Chemokine receptor CXCR3 and its ligands CXCL9and CXCL10 are required for the developmentof murine cerebral malariaGabriele S. V. Campanella, Andrew M. Tager, Joseph K. El Khoury, Seddon Y. Thomas*, Tabitha A. Abrazinski,Lindsay A. Manice, Richard A. Colvin, and Andrew D. Luster†

Center for Immunology and Inflammatory Diseases, Division of Rheumatology, Allergy, and Immunology, Massachusetts General Hospital,Harvard Medical School, Charlestown, MA 02129

Communicated by Jeffrey Ravetch, The Rockfeller University, New York, NY, February 15, 2008 (received for review December 18, 2007)

Cerebral malaria is a significant cause of global mortality, causingan estimated two million deaths per year, mainly in children. Thepathogenesis of this disease remains incompletely understood.Chemokines have been implicated in the development of cerebralmalaria, and the IFN-inducible CXCR3 chemokine ligand IP-10(CXCL10) was recently found to be the only serum biomarker thatpredicted cerebral malaria mortality in Ghanaian children. Weshow that the CXCR3 chemokine ligands IP-10 and Mig (CXCL9)were highly induced in the brains of mice with murine cerebralmalaria caused by Plasmodium berghei ANKA. Mice deficient inCXCR3 were markedly protected against cerebral malaria and hadfar fewer T cells in the brain compared with wild-type mice. Incompetitive transfer experiments, CXCR3-deficient CD8� T cellswere 7-fold less efficient at migrating into the infected brains thanwild-type CD8� T cells. Adoptive transfer of wild-type CD8� effec-tor T cells restored susceptibility of CXCR3-deficient mice to cere-bral malaria and also restored brain proinflammatory cytokine andchemokine production and recruitment of T cells, independent ofCXCR3. Mice deficient in IP-10 or Mig were both partially protectedagainst cerebral malaria mortality when infected with P. bergheiANKA. Brain immunohistochemistry revealed Mig staining of en-dothelial cells, whereas IP-10 staining was mainly found in neu-rons. These data demonstrate that CXCR3 on CD8� T cells isrequired for T cell recruitment into the brain and the developmentof murine cerebral malaria and suggest that the CXCR3 ligands Migand IP-10 play distinct, nonredundant roles in the pathogenesis ofthis disease.

T cells � Plasmodium � IP-10 � brain � trafficking

Malaria remains one of the most common infectious diseasesin the world with an estimated 500 million clinical cases

per year. One of the most lethal complications of infection withPlasmodium falciparum is cerebral malaria (CM), which causesmore than one million deaths per year, mainly in Africanchildren under the age of 5. Although CM has been studiedextensively, the pathogenesis of this disease is not fully under-stood. Sequestration of parasitized red blood cells in the micro-vasculature of the brain has long been thought to be the mainfactor leading to CM. However, sequestration does not seem tobe the sole cause of CM. Instead, loss of blood–brain barrierintegrity and activation of the inflammatory response have beenwell documented in CM and likely contribute to the complexpathogenesis of this disease (1).

Mouse models of CM have been widely used to gain insightinto the pathogenesis of CM. Although no animal model fullyrecapitulates the human disease, infection of certain mousestrains with Plasmodium berghei ANKA induces murine CM withmany characteristics of the human disease. Infected mice de-velop cerebral complications, including paralysis, ataxia, con-vulsion, coma, and death. CD8� T cell sequestration in the brainhas been shown to play a critical role in the development ofmurine CM (2–4). However, the mechanisms directing these

cells to the brain have not been fully elucidated, with two studiesshowing that the chemokine receptor CCR5 partially guidesCD8� T cells into the brain (3, 4).

Chemokines are a superfamily of chemotactic cytokines thatplay important roles in the generation and delivery of immuneand inflammatory responses. They are also involved in mostdisease processes, including infectious, autoimmune, inflamma-tory, and malignant diseases (5). Chemokines orchestrate themovement of leukocytes and other cells by activating specificseven transmembrane-spanning G protein-coupled receptorsexpressed on responsive cells. IP-10 (IFN-induced protein of 10kDa), or CXCL10, one of the very first chemokines identified,directs the trafficking of activated effector CD4� and CD8� Tlymphocytes and other effector lymphocytes, such as naturalkiller (NK) and natural killer T (NKT) cells. It does so by bindingto its high-affinity receptor CXCR3, which it shares with twoother ligands, monokine-induced by IFN-� (Mig/CXCL9) andIFN-inducible T cell-� chemoattractant (I-TAC/CXCL11).CXCR3 and its ligands have been implicated in neurologicaldiseases caused by other pathogens, including Toxoplasma gondii(6), West Nile virus (7), and HIV (8). To date, little has beendone to investigate the role of CXCR3 and its ligands in CM. Arecent study measuring 36 biomarkers found that IP-10 was theonly serum marker independently associated with CM mortalityin Ghanaian children (9). In addition, IP-10 was one of eightbiomarkers in the cerebral spinal f luid (CSF) significantly up-regulated in the CM group. Furthermore, in a mouse model ofCM, NK cells were shown to induce the recruitment of CXCR3�

T cells to the brain (10). However, the role of IP-10 and CXCR3in CM pathogenesis has not been fully established. We thereforeinvestigated chemokine expression in the brain and spleen ofmice infected with P. berghei ANKA and studied the effects ofIP-10, Mig, and CXCR3 in murine CM, by using mice geneticallydeficient for CXCR3, IP-10, and Mig.

ResultsCXCR3-Binding Chemokines Are Highly Up-Regulated in the Brains ofMice with CM. We examined the expression of 21 chemokines inthe brain and spleen during murine CM induced by P. bergheiANKA in C57BL/6 mice. Eight days after infection, when themice displayed signs of CM, the two CXCR3 ligands, IP-10 andMig, were the most highly expressed of the tested chemokines inthe brain in terms of total mRNA levels (Fig. 1A). In addition,

Author contributions: G.S.V.C., A.M.T., and A.D.L. designed research; G.S.V.C., T.A.A., andL.A.M. performed research; G.S.V.C., J.K.E.K., S.Y.T., and R.A.C. analyzed data; and G.S.V.C.,A.M.T., and A.D.L. wrote the paper.

The authors declare no conflict of interest.

*Present address: Howard Hughes Medical Institute, Committee on Immunology, Depart-ment of Pathology, University of Chicago, Chicago, IL 60637.

†To whom correspondence should be addressed at: Massachusetts General Hospital, Build-ing 149, 13th Street, Charlestown, MA 02129. E-mail: aluster@mgh.harvard.edu.

© 2008 by The National Academy of Sciences of the USA

4814–4819 � PNAS � March 25, 2008 � vol. 105 � no. 12 www.pnas.org�cgi�doi�10.1073�pnas.0801544105

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

they were also the most highly induced chemokines. Mig wasinduced 500-fold and IP-10 220-fold after infection (Fig. 1B).Although a small increase in I-TAC mRNA expression wasobserved, we did not study further the effect of I-TAC on thedevelopment of murine CM. C57BL/6 mice do not producefunctional I-TAC protein because of a mutation that introducesa stop codon in the I-TAC leader sequence (GenBank accessionnumbers NM�019494 and AA174767). Sequencing of PCR prod-ucts obtained from our IP-10- and Mig-deficient mice in theC57BL/6 background revealed the same mutations, demonstrat-ing that the IP-10- and Mig-deficient mice do not expressfunctional I-TAC protein (R.A.C., G.S.V.C., and A.D.L., un-published data). In addition to the CXCR3 ligands, the CCchemokines MCP-1, MCP-3, RANTES, MIP-1�, MIP-1�, andthe CXC chemokine MIP-2 were also induced but to lower levelsthan the CXCR3 ligands. Chemokine levels in the spleen ingeneral did not increase as much as in the brain during murineCM (Fig. 1 D and E). In the spleen, Mig increased 1.8-fold andIP-10 2.4-fold, whereas the mRNA levels of a number of otherchemokines, including RANTES, SDF-1, and ELC decreasedafter P. berghei ANKA infection.

IP-10 and Mig were the initial chemokines induced in the brain6 days after infection (50-fold and 165-fold, respectively) andincreased further by day 8 (Fig. 1C). RANTES, MIP-1�, MCP-1,and IFN-� levels were still low on day 6 and increased by day 8.In contrast, in the spleen, the highest chemokine levels were seen6 days after infection, with MCP-1 and IFN-� showing thehighest induction (Fig. 1F). By day 8 after infection, spleen RNAlevels had decreased for all chemokines analyzed.

IP-10 and Mig protein levels in the brain and spleen wereanalyzed by ELISA. Consistent with RNA data, IP-10 and Migprotein levels increased by day 6 after infection (Fig. 1C Inset)and increased further on day 8. Spleen IP-10 and Mig proteinlevels were increased day 6 after infection and declined by day8 (Fig. 1F Inset).

CXCR3 KO Are Markedly Protected from CM. To determine the roleof the CXCR3 ligands in murine CM, we infected CXCR3-deficient (CXCR3 KO) mice with P. berghei ANKA. Wild-typemice developed murine CM between days 6 and 10 afterinfection, and their mortality was 80–100% (Fig. 2A). In con-

trast, CXCR3 KO mice were markedly protected from CM, withmortality of only 10–30% (Fig. 2 A). Levels of parasitemia forwild-type and CXCR3 KO mice were found to be similar throughday 10 after infection, by which time most wild-type mice haddied (Fig. 2B). CXCR3 KO mice that did not develop CM hadincreasing levels of parasitemia, developed anemia, and werekilled at day 21 after infection. Typical signs of CM were evidentin H&E-stained brain sections from wild-type mice, with oc-cluded and disrupted blood vessels and leukocyte accumulation,whereas these findings were absent in CXCR3 KO mice(Fig. 2C).

T Cell Recruitment to the Brain Is Reduced in CXCR3 KO Mice.Brain-infiltrating CD8� T cells have been shown to be crucialmediators of CM induced by P. berghei ANKA. We thereforecompared the recruitment of CD3� T cells and CD3�CD8� T

Fig. 1. Chemokine mRNA and protein levels in brain and spleen during murine CM. C57BL/6 mice were infected with P. berghei ANKA (2 � 106 parasitized redblood cells), and their brains (A–C) and spleens (D–F) were harvested at the indicated time points. mRNA levels were analyzed by qPCR and are shown normalizedto GAPDH (A and D) and as fold induction compared with uninfected mice (B, C, E, and F). Brain and spleen IP-10 and Mig protein levels were determined byELISA (C and F, Inset). Results are representative of two independent experiments with two or three mice per group.

Fig. 2. CXCR3 KO mice are protected from murine CM. Wild-type and CXCR3KO C57BL/6 mice were infected with P. berghei ANKA. (A) Mortality waschecked twice daily. The experiment is representative of four separate infec-tions, each with at least eight mice per group. (B) Parasitemia was monitoredevery other day by Giemsa-stained thin blood smears. The experiment isrepresentative of two independent infections. (C) Brain histology was assessedby H&E staining. Brains were removed after heart perfusion with 10% forma-lin of uninfected mice or mice 8 days after infection, as indicated. (Scale bars:20 �m.)

Campanella et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4815

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

cells to the brains of infected wild-type and CXCR3 KO mice.T cell infiltration to the brain peaked at day 8, when micedeveloped signs of severe CM (Fig. 3). Infiltration of both CD3�

and CD8� T cells into the brain was higher in wild-type mice thanin CXCR3 KO mice. The number of brain-sequestered CD8� Tcells at day 8 after infection when CM was evident was 14-foldhigher in wild-type mice than on day 0, with overall CD3� T cellsbeing increased 7-fold (Fig. 3 Left). This response was markedlyattenuated in CXCR3 KO mice, which only had a 4-fold increasein CD3�CD8� T cell and 2.7-fold increase in CD3� T cellrecruitment compared with uninfected mice. CD8� T cell re-cruitment to the infected brain in CXCR3 KO mice was reduced300% compared with wild-type mice.

In addition to T lymphocytes, NK and NKT cells have beenshown to play an important role in the development of murineCM. Recruitment of NK cells to the brain in CXCR3 KO micewas 2-fold lower than for wild-type mice, whereas NKT cells werenot significantly different (Fig. 3 Bottom).

CXCR3 Is Required for CD8� T Cell Trafficking to the Brain. Todetermine directly the role for CXCR3 in CD8� T cell homingto the brain in murine CM, we used competitive adoptivetransfer experiments. In these studies, in vitro-generated wild-type and CXCR3 KO effector CD8� T cells were cotransferredinto wild-type mice that were infected 5 days earlier with P.berghei ANKA. We used CD8� T cells from ovalbumin-specificT cell receptor-transgenic mice (OT-I) crossed with wild-typeThy1.1 or CXCR3 KO Thy1.2 mice. When transferred intowild-type Thy1.1�Thy1.2 mice, both wild-type (Thy1.1) andCXCR3 KO (Thy1.2) OT-1 CD8� T cells can be analyzed in thesame recipient mouse. In addition, the use of OT-I cells allowedus to examine the role of CXCR3 in migration to the brainindependent of antigen-induced proliferation. Both wild-typeand CXCR3 KO OT-I effector CD8� T cells were confirmed tobe similarly activated, with high CD25 expression and interme-diate CD62L expression, as determined by FACS analysis (data

not shown). When recipient mice developed CM, brain andspleen were harvested and analyzed for the presence of trans-ferred cells by using the Thy1.1 and Thy1.2 markers. In thespleen, wild-type and CXCR3 KO cells were present in similarnumbers (Fig. 4A). In contrast, in the brain there were 7-foldfewer CXCR3 KO OT-I cells than wild-type cells (Fig. 4B),which clearly demonstrates that CXCR3-deficient CD8� T cellscould not traffic to the brain as efficiently as wild-type CD8� Tcells.

Wild-Type CD8� T Cells Restore CXCR3KO Mice CM Mortality. Todefine more specifically the involvement of CXCR3 in thisdisease, splenocytes from wild-type or CXCR3 KO mice infected5 days earlier with P. berghei ANKA were adoptively transferredi.v. into CXCR3 KO recipients. Two hours later, recipient andcontrol CXCR3 KO mice were infected with P. berghei ANKA.CXCR3 KO mice that had received wild-type splenocytes de-veloped CM at nearly the same rate as wild-type mice, whereasonly a very low proportion of CXCR3 KO mice that had receivedCXCR3 KO splenocytes developed CM, similar to CXCR3 KOmice that did not receive any transferred cells (Fig. 5A). Thisfinding demonstrates that CXCR3 on leukocytes is required forthe development of murine CM and that wild-type leukocytescan restore CM mortality in CXCR3 KO mice. Because CXCR3is expressed on different leukocyte subsets among splenocytes,we next isolated CD3�CD8� T cells from wild-type splenocytes5 days after infection by FACS sorting and then adoptivelytransferred them into CXCR3 KO mice before infection. Micethat had received CD8� T cells developed CM at rates similar tothose of wild-type mice (Fig. 5A), indicating that CXCR3 onCD8� T cells is required to induce CM.

Using the Thy1.1 allele, we found that adoptively transferredwild-type CD8� T cells (Thy1.1) were recruited to the brain ofCXCR3 KO mice (Thy1.2) at the CM stage (0.72 � 0.20% oflymphocyte gate) (Fig. 5 B and C). CXCR3 KO mice that hadreceived wild-type CD8� T cells also had a marked increase inthe numbers of their own CXCR3-deficient CD8� T cells(Thy1.2) in the brain compared with CXCR3 KO mice that hadnot received wild-type CD8� T cells (Fig. 5C). The number ofCXCR3KO CD8� T cells (Thy1.2) in the brain of CXCR3 KOmice that had received adoptively transferred wild-type CD8� Tcells was 3-fold higher than in CXCR3 KO mice withouttransferred cells. To investigate the mechanism of this CXCR3-independent recruitment of CD8� T cells, we measured chemo-kine and IFN-� expression in the brains of wild-type, CXCR3KO, and CXCR3 KO mice that had received adoptively trans-

Fig. 3. CD3�CD8� T cell sequestration is reduced in the brains of CXCR3 KOmice infected with P. berghei ANKA. Flow cytometric analysis of brain-sequestered T lymphocytes and NK cells is shown. Wild-type and CXCR3 KOmice were infected with P. berghei ANKA, and brains were harvested atdifferent days after infection. (Left) Number of brain sequestered CD3� (Top),CD3�CD8� (Middle), and NK (Bottom) cells were averaged over four indepen-dent experiments and are shown � SD. (Right) Representative primary flowcytometric dot plots are shown after lymphocyte gating for T cells (Top andMiddle) and NK/NKT cells (Bottom). The percentage of lymphocyte gate isshown for the indicated cell populations.

Fig. 4. CXCR3 is required for efficient CD8 T cell trafficking to the brain.Wild-type (Thy1.1) and CXCR3 KO (Thy1.2) OT-I cells were activated in vitro for5 days and were adoptively cotransferred (5 � 106 each) into the sameThy1.1�Thy1.2 wild-type mouse by i.v. injection. Mice were infected 5 daysbefore adoptive transfer with P. berghei ANKA. Eight days after infection,when mice developed signs of murine CM, spleens (A) and brains (B) wereharvested, and Thy1.1 and Thy1.2 single positive cells were analyzed by flowcytometry as shown. Results are representative of three independent exper-iments. *, P � 0.05 compared with wild-type OT-I cells. (Right) Representativeprimary flow cytometric dot plots are shown after lymphocyte gating. Thepercentage of lymphocyte gate is shown for the indicated cell populations.

4816 � www.pnas.org�cgi�doi�10.1073�pnas.0801544105 Campanella et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

ferred wild-type CD8� T cells (Fig. 5D). The marked inductionof IP-10, Mig, RANTES, MCP-1, and IFN-� seen in wild-typemice was reduced in CXCR3 KO mice. Adoptive transfer of

wild-type CD8� T cells restored the expression of these chemo-kines and IFN-� in CXCR3 KO mice to levels observed inwild-type mice. This finding suggests that CXCR3 guides CD8�

T cells into the brain, which results in the production ofadditional chemokines that then amplifies CD8� T cell recruit-ment to the brain independently of CXCR3.

IP-10 KO and Mig KO Are Partially Protected from CM. Because wefound that both IP-10 and Mig were highly induced in the brainby murine CM, we sought to evaluate the individual contribu-tions of these two ligands to the development of this syndrome.To do so, IP-10- and Mig-deficient mice were infected with P.berghei ANKA and observed for mortality from CM. Both IP-10-and Mig-deficient mice were partially protected from CM, withonly 30–40% mortality (Fig. 6 A and B) compared with 80–100%for wild-type mice, suggesting that both ligands are indepen-dently involved in the disease. To explore further the respectiveroles of IP-10 and Mig in CM, we performed immunhistochem-istry on brains of mice on days 0 and 8 after infection. Uninfectedmice showed very little staining for either ligand. Mig- andIP-10-deficient mice that had developed signs of CM were usedas controls to ensure specific antibody staining. After infectionand the development of CM, brains of wild-type mice showedmarked Mig staining on endothelial vessels (Fig. 6C). In con-trast, IP-10 staining was found widely disseminated in neuronsthroughout the brain parenchyma and only occasionally onendothelial cells (Fig. 6D). Immunofluorescence double stainingwith antibodies specific for neurons (TuJ-1) and astrocytes(GFAP) confirmed that IP-10 was expressed by neurons but notastrocytes (Fig. 6E). These data suggest that IP-10 and Mig havedistinct functions in the pathogenesis of CM.

DiscussionMortality rates in CM remain very high (10–30%) because of thelack of timely and effective treatments, which relates, in part, toour lack of understanding of the pathogenesis of this syndrome.To advance the understanding of this complex disease, bothhuman and animal studies are required, which complement anddirect each other. In recent years, human studies have foundassociations between levels of certain chemokines and CMmortality, which we are now studying further in a mouse modelof this disease.

In particular, a recent study found that among 36 biomarkers,serum IP-10 levels were the only variable independently associ-

Fig. 5. Wild-type CD8 T cells restore CXCR3 KO murine CM mortality. (A)Wild-type or CXCR3 KO splenocytes (Left) or CD8� T cells (Right) were pre-pared from P. berghei ANKA-infected mice 5 days after infection and i.v.transferred into CXCR3 KO mice 2 h before infection with P. berghei ANKA. (B)Brains of wild-type or CXCR3 mice with or without CD8� T cell transfer wereharvested on day 8 after infection, and brain-sequestered recipient CD8� T(Thy1.2) cells were analyzed by flow cytometry. (C) Representative primaryflow cytometric dot plots are shown after lymphocyte gating. The percentageof lymphocyte gate is shown for the indicated cell populations. (D) Chemokineand IFN-� expression in the brain was analyzed for mice from experimentsshown in C.

Fig. 6. IP-10 and Mig KO are protected from murine CM and are expressed on different cells in the brain. (A and B) Wild-type and Mig KO (A) or IP-10 KO (B)in the C57BL/6 background were infected with P. berghei ANKA, and mortality from murine CM was monitored twice daily. One representative experiment fromthree independent infections is shown with at least eight mice per group. (C–E) Representative immunohistochemistry (C and D) and immunofluorescence (E)of mouse brains harvested day 0 or day 8 after infection, staining for Mig (C) or IP-10 (D and E). Arrows show endothelial cells staining, arrowheads show neuronalstaining. (Scale bars: 20 �m.)

Campanella et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4817

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

ated with CM mortality and that CSF IP-10 concentrations werealso up-regulated in Ghanaian children who died of CM (9).Although the mouse model of CM caused by P. berghei ANKAdoes not reproduce all of the features of human CM, thesehuman findings are very similar to what we report here for IP-10and the other CXCR3 ligand Mig, which were the most highlyexpressed chemokines in the brains of mice with CM. Inaddition, we also found that splenic levels of both chemokineswere up-regulated 2-fold, similar to what was found for serumlevels in Ghanaian children. Interestingly, the human study alsofound that CM induced a greater fold increase in IP-10 levels inthe CSF than in the serum, which is also similar to what we foundin our murine study. Different human studies have reported thatdecreased RANTES serum levels were independently associatedwith CM mortality in Uganda (11), whereas another studyreported increased RANTES expression in the brain in childrenwho died of CM in Ghana (12). We similarly found decreasedRANTES expression in the spleen and increased expression inthe brain during murine CM. Taken together, these data suggestthat chemokine regulation is similar in the P. berghei ANKAmouse model to what has been described for children dyingof CM.

Although the human studies have shown that chemokine levelsare associated with mortality from CM, they cannot provideinformation about the functional roles of different chemokinesin the pathogenesis of this disease. Employing the mouse modelhas allowed us to interrogate the potential contribution ofspecific chemokines and chemokine receptors. We found thatmice deficient in CXCR3 were highly protected against murineCM. CD3� and CD3�CD8� T cell recruitment to the brain wasmarkedly reduced in CXCR3 KO mice, whereas NK cell recruit-ment was only modestly reduced. To establish whether CXCR3is directly involved in directing CD8� T cells to the brain duringmurine CM, we used a competitive transfer model in which bothwild-type and CXCR3 KO CD8� T cells were i.p. injected intothe same mouse. Similarly to what we found in the CXCR3 KOmice, a small number of CXCR3 KO OT-I cells were foundsequestered in the brain, but wild-type cells were 7-fold moreefficient in migrating to the brain. Adoptive transfers of in vivoactivated wild-type CD8� T cells restored murine CM suscep-tibility to CXCR3 KO mice and induced CXCR3-independentrecruitment CD8� T cells to the brain. Early IP-10 and Migexpression in the brain might be induced by the innate immunesystem, possibly in response to Toll-like receptor signaling, asone study reported that brain IP-10 and Mig expression de-pended on MyD88 signaling after P. berghei ANKA infection(13). IP-10 and Mig were the earliest chemokines up-regulatedin the brain, 6 days after infection, before CD8� T cell infiltra-tion to the brain. Our findings therefore suggest that CD8� Tcells recruited to the brain by IP-10 and Mig amplify theinduction of cytokines and additional chemokines that leads tothe recruitment of CD8� T cells independent of CXCR3.

One intriguing question in the chemokine field is the apparentredundancy of the chemokine system, with most receptors beingactivated by more than one chemokine and many chemokinesbinding to more than one receptor. To determine the respectiveroles of the CXCR3 ligands in CM, we analyzed IP-10 and MigKO mice in the P. berghei ANKA model. We found that IP-10and Mig KO mice were partially protected from CM, demon-strating that both of these CXCR3 ligands were needed for thedevelopment of CM. Immunohistochemistry revealed that thetwo chemokines may play distinct roles. Mig staining was foundmainly on brain endothelial cells, whereas IP-10 was expressedmainly on neurons. While our article was in review, Miu et al.(14) reported similar findings regarding the role of CXCR3 inmurine CM induced by P. berghei ANKA. However, by using insitu RNA hybridization, Miu et al. detected Mig expression inmicroglia and endothelial cells and IP-10 expression in astrocytes

and endothelial cells, but not neurons. Using immunohistochem-istry, we also detected Mig and IP-10 expression in endothelialcells. However, we detected IP-10 expression in neurons but notastrocytes, which we confirmed by using immunofluorescencedouble staining. In West Nile virus infection, IP-10 is expressedin neurons and directs the migration of CD8� T cells into thebrain (7). In HIV infection, IP-10 is expressed by neurons andhas a direct neurotoxic effect, and its CSF levels were found tocorrelate with HIV-associated dementia (15, 16). It is possiblethat IP-10 may contribute to the neurological impairment seenin human CM in an analogous manner through direct neuro-toxicity in addition to its role in CD8� T cell recruitment.

In summary, we have shown that CXCR3 and its two ligandsIP-10 and Mig are required for the development of murine CM.CXCR3 is critically required for early CD8� T cell sequestrationin the brain of infected mice. CXCR3� T cells induce theproduction of a second wave of chemokines in the brain, whichin turn induces further recruitment of CD8� T cells indepen-dently of CXCR3. Our findings suggest that CXCR3 and itsligands link innate immune cell activation and the delivery of animmunopathogenic adaptive immune response to the brain.

MethodsMaterials and Mice. C57BL/6 wild-type mice were purchased from the NationalCancer Institute, National Institutes of Health (NIH), Bethesda, MD. Breederpairs of CXCR3-deficient (CXCR3 KO) mice were a kind gift from G. Gerard(Children’s Hospital, Harvard Medical School, Boston, MA), and breeder pairsof Mig KO mice were a kind gift from J. M. Farber (National Institute of Allergyand Infectious Diseases, NIH, Bethesda, MD). IP-10 KO mice were generated inour laboratory and backcrossed into C57BL/6 for nine generations (17); CXCR3KO and Mig KO mice were backcrossed into C57BL/6 mice 10 and ninegenerations, respectively. All three KO mice strains were rederived to be freeof Helicobacter and Pasteurella pneumotropica. The OT-I TCR mice in theC57BL/6 background were obtained from Jackson Immunoresearch Labora-tories and crossed with CXCR3 KO mice. All protocols were approved by theMassachusetts General Hospital Subcommittee on Research and Animal Care.Antibodies were from BD Biosciences, unless stated otherwise.

Infections. Six- to 8-week-old female mice were injected i.p. with 2 � 106 P.berghei ANKA-infected red blood cells. Mortality was checked twice daily, andmice showing signs of severe CM (ataxia, convulsion, coma) between days 6and 10 after infection were killed. Parasitemia was assessed by tail bloodsmears stained with Giemsa every other day. Mice without CM developedanemia because of rising parasitemia and were killed on day 21.

Quantitative PCR. Total RNA from spleens and brains of mice either uninfectedor at different points after infection with P. berghei ANKA was isolated byusing TRIzol and subsequently with an RNAeasy kit (Qiagen). RNA was con-verted to cDNA and analyzed by qPCR as described before (18) by using theMX4000 multiplex quantitative PCR system (Stratagene).

Brain and Spleen Chemokine ELISA. Brains and spleens were homogenized inHBSS and protease inhibitor mixtures (Roche Applied Science) and centrifugedat 11,000 � g for 20 min. IP-10 and Mig levels in the supernatants weredetermined by using murine Quantikine immunoassay kits (R&D Systems).

Brain-Sequestered Leukocyte Isolation and Spleen Cell Isolation. Brain andspleen cells were obtained from mice at different times after infection. Brainswere removed after intracardial perfusion with 35 ml of PBS; they wereminced and digested for 45 min with Blendzyme (Roche) and DNase (Sigma–Aldrich) at 37°C with shaking. The cells were passed through a 70-�m cellstrainer and centrifuged on a 35% Percol for 40 min at 450 � g. The pelletswere washed, and the red blood cells were lysed (Sigma), counted, and stainedfor flow cytometry. Spleens were removed, passed through a 70-�m strainer,and red blood cells lysed and counted. For flow cytometry, cells were incu-bated for 10 min with 2.4G2 anti-Fc�III/II receptor (BD PharMingen) andstained with the following antibodies: FITC-conjugated anti-murine CD3,PE-Cy5.5 or allophycocyanin (APC)-conjugated anti-murine CD8, APC-conjugated anti-murine NK1.1, PE- or PercP-conjugated anti-murine Thy1.1,or APC-conjugated anti-murine Thy1.2 at 4°C for 20 min. Cells were fixed with1% paraformaldehyde, and cytofluorometry was performed by using a FACS

4818 � www.pnas.org�cgi�doi�10.1073�pnas.0801544105 Campanella et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

Caliber Cytometer (Becton–Dickinson) and analyzed with CellQuest or Flowjosoftware.

In Vivo OT-I Recruitment. OT-I cells from OT-I wild-type Thy1.1 and CXCR3 KOOT-I Thy1.2 were prepared and cultured with IL-2, IL-12, anti-CD28 andovalbumin antigen as described in ref. 19, and harvested after 5 days of in vitrostimulation with Lympholyte (Cedarlane). FACS analysis confirmed equalactivation of both cell types. Wild-type and CXCR3 KO cells were mixed (5 �106 each per mouse) and coinjected i.p. in 500 �l of HBSS into Thy1.1�Thy1.2mice infected 5 days earlier with P. berghei ANKA. When mice were at CMstage, the spleen and brain were harvested and processed as above. Trans-ferred cells were detected with Thy1.1 PercP and Thy1.2 APC antibodies.

Splenocyte and CD8� T Cell Transfer. Spleens of wild-type Thy1.1 mice or CXCR3KO mice were removed 5 days after infection with P. berghei ANKA and putthrough a 70-�m cell strainer. Red blood cells were lysed, and splenocyteswere washed three times. For CD8� T cell transfer, splenocytes were stainedwith FITC-conjugated anti-murine CD3 and APC-conjugated anti-murine CD8as described above. CD3�CD8� cells in the lymphocyte gate were sorted on afluorescence-activated Aria cell-sorting instrument (BD Biosciences) and were99% pure after sorting. Splenocytes (107) or CD8� T cells (106) were injected bytail vein into CXCR3 KO mice. Mice were infected with P. berghei ANKA 2 hafter cell transfer as described above.

Immunohistochemistry (IHC). Brains were removed after intracardial perfusionwith 35 ml of buffered formalin, sectioned (5 �m), and stained for H&E. For

IP-10 and Mig staining, paraffin-embedded 5-�m sections were deparaf-finized, and for IHC they were incubated in 3% H2O2 in methanol andsubsequently in 70% ethanol. Slides were pretreated with antigen decloakerin the decloaker (Biocare Medical) and blocked with horse serum/avidin(Vector Laboratories). The anti-mIP-10 or anti-mMig antibodies (both fromR&D Systems) were added at 1:200 dilution overnight. Slides were incubatedfirst with biotin, then biotinylated horse anti-goat IgG (1:200, 35 min; VectorLaboratories). For IHC, slides were treated with Elite (Vector Laboratories) for1 h, followed by Perm AEC solution (Biocare Medical), and counterstainedwith hematoxylin and lithium bicarbonate. For immunofluorescence, slideswere incubated with TuJ-1 antibody (1:200, overnight; R&D Systems) forneuronal staining, employing the MoM kit (Vector Laboratories), then rabbitanti-mouse Alexa Fluor 488 (1:500, 20 min; Invitrogen). For astrocyte staining,slides were incubated with anti-GFAP (1:200, 1 h; Dakocytomation) followedby donkey anti-rabbit Alexa Fluor 488 (1:500, 20 min). IP-10 staining wasperformed as above, employing rabbit serum, biotinylated rabbit anti-goatIgG (Vector Laboratories), and strep-Alexa Fluor 555 (1:250, 20 min; Invitro-gen). Slides were counterstained with DAPI.

Statistical Analysis. Statistical analysis was performed by Student’s two-tailedt test (unpaired) for means and by Cox–Mantel log-rank test for differences inmortality. A P value of �0.05 was considered significant.

ACKNOWLEDGMENTS. We thank Nicole Brousaides for technical assistance.This work was supported by National Institutes of Health Grant R01-CA69212(to A.D.L.).

1. Coltel N, Combes V, Hunt NH, Grau GE (2004) Cerebral malaria: A neurovascularpathology with many riddles still to be solved. Curr Neurovasc Res 1:91–110.

2. Hermsen C, van de Wiel T, Mommers E, Sauerwein R, Eling W (1997) Depletion of CD4�

or CD8� T cells prevents Plasmodium berghei-induced cerebral malaria in end-stagedisease. Parasitology 114:7–12.

3. Nitcheu J, et al. (2003) Combadiere B Perforin-dependent brain-infiltrating cytotoxicCD8� T lymphocytes mediate experimental cerebral malaria pathogenesis. J Immunol170:2221–2228.

4. Belnoue E, et al. (2003) CCR5 deficiency decreases susceptibility to experimentalcerebral malaria. Blood 101:4253–4259.

5. Viola A, Luster AD (2007) Chemokines and their receptors: Drug targets in immunityand inflammation. Annu Rev Pharmacol Toxicol 48:171–197.

6. Khan IA, et al. (2000) IP-10 is critical for effector T cell trafficking and host survival inToxoplasma gondii infection. Immunity 12:483–494.

7. Klein RS, et al. (2005) Neuronal CXCL10 directs CD8� T cell recruitment and control ofWest Nile virus encephalitis. J Virol 79:11457–11466.

8. Sui Y, et al. (2004) Neuronal apoptosis is mediated by CXCL10 overexpression in simianhuman immunodeficiency virus encephalitis. Am J Pathol 164:1557–1566.

9. Armah HB, et al. (2007) Cerebrospinal fluid and serum biomarkers of cerebral malariamortality in Ghanaian children. Malar J 6:147 (abstr).

10. Hansen DS, Bernard NJ, Nie CQ, Schofield L (2007) NK cells stimulate recruitment ofCXCR3� T cells to the brain during Plasmodium berghei-mediated cerebral malaria.J Immunol 178:5779–5788.

11. John CC, Opika-Opoka R, Byarugaba J, Idro R, Boivin MJ (2006) Low levels of RANTESare associated with mortality in children with cerebral malaria. J Infect Dis 194:837–845.

12. Sarfo BY, et al. (2004) The cerebral malaria-associated expression of RANTES, CCR3, andCCR5 in post-mortem tissue samples. Ann Trop Med Parasitol 98:297–303.

13. Griffith JW, et al. (2007) Toll-like receptor modulation of murine cerebral malaria isdependent on the genetic background of the host. J Infect Dis 196:1553–1564.

14. Miu J, et al. (2008) Chemokine gene expression during fatal murine cerebral malariaand protection due to CXCR3 deficiency. J Immunol 180:1217–1230

15. Kolb SA, et al. (1999) Identification of a T cell chemotactic factor in the cerebrospinalfluid of HIV-1-infected individuals as interferon-�-inducible protein 10. J Neuroimmu-nol 93:172–181.

16. Cinque P, et al. (2005) Cerebrospinal fluid interferon-�-inducible protein 10 (IP-10,CXCL10) in HIV-1 infection. J Neuroimmunol 168:154–163.

17. Dufour JH, et al. (2002) IFN-�-inducible protein 10 (IP-10; CXCL10)-deficient micereveal a role for IP-10 in effector T cell generation and trafficking. J Immunol168:3195–3204.

18. Means TK, Hayashi F, Smith KD, Aderem A, Luster AD (2003) The Toll-like receptor 5stimulus bacterial flagellin induces maturation and chemokine production in humandendritic cells. J Immunol 170:5165–5175.

19. Medoff BD, et al. (2005) BLT1-mediated T cell trafficking is critical for rejection andobliterative bronchiolitis after lung transplantation. J Exp Med 202:97–110.

Campanella et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4819

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020