and Immunity

lsevier .com/locate /ybrbilearning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,do not function autonomously, but are importantly inuenced bythe endocrine and nervous systems, demonstrating the importanceof neuroimmune interactions for tissue maintenance and remod-eling even in conditions that do not involve infection or injury(Bajayo et al., 2005; Yirmiya and Bab, 2009; Yirmiya et al., 2006).Similar immune-mediated remodeling processes also occur inother tissues, such as the muscles, fat, and reproductive organs,particularly when these tissues encounter and have to adapt tomajor environmental challenges, e.g., during or following muscleexertion, obesity, ovulation, and menses.

The healthy brain provides a classic example for the necessity oftissue remodeling to adaptive coping, since neural cells and net-works are constantly altered by experience. During development,neurons and other cells are added to the evolving brain structures,however, many other cells (as much as 5060% of the neurons thatare formed in the brain during development) die before birth(Oppenheim, 1991). Moreover, a large percentage of the processesof the developing neurons undergo dynamic and dramatic pruning,i.e., selective degeneration of whole or parts of the dendrites, axoncollaterals or terminals. Neuronal (and other cellular) death andpruning ensures the formation of accurate, ne-tuned, and ef-ciently functioning neural circuits (Luo and OLeary, 2005). In ani-mals and humans these processes continue into adulthood, albeitin a less dramatic manner, i.e., brain cells still undergo apoptosisand neurogenesis (at least in specic brain locations), axons anddendrites are still formed and get pruned, and most importantly,individual synapses (and associated structural elements) areformed, retracted and get modied throughout life (Luo andOLeary, 2005). These processes, collectively termed neural plastic-ity, underlie the most amazing and wonderful capacity of the brainto adapt to the ever changing environment via learning andmemory.

Similarly to its role in remodeling of bodily tissues, the immunesystem participates in modulating and sculpting the brain. It has tobe particularly involved when cellular events in the nervous sys-tem lead to apoptosis, as well as degradation of processes and evenindividual synapses. Cellular corpses, neuritic debris, and remainsof other cells (e.g., myelin and associated proteins that remain afternormal pruning of axonal processes) cannot stay in the tissue with-out interfering with its normal functioning. Therefore, the processof neural plasticity must be exquisitely coordinated with and reg-ulated by immune mechanisms that ensure the quality and ef-ciency of this process. As will be discussed below, immune-mediated brain remodeling processes may be initiated by neuronalactivity, but they primarily involve various non-neuronal cellswithin the brain parenchyma (mainly microglia, but also astro-cytes and possibly mast cells), as well as cells within and aroundthe brain vasculature, choroids plexus and meninges (includingendothelial cells, perivascular macrophages, and T cells). Other im-mune molecules that were also found to be important for normalneural and synaptic functioning include the major histocompatibil-ity complex (MHC) class I (Boulanger and Shatz, 2004), and thecomplement system (Stevens et al., 2007). Together, the brain-associated immune cells and the molecules secreted by these cellstake part in promoting plasticity-related structural changes, andmay be directly involved in the neurophysiological processesunderlying the plastic changes.

Interestingly, the activation of neuroimmune responses byphysiological neural activity (in the absence of infection or overtinjury) can also activate brain-to-body communication pathways,such as the hypothalamuspituitaryadrenal (HPA) axis and theautonomic nervous system (Besedovsky and Del Rey, 2007). Theresultant peripheral hormonal and neurochemical alterations

2 R. Yirmiya, I. Goshen / Brain, Behav(e.g., the elevation in blood levels of glucocorticoids, adrenalineand norepinephrine) feed back into the brain and exert powerfulmodulatory effects on neural plasticity and neurogenesis. These

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015neuro-hormonal processes can also involve alterations in periphe-ral immune parameters, which in turn inuence central immuneresponses, creating a brain-to-body-to-brain reverberating feed-back loops.

The idea that the immune system is involved in normal neuro-behavioral processes was suggested more than a decade ago,although initially it did not receive much attention, probably be-cause of the overwhelming evidence demonstrating that immuneprocesses during infection, injury and stress produce sicknessbehavior, debilitation and impaired neurobehavioral plasticity.Based on the observations that cytokines and their receptors areexpressed, albeit at low levels, in the healthy brain, that neurons(in addition to glia) can produce and respond to inammatorycytokines, and that neuronal activity can regulate the productionand secretion of cytokines, it was suggested that cytokines act asneuromodulators in the normal healthy brain (i.e., without anyovert pathophysiological stimuli) (Vitkovic et al., 2000). Additionalsupport for this notion came from studies demonstrating theinvolvement of cytokines in specic normal neurobehavioral func-tions, including sleep (Krueger et al., 2001; Opp, 2005), pain (Wolfet al., 2003) and responsiveness to various psychological stressors(Goshen and Yirmiya, 2009). The following section demonstratesthat immune processes and particularly pro-inammatory cyto-kines play an important role in behavioral and neural plasticity.

1.1. The role of the immune system in promoting learning and memory

1.1.1. The role of T cellsBased on their previous ndings that CD4+ T cells targeted

against brain self antigens can be neuroprotective (a phenomenontermed protective autoimmunity) (Moalem et al., 1999), M. Sch-wartz, J. Kipnis and their colleagues raised the notion that circulat-ing T cells play a general supportive role in brain and mindfunctioning, including cognitive abilities and neurogenesis (Kipniset al., 2008; Schwartz and Shechter, 2010). Experimental evidencefor this notion was rst provided by demonstrating that mice withsevere combined immune deciency (SCID, devoid of both T and Bcells) as well as nude mice (decient only in mature T cells) displaydramatic impairments in hippocampal-dependent spatial learningand memory in the water maze (Kipnis et al., 2004; Ron-Harelet al., 2008). SCID mice also exhibited impaired learning and mem-ory in three other paradigms measuring hippocampal functioning the water-free Barnes maze, the radial arm water maze(Brynskikh et al., 2008) and recognition of novel spatial arrange-ment of familiar objects (Ron-Harel et al., 2008). Furthermore,replenishing T cells in nude mice markedly improved their learningand memory in the water maze. Consistently, boosting T cell acti-vation by vaccination with copolymer-1, a weak agonist of variousself-reactive T cells, counteracted learning decits induced by neu-rotransmitter abnormalities (Kipnis et al., 2004). Replenishmentwith T cells derived from wild type (WT) mice (but not withSCID-derived T cells or with T-cell depleted splenocytes) also im-proved the memory of SCID mice in the water maze (Ron-Harelet al., 2008) or the radial arm water maze (Brynskikh et al.,2008). A recent study corroborated these ndings by demonstrat-ing that T cell depleted mice displayed impaired performance inreversal training in the water maze paradigm (although their per-formance during the learning phase or the probe test was similar tocontrol) (Wolf et al., 2009a).

The learning and memory impairments in SCID mice do not re-sult from health problems related to their lifelong immune de-ciency following complete depletion of their immune systemby lethal irradiation, adult SCID mice that received bone marrow

and Immunity xxx (2010) xxxxxxtransplantation of WT immune cells exhibited normal memoryfunctioning in the water maze and the novel location task, whereasmice transplanted with bone marrow cells derived from SCID mice

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior, aexhibited marked memory decits (Brynskikh et al., 2008; Ron-Harel et al., 2008). Consistent with all of the above ndings, trans-genic mice with excess of monospecic T cells directed towards abrain antigen exhibited better learning and memory in the watermaze than their controls, whereas the performance of mice withtransgenic excess of T cells directed against an irrelevant (non-self)antigen was worse than their WT controls (Ziv et al., 2006).

The meningeal spaces constitute an important location for Tcell-based support of behavioral plasticity, providing an explana-tion for such effects of T cells without their presence in the CNSparenchyma (Derecki et al., 2010; Schwartz and Shechter, 2010).Specically, performance of cognitive tasks led to increased num-ber of T cells in the meninges, and depletion of T cells from menin-geal spaces resulted in learning and memory impairments.Moreover, this depletion skewed meningeal myeloid cells towarda pro-inammatory phenotype, which may have interfered withthe learning process (Derecki et al., 2010). The study further dem-onstrates an important role for T cell-derived IL-4 in the regulationof cognitive functioning. Specically, the T cells that accumulatedin the meninges following learning in the water maze expressedhigh levels of IL-4. Furthermore, IL-4-decient mice, as well as irra-diated wild-type recipient mice that were transplanted with IL-4-decient bone marrow, exhibited cognitive impairments, concom-itantly with a skewed pro-inammatory meningeal myeloid cellphenotype. Moreover, adoptive transfer of T cells from wild-typeinto IL-4-decient mice reversed cognitive impairment and atten-uated the pro-inammatory character of meningeal myeloid cells.Thus, this study clearly demonstrates that T cell-derived IL-4 regu-lates learning and memory, possibly by inuencing meningealmyeloid cell activation (Derecki et al., 2010).

The importance of properly functional and regulated T cellimmunity in cognitive functioning is supported not only by theabove-mentioned studies on SCID, nude and irradiated/immune-depleted mice, but also by more common and natural conditionsassociated with both diminished T cell activity and cognitiveimpairments, such as aging, HIV infection and chemotherapy (Kip-nis et al., 2008; Ron-Harel and Schwartz, 2009). Indeed, manipula-tion of T cells in aged mice (e.g., bone marrow transplantationfollowing irradiation) reversed some aspects of the spatial memorydecits exhibited by these animals (Ron-Harel et al., 2008). More-over, the procedures that have so far been shown to alleviate brainaging and the associated memory loss, including physical exerciseand calorie restriction, enhance T cell immunity, suggesting thatboosting T cell immunity might be benecial for aging-associatedmemory impairments (Ron-Harel and Schwartz, 2009).

1.1.2. The role of inammatory cytokinesMost of the research on the role of the immune system in learn-

ing and memory processes focused on the involvement of inam-matory cytokines, particularly IL-1, IL-6, and tumor necrosisfactor TNF-a.

1.1.2.1. IL-1. Several lines of evidence indicate that IL-1 is requiredfor some learning and memory processes, particularly for the con-solidation of memory that depends on proper functioning of thehippocampus:

1.1.2.1.1. Induction of IL-1 during the learning process. Since un-der normal quiescent conditions brain levels of IL-1 are very low(usually at or below the threshold of IL-1 protein detection), IL-1should be induced during learning and memory consolidation inorder to exert effects on these processes. We assessed this hypoth-esis by measuring the expression of IL-1b mRNA at various timepoints following contextual fear conditioning. We reported that

R. Yirmiya, I. Goshen / Brain, BehavIL-1b gene expression in the hippocampus showed a signicant in-crease 24 h, but not 1.5 or 4 h after contextual learning (Goshenet al., 2007). This increase could not be attributed to the exposure

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015to the stress caused by the electrical shock per se, but only to thelearning experience, as shock administration in the home cagedid not affect IL-1b gene expression. Interestingly, in two geneti-cally manipulated mouse strains with decient IL-1 signaling, micewith deletion of the IL-1 receptor type I (IL-1rKO) or mice withtransgenic over-expression of IL-1 receptor antagonist (IL-1raTG),the levels of IL-1b gene expression were not increased 24 h afterfear-conditioning. Because IL-1 itself is one of the major triggersfor the induction of further IL-1 production (Dinarello, 1996), it ispossible that an initial small increase in IL-1 secretion and activa-tion of IL-1R1 are required to induce the relatively high levels ofexpression at 24 h post-conditioning. A recent study corroboratedthese ndings by demonstrating the induction of IL-1b mRNAexpression following spontaneous spatial recognition in theY-maze paradigm (Labrousse et al., 2009). Interestingly, mice withgenetic deletion of P2X7 ATP receptors, which are critical for IL-1bproduction by hippocampal glia, displayed no IL-1b expression fol-lowing exposure to this paradigm, concomitantly with memoryimpairment and abrogation of hippocampal neural activation.These ndings support a crucial role for P2X7 receptor-mediatedIL-1b expression in memory processes involving the hippocampus(Labrousse et al., 2009). In another study, the gene expression ofIL-1a, IL-1b, and IL-1ra was examined in the hippocampus at dif-ferent time points (1, 4, 6, and 9 h) following a single acquisitiontrial of step-down passive-avoidance. In that study, IL-1a geneexpression was increased 4 h after acquisition, but no change inthe expression of IL-1b and IL-1ra were observed at any time point(Depino et al., 2004). The somewhat different results obtained inthis study may stem from a procedural difference because the con-trol group in this study consisted of animals that received a shockin the conditioning context but did not perform the step-down ac-tion, thus the difference between the experimental and controlgroups is not in contextual learning per se, but in its associationwith the performance of a motor activity.

1.1.2.1.2. Facilitation of learning and memory by IL-1 administra-tion. Although under most circumstances exogenous administra-tion of IL-1 produces learning and memory impairments (seeSection 2.1.1.1 below), in other situations (which probably dependon specic combinations of the timing, dose and route of adminis-tration, as well as on the particular memory paradigm), adminis-tration of IL-1 produces memory facilitation.

Clear evidence for IL-1-induced memory facilitation was ob-tained using the passive and active avoidance paradigms. In onestudy, we demonstrated that i.c.v. administration of a relativelylow dose of IL-1b, immediately following passive avoidance train-ing, resulted in better memory 58 days later (Yirmiya et al., 2002).This nding was later corroborated by showing that i.c.v. adminis-tration of IL-1b shortly before passive avoidance acquisition, aswell as before the memory tests (conducted 24 and 48 h later) re-sulted in improved memory (Song et al., 2003). Although in an-other study very low doses of IL-1a had no effect on passiveavoidance memory, similar doses of IL-1a attenuated the amnesiceffect of scopolamine (an anti-cholinergic drug) on memory in thistest (Bianchi et al., 1998). In the active avoidance paradigm, inwhich rats can press a lever to prevent an imminent shock (avoid-ance response) or to terminate the shock after it had begun (escaperesponse), low doses of IL-1b, administered 24 h before training,increased the number of avoidance responses (which critically de-pend on hippocampal functioning), but had no effect on the num-ber of escape responses (which are hippocampal-independent)(Brennan et al., 2003, 2004).

Under some conditions, IL-1bwas also found to facilitate spatialand contextual fear memories, which also depend on hippocampal

nd Immunity xxx (2010) xxxxxx 3functioning. In one study, IL-1b administration facilitated spatialmemory in the water maze (Gibertini, 1998). In another study,we found that i.c.v. administration of a relatively low dose of

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,IL-1b immediately following the association of a specic contextwith mild foot-shock improved contextual fear conditioning,tested 48 h later (Goshen et al., 2007). The same IL-1 dose had noeffect on auditory cued fear conditioning, which does not dependon hippocampal functioning. Finally, peripheral treatment withIL-b led to facilitated acquisition of the classically conditionedeye blink response (Servatius and Beck, 2003).

1.1.2.1.3. Learning and memory are impaired following pharma-cological blockade of IL-1 signaling via IL-1ra administration. The ef-fects of IL-1ra were studied in several memory paradigms. In thepassive avoidance paradigm, intracerebral administration of IL-1raat the end of training signicantly impaired memory, tested58 days later (Yirmiya et al., 2002). In the fear conditioning para-digm, similar administration of IL-1ra also induced contextual (butnot auditory-cued) fear conditioning impairment (Goshen et al.,2007). It should be noted that in contrast with these ndings, in-tra-hippocampal injection of an adenovector containing the ratIL-1ra gene was found to enhance short- and long-term memoryin the step-down passive avoidance test (Depino et al., 2004).The use of an adenovector vs. pharmacological IL-1ra administra-tion, and differences in the species (mice vs. rats) and learning pro-cedures may explain the discrepancy between the results of thesestudies.

1.1.2.1.4. Learning and memory are disturbed in mice with geneticimpairments in IL-1 signaling. To corroborate the studies demon-strating that pharmacological blockade of IL-1 impairs memoryfunctioning, we examined spatial memory and fear conditioningin IL-1rKO and IL-1raTG mice. In both models, IL-1 signaling iscompletely abolished, as reected by lack of responsiveness toexogenous IL-1a or IL-1b administration. Both IL-1rKO andIL-1raTG mice displayed a slower rate of learning in the spatialmemory paradigm (Avital et al., 2003; Goshen et al., 2007). In con-trast, in the non-spatial memory paradigm (visible platform) bothIL-1-signaling decient strains showed no differences from con-trols. Similar ndings were obtained using the fear conditioningparadigm: both IL-1rKO and IL-1raTG mice exhibited impairedcontextual, but normal auditory-cued fear conditioning (Avitalet al., 2003; Goshen et al., 2009, 2007). A recent study, corrobo-rated these ndings, demonstrating that IL-1raTG mice displayedlearning and long (but not short) term memory decits in thewater T maze paradigm (Spulber et al., 2009), which also dependson hippocampal functioning. Another recent study reported con-tradictory ndings, i.e., increased freezing in both contextual andauditory-cued fear conditioning testing (Koo and Duman, 2009).The reason for this discrepancy is not clear, and may be relatedto minor procedural and measurement differences, as well as a dif-ferent genetic background of the mice.

In conclusion, the data presented above demonstrates that IL-1is induced within the hippocampus during the learning process,that under some conditions exogenous administration of low dosesof IL-1 can improve hippocampal-dependent memory functioning,whereas blockade of IL-1 signaling, either by pharmacological or bygenetic manipulations, can impair memory functioning. Thus, itcan be concluded that low, physiological levels of IL-1 in the hip-pocampus play an important role in learning and memoryprocesses.

1.1.2.2. IL-6. Ample research indicates that high levels of IL-6, par-ticularly in the context of aging, are associated with poor memoryfunctioning or memory decline over time (see Section 2.1.1.2 be-low). However, it is now clear that the role of IL-6 in memory isquite complex and that IL-6 may inuence learning and memoryin different, and even opposite ways under various conditions. In

4 R. Yirmiya, I. Goshen / Brain, Behavthe only study on the effects of exogenous IL-6 on memory func-tioning in humans, IL-6 was administered to 19 chronic fatiguesyndrome (CFS) patients and 10 control subjects, and memory, as

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015well as other neuropsychological functions, was assessed 6.5 hlater. Surprisingly, IL-6 did not produce any memory disturbance.In fact, both CFS patients and controls demonstrated improved per-formance in this test (Arnold et al., 2002). Although the investiga-tors ascribed this improvement to practice effect, it is possible thatthe improved cognitive functioning was induced by IL-6 (in the ab-sence of a saline-administered control group, it is impossible todistinguish between these two possibilities).

The possibility that under some conditions IL-6 may be associ-ated with protective effects on cognition was suggested by twoadditional studies in humans, examining the role of IL-6 in sys-temic lupus erythematosus and surgical patients. In Lupus pa-tients, a negative association between plasma IL-6 and cognitivedecline was obtained, i.e., higher levels of IL-6 in the plasma wereassociated with higher learning scores (Kozora et al., 2001). Simi-larly, in a recent study examining the effects of surgical stress onimmune and cognitive functioning, we found that the elevationin IL-6 levels one day after surgery was inversely associated withcognitive deterioration, i.e., patients with elevated IL-6 levelsexhibited smaller declines in declarative memory regardless ofinterpersonal differences in age, gender, pain experienced, or base-line ability (Shapira-Lichter et al., 2008).

Consistently with these ndings, a protective role for IL-6administration was also found in two studies using animal models.Specically, i.p. administration of IL-6 partially blocked the effectof the amnesic drug scopolamine in the passive avoidance task(Bianchi et al., 1997; Bianchi et al., 1998). Moreover, chronicadministration of IL-6 for a week, commencing 2 h before ischemia,resulted in improved memory in a passive avoidance task at theend of this week (Matsuda et al., 1996).

1.1.2.3. TNFa. In addition to the extensive literature on the detri-mental effects of TNFa in learning and memory (see Sec-tion 2.1.1.1.3), there is some evidence for its benecial effectsunder certain conditions. For example, in the passive avoidanceparadigm a single i.p. injection of TNFa 24 h before training re-sulted in increased number of avoidance as well as escape re-sponses (Brennan et al., 2004). TNFa also seems to have apositive effect on the recovery of memory functioning followinginfection; after surviving pneumococcal meningitis, mice with tar-geted deletion of the TNFa gene (TNFaKO) mice demonstrated im-paired water maze performance compared to surviving WTcontrols, suggesting a benecial role for TNFa in memory recovery(Gerber et al., 2004). This nding suggests that similarly to IL-6,when brain homeostasis is maintained TNFa has a detrimental ef-fect on memory processes, whereas when this balance is violated,TNFa may play a protective role.

1.1.3. ProstaglandinsProstaglandins (PGs) are important inammatory mediators,

synthesized from arachidonic acid by the enzyme cyclooxygenase(COX). Within the brain, PGs are mainly produced following theinduction of COX-2, which is expressed in neurons, glia and endo-thelial cells. Various stimuli can induce COX-2 expression in thebrain, including inammatory challenges, particularly IL-1, as wellas synaptic activity (Laamme et al., 1999; OBanion et al., 1996). Arole for PGs in memory functioning has been suggested by the nd-ings that peripheral or i.c.v. administration of the non-selectiveCOX inhibitors indomethacin and ibuprofen, as well as the selec-tive COX-2 inhibitors NS-398 and celecoxib, impaired spatial butnot visually guided water maze performance (Cowley et al.,2008; Shaw et al., 2003; Teather et al., 2002). A similar attenuationof water maze acquisition was found when the COX-2 specic

and Immunity xxx (2010) xxxxxxinhibitor celecoxib was administrated directly into the hippocam-pus (Rall et al., 2003; Sharifzadeh et al., 2005). Furthermore, directintra-hippocampal administration of the non-selective COX inhib-

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior, aitor naproxen signicantly impaired memory in the contextual fearconditioning paradigm (Hein et al., 2007). It should be noted, how-ever, that genetic deciency of either COX-1 or COX-2 had no effecton spatial learning and memory in the water maze (Kelso et al.,2009).

1.2. The role of immune processes in the induction and maintenance ofneural plasticity and LTP

The idea that synaptic changes underlie experience-inducedadaptations in brain and behavior was proposed more than60 years ago by Donald Hebb, who suggested that the synapse be-tween two neurons is strengthened if they are active simulta-neously (Hebb, 1949). Empirical support for this proposal wasprovided by the demonstration that brief high-frequency stimula-tion (HFS) of hippocampal afferents resulted in persistent augmen-tation of their synaptic strength within the dentate gyrus (DG), aphenomenon termed long-term potentiation (LTP) (Bliss and Lomo,1973). Ample research demonstrated that synaptic plasticity andLTP underlie memory storage in the hippocampus and other brainareas (Lynch, 2004). It should be noted, however, that althoughsynaptic plasticity was proved in many studies to be necessaryfor learning and memory, it is not clear whether it is sufcientfor memory formation (see (Martin et al., 2000), for review).

As will be discussed below, ample evidence demonstrates thatimmune processes in the brain produce detrimental effects on neu-ral plasticity, specically by reducing hippocampal LTP inductionand maintenance (reviewed in (Goshen and Yirmiya, 2007; Lynch,2002; OConnor and Coogan, 1999). However, in 1998 Hugo Bese-dovsky and his colleagues (Schneider et al., 1998) published a sem-inal paper, which demonstrated for the rst time that aninammatory-like process (the secretion of the pro-inammatorycytokine IL-1 in the hippocampus) accompanies LTP inductionand is critically involved in maintaining LTP. Since then, manyadditional studies veried the role of inammatory cytokines andother inammatory mediators in normal, physiological neuralplasticity.

1.2.1. The role of inammatory cytokines1.2.1.1. IL-1. As noted above, the nding that IL-1 gene expressionis substantially increased concomitantly with the development ofLTP provided the rst evidence for the involvement of an inam-matory-like process in normal/physiological neural plasticity. Theincreased IL-1b gene expression commenced rapidly (15 min afterstimulation in hippocampal slices and 8 h after LTP induction infreely moving rats), and it was long lasting, specic to potentiation(i.e., was observed only in the stimulated, but not the contralateral,hippocampus) and could be prevented by blockade of potentiationwith the NMDA receptor antagonist AP-5 (Balschun et al., 2003;Schneider et al., 1998). A recent study demonstrated that LTP-asso-ciated induction of IL-1b expression is not restricted to the hippo-campus: HFS to the spinal cord, which induced a robust LTP ofC-ber responses within the dorsal horn, was associated with a sig-nicant increase in the expression of mRNA for IL-1b, measured 6 hfollowing HFS administration (Pedersen et al., 2009).

The LTP-associated elevation in IL-1b gene expression seems toplay a causal physiological role in LTP maintenance, as blocking IL-1 receptors by i.c.v. administration of IL-1ra 90 min after the induc-tion of LTP impaired its maintenance (Schneider et al., 1998). Theeffects of IL-1ra administration before LTP induction are less clear;in vivo administration of IL-1ra 30 min before stimulation had noeffect in one study (Schneider et al., 1998), but in two other studiesIL-1ra administration 30 or 60 min before HFS signicantly atten-

R. Yirmiya, I. Goshen / Brain, Behavuated the initial potentiation and impaired the maintenance ofLTP (Loscher et al., 2003; Schmid et al., 2009). The critical role ofIL-1 in LTP was also demonstrated in hippocampal slices in vitro:

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015Application of IL-1ra 30 min after the induction of LTP in the ratDG in vitro reduced synaptic activity back to baseline levels(Coogan et al., 1999), and when IL-1ra was applied to hippocampalsliced in a physiological temperature for 40 min before stimula-tion, the initial small increase in synaptic activity subsided within30 min (Ross et al., 2003). The mechanisms underlying the detri-mental effect of IL-1ra on LTP were studied using hippocampalsynaptosomes preparation (Loscher et al., 2003), revealing de-creased glutamate release and increased JNK phosphorylation fol-lowing exposure to IL-1ra. Interestingly, these effects were notmediated by the IL-1 receptor type I (IL-1RI), as they appeared bothin the presence of neutralizing antibodies against IL-1RI, and insynaptosomes prepared from IL-1rKO mice (Loscher et al., 2003).

Using a different approach to study the necessity of physiolog-ical IL-1 levels for hippocampal plasticity, we found that anesthe-tized IL-1rKO mice exhibited no LTP in the DG following HFS ofthe perforant path (Avital et al., 2003; Goshen et al., 2009). Simi-larly, a complete absence of LTP was observed in the CA1 regionof hippocampal slices taken from IL-1rKO mice compared withWT controls (Avital et al., 2003). In contrast with these ndings,no impairment in either LTP or long term depression (LTD) was ob-served in mice with knockout of the IL-1a and IL-1b genes (Ikegayaet al., 2003). Short-term plasticity was also affected by IL-1 signal-ing impairment, evidenced by enhanced paired-pulse inhibition inthe DG of IL-1rKO mice in vivo, and decreased paired-pulse poten-tiation in hippocampal slices from IL-1rKO mice (Avital et al.,2003).

Together, the increase in IL-1 gene expression following LTPinduction, and the blockade of LTP by pharmacological or geneticmanipulations in IL-1 signaling demonstrate the requirement ofphysiological IL-1 levels for the induction and maintenance of LTP.

1.2.1.2. IL-6. As detailed below (Section 2.2.1.2), many studiesdemonstrated a detrimental effect of elevated IL-6 levels on long-term synaptic plasticity. However, recent studies suggest thatendogenous IL-6 may actually have a physiological role in LTP inhi-bition. In one study, a 20-fold increase in IL-6 gene expression wasmeasured 4 h following in vivo LTP induction by HFS (Jankowskyet al., 2000). This nding was corroborated by another study, dem-onstrating increased IL-6 gene expression 8 h after HFS in freelymoving rats as well as 13 h after HFS in hippocampal slices (Balsc-hun et al., 2004). This increase was found only in rats in which LTPwas robust and maintained for 8 h, but not when it subsided with-in 3 h following HFS. Immunoneutralization of IL-6 by i.c.v. admin-istration of anti-IL-6 antibodies 90 min following the induction ofLTP resulted in longer maintenance of LTP; i.e., LTP was still pre-served in anti-IL-6 injected rats after it subsided in control rats.In contrast, anti-IL-6 antibodies that were injected 30 min beforeor 5 min after the induction of LTP had no effect (Balschun et al.,2004), suggesting that IL-6 is selectively involved in a specicphase of LTP consolidation. No other endogenous protein is knownto interfere with LTP maintenance in such a late phase, withoutinuencing LTP induction.

The nding that LTP increases IL-6 gene expression, togetherwith the nding of longer preservation of LTP following IL-6 neu-tralization, suggest that IL-6 may have a physiological role in thetermination of LTP.

1.2.1.3. TNFa. Several lines of evidence implicate TNFa in synapticfunctioning, in general, and in some forms of synaptic plasticity, inparticular. Studies in both hippocampal cultures and slices demon-strated that TNFa selectively secreted by astrocytes enhances syn-aptic efcacy by increasing surface expression of AMPA receptors.

nd Immunity xxx (2010) xxxxxx 5Conversely, blocking TNFa signaling by TNF soluble receptors re-duces synaptic strength and decreases AMPA expression (Beattieet al., 2002). It should be noted, however, that the newly expressed

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,AMPA receptors have abnormal stoichiometry, as they lack theGluR2 subunit; thus, they become Ca2+ permeable and may con-tribute to neurotoxicity (Stellwagen et al., 2005).

TNFa does not seem to be involved in acute plasticity, reectedby the normal LTP exhibited by mice with genetically impaired TNFsignaling (Albensi and Mattson, 2000; Kaneko et al., 2008; Stellwa-gen and Malenka, 2006) (in fact, under certain conditions it hasdetrimental effects on LTP, see Section 2.2.1.3). Although in onestudy a role for TNFa in LTD has been suggested by demonstratingthat in hippocampal slices low frequency stimulation of Schaffercollateral axons produced LTD in the CA1 synapses of wild-typebut not TNF receptor knockout mice (Albensi and Mattson, 2000),in other preparations TNF signaling-decient mice displayed nor-mal LTD (Kaneko et al., 2008; Stellwagen and Malenka, 2006). Incontrast with its questionable role in acute plasticity, a clear rolefor TNFa has been shown in two forms of long-term plasticity inthe adult brain: synaptic scaling and structural plasticity inducedin the visual cortex by monocular deprivation.

Synaptic scaling refers to a set of homeostatic plasticity mech-anisms that dynamically adjust the strengths in all synapses on acell in response to prolonged changes in the cells electrical activ-ity, thereby stabilizing neural networks functioning. A role forTNFa in synaptic scaling was demonstrated by the nding that fol-lowing chronic blockade of neuronal activity, TNFa is necessary forthe increases in surface AMPAR levels and synaptic strength. Fur-thermore, it was found that the TNFa required for homeostaticsynaptic strengthening is exclusively produced by astrocytes(Stellwagen and Malenka, 2006).

Ample studies demonstrate that monocular deprivation in theadult mammalian brain results in weakening and pruning of inputsfrom the deprived eye, along with a gradual strengthening andexpansion of inputs from the open eye (Wiesel, 1982). A role forTNFa in this phenomenon has been demonstrated by showing thatmice decient in TNFa, either by genetic deletion of the TNFa geneor due to pharmacological blockade of TNFa signaling display nor-mal initial loss of deprived-eye responses, but complete absence ofsubsequent increase in response to the open eye. This mutationalso blocks homeostatic synaptic scaling of mEPSCs in visual cortexin vitro, without affecting LTP. These ndings suggest that TNFasignaling is essential for experience-dependent homeostatic syn-aptic scaling process (Kaneko et al., 2008).

1.2.2. ProstaglandinsIn accordance with their role in memory functioning, basal lev-

els of PGs also have a role in neural plasticity. This conclusion isbased on several lines of research demonstrating that: (1) COX-2activity is upregulated following HFS that is associated with LTPinduction (Yamagata et al., 1993). (2) PGs production by COX-2can occur in postsynaptic dendritic spines (Kaufmann et al.,1996). (3) Inhibition of COX activity by ibuprofen resulted in im-paired hippocampal LTP (Shaw et al., 2003). Furthermore, selectiveinhibition of COX-2 (but not COX-1) reduced postsynaptic mem-brane excitability and LTP induction in hippocampal dentate gran-ule neurons. This reduction, as well as the reduction is downstreamsignaling mechanisms (extracellular signaling-regulated kinase(ERK)-phosphorylation and c-FOS expression) could be completelyrescued by exogenous application of PGE2 (Chen et al., 2002; Cow-ley et al., 2008). These ndings indicate that endogenous basal lev-els of PGE2 resulting from COX-2 but not COX-1 activity are criticalfor long-term hippocampal synaptic plasticity.

1.2.3. MHC Class I proteinsMHC Class I (MHC-I) proteins are expressed by almost all nucle-

6 R. Yirmiya, I. Goshen / Brain, Behavated cells of the body, which bind and display endogenous cellularpeptides on the membrane surface and thereby permit immunerecognition of self vs. non self (foreign) antigens generated

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015by infected or cancerous tissues. Recent evidence demonstratesthat MHC-I proteins, as well as their immunoreceptors, play animportant role in the developing nervous system, in which theymay be required for the normal elimination of inappropriate pro-jections, similarly to their immunological role (recognition and re-moval of unwanted cells expressing non self antigens).Furthermore, neuronal MHC-I molecules regulate various aspectsof basal synaptic functioning (Boulanger, 2009), and have beenimplicated in neural plasticity. Specically, mice with genetic de-ciency of MHC-I or the MHC-I receptor component, CD3f, displayenhanced LTP and an absence of LTD in the hippocampus (Boulan-ger and Shatz, 2004; Huh et al., 2000), as well as lower thresholdfor induction of LTD in the cerebellum, associated with improvedmotor learning acquisition (McConnell et al., 2009).

Interestingly, mice lacking the polypeptide DAP12, which isselectively expressed only in microglia, is structurally and func-tionally related to CD3f, and associates with cell-surface receptorslike MHC-I, also exhibit enhanced LTP. This enhancement wasaccompanied by decreases in the expression of the AMPA receptorGluR2 in the postsynaptic densities, as well as decreased synapticBNDF-tyrosine kinase receptor B (TrkB) signaling (Roumier et al.,2004). These ndings suggest that microglialneuron interactioncan be involved in the regulation of synaptic functioning and neu-ral plasticity.

1.3. The role of the immune system and inammatory cytokines inneurogenesis

In addition to the biochemical and structural processes of syn-aptic plasticity, the brain is capable of plasticity in the whole neu-ronal level, i.e., via the formation of new neurons, a process termedneurogenesis (Deng et al., 2010; Leuner and Gould, 2010). In thehealthy brain, neurogenesis is restricted to two dened anatomicallocations in which stem cells can proliferate and differentiate intoneurons and migrate to their nal destination the subventricularzone of the lateral ventricle (SVZ), from which new neurons mi-grate to the olfactory bulb, and the subgranular zone of the hippo-campal DG (SGZ), from which new neurons migrate to the entireDG. Several lines of evidence indicate that neurogenesis plays animportant role in learning, memory and neural plasticity (Denget al., 2010; Leuner and Gould, 2010): (1) Neurogenesis increasesfollowing specic forms of learning and memory formation. (2)The rate of neurogenesis and hippocampal-dependent memory for-mation are positively correlated. (3) Conditions that increase mem-ory abilities, such as environmental enrichment and exercise, alsoincrease neurogenesis. (4) The level of hippocampal neurogenesisplays a role in determination of the hippocampus-dependent per-iod of memory. (5) Ablation of neurogenesis induces learning andmemory impairments.

The neurobiological basis for the role of neurogenesis in learn-ing and memory is probably related to the hyper-plasticity exhib-ited by these cells. Thus, young granule cells in the adulthippocampus exhibit substantially different active and passivemembrane properties than mature granule cells, making thesecells hyper-excitable. Moreover, associative long-term potentiationcan be induced more easily in young neurons than in mature neu-rons under identical conditions. Thus, newly generated neuronsexpress unique mechanisms to facilitate synaptic plasticity, whichmay be important for the formation of new memories (Ge et al.,2007; Schmidt-Hieber et al., 2004). Similarly, recently generatedadult-born olfactory interneurons undergo different experience-dependent synaptic modications compared with their pre-exist-ing mature neighbors and provide a possible substrate for adult

and Immunity xxx (2010) xxxxxxneurogenesis-dependent olfactory learning (Nissant et al., 2009).Several studies indicate that various components of the im-

mune system are involved in the process of neurogenesis.

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

2006). Further evidence for a role of microglia in neurogenesiswas provided using a culture system of the SVZ, in which neuropoi-

ior, a1.3.1. T cellsInitial evidence for a role of T cells in neurogenesis was pro-

vided by the demonstration that microglia activated by the cyto-kines IL-4 or low level of IFN-c, which are known to be producedby T-helper cells, promoted neurogenesis in vitro (Butovsky et al.,2006). More direct evidence was provided by the ndings thatT-cell decient mice (either SCID or nude) display a severe impair-ment in neurogenesis within the hippocampus, which can be res-cued by adaptive transfer of WT splenocytes (Ziv et al., 2006).Consistently, transgenic mice in which most T cells express aCNS-specic receptor were found to display enhanced neurogene-sis, whereas mice with transgenic excess of T cells with a receptorfor a non-CNS protein displayed reduced neurogenesis. Moreover,environmental enrichment, which markedly increased the numberof new neurons in the hippocampus of WT mice, induced theappearance of T cells in the parenchyma of the hippocampus, butit had no effect on neurogenesis in SCID mice (Ziv et al., 2006).

Two recent studies corroborated these ndings by demonstrat-ing that systemic depletion of CD4-positive T lymphocytes (byeither genetic or immunological means) led to signicantly re-duced hippocampal neurogenesis. The specicity of this effect toT helper cells was attested by the ndings that no such effectwas observed after depletion of CD8 or B cells, and that repopula-tion of a strain of mice which lacks both T and B cell function withCD4, but not with CD8 cells rescued the suppressed neurogenesis(Wolf et al., 2009a). Furthermore, specic peripheral T-cell activa-tion (by either antigen-induced arthritis in the knee joint or staph-ylococcus enterotoxin) was found to be associated with a transientincrease in hippocampal precursor cell proliferation and neurogen-esis (in contrast with innate immune activation by lipopolysaccha-ride (LPS), which caused neurogenesis suppression) (Wolf et al.,2009b).

1.3.2. MicrogliaMicroglia are now known to play both benecial, neuroprotec-

tive effects, as well as detrimental, neurotoxic effects under variousconditions. Although initially microglia were shown to be involvedin neurogenesis suppression, it is now clear that under quiescentconditions they play a role in supporting neurogenesis (Ekdahlet al., 2009; Hanisch and Kettenmann, 2007; Ziv and Schwartz,2008). Evidence for a benecial role of microglia in neurogenesiswas rst provided by showing that exposure of rats to environ-mental enrichment induces not only increased neurogenesis, butalso a signicant increase in the number of hippocampal microglia(probably due to increased microglial proliferation) (Ziv et al.,2006). The microglia in the enriched animals assumed a neuropro-tective phenotype, expressing MHC class II and the neurotrophicfactor insulin growth factor (IGF)-1 (which is known to promoteneurogenesis). The inuence of microglia on neurogenesis maybe related to their interactions with T cells, since in transgenicmice with excess CNS-specic T cells the increased neurogenesiswas attenuated by chronic treatment with the microglial inhibitorminocycline (Ziv et al., 2006). As mentioned above, in vitro data onT cellmicroglial interactions supports this notion, demonstratingthat microglia activated by the T cells-derived cytokine interferon(IFN)-c induce neuronal differentiation (Butovsky et al., 2006).Consistently, transgenic mice with brain-specic expression ofIFN-c display elevated levels of neurogenesis (Baron et al., 2008).In contrast with these ndings, a recent study demonstrated thatexposure of mice to spontaneous activity in a running wheel re-sulted in dramatic neurogenesis, with no signs of microglial prolif-eration or activation, and no indication of T cellmicroglialinteractions (i.e., no MHC class II expression or presence of T cells

R. Yirmiya, I. Goshen / Brain, Behavin the hippocampus) (Olah et al., 2009).Microglia were also shown to play a neurogenesis supportive

role in adrenalectomized rodents. This surgical manipulation,

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015etic cells continue to proliferate and differentiate, but progres-sively lose this ability with continued culture. In this model,neurogenesis can be rescued by co-culture with microglia ormicroglia-derived conditioned medium, indicating that microgliaprovide secreted factor(s) essential for neurogenesis (Waltonet al., 2006). Further work with adult precursor cell cultures dem-onstrated that activated microglia cells release soluble factors thatdirect the differentiation of neural precursor cells toward a neuro-nal phenotype (Aarum et al., 2003).

A role for innate immunity-like process in neurogenesis wasalso suggested by the nding that neural precursor cells (NPCs) ex-press toll-like receptor (TLR) 2 (which in the brain is usually asso-ciated with microglia and other innate immune cells). Moreover,signaling via this receptor was found to positively regulate NPCsdifferentiation (Rolls et al., 2007).

1.3.3. Pro-inammatory cytokines and prostaglandinsAlthough most studies in this area implicated pro-inamma-

tory cytokines in suppression of neurogenesis, there is evidencethat at least under some conditions TNFa may have a pro-neuro-genic role. Specically, exposure of neural stem cells (NSCs) toTNFa produced a doseresponse related increase in proliferation,with no effect on NSCs differentiation. The signaling cascade in-volved in this effect was found to depend on TNFa-inducedIKK-a/b-complex, which activates the NFjB pathway, which inturn activates the TGFb activated kinase-1 (TAK-1) signaling cas-cade (Widera et al., 2006). These ndings corroborate a previousstudy, demonstrating that i.p. injection of TNFa increases the pro-liferation of neural progenitors in the subventricular zone (Wuet al., 2000).

The nding that COX-2 inhibitors are potent suppressors ofneurogenesis (producing 4090% reductions in the number of pro-liferating neurons in the hippocampus and SVZ) strongly impli-cates prostaglandins in this process. Because COX-2 is notexpressed by the proliferating progenitors in vivo, and COX-2inhibitors do not affect the growth rate of cultured progenitor cells,it may be suggested that the effect of these drugs is indirect. SinceCOX-2 is highly expressed by resting microglia that closely associ-which eliminates endogenous glucocorticoids, resulted in triplingneurogenesis rate along with an 8-fold increase in the number ofactivated microglia. Interestingly, the number of activated (butnot resting) microglia highly correlated with the number ofBrdU-labeled new neurons. The activated microglia seemed to ex-ert their benecial effect via expression of the anti-inammatory/neuroprotective cytokine transforming growth factor-beta (TGF-b1), which was documented to be essential for adrenalectomy-in-duced increased neurogenesis (Battista et al., 2006). Consistentlywith the latter, TGF-b over-expression (induced by injection of anadenoviral vector expressing this anti-inammatory cytokine) in-creased neurogenesis in the subventricular zone (Mathieu et al.,2010b).

Additional evidence for a role of microglia in neurogenesis is de-rived from a study reporting that under quiescent conditionsmicroglia in the SVZ are in an intermediate state of constitutiveactivation compared with non-neurogenic cortical areas. More-over, the basal level of proliferation of SVZ microglia is much high-er than in surrounding forebrain areas. Although these ndingssupport an association between microglial activation and neuro-genesis, it should be noted that no such relationship was foundin the other neurogenic area of the brain (the SGZ) (Goings et al.,

nd Immunity xxx (2010) xxxxxx 7ate with the proliferating precursor cells, it is likely that COX-2inhibitors produce their effects by acting on these cells (Goncalveset al., 2010).

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,1.4. Possible mechanisms underlying the benecial effects of immuneprocesses in neurobehavioral plasticity

1.4.1. Neuroendocrineimmune interactionsEmotional arousal or mild stress reactions are an essential com-

ponent of many types of neurobehavioral plasticity. The formationof long-termmemory and neural plasticity depends on a process ofconsolidation, in which newmemories, which in the short term ex-ist in a relatively fragile state, undergo a process of stabilization.During this process, which depends on protein synthesis, memorybecomes much less amenable to interference (McGaugh, 2000).Consolidation of memories enables the modulation of memorystrength by endogenous processes activated by an experience.Such modulation is particularly relevant for memory associatedwith emotionally arousing experiences, which are modulated bystress-induced activation of the HPA axis (and the secretion of glu-cocorticoids), as well as the activation of the SNS in the peripheryand activation of monoaminergic neurotransmitter systems in thebrain. Specically, the secretion and actions of glucocorticoids, nor-epinephrine, dopamine and serotonin can facilitate memory con-solidation, and the actions of these compounds on memorystorage circuits can induce or strengthen LTP and its molecularsubstrate (McGaugh, 2000; Schwabe et al., 2010). Interestingly,both LTP and neurogenesis can also be strongly facilitated bymonoamines (Bliss et al., 1983; Brezun and Daszuta, 1999). Itshould be noted, however, that under severe stressful conditionsover-activation of the HPA axis and monoaminergic neurotrans-mission can disrupt memory consolidation (McEwen, 1999; Roo-zendaal, 2000; Schwabe et al., 2010) (see Section 2.4.2).

In view of the multiple levels of interactions between immuneparameters, the HPA axis and monoaminergic neurotransmission,these systems are excellent candidates for mediating the effectsof immune processes on memory functioning and neural plasticity.These interactions are bi-directional: on the one hand emotionaland stressful conditions along with the resultant HPA axis andmonoaminergic stimulation induce activation of immune-like pro-cesses, and on the other hand, immune-like processes play animportant role in mediating the neuro-hormonal stress responses.Most research on these interactions was conducted in the contextof the detrimental role of inammation in neurobehavioral plastic-ity. However, interactions between immunity, HPA and SNS activa-tion occur also with respect to mild emotional stimulation/stress,which promotes learning and plasticity. For example, the emo-tional stimulation that is an integral part of many cognitive pro-cesses (such as examination or public speaking in humans orvarious learning paradigms in rodents) is associated with cytokines(particularly IL-1) production (Brydon et al., 2005; Depino et al.,2004; Goshen et al., 2007; Heinz et al., 2003; Labrousse et al.,2009; Steptoe et al., 2007). On the other hand, cytokine production,particularly in the hippocampus and hypothalamus, is importantfor mild stress-induced activation of the HPA axis and SNS (Goshenand Yirmiya, 2009) (Fig. 1). As detailed below, under certain condi-tions these interactions may be benecial for memory andplasticity.

Many research studies examined the effects of endogenous(stress-induced) or exogenous glucocorticoids (GCs) administra-tion on hippocampal-dependent memory, neural plasticity andneurogenesis, reporting that similarly to the nding on theinvolvement of cytokines in these parameters the effects of GCscan be either benecial/facilitatory (particularly at low levels) ordetrimental/inhibitory (particularly at high levels) (Conrad et al.,1999; de Kloet et al., 1999; Kim and Diamond, 2002; McEwenand Sapolsky, 1995; Wolf et al., 2009b). The role of GCs in mediat-

8 R. Yirmiya, I. Goshen / Brain, Behaving the benecial effects of immunity was particularly examinedby testing the effects of GC receptor blockade on IL-1-inducedmemory improvement. Rats that were injected i.c.v. with IL-1b dis-

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015played improved contextual passive avoidance responses concom-itantly with increased corticosterone secretion, however, whenIL-1b was co-administered together with the GC antagonistRU486, the benecial effect on memory was eliminated, as wasthe increase in corticosterone levels (Song et al., 2003). Consistentwith these ndings, IL-1rKO mice, which displayed impaired mem-ory performance (Avital et al., 2003), also showed diminished cor-ticosterone secretion in response to mild stressors (Goshen et al.,2003), suggesting that impaired HPA axis activation may mediatethe poor memory performance of these mice.

A recent study has provided direct evidence for a dual role ofhippocampal glucocorticoids in neurogenesis, by demonstratingthat in vitro exposure of murine neuronal precursor cells to cortico-sterone induced either proliferation at low concentrations or celldeath at high concentrations. Further in vivo experimentationdemonstrated that specic peripheral activation of T cells, whichproduced a small increase in hippocampal corticosterone levels(1- to 2-fold over the physiological amount), increased neurogene-sis, whereas treatment with LPS, which induced a 5-fold increase inhippocampal corticosterone levels, produced neurogenesis sup-pression (Wolf et al., 2009b).

As mentioned above, learning, memory and LTP are importantlymodulated by monoamines, including NE, DA and 5-HT (McGaugh,2000). These neurotransmitters can not only inuence neurons,but can also inuence the production and secretion of inamma-tory mediators by both microglia and astrocytes, via activation ofspecic monoaminergic receptors expressed by these cells (Pocockand Kettenmann, 2007). The role of NE in glial-induced cytokinesecretion is complex: on the one hand acute exposure to NEin vivo induces microglial activation and secretion IL-1 (Johnsonet al., 2008; Maruta et al., 1997; McNamee et al., 2010a,b). In con-trast, endogenous and exogenous adrenergic stimulation duringinammatory conditions, including LPS administration, demyelin-ating and neurodegenerative diseases, results in decreased secre-tion of IL-1 and other inammatory mediators, and in theproduction of anti-inammatory cytokines, such as IL-10, IL-1ra,and IL-1 type II receptors (Feinstein et al., 2002; Heneka et al.,2002; McNamee et al., 2010a,b). To date, no studies directly as-sessed the involvement of monoamines in the benecial effectsof immune activation on memory functioning and neural plasticity.

1.4.2. Neuroglial interactionsNeural and behavioral plasticity result from specic patterns of

neuronal activity. It is now well established that both brain andperipheral immune processes can inuence, and be directly inu-enced by neuronal activity, providing a functional neuroimmunebasis for immune modulation of plasticity.

Similarly to the mechanisms underlying the effects of neuronson other neurons, neuroimmune interactions are mediated bythe secretion of neurotransmitters and neuromodulators fromaxon terminals, as well as indirectly by the secretion of hormonesfrom the pituitary and other endocrine glands, and their effects viaspecic receptors expressed by immune cells. For example, in theperiphery immune cells, including T, B, NK cells and monocytes/macrophages, express noradrenergic receptors, whose activationmarkedly alters the functioning of these cells (Nance and Sanders,2007). Macrophages/monocytes may also express nicotinic a-7receptors, allowing modulation of immune functioning by para-sympathetic activation (Tracey, 2002). Peripheral immune cellsalso express receptors for various neuropeptides and hormonesthat are derived from autonomic nervous system axon terminalsor from various endocrine systems (Besedovsky and Del Rey,2007; Friedman and Irwin, 1997), as well as receptors for the

and Immunity xxx (2010) xxxxxxmonoamines serotonin and dopamine (although in the peripherythese compounds are not secreted by nerve cells). Whereas theinvolvement of peripheral immune cells in normal neurobehavior-

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior, aR. Yirmiya, I. Goshen / Brain, Behaval plasticity is probably indirect, within the brain direct interac-tions between astrocytes, microglia, neurons and neural precursorcells are critical for various forms of plasticity. Furthermore, astro-cytes, microglia and possibly other brain cells (e.g., endothelialcells and perivascular macrophages) also interact with peripher-ally-derived immune cells located in various structures surround-ing the brain, including the perivascular space, meninges andchoroid plexus. These interactions activate myeloid and T cells,

Fig. 1. A systemic model of the benecial role of immune processes in behavioral and nehippocampal circuits by glutamatergic inputs that originate mainly in multiple cortical(particularly of the amygdala and hypothalamus), inducing a mild stressful condition, wThe peripheral organs that are the targets of these systems (e.g., the adrenal glad, heart, bthat culminate in stimulation of receptors for glucocorticoids, norepinephrine, dopamconsolidation, neural plasticity and neurogenesis. Furthermore, these inputs induce the pin the hippocampus, as well as in other brain areas (such as the hypothalamus and braincytokines, in turn further activate the HPA axis and SNS, thus participating in a brain-to

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015nd Immunity xxx (2010) xxxxxx 9which in turn, feedback and send regulatory signals affecting thebrain cells (Kipnis et al., 2008; Schwartz and Shechter, 2010). Ageneral scheme of these interactions is presented in Fig. 2, andthe hypothesized immune-like roles of astrocytes and microgliain memory and neural plasticity are discussed below.

1.4.2.1. Astrocytes. Although astrocytes are not considered as im-mune cells, they do have some immune-like properties, including

ural plasticity. Learning, memory and synaptic plasticity involve neural activation ofareas. Long-term memory consolidation also requires emotional (limbic) activationhich in turn results in HPA axis and sympathetic nervous system (SNS) stimulation.lood vessels and gastrointestinal (GI) tract), in turn, send afferent inputs to the brainine and serotonin on hippocampal cells. These inputs are critical for memory

roduction of IL-1, and possibly other cytokines, chemokines and immune mediatorsstem) that are critically important for neurobehavioral plasticity. Moreover, these-body-to-brain reverberating feedback loops.

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,10 R. Yirmiya, I. Goshen / Brain, Behavtheir ability to respond to inammatory cytokines (particularly toIL-1), to secrete inammatory cytokines (particularly TNFa andIL-6) (Lee et al., 1993) and to phagocytose cellular processes anddebris (Bechmann and Nitsch, 1997). The immune-like nature ofastrocytes is particularly notable considering that recent studiesindicate that these cells are not merely the supportive cells ofthe brain, but that they also play an important integral role in neu-ral and synaptic functioning (Halassa and Haydon, 2010; Henne-berger et al., 2010; Volterra and Meldolesi, 2005). Specically,astrocytic processes ensheath most synapses in the brain, and ex-press receptors for several neurotransmitters. Signaling via thesereceptors evokes elevations in astrocytic Ca2+ concentration,resulting in the regulated secretion of various gliotransmitters,which modulate neuronal excitability and synaptic strength (Ha-lassa and Haydon, 2010; Perea and Araque, 2007). Astrocytes-to-neurons communication also plays a critical role in neurobehavior-

Fig. 2. A molecular/cellular model of the benecial role of immune processes in behavioinduces LTP, the external glutamatergic, monoaminergic and adrenocortical input, along whippocampal neurons, but also hippocampal microglia and astrocytes (blue arrows). Sigvarious mediators. For example, glutamatergic activation, along with purinergic ATP signmediators) by microglia (red arrows). IL-1 can in turn further activate astrocytes, inducsynaptic plasticity, such as D-serine, BDNF, TNFa and additional glutamate (green arrowsby astrocytes, which are important for long-term memory consolidation. Microgliafunctioning and neural precursor cells (NPCs) (which underlie hippocampal neurogenesisIL-1 can also directly inuence neurons by upregulating NMDA receptor functioning. Imneuronal-derived factors, including GABA, CD200 and fractalkine. Microglial expressionimportant role in learning and neurogenesis, possibly via IL-4- and IFNc-medited interactactivate endothelial cells, which produce various trophic factors, such as VEGF and IGF-1(For interpretation of the references to colour in this gure legend, the reader is referre

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015and Immunity xxx (2010) xxxxxxal plasticity (Bains and Oliet, 2007). For example, female mice inwhich the transcription nuclear factor-kappa B (NFjB) was inhib-ited specically in astrocytes displayed decits in learning,memory and LTP (Bracchi-Ricard et al., 2008). Moreover, pharma-cologic blockade of astrocytic glutamate uptake in rats was alsofound to impair spatial memory (Bechtholt-Gompf et al., 2010),and motor skill learning was reported to be associated with astro-cytic hypertrophy, which was reversed in the absence of continuedtraining (Kleim et al., 2007). In addition, several studies establishedthat astrocytic energy metabolism is involved in memory consoli-dation and in the inuence of noradrenergic mechanisms on hip-pocampal-dependent memory (Gibbs et al., 2006). Thesemetabolic effects may be related to the ndings that IL-1 facilitatesglucose uptake and the astrocytic production of lactate (Del Reyet al., 2006; Vega et al., 2002), which is important for long-termmemory consolidation (McNay et al., 2000).

ral and neural plasticity. During learning or high frequency stimulation (HFS) thatith glutamate secreted from neurons within the hippocampus, can activate not only

naling via specic receptors expressed on these glia cells induces the production ofaling can direct the production and secretion of IL-1 (as well as other inammatorying the secretion of several compounds that are critical for memory formation and). IL-1 has also been shown to facilitate glucose uptake and the production of lactateand astrocytes also secrete various compounds that directly inuence neuronal), including BDNF, IGF-1, TGFb, and low levels of TNFa and PGE2. Microglial-derivedportantly, the production of IL-1 and other glial mediators is tightly regulated byof IL-1, MHC class II and various chemokines can inuence T cells, which play anions with microglia and meningeal myeloid cells (light blue arrows). Finally, IL-1 canthat are important for memory, neural plasticity and neurogenesis (purple arrows).d to the web version of this article.)

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior, aA role for astrocytes in LTP was demonstrated in several studiesthat implicated astrocytic GFAP and S-100b in regulation of LTP(Nishiyama et al., 2002; Tanaka et al., 2002). Furthermore, in bothhippocampal cell cultures and slices, the induction of LTP wasfound to depend on the presence of astrocytes and the secretionof D-serine by these cells, which in turn binds to the glycine-siteon neuronal NMDA receptors and enables their critical role inLTP (Henneberger et al., 2010; Yang et al., 2003). Astrocytes alsomediate other forms of plasticity, such as homeostatic synapticscaling following prolonged inhibition of neuronal activity, viasecretion of the pro-inammatory cytokine TNFa (Kaneko et al.,2008; Stellwagen and Malenka, 2006), a known synaptic strengthenhancer, which increases the surface expression of AMPA gluta-matergic receptors (Beattie et al., 2002).

To test directly the role of astrocytes, and their interaction withimmune mechanisms, in neurobehavioral plasticity we recentlyinvestigated the involvement of astrocytes in memory and LTP,using IL-1rKO mice as a model of impaired learning and synapticplasticity. NPCs derived from either WT or IL-1rKO neonatal mice,were labeled with BrdU and transplanted into the hippocampus ofeither IL-1rKO or WT adult host mice. Transplanted NPCs showedlong-term survival and differentiated into astrocytes (expressingGFAP and S100b), but not to neurons. Moreover, several weekspost-transplantation IL-1rKO mice transplanted with IL-1rKO cellsor sham operated displayed severe memory disturbances and amarked impairment in LTP, however, IL-1rKO mice transplantedwith WT NPCs (expressing IL-1R1) displayed a complete rescueof the impaired memory functioning, as well as partial restorationof LTP (Ben Menachem-Zidon et al., in press). These ndings notonly indicate that astrocytes play a critical role in memory func-tioning and LTP, but they specically implicate astrocytic IL-1 sig-naling in these processes.

1.4.2.2. Microglia. Within the brain, the prominent representativeof the immune system is the microglia cell. These resident, macro-phage-like cells, which comprise about 15% of brain cells, wereshown to play a critical role in developmental neuronal death inthe hippocampus (Wakselman et al., 2008), as well as in the clear-ance of apoptotic neurons (Takahashi et al., 2005). Moreover, rest-ing microglial processes were found to be highly motile(Nimmerjahn et al., 2005), and to continuously and dynamicallymonitor and respond to the functional status of synapses (Wakeet al., 2009).

In the normal, quiescent brain, microglia are controlled by bothintrinsic and extrinsic systems. Their functioning is modulated byneuronal activity, via neurotransmitter receptors, particularly glu-tamate, as well as specic regulatory molecules, such as fractalkineand CD200 (Biber et al., 2007; Hung et al., 2010). They are alsomodulated by astrocytes, e.g., via the secretion of ATP and its sig-naling via P2X receptors, particularly the p2X7 receptors, whoseactivation together with glutamatergic inputs is essential for thesecretion of some cytokines, such as IL-1 (Ferrari et al., 2006).Microglia can also directly inuence neuronal activity. For exam-ple, microglial-derived IL-1b can facilitate NMDA receptor activa-tion in neurons via ceramide-src pathway, as detailed below(Viviani et al., 2003; Yang et al., 2005). Microglial-derived IL-1can also activate endothelial cells, which produce various trophicfactors, such as VEGF and IGF-1 that are important for memory,neural plasticity and neurogenesis (Anderson et al., 2002; Caoet al., 2004).

Despite recent evidence for neuralmicroglial interactions, andthe ndings that microglia secrete various plasticity related com-pounds (e.g., glutamate, BDNF and other neurotrophins, as well

R. Yirmiya, I. Goshen / Brain, Behavas inammatory cytokines such as TNFa and IL-1), there is minimalevidence for a direct role for microglia in learning, memory andLTP. The only exception is recent work demonstrating an involve-

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015ment of microglia in the LTP of C-ber-evoked eld potentials inspinal dorsal horn. Specically, HFS-induced LTP was convertedto LTD when rats were pre-treated with microglial inhibitors, suchas minocycline (Zhong et al., 2010). Moreover, spinal LTP wasfound to depend on the activation of microglial Src-family kinases(SFKs), evidenced by the ndings that phosphorylated SFK was re-stricted to microglia and was up-regulated by HFS. Moreover, SFKsinhibitors also converted LTP to LTD. Microglial-derived TNFaseemed to play a role in spinal LTP, since the inhibitory effects ofminocycline on spinal LTP were reversed by spinal application ofTNFa, and HFS failed to induce LTP in TNF receptor-1 knockoutmice or in rats pre-treated with TNFa neutralization antibody(Zhong et al., 2010). HFS-induced spinal LTP was also found to de-pend of the activation of microglial P2X7 receptors (Chu et al.,2010), which are critical for the production and secretion of IL-1.Indeed, blockade of P2X7 receptors by various methods preventedthe induction of spinal LTP, as well as the production of IL-1, andadministration of IL-1ra also prevented this type of LTP (Chuet al., 2010). Microglia were also implicated in the induction ofLTP of C-ber-evoked eld potentials in the spinal dorsal horn byexposure to ATP. LTP in this paradigm was found to be criticallydependent on the activation of P2X4 receptors, which are exclu-sively expressed by microglia. Moreover, following LTP inductionmicroglial expression of these receptors was upregulated and theysignaled via the p38 mitogen-activated protein (MAP) kinase(Gong et al., 2009).

These ndings are not only important for explaining injury-associated sensitization in pain pathways that can contribute tochronic neuropathic pain, but they may also directly demonstratethe importance of microglia for neural plasticity in general.

1.4.3. Immune-induced alterations in plasticity-related molecular andcellular processes

Memory and LTP are accompanied by molecular and morpho-logical changes within the participating neurons, includingchanges in intracellular signaling, expression of immediate-earlyand then structural genes, changes in receptor presentation andspine size modication. Some of these plasticity-related processeswere shown to be inuenced by immune mechanisms, as detailedbelow:

1.4.3.1. The immediate early genes (IEG) activity-regulated cytoskel-eton-associated protein (Arc). Hippocampal-dependent learning in-duces the expression of Arc, possibly in relation to NMDA receptoractivation and secretion of BDNF (another protein essential formemory consolidation, see below). Moreover, Arc levels are corre-lated with performance in this task (Guzowski et al., 2001). A rolefor Arc in the benecial effects of IL-1 on memory has been re-cently suggested by demonstrating that basal hippocampal Arcexpression is lower in IL-1raTGmice (which display poor memory),and the levels of hippocampal Arc protein in these mice are not in-creased following exposure to novelty, as they do in WT controlmice (Spulber et al., 2009). As will be discussed below, chronic highlevels of IL-1 produce opposite effects on Arc (i.e., suppression ofhippocampal expression), which are associated with suppressionof memory (Frank et al., 2010; Hein et al., 2010).

1.4.3.2. Synaptic proteins. Synaptic proteins, which may contributeto memory functioning and plasticity, have been suggested tounderlie the role of T cell immunity in memory functioning. Specif-ically, T cell decient SCID mice transplanted with bone marrowisolated from SCID mice displayed severe memory decit in the

nd Immunity xxx (2010) xxxxxx 11water maze, as well as markedly reduced levels of expression oftwo presynaptic proteins (Syt10 and Cplx2). In contrast, SCID micetransplanted with WT-derived bone marrow displayed normal

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior,memory functioning and expression of the two proteins (Ron-Harel et al., 2008).

1.4.3.3. Neurotrophins. Neurotrophic factors, such as BDNF, IGF-1,NGF, GDNF, and VEGF are essential regulators of various forms ofneural and cellular plasticity, not only during development, butalso in the adult organism (McAllister et al., 1999). All of these fac-tors can be secreted by several types of immune cells, including Tcells, macrophages, mast cells and microglia (Elkabes et al., 1996;Nakajima et al., 2001), particularly after exposure of these cellsto various cytokines, including IL-1, IL-6, and TNF-a (Gadientet al., 1990; Schulte-Herbruggen et al., 2005). Once secreted, theseneurotrophic factors can serve as mediators of the benecial effectsof immunity on neural and behavioral plasticity.

1.4.3.3.1. BDNF. Most research on the relations between im-mune mechanisms and neurotrophins in general, and with respectto neurobehavioral plasticity in particular, focused on brain de-rived neurotrophic factor (BDNF). This is understandable given thatthe production and signaling of BDNF via the TrkB receptor hasbeen implicated in almost every aspect of neural and behavioralplasticity, including hippocampal-dependent memory (Barnesand Thomas, 2008; Heldt et al., 2007), LTP (Lu et al., 2008) and neu-rogenesis (Li et al., 2008). BDNF and its Trk-B receptors are ex-pressed not only by neurons, but also by astrocytes andmicroglia. For example, BDNF secreted by activated microglia caninuence neuronal functioning (e.g., produce neuroprotective ef-fects (Batchelor et al., 1999)), but it can also act in an autocrinemanner to promote proliferation and survival of microglia (Elkabeset al., 1996; Zhang et al., 2003). BDNF, acting via Trk-B receptors onmicroglia can also induce sustained elevation of intracellular Ca2+elevation in these cells (which in turn can have anti-inammatoryeffects) (Mizoguchi et al., 2009).

Several lines of evidence suggest that enhanced BDNF signalingunderlies the benecial inuence of immune processes on learning,memory, LTP and neurogenesis: (1) Transgenic mice with excess ofT cells directed towards a brain self antigen, which exhibit en-hanced learning, memory and hippocampal neurogenesis, exhibitelevated levels of hippocampal BDNF expression (Ziv et al.,2006). (2) Cop-1 vaccination, which abolishes the learning decitsinduced by neurotransmitter imbalance via boosting T cell activity,was found to induce the production of BDNF by T cells. This re-sponse was even greater when the Cop-1 reactive T cells encoun-tered the CNS-related self-antigen myelin basic protein, but notthe non-CNS-related ovalbumin, further suggesting the involve-ment of BDNF in the benecial effect of immune stimulation onlearning and memory (Kipnis et al., 2004). (3) The secretion ofBDNF by cultured astrocytes is markedly enhanced by IL-4 (whichseems to be essential for the benecial effects of T cells on learningand memory) (Derecki et al., 2010). Moreover, following watermaze training, mice exhibit increased expression of BDNF, whereasIL-4 decient mice, which display impaired performance in thistask, do not show such an increase (Derecki et al., 2010). (4) Sys-temic depletion of CD4-positive T lymphocytes, which led to signif-icantly reduced hippocampal neurogenesis and impaired reversallearning in the Morris water maze, also decreased BDNF expressionin the brain (Wolf et al., 2009a). (5) Memory-impaired mice withtransgenic expression of IL-1ra exhibit lower levels of hippocampalBDNF. Moreover, whereas in WT mice learning of a novel environ-ment signicantly increased hippocampal BDNF levels as well asERK1/2 activation (which underlies BDNFs effects on memory con-solidation) no such increases were shown in IL-1raTG mice (Spul-ber et al., 2009). (6) The ibuprofen-induced inhibition ofprostaglandin synthesis, which disturbs learning, memory and

12 R. Yirmiya, I. Goshen / Brain, BehavLTP, is also accompanied by diminished production of PGE2 andBDNF following spatial learning and LTP. Moreover, elevation ofBDNF levels by prior exercise in a running wheel increased endog-

Please cite this article in press as: Yirmiya, R., Goshen, I. Immune modulation of(2010), doi:10.1016/j.bbi.2010.10.015enous BDNF levels sufciently to reverse the detrimental effect ofibuprofen on spatial learning and LTP, and restored a parallel in-crease in LTP and learning-related BDNF and PGE2 (Shaw et al.,2003).

1.4.3.3.2. Insulin growth factor (IGF)-1. IGF-1 was shown to beinvolved in memory, plasticity and neurogenic processes, espe-cially in the aging brain (Maher et al., 2006; Sonntag et al.,2005). For example, IGF-1 deciency results in both memory andLTP impairments (Trejo et al., 2007). A role for IGF-1 in the bene-cial cognitive effects of T cells was suggested by the ndings that inaddition to their decit in BDNF, SCID mice also exhibit reducedexpression of the igf-1 gene, concomitantly with disturbances inmemory functioning and neurogenesis. Transplantation of WTbone marrow rescued the impairment in IGF-1 expression and alsorestored neurocognitive functioning (Ron-Harel et al., 2008).

1.4.3.4. Glutamatergic neurotransmission. Changes in glutamatergicneurotransmission may also mediate the benecial effects of IL-1signaling on memory and LTP, as suggested by the ndings that:(1) microglial-derived IL-1 can alter neuronal functioning byinducing Src family of kinases-mediated tyrosine phosphorylationof N-methyl-D-aspartate (NMDA) receptor NR2B subunit inhippocampal neurons (Viviani et al., 2003; Yang et al., 2005). ThisIL-1-mediated phosphorylation upregulates the function of NMDAreceptors, resulting in greater stimulation-induced increases inintracellular Ca2+ in neurons, which is a critical step in the forma-tion of LTP and long-term memory (Viviani et al., 2003). Althoughthis was hypothesized to mediate the neurotoxic effects of IL-1,this pathway may also mediate the benecial effects of low levelsof IL-1 on memory and LTP (Coogan et al., 1999). (2) Memory im-paired IL-1raTGmice display lower basal levels of GluR1 and GluR2expression (Spulber et al., 2009), though no difference in GluR1levels was observed in IL-1rKO mice (Goshen et al., 2009). Thesendings suggest that constitutive hippocampal IL-1 productionmay be important for maintenance of glutamatergic neuro-transmission.

1.4.3.5. Changes in dendritic morphology. The accumulation of manystudies from the last decade suggests that morphological changesin dendritic spines may be related to memory consolidation andstorage (Segal, 2005). For example, spatial learning increases thenumber and density of spines in the DG (OMalley et al., 2000),as well as the number of mushroom shaped spines in the CA1 re-gion (Hongpaisan and Alkon, 2007). LTP also induces the formationof new spines (e.g., Engert and Bonhoeffer, 1999). Consistently,memory and plasticity deciencies are accompanied by spine sizereduction (von Bohlen und Halbach et al., 2006), whereas memoryimprovement is accompanied by spine size increase (Zhou et al.,2008).

In a recent study (Goshen et al., 2009), we reported that out ofseveral memory and plasticity-related molecular and morphologi-cal parameters that were investigated in the brains of IL-1rKO andWT mice, reduction in spine size was the only parameter that dif-fered between the two strains. This nding suggests that normallyIL-1 is involved in regulation of spine size, and therefore reducedspine size in IL-1rKO mice may contribute to their decits in mem-ory and LTP. Alternatively, IL-1 signaling may be needed to pro-mote learning and LTP-induced increases in spine size. Thishypothesis is consistent with the impairments in learning andLTP produced by acute administration of IL-1ra (Goshen et al.,2007; Ross et al., 2003; Schneider et al., 1998; Yirmiya et al.,2002). Interestingly, IL-1rKO mice that were raised in an enrichedenvironment (EE) did not differ in spine size from their WT con-

and Immunity xxx (2010) xxxxxxtrols. The marked increase in spine size in enriched IL-1rKO mice,despite the persistent lack of IL-1 signaling in these mice, suggeststhat mechanisms other than IL-1 signaling are also responsible for

learning, memory, neural plasticity and neurogenesis. Brain Behav. Immun.

ior, a2.1.1.1. Effects of inammatory cytokines. Because the immune re-sponse to medical conditions and the sickness behavior thataccompanies them are mediated by pro-inammatory cytokines,many research groups sought to directly assess the involvementof pro-inammatory cytokines in memory processes. The data col-lected in these studies, specically regarding the cytokines IL-1, IL-6, and TNFa, will be presented in this section.

2.1.1.1.1. IL-1. (i) Effects of IL-1 administration on learning andmemory: The detrimental effect of IL-1 on memory functioningwas rst demonstrated by showing that i.c.v. administration ofIL-1b 1 h before the beginning of spatial water maze trainingcaused a transient memory impairment in the rst trial of the fol-lowing day. When IL-1b was injected immediately before training,no effect was found, suggesting that IL-1b does not affect theacquisition of spatial memory, but rather the retention of thisspine size regulation, and that EE activates mechanisms that by-pass the need for IL-1 in controlling spine size.

It should be noted that the reported effect of IL-1 signaling de-ciency on spine size, as well as some of the behavioral and physi-ological data, was obtained using genetically manipulated mice.Because in these animals the modied signaling is presentthroughout their life span, from conception to maturity, it mayinuence memory and plasticity either by affecting brain develop-ment in a way that will alter its functioning in adulthood, and/orby directly inuencing ongoing memory and plasticity processesin the adult brain at the time of testing. Thus, the effects of IL-1on spine size may also be exerted during pre- or peri-natal devel-opment. Indeed, we have shown that exposure to IL-1ra in uterocan produce memory decits similarly to those observed in IL-1rKO mice (Goshen et al., 2007).

2. The role of the immune system in learning, memory, neuralplasticity and neurogenesis under inammatory conditions

2.1. Detrimental effects of immune activation on learning and memory

2.1.1. Studies in experimental animalsExposure to pathogens that stimulate the immune system re-

sults in altered memory performance, as part of the general sick-ness behavior syndrome. In particular, viral, bacterial, andparasitic infections, as well as exposure to viral coat proteins orbacterial endotoxin, has been shown to produce impaired memoryfunctioning in rodents, evidenced using various paradigms forassessment of different forms and phases of learning and memory,including the water maze, active and passive avoidance, and fearconditioning tests (e.g., Braithwaite et al., 1998; Cunninghamet al., 2009; Gibertini et al., 1995a,b; Kamitani et al., 2003; Leeet al., 2000; Li et al., 2004; Murray et al., 1992; Pugh et al., 2000,1998; Shaw