ADAM17 Deletion in Thymic Epithelial Cells Alters Aire Expression without Affecting T Cell Developmental Progression

Background Cellular interactions between thymocytes and thymic stromal cells are critical for normal T cell development. Thymic epithelial cells (TECs) are important stromal niche cells that provide essential growth factors, cytokines, and present self-antigens to developing thymocytes. The identification of genes that mediate cellular crosstalk in the thymus is ongoing. One candidate gene, Adam17, encodes a metalloprotease that functions by cleaving the ectodomain of several transmembrane proteins and regulates various developmental processes. In conventional Adam17 knockout mice, a non-cell autonomous role for ADAM17 in adult T cell development was reported, which strongly suggested that expression of ADAM17 in TECs was required for normal T cell development. However, knockdown of Adam17 results in multisystem developmental defects and perinatal lethality, which has made study of the role of Adam17 in specific cell types difficult. Here, we examined T cell and thymic epithelial cell development using a conditional knockout approach. Methodology/Principal Findings We generated an Adam17 conditional knockout mouse in which floxed Adam17 is deleted specifically in TECs by Cre recombinase under the control of the Foxn1 promoter. Normal T cell lineage choice and development through the canonical αβ T cell stages was observed. Interestingly, Adam17 deficiency in TECs resulted in reduced expression of the transcription factor Aire. However, no alterations in the patterns of TEC phenotypic marker expression and thymus morphology were noted. Conclusions/Significance In contrast to expectation, our data clearly shows that absence of Adam17 in TECs is dispensable for normal T cell development. Differentiation of TECs is also unaffected by loss of Adam17 based on phenotypic markers. Surprisingly, we have uncovered a novel genetic link between Adam17and Aire expression in vivo. The cell type in which ADAM17 mediates its non-cell autonomous impact and the mechanisms by which it regulates intrathymic T cell development remain to be identified.


Introduction
Proper intercellular communication is critical for the development of T cells in the thymus. Hematopoietic progenitors must receive signals from the thymic microenvironment to proliferate, differentiate, and undergo positive and negative selection [1]. Thymocyte progenitors enter the thymus at the corticomedullary junction and are identified as CD4 2 CD8 2 ''double negative'' (DN) stage. During the generation of TCRab + T cells, they then progress through four DN stages characterized by the expression of CD44 and CD25 (DN1: CD44 + CD25 2 , DN2: CD44 + CD25 + , DN3: CD44 2 CD25 + , and DN4: CD44 2 CD25 2 ) [2]. After TCRb selection in the DN3 stage thymocytes progress through the transitory DN4 stage to the CD4 + CD8 + ''double positive'' (DP) stage, and then undergo both positive and negative selection prior to being released as single positive (SP) CD4 or CD8 T cells. The thymocytes migrate through anatomically distinct thymic microenvironments as they progress through these developmental stages (reviewed in [1]).
The identification of genes that mediate cellular crosstalk in the thymus during T cell development is an area of active investigation [3]. Adam17 (Entrez GeneID: 11491; also known as TNFaconverting enzyme (TACE)), for example, has been proposed as one of these genes. The metalloprotease ADAM17 has a well characterized ''sheddase'' activity, in which it releases membrane bound proteins from the cell surface, often initiating their signaling ability [4,5,6,7]. ADAM17 can also participate in the process of regulated intramembrane proteolysis in which it cleaves the extracellular portion of a transmembrane receptor, which leads to intracellular cleavage by the c-secretase complex and resultant cellular signaling [8]. By analyzing the temporal and spatial activities of ADAM17, we can gain insights into the regulation of cellular signaling in the thymus. ADAM17 is capable of cleaving a large number of immunologically relevant substrates (reviewed in [9]). It was initially described for its role in shedding the transmembrane precursor of soluble TNFa [4,10]. Subsequently, it was described to regulate cell surface expression of TNFR-I [11], TNFR-II [12], Notch [13], Delta-like-1 [14], SCF [15], CD40 [16], L-selectin [17], CX3CL1 [18], LAG3 [19], Flt3L [20] and several other transmembrane proteins that are relevant to immune system development. Through a conventional knockout (KO) approach, it became clear that ADAM17 is critically important for the development of epithelial tissues via its regulation of the EGFR signaling pathway [17]. ADAM17 acts to shed EGF family ligands, a mechanism that is implicated in the development of mammary epithelium, heart valves, skin, and hair [17,21,22]. However, due to perinatal lethality of Adam17 DZn/DZn conventional knockout mice (in which mutated ADAM17 that lacks its critical zinc binding ability is expressed), analyses of adult physiology could not be assessed. Conditional knockout of Adam17 using the Cre/Lox system has recently solved this issue and revealed Adam17's importance in inflammation [23].
The role of ADAM17 in adult T cell development is largely unknown at a mechanistic level. The current literature describes a block from the DN to DP stage of thymopoiesis in Adam17 DZn/DZn mice that survive to adulthood on a mixed B6 and 129S3 background [24,25]. Importantly, through transplantation of Adam17 DZn/DZn bone marrow into RAG 2/2 recipient mice, this block was shown to be non-cell autonomous to the hematopoietic compartment [25]. This line of experimentation indicated that the likely location of ADAM17's role in T cell development as the thymic stroma. Through in vitro assays analyzing the effect of the cleavage of the Notch ligand Dll1, our laboratory determined that modulation of the Notch signaling pathway is not likely the causative mechanism of the defect in Adam17 DZn/DZn mice [26].
Thymic epithelial cells (TECs) are thought to be a principal component of the thymic stroma that provides signals to developing thymocytes, and we therefore hypothesized that ADAM17 in TECs plays a key non-cell autonomous role on development of T cells. In order to assess the role of ADAM17 specifically in adult TECs, we devised a conditional knockout strategy in which Adam17 ''floxed'' mice [23] were crossed with Foxn1-Cre mice [27]. In these ADAM17 conditional knockout mice (Adam17/Foxn1), Adam17 was effectively deleted from TECs. Surprisingly, T cell development was unaffected in Adam17/Foxn1 mice. Interestingly, analysis of TECs revealed an unaltered TEC population structure, but a strong reduction of Aire mRNA expression. The lack of T cell phenotype in these mice requires a re-assessment of the proposed role of ADAM17 in the thymus.

Results
Generation of TEC-specific Adam17 knockout mice Mice expressing Cre under the control of the Foxn1 promoter [27] were crossed with Adam17 exon2-floxed mice [23]. Specificity of Cre expression in TECs has previously been demonstrated and its utility in in vivo functional assays has been validated [28]. To further confirm Cre restriction to the TEC compartment, we performed TEC digestions from 1 week old Foxn1-Cre mice. CD45 2 and CD45 + fractions were separated by MACS and the CD8 + fraction of thymocytes was further purified by FACS. RT-PCR for Cre expression revealed high levels of Cre in the CD45 2 fraction enriched for TECs, but low to no expression in other sorted populations ( Figure 1A). Mice of the genotype Foxn1-Cre/Adam17 flox/flox (referred to as Adam17/Foxn1) were present in expected Mendelian ratios compared to littermate controls and had no overt phenotypes or distinctive external features (data not shown), unlike Adam17 DZn/DZn mice [17].
To confirm knockdown of gene expression, Adam17 levels were assessed by qRT-PCR. TECs were digested, MACS enriched for CD45 2 cells, and further sorted by FACS based on EpCAM expression. Quantitative Adam17 levels were determined using primers specific to the floxed region of Adam17. EpCAM + TECs from Adam17/Foxn1 mice displayed less than 2% the Adam17 levels of littermate controls ( Figure 1B). Furthermore, RT-PCR confirmed the presence of the truncated, null form of the mRNA transcript expected after recombination in Adam17/Foxn1 mice, while the wildtype form of the transcript was present in both Adam17 flox/+ and Foxn1-Cre/Adam17 flox/+ mice ( Figure 1C). Thymocyte development is normal in Adam17/Foxn1 mice Conventional Adam17 KO mice exhibit a non-cell autonomous block in T cell development from the DN to DP stage, which indicated a likely role for ADAM17 on TECs [25]. However, in our TEC-specific Adam17/Foxn1 knockout mice, the thymus and spleen are of normal size ( Figure 2A). Likewise, analysis of intrathymic populations revealed indistinguishable numbers of DN, DP, CD4SP, and CD8SP thymocytes between control and Adam17/Foxn1 mice ( Figure 2B). Furthermore, the frequencies and absolute numbers of DN1 through DN4 stages of thymocyte development were unaltered in the absence of ADAM17 on TECs ( Figure 2C, and data not shown). Likewise, analysis of regulatory T cells showed no alterations in the frequency of CD4SP thymocytes expressing FoxP3 or CD25 nor CD4SP splenocytes expressing FoxP3 ( Figure S1). These results are not consistent with the hypothesis that the ADAM17 non-cell autonomous role in T cell development is localized to the TECs.

Aire expression is diminished in Adam17/Foxn1 TECs
We next examined whether deletion of ADAM17 in TECs affected their development and expression of TEC-specific genes in a cell-autonomous fashion. Analysis of gene expression in FACS-sorted EPCAM + TECs by qRT-PCR yielded the unexpected finding that the autoimmune regulator transcription factor, Aire, was diminished in Adam17/Foxn1 mice ( Figure 3A). Aire mutations have been implicated in the human disease APECED, which may result from defective medullary TEC development, and/or lack of expression of tissue-specific antigens (TRAs) that result in alterations in negative selection of thymocytes (reviewed in [29]). Aire mRNA expression in TECs in Adam17/Foxn1 mice was less than 10% of control mice. TRAs whose expression is dependent or independent of Aire have been previously described [30]. The Aire-dependent TRA, Casein a (Csna), was significantly reduced in Adam17/Foxn1 mice relative to controls ( Figure 3B). However, other Aire-dependent TRAs, Casein c (Csnc), Salivary protein 1 (Spt1), and Insulin 2 (Ins2), were not reduced. As expected, Aire-independent TRAs, Casein b (Csnb), Casein k (Csnk), and Glutamic acid decarboxylase-67 (Gad67), were expressed similarly between Adam17/Foxn1 and control mice ( Figure 3B).
To assess why there was not a more pronounced reduction in Aire-dependent TRAs in Adam17/Foxn1 thymi, we analyzed Aire protein levels by immunohistochemistry and flow cytometry. Aire protein was clearly present in Adam17/Foxn1 thymi based on immunofluorescence analysis of thymic tissue sections ( Figure 4A). Furthermore, thymi were properly organized into outer cortical and inner medullary regions as assessed by staining with the predominantly medullary marker Keratin-5 and the predominantly cortical marker Keratin-8 ( Figure 4A) [31]. Morphometrical analyses of the thymic sections did not did not reveal any quantitative differences in the size of the medullary regions or the numbers of Aire + cells in the medulla between controls and Adam17/Foxn1 thymi (data not shown). Flow cytometric analysis of TECs, to measure the percentage of Aire + events within the EpCAM + fraction, indicated a reduction in the frequency of TECs that express Aire in Adam17/Foxn1 thymi ( Figure 4B). From two independent experiments, the frequency of Aire + TECs in Adam17/Foxn1 thymi was 65% and 85% of control mice ( Figure 4B and data not shown). Furthermore, the mean fluorescence intensity of the events in the EpCAM + Aire + gate was reduced to 92% and 96% of control levels in two independent experiments ( Figure 4B and data not shown). The more modest reduction in Aire protein levels relative to the more pronounced reduction in mRNA levels may explain why Aire-dependent TRAs are not more significantly altered in Adam17/Foxn1 thymi. To confirm the identity of Aire + TECs, we stained for CD80 and observed that all Aire + cells in both control and Adam17/Foxn1 mice expressed CD80 ( Figure 4C), consistent with the previous literature [32].
Thymic stromal populations are unaltered in Adam17/ Foxn1 mice In order to determine if the observed reduction in Aire levels were the result of a reduction in mature medullary TECs (mTECs) or an alteration in another thymic population, we stained for markers of thymic stromal cells. Critically, EpCAM + TECs were unaltered both in number and frequency in Adam17/Foxn1 mice ( Figure 5A, D, E). mTECs at the terminal stage of differentiation upregulate Aire expression and can be identified by several surface markers [33]. To address whether lack of ADAM17 results in a loss of mature Aire + mTECs, we stained with the mTEC markers UEA1, CD80, and MHCII [34]. TEC subpopulation analysis revealed similar percentages of medullary TECs, as shown by the marker UEA1 ( Figure 5B, F), and normal proportions of mature and immature medullary TEC subsets as assessed by high and intermediate MHCII levels, respectively ( Figure 5C). Also, Aire + TECs in both control and Adam17/Foxn1 thymi also highly expressed CD80, corresponding to a mature medullary TEC phenotype, while Aire 2 TECs were predominantly CD80 low ( Figure 4C). These combined data indicate that, based on phenotypic TEC markers and histological examination, the medullary compartment in Adam17/Foxn1 mice is normal despite the clear reduction in Aire mRNA expression. Furthermore, staining with the antibody MTS15 [35] revealed normal proportions of MTS15 + thymic fibroblasts in Adam17/Foxn1 mice ( Figure 5A,G).

ADAM17 and EGF substrates are expressed on non-TEC thymic stroma
Since Adam17/Foxn1 thymi do not display the T cell developmental block observed in Adam17 conventional knockout mice [25], it is possible that ADAM17 in another thymic stromal cell type is used during T cell development. To assess this possibility, we examined the non-TEC stroma for presence of ADAM17 and EGF family ligands, which have been shown to be key substrates for ADAM17 in vivo. First, to confirm ADAM17 expression on non-TEC stroma, we FACS sorted CD45 2 Ter119 2 EpCAM 2 cells from normal B6 mice. Quantitative PCR revealed levels of Adam17 expression in these non-TECs, at least as high as that of TECs ( Figure 6A). We then proceeded to examine more specific stromal cell populations, including CD45 2 Ter119 2 EpCAM 2 MTS15 + thymic fibroblasts and the CD45 2 Ter119 2 EpCAM 2 MTS15 2 populations containing thymic endothelial cells and other mesenchyme [35] ( Figure 5A). Notably, in thymic fibroblasts, we observed levels of TGFa and HB-EGF that were elevated relative to those of TECs, as well as the presence of the EGF receptor ( Figure 6B). This indicates that ADAM17 and EGF pathway substrates shown to be critical in other developmental systems are present on the non-TEC populations in the thymus.
Adam10 expression is not altered on Adam17/Foxn1 thymic stroma ADAM10 has a high degree of substrate overlap with ADAM17 and may function to compensate for ADAM17 deficiency in vivo. ADAM10 has been shown to be tightly regulated at the transcriptional level, whereas ADAM17 is predominantly activated in a post-transcriptional manner [6,19]. To assess whether Adam10 expression is altered in the thymus following deletion of Adam17, we performed qRT-PCR on sorted EpCAM + TECs and EpCAM 2 stroma. We did not see any alteration in Adam10 levels in TECs or non-TEC stroma in Adam17/Foxn1 mice ( Figure S2).

Discussion
Loss of Adam17 results in a profound non-cell autonomous T cell developmental block in conventional knockout mice, and the underlying assumption from the previous literature was that absence of ADAM17 in thymic epithelial cells was the key cause of this block [25]. In the present study, we have conditionally knocked out Adam17 in thymic epithelial cells expressing the transcription factor Foxn1 [27]. We have convincingly shown that wildtype Adam17 is eliminated at the mRNA level by both quantitative and standard RT-PCR strategies. Surprisingly, Adam17/Foxn1 conditional KO mice do not possess a similar block in T cell development to Adam17 conventional knockout mice. No hematopoietic lineage was statistically different in percentage or total cell numbers in the Adam17/Foxn1 thymus compared to controls. We have also performed a limited analysis of embryonic day 17.5 thymi in conventional Adam17 knockout mice, and have not observed any differences in thymocyte subset or TEC subset frequencies at this earlier developmental stage (Gravano et al., unpublished observations). Therefore, we conclude that TEC subsets are unaltered at either the fetal or adult stages in the absence of Adam17. In addition, our results show that unlike the thymus in the postnatal conventional Adam17 knockout mice [25], the thymus in TEC-specific, Adam17 knockout mice is not defective in the generation of T cells. Our data support the idea that Adam17 is a regulator of adult, but not fetal, thymic development, but is not dependent on Aire levels. Our results, which show that Adam17 in TECs is not necessary for thymic development, were contrary to expectation, but could be explained by several possible reasons, described below.
First, ADAM17 may be acting within the thymic microenvironment to regulate T cell development, but on a non-epithelial cell type (see conceptual model shown in Figure 7). Consistent with this idea, we have observed high levels of Adam17 expression on the CD45 2 EpCAM 2 population of cells in the thymus, which contain fibroblasts, endothelial cells, and other mesenchymal cells. Furthermore, these non-epithelial cell types are unaffected in Adam17/Foxn1 mice. In addition, TGFa and HB-EGF, two EGF pathway ligands shed by ADAM17 during the development of other epithelial tissues, are expressed at high levels in adult thymic fibroblasts relative to TECs. It is has been well established that growth factors such as FGF7 and FGF10 are secreted from the neural crest-derived mesenchyme that surrounds the thymic epithelium in fetal development [36]. The interaction of mesenchyme regulates TEC development, proliferation and ability to support thymopoiesis. Through fetal thymic organ culture experiments it has been shown that this mesenchymal support can be replaced by addition of the soluble ADAM17 substrate TGFa [37]. Thus, ADAM17 may potentially be produced by non-TEC stroma during the fetal and/or adult stages to promote the development and proliferation of TECs, facilitating proper thymocyte development. Alternatively, signals shed by ADAM17 on non-TEC stroma may target the thymocytes directly. It has previously been shown that thymic mesenchyme and fibroblasts can play a direct role in early T cell development [38,39]. Recently, intrathymic PDGFR + pericytes of neural crest origin have been show to promote thymocyte egress, highlighting the importance of non-TEC stroma in thymus physiology [40,41].
Second, it is possible that there is an extrathymic requirement for ADAM17 for proper T cell development (Figure 7). Since ADAM17 sheds factors that can act systemically [17], it is possible that extrathymically derived factors influence intrathymic development. The multiple systemic defects in Adam17 DZn/DZn mice also support this hypothesis [17]. For example, the DP thymocyte subset, which is the population reduced in Adam17 DZn/DZn mice [25], is highly sensitive to levels of systemic factors such as glucocorticoids, which are involved in metabolism and stress response [42]. Adult Adam17 DZn/DZn mice have a lean and hypermetabolic phenotype characterized by reduced fat mass despite normal food intake. However, corticosterone (the major glucocorticoid in mice) was not altered in Adam17 DZn/DZn mice [24]. Interestingly, plasma levels of leptin, a key factor in the energy homeostasis system, were reduced approximately 2-fold in Adam17 DZn/DZn mice. Leptin-deficient ob/ob mice possess a high level of DP thymocyte apoptosis, which can be reversed by addition of recombinant leptin [43]. Leptin has been shown to act by protecting thymocytes from glucocorticoid induced apoptosis [44]. The phenotype of ob/ob DP thymocytes is very similar to Adam17 DZn/DZn mice. It remains to be determined if changes in these systemic factors (which would not be deleted in the Adam17/ Foxn1 mice) are the cause of the thymic hypocellularity in Adam17 DZn/DZn mice.
Our data show that Adam17/Foxn1 mice have reduced expression of the transcription factor Aire in TECs, but this reduction does not affect TEC population composition. Mutations in Aire have been associated with autoimmune disease in humans and in mouse models [29]. Adam17/Foxn1 mice do not display any overt phenotypes that would indicate autoimmune disease, which is consistent with the normal thymic phenotype and production of mature T cells observed in two-to-three-month-old Aire KO mice [45]. The extent of autoimmune disease in Aire KO mice is dependent on the genetic background [46], with very mild disease observed on the mixed 129xB6 background [47] (which incidentally, is the same background as the Adam17/Foxn1 mice used in our studies). Therefore, the unaltered thymus development observed in Adam17/Foxn1 mice, despite the reduced Aire mRNA expression in these mice, is of similar phenotype to Aire KO mice. Little is known about the factors that regulate Aire expression. Aire expression is highly specific to mature medullary TECs only at their latest stages of differentiation [33]. Lymphotoxin b receptor signaling was initially thought to regulate the expression of Aire [48], but the current consensus now favors the conclusion that lymphotoxin b receptor signals are required for TEC differentiation but do not impact Aire expression directly [29,49]. We are convinced that our PCR primers should detect all known Aire splice variants [50], strongly supporting the idea that Adam17 promotes the transcription of the Aire gene. Aire protein expression was measured both by flow cytometry and immunohistochemistry, using different antibodies but with similar results, demonstrating that the Aire protein levels are close to normal in Adam17/Foxn1 mice. Therefore, one explanation is that low level of Aire mRNA transcript is sufficient to produce normal Aire protein levels in Adam17/Foxn1 mice. Alternatively, Adam17 may also play a role in the turnover of Aire protein. If ADAM17 degrades Aire protein as part of its normal function, then we might indeed expect to see the levels of Aire protein to be normal in Adam17/Foxn1 mice even if Aire mRNA levels are low, due to an increased lifespan of the Aire protein in the absence of Adam17. The role of ADAM17 as sheddase and an activator of regulated intramembrane proteolysis is well documented [5,7,8,9], but to date, the substrates of ADAM17 have been restricted to extracellular transmembrane receptor and ligands. In contrast, no direct link between ADAM17 and the regulated turnover of intracellular nuclear proteins, such as Aire, has been reported. Therefore, further experimentation is required to confirm whether or not ADAM17 can facilitate Aire protein turnover in mTECs.
ADAM17 has recently been shown to cleave the RANK receptor in macrophages [51]. RANK signaling is a key signaling pathway in the development of mature medullary Aire-expressing EGF pathway gene expression levels on the indicated populations were calculated relative to levels on TECs with Gapdh serving as the internal control. Data are from 2 independent experiments with each experiment consisting of n$3 pooled mice. Data report mean + SD. In these experiments, the cell types were identified as follows: TEC: CD45 -Ter119 -EpCAM + MTS15 -; Fibroblast: CD45 -Ter119 -EpCAM -MTS15 + ; Non TEC/Fibroblast: CD45 -Ter119 -EpCAM -MTS15 -. TECs were sorted to an average of 92% purity, fibroblasts to 68.73% purity, and non-TEC/ fibroblast to 95.84% purity. Despite lower fibroblast purity, notably, no sorted fibroblasts were detected within the TEC gate upon re-analysis (data not shown). doi:10.1371/journal.pone.0013528.g006 TECs at the fetal stage [52,53,54]. Although RANKL is necessary for the induction of Aire expression and mTEC development in fetal thymic organ cultures, the dependence of RANKL on mTEC after the embryonic stage is not as severe [52]. Our data show that adult mTEC frequencies are not affected, and Aire + TECs appear to be mature in the absence of Adam17. Furthermore, although Aire mRNA levels are severely reduced in adult Adam17/Foxn1 mice, Aire protein levels are not (Figure 4). Therefore, despite the link between ADAM17 and RANK signaling, our observation that TEC populations are normal in Adam17/Foxn1 mice does not support the hypothesis that ADAM17 is necessary to facilitate RANK signaling in the adult thymus.
In conclusion, our findings reveal that ADAM17 is dispensable in TECs for the ability to support T cells to progress through the characterized stages of ab T cell development. This contrasts with conventional Adam17 DZn/DZn mice and suggests a novel role for ADAM17 in non-TEC stroma or extrathymically-derived ADAM17 substrates on T cell development. Adam17 has been deleted under the control of other promoters including Sox9, resulting in an osteoporotic phenotype due to G-CSF dysregulation [55] and Tie2, revealing a role for Adam17 in pathological neovascularization [56]. New conditional knockout strategies are necessary to identify the cell type in which ADAM17 nonautonomously regulates T cell development. In addition, our findings indicate that ADAM17 in TECs regulates the expression of the transcription factor Aire. Further experiments will be required to identify the exact ADAM17 substrates involved in Aire regulation and whether and the ADAM17 deficit in TECs leads to breaches in the regulation of negative selection in thymocytes and self-tolerance.

Ethics Statement
All animal procedures were approved by the UC Merced Institutional Animal Care and Use Committee (Protocol # UCM 185).

Thymocyte and TEC isolation
Thymi were harvested and placed in Medium 199 (M199) (Invitrogen) containing 2% heat inactivated fetal calf serum (FCS) (Atlanta Biologicals). For thymocyte analysis, thymocytes were extracted by mechanical disruption with the base of a 5 ml syringe followed by filtration through 70 mm nylon mesh. For TEC analysis by flow cytometry [57], thymi were trimmed of excess fat and connective tissue. Thymi were placed in M199 and the capsule was nicked in several places with fine scissors. Thymi were agitated with a magnetic stirrer at 4uC for 20 minutes to remove a large fraction of thymocytes. Thymic fragments were then transferred into an enzymatic digestion buffer containing 0.125% (w/v) Collagenase D (Roche) with 0.1% (v/v) DNAse I (Roche) and gently rotated on an orbital shaker at 37uC for 15 minutes. Thymic fragments were pelleted by centrifugation at 3006 g and the digestion buffer was exchanged. After 4 digestions, 0.125% (w/v) dispase (Worthington Biochemical) was added to the digestion buffer and thymi were incubated for an additional 15 minutes. After the final digestion, cells were incubated in PBS containing 5 mM EDTA + 1% FCS + 0.02% (w/v) NaN 3 for 10 minutes on ice. Cells were passed several times through a 21G needle and filtered through a 70 mm nylon mesh. Red blood cells were lysed with ACK lysis buffer (0.15 M NH 4 Cl + 10 mM KHCO 3 + 0.1 mM Na 2 EDTA, pH 7.2). Cells were counted by hemocytometer and population numbers determined by calculating based on flow cytometric gating.

RT-PCR and qRT-PCR
Total RNA was extracted from FACS or MACS sorted cells using the RNeasy kit (Qiagen) and cDNA was generated using the Superscript III RT kit (Invitrogen) according to the manufacturers' instructions. Reverse transcriptase PCR (RT-PCR) primers to Adam17 were designed to span the floxed exon2 with the forward primer 59-CGGAGGAAGCAGGCTCTG-39 annealing to exon1 and the reverse primer 59-GAGTCAGGCTCACCAACCAC-39 annealing to exon3. This allows discrimination of a wildtype transcript and a mutant, null transcript resulting from recombination. After one minute dissociation at 95uC, Adam17 was amplified from the cDNA over 40 cycles at 95uC for 30 seconds, 57uC for 30 seconds, and 72uC for 1 minute. Quantitative RT-PCR (qRT-PCR) was performed using SYBR Green PCR master mix (Applied Biosystems) and samples run on an Applied Biosystems 7300 Real-Time system. Samples were denatured at 95uC for 2 minutes followed by 40 cycles of 95uC for 15 seconds and 60uC for 1 minute. The DDCt method was utilized to determine relative expression levels of the gene of interest between samples using ribosomal protein L7 (RPL7) as the internal control. For Adam17 qRT-PCR the above noted forward primer was used in conjunction with the reverse primer 59-GTTTCTAAGT-GTGTCGCAGACTG-39 that anneals to exon2. Therefore, amplification only occurs off the wildtype transcript. Unless noted otherwise, primer sequences were obtained from Primerbank (http://pga. mgh.harvard.edu/primerbank/) or from previous literature [30]. All qRT-PCR primers were validated to be between 90-110% efficient by a template serial dilution standard curve and to have a single peak when a dissociation curve was performed.

Immunohistochemistry
Thymi were extracted and snap frozen in OCT compound (Sakura Finetek) over dry ice. Sectioning of the thymus, tissue processing protocols, keratin-5, keratin-8 and Aire staining and morphometry data analysis were performed as described [59].