Ndfip1 mediates peripheral tolerance to self andexogenous antigen by inducing cell cycle exit inresponding CD4+ T cellsJohn A. Altina, Stephen R. Daleya, Jason Howittb, Helen J. Rickardsa, Alison K. Batkina, Keisuke Horikawaa,Simon J. Prasada, Keats A. Nelmsa, Sharad Kumarc, Lawren C. Wud, Seong-Seng Tanb, Matthew C. Cooka,1,and Christopher C. Goodnowa,1,2

aJohn Curtin School of Medical Research, Australian National University, Canberra, ACT 2601, Australia; bFlorey Neurosciences Institute, Universityof Melbourne, Parkville, VIC 3010, Australia; cCentre for Cancer Biology, University of South Australia, Adelaide, South Australia 5000, Australia;and dDepartment of Immunology, Genentech, South San Francisco, CA 94080

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2013.

Contributed by Christopher C. Goodnow, December 17, 2013 (sent for review October 13, 2013)

The NDFIP1 (neural precursor cell expressed, developmentally down-regulated protein 4 family-interacting protein 1) adapter for theubiquitin ligase ITCH is genetically linked to human allergic and auto-immune disease, but the cellular mechanism by which these proteinsenable foreign and self-antigens to be tolerated is unresolved. Here,we use two unique mouse strains—an Ndfip1-YFP reporter and anNdfip1-deficient strain—to show that Ndfip1 is progressively in-duced during T-cell differentiation and activation in vivo and thatits deficiency causes a cell-autonomous, Forkhead box P3-indepen-dent failure of peripheral CD4+ T-cell tolerance to self and exogenousantigen. In small cohorts of antigen-specific CD4+ cells respondingin vivo, Ndfip1 was necessary for tolerogen-reactive T cells to exitcell cycle after one to five divisions and to abort Th2 effector differ-entiation, defining a step in peripheral tolerance that providesinsights into the phenomenon of T-cell anergy in vivo and is distinctfrom the better understood process of Bcl2-interacting mediatorof cell death-mediated apoptosis. Ndfip1 deficiency precipitatedautoimmune pancreatic destruction and diabetes; however, thisdepended on a further accumulation of nontolerant anti-self T cellsfrom strong stimulation by exogenous tolerogen. These findingsilluminate a peripheral tolerance checkpoint that aborts T-cellclonal expansion against allergens and autoantigens and demon-strate how hypersensitive responses to environmental antigens maytrigger autoimmunity.

immunological tolerance | allergy | T lymphocyte | Interleukin-4 |Aire (Autoimmune Regulator)

In healthy individuals, mature T cells in peripheral lymphoidtissues proliferate and acquire effector functions in response to

antigens from pathogenic microbes but remain tolerant to self-antigens and innocuous environmental antigens. Defects in thisphenomenon of “peripheral T-cell tolerance” are thought tocontribute to the burden of autoimmune and allergic disease, butthere is only a fragmented understanding of its cellular basis, itsconnection to specific genetic circuits, and the interconnectionbetween autoimmunity and hypersensitivity to exogenous anti-gens (1). This problem is exemplified by the genetic circuitencoding Ndfip1 [neural precursor cell expressed, developmentallydown-regulated protein 4 (NEDD4) family-interacting protein 1],a transmembrane protein localized to the Golgi and intracellularvesicles that recruits and activates the HECT-type E3 ubiquitinligase Itch (2–7). Human genetic studies have associated NDFIP1and ITCH with allergic and autoimmune diseases. Inherited ITCHdeficiency results in asthma-like chronic lung disease with non-fibrotic lymphocytic pneumonitis (90% cases) and organ-specificautoimmunity (60% cases) variably involving the thyroid, liver,intestine, or pancreatic islets (8). Inherited NDFIP1 polymorphismsare associated with inflammatory bowel disease (9, 10), asthma

(11), rheumatoid arthritis (12), and multiple sclerosis (13). Itremains unclear which cellular mechanisms of tolerance aredisrupted by these genetic variants to result in allergic andautoimmune disease.Ndfip1 and Itch were first revealed as important immune

regulators in mouse genetic studies. Homozygous inactivatingmutations in the Itchy strain cause dermatitis, lung mononuclearinflammation, lymphadenopathy with follicular hyperplasia, in-creased activated T cells (notably IL-4–producing Th2 cells),expansion of B1b cells in the peritoneal cavity, and early death(5, 6, 14, 15). Although the murine pathology has often beendescribed as autoimmune because of its spontaneous de-velopment, there is currently little direct evidence of T-cell au-toimmunity, and the predominant inflammation of skin andmucosal surfaces suggests an exaggerated response to innocuousenvironmental antigens. Indeed, elegant studies showed that Itchdeficiency prevents high-zone tolerance in an experimentalmodel of respiratory exposure to an egg protein allergen (16).An almost identical skin and lung inflammatory syndrome occursin mice inheriting a homozygous gene-trap insertion that greatlyreduces Ndfip1 mRNA and protein (2). Although much progresshas been made elucidating diverse biochemical functions of Itch

Significance

Advances in organ transplantation and treatment of allergyand autoimmune disease hinge upon harnessing a physiologi-cal switch that allows T cells to decide between proliferatingextensively or actively becoming tolerant. The experimentspresented here illuminate a critical element of this natural switch,Ndfip1 (neural precursor cell expressed, developmentally down-regulated protein 4 family-interacting protein 1), a partner pro-tein of ubiquitin ligases induced during the first several divisionsafter T cells encounter antigen. They define the cellular action ofNdfip1 in vivo, acting within dividing helper T cells that haveresponded to innocuous foreign or self-antigen that shouldnormally be tolerated, to force their exit from cell cycle beforethey have divided so many times that they acquire tissue-dam-aging effector functions.

Author contributions: J.A.A., S.R.D., J.H., H.J.R., A.K.B., K.H., S.J.P., K.A.N., L.C.W., S.-S.T.,M.C.C., and C.C.G. designed research; J.A.A., S.R.D., J.H., H.J.R., A.K.B., K.H., S.J.P., K.A.N.,and M.C.C. performed research; J.H., S.K., and S.-S.T. contributed new reagents/analytictools; J.A.A., S.R.D., J.H., H.J.R., A.K.B., K.H., K.A.N., L.C.W., S.-S.T., M.C.C., and C.C.G.analyzed data; and J.A.A., S.R.D., M.C.C., and C.C.G. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1M.C.C. and C.C.G. contributed equally to this work.2To whom correspondence should be addressed. E-mail: chris.goodnow@anu.edu.au.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1322739111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1322739111 PNAS | February 11, 2014 | vol. 111 | no. 6 | 2067–2074

IMMUNOLO

GY

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

and Ndfip1 in many cell types (3, 17), the cellular basis for im-mune dysregulation in their absence is unresolved, and their rolein T-cell tolerance to self-antigens has yet to be examined.Defects in several different cellular mechanisms for peripheral

T-cell tolerance have been implicated in the inflammatory dis-ease caused by defects in the Itch-Ndfip1 genetic circuit. T-cellanergy is a mechanism defined initially in tissue culture thatprevents initiation of T-cell proliferation when T cells are stim-ulated without a CD28 costimulus (18). Itch was required forT-cell anergy in cultured cells rendered anergic by prolonged invitro treatment with ionomycin or harvested from TCR trans-genic (Tg) mice 10 d after exposure to a high-tolerogenic dose offoreign antigen. An intact Itch gene was correlated with di-minished TCR signaling and proteolytic degradation of proteinkinase C (PKC)-θ, phospholipase C (PLC)-γ, JunB, and c-Junproteins (16, 19). A role for Itch in nondegradative ubiquitina-tion of the TCR CD3 ζ subunit to inhibit its phosphorylation andthe activation of Zap-70 has also been shown (20). Likewise,Ndfip1 deficiency causes JunB accumulation (2) and allows Tcells to make IL-2 for a sustained period in vitro without theneed for CD28 costimulation (21). Itch-deficient T cells alsodisplay exaggerated NF-κB signaling in response to TNF-α, be-cause Itch forms a complex that recruits Tnfaip3 to terminateNF-κB signaling (22). Itch has also been implicated in ubiquiti-nation and degradation of Bcl-10, an adaptor protein with anessential role in TCR and CD28 signaling to NF-κB (23). TCR-activated T cells that lack Itch or Ndfip1 form a diminishedpercentage of Forkhead box P3 (Foxp3+) induced T-regulatorycells (iTregs) when cultured with TGF (24, 25).Increased differentiation of Th2 effector cells is a prominent

feature of Itch or Ndfip1 deficiency that is partly explained bytheir role in ubiquitination and degradation of JunB, an Il4 genetranscription factor preferentially expressed in Th2 cells (2, 14,26). However, Tg mice that express equivalently high levels ofJunB in T cells do not exhibit a similar spontaneous accumula-tion of Th2 effector cells or inflammatory disease (26, 27). Ab-normal Notch signaling within activated T cells might alsocontribute (28–30) because the Itch ortholog in Drosophila,Suppressor of Deltex, was discovered as a negative regulator ofNotch signaling, and the Drosophila Ndfip1 ortholog has similargenetic effects on this pathway (31, 32). Itch ubiquitinates andterminates ligand-independent Notch signaling in endosomes,requiring an adaptor protein that may be Ndfip1 based onDrosophila studies (31, 33, 34).Although there are many possible cellular explanations for

how allergy and autoimmunity may result from defects in theItch-Ndfip1 genetic circuit, resolving these alternatives to placethe biochemical circuit in its correct cellular context awaitsa systematic comparison of the fates of mutant and wild-type Tcells responding to normally tolerogenic antigens in vivo. Here,we perform this comparison and reveal that the critical functionfor Ndfip1 in peripheral tolerance is as an induced, cell-auton-omous brake against effector CD4+ cell differentiation.

ResultsLethal Immune-Mediated Th2 Disease Caused by a TruncatingMutation of Ndfip1. In a C57BL/6 × C57BL/10 mouse pedigreesegregating point mutations induced by ethylnitrosourea, off-spring exhibited a Mendelian recessive syndrome of spontaneousmast cell-rich dermatitis with median onset age 90 d, followedby weight loss and premature mortality at a median age of 160 d(Fig. 1 and Fig. S1 A and B). This was accompanied by lymph-adenopathy, splenomegaly, increased CD86 and CD23 activationmarkers on B cells, expansion of the activated/memory subset ofCD4+ and CD8+ T cells, greatly elevated serum IgE, and for-mation of a prominent population of IL-4+ and IFN-γ+ CD4+cells (Fig. S1 C–E). The mutation was given the allele name“krusty” and mapped in a genome-wide scan of C57BL/6 ×C57BL/10 SNPs to a chromosome 18 region containing a strongcandidate gene, Ndfip1, based on a similar phenotype in a strainbearing a gene-trap insertion (2). Sequencing the Ndfip1 exons

revealed a G-to-A substitution in the first nucleotide of the intronfollowing exon 5, disrupting the mRNA splice donor sequence(Fig. 1A, allele denoted “Ndfip1kru”). Genotyping for this muta-tion showed perfect concordance between Ndfip1kru homozygos-ity and the lethal dermatitis phenotype in >100 affected offspring(>800 meioses) generated by successively backcrossing to wild-type C57BL/6 mice and intercrossing heterozygous progeny.Amplification and sequencing of cDNA from Ndfip1wt/wt ver-

sus Ndfip1kru/kru splenocytes revealed the mutation abolishednormal splicing and resulted in two aberrantly spliced mRNAproducts, encoding frame shifts and premature stop codonstruncating the Ndfip1 protein either immediately preceding orfollowing the first of the three transmembrane domains (Fig. 1 Aand B and Fig. S2 A and B). In transfected HEK293 cells, Ndfip1protein was detected from only the aberrantly spliced cDNAretaining the first transmembrane sequence and from the wild-type cDNA (Fig. S2C). Western blotting of activated T-celllysates from Ndfip1kru/kru mice with antibody to the Ndfip1 Nterminus (35) confirmed the absence of the ∼27-kDa wild-typeNdfip1 protein and only trace amounts of ∼20-kDa proteincorresponding to the aberrantly spliced mRNA retaining the firsttransmembrane sequence (Fig. 1C). This truncated residualprotein included the N-terminal cytoplasmic PY motifs that bindand activate Itch and Nedd4 (3) but was evidently insufficient tosuppress lethal inflammation.Because Ndfip1 is expressed widely and has been implicated

in functions outside of the adaptive immune system (36), wetested whether or not the dermatitis or premature lethality wassuppressed by the selective elimination of T and B cells inRag1−/− Ndfip1kru/kru mice. In contrast to lymphocyte-sufficientNdfip1kru/kru littermate controls, dermatitis and mortality did notdevelop in Rag1−/− Ndfip1kru/kru mice (Fig. 1D), demonstratingthat the spontaneous pathology absolutely requires T and/orB lymphocytes.

Ndfip1 Deficiency Causes Cell Autonomous Accumulation of CD4+

Effectors Not Corrected by Normal Foxp3+ Cells. To determinewhich subsets of T or B cells are regulated by Ndfip1, 50:50 mixedhematopoietic chimeras were constructed by transplanting bonemarrow from Ndfip1kru/kru or wild-type donors, distinguished byCD45.1 or CD45.2 allotypic markers, into irradiated Rag1−/−

recipients. As observed in unmanipulated Ndfip1kru/kru mice, B cells

D

A

N

C

nt

-58 nt

-125

PY1 PY2 PY?

wild

-type

mut

ant

KO

Ndfip1

tubulin

B CExon 4 Exon 5 Exon 6

GTATG…….GG gtGTATG…….GG at

-58 nt

-125 nt

Wt: Mut:

0 100 2000

20406080

100

(n=13)(n=27)

age (days)

%de

rmat

itis-

free

0 100 2000

20406080

100

age (days)

%su

rviv

ing

(n=13)(n=27)

Ndfip1kru/kru Rag+/+ , +/– (n=13)Ndfip1kru/kru Rag–/– (n=27)

Fig. 1. Immune-mediated lethal inflammatory syndrome in mice witha truncating Ndfip1 splice site mutation. (A) Schematic showing the locationof the Ndfip1kru mutation within the Ndfip1 exon 5 splice donor sequenceand the resulting two aberrant splice products. (B) Schematic showing thetopology of the Ndfip1 protein and position of the truncating mutations. (C)Western blot of primary T-cell lysates from wild-type, Ndfip1kru/kru (mutant),and Ndfip1−/− (KO) mice, probed with an antibody raised to a conserved N-terminal peptide of Ndfip1, and then stripped and reprobed with antibodyto tubulin to assess loading. (D) Dermatitis and survival of Ndfip1kru/kru micewith normal or null Rag1 genes, aged to 250 d or until moribund necessi-tating the animal to be killed.

2068 | www.pnas.org/cgi/doi/10.1073/pnas.1322739111 Altin et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

and CD8 T cells were both dramatically altered in chimerasreconstituted with only 50% Ndfip1kru/kru hematopoietic cells (Fig.S3 A–C), but these effects were an indirect consequence ofNdfip1kru/kru in other lymphocytes because they occurred to an equalextent in the CD45.1 B and CD8 T cells bearing wild-type Ndfip1developing alongside Ndfip1kru/kru lymphocytes. By contrast, therewas a direct, cell-autonomous requirement for Ndfip1 in CD4+ Tcells (Fig. 2). Ndfip1 deficiency resulted in greatly increasedfrequencies of CD44hi IFN-γ+ and IL-4+ cells among CD4+ cellsin chimeras reconstituted with 100% Ndfip1kru/kru marrow. In mixedchimeras bearing 50% Ndfip1kru/kru marrow, expansion of IL-4+

cells, CD44hi cells, and IFN-γ+ CD4+ cells was limited to theCD45.2+ T cells lacking normal Ndfip1 and did not occur in theCD45.1+ wild-type CD4+ cells that developed in the same ani-mals. Analysis of Foxp3+ CD4+ cells in the mixed chimerasshowed that Ndfip1kru/kru neither increased nor decreased thefrequency of Foxp3+ cells even under competitive reconstitutionconditions where wild-type marrow was also present (Fig. 2).Despite the presence of large numbers of wild-type Foxp3+ cellsin these 50:50 mixed chimeras, accumulation of mutant CD44hi

Th2 and Th1 CD4+ effector cells was not corrected.

Ndfip1-YFP Is Progressively Induced During T-Cell Maturation andActivation. To understand the cell-autonomous requirement forNdfip1 in preventing pathological accumulation of effector CD4+T cells, Ndfip1-YFP Tg reporter mice were constructed by pro-nuclear injection of a mouse bacterial artificial chromosome(BAC) containing the Ndfip1 genomic locus with an eYFP cDNAinserted at the Ndfip1 translation start site (Ndfip1-YFP Tgmice). In the thymus, YFP fluorescence of CD4+CD8+ double-positive (DP) cells was only marginally higher than control thy-mocytes from non-Tg mice, but fluorescence was ∼100-foldhigher in CD4+CD8– and CD4–CD8+ single-positive (SP) cells,corresponding to the stage of positive selection by TCR-pMHCligation (Fig. 3A). A further ∼fivefold higher YFP fluorescencewas observed in the CD25+GITR+ subset of CD4+ SP cells,which comprise cells receiving a stronger TCR signal, includingnascent Foxp3+ regulatory T cells (Tregs) and cells undergoingnegative selection (37). In the spleen, YFP fluorescence in mature

CD4+ and CD8 T+ cells was comparable to their thymic SP coun-terparts, whereas lower heterogeneous levels were observed in Bcells (Fig. 3B). Within splenic CD4+ cells, CD44hi CD25– ef-fector/memory and CD44int CD25+ Treg fractions had ∼fivefoldincreased YFP above CD44lo CD25– naïve CD4+ T cells. Theseresults, measured at single-cell resolution, mirror the pattern ofNdfip1mRNA expression observed in sorted T-cell subsets in theImmgen database (38). When splenocytes from Ndfip1-YFP Tganimals were cultured with plate-bound antibodies against CD3and CD28, YFP increased ∼30-fold above the level in restingCD4+ T cells over the course of 2 d, with ∼2-fold induction by6 h and ∼10-fold induction by 24 h (Fig. 3C). Thus, Ndfip1 isprogressively induced by antigen-receptor signaling as T cellsmature and then become activated.

Ndfip1 Aborts Antigen-Induced Division and Differentiation of CD4+

T Cells After Several Days. The findings above indicated that Ndfip1may regulate either the initial activation of naive CD4+ cells byinnocuous foreign- or self-antigens or their subsequent clonalaccumulation as effector cells. To resolve these alternatives, theNdfip1kru mutation was crossed into the 3A9 TCR Tg strain (37,39), providing a source of naïve CD4+ T cells recognizing henegg lysozyme (HEL) peptide presented by the MHC class IIprotein I-Ak. Equal numbers of naïve HEL-specific CD4+ T cellsfrom wild-type 3A9 Tg donors (allotypically marked as hetero-zygous CD451/2) and Ndfip1kru/kru 3A9 Tg donors (allotypicallymarked as homozygous CD452/2) were mixed, labeled with car-boxyfluorescein diacetate succinimidyl ester (CFSE), and trans-ferred into normal congenic B10.BR CD451/1 recipients. In theseexperiments, normal Ndfip1 was present in all cells other thanthe small subset of transferred CD452/2 CD4+ TCR Tg cells. Toanalyze the response to innocuous foreign antigen, the recipientmice were injected i.p. with 100 μg of HEL protein in salinewithout adjuvant, and clonal expansion of the transferred T cellsmonitored 3 or 6 d later using antibodies against the CD45allotypes and by CFSE fluorescence (Fig. 4A). After 3 d, donor Tcells in control recipients that did not receive HEL persisted inequal frequencies and had not diluted CFSE, indicating thatabsence of Ndfip1 did not result in spontaneous activation of

0 102 103 104 1050

20

40

60

80

100

28.3

0 102 103 104 1050

20

40

60

80

100

25.1

0 102 103 104 1050

20

40

60

80

100

5.42

0 102 103 104 1050

20

40

60

80

100

% o

f Max 5.25

0 102 103 104 105

0

102

103

104

105

IFN

g 1.83

0 102 103 104 105

0

102

103

104

105

3.09

0 102 103 104 105

0

102

103

104

105

5.76

0 102 103 104 105

0

102

103

104

105

17.5

0 102 103 104 105

0

102

103

104

105

13.4

0 102 103 104 105

0

102

103

104

105

7.71

0 102 103 104 105

0

102

103

104

105

0.037

0 102 103 104 105

0

102

103

104

105

IL-4

0.26

0 102 103 104 105

0

102

103

104

105

Foxp

3

6.44

0 102 103 104 105

0

102

103

104

105 6.48

0 102 103 104 105

0

102

103

104

105 9.23

0 102 103 104 105

0

102

103

104

105 5.12

Chimeras:

CD45.1 cells:CD45.2 cells:

Gated on:+ CD4 cells

100% wt 50% kru 100% kru

wtwt

wtkru

nonekru

CD45.2 CD45.1 CD45.2 CD45.2

100% wt

50% wt

100% kru

wtwt

wtkru

wtkru

CD

45.1

CD

45.2

CD

45.1

CD

45.2

CD

45.2

wtkru

5.1

5.2

5.1

5.2

01020304050

% C

D44

hi(F

oxp3

–C

D4+ )

05101520

IFN

g+( %

ofC

D4+ )

105101520

IL-4

+(%

ofC

D4+

)

2 2 20510152025

Foxp

3+

(%of

CD

4+)

nsns

*****

****

****

CD44

CD4 cells CD4 cells CD4 cells+ + +

Fig. 2. Ndfip1 deficiency causes cell-autonomousaccumulation of activated/effector CD4+ T cells thatis not corrected by the presence of wild-type Foxp3+

Tregs. Three groups of mixed bone marrow chi-meras were generated by transplanting irradiatedRag1−/− mice: “100% kru,” transplanted with 100%Ndfip1kru/kru CD45.2 bone marrow; “50% kru,”transplanted with a 50:50 mixture of Ndfip1kru/kru

CD45.2 bone marrow and Ndfip1+/+CD45.1 bonemarrow; “100% wt,” transplanted with a 50:50mixture of Ndfip1+ /+CD45.2 bone marrow andNdfip1+/+CD45.1 bone marrow. Flow cytometry wasused to analyze lymph node CD4+ cells, gated intoeither CD45.2+ or CD45.1+ cells, measuring cell sur-face expression of CD44, intracellular IFN-γ or IL-4after 3 h ex vivo phorbol myristate acetate (PMA)/ionomycin stimulation, and intracellular Foxp3. Rep-resentative plots and quantitation in multiple animalsare shown. Lines link data points from the same re-cipient animal. *P < 0.05; **P < 0.01; ***P < 0.001 byStudent t test.

Altin et al. PNAS | February 11, 2014 | vol. 111 | no. 6 | 2069

IMMUNOLO

GY

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

naïve CD4+ T cells. Three days after HEL injection, however,both mutant and wild-type donor T cells had increased in fre-quency 8- to 10-fold and had undergone multiple rounds of celldivision, reported by CFSE dilution (Fig. 4 A and B). At thisinitial stage of the response, there was little difference in theproliferation of mutant and wild-type TCR Tg CD4+ cells, al-though the Ndfip1kru/kru T cells were ∼1.5-fold as frequent.A remarkable divergence between wild-type and Ndfip1kru/kru

CD4+ T cells developed by day 6 after exposure to HEL antigenin saline, when the frequency of wild-type Tg donor cells haddecreased ∼fivefold compared with their peak at day 3, and theremaining cells retained appreciable CFSE (Fig. 4 A and B). Incontrast, the Ndfip1kru/kru CD4+ T cells had continued to increasein frequency a further ∼fourfold from day 3 and had diluted theirCFSE label beyond the limit of detection. Consequently, at day6, the Ndfip1kru/kru CD4+ progeny were ∼30- to 50-fold moreabundant than their cotransferred wild-type counterparts. Intra-

cellular flow cytometric staining at this time point revealed moreJunB in Ndfip1kru/kru CD4+ Tg cells than in coresident wild-typecells and that most Ndfip1kru/kru cells produced high levels of IL-4upon brief restimulation ex vivo, whereas few of the wild-type Tgcells did (Fig. 4 C–E).The continued expansion and effector differentiation of

Ndfip1kru/kru CD4+ T cells observed at day 6 of the response toexogenous HEL “tolerogen” was not attributable to a deficiencyin transacting suppression by Tregs, because it occurred cell-autonomously in recipient mice that had a normal T-cell reper-toire. It was nevertheless possible that differences in Foxp3induction within the transferred mutant and wild-type CD4+ Tcells might explain their divergent behavior. To test this possi-bility, we repeated the cotransfer experiments with a parallelgroup of recipients that received a mixture of CD452/2 Foxp3-deficient 3A9 TCR Tg CD4 cells and CD451/1 wild-type Tg cells(Fig. 5). Six days after transfer and administration of HEL insaline, Ndfip1kru/kru CD4+ T cells had expanded dramatically asbefore, but the expansion of Foxp3-deficient CD4+ T cells wasindistinguishable from their cotransferred wild-type CD4+ cells(Fig. 5A). Hence, Foxp3 is not required within tolerogen-responding CD4+ T cells to curtail their proliferative response invivo, whereas Ndfip1 is essential.Mitochondrial apoptosis regulated by the Bcl-2 and Bcl2-

interacting mediator of cell death (Bim) proteins is a well-established pathway contributing to peripheral T-cell tolerance

0 102 103 104 105

Ndfip1-YFP0

20

40

60

80

100

% o

f Max

0 102 103 104 105

Ndfip1-YFP

0

20

40

60

80

100

% o

f Max

0 102 103 104 105

Ndfip1-YFP

0

20

40

60

80

100

% o

f Max

0 102 103 104 105

Ndfip1-YFP

0

20

40

60

80

100

% o

f Max

0 102 103 104 1050

20

40

60

80

100

% of

Max

6 hr

unstimulatedstimulated

0 102 103 104 1050

20

40

60

80

100

% of

Max

24 hr

0 102 103 104 105

Ndfip1-YFP0

20406080

100

% of

Max

48 hr

A C

B220– cells

CD4+CD8– cells

thymocytes

splenocytes

CD4+ cells

B

0 102 103 104 105

CD8

0

102

103

104

105

CD

4

5.69

0.74

86.5

0 102 103 104 105

CD25

0

102

103

104

105

GIT

R 3.2

91.2

0 102 103 104 105

B220

0

102

103

104

105

CD

11b

50.438.1

0 102 103 104 105

CD44

0

102

103

104

105

CD

25

9.3177.5

9.64

0 102 103 104 105

CD8

0

102

103

104

105

CD

4

37.2

25.9

Fig. 3. Progressive induction of an Ndfip1 reporter gene during T-cell se-lection and activation. Flow cytometric analysis of Tg Ndfip1 reporter(Ndfip1-YFP Tg) mice bearing a BAC with eYFP integrated in place of theNdfip1 translation initiation codon. (A) YFP fluorescence of CD4+CD8+, CD4+

CD8–, CD4–CD8+, CD4+CD8–CD25–GITR–, and CD4+CD8–CD25+GITR+ thymo-cyte populations from Ndfip1-YFP Tg mice, gated as indicated. (B) YFPfluorescence of CD4+, CD8+, B220+, CD4+CD44loCD25–, CD4+CD44hiCD25–,and CD4+CD44intCD25+ splenocyte populations from Ndfip1-YFP Tg mice,gated as indicated. (C) YFP fluorescence of gated CD4+ cells among sple-nocytes from Ndfip1-YFP Tg mice that were incubated with or without plate-bound antibodies against CD3 and CD28 for the indicated periods. In allcases, the shaded gray histogram shows the YFP fluorescence of cells fromnon-Tg control mice.

–HEL day 3A B

C

D E

CD45.2

CD

45.1

CFSE

+HEL day 3 +HEL day 6

0 102 103 104 105

0

102

103

104

105 13 0

0.3587

0 102 103 104 105

0

102

103

104

105 1.0 0.29

70280 102 103 104 105

0

102

103

104

105

g

4 0.2

2.993

0 102 103 104 105

0

102

103

104

105 5.8 0

094

IL-4

IFN

g

CD45 CD452/

JunB

wt CD451/2wt CD452/2

wt CD451/2kru CD452/2

wt CD451/2 kru CD452/2

02468

10

HEL: – +3 63day:

+

dono

rcel

ls(%

)

020406080

HEL: – +3 63day:

+

IL-4

+do

norc

ells

(%)

0

2000

4000

6000

8000

wt + wt wt + kru

JunB

MFI

0 102 103 104 1050

20

40

60

80

100

0 102 103 104 1050

20

40

60

80

100

0 102 103 104 1050

20

40

60

80

100

CD4 20 102 103 104 105

0

102

103

104

105 951.5

2.75

0 102 103 104 1050

20

40

60

80

100

0 102 103 104 1050

20

40

60

80

100

0 102 103 104 105

0

102

103

104

105 90.40.32

9.06

0 102 103 104 105

0

102

103

104

105 90 0.3

9.06

0 102 103 104 105

0

102

103

104

105 99.40.23

0.29

0 102 103 104 105

0

102

103

104

105 99 0.2

0.29

1 2/ 2

Fig. 4. Ndfip1 is required in activated CD4+ T cells to prevent effector ac-cumulation and differentiation in response to exogenous antigen withoutadjuvant. (A) An equal mixture of HEL-specific CD4+ T cells was preparedfrom 3A9 TCRHEL Tg mice that were Ndfip1kru/kru CD452/2 or wild-type CD451/2,labeled with CFSE, and 2 × 106 total CD4+ cells injected into the circulation ofnormal B10.BR CD451/1 recipient mice. A subset of the recipient mice wasgiven 100 μg of HEL antigen in saline i.p. at the time of cell transfer. Shown isrepresentative flow cytometric analysis of recipient spleen cells on the in-dicated day after T-cell transfer. Upper graphs are gated on CD4+ T cells andshow the percentage derived from the two donors and the recipient. Lowergraphs show CFSE fluorescence of the respective donor and recipient cells.(B) Percentage of donor cells in multiple recipients analyzed as in A. Lineslink data points from the same recipient animal. (C) Percentage of donor Tcells expressing IL-4 measured by intracellular staining and gated as in Aafter 3 h ex vivo PMA/ionomycin stimulation. (D) Intracellular staining forJunB, IL-4, and IFN-γ on gated CD451/2 or CD452/2 donor CD4+ cells at day 6after transfer and injection of HEL in saline. (E) Quantitation of the JunB MFIof gated donor cells in multiple recipients analyzed as in D.

2070 | www.pnas.org/cgi/doi/10.1073/pnas.1322739111 Altin et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

(1, 40). To test whether a defect in this pathway could replicatethe failure of Ndfip1kru/kru cells to abort clonal expansion, weincluded a parallel group of recipients that received a mixture ofCD452/2 Bim-deficient 3A9 TCR Tg cells and CD451/1 wild-type3A9 Tg cells (Fig. 5). In contrast to the >50-fold greater ex-pansion of Ndfip1kru/kru cells, Bim-deficient cells exposed toexogenous HEL in saline accumulated an average ∼eightfoldhigher than the wild-type internal controls (Fig. 5A).To compare how defects in Ndfip1 or Bim affected the pro-

liferative state of the responding cells, we labeled the transferredT cells with cell trace violet (CTV) and measured division-linkedCTV dilution together with expression of the cell cycle markerKi-67 at day 6 of the response (Fig. 5 B and C). Analysis of CTVdilution indicated that, like their wild-type counterparts and alsothe Foxp3-deficient cells, most of the Bim-deficient CD4+ cellshad responded to exogenous HEL tolerogen by dividing oneto five times. However, the resulting CTVlow progeny expressedlittle Ki-67 on day 6, indicating that they had exited cell cycle bythis time point. By contrast, the transferred Ndfip1kru/kru cellsremained predominantly Ki-67+, revealing a failure to exit cellcycle (Fig. 5 B and C). These results establish that peripheraltolerance to innocuous exogenous antigen involves two geneti-cally separate cellular mechanisms within responding CD4+ Tcells: (i) Bim-dependent apoptosis; and (ii) Ndfip1-dependentexit from cell cycle that occurs even when apoptosis is defective.

Ndfip1 Prevents Autoimmune Disease by Aborting CD4+ CellProliferation and Differentiation. Peripheral tolerance mechanismsbecome critical backups when thymic tolerance fails (41); forexample, because of inherited Aire mutations that disrupt thy-mic deletion of organ-specific T cells (39, 42). In B10.BR micewith intact peripheral tolerance mechanisms, Aire deficiencycaused little clinical autoimmune disease on its own. We thereforecombined Aire and Ndfip1 deficiency by transplanting wild-type or

Aire−/− B10.BR mice with Ndfip1kru/kru bone marrow. WhereasAire-deficient recipients of wild-type marrow showed no signsof disease or histological evidence of pancreatitis, recipients ofNdfip1kru/kru marrow lost weight and became moribund 30–40 dafter transplant with extensive lymphocytic infiltration and auto-immune destruction of the exocrine pancreatic cells (Fig. S4). Thisgenetic cooperation mirrored that observed between Aire andCblb, the latter encoding another ubiquitin ligase with a role inperipheral tolerance (41).To analyze peripheral tolerance to a defined self-antigen made

in the pancreas, we crossed Ndfip1kru 3A9 TCRHEL Tg mice withmice expressing HEL protein controlled by the rat insulin genepromoter (insHEL Tg mice). Despite greatly increased thymicproduction of CD4+ T cells specific for the pancreatic islets, themajority of TCRHEL:insHEL double Tg mice with one or twowild-type Ndfip1 alleles were protected from diabetes (Fig. 6A).This protection has previously been shown to be attributable toAire-dependent thymic deletion (39) and Cblb-dependent pe-ripheral tolerance (43). By contrast, 100% of Ndfip1kru/kru doubleTg animals developed diabetes by 60 d of age (Fig. 6A). Analysisof the thymi nevertheless showed no effect of the Ndfip1kru/kru

mutation upon thymic deletion of 3A9 TCRHEL CD4+ cells ininsHEL Tg animals. Thus, the enhanced progression to diabetescaused by Ndfip1 deficiency likely reflected the fact that someTCRHEL CD4+ cells escape thymic deletion and must be con-trolled by Ndfip1-mediated peripheral tolerance.The role of Ndfip1 in peripheral tolerance to pancreatic islets

was pinpointed by transferring mature naïve CD4+ T cells from3A9 TCRHEL Tg wild-type or Ndfip1kru/kru mice into wild-typeinsHEL Tg recipients. This experimental design restricted Ndfip1deficiency to the small population of transferred cells, whichwere admixed into a normal, diverse repertoire of T cells in-cluding Tregs. Neither the wild-type nor the Ndfip1kru/kru CD4+ Tcells precipitated diabetes in insHEL Tg recipients unless 100 μgof HEL in saline was also given at the time of T-cell transfer(Fig. 6B). Within 8 d, all of the mice that had receivedNdfip1kru/kru

T cells and exogenous HEL became diabetic, whereas none of themice that received wild-type T cells and exogenous HEL becamediabetic. Diabetes in exogenous HEL-exposed mice receivingNdfip1kru/kru T cells was accompanied by severe lymphocytic in-filtration and destruction of most pancreatic islets (Fig. 6C).To test whether Ndfip1 deficiency affected the peripheral re-

sponse to islet-derived self-antigen alone, we tracked congeni-cally marked mixtures of Ndfip1kru/kru and wild-type 3A9TCRHEL Tg T cells in insHEL Tg recipients. In insHEL Tgrecipients not given exogenous HEL, a subset of Ndfip1kru/kru andwild-type donor CD4+ T cells were stimulated to divide in re-sponse to low levels of self-HEL, but only the Ndfip1kru/kru di-vided extensively based on analysis of CTV dilution on day 6,remained Ki-67+, and had differentiated into GATA-3+ effectorcells (Fig. 6D). In control recipients that lacked the insHELtransgene, neither mutant nor wild-type T cells were induced todivide or become GATA-3+. Thus, Ndfip1 is critical to abortCD4+ T-cell proliferation and differentiation induced by pancreas-derived self-antigen.Although responding Ndfip1kru/kru T cells remained in cell

cycle after stimulation by self-antigen alone, their net expansionwas only ∼threefold more than the cotransferred wild-typecounterparts (Fig. 6E). In insHEL Tg recipients that also re-ceived 100 μg of exogenous HEL in saline, however, theNdfip1kru/kru T cells accumulated to >10-fold higher frequenciescompared either to cotransferred wild-type T cells or to mutantT cells encountering self-antigen alone (Fig. 6 D and E). Becauselimited amounts of HEL are made and presented to 3A9 TCRTg CD4+ cells in the pancreas-draining nodes of insHEL Tgrecipients (44), we hypothesized that the amount of tolerogenpresented (from either self or foreign source) also governed thenet accumulation of Ndfip1kru/kru T cells. We tested this hy-pothesis by including a recipient group that lacked the insHELtransgene and was instead given a 1,000-fold lower dose ofexogenous HEL (0.1 μg) in saline (Fig. 6 D and E). Like the cells

CTV

Ki-6

7

+HEL

wt CD451/1

kru CD452/2

kru CD452/2

Foxp3ko

CD452/2BimkoCD452/2

–HEL

kru

foxp3ko

bimko

0.1

1

10

100

1000fo

ld e

xpan

sion

(rela

tive

to c

o-tra

nsfe

rred

wt)

kru

foxp3ko

bimko

0

20

40

60

80

100

Ki-6

7+ (%

am

ong

divi

ded

CD

4+ )

wtko

A C

B

Fig. 5. Ndfip1 mediates tolerance distinct from mitochondrial apoptosis orFoxp3, by aborting the proliferative response of CD4+ T cells. Cotransfer ofallotypically distinguished 3A9 TCRHEL Tg cells was performed as in Fig. 4 butincluding parallel groups of recipients that received cell mixtures from wild-type and Foxp3-deficient, or wild-type and Bim-deficient mice (all mice 3A9TCRHEL Tg), and using the cell division label CTV. (A) Quantitation of the fold-expansion (relative to cotransferred wild-type cells) of wild-type, Ndfip1kru/kru,Foxp3-deficient or Bim-deficient CD4+ cells in recipient mice at day 6 fol-lowing injection of HEL in saline, compared with the cell inputs at day 0. (B)Representative Ki-67 and CTV data from CD451/2 recipients on day 6, gatedon the cotransferred CD452/2 (red plots) or CD451/1 (blue plots) 3A9 TCRHEL

CD4+ T cells of the indicated genotypes. (C) Percentage of Ki-67+ cells remainingin cell cycle among donor CD4+ cells that had previously divided (CTVlo) inresponse to 100ug HEL in saline, measured in individual recipients as in B.

Altin et al. PNAS | February 11, 2014 | vol. 111 | no. 6 | 2071

IMMUNOLO

GY

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

exposed to tolerogenic HEL in other formats (insHEL and/or100 μg of exogenous HEL), the Ndfip1kru/kru donor CD4+ cells inrecipients given the low dose of exogenous HEL differentiatedinto GATA-3+ effectors and remained Ki-67+, but their accu-mulation was ∼10-fold less than in animals given the high dose.

DiscussionThe experiments above illuminate cellular mechanisms for pe-ripheral T-cell tolerance, which remain to be harnessed for al-lergy desensitization, transplantation tolerance, or autoimmunedisease therapy. As reviewed in ref. 1, many in vivo studies haverevealed that peripheral T-cell tolerance occurs by a burst oftolerogen-induced proliferation followed by Bim-mediated apo-ptotic elimination. The results here reveal a separate mechanism,showing that Ndfip1 acts within individual CD4+ T cells thathave begun proliferating in response to innocuous self or foreignantigens, forcing them to exit cell cycle after one to five divisionsin vivo. Because CD4+ cell differentiation into IL-4–producingcells is a function of the number of times they have divided (45,46), the failure of tolerogen-responding Ndfip1-deficient T cellsto exit cell cycle may explain their differentiation into GATA-3+

IL-4–producing Th2 effector cells. The failure to exit cyclewithout Ndfip1 contrasted with the peripheral tolerance defectcaused by Bim deficiency, where responding progeny still exitedcycle after dividing but resisted elimination by apoptosis. Theseresults identify an activation-induced partner of HECT ubiquitinligases, Ndfip1, to explain previous observations that Bim-deficient CD4+ T cells are still able to cease dividing in re-sponse to self-antigen in vivo despite their inability to undergoapoptosis (47). Our findings provide a cellular explanation forthe association of NDFIP1 with both allergic and autoimmunedisease (9–13), the co-occurrence of autoimmunity and asthma-like disease in humans with ITCH deficiency (8), and how cross-reactive environmental antigens may help to trigger autoimmunedisease (48, 49). As discussed below, Ndfip1-mediated exit fromcycle may identify a critical in vivo counterpart of T-cell anergyin vitro.The findings here address the unresolved role of Ndfip1 in

peripheral T-cell tolerance. Previous studies have establishedthat Ndfip1 deficiency in mice causes a T-cell driven, lethal

inflammatory disorder affecting particularly the skin, gastroin-testinal tract, and lungs, which could be recapitulated by selectiveNdfip1 deficiency in the T-cell lineage and suppressed by skew-ing the TCR repertoire to an innocuous ovalbumin OT-II TCRspecificity (2, 21, 24, 25). In the earlier studies, mixed bonemarrow chimera experiments found a higher fraction of Ndfip1-deficient CD4+ and CD8+ T cells were activated, but analysis waslimited because the mutant animals were on a heterogeneousstrain background derived variably from 129 and B6 strains, sothat complicated cell sorting and permeabilization was needed todistinguish mutant cells from the homogeneous B6-strain con-trols. The findings here extend those mixed chimera results usingmatched congenic B6.CD45.2/CD45.1 strains, revealing that ac-tivation of many CD8+ cells and B cells is a reaction to Ndfip1deficiency in other hematopoietic cells, whereas CD4+ cellslacking Ndfip1 undergo cell-autonomous expansion into Th1 andTh2 effector cells despite the presence of wild-type Tregs andother cells.Foxp3+ Tregs that develop during negative selection in the

thymus (nTregs) or during T-cell activation in the periphery(iTregs) mediate a non–cell-autonomous mechanism of periph-eral tolerance, and previous studies established that Itch- orNdfip1-deficient T cells are crippled in their differentiation intoiTregs (24, 25). The experiments here showing that Ndfip1 mu-tation induced autoimmune and inflammatory disease in con-genically matched mixed bone marrow chimeras and in celltransfer recipients despite numerous normal Foxp3+ cells, andfailure of Ndfip1-deficient T cells to abort proliferation that wasnot recapitulated by Foxp3-deficient cells, excludes deficits inFoxp3-expressing cells or their function as a primary explanation.The results here are consistent with Ndfip1 acting as a mech-

anism to steadily “raise the stakes” for each T cell to proceedinto successive rounds of division, thereby forcing cells that havereceived a suboptimal (tolerogenic) stimulus to abort ongoingcell division several days into the response. Ndfip1-YFP Tg re-porter mice revealed that TCR/CD28 stimulation causes a 30-fold increase in YFP fluorescence occurring gradually over thecourse of 2 d. This result is consistent with the Ndfip1 inductionpreviously reported for TCR/CD28-stimulated T cells: at theprotein level in cultured unfractionated T cells (2) and at the

0 100 200 3000

20406080

100

Ndfip1+/+Ndfip1kru/+Ndfip1kru/kru

age (days)

% n

on-d

iabe

tic

(n=5)(n=20)(n=14)A B C

D E

0.01

0.1

1

10

100

insHEL: + +g HEL PBS: 100 0.1

dono

r cel

ls(%

, nor

mal

ised

)

GATA-3

CTV

Ki-6

7

–HEL insHEL + insHEL 0.1ug HEL100ug HEL

kru Cd45 wt Cd45

2/2 1/1

10.9 77.6 99.5 28.3

7.26 15.9 89.7 8.24

020406080

100

wild-type Ndfip1kru

% p

ancr

eatic

isle

ts

T cells:

nonemoderatesevere

insulitis:

0 5 10 15 2005

1015202530

diabetic

days post cell transferbl

ood

gluc

ose

(mM

)

T cells:HEL PBS:

wt wt kru kru- + - +

Fig. 6. Ndfip1 deficiency disrupts peripheral T-cell toler-ance to pancreatic islets. (A) Incidence of diabetes in a co-hort of mice with the indicated Ndfip1 genotypes thatwere all doubly Tg for the 3A9 TCRHEL transgene and aninsulin-promoter HEL transgene (insHEL Tg). (B) Daily bloodglucose readings in singly insHEL Tg mice injected i.v. with106 CD4+ T cells from 3A9 TCRHEL Tg donors that were ei-ther Ndfip1kru/kru or wild-type. Half of the recipients alsoreceived 100 μg of HEL in saline i.p. at the time of T-celltransfer. (C) Quantitation of insulitis in ≥10 individual isletsin individual recipients (denoted by columns) 3 wk aftertransfer of wild-type or Ndfip1kru/kru T cells together withHEL in saline. (D) A mixture of HEL-specific CD4+ T cellsfrom 3A9 TCRHEL Tg mice that were Ndfip1kru/kru CD452/2 orwild-type CD451/2 was CTV-labeled and injected into thecirculation of non-Tg or insHEL Tg B10.BR CD451/1 recipientmice. One group of insHEL Tg recipients was given 100 μgof HEL in saline i.p. at the time of cell transfer, whereasanother group of non-Tg recipients received 0.1 μg of HELin saline i.p. The donor CD4+ cells were analyzed 6 d laterin the spleen for CTV dilution and expression of Ki-67 andGATA-3. Shown are representative plots of CTV fluores-cence versus Ki-67 staining (upper plots) or GATA-3 stain-ing (lower plots) on the cotransferred kru CD452/2 CD4+cells (red) and wt CD451/2 CD4+ cells (blue). (E) Frequencyamong CD4+ cells of wild-type and Ndfip1kru/kru donor cellsin individual recipient mice described in D, normalized tofrequencies in a non-Tg recipient that received no HEL.Lines link data points from the same recipient animal.

2072 | www.pnas.org/cgi/doi/10.1073/pnas.1322739111 Altin et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

mRNA level in sorted naïve T cells cultured with TGF-β (25).The induction of Ndfip1-YFP in all CD44hi activated/memoryT cells contrasts with models of T-cell anergy that would postulateselective induction of Ndfip1 by tolerizing stimuli. Instead, in-duction of Ndfip1 in actively dividing T cells may progressivelyraise the signaling threshold needed to keep the cells in cycle,and in the absence of sufficient microbial costimuli, this mayforce cell cycle exit. A feature of the “rising stakes” model ofNdfip1-mediated peripheral tolerance is that it defers the de-cision of whether to attack or disarm until after an initial pro-liferative response is made, providing more time for T cellsto sample the environment without compromising their speedof mobilization.Ndfip1-induced exit from cell cycle in CD4+ cells responding

to tolerogens may be an unexpected in vivo manifestation of T-cell clonal anergy. T-cell anergy has been best defined in T cellsstimulated in vitro through their TCR without CD28 cos-timulation, and is mostly viewed as a mechanism blocking initi-ation of T-cell IL-2 production and proliferation (18). Anergyhas not been considered as a mechanism to trigger dividing Tcells to exit cycle. T-cell responses to antigen in the absence ofCD28 costimuli in vivo have nevertheless not recapitulated the invitro models: instead, proliferation was initiated but followed byBim-mediated apoptosis, although the remaining cells have beenshown to be anergic to reinitiation of proliferation or IL-2 pro-duction provided sufficient tolerogen was present (1, 47). Itchhas been shown to be required in T cells for in vitro and in vivoT-cell anergy, as measured by reinitiation of cytokine productionand cell proliferation (16, 19). Ndfip1-deficient CD4+ cellsstimulated in culture with antibodies to CD3 in the absence ofCD28 have recently been shown to make more IL-2 for an ex-tended period (21). This difference could nevertheless be sec-ondary to a failure to exit cell cycle without Ndfip1, because IL-2production normally increases on a per cell basis as a function ofthe number of TCR-induced cell divisions (45, 46), and the IL-2measurements did not control for cell division history. Decreasedproduction of IL-2 without apparent loss of proliferative re-sponse during in vitro restimulation has been observed in anergicDO11 TCR Tg CD4+ T cells that have exited cycle in response totolerogenic self antigen but failed to undergo apoptosis becauseof Bim deficiency (47). The extent to which Ndfip1 forces cellcycle exit by down-regulation of IL-2 synthesis or by independenteffects on TCR-induced cell proliferation will require celldivision-based analysis with Il2/Ndfip1 double-deficient T cellsin future studies.Another future question raised by the findings here is which

biochemical targets of Ndfip1 result in exit from cell cycle inCD4+ T cells stimulated by self- or foreign-antigen in the ab-sence of adjuvant. Induction of Ndfip1 in actively dividing T cellsmay impose a sustained and elevated TCR-CD28 costimulatoryrequirement by downregulating TCR-ζ (20), PKC-θ, PLC-γ,JunB and c-Jun proteins (16, 19), Bcl-10, and NF-κB (22, 23).This is supported by the demonstration that Ndfip1-deficient Tcells make more IL-2 than wild-type T cells even when CD28 isgenetically ablated from both (21). However, normal T cellsabort their proliferation to tolerogenic stimuli in vivo even whenCD28 signals have been received (50–56). The presence of ad-ditional cytokines is also normally required to promote sustainedrounds of T-cell division and effector differentiation, notablyIL-12, IFN-α and -β, IL-1, and IL-4 (55, 57–59). These cyto-kines are typically produced extrinsically to the responding Tcells in response to infection, adjuvants, or cell damage, althoughautocrine or paracrine sources arise if the T cells divide enoughtimes to differentiate into effector cells that produce IFN-γ orIL-4. Ndfip1 may suppress the potential for autocrine productionof IL-2 (21) or IL-4 in actively dividing CD4+ cells by degradingc-Jun and JunB (2, 14), and by inhibiting Notch (31, 32). Ndfip1deficiency may also allow tolerogen-stimulated CD4+ T cells toremain in cycle by crippling the TGFβ signaling pathway (24, 25),which normally delivers an important anti-proliferative signal forT cells (60). The findings here provide the in vivo cellular context

to understand the integration of these diverse biochemicalpathways in future studies.A key finding from the experiments is that large numbers of

autoimmune effector T cells and autoimmune islet destructiononly developed when an intrinsic peripheral tolerance defect wascombined with a sufficiently large pool of organ-specific CD4+

cells that had escaped thymic deletion and also a large exogenousantigen trigger. The experiments highlight a third controlmechanism—limiting amount of tolerogenic antigen stimulus—that has also frequently been observed as a key variable in pe-ripheral T-cell tolerance (1). A high density of pMHC maysimply drive more rapid progress through successive cell cycles ordecrease apoptotic loss of daughter cells. One can envisage twosituations in which self-reactive T cells that have evaded thymicdeletion might be strongly stimulated: (i) when an exogenousfood, environmental, or microbial protein happens to containpeptides of similar sequence to the self-antigen (48, 49); and (ii)where damage to an organ releases much more self-antigen forpresentation in the draining lymph node. Given the associationof NDFIP1 polymorphisms with a number of inflammatory dis-eases involving exogenous or self-antigens (9–13), it is not in-conceivable that the cellular defect in peripheral tolerancedefined here could arise through polygenic inheritance patternsinvolving NDFIP1, HLA, and other T-cell regulatory genes. Thefindings here imply that “high-zone tolerance” intervention withexogenous peptides or whole antigens, as practiced in some allergy-desensitization regimes, may in fact exacerbate disease in indi-viduals where there is an intrinsic failure of tolerogen-reactiveCD4+ cells to exit cycle and abort effector differentiation.

Materials and MethodsMice. The 3A9 TCR Tg, insHEL Tg, Aire−/−, Bim−/−, and Foxp3−/− mice were asdescribed previously (37, 39, 61). Mice were either on the C57BL/6 back-ground, the congenic C57BL/6.SJL-PtprcCD45.1 background, or the congenicB10.BR and B10.BR.SJL-PtprcCD45.1 strains, as indicated. All procedureswere approved by the Australian National University Animal Ethics andExperimentation Committee.

Ndfip1 cDNA Sequencing and Western Blotting. TRIzol-isolated RNA fromsplenocytes was reverse transcribed with oligo-dT, amplified using primersflanking the Ndfip1 coding region, separated by agarose gel electrophoresis,cloned into pMXs-IRES-GFP retroviral vector, and sequenced on an ABI3730capillary sequencer. Naïve T cells (CD4+CD44loCD62LhiCD25–) were sortedfrom wild-type, Ndfip1kru/kru, and Ndfip1null/null mice, stimulated 48 h with5 μg/mL plate-bound anti-CD3 (145-2C11) and 2 μg/mL soluble anti-CD28(37.51), and lysed in 1× SDS sample buffer with sonication and boiling, andlysates were resolved on a polyacrylamide gel. Proteins were blotted toPVDF membrane, blocked with 5% (wt/vol) milk, probed with rat mono-clonal antibody against an N-terminal peptide of mouse Ndfip1 (35), anddeveloped with HRP-conjugated anti-rat IgG (Millipore) and Enhanced Chem-iluminescence Substrate (Perkin-Elmer).

Flow Cytometry. In vitro stimulation and staining for cell surface and in-tracellular proteins and CFSE dilution was performed as previously described(61). CTV labeling was done at room temperature as described (62) withslight modifications as described in SI Materials and Methods.

Bone Marrow Chimeras. RAG1-deficient mice were X-irradiated with 1 × 500cGy, and wild-type or Aire-deficient mice were given 2 × 500 cGy 4 h apart,before i.v. injection of 2 × 106 T cell-depleted bone marrow cells per mouse.Recipients were maintained on polymixin B and neomycin for 6 wk.

T-Cell Responses to HEL Protein in Vivo. CD45 and CFSE/CTV-labeled splenocytesuspensions containing 0.5–2 × 106 TCR3A9-expressing CD4+ T cells weretransferred by i.v. injection to B10.BR-congenic recipients. Where indicated,recipients received 0.1 or 100 μg of HEL in PBS by i.p. injection on the sameday. For cellular analysis, spleens were collected on days 3 or 6 for flowcytometry. For analysis of pathology, recipients were monitored daily forhyperglycemia using a blood glucose meter (Medisense Optium), and at3 wk posttransfer, pancreata were formalin-fixed, paraffin-embedded, andstained with hematoxylin/eosin, and the extent of lymphocytic infiltration ofindividual islets was scored blinded to the sample identity.

Altin et al. PNAS | February 11, 2014 | vol. 111 | no. 6 | 2073

IMMUNOLO

GY

INAUGURA

LART

ICLE

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0

ACKNOWLEDGMENTS. We thank the Genotyping Team of the AustralianPhenomics Facility (APF) for expert mapping, sequencing, and genotyp-ing; the staff of the APF for husbandry; Debbie Howard for experttechnical assistance; and Dr. Ian Parish for advice on the manuscript. This

work was supported by National Institutes of Health Grants AI100627 andAI054523, Wellcome Trust Grant 082030/B/07/Z, and National Health andMedical Research Council Grants 585490, 1016953, 427620, 1009190,and 1002863.

1. Goodnow CC, Ohashi PS (2013) Immunologic Tolerance. Fundamental Immunology,ed Paul WE, 7th Ed (Wolters Kluwer Health/Lippincott Williams and Wilkins, Phila-delphia), pp 765–794.

2. Oliver PM, et al. (2006) Ndfip1 protein promotes the function of itch ubiquitin ligaseto prevent T cell activation and T helper 2 cell-mediated inflammation. Immunity25(6):929–940.

3. Mund T, Pelham HR (2009) Control of the activity of WW-HECT domain E3 ubiquitinligases by NDFIP proteins. EMBO Rep 10(5):501–507.

4. Harvey KF, Shearwin-Whyatt LM, Fotia A, Parton RG, Kumar S (2002) N4WBP5, a po-tential target for ubiquitination by the Nedd4 family of proteins, is a novel Golgi-associated protein. J Biol Chem 277(11):9307–9317.

5. Hustad CM, et al. (1995) Molecular genetic characterization of six recessive viablealleles of the mouse agouti locus. Genetics 140(1):255–265.

6. Perry WL, et al. (1998) The itchy locus encodes a novel ubiquitin protein ligase that isdisrupted in a18H mice. Nat Genet 18(2):143–146.

7. Rotin D, Kumar S (2009) Physiological functions of the HECT family of ubiquitin li-gases. Nat Rev Mol Cell Biol 10(6):398–409.

8. Lohr NJ, et al. (2010) Human ITCH E3 ubiquitin ligase deficiency causes syndromicmultisystem autoimmune disease. Am J Hum Genet 86(3):447–453.

9. Ramon HE, et al. (2011) The ubiquitin ligase adaptor Ndfip1 regulates T cell-mediatedgastrointestinal inflammation and inflammatory bowel disease susceptibility. Muco-sal Immunol 4(3):314–324.

10. Franke A, et al. (2010) Genome-wide meta-analysis increases to 71 the number ofconfirmed Crohn’s disease susceptibility loci. Nat Genet 42(12):1118–1125.

11. Ferreira MA, et al.; Australian Asthma Genetics Consortium (2011) Identification ofIL6R and chromosome 11q13.5 as risk loci for asthma. Lancet 378(9795):1006–1014.

12. Hu X, et al. (2011) Integrating autoimmune risk loci with gene-expression dataidentifies specific pathogenic immune cell subsets. Am J Hum Genet 89(4):496–506.

13. Sawcer S, et al.; International Multiple Sclerosis Genetics Consortium; Wellcome TrustCase Control Consortium 2 (2011) Genetic risk and a primary role for cell-mediatedimmune mechanisms in multiple sclerosis. Nature 476(7359):214–219.

14. Fang D, et al. (2002) Dysregulation of T lymphocyte function in itchy mice: A role forItch in TH2 differentiation. Nat Immunol 3(3):281–287.

15. Parravicini V, Field AC, Tomlinson PD, Basson MA, Zamoyska R (2008) Itch-/- alphabetaand gammadelta T cells independently contribute to autoimmunity in Itchy mice.Blood 111(8):4273–7282.

16. Venuprasad K, et al. (2006) Convergence of Itch-induced ubiquitination with MEKK1-JNK signaling in Th2 tolerance and airway inflammation. J Clin Invest 116(4):1117–1126.

17. Liu YC (2007) The E3 ubiquitin ligase Itch in T cell activation, differentiation, andtolerance. Semin Immunol 19(3):197–205.

18. Mueller DL, Jenkins MK, Schwartz RH (1989) Clonal expansion versus functional clonalinactivation: A costimulatory signalling pathway determines the outcome of T cellantigen receptor occupancy. Annu Rev Immunol 7:445–480.

19. Heissmeyer V, et al. (2004) Calcineurin imposes T cell unresponsiveness through tar-geted proteolysis of signaling proteins. Nat Immunol 5(3):255–265.

20. Huang H, et al. (2010) K33-linked polyubiquitination of T cell receptor-zeta regulatesproteolysis-independent T cell signaling. Immunity 33(1):60–70.

21. Ramos-Hernandez N, et al. (2013) Ndfip1 enforces a requirement for CD28 cos-timulation by limiting IL-2 production. J Immunol 191(4):1536–1546.

22. Shembade N, et al. (2008) The E3 ligase Itch negatively regulates inflammatory sig-naling pathways by controlling the function of the ubiquitin-editing enzyme A20. NatImmunol 9(3):254–262.

23. Scharschmidt E, Wegener E, Heissmeyer V, Rao A, Krappmann D (2004) Degradationof Bcl10 induced by T-cell activation negatively regulates NF-kappa B signaling. MolCell Biol 24(9):3860–3873.

24. Venuprasad K, et al. (2008) The E3 ubiquitin ligase Itch regulates expression oftranscription factor Foxp3 and airway inflammation by enhancing the function oftranscription factor TIEG1. Nat Immunol 9(3):245–253.

25. Beal AM, Ramos-Hernández N, Riling CR, Nowelsky EA, Oliver PM (2012) TGF-β in-duces the expression of the adaptor Ndfip1 to silence IL-4 production during iTreg celldifferentiation. Nat Immunol 13(1):77–85.

26. Li B, Tournier C, Davis RJ, Flavell RA (1999) Regulation of IL-4 expression by thetranscription factor JunB during T helper cell differentiation. EMBO J 18(2):420–432.

27. Szremska AP, et al. (2003) JunB inhibits proliferation and transformation in B-lym-phoid cells. Blood 102(12):4159–4165.

28. Fang TC, et al. (2007) Notch directly regulates Gata3 expression during T helper 2 celldifferentiation. Immunity 27(1):100–110.

29. Amsen D, et al. (2007) Direct regulation of Gata3 expression determines the T helperdifferentiation potential of Notch. Immunity 27(1):89–99.

30. Tu L, et al. (2005) Notch signaling is an important regulator of type 2 immunity. J ExpMed 202(8):1037–1042.

31. Dalton HE, et al. (2011) Drosophila Ndfip is a novel regulator of Notch signaling. CellDeath Differ 18(7):1150–1160.

32. Cornell M, et al. (1999) The Drosophila melanogaster Suppressor of deltex gene,a regulator of the Notch receptor signaling pathway, is an E3 class ubiquitin ligase.Genetics 152(2):567–576.

33. Wilkin MB, et al. (2004) Regulation of notch endosomal sorting and signaling byDrosophila Nedd4 family proteins. Curr Biol 14(24):2237–2244.

34. Chastagner P, Israël A, Brou C (2008) AIP4/Itch regulates Notch receptor degradationin the absence of ligand. PLoS ONE 3(7):e2735.

35. Howitt J, et al. (2012) Ndfip1 regulates nuclear Pten import in vivo to promoteneuronal survival following cerebral ischemia. J Cell Biol 196(1):29–36.

36. Sang Q, et al. (2006) Nedd4-WW domain-binding protein 5 (Ndfip1) is associated withneuronal survival after acute cortical brain injury. J Neurosci 26(27):7234–7244.

37. Daley SR, Hu DY, Goodnow CC (2013) Helios marks strongly autoreactive CD4+ T cellsin two major waves of thymic deletion distinguished by induction of PD-1 or NF-κB.J Exp Med 210(2):269–285.

38. Heng TS, Painter MW; Immunological Genome Project Consortium (2008) The Im-munological Genome Project: Networks of gene expression in immune cells. NatImmunol 9(10):1091–1094.

39. Liston A, Lesage S, Wilson J, Peltonen L, Goodnow CC (2003) Aire regulates negativeselection of organ-specific T cells. Nat Immunol 4(4):350–354.

40. Strasser A (2005) The role of BH3-only proteins in the immune system. Nat Rev Im-munol 5(3):189–200.

41. Teh CE, Daley SR, Enders A, Goodnow CC (2010) T-cell regulation by casitas B-lineagelymphoma (Cblb) is a critical failsafe against autoimmune disease due to autoimmuneregulator (Aire) deficiency. Proc Natl Acad Sci USA 107(33):14709–14714.

42. Anderson MS, et al. (2002) Projection of an immunological self shadow within thethymus by the aire protein. Science 298(5597):1395–1401.

43. Hoyne GF, et al. (2011) Visualizing the role of Cbl-b in control of islet-reactive CD4 Tcells and susceptibility to type 1 diabetes. J Immunol 186(4):2024–2032.

44. Silva DG, et al. (2011) Anti-islet autoantibodies trigger autoimmune diabetes in thepresence of an increased frequency of islet-reactive CD4 T cells. Diabetes 60(8):2102–2111.

45. Gett AV, Hodgkin PD (1998) Cell division regulates the T cell cytokine repertoire,revealing a mechanism underlying immune class regulation. Proc Natl Acad Sci USA95(16):9488–9493.

46. Gudmundsdottir H, Wells AD, Turka LA (1999) Dynamics and requirements of T cellclonal expansion in vivo at the single-cell level: Effector function is linked to pro-liferative capacity. J Immunol 162(9):5212–5223.

47. Barron L, Knoechel B, Lohr J, Abbas AK (2008) Cutting edge: Contributions of apo-ptosis and anergy to systemic T cell tolerance. J Immunol 180(5):2762–2766.

48. Oldstone MB (1987) Molecular mimicry and autoimmune disease. Cell 50(6):819–820.49. Sollid LM, Jabri B (2011) Celiac disease and transglutaminase 2: A model for post-

translational modification of antigens and HLA association in the pathogenesis ofautoimmune disorders. Curr Opin Immunol 23(6):732–738.

50. Wells AD, et al. (1999) Requirement for T-cell apoptosis in the induction of peripheraltransplantation tolerance. Nat Med 5(11):1303–1307.

51. Mittrücker HW, Shahinian A, Bouchard D, Kündig TM, Mak TW (1996) Induction ofunresponsiveness and impaired T cell expansion by staphylococcal enterotoxin B inCD28-deficient mice. J Exp Med 183(6):2481–2488.

52. Kündig TM, et al. (1996) Duration of TCR stimulation determines costimulatory re-quirement of T cells. Immunity 5(1):41–52.

53. Kearney ER, et al. (1995) Antigen-dependent clonal expansion of a trace populationof antigen-specific CD4+ T cells in vivo is dependent on CD28 costimulation and in-hibited by CTLA-4. J Immunol 155(3):1032–1036.

54. Khoruts A, Mondino A, Pape KA, Reiner SL, Jenkins MK (1998) A natural immuno-logical adjuvant enhances T cell clonal expansion through a CD28-dependent, in-terleukin (IL)-2-independent mechanism. J Exp Med 187(2):225–236.

55. Hernández J, Aung S, Marquardt K, Sherman LA (2002) Uncoupling of proliferativepotential and gain of effector function by CD8(+) T cells responding to self-antigens.J Exp Med 196(3):323–333.

56. Townsend SE, Goodnow CC (1998) Abortive proliferation of rare T cells induced bydirect or indirect antigen presentation by rare B cells in vivo. J Exp Med 187(10):1611–1621.

57. Curtsinger JM, Lins DC, Mescher MF (2003) Signal 3 determines tolerance versus fullactivation of naive CD8 T cells: Dissociating proliferation and development of effectorfunction. J Exp Med 197(9):1141–1151.

58. Ben-Sasson SZ, et al. (2009) IL-1 acts directly on CD4 T cells to enhance their antigen-driven expansion and differentiation. Proc Natl Acad Sci USA 106(17):7119–7124.

59. Vella AT, Dow S, Potter TA, Kappler J, Marrack P (1998) Cytokine-induced survival ofactivated T cells in vitro and in vivo. Proc Natl Acad Sci USA 95(7):3810–3815.

60. Yoshimura A, Muto G (2011) TGF-β function in immune suppression. Curr Top Mi-crobiol Immunol 350:127–147.

61. Altin JA, et al. (2011) Decreased T-cell receptor signaling through CARD11 differen-tially compromises forkhead box protein 3-positive regulatory versus T(H)2 effectorcells to cause allergy. J Allergy Clin Immunol 127(5):1277–1285.

62. Quah BJ, Parish CR (2010) The use of carboxyfluorescein diacetate succinimidyl ester(CFSE) to monitor lymphocyte proliferation. J Vis Exp (44):e2259.

2074 | www.pnas.org/cgi/doi/10.1073/pnas.1322739111 Altin et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 8,

202

0