LETTERS

Regulation of IgA production by naturally occurringTNF/iNOS-producing dendritic cellsHiroyuki Tezuka1, Yukiko Abe1, Makoto Iwata2, Hajime Takeuchi2, Hiromichi Ishikawa3, Masayuki Matsushita4,Tetsuo Shiohara5, Shizuo Akira6 & Toshiaki Ohteki1

Immunoglobulin-A has an irreplaceable role in the mucosaldefence against infectious microbes1–6. In human and mouse,IgA-producing plasma cells comprise 20% of total plasma cellsof peripheral lymphoid tissues, whereas more than 80% of plasmacells produce IgA in mucosa-associated lymphoid tissues(MALT)1–6. One of the most biologically important and long-standing questions in immunology is why this ‘biased’ IgAsynthesis takes place in the MALT but not other lymphoid organs.Here we show that IgA class-switch recombination (CSR) isimpaired in inducible-nitric-oxide-synthase-deficient (iNOS2/2;gene also called Nos2) mice. iNOS regulates the T-cell-dependentIgA CSR through expression of transforming growth factor-breceptor, and the T-cell-independent IgA CSR through productionof a proliferation-inducing ligand (APRIL, also called Tnfsf13) anda B-cell-activating factor of the tumour necrosis factor (TNF)family (BAFF, also called Tnfsf13b). Notably, iNOS is preferen-tially expressed in MALT dendritic cells in response to the recog-nition of commensal bacteria by toll-like receptor. Furthermore,adoptive transfer of iNOS1 dendritic cells rescues IgA productionin iNOS2/2 mice. Further analysis revealed that the MALT dend-ritic cells are a TNF-a/iNOS-producing dendritic-cell subset, ori-ginally identified in mice infected with Listeria monocytogenes7,8.The presence of a naturally occurring TNF-a/iNOS-producingdendritic-cell subset may explain the predominance of IgA pro-duction in the MALT, critical for gut homeostasis.

IgA is the most abundant immunoglobulin in the body and has acritical role in mucosal immunity1,2. Secretory IgA is constitutivelyreleased into the lumen via the transport system of the intestinalepithelia3; it prevents invading pathogens from binding to mucosalepithelial cells and neutralizes their toxins4–6. Natural secretory IgA isinduced by constant antigenic stimulation by the intestinal com-mensal bacteria, and dendritic cells have an important function inthis process9,10. IgA-specific CSR occurs in the constant region of theimmunoglobulin heavy chain (CH), from Cm to Ca, in naive B cells,and is accomplished by T-cell-dependent and T-cell-independentCSR5. T-cell-dependent CSR is induced by TGF-b1 and CD40Lexpressed on activated T cells. On the other hand, two dendritic-cell-derived cytokines, APRIL and BAFF, directly mediate T-cell-independent CSR11,12. Notably, the IgA CSR mainly takes place inthe MALT, particularly in Peyer’s patches and lamina propria, thenIgA1 B cells re-circulate through other lymphoid organs and thebloodstream5,13,14. In this context, one of the more enigmatic ques-tions is why IgA CSR and IgA secretion take place largely in theMALT but not other lymphoid organs.

iNOS is upregulated by bacterial components and modulatesimmune responses15. However, little is known about the role of

iNOS in immunoglobulin synthesis. Under steady-state conditions,the serum IgA and IgG2b levels in iNOS2/2 mice were significantlylower compared with wild-type mice, whereas the serum levels ofother immunoglobulin subclasses were unaffected in iNOS2/2 mice(Fig. 1a). iNOS2/2 mice also showed impaired secretory IgA produc-tion in both intestinal contents and faecal pellets (Fig. 1a). Notably,the impaired IgA production strictly correlates with the lack of theiNOS gene even in F2 ((wild type 3 iNOS2/2) F1 3 iNOS2/2) off-spring (Supplementary Fig. 1a), and the level of IgA in the sera andintestinal contents of wild-type mice treated with selective iNOSinhibitors (aminoguanidine (AMG) and L-N6-(1-iminoethyl)-lysine(L-NIL)) and a NO scavenger (carboxy-PTIO (C-PTIO)) was signifi-cantly lower than in untreated wild-type mice (Fig. 1b). In contrast,the number of intestinal T-cell-receptor (TCR)-cd1 cells and theexpression level of polymeric immunoglobulin receptor (pIgR) mes-senger RNA in intestinal epithelial cells—critical regulators for thetransportation of secretory IgA into the lumen16,17—were unaffectedin iNOS2/2 mice (Supplementary Fig. 1b, c). As expected from theseobservations, the number of post-CSR IgA1 B cells (IgA1B2201) andIgA1 plasma cells (IgA1B2202), and the in vitro production of IgA ofthese cells in the absence or presence of interleukin (IL)-5, weresignificantly reduced in iNOS2/2 mice (Fig. 1c, f and Supplemen-tary Fig. 1d) and wild-type mice treated with AMG, L-NIL andC-PTIO (Fig. 1d, e). This effect was probably not due to the impairedtrafficking and survival of IgA1 B cells (Supplementary Fig. 1e), andthe expression level of IL-5 receptor and IL-5-dependent prolifera-tion were unaffected in B cells of iNOS2/2 mice (Supplementary Fig.1f). Notably, impaired ovalbumin (OVA)-specific IgA response wasalso apparent in iNOS2/2 mice (Fig. 1g). Thus, iNOS-derived NO isrequired for IgA synthesis in vivo.

IgA CSR is mediated by both T-cell-dependent and T-cell-independent pathways5. To test whether the T-cell-dependent IgACSR is impaired in iNOS2/2 B cells, naive IgD1 B cells of wild-typeand iNOS2/2 mice were stimulated with TGF-b1 and anti-CD40monoclonal antibody, an inducer for IgA CSR5. Additional stimu-lation, such as anti-CD40 plus interferon-c and anti-CD40 plus IL-4,was also included to induce IgG2c and IgG1/IgE CSR, respectively.Not only immunoglobulin production but also the expression ofactivation-induced cytidine deaminase (AID, an essential enzymefor CSR18) and a-germinal transcript (a-GT, a prerequisite for Cmto Ca CSR) were impaired in iNOS2/2 B cells under IgA-CSR-inducing conditions (Fig. 2a–c), indicating that the IgA CSR isimpaired in iNOS2/2 B cells. The TGF-b receptor, an essential com-ponent of T-cell-dependent IgA CSR, is composed of TGF-bRI andTGF-bRII19,20. Of note, the expression of TGF-bRII, but not TGF-bRI, in naive B cells prepared from the indicated organs of iNOS2/2

1Department of Immunology, Akita University Graduate School of Medicine, Akita 010-8543, Japan. 2Faculty of Pharmaceutical Sciences at Kagawa Campus, Tokushima BunriUniversity, Kagawa 769-2193, Japan. 3Department of Microbiology and Immunology, Keio University School of Medicine, Tokyo 160-8582, Japan. 4Mitsubishi Kagaku Institute of LifeSciences, Tokyo 194-8511, Japan. 5Department of Dermatology, Kyorin University School of Medicine, Tokyo 181-8611, Japan. 6Department of Host Defense, Research Institute forMicrobial Diseases, Osaka University, Osaka 565-0871, Japan.

Vol 448 | 23 August 2007 | doi:10.1038/nature06033

929Nature ©2007 Publishing Group

mice was much lower compared with that of wild-type mice (Fig.2d, e). Thus, the TGF-b1- and anti-CD40-mediated induction ofdownstream molecules of the TGF-b receptor, such as Smad3(ref. 21) and Runx3 (ref. 22), was decreased in iNOS2/2 B cells(Fig. 2f). In contrast, neither CD40 expression nor CD40-dependentproliferation was affected in these cells (Supplementary Fig. 2a).Notably, the B-cell expression of TGF-bRII was restored whennaive iNOS2/2 B cells were cultured with NONOate, a NO donor,in a phosphatidylinositol-3 kinase/AKT-dependent manner in vitro(Supplementary Fig. 2b) and transferred into mMt (known as Igh-6)mice (Fig. 2g). The latter was coincident with the restoration ofserum IgA level (Fig. 2h), indicating that iNOS2/2 B cells retainthe ability to secrete IgA in iNOS1/1 host. These results demonstratethat the critical role for iNOS in the T-cell-dependent IgA CSR is toregulate the B-cell expression of TGF-bRII.

To examine further whether iNOS2/2 T cells are functional inT-cell-dependent IgA production in vivo, T cells isolated fromwild-type and iNOS2/2 mice were transferred into T-cell-deficientTcrb2/2Tcrd2/2 mice, in which systemic and secretory IgA produc-tion was impaired owing to the lack of T-cell-dependent CSR(Supplementary Fig. 3a)10. With time, the serum and intestinal con-tent IgA production was restored in Tcrb2/2Tcrd2/2 mice thatreceived either wild-type or iNOS2/2 T cells, at a comparablelevel (Supplementary Fig. 3b). Furthermore, in the presence ofTGF-b1, naive wild-type B cells co-cultured with activatediNOS2/2 T cells produced IgA at a similar level to those co-culturedwith activated wild-type T cells (Supplementary Fig. 3c). These

results indicate that iNOS2/2 T cells are functional in T-cell-dependent IgA synthesis.

We next examined T-cell-independent IgA production, which ismediated by dendritic-cell-derived cytokines APRIL and BAFF, aswell as B-1 cells10–12. Of note, iNOS was co-expressed with APRILand BAFF in Peyer’s patches and lamina propria dendritic cells ofwild-type mice, whereas the expression level of these cytokines wasdecreased in the equivalent dendritic cells of iNOS2/2 mice (Fig. 3a, band Supplementary Fig. 4a). In this context, on CpG stimulation,expression of APRIL and BAFF was elevated in wild-type dendriticcells but not in iNOS2/2 dendritic cells nor wild-type dendritic cellstreated with L-NIL (Supplementary Fig. 4b). In contrast, expressionof transmembrane activator and CAML interactor (TACI), a receptorfor APRIL and BAFF, was unaffected in iNOS2/2 B cells (Supplemen-tary Fig. 4c). As expected from these findings, IgA production fromiNOS2/2 B cells was impaired when co-cultured with Peyer’s patchesand lamina propria dendritic cells of iNOS2/2 mice, whereas it wassignificantly restored when co-cultured with the equivalent dendriticcells of wild-type mice or those of iNOS2/2 mice in the presence ofAPRIL (Fig. 3c, d). The T-cell-independent IgA production wasefficiently blocked in the presence of BCMA- and TACI-Ig(Supplementary Fig. 4d). In this context, iNOS2/2 dendritic-cellexpression of APRIL was substantially restored when cultured withNONOate in a NFAT/NF-kB-dependent manner, whereas express-ion of TGF-bRII by iNOS2/2 B cells remained impaired even afterAPRIL stimulation (Supplementary Fig. 4e, f). In addition, the levelof APRIL-induced IgA synthesis was unaffected in the presence of

0

800

1,600 ControlAMG

C-PTIO

L-NIL

ControlAMG

C-PTIO

IgALP

** *

b

d e

IgMIgA

SerumICSerum Serum

IgG

80

40

120

0

80

40

120* * * * * *

0

200

100

0

800

400

0

c

iNOS–/–

0.8 1.4

WT

Spl. LPPPMLN

IgA

1.3

6.7

1.4

0.7

23.9

0.8

0.2

5.7

0.3 0.1B22

0

g

a

0

200

400

600

IgG1Serum

WTiNOS–/–

IgG3Serum

0

200

400

600

800

0

10

20

30

IgG2cSerum

0

200

400

IgG2bSerum

*

IgMSerumSerum

*

200

100

0

200

100

0

200

100

0

200

100

0

IC

*

FP

*

IgA

Imm

unog

lob

ulin

(µg

ml–1

)

Imm

unog

lob

ulin

(ng

ml–1

)

Imm

unog

lob

in (µ

g m

l–1)

0

2

4

6

8

10

IgG

*

Spl. IgG2cIgG2bIgG1IgMIgA IgG30.2

0

0.1

Ant

i-O

VA

Ig (a

ntib

odie

s)

0.4

0.6

0

0.2

0.8

0

0.8

0.4

0

0.12

0.8

0.4

0

0.08

0.04

0

0.4

WTiNOS–/–

WTiNOS–/–

–0

200

IgM

Spl.

–

f

Spl

*0

800

1,600

0

Imm

unog

lob

in (n

g m

l–1)

800

1,600

*

MLN

**

PP

0

800

1,600

**0

800

1,600

*

LP

IgA

– –––+ ++IL-5

LP

PP

IgA

B22

0

Control C-PTIOAMG

26.6 15.0 16.5

2.93.16.2

L-NIL

L-NIL

2.2

14.0

Figure 1 | IgA reduction in iNOS2/2 mice. a, b, Samples from the sera,intestinal content (IC) and faecal pellets (FP) of control wild-type andiNOS2/2 mice (a), and wild-type mice given 2.5% (w/v) AMG for 7 days,4.5 mM L-NIL for 14 days, and 3 mg day21 C-PTIO for 14 days (b), weretested for immunoglobulin levels by ELISA. Each group includes three tofour mice and data are representative of three experiments. c, d, Mononuclearcells from the spleen (Spl.), mesenteric lymph node (MLN), Peyer’s patches(PP) and lamina propria (LP) cells of indicated mice were stained for IgA andB220. Numbers in each rectangle indicate the proportion of the cells. Dataare representative of three experiments. e, Lamina propria cells of AMG-,L-NIL- and C-PTIO-treated wild-type mice were cultured for 3 days. f, B cells

from the spleen, mesenteric lymph node and Peyer’s patches, and total LPcells, of indicated mice were cultured in the absence or presence of100 ng ml21 IL-5 for 7 days. Immunoglobulin level in the culture medium isshown in e and f. Data in e, f, show mean 6 s.d. of triplicate cultures and arerepresentative of five experiments. g, Mice were immunized with 100mgOVA, boosted with 50 mg OVA 21 days after immunization, and OVA-specific immunoglobulin levels were determined 10 days later by ELISA.Each group includes three mice and data are representative of twoexperiments. Asterisks in relevant panels indicate unpaired Student’s t-test;P value is ,0.05. Error bars represent s.d.

LETTERS NATURE | Vol 448 | 23 August 2007

930Nature ©2007 Publishing Group

anti-TGF-b neutralizing antibody or TGF-b1 (Supplementary Fig.4g), showing that APRIL regulates IgA synthesis independently ofTGF-b/TGF-bR in mice. Furthermore, the number and in vitro IgAproduction of B-1 cells were reduced in iNOS2/2 mice (Supple-mentary Fig. 4h, i).

These results indicate that iNOS might be expressed in the dend-ritic cells residing in the sites where IgA CSR predominates in vivo.Notably, iNOS was preferentially expressed in CD11clo dendritic cellsin the MALT. The number of iNOS1 dendritic cells was significantlyhigher in the lamina propria than in the Peyer’s patches and mes-enteric lymph node, and negligible in the spleen, of wild-type mice(Fig. 4a), which correlates well with the IgA production in each site(Fig. 1f). To identify the induction mechanism of iNOS1 dendriticcells in the MALT under steady-state conditions, we further analysedMyd882/2, germ-free and Tlr22/2Tlr42/2Tlr92/2 mice. The num-ber of iNOS1 dendritic cells was much reduced in the MALT ofthese mice (Fig. 4a and Supplementary Fig. 5a), suggesting that thisdendritic cell subset is induced by TLR-dependent recognition ofcommensal bacteria. The lamina propria dendritic cells alsoexpressed gaseous NO, suggesting that NO is produced by iNOS1

dendritic cells (Fig. 4b). The iNOS1 dendritic cells were TNF-a1

CD11b1MHC class II1CD801CD86loLy6C1Ly6G2Gr-11Mac31,revealing that they are presumably a naturally occurring TNF-a/iNOS-producing dendritic cell subset (Fig. 4c and SupplementaryFig. 5b)7,8. The absence of iNOS1 dendritic cells in iNOS2/2,Myd882/2, germ-free and Tlr22/2Tlr42/2Tlr92/2 mice correlateswell with the impaired IgA production in the sera and intestinalcontents of these mice (Fig. 4d and Supplementary Fig. 5c).Furthermore, IgA production, B-cell expression of TGF-bRII, andthe numbers of IgA1 B cells and IgA-producing plasma cells were

a

c

b fe

d

g h

TGF-βIFN-γ

IL-4 +––+–––+––+–––+––+++++++

Aid

WT

Gadph

Anti-CD40

0

100

200

300

*

+–TGF-β1+ anti-CD40

IgA

(ng

ml–1

)

WTiNOS–/–

Gapdh

WTSpl. B cells

Tgfbr1

Tgfbr2

WTMLN B cells

iNOS–/–iNOS–/–iNOS–/–

iNOS–/–

iNOS–/–

iNOS–/–

WTPP B cells

0

0.5

1.0

Spl. B PP BMLN BWT

Rel

ativ

e Tg

fbr2

m

RN

A le

vel

** *0

0.5

1.0

0

0.5

1.0

0

0.5

1.0WT

B cells stimulated with TGF-β1 + anti-CD40

Rel

ativ

e m

RN

A le

vel

*

Smad3

0

0.5

1

*

Runx3

µMtWT B µMt

WT

iNOS–/– B µMt

*

Ser

um Ig

A (µ

g m

l–1)

0

50

100

150* *

Gadph

B cells stimulated withTFG-β1 + anti-CD40

WTα-GT

iNOS–/–

Rel

ativ

e Tg

fbr2

m

RN

A le

vel

0

0.25

0.5

0.75

1.0

1.25 * ** iNos–/–

WT B µMt

WT

iNOS–/– B µMtNS

NS

Figure 2 | Impaired T-cell-dependent IgA CSR in iNOS2/2 mice. Naive IgD1

B cells isolated from the spleen of wild-type and iNOS2/2 mice were culturedwith the indicated stimuli for 7 days (a) or 3 days (b, c, f). The cytokines andmonoclonal antibody used in this experiment were: 1 ng ml21 TGF-b1,100 ng ml21 interferon-c, 100 ng ml21 IL-4, 2mg ml21 anti-CD40 monoclonalantibody. a, IgA production by stimulated B cells was determined by ELISA.Data show mean 6 s.d. of triplicate cultures. b, c, Aid (also called Aicda) anda-GT (an intermediate germinal transcript for Cm to Ca CSR) mRNAexpression in stimulated B cells was examined by RT–PCR. d, e, TGF-bRI and

TGF-bRII expression levels (Tgfbr1 and Tgfbr2, respectively) in naive B cells ofthe indicated tissues were evaluated by RT–PCR (d) and real-time PCR(e). f, Expression of Smad3 and Runx3 mRNA in stimulated B cells wasassessed by real-time PCR. g, h, Wild-type and iNOS2/2 naive IgD1 B cellswere transferred into mMt2/2 mice. Three weeks after the transfer, naive B cellswere tested for TGF-bRII expression by real-time PCR (g) and serum IgA levelwas determined by ELISA (h). NS, not significant. Data are representative offive experiments (b–d) or show mean 6 s.d. of three samples (e–h). Asterisksin relevant panels indicate unpaired Student’s t-test; P value is ,0.05.

a b

c d

1.0

Rel

ativ

e m

RN

A le

vel

Rel

ativ

e m

RN

A le

vel

BaffApril

April April

April Baff

0.5

0

1.0

0.5

0

1.0

0.5

0

1.0

0.5

0

WT

500

0

1,000

IgA

(ng

ml–1

)

IgA

(ng

ml–1

)

1,500

500

0

1,000

1,500

2,000

WTKOKO

KOWT

KOKO

*

* **

* **

**

–

B cellsDCs

– – +

WTWT

KOKO

KOWT

KOKO

–

B cellsDCs

– – +

WTiNOS–/–

Figure 3 | Impaired T-cell-independent CSR in iNOS2/2 mice. a, b, Peyer’spatches (a) and lamina propria (b) dendritic cells of wild-type and iNOS2/2

mice were examined for April and Baff expression by real-time PCR. Datashow mean 6 s.d. of three samples. c, d, Splenic IgD1 B cells of wild-type andiNOS2/2 (KO) mice were cultured with Peyer’s patches (c) and laminapropria (d) dendritic cells (DCs) of wild-type and iNOS2/2 (KO) mice in thepresence or absence of 100 ng ml21 APRIL for 7 days, and the IgA level wasevaluated by ELISA. Data show mean 6 s.d. of triplicate cultures and arerepresentative of five experiments. Asterisks in relevant panels indicateunpaired Student’s t-test; P value is ,0.05.

NATURE | Vol 448 | 23 August 2007 LETTERS

931Nature ©2007 Publishing Group

restored in iNOS2/2 mice when transferred with a non-T/B-cellfraction of iNOS1/1 lamina propria cells and iNOS1/1 lamina pro-pria dendritic cells (Fig. 4e, f and Supplementary Fig. 5d). In contrast,these factors (IgA production, B-cell expression of TGF-bRII, and thenumbers of IgA1 B cells and IgA-producing plasma cells) remainedlow in iNOS2/2 mice when transferred with a dendritic-cell-depletednon-T/B-cell fraction of iNOS1/1 lamina propria cells, a non-T/B-cell fraction of iNOS2/2 lamina propria cells, iNOS2/2 lamina pro-pria dendritic cells and iNOS1/1 spleen dendritic cells (Fig. 4e).Collectively, these results demonstrate that TNF-a/iNOS-producingdendritic cells are essential for IgA production in vivo.

Our results show that IgA production is selectively reduced iniNOS2/2 mice. iNOS controls B-cell expression of TGF-bRII inT-cell-dependent IgA CSR, as well as dendritic-cell-derived APRILand BAFF, and B-1-cell-derived IgA production in T-cell-independentIgA CSR. Consistent with these observations, mice lacking TGF-bRII in B cells show a selective reduction of IgA in the sera andexternal secretions19,20, and the serum IgA level and IgA responseto mucosal antigen exposure are reduced in April2/2 mice23 andTaci2/2 mice24. With respect to the molecular mechanism, NO reg-ulates expression of TGF-bRII on B cells in a phosphatidylinositol-3kinase/AKT-dependent manner, whereas it influences APRIL secre-tion by dendritic cells through NFAT- and NF-kB-dependent

pathways. Supporting our results, NF-kB is critical for BAFF induc-tion25. In addition, we also found that the naturally occurring TNF-a/iNOS-producing dendritic cells comprised a portion of retinaldehydrogenase (RALDH1) dendritic cells in the lamina propria(Supplementary Fig. 6a–c), which imprint gut-homing specificityon T and B cells by inducing the integrin a4b7 (refs 26, 27); thenumber of a4b71 T and B cells was unaffected in TNF-a/iNOS-producing dendritic-cell-deficient iNOS2/2 mice (SupplementaryFig. 6d, e). Accordingly, we propose that TNF-a/iNOS-producingdendritic cells directly govern IgA class-switching of naive B cells,whereas RALDH1 dendritic cells imprint gut-homing specificity ofthe post-CSR IgA1 B cells.

Impaired IgA production and intestinal epithelial cell homeostasiswere observed in germ-free mice under steady-state conditions10,28,and dendritic cells that reside between the intestinal epithelial cellssample commensal bacteria at the epithelial surface and participate inIgA induction in the MALT9,29,30. Notably, we observed that iNOS isexpressed in a subset of MALT dendritic cells, this expression isaffected by toll-like-receptor-dependent recognition of commensalbacteria, and it has a critical role in IgA production. Our findingsshed light on the important question of why IgA production, which iscritical for the maintenance of gut homeostasis, takes place pre-dominantly in the MALT.

a

e

b

d

f

iNOS+CD11clo LP cells

Cel

l num

ber

TNF-α CD11b

Mac3Ly6C

B6.iNOS–/–

C57BL/6

MLN LPPP

iNO

S

Non-T, B-cell fraction

6.5

< 0.1

< 0.1 0.2 0.2 0.3 B6.Myd88–/–

0.10.1

0.2

< 0.1

1.5

< 0.1 < 0.1

BALB/c

BALB/c-GF< 0.1

CD11c

< 0.1

0.1 1.8 5.61.4

2.2 60

40

0

100

Serum IC

MLN B cells

Rea

ltive

Tgf

br2

m

RN

A le

vel

PP B cells

iNOS–/–

WT LP DCsiNOS–/– LP DCs

80604020

0

120

1.00

0.50

0

1.00*

* **

**

WT

0.50

0

100806040200

120

120Serum

IgA

(µg

ml–1

)Ig

A (µ

g m

l–1)

IC

iNOS

–/–

iNOS–/–

Myd

88–/

–

BALB/c

C57BL/

6

BALB/c

-GF

iNOS

–/–

Myd

88–/

–

BALB/c

C57BL/

6

BALB/c

-GF

60

40

0

120

DAF/CD11c c

WT

*

**

* * * **

* * *

* ** *

*

iNOS–/–

iNOS–/–

Spl.

LP

DC– L

P

LP D

Cs

iNOS

–/– L

P DCs

iNOS

–/– L

P

Spl. DCs

None

iNOS

–/– S

pl. DCs

LP

DC– L

P

LP D

Cs

iNOS

–/– L

P DCs

iNOS

–/– L

P

Spl. DCs

None

iNOS

–/– S

pl. DCs

Figure 4 | Identification of iNOS1 dendritic cells in the MALT.a, CD3e2B2202 cells from the indicated tissues of C57BL/6, B6.iNOS2/2,B6.Myd882/2, BALB/c and BALB/c germ-free mice were stained for CD11cand intracellular iNOS. b, NO detection in lamina propria dendritic cells bydiaminofluorescein (DAF) staining. Original magnification, 3400.c, Characterization of lamina propria iNOS1 dendritic cells. Lamina propriaiNOS1 dendritic cells were stained for the indicated surface markers. Solidline, each marker staining; dashed line, isotype control staining. d, IgA levelin the sera and intestinal content of C57BL/6 (n 5 5), B6.iNOS2/2 (n 5 5),B6.Myd882/2 (n 5 8), BALB/c (n 5 5) and BALB/c germ-free (n 5 12) micewas measured by ELISA. e, IgA level in the sera and intestinal content ofiNOS2/2 mice transferred with a non-T/B-cell fraction of iNOS1/1 laminapropria cells (LP), a dendritic-cell-depleted non-T/B-cell fraction of

iNOS1/1 lamina propria cells (DC2 LP), a non-T/B-cell fraction of iNOS2/2

lamina propria cells (iNOS2/2 LP), iNOS1/1 lamina propria dendritic cells(LP DCs), iNOS2/2 lamina propria dendritic cells (iNOS2/2 LP DCs),iNOS1/1 spleen dendritic cells (Spl. DCs), and iNOS2/2 spleen dendriticcells (iNOS2/2 Spl. DCs). Each group includes three mice. Two weeks afterthe transfer, the IgA level was determined by ELISA. f, TGF-bRII expressionin naive B cells of C57BL/6, iNOS2/2, iNOS2/2 mice transferred with wild-type lamina propria dendritic cells (WT LP DCs) and iNOS2/2 laminapropria dendritic cells (iNOS2/2 LP DCs). Two weeks after the transfer, theTGF-bRII expression was evaluated by real-time PCR. Data arerepresentative of three experiments. Asterisks in relevant panels indicateunpaired Student’s t-test; P value is ,0.05. Error bars represent s.d.

LETTERS NATURE | Vol 448 | 23 August 2007

932Nature ©2007 Publishing Group

METHODS SUMMARYMice. C57BL/6 mice (B6 mice) and BALB/c mice were obtained from Clea.

B6.iNOS2/2 mice were obtained from Taconic Farms, Inc. B6.Rag22/2 mice

were provided by Central Laboratories for Experimental Animals. F2 ((C57BL/

6 3 iNOS2/2) F1 3 iNOS2/2) offspring were made by crossing C57BL/6 mice

and iNOS2/2 mice. B6.iNOS2/2Rag22/2 mice were generated by crossing

B6.iNOS2/2 mice and B6.Rag22/2 mice. BALB/c germ-free mice were

obtained from Yakult Central Institute for Microbiological Research.

B6.Tlr22/2Tlr42/2Tlr92/2 mice were made by crossing each Tlr2/2 mouse.

All mice except for the germ-free mice were maintained in our SPF animal facility

using the Micro-VENT system (Allentown Caging Equipment Company), and

all animal experiments were done with the approval of the Institutional Animal

Care Committee of the Akita University.

Reagents and in vivo treatment. To examine the requirement of iNOS and NO

for IgA production, wild-type mice were given 2.5% (w/v) aminoguanidine

(Calbiochem), 4.5 mM L-N6-(1-iminoethyl)-lysine (Wako), and 3 mg day21 car-

boxy-PTIO (2-(4-carboxyphenyl)-4,4,5,5-tetramethylimidazoline-1-oxyl-3-

oxide) (Dojindo) intraperitoneally and in drinking water for 7–14 days.

Immunization. Mice were immunized intraperitoneally with 100mg alum-pre-

cipitated ovalbumin (Sigma) (OVA/alum), boosted intraperitoneally with 50mg

OVA/alum 21 days later, and samples from the sera were collected 10 days after

booster for estimation of anti-OVA antibodies.

Measurement of immunoglobulin levels. Samples were collected from sera,

intestinal content, faecal pellets and culture supernatants. The levels of IgM,

IgG, IgG1, IgG2b, IgG2c and IgG3 were determined using the mouse ELISA

quantification kit (Bethyl).

PCR. For semi-quantitative RT–PCR analysis, serial dilutions of the cDNA

templates were amplified using specific primers on a DNA thermal cycler (MJ

Research). Primers for pIgR, AID, a-GT, TGF-bRI, TGF-bRII, iNOS and

GAPDH were obtained from Hokkaido System Science Co. Real-time PCR

analysis was performed using LightCycler (Roche) to measure SYBR green

(Roche) incorporation. Primers for TGF-bRII, Smad3, Runx3, APRIL, BAFF

and b-actin were obtained from Nihon gene research Laboratories Inc.

Relative expression of mRNA was normalized to b-actin mRNA levels within

each sample.

Statistical analysis. The statistical significance of the obtained values was eval-

uated by Student’s t-test. A P value ,0.05 was considered significant.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 22 May; accepted 19 June 2007.

1. Brandtzaeg, P. et al. Regional specialization in the mucosal immune system: whathappens in the microcompartments? Immunol. Today 20, 141–151 (1999).

2. van Egmond, M. et al. IgA and the IgA Fc receptor. Trends Immunol. 22, 205–211(2001).

3. Mostov, K. E. Transepithelial transport of immunoglobulins. Annu. Rev. Immunol.12, 63–84 (1994).

4. Matsunaga, T. & Rahman, A. What brought the adaptive immune system tovertebrates?—The jaw hypothesis and the seahorse. Immunol. Rev. 166, 177–186(1998).

5. Fagarasan, S. & Honjo, T. Intestinal IgA synthesis: regulation of front-line bodydefences. Nature Rev. Immunol. 3, 63–72 (2003).

6. Suzuki, K. et al. Aberrant expansion of segmented filamentous bacteria in IgA-deficient gut. Proc. Natl Acad. Sci. USA 101, 1981–1986 (2004).

7. Serbina, N. V., Salazar-Mather, T. P., Biron, C. A., Kuziel, W. A. & Pamer, E. G. TNF/iNOS-producing dendritic cells mediate innate immune defense against bacterialinfection. Immunity 19, 59–70 (2003).

8. Serbina, N. V. & Pamer, E. G. Monocyte emigration from bone marrow duringbacterial infection requires signals mediated by chemokine receptor CCR2. NatureImmunol. 7, 311–317 (2006).

9. Macpherson, A. J. & Uhr, T. Induction of protective IgA by intestinal dendritic cellscarrying commensal bacteria. Science 303, 1662–1665 (2004).

10. Macpherson, A. J. et al. A primitive T cell-independent mechanism of intestinalmucosal IgA responses to commensal bacteria. Science 288, 2222–2226 (2000).

11. Litinskiy, M. B. et al. DCs induce CD40-independent immunoglobulin classswitching through BLyS and APRIL. Nature Immunol. 3, 822–829 (2002).

12. Castigli, E. et al. TACI and BAFF-R mediate isotype switching in B cells. J. Exp. Med.201, 35–39 (2005).

13. Mestecky, J. & McGhee, J. R. Immunoglobulin A (IgA): molecular and cellularinteractions involved in IgA biosynthesis and immune response. Adv. Immunol.40, 153–245 (1987).

14. Fagarasan, S., Kinoshita, K., Muramatsu, M., Ikuta, K. & Honjo, T. In situ classswitching and differentiation to IgA-producing cells in the gut lamina propria.Nature 413, 639–643 (2001).

15. Bogdan, C. Nitric oxide and the immune response. Nature Immunol. 2, 907–916(2001).

16. Fujihashi, K. et al. c/d T cell-deficient mice have impaired mucosalimmunoglobulin A responses. J. Exp. Med. 183, 1929–1935 (1996).

17. Shimada, S.-I. et al. Generation of polymeric immunoglobulin receptor-deficientmouse with marked reduction of secretory IgA. J. Immunol. 163, 5367–5373(1999).

18. Muramatsu, M. et al. Class switch recombination and hypermutation requireactivation-induced cytidine deaminase (AID), a potential RNA editing enzyme.Cell 102, 553–563 (2000).

19. Cazac, B. B. & Roes, J. TGF-b receptor controls B cell responsiveness andinduction of IgA in vivo. Immunity 13, 443–451 (2000).

20. Borsutzky, S., Cazac, B. B., Roes, J. & Guzman, C. A. TGF-b receptor signaling iscritical for mucosal IgA responses. J. Immunol. 173, 3305–3309 (2004).

21. Park, S. R., Lee, J. H. & Kim, P. H. Smad3 and Smad4 mediate transforming growthfactor-b1-induced IgA expression in murine B lymphocytes. Eur. J. Immunol. 31,1706–1715 (2001).

22. Fainaru, O. et al. Runx3 regulates mouse TGF-b-mediated dendritic cell functionand its absence results in airway inflammation. EMBO J. 23, 969–979 (2004).

23. Castigli, E. et al. Impaired IgA class switching in APRIL-deficient mice. Proc. NatlAcad. Sci. USA 101, 3903–3908 (2004).

24. von Bulow, G.-U., van Deursen, J. M. & Bram, R. J. Regulation of the T-independenthumoral response by TACI. Immunity 14, 573–582 (2001).

25. He, B., Raab-Traub, N., Casali, P. & Cerutti, A. EBV-encoded latent membraneprotein 1 cooperates with BAFF/BLyS and APRIL to induce T cell-independent Igheavy chain class switching. J. Immunol. 171, 5215–5224 (2003).

26. Iwata, M. et al. Retinoic acid imprints gut-homing specificity on T cells. Immunity21, 527–538 (2004).

27. Mora, J. R. et al. Generation of gut-homing IgA-secreting B cells by intestinaldendritic cells. Science 314, 1157–1160 (2006).

28. Rakoff-Nahoum, S., Paglino, J., Eslami-Varzaneh, F., Edberg, S. & Medzhitov, R.Recognition of commensal microflora by toll-like receptors is required forintestinal homeostasis. Cell 118, 229–241 (2004).

29. Rescigno, M. et al. Dendritic cells express tight junction proteins and penetrate gutepithelial monolayers to sample bacteria. Nature Immunol. 2, 361–367 (2001).

30. Niess, J. H. et al. CX3CR1-mediated dendritic cell access to the intestinal lumenand bacterial clearance. Science 307, 254–258 (2005).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank K. Onodera for animal care; K. Yamashita forexperimental support; K. Honda, M. Muramatsu and M. Miyasaka for discussions;M. Nanno for BALB/c germ-free mice; T. Tsubata for B6.mMt mice; and K. Takatsufor monoclonal antibody H-7. This work was supported by Yakult Bio-ScienceFoundation (to T.O.), a Sasakawa Scientific Research Grant from The JapanScience Society (to H. Tezuka), a Grant-in-Aid for Scientific Research on PriorityAreas from the Ministry of Education, Science, Sports and Culture of Japan (T.O.),and a Grant-in-Aid for Young Scientists (B) (to H. Tezuka.).

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Correspondence and requests for materials should be addressed to T.O.(tohteki@med.akita-u.ac.jp).

NATURE | Vol 448 | 23 August 2007 LETTERS

933Nature ©2007 Publishing Group

METHODSMice. B6.Tcrb2/2Tcrd2/231, B6.Myd882/232 and B6.mMt33 mice were as

described.

Measurement of immunoglobulin levels. The level of IgA was measured by

enzyme-linked immunosorbent assay (ELISA) as described previously31.

Cell preparation. To prepare naive B cells, spleen cells were labelled with PE-

conjugated anti-mouse IgD mAb (eBioscience) and anti-PE MicroBeads

(MiltenyiBiotec). IgD1 B cells were isolated using AutoMACS (MiltenyiBiotec)

and used as naive B cells (.97% IgD1). Likewise, B2201 cells isolated from

spleen, MLN, and PP were used as conventional B cells (.98% B2201). Toprepare B-1 cells, peritoneal cells were labelled with PE-BM8 (anti-F4/80), PE-

B3B4 (anti-CD23), and Thy1.2 (anti-CD90.2) MicroBeads and subsequently with

anti-PE MicroBeads. The negative fraction was used as B-1 cells. To prepare BM

DCs, BM cells were cultured in the presence of 10 ng ml21 recombinant mouse

GM-CSF (R&D Systems) for 6 days by replacing medium every 3 days. After

6 days of culture, BM cells were further stimulated with 20 ng ml21 recombinant

mouse TNF-a for 2 days (R&D Systems). The cells were then labelled with N418-

MicroBeads (anti-mouse CD11c) (MiltenyiBiotec) and isolated using

AutoMACS.

To prepare the intestinal intraepithelial lymphocytes (iIELs), PPs, fatty tissues

and mesentery were removed from the small intestine. The small intestine was

opened longitudinally and cut into 5-mm pieces. The pieces were washed gently

and shaken for 45 min at 37 uC in RPMI 1640 medium containing 5% FCS,

100 U ml21 penicillin G, and 100mg ml21 streptomycin. Whole-intestinal-cell

suspensions, which include iECs and iIELs, were applied to discontinuous

Percoll density gradients (Amersham Bioscience) of 25%, 40% and 75%, and

cells at the interface between 40% and 75% were collected and used as IELs. After

removal of iECs and iIELs, the remaining tissue was further digested with col-lagenase (type IV, Sigma), and cells at the interface between 30% and 60% of

Percoll gradient were collected and used as LP cells. To purify DCs, collagenase

digested spleen, MLN, PP and LP cells were labelled with N418-MicroBeads

(anti-mouse CD11c) (MiltenyiBiotec) and isolated using AutoMACS. These

purified CD11c1 cells (.95% CD11c1 I-A1) were used as BM, spleen, MLN,

PP and LP DCs.

Cell culture. To examine Ig secretion, B cells were cultured with anti-CD40 mAb

(2mg ml21; eBioscience), LPS (2mg ml21, Sigma), TGF-b1 (1 ng ml21, RDI),

mouse IL-5 (100 ng ml21, R&D Systems), mouse IL-15 (100 ng ml21, RDI),

mouse APRIL (100 ng ml21, RDI), and anti-TGFb1 mAb (10mg ml21, R&D

Systems), and the culture supernatants were collected for ELISA. A cell prolif-

eration assay was also performed in some experiments, in which the cells were

pulsed with 1 mCi per well 3H-thymidine (PerkinElmer Life Science) for the

last 16 h of a 72-h culture, and 3H-thymidine incorporation was measured by

a b-counter (Packard). In some experiments, 1 3 106 T cells that had been

activated with anti-CD3e mAb (1 mg ml21) and anti-CD28 mAb (1mg ml21)

(eBioscience), were co-cultured with 5 3 105 splenic naive B cells in the presence

of TGF-b1 for 7 days. Likewise, 5 3 105 B cells were co-cultured with 1.25–2.5 3 105 DCs in the presence or absence of APRIL, TACI-Ig (3.3mg ml21,

R&D Systems) or BCMA-Ig (3.3mg ml21, R&D Systems) for 7 days. Splenic

IgD1 B cells were cultured with LY294002 (PI3 kinase inhibitor, Calbiochem)

or Akt inhibitor (Calbiochem) for 1 h before addition of 100mM spermine

NONOate (Alexis Biochemicals) for 18 h. LP DCs were cultured with 5 mM

BAY11-7082 (NF-kB inhibitor, Calbiochem), 5mM 11R-VIVIT (NFAT inhibitor

peptide, RRRRRRRRRRR-GGG-MAGPVIVITGPHEE, Sigma Genosys)34, 5mM

11R-VEET (control peptide, RRRRRRRRRRR-GGG-MAGPPHIVEETGPHVI,

Sigma Genosys)34, or 500mM L-NIL for 1 h before the addition of 100mM

spermine NONOate or 1 mM CpG-A (59-TCCATGACGTTCCTGATGCT-39,

Hokkaido System Science) for 18 h.

Antibodies and flow-cytometric analysis. Cells were stained with the following

mAbs: fluorescein isothiocyanate (FITC)-N418 (anti-CD11c), phycoerythrin

(PE)-RA3-B2 (anti-B220), PE-53-6.7 (anti-CD8a), PE-B3B4 (anti-CD23), PE-

BM8 (anti-F4/80), biotin-UC7-13D5 (anti-TCRcd), biotin-1C10 (anti-CD40),

biotin-N418, biotin-M1/70 (anti-CD11b), biotin-16-10A1 (anti-CD80), biotin-

GL1 (anti-CD86), biotin-RB6-8C5 (anti-Gr-1), biotin-53-6.7 (anti-CD8a),

biotin-MB19-1 (anti-CD19), biotin-RA3-B2, biotin-53-2.1 (anti-Thy1.2),biotin-2E7 (anti-CD103), biotin-DX5 (anti-DX5), biotin-NLDC-145 (anti-

CD205), and biotin-4B12 (anti-CCR7). These mAbs were purchased from

eBioscience. FITC-C10-3 (anti-IgA), FITC-1A8 (anti-Ly6G), and biotin-AL-21

(anti-Ly6C) were purchased from BD PharMingen. Biotin-H-7 (anti-IL-5Ra)

was provided by T. Kouro and K. Takatsu of Tokyo University. Biotinylated

mAbs were detected with streptavidin-PE-Cy5 (eBioscience). For intracellular

iNOS and Mac-3 staining, CD3e2B2202 cells of the spleen, MLN, PP and LP

were stained with biotin N418 and streptavidin-PE-Cy5, and were fixed and

permeabilized with Cytofix/Cytoperm solution (BD Pharmingen). The cells

were then stained intracellularly with PE-N20 (anti-iNOS, Santa Cruz) and

FITC-M3/84 (anti-Mac-3) (eBioscience). For intracellular TNF-a staining,

CD3e2B2202 cells of the LP were cultured with 50 ng ml21 PMA (Wako Pure

Chemical Industries), 500 ng ml21 ionomycin (Sigma), and GolgiPlug (BD

Pharmingen) for 4 h, after which they were stained with biotin N418 and PE-

streptavidin-Cy5 (eBioscience), and fixed and permeabilized as described above.

The cells were then stained intracellularly with PE-N20 and FITC-MP6-XT22

(anti-TNF-a). FITC-rat IgG1 (eBioscience) was also included as isotype controls

for Mac-3 and TNF-a. Samples were analysed on a FACSCalibur using the

CELLQuest program (Becton Dickinson).

Immunocytochemical analysis. LP DCs and splenic DCs were fixed and per-

meabilized with Cytofix/Cytoperm solution (BD Pharmingen). These cells were

stained with rabbit anti-mouse RALDH1 or RALDH2 (prepared in-house), and

PE-N-20 (anti-iNOS). For RALDH staining, cells were further stained with

FITC-conjugated anti-rabbit IgG (Jackson ImmunoResearch) and AlexaFluor

488-conjugated anti-FITC (Molecular Probes). To detect intracellular NO, LP

DCs were incubated with 10mM diaminofluorescein (DAF)-FM diacetate

(Daiichi Pure Chemicals) and then stained with PE-N418 (anti-CD11c).

Stained cells were mounted on slides and analysed by fluorescence microscopy.

RT–PCR. Specific primer pairs used for RT–PCR were as follows: pIgR sense (59-

TTGTTCACGCTCTTGGTAACTG-39) and pIgR antisense (59-ACAGGCC-

TCGGTTACTGGTACC-39); AID sense (59-ATATGGACAGCCTTCTGATG-

AAGC-39) and AID anti-sense (59-TCAAAATCCCAACATACGAAATGC-39);

aGT sense (59-CTACCATAGGGGAAGATAGCCT-39) and aGT anti-sense (59-

TAATCGTGAATCAGGCAG-39); TGF-bRI sense (59-ATCCATCACTAG-

ATCGCCCT-39) and TGF-bRI anti-sense (59-CGATGGATCAGAAGGTA-

CAAGA-39); TGF-bRII sense (59-CGTGTGGAGGAAGAACAACA-39) and

TGF-bRII anti-sense (59-TCTCAAACTGCTCTGAGGTG-39); iNOS sense (59-

CGTTGGATTTGGAGCAGAAGTG-39) and iNOS anti-sense (59-CATGC-

AAAATCTCTCCACTGCC-39); GAPDH sense (59-ACCACAGTCCATGCC-

ATCAC-39) and GAPDH anti-sense (59-TCCACCACCCTGTTGCTGTA-39).

Specific primer pairs used for real-time PCR analysis were as follows: TGF-

bRII sense (59-CCTACTCTGTCTGTGGATGA-39) and TGF-bRII anti-sense

(59-GCTCGTAATCCTTCACTTCTC-39); Smad3 sense (59-GCACAGCCAC-

CATGAATTAC-39) and Smad3 anti-sense (59-GCACCAACACTGGAGG-

TAGA-39); Runx3 sense (59-ACCACGAGCCACTTCAGCAG-39) and Runx3

anti-sense (59-CGATGGTGTGGCGCTGTA-39); APRIL sense (59-TCACAA-

TGGGTCAGGTGGTATC-39) and APRIL anti-sense (59-TGTAAATGAAA-

GACACCTGCACTGT-39); BAFF sense (59-TGCTATGGGTCATGTCATCCA-

39) and BAFF anti-sense (59-GGCAGTGTTTTGGGCATATTC-39); b-actin

sense (59-GATGACGATATCGCTGCGCTG-39) and b-actin anti-sense (59-

GTACGACCAGAGGCATACAGG-39). Relative expression of mRNA was nor-

malized to b-actin mRNA levels within each sample.

Adoptive transfer. To examine the IgA productivity of iNOS2/2 B cells, naive B

cells (a mixture of 2 3 107 splenic B cells, 1.5 3 106 MLN B cells, and 3 3 106 PP

B cells) of WT and iNOS2/2 mice were intravenously transferred into mMt

hosts. To examine the function of iNOS2/2 T cells, 1 3 107 spleen T cells of

WT and iNOS2/2 mice were transferred into Tcrb2/2Tcrd2/2 mice. To examine

the role of iNOS1/1DCs in IgA production, 2 3 106 of a non-T/B fraction of

iNOS1/1 LP cells, a DC-depleted non-T/B fraction of iNOS1/1 LP cells, a non-

T/B fraction of iNOS2/2 LP cells, and 2 3 106 iNOS1/1 LP DCs, iNOS2/2 LP

DCs, iNOS1/1 SPL DCs were transferred into iNOS2/2 mice. Two to four weeks

after the transfer, the IgA level was determined by ELISA. To examine the traf-

ficking and survival of IgA1 B cells in iNOS2/2 hosts, PP IgA1 B cells (4 3 106

cells) or LP IgA1 B cells (2 3 106 cells) of B6.SJL (CD45.11) mice were intrave-

nously transferred into iNOS1/1Rag22/2 (WT), and iNOS2/2Rag22/2

(iNOS2/2) hosts (CD45.21). Sixteen hours (migration of the PP IgA1 B cells)

and 6 days (survival of the LP IgA1 B cells) after the transfer, LP cells isolated

from hosts were stained with FITC-104 (anti-CD45.2) and PE-A20 (anti-

CD45.1) and analysed on FACSCalibur using the CELLQuest program

(Becton Dickinson).

31. Nishiyama, Y. et al. Homeostatic regulation of intestinal villous epithelia by Blymphocytes. J. Immunol. 168, 2626–2633 (2002).

32. Adachi, O. et al. Targeted disruption of the MyD88 gene results in loss of IL-1- andIL-18-mediated function. Immunity 9, 143–150 (1998).

33. Kitamura, D., Roes, J., Kuhn, R. & Rajewsky, K. A B cell-deficient mouse by targeteddisruption of the membrane exon of the immunoglobulin m chain gene. Nature350, 423–426 (1991).

34. Noguchi, H., Matsushita, M., Okitsu, T. & Matsui, H. A new cell-permeablepeptide allows successful allogeneic islet transplantation in mice. Nature Med. 10,305–309 (2004).

doi:10.1038/nature06033

Nature ©2007 Publishing Group