Preferential localization of IgG memory B cellsadjacent to contracted germinal centersYuichi Aibaa, Kohei Kometania, Megumi Hamadatea, Saya Moriyamaa,b, Asako Sakaue-Sawanoc, Michio Tomurad,Hervé Luchee, Hans Jörg Fehlinge, Rafael Casellasf, Osami Kanagawad, Atsushi Miyawakic, and Tomohiro Kurosakia,b,g,1

aLaboratory for Lymphocyte Differentiation, RIKEN Research Center for Allergy and Immunology, Yokohama, Kanagawa 230-0045, Japan; bLaboratory forLymphocyte Differentiation, Graduate School of Frontier Biosciences, Osaka University, Suita, Osaka 565-0871, Japan; cLaboratory for Cell Function andDynamics, RIKEN Brain Science Institute, Wako, Saitama 351-0198, Japan; dLaboratory for Autoimmune Regulation, RIKEN Research Center for Allergy andImmunology, Yokohama, Kanagawa 230-0045, Japan; eInstitute of Immunology, University Clinics Ulm, 89081 Ulm, Germany; fGenomics and Immunity,National Institute of Arthritis and Musculoskeletal and Skin Diseases, and Center of Cancer Research, National Cancer Institute, National Institutes of Health,Bethesda, MD 20892; and gLaboratory of Lymphocyte Differentiation, World Premier International Immunology Frontier Research Center, Osaka University,Suita, Osaka 565-0871, Japan

Edited* by Jeffrey V. Ravetch, The Rockefeller University, New York, NY, and approved May 21, 2010 (received for review April 21, 2010)

It has long been presumed that after leaving the germinal centers(GCs), memory B cells colonize the marginal zone or join therecirculating pool. Here we demonstrate the preferential localiza-tion of nitrophenol-chicken γ-globulin-induced CD38+IgG1+ mem-ory B cells adjacent to contracted GCs in the spleen. The memoryB cells in this region proliferated after secondary immunization,a response that was abolished by depletion of CD4+ T cells. Wealso found that these IgG1+ memory B cells could present antigenon their surface, and that this activity was required for their acti-vation. These results implicate this peri-GC region as an importantsite for survival and reactivation of memory B cells.

immunoglobulin | immunological memory

Appropriate interactions between antigen-specific B and Tlymphocytes are essential for humoral immune responses to

T-dependent antigens (1, 2). After their initial exposure to anti-gen, antigen-binding IgM+ B cells migrate from random locationswithin the B-cell follicles to the border between the follicles andthe T-cell–rich areas, where cognate interactions with antigen-specific CD4 T cells occur and subsequent B-cell proliferation isinduced (3, 4). Then, shortly after the appearance of extra-follicular foci of antibody (Ab)-secreting plasma cells, clusters ofisotype-switched cells such as IgG+ B cells, which can be detectedby staining with peanut agglutinin and the GL7 mAb, appear ingerminal centers (GCs) within the areas occupied by folliculardendritic cells (5–7). Because the Ig variable-region genes ofIgG+ long-lived memory B cells contain somatic mutations, andbecause these mutations occur primarily in GCs, it is thought thatIgG+ memory B cells are mainly derived from the GC.Despite the importance of IgG+ memory B cells in long-term

humoral memory (8, 9), their sites of residency and activationremain elusive, in part, because of technical difficulties associ-ated with in situ detection of the rare IgG+ memory B cellsspecific for a given antigen. Thus, to circumvent this problem,a transgenic mouse line harboring an IgM-type B-cell antigenreceptor (BCR) has been used; these studies showed that long-lived IgM+ memory B cells reside not just in the marginal zone(MZ), as had been thought, but also in splenic follicles (10).However, given the finding that antigen-experienced IgM B cellsand switched IgG2a B cells differentially localize during primaryimmune responses (11), extrapolation of the above scenario toresidency sites of IgG+ memory B cells in physiological settingsneeds to be done with great caution.Here we focus on where IgG+ memory B cells reside and how

these memory B cells are activated upon secondary antigenchallenge.

ResultsCD38+IgG1+ Memory B Cells Localize Around GCs. TheAb response tothe hapten nitrophenol (NP) has been characterized extensively

(12, 13). Thus, to determine where IgG-typememory B cells reside,we used this model system. The response to NP in C57BL/6 (B6)mice is dominated by Abs composed of the VH186.2 heavy chainand an Igλ light chain (12, 13). Consistent with these reports, byflow cytometry analysis, ∼60–80% of NP-specific IgG1+ B cellsexpressed λ light chains on day 60 after alum-precipitated NP-conjugated chicken γ-globulin (NP-CGG) immunization (Fig. S1A).We then used immunohistochemical analyses to determine thelocalization of IgG1+λ+ B cells, which are mixtures of memoryandGCB cells as described below, in the spleen on day 30 and day60 postimmunization. As shown in Fig. S1B, IgG1+ cells werelocated in several areas including the red pulps and follicles,but the λ-expressing IgG1+ cells were found mainly as clustersresiding in the centers of follicles both at day 30 and day 60.Although we expended considerable effort attempting to detectNP-reactive cells in sections using NP-labeled fluorochromes,in our hands the data obtained using this approach were un-convincing.Previous studies have demonstrated that CD38, used in con-

junction with GL7, is a good marker for distinguishing betweenmemory and GC B cells: The former are CD38+IgG1+ and thelatter are CD38−IgG1+GL7+ (14, 15). Therefore, to localizeIgG1-type memory B cells, we stained sections of spleen fromNP-CGG-immunized mice with anti-CD38, -IgG1, and -GL7 Abs. Onday 30, a small number of IgG1+CD38+ B cells were detected inthe follicles adjacent to clusters of CD38−GL7+ GC B cells (Fig.1A). On day 60, these IgG1+CD38+ B cells were still observednear the clusters of CD38−GL7+ GC B cells; the GC B cells werestill present, despite a significant reduction in their numberscompared with day 30 by both flow cytometry and immunohis-tology (Fig. 1B). Results using PNA, another widely used markerfor GC B cells, were similar to those with the GL7 mAb (Fig. S2).Moreover, consistent with the above results, ∼50–70% of IgG1+

cells clustering near the GCs on day 60 were λ+ (Fig. 1A). To-gether, these observations suggest that the majority of IgG1+

memory B cells localized near the contracted GCs on day 60 werespecific for NP.

IgM- and IgG-type Memory B Cells Localize in Distinct Areas of theSpleen. Several previous reports have described a different lo-calization of memory B cells from the one we describe above,that is, scattered in follicles or within the marginal zones (10, 16).

Author contributions: Y.A. and T.K. designed research; Y.A., K.K., M.H., and S.M. per-formed research; A.S.-S., M.T., H.L., H.J.F., R.C., O.K., and A.M. contributed new re-agents/analytic tools; Y.A. and K.K. analyzed data; and Y.A. and T.K. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. E-mail: kurosaki@rcai.riken.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1005443107/-/DCSupplemental.

12192–12197 | PNAS | July 6, 2010 | vol. 107 | no. 27 www.pnas.org/cgi/doi/10.1073/pnas.1005443107

Dow

nloa

ded

by g

uest

on

May

3, 2

020

Because previous reports mainly focused on memory B cellsexpressing IgM, we next examined the possibility that IgM- andIgG-type memory B cells might differentially localize in thespleen. For this purpose, we used mice that have a heavy-chainlocus targeted with the B1-8hi IgH (Igha) gene. Because the lightchains in B1-8hi mice are not fixed, only 3–5% of their B cellsexpress λ and bind NP (17). F1 offspring of CD45.1 and CD45.2congenic mice (CD45.1-CD45.2 F1) were adoptively transferredwith B cells from B1-8hi IgH knock-in mice (CD45.2) and im-munized with alum-precipitated NP-CGG. In this experimentalsetting, almost all donor-derived B cells present on day 30 afterimmunization can be considered as antigen-experienced, becauseof the following two lines of evidence. First, on day 30, more than90% of the transferred B cells bound NP, in contrast to about5% of naïve B cells (Fig. S3). Second, more than 90% of trans-ferred B cells labeled with carboxyfluorescein diacetate succini-midyl ester (CFSE) before transfer became CFSE-negative by day30 after immunization, indicating that they had undergone ex-tensive proliferation (Fig. S3). As shown in Fig. 2A, two distinctsubpopulations of donor-derived B cells were observed in immu-nized mice on day 60. The largest subpopulation was CD38+GL7−,

the phenotype of memory B cells, and most of these cells expressedIgM. As expected, those IgM+ cells could bind NP, indicating thatNP-reactive IgM-type memory B cells are generated in this ex-perimental system. Histological analysis revealed that the IgM-typememoryB cells detected with anti-IgMamAbweremainly scatteredin follicles (Fig. 2A and Fig. 2C Left).We next attempted to determine the localization of IgG1-type

memory B cells in the same adoptive transfer experiment as de-scribed above. However, despite tremendous efforts, the anti-allotypic IgG1a mAb did not work well, especially for histologicalanalysis. Thus, we tried another approach to detect donor-derivedIgG1-type memory B cells. We used a transgenic mouse expressinga cell-cycle-sensitive probe, fluorescent indicator for cell-cycle pro-gression (Fucci), in which cells become reversibly fluorescent de-pending on their cell-cycle status; they are red in the G1, but notS/G2/M, phases (Fucci-red) (18), therefore being suitable for la-beling resting cells such as memory B cells. CD45.1-CD45.2 F1mice were adoptively transferred with B cells from Fucci-red

A

B

Fig. 1. Detection of CD38+IgG1+ memory B cells adjacent to contracted GCs.(A) B6 mice were immunized with alum-precipitated NP-CGG. After 30 and60 days, spleens were harvested and subjected to flow cytometric (Upper)and immunohistological (Lower) analyses. B220+ gated populations areshown in the upper panels. A fluorescence-activated cell sorting (FACS)profile of an unimmunized B6 mouse (day 0) is also shown. White arrows inthe first and second rows of the second section indicate cells stained withboth anti-CD38 and -IgG1 Abs, and in the third row indicate cells that areIgG1+λ+. [Scale bars, 150 μm (Left), 50 μm (Center).] (B) Quantification of theabsolute number of memory (CD38+IgG1+) and GC (CD38−IgG1+) B cells byflow cytometric (Left) and immunohistological (Right) analyses.

None, Day 30

CD

45

.1

B220

NP-CGG + Alum, Day 60

CD

45

.1

B220

CD

38

IgG

1

IgMa

NP

-bin

din

g

GL7 IgMa

CD38, IgMa, GL7

CD38, IgMaCD38IgMa

A

B

N. o

f cells

/mm

2

0

2

4

6

8

10

0

2

4

6

8

10

RedPulp

Fo.Scat.

Fo.Close to GC

RedPulp

Fo.Scat.

Fo.Close to GC

IgMa+ cells IgG1+Fucci-red+ cellsC

IgG1, Fucci-redCD38, Fucci -red, GL7 CD38, Fucci-red

None

B2

20

Fucci-red

NP

-bin

din

g

IgG1

CD

38

IgG1

NP-CGG + Alum, Day 60

B2

20

Fucci-red

Fig. 2. IgM- and IgG-type memory B cells differentially localize in the spleen.(A) B cells from B1-8hi IgH knock-in mice were adoptively transferred toCD45.1-CD45.2 F1 mice. The mice were immunized with alum-precipitatedNP-CGG or left untreated. After 30 days, unimmunized mice were killed andanalyzed for the presence of donor-derived, B220+CD45.1− cells by flow cyto-metric analysis (Upper Left; None, day 30). Immunized mice were killed at60 days after immunization (NP-CGG + Alum, day 60). Spleens were subjectedtoflow cytometric (Upper) or immunohistological (Lower) analyses. [Scale bars,300 μm (Left), 20 μm (Right).] (B) B cells purified from double-transgenic mice(B1-8hi IgH knock-in and Fucci-red) were adoptively transferred to CD45.1-CD45.2 F1 mice. The mice were immunized with alum-precipitated NP-CGG orleft untreated. After 60 days, spleens fromunimmunized (None) or immunized(NP-CGG + Alum, day 60) mice were subjected to flow cytometric (Upper) andimmunohistological (Lower) analyses. The panels of histological analysis rep-resent the same region in the same follicle in different sections. [Scale bars,300 μm (Left), 50 μm (Center), 100 μm (Right).] (C) Quantification of IgMa+ cellsinA and IgG1+Fucci-red+ cells in B are shown. Average number± SD from threemice are shown. Cells in the red pulp were probably plasma cells, as judgedby low expression of CD38. Fo. Scat., IgG1+ cells scattered among follicles.

Aiba et al. PNAS | July 6, 2010 | vol. 107 | no. 27 | 12193

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

May

3, 2

020

transgenic B1-8hi IgH knock-in mice, and immunized with NP-CGG precipitated in alum. As shown in Fig. 2B, donor-derivedFucci-red-labeled cells were observed in mice on day 60 after im-munization. Flow cytometric analysis revealed that IgG1+ cellswere present in the Fucci-red-labeled cell population. Because wegated out Fucci-red-negative cells, probably including GC B cells,most of these IgG1+ cells expressed CD38 and were able to bindNP, suggesting that NP-specific IgG1-type memory B cells werelabeled with Fucci-red. Localization of Fucci-red-labeled cellswas examined by immunohistological analysis (Fig. 2B Lower).Clustering of Fucci-red-labeled CD38+ cells was observed nearGL7+CD38− GC cells, in addition to cells scattered in follicles.Staining the sections with anti-IgG1 Ab revealed that IgG1+

cells were predominantly located in clusters near GCs (Fig. 2 Band C Right). Together, these data demonstrate differential lo-calization of IgM- and IgG1-type memory B cells.

Detection of Memory B Cells Based on the AID-Cre–MediatedExpression of Red Fluorescent Protein. In addition to B cells, follic-ular dendritic cells are located within and/or surrounding GCs,where they can capture antigen–IgG complexes by virtue of Fcγ(19) and complement (20) receptors during immune responses.Because the above assay relies on anti-IgG1 staining, one concernis that this method might not be detecting B cells but instead cellsthat had passively acquired secreted IgG1 Abs. To eliminate thispossibility, we used a genetic approach in which the Cre recombi-nase gene is expressed under the control of the Aicda (activation-induced cytidine deaminase; AID) promoter and red fluorescentprotein (RFP) is only expressed upon Cre-mediated deletion ofa floxed neomycin gene (AID-cre/RFP-ROSA) (Fig. S4) (21, 22).In these mice, the progeny of AID-expressing cells, includingmemory B cells, are permanently RFP+ (Fig. S5A).Localization of the red fluorescent cells was determined by

immunohistological analysis (Fig. S5B). Consistent with the aboveimmunohistological analysis using anti-IgG1 Ab, RFP+CD38+

B cells were also located near the GL7+GCB cells (Fig. S5B Left).Most of these RFP+CD38+ B cells were IgG1+ (Fig. S5B Right).Based on these findings, we conclude that most IgG1 memoryB cells in the spleen are found in clusters near the contracted GCs.

Location of the CD38+IgG1+ Memory B Cells That ProliferateFollowing Antigen Rechallenge. To next examine where memoryB cells are activated upon secondary antigen challenge (NP-CGGwithout alum), we used another strain of Fucci transgenic mice, inwhich cells become reversibly fluorescent (green) in the S/G2/M,but not G1, phases (Fucci-green) (18). As expected, flow cyto-metric analysis indicated that almost all of the CD38+IgG1+NP+

memory B cells on day 60 after primary challenge were in theresting state. However, by 2 days after secondary challenge, manyof them had entered the cell cycle and become green (Fig. 3A). Insitu studies were performed to identify the location of the greencells in the spleen (Fig. 3B). Before secondary challenge, a fewgreen B cells were detectable, consistent with the flow cytometricdata (Fig. 3A). Two days after secondary challenge, many IgG1B cells entered the S/G2/M phase. Among these green cells, abouttwo-thirds of them were CD38+ memory cells and the remainingwere CD38− GC B cells. These results suggest that IgG1 memoryB cells near the contracted GCs, together with the GC B cells,start to proliferate upon secondary challenge.

CD4+ T Cells Reside Close to IgG1+ Memory B Cells in the Follicles.Considering the recent evidence that some T cells, particularlyfollicular helper T cells (TFH), are localized inside or surroundingGCs during primary humoral responses (23–25), it seemed possiblethat helper T cells for activating memory B cells might also residenear the contracted GCs. If so, in contrast to the requirement formigration of naïve B and naïve T cells toward the T-B border areafor their initial cognate interactions, such active migration might

not necessarily be required for activating memory B cells. Thispossibility was tested by immunohistological analysis of spleensections onday 60 after primary immunization.As shown inFig. 4A,CD4+ T cells were found near the region where CD38+IgG1+ andCD38−IgG1+ B cells were localized in the follicles. To further ex-amine whether these CD4+ T cells express TFH markers, we usedanti-PD-1mAb (26) and found that some, but not all, CD4+ T cellsin the follicles on day 60 after primary immunization expressedPD-1 (Fig. 4B).

Cognate Interaction of IgG1+ Memory B Cells with CD4+ T Cells IsRequired for Their Activation. The presence of T cells near theIgG1 memory B cells prompted us to examine the functional re-quirement for such helperT cells in humoralmemory responses. Toaddress this question, B6 or Fucci-green transgenic mice that hadbeen immunized with alum-precipitated NP-CGG were treatedwith anti-CD4 mAb and control Abs before secondary challenge(NP-CGGwithout alum). The initial proliferation of thememory Bcells, as judged by the expression of the Fucci-green probe (Fig.4C), as well as the production of secondary anti-NP Abs (Fig. 4D)were almost completely abolished by the anti-CD4 treatment.Having demonstrated the importance of CD4+ T cells for

activating IgG1 memory B cells, we wished to address whethercognate interactions between the B and T cells are required.Before addressing this question, we examined whether the IgG1memory B cells are capable of presenting antigen. To do this,NP-CGG-primed B6 mice were boosted with an NP-conjugatedfusion protein composed of GFP and amino acids 46–74 of the

Fucci-green IgG1

CD38

0

5

10

15

20

25

N. o

f ce

lls/m

m2

N. o

f ce

lls/1

06

sp

len

ocyte

s

P P+C P P+C P P+C P P+C P P+C P P+C

CD38+

IgG1+

NP-binding

CD38+

IgG1+

NP-binding

FG-

CD38+

IgG1+

NP-binding

FG+

CD38+

IgG1+

CD38+

IgG1+

FG-

CD38+

IgG1+

FG+

0

20

40

60

80

100

120

A

B

C

Prime alone Prime + Challenge, Day 2

Fucci-green

IgG1

CD38

Fucci-green IgG1

CD38 Fucci-green

IgG1

CD38

NP

-bin

din

g

IgG1 Fucci-green

CD

38

NP

-bin

din

g

IgG1 Fucci-green

CD

38

Prime alone Prime + Challenge, Day 2

88.8 0.7 55.7 20.9

P < 0.05

P < 0.01

Fig. 3. Activation of CD38+IgG1+ cells after antigen rechallenge. Fuccitransgenic mice were immunized with alum-precipitated NP-CGG. After 60days, mice were injected with NP-CGG without adjuvant (Prime + Chal-lenged, day 2) or left untreated (Prime alone) and were killed 2 days later.Spleens were subjected to flow cytometric (A) and histological analysis (B).Percentages of CD38+Fucci− and CD38+Fucci+ cells among NP-binding IgG1+

cells are indicated in the FACS profiles (A). White arrows in the right panel ofB indicate the Fucci probe-positive cells stained with anti-CD38 and -IgG1Abs. (Scale bars, 50 μm.) (C) Quantification of CD38+IgG1+ cells as eithernegative (FG−) or positive (FG+) for the Fucci probe was performed as de-scribed in Fig. 1B. P values were calculated with a two-tailed Student’s t test.P, primed alone; P+C, primed and rechallenged.

12194 | www.pnas.org/cgi/doi/10.1073/pnas.1005443107 Aiba et al.

Dow

nloa

ded

by g

uest

on

May

3, 2

020

I-Edα MHC II subunit (NP-EαGFP) as a secondary challengeantigen. In this setting, antigen presentation can be monitoredwith the Y-Ae mAb, which is specific for pEα:I-Ab complexeson the surface of antigen-presenting cells (27, 28). As demon-strated in Fig. 5A, CD38+IgG1+ memory B cells were able topresent exogenous antigens in the context of MHC class IImolecules. In these experiments, IgM+ and IgD+ cells were notdepleted before flow cytometric analysis for technical reasons,and thus the large number of NP-binding IgG1− cells is prob-ably IgM+ B cells.Next, to test the requirement for antigen presentation by mem-

ory B cells in their activation, sorted CD38+IgG1+NP+ memoryB cells were transferred together with CGG-primed CD4+ T cells

into Rag1−/− mice. The memory B cells were able to mount a sec-ondary Ab response when the recipient mice were administeredNP-CGG, but not NP-chicken ovalbumin (NP-OVA) (Fig. 5B).The simplest explanation for these results is that IgG1+NP+

memory B cells efficiently differentiated only when expression ofan NP-specific IgG1 BCR allowed efficient uptake of CGG andpresentation of CGG peptides to the CGG-primed T cells. NPconjugated with a different carrier, OVA, would be unable to elicitthe requisite T-cell help in this system. This idea was substantiatedby our findings usingMHC class II−/−Rag1−/−mice as recipients, anexperimental setting in whichMHC class II is only expressed on thetransferred cells. As shown in Fig. 5C, transfer of memory B cellstogether with primed T cells into the MHC class II−/− Rag1−/− mice

C

D

CD

38

Fucci-green

Control Ig treated Anti-CD4 mAb treated

Prime alone Prime + Challenge Prime alone Prime + Challenge

1.0 19.3 0.9 1.8

0

10

20

30

40

0

2

4

6

8

10

N.o

fce

lls/m

m2

N.o

fce

lls/1

06sp

len

ocyte

s

P P+C P P+C

Control Ig Anti-CD4 mAb

P< 0.05P< 0.05

P P+C

Tite

ro

fa

nti-N

PhiIg

G1

inse

ra(A

rbitra

ryU

nits)

1

10

102

103

Control Rat Ig treated

Anti-CD4 mAb treated

P< 0.05

P P+C P P+C

Control Ig Anti-CD4 mAb

CD38 IgG1

CD4 CD38IgG1CD4

PD-1

CD4

IgG1

PD-1IgG1CD4

A B

Fig. 4. Depletion of CD4+ cells prevents activation of memory B cells and abolishes the secondary Ab response. B6 mice were immunized with alum-precipitated NP-CGG. On day 60, spleens harvested from the mice were subjected to immunohistological analysis. (A) A section stained with anti-CD38, -IgG1,and -CD4 mAbs. (Scale bar, 50 μm.) (B) A section stained with anti-PD-1, -IgG1, and -CD4 Abs. White arrows in B indicate cells stained with both anti-CD4and -PD-1 mAbs. (Scale bar, 50 μm.) Fucci-green transgenic (C ) or B6 (D) mice were immunized with alum-precipitated NP-CGG. After 60 days, mice weretreated daily with anti-CD4 mAb or control rat Ab before rechallenge. Three days after the first Ab treatment, mice were rechallenged with NP-CGGwithout adjuvant (Prime + Challenge) or left untreated (Prime alone). (C ) Spleens from Fucci-green transgenic mice were harvested 2 days afterrechallenge and subjected to flow cytometric and histological analyses. Percentages of CD38+Fucci+ cells among NP-binding IgG1+ cells are indicated inthe FACS profiles. (D) Anti-NP IgG1 titers in sera from immunized B6 mice were measured by ELISA 7 days after rechallenge. P, primed alone; P+C, primedand rechallenged. P values were calculated with a two-tailed Student’s t test.

Aiba et al. PNAS | July 6, 2010 | vol. 107 | no. 27 | 12195

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

May

3, 2

020

could elicit the secondary NP response, whereas administration ofFab fragments of anti-MHC class II mAb blocked this response,demonstrating the importance of MHC class II expression onmemory B cells.

DiscussionPrevious work has shown that NP-specific memory B cells appearin the MZ area shortly after primary immunization in rats, and

then emerge in both the follicle and the MZ at later time points(16). Although the study using rats did not distinguish whetherthese memory B cells were of the IgM or class-switched type,given that the same localization sites was observed when NP-specific IgM transgenic mouse models were immunized (10) it ismost likely that the NP-specific memory B cells observed in therat system were of the IgM type. This possibility is further sup-ported by our observations that antigen-experienced IgM+ cellsare scattered in the follicles on day 60 after primary immuniza-tion (Fig. 2 A and C).In contrast to the localization of IgM-type memory B cells, we

have shown here that IgG1-type memory B cells (IgG1+CD38+)are mainly located near the contracted GC-like structures stillpresent on day 60 after primary immunization (NP-CGG withalum). Our histochemical resolution did not suffice to concludewhether the IgG1+CD38+ memory B cells are localized near theGC light or dark zone. The presence of these GC-like structureson day 60, albeit much smaller than on day 30, is quite consistentwith a recent report demonstrating that GC-like structures per-sist for up to 8 months after being challenged with sheep redblood cells (SRBCs) twice (29). Because SRBCs induce a verypotent polyclonal B-cell response, the persistence of GC-likestructures for longer periods in the case of SRBCs probablyreflects the fact that a steady-state level of newly activated B-cellclones is high, thereby being continuously recruited into the GCfractions. Together with our data, it now seems clear that GC-like structures can persist longer than previously appreciated,and that the duration of such structures is dependent, at leastpartly, on the nature of the immunogen and adjuvants.By using lymphocytes harboring the Fucci cell-cycle tracker,

we have demonstrated here that IgG1+CD38+ memory B cellsnear GCs, in addition to the IgG1+CD38− (GC B cells from ourcriteria), begin to proliferate upon secondary challenging on day60. As discussed above, in the NP-CGG immunization protocol,some IgG1+CD38− GC B cells still remain at day 60 and theirproliferation appears to be enhanced upon secondary challenge.Because recent reports have suggested that IgG-type memoryB cells undergo differentiation into plasma cells rather thanentering into GC pools (29, 30), we favor the idea that theprecursors of proliferating IgG1+CD38− B cells on day 2 aftersecondary challenge are GC, but not memory, B cells. However,at present, we cannot completely exclude the possibility thatIgG1+CD38+ memory B cells differentiate into IgG1+CD38−

cells, or vice versa, during the 2 days after secondary challenge.Assuming that such preferential localization and activation ofmemory B cells near the contracted GCs also occur in draininglymph nodes, our observations would explain the previous find-ings that the ipsilateral lymph node transfers a significantlyhigher humoral memory response than does the contralateralnode at all intervals after a primary challenge (31).Proliferation of IgG1+CD38+ memory B cells and their sub-

sequent differentiation into plasma cells are likely to requirecognate interactions between B and T cells. This notion is sup-ported by the following three lines of evidence: (i) the re-quirement for T cells for the proliferation and differentiation ofIgG1 memory B cells; (ii) IgG1+CD38+ memory B cells are ableto present antigens; and (iii) the requirement for MHC class IIon B cells for differentiation of IgG1+CD38+ memory B cells.Given that CD4+ T cells exist in close proximity to IgG1 memoryB cells near the contracted GCs on day 60, we speculate thatsome of these CD4+ T cells are long-lived follicular helpermemory T cells and are responsible for activation of IgG1memory B cells. If so, we would propose that this close proximityof memory B cells to memory T cells can explain, at least partly,the more rapid kinetics of memory responses because, duringprimary responses, movement of antigen-specific B cells andantigen-specific T cells toward the T-B border area is required.

None NP-CGG NP-OVA

N. o

f A

FC

s/1

07

sp

len

ocyte

s

Rechallengedwith

1

102

104

106

MHC II

Rag1

+/+

-/-

-/-

-/-

106

102

1

N. o

f A

FC

s/1

07

sp

len

ocyte

s

Recipient genotype

104

-/-

-/-

Anti-MHC class II Fab

- - +

B

C

3.3

53.1

Y-Ae

Y-Ae

CD

38

CD

38

IgG1

IgG1

NP

-bin

din

gN

P-b

ind

ing

Prime alone

Prime + NP-EaGFP

A

Fig. 5. Cognate interaction with antigen-specific T cells is required for theresponse of IgG-type memory B cells. (A) B6 mice were immunized with alum-precipitated NP-CGG. After 60 days, mice were injected i.p. with NP-EαGFP(Prime + NP-EαGFP) or left untreated (Prime alone). After 24 h, splenocyteswere analyzed by flow cytometry with Y-Ae mAb for estimating antigen pre-sentation in the context ofMHC class II. Percentages of CD38+Y-Ae+ cells amongNP-binding IgG1+ cells are indicated in FACS profiles. (B and C) Memory B cellsand CGG-primed T cells were purified from B6 mice immunized with alum-precipitated NP-CGG. (B) Cells were transferred to Rag1−/−mice. One day later,the mice were left untreated (None) or rechallenged with NP-CGG or NP-OVAwithout adjuvant. (C) Cells were transferred to Rag−/− or Rag−/− MHC class II−/−

mice. After 1 day, mice were immunized with NP-CGG without adjuvant. Onday 0 and day 3 after cell transfer, somemice were injected with 200 μg of Fabfragments of anti-MHC class IImAb. In bothB and C, splenocyteswerepreparedand subjected to enzyme-linked immunospot (ELISPOT) assays for measuringthe number of antibody-forming cells (AFCs) 7 days after cell transfer.

12196 | www.pnas.org/cgi/doi/10.1073/pnas.1005443107 Aiba et al.

Dow

nloa

ded

by g

uest

on

May

3, 2

020

In addition to this spatial advantage, intrinsic properties ofIgG-type memory B cells likely contribute to their more rapidresponse as demonstrated in previous studies. For example, IgG1-type memory B cells are more apt to differentiate into plasmacells than IgM-type memory and/or naïve cells (29, 30). At themolecular level, the cytoplasmic domain of the IgM-type BCR isalmost nonexistent, whereas the IgG-type BCR contains an ex-tended cytoplasmic domain which has been suggested to generateunique signals for conferring a memory phenotype (32, 33).Although our study has focused on IgG-type memory B cells,

we do not argue that antigen-specific IgM memory B cells aredispensable for T-dependent memory IgG responses. In regardto their contribution, it has been recently proposed thatmemory IgM B cells, by virtue of their rapid mobilization inGCs and switching to IgG after antigen rechallenge, ensurereplenishment of the memory pool, probably including bothIgM and IgG types (29).Our data, together with recent work (29), strongly suggest the

importance of persisting GC-like structures for normal humoralmemory responses. However, previous studies involving geneticdeficiency of CD40 and Bcl6 in human and mouse, respectively(34, 35), demonstrated that in the absence of GCs, an unmutatedbut functional memory B-cell compartment could develop, im-plying that there may exist at least one, if not multiple, GC-independent pathways of memory B-cell development. However,the IgG memory B cells that develop in the absence of Bcl6appear not to give rise to long-lived plasma cells (34). Thus, it

is possible that there are functional differences between GC-dependent and -independent IgG memory B cells.

MethodsDetailed descriptions of all materials and methods are provided inSI Methods.

Mice, Immunization, and Treatment with Anti-CD4 Monoclonal Antibody.C57BL/6 mice were purchased from CLEA Japan. B1-8hi IgH knock-in micewere described in ref. 17. AID-cre mice and RFP-ROSA mice have been de-scribed previously (21, 22). MHC class II-deficient mice (36) were crossed withRag1−/− mice to obtain Rag1−/− MHC class II−/− mice. Fucci-red and -greentransgenic mice were established as described in a previous report (18). Linesof transgenic mice were screened for the expression of the Fucci probe inperipheral blood leukocytes. For primary immunization, mice were injectedi.p. with 100 μg of NP-CGG in 200 μL of alum (Thermo Scientific) according tothe manufacturer’s instructions. For secondary immunization, mice wereinjected i.p. with 50 μg of NP-CGG or EαGFP-NP in PBS without any adjuvant.For in vivo depletion of CD4+ cells, mice were injected i.p. with 200 μg ofanti-CD4 mAb (clone; YTS191.1) every day for 3 dsya before rechallenge withNP-CGG. All of the protocols for animal experiments were approved by theRIKEN Animal Research Committee.

ACKNOWLEDGMENTS. We thank Dr. K. A. Pape and Dr. M. K. Jenkins(Department of Microbiology, Center for Immunology, University of Minne-sota Medical School, Minneapolis) for providing us with the EαGFP construct,Dr. T. Okada for discussions, and Dr. P. D. Burrows for critical reading of ourmanuscript. This work was supported by grants to Y.A. and T.K. from theMinistry of Education, Culture, Sports, Science, and Technology in Japan andby a grant to T.K. from Japan Science and Technology Agency, Core Researchof Evolutional and Technology.

1. Catron DM, Itano AA, Pape KA, Mueller DL, Jenkins MK (2004) Visualizing the first50 hr of the primary immune response to a soluble antigen. Immunity 21:341–347.

2. Allen CD, Okada T, Cyster JG (2007) Germinal-center organization and cellulardynamics. Immunity 27:190–202.

3. Garside P, et al. (1998) Visualization of specific B and T lymphocyte interactions in thelymph node. Science 281:96–99.

4. Okada T, et al. (2005) Antigen-engaged B cells undergo chemotaxis toward the T zoneand form motile conjugates with helper T cells. PLoS Biol 3:e150.

5. Allen CD, Okada T, Tang HL, Cyster JG (2007) Imaging of germinal center selectionevents during affinity maturation. Science 315:528–531.

6. Hauser AE, et al. (2007) Definition of germinal-center B cell migration in vivo revealspredominant intrazonal circulation patterns. Immunity 26:655–667.

7. Schwickert TA, et al. (2007) In vivo imaging of germinal centres reveals a dynamicopen structure. Nature 446:83–87.

8. Radbruch A, et al. (2006) Competence and competition: The challenge of becominga long-lived plasma cell. Nat Rev Immunol 6:741–750.

9. Tarlinton D (2006) B-cell memory: Are subsets necessary? Nat Rev Immunol 6:785–790.

10. Anderson SM, Tomayko MM, Ahuja A, Haberman AM, Shlomchik MJ (2007) Newmarkers for murine memory B cells that define mutated and unmutated subsets. J ExpMed 204:2103–2114.

11. Pape KA, et al. (2003) Visualization of the genesis and fate of isotype-switched B cellsduring a primary immune response. J Exp Med 197:1677–1687.

12. Cumano A, Rajewsky K (1985) Structure of primary anti-(4-hydroxy-3-nitrophenyl)acetyl (NP) antibodies in normal and idiotypically suppressed C57BL/6 mice. EurJ Immunol 15:512–520.

13. Bothwell AL, et al. (1981) Heavy chain variable region contribution to the NPb familyof antibodies: Somatic mutation evident in a γ 2a variable region. Cell 24:625–637.

14. Takahashi Y, Ohta H, Takemori T (2001) Fas is required for clonal selection in germinalcenters and the subsequent establishment of the memory B cell repertoire. Immunity14:181–192.

15. Ridderstad A, Tarlinton DM (1998) Kinetics of establishing the memory B cellpopulation as revealed by CD38 expression. J Immunol 160:4688–4695.

16. Liu YJ, Oldfield S, MacLennan IC (1988) Memory B cells in T cell-dependent antibodyresponses colonize the splenic marginal zones. Eur J Immunol 18:355–362.

17. Shih TA, Roederer M, Nussenzweig MC (2002) Role of antigen receptor affinity inT cell-independent antibody responses in vivo. Nat Immunol 3:399–406.

18. Sakaue-Sawano A, et al. (2008) Visualizing spatiotemporal dynamics of multicellularcell-cycle progression. Cell 132:487–498.

19. Ravetch JV, Bolland S (2001) IgG Fc receptors. Annu Rev Immunol 19:275–290.

20. Carroll MC (1998) The role of complement and complement receptors in inductionand regulation of immunity. Annu Rev Immunol 16:545–568.

21. Luche H, Weber O, Nageswara Rao T, Blum C, Fehling HJ (2007) Faithful activation ofan extra-bright red fluorescent protein in “knock-in” Cre-reporter mice ideally suitedfor lineage tracing studies. Eur J Immunol 37:43–53.

22. Crouch EE, et al. (2007) Regulation of AID expression in the immune response. J ExpMed 204:1145–1156.

23. Kim CH, et al. (2001) Subspecialization of CXCR5+ T cells: B helper activity is focused ina germinal center-localized subset of CXCR5+ T cells. J Exp Med 193:1373–1381.

24. Breitfeld D, et al. (2000) Follicular B helper T cells express CXC chemokine receptor 5,localize to B cell follicles, and support immunoglobulin production. J Exp Med 192:1545–1552.

25. Vinuesa CG, Tangye SG, Moser B, Mackay CR (2005) Follicular B helper T cells inantibody responses and autoimmunity. Nat Rev Immunol 5:853–865.

26. Haynes NM, et al. (2007) Role of CXCR5 and CCR7 in follicular Th cell positioning andappearance of a programmed cell death gene-1high germinal center-associatedsubpopulation. J Immunol 179:5099–5108.

27. Pape KA, Catron DM, Itano AA, Jenkins MK (2007) The humoral immune response isinitiated in lymph nodes by B cells that acquire soluble antigen directly in the follicles.Immunity 26:491–502.

28. Itano AA, et al. (2003) Distinct dendritic cell populations sequentially present antigento CD4 T cells and stimulate different aspects of cell-mediated immunity. Immunity19:47–57.

29. Dogan I, et al. (2009) Multiple layers of B cell memory with different effectorfunctions. Nat Immunol 10:1292–1299.

30. Benson MJ, et al. (2009) Distinction of the memory B cell response to cognate antigenversus bystander inflammatory signals. J Exp Med 206:2013–2025.

31. Baine Y, Thorbecke GJ (1982) Induction and persistence of local B cell memory in mice.J Immunol 128:639–643.

32. Martin SW, Goodnow CC (2002) Burst-enhancing role of the IgG membrane tail asa molecular determinant of memory. Nat Immunol 3:182–188.

33. Kaisho T, Schwenk F, Rajewsky K (1997) The roles of γ 1 heavy chain membraneexpression and cytoplasmic tail in IgG1 responses. Science 276:412–415.

34. Toyama H, et al. (2002) Memory B cells without somatic hypermutation are generatedfrom Bcl6-deficient B cells. Immunity 17:329–339.

35. Weller S, et al. (2001) CD40-CD40L independent Ig gene hypermutation suggestsa second B cell diversification pathway in humans. Proc Natl Acad Sci USA 98:1166–1170.

36. Madsen L, et al. (1999) Mice lacking all conventional MHC class II genes. Proc NatlAcad Sci USA 96:10338–10343.

Aiba et al. PNAS | July 6, 2010 | vol. 107 | no. 27 | 12197

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

May

3, 2

020