RESEARCH ARTICLE SUMMARY◥

IMMUNOLOGY

Differential IL-2 expression definesdevelopmental fates of follicularversus nonfollicular helper T cellsDaniel DiToro*, Colleen J. Winstead*, Duy Pham, Steven Witte, Rakieb Andargachew,Jeffrey R. Singer, C. Garrett Wilson, Carlene L. Zindl, Rita J. Luther, Daniel J. Silberger,Benjamin T. Weaver, E. Motunrayo Kolawole, Ryan J. Martinez, Henrietta Turner,Robin D. Hatton, James J. Moon, Sing SingWay, Brian D. Evavold, Casey T. Weaver†

INTRODUCTION: The adaptive immune sys-tem has evolved to mount different types ofresponses that are matched to the type of in-vading pathogen. For CD4+ T cells, this is pre-dicated on the multipotentiality of clonallyrestricted naïve T cells, which differentiateinto distinct subsets of effector T cells con-tingent on recognition of cognate antigen andcytokine cues from cells of the innate immunesystem.There are twobroaddivisions of effectorCD4+ T cells: T follicular helper (TFH) cells,

which are programmed to interact with B cellswithin lymphoid tissues to support produc-tion of high-affinity, class-switched antibodies,and non-TFH effector cells, including T helper1 (TH1), TH2, and TH17 cells, which are pro-grammed to egress from lymphoid tissues toorchestrate heightened innate immune cellfunction at sites of pathogen entry. The mech-anisms controlling bifurcation into TFH versusnon-TFH effector cell pathways are incom-pletely understood.

RATIONALE: An impediment to understand-ing mechanisms controlling TFH–non-TFH celldivergence is an absence of early markers todefine cells destined for these alternative fates.Unlike effector CD4+ T cells, which produce adiversity of cytokines that define their pheno-type and function, naïve CD4+ T cells are largelylimited to the rapid production of interleukin-2(IL-2) when activated by antigen. IL-2 is onlyproduced by a subset of activated naïve T cells,suggesting a possible relationshipbetween IL-2production and effector cell fate determination.To explore this,we developed two IL-2 reporter

mice strains with com-plementary features thatenabled the tracking anddeletion of T cells on thebasis of differential IL-2expression. This allowedus to determine whether

naïve T cells that do, or do not, produce IL-2are biased in their developmental program-ming and, if so, how.

RESULTS: RNA sequencing of naïve T cellssorted on the basis of IL-2 reporter expressionidentified cosegregation of transcripts encod-ing IL-2 and Bcl6—the signature transcriptionfactor of TFH cells. Conversely, IL-2–negative(IL-2–) cells preferentially expressed the genePrdm1, which encodes the transcriptional re-pressor Blimp1. Blimp1, in turn, antagonizesBcl6 and the TFH developmental program. Thissuggested that IL-2 producers give rise to TFHcells, whereas IL-2 nonproducers give rise tonon-TFH effector cells. Moreover, the fact thatIL-2 receptor signaling induces expression ofPrdm1 via Stat5 suggested that IL-2 producersresisted IL-2 signaling and activated IL-2 signal-ing in nonproducers in trans. Indeed, in vivostudies established that IL-2 signaling wasmostly paracrine and that depletion of IL-2–producing cells selectively impaired TFH celldevelopment. Finally, IL-2 expressionwas lim-ited to a subset of naïve T cells that received thestrongestT cell receptor (TCR) signals, establishinga link between TCR signal strength, IL-2 prod-uction, and TFH versus non-TFH differentiation.

CONCLUSION: Thisstudyprovidesnewinsightsinto themechanisms thatcontrol earlybifurcationof CD4+ T cells into TFH and non-TFH effectors.Naïve T cells that receive differing strengths ofTCR signals stratify into those that exceed athreshold predisposing them to IL-2 productionand early TFH commitment and those that donotexpress IL-2 yet receive IL-2 signaling, whichreinforces non-TFH effector commitment.▪

RESEARCH

DiToro et al., Science 361, 1086 (2018) 14 September 2018 1 of 1

The list of author affiliations is available in the full article online.*These authors contributed equally to this work.†Corresponding author. Email: cweaver@uab.eduCite this article as D. DiToro et al., Science 361, eaao2933(2018). DOI: 10.1126/science.aao2933

CXCL13

CD28

ICOS

IL-2rβ IL-2rα

TCR

CD80/86

MHCII

Non-TNon-TFHFHEffectorEffectorNon-TFHEffector

NaïveCD4+

T cell

IL-2

S1PEndothelial cell

Non-TFHBlimp1

T cell zone

B cell zone

ICOS

CD40L

CXCR5

S1PR1

GCB Cell

B Cell

GC-TFH

Germinal center

Efferent lymphatic

Antigenpresenting

cell

Treg

IL-2Bcl6

IL-2rαBlimp1

Longdwelltime

Shortdwelltime

TFHBcl6

IL-2–producing CD4+ Tcells become TFH cells,whereas IL-2 nonproducersbecome non-TFH cells. (Left) Strong TCR signaling via an antigen presentingcell induces Il2 and Bcl6 gene expression (red pathway); weakersignaling induces expression of non-TFH genes (blue pathway), includingPrdm1 and S1pr1, which encodes the S1P receptor S1PR1. Bcl6+ cells(red) secrete IL-2 in trans to Tregulatory (Treg) cells (yellow) andrecently activated IL-2– cells (blue), up-regulating IL-2 receptor IL2raand reinforcing Prdm1. (Top right) Bcl6+ cells engage cognateB cells (green) and migrate to germinal centers (GCs); Bcl6+

TFH cells mature into GC-TFH cells. (Bottom right) Prdm1+

cells migrate to efferent lymphatics and mature intonon-TFH effectors in nonlymphoid tissues. MHCII, majorhistocompatibility complex class II; ICOS, inducible costim-ulator; CXCR5, a receptor for chemokine CXCL13.

ON OUR WEBSITE◥

Read the full articleat http://dx.doi.org/10.1126/science.aao2933..................................................

on February 9, 2020

http://science.sciencem

ag.org/D

ownloaded from

RESEARCH ARTICLE◥

IMMUNOLOGY

Differential IL-2 expression definesdevelopmental fates of follicularversus nonfollicular helper T cellsDaniel DiToro1*, Colleen J. Winstead1*†, Duy Pham1, StevenWitte1, Rakieb Andargachew2,Jeffrey R. Singer1, C. Garrett Wilson1, Carlene L. Zindl1, Rita J. Luther1‡, Daniel J. Silberger1,Benjamin T.Weaver3, E. Motunrayo Kolawole2§, Ryan J. Martinez2, Henrietta Turner1,Robin D. Hatton1, James J. Moon4, Sing SingWay5, Brian D. Evavold2§, Casey T.Weaver1||

In response to infection, naïve CD4+ T cells differentiate into two subpopulations:T follicular helper (TFH) cells, which support B cell antibody production, and non-TFH cells,which enhance innate immune cell functions. Interleukin-2 (IL-2), the major cytokineproduced by naïve Tcells, plays an important role in the developmental divergence of thesepopulations. However, the relationship between IL-2 production and fate determinationremains unclear. Using reporter mice, we found that differential production of IL-2 bynaïve CD4+ Tcells defined precursors fated for different immune functions. IL-2 producers,which were fated to become TFH cells, delivered IL-2 to nonproducers destined tobecome non-TFH cells. Because IL-2 production was limited to cells receiving the strongestT cell receptor (TCR) signals, a direct link between TCR-signal strength, IL-2 production,and T cell fate determination has been established.

NaïveCD4+T cells aremultipotent precursorsthat differentiate into functionally distincteffector subsets to coordinate different as-pects of immunity. T helper 1 (TH1), TH2,and TH17 cells are products of develop-

mental pathways induced by different classes ofpathogens. They are programmed to egress fromT cell zones of secondary lymphoid tissues soonafter induction to orchestrate heightened innateimmune cell function at sites of pathogen entry.T follicular helper (TFH) cells develop concurrentlywithTH1, TH2, andTH17 cells but are programmedto migrate to B cell zones within secondary lym-phoid tissues. They provide help to B cells tosupport the production of high-affinity, class-switched antibodies. TFH- and non-TFH effectorcell development diverges early in evolving adapt-ive responses. However, the type of immuneresponse (type 1, 2, or 3) is linked such thatpathogen-clearance mechanisms mediated byinnate immune cells are amplified by coordinated

help from non-TFH cell effectors and the anti-bodies that result from TFH cell–mediated B cellhelp. Cytokines elicited from innate immune cellsby pathogens appear to be dominant in determin-ing the type of adaptive response (1), whereasthe intensity of T cell receptor (TCR) signalingappears to contribute to TFH–non-TFH cell spe-cification (2) by mechanisms that are incom-pletely understood.An impediment to understanding the mecha-

nisms controlling TFH–non-TFH cell divergence isthe absence of reliable early markers to definecells destined for these alternative fates. Unlikeeffector CD4+ T cells, which are distinguished bya diversity of cytokines that define their pheno-type and function, naïve CD4+ T cells are largelylimited to the production of interleukin-2 (IL-2),which is produced rapidly by a subset of antigen-activated cells (3). Through the activation ofStat5 and induction of Blimp1 (4, 5), IL-2 sup-presses Bcl6—a central TFH transcription factor—and, consequently, TFH development (6). Thisimplies a direct relationship between the prod-uction of IL-2 by naïve CD4+ T cells and theirdevelopment into either non-TFH or TFH effectorcells. We explored this relationship using trans-genic mice engineered to report the expressionof IL-2.

IL-2 and Bcl6 expression cosegregatewithin hours of naïve T cell activation

IL-2.eGFP reporter mice were generated by thetargeted insertion of an IRES-eGFP (internal ri-bosome entry site–enhanced green fluorescentprotein) expression cassette into the fourth exon

of the endogenous Il2 gene (Fig. 1A). Naïve CD4+

T cells from IL-2.eGFP mice stimulated undernonpolarizing conditions in vitro diverged intoCD69+IL-2+ (GFP+) and CD69+IL-2– (GFP–) sub-populations within hours of activation and beforecell division (Fig. 1, B to E). Reporter expressionwas rapidly detectable and peaked at approx-imately 24 hours before declining. This declinelagged production of IL-2 because of the rela-tively long half-life of the reporter. To definegenes differentially expressed by IL-2 producersand nonproducers, CD69+IL-2+ and CD69+IL-2–

cells were analyzed by RNA-sequencing (RNA-seq) (Fig. 1C). Among the 151 genes that werepreferentially expressed by IL-2+ cells were Bcl6and the tumor necrosis factor (TNF) super-family member Cd40l, which are important inTFH cell development or function, respectively.Also enriched in IL-2+ cells was Zbtb32, which,like Bcl6, encodes a member of the POK/ZBTBfamily of transcription factors and has beenshown to restrict the expression of TH1 and TH2cell cytokines (7). By contrast, among the 210genes preferentially expressed by IL-2– cells weremultiple genes characteristic of non-TFH effectorcell differentiation, including Prdm1, which en-codes Blimp1, as well as S1pr1 and Klf2. Similarresults were obtained from an analysis of naïveSMARTA TCR-transgenic IL-2.eGFP CD4+ T cellsstimulated with antigen (fig. S1). S1pr1 is re-quired for the egress of non-TFH effector CD4+

T cells from secondary lymphoid tissues (8), andits expression inhibits TFH development in vivo(9, 10). Klf2 suppresses TFH differentiation whilepromoting non-TFH effector cell differentiation,at least in part via the induction of Blimp1 (9).These findings, which were independently val-idated by quantitative polymerase chain reac-tion (qPCR) (Fig. 1, D and E), suggested thatIL-2 producers may be fated to become TFHcells, whereas IL-2 nonproducers may be fatedto become non-TFH effector cells. Akin to find-ings in CD4+ T cells, the differential expression ofBcl6 and Blimp1 was found in IL-2+ and IL-2–

subsets isolated from activated naïve CD8+ T cells(fig. S2). This suggests that, despite their lowerproduction of IL-2 relative to that of CD4+ T cells,early divergence of CD8+ T cells destined tobecome Blimp1+ short-lived effector cells orBcl6+ memory precursor effector cells (11) maybe similarly linked to the differential expressionof IL-2.The differential expression kinetics ofBcl6 and

Prdm1 by CD4+ T cells were discordant (Fig. 1D).Bcl6 expression tracked with Il2 expression anddecayed to background levels as Prdm1 expres-sion increased. Indeed, at the peak of differentialBcl6 expression (8 hours), Prdm1 expression re-mained at background levels in both IL-2+ andIL-2– cells. Thus, although these transcriptionfactors are believed to be directly antagonisticin the specification of TFH versus non-TFH ef-fectors (12), the rapid, reciprocal expression ofBcl6 in IL-2+ and IL-2– fractions was not con-trolled by Blimp1. Instead, we found differentialexpressionof the gene encodingMxd1 (also knownas Mad1) (Fig. 1C and fig. S3), which has been

RESEARCH

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 1 of 13

1Department of Pathology, University of Alabama atBirmingham, Birmingham, AL 35203, USA. 2Department ofMicrobiology and Immunology, Emory University, Atlanta, GA30322, USA. 3HudsonAlpha Institute for Biotechnology,Huntsville, AL 35806, USA. 4Center for Immunology andInflammatory Diseases, Massachusetts General Hospital andHarvard Medical School, Boston, MA 02129, USA. 5Division ofInfectious Diseases and Perinatal Institute, CincinnatiChildren’s Hospital, Cincinnati, OH 45229, USA.*These authors contributed equally to this work. †Present address:Merck and Company, Inc., Kenilworth, NJ 07033, USA. ‡Presentaddress: Valencia College, Orlando, FL 32825, USA. §Presentaddress: Department of Pathology, University of Utah, Salt LakeCity, UT 35203, USA.||Corresponding author. Email: cweaver@uab.edu

on February 9, 2020

http://science.sciencem

ag.org/D

ownloaded from

shown to directly bind and down-regulate Bcl6during the differentiation of germinal center Bcells into plasma cells (13). The contemporane-ous, reciprocal expression of Mxd1 and Bcl6antecedent to the expression of Prdm1 suggests

that repression of Bcl6 by Mxd1, rather than byBlimp1, may contribute to the early bifurcationof TFH and non-TFH effectors (Fig. 1D and fig. S3).Although Bcl6, like Blimp1, often acts as a

transcriptional repressor, the parallel kinetics

of Il2 and Bcl6 expression suggested that Bcl6may positively regulate Il2 expression. Thus,we performed chromatin immunoprecipitation(ChIP) analysis of conserved noncoding sequencesin the Il2 promoter and 35 kb upstream that were

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 2 of 13

1 kb

4321chromosome 3:

37,120,523 – 37,125,959

Exon 4(split) IRES eGFP loxP loxPPGK EM7 NEO bGHpA

Exon 4(split)A

C

IL-2+

IL-2–

EBcl6 Zbtb32Il2

0 102 103 104 105

0

102

103

104

105

IL-2

.eG

FP

CD

44

IL-2

.eG

FP

CD

69

0 102 103 104 105

0

102

103

104

105

88.9

0 102 103 104 105

0

102

103

104

105

57.0

0 102 103 104 105

0

102

103

104

105

39.8

58.5

B

0 103 104 105 1060

20

40

60

80

100

CD69+ IL-2–CD69+ IL-2+

CD69–

Cell Trace Violet

8 - 36 hrs

24 hours

Il2 gene

D

0.0

0.5

1.0

1.5

2.0

2.5

0.00

0.01

0.02

0.03

0.04

0.05

0.00

0.01

0.02

0.03

0.04

Cd40l Mxd1

log2 Fold Change

-log 10

(pad

j)

Il2x=3.5y=165

S1pr1

Cd2Mxd1

Klf2

Il2ra

Ltb

Socs3

Icam1

Prdm1

Cd40l

Zbtb32

Il1r2

Cxcr3

Bcl6

0

5

10

15

20

25

30

35

–2 –1 0 1 2 3

IL-2–

IL-2+

IL-2

.eG

FP

Naive 8 hours 12 hours 16 hours 20 hours 24 hours 36 hours

0 102 103 104 105

24.6

74.2

0 102 103 104 105

33.1

65.4

0 102 103 104 105

39.8

58.5

0 102 103 104 105

42.8

55.6

0 102 103 104 105

50.6

47.3

0 102 103 104 105

49.2

48.8

0 102 103 104 105

0

102

103

104

105

0

100

CD4

0.00

0.05

0.10

0.15

2^(C

TB

2M-C

TG

OI)

Time (h): 0 8 12 16 20 24 360.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0 8 12 16 20 24 360.000

0.001

0.002

0.003

0.004

0.005

0.006

0 8 12 16 20 24 36

IL-2–

IL-2+

Naive

Bcl6 Prdm1Il2

EIIa-Cre deleted

Gadd45g

CD62LCD4 CD4 CD42^

(CT

B2M

-CT

GO

I)

0.000

0.002

0.004

0.006

*** *** ***

****

S1pr1 Prdm1 Klf2

0.0000

0.0005

0.0010

0.0015

0.0020

0.000

0.005

0.010

0.005

0.020

0.000

0.001

0.002

0.003

0.000

0.002

0.004

0.006

0.008

0.000

0.002

0.004

0.006

0.008

Ltb Gadd45g

2^(C

TB

2M-C

TG

OI)

****** **

**

**

0.0000

0.0005

0.0010

0.0015

0.0020 **

Fig. 1. Differential expression of Bcl6 and Blimp1 by IL-2+ and IL-2–

Tcells (A) Gene-targeting strategy for the generation of IL-2.eGFP knockinreporter mice. The loxP-flanked neomycin resistance cassettewas deleted by crossing founders to EIIa-Cre transgenic mice. (B) Sortednaïve (GFP–CD44–CD62L+) IL-2.eGFP CD4+ T cells were labeled withCellTrace Violet (CTV), stimulated in vitro with soluble anti-CD3 (5 mg/ml)and irradiated CD4-depleted feeder cells, and then examined for expres-sion of CD69 and IL-2.eGFP by flow cytometry at the indicated timepoints. Data are representative of four experiments with at least threereplicates per condition. CTV staining was performed in two of fourexperiments. (C) Total RNA isolated from naïve IL-2.eGFP CD4+ T cellsstimulated for 18 to 24 hours as in (B) and fluorescence-activated cell

sorter (FACS)–purified into CD69+GFP– (IL-2–) or CD69+GFP+ (IL-2+)fractions was analyzed by comparative expression profiling with RNA-seq.Data depict two biological replicates per condition. padj, adjusted P value.(D) RNA isolated from IL-2.eGFP CD4+ T cells stimulated and FACS-purified as in (C) was analyzed by qPCR for expression of Il2, Bcl6, andPrdm1 at the indicated time points. Error bars represent SEM of threetechnical replicates per sample. Data are representative of fourexperiments. (E) Validation of selected transcript expression with RNAisolated from IL-2.eGFP CD4+ T cells stimulated and FACS-purified as in(C). Three technical replicates per sample are shown. Data were analyzedby using Student’s t tests and are representative of two experiments.**P < 0.01; ***P < 0.001; ****P < 0.0001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

identified by ATAC-seq (assay for transposase-accessible chromatin–sequencing) analysis asaccessible in IL-2+ cells compared with naïve andIL-2– cells (Fig. 2A). Bcl6 preferentially boundthese sites in IL-2–producing cells relative to IL-2nonproducers, at a time point (20 hours) whenthe expression of Bcl6 and Blimp1 overlapped(Fig. 2B). Blimp1 preferentially bound these sitesin IL-2– cells, as did Foxo1, which was recentlyshown to suppress TFH differentiation (14). Thepermissive histone modification trimethylatedhistone H3 lysine 4 (H3K4me3) was significantlyenriched in IL-2+ cells at the sites of Bcl6 bind-ing, whereas repressive H3K27me3 histone markswere reduced in both IL-2+ and IL-2– cells rel-ative to naïve cells. Thus, the expression of Il2correlated positively with Bcl6 binding at sitesof induced chromatin accessibility in the Il2gene locus and negatively with binding of Blimp1(and Foxo1) at the same sites. Because the ex-pression of Prdm1 significantly trailed the peakof differential Il2 expression (Fig. 1D), occupancyof these sites by Blimp1 did not appear to berequired for the repression of Il2 early in IL-2–

cells. Rather, Blimp1 appeared to act primarily toreinforce the lack of Il2 expression in the IL-2–

fraction of activated naïve T cells at later timepoints. This is consistent with Blimp1’s reportedrole as a feedback inhibitor of IL-2 (15, 16). Thesefindings indicate that, in addition to its predictivevalue in defining early precursors of TFH andnon-TFH effector cells, expression of IL-2 may bedirectly regulated by the antagonistic actionsof Bcl6 versus Blimp1 and Foxo1 at conservedcis-regulatory elements in the Il2 gene locus.Because TFH cell development occurs concur-

rently with each of the CD4+ T effector cell

pathways, we determined if the correlation be-tween reciprocal expression of Bcl6-Blimp1 andIL-2 occurred under TH1, TH2, and TH17 cellpolarizing conditions (Fig. 3A), as it did for TH0cells (Figs. 1 and 3). Under each of these activa-tion conditions, the expression of the IL-2.eGFPreporter was limited to a subset of cells ex-pressing the highest amounts of CD69 (Fig. 3A).GFP expression also correlated positively withBcl6 expression and negatively with Prdm1 ex-pression (Fig. 3B). Cells activated under TH17cell conditions expressed the highest frequencyand single-cell levels of IL-2, despite the reportedsuppression of TH17 cell differentiation by IL-2signaling (17). Thus, Il2 and Bcl6 expressionmirrored one another under each of the con-ditions examined (Fig. 3B), and the highestamounts of Il2 and Bcl6 were found in IL-2+

cells activated under TH17 cell conditions. Thismay reflect the shared requirement for IL-6 inboth TH17 and TFH cell developmental programs.

IL-2 signaling is predominantly paracrine

Although the foregoing studies suggested a linkbetween Il2 gene expression and TFH–non-TFH

cell fate determination, the differentiation ofphysiologic TFH cells ex vivo has not yet beenestablished. Thus, we examined this relationshipin vivo, in the context of infection with ActA-deficient Listeria monocytogenes (ActA-Lm)(2, 18). This attenuated type 1 bacterial pathogenwas engineered to express peptide antigens, whichenabled the tracking of endogenous antigen-specific CD4+ T cell responses with peptide-loaded major histocompatibility complex classII (MHCII) (p:MHC) tetramers or transferredTCR-transgenic T cells (2, 19). Naïve CD4+ T cells

from CD45.2+ IL-2.eGFP-SMARTA TCR-transgenicmice were transferred into wild-type (WT) CD45.1+

mice and infected with ActA-Lm that expressovalbumin (OVA) peptide and the gp66 peptiderecognized by the SMARTA TCR (ActA-Lm-OVA-gp66). Antigen-activated SMARTA T cellswere recovered near the peak of IL-2 expressionand sorted into IL-2+ and IL-2– fractions for dif-ferential gene expression analysis by RNA-seq(Fig. 4A). In agreement with our in vitro findings(Fig. 1C), IL-2+ cells were significantly enrichedfor expression of Il2, Bcl6, Zbtb32, and CD40l,whereas IL-2– cells were significantly enrichedfor Prdm1, S1pr1, and Klf2. Multiple TH1 cell–associated transcripts (e.g., Ifng, Il12rb2, Ltb, andGzmb) were identified in IL-2– cells, consistentwith the induction of type 1 immunity by ActA-Lm.Gene set enrichment analysis (GSEA) identifiedenhanced activity of multiple effector signalingpathways in IL-2– cells. Of these, two of the mostsignificant were interferon- and inflammatory-signaling gene sets (Fig. 4B). By contrast, themost significantly enhanced gene sets in IL-2+

cells were those of Myc and the E2F family oftranscription factors (Fig. 4B and fig. S4, A andB). Both are involved in cell-cycle regulationand are suppressed by Blimp1 in germinal centerB cells (20, 21). Mxd1, the transcript for whichwas again one of the most highly enriched in IL-2– cells, antagonizesMyc by competing for bind-ing to a shared dimerization partner, Max (22).Myc expression by T cells correlates directlywith strength of activation (23, 24), implicatinga role for Mxd1 in restraining the actions ofMyc in less strongly activated T cells and con-sistent with expression of IL-2 by more stronglyactivated naïve CD4+ T cells.

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 3 of 13

A Il2 Il21

CNS-35Il2p

naive

IL-2+

IL-2–

B NaiveIL-2–

IL-2+

0

5

10

15

20H3K27me3

ns% In

put

Il2p CNS-35

ns

0.0

0.2

0.4

0.6

0.8Bcl6

% In

put

Il2p CNS-35

* ***

0.0

0.2

0.4

0.6Blimp1

% In

put

Il2p CNS-35

****

*

0.0

0.2

0.4

0.6

0.8Foxo1

% In

put

Il2p CNS-35

***

**

0

2

4

6

8H3K4me3

% In

put

Il2p CNS-35

*

**

Fig. 2. Differential chromatin accessibility and transcription factorbinding at the Il2 locus in IL-2+ and IL-2– Tcells. (A) ATAC-seq wasperformed on nuclei isolated from naïve (GFP–CD44–CD62L+) IL-2.eGFPCD4+ Tcells and FACS-purified CD69+GFP+ (IL-2+) and CD69+GFP– (IL-2–)fractions treated as in Fig. 1C. Chromatin accessibility peaks werevisualized by using Integrated Genome Browser (IGB) and are shownaligned against a VISTA plot of syntenic regions of mouse and humanchromosomes corresponding to Il2-Il21 and IL2-IL21 gene loci, respectively.Data are representative of two experiments. (B) Naïve IL-2.eGFP CD4+

T cells were treated as in Fig. 1C, and the IL-2 promoter region (Il2p) andconserved noncoding sequence 35 kb upstream of the Il2 transcriptionstart site (CNS-35) of CD69+GFP+ (IL-2+) and CD69+GFP– (IL-2–) fractionswere analyzed by quantitative ChIP-PCR for the presence of Bcl6, Blimp1,and Foxo1 binding or H3K4me3 and H3K427me3 histone modifications,normalized to total DNA input. Three technical replicates per group areshown. Data for each region were analyzed separately by one-way analysisof variance (ANOVA). Not significant (ns), P > 0.05; *P < 0.05; **P < 0.01;***P < 0.001; ****P < 0.0001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

Notably, Il2ra, which encodes the inducible,high-affinity component of the IL-2 receptor(IL-2ra or CD25) that is up-regulated on ac-tivated T cells, was enriched in IL-2– cells (Fig.4A). Accordingly, the hallmark IL-2–Stat5 signal-ing gene set was significantly enriched in IL-2–

cells (Fig. 4B). Among a manually curated con-sensus list of 23 gene sets modulated by IL-2signaling (fig. S4C), those up-regulated in responseto IL-2 were enriched in IL-2– cells, many ofwhich include Il2ra (fig. S4, D and E). By con-trast, genes down-regulated in response to IL-2were enriched in IL-2+ cells.Collectively, these findings support a model in

which highly activated naïve T cells up-regulateBcl6, produce IL-2, and are fated to become TFHeffectors. IL-2 producers deliver IL-2 to non-producers, inducing the latter’s up-regulation ofBlimp1 and differentiation into non-TFH effec-tors. To examine the relationship between IL-2production and utilization in vivo and directlyaddress the fate of IL-2 producers and non-producers, we generated a second transgenicIL-2 reporter mouse line with features com-plementary to those of the IL-2.eGFPmice (Fig.4C). IL-2.BAC-in Thy1.1 (2BiT) reportermicewereengineered to express high amounts of the sur-face molecule Thy1.1 under control of the Il2gene locus to facilitate intracellular costainingby flow cytometry and enable the in vivo dele-tion of IL-2–producing cells (25). As with T cellsfrom IL-2.eGFP mice, activated (CD69+) 2BiTT cells rapidly bifurcated into Thy1.1+ (IL-2+) andThy1.1– (IL-2–) fractions (Fig. 4D and fig. S5).To determine whether IL-2 production and sig-naling segregate in antigen-activated naïve CD4+

T cells, 2BiT mice were infected with ActA-Lmand analyzed for the expression of IL-2 versusintracellular phospho-Stat5 (p-Stat5) at the peakof IL-2 expression (Fig. 4E). Reciprocal IL-2 ex-pression and IL-2 signaling were observed; Thy1.1(IL-2) was almost exclusively expressed by p-Stat5–

CD4+T cells (Fig. 4E),whereas p-Stat5was limitedto Thy1.1– cells. Consistent with gene expressionresults (Fig. 4A), nearly all Thy1.1+ cells wereCD25– at this timepoint,whereasnearly all p-Stat5+

cells were CD25+. Thus, IL-2 signals predominantlyin a paracrine, not autocrine, manner (26). More-over, IL-2 producers are initially resistant to IL-2signaling, in accord with their lack of CD25up-regulation.Most endogenous p-Stat5+ CD4+ T cells im-

mediately after infection are Foxp3+ regulatoryT (Treg) cells (26), owing to their constitutive ex-pression of CD25 and relative abundance com-pared with that of naïve clonal precursors. Toexamine IL-2–induced Stat5 signaling in naïvepathogen-specific non-Treg cells, naïve CD45.2

+

2BiT-SMARTA T cells were transferred intoCD45.1+ WT mice infected with ActA-Lm ex-pressing the gp66 peptide (ActA-Lm-OVA-gp66)(Fig. 4F). Analysis of transferred 2BiT-SMARTA(clonotypic) and endogenous CD4+ T cells showedthat the majority of endogenous p-Stat5+ cellswere Foxp3+, whereas p-Stat5+ clonotypic T cellswere Foxp3–. Thus, the paracrine model of IL-2signaling applies to both “bystander” Treg cells

as well as naïve CD4+ T cells responding toinfection.

TFH cells are derived from IL-2 producers

To examine the fate of antigen-activated IL-2–producing and –nonproducing T cells in vivo,2BiT mice were treated with a depleting anti-Thy1.1 or nondepleting control antibody (25)immediately before infection with ActA-Lm co-expressing OVA and the antigenic peptides gp66,2W1S, Cbir1, or FliC (Fig. 5 and fig. S6). MHCII-tetramer analysis of endogenous CD4+ T cellsspecific for each of these peptides showed thatIL-2 (Thy1.1) expression was restricted to CXCR5+

cells and that the depletion of IL-2–expressingcells preferentially eliminated TFH cells and sparednon-TFH (TH1) effectors, as defined by expressionof CXCR5 and PD-1 (Fig. 5A and fig. S6) orCXCR5 and Bcl6 (Fig. 5B). Notably, the numberof non-TFH effectors was not compromised by thedepletion of IL-2 producers (Fig. 5C). This sug-gested that IL-2 was not required for the clonalexpansionof non-TFHeffectors.However, this likelyreflects the discordant kinetics of IL-2 secretionrelative to reporter expression and antibody-mediated cell depletion, because IL-2 reporterexpression and Stat5 phosphorylation were onlypartially decreased at the peak of IL-2 expression(fig. S7). Thus, TFH effector cells developed fromIL-2–expressing precursors, whereas non-TFH

effectors did not. Accordingly, IL-2 was a reliablemarker with which to distinguish precursorsfated to become TFH or non-TFH effector cells.Although surface markers define TFH cells

capable of providing B cell help, TFH cell functionis predicated on a subset of TFH cells that localizeto the germinal center after productive interac-tions with cognate B cells. Referred to as germinalcenter (GC-TFH) T cells, these cells express highamounts of PD-1, Bcl6, and CXCR5 and supportthe germinal center response and productionof high-affinity, class-switched antibodies (12).To examine the effects of depletion of IL-2–expressing precursors on the development andfunction of this TFH cell subset, we characterizedthe effects of anti-Thy1.1 depletion on antibodyresponses and the generation of GC-TFH cells.2BiTmice were immunized with heat-killed Lm(HKLm) (Fig. 5, D to F), because infection withlive Lmdoes not induce good antibody responses(27). Anti-Thy1.1 depletion of IL-2–producingT cells reduced the production of Lm-specificimmunoglobulin G (IgG) by more than 90% com-pared with treatment with an isotype controlantibody (Fig. 5D). Similarly, anti-Thy1.1 treat-ment of 2BiTmice immunized with OVA undertype 1 conditionsmarkedly impaired the anti-OVAIgG response, in association with the depletionof endogenous OVA-specific TFH cells and reduc-tion of germinal center B cells (fig. S8).

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 4 of 13

0 103 104 105 106

0

102

103

104

105

106

90.7

31.4

0 103 104 105 106

93.7

30.8

0 103 104 105 106

87.6

29.3

0 103 104 105 106

91.9

47.8

IL-2.eGFP

CD

69

Th1IL-12

Th2IL-4

Th17IL-6 + TGFβ

Th00

A

0 103 104 105 106

Th0Th1Th2Th17

0

20

40

60

80

100

% C

D69

+

0

5

10

15

20

25

GF

P g

MF

I (x1

03 )

0

20

40

60

% G

FP

+

0

5

10

15

20

25

CD

69 g

MF

I (x1

03 )

0.000

0.005

0.010

0.015Bcl6

Th0 Th1 Th2 Th17N0.0

0.5

1.0

1.5Il2

2^(C

TB

2M-C

TG

OI)

IL-2–

IL-2+

Naive

Th0 Th1 Th2 Th17N0.0000

0.0005

0.0010

0.0015Prdm1

Th0 Th1 Th2 Th17N

IL-2.eGFP gMFI

Th 0 1 2 17 0 1 2 17 0 1 2 17 0 1 2 17

B

ThTh Th

Fig. 3. Bcl6 and IL-2 cosegregate early in each Teffector cell developmental program. (A) Naïve(GFP–CD44–CD62L+CD69–CD25–) IL-2.eGFP CD4+ T cells were stimulated in vitro underTH0, TH1, TH2, and TH17 cell conditions for 20 hours and examined by flow cytometry for CD69and IL-2.eGFP expression. Data are representative of two experiments. Flow plots depict cellnumber–controlled concatenated averages of three samples per group. Error bars depict SD. gMFI,geometric mean fluorescence intensity. (B) Experiment performed as in (A), with CD69+ IL-2.eGFP+

and IL-2.eGFP– CD4+ T cells sorted 20 hours after activation. RNA was isolated and analyzed byqPCR for expression of Il2, Bcl6, and Prdm1. Three technical replicates per condition are shown. N, naïveT cells. Error bars depict SEM. Data for (A) and (B) are representative of two experiments each.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 5 of 13

C Il21 Il2 Tenr

4321

BAC

20 kbThy1.1 SV40pA

1 kb

4321

2BiTTransgene

INTERFERON_ALPHA_RESPONSETNFA_SIGNALING_VIA_NFKBINTERFERON_GAMMA_RESPONSETGF_BETA_SIGNALINGIL6_JAK_STAT3_SIGNALINGP53_PATHWAYHYPOXIAEPITHELIAL_MESENCHYMAL_TRANSITIOAPOPTOSISINFLAMMATORY_RESPONSEALLOGRAFT_REJECTIONUV_RESPONSE_DNREACTIVE_OXIGEN_SPECIES_PATHWAYUV_RESPONSE_UPMYOGENESISANGIOGENESISAPICAL_SURFACEIL2_STAT5_SIGNALINGCOMPLEMENTHEDGEHOG_SIGNALINGESTROGEN_RESPONSE_EARLYNOTCH_SIGNALINGESTROGEN_RESPONSE_LATEANDROGEN_RESPONSEKRAS_SIGNALING_UPAPICAL_JUNCTIONKRAS_SIGNALING_DNPANCREAS_BETA_CELLSCHOLESTEROL_HOMEOSTASISPI3K_AKT_MTOR_SIGNALINGHEME_METABOLISMCOAGULATIONWNT_BETA_CATENIN_SIGNALINGXENOBIOTIC_METABOLISMADIPOGENESISFATTY_ACID_METABOLISMBILE_ACID_METABOLISMGLYCOLYSISOXIDATIVE_PHOSPHORYLATIONPROTEIN_SECRETIONHALLMARK_SPERMATOGENESISPEROXISOMEMITOTIC_SPINDLEDNA_REPAIRUNFOLDED_PROTEIN_RESPONSEMTORC1_SIGNALINGMYC_TARGETS_V2G2M_CHECKPOINTMYC_TARGETS_V1E2F_TARGETS

123456789

1011121314151617181920212223242526272829303132333435363738394041424344454647484950

–0.2

–0.1

0.0

0.1

0.2

0.3

2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

B

E

Mea

n Lo

g 2 (FPK

M (G

FP– )

FPK

M (G

FP+ )

) for a

ll ge

nes

with

in s

et

IL-2–

IL-2+

log2FoldChange

-log 10

(pad

j)

AMxd1

Ltb

Gadd45g

GzmbKlf2

Il2ra

Socs3

Icam1

Ifng

Fas

S1pr1

Cxcr4

Prdm1

Il12rb2

Zbtb32

Bcl2

Il1r2

Cxcr3Cd40lg

Bcl6

Cd2

0

5

10

15

20

25

30

35

-3 -2 -1 0 1 2 3

Il2x=3.1y=199

D

CD

690 102 103 104 105

0

102

103

104

105 82.2

IL-2

.Thy

1.1

0 102 103 104 105

0

102

103

104

105

44.2

52.4

0 102 103 104 105

0

102

103

104

105

82.3

CD

44 Fol

d C

hang

e

Il2 mRNA0

5

10

15

20 Thy1.1+

Thy1.1–

F

IL-2

.Thy

1.1

0 102 103 104 105

5.5 4.6

1.288.70 102 103 104 105

6.4 3.9

0.889.00 102 103 104 105

9.6 2.1

0.487.90 102 103 104 105

0

103

104

105 10.3 1.2

0.288.3Fox

p3

Clo

noty

pic

End

ogen

ous

ActA-Lm-gp66

CD45.1 WT

24 hrs

CD45.2 SMARTA2BiT CD4 T cells

CD45.1 WT

4 - 24 hrs

7.3 0.2

2.789.8

23.8 0.9

6.568.8

10.9 1.3

5.682.2

Naive 8 hours 16 hours

0

102

103

104

105 3.0 0.0

0.396.6

24 hours

0 102 103 104 105 0 102 103 104 105 0 102 103 104 1050 102 103 104 105

0 102 103 104 105 0 102 103 104 105 0 102 103 104 1050 102 103 104 105

0.3 0.3

2.497.0

0.4 0.2

6.593.0

1.1 1.3

4.293.40

103

104

105 1.3 0.2

0.298.3Fox

p3

0

103

104

105

106

1.4 0.1

1.497.00 103 104 105 10

0 103 104 105 10

IL-2

.Thy

1.1

p-Stat5

0

103

104

105

106

1.4 0.3

11.087.30 103 104 105 10

IL-2

.Thy

1.1

CD25

0

103

104

105

106

0.5 1.3

10.587.70 103 104 105 106

p-S

tat5

CD25

ActA-Lm

2BiT

24 hrs

Sac

CD45.2 SMARTAIL-2.eGFP CD4 T cells

ActA-Lm-gp66

CD45.1 WT

24 hrs 20-24 hrs

GFP+

Sac

Sac

CD62L CD4 CD4

p-Stat5

p-Stat5

p-Stat5

GFP–

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49

<0.

001

<0.

005

<0.

01<

0.05

>0.

05<

0.05

<0.

01<

0.00

5<

0.00

1

p-Values

IL2_STAT5_SIGNALING

MYC_TARGETS_V1

E2F_TARGETS

Fig. 4. IL-2+ Tcells activate IL-2–Tcells via paracrine IL-2 signalingto drive differential gene expression in vivo. (A) Sorted naïve(GFP–CD44–CD62L+) IL-2.eGFP CD45.2+ SMARTA CD4+ T cells weretransferred into CD45.1+ WTmice infected with ActA-Lm-gp66 24 hoursbefore transfer.Total RNA was isolated by FACS-purified CD45.2+ CD69+GFP+

(IL-2+) and CD69+GFP– (IL-2–) CD4+ T cells 20 to 24 hours after transferand analyzed by RNA-seq. Data depict three biological replicates percondition from three separate experiments. Sac, sacrificed. (B) HallmarkGSEA of IL-2+ and IL-2– T cells from (A). For each pathway, mean and 95%confidence intervals are plotted and then color coded to indicate falsediscovery rate–corrected P values. FPKM, fragments per kilobase million.(C) Schematic of targeting strategy to generate IL-2.BAC-in Thy1.1 (2BiT)transgenic reporter mice. BAC, bacterial artificial chromosome. (D) Sortednaïve (Thy1.1–CD44–CD62L+) 2BiT CD4+ T cells were stimulated in vitro

with soluble anti-CD3 (5 mg/ml) and irradiated CD4-depleted feeder cells for24 hours and then examined by flow cytometry for expression of CD69and Thy1.1. RNA isolated from CD69+Thy1.1+ (IL-2+) and CD69+Thy1.1– (IL-2–)CD4+ T cells was analyzed by qPCR for expression of Il2 mRNA. Error barsrepresent SEM of three technical replicates per sample. Data are represent-ative of two experiments. (E) 2BiTmice were infected with ActA-Lm.After 18 hours, mice were sacrificed and splenic CD4+ T cells were analyzedby flow cytometry for the expression of IL-2.Thy1.1, CD25, and tyrosinephosphorylation of Stat5 (p-Stat5). Data are representative of twoexperiments. (F) Congenic CD45.1+ WTmice were infected with ActA-Lm-gp66. After 24 hours, naïve CD45.2+ SMARTA 2BiT CD4+ T cells weretransferred into infected CD45.1+ recipients. Mice were sacrificed at theindicated times, and splenic CD4+ T cells were analyzed for expression ofThy1.1, Foxp3, and p-Stat5. Data are representative of two experiments.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 6 of 13

F

3.4

3.0

GL7

Fas

B220+ IgDlo

Isotype Ctrl anti-Thy1.1

GFP CD45.1 IgD Ki67

0.00

0.01

0.02

0.03

0.04

0.05

T C

ells

/GC

(m

m2 )

Per

Gen

otyp

e

0.00

0.01

0.02

0.03

0.04

0.05

T C

ells

/GC

(m

m2 )

Tot

al

Control Treated

WT2BiT

ns

ns

****

D

11.6 31.1

13.943.4

11.3 7.0

9.172.6

B

Bcl

6

CXCR5

0 6 12 18 24 30102

103

104

105

106

Ant

i-Lm

IgG

(ug/

mL)

Days post HKLm

HKLm

2BiT

Isotypeor

Anti-Thy1.1

89.3

Isot

ype

Ctr

lA

nti-T

hy1.

1

1.0

0.4

37.2

7.4

7.9 19.4

26.746.0

2.2 1.5

11.584.8Ly6C

IAb-gp66

PD

-1

CXCR5IL

-2.T

hy1.

1CXCR5

57.5

ActA-Lm-gp662BiT

Isotypeor

Anti-Thy1.1Day 3

HKLm

Tcrb -/-

Isotypeor

anti-Thy1.1

WTCD45.2

2BiTCD45.1

E WT 2BiT

72.4

24.8

47.5

51.7A

nti-T

hy1.

1Is

otyp

e ct

rlC

D45

.1

PD

-1 21.5

4.7

40.8

15.3

40.5

14.8

36.2

15.5

CXCR5CD45.2

38.738.7

64.343.0

Day 5

CControl Anti-Thy1.1

Tota

l tet

+ T

hy1.

1+

(x 1

03 )

Tota

l tet

+ C

D4+

(x 1

03 )

% T

et+ C

D4+

%

Tet+

Thy

1.1+

0

20

40

60

80

100 ****

****

0.1

1

10

100ns *

non-Tfh Tfh0

10

20

30

40

50**** ****

non-Tfh Tfh0.1

1

10

100 ***

***

Ctrl αThy1.1

WT Tfh2BiT Tfh

WT GC-Tfh2BiT GC-Tfh

Total WT Tfh

Total 2BiT Tfh0

10

20

30

40

50

60

70

% C

XC

R5+

-PD

-1+ ns

ns

0

10

20

30

40

50

% T

fh v

s G

C-T

fh

0

10

20

30

40

50

2.9x3.2x

1.9x1.6x

******

**

***

ns

***

A

ns

Ctrl αThy1.1

% G

L7+-F

as+

0

1

2

3

4

5

Fig. 5. IL-2 producers are precursors of TFH cells. (A to C) 2BiTmicewere injected with 250 mg of anti-Thy1.1 or isotype control (Ctrl)monoclonal antibody (mAb), then infected 1 day later with ActA–Lm-gp66.Endogenous CD4+ T cells specific for IAb-gp66 were enriched from lymphnodes and spleens 3 days after infection by using tetramer-basedmagnetic sorting and analyzed by flow cytometry for IAb-gp66 tetramerbinding (tet+) and expression of Ly6C, CXCR5, IL-2.Thy1.1, and PD-1 (A)or Bcl6 (B). Flow plots depict cell number–controlled concatenatedaverages of all samples within a group. Data for (A) and (B) arerepresentative of two experiments each. (C) Data from the experimentsdepicted in (A) and fig. S6 were analyzed by two-way ANOVA. A total ofeight control and eight treatment animals from two separate experimentsare shown. (D) 2BiTmice were injected with 250 mg of anti-Thy1.1 orisotype control mAb and immunized with 2 × 1010 colony-forming units(CFU) of HKLm. Mice were bled every 6 days for 24 days, and serum anti-Lm IgG was measured by enzyme-linked immunosorbent assay (ELISA).n = 7 mice per group. Data are representative of two experiments.(E) Magnetically enriched WT CD45.1+ and 2BiT CD45.2+ CD4+ Tcells were

transferred into Tcrb–/– mice. After 24 hours, mice were immunized with2 × 1010 CFU HKLm and injected with 250 mg anti-Thy1.1 or isotype controlmAb. Mice were sacrificed 5 days after immunization, and splenic CD4+

T cells were analyzed by flow cytometry for expression of CD44, CD45.1,CD45.2, PD-1, and CXCR5. Results were analyzed by two-way ANOVA.n = 3 mice per group. Data are representative of three experiments.(F) CD4+ Tcells magnetically enriched from WT CAG-eGFP (CD45.2) miceand CD45.1+ 2BiTmice were adoptively transferred into TCRb-deficientrecipients. After 24 hours, the mice were immunized with 2 × 1010 CFUHKLm and injected with 250 mg of anti-Thy1.1 or isotype control mAb. Twoweeks after immunization, spleens were collected and analyzed byconfocal microscopy for the expression of GFP (WT), CD45.1 (2BiT), Ki67,and IgD. Quantitation of WT (GFP+), 2BiT (CD45.1+), and total T cellnumbers in germinal centers (GCs) was performed by computer-assistedcounting. Splenic B cells were analyzed by flow cytometry for theexpression of IgD, B220, GL7, and Fas in an IgDlo B cell gate. n = 3 mice pergroup. Data are representative of three experiments. ns, P > 0.05; *P <0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

To examine the effects on GC-TFH cell differ-entiation, we transferred equivalent numbers ofcongenically marked 2BiT and GFP-expressingWT CD4+ T cells into T cell–deficient (Tcrb–/–)mice, which were immunized withHKLm (Fig. 5,E and F). Anti-Thy1.1 treatment selectively de-pleted CD45.1+ 2BiT PD-1+CXCR5+ T cells andhad no significant effect on CD45.2+ GFP+ con-trol T cells (Fig. 5E). The magnitude of reduc-tion was highest among PD-1hiCXCR5hi cells.Immunohistology revealed that Thy1.1 depletiondramatically reduced the number of 2BiT CD4+

T cells found within germinal centers, with acompensatory increase in the numbers of WTGFP+ T cells, resulting in no change in totalnumbers of germinal center T cells or B cells (Fig.5F). These data establish that functional TFHeffectors that populate germinal centers andprovidehelp for class-switchedantibody responsesin response to type 1 pathogens develop fromIL-2–producing precursors.To extend these findings, we determined

whether the in vivo depletion of IL-2+ cells alsoselectively targeted TFH cells under conditions oftype 2 (TH2) and type 3 (TH17) immune induction(Fig. 6). Anti-Thy1.1 treatment of 2BiT mice im-munized with OVA by using the TH2-inducingadjuvant alum resulted in the specific depletionof OVA-specific TFH cells and ablation of germinalcenter B cell and anti-OVA antibody responses,while sparing OVA-specific non-TFH effectors(Fig. 6A). Similarly, anti-Thy1.1 depletion of 2BiTmice challenged with the TH17 cell–inducingenteric pathogen Citrobacter rodentium (28) re-sulted in a loss of bacterial clearance, the kineticsof which were characteristic of an impaired anti-Citrobacter antibody response (29, 30). This wasassociated with depletion of TFH cells specificfor the immunodominant Citrobacter antigenintimin, as well as a markedly impaired ger-minal center B cell response (Fig. 6B). There wasno significant decrease in IL-17A– or interferon-g(IFNg)–producing cells in infected spleens (Fig.6C), both of which are characteristic of the T cellresponse against Citrobacter (28). However, amodest, but significant, decrease in TH17 cells,but not TH1 cells, was observed in themesentericlymph nodes of mice depleted of IL-2 producers(Fig. 6D). These findings indicate that, in thecontext of type 3 responses, there is some overlapin the developmental fate of IL-2–producing pre-cursors of TFH and TH17 effector cells. They alsoestablish that IL-2+ T cells are precursors for TFHcells in the context of type1, type 2, and type 3effector responses.

IL-2 production and TFH celldifferentiation correlate with TCRsignal strength

The developmental divergence of TFH andnon-TFHcells is influenced by a combination of cell-intrinsicfactors, including TCR affinity, and cell-extrinsicfactors, such as antigen availability, strength ofcostimulation, and cytokine milieu. Given its di-rect correlationwith TFH–non-TFH cell fate deter-mination, we examined whether IL-2 expressionshared similar mechanistic underpinnings. To

examine the relationship between antigen dose,IL-2 expression, and TFH–non-TFH cell specifica-tion on T cells of uniform TCR affinity, naïveCD45.2+ IL-2.eGFP-SMARTAT cells were transfer-red into WT CD45.1+ recipients, which were in-fectedwith various doses of ActA-Lm expressingthe gp66 peptide (ActA-Lm-OVA-gp66; conditionLm-gp66). Nonspecific effects of Lm-inducedinflammatory signals were excluded by keepingthe total dose of ActA-Lm constant via coinfec-tion with ActA-Lm expressing an irrelevantspecificity (ActA-Lm-OVA-2W1S; condition Lm-2W1S) (Fig. 7, A andB). Splenic CD45.2+ cells wereanalyzed for CD69 and GFP (IL-2) expressionaround the peak of IL-2 expression (Fig. 7A). Thefrequencies and absolute numbers of endogenousgp66-specific TFH and non-TFH cells were quan-tified near the peak of the effector T cell response(Fig. 7B). The expression of both CD69 and GFPcorrelated tightly with pathogen-expressed anti-gendose, as did themagnitude of clonal expansionand reciprocal TFH versus non-TFH differentiation.There was a similar correlation over a broaderdose range of ActA-Lm-OVA-gp66 administeredalone (fig. S9). Thus, the frequencies of CD69+

and IL-2+ cells correlated positively with Lm-gp66dose, as did the generation of TFH cells. Thisindicates that antigen dose—and consequentlyTCR signal strength—is amajor determinant ofthe fraction of clonal precursors that expressIL-2 and are fated to become TFH cells.The relative development of TFH and non-TFH

effectors is also influenced by TCR affinity, whichis invariant on individual T cell clones but variesbetween different clones within the repertoire(2). In agreement with the strong correlationbetween differential IL-2 expression and TFH–non-TFH cell bifurcation, we found that two clonalpopulations of the same precursor frequency,which had differing TCR specificities (OVA versusgp66) and affinities, produced significantly dif-ferent frequencies of IL-2–expressing T cells inresponse to the same antigen dose despite nodifference in the frequency of cells expressingCD69 (Fig. 7C). Similarly, the relative frequenciesof endogenous TFH and non-TFH cells generatedby two TCR specificities of differing TCR af-finities diverged in response to the same antigendose (Fig. 7D). In accord with these results, TFHcells were characterized by a significantly greatertwo-dimensional (2D) TCR affinity than non-TFHcells that developed in a polyclonal T cell re-sponse to the same antigen (gp66) (Fig. 7E). To-getherwith our antigen-dose results, these findingssupport a deterministic function of TCR signalstrength in driving TFH versus non-TFH celldevelopment (2), such that higher TCR signalingfavors TFH cell development.To calculate 2D affinity values, normalized

adhesion-bond measurements must be adjustedto control for TCR density (31). Although equiv-alent in terms of cell size, gp66-specific TFH cellsexpressed more TCRmolecules per cell than non-TFH cells (Fig. 7E). Similar results were found forOVA-specific and total TFH cells. Accordingly, thedifference in the normalized adhesion bond wasaccounted for by differences in both TCR affinity

and TCR number per cell. Thus, we examined theinfluence of variation in TCR number on T cellactivation and IL-2 production (fig. S10A). Asexpected, there was no difference in 2D TCRaffinity of CD69+IL-2+ SMARTA IL-2.eGFP T cellscompared with that of CD69+IL-2– cells. However,when naïve SMARTA IL-2.eGFP T cells werestimulated with limiting concentrations of gp66peptide, only cells expressing the highest amountsof the TCR component Va2 up-regulated CD69and IL-2 (Fig. 7F). Moreover, when naïve T cellswere sorted on the basis of high or low TCRbexpression and stimulated with a range of anti-CD3 concentrations, cells with higher TCR num-bers showed higher CD69 expression across thefull range of anti-CD3 concentrations and ex-pressed increased IL-2 (fig. S10B). Thus, in addi-tion to intrinsic TCR affinity differences betweenT cell clones, variation in TCR number within aclonal population may influence the probabilitythat a given cell will exceed a threshold for acti-vation and expression of IL-2 and, therefore, TFHversus non-TFH cell differentiation.The expression of CD69 by activated naïve

T cells has been shown to correlate linearlywith the expression of Myc and Nur77, provid-ing an indicator of graded TCR signal intensity(23, 32). Yet, on the basis of the observed lim-itation of IL-2 expression to a subset of thehighest CD69 expressors (Figs. 3A and 7A), ourfindings suggested that only those T cells thatexceeded a minimum TCR signaling intensityproduced IL-2, linking TFH–non-TFH cell bifurca-tion to a threshold of TCR signaling.To explore this further, the relative expression

of CD69 and IL-2 by naïve T cells was assessedunder conditions in vitro for which only theintensity of TCR stimulationwas varied (Fig. 8A).The percentage of cells expressing CD69 and IL-2correlated positively with the dose of anti-CD3,as did expression amounts of CD69 and IL-2,consistent with our in vivo antigen-dose experi-ments (Fig. 7A and fig. S8A). However, althoughthe distribution and mean expression of CD69varied with the intensity of TCR stimulation,CD69 expression among IL-2+ cells was con-stant, and only cells that exceeded an invariant,high magnitude of CD69 expressed IL-2. Asimilar effect was seen with the induction ofinducible costimulator (ICOS) expression (fig.S11A), which is a functional marker of TFH celldifferentiation (33). ICOS expression was highestamong IL-2+ cells across a range of stimulationconditions. Although the mean expression ofICOS varied with the intensity of TCR stimula-tion among IL-2– cells, its expression by IL-2+

cells was constant. The expression of ICOS ligandby antigen-presenting cells was not requiredfor the induction of IL-2 (fig. S11B). Thus, underconditions for which only TCR signaling strengthis varied, there is a minimum threshold forIL-2 expression and, by extension, TFH celldifferentiation.The expression of Myc, which has been di-

rectly correlated with the number of cell di-visions that T cells are fated to undergo (24), wastightly correlatedwith CD69 expression and thus

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 7 of 13

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

TCR signal intensity and IL-2 expression (Fig.8B). The early expression of E2F familymembers,which are associated with cell-cycle entry, waslimited to cells that exceeded a threshold for IL-2expression. This was consistent with the GSEAdata (Fig. 4B and fig. S4), which indicated thatE2F family targets are strongly enriched in IL-2+

cells compared with IL-2– cells. This suggests

that IL-2+ cells enter the cell cycle more rapidlythan IL-2– cells and are likely to undergo agreater number of cell divisions before exitingthe cell cycle. Accordingly, TFH effectors demon-strated an increase in the average number of celldivisions relative to non-TFH effectors in vivo(Fig. 8C). Thus, although IL-2 has traditionallybeen viewed as a T cell growth factor, precursors

of TFH cells, which do not respond to IL-2 despiteproducing it, appear programmed for earlier cell-cycle entry and a greater number of cell divisionsthan precursors of non-TFH effectors, which dorespond to IL-2.Our results support a model whereby cell-

intrinsic and -extrinsic variables that influenceTCR signal strength contribute to a threshold

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 8 of 13

A

OVA/Alum

12 days

Isotypeor

Anti-Thy1.1

2BiT89.3

12.0

81.6

5.6

0.6

0.3 1.4

12.7

IAb-OVA

CD

44

CXCR5

PD

-1

Fas

GL7

Isot

ype

Ctr

lA

nti-T

hy1.

1

B220+ IgDlo

0.0

0.2

0.4

0.6

0.8

% IA

b -O

VA

+

0

5

10

15

*

% G

C B

Cel

l

0

5

10

15

20

25

% T

fhns *

0.0

0.5

1.0

1.5

IAb -

OV

A+ (

x107

)

2

4

6

8

10

00.0

0.5

1.0

1.5

2.0

2.5

**

0.0

0.5

1.0

1.5

GC

B c

ells

(x1

06)

Tfh

(x1

06)

Non

-Tfh

(x1

06)

ns

ns

*

B

C. rodentium

5-28 days

Isotypeor

Anti-Thy1.1

2BiT

0.05

0.01

20.2

9.4

CXCR5

Isot

ype

Ctr

lA

nti-T

hy1.

1

PD

-1IAb-Int884C

CD

44

*0

1

2

3

4

5

Tfh

(x1

04 )

***

0

10

20

30

40

50

GC

B c

ell (

x104 )

Isot

ype

Ctr

lA

nti-T

hy1.

1

7 215000

1000

2000

4000

3000

Day:

Control

Anti-Thy1.1

0 7 14 21 28Lum

ines

cenc

e (c

ount

s/s)

102

103

104

105

106

101

C

C. rodentium

14 days

Isotypeor

Anti-Thy1.1

2BiT

ns

0

5

10

15

IFN

γ+ (

x105 )

0

10

20

30

40

50

% IF

Nγ+

ns

0

5

10

15

IL-1

7A+

(x1

04 )

0

2

4

6

% IL

-17A

+

nsns

Isot

ype

Ctr

lA

nti-T

hy1.

1

36.3 0.6

4.059.135.7 1.0

3.060.2

IL-17A

IFN

γ

IFN

γ

Spleen

0

100

200

300

Ant

i-OV

A Ig

G(n

g/m

l)

***

ns

Isot

ype

Ctr

lA

nti-T

hy1.

1

13.3 0.6

6.479.7

13.4 0.7

5.280.7

IL-17A

MLND

0

5

10

15

IL-1

7A+

(x1

04 )

*

0

2

4

6

8

10

% IL

-17A

+ ns

0

5

10

15

20

% IF

Nγ+

IFN

γ+ (

x104 ) ns

0

5

10

15

20

Fig. 6. IL-2 producers are fated to become TFH cells in type 2 and type3 immune responses. (A) 2BiTmice were injected with 250 mg ofanti-Thy1.1 or isotype control. After 24 hours, they were immunizedwith OVA emulsified in alum. Mice were bled and sacrificed on day 12.Splenic IAb-OVA tetramer-specific CD4+ T cells were analyzed by flowcytometry for the expression of CD44, PD-1, and CXCR5. Splenic B cellswere analyzed for the expression of B220, IgD, GL7, and Fas. Serumanti-OVA IgG was measured by ELISA. n = 5 or 6 mice per group. (B) 2BiTmice were injected with 250 mg of anti-Thy1.1 or isotype control. After24 hours, they were orally gavaged with 1 × 109 to 2 × 109 CFU C.rodentium strain DBS100 (ATCC 51459) or the bioluminescent ICC180derivative. Whole-body bioluminescence was tracked and quantified afterinfection. Splenic IAb-Int884C tetramer-specific CD4+ T cells harvested

on day 14 were analyzed by flow cytometry for the expression of CD44,PD-1, and CXCR5. Splenic B cells were analyzed for the expressionof B220, IgD, GL7, and Fas. n = 5 or 6 mice per group. Flow plotsdepict cell number–controlled concatenated averages. Data arerepresentative of two experiments. (C and D) Splenic (C) andmesenteric lymph node (MLN) (D) CD4+ T cells harvested from micetreated as in (B) were isolated, restimulated with phorbol myristateacetate (PMA) and ionomycin and then analyzed by flow cytometryfor the expression of CD44, Foxp3, IFNg, and IL-17A. Flow plotsand bar graphs are gated on CD4+CD44+Foxp3– cells. Flow plotsdepict cell number–controlled concatenated averages. n = 5 or 6 miceper group. Data were analyzed by using Student’s t tests. ns, P > 0.05;*P < 0.05; **P < 0.01; ***P < 0.001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 9 of 13

A

0 103 104 1050 103 104 105

0

102

103

104

105

0 103 104 105 0 103 104 105 0 103 104 1050 103 104 105

85.953.1

13.63.4

75.743.3

90.357.3

90.259.5

59.924.2

IL-2.eGFP

CD

690Lm-gp66:

Lm-2W1S:5x105 5x1055x106 4.95x107 5x1074.5x107

5x107 4.95x107 4.5x107 0.05x107 00.5x107

a b c d e f

f0

20

40

60

80

100

0

20

40

60

80

100

% C

D69

+

SmartaEndogenous

a b c d e0

20

40

60

80

100

% C

D69

+IL

-2.e

GF

P+

CD45.2+ SMARTA

a b c d e f

CD45.1

24 hrsCD45.2+SMARTA IL-2.eGFP CD45.2 + ActA-Lm-OVA-gp66/ActA-Lm-OVA-2W1S

Condition:

Condition Condition

B5 Days

ActA-Lm-OVA-gp66/ActA-Lm-OVA-2W1S CD44+ IAb-gp66+

0 103 104 105 106

15.0

0 103 104 105 106

18.9

0 103 104 105 106

41.7

0 103 104 105 106

43.9

0 103 104 105 106

43.6

0 103 104 105 106

0

103

104

105

106

N.D.

CXCR5

PD

-1

a b c d e f

0 103 104 105 106

0

103

104

105

36.1 <0.01

<0.010 103 104 105 106

36.7 0.01

<0.010 103 104 105 106

33.2 0.1

<0.010 103 104 105 106

28.9 0.2

<0.010 103 104 105 106

33.1 0.2

<0.010 103 104 105 106

36.0 0.2

<0.01

IAb-gp66

CD

44

106

0

10

20

30

40

50

% T

fh

a b c d e fCondition

0

2

4

6

8

Tota

l Tfh

(x

105 )

a b c d e fCondition

0.0

0.2

0.4

0.6

0.8 CD4+CD44+

CD4–

% IA

b -gp

66+

ns

****

*

**** ****

****

a b c d e fCondition

0.0

0.5

1.0

1.5

2.0

Tota

l IA

b -gp

66+ (

x 10

6 )

a b c d e fCondition

E

****

****

CXCR5

PD

-1

CXCR5

PD

-1

0 103 104 105 106

0

103

104

105

106

19.3

0 103 104 105 106

0

103

104

105

106

5.0

CD

44C

D44

0 103 104 105 106

0

103

104

105

106

0.9

0 103 104 105 106

0

103

104

105

106

0.1

IAb-OVA

IAb-gp66

0.0

0.2

0.4

0.6

0.8

1.0

% T

etra

mer

+

0

10

20

30

% T

fh

5 DaysActA-Lm-OVA-gp66

IAb-OVA

IAb-gp66

0 103 104 105 106

0

103

104

105

106

56

44

CD4

Vβ5

CD45.2+

D

0

20

40

60

80

100 ns

% C

D69

+

0

20

40

60 ****

% G

FP

+(o

f CD

69+

)

IL-2.eGFP

CD

69

0

103

104

105

106

90.09.9

OT-II

0 103 104 105 106

0

103

104

105

106

90.554.9

SMARTA

C

F

0 103 104 105 106

0

103

104

105

106

0.9

IAb-gp66

CD

44

0 103 104 105 106

72.4

20.1

CXCR5

PD

-1

0 103 104 105 106

63.5

28.6

0

103

104

105

106

5 DaysActA-Lm-OVA-gp66

-2.5

-2.0

-1.5

-1.0

-0.5

Nor

mal

ized

adhe

sion

bon

d ***

-4.5

-4.0

-3.5

-3.0

-2.5

Affi

nity

(A

c K

a µm

4 )

**

0

5

10

15

20

25

TC

Rs/

Cel

l (x

103 ) **

0 102 103 104 105

0

102

103

104

105

3.4

1ug/mL gp66

CD69

Vα2

0 102 103 104 105

0

102

103

104

105

CD69+/GFP–CD69+/GFP+

CD69–

IL-2.eGFP

Vα2

20.9(CD69+)

2D Affinity Measurement SMARTA IL-2.eGFP CD4+ T cells

****

-5.0

-4.5

-4.0

-3.5

-3.0

-2.5

Affi

nity

(A

c K

a µm

4 )

OT-II IL-2.eGFP+

SMARTA IL-2.eGFP+

ActA-Lm-OVA-gp66

CD45.1

24 hrs

Fig. 7. IL-2 production and TFH differentiation correlate with TCR signalstrength. (A) WTCD45.1+ recipient mice were infected with ActA-Lm-OVA-gp66and/or ActA-Lm-OVA-2WIS at the indicated doses. After 24 hours, 106 naïve(GFP–CD44–CD62L+) SMARTA IL-2.eGFP CD4+ Tcells were adoptively transferredinto infected hosts. Splenic CD4+ Tcells were harvested 15 hours after transfer andanalyzed forexpressionof CD69and IL-2.eGFPby flowcytometry.Values in the largerboxesof flowcytometricplots representpercentagesofCD69+cells, andvalues in thesmaller boxes represent percentages IL-2.eGFP+ cells within the CD69+ fraction.n = 4mice per group. Data are representative of two experiments. (B) WTmicewere infected with ActA-Lm-OVA-gp66 and/or ActA-Lm-OVA-2WIS at the indicateddoses. After 5 days, magnetically enriched endogenous splenic CD4+ Tcells wereanalyzed by flow cytometry for binding of IAb-gp66 tetramer and expression ofCD44, CXCR5, and PD-1. n = 3mice per group. Data are representative of twoseparate experiments. (C) 2D affinity measurements were performed on splenicTCR-transgenic CD4+ Tcells via micropipette adhesion frequency assays withbiotinylated p:MHC IAb-gp66 and IAb-OVA3Cmonomers. Log-normalized data areshown.WTCD45.1+ recipient mice were infected with ActA-Lm-OVA-gp66. After24 hours, 0.5 × 106 naïve (GFP–CD44–CD62L+CD69–CD25–) SMARTA IL-2.eGFPand OTII IL-2.eGFP CD4+ Tcells were pooled and adoptively transferred into infectedhosts. Splenic CD4+ Tcells were harvested 18 hours after transfer and analyzed for

expression of CD45.1, CD45.2,Vb5, CD69, and IL-2.eGFP by flow cytometry.Valuesin the larger boxes of flow cytometric plots depicting CD69 and IL-2.eGFP representpercentages of CD69+ cells, and values in the smaller boxes represent thepercentages of IL-2.eGFP+ cellswithin theCD69+ fraction. n=3mice per group. Dataare representative of two experiments. Ac Ka, is the 2D affinity (seemethods fordetails). (D)WTmicewere infectedwithActA-Lm-OVA-gp66. Five days after infection,magnetically enriched endogenous splenicCD4+Tcellswere costainedwith IAb-gp66and IAb-OVA3C tetramers and analyzed by flow cytometry for expression of CD44,CXCR5, and PD-1. Data are representative of four experiments. (E) WTmice wereinfected with 2.5 × 107 CFUActA-Lm-OVA-gp66. Enriched splenic CD4+ Tcellsharvested 5 days after infection were stained for IAb-gp66, CD44,TCRb, PD-1, andCXCR5. Log-normalized 2D affinitymeasurements were performed on FACS-purifiedIAb-gp66 tetramer-positive splenicTFH and non-TFH cells pooled from three to fiveanimals.TCRb quantifications were performed on unsorted aliquots stainedseparately. Data from two experiments are shown. (F) Naïve (GFP–CD44–CD62L+)SMARTA IL-2.eGFP CD4+ Tcells were stimulated for 16 hours with irradiatedCD4-depleted feeder cells and 1 mg/ml of gp66 and analyzed by flow cytometryfor expression of CD69,Va2, and IL-2.eGFP. Data are representative of twoexperiments. Data were analyzed by using Student’s t tests. ns, P > 0.05; *P < 0.05;**P < 0.01; ***P < 0.001; ****P < 0.0001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

that is tightly associated with IL-2 productionand TFH cell differentiation. Although TCR af-finity and antigen dose clearly contribute to theprobability that a cell will attain this threshold,costimulation and cytokine signals can also con-tribute. It has been shown that transforminggrowth factor–b (TGFb) signaling can attenuateTCR signaling (34) and limit T cell responses tohigh-affinity antigens (35). Indeed, the addition ofTGFb significantly reduced the mean expressionof CD69 by activated naïve T cells, the propor-tion of CD69+ cells that produced IL-2, andthe mean expression of IL-2, whereas TGFb block-ade had the opposite effect (fig. S12). However,the percentage of cells that expressed CD69 wasunaltered by either intervention. Thus, in addi-tion to modulating TCR-independent signalingcascades that control effector T cell specification,TGFb may also influence IL-2 production bylimiting TCR-signal strength without limitingthe frequency of naïve T cells that receiveactivating TCR signals (36). In this regard, it isnotable that the addition of IL-6 overrode therepressive effects of TGFb (TH17 cell conditions;Fig. 3A), resulting in significantly higher IL-2expression. Indeed, this was greater than anyother T effector–polarizing condition. Becausethe major effects of IL-6 are independent ofTCR-signal strength and its actions contributeto both TFH and TH17 cell development, clearly,TCR-independent factors that modulate IL-2production may affect TFH–non-TFH cell devel-opmental decisions. Furthermore, in view of theshared requirement for IL-6–induced Stat3 sig-

naling and high production of IL-21 by both TH17and TFH cells, these data suggest a considerableoverlap in the developmental programming ofthese two subsets and, perhaps, shared regu-lation of the tightly syntenic Il2 and Il21 loci(Fig. 2A).

Discussion

The findings in this study provide new insightsinto themechanics that control the early bifurca-tion of CD4+T cells intoTFH andnon-TFH effectors,placing reciprocal production and utilization ofIL-2 at the center of this key developmentaldecision. Because divergent IL-2 signaling andBcl6 expression have been linked to effector ver-sus central memory, respectively (37), these re-sults may also have implications for alternateprogramming of CD4+ T cell memory. Our find-ings predict that IL-2 nonproducers are fatedfor effector memory, whereas IL-2 producers arefated for central memory.It has been proposed that asymmetric cell di-

vision results in the partitioning of factors thatguide the divergent development of progeny ofactivated naïve T cells (38). Although the observa-tion herein that TFH–non-TFH cell fate determina-tion is initially encoded well before cell divisiondoes not preclude a role for asymmetric celldivision, it does suggest that signaling betweenT cells that receive differing activation signalslikely plays a dominant role. It has been shownthat homotypic T cell conjugation mediated byLFA-1–ICAM interactions between activatedT cells facilitates directional, paracrine delivery

of IL-2 via multifocal synapses (39). The expres-sion of Icam1 was enhanced on IL-2– cells in thecurrent study (Figs. 1C and 4A). Because thekinetics of IL-2 production are within the averagedwell time of T cells that form long-lived inter-actions on an activating dendritic cell (40–42),our findings suggest that T cell–T cell interac-tions that result in directional IL-2 signaling be-tween IL-2 producers and nonproducers mayoccur on the same dendritic cell, although thiswill necessitate further study. The role of Treg cellsin buffering IL-2 availability to non–IL-2 pro-ducers, owing to the constitutive expression of thehigh-affinity IL-2 receptor and high-avidity LFA-1by Treg cells will also require additional inquiry.Results in this report indicate that naïve CD4+

T cells that receive differing strengths of TCRsignals—whether the result of intrinsic TCR af-finity or expression differences or receipt of con-temporaneous non-TCR signals that augment orrepress TCR signal strength—stratify into thosethat exceed a threshold that predisposes to IL-2production and early TFH cell commitment andthose that fail to express IL-2 yet are program-med to receive IL-2 signaling that reinforces non-TFH effector commitment. However, althoughthe exceedance of this threshold appears neces-sary, it is not always sufficient, because somecells that express comparable levels of CD69 donot express IL-2, implying the contribution ofadditional factors yet to be defined. Moreover,although in a primary response, IL-2 expres-sion strongly correlates with the developmentof TFH versus non-TFH cells, this correlation

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 10 of 13

0.00

0.05

0.10

0.15

2^(C

TB

2M–C

TG

OI)

Myc

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

E2f3a

0.000

0.002

0.004

0.006

E2f5

0.000

0.005

0.010

0.015

0.020

E2f6

0 102 103 104 105

0

102

103

104

105

48.2 49.4

IL-2.eGFP

CD

69

20 hours

2.5µg/mL αCD3

C

0 103 104 105 1060

20

40

60

80

100

CTV

Tfhnon-Tfh

0 103 104 105 106

0

103

104

105

106

40.7

51.0

CXCR5

PD

-1

0 103 104 105 106

0

103

104

105

106

CTV

CX

CR

5

CD45.2+

CD45.1

SMARTA CD45.2 + CTV

ActaA-Lm-OVA-gp66

72 hrsB

********

****

*****

ns

****

ns

*****

ns

CD69+ GFP+ 0.1 µg/mL

CD69+ GFP– 0.1 µg/mL

CD69– GFP– 0.1 µg/mL

CD69+ GFP+ 10 µg/mL

CD69+ GFP– 10 µg/mL

CD69– GFP– 10 µg/mL

0 103 104 105 106

CD69 MFI

αCD3(μg/ml) 0.1 0.5 1 5 10

0

10000

20000

30000

40000

50000

0.1 0.5 1 5 10

0

50

100

0.1 0.5 1 5 10

0

5

10

15

20

0.1 0.5 1 5 10

0

1000

2000

3000

CD

69

0 103 104 105 106

0

103

104

105

106

0.1µg/mL αCD3

45.46.3

10µg/mL αCD3

95.718.6

0 103 104 105 106

IL-2.eGFP

% C

D69

+

IL-2

.eG

FP

MF

I

CD

69 M

FI (

CD

69+)

% IL

-2.e

GF

P+

A

Fig. 8. IL-2 producers and TFH exhibit enhanced cell-cycle progression.(A) Naïve (GFP–CD44–CD62L+CD69–CD25–) IL-2.eGFP CD4+ T cellswere stimulated for 18 hours with indicated concentrations ofplate-bound anti-CD3 and 1 mg/ml of soluble anti-CD28. They werethen analyzed for the expression of CD69 and IL-2.eGFP by flowcytometry. The MFI of CD69 expression within the CD69+GFP– andCD69+GFP+ gates was quantitated for the indicated concentrations ofanti-CD3 (right). Three technical replicates per condition are shown.This experiment was performed three times. (B) Naïve (GFP–CD44–CD62L+

CD69–CD25–) IL-2.eGFP CD4+ T cells were stimulated in vitro withsoluble anti-CD3 (2.5 mg/ml), soluble anti-CD28 (0.5 mg/ml), and irradiated

CD4-depleted feeder cells. qPCR was performed on CD69+ IL-2.eGFP+

and IL-2.eGFP– CD4+ Tcells sorted 20 hours after activation. Three technicalreplicates per condition are shown. Data are representative of twoexperiments and were analyzed by one-way ANOVA. (C) WT CD45.1+

recipient mice were infected with ActA-Lm-OVA-gp66. After 24 hours, 5 × 104

CTV-labeled naïve (GFP–CD44–CD62L+CD69–CD25–) SMARTA CD4+

T cells were adoptively transferred into infected hosts. Splenic CD4+ T cellsharvested 3 days after transfer were analyzed for expression of CD44,PD-1, and CXCR5 by flow cytometry. n = 4 mice per experiment. Thisexperiment was performed three times. ns, P > 0.05; **P < 0.01;***P < 0.001; ****P < 0.0001; error bars depict SEM.

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

is not fixed for subsequent responses (fig. S13).Thus, IL-2 expression by TFH cell precursorsdoes not ensure IL-2 expression by TFH ef-fectors in a recall response nor does lack ofIL-2 expression by non-TFH cell precursors pre-clude IL-2 expression by non-TFH cell effectors.Nevertheless, the utility of IL-2 as an earlymarker for cells fated to these different effectorprograms is established herein. Thiswill offer theopportunity to discover new factors that deter-mine the bifurcation into TFH and non-TFHeffectors, as exemplified by the finding of apossible Mxd-Myc-Max axis in controlling theearly differential expression of Bcl6. We proposethat this should provide a basis for strategiesto modulate the balance of effector T cell re-sponses for therapeutic ends.

Materials and methodsMice

B6.Cg-Tg-IL2tm1(eGFP)Weav (IL-2.eGFP) and B6.IL-2.BAC-inThy1.1 (2BiT) were generated usingstrategies previously described (25) and bred atthe University of Alabama at Birmingham (UAB)animal facility. B6N-Tyrc-Brd/BrdCrCrl (albino B6)and B6-LY5.2/Cr (congenic B6 CD45.1) were pur-chased from Frederick Cancer Center and inter-crossed to produce albino B6.CD45.1. C57BL/6(WT B6), Tcrb–/– (B6.129P2-Tcrbtm1Mom/J), OT-II(B6.Cg-Tg(TcraTcrb)425Cbn/J) mice and micetransgenic for constitutive eGFP expression(C57BL/6-Tg(CAG-EGFP)1310sb/LeySop/J) were pur-chased from The Jackson Laboratory. SMARTATg [Tg(TcrLCMV)1Aox] (43) on a B6 backgroundwere a generous gift from A. Zajac (Departmentof Microbiology, UAB). All intercrosses to gen-erate additional strains, such as SMARTA IL-2.eGFP, SMARTA2BiT, SMARTA IL-2.eGFPThy1.1+,OT-II IL-2.eGFP, and 2BiT CD45.1 were generatedby crosses inUAB’s breeding facility. Animalswerebred andmaintained under specific pathogen-freeconditions in accordancewith institutional animalcare and use committee regulations.

Tissue processing and flowcytometric analysis

Mice were sacrificed by isoflurane euthanisia be-fore removal of spleen and/or lymph nodes.Secondary lymphoid tissues were disrupted bymashing with a syringe in complete RPMI-1640over a 70-mm filter. Surface staining was per-formed in PBS with 2% FBS and 0.1% sodiumazide. T cells from 2BiT animals were directlystained for surface Thy1.1 (clone HIS5-1) withoutsecondary stimulation. For identification of TFH,cells were incubated with biotinylated anti-CXCR5 for 1 hour at room temperature, thenwashed and incubated with streptavidin-APCor PE-Cy7 and additional surface markers for20 min at 4°C. Intracellular staining for tran-scription factors was performed using either BDFix/Perm or eBioscience Foxp3 staining kits. Forex vivo phospho-Stat staining, freshly harvestedsplenocytes were fixed for 10 min at 37°C in 4%PFA in PBS, stained with eFluor450- or PacBlue-conjugated anti-Thy1.1, refixed with 4% PFA inPBS and permeabilized in 90%MeOH for 30min

on ice. Following this, cells were stained forphosphorylated Stat5 and additional markers atroom temperature for 1 hour. Alternatively, cellswere fixed in 4% PFA in PBS for 10 min at 37°C,permeabilized in 90% MeOH for 30 min on ice,and then stained for 1 hour at room temperature.Absolute T and B cell numberswere calculated

using PKH reference beads (Sigma- Aldrich) orconcentration values (events per microliter) ob-tained on the Attune NxT. Absolute TCRb num-bers were calculated using BDQuantibrite BeadsFluorescence Quantification Kits. Briefly, cellswere stained at saturating concentrations ofPE-labeled anti-TCRb (clone H57-597). A stan-dard curve generated from PEQuantibrite Beadswas then used to transform TCRb MFI measure-ments into absolute quantifications.All flow cytometric data were acquired on an

Attune NxT (Thermo Fisher Scientific), LSRII, oran Aria II (BD Immunocytometry Systems, SanJose, CA) and analyzed with FlowJo software(Tree Star, Eugene, OR).

T cell isolationNaïve T cell isolation

Polyclonal CD62LhiCD25–CD44loCD4+ and CD8+

T cells were purified from single-cell suspensionsof secondary lymphoid tissues (spleen with orwithout axillary, brachial, cervical, mesenteric,inguinal, and medial iliac lymph nodes) in twostages. First, CD8+ or CD4+ T cells were isolatedusing Dynabeads (ThermoFisher 11445D), MACSCell Separation (Miltyeni Biotec 130-104-454) orprepared by negative selection against CD8 orCD4, MHC class II, CD11b, B220, and CD25 (allAbs labeled with FITC) using anti-FITC BioMagparticles (Polysciences, Warrington, PA). Second,cells were purified by sorting on a FACSAria II,gating on the CD4+CD25–CD62LhiCD44lo (andin some cases IL-2.Thy1.1–/IL-2.eGFP–) fraction.Naïve SMARTA TCR Tg cells were sorted di-rectly as above from lymph node and splenictissue.

Activated T cell isolation

Cells cultured in vitro were harvested at varioustime points, resuspended in labeling buffer (2%FBS in PBS), and FAC-sorted a second time asCD4+ or CD8a+ CD69+ and either IL-2.eGFP+/Thy1.1+ or as IL-2.eGFP–/Thy1.1–. SMARTA IL-2.eGFP T cells isolated ex vivo from acutely ac-tivated recipient mice were processed from tissueby negative selection with biotinylated antibodiesto CD11b, CD11c, and B220, streptavidin-conjugatedmicrobeads, and LS columns (Miltenyi Biotech).Column flow-through fractions were then stainedfor a congenic marker (Thy1.1, Thy1.2, or CD45.2)before sorting as above.

In vitro T cell activation

Sorted naïve T cells were activated in completeRPMI-1640 (RPMI medium containing 10% FBS,100 IU/ml penicillin, 100 mg/ml streptomycin,1 mM sodium pyruvate, nonessential aminoacids, 50 mM b-mercaptoethanol and 2 mMl-glutamine) for 4 to 36 hours with anti-CD3(2.5 mgml−1) or 1 mgml−1 LCMV glycoprotein pep-

tide 66-77, anti-CD28 (1 mgml−1), and irradiatedsplenocytes at a 5:1 ratio of splenocytes to T cellsunder nonpolarizing conditions (i.e., without ad-ditional cytokines or antibodies). In some experi-ments, sorted naïve T cells were activated with arange of plate-bound anti-CD3 concentrationsand 0.5 mg/ml anti-CD28 or a range of solubleanti-CD3 concentrations and irradiated feedersat a CD4:feeder ratio of 1:5.For restimulation of splenic TFH and non-TFH

cells, magnetically enriched splenic CD4+ T cellswere stained simultaneously with tetramer andbiotin-labeled anti-CXCR5 (see table above) for1 hour at room temperature. Cells were thenwashed and stained with fluorophore-labeledstreptavidin and PD1 for 20 min at 4°C. Labeledcells were then incubated for 4 hours in completeRPMIwith 2 mg/ml anti-CD28 on flat-bottom 96-well plates precoated with 5 mg/ml anti-CD3.Following restimulation, cells were stained forCD69, CD44, CD4, additional surface markers,and viability dye for 15 to 20 min at 4°C.

Adoptive transfer and Ab-mediatedin vivo depletion

For adoptive transfer experiments examiningIL-2.eGFP or IL-2.Thy1.1 expression at earlytime points, 1 × 106 to 2.5 × 106 naïve cells wereinjected i.v. into congenic recipientmice infected24 hours before transfer unless otherwise in-dicated. For experiments involving cotransferof Smarta.IL-2.eGFP and OT-II.IL-2.eGFP donorcells, 5 × 105 sorted naïve cells of each donorstrain were injected retro-orbitally (RO) intomiceinfected 24 hours before transfer. For adoptivetransfer experiments examining TFH differentia-tion three or more days after transfer, 5 × 104

sorted naive donor cells were injected RO intocongenic recipients infected 24 hours beforetransfer unless otherwise indicated. For cotransferinto TCRb KO recipients, 1 × 106 magneticallyenriched bulk CD4+ T cells from wild-type B6 ortransgenic eGFP and CD45.1 2BiT congenic micewere injected RO into TCRb-deficient mice fol-lowed by infection 1 day after transfer. For de-pletion of IL-2.Thy1.1 (2BiT) cells in vivo, micewere given a single intraperitoneal injection of250 mg of anti-Thy1.1 or isotype monoclonal anti-body 24 hours before infection or immunization.

Infections and protein immunizationsLm

Mice were immunized i.v. with 200 ml of PBScontaining live (dose as indicated) or heat-killedActin A–deficient Listeria monocytogenes (ActA–

Lm; 2 × 109 to 2 × 1010) (18). All Lm strains usedwere transformed by a plasmid containingOVA250-387 and one of four different IAb–specific“foreign” peptides: (i) a mutant epitope ofI-Ea (“2W1S”); (ii) flagellin peptide 456-475 fromClostridium (“Cbir1”); (iii) glycoprotein 66-77peptide of LCMV (“gp66”); or (iv) flagellin pep-tide 427-441 from Salmonella typhimurium(“FliC”) all expressed under the control of the hly(listeriolysin O) promoter. All Lm strains wereproduced in the laboratory of S.-S. Way as pre-viously described (44). Bacteria were grown in

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 11 of 13

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

brain-heart infusion (BHI)mediumwith 15 mg/mlchloramphenicol to an absorbance of >0.1 at600 nm, and doses varied as indicated. The ac-tual number of live bacteria injected was con-firmed by dilution and growth on BHI agarplates containing chloramphenicol.

OVA/CFA

Mice were immunized i.p. with 100 ml of a 100-mgchicken egg ovalbumin emulsion in CFA.

OVA/Alum

Mice were injected i.p. with 200 ml of a 0.5-mg/ml emulsion of chicken egg ovalbumin in alum.The emulsion was prepared by mixing 1 mg/mlovalbumindissolved inwater (Invivogen vac-pova)1:1 with alum (FisherScientific Imject Alum 77161).

Citrobacter rodentium

Mice were orally gavaged with 1 × 109 to 2 × 109

CFU of Citrobacter rodentium strain DBS100(ATCC 51459) or the bioluminescent ICC180derivative (generously provided by G. Frankeland S. Wiles, Imperial College London). Miceinfected with the ICC180 derivative were shavedand imaged with an IVIS 100 Imaging System(Xenogen, Inc.) as previously described (28).

RNA-seq and analysis

For sample preparation and hybridization, totalRNA was isolated from purified naïve (CD4+ orCD8a+, CD25–CD69–CD44loCD62L+ IL-2.eGFP–)and activated (CD4+ or CD8a+, CD69+ IL-2.eGFP+

or IL-2.eGFP–) T cells with Qiazol and miRNeasymicro kits according to manufacturer’s recom-mendations (Qiagen). Library preparation wasperformed using Illumina TruSeq techonology.Samples were processed at UAB Heflin Centerfor Genomic Science for Next Generation Se-quencing (NGS) or La Jolla Institute (LJI) usingthe Illumina HiSeq 2000 Sequencing System.Reads were mapped to the mm10 genome usingTopHat (version 2.0.12) (45). BAM files weresorted using SAMtools (version 0.1.19) (46), andreads were counted for each gene using HTSeq(version 0.6.1) (47) and NCBI Mus musculusAnnotation Release 106 (GRCm38.p4). RNA ex-pression was normalized using the rlog functionfrom the DEseq2 R package (version 1.8.2) (48).Differential gene expression analysis was per-formed using DEseq2, and P values were cor-rected with the Benjamini-Hochburg procedure.Volcano plots were created using the ggrepel Rpackage. To calculate gene set enrichment, adifferential expression probability density func-tion (PDF) was determined for each gene usingQuantitative Set Analysis for Gene Expression(QuSAGE) (49). PDFs were combined for eachgene set to calculate gene set activity after cor-recting for gene-gene correlation. Gene set PDFswere compared usingWelch’s t test, and P valueswere adjusted using the Benjamini-Hochbergprocedure.

Statistical analysis

Experimental P values were calculated usingunpaired Student’s t tests, Welch’s t tests, or

one-way or two-way ANOVA tests with Tukey’spost hoc multiple comparisons analysis. A P valueof <0.05 was considered significant. See figurelegends for details.

REFERENCES AND NOTES

1. J. Zhu, H. Yamane, W. E. Paul, Differentiation of effector CD4T cell populations. Annu. Rev. Immunol. 28, 445–489(2010). doi: 10.1146/annurev-immunol-030409-101212;pmid: 20192806

2. N. J. Tubo et al., Single naive CD4+ T cells from a diverserepertoire produce different effector cell types during infection.Cell 153, 785–796 (2013). doi: 10.1016/j.cell.2013.04.007;pmid: 23663778

3. R. P. Bucy et al., Single cell analysis of cytokine genecoexpression during CD4+ T-cell phenotype development.Proc. Natl. Acad. Sci. U.S.A. 92, 7565–7569 (1995).doi: 10.1073/pnas.92.16.7565; pmid: 7638231

4. R. J. Johnston et al., Bcl6 and Blimp-1 are reciprocal andantagonistic regulators of T follicular helper cell differentiation.Science 325, 1006–1010 (2009). doi: 10.1126/science.1175870;pmid: 19608860

5. R. I. Nurieva et al., STAT5 protein negatively regulatesT follicular helper (TFH) cell generation and function.J. Biol. Chem. 287, 11234–11239 (2012). doi: 10.1074/jbc.M111.324046; pmid: 22318729

6. A. Ballesteros-Tato et al., Interleukin-2 inhibits germinal centerformation by limiting T follicular helper cell differentiation.Immunity 36, 847–856 (2012). doi: 10.1016/j.immuni.2012.02.012; pmid: 22464171

7. I. Bilic, W. Ellmeier, The role of BTB domain-containing zincfinger proteins in T cell development and function.Immunol. Lett. 108, 1–9 (2007). doi: 10.1016/j.imlet.2006.09.007; pmid: 17084908

8. J. G. Cyster, Chemokines, sphingosine-1-phosphate, and cellmigration in secondary lymphoid organs. Annu. Rev. Immunol.23, 127–159 (2005). doi: 10.1146/annurev.immunol.23.021704.115628; pmid: 15771568

9. J.-Y. Lee et al., The transcription factor KLF2 restrains CD4+

T follicular helper cell differentiation. Immunity 42, 252–264(2015). doi: 10.1016/j.immuni.2015.01.013; pmid: 25692701

10. J. P. Weber et al., ICOS maintains the T follicular helpercell phenotype by down-regulating Krüppel-like factor 2.J. Exp. Med. 212, 217–233 (2015). doi: 10.1084/jem.20141432;pmid: 25646266

11. W. Cui, S. M. Kaech, Generation of effector CD8+ T cellsand their conversion to memory T cells. Immunol. Rev.236, 151–166 (2010). doi: 10.1111/j.1600-065X.2010.00926.x;pmid: 20636815

12. S. Crotty, Follicular helper CD4 T cells (TFH). Annu. Rev. Immunol.29, 621–663 (2011). doi: 10.1146/annurev-immunol-031210-101400; pmid: 21314428

13. S. C. Lee, A. Bottaro, L. Chen, R. A. Insel, Mad1 is atranscriptional repressor of Bcl-6. Mol. Immunol. 43,1965–1971 (2006). doi: 10.1016/j.molimm.2005.11.017;pmid: 16423395

14. E. L. Stone et al., ICOS coreceptor signaling inactivates thetranscription factor FOXO1 to promote TFH cell differentiation.Immunity 42, 239–251 (2015). doi: 10.1016/j.immuni.2015.01.017; pmid: 25692700

15. D. Gong, T. R. Malek, Cytokine-dependent Blimp-1expression in activated T cells inhibits IL-2 production.J. Immunol. 178, 242–252 (2007). doi: 10.4049/jimmunol.178.1.242; pmid: 17182561

16. G. A. Martins, L. Cimmino, J. Liao, E. Magnusdottir,K. Calame, Blimp-1 directly represses Il2 and theIl2 activator Fos, attenuating T cell proliferation andsurvival. J. Exp. Med. 205, 1959–1965 (2008). doi: 10.1084/jem.20080526; pmid: 18725523

17. A. Laurence et al., Interleukin-2 signaling via STAT5 constrainsT helper 17 cell generation. Immunity 26, 371–381 (2007).doi: 10.1016/j.immuni.2007.02.009; pmid: 17363300

18. R. A. Brundage, G. A. Smith, A. Camilli, J. A. Theriot,D. A. Portnoy, Expression and phosphorylation of theListeria monocytogenes ActA protein in mammalian cells.Proc. Natl. Acad. Sci. U.S.A. 90, 11890–11894 (1993).doi: 10.1073/pnas.90.24.11890; pmid: 8265643

19. M. M. Curtis et al., Fidelity of pathogen-specific CD4+ T cells tothe Th1 lineage is controlled by exogenous cytokines,interferon-gamma expression, and pathogen lifestyle. Cell HostMicrobe 8, 163–173 (2010). doi: 10.1016/j.chom.2010.07.006;pmid: 20709293

20. Y. Lin, K. Wong, K. Calame, Repression of c-myc transcriptionby Blimp-1, an inducer of terminal B cell differentiation.Science 276, 596–599 (1997). doi: 10.1126/science.276.5312.596; pmid: 9110979

21. A. L. Shaffer et al., Blimp-1 orchestrates plasma celldifferentiation by extinguishing the mature B cell geneexpression program. Immunity 17, 51–62 (2002). doi: 10.1016/S1074-7613(02)00335-7; pmid: 12150891

22. C. Grandori, S. M. Cowley, L. P. James, R. N. Eisenman, TheMyc/Max/Mad network and the transcriptional control of cellbehavior. Annu. Rev. Cell Dev. Biol. 16, 653–699 (2000).doi: 10.1146/annurev.cellbio.16.1.653; pmid: 11031250

23. G. C. Preston et al., Single cell tuning of Myc expression byantigen receptor signal strength and interleukin-2 inT lymphocytes. EMBO J. 34, 2008–2024 (2015).doi: 10.15252/embj.201490252; pmid: 26136212

24. S. Heinzel et al., A Myc-dependent division timer complementsa cell-death timer to regulate T cell and B cell responses.Nat. Immunol. 18, 96–103 (2017). doi: 10.1038/ni.3598;pmid: 27820810

25. L. E. Harrington, K. M. Janowski, J. R. Oliver, A. J. Zajac,C. T. Weaver, Memory CD4 T cells emerge from effector T-cellprogenitors. Nature 452, 356–360 (2008). doi: 10.1038/nature06672; pmid: 18322463

26. M. Long, A. J. Adler, Cutting edge: Paracrine, but not autocrine,IL-2 signaling is sustained during early antiviral CD4 T cellresponse. J. Immunol. 177, 4257–4261 (2006). doi: 10.4049/jimmunol.177.7.4257; pmid: 16982857

27. B. T. Edelson, P. Cossart, E. R. Unanue, Cutting edge:Paradigm revisited: Antibody provides resistance to Listeriainfection. J. Immunol. 163, 4087–4090 (1999).pmid: 10510340

28. P. R. Mangan et al., Transforming growth factor-b inducesdevelopment of the TH17 lineage. Nature 441, 231–234 (2006).doi: 10.1038/nature04754; pmid: 16648837

29. R. Basu et al., Th22 cells are an important source of IL-22 forhost protection against enteropathogenic bacteria. Immunity37, 1061–1075 (2012). doi: 10.1016/j.immuni.2012.08.024;pmid: 23200827

30. D. J. Silberger, C. L. Zindl, C. T. Weaver, Citrobacter rodentium:A model enteropathogen for understanding the interplay ofinnate and adaptive components of type 3 immunity.Mucosal Immunol. 10, 1108–1117 (2017). doi: 10.1038/mi.2017.47; pmid: 28612839

31. J. Huang et al., The kinetics of two-dimensional TCRand pMHC interactions determine T-cell responsiveness.Nature 464, 932–936 (2010). doi: 10.1038/nature08944;pmid: 20357766

32. B. D. Stadinski et al., Hydrophobic CDR3 residues promotethe development of self-reactive T cells. Nat. Immunol. 17,946–955 (2016). doi: 10.1038/ni.3491; pmid: 27348411

33. Y. S. Choi et al., ICOS receptor instructs T follicular helper cellversus effector cell differentiation via induction of thetranscriptional repressor Bcl6. Immunity 34, 932–946 (2011).doi: 10.1016/j.immuni.2011.03.023; pmid: 21636296

34. C.-H. Chen et al., Transforming growth factor b blocks Teckinase phosphorylation, Ca2+ influx, and NFATc translocationcausing inhibition of T cell differentiation. J. Exp. Med.197, 1689–1699 (2003). doi: 10.1084/jem.20021170;pmid: 12810687

35. W. Z. Mehal, S. Z. Sheikh, L. Gorelik, R. A. Flavell, TGF-bsignaling regulates CD8+ T cell responses to high- and low-affinity TCR interactions. Int. Immunol. 17, 531–538 (2005).doi: 10.1093/intimm/dxh233; pmid: 15824070

36. S. C. McKarns, R. H. Schwartz, N. E. Kaminski, Smad3 isessential for TGF-b1 to suppress IL-2 production and TCR-induced proliferation, but not IL-2-induced proliferation.J. Immunol. 172, 4275–4284 (2004). doi: 10.4049/jimmunol.172.7.4275; pmid: 15034041

37. M. Pepper, A. J. Pagán, B. Z. Igyártó, J. J. Taylor, M. K. Jenkins,Opposing signals from the Bcl6 transcription factor and theinterleukin-2 receptor generate T helper 1 central and effectormemory cells. Immunity 35, 583–595 (2011). doi: 10.1016/j.immuni.2011.09.009; pmid: 22018468

38. J. T. Chang et al., Asymmetric T lymphocyte divisionin the initiation of adaptive immune responses. Science315, 1687–1691 (2007). doi: 10.1126/science.1139393;pmid: 17332376

39. C. A. Sabatos et al., A synaptic basis for paracrine interleukin-2signaling during homotypic T cell interaction. Immunity29, 238–248 (2008). doi: 10.1016/j.immuni.2008.05.017;pmid: 18674934

40. S. Stoll, J. Delon, T. M. Brotz, R. N. Germain, Dynamic imagingof T cell-dendritic cell interactions in lymph nodes. Science

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 12 of 13

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

296, 1873–1876 (2002). doi: 10.1126/science.1071065;pmid: 12052961

41. G. Shakhar et al., Stable T cell-dendritic cell interactionsprecede the development of both tolerance and immunity invivo. Nat. Immunol. 6, 707–714 (2005). doi: 10.1038/ni1210;pmid: 15924144

42. V. Hurez et al., Restricted clonal expression of IL-2 by naiveT cells reflects differential dynamic interactions with dendriticcells. J. Exp. Med. 198, 123–132 (2003). doi: 10.1084/jem.20022230; pmid: 12835480

43. A. Oxenius, M. F. Bachmann, R. M. Zinkernagel,H. Hengartner, Virus-specific major MHC classII-restricted TCR-transgenic mice: Effects onhumoral and cellular immune responses after viralinfection. Eur. J. Immunol. 28, 390–400 (1998).doi: 10.1002/(SICI)1521-4141(199801)28:01<390::AID-IMMU390>3.0.CO;2-O; pmid: 9485218

44. J. M. Ertelt et al., Selective priming and expansion of antigen-specific Foxp3−CD4+ T cells during Listeria monocytogenesinfection. J. Immunol. 182, 3032–3038 (2009). doi: 10.4049/jimmunol.0803402; pmid: 19234199

45. D. Kim et al., TopHat2: Accurate alignment of transcriptomesin the presence of insertions, deletions and gene fusions.Genome Biol. 14, R36 (2013). doi: 10.1186/gb-2013-14-4-r36;pmid: 23618408

46. H. Li et al., The Sequence Alignment/Map format andSAMtools. Bioinformatics 25, 2078–2079 (2009).doi: 10.1093/bioinformatics/btp352; pmid: 19505943

47. S. Anders, P. T. Pyl, W. Huber, HTSeq—A Python framework towork with high-throughput sequencing data. Bioinformatics 31,

166–169 (2015). doi: 10.1093/bioinformatics/btu638;pmid: 25260700

48. M. I. Love, W. Huber, S. Anders, Moderated estimationof fold change and dispersion for RNA-seq data with DESeq2.Genome Biol. 15, 550 (2014). doi: 10.1186/s13059-014-0550-8;pmid: 25516281

49. G. Yaari, C. R. Bolen, J. Thakar, S. H. Kleinstein, Quantitative setanalysis for gene expression: A method to quantify gene setdifferential expression including gene-gene correlations.Nucleic AcidsRes. 41, e170 (2013). doi: 10.1093/nar/gkt660; pmid: 23921631

ACKNOWLEDGMENTS

The authors thank members of the Weaver Lab, L. Harrington,H. Hu, S. Kaech, A. Weinmann, and A. Zajac for helpful discussions.We thank D. Wright and B. Dale for mouse breeding andgenotyping. Funding: This work was supported by NIH grantsR01 AI035783 and R01 DK113739 (C.T.W.), R01 AI110113 (B.D.E.),R01 AI107120 (J.J.M.), P30 DK04335 (J.J.M.), R21 AI124143(J.J.M.), and DP1 AI131080 (S.S.W.). Trainee support was providedby NIH T32 AI007051 to C.J.W., D.D., D.P., and D.J.S. and byNIH F30 DK105680 (J.R.S.). Additional support was provided byUAB Institutional Funds (C.T.W.), March of Dimes Foundation(S.S.W.), HHMI Scholar’s Program (S.S.W.), Burroughs WellcomeFund (S.S.W.), and the Milton Fund (J.J.M.). D.D., S.W., J.R.S., andC.G.W. are members of the UAB Medical Scientist TrainingProgram (MSTP), supported by NIH T32 GM008361. Weacknowledge the UAB Epitope Recognition and ImmunoreagentCore Facility for provision of some antibodies used in this study.Author contributions: Author contributions are as follows: project

conception, experimental design, and data interpretation byD.D., C.J.W., and C.T.W.; reporter mouse design, construction,and validation by C.J.W., D.D., R.J.L., H.T., and C.T.W.; RNA-seqstudies and data analysis by C.J.W., S.W., R.D.H., and B.T.W.;ChIP-PCR and ATAC-seq experiments and data analysis by D.P.and R.D.H.; in vivo Listeria, Citrobacter, and immunization datacollection, analysis, and interpretation by C.J.W., D.D., D.J.S., C.G.W.,J.R.S., and C.T.W.; confocal microscopy by C.L.Z.; TCR-affinityand stimulation-strength experiments and data interpretation byD.D., R.A., E.M.K., R.J.M., and B.D.E.; additional ex vivo experimentsby D.D., C.J.W., C.G.W., and J.R.S.; design, construction, andvalidation of MHCII tetramers and data interpretation by J.J.M.;design, construction, and validation of ActA Listeria strains anddata interpretation by S.S.W.; and manuscript preparation andediting by D.D., C.J.W., and C.T.W. Competing interests: Theauthors declare no competing interests. Data and materialsavailability: RNA-seq data are deposited in the NCBI GeneExpression Omnibus under accession number GSE116608. Anyadditional data needed to evaluate the conclusions in this paperare present either in the main text or the supplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/361/6407/eaao2933/suppl/DC1Materials and MethodsFigs. S1 to S13References (50–57)

4 July 2017; resubmitted 25 March 2018Accepted 6 August 201810.1126/science.aao2933

DiToro et al., Science 361, eaao2933 (2018) 14 September 2018 13 of 13

RESEARCH | RESEARCH ARTICLEon F

ebruary 9, 2020

http://science.sciencemag.org/

Dow

nloaded from

helper T cellsDifferential IL-2 expression defines developmental fates of follicular versus nonfollicular

Turner, Robin D. Hatton, James J. Moon, Sing Sing Way, Brian D. Evavold and Casey T. WeaverCarlene L. Zindl, Rita J. Luther, Daniel J. Silberger, Benjamin T. Weaver, E. Motunrayo Kolawole, Ryan J. Martinez, Henrietta Daniel DiToro, Colleen J. Winstead, Duy Pham, Steven Witte, Rakieb Andargachew, Jeffrey R. Singer, C. Garrett Wilson,

DOI: 10.1126/science.aao2933 (6407), eaao2933.361Science 

, this issue p. eaao2933ScienceIL-2.

T cells that received the highest T cell receptor signals were able to produce+only those naïve CD4−−receptor strength fate decision was linked to T cell− T cell+ cells. The CD4FHwhereas nonproducers, which receive IL-2, become non-T

cells,FH T cells that produce IL-2 are fated to become T+ show that naïve CD4et al.Using IL-2 reporter mice, DiToro poorly understood, though there is evidence to suggest that the T cell growth factor interleukin-2 (IL-2) may play a role.

cell remainFH or non-TFHcell functions at sites of pathogen encounter. The factors underlying differentiation into a T cells orchestrate enhanced innate immuneFHproduction and the establishment of B cell memory. By contrast, non-T

T cells that support B cell antibody+) cells are a subpopulation of CD4FHImmunological T follicular helper (T(IL-)2 be or not to be?

ARTICLE TOOLS http://science.sciencemag.org/content/361/6407/eaao2933

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2018/09/12/361.6407.eaao2933.DC1

REFERENCES

http://science.sciencemag.org/content/361/6407/eaao2933#BIBLThis article cites 57 articles, 19 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2018 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on February 9, 2020

http://science.sciencem

ag.org/D

ownloaded from