Aire’s plant homeodomain(PHD)-2 is criticalfor induction of immunological toleranceSiyoung Yang1, Kushagra Bansal1, Jared Lopes2, Christophe Benoist3, and Diane Mathis3

Division of Immunology, Department of Microbiology and Immunobiology, Harvard Medical School, Boston, MA 02115

Contributed by Diane Mathis, December 18, 2012 (sent for review October 24, 2012)

Aire impacts immunological tolerance by regulating the expres-sion of a large set of genes in thymic medullary epithelial cells,thereby controlling the repertoire of self-antigens encountered bydifferentiating thymocytes. Both humans and mice lacking Airedevelop multiorgan autoimmunity. Currently, there are few mo-lecular details on how Aire performs this crucial function. The moreamino-terminal of its two plant homeodomains (PHDs), PHD1,helps Aire target poorly transcribed loci by “reading” the methyla-tion status of a particular lysine residue of histone-3, a process thatdoes not depend on the more carboxyl-terminal PHD-2. This studyaddresses the role of PHD2 in Aire function by comparing the be-havior of wild-type and PHD2-deleted Aire in both transfected cellsand transgenic mice. PHD2 was required for Aire to interact withsets of protein partners involved in chromatin structure/binding ortranscription but not with those implicated in pre-mRNA process-ing; it also was not required for Aire’s nuclear translocation or re-gional distribution. PHD2 strongly influenced the ability of Aire toregulate the medullary epithelial cell transcriptome and so wascrucial for effective central tolerance induction. Thus, Aire’s twoPHDs seem to play distinct roles in the scenario by which it assuresimmunological tolerance.

negative selection | protein–protein interactions | thymus |transcription factor

An important facet of immunological tolerance is the elimi-nation of potentially self-reactive T cells during their dif-ferentiation in the thymus (reviewed in ref. 1). The transcriptionfactor, Aire, a member of the zinc-finger family, is responsiblefor purging the T-cell repertoire of a broad swath of self-reactivespecificities (reviewed in refs. 2 and 3). Consequently, bothhumans and mice with a mutation in the gene encoding Airedevelop multiorgan autoimmune disease. Aire is expressed ina subset of thymic MECs, where it regulates the transcription ofa large set of genes, in particular loci encoding peripheral-tissueantigens (PTAs). Peptides derived from these PTAs are loadedonto major histocompatibility complex (MHC) molecules anddisplayed at the medullary epithelial cell (MEC) surface, wherethey are encountered by differentiating T cells percolating throughthe thymus. Those thymocytes whose T-cell receptors avidly rec-ognize MEC MHC:peptide complexes undergo clonal deletion,thereby preserving immunological tolerance (4–8).The molecular mechanisms underlying Aire’s control of PTA

gene expression are only vaguely understood at present (2, 3). Itseems not to operate like a traditional transcription factor, bindingto a promoter and/or enhancer element and inducing transcriptinitiation, but instead to somehow release RNA polymerase(RNA-Pol)II pausing and promote transcript elongation (9, 10).Aire targets and regulates genes as a component of one or morelarge multiprotein complexes (11–13). Its numerous associatesinclude sets of proteins involved in nuclear transport, chromatinbinding/structure, transcription, and pre-mRNA processing (12).The geographical features of these Aire-containing complexesremain to be elucidated—i.e., which protein(s) Aire binds di-rectly to, which regions of Aire are responsible for these asso-ciations, and which of the interactions have what functionalconsequences.

Among Aire’s structural domains are two plant homeodomains(PHDs) (Fig. 1A), which generally function in protein:proteininteractions, notably as histone “readers” or as docking domainsinvolved in stabilization of multiprotein complexes in variousstages of transcription (14). The more amino-terminal PHD1 isa functionally critical region of Aire, acting as a reader of histone-3 molecules unmethylated at the lysine-4 residue (H3K4me0) andthereby promoting targeting to transcriptionally dormant genes(15–20). PHD1 has also been reported to be a vital link in Aire’sinteraction with DNA–PKcs (20) and to harbor E3 ubiquitin–li-gase activity essential for transcriptional induction (21), althoughthe NMR solution structure and subsequent functional assess-ments did not support this latter role (22). The function of Aire’smore carboxyl-terminal PHD2 remains an enigma, with argu-ments both for and against a critical role in transcriptionaltransactivation (11, 13, 16, 21, 23–25).The focus of this report is the function of Aire–PHD2. Does

removal of this domain compromise Aire’s access to and locali-zation within the nucleus? Its interaction with other proteins? Itsregulation of PTA gene expression in thymic MECs? Its controlover autoimmunity? These questions are addressed throughanalyses of cells transfected with and mice transgenic for an Airegene devoid of PHD2.

ResultsWithout Its PHD2, Aire No Longer Interacts with a Subset of ItsPartners. As a first approach to elucidating the function of thePHD2 of Aire, we investigated whether it is required for Aire’sinteractions with its known structural associates. Thus, we gen-erated an expression construct encoding a FLAG-tagged, PHD2-deleted form of Aire (Aire-ΔPHD2) and an analogous constructspecifying FLAG-tagged, wild-type Aire (Aire-WT), and wecompared the outcome of transfecting each into 293T cells,which are widely used for analyses of protein:protein (includingAire) interactions (e.g., refs. 12 and 13). Deletion of PHD2 hadno evident effect on Aire’s localization in the nucleus or on itsspeckled disposition therein (Fig. 1B).Experiments entailing Aire-targeted coimmunoprecipitation

(co-IP) followed by mass spectrometry, as well as multiple RNAinterference-mediated knockdown approaches, have identifieda large set of proteins that associate with Aire (e.g., refs. 12 and13). These proteins were classified into four major functionalclasses: nuclear transport, chromatin binding/structure, transcrip-tion, and pre-mRNA processing (12) (Fig. 1C). We chose repre-sentative members of each class (except for the nuclear transportgroup, given that nuclear localization was unaffected by deletion

Author contributions: S.Y., K.B., C.B., and D.M. designed research; S.Y., K.B., and J.L. per-formed research; S.Y., K.B., C.B., and D.M. analyzed data; and S.Y., K.B., C.B., and D.M.wrote the paper.

The authors declare no conflict of interest.1S.Y. and K.B. contributed equally to this work.2Present address: Alkermes, PLC, Waltham, MA 02451.3To whom correspondence may be addressed. E-mail: cbdm@hms.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222023110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1222023110 PNAS Early Edition | 1 of 6

IMMUNOLO

GY

mailto:cbdm@hms.harvard.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222023110/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222023110/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1222023110

of PHD2), and assayed their interaction with Aire by co-IPexperiments on 293T cells transfected with either the Aire-WT orAire-ΔPHD2 construct. Interestingly, tested members of the pre-mRNA processing group (SFRS3 and DDX5) still coimmuno-precipitated with Aire lacking PHD2, whereas all members of thechromatin binding/structure and transcription groups (DNA–PKcs, Ku80, PARP1, RNA-PolII, TOP2A, DSIF, P-TEFb,HEXIM1, and MCM6) coimmunoprecipitated poorly or not atall (Fig. 1 D and E). These results suggest that PHD2 has a spe-cific function in particular aspects of Aire’s role in regulatingtranscription.

Construction of Transgenic Mice Expressing Aire with the ΔPHD2Mutation. To address the function of Aire’s PHD2 in vivo, weused a described system (26) to generate lines of transgenicmice expressing either WT Aire (iA-WT) or PHD2-deleted Aire(iA-ΔPHD2) quasi-specifically in thymic MECs, on an Aire-knockout (Aire-KO) nonobese diabetic (NOD) background.Mice generated by using this system faithfully recapitulate thecritical features of Aire control of immunological tolerance andhave been used previously to examine the role of PHD1 in Aire’sactivities (19), as well as temporal aspects of Aire’s impact ontolerance (27).We first compared thymic epithelial cell compartments in the

transgenic mice. Aire-expressing cells are a component of the so-

calledMEChi subset, which displays low levels of the Ly51 markerand high levels of MHC class II molecules. As has been reported(28), Aire-KO mice had a significantly higher fraction and num-ber of MEChi cells than did Aire-WT animals; similarly, iA-ΔPHD2 mice had more MEChi cells than did iA-WT animals(Fig. 2 A–C). Both transgenic lines had a lower fractional rep-resentation of Aire+ cells in the MEChi compartment than didAire-WT mice, although levels of Aire expression were in-distinguishable in the three lines. However, most importantly,the expression of Aire was very similar in the two transgeniclines, whether assessed as % Aire+ of MEChi (Fig. 2D), as themean fluorescence intensity (MFI) of Aire in Aire+MEChi (Fig.2E), or by the distribution of Aire on thymus sections (Fig. 2F).All in all, the thymus of iA-ΔPHD2 mice appeared grosslynormal (including normal-looking thymocyte populations), buttheir increase inMEChi, similar to what is routinely seen in Aire-KO mice, portended a possible deficiency in Aire function.

Aire’s PHD2 Is Required for Most of Its Effects on MEC GeneExpression. Aire controls immunological tolerance in great partby regulating transcription of a large set of genes in thymicMECs, above all loci encoding PTAs. To investigate the role ofPHD2 in this function, we compared the transcriptomes ofMEChi isolated from various mouse lines. Fig. 3A presents a di-rect comparison of the impact of the WT vs. ΔPHD2 versions of

E

CDK9 (P-TEFb)SPT5 (DSIF)

DNA-PKcsKu80

RNA- PolIIPARP1

TOP2A

HEXIM1

DDX5

MCM6SFRS3

0 50 100 150 200 250

% IP ΔPHD2 vs WT

FLAG (Aire)

SPT5 (DSIF)

DNA-PKcs

Ku80

PARP1

TOP2A

HEXIM1

SFRS3

CDK9 (P-TEFb)

MCM6

DDX5

IP: anti-FLAGDC WT Δ PHD2

Input

RNA-PolII

C WT Δ PHD2

C

SFRS1

SFRS2

DDX17

Gemin5

snrpB

PABPC1

SFRS3

NOP56

EFTUD2

MYBBPsnrpD3

DDX5

Cq1BP

SMC1

MCM2

MSH2

MCM5

NASP

RAD21

TRIM28

MSH6

TOP2A

RNA-PolII

DNA-PKcs

DSIF

PARP1

P-TEFbHEXIM1

FACT

PCNA

RuvBL2

MYBBP

MCM6

SMC3

LMNB1

CHD6

Histone3

Histone2ac

Histone2bo

AIRE

transcription

pre-mRNA processing

chromatin binding/structure

IPO7

NUP93

XPO1

XPOT

KPNB1

RanBP2

RanBP9

nuclear transport

Ku80

WT ΔPHD2AireDAPI

BA

CARD/HSR SAND PHD1 PHD2PRRNLS

1 106 182 282 298 345 434 475

Fig. 1. Aire’s PHD2 is required for interaction with some, but not all, of its partners. (A) Domain structure of mouse Aire. Numbers represent the amino acidresidues, and boxes represent the functional domains. HSR, homogenously staining region; NLS, nuclear localization signal; SAND, Sp100, Aire, NucP41/75,DEAF1 domain; PRR, proline-rich region (3). (B) Representative microscopic images of HEK-293 cells expressing Aire-WT (Left) or Aire-ΔPHD2 (Right). Nucleiwere counterstained with DAPI. (C) Classification of Aire partners into functional groups, as per ref. 12. (D) Co-IP of known Aire partners. HEK-293 cells,transfected with vector alone (C) or with the Aire-WT or -ΔPHD2 construct were lysed, and the nuclear extracts were immunoprecipitated with anti-FLAG Ab,followed by immunoblotting for the indicated proteins. (E) Summary data from four experiments. Quantitative analysis of band intensities was performed byusing Fujifilm Multi Gauge (Version 2).

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222023110 Yang et al.

www.pnas.org/cgi/doi/10.1073/pnas.1222023110

Aire on MEC transcription in the absence of Aire, i.e., a foldchange/fold change (FC/FC) plot for iA-WT vs. Aire-KO vis-à-vis iA-ΔPHD2 vs. Aire-KO. Clearly, deletion of PHD2 greatlycompromised Aire function because most of the transcripts in-duced by Aire-WT were only slightly up-regulated by Aire-PHD2.However, the fact that the cloud of Aire-induced transcripts wastilted away from the x-axis suggested that the mutation was lessdramatic then the knockout of Aire. This finding was particularlytrue for a small subset of highly induced transcripts highlightedin Fig. 3B. In brief, then, PHD2 is required for most of Aire’scontrol over gene expression in MECs.

Aire-PHD2 Is Necessary for Imprinting Immunological Tolerance. Twoassays were used to evaluate the importance of PHD2 for Aire’sinfluence on central tolerance, directly comparing cohorts ofAire-WT, Aire-KO, iA-WT, and iA-ΔPHD2 mice. First, at 14 wkof age (or when 15–20% of body weight had been lost), micewere killed, selected organs were removed, and hematoxylin andeosin (H+E) histology was performed. In general, both forfemales and males, iA-ΔPHD2 and Aire-KO mice presenteda very similar picture, whether the criterion evaluated was thenumber and type of tissues infiltrated (Fig. 4A), average tissue

infiltration score (Fig. 4B), or histological presentation (Fig. 4Cand Fig. S1A). Second, we assayed the range of autoantibody(autoAb) specificities by Western blotting whole-tissue extractsfrom organs known to be autoAb targets in Aire-KO mice onthe NOD background (eye, lung, pancreas, and stomach) andprobing them with sera from the different lines of mice (Fig. 4Dand Fig. S1B). Profiles of serum autoAbs were very similar in iA-ΔPHD2 and Aire-KO individuals, whether male or female. Inshort, then, Aire requires its PHD2 domain to effectively imposeimmunological tolerance.

DiscussionAire has two PHDs. Several lines of evidence argue that PHD1binds preferentially to unmethylated H3K4 residues—a mark ofpoorly transcribed chromatin—and is thereby an element of themechanism by which Aire chooses its gene targets (15–17, 19,29). PHD2’s function has been less studied and thus remains inquestion. Early results on Aire-transfected cell lines suggesteda role in transcriptional transactivation (11, 21, 25), although thisresult proved not to be a general observation (16). In fact,Xenopus Aire lacks a PHD2, and chicken Aire has a substantiallytruncated version, casting doubt on a required function (30).Here we demonstrated that the PHD2 of Aire does indeed playa critical role in its control of thymic MEC gene transcriptionand, as a consequence, in immunological tolerance.Three major points emerge from our studies. First, PHD2 is

not required for Aire to localize to the nucleus or for it toconcentrate within punctate structures therein. This finding issomewhat surprising given a recent report that a point mutation

EB C D0.003

0.028

0

5

10

15

20

WT

KO

iA-W

TiA

- ΔPH

D2

% M

EC

hi o

f CD

45- c

ells

F iA -WT iA-ΔPHD2Aire DAPI

0.001

0.001

0

10

20

30

# M

EC

hi p

er t

hym

us (x

103 )

WT

KO

iA-W

TiA

-ΔPH

D2

0

20

40

600.0001

% A

ire+

of M

EC

hi n.s

WT

iA-W

TiA

- ΔPH

D2

10

20

0

n.s

Aire

MFI

of A

ire+

ME

Chi

(x10

3 )

WT

iA-W

TiA

-ΔPH

D2

iA-WTWT

Ly51

MH

C c

lass

II

A KOiA-ΔPHD29.03 9.11 12.5 14.2

Fig. 2. Aire-expressing MEChi compartments in the different mouse lines.(A) Representative flow-cytometric analyses of CD45− thymic stromal cellsfrom three independent experiments. MEChi was gated as Ly51-/loMEC classIIhi;numbers indicate the percentage of MEChi cells. (B) Summary data for MEChi

as a fraction of CD45− cells (n = 3). P values are from Student’s t test. (C)Summary data for total numbers of MEChi per thymus. (D) Summary data forfraction of Aire+ cells within the MEChi fraction. (E) Summary data for AireMFI in Aire+MEChi cells. (F) Aire staining on 5-μm frozen sections of thymifrom 5-wk-old iA-WT (Left) or iA-ΔPHD2 (Right) mice, counterstained withDAPI. Magnification: 100×.

A

1

10

0.2

1 5 10 300.2

Spt1

Mup7

Mup1

Mup2

Mup4

FC iA-WT vs KO

FC iA

-ΔP

H2

vs K

O

B Aire KO/ΔPHD2 dependent gene listsoitaROK/TWCFemosomorhCeneG

Spt1 77.906 6.061 12.854Mup7 41.572 3.333 12.472Mup2 31.286 3.165 9.819Mup1 29.535 3.367 8.772Ly6g 29.595 5.618 5.268Ifit3 8.728 2.160 4.041Car4 9.340 2.364 3.951

Hsd3b6 17.525 4.762 3.680Calb2 11.627 3.521 3.302

BB287469 8.394 2.558 3.282Aspn 13.225 4.115 3.214Mup4 10.106 3.322 3.042

FC iA-WT/

iA-ΔPHD2

Chr 15 Chr 4 Chr 4 Chr 4

Chr 15 Chr 19 Chr 11 Chr 3 Chr 8 Chr 12 Chr 13Chr 4

Fig. 3. Aire-PHD2 is essential for most Aire-induced gene expression inMECs. (A) Comparing Aire-WTs with Aire-ΔPHD2’s impact on the Aire-lesstranscriptome. FC/FC plots of triplicate data comparing iA-WT with Aire-KO (x axis) and iA-ΔPHD2 with Aire-KO (y axis). Red dots indicate some ofthe transcripts highlighted in B. (B) List of genes impacted less differen-tially (more than threefold less) by the ΔPHD2 than the KO mutation.

Yang et al. PNAS Early Edition | 3 of 6

IMMUNOLO

GY

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222023110/-/DCSupplemental/pnas.201222023SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222023110/-/DCSupplemental/pnas.201222023SI.pdf?targetid=nameddest=SF1

within PHD2 (human Aire C446G) results in multiorgan auto-immune disease because it leads to aberrant Aire localization

confined to the cytoplasm and, consequently, a reduction inAire-dependent gene transcripts (13). It is likely that the point

Dgnuleye hcamotssaercnap

W W i i i i i Δ Δ Δ Δ Δ K K K K K W W i i i i i Δ Δ Δ Δ Δ K K K K K W W i i i i i Δ Δ Δ Δ Δ K K K K KW W i i i i i Δ Δ Δ Δ Δ K K K K K

C

A

retina stomach lung pancreasovary

WT

B

i : insulitis

WT

iA-WT

iA-ΔPHD2

KO

i i i i i

retinalung

salivary gland

stomachpancreas

liver

prostate/ovary

lachrymal gland

i i i i i

i i i i i

i i i

Male Female

iA-WT

iA-ΔPHD2

KO

0

1

2

3

4

5

0

1

2

3

4

5

WTiA-WT

Infil

tratio

n sc

ore

liver stomach salivarygland

ovary lachrymalgland

lung retina pancreas

Male

Female

iA-ΔPHD2KO

Infil

tratio

n sc

ore WTiA-WT

iA-ΔPHD2KO

liver stomach salivarygland

prostate lachrymalgland

lung retina pancreas

i i

Fig. 4. Critical role of Aire-PHD2 in imparting immunological tolerance. (A) Range of infiltrates. Histological analysis of WT, KO, iA-WT, and iA-ΔPHD2 micevia H+E staining of fixed tissues. Mice were taken at 14 wk of age or at 15–20% of body-weight loss. i, insulitis typical of Aire-positive NOD mice. (B) Eachorgan (n = 10) was scored as described in Materials and Methods. (C) Representative infiltrates for female mice. White arrows indicate areas of infiltration.Magnification: 20×, except for the stomach (5×). (D) AutoAbs in female mouse serum. Each lane of the multiscreen immunoblot shows autoAbs for an in-dividual mouse. W, Aire-WT; i, iA-WT; Δ, iA-ΔPHD2; K, Aire-KO. Dilutions, 1:25,000.

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222023110 Yang et al.

www.pnas.org/cgi/doi/10.1073/pnas.1222023110

mutation provokes abnormal folding and aggregation that doesnot occur when the entire domain is absent.Second, deletion of PHD2 greatly compromised Aire’s in-

teraction with certain of its partners (members of the chromatinstructure/binding and transcription groups) but not with others(members of the pre-mRNA processing and likely nuclear trans-port groups). This disjunction is reminiscent of our previousfinding that shRNA-driven knockdown of TOP2A (a member ofthe transcription group) did not destroy the interaction betweenAire and SFRS2 (a member of the pre-mRNA processing group)(12). Such observations suggest that Aire might participate inmultiple multiprotein complexes, differentially dependent on itsindividual domains. This notion is consistent with recent 3Dstructural data on a tandem Aire PHD1–PHD2 fragment, whichrevealed the two domains to be independent and noninteracting(13). This study also showed that point mutations in PHD1 thathave a major impact on PTA gene expression had relatively minoreffects (less than two-fold) on the interaction of Aire with proteinpartners involved in chromatin binding/structure or transcription(e.g., RuvBL2, SMC1A, DNA-PKcs, and Msh6).Third, removal of PHD2 greatly dampened Aire’s impact on

the MEC transcriptome. This effect is similar to that reported fora debilitating point mutation in PHD1 (19), and both alterationsled to a multiorgan autoimmune disease essentially indistinguish-able from that ofAire-KOmice.At present, we have noexplanationfor why a small subset of genes was notably less affected by theΔPHD2 than the Aire-null mutation. However, it may be relevantthat several of the genes highlighted as being less responsive to theformer than the latter are members of multigene families (e.g.,Mup) whose members are located together on the chromosome.Interestingly, some of the same loci were previously found to be lessaffected by a debilitating PHD1 point mutation than the Aire-nullmutation (19).On the basis of these findings (and others), a model worth

considering is that PHD1 helps target Aire to poorly transcribedgenes by preferentially binding to H3K4Me0 residues (or by pref-erentially avoiding H3K4Me3 residues) and that PHD2 serves asa docking site or adaptor for one or more multiprotein complexesthat act to release RNA–PolII pausing and promote transcrip-tional elongation. Once these processes have been set in motion,a different Aire-containing complex—one that does not depend onPHD2—may come into play to regulate mRNA processing.However, given this potentially central role for Aire-PHD2,

how can Xenopus and chickens be immunologically tolerant in itsabsence or truncation? One possibility is that in these two speciesAire-PHD1 and/or the typical binding partners of Aire-PHD2have evolved to permit PHD1 to take over PHD2’s criticalfunction(s). Another possibility is that these two species are in-deed compromised in their ability to purge the thymus of po-tentially autoreactive T cells, but that other tolerance mechanismscan come into play effectively enough to dampen autoimmunity.Indeed, we have only a very limited view of the relative degree towhich Xenopus and chickens succumb to autoimmunity vis-à-visother vertebrate species.

Materials and MethodsCell Culture and Transfection. HEK-293 cells were cultured in DMEM supple-mented with 10% (vol/vol) FBS, L-glutamate and pen/strep antibiotics andmaintained in a humidified atmosphere at 37 °C with 5% CO2. For trans-fection, the cells were counted and seeded in six-well or 10-cm tissue cultureplates and transfected with the specified plasmids using TransIT reagent(Mirus) according to the manufacturer’s instructions.

Abs and Plasmids. Abs recognizing the following protein or protein-tags werepurchased:FLAG-tag(SigmaAldrich);DNA-PKcsandSFRS3(Abnova);PARP1(CellSignaling Technology); Ku80, TOP2A, and HEXIM1 (Abcam); RNA-PolII, CDK9,SPT5,DDX5,andMCM6(SantaCruzBiotechnology); andAire (eBioscience).Anti-mouse and anti-rabbit immunoglobulin(Ig)G secondary Abs conjugated withhorseradish peroxide or fluorescein isothiocyanate (FITC) were purchased from

Jackson Immunoresearch. Mammalian expression plasmids carrying FLAG-tag-ged Aire-WT or Aire-ΔPHD2 cDNAs were constructed by in-frame insertion ofeach between the BglII and SalI sites of the pCMV-tag1 vector (Clontech).

Immunofluorescence. HEK-293 cells, seeded on coverslips and transiently trans-fected with Aire-WT or Aire-ΔPHD2 constructs, were fixed with 4% (vol/vol)paraformaldehyde for 10 min followed by permeabilization with 0.5% TritonX-100 [in phosphate-buffered saline (PBS), pH 7.4] for 10 min. Cells wereblocked with 1% BSA in Tween buffer (PBST; 0.05% Tween 20 in PBS, pH 7.4)for 1 h and stained with anti-FLAG Abs for 2 h, followed by incubation withFITC-conjugated anti-mouse IgG secondary Abs in the dark for 1 h at roomtemperature. Cells were counterstained with DAPI for visualization of nuclei.Coverslips with cells were mounted on a slide with fluoromount G, and im-munofluorescent images were acquired by a fluorescence microscope (ZeissAxio Imager M1) equipped with filters matching the spectral excitation andemission characteristics of DAPI and FITC. For frozen sections, thymi from 5-wk-old mice were washed and fixed in 4% (vol/vol) paraformaldehyde. Sectionswere stained by using anti-mouse Aire (5H12 clone) and counterstained withDAPI. Immunofluorescence was visualized by laser scanning confocal micros-copy (Olympus).

Immunoprecipitation. HEK-293 cells, transfected with empty vector (C) or withan Aire-WT or Aire-ΔPHD2 construct for 48 h, were harvested and lysed ina hypotonic lysis buffer [0.05% Nonidet P-40, 10 mM Hepes, 1.5 mM MgCl2,10 mM KCl, 5 mM EDTA (30 × 106 cells per mL)] plus complete proteaseinhibitor mixture (Roche), pH 7.4, and incubated on ice for 15 min. Cellnuclei were separated from the cytosolic fraction and incubated on ice for1 h in a native nuclear extraction buffer [50 mM Bis-Tris, 750 mM 6-amino-caproic acid, 3 mM CaCl2, 10% Glycerol, EDTA-free complete protease in-hibitor mixture (Roche) and micrococcal nuclease (Nuclease S7; Roche), pH7.4, (60 × 106 cells per mL)]. Nuclear extracts were incubated with 20 μL ofProtein G Sepharose beads (GE Healthcare) conjugated to anti-FLAG Abs(Sigma) for 3 h with rotation at 4 °C. Beads were washed three times withice-cold PBS containing 0.05% Nonidet P-40, and once with pure ice-coldPBS. Immunoprecipitated proteins were eluted by boiling in sample bufferfor 15 min and separated on 10% SDS/PAGE followed by transfer of pro-teins to polyvinylidene difluoride membranes (Millipore). Membranes wereblocked in PBST buffer containing 5% (wt/vol) nonfat dried milk and probedwith primary Abs overnight at 4 °C. After a wash with PBST, membraneswere incubated with secondary Abs linked to horseradish peroxidase. Theblots were then developed with an enhanced chemiluminescence detectionsystem (Thermo Scientific) as per the manufacturer’s instructions.

Mice. NOD/Lt J (Aire-WT) mice were originally purchased from the JacksonLaboratory. Aire-KO and iA-WT mice on the NOD genetic background havebeen described (19, 27). An Aire cDNA carrying PHD2-domain deletion (aminoacids 434–475) was constructed by QuikChangeII XL (Stratagene) and in-troduced into the ClaI site of the TOA construct (19). A fragment containingthe Aire-ΔPHD2 cDNA intron and polyadenylation site from the rabbit β-glo-bin gene were AatII and AfeI digested for the TOP construct, purified, andmicroinjected into NOD embryos as described (27). Progeny were screened bya PCR assay specific for the TOA and PHD2 transgene. All mice were bred andhoused under specific-pathogen-free conditions at the Harvard Medical SchoolCenter for Animal Resources and Comparative Medicine following In-stitutional Animal Care and Use Committee Protocol 2954.

Isolation of Thymic Epithelial Cells and Sorting of MECs. Thymi of 5-wk-oldindividual Aire WT, Aire-KO, iA-WT, or iA-ΔPHD2 mice were routinely used ina four-way comparison. Thymic lobes underwent a small cut and were agi-tated in RPMI to release thymocytes. The fragments were digested withcollagenase (Roche) and DNase (Sigma) for 30 min and then with collage-nase/dispase (Roche) for 30 min, as described (10). The released cells werestained for flow-cytometric sorting or analysis, by using a MoFlo (Dako) orLSRII (Becton Dickinson), respectively, as described (31). Data were analyzedwith FlowJo software (Tree Star).

Microarray Analysis. Total RNA was prepared from MECs of individual 5-wk-old Aire-WT, Aire-KO, iA-WT, or iA-ΔPHD2 mice. RNA amplification andmicroarray hybridization were performed as described (32). Briefly, RNA wasamplified using a T7 polymerase-based method, and cDNA was hybridizedto random primers. The cDNA was purified, fragmented, and terminally la-beled by using the Affymetrix terminal labeling kit. Labeled DNA was hy-bridized to 1.0 ST Affymetrix arrays, washed, stained, and scanned.Microarray data were normalized by robust multiarray average (33) andanalyzed by the multiplot module of GenePattern 3.4.

Yang et al. PNAS Early Edition | 5 of 6

IMMUNOLO

GY

Autoimmune Disease Monitoring. Individuals were killed at 14 wk of age orwhen they had lost 15–20% body weight relative to that of littermates, asdescribed (19, 34). Designated tissues were removed, fixed in 10% formalin,and embedded in paraffin. Tissue sections were stained with H+E, and in-filtration of various organs was scored. In general, scores of 0, 0.5, 1, 2, 3,and 4 indicated no, trace, mild, moderate, or severe lymphocytic infiltration,and complete destruction, respectively. For retinal degeneration, scores wereas follows: 0, lesion present without any photoreceptor layer lost; 1, lesionpresent, but less than half of the photoreceptor layer lost; 2, more than halfof the photoreceptor layer lost; 3, entire photoreceptor layer lost without orwith mild outer nuclear layer attack; and 4, the entire photoreceptor layerand most of the outer nuclear layer destroyed. All of the infiltrated sampleswere scored blindly and independently by two investigators.

The appearance of autoAbs was monitored as described (34). Briefly,the indicated organs were removed from a 4-wk-old WT NOD mouse andhomogenized and centrifuged at 15,700 × g for 10 min. Supernatants

were mixed with protein sample buffer, extracts were size-fractionatedby 10% SDS/PAGE gel electrophoresis, gel-separated protein were trans-ferred to a poly(vinylidene fluoride) membrane, and the membrane wasblocked and incubated with a 1:25,000 dilution of serum from individ-ual mice. After washing with PBST, the bound Abs were reacted withhorseradish-peroxidase-conjugated anti-mouse IgG for 1 h and revealedwith an enhanced chemiluminescence reagent (Thermo Scientific) andautoradiography.

ACKNOWLEDGMENTS. We thank Drs. N. Fujikado, H. Yoshida, andE. Wakamatsu for insightful discussions; and A. Ortiz-Lopez and K. Hattorifor technical assistance. This work was supported by National Institutes ofHealth Grant AI088204 (to D.M. and C.B.). S.Y., K.B., and J.L. were supportedby fellowships from National Research Foundation of Korea Grant Fund NRF-2011-357-C00102, American Diabetes Association Grant 7-12-MN-51, and theCanadian Diabetes Association, respectively.

1. Hogquist KA, Baldwin TA, Jameson SC (2005) Central tolerance: Learning self-controlin the thymus. Nat Rev Immunol 5(10):772–782.

2. Peterson P, Org T, Rebane A (2008) Transcriptional regulation by AIRE: Molecularmechanisms of central tolerance. Nat Rev Immunol 8(12):948–957.

3. Mathis D, Benoist C (2009) Aire. Annu Rev Immunol 27:287–312.4. Derbinski J, Schulte A, Kyewski B, Klein L (2001) Promiscuous gene expression in

medullary thymic epithelial cells mirrors the peripheral self. Nat Immunol 2(11):1032–1039.

5. Gervois N, et al. (1992) Structure of the TCR-Ag-MHC complex. T lymphocytes:Structure, Functions, Choices, eds Celada F, Pernis B (Plenum, New York), pp 17–23.

6. Gronski MA, et al. (2004) TCR affinity and negative regulation limit autoimmunity.Nat Med 10(11):1234–1239.

7. Krimpenfort P, Ossendorp F, Borst J, Melief C, Berns A (1989) T cell depletion intransgenic mice carrying a mutant gene for TCR-beta. Nature 341(6244):742–746.

8. Lang P, et al. (2001) TCR-induced transmembrane signaling by peptide/MHC class II viaassociated Ig-alpha/beta dimers. Science 291(5508):1537–1540.

9. Oven I, et al. (2007) AIRE recruits P-TEFb for transcriptional elongation of target genesin medullary thymic epithelial cells. Mol Cell Biol 27(24):8815–8823.

10. Giraud M, et al. (2012) Aire unleashes stalled RNA polymerase to induce ectopic geneexpression in thymic epithelial cells. Proc Natl Acad Sci USA 109(2):535–540.

11. Halonen M, et al. (2004) APECED-causing mutations in AIRE reveal the functionaldomains of the protein. Hum Mutat 23(3):245–257.

12. Abramson J, Giraud M, Benoist C, Mathis D (2010) Aire’s partners in the molecularcontrol of immunological tolerance. Cell 140(1):123–135.

13. Gaetani M, et al. (2012) AIRE-PHD fingers are structural hubs to maintain the integrityof chromatin-associated interactome. Nucleic Acids Res 40(22):11756–11768.

14. Sanchez R, Zhou MM (2011) The PHD finger: A versatile epigenome reader. TrendsBiochem Sci 36(7):364–372.

15. Org T, et al. (2008) The autoimmune regulator PHD finger binds to non-methylatedhistone H3K4 to activate gene expression. EMBO Rep 9(4):370–376.

16. Koh AS, et al. (2008) Aire employs a histone-binding module to mediate immuno-logical tolerance, linking chromatin regulation with organ-specific autoimmunity.Proc Natl Acad Sci USA 105(41):15878–15883.

17. Chignola F, et al. (2009) The solution structure of the first PHD finger of autoimmuneregulator in complex with non-modified histone H3 tail reveals the antagonistic roleof H3R2 methylation. Nucleic Acids Res 37(9):2951–2961.

18. Ge Q, Bai A, Jones B, Eisen HN, Chen J (2004) Competition for self-peptide-MHCcomplexes and cytokines between naive and memory CD8+ T cells expressing thesame or different T cell receptors. Proc Natl Acad Sci USA 101(9):3041–3046.

19. Koh AS, Kingston RE, Benoist C, Mathis D (2010) Global relevance of Aire binding tohypomethylated lysine-4 of histone-3. Proc Natl Acad Sci USA 107(29):13016–13021.

20. �Zumer K, Low AK, Jiang H, Saksela K, Peterlin BM (2012) Unmodified histone H3K4and DNA-dependent protein kinase recruit autoimmune regulator to target genes.Mol Cell Biol 32(8):1354–1362.

21. Uchida J, et al. (2004) The innate mononuclear phagocyte network depletes B lym-phocytes through Fc receptor-dependent mechanisms during anti-CD20 antibodyimmunotherapy. J Exp Med 199(12):1659–1669.

22. Bottomley MJ, et al. (2005) NMR structure of the first PHD finger of autoimmuneregulator protein (AIRE1). Insights into autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED) disease. J Biol Chem 280(12):11505–11512.

23. Björses P, et al. (2000) Mutations in the AIRE gene: effects on subcellular location andtransactivation function of the autoimmune polyendocrinopathy-candidiasis-ecto-dermal dystrophy protein. Am J Hum Genet 66(2):378–392.

24. Pitkänen J, et al. (2000) The autoimmune regulator protein has transcriptionaltransactivating properties and interacts with the common coactivator CREB-bindingprotein. J Biol Chem 275(22):16802–16809.

25. Meloni A, Incani F, Corda D, Cao A, Rosatelli MC (2008) Role of PHD fingers andCOOH-terminal 30 amino acids in AIRE transactivation activity. Mol Immunol 45(3):805–809.

26. Witherden D, et al. (2000) Tetracycline-controllable selection of CD4(+) T cells: Half-life and survival signals in the absence of major histocompatibility complex class IImolecules. J Exp Med 191(2):355–364.

27. Guerau-de-Arellano M, Martinic M, Benoist C, Mathis D (2009) Neonatal tolerancerevisited: A perinatal window for Aire control of autoimmunity. J Exp Med 206(6):1245–1252.

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

29. Chakravarty S, Zeng L, Zhou MM (2009) Structure and site-specific recognition ofhistone H3 by the PHD finger of human autoimmune regulator. Structure 17(5):670–679.

30. Saltis M, et al. (2008) Evolutionarily conserved and divergent regions of the auto-immune regulator (Aire) gene: A comparative analysis. Immunogenetics 60(2):105–114.

31. Gray DHD, Abramson J, Benoist C, Mathis D (2007) Proliferative arrest and rapidturnover of thymic epithelial cells expressing Aire. J Exp Med 204(11):2521–2528.

32. Fu W, et al. (2012) A multiply redundant genetic switch ‘locks in’ the transcriptionalsignature of regulatory T cells. Nat Immunol 13(10):972–980.

33. Irizarry RA, et al. (2003) Exploration, normalization, and summaries of high densityoligonucleotide array probe level data. Biostatistics 4(2):249–264.

34. Jiang W, Anderson MS, Bronson R, Mathis D, Benoist C (2005) Modifier loci conditionautoimmunity provoked by Aire deficiency. J Exp Med 202(6):805–815.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1222023110 Yang et al.

www.pnas.org/cgi/doi/10.1073/pnas.1222023110

Supporting InformationYang et al. 10.1073/pnas.1222023110

gnuleyeW W i i i i i K K K K K W W i i i i i K K K K K

hcamotssaercnapW W i i i i i K K K K KW W i i i i i K K K K K

retina stomach lung pancreasprostateA

B

WT

iA-WT

iA- PHD2

KO

Fig. S1. Additional tolerance data on male mice. Figure is as per Fig. 4 C and D except that material from males was examined.

Yang et al. www.pnas.org/cgi/content/short/1222023110 1 of 1

www.pnas.org/cgi/content/short/1222023110