Article available online at http://www.idealibrary.com on1 Microbial Pathogenesis 2002; 33: 225±237

doi:10.1006/mpat.2002.0531

Membrane sorting during swimminginternalization of Brucella is required forphagosome traf®cking decisionsSuk Kim, Masahisa Watarai*, Sou-ichi Makino &Toshikazu Shirahata

Department of Veterinary Microbiology, Obihiro University of Agriculture and Veterinary Medicine,Obihiro, Hokkaido 080-8555, Japan

(Received July 24, 2000; accepted in revised form September 5, 2002)

Brucella infects macrophages by swimming internalization, after which it is enclosed inmacropinosomes. We investigated the role of the uptake pathway in phagosome traf®cking, whichremains unclear. This study found membrane sorting during swimming internalization and is essentialin intracellular replication of Brucella. The B. abortus virB mutant replicated intracellularly when it wasin the macropinosome established by wild-type B. abortus that retained its ability to alter phagosometraf®cking. Lipid rafts-associated molecules, such as GM1 ganglioside, were selectively includedinto macropinosomes, but Rab5 effector early endosome autoantigen (EEA1) and lysosomalglycoprotein LAMP-1 were excluded from macropinosomes containing B. abortus induced byswimming internalization. In contrast, when the swimming internalization was bypassed by phorbolmyristate acetate (PMA)-induced macropinocytosis, lipid raft-associated molecules were excluded,and EEA1 and LAMP-1 were included into macropinosomes containing bacteria. The phosphati-dylinositol 3-kinase inhibitor wortmannin that inhibits PMA-induced macropinocytosis blockedinternalization of virB mutant, but not of wild-type of B. abortus and wortmannin treatment did notaffect intracellular replication. Our results suggest that membrane sorting requires swimminginternalization of B. abortus and decides the intracellular fate of the bacterium, and that Brucella-induced macropinosome formation is a different mechanism from PMA-induced macropinocytosis.

& 2002 Elsevier Science Ltd. All rights reserved.

Key words: Membrane sorting, macrophage, brucellosis, intracellular replication.

Introduction

Phagocytosisby`professional'phagocytes,whichare macrophages and neutrophils, is central tothe pathogenesis of intracellular pathogens,

0882±4010/02/$ ± See front matter

*Author for correspondence. E-mail:watarai@obihiro.ac.jp

such as Brucella, and is the initial step in thedegradation of dying cells, inert particles andlive infectious agents. It has a critical part inessential biological functions, such as in¯amma-tion, immunity and development [1]. Afteruptake by professional phagocytes, inert parti-cles are found in a membrane-derived phago-some, which undergoes maturation to a

& 2002 Elsevier Science Ltd. All rights reserved.

226 S. Kim et al.

hydrolase-rich phagolysosome. Intracellularpathogens control the membrane-derived pha-gosomes, circumventing host defences andfurther degrading and transforming their com-partments into replicative phagosomes [2, 3].

Brucella species are facultative intracellularpathogens that survive in a variety of cells,including macrophages, and their virulence andchronic infections are thought to be due to theirability to avoid the killing mechanisms withinmacrophages [4]. The molecular mechanisms oftheir virulence and chronic infections are incom-pletely understood. Recent studies with HeLacells have con®rmed that Brucella inhibits phago-some-lysosome fusion and transits through anintracellular compartment that resembles auto-phagosomes. Bacteria replicate in a differentcompartment, containing protein markers nor-mallyassociatedwith theendoplasmicreticulum,shown by confocal microscopy and immunogoldelectron microscopy [5±7]. Brucella internalizesinto macrophages by swimming on the cellsurface with generalized membrane ruf¯ing forseveral minutes, after which the bacteria areenclosed by macropinosomes [8]. Lipid raft-associated molecules, such as glycosylphospha-tidylinositol (GPI)-anchored proteins, GM1ganglioside and cholesterol, are selectivelyincorporated into macropinosomes containingB. abortus. The disruption of lipid rafts on macro-phagesmarkedlyinhibitsVirB-dependentmacro-pinocytosis and intracellular replication [8]. Thusreplicative phagosome formation of B. abortusis modulated by lipid raft microdomains.

We here describe the swimming internaliza-tion of B. abortus into macrophages is essential fortargeting properly and formation of replicativephagosome. Macropinosomes induced by PMA,which cannot alter phagosome traf®cking, aredistinct from macropinosomes formed after theswimming internalization. This indicates that theinternalization strategy of B. abortus into macro-phages contributes to membrane sorting and isan important role in phagosome traf®cking.

Results

Macropinosomes containing B. abortusare replicative phagosomes

Our previous results showed that B. abortusaltered phagosomes into a specialized organellepermissive for bacterial replication through

speci®c actions of the VirB system [8]. To furtherinvestigate if replicative phagosomes formed bywild-type B. abortus can support the intracellularreplication of virB mutant, bone marrow-derivedmacrophages were co-infected simultaneouslywith 544 (wild-type, GFP-negative) and Ba604(DvirB4 GFP-positive), and then macropino-somes containing both bacteria were identi®ed.After 15 min infection, macropinosomes contain-ing bacteria were analyzed by ¯uorescencemicroscopy. Seventy-nine percent of macropino-somes contained 544 (wild-type), and 18%macropinosome contained both 544 (wild-type)and Ba604 (DvirB4) [Figs 1(A)±(D), (I)]. To exam-ine in detail the distribution of bacteria insidemacropinosomes containing both strains, thenumber of bacteria inside macropinosomes wasscored. Fifty-two percent of the macropinosomescontained single bacteria of 544 (wild-type) andBa604 (DvirB4), and 21±25% of the macropino-somes contained single or double bacteria ofthese strains; macropinosomes containing morethan double bacteria of both strains were hardlydetectable in this procedure [Fig. 1(J)].

To determine whether synchronous uptake ofwild-type and mutant strains can rescue thereplication of the mutant strain, macrophageswere co-infected simultaneously with 544 (wild-type) and Ba604 (DvirB4) and then replicativephagosomes containing both strains were iden-ti®ed. After 24 h infection, macrophages harbor-ing bacteria were analyzed by ¯uorescencemicroscopy. Thirty-®ve percent of macrophagescontained no bacteria, and 25±27% of themacrophages contained 544 (wild-type) or/andBa604 (DvirB4) [Fig. 1(K)]. Macrophages contain-ing more than six bacteria of both strains wereobserved (18%), indicating that the replicatingmutant strain was rescued [Figs 1(E)±(H), (K)].Replication of virB mutant was only supportedin the macropinosome containing wild-typestrain, because it was not observed replicationof virB mutant only.

Internalization of B. abortus byPMA-induced macropinocytosis

To investigate if the swimming internalization ofB. abortus requires replicative phagosomeformation, PMA-treated macrophages wereinfected with bacteria. As PMA stimulatesboth membrane ruf¯ing and macropinocytosisin macrophages [9, 10], bacteria mayinternalize accidentally into macrophages by

Figure 1. Brucella abortus creates an organelle permissive for intracellular replication by VirB-dependentmacropinocytosis. (A±H) Mouse bone marrow-derived macrophages were infected simultaneously with 544(wild-type, GFP-negative) and Ba604 (DvirB4, GFP-positive), and ®xed after 15 min (A±D, I and J) or 24 h (E±Hand K) incubation. Red, anti-B. abortus staining; green, GFP-DvirB4 mutant. The DvirB4 mutant (green, panel B)internalized in a macropinosome that corresponds to phase contrast image (D) were localized by GFP¯uorescence and anti-B. abortus staining (yellow) with wild-type strain (red) in the merged image (C). TheDvirB4 mutant (green, panel F) replicating intracellularly that corresponds to the phase contrast image (H)were localized by GFP ¯uorescence and anti-B. abortus staining (yellow) with wild-type strain (red) in themerged image (G). White arrows point to both wild-type and DvirB4 mutant (A±H), and the blue arrow pointswild-type strain in macropinosome (A±D). (I±K) Number of B. abortus wild-type and DvirB4 mutant in co-infected macrophages were scored by ¯uorescence microscopy (see Experimental procedures). (I) Numbers ofmacropinosomes containing wild-type B. abortus (WT), DvirB4 mutant (virB4) and both strains (WT/virB4)were scored. Data for macropinosomes are from triplicate coverslips representing 50 macropinosomes percoverslip, and the error bars represent the standard deviation. (J) Distribution of bacterial numbers inmacropinosomes containing both strains. Numbers of bacteria in macropinosomes containing both strainsare shown in parentheses. Data for macropinosomes are from triplicate coverslips representing 30macropinosomes containing both strains per coverslip, and the error bars represent the standard deviation.(K) After 24 h infection, bacterial phagosomes were scored. Macrophages that were uninfected (no bacteria),harboring phagosomes containing only a single strain B. abortus (WT or virB4), harboring both strains (WT andvirB4) or containing replicating DvirB4 mutant (virB4 rescue) were scored. One hundred macrophages wereexamined per coverslip. Data are the average of triplicate samples from three identical experiments, and theerror bars represent the standard deviation.

Membrane sorting of Brucella 227

228 S. Kim et al.

macropinocytosis without the swimming intern-alization. Time-lapse videomicroscopy was usedto follow the internalization of Ba600 (wild-type)into macrophages. Bacteria moved from the siteof initial bacterial contact with PMA-untreatedmacrophages for several minutes [Fig. 2(A)].Internalization of Ba600 (wild-type) into PMA-treated macrophages, in contrast, was muchquicker and bacteria were enclosed by macro-pinosomes within a minute without swimming[Fig. 2(B)]. Some bacteria, however, showed thesame swimming internalization with PMA-treated macrophages as with PMA-untreatedmacrophages. From analysis of time-lapsevideomicroscopy, some Ba600 (wild-type) inter-nalized into PMA-treated macrophages bymacropinocytosis without swimming. To ana-lyze this further, samples were ®xed after briefincubation of macrophages with B. abortus. Thedifferences in the rate of phagocytosis and theformation of macropinosomes for PMA-treatedand -untreated macrophages were quantitated-microscopically at various times of incubation,using strategies that allowed as synchronous aninfection as possible (see Experimental proce-dures). In PMA-untreated macrophages, Ba604(DvirB4) was rapidly internalized, with most ofthe associated bacteria internalized before fur-ther incubation, but the internalization of Ba600(wild-type) was delayed [Fig. 2(C) and (E)]. By15 min after deposition of Ba600 (wild-type),55.6% of the bacteria were in macropinosomes.Phagosomes containing Ba604 (DvirB4) wererelatively devoid of the ¯uid phase marker[Fig. 2(D)]. In PMA-treated macrophages, incontrast, Ba600 (wild-type) rapidly internalizedand showed similar intracellular bacterial num-bers as Ba604 (DvirB4) [Fig. 2(F) and (H)]. Inmacropinosome formation, more than 20% ofboth wild-type and mutant strains were inmacropinosomes at 0 and 5 min [Fig. 2(G)].These results at 0 and 5 min after infectionindicated that PMA-induced macropinocytosiscan take up extracellular bacteria more ef®cientlyand that about 20±40% of bacteria internalizedinto macrophages by PMA-induced macropino-cytosis without swimming internalization.

Localization of membrane markers onB. abortus or PMA-inducedmacropinocytosis

To investigate differences in phagosome matur-ation between normal (PMAÿ) or abnormal

(PMA�) uptake pathways, the infected PMA-treated and -untreated macrophages werestained with several membrane markers. Wepreviously showed that lipid raft-associatedmolecules, such as GPI-anchored proteins,GM1 ganglioside and cholesterol, are selectivelyincorporated into macropinosomes containingBa600 (wild-type) [8]. To determine if GM1ganglioside is incorporated into PMA-inducedmacropinosomes containing B. abortus, themacropinosomes were probed with biotin-labeled CTB. In PMA-untreated macrophages,the kinetics and degree of association of CTB-labeled GM1 ganglioside with internalizedBa600 (wild-type) showed maximal associationafter 15 min incubation at 37�C [41%, Figs 3(A),4(A)]. The association of CTB-labeled GM1ganglioside with internalized Ba600 (wild-type)showed a marked high ef®ciency in macropino-somes [> 78%, Figs 3(A), 4(C)]. Colocalizationof CTB-labeled GM1 ganglioside with Ba604(DvirB4) was much less pronounced [Figs 3(A),4(A) and (C)]. In PMA-treated macrophages, thedegree of association of CTB-labeled GM1 gang-lioside with internalized Ba600 (wild-type) wasless than 23.3% after 15 min incubation at 37�C[Figs 3(A), 4(B)]. The association of CTB-labeledGM1 ganglioside with internalized Ba600 (wild-type) showed a lower ef®ciency in macropino-somes [542%, Figs 3(A), 4(D)]. Similar resultswere obtained from aerolysin-labeled GPI-anchored proteins and ®lipin-labeled cholesterol(data not shown). These results indicated thatabout 36% macropinosomes containing Ba600(wild-type) targeted improperly in PMA-treatedmacrophages, and thus membrane sorting mightcorrelate with the uptake pathway.

To con®rm these indications, acquisition ofRab5 effector early endosome autoantigen(EEA1) in macropinosomes was tested. In PMA-untreated macrophages, more than 88.2% of thephagosomes and 96.4% of the macropinosomescontaining Ba600 (wild-type) failed to colocalizewith EEA1, whereas Ba604 (DvirB4) was foundpredominantly colocalized with EEA1 [Figs 3(B),4(E) and (G)]. In PMA-treated macrophages, incontrast, acquisition of EEA1 of the phagosomesor macropinosomes containing Ba600 (wild-type) increased (10.3±22.4% or 32.1±46.3%,respectively) [Figs 3(B), 4(F) and (H)]. Similarresults were obtained from LAMP-1 acquisitionin PMA-treated and -untreated macrophages[Figs 3(C) and 4(I±L)]. In PMA-untreatedmacrophages, Ba600 (wild-type) preventedphagosome-lysosome fusion, and therefore

Figure 2. Effect of PMA treatment on delayed phagocytosis and macropinosome formation of B. abortus.(A and B) Selected time-lapse videomicroscopic images of wild-type B. abortus entry into mouse bone marrow-derived macrophages with (B) without (A) PMA treatment. Elapsed time in minutes is indicated at the bottomof each frame. Arrows point to bacteria. (C±H) Uptake and macropinosome formation increased by PMAtreatment. Bacteria were deposited onto bone marrow-derived macrophages with (F±H) or without (C±E)PMA treatment and incubated at 37�C for the periods indicated. Uptake (C, E, F and H) or macropnosomeformation (D and G) were quantitated as described (see Experimental procedures). C, D, F and G,100 macrophages were examined per coverslip; E and H, uptake ef®ciency by macrophages was determinedby protection of internalized bacteria from gentamicin killing. Black bars, Ba600 (wild-type); open bars,Ba604 (DvirB4). Data are the average of triplicate samples from three identical experiments, and the errorbars represent the standard deviation.

Membrane sorting of Brucella 229

Figure 3. Association of membrane markers with B. abortus, macropinosomes. Bone marrow-derivedmacrophages with (PMA�) or without (PMAÿ) PMA treatment were incubated with wild-type or virB4mutant of B. abortus, and GM1 ganglioside (A), EEA1 (B) or LAMP-1 (C) was localized byimmuno¯uorescence, as described (see Experimental procedures). Shown are merged images of the GFP(bacteria) and TRITC (GM1 ganglioside, EEA1 or LAMP-1) channels up-and-down with phase contrast imagesof the identical cells. Arrows point to bacteria.

230 S. Kim et al.

Figure 4. Kinetics of colocalization membrane markers with phagosomes containing B. abortus. Ba600 (wild-type) (black bars) or Ba604 (DvirB4) (open bars) was deposited onto macrophages with (Phagosome: PMA�) orwithout (Phagosome: PMAÿ) PMA treatment, then incubated for the periods indicated at 37�C before ®xationand probing with biotin-labeled CTB, anti-EEA1 or anti-LAMP-1 antibody (see Experimentalprocedures). `% GM1ganglioside, EEA1 or LAMP-1 positive' refers to percentage of internalized bacteriathat showed co-staining with the GM1 ganglioside (A and B), EEA1 (E and F) or LAMP-1 (I and J) based onobservation of 100 bacteria per coverslip. Data are the average of triplicate samples from three identicalexperiments, and the error bars represent the standard deviation. Macrophages with (Macropinosome:PMA�) or without (Macropinosome PMAÿ) PMA treatment incubated with Ba600 (wild-type) (black bars)or Ba604 (DvirB4) (open bars) for the indicated time at 37�C were probed for GM1 ganglioside, andmacropinosomes harboring bacteria were identi®ed by ¯uorescence and phase microscopy. Themacropinosomes were then observed for the presence of GM1 ganglioside (C and D), EEA1 (G and H) orLAMP-1 (K and L). `% GM1 ganglioside, EEA1 or LAMP-1 positive' refers only to those phagosomes with amacropinocytotic morphology, and represents the percentage of macropinosomes that show costaining withthe GM1 ganglioside, EEA1 or LAMP-1. Data for macropinosomes are from triplicate coverslips representing50 macropinosomes per coverslip, and the error bars represent the standard deviation. ND, not detectable.

Membrane sorting of Brucella 231

phagosomes containing Ba600 (wild-type) do nothave endocytic and lysosomal marker proteinLAMP-1 in macrophages [Figs 3(C), 4(I) and (K)].In PMA-treated macrophages, in contrast, Ba600

(wild-type) failed to block phagosome matur-ation comparatively, as shown by colocalizationof phagosomes or macropinosomes containingthe bacteria and the LAMP-1 at 35 min after

232 S. Kim et al.

infection (35.5 or 49.2% positive, respectively)[Figs 3(C), 4(J) and (L)]. Bacteria in LAMP-1positive phagosomes cannot replicate [11]. Thus,these results suggest that replicative phagosomeformation requires an uptake pathway associatedwith swimming internalization.

B. abortus-induced macropinocytosis is aPI3-kinase-independent mechanism

To further investigate if B. abortus-inducedmacropinocytosis requires replicative phago-some formation, the role of phosphoinositide3-kinase (PI3-kinase) in the internalization ofB. abortus was assessed. PI3-kinase is not neces-sary for receptor-mediated stimulation ofpseudopod extension, but rather functions inthe closure of macropinosomes and phagosomesinto intracellular organelles [12]. Ba600 (wild-type) or Ba604 (DvirB4) was infected with macro-phages treated with or without wortmannin, aninhibitor of PI3-kinase [13]. The internalizationand macropinosome formation of Ba600 (wild-type) were not affected markedly by wortmannin[Fig. 5(A)±(C)]. However, wortmannin treatmentappeared to decrease the internalization ability ofBa604 (DvirB4) [Fig. 5(A)±(C)]. Consistent withthese results, LAMP-1 acquisition of both Ba600(wild-type) and Ba604 (DvirB4) was not affectedmarkedly by wortmannin [Fig. 5(D)]. To deter-mine if PI3-kinase has a role in bacterial replica-tion in macrophages, macrophages were treatedwith wortmannin and then were infected withBa600 (wild-type) or Ba604 (DvirB4). Ba600 (wild-type) replicated in macrophages without drugtreatment. The intracellular replication of Ba600(wild-type) was not affected markedly by wort-mannin [Fig. 5(E)]. In contrast, Ba604 (DvirB4)failed to replicate in macrophages with or with-out wortmannin treatment [Fig. 5(E)].

Discussion

This study showed membrane sorting is duringthe swimming internalization of Brucella abortusinto mouse bone marrow-derived macrophagesand is essential for intracellular replication.These events were dependent on the presenceof the VirB system. As our previous studyshowed that B. abortus induced macropinosomeformation [8], we speculated that co-infected virBmutant with wild-type B. abortus was easy todetect in macropinosomes containing wild-type

B. abortus by microscopy. As expected, the virBmutant of B. abortus grew intracellularly whenthe bacteria were internalized by VirB-dependent macropinocytosis, suggesting thatB. abortus-induced macropinosomes supportintracellular replication of bacteria. However,this result suggested the important question ofthe role of macropinosomes in intracellularreplication of B. abortus. Macropinosomes aregenerally large organelles, and their surface-to-volume ratio is lower than for smaller vesicles.Macropinosomes internalize true ¯uid-phasesolutes more ef®ciently for each unit area ofmembrane than smaller vesicles [14]. Therefore,macropinosomes provide a comparatively ef®-cient, though non-selective, mechanism tointernalize extracellular macromolecules [15].

PMA and macrophage colony-stimulatingfactor stimulate macropinocytosis in macro-phages [9, 10]. To clarify our question of therole of macropinosomes in intracellular replica-tion of B. abortus, we investigated if B. abortusenclosed in PMA-induced macropinosomes canalter phagosome traf®cking. Our results showedthat PMA-induced macropinosomes containingwild-type B. abortus could not sort a phagoso-mal membrane, suggesting that the uptakepathway is more important than macropino-some formation for the alteration of phagosometraf®cking. Brucella abortus internalizes intomacrophages by swimming internalization [8].Presumably, the membrane sorting is duringswimming internalization, and then replicativephagosomes are established. Indeed, in thisstudy, membrane sorting did not occur whenbacteria internalized into macrophages bybypassing the swimming internalization. Brucellaabortus-induced macropinosomes are thought toresult from swimming internalization, becausebacteria moving round from the site of initialbacterial contact with the macrophages wereobserved. Macropinosome formation inducedby B. abortus is clearly distinct from thatinduced by PMA. The swimming of the bacteriaon the macrophage surface often lasted forseveral minutes with generalized plasma mem-brane ruf¯ing before eventual enclosure inmacropinosomes. In contrast, bacterial internal-ization into PMA-treated macrophages is muchquicker. After stimulation by macrophage col-ony stimulating factor or phorbol esters, circularruf¯es appear in macrophages. Membrane ruf-¯ing becomes longer and broader, and frequentlyclose into large macropinosomes [16]. Thecircular ruf¯es are clearly different from the

Figure 5. Effect of wortmannin treatment on B. abortus infection into macrophages. (A±C) Bacteria weredeposited onto macrophages in the presence or absence of wortmannin and incubated at 37�C for 15 min.Uptake (A), macropinosome formation (B) or gentamicin protection (C) was quantitated as described (seeExperimental procedures). Black bars: Ba600 (wild-type); open bars: Ba604 (DvirB4). One hundredmacrophages were examined per coverslip (A and B). Data are the average of triplicate samples from threeidentical experiments, and the error bars represent the standard deviation. (D) Kinetics of colocalization ofLAMP-1 with phagosomes containing B. abortus. Ba600 (wild-type) (black bars) or Ba604 (DvirB4) (open bars)was deposited onto macrophages with or without wortmannin treatment, then incubated for the periodsindicated at 37�C before ®xation and probing with anti-LAMP-1 antibody (see Experimental procedures).`% LAMP-1 positive' refers to percentage of internalized bacteria that showed co-staining with the LAMP-1,based on observation of 100 bacteria per coverslip. Data are the average of triplicate samples from threeidentical experiments, and the error bars represent the standard deviation. (E) Intracellular replication ofB. abortus in macrophages. Macrophages in the presence or absence of wortmannin were infected with Ba600(wild-type) or Ba604 (DvirB4) as described in Experimental procedures. Data points and error bars representthe mean CFU of triplicate samples from a typical experiment (performed at least four times) and theirstandard deviation.

Membrane sorting of Brucella 233

generalized plasma membrane ruf¯ing inducedby B. abortus.

Consistent with these observations, regulationof macropinosome formation induced byB. abortus was different from that induced byPMA. Inhibitor of PI3-kinase inhibits macropi-nocytosis, not by interfering with the initiationof the process but rather by preventing itscompletion [12]. PI3-kinase is necessary for

macropinocytosis and phagocytosis, but not formicropinocytosis or receptor-mediated stimula-tion of pseudopod extension [12]. PI3-kinasecontributes to a late step in the formation ofmacropinosomes and phagosomes, probablythe closure of pseudopodia to form intracellularvesicles [12]. Our results showed that regula-tion of macropinosome formation induced byB. abortus was a mechanism independent of

234 S. Kim et al.

PI3-kinase. Therefore, we concluded thatB. abortus stimulates macropinosome-like novellarge phagosome formation after swimminginternalization.

Macropinosome formation has been describedfor Legionella pneumophila, and macropinosomescontaining L. pneumophila included in lipid raft-associated molecules [17]. Legionella pneumophilamodulates traf®cking of phagosomes in whichthey reside [18]. Legionella pneumophila resides inphagosomes that restrict its fusion with hostendosomes and lysosomes [19] and its intracel-lular fate is decided by the type IV transportsystem, Dot/Icm apparatus [20]. PI3-kinaseinhibitor blocks phagocytosis of avirulent mu-tants, but not wild-type L. pneumophila, withoutaffecting membrane ruf¯ing and actin polymer-ization [21]. The mechanism of macropinosomeformation by B. abortus and L. pneumophila maybe similar, but swimming internalization has notbeen observed in L. pneumophila internalizationand the process of initial internalization ofL. pneumophila into mouse bone marrow-derivedmacrophages is still unclear.

This study showed that the early endosomalmembrane-tethering molecule EEA1 was ex-cluded in B. abortus-induced macropinosomes.The presence of EEA1 on phagosomes is arequisite for the subsequent acquisition of lateendocytic characteristics. As EEA1 has beenimplicated in traf®cking between the trans-Golgi network and endosomes [22] and inhomotypic fusion with early endocytic compart-ments [23], the exclusion of this Rab5 effectorfrom B. abortus-induced macropinosomes isresponsible for an orderly acquisition of lateendosomal membrane proteins. Consistent withthis, B. abortus-induced macropinosomes failedto colocalize with endocytic and lysosomalmarker LAMP-1. The intracellular pathway inHeLa cells of the virulent B. abortus strain 2308and the attenuated strain 19 is showed [6]. Bothbacterial strains are transiently detected inphagosomes characterized by the presence ofearly endosomal markers such as EEA1 [6].Differences between this study and that study[6] may be caused by using different types ofcell. Intracellular traf®cking of B. abortus phago-somes is different between professional phago-cytes and non-professional phagocytes [24].

EEA1 associates with oligomeric structurescontaining N-ethylmaleimide-sensitive factor(NSF) and Rabaptin-5/Rabex-5 on endosomalmembranes, and interacts directly with sintaxin13 [25]. The function of the soluble NSF

attachment protein receptor (SNARE) in mem-brane transport is regulated by Rab proteins andtheir effectors [26]. Rab effectors mediate initialdocking of vesicles to their target compartment,which must be synchronized with the priming ofSNAREs and formation of trans-paired SNAREcomplexes, ultimately resulting in lipid bilayerfusion [27]. Whether SNAREs are excluded fromB. abortus-induced macropinosomes is greatinterest, and needs further investigation.

Material and Methods

Reagents

Tetramethyl rhodamine isothiocyanate (TRITC)-dextran of molecular weight 155,000(TRD6155), phorbol myristate acetate (PMA),®lipin, and wortmannin were obtained fromSIGMA (St. Louis, MO, USA). Cholera toxin Bsubunit (CTB)-biotin conjugate was obtainedfrom List Biological Laboratories (Campbell,CA, USA). Alexa Fluor 594-streptavidin,Cascade blue-goat anti-rabbit IgG, Texas Red-goat anti-rat IgG were obtained from MolecularProbes, Inc (Eugene, OR, USA). Rhodamine-goatanti-rabbit IgG was obtained from ICN Phama-ceuticals (Aurora, OH, USA). TRITC-rabbit anti-goat IgG was obtained from Chemicon (Teme-cula, CA, USA). Anti-EEA1 goat polyclonalantibody was obtained from Santa Cruz Biotech-nology, Inc. (Santa Cruz, CA, USA). Anti-B.abortus polyclonal rabbit serum was describedpreviously [8, 28]. Anti-LAMP-1 rat monoclonalantibody 1D4B was obtained from Developmen-tal Studies Hybridoma Bank, of the Departmentof Pharmacology and Molecular Sciences, JohnsHopkins University School of Medicine, Balti-more, MD, USA and the Department of Biology,University of Iowa, Iowa City, IA, USA.

Bacterial strains and media

All B. abortus derivatives were from 544(ATCC23448), smooth virulent B. abortus biovar1 strains. Ba598 (544 DvirB4), Ba600 (544 GFP�)and Ba604 (Ba598 GFP�) were described previ-ously [8, 28]. Brucella abortus strains were main-tained as frozen glycerol stocks and werecultured on Brucella broth (Becton Dickinsonand Company, Cockeysville, MD, USA) or

Membrane sorting of Brucella 235

Brucella broth containing 1.5% agar. Kanamycinwas used at 40mg/ml.

Cell culture

Bone marrow-derived macrophages from femaleBALB/c mice were prepared as described[29]. After culturing in L-cell conditioned me-dium, the macrophages were replated for use bylifting cells in PBS on ice for 5±10 min, harvestingcells by centrifugation, and resuspending cells inRPMI 1640 containing 10% fetal bovine serum.The macrophages were seeded (2±36105 perwell) in 24-well tissue culture plates for allassays.

Time lapse video microscopy

Bone marrow-derived macrophages were platedin a Lab-Tek Chambered coverglass (NalgeNunc, Naperville, IL, USA) and were allowedto incubate overnight in RPMI 1640 containing10% FBS at 37�C in 5% CO2. Bacteria (26106/ml)were added to the chamber, which was thenplaced on a heated microscope stage set to 37�Cfor observation by using an Olympus IX70inverted phase microscope with 100X UPlanApolens ®tted with phase contrast optics. Thebacteria were allowed to pellet passively ontothe macrophages, and images were capturedover a 30 min period. If further observationswere desired, new samples were prepared andthe procedure was initiated again.

To capture images, the lens was focused on theupper surface of the macrophage. When bacteriabegan to appear within the focal plane of theimage, the images were captured every 15 susing a cooled charge-coupled device camera(CoolSNAP; Roper Scienti®c, Trenton, NJ, USA),and were processed using Openlab software(Improvision, Lexington, MA, USA) on a PowerMacintosh G4 computer.

Detection of intracellular bacteria andmacropinosome formation by using¯uorescence microscopy

Brucella abortus strains were grown to A600� 3.2in Brucella broth and used to infect mouse bonemarrow-derived macrophages for variousperiods at the indicated multiplicity of infection

(MOI). Bacteria in 250ml of RPMI 1640 contain-ing 1 mg/ml of TRD6155 were deposited ontothe macrophages by 1506g centrifugation for5 min at room temperature, and were thenincubated at 37�C, for 0 (no incubation), 5, 15,25, and 35 min. Infected cells were ®xed inperiodate-lysine-paraformaldehyde (PLP) [30]containing 5% sucrose for 1 h at 37�C. Thesamples were washed three times in PBS, andwere incubated three times for 5 min each inblocking buffer (2% goat serum in PBS) at roomtemperature. The samples were stained withanti-B. abortus polyclonal rabbit serum diluted1:1000 in blocking buffer to identify extracellularbacteria. After incubating for 1 h at 37�C, thesamples were washed three times for 5 min eachwith blocking buffer, were stained with Cascadeblue-conjugated goat anti-rabbit IgG diluted1:500 in blocking buffer, and were incubatedfor 1 h at 37�C. The samples were washed threetimes and were placed in mounting medium.One hundred macrophages were examined percoverslip to determine the total number ofintracellular bacteria, macropinosome forma-tion, and the total number of bacteria withinmacropinosome. Macropinosomes were de®nedas TRD6155-labeled phagosomes in whichdetectable TRD6155 surrounding the bacteriawas observed by ¯uorescence microscopy.

Co-infection of wild-type and �virB4strain of B. abortus

Ba604 (DvirB4 GFP�) and 544 (wild-type GFPÿ)were deposited onto bone marrow-derivedmacrophages by 1506g centrifugation at anMOI of 20 for 5 min at room temperature andwere then incubated at 37�C for 15 min or 24 h.Infected macrophages were ®xed in PLP-sucrosefor 1 h at 37�C. The samples were washed threetimes in PBS, and were successively incubatedthree times for 5 min each in blocking buffer (2%goat serum in PBS) at room temperature. Then,the samples were permeabilized in 0.1% TritonX-100, were washed three times with blockingbuffer and were incubated with anti-B. abortuspolyclonal rabbit serum diluted 1:1000 in block-ing buffer. After incubating for 1 h at 37�C, thesamples were washed three times for 5 min eachwith blocking buffer, were stained withrhodamine-goat anti-rabbit IgG diluted 1:500 inblocking buffer, and were incubated for 1 h at37�C. After three washes with PBS, the samples

236 S. Kim et al.

were placed in mounting medium and visua-lized by ¯uorescence microscopy.

Measurement of the ef®ciency ofbacterial uptake by culturedmacrophages

To measure the uptake of bacteria, mouse bonemarrow-derived macrophages were infectedwith B. abortus as described in the previoussection. After 0, 5, 15, 25 and 35 min incubationat 37�C, macrophages were washed once withmedia and were incubated with 30 mg/ml gen-tamicin for 30 min. Macrophages were thenwashed three times with fresh media and werelysed with distilled water. Colony-forming units(CFU) were determined by serial dilutions onBrucella plates. Percentage protection was cal-culated by dividing the number of bacteriasurviving the assay by the number of bacteriain the infectious inoculum, as determined byviable counts.

Measurement of the ef®ciency ofintracellular growth of bacteria

Bacteria were deposited onto macrophages at anMOI of 20 by centrifugation at 1506g for 5 minat room temperature, and were then incubatedat 37�C in 5% CO2 for 1 h. The macrophageswere then washed once with medium and wereincubated with 30 mg/ml gentamicin. At differ-ent time points, cells were washed and lysedwith distilled water, and the number of bacteriawas counted on plates of a suitable dilution.

LAMP-1 staining

Infected macrophages were ®xed in PLP-sucrosefor 1 h at 37�C and stained for extracellularbacteria as described above. All antibody-probing steps were for 1 h at 37�C. Sampleswere washed three times in PBS for 5 min andthen permeabilized in ÿ20�C methanol for 10 s.After incubating three times for 5 min withblocking buffer, samples were stained withanti-LAMP-1 rat monoclonal antibody 1D4Bdiluted 1:100 in blocking buffer [31]. Afterwashing three times for 5 min in blocking buffer,samples were stained simultaneously with TexasRed-goat anti-rat IgG. Samples were placed in

mounting medium and were visualized by¯uorescence microscopy. Intracellular bacteriawere detected by GFP ¯uorescence and byabsence of staining with Cascade blue.

Colocalization of proteins withmacropinosomes

To detect localization of GM1 ganglioside inmacropinosomes, macrophage monolayerswere incubated for 5 min with biotin-CTB(10mg/ml), were rinsed three times in RPMI1640 and were incubated with either Ba600(wild-type) and Ba604 (DvirB4) for the indicatedtime periods at 37�C [8]. The macrophages werewashed once, were ®xed in PLP-sucrose, wereprobed for extracellular bacteria, as above, andwere permeabilized in 0.05% saponin for 10 minat room temperature. After three washes withPBS and incubation in blocking buffer, the biotin-CTB was detected by using Alexa Fluor 594-streptavidin (1:500 in blocking buffer). To detectGPI linkages, the samples ®xed as above werepermeabilized in methanol at ÿ20�C for 10 s andwere probed with puri®ed aerolysin (2.5 mg/ml)for 1 h at 37�C. The antibody-probing steps ofaerolysin (1:1000), and EEA1 (1:100) were thesame as for LAMP-1 staining. To detect choles-terol, samples ®xed as above were incubated withthe ¯uorescent cholesterol-binding drug ®lipin(50 mg/ml) for 2 h at room temperature [8].

Drug treatment

PMA and wortmannin treatments were done bythe method of Araki et al. [12]. Brie¯y, mousebone marrow-derived macrophages were incu-bated with RPMI 1640 containing 30 ng/mlPMA or 100 nM wortmannin for 30 min at37�C. After washing with media containingPMA or wortmaninn, the macrophages wereinfected with bacteria as described in theprevious section.

Acknowledgements

This work was supported, in part, by grants fromGrant-in-Aid for Scienti®c Research (12575029 and13770129), Japan Society for the Promotion of Sci-ence, and by the Sasakawa Scienti®c Research Grantfrom The Japan Science Society.

Membrane sorting of Brucella 237

References

1 Lopes MF, Freire-de-Lima CG, DosReis GA. The macro-phage haunted by cell ghosts: a pathogen grows.Immunol Today 2000; 21: 489±94.

2 Dorn BR, Dunn WA Jr, Progulske-Fox A. Bacterial inter-actions with the autophagic pathway. Cell Microbiol2002; 4: 1±10.

3 MeÂresse S, Steele-Mortimer O, Moreno E, Desjardins M,Finlay B, Gorvel J. Controlling the maturation ofpathogen-containing vacuoles: a matter of life anddeath. Nat Cell Biol 1919; 1: 183±8.

4 Baldwin CL, Winter AJ. Macrophages and Brucella.Immunol Ser 1994; 60: 363±80.

5 Comerci DJ, Martinez-Lorenzo MJ, Sieira R, Gorvel J,Ugalde RA. Essential role of the VirB machinery inthe maturation of the Brucella abortus-containingvacuole. Cell Microbiol 2001; 3: 159±68.

6 Pizarro-Cerda J, MeÂresse S, Parton RG et al. Brucellaabortus transits through the autophagic pathway andreplicates in the endoplasmic reticulum of nonprofes-sional phagocytes. Infect Immun 1998; 66: 5711±24.

7 Pizarro-Cerda J, Moreno E, Sanguedolce V, Mege JL,Gorvel JP. Virulent Brucella abortus prevents lysosomefusion and is distributed within autophagosome-likecompartments. Infect Immun 1998; 66: 2387±92.

8 Watarai M, Makino S-I, Fujii Y, Okamoto K, Shirahata T.Modulation of Brucella-induced macropinocytosis bylipid rafts mediates intracellular replication. CellMicrobiol 2002; 4: 341±56.

9 Racoosin EL, Swanson JA Macrophage colony-stimulating factor (rM-CSF) stimulates pinocytosis inbone marrow-derived macrophages. J Exp Med 1989:170, 1635-48.

10 Swanson JA. Phorbol esters stimulate macropinocytosisand solute ¯ow through macrophages. J Cell Sci 1989;94: 135±42.

11 Watarai M, Makino S-I, Michikawa M, Yanagisawa K,Murakami S, Shirahata T. Macrophage plasma mem-brane cholesterol contributes to Brucella abortus infectionmice. Infect Immun 2002; 70: 4818±25.

12 Araki N, Johnson MT, Swanson JA. A role for phosphoi-nositide 3-kinase in the completion of macropinocytosisand phagocytosis by macrophages. J Cell Biol 1996; 135:1249±60.

13 Arcaro A, Wymann MP. Wortmannin is a potent phos-phatidylinositol 3-kinase inhibitor: the role ofphosphatidylinositol 3,4,5-trisphosphate in neutrophilresponses. Biochem J 1993; 296: 297±301.

14 Racoosin EL, Swanson JA. M-CSF-induced macropino-cytosis increases solute endocytosis but not receptor-mediated endocytosis in mouse macrophages. J CellSci 1992; 102: 867±80.

15 Watts C, Marsh M. Endocytosis: what goes in and how? JCell Sci 1992; 103: 1±8.

16 Swanson JA, Watts C. Macropinocytosis. Trend Cell Biol1995; 5: 424±8.

17 Watarai M, Derre I, Kirby J, Growney JD, Dietrich WF,Isberg RR. Legionella pneumophila is internalized by amacropinocytotic uptake pathway controlled by theDot/Icm system and the mouse Lgn1 locus. J Exp Med2001; 194: 1081±95.

18 Vogel JP, Isberg RR. Cell Biology of Legionella pneumo-phila. Curr Opinions Microbiol 1999; 2: 30±4.

19 Horwitz MA. The Legionnaires' disease bacterium(Legionella pneumophila) inhibits phagosome-lysosomefusion in human monocytes. J Exp Med 1983; 158:1319±31.

20 Christie PJ, Vogel JP. Bacterial type IV secretion: conju-gation systems adapted to deliver effector molecules tohost cells. Trend Microbiol 2000; 8: 354±60.

21 Khelef N, Shuman HA, Max®eld FR. Phagocytosis ofwild-type Legionella pneumophila occurs through awortmannin-insensitive pathway. Infect Immun 2001;69: 5157±16.

22 Simonsen A, Gaullier JM, D'Arrigo A, Stenmark H. TheRab5 effector EEA1 interacts directly with syntaxin-6.J Biol Chem 1999; 274: 28857±60.

23 Simonsen A, Lippe R, Christoforidis S et al. EEA1 linksPI(3)K function to Rab5 regulation of endosome fusion.Nature 1998; 394: 494±8.

24 Arenas GN, Staskevich AS, Aballay A, Mayorga LS.Intracellular traf®cking of Brucella abortus in J774 macro-phages. Infect Immun 2000; 68: 4255±63.

25 McBride HM, Rybin V, Murphy C, Giner A, Teasdale R,Zerial M. Oligomeric complexes link Rab5 effectorswith NSF and drive membrane fusion via interactionsbetween EEA1 and syntaxin 13. Cell 1999; 98: 377±86.

26 Novick P, Zerial M. The diversity of Rab proteins invesicle transport. Curr Opin Cell Biol 1990; 9: 496±504.

27 Weber T, Zemelman BV, McNew JA et al. SNAREpins:minimal machinery for membrane fusion. Cell 1998; 92:759±72.

28 Watarai M, Makino S-I, Shirahata T. An essential viru-lence protein of Brucella abortus, VirB4, requires an intactnucleoside triphosphate-binding domain. Microbiology2002; 148: 1439±46.

29 Watarai M, Andrews HL, Isberg RR. Formation of a®brous structure on the surface of Legionella pneumophilaassociated with exposure of DotH and DotO proteinsafter intracellular growth. Mol Microbiol 2001: 39: 313±29.

30 McLean IW, Nakane PK. Periodate-lysine-paraformal-dehyde ®xative: a new ®xative for immunoelectronmicroscopy. J Histochem Cytochem 1974; 22: 1077±83.

31 Swanson MS, Isberg RR. Identi®cation of Legionella pneu-mophila mutants that have aberrant intracellular fates.Infect Immun 1996; 64: 2585±94.