Activation of innate immunity in the CNS trig - gers neurodegeneration through a Toll - like receptor 4 - dependent path - way

Activation of innate immunity in the CNS triggersneurodegeneration through a Toll-like receptor4-dependent pathwaySeija Lehnardt*, Leon Massillon†, Pamela Follett†, Frances E. Jensen†‡, Rajiv Ratan*‡, Paul A. Rosenberg†‡,Joseph J. Volpe†‡, and Timothy Vartanian*‡§

*Department of Neurology, Beth Israel Deaconess Medical Center, †Department of Neurology, Children’s Hospital, and ‡Center for Neurodegeneration andRepair and the Program in Neuroscience, Harvard Medical School, Boston, MA 02115

Communicated by Richard L. Sidman, Harvard Medical School, Boston, MA, May 1, 2003 (received for review February 14, 2003)

Innate immunity is an evolutionarily ancient system that providesorganisms with immediately available defense mechanismsthrough recognition of pathogen-associated molecular patterns.We show that in the CNS, specific activation of innate immunitythrough a Toll-like receptor 4 (TLR4)-dependent pathway leads toneurodegeneration. We identify microglia as the major lipopoly-saccharide (LPS)-responsive cell in the CNS. TLR4 activation leads toextensive neuronal death in vitro that depends on the presence ofmicroglia. LPS leads to dramatic neuronal loss in cultures preparedfrom wild-type mice but does not induce neuronal injury in CNScultures derived from tlr4 mutant mice. In an in vivo model ofneurodegeneration, stimulating the innate immune response withLPS converts a subthreshold hypoxic-ischemic insult from no dis-cernable neuronal injury to severe axonal and neuronal loss. Incontrast, animals bearing a loss-of-function mutation in the tlr4gene are resistant to neuronal injury in the same model. Thepresent study demonstrates a mechanistic link among innateimmunity, TLRs, and neurodegeneration.

microglia � neuronal injury � lipopolysaccharide � infection

Systemic infection is associated with sustained worsening inmany diseases of the CNS, yet the molecular and cellular

relationship between infection outside the CNS and potentialneuronal loss within the CNS is elusive. Activation of microglia,bone marrow-derived macrophage-like cells that function as theresident immune defense system of the brain (1), is a charac-teristic feature of most neurodegenerative diseases includingAlzheimer’s disease, Parkinson’s disease, multiple sclerosis,AIDS dementia complex, and amyotrophic lateral sclerosis aswell as ischemia and posttraumatic brain injury (2–4). Neuro-toxicity induced by �-amyloid or HIV proteins in mixed CNScultures depends on the presence and activation of microglia (5,6). Liberatore et al. (7) demonstrated in vivo that microglialinducible nitric oxide synthase plays a crucial role in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced dopami-nergic neurodegeneration in the MPTP mouse model of Par-kinson’s disease.

The evolutionarily ancient innate immune system provides thefirst line of host defense against a large variety of pathogens andalso controls many aspects of the adaptive immune response (8).Cells of the innate immune system recognize invariant molecularstructures of pathogens termed pathogen-associated molecularpatterns through a series of genetically conserved and stablecell-surface receptors related to the Drosophila gene toll that thusare referred to as Toll-like receptors (TLRs) (9).

TLR4 functions as the signal-transducing receptor for theendotoxin lipopolysaccharide (LPS) (10), which is a majorcomponent of the outer membrane of Gram-negative bacteria.LPS binds to the serum protein LPS-binding protein and thesoluble or glycosylphosphatidylinositol-anchored CD14. Thiscomplex in turn binds to TLR4 (11) and initiates an intracellularsignaling pathway that regulates gene expression through dere-

pression of the transcriptional activator nuclear factor �B (12).To respond efficiently to LPS, TLR4 requires an accessoryprotein, MD-2, that binds to the extracellular domain of TLR4and enhances its surface expression (13).

In the CNS, constitutive expression of TLR4 and CD14transcripts was described in distinct anatomical areas of the brain(14, 15). We recently showed that microglia but not astrocytes oroligodendrocytes express TLR4 and that TLR4 is required forLPS-induced oligodendrocyte death in vitro (16).

The present study elucidates the mechanistic link amongactivation of innate immunity, TLRs, and neuronal injury in theneonatal CNS. We show that microglia are the major cells in theCNS that express TLR4 and that cortical neurons do not expressthis receptor. Fluorescently tagged LPS binds to microglia butnot to neurons. LPS-induced neuronal cytotoxicity is not cell-autonomous but rather requires microglia. In mixed CNS cul-tures prepared from mice lacking functional TLR4, LPS does notinduce death of cortical neurons. Finally, we demonstrate thatTLR4 is necessary for neuronal injury triggered by LPS incombination with a subthreshold hypoxic-ischemic insult in vivo.

Materials and MethodsAnimals. C.C3H-Tlr4lps-d (lpsd�tlr4 mutant) and BALB�cJ (wild-type) mice were provided by The Jackson Laboratory.

RT-PCR. Relative levels of TLR4 and CD14 mRNA in neuronsand microglia were determined by RT-PCR as described (16).

Cell Culture. Primary cultures of cortical neurons were generatedfrom forebrains of embryonic day-17 (E17) BALB�cJ mice, Lpsd

mice, or Sprague–Dawley rats (17).Cortices were triturated and dissociated with papain in Earle’s

balanced salt solution for 5 min at 37°C, resuspended in 0.25%trypsin inhibitor�0.25% BSA in Earle’s balanced salt solution(GIBCO), and incubated at 37°C for 5 min. Cells were harvestedby centrifugation at 1,000 � g for 5 min. Cells (1 � 106) in MEM(GIBCO) plus 10% FBS were plated onto poly-D-lysine-coatedglass slides (BD Biosciences). Cultures were grown in 5% CO2at 37°C. Immediately after plating, staining revealed 90–95%purity for neurons. To obtain mixed glia, cells were maintainedfor 10 days until a mixed glial layer was achieved.

Spinal cord from E17 mice and E13 chicken or dorsal rootganglion cells from E13 chicken were triturated and dissociatedwith trypsin for 20 min at 37°C, passed through 70-�m meshfilters, and plated on poly-D-lysine-coated glass coverslips inDMEM plus 10% FBS.

Purified oligodendrocytes, microglia, and astrocytes were

Abbreviations: MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; TLR, Toll-like recep-tor; LPS, lipopolysaccharide; En, embryonic day n; Pn, postnatal day n; NeuN, neuronalnuclei; HI, hypoxia-ischemia; DAPI, 4�,6-diamidino-2-phenylindole.

§To whom correspondence should be addressed at: Harvard Institutes of Medicine, 77Avenue Louis Pasteur, Boston, MA 02115. E-mail: [email protected].

8514–8519 � PNAS � July 8, 2003 � vol. 100 � no. 14 www.pnas.org�cgi�doi�10.1073�pnas.1432609100

generated from forebrains of 2-day-old Sprague–Dawley rats asdescribed (16, 18).

Immunfluorescence Microscopy. Cells were fixed and stained withantibodies against neuronal nuclei (NeuN), neurofilament,microtubule-associated protein 2, glial fibrillary acidic protein(Chemicon), O4 (American Type Culture Collection), andF4�80 (Serotec) as described (16). Secondary antibodies werefrom Jackson ImmunoResearch. Microglia were stained withisolectin IB4 and Alexa 488-conjugated LPS (Molecular Probes).

LPS in Combination with Hypoxia-Ischemia (HI) in Vivo. LPS treat-ment and HI were carried out as described (19) with somealterations. BALB�cJ and lpsd mouse pups received a single i.p.dose of LPS (Escherichia coli 0111:B4, List Biological Labora-tories, Campbell, CA; 0.3 mg�kg) or vehicle (PBS). This dose wasused to simulate a subclinical infection, and it induced no otherapparent impairment in the animals. To induce HI in combina-tion with LPS treatment, LPS or vehicle was administered to7-day-old mice. One hour later pups were subjected to unilateralHI. Ligation of the right common carotid artery was carried outunder ether anesthesia. After isolation of the artery, it wascauterized and bisected. The incision was sealed with nylonsutures. Mice were allowed to recover for 1.5 h and then wereexposed to 7.7% oxygen in nitrogen at 33°C for 30 min beforebeing returned to room air. The rectal temperature was mea-sured before the injections, before and after surgery, and afterHI was induced.

On postnatal day 11 (P11), mice were anaesthetized andperfused intracardially with PBS followed by 4% paraformalde-hyde in PBS. Forebrains were postfixed in 4% paraformaldehydein PBS for 4 h and then cryoprotected in 30% sucrose. Coronalsections (15 �m) were cut through the brain with a cryostat.Immunohistochemical analysis was performed as described (16)by using antibodies against neurofilament or NeuN.

ResultsLPS Induces Loss of Axons, Oligodendrocytes, and Microglia in MixedCNS Cultures. To investigate the effect of LPS on different CNScell types, we generated mixed CNS cultures from rat forebrainsand treated them with 10 �g�ml LPS in combination with 20 mMKCl for 5 days. Cells then were stained with antibodies againstneurofilament, O4, glial fibrillary acidic protein, or isolectin B4to analyze the effects of LPS on axons, oligodendrocytes, astro-cytes, and microglia, respectively (Fig. 1). Because extracellularpotassium enhances LPS-induced cytotoxicity (20), LPS treat-ment of cell cultures in further experiments was combined withthe addition of 20 mM KCl. Control cultures were treated withKCl alone. Data were analyzed with respect to the total numberof cells in culture as assessed by parallel 4�,6-diamidino-2-phenylindole (DAPI) staining. LPS plus 20 mM KCl induced adramatic reduction in numbers of axons, oligodendrocytes, andmicroglia. Astrocyte numbers were not altered by LPS toxicity.LPS-treated astrocytes showed a hypertrophic shape comparedwith astrocytes under control conditions. Neuronal, oligoden-drocyte, and microglial cultures treated with KCl alone showedno cell damage or cell loss when compared with untreatedcultures.

The Cytotoxic Effect of LPS on Neurons Does Not Depend on AnimalSpecies or Neuronal Type. We next addressed the question ofwhether the effects of LPS on neurons depends on the neuronalcell type or species. We prepared neuronal cultures from mouseand chicken spinal cord and chicken dorsal root ganglion cells.Because dorsal root ganglion cell cultures do not contain glia, weadded mixed glial cells from rat to these cultures before treat-ment with LPS to obtain comparable conditions. Cultures weretreated with 10 �g�ml LPS and 20 mM KCl for 5 days and stained

with antibodies against microtubule-associated protein 2, neu-rofilament, and DAPI to analyze cytotoxic effects on neurons(Fig. 2). All LPS-treated cultures showed striking damage andloss of axons and dendrites. These experiments show that thetoxic effect of LPS on neurons is a general phenomenon,independent of neuronal subtype or species.

The Effect of LPS on Neurons Is Not Cell-Autonomous but RequiresMicroglia. Treatment of neuron�glia cocultures with LPS leads toloss of neurons (21). However, it is not clear whether thisphenomenon is cell-autonomous.

We generated purified cultures of cortical neurons. Thesecultures were used for studies only if the staining for oligoden-drocytes, astrocytes, and microglia demonstrated �5% nonneu-ronal cells. Purified cultures of cortical neurons were treatedwith 10 �g�ml LPS and 20 mM KCl for 2 days and then stainedwith antibodies against NeuN and microglia (Fig. 3A). LPS hadno effect on neuronal survival in the absence of microglia. Incontrast, when purified microglia were added to highly enrichedneurons, we observed an almost complete loss of neurons. Thestatistical analysis of the ratio of NeuN� cells per mm2 confirmedthis result (Fig. 3B). Microglial cells themselves did not survivetreatment with LPS. These data indicate that the toxic effect ofLPS on neurons is cell-extrinsic and requires microglia.

Neurons Do Not Express TLR4 or CD14 and Do Not Bind LPS in Vitro.Microglia express TLR4 and CD14 (16, 22). Both receptors arerequired for the effects of LPS on circulating monocytes (9). Our

Fig. 1. LPS induces axonal, oligodendrocyte, and microglial death. MixedCNS cultures from rat forebrains were incubated with 10 �g�ml LPS in com-bination with 20 mM KCl or with PBS for 5 days. Staining with antibodiesagainst neurofilament, O4, and glial fibrillary acidic protein marked axons,oligodendrocytes, and astrocytes, respectively. Microglia were stained withisolectin B4. DAPI staining revealed the total number of cells by staining allnuclei. (Scale bar, 50 �m.)

Lehnardt et al. PNAS � July 8, 2003 � vol. 100 � no. 14 � 8515

NEU

ROSC

IEN

CE

PCR studies determined that astrocytes and oligodendrocyteprecursors express CD14 but not TLR4 (16). To investigatewhether neurons express TLR4, CD14, or both, we preparedpurified cultures of cortical neurons and microglia from newbornrats. To avoid contaminating signals from microglia, oligoden-drocytes, or astrocytes, only neuronal cultures that contained�98% neurons as determined by immunohistochemistry wereused. RNA was isolated and reverse-transcribed. PCR wascarried out by using primers specific for rat TLR4 and CD14. Notranscripts for TLR4 or CD14 were detected in the corticalneurons, whereas microglia expressed both (Fig. 3C).

Microglia bind Alexa-tagged LPS in vitro, whereas oligoden-drocytes and astrocytes do not (16). Thus, we next asked whetherneurons bind LPS. Primary cultures of purified rat corticalneurons and microglia were incubated with 10 �g�ml LPS-Alexafor 1 h at 37°C. Cells were visualized by fluorescence microscopyafter washing with PBS (Fig. 3D). Cortical neurons did not showlabeling with LPS-Alexa. The presence of neurons was con-firmed by staining with NeuN. In contrast, microglia werestained intensely with LPS-Alexa.

TLR4 Is Necessary for LPS-Induced Neuronal Injury in Vitro. Todetermine whether TLR4 is necessary for LPS-mediated neu-ronal injury and death we made use of C.C3H-TLR4lps-d (lpsd)mice. The lpsd mutation originated from the C3H�HeJ mouseand was introduced into the BALB�cJ background by backcross-ing C3H�HeJ mice to BALB�cJ mice. The resulting C.C3-TLR4lps-d mice were used in our experiments. The tlr4 mutantmouse C3H�HeJ is characterized by hyporesponsiveness to LPSas a consequence of a functionally defective TLR4 membraneprotein due to a mutation of the C-terminal part of TLR4 thatinterferes with LPS-induced signaling. Macrophages from thisstrain fail to induce inflammatory cytokines such as tumornecrosis factor �, IL-1, and IL-6 (10, 23, 24).

Cortical neurons from E17 forebrains were generated from

lpsd and BALB�cJ mice. Staining after 10 days in culturerevealed the presence of neurons and astrocytes, oligodendro-cytes, and microglia. Cultures were treated with 10 �g�ml LPSor PBS for 5 days. Axons and NeuN were visualized by antineu-rofilament (Fig. 4A) and NeuN staining (Fig. 4B), respectively.No difference in axonal or neuronal number was observedbetween control-treated cultures from lpsd and those fromBALB�cJ mice. LPS treatment led to a major loss of axons andneurons in mixed CNS cultures derived from BALB�cJ fore-brains. LPS treatment had no effect on the number of axons andneurons of lpsd mice. DAPI staining and statistical analysis of theratio of neuron to DAPI� cells confirmed these results (Fig. 4C).

Fig. 2. LPS-induced neurotoxicity does not depend on cell type or species.Mixed CNS cultures were prepared from mouse and chicken spinal cords.Dorsal root ganglion cells (DRG) were generated from chicken, and mixed gliafrom rat were added. Cultures were treated with 10 �g�ml LPS in combinationwith 20 mM KCl or PBS for 5 days. Dendrites and axons were stained withmicrotubule-associated protein 2 and antibody against neurofilament. Nucleiwere stained with DAPI. (Scale bar, 50 �m.)

Fig. 3. LPS-induced neurotoxicity is not cell-autonomous but requires mi-croglia. Cortical neurons and microglia were prepared from rat forebrains. (A)Purified neurons and neurons in combination with microglia were incubatedwith 10 �g�ml LPS�KCl or PBS. After 2 days, cultures were fixed and stainedwith NeuN and F4�80 to mark neurons and microglia, respectively. (Scale bar,50 �m.) (B) Quantitation of NeuN� neurons in purified and microglia-enrichedcultures in the presence or absence of LPS�KCl. Results are presented asmean � SE. P � 0.001. (C) PCR. TLR4 is expressed in microglia but not inneurons. RNA was isolated from purified neuronal and microglial cultures andamplified by RT-PCR with specific primers for TLR4, CD14, or �-actin. (D)Fluorescently labeled LPS binds to microglia but not to neurons. Enrichedcultures of microglia and cortical neurons were incubated with Alexa-taggedLPS and analyzed by immunofluorescence. Parallel cultures were stained toidentify microglia (F4�80) and neurons (NeuN). (Scale bar, 100 �m.)

8516 � www.pnas.org�cgi�doi�10.1073�pnas.1432609100 Lehnardt et al.

At baseline, no differences in microglial numbers were detect-able between lpsd and BALB�cJ mice (data not shown).

TLR4 Is Necessary for LPS-Sensitized Hypoxic-Ischemic Neuronal Injuryin Vivo. We addressed the question of whether the effect of LPSon neurons observed in vitro is relevant to models of neurode-generation in vivo. We injected LPS or vehicle i.p. into 7-day-oldlpsd and BALB�cJ mice and combined this treatment withexposure to HI. Three lpsd mice were treated with HI alone, andsix lpsd mice received LPS i.p. before HI. In parallel, fiveBALB�cJ mice were treated with HI, and five BALB�cJ micereceived an injection of LPS before HI. All 19 animals werestudied after 4 days. There was no mortality between experi-mental treatment on P7 and fixation on P11. A tendency towardlower rectal temperatures after carotid ligation in animals fromboth strains was observed. The level of starting temperature wasalways achieved when the animals were allowed to recover afterexposure to 7.7% oxygen.

In vehicle-treated animals combined with 30 min HI, noaxonal or neuronal damage was observed in lpsd mice orBALB�cJ mice (Fig. 5). Of the total of five animals from theBALB�cJ strain treated with HI in combination with LPS, fourshowed axonal (Fig. 5A) and neuronal (Fig. 5B) loss in the corpuscallosum and underlying structures ipsilateral to the carotidligation. The contralateral hemisphere revealed neurofilamentand NeuN staining comparable to animals that did not receiveLPS. In contrast, of the total of six lpsd mice that received LPSin addition to HI, none showed axonal or neuronal damage. Nodifference in neurofilament and NeuN staining was observedbetween the ipsilateral and contralateral hemispheres. Quanti-tative analysis of the pericallosal area of the four affected

BALB�cJ mice and six lpsd mice treated with HI and LPSconfirmed these results. The total number of NeuN� cells perfield in the ipsilateral hemisphere of BALB�cJ mice was reduced2.5-fold compared with the contralateral hemisphere, whereasnumbers of NeuN� cells of the ipsilateral hemisphere in lpsd

mice showed no significant change compared with the contralat-eral side. The total number of NeuN� cells of the contralateralside of BALB�cJ mice did not differ from the numbers of bothhemispheres of lpsd mice. The results discussed above indicate anobligatory role for TLR4 in LPS-mediated neuronal injury invivo.

DiscussionIn most organ systems, inflammation causes bystander injurythat is typically reversible due to the inherent regenerativecapacity of the cellular elements of that tissue. However, in theCNS, the stakes are higher. The common consequence ofbystander injury in the CNS is irreversible neuronal loss andatrophy due to regenerative failure. Through evolution there isno clear system that has developed to protect the CNS from theeffects of inflammation as compensation for its poor regenera-tive capacity. In this report we elucidated a pathway that links aspecific molecular component of Gram-negative infections, LPS,to the resident macrophage-like cell within the CNS, the micro-glial cell, through TLR4. We showed that activation of microgliaby LPS results in neuronal and axonal loss both in vitro and invivo. In the absence of functional TLR4, neurons and axons wereresistant to LPS-mediated injury.

The relationship between neurodegeneration and microglialactivation is complicated and unresolved. Traditionally, micro-glial activation is the normal response to CNS injury. However,

Fig. 4. TLR4 is necessary for LPS-induced neuronal injury in vitro. Mixed CNS cultures prepared from BALB�cJ and lpsd mouse forebrains were incubated with10 �g�ml LPS alone or in combination with 20 mM KCl for 5 days. (A) Cultures were stained with antibody against neurofilament to mark axons and with DAPI.(Scale bar, 50 �m.) (B) Cultures were stained with antibody NeuN (neurons) and DAPI. (Scale bar, 50 �m.) (C) Quantitation of NeuN� neurons from BALB�cJ andlpsd mice in the presence or absence of LPS. Experiments were repeated three times. Results are presented as mean � SE. P � 0.001.

Lehnardt et al. PNAS � July 8, 2003 � vol. 100 � no. 14 � 8517

NEU

ROSC

IEN

CE

in some cases activation of microglia is believed to contribute to,if not cause, neurodegeneration by releasing proinflammatoryand cytotoxic factors, including nitric oxide and IL-1 (25, 26).Inhibition of microglial activation by minocycline protectsMPTP-induced neuronal death in the MPTP mouse model ofParkinson’s disease (27). Activated microglia are critical forN-methyl-D-aspartate-induced neurodegeneration in slice cul-tures (28). �-Amyloid in combination with IFN-� triggers theproduction of reactive nitrogen intermediates and tumor necro-sis factor � from microglia with consequent neuronal injury (29).Elevated levels of cytokines such as IL-1� and tumor necrosisfactor � mainly produced by microglia are frequently detected inthe serum and cerebrospinal f luid of patients with Alzheimer’sdisease, Parkinson’s disease, multiple sclerosis, and amyotrophiclateral sclerosis (30). Activation of microglia is interpreted as aninflammatory response to infection and microglial cytokines areknown to induce neuronal death in models of neurodegenera-tion. Therefore, the question of whether infections may play arole in the aforementioned diseases arises.

TLRs play a central role in the initiation of cellular innateimmune responses and serve as pathogen-associated molecularpattern receptors that bind microbial molecular motifs with highspecificity. TLR4 is required for transducing LPS signals onmonocytes and macrophages (24). We showed that microgliaplay a role in innate immune actions in the CNS and thatmicroglia are the only nonneuronal cell type that expresses TLR4(16). In the present study we showed that microglia are not onlythe sole glial cells that express TLR4 mRNA but that thisproperty also distinguishes microglia from neurons. Consistent

with these results and in contrast to microglia, cortical neuronsdid not bind fluorescently tagged LPS. LPS induced axonal andneuronal loss in different neuronal populations derived fromwild-type animals. Axons and neurons, however, were unaf-fected by LPS in cultures derived from tlr4 mutant mice. BecauseLPS functions through TLR4 and microglia but not neuronsexpress TLR4, it is consistent that LPS-induced neurotoxicityrequires the presence of microglia. It is possible that underconditions of stress as occur in pathologic states, TLR4 expres-sion may be induced in cells other than microglia. Because we didnot investigate expression of TLR4 in other neuronal cell types,the possibility that LPS-induced neurotoxicity in other types ofneurons is based on a different mechanism than we haveobserved cannot be excluded.

We observed a rapid loss of microglia in purified culturesduring LPS treatment as well as in CNS cultures, although to alesser extent. It is possible that in vitro microglia ‘‘sacrifice’’themselves during their normal response to injury or infection.Microglia can be divided into several subtypes based on theirmorphology and localization (31). Parenchymal microglia dif-ferentiate from monocytes that migrate into the CNS duringembryonic development (32) and are maintained as a pool witha low turnover rate and limited, if any, replication duringadulthood, although the number of microglia increases at sites ofCNS reaction after injury. The origin of reactive microglia couldeither be migrating parenchymal microglia or monocytes in-duced to differentiate into parenchymal microglia. Other sourcesof microglia include perivascular microglia or leptomeningealmacrophages derived from circulating monocytes (33). It is also

Fig. 5. TLR4 is necessary for LPS-induced neuronal injury in vivo. Coronal brain sections of P11 neonatal BALB�cJ and lpsd mice that received a single dose ofeither LPS or vehicle i.p. before being exposed to HI treatment on P7 were stained with antibodies against neurofilament (A) or NeuN (B). (Scale bar, 50 �m.)

8518 � www.pnas.org�cgi�doi�10.1073�pnas.1432609100 Lehnardt et al.

possible that function and survival of microglia require distinctyet unidentified molecules produced by the environment as itexists in vivo but not in vitro.

In some cases, activation of innate immunity may promoteCNS regeneration. Zymosan, a yeast-wall extract and a potentpathogen-associated molecular pattern, promotes regenerationof crushed optic nerves (34). Overexpression of transforminggrowth factor �1 decreases neurotoxic �-amyloid plaque load ina mouse model of Alzheimer’s disease through activation ofmicroglia (35). Microglia introduced into the injured spinal cordsupport growth of injured axons (36). The differences betweeninnate immunity-induced injury and protection may reside in adelicate balance of extracellular signals acting on microglia. Withrespect to injury, the context within the CNS in which microgliaare activated may be critical. In amyotrophic lateral sclerosis,Alzheimer’s disease, and stroke, where neurons experience aprimary insult that predisposes them to cell death probablythrough oxidative stress (37, 38), the impact of a secondary insultmediated by microglia may tip the balance to irreversible injury.Consistent with this hypothesis, we show in this report that asubthreshold hypoxic-ischemic insult in mice can be converted toneurodegeneration by activating innate immunity. The CNS isparticularly vulnerable to HI (39). Several studies suggest thatdefects in mitochondrial energy metabolism and excitotoxicneuronal death play a critical role in neurodegenerative diseases.To study the etiology of these diseases, animal models based onHI have been developed. The MPTP model of Parkinson’sdisease involves a specific defect in mitochondrial energy me-tabolism induced by MPP�, whereas models of HD are based ondefects in oxidative phosphorylation (40, 41). To model the roleof innate immunity in neurodegeneration we made use of aparadigm in which brief HI leads to a subthreshold injury in theCNS. When coupled with a specific activator of innate immunitysuch as LPS, this subthreshold injury is converted to prominentneuronal injury (19). We adapted this model of HI originally

designed for the use in rats to 7-day-old mice. We used a low doseof i.p.-injected LPS to stimulate innate immunity, but not causesystemic hypotension, and a limited duration of HI. Although wedetected extensive neuronal and axonal loss in the cortex andunderlying structures in LPS-treated control animals, no sucheffect was seen in tlr4 mutant mice. Because we administeredLPS systemically we cannot exclude the possibility that inBALB�cJ mice peripheral macrophages became activated andentered the brain through the damaged blood–brain barrier. Itis possible that instead of brain-resident microglia, activatedperipheral macrophages contributed to neuronal injury by pro-ducing neurotoxic molecules in response to activation of theTLR4 pathway.

LPS is only one potential ligand for TLR4. Other exogenousor endogenous ligands may contribute to neurotoxicity andneurodegeneration in disorders not caused by microorganisms.Therefore, LPS should be viewed as an exemplary ligand forTLR4 rather than the sole mechanism to activate microglia andlead to neurodegeneration. Our data demonstrate a key role forTLR4 in LPS-induced neuronal injury in vivo and support theconcept that neurons experiencing a single reversible insult (HI)undergo irreversible injury after a second insult (microglialactivation).

In summary, this study establishes a general mechanistic linkbetween innate immunity and neuronal injury. Microglia are theonly cell type in the resting CNS that expresses TLR4. TLR4plays a crucial role in LPS-induced neuronal damage and celldeath in vitro and in vivo. Further understanding the mechanismsmediated by TLRs may provide insights into potential thera-peutic interventions for inflammation-related neurodegenera-tive diseases.

We thank Drs. Clifford Saper, Amyn Habib, Eckart Schott, and theVartanian Laboratory for helpful comments. This work was supportedby National Institute of Neurological Disorders and Stroke GrantsP01NS38475 (to T.V., J.J.V., P.A.R., and F.E.J.), K02N502028 (to T.V.),and P30HD18655 and a grant from The Dalis Foundation.

1. Kreutzberg, G. W. (1996) Trends Neurosci. 19, 312–318.2. Dickson, D. W., Lee, S. C., Mattiace, L. A., Yen, S. H. & Brosnan, C. (1993)

Glia 7, 75–83.3. McGeer, P. L., Itagaki, S., Boyes, B. E. & McGeer, E. G. (1988) Neurology 38,

1285–1291.4. Trapp, B. D., Bo, L., Mork, S. & Chang, A. (1999) J. Neuroimmunol. 98, 49–56.5. Giulian, D., Wendt, E., Vaca, K. & Noonan, C. A. (1993) Proc. Natl. Acad. Sci.

USA 90, 2769–2773.6. McDonald, D. R., Brunden, K. R. & Landreth, G. E. (1997) J. Neurosci. 17,

2284–2294.7. Liberatore, G. T., Jackson-Lewis, V., Vukosavic, S., Mandir, A. S., Vila, M.,

McAuliffe, W. G., Dawson, V. L., Dawson, T. M. & Przedborski, S. (1999) Nat.Med. 5, 1403–1409.

8. Medzhitov, R. & Janeway, C. A., Jr. (1997) Curr. Opin. Immunol. 9, 4–9.9. Medzhitov, R. & Janeway, C., Jr. (2000) Immunol. Rev. 173, 89–97.

10. Hoshino, K., Takeuchi, O., Kawai, T., Sanjo, H., Ogawa, T., Takeda, Y.,Takeda, K. & Akira, S. (1999) J. Immunol. 162, 3749–3752.

11. Wright, S. D. (1999) J. Exp. Med. 189, 605–609.12. Medzhitov, R., Preston-Hurlburt, P., Kopp, E., Stadlen, A., Chen, C., Ghosh,

S. & Janeway, C. A., Jr. (1998) Mol. Cell 2, 253–258.13. Shimazu, R., Akashi, S., Ogata, H., Nagai, Y., Fukudome, K., Miyake, K. &

Kimoto, M. (1999) J. Exp. Med. 189, 1777–1782.14. Lacroix, S., Feinstein, D. & Rivest, S. (1998) Brain Pathol. 8, 625–640.15. Laflamme, N. & Rivest, S. (2001) FASEB J. 15, 155–163.16. Lehnardt, S., Lachance, C., Patrizi, S., Lefebvre, S., Follett, P. L., Jensen, F. E.,

Rosenberg, P. A., Volpe, J. J. & Vartanian, T. (2002) J. Neurosci. 22,2478–2486.

17. Murphy, T. H., Schnaar, R. L. & Coyle, J. T. (1990) FASEB J. 4, 1624–1633.18. Vartanian, T., Li, Y., Zhao, M. & Stefansson, K. (1995) Mol. Med. 1, 732–743.19. Eklind, S., Mallard, C., Leverin, A. L., Gilland, E., Blomgren, K., Mattsby-

Baltzer, I. & Hagberg, H. (2001) Eur. J. Neurosci. 13, 1101–1106.20. Chang, R. C., Hudson, P. M., Wilson, B. C., Liu, B., Abel, H. & Hong, J. S.

(2000) Neuroscience 97, 757–764.21. Jeohn, G. H., Kim, W. G. & Hong, J. S. (2000) Brain Res. 880, 173–177.

22. Becher, B., Fedorowicz, V. & Antel, J. P. (1996) J. Neurosci. Res. 45, 375–381.23. Poltorak, A., He, X., Smirnova, I., Liu, M. Y., Huffel, C. V., Du, X., Birdwell,

D., Alejos, E., Silva, M., Galanos, C., et al. (1998) Science 282, 2085–2088.24. Qureshi, S. T., Lariviere, L., Leveque, G., Clermont, S., Moore, K. J., Gros, P.

& Malo, D. (1999) J. Exp. Med. 189, 615–625.25. Chao, C. C., Hu, S., Molitor, T. W., Shaskan, E. G. & Peterson, P. K. (1992)

J. Immunol. 149, 2736–2741.26. Lee, S. C., Liu, W., Dickson, D. W., Brosnan, C. F. & Berman, J. W. (1993)

J. Immunol. 150, 2659–2667.27. Wu, D. C., Jackson-Lewis, V., Vila, M., Tieu, K., Teismann, P., Vadseth, C.,

Choi, D. K., Ischiropoulos, H. & Przedborski, S. (2002) J. Neurosci. 22,1763–1771.

28. Ullrich, O., Diestel, A., Eyupoglu, I. Y. & Nitsch, R. (2001) Nat. Cell Biol. 3,1035–1042.

29. Meda, L., Cassatella, M. A., Szendrei, G. I., Otvos, L., Jr., Baron, P., Villalba,M., Ferrari, D. & Rossi, F. (1995) Nature 374, 647–650.

30. Gonzalez-Scarano, F. & Baltuch, G. (1999) Annu. Rev. Neurosci. 22, 219–240.31. Rinner, W. A., Bauer, J., Schmidts, M., Lassmann, H. & Hickey, W. F. (1995)

Glia 14, 257–266.32. Richardson, A., Hao, C. & Fedoroff, S. (1993) Glia 7, 25–33.33. Hickey, W. F., Vass, K. & Lassmann, H. (1992) J. Neuropathol. Exp. Neurol. 51,

246–256.34. Leon, S., Yin, Y., Nguyen, J., Irwin, N. & Benowitz, L. I. (2000) J. Neurosci.

20, 4615–4626.35. Wyss-Coray, T., Lin, C., Yan, F., Yu, G. Q., Rohde, M., McConlogue, L.,

Masliah, E. & Mucke, L. (2001) Nat. Med. 7, 612–618.36. Prewitt, C. M., Niesman, I. R., Kane, C. J. & Houle, J. D. (1997) Exp. Neurol.

148, 433–443.37. Coyle, J. T. & Puttfarcken, P. (1993) Science 262, 689–695.38. Beal, M. F. (1995) Ann. Neurol. 38, 357–366.39. Lee, J. M., Grabb, M. C., Zipfel, G. J. & Choi, D. W. (2000) J. Clin. Invest. 106,

723–731.40. Beal, M. F. (1992) Ann. Neurol. 31, 119–130.41. Lee, J. M., Zipfel, G. J. & Choi, D. W. (1999) Nature 399, A7–A14.

Lehnardt et al. PNAS � July 8, 2003 � vol. 100 � no. 14 � 8519

NEU

ROSC

IEN

CE