ARTICLEdoi:10.1038/nature11531

Compensatory dendritic cell developmentmediated by BATF–IRF interactionsRoxane Tussiwand1*, Wan-Ling Lee1*, Theresa L. Murphy1*, Mona Mashayekhi1, Wumesh KC1, Jorn C. Albring1,Ansuman T. Satpathy1, Jeffrey A. Rotondo1, Brian T. Edelson1, Nicole M. Kretzer1, Xiaodi Wu1, Leslie A. Weiss2, Elke Glasmacher3,Peng Li4, Wei Liao4, Michael Behnke2, Samuel S. K. Lam1, Cora T. Aurthur1, Warren J. Leonard4, Harinder Singh3,Christina L. Stallings2, L. David Sibley2, Robert D. Schreiber1 & Kenneth M. Murphy1,5

The AP1 transcription factor Batf3 is required for homeostatic development of CD8a1 classical dendritic cells that primeCD8 T-cell responses against intracellular pathogens. Here we identify an alternative, Batf3-independent pathway inmice for CD8a1 dendritic cell development operating during infection with intracellular pathogens and mediated by thecytokines interleukin (IL)-12 and interferon-c. This alternative pathway results from molecular compensation for Batf3provided by the related AP1 factors Batf, which also functions in Tand B cells, and Batf2 induced by cytokines in responseto infection. Reciprocally, physiological compensation between Batf and Batf3 also occurs in T cells for expression ofIL-10 and CTLA4. Compensation among BATF factors is based on the shared capacity of their leucine zipper domains tointeract with non-AP1 factors such as IRF4 and IRF8 to mediate cooperative gene activation. Conceivably, manipulatingthis alternative pathway of dendritic cell development could be of value in augmenting immune responses to vaccines.

BATF1 and BATF3 are activator protein 1 (AP1)2 transcriptionfactors3,4 with immune-specific functions5–8. Batf is required fordevelopment of T-helper cells producing IL-17 (TH17) and follicularhelper T (TFH) cells5, and class-switch recombination (CSR) in Bcells6,9. Batf3 is required for development of CD8a1 classical dendriticcells (cDCs) and related CD1031 dendritic cells8 that cross-presentantigens to CD8 T cells7 and produce IL-12 in response to pathogens10.

We recently recognized a heterozygous phenotype for Batf3 of 50%fewer CX3CR12CD8a1 cDCs in Batf31/2 mice11, indicating thatlevels of BATF3 are limiting for CD8a1 dendritic cell developmentunder homeostatic conditions. While analysing Toxoplasma gondiiinfection in Batf32/2 mice, we observed evidence indicating Batf3-independent CD8a1 cDC development10 based on apparent primingof pathogen-specific CD8 T cells in Batf32/2 mice treated with IL-12.Here we report a Batf3-independent pathway of CD8a1 cDC develop-ment that functions physiologically during infection by intracellularpathogens and describe its molecular basis, which involves com-pensatory BATF factors.

Pathogens and IL-12 restore CD8a1 cDCs in Batf32/2 miceIL-12 administration to Batf32/2 mice before infection with type IIPrugniaud (Pru) T. gondii reversed their susceptibility by inducinginterferon (IFN)-c production not only from natural killer (NK) cellsbut also from CD8 T cells10, indicating potentially restored cross-priming. To test this idea, we infected Batf32/2 mice with the atte-nuated12 RHDku80Drop5 strain of T. gondii and examined CD8a1

cDCs (Supplementary Fig. 1a). Notably, CD8a1 cDCs reappeared inspleens of Batf32/2 mice by day 10 after infection. Infection byListeria monocytogenes also restored CD8a1 cDC production inBatf32/2 mice (Supplementary Fig. 1b). Aerosol-mediated infectionwith Mycobacterium tuberculosis caused a progressive restorationof CD8a1 cDCs in Batf32/2 mice and expanded CD8a1 cDCs in

wild-type mice (Fig. 1a). Mycobacterium tuberculosis infection ofBatf32/2 mice restored the missing lung-resident CD1031 cDCs(DEC2051CD241CD42SIRP-a2CD11b2)8,13 (Supplementary Fig. 1c).Batf32/2 mice showed no difference in survival to M. tuberculosisinfection compared to wild-type mice (Supplementary Fig. 1d),suggesting that the initial lack of CD8a1 cDCs and peripheralCD1031 cDCs was not sufficient for lethality with M. tuberculosis.Because IL-12 is important in control of M. tuberculosis14, we measuredserum IL-12 levels in M. tuberculosis-infected wild-type and Batf32/2

mice (Fig. 1b). IL-12 levels were reduced in Batf32/2 mice in thefirst 3 weeks, but increased after 6 weeks to approximately 50% ofM. tuberculosis-infected wild-type mice.

We found that IL-12 administration restored CD8a1 cDC productionin all backgrounds of Batf32/2 mice (Fig. 1c and Supplementary Fig. 2a).Restored CD8a1 cDCs were functional for cross-presentation7,15,16

(Fig. 2a and Supplementary Fig. 2b). Purified splenic CD8a1 andCD41 cDCs from wild-type or Batf32/2 mice, treated with vehicle orIL-12, were tested for cross-presentation. As a control, OT-I T cellsproliferated in response to wild-type CD8a1 dendritic cells, but notCD41 dendritic cells, co-cultured with ovalbumin-loaded cells with orwithout IL-12 treatment (Fig. 2a). Importantly, OT-I T cells proliferatedsimilarly in response to IL-12-induced Batf32/2 CD8a1 dendritic cells,but not Batf32/2 CD41 dendritic cells, as to wild-type CD8a1 cDCs.The CD8a1 cDCs restored by IL-12 administration in Batf32/2 miceshowed a gene expression profile that was more similar to wild-typeCD8a1 dendritic cells than to CD41 cDCs (Supplementary Fig. 2c). Atotal of 1,855 genes differed more than fourfold in expression betweenIL-12-induced CD8a1 cDCs and CD41 cDCs within Batf32/2 mice.However, in comparing IL-12-induced CD8a1 cDCs in Batf32/2 micewith the rare CD8a1 cDCs from Batf32/2 C57BL/6 mice, or withCD8a1 cDCs in wild-type mice, there were only 34 and 206 genes,respectively, differing more than fourfold in expression.

*These authors contributed equally to this work.

1Department of Pathology and Immunology, Washington University School of Medicine, 660 South Euclid Avenue, St Louis, Missouri 63110, USA. 2Department of Molecular Microbiology, WashingtonUniversity School of Medicine, St Louis, Missouri 63110, USA. 3Department of Discovery Immunology, Genentech, Inc., 1 DNA Way, South San Francisco, California 94080, USA. 4Laboratory of MolecularImmunology and Immunology Center, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland 20892-1674, USA. 5Howard Hughes Medical Institute, WashingtonUniversity School of Medicine, 660 South Euclid Avenue, St Louis, Missouri 63110, USA.

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2012

The IL-12-induced restoration of CD8a1 cDCs in Batf32/2 micewas dependent on IFN-c in vivo (Supplementary Fig. 2d). Withadministration of control antibody, IL-12 induced a threefold increasein CD8a1 cDCs in wild-type mice and restored CD8a1 cDC produc-tion in Batf32/2 mice. Both effects were blocked by anti-IFN-cantibody. IL-12 induced IFN-c production from NK cells but not Tcells (Supplementary Fig. 2e), and IL-12 treatment was able to restoreCD8a1 cDCs in Rag22/2Batf32/2 mice (Supplementary Fig. 2f),indicating that T cells and B cells are not required for this effect.

Batf32/2 mice fail to reject highly immunogenic H31m1 and D42m1fibrosarcomas7. However, Batf32/2 mice pre-treated with IL-12 eitherfully rejected or showed reduced growth of these fibrosarcomas (Fig. 2band Supplementary Fig. 2g). Rejection was not simply due to NK cellactivation, as IL-12 treatment failed to alter tumour growth in Rag22/2

mice. Notably, IL-12-treated Batf32/2 mice re-established priming ofCD8 T cells capable of infiltrating tumours similar to wild-type mice(Fig. 2c). Mice with the inactivating IRF8 R249C mutation17–19 failedto restore CD8a1 cDCs upon IL-12 administration (SupplementaryFig. 3a), indicating that restoration requires functional IRF8 protein.Furthermore, these mice were highly susceptible to infection by aerosol-administered M. tuberculosis20 (Supplementary Fig. 3b), indicating thatBatf32/2 mice may resist M. tuberculosis infection by restoration ofCD8a1 cDCs. Thus, physiological Batf3-independent development ofCD8a1 cDCs is induced by pathogens, mediated by IL-12 and IFN-c.

Batf, Batf2 and Batf3 cross-compensateWe next investigated whether Batf could replace Batf3 for in vitro cDCdevelopment7,21 (Fig. 3a). CD1031SIRP-a2 cDCs do not develop inFlt3L-treated Batf32/2 bone marrow cultures. Retroviral expression ofBatf3 into Batf32/2 bone marrow progenitors restored CD1031 cDCdevelopment. Notably, Batf expression fully and cell-intrinsicallyrestored CD1031SIRP-a2 cDC development, whereas Fos was

inactive (Supplementary Fig. 3c). CD1031 cDCs restored by Batfand Batf3 expression in vitro were functional, showing features ofmature CD1031 cDCs, including loss of SIRP-a and CD11b expres-sion, upregulation of CD24, and selective production of IL-12 in res-ponse to T. gondii antigen (Supplementary Fig. 3c, d). Reciprocally,Batf3 but not Fos restored cell-intrinsic IL-17a production by Batf2/2

T cells and CSR in Batf2/2 B cells (Supplementary Fig. 3e, f). Thus,Batf and Batf3 can molecularly cross-compensate for several distinctlineage-specific functions, activities not shared by Fos.

We also identified in vivo compensation between Batf and Batf3 indendritic cells. On the 129SvEv and BALB/c backgrounds, Batf32/2

mice completely lack CD8a1 cDCs (Fig. 3b and Supplementary Fig. 2a).In contrast, C57BL/6 Batf32/2 mice retain a 3–7% population of CD8a1

cDCs19,22 in spleen and unexpectedly retain a completely normal popu-lation of CD8a1 cDCs in skin-draining inguinal lymph nodes (ILNs)(Fig. 3b). These CD8a1 cDCs are functional, as C57BL/6 Batf32/2 miceshowed robust priming of CD8 T cells against herpes simplex virus(HSV) infection in the footpad, whereas 129SvEv Batf32/2 mice, lackingCD8a1 cDCs in ILNs, did not (Supplementary Fig. 3g). However,C57BL/6 Batf2/2Batf32/2 mice lack CD8a1 dendritic cells in ILNsand fail to prime T cells against HSV. Thus, retention of functionalCD8a1 cDCs in ILNs of C57BL/6 Batf32/2 mice is Batf-dependent.

Batf and Batf3 also compensate in expression of genes by T cells.IL-4 and IL-10 production was not substantially affected in eitherBatf2/2 or Batf32/2 TH2 cells (Fig. 3c and Supplementary Fig. 3i–k).However, IL-10 is reduced at least eightfold in Batf2/2Batf32/2 TH2

0.78.0 7.3

CD8

Thy1.2

WT

Batf3–/–

PBS IL-12

Tum

ou

r siz

e (

mm

)

Days after tumour inoculation

0

5

10

15

20

0 10 20 30 0 10 20 30 0 10 20 30

Fibrosarcoma H31m1

WT Batf3–/–

PBS IL-12

0 5 10 15 20 250

20

40

60

80

CD4DC WT

CD4DC WT + IL-12

CD8DC WT + PBS

CD8DC WT + IL-12

CD4DC Batf3–/–

CD4DC Batf3–/– + IL-12

CD8DC Batf3–/– + IL-12

50 75 100 125

Ova-loaded splenocytes (×103)

Pro

life

rati

ng

OT

-1 (%

)

a

c

b

0

102

103

104

105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 105

Figure 2 | IL-12-induced CD8a1 cDCs in Batf32/2 mice can cross-presentand mediate tumour rejection. a, From mice in Fig. 1c, dendritic cells werepurified by sorting as CD32DX52MHCII1CD11c1SIRP-a2CD241DEC2051

dendritic cells (CD8DC) and CD32DX52MHCII1CD11c1SIRP-a1CD242DEC2052 dendritic cells (CD4DC) and assayed for cross-presentation7. OT-I proliferation in response to cDCs mixed with the indicatednumber of MHC-class-I-deficient ovalbumin (Ova)-loaded splenocytes isshown. b, Wild-type or Batf32/2 mice treated with vehicle (PBS) or with IL-12were inoculated with 1 3 106 H31m1 fibrosarcomas. Tumour size in individualmice is shown. c, Mice in b were analysed by FACS 11 days after H31m1inoculation for CD8 T-cell infiltration into tumours7.

IL-1

2p

40

01

2

3

4

5

6 WTBatf3–/–

Weeks after infection

(pg

ml–

1) (×

10

2)

b

3 wk

100

101

102

103

104

DE

C2

05

WT

Batf

3–/–

a Ctrl 1 wk

CD8

4.5

0.6

9 wk

4.4

0.8

13

5.7

27

18

17

76

5.2

89

0.1

94

9.4

83

PBS IL-12

DE

C205

CD8

WT

Batf

3–/–

c

100

101

102

103

104

100 101 102 103 104 100 101 102 103 104 100 101 102 103 104 100 101 102 103 104

1 10 2 2 3 3 6 6

100

101

102

103

104

100

101

102

103

104

100 101 102 103 104 100 101 102 103 104

Figure 1 | Intracellular pathogens or IL-12 restore lymphoid CD8a1 cDCsand tissue-resident CD1031 cDCs in Batf32/2 mice. a, Wild type (WT) andBatf32/2 129SvEv mice were uninfected (Ctrl) or infected with M. tuberculosis,and spleens collected and analysed by FACS at the indicated time. Histograms forindicated markers are gated as autofluorescent2MHCIIhighCD11c1 cells.Numbers are per cent of cells in the gate. b, Serum IL-12 level was measured fromindividual mice (a) at the indicated time. c, Wild-type and Batf32/2 129SvEv micewere treated with vehicle (PBS) or IL-12 and analysed by FACS after 3 days as in a.

RESEARCH ARTICLE

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

cells. Also, expression of the inhibitory receptor CTLA4 is partiallyreduced in Batf2/2 TH2 cells but reduced threefold further inBatf2/2Batf32/2 TH2 cells. In contrast, IFN-c production by TH1 cellsis unaffected by loss of Batf, Batf3 or both.

We asked whether IL-12-induced restoration of CD8a1 cDCsin Batf32/2 mice was due to compensation by Batf (SupplementaryFig. 3h). Restoration of splenic CD8a1 cDCs in IL-12-treatedBatf2/2Batf32/2 mice was reduced to 5% from 11% in IL-12-treatedBatf32/2 mice. Whereas Batf seems to be responsible for roughly halfof the IL-12-induced restoration of CD8a1 cDCs in Batf32/2 mice,some residual compensation remained in Batf2/2Batf32/2 mice. Weasked if a third factor might compensate for Batf and Batf3. Batf2 (alsocalled SARI)23 is closely related to Batf and Batf3 and is induced bylipopolysaccharide (LPS) and IFN-c in macrophages and CD1031

dendritic cell populations (Supplementary Fig. 4a–c). We found thatBatf2 was induced by IFN-c in wild-type and Batf32/2 dendritic cellsderived from Flt3L-cultured bone marrow (Supplementary Fig. 4d, e)and in vivo by IL-12 in dendritic cells in Batf32/2 mice (Supplemen-tary Fig. 4f). Induction of Batf2 by IFN-c in cDCs made it a potentialcandidate to mediate IFN-c-dependent compensation for Batf3.

We generated Batf22/2 mice by targeting exons 1 and 2, eliminat-ing BATF2 protein expression (Supplementary Fig. 5). Batf22/2 mice

had normal development of NK, T and B cells, pDCs, neutrophils,resting cDCs, and peritoneal, liver and lung macrophages (Sup-plementary Fig. 6a–c). However, Batf22/2 mice displayed signifi-cantly decreased survival after infection by T. gondii (Pru) (Fig. 4a),although parasite burden and serum cytokine levels were similar towild-type mice (Supplementary Fig. 7a, b). Notably, Batf22/2 miceshowed significantly decreased numbers of lung-residentCD1031CD11b2 dendritic cells and CD1032CD11b2 macrophagesafter infection (Fig. 4b and Supplementary Fig. 7c). These changeswere specific, as there were no differences in other myeloid subsets(Supplementary Fig. 7d, e). Thus, Batf2 has a role in maintainingnumbers of Batf3-dependent CD1031 dendritic cells in the lung afterinfection with T. gondii.

We next investigated whether Batf2 could compensate for dendriticcell defects in Batf32/2 mice and for T and B cell defects in Batf2/2

mice (Fig. 4c and Supplementary Fig. 8a, b). Like Batf, retroviralexpression of Batf2 restored development of CD1031SIRP-a2 dendriticcells in Flt3L-treated Batf32/2 bone marrow. Unlike Batf and Batf3,expression of Batf2 did not restore TH17 development in Batf2/2 T cellsand only weakly restored CSR in Batf2/2 B cells. Thus, Batf2 selectivelycompensates for Batf and Batf3 in cDCs but not in T or B cells. We next

WT WT

Uninfected T. gondii

CD

45

.2+ c

ells

(%

)

0 5 10 15 20 25 30

0

25

50

75

100

Time after infection (days)

Surv

ival (%

)

WT Batf2–/–P=0.0009

a b

c

31 0.8

Empty

49

BATF

62

BATF2

CD

10

3

Batf2–/–

WT

Batf2–/– Batf2–/–

SIRP-α

d

CD103+ DCs

Batf3–/–

T. gondii infection

0.0

0.1

0.2

0.3

0.4

1.5

2.0

2.5

3.0

***

P=0.0006

CD

8α+

cD

Cs (%

)

PBS IL-120

5

10

15

20

25WT

Batf2–/–

Batf3–/–

Batf2–/–Batf3–/–

**P=0.0068

Empty

WT

****P<0.0001

0

102

103

104

105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 105 0 102 103 104 105

Figure 4 | Batf2 compensates for Batf3 in CD8a1 and CD1031 cDCdevelopment during T. gondii infection. a, Wild-type and Batf22/2 micewere infected with T. gondii and monitored for survival. n 5 29 for wild-type(dashed line) and Batf22/2 (solid line) mice. b, Shown are percentages of lungCD1031 dendritic cells of total CD45.21 cells for uninfected and infected (T.gondii) wild-type or Batf22/2 mice on day 10. n 5 5 from one of threeexperiments. c, Wild-type or Batf32/2 bone marrow cells were infected with theindicated retrovirus, cultured with Flt3L and analysed by FACS on day 10.d, Groups of five mice each of the indicated genotypes were treated with vehicle(PBS) or IL-12 and analysed by FACS after 3 days. Shown are percentages ands.e.m. of CD8a1 cDCs as a total of splenic cDCs.

24 3.3 5.2

34 36 3.1

CD8

DE

C205

WT Batf3–/– Batf–/–Batf3–/–

129 SvEV

C57BL/6

c

0.1

Empty

39

BATF

37

BATF3

15

SIRP-α

CD

103

a

b Inguinal lymph node

WT Batf3–/–

70 49 76 9

100

20

40

60

80

0

IL-10

WT Batf3–/– Batf–/–Batf3–/–Batf–/–

Empty

0

102

103

104

105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 105 0 102 103 104 105

0

102

103

104

105

0

102

103

104

105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 105 0 102 103 104 105

Figure 3 | Batf compensates for CD8a1 cDC development in Batf32/2

mice. a, Wild-type or Batf32/2 bone marrow cells were infected with GFP-RV35 (empty) or retrovirus expressing the indicated cDNA and cultured withFlt3L7. Histograms for the indicated markers are for B2202CD11c1 cells on day10. Numbers are the per cent of cells in the gate. b, Inguinal lymph nodes fromwild-type, Batf32/2 or Batf2/2Batf32/2 mice on 129SvEv or C57BL/6backgrounds were analysed by FACS. Shown are histograms for DEC205 andCD8a. c, CD4 T cells of the indicated genotype were differentiated twice underTH2 conditions5 and analysed by FACS for intracellular IL-10.

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 3

Macmillan Publishers Limited. All rights reserved©2012

examined in vivo IL-12-induced restoration of CD8a1 cDCs inBatf22/2, Batf32/2 and Batf22/2Batf32/2 mice (Fig. 4d and Sup-plementary Fig. 8c). IL-12 treatment restored CD8a1 cDCs from,1% to .5% of total cDCs in Batf32/2 mice, but this was significantlyreduced to ,2% in Batf22/2Batf32/2 mice. This suggests that Batf2 isresponsible for roughly half of IL-12-induced CD8a1 cDC restorationin Batf32/2 mice. We found similar results with in vitro CD1031 cDCdevelopment. Whereas granulocyte–macrophage colony-stimulatingfactor (GM–CSF) restored only CD103 and not DEC205 expression inFlt3L-treated Batf32/2 bone marrow cultures19, adding IFN-c withGM–CSF restored DEC2051CD1031CD11b2 dendritic cells (Sup-plementary Fig. 8d, e). We used this system to analyse compensationbetween Batf, Batf2 and Batf3 (Supplementary Fig. 8d, e). Relative towild-type bone marrow, CD1031DEC2051CD11b2 cDCs werepartially reduced in Batf2/2, Batf22/2 and Batf32/2 singly deficientbone marrow, but were reduced to an even greater extent in bothBatf2/2Batf32/2 and Batf22/2Batf32/2 bone marrow relative to allsingle-deficient bone marrow cultures. Collectively, these resultssuggest that Batf and Batf2 both act in the cytokine-dependent rescueof CD8a1 cDC development in Batf32/2 mice.

The BATF LZ domain interacts with IRF4 and IRF8To understand BATF cross-compensation, we analysed chimaericproteins containing fused domains of BATF, BATF2 and FOS(Supplementary Fig. 9). FOS inactivity could result from inhibitory

actions of amino- and carboxy-terminal domains missing in BATFand BATF3 (Fig. 5a). However, removing these domains from FOS(D59FOSD39) or fusing the BATF DNA-binding domain (DBD) withthe FOS leucine zipper (LZ) (BBRFFD39) failed to restore CD1031

cDC development (Fig. 5a), TH17 development or CSR activity(Supplementary Fig. 10a, b). However, fusing the BATF LZ with theFOS DBD (D59FFQBB) fully restored CD1031 cDC development inFtl3L-treated Batf32/2 bone marrow cultures, fully restored CSR inBatf2/2 B cells and exhibited 20% of wild-type activity for restoringIL-17 production. Furthermore, BATF with a C-terminal GFP(BATF–GFP) or FOS (BBBF) domain had .40% of wild-typeBATF activity in all three assays (Supplementary Fig. 11). Thus, theLZ domain, not the DBD, determines BATF specificity.

Removing the BATF2 C-terminal domain (BATF2 DM) allowedfor full restoration of CD1031 cDC development, but led to onlypartial restoration of CSR activity (,15%), and no restoration ofTH17 development (Supplementary Fig. 10c–h). Fusing the BATFDBD with the BATF2 LZ (B1HB2 DM) allowed for approximately50% of wild-type BATF activity for CSR and TH17 developmentin Batf2/2 B and T cells. Fusing the BATF2 DBD with BATF LZ(B2QB1) allowed for 50% of wild-type BATF activity in CD1031

cDC development, but provided less than 5% wild-type BATFactivity in TH17 development. Thus, the BATF2 DBD restricts activityin T and B cells, but its LZ functions for all BATF lineage-specificactivities.

0.0

Δ5′FOSΔ3′0.4

BBRFFΔ3′38 Δ5′FFQBB

0.3

Empty 51 BATF

SIRP-α

CD

103

a b

BATF

BBRFFΔ3′

Δ5′FFQBB

Δ5′FOSΔ3′

5′

Batf3–/–

BATF

AICE1 EICE (PU.1)

IRF4BAT

F

Δ5′FO

SΔ3′

Δ5′FF

QBB

BBRFFΔ3′c

BAT

F

Non

e

Non

e

Non

e

Eα AICE1

M1

M2Comp:

Probe:d

0.3 Empty 53 BATF 15 H55Q 14 H55Q

L56A

2.8 H55Q

L56A

K63D

E77K

SIRP-α

CD

10

3

e

AICE1

BAT

F

H55Q

H55Q

L56A

K63D

E77K

H55Q

L56A

K63D

E77K

BAT

F

K63D

R69Q

K70T

E77K

E59D

D60Q

K63D

E77K

f

No

Ab

Anti-F

OS

Anti-B

ATF

Anti-I

RF4

No

Ab

Anti-F

OS

Anti-B

ATF

Anti-I

RF4

AP1 AICE1Probe:

FOS

15

Empty

SIRP-α

CD

103

WT

IRF4

AP1Probe:

3′DB H LZ

1 2 3 4 5 6 7 8

––

–+ –

+ ++

––

–+ –

+ ++

––

–+ –

+ ++

1 2 3 4 5 6 7 8 9 10 11 12

– + + + +

1 2 3 4 5

M1 M2 AICE1

1 2 3 4 5 6 7

– + + + + + + +

1 2 3 4 5 6 7 8

Figure 5 | BATF leucine zipper interactions with non-AP1 factors mediatelineage-specific actions. a, Structures of chimaeric proteins are shown below adiagram of FOS. DNA binding domain (DB), hinge (H), leucine zipper (LZ), N(59) and C termini (39) are indicated. Flt3L-treated wild-type or Batf32/2 bonemarrow cells infected with the indicated retrovirus were analysed after 10 days.b, 293FT cells expressing both Batf and Irf4 (upper panel) or Batf and Irf4 asindicated (lower panel) were analysed by EMSA with the indicated probes and

antibodies c, 293FT cells expressing IRF4 (1) and the indicated BATFchimaera were analysed by EMSA with the AICE1 probe. d, B cells wereanalysed by EMSA with the indicated probe and competitor oligonucleotides(comp). e, Batf32/2 bone marrow cells infected with the indicated BATFretroviruses encoding BATF were analysed as in a. f, 293FT cells expressing theindicated BATF mutants were analysed by EMSA with the AICE1 probe.

RESEARCH ARTICLE

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

A recent study24 proposed that IRF4 and BATF may interact.Correspondence between the phenotypes of Batf2/2 with Irf42/2

mice, lacking TH17 development and CSR, and Batf32/2 withIrf82/2 mice, lacking CD8a1 cDC development (SupplementaryFig. 12)5–7,25–28, suggested that BATF and BATF3 may cooperate withboth IRF4 and IRF8. Furthermore, ChIP-seq analysis of BATF andIRF4 revealed coincident binding to composite AP1 and IRFmotifs29,30. By analogy with the ETS–IRF composite elements(EICE)31, the AP1–IRF composite elements are designated asAICE29,30. Recalling the FOS LZ interaction with NF-AT that mediatescooperative binding to Il2 regulatory regions32, we therefore askedwhether the BATF LZ interacted with non-AP1 factors, includingIRF4.

Electrophoretic mobility shift assays (EMSAs) demonstratedinteractions between BATF and both IRF4 and IRF8 (Fig. 5 andSupplementary Figs 13 and 14). The BATF–JUN complex that formedon an AP1 consensus probe1,2 was unchanged by addition of IRF4 orIRF8. Its abundance was increased by additional JUNB (Supplemen-tary Fig. 13a). However, using an AICE from the Ctla4 locus, a slowermobility complex formed with addition of either Irf4 or Irf8, whichrequired the IRF consensus element (Fig. 5b and SupplementaryFig. 13b). This BATF–JUN–IRF4 complex required both BATF andIRF protein, and IRF4 was unable to bind the AICE1 probe withoutBATF (Fig. 5b). In contrast, IRF4 was able to bind to an ETS–IRFconsensus element33 (EICE) in the presence of PU.1 independently ofBATF (Fig. 5b). The BATF–JUN–IRF4 complex had similar mobilityto the FOS–JUN complex in activated cells, but could be distinguishedby antibody supershifts (Supplementary Fig. 13c). Spatial constraintson the BATF–IRF4 interaction are suggested by lack of BATF–IRF4complex formation on the AICE2 probe with reversed orientation ofthe AP1 and IRF elements (Supplementary Fig. 13a, c).

The ability of chimaeric BATF proteins to interact with IRF4correlated with their functional activity (Fig. 5c). The activeD59FFQBB interacted with IRF4, but the inactive D59FOSD39 andBBRFFD39 did not (Fig. 5c), although all three bound an AP1consensus (Supplementary Fig. 13d). A BATF–IRF complex alsoformed in B and T cells, requiring both the IRF and AP1 consensuselements (Fig. 5d and Supplementary Fig. 14c), and involved contri-bution by endogenous IRF4 and IRF8 (Supplementary Fig. 14a, b, d).The BATF–IRF complex was completely absent in Batf2/2 B cells, butwas partially retained in single-deficient Irf42/2 or Irf82/2 B cellextracts (Supplementary Fig. 14a, b). Supershifts using anti-IRF4and anti-IRF8 antibodies in Irf42/2 or Irf82/2 B cell nuclear extractsshowed that the endogenous BATF–IRF complex is composed of amixture of IRF4 and IRF8 (Supplementary Fig. 14b). Similarly, super-shifts of primary TH17 cells demonstrate an interaction betweenBATF and both IRF4 and IRF8 (Supplementary Fig. 14d).

We identified BATF LZ mutations which abrogated both functionand IRF4 interactions (Fig. 5e, f). Positions b, c and f of the FOS LZface away from the parallel coiled-coil of the JUN LZ and mediateinteractions with NF-AT32 (Supplementary Fig. 9d). Several BATFmutations in these positions had no effect on activity (Supplemen-tary Table 1 and Supplementary Fig. 9c). However, four residuestogether controlled nearly all BATF-specific activity (Fig. 5e andSupplementary 15). BATF(L56A), BATF(K63D) and BATF(E77K)showed no reduction in CD1031 cDC development, and retained.60% of wild-type BATF activity in CSR. In contrast, BATF(H55Q)activity was reduced by nearly 70% (Fig. 5e). Double mutantsBATF(K63D/E77K) and BATF(H55Q/L56A) were reduced by 50%and 75%, but quadruple mutant BATF(H55Q/L56A/K63D/E77K)had less than 10% of wild-type BATF activity in CD1031 cDC develop-ment (Fig. 5e). This loss of activity was specific, as two other quadruplemutants—BATF(E59Q/D60Q/K63D/E77K) and BATF(K63D/R69Q/K70D/E77K)—maintained .50% of wild-type BATF activity(Supplementary Fig. 15c, d). A similar pattern was seen for CSR inBatf2/2 B cells (Supplementary Fig. 15b). The activity of these mutants

correlated with IRF4 interaction (Fig. 5f). Wild-type BATF and func-tional mutant BATF(K63D/E77K), and both functional quadruplemutants, formed a complex with IRF4, whereas the less activeBATF(H55Q) and BATF(H55Q/L56A), and the completely inactiveBATF(H55Q/L56A/K63D/E77K), did not, although all were stablyexpressed and could bind an AP1 site (Supplementary 13e, f).

DiscussionWe have uncovered a cytokine-driven pathway for expansion of func-tional CD8a1 cDCs occurring physiologically during infection byintracellular pathogens. This pathway relies on functional compensa-tion for Batf3 provided by Batf and Batf2 through a shared specificitydefined by the BATF LZ domain to support CD8a1 cDC and TH17development, and CSR in B cells. This compensatory pathway ofCD8a1 cDC development may provide a basis for augmentingtherapeutic immune responses. The basis of this shared specificityis the LZ interaction with non-AP1 factors, including IRF4 andIRF8, extending the repertoire of AP1 interactions beyond therecognized association of FOS with NF-AT32. The ability of bothBATF and BATF3 to support both IRF4- and IRF8-dependent lineageactivities implies that the specificity for gene activation is determinedlargely by the IRF factors. An important next step will be to determinehow regulatory elements discriminate between these distinct complexes.

METHODS SUMMARYMice. Batf2/2, Batf22/2 and Batf32/2 mice and intercrosses were bred in ourspecific pathogen-free animal facility according to institutional guidelines.Irf82/2 mice from the European Mutant Mouse Archive and BXH2/TyJ micefrom The Jackson Laboratory were maintained on the C57BL/6 background.Germline-deleted Irf42/2 mice were produced from crossing B6.129S1-Irf4tm1Rdf/J mice obtained from The Jackson Laboratory with a CMV-Cre deleterstrain on the C57BL/6 background. Experiments were in accordance withprocedures approved by the AAALAC accredited Animal Studies Committeeof Washington University in St Louis.In vivo IL-12 treatment. Recombinant murine IL-12 (Peprotech and Pfizer) wassuspended in pyrogen-free saline at 2.5mg ml21. Mice were injected intraperitoneallywith 0.5mg of IL-12 two times 3 days before analysis. For tumour transplantation andtumour collection experiments, mice were treated on days 4, 3 and 1 before tumourinoculation and 24 h after tumour inoculation.

Full Methods and any associated references are available in the online version ofthe paper.

Received 31 May; accepted 31 August 2012.

Published online 19 September 2012.

1. Dorsey, M. J. et al. B-ATF: a novel human bZIP protein that associates withmembers of the AP-1 transcription factor family. Oncogene 11, 2255–2265(1995).

2. Wagner, E. F. & Eferl, R. Fos/AP-1 proteins in bone and the immune system.Immunol. Rev. 208, 126–140 (2005).

3. Williams, K. L. et al. Characterization of murine BATF: a negative regulator ofactivator protein-1 activity in the thymus. Eur. J. Immunol. 31, 1620–1627 (2001).

4. Echlin, D. R., Tae, H. J., Mitin, N. & Taparowsky, E. J. B-ATF functions as a negativeregulator of AP-1 mediated transcription and blocks cellular transformation byRas and Fos. Oncogene 19, 1752–1763 (2000).

5. Schraml, B. U. et al. The AP-1 transcription factor Batf controls TH17differentiation. Nature 460, 405–409 (2009).

6. Ise, W. et al. The transcription factor BATF controls the global regulators of class-switch recombination in both B cells and T cells. Nature Immunol. 12, 536–543(2011).

7. Hildner, K. et al. Batf3 deficiency reveals a critical role for CD8a1 dendritic cells incytotoxic T cell immunity. Science 322, 1097–1100 (2008).

8. Edelson, B. T. et al. Peripheral CD1031 dendritic cells form a unified subsetdevelopmentally related to CD8a1 conventional dendritic cells. J. Exp. Med. 207,823–836 (2010).

9. Betz, B. C. et al. Batf coordinates multiple aspects of B and T cell function requiredfor normal antibody responses. J. Exp. Med. 207, 933–942 (2010).

10. Mashayekhi, M. et al. CD8a1 dendritic cells are the critical source of interleukin-12that controls acute infection by Toxoplasma gondii tachyzoites. Immunity 35,249–259 (2011).

11. Bar-On, L. et al. CX3CR11 CD8a1 dendritic cells are a steady-state populationrelated to plasmacytoid dendritic cells. Proc. Natl Acad. Sci. USA 107,14745–14750 (2010).

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 2 | V O L 0 0 0 | N A T U R E | 5

Macmillan Publishers Limited. All rights reserved©2012

12. Behnke, M. S. et al. Virulence differences in Toxoplasma mediated by amplificationof a family of polymorphic pseudokinases. Proc. Natl Acad. Sci. USA 108,9631–9636 (2011).

13. Ginhoux, F. et al. The origin and development of nonlymphoid tissue CD1031 DCs.J. Exp. Med. 206, 3115–3130 (2009).

14. Modlin, R. L. & Barnes, P. F. IL12 and the human immune response tomycobacteria. Res. Immunol. 146, 526–531 (1995).

15. denHaan, J.M., Lehar, S.M.&Bevan,M. J.CD81 butnotCD82 dendritic cellscross-prime cytotoxic T cells in vivo. J. Exp. Med. 192, 1685–1696 (2000).

16. Wilson, N. S. et al. Systemic activation ofdendritic cellsby Toll-like receptor ligandsor malaria infection impairs cross-presentation and antiviral immunity. NatureImmunol. 7, 165–172 (2006).

17. Aliberti, J.et al. Essential role for ICSBP in the in vivodevelopment ofmurineCD8a1

dendritic cells. Blood 101, 305–310 (2003).18. Tailor, P., Tamura, T., Morse, H. C. & Ozato, K. The BXH2 mutation in IRF8

differentially impairs dendritic cell subset development in the mouse. Blood 111,1942–1945 (2008).

19. Edelson, B. T.et al. Batf3-dependentCD11blow/2 peripheral dendritic cells areGM-CSF-independent and are not required for Th cell priming after subcutaneousimmunization. PLoS ONE 6, e25660 (2011).

20. Marquis, J. F., Lacourse,R., Ryan, L.,North, R. J.& Gros,P.Disseminatedand rapidlyfatal tuberculosis in mice bearing a defective allele at IFN regulatory factor 8.J. Immunol. 182, 3008–3015 (2009).

21. Jackson, J. T. et al. Id2 expression delineates differential checkpoints in the geneticprogram of CD8a1 and CD1031 dendritic cell lineages. EMBO J. 30, 2690–2704(2011).

22. Edelson, B. T. et al. CD8a1 dendritic cells are an obligate cellular entry point forproductive infection by Listeria monocytogenes. Immunity 35, 236–248 (2011).

23. Su, Z. Z. et al. Cloning and characterization of SARI (suppressor of AP-1, regulatedby IFN). Proc. Natl Acad. Sci. USA 105, 20906–20911 (2008).

24. Ravasi, T. et al. An atlas of combinatorial transcriptional regulation in mouse andman. Cell 140, 744–752 (2010).

25. Brustle, A. et al. The development of inflammatory TH17 cells requires interferon-regulatory factor 4. Nature Immunol. 8, 958–966 (2007).

26. Klein, U. et al. Transcription factor IRF4 controls plasma cell differentiation andclass-switch recombination. Nature Immunol. 7, 773–782 (2006).

27. Tamura, T. & Ozato, K. ICSBP/IRF-8: its regulatory roles in the development ofmyeloid cells. J. Interferon Cytokine Res. 22, 145–152 (2002).

28. Sciammas, R. et al. Graded expression of interferon regulatory factor-4coordinates isotype switching with plasma cell differentiation. Immunity 25,225–236 (2006).

29. Glasmacher, E. et al. A genomic regulatory element that directs assembly andfunction of immune-specific AP-1-IRF complexes. Science. (in the press).

30. Li, P. et al. BATF–JUN is critical for IRF4-mediated transcription in T cells. Naturehttp://dx.doi.org/10.1038/nature11530 (19 September 2012).

31. Brass, A. L., Zhu, A. Q. & Singh, H. Assembly requirements of PU.1-Pip (IRF-4)activator complexes: inhibiting function in vivo using fused dimers. EMBO J. 18,977–991 (1999).

32. Chen, L., Glover, J. N., Hogan, P. G., Rao, A. & Harrison, S. C. Structure of the DNA-binding domains from NFAT, Fos and Jun bound specifically to DNA. Nature 392,42–48 (1998).

33. Eisenbeis, C. F., Singh, H. & Storb,U. Pip, a novel IRF family member, is a lymphoid-specific, PU.1-dependent transcriptional activator. Genes Dev. 9, 1377–1387(1995).

34. Heng, T. S. & Painter, M. W. The Immunological Genome Project: networks of geneexpression in immune cells. Nature Immunol. 9, 1091–1094 (2008).

35. Ranganath, S. et al. GATA-3-dependent enhancer activity in IL-4 gene regulation. J.Immunol. 161, 3822–3826 (1998).

Supplementary Information is available in the online version of the paper.

AcknowledgementsThisworkwas supportedby theHowardHughesMedical Institute,National Institutes of Health (AI076427-02) and Department of Defense(W81XWH-09-1-0185) (K.M.M.), the American Heart Association (12PRE8610005)(A.T.S.), German Research Foundation (AL 1038/1-1) (J.C.A), American Society ofHematology Scholar Award and Burroughs Welcome Fund Career Award for MedicalScientists (B.T.E.), and Cancer Research Institute predoctoral fellowship (W.-L.L.). Wethank the ImmGen consortium34, M. White for blastocyst injections and generation ofmouse chimaeras, the Alvin J. Siteman Cancer Center at Washington University Schoolof Medicine for use of the Center for Biomedical Informatics and Multiplex GeneAnalysis Genechip Core Facility. The Siteman Cancer Center is supported in part by theNCI Cancer Center Support Grant P30 CA91842. IL-12 was a gift from Pfizer.

Author Contributions R.T., W.-L.L., T.L.M. and M.M. performed experiments withBatf2/2, Batf22/2, Batf32/2 and double-knockout mice; T.L.M. made and analysed allBATF mutants; J.A.R. aided with EMSAs; W.KC, J.C.A. and A.T.S. were involved withmicroarray analysis andgeneration of mutant mice. J.C.A. was involved in generating ofBatf22/2 mice. B.T.E. was involved with L. monocytogenes analysis; N.M.K was involvedwith cross-presentation analysis. X.W. was involved with bioinformatic analysis; L.A.W.and C.L.S. were involved with M. tuberculosis infection; E.G. and H.S. helped to identifyEMSA probes; P.L., W.L. and W.J.L. performed ChIP-seq and helped to identify EMSAprobes; M.B. and L.D.S. aided with T. gondii infections; S.S.K.L., C.T.A. and R.D.S. aidedwith tumour models. H.S. and W.J.L. provided helpful discussions. K.M.M. directed theworkandwrote themanuscript. All authorsdiscussed the results andcontributed to themanuscript.

Author Information All original microarray data have been deposited in NCBI’s GeneExpression Omnibus under accession number GSE40647. Reprints and permissionsinformation is available at www.nature.com/reprints. The authors declare nocompeting financial interests. Readers are welcome to comment on the online versionof the paper. Correspondence and requests for materials should be addressed toK.M.M. (kmurphy@wustl.edu).

RESEARCH ARTICLE

6 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

METHODSGeneration of Batf22/2 mice. The targeting construct was assembled byGateway recombination cloning system (Invitrogen). To construct pENTRloxFRT rNEO, a floxed PGK-neor (phosphoglycerate kinase promoter-neomycinphosphotransferase) gene cassette (1,982 bp) was excised from the pLNTKtargeting vector36 using SalI and XhoI. After incubation with Easy-A cloningenzyme (Stratagene) and dNTPs to generate 39-A overhangs, the PGK-neor

cassette was ligated into the multiple cloning site of pGEM-T Easy (Promega).The resulting 2,022-bp PGK-neor gene cassette was released using NotI andligated into the 2,554-bp backbone of NotI-digested pENTR lox-Puro37.Flippase Recognition Target (FRT) sites were sequentially inserted at the SacIand HindIII sites using DNA fragments generated by following annealingoligonucleotides: SacII-FRT-A (59-GAAGTTCCTATTCCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCCGC-39) and SacII-FRT-B (59-GGAAGTTCCTATACTTTCTAGAGAATAGGAACTTCGGAATAGGAACTTCGC-39),or HindIII-FRT-A (59-AGCTTGAAGTTCCTATTCCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTC-39) and HindIII-FRT-B (59-AGCTGAAGTTCCTATACTTTCTAGAGAATAGGAACTTCGGAATAGGAACTTCA-39).

To construct pENTR-BATF2-5HA, the 59 homology arm was generated byPCR from genomic DNA using the following oligonucleotides, which contain theattB4 and attB1r site: 59-GGGGACAACTTTGTATAGAAAAGTTGGCAGGCTGAAGCAGGGACAC-39, and 59-GGGGACTGCTTTTTTGTACAAACTTGCCCTAGCACACCCAGTTTCAGTTTC-39. The attB4-attB1r PCR fragmentwas ligated into pDONR(P4-P1R) plasmid (Invitrogen) by BP recombinationreaction. To construct pENTR-BATF2-3HA, the 39 homology arm was generatedby PCR from genomic DNA using the following oligonucleotides which containattB2 and attB3site: 59-GGGGACAGCTTTCTTGTACAAAGTGGGCAGTCCCAGCATGACCCAGCCCCAGCATGACCCAGCATCTGC-39, and 59-GGGGACAACTTTGTATAATAAAGTTGCCGTCTCACTCAGTTCTGTGTGTG-39.The attB2-attB3 PCR fragment was ligated into pDONR(P2-P3) plasmid(Invitrogen) by BP recombination reaction, following by the LR recombinationreaction to generate final targeting construct by using pENTR-BATF2-5HA,pENTR-BATF2-3HA, pENTR-loxFRT rNEO, and pDEST DTA-MLS. Thelinearized vector was electroporated into EDJ22 embryonic stem cells, on the129 SvEv background, and targeted clones were identified by Southern blotanalysis with 59 and 39 probes. Blastocyst injections were performed, and malechimaeras were bred to female 129S6/SvEv mice. Probes for Southern blotanalysis were amplified from genomic DNA using the following primers:59-GTTGGTGTGAGGTGATGAGGTCCC-39 and 59-CTTGACTTCCTAGACCAGGGGC-39 for the 59 probe; 59-GGACCATATACTCTTACTGTTGAAACC-39 and 59-GGGCTGGGGACACACATG-39 for the 39 probe. Southern blotanalysis and genotyping PCR were performed to confirm germline transmissionand the genotype of the progeny. Primers used for genotyping PCR are shown asfollows: knockout screen forward, 59-GAAACTGAAACTGGGTGTGCTAGGG-39; knockout screen reverse, 59- CGCCTTCTTGACGAGTTCTTCTGAG-39; wild-type screen reverse, 59-GCCTTCCTTGTCTCTCTCCATAGCG-39.Generation of BATF2-specific rabbit polyclonal antibody. Murine full-lengthBatf2 cDNA was cloned into the pET-28a (1) expression vector (Novagen) togenerate recombinant BATF2 protein with a His tag. Recombinant BATF2 wasthen expressed using Escherichia coli BL21 (Invitrogen). Purified recombinantBATF2 was used to immunize New Zealand white rabbits (Harlan), and rabbitanti-mouse BATF2 sera were collected and tested by ELISA and western blotanalysis.Pathogen infections. Listeria monocytogenes and T. gondii infections werecarried out as previously described38,39. The type II avirulent Prugniaud (Pru)strain of T. gondii expressing a firefly luciferase and GFP transgene (provided byJ. Boothroyd) was used for infections of Batf22/2 mice for the experiments shownin Fig. 4 and Supplementary Fig. 7. Briefly, 800 tachyzoites were injectedintraperitoneally (i.p.) into mice. For herpes simplex virus 1 (HSV-1) infections,mice were infected with the KOS strain subcutaneously (s.c.) with 1.5 3 105

plaque-forming units (p.f.u.) per mouse in the footpad. Viral supernatant andHSV peptide were provided by M. Colonna. For M. tuberculosis infections,Erdman strain was grown to a density of 8 3 106 colony-forming units (c.f.u.) perml, and mice were exposed to aerosol infection for 40 min using a Glas-Colaerosol exposure system. After 24 h, two mice per group were killed, and lungswere collected to determine infection efficiency, which was about 100 c.f.u. perlung. Lungs and spleens were collected at 1, 3 and 9 weeks after infection. The leftlobe of the lung and one-third of the spleen were used for histological examina-tion upon formalin fixation. The right lobe of the lung and two-thirds of thespleen were divided into c.f.u. analysis and FACS analysis. For c.f.u. analysis,organs were homogenized and plated at various dilutions on 7H-10 culture plates,and after 3 weeks, colonies were counted. For FACS analysis, organs wereprepared as described40. For lungs, Ficoll gradient purification was performed

before surface staining. All samples were fixed upon staining for 20 min in 4%formalin. Serum was collected by retro-orbital bleeding at week 1, 2, 3 and 6 anddecontaminated by double filtering through a 0.2-mm filter.Bioluminescence imaging. Imaging was done as previously described41. In brief,mice were injected intraperitoneally with D-luciferin (Biosynth AG) at 150 mg kg21

and allowed to remain active for 5 min before being anaesthetized with 2%isoflurane for 5 min. Animals were than imaged with a Xenogen IVIS200 machine(Caliper Life Sciences). Data were analysed with the Living Image software (CaliperLife Sciences).In vitro T-cell re-stimulation after HSV infection. One week after infection,spleens were collected and re-stimulated with the HSV peptide HSV-gB2 (498-505)(Anaspec). Briefly, 2 3 106 splenocytes were re-stimulated for 5 h in the presence ofbrefeldin A at 1 mg ml21 and analysed by FACS for intracellular IFN-c and TNF-aproduction as described later.In vitro cross-presentation. Dendritic cell cross-presentation of antigen toCD81 OT-I T cells was assessed as previously described42. Briefly, spleens fromnaive, vehicle or IL-12-treated wild-type or Batf32/2 mice were digested withcollagenase B (Roche) and DNase I (Sigma-Aldrich). CD11c1 dendritic cells wereobtained by negative selection using B220, Thy1.2, and DX5 microbeads followedby positive selection with CD11c microbeads by MACS purification (MiltenyiBiotec). CD8a1 and CD41 cDCs were cell sorted on a FACSAria II flowcytometer (post-sort purity .96%). Splenocytes from Kb2/2Db2/2b2m2/2 micewere prepared in serum-free medium, loaded with 10 mg ml21 ovalbumin (EMD)by osmotic shock, and irradiated (13.5 Gy) as described previously42. OT-IT cells were purified from OT-I/Rag22/2 mice by CD11c and DX5 negativeselection, followed by positive selection with CD8a microbeads (purity .96%).T cells were fluorescently labelled by incubation with 1mM CFSE (Sigma-Aldrich)for 9 min at 25 uC at a density of 2 3 107 cells ml21. For the cross-presentationassay, 5 3 104–105 purified dendritic cells were incubated with 5 3 104–105

CFSE-labelled OT-I T cells in the presence of increasing numbers of irradiated,ovalbumin-loaded Kb2/2Db2/2b2m2/2 splenocytes (5 3 103 to 1 3 105). After3 days, OT-I T cell proliferation was analysed by CSFE dilution gating onCD31CD81CD45.11 cells.Tumour transplantation. Methylcolanthrine-induced fibrosarcomas werederived from 129SvEv strain Rag22/2 or wild-type mice as described previ-ously43. Tumour cells were propagated in vitro and 106 tumour cells (129SvEvfibrosarcoma tumour line H31m1 or D42m1) were injected s.c. in a volume of150ml endotoxin-free PBS into the shaved flanks of naive or IL-12-conditioned(as described above) wild-type Batf32/2 and Rag2 2/2 recipient mice as describedpreviously42. Injected cells were .90% viable as assessed by trypan blue exclusion.Tumour size was measured on the indicated days and is presented as the mean oftwo perpendicular diameters.Tumour harvest. Wild-type and naive or IL-12-treated Batf32/2 mice wereinoculated with the H31m1 fibrosarcoma tumour line. Eleven days after, tumourswere removed, minced and treated with 1mg ml21 type A collagenase (Sigma) inHBSS (Hyclone) for 1 h at 37 uC. Cell suspension was analysed by FACS for CD8T-cell recruitment at the tumour site gating on CD451, Thy1.21 and CD8a1

cells.Quantitative RT–PCR. For gene expression analysis, RNA was prepared fromvarious cell types with either RNeasy Mini kit (Qiagen) or RNeasy Micro kit(QIAGEN), and cDNA was synthesized with Superscript III reverse transcription(Invitrogen). Real-time PCR and a StepOnePlus real-time PCR system (AppliedBiosystems) were used according to the manufacturer’s instructions, with theQuantitation Standard-Curve method and HotStart-IT SYBR Green qPCRMaster Mix (Affymetrix/USB). PCR conditions were 10 min at 95 uC, followedby 40 two-step cycles consisting of 15 s at 95 uC and 1 min at 60 uC. Primers usedfor measurement of Batf2 expression are as follows: Batf2 qRT forward,59-GGCAGAAGCACACCAGTAAGG-39; Batf2 qRT reverse, 59-GAAGGGCGTGGTTCTGTTTC-39. Hprt was used as normalization control, and primersused are as follows: Hprt forward, 59-TCAGTCAACGGGGGACATAAA-39;Hprt reverse, 59-GGGGCTGTACTGCTTAACCAG-39.Dendritic cell preparation. Dendritic cells from lymphoid organs and non-lymphoid organs were collected and prepared as described40. Briefly, spleens,MLNs, SLNs (inguinal) and liver were minced and digested in 5 ml Iscove’smodified Dulbecco’s media plus 10% FCS (cIMDM) with 250mg ml21 collagenaseB (Roche) and 30 U ml21 DNase I (Sigma-Aldrich) for 30 min at 37 uC withstirring. Cells were passed through a 70-mm strainer before red blood cells werelysed with ACK lysis buffer. Cells were counted on a Vi-CELL analyser, and5–10 3 106 cells were used per antibody staining reaction. Lung cell suspensionswere prepared after perfusion with 10 ml Dulbecco’s PBS (DPBS) via injectionsinto the right ventricle after transection of the lower aorta. Dissected and mincedlungs were digested in 5 ml cIMDM with 4 mg ml21 collagenase D (Roche) for 1 h

ARTICLE RESEARCH

Macmillan Publishers Limited. All rights reserved©2012

at 37 uC with stirring. Cells from the peritoneal cavity were collected by washingthe peritoneal cavity with 10 ml HBSS plus 2% FCS and 2 mM EDTA.Bone-marrow-derived dendritic cells and macrophages. Bone-marrow-deriveddendritic cells were generated from bone marrow with 100 ng ml21 of recombinantFlt3L as described42. Bone-marrow-derived macrophages were generated frombone marrow with 20 ng ml21 of recombinant M-CSF (Peprotech) for 7 days,and rested in cIMDM without M-CSF for 24 h before stimulation with variousconditions as described in the figure legends.Preparation of T. gondii lysate antigen (STAg). Toxoplasma gondii antigen wasprepared by lysing tachyzoites of the Pru strain in foreskin fibroblast culturesthrough two cycles of freeze–thaw in liquid nitrogen. Lysate was then filtered,aliquoted at a concentration of 1 mg ml21 in PBS and stored at 280 uC. Fordendritic cell stimulation, 0.05 mg ml21 were used and intracellular IL-12 pro-duction was analysed by FACS.Cytokine-induced CD8a1 dendritic cells. Wild-type and Batf32/2 mice bonemarrow was prepared as described above. GM–CSF (10 ng ml21) and or IFN-c(0.1 ng ml21) was added to the cultures between days 8 and 10. Cells were ana-lysed 2 days after cytokine addition.Antibodies and flow cytometry. Staining was performed at 4 uC in the presenceof Fc block (anti-CD16/32 clone 2.4G2, BioXCell) in FACS buffer (DPBS plus0.5% BSA plus 2 mm EDTA). The antibodies used for dendritic cell analysis wereas recently described44. Cells were analysed on BD FACSCanto II or FACSAria IIand analysed with FlowJo software (Tree star, Inc.).Intracellular cytokine staining. For intracellular cytokine staining, cellswere surface stained, then fixed in 2% paraformaldehyde for 15 min at 4 uC,permeabilized in DPBS plus 0.1% BSA plus 0.5% saponin, and stained forintracellular cytokines as previously described45. Additional antibodies includedanti-TNF-a (MP6-XT22) and anti-IFN-c (XMG1.2) from BioLegend.ELISA and CBA. IL-12p40 concentration was measured from serum sampleswith the mouse IL-12p40 OptEIA ELISA set (BD Bioscience) according to themanufacturer’s instructions. The concentration of inflammatory cytokines wasmeasured in the serum with the BD CBA mouse inflammation kit (BDBiosciences), and data were analysed with FCAP Array software (Soft Flow, Inc.).Statistical analysis. Differences between groups in survival were analysed by thelog-rank test. Analysis of all other data was done with an unpaired, two-tailedStudent’s t-test with a 95% confidence interval (Prism; GraphPad Software, Inc.).P values less than 0.05 were considered significant. *0.01 , P , 0.05;**0.001 , P , 0.01; ***P , 0.001; ****P , 0.0001.Expression microarray analysis. Total RNA was isolated from cells using theAmbion RNAqueous-Micro kit. For mouse genome 430 2.0 arrays, RNA wasamplified, labelled, fragmented, and hybridized using the 39 IVT Express kit(Affymetrix). Data were normalized and expression values were modelled usingDNA-Chip analyser (dChip) software (http://www.dChip.org)46. For mouse gene1.0 ST arrays, RNA was amplified with the WT expression kit (Ambion) andlabelled, fragmented and hybridized with the WT terminal labelling andhybridization kit (Affymetrix). Data were processed using RMA quantilenormalization and expression values were modelled using ArrayStar software(DNASTAR). All original microarray data have been deposited in NCBI’sGene Expression Omnibus under accession number GSE40647.CD41 T-cell cultures. CD41 T cells were purified using the DynaBeadsFlowComp mouse CD4 kit (Invitrogen) and were activated on anti-CD3/anti-CD28 coated plates under the following culture conditions: TH1 conditions:anti-IL-4 10 mg ml21 (11B11, BioXcell), IFN-c 0.1mg ml21 (Peprotech), IL-1210 U ml21, IL-2 40 U ml21; TH2 conditions: anti-IL-12 10mg ml21 (Tosh,BioXCell), anti-IFN-c 10mg ml21 (XMG1.2, BioXCell), IL-4 10 ng ml21

(Peprotech), IL-2 40 U ml21; TH17 conditions: IL-6 25 ng ml21 (Peprotech),TGF-b 2 ng ml21 (Peprotech), IL-1b 10 ng ml21 (Peprotech), anti-IL-4 10 mgml21 (11B11, BioXCell), anti-IFN-c 10 mg ml21 (XMG1.2, BioXCell), and anti-IL-12 10 mg ml21 (Tosh, BioXCell). IL-2 (40 U ml21) was added for the TH17cultures in Supplementary Fig. 12a. Cells were diluted three-fold in fresh mediaon day 3. On day 5, cells were activated with PMA/ionomycin for analysis ofcytokines and surface markers by FACS. For some experiments, cells werere-stimulated on day 7 under the same conditions and analysed 5 days later.Retroviral analysis of BATF functional activity. For CD8a1 dendritic celldevelopment, bone marrow was cultured with Flt3L as described42, infected withretrovirus and 2 mg ml21 polybrene on day 1, and analysed for development ofcDCs (CD11c1B2202) on day 10.

CD41 T cells were cultured under TH17 conditions45,47 and infected withretrovirus and 6mg ml21 polybrene on day 1. Cells were analysed on day 5 forcytokine expression by intracellular staining. For some experiments, cells werere-stimulated on day 7 under the same conditions and analysed 5 days later.

For analysis of class switch recombination, B220 cells purified byaB220 MACSmicrobeads (Miltenyi) were cultured with LPS and IL-4 as described48, infected

with retrovirus and 6mg ml21 polybrene on day 1, and analysed for switching toIgG1 on day 4 by FACS.

For infection of primary mouse cells, GFP-RV49 containing cDNAs fortranscription factors, chimaeric proteins and mutated BATF were transfectedinto Phoenix E cells as described previously50. Viral supernatants were collected2 days later and concentrated by centrifugation51. Cells were infected with viralsupernatants by spin infection at 1,800 r.p.m. for 45 min at room temperature.Electromobility shift assays (EMSAs). For stable expression in 293FT cells,retroviruses (GFP-RV containing BATF, chimaeric proteins or BATF mutants,and/or truncated hCD4-RV49 containing IRF4 or IRF8 cDNA) were packaged inPhoenix A cells and concentrated by centrifugation51. Infected 293FT cells weresorted for retroviral marker expression and when indicated were transientlytransfected with JUNB-GFP-RV using calcium phosphate. For transient expres-sion, 293FT cells were transfected with retroviral constructs using calciumphosphate. Nuclear extracts were prepared 48 h after transfection. For 293FTcells, nuclei were obtained after cellular lysis with buffer A containing 0.2%NP40 and nuclear extracts in buffer C were dialysed against buffer D asdescribed52. Nuclear extracts from T and B cells were prepared as described48.

EMSA was as described53 using 3mg of nuclear extract, 1 mg poly(dIdC) (Sigma)and 7%T 3.3%C polyacrylamide gels. For competition assays, extracts were incu-bated with excess unlabelled competitor DNA for 20 min before addition of labelledprobe. For supershifts, extracts were incubated with antibodies goat anti-FOS (4)(Santa Cruz) for 293FT cells, or anti-FOS (2G9C3) (Abcam) for mouse T cells, anti-IRF4 (H-140) (Santa Cruz), anti-ICSBP (C-19) (Santa Cruz), anti-JUNB (C-11)(Santa Cruz) and rabbit anti-BATF45, rabbit anti-BATF2 (described above) andrabbit anti-BATF3 (W. KC et al., manuscript submitted) for 1 h on ice beforeaddition of labelled probe and incubation at room temperature for 35 min.

The following pairs of oligonucleotides were annealed to generate probes whichwere labelled with 32P-dCTP using Klenow polymerase. The AP1consensus bind-ing site is underlined; the IRF consensus binding site is in bold; mutated bases are inlower case.

AP1 59-GATCAGCTTCGCTTGATGAGTCAGCCGG-39/59-GATCCGGCTGACTCATCAAGCGAAG-39; AICE1 (from third intron of mouse Ctla4)59-CTTGCCTTAGAGGTTTCGGGATGACTAATACTGTA-39/59-TCACGTACAGTATTAGTCATCCCGAAACCTCTAAGG-39; AICE1 M1 (mutates theIRF consensus binding site) 59-CTTGCCTTAGAGGccaCGGGATGACTAATACTGTA-39/59-TCACGTACAGTATTAGTCATCCCGtggCCTCTAAGG-39;AICE M2 (mutates the AP1 consensus binding site) 59-CTTGCCTTAGAGGTTTCGGGAgacCTAATACTGTA-39/59-TCACGTACAGTATTAGgtcTCCCGAAACCTCTAAGG-39; AICE2 (from third intron of Il23r): 59-GATGTTTCAGGGAAAGCACTGACTCACTGGCTCTCCA-39/ 59-GGTGGAGAGCCAGTGAGTCAGTGCTTTCCCTGAAA-39; EICE54 59-GAAAAAGAGAAATAAAAGGAAGTGAAACCAAG-39/59-GATCCTTGGTTTCACTTCCTTTTATTTCTCTTT-39; Ea 59-TCGACATTTTTCTGATTGGTTAAA-39/59-GACTTTTAACCAATCAGAAAAATG-39.Plasmids. cDNAs for Batf, Batf2, Batf3, Irf8 and JunB were generated by PCRfrom primary cells. MigR1 containing HA-tagged IRF4 was from H. Singh(Genentech). HA–IRF4 and IRF8 cDNAs were subcloned into hCD4-RV49.Mutations in Batf were generated using QuickChange mutagenesis with BATF-GFP-RV as the template. cDNAs for chimaeric proteins were generated by over-lap extension and PCR and cloned into GFP-RV.

36. Gorman, J. R. et al. The Igk enhancer influences the ratio of Igk versus Igl Blymphocytes. Immunity 5, 241–252 (1996).

37. Iiizumi, S. et al. Simple one-week method to construct gene-targeting vectors:application to production of human knockout cell lines. Biotechniques 41,311–316 (2006).

38. Edelson, B. T. et al. CD8a1 dendritic cells are an obligate cellular entry point forproductive infection by Listeria monocytogenes. Immunity 35, 236–248 (2011).

39. Mashayekhi, M. et al. CD8a1 dendritic cells are the critical source of interleukin-12that controls acute infection by Toxoplasma gondii tachyzoites. Immunity 35,249–259 (2011).

40. Edelson, B. T. et al. Peripheral CD1031 dendritic cells form a unified subsetdevelopmentally related to CD8a1 conventional dendritic cells. J. Exp. Med. 207,823–836 (2010).

41. Saeij, J. P., Boyle, J. P., Grigg, M. E., Arrizabalaga, G. & Boothroyd, J. C.Bioluminescence imaging of Toxoplasma gondii infection in living mice revealsdramatic differences between strains. Infect. Immun. 73, 695–702 (2005).

42. Hildner, K. et al. Batf3 deficiency reveals a critical role for CD8a1 dendritic cells incytotoxic T cell immunity. Science 322, 1097–1100 (2008).

43. Matsushita,H. et al.Cancer exome analysis reveals a T-cell-dependent mechanismof cancer immunoediting. Nature 482, 400–404 (2012).

44. Satpathy, A. T. et al. Zbtb46 expression distinguishes classical dendritic cells andtheir committed progenitors from other immune lineages. J. Exp. Med. 209,1135–1152 (2012).

45. Schraml, B. U. et al. The AP-1 transcription factor Batf controls TH17differentiation. Nature 460, 405–409 (2009).

RESEARCH ARTICLE

Macmillan Publishers Limited. All rights reserved©2012

46. Li, C. & Wong, W. H. Model-based analysis of oligonucleotide arrays: expressionindex computation and outlier detection. Proc. Natl Acad. Sci. USA 98, 31–36(2001).

47. Bending, D. et al. Highly purified Th17 cells from BDC2.5NOD mice convert intoTh1-like cells in NOD/SCID recipient mice. J. Clin. Invest. 119, 565–572 (2009).

48. Ise, W. et al. The transcription factor BATF controls the global regulators of class-switch recombination in both B cells and T cells. Nature Immunol. 12, 536–543(2011).

49. Ranganath, S. et al. GATA-3-dependent enhancer activity in IL-4 gene regulation.J. Immunol. 161, 3822–3826 (1998).

50. Sedy, J. R. et al. B and T lymphocyte attenuator regulates T cell activation throughinteraction with herpesvirus entry mediator. Nature Immunol. 6, 90–98 (2005).

51. Kanbe, E. & Zhang, D. E. A simple and quick method to concentrate MSCVretrovirus. Blood Cells Mol. Dis. 33, 64–67 (2004).

52. Dignam, J. D., Lebovitz, R. M. & Roeder, R. G. Accurate transcription initiation byRNA polymerase II in a soluble extract from isolated mammalian nuclei. NucleicAcids Res. 11, 1475–1489 (1983).

53. Szabo, S. J., Gold, J. S., Murphy, T. L. & Murphy, K. M. Identification of cis-actingregulatory elements controlling interleukin-4 gene expression in T cells: roles forNF-Y and NF-ATc. Mol. Cell. Biol. 13, 4793–4805 (1993); erratum 13, 5928(1993).

54. Eisenbeis, C. F., Singh, H. & Storb,U. Pip, a novel IRF family member, is a lymphoid-specific, PU.1-dependent transcriptional activator. Genes Dev. 9, 1377–1387(1995).

ARTICLE RESEARCH

Macmillan Publishers Limited. All rights reserved©2012