The Journal of Immunology

Homeodomain-Interacting Kinase 2, a Novel Autoimmune Regulator Interaction Partner, Modulates Promiscuous Expression in Medullary Thymic Epithelial Cells

Kristin Rattay,* Janine Claude,*,1 Esmail Rezavandy,* Sonja Matt,† Thomas G. Hofmann,† Bruno Kyewski,* and Jens Derbinski*,2

Promiscuous expression of a plethora of tissue-restricted Ags (TRAs) by medullary thymic epithelial cells (mTECs) plays an essential role in tolerance. Although the cellular mechanisms by which promiscuous (pGE) imposes T cell tolerance have been well characterized, the underlying molecular mechanisms remain poorly understood. The autoimmune regulator (AIRE) is to date the only validated molecule known to regulate pGE. AIRE is part of higher-order multiprotein complexes, which promote , elongation, and splicing of a wide range of target . How AIRE and its partners mediate these various effects at the molecular level is still largely unclear. Using a yeast two-hybrid screen, we searched for novel AIRE-interacting and identified the homeodomain-interacting protein kinase 2 (HIPK2) as a novel partner. HIPK2 partially colocalized with AIRE in nuclear bodies upon cotransfection and in human mTECs in situ. Moreover, HIPK2 phos- phorylated AIRE in vitro and suppressed the coactivator activity of AIRE in a kinase-dependent manner. To evaluate the role of Hipk2 in modulating the function of AIRE in vivo, we compared whole-genome gene signatures of purified mTEC subsets from TEC-specific Hipk2 knockout mice with control mice and identified a small set of differentially expressed genes. Unexpectedly, most differentially expressed genes were confined to the CD80lo mTEC subset and preferentially included AIRE-independent TRAs. Thus, although it modulates gene expression in mTECs and in addition affects the size of the medullary compartment, TEC-specific HIPK2 deletion only mildly affects AIRE-directed pGE in vivo. The Journal of Immunology, 2015, 194: 921–928.

he is the exquisite site of T cell development, Intrathymically presented self-Ags are of intrathymic and extra- a process that consists in the generation of a highly diverse thymic origin. Extrathymic self-Ags enter the thymus via the blood T TCR repertoire and subsequent selection (“quality con- or via immigrating DCs or are generated within the thymus by one trol”) of those TCRs (i.e., ), which are both self-MHC particular APC type, mTECs, by a process known as promiscuous restricted and self-tolerant (1). Intrathymic self-tolerance induc- gene expression (pGE). MTECs transcribe and translate .3000 so- tion () entails the probing of a diverse array of called tissue-restricted Ags (TRAs) (3–5). These TRAs are pre- self-MHC–restricted TCRs against a plethora of self-peptide/ sented to thymocytes either autonomously by mTECs themselves or MHC complexes displayed by various APC lineages and subsets are handed over to neighboring DCs, which, in turn, process and/or thereof. These include subsets of thymic dendritic cells (DCs), present these transferred Ags. Both mTECs and DCs concurrently cortical and medullary thymic epithelial cells (cTECs and mTECs, mediate the two major modes of self-tolerance imposition, namely, respectively), and thymic B cells (2). This diversity of thymic induction of (recessive tolerance) or fate diversion into APCs with each subset presumably displaying partly nonover- the regulatory T cell lineage (dominant tolerance) (2, 6). The scope lapping self-peptides (MHC ligandomes) ensures a maximal intra- of pGE represents essentially all tissues of the body, and thus ensures thymic representation of the immunological “self” of the body. a maximal coverage of central tolerance. Failure of pGE results in

*Division of Developmental Immunobiology, Tumor Immunology Program, German Address correspondence and reprint requests to Prof. Bruno Kyewski or Dr. Thomas G. Cancer Research Center, 69120 Heidelberg, Germany; and †Zellula¨re Seneszenz- Hofmann, Division of Developmental Immunobiology, Tumor Immunology Program, Gruppe, Deutsches Krebsforschungszentrum-Zentrum fur€ Molekulare Biologie German Cancer Research Center, Im Neuenheimer Feld 280, 69120 Heidelberg, Allianz, Deutsches Krebsforschungszentrum, 69120 Heidelberg, Germany Germany (B.K.) or Zellula¨re Seneszenz-Gruppe, Deutsches Krebsforschungszentrum- Zentrum fur€ Molekulare Biologie Allianz, Deutsches Krebsforschungszentrum, 1Current address: Pharma-Akademie, Munchen,€ Germany. 69120 Heidelberg, Germany (T.G.H.). E-mail addresses: b.kyewski@dkfz-heidelberg. 2Current address: Cytonet GmbH & Co. KG, Weinheim, Germany. de (B.K.) or [email protected] (T.G.H.) Received for publication October 24, 2014. Accepted for publication November 25, The online version of this article contains supplemental material. 2014. Abbreviations used in this article: Aire, autoimmune regulator; cTEC, cortical thymic This work was supported by the German Cancer Research Center, the Deutsche epithelial cell; DC, dendritic cell; DeSi-1, desumoylating isopeptidase-1; DKFZ, Forschungsgemeinschaft (Grant SFB1036 to T.G.H., Grant SFB 405 to B.K. and German Cancer Research Center; FRT, flippase recognition target; GO, gene ontol- J.D.), the Monika-Kutzner-Stiftung (to T.G.H.), and the European Research Council ogy; HIPK2, homeodomain-interacting protein kinase 2; ko, knockout; mTEC, med- (Grant ERC-2012-AdG to B.K.) ullary ; NB, nuclear body; pGE, promiscuous gene expression; PIAS1, protein inhibitor of activated STAT 1; PML, promyelocytic leukemia; qPCR, J.D., K.R., T.G.H., and B.K. designed the project; J.D., K.R., S.M., and J.C. per- quantitative PCR; SUMO, small ubiquitin-like modifier; TRA, tissue-restricted Ag; formed the experiments; E.R. provided technical assistance; J.D. and K.R. analyzed Y2H, yeast two-hybrid. the data; and K.R., T.G.H., and B.K. wrote the manuscript. The sequences presented in this article have been submitted to the Gene Expression Copyright Ó 2015 by The American Association of Immunologists, Inc. 0022-1767/15/$25.00 Omnibus database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE63432) under accession number GSE63432. www.jimmunol.org/cgi/doi/10.4049/jimmunol.1402694 922 HIPK2, A NOVEL AIRE INTERACTION PARTNER a severe multiorgan in mice and humans (7). incubation of GST and GST-AIRE with in vitro–translated, [35S]methio- How a specialized, terminally differentiated epithelial cell can nine radiolabeled HIPK2 or HIPK1. Subsequently, pull-downs were ana- override the tight regulation of tissue-specific gene expression lyzed by reducing SDS-PAGE and autoradiography as reported previously (19). is still poorly understood. Progress in understanding the molecular regulation of such Immunofluorescence stainings a diverse set of genes in one specialized cell type has been slow to H1299 cells were transfected to introduce cloned -AIRE and Flag- come. The thymic transcriptional cofactor autoimmune regulator HIPK2. For immunofluorescence analysis, cells were fixed in 4% parafor- (Aire) is to date still the only identified molecular regulator spe- maldehyde for 20 min at room temperature, followed by permeabilization for cifically dedicated to control a sizeable fraction of the promiscu- 5 min with 0.5% Triton X-100. The primary Abs anti-myc (1 mg/ml; mouse IgG1; Santa Cruz Biotech) and rabbit anti-flag (4 mg/ml; rabbit IgG; Sigma- ously expressed gene pool (7–9). It is currently thought that one Aldrich) were added and incubated for 1 h at room temperature. The major mode by which Aire targets such a wide array of genes is by secondary Ab staining with goat Alexa Fluor 488 anti-mouse (10 mg/ml; relieving stalled polymerase II, thus allowing full-length tran- Molecular Probes) and goat Alexa Fluor 594 anti-rabbit (10 mg/ml; scription (10). Other general transcription factors like Myc (11) Molecular Probes) was performed for 40 min. Hoechst 33342 stain- ing was used for nuclear staining. have been shown to also operate via elongation rather than initi- On human thymus samples, cryosections were performed, which were ation of transcription, although the target range of Aire seems to fixed in paraformaldehyde. Blocking was performed using goat serum. surpass that of other “classical” transcription factors (4, 5). Stalling AIRE was stained by using the primary Ab clone 6.1; HIPK2 was stained polymerase II, which binds genomewide to promoters, would be using an affinity-purified HIPK2 Ab from rabbit (19). Goat anti-mouse IgG- one way to silence transcription of TRAs in mTECs (and other cell Alexa 488 and goat anti-rabbit Cy3 were used as secondary reagents. Hoechst33342 served as a nuclear counterstain. types), which are not supposed to be expressed in these cells. This mechanism would, at least in part, explain the significant enrich- In vitro phosphorylation assay ment of TRAs among the targets of Aire. In addition, Aire promotes In vitro phosphorylation of bacterially expressed and purified GST-AIRE induction of double-strand breaks and splicing of nascent mRNA was performed using 6xHis-HIPK2 purified from E. coli essentially as transcripts (7–10, 12). Yet, there have to be other levels of Aire- described previously (20). In brief, 1.5 mg purified GST-AIRE was incu- dependent and -independent gene regulation, given the intricacies of bated with 100 ng 6xHis-HIPK2 protein in kinase reaction buffer con- taining g-[32P]-ATP as phosphate donor. In vitro phosphorylation reactions pGE, that is, the mosaic expression pattern of TRAs at the pop- were analyzed by SDS-PAGE and autoradiography as described previously ulation level and the TRA coexpression patterns at the single-cell (20). level (13–15). Obviously any advance in understanding the mo- Luciferase assay lecular regulation of pGE by Aire and beyond would help in un- derstanding how the mTEC-specific MHC ligandome is generated Gal4-reporter assays were essentially performed as described previously in the first place and importantly how dysregulation at the genetic (21). In brief, 300 ng of an expression vector coding for the DNA binding domain of Gal4 or encoding a Gal4-AIRE fusion protein was transfected in and epigenetic levels could undermine central tolerance, and thus 293 cells along with 500 ng Gal4-luciferase reporter gene, 50 ng RSV-lacZ explain autoimmune pathologies. (for normalization of the transfection efficiency), and either 750 ng Flag- Likewise, little is known about how Aire expression itself is HIPK2, Flag-HIPK2 K221A (kinase-deficient mutant), or empty Flag vector regulated along mTEC differentiation and which posttranslational as indicated. Total DNA amounts were kept equal in all transfections by modifications ensure the function of Aire as part of a multiprotein addition of empty Flag vector. Cells were harvested 24 h posttransfection, and analyzed and normalized using luminometric measurement as described transcriptional complex. One way to approach this complex issue is previously (21). the identification of Aire’s partners within the nucleus either by biochemical or functional screens. A series of studies using either Mice approach identified .40 factors binding either directly or indi- Hipk2 conditional knockout (ko) mice (Hipk2flox/flox; B6.CgHipk2tm1Tgh) rectly to Aire or interfering with its transactivation (7–9, 12, 16, were generated by Taconic Artemis GmbH (Ko¨ln, Germany). In brief, 17). Yet, the functional hierarchy of the various components exons 3 and 4 of Hipk2 were targeted. The targeted region is flanked with loxP sites and a selection marker flanked with flippase recognition target through which Aire acts in such a complex remains completely (FRT) sites. C57BL/6NTac embryonic stem cells were electroporated with unknown, and it is likely that the current list is still incomplete. the targeting vector, and positive embryonic stem cell clones were vali- In this study, we used a yeast two-hybrid (Y2H) screen to dated by Southern blot. Validated clones were injected into mouse blas- identify additional potential Aire interaction partners. We identified tocysts to generate chimeric mice. To obtain germline transmission and to homeodomain-interacting protein kinase 2 (HIPK2), a protein eliminate the selection marker, we bred the resulting chimeras to mice carrying a ubiquitously expressed Flp transgene (C57BL/6-Tg(CAG-Flpe) kinase involved in transcriptional regulation and cell fate decision 2Arte). Animals heterozygous for the transgene cassette were backcrossed as a novel partner. It directly binds to Aire, and the interaction has onto the C57BL6/N background. To generate epithelial cell–specific Hipk2 a functional impact on the cotranscriptional activity of Aire in ko mice (B6-Tg(Foxn1-cre)1Tbo Hipk2tm1Tgh), we crossed Hipk2flox/flox in vitro assays. Moreover, HIPK2 also influences pGE in mTECs mice with Foxn1-cre mice (B6-Tg(Foxn1-cre)1Tbo) (22). All breedings and cohort maintenance were performed in the central animal laboratory of in vivo, affecting transcription of both Aire-dependent and the DKFZ under approved conditions in accordance with the European -independent genes. Convention for the Protection of Vertebrate Animals used for Experimental and Other Specific Purposes and the German Legislation. Materials and Methods Genotyping of Hipk2 deletion Yeast two-hybrid screen The Hipk2 gene was analyzed in a three-primer PCR, using genomic The Yeast two-hybrid screen was performed by the Genome and Proteome DNA and the following primers: hipk2-1: 59-GAATTCTGGTTGACTCT- Core Facility at the German Cancer Research Center (DKFZ) as described CAGG-39; hipk2-2: 59- CTCATCCCAATGATCTTTGTCC-39; and hipk2- previously in Albers et al. (18). 3: 59- CCCAGCACAAATATTGCTCC -39. The PCR conditions were as GST pull-down assays follows: initial activation of the polymerase for 3 min at 94˚C, followed by 33 cycles of 30-s denaturation at 94˚C, annealing for 30 s at 72˚C and The GST-AIRE expression vector was generated using standard PCR-based extension for 45 s at 72˚C, and finally with a 5-min fill in step at 72˚C. A cloning procedure, and the AIRE cDNA sequence was confirmed by DNA band size of 174 bp was diagnostic for the wild-type allele, a band size of sequencing. GST-AIRE fusion protein and GST control protein were 402 bp for the floxed allele, and a band size of 295 bp for the deleted allele. expressed in Escherichia coli BL21 and purified as described using glu- Floxed Hipk2 mice were crossed with FoxN1 Cre mice to generate a thymic tathione Sepharose beads (19). GST-pull-down assays were performed by epithelial cell–specific ko for Hipk2. The Journal of Immunology 923 mTEC preparation ological duplicates for mTEChi and mTEClo fractions on the Hipk2ko and Hipk2wt background. Genes with a fold change of .2or,0.5 and with a Primary mTECs were isolated by sequential fractionated enzyme digestion p value #0.0005 were considered to be differentially expressed. Statistical as previously described (14). Thymi were cut into pieces and digested in t test calculation was performed in R. Chipster software was used for collagenase under magnetic stirring for 15 min at room temperature, fol- further data analysis. lowed by several digestion rounds in collagenase/dispase enzyme mix for TRAs were defined using the public database (http://symatlas.gnf.org) 25 min at 37˚C in a water bath under magnetic stirring. The collagenase/ (25). A gene was defined as a TRA if its expression was five times more dispase cell fractions were pooled and filtered through a 40-mm cell than the median expression over all tissues in less than five tissues (26). strainer. After digestion, the single-cell fraction was pre-enriched for 2 + The significance of TRA enrichment (p value) was calculated using x thymic stromal cells by depleting CD45 cells with anti-CD45 magnetic tests. beads and the AutoMACS (Miltenyi Biotec). (GO) analysis on the differentially expressed genes was Pre-enriched fractions were stained using the following Abs: 1) performed using GeneCodis (27–29). The p values were computed using anti–CD45-PE-Cy5 (clone 30-F11; BD), anti–EpCAM-A647 (G8.8 hybrid- the hypergenomic distribution method (for further information, see http:// oma), anti-CD80 (clone 16-10A1; BD Pharmingen), anti–Ly51-FITC (clone genecodis.dacya.ucm.es) in combination with the false discovery rate 6C3; BD Pharmingen). Dead cell staining was performed using propidium p value correction (30). iodide in a final concentration of 0.2 mg/ml. Cells were sorted on an Aria ΙΙ cell sorter from BD. mTECs were defined as CD452Ly512EpCAM+,of which the CD80low and the CD80high subsets were sorted separately; 2) Aire Results staining: anti–CD45-PerCP (clone 30-F11; BD Pharmingen), anti–EpCAM- HIPK2 interacts with AIRE, and both factors colocalize at A647 (G8.8 hybridoma), anti–CD80-PE (clone 16-10A1; BD Pharmingen), nuclear bodies anti–CDR1-Pacific Blue (cell culture supernatant (23), Pacific Blue Protein Labeling Kit from Life Technologies), anti–Aire-FITC (clone 5H12) (24), To identify novel AIRE-interacting proteins, we performed a Y2H Fixable Viability Dye eFluor 780 (eBioscience). screen. We used an AIRE deletion mutant lacking its potential auto- transactivating PHD2 domain as bait to screen a murine testis and Total RNA preparation total mouse embryo (E17) cDNA library. Note that testis (i.e., Total RNA of primary sorted mTECs was isolated and purified using the spermatogonia and spermatocytes) is one of the few tissues ex- High Pure RNA Isolation Kit (Roche Diagnostics) according to the man- pressing Aire at the mRNA and protein level and displaying pGE ufacturer’s protocol. For quantitative PCR (qPCR), the isolated RNA was reverse transcribed (31, 32). Using this approach, we identified cDNA clones encoding into cDNA using random primers and Superscript II Reverse Transcriptase Ubc9 (also called Ube2i), protein inhibitor of activated STAT 1 (Invitrogen) following the manufacturer’s protocol. (PIAS1), desumoylating isopeptidase-1 (DeSi-1), ZFP451, and HIPK1 as AIRE-interacting factors (Supplemental Fig. 1A–E). Quantitative real-time PCR Of note, PIAS1 (33) and Ubc9 (34) have been previously de- Quantitative real-time PCR was performed in a total volume of 20-ml scribed as AIRE binding proteins, testifying to the specificity of reactions using Power Sybr Green Mix (Applied Biosystem) and the GeneAmp our Y2H screen. 7300 (Applied Biosystem). Intron spanning primers were designed using Primer3 software. Reactions were performed in technical duplicates and We decided to focus on the HIPK family kinases as potential normalized to Actin expression and relative to total thymus cDNA using novel AIRE binding proteins, because, similar to what has been the dd cycle threshold method. reported for the AIRE protein (35), HIPKs localize to nuclear bodies (NBs), which overlap with promyelocytic leukemia (PML) Microarray analysis NBs (36). Because HIPK family members are highly conserved Isolated purified RNA samples were run on Illumina MouseWG-6 v2.0 showing a largely identical domain structure and amino acid Expression Bead Chip Sentrix arrays. Labeling of the samples and hybridization composition (37), we analyzed potential interactions of AIRE with were performed by the Genome and Proteome Core Facility of the DKFZ in two HIPK family members, HIPK1 and HIPK2. GST pull-down Heidelberg. The microarray values were quantile normalized followed by Limma assays revealed a weak interaction of AIRE with HIPK1 in vitro analysis to identify differentially expressed genes between wild-type and (Supplemental Fig. 1F). In contrast, a strong interaction of AIRE HIPK2-deficient animals. Each microarray analysis was performed in bi- and the HIPK family member HIPK2 was detected, and this in-

FIGURE 1. Interaction of HIPK2 and Aire in vitro and colocalization in situ. (A) In vitro interaction of AIRE and HIPK2. The GST-pull-down assay was performed with GST-Aire, GST only as negative control, and a 10% input control. The autoradiography for HIPK2 is shown at the top and the corre- sponding Coomassie staining on the bottom. (B) Colocalization of ectopically expressed AIRE and HIPK2. Immunohistochemistry staining of Myc-Aire (red) and HIPK2-GFP (green) in cotransfected H1299 cells indicating colocalization. Aire was detected by an anti-Myc Ab (red) along with nuclear stain (DAPI, blue). (C) Coexpression and colocalization of endogenously expressed AIRE and HIPK2 in human thymus. Endogenous colocalized expression of AIRE (red) and HIPK2 (green) in human thymus cryosections along with nuclear stain (DAPI, blue). Scale bars, 5 mm(B), 2 mm(C). 924 HIPK2, A NOVEL AIRE INTERACTION PARTNER teraction was comparable in strength with the interaction observed pression of wild-type HIPK2 resulted in a dose-dependent reduction between HIPK2 and its known partner protein (36) (Fig. 1A). of the AIRE-driven reporter gene activity, indicating that HIPK2 Thus, we focused our analysis on a potential interplay between represses AIRE-controlled transactivation, as has been previously AIRE and HIPK2. Because AIRE and HIPK2 are known to lo- reported for DAXX (38). Of note, this repressive effect of HIPK2 calize to NBs, we asked whether both factors might colocalize in required its kinase activity, as a kinase-deficient HIPK2 point mu- the cell nucleus. In fact, immunofluorescence staining and con- tant (HIPK2 K221A) failed to suppress the coactivator function of focal microscopy of ectopically expressed AIRE and HIPK2 AIRE (Fig. 2B). These results indicate that HIPK2 modulates revealed a partial colocalization of both factors in the nucleus and AIRE-regulated gene expression in a kinase-dependent fashion. at NBs (Fig. 1B). This subcellular localization is compatible Phenotype of TEC-specific HIPK2 deficiency in vivo with previous reports showing independently that each molecule localizes to PML NBs, a subtype of NBs (35, 36). To substantiate To assess whether the AIRE–HIPK2 interaction had an in vivo AIRE-HIPK2 colocalization under physiological conditions, we effect on the regulation of gene expression in mTECs in general performed immunofluorescence staining of endogenous AIRE and and the regulation of pGE in particular, we analyzed primary HIPK2 proteins in sections from human thymus. Remarkably, mouse mTECs. A conditional HIPK2 ko mouse line containing confocal microscopy identified colocalization of AIRE and HIPK2 LoxP-flanked Hipk2 alleles was generated and crossed with a in part of the mTECs (Fig. 1C), compatible with a functional in- mouse strain expressing Cre recombinase under control of the terplay between AIRE and HIPK2 under physiological conditions TEC-specific promoter of the FoxN1 gene to inactivate Hipk2 in ko in the thymus. Taken together, these results indicate that HIPK2 is both TEC lineages (Fig. 3A). MTECs were isolated from HIPK2 2 a novel AIRE-interacting protein that partially colocalizes with and mating wild-type control (floxed, Cre ) (referred to as control AIRE in the cell nucleus and in NBs. in the following) thymi by sequential enzymatic digestion fol- lowed by MACS and FACS. Efficient HIPK2 depletion in mTECs HIPK2 phosphorylates AIRE in vitro and represses its was confirmed by PCR on sorted mTECs (Fig. 3B). Total RNA transcriptional coactivator function in a kinase-dependent was isolated from sorted CD80lo and CD80hi mTECs from both manner HIPK2ko and control mice, and was analyzed by Illumina bead We next analyzed whether AIRE might be a substrate for HIPK2. microarrays (deposited in the Gene Expression Omnibus under To this end, we performed in vitro kinase assays in the presence of accession number GSE63432). g-[32P]-ATP using bacterially expressed, recombinant His-HIPK2 Within the biological duplicates of each experiment, the two and GST-AIRE as substrate. AIRE phosphorylation was analyzed different mTEC subsets, that is, mTECs CD80low versus mTECs by SDS-PAGE and subsequent autoradiography. His-HIPK2 was CD80high, clustered together (Fig. 4A). Furthermore, within the found to be strongly auto-phosphorylated (Fig. 2A), which is CD80high populations, there was a close correlation between the a prerequisite for its proper activity (20). Interestingly, the full- wild-type and HIPK2-ko mTECs of the biological duplicates. In length AIRE protein (and truncated AIRE products) was clearly the CD80low fraction, the variation between the biological dupli- phosphorylated by HIPK2. These data indicate that HIPK2 di- cates was found to be higher. The top 139 differentially regulated rectly phosphorylates AIRE in vitro. genes (filtered according to 3 SD = 99.7%) are displayed by the heat We next addressed the functional consequence of HIPK2- map (Fig. 4B). The biological significance cutoff was set to $2-fold mediated AIRE phosphorylation. AIRE is essential for regulat- differential expression in mTECs of ko mice compared with the ing pGE in the thymus (8). To address a potential impact of HIPK2 controls. To compare changes in gene expression levels between on AIRE-mediated transcriptional activity, we fused AIRE to Gal4 enriched immature and mature mTECs in Hipk2ko and control mice, and measured its transactivating activity using a Gal4-dependent the differential expression was analyzed and represented using luciferase reporter gene assay. Gal4-AIRE expression resulted in a scatterplots. As previously reported, CD80low and CD80high control profound increase in reporter gene activity, indicating that Gal4- mTECs clearly differed in their mRNA transcriptome (26) (genes AIRE functions as a transcriptional coactivator (Fig. 2B). Coex- plotted outside the red lines, indicating a $2-fold differential ex-

FIGURE 2. HIPK2 phosphorylates AIRE and regulates AIRE-dependent cotranscriptional activity. (A) AIRE serves as a HIPK2 substrate in vitro. In vitro phosphorylation assays were performed using recombinant His-HIPK2 and GST-AIRE proteins in the presence of radioactively labeled ATP. Phosphorylation was visualized by autoradiography. (Left panel) autoradiogram; (right panel) corresponding Coomassie Brilliant Blue staining. Asterisks indicate truncated Aire products. (B) HIPK2 regulates AIRE-controlled cotranscriptional activity in a kinase-dependent fashion. Luciferase reporter gene assays are shown. The luciferase activity of a 5xGal4-luciferase reporter was measured. The Gal4-Luciferasereporter was transfected as indicated with Gal4, Gal4-Aire, HIPK2, and a kinase-deficient HIPK2K221A point mutant. The means and SDs of three independent experiments are shown. The Journal of Immunology 925

FIGURE 3. Conditional targeting of the mouse HIPK2 gene locus. (A) The mouse HIPK2 gene locus consists of 15 exons; exon 2 contains the translation initiation site. Exons 3 and 4 were targeted and flanked by loxP sites. The selection marker (PuroR) has been flanked by FRT sites and was inserted into intron 2. The Neomycin cassette flanked by FRT sites was inserted into intron 4. The conditional ko allele was generated by Flp-mediated removal of the Puro and Neo cassettes. The constitutive ko allele was generated by in vivo Cre-mediated deletion of exons 3–4, resulting in a frameshift. (B) Genotyping on FACS sorted cTECs, mTECs, and CD45+ cells of the Hipk2ko and Hipk2flox control mice is shown. A band size of 174 bp denotes the wild-type allele, a band size of 402 bp the floxed allele, and a band size of 295 bp the deleted allele. pression). This was also the case for Hipk2ko mice (Fig. 4C, 4D). A only a limited effect on the steady-state transcriptome of mTECs comparison of the CD80lo mTECs between Hipk2ko and control irrespective of their maturation stage (Supplemental Table I). mice showed differential expression of a limited set of genes be- Interestingly, the array analysis revealed that HIPK2 absence tween these two groups (Fig. 4E). In contrast, in the CD80hi mTECs, resulted in downregulation of the majority of differentially expressed only seven genes were differentially expressed when comparing genes (not predicted by the suppressive effect of HIPK2 on the control versus Hipk2ko mice (Fig. 4F). Thus, HIPK2 deficiency had coactivating activity of Aire; Fig. 2). This observation was validated

FIGURE 4. Transcriptome analysis of mTECs from HIPK2 ko and control mice. (A) Dendrogram scheme of the array samples; clustering is indicated by Pearson correlation (height). (B) Hierarchical clustering of samples of mTEC low and mTEC high in HIPK2-deficient and control animals. Blue, downregulated; yellow, upregulated genes. Experiments were performed in biological duplicates. (C–F) Scatterplots of grouped differential analysis of mTEC low and mTEC high of HIPK2-deficient and control mice. Red lines indicate threshold of 2-fold expression changes. Data have been generated with Chipster and R software. 926 HIPK2, A NOVEL AIRE INTERACTION PARTNER

FIGURE 5. HIPK2 target genes in mTECs. Shown is the fold reduction of gene expression levels in mTECs of HIPK2 ko versus control animals. Black bars represent the array fold changes; white bars show the qPCR results for validation. Array experiments have been performed in biological duplicates and qPCR in biological triplicates. Shown are the means and SEM. by quantitative real-time PCR analysis of the top 10 gene hits of the Discussion array list, with both assays showing highly concordant results (Fig. 5). In this study, we used a protein–protein interaction screening approach to identify novel AIRE binding factors. We identified HIPK2 deficiency affects the expression of TRAs in mTECs five interaction partners, two of which, PIAS1 and Ubc9, had been Next, we asked whether the differentially expressed gene pool previously reported to interact with Aire (33, 34). We provide displayed features of pGE, that is, enrichment for Aire dependency evidence that HIPK2, a known regulator of cell fate and tran- and TRA content as insinuated by the in vitro effect of HIPK2 scription, is a novel Aire binding protein, adding to the growing on the function of AIRE (Fig. 2). The number of differentially list of Aire binding partners and Aire “allies” (7–9, 12, 16, 17). In expressed genes and TRAs were normalized to the relative rep- accordance with the physical interaction observed between both resentation of TRAs on Illumina bead array. We observed a sig- factors, AIRE colocalizes with HIPK2 in the cell nucleus in NBs nificant increase ($2-fold increase) in the representation of TRAs both upon ectopic expression and also under physiological con- among the differentially downregulated genes when compared ditions in situ in thymic sections. In addition, HIPK2 regulates the with the total gene pool (Fig. 6A). transactivating function of AIRE in a kinase-dependent fashion. Further, we investigated whether HIPK2 deficiency had a se- Because our data indicate that AIRE is a HIPK2 substrate, which lective effect on Aire-dependent genes. We therefore compared the is directly phosphorylated by HIPK2 in vitro, HIPK2 presumably number of differentially downregulated Aire-dependent and -in- regulates the transcriptional activity of AIRE through phosphor- dependent genes with their normal distribution on the Illumina bead ylation. array. This calculation gives the relative extent of Aire-dependently HIPK2 is known to be functionally modulated through covalent regulated genes affected by the Hipk2 ko phenotype (Fig. 6B). modification with small ubiquitin-like modifier (SUMO)-1 (39, There was no specific influence of the Hipk2 ko on Aire-dependent 40). Interestingly, three other factors identified in our Y2H screens, gene expression observed. Thus, HIPK2 does not selectively affect namely, Ubc9, PIAS1, and DeSi-1, are known enzymatic regulators the Aire-dependent gene pool. of the posttranslational SUMO modification pathway. The SUMO To further characterize the HIPK2 target gene pool in mTECs, conjugating enzyme Ubc9 and the SUMO E3 ligase PIAS1 actively we performed a GO analysis using the GeneCodis3 software. The catalyze the SUMOylation of substrates, whereas DeSi-1 functions differentially expressed genes were tested for their annotation as an antagonist of SUMOylation by catalyzing deconjugation of classification concerning the cellular component, biological pro- SUMO from substrate proteins. Yet, PIAS, previously identified as cess, and molecular function (Supplemental Fig. 2). The cellular an AIRE interacting protein, does not lead to SUMO modification component analysis provided mainly hits in transmembrane, of AIRE (33). However, ectopic expression of PIAS1 attracted membrane coupled, or associated hits, followed by the hetero- AIRE to SUMO-1–containing nuclear complexes, which suggests trimeric G protein complex annotation. The biological process that AIRE function may be controlled by the SUMO pathway, analysis identified ion transport and cation transport as main hits, which can modulate protein localization and plays a prominent followed by ATP biosynthetic processes. The most prominent hits role in transcriptional regulation (41). of the molecular function annotation analysis were voltage-gated Notably, AIRE and HIPK2 share some of their interacting ion channel activity and ATPase activity coupled to transmem- proteins including the transcriptional cofactors CBP and DAXX, brane movements of ions. Taken together, the GO annotation two factors shown to be recruited, similar to AIRE and HIPK2, analysis of the HIPK2 target gene pool implies selective enrich- to PML NBs in response to cellular stress (36, 38, 40, 42–44). ment for genes involved in signaling transduction pathways. HIPK2 interacts with CBP and stimulates the acetyltransferase Interestingly, TEC-specific HIPK2 deficiency also resulted in activity of CBP/p300 (36, 45). Remarkably, acetylation has a significant contraction of the medullary, but not of the cortical been shown to affect different aspects of the function of Aire, compartment. Thus, the number of mTECs was reduced by 62% that is, the selection of target genes (46), protein stability, and (58% in CD80hi; 70% in CD80lo), whereas cTEC numbers re- nuclear localization (34). The tripartite interaction between Aire, mained unaltered (Supplemental Fig. 3A–C). The mTEC/cTEC HIPK2, and CBP may thus stimulate Aire acetylation and phos- ratio was reduced by 50%, whereas the mTEChi/mTEClo ratio phorylation, with the latter leading to suppression of coactivator was unaltered between Hipk2ko and control mice (Supplemental activity in vitro. Fig. 3D). The frequency of Aire-positive mTECs was comparable Immunohistochemical staining of sections from human thymus in Hipk2ko compared with control mice (Supplemental Fig. 3E), revealed colocalization of AIRE and HIPK2 in a subset of mTECs analyzing 4-wk-old (control: 16.0%; Hipk2ko: 21.5%) and 7-wk- under physiological conditions. Of note, ectopic expression of old mice (control: 12.6%; Hipk2ko: 11.4%). Taken together, thy- AIRE has been shown to stimulate the induction of DNA double- mic deletion of HIPK2 results in an overall reduction of the mTEC strand breaks (12). Interestingly, DNA double-strand breaks are compartment without altering its major subset composition. a well-known trigger of HIPK2 stabilization and activation by The Journal of Immunology 927

FIGURE 7. Model of HIPK2 function on pGE regulation. (A) Initial results (see Fig. 2) predicted HIPK2 to be a negative modulator of Aire’s transactivation activity. The outcome hypothesis includes the potential compensatory effects of other family members on the observed HIPK2 ko phenotype. (B) Suggested model of HIPK2 action on yet unidentified tran- scription factor(s) that upregulate(s) TRA expression upon HIPK2-mediated phosphorylation. TRAR, TRA regulator.

its transcriptional activity, which, in turn, would lead to promiscuous expression of certain TRAs. In the absence of HIPK2, these particular of TRAs would be downregulated. Among the affected genes, we noted components of signaling cascades regulating cell migration, differentiation, or motility. Interestingly, we could also observe an effect of HIPK2 deficiency A on the overall size of the mTEC compartment; that is, the number of FIGURE 6. HIPK2 deletion affects TRA expression. ( ) Percent rep- lo hi resentation of TRAs and non-TRA genes among downregulated genes CD80 and the mature CD80 mTECs was substantially reduced, compared with unaffected genes in HIPK2 ko mice; the enrichment for whereas their ratio (i.e., their developmental progression) was not lo TRA genes was significant (x2 = 10.249 with 1 degree of freedom; two- altered. Given the heterogeneity of CD80 mTECs with respect to tailed **p = 0.0014). (B) Percentage of Aire-dependent and -independent CCL-21 expression and a pre- and post-Aire stage (48–50), we genes among downregulated genes compared with unchanged genes by cannot exclude that the transcriptional changes observed in this array analysis. x2 = 0.065 with 1 degree of freedom; two-tailed p = 0.7995. study were confined to either of these subsets. Aire-dependent genes were not significantly (ns) enriched. It is currently unclear why the medulla in contrast to the cortex was selectively affected by HIPK2 deficiency and how this relates stimulating activation of the DNA damage checkpoint kinase to the aforementioned changes in gene expression of mTECs. In ATM (19, 47). Through engaging the DNA damage checkpoint summary, our data add HIPK2 to the growing list of molecules that kinase ATM, AIRE may provoke HIPK2 activation and stabili- modify Aire posttranslationally and regulate pGE and mTECs zation, thereby leading to the observed coexpression/colocalization development, two intricately connected processes. of AIRE and HIPK2 in mTECs. Our results show that the genetic ablation of HIPK2 modulates Acknowledgments the promiscuous expression of TRAs in mTECs, though the af- We thank the DKFZ Core Facilities Imaging and Cytometry for assistance fected number of genes was relatively low. Moreover, analysis of with cell sorting, Core Facilities Genome and Proteome for microarray anal- the Aire dependency of the differentially expressed genes showed ysis and Y2H screen performance, the Central Animal Facility for animal the Hipk2ko effect not to be restricted to Aire-dependent genes, but caretaking, and Eva Krieghoff-Henning for help with the reporter assays and instead other genes were also regulated. Our in vitro data indicated in vitro phosphorylation experiments. H. Scott (Centre for Cancer Biology) that HIPK2 had a repressive effect on Aire’s transcriptional reg- kindly provided the anti–Aire-FITC Ab. ulatory capability predicting that the removal of HIPK2 would thereby lead to an upregulation of target genes. Yet, we observed Disclosures that the majority of differentially regulated genes in Hipk2ko mice The authors have no financial conflicts of interest. were downregulated. Hence the effect of HIPK2 deficiency on Aire-dependent regulation of TRAs might be compensated for by References another regulatory component. It is possible that another HIPK 1. Boehm, T. 2011. Design principles of adaptive immune systems. Nat. Rev. family member such as HIPK1 might instead phosphorylate Aire Immunol. 11: 307–317. in place of HIPK2 (Fig. 7A). 2. Klein, L., B. Kyewski, P. M. Allen, and K. A. Hogquist. 2014. Positive and Furthermore, we did not observe an exclusive effect on Aire- negative selection of the T cell repertoire: what thymocytes see (and don’t see). Nat. Rev. Immunol. 14: 377–391. dependent, but rather Aire-independent genes were also affected 3. Kyewski, B., and L. Klein. 2006. A central role for central tolerance. Annu. Rev. by HIPK2 deficiency, implying that HIPK2 might modify another Immunol. 24: 571–606. 4. St-Pierre, C., S. Brochu, J. R. Vanegas, M. Dumont-Lagace´, S. Lemieux, and factor(s) regulating pGE in mTECs (Fig. 7B). We hypothesize that C. Perreault. 2013. Transcriptome sequencing of neonatal thymic epithelial cells. HIPK2 regulates this factor(s) by phosphorylation, thereby activating Sci Rep 3: 1860. 928 HIPK2, A NOVEL AIRE INTERACTION PARTNER

5. Sansom, S. N., N. Shikama, S. Zhanybekova, G. Nusspaumer, I. C. Macaulay, gene lists through enrichment analysis and integration of diverse biological in- M. E. Deadman, A. Heger, C. P. Ponting, and G. A. Holla¨nder. 2014. Population formation. Nucleic Acids Res. 37: W317–W322. and single cell genomics reveal the Aire-dependency, relief from Polycomb si- 29. Tabas-Madrid, D., R. Nogales-Cadenas, and A. Pascual-Montano. 2012. Gene- lencing and distribution of self- expression in thymic epithelia. Genome Codis3: a non-redundant and modular enrichment analysis tool for functional Res. 24: 1918–1931. genomics. Nucleic Acids Res. 40: W478–W483. 6. Perry, J. S. A., C.-W. J. Lio, A. L. Kau, K. Nutsch, Z. Yang, J. I. Gordon, 30. Boyle, E. I., S. Weng, J. Gollub, H. Jin, D. Botstein, J. M. Cherry, and K. M. Murphy, and C.-S. Hsieh. 2014. Distinct contributions of Aire and antigen- G. Sherlock. 2004. GO:TermFinder—open source software for accessing Gene presenting-cell subsets to the generation of self-tolerance in the thymus. Im- Ontology information and finding significantly enriched Gene Ontology terms munity 41: 414–426. associated with a list of genes. Bioinformatics 20: 3710–3715. 7. Peterson, P., T. Org, and A. Rebane. 2008. Transcriptional regulation by AIRE: 31. Ruan, Q. G., C. Y. Wang, J. D. Shi, and J. X. She. 1999. Expression and alter- molecular mechanisms of central tolerance. Nat. Rev. Immunol. 8: 948–957. native splicing of the mouse autoimmune regulator gene (Aire). J. Autoimmun. 8. Mathis, D., and C. Benoist. 2009. Aire. Annu. Rev. Immunol. 27: 287–312. 13: 307–313. 9. Zumer, K., K. Saksela, and B. M. Peterlin. 2013. The mechanism of tissue- 32. Schaller, C. E., C. L. Wang, G. Beck-Engeser, L. Goss, H. S. Scott, restricted antigen gene expression by AIRE. J. Immunol. 190: 2479–2482. M. S. Anderson, and M. Wabl. 2008. Expression of Aire and the early wave of 10. Giraud, M., H. Yoshida, J. Abramson, P. B. Rahl, R. A. Young, D. Mathis, and apoptosis in spermatogenesis. J. Immunol. 180: 1338–1343. C. Benoist. 2012. Aire unleashes stalled RNA polymerase to induce ectopic gene 33. Ilmarinen, T., H. Kangas, T. Kyto¨maa, P. Eskelin, J. Saharinen, J.-S. Seeler, expression in thymic epithelial cells. Proc. Natl. Acad. Sci. USA 109: 535–540. K. Tanhuanpa¨a¨, F. Y. Chan, R. M. Slattery, K. Alakurtti, et al. 2008. Functional 11. Rahl, P. B., C. Y. Lin, A. C. Seila, R. A. Flynn, S. McCuine, C. B. Burge, interaction of AIRE with PIAS1 in transcriptional regulation. Mol. Immunol. 45: P. A. Sharp, and R. A. Young. 2010. c-Myc regulates transcriptional pause re- 1847–1862. lease. Cell 141: 432–445. 34. Incani, F., M. Serra, A. Meloni, C. Cossu, L. Saba, T. Cabras, I. Messana, and 12. Abramson, J., M. Giraud, C. Benoist, and D. Mathis. 2010. Aire’s partners in the M. C. Rosatelli. 2014. AIRE acetylation and deacetylation: effect on protein molecular control of immunological tolerance. Cell 140: 123–135. stability and transactivation activity. J. Biomed. Sci. 21: 85. 13. Villasen˜or, J., W. Besse, C. Benoist, and D. Mathis. 2008. Ectopic expression of 35. Bjo¨rses, P., M. Pelto-Huikko, J. Kaukonen, J. Aaltonen, L. Peltonen, and peripheral-tissue in the thymic epithelium: probabilistic, monoallelic, I. Ulmanen. 1999. Localization of the APECED protein in distinct nuclear misinitiated. Proc. Natl. Acad. Sci. USA 105: 15854–15859. structures. Hum. Mol. Genet. 8: 259–266. 14. Derbinski, J., S. Pinto, S. Ro¨sch, K. Hexel, and B. Kyewski. 2008. Promiscuous 36. Hofmann, T. G., A. Mo¨ller, H. Sirma, H. Zentgraf, Y. Taya, W. Dro¨ge, H. Will, gene expression patterns in single medullary thymic epithelial cells argue for and M. L. Schmitz. 2002. Regulation of p53 activity by its interaction with a stochastic mechanism. Proc. Natl. Acad. Sci. USA 105: 657–662. homeodomain-interacting protein kinase-2. Nat. Cell Biol. 4: 1–10. 15. Pinto, S., C. Michel, H. Schmidt-Glenewinkel, N. Harder, K. Rohr, S. Wild, 37. Kim, Y. H., C. Y. Choi, S. J. Lee, M. A. Conti, and Y. Kim. 1998. Homeodomain- B. Brors, and B. Kyewski. 2013. Overlapping gene coexpression patterns in interacting protein kinases, a novel family of co-repressors for homeodomain human medullary thymic epithelial cells generate self-antigen diversity. Proc. transcription factors. J. Biol. Chem. 273: 25875–25879. Natl. Acad. Sci. USA 110: E3497–E3505. 38. Meloni, A., E. Fiorillo, D. Corda, F. Incani, M. L. Serra, A. Contini, A. Cao, and 16. Giraud, M., N. Jmari, L. Du, F. Carallis, T. J. F. Nieland, F. M. Perez-Campo, M. C. Rosatelli. 2010. DAXX is a new AIRE-interacting protein. J. Biol. Chem. O. Bensaude, D. E. Root, N. Hacohen, D. Mathis, and C. Benoist. 2014. An RNAi 285: 13012–13021. screen for Aire cofactors reveals a role for Hnrnpl in polymerase release and Aire- 39. de la Vega, L., I. Grishina, R. Moreno, M. Kruger,€ T. Braun, and M. L. Schmitz. activated ectopic transcription. Proc. Natl. Acad. Sci. USA 111: 1491–1496. 2012. A redox-regulated SUMO/acetylation switch of HIPK2 controls the sur- 17. Waterfield, M., I. S. Khan, J. T. Cortez, U. Fan, T. Metzger, A. Greer, K. Fasano, vival threshold to oxidative stress. Mol. Cell 46: 472–483. M. Martinez-Llordella, J. L. Pollack, D. J. Erle, et al. 2014. The transcriptional 40. Hofmann, T. G., E. Jaffray, N. Stollberg, R. T. Hay, and H. Will. 2005. Regu- regulator Aire coopts the repressive ATF7ip-MBD1 complex for the induction of lation of homeodomain-interacting protein kinase 2 (HIPK2) effector function immunotolerance. Nat. Immunol. 15: 258–265. through dynamic small ubiquitin-related modifier-1 (SUMO-1) modification. J. 18. Albers, M., H. Kranz, I. Kober, C. Kaiser, M. Klink, J. Suckow, R. Kern, and Biol. Chem. 280: 29224–29232. M. Koegl. 2005. Automated yeast two-hybrid screening for nuclear - 41. Girdwood, D. W. H., M. H. Tatham, and R. T. Hay. 2004. SUMO and tran- interacting proteins. Mol. Cell. Proteomics 4: 205–213. scriptional regulation. Semin. Cell Dev. Biol. 15: 201–210. 19. Winter, M., D. Sombroek, I. Dauth, J. Moehlenbrink, K. Scheuermann, J. Crone, 42. Hofmann, T. G., N. Stollberg, M. L. Schmitz, and H. Will. 2003. HIPK2 reg- and T. G. Hofmann. 2008. Control of HIPK2 stability by ubiquitin ligase Siah-1 ulates transforming growth factor-beta-induced c-Jun NH(2)-terminal kinase and checkpoint kinases ATM and ATR. Nat. Cell Biol. 10: 812–824. activation and apoptosis in human hepatoma cells. Cancer Res. 63: 8271–8277. 20. Bitomsky, N., E. Conrad, C. Moritz, T. Polonio-Vallon, D. Sombroek, 43. Mo¨ller, A., H. Sirma, T. G. Hofmann, H. Staege, E. Gresko, K. S. Ludi,€ K. Schultheiss, C. Glas, V. Greiner, C. Herbel, F. Mantovani, et al. 2013. E. Klimczak, W. Dro¨ge, H. Will, and M. L. Schmitz. 2003. Sp100 is important Autophosphorylation and Pin1 binding coordinate DNA damage-induced HIPK2 for the stimulatory effect of homeodomain-interacting protein kinase-2 on p53- activation and cell death. Proc. Natl. Acad. Sci. USA 110: E4203–E4212. dependent gene expression. Oncogene 22: 8731–8737. 21. Fleischer, S., S. Wiemann, H. Will, and T. G. Hofmann. 2006. PML-associated 44. Pitka¨nen, J., V. Doucas, T. Sternsdorf, T. Nakajima, S. Aratani, K. Jensen, H. Will, repressor of transcription (PAROT), a novel KRAB-zinc finger repressor, is regu- P. Va¨ha¨murto, J. Ollila, M. Vihinen, et al. 2000. The autoimmune regulator protein lated through association with PML nuclear bodies. Exp. Cell Res. 312: 901–912. has transcriptional transactivating properties and interacts with the common 22. Soza-Ried, C., C. C. Bleul, M. Schorpp, and T. Boehm. 2008. Maintenance of coactivator CREB-binding protein. J. Biol. Chem. 275: 16802–16809. thymic epithelial phenotype requires extrinsic signals in mouse and zebrafish. J. 45. Aikawa, Y., L. A. Nguyen, K. Isono, N. Takakura, Y. Tagata, M. L. Schmitz, Immunol. 181: 5272–5277. H. Koseki, and I. Kitabayashi. 2006. Roles of HIPK1 and HIPK2 in AML1- and 23. Rouse, R. V., L. M. Bolin, J. R. Bender, and B. A. Kyewski. 1988. Monoclonal p300-dependent transcription, hematopoiesis and blood vessel formation. EMBO antibodies reactive with subsets of mouse and human thymic epithelial cells. J. J. 25: 3955–3965. Histochem. Cytochem. 36: 1511–1517. 46. Saare, M., A. Rebane, B. Rajashekar, J. Vilo, and P. Peterson. 2012. Autoim- 24. Hubert, F.-X., S. A. Kinkel, K. E. Webster, P. Cannon, P. E. Crewther, mune regulator is acetylated by transcription coactivator CBP/p300. Exp. Cell A. I. Proeitto, L. Wu, W. R. Heath, and H. S. Scott. 2008. A specific anti-Aire Res. 318: 1767–1778. antibody reveals aire expression is restricted to medullary thymic epithelial cells 47. Dauth, I., J. Kruger,€ and T. G. Hofmann. 2007. Homeodomain-interacting pro- and not expressed in periphery. J. Immunol. 180: 3824–3832. tein kinase 2 is the ionizing radiation-activated p53 serine 46 kinase and is 25.Su,A.I.,T.Wiltshire,S.Batalov,H.Lapp,K.A.Ching,D.Block,J.Zhang, regulated by ATM. Cancer Res. 67: 2274–2279. R. Soden, M. Hayakawa, G. Kreiman, et al. 2004. A gene atlas of the mouse 48. Lkhagvasuren, E., M. Sakata, I. Ohigashi, and Y. Takahama. 2013. Lymphotoxin and human protein-encoding transcriptomes. Proc. Natl. Acad. Sci. USA 101: b receptor regulates the development of CCL21-expressing subset of postnatal 6062–6067. medullary thymic epithelial cells. J. Immunol. 190: 5110–5117. 26. Derbinski, J., J. Ga¨bler, B. Brors, S. Tierling, S. Jonnakuty, M. Hergenhahn, 49. Metzger, T. C., I. S. Khan, J. M. Gardner, M. L. Mouchess, K. P. Johannes, L. Peltonen, J. Walter, and B. Kyewski. 2005. Promiscuous gene expression in A. K. Krawisz, K. M. Skrzypczynska, and M. S. Anderson. 2013. Lineage thymic epithelial cells is regulated at multiple levels. J. Exp. Med. 202: 33–45. tracing and cell ablation identify a post-Aire-expressing thymic epithelial cell 27. Carmona-Saez, P., M. Chagoyen, F. Tirado, J. M. Carazo, and A. Pascual- population. Cell Reports 5: 166–179. Montano. 2007. GENECODIS: a web-based tool for finding significant con- 50. Nishikawa, Y., H. Nishijima, M. Matsumoto, J. Morimoto, F. Hirota, current annotations in gene lists. Genome Biol. 8: R3. S. Takahashi, H. Luche, H. J. Fehling, Y. Mouri, and M. Matsumoto. 2014. 28. Nogales-Cadenas, R., P. Carmona-Saez, M. Vazquez, C. Vicente, X. Yang, Temporal lineage tracing of Aire-expressing cells reveals a requirement for Aire F. Tirado, J. M. Carazo, and A. Pascual-Montano. 2009. GeneCodis: interpreting in their maturation program. J. Immunol. 192: 2585–2592.