Neonatal tolerance revisited: a perinatal window for Aire control of autoimmunity

To this end, we attempted to develop a regulable Aire transgenic system that could complement an Aire deficiency in a time- and dose-controlled manner. The TA transgene drives expression of the tetTA (tTA) transcriptional activator in thymic epithelium (14); its product activates the tetOAire (TOA) reporter transgene, which carries the mouse Aire complementary DNA (cDNA) under the dictates of a tTA-responsive promoter (Fig. 1 A). When both transgenes are present, Aire is expressed in thymic epithelium, but administration of doxycycline (Dox) blocks the interaction of tTA with the reporter gene promoter, turning Aire off. The transgenes were crossed to the Aire KO mutation (15) on the nonobese diabetic (NOD) genetic background to generate Aire^−/−TA⁺TOA⁺ mice, which are referred to as inducible-Aire (iA) mice.

RT-PCR analysis of Aire transcripts in the thymus showed that the iA system rescued Aire expression to values close to those of WT mice (Fig. 1 B). Similarly, the usual punctate staining of Aire in nuclei from thymic epithelial cells was recovered in iA mice (Fig. 1 C). Transgene-encoded Aire was expressed in medullary and cortical epithelial cells, with little to no expression in DCs and macrophages, as expected from the properties of the driver construct (Fig. 1 D). Expression in these thymic populations was dependent on the presence of both the driver and reporter transgenes, as the reporter alone yielded no detectable expression (Fig. 1 D). Flow cytometry after intracellular staining confirmed the presence of Aire in MECs from iA mice, in quantities slightly higher than those of WT littermates and of an Aire⁺ MEC population in close to normal proportions (Fig. 1 E).

Aire expression in the thymus of iA mice proved sensitive to Dox, with a complete loss of Aire transcripts within the first 24 h of Dox treatment and a disappearance of Aire⁺ MECs over the subsequent 3 and 5 d (Fig. 1 F), which is consistent with the rapid turnover of Aire⁺ MHCII^hi MECs (16). Reexpression of Aire after a turnoff period was more sluggish but could be achieved after 1–2 wk of Dox withdrawal, particularly when low doses were initially used (Fig. 1 G and not depicted). Because the turnoff with Dox treatment was sharper than the turnon in this model, as in other Dox-controlled transgenic settings, regulated Aire extinction was the strategy primarily used in the studies described in the subsequent paragraphs.

It was recently reported that Aire is expressed in the stromal cells of peripheral lymphoid organs (17, 18). Aire-encoding transcripts from the iA transgenes were detected almost exclusively in the thymus, with very low expression in the spleen and LNs (<3–5% of thymus values; Fig. 1 H, top). In the CD45⁻gp38⁺UEA-1⁺ stromal cell population described by Lee et al. (17), Aire transcripts were 10× more abundant in iA^ON than in WT mice but were maintained in iA^OFF animals, likely because of the leakiness occasionally observed with Dox-controlled systems (Fig. 1 H, bottom left). Similarly, significant transgene-driven Aire transcripts were detected in CD45⁻ spleen cells that correspond to the eTAC cells of Gardner et al. (18), but this expression was also only partially affected by Dox, with residual expression in treated iA mice comparable to that of WT (Fig. 1 H, bottom right). No expression was detected in other tissues, including the lung, kidney, heart, or brain (Fig. 1 H, top, and not depicted).

The key test of complementation was whether the iA transgenes could protect, in a Dox-sensitive manner, from Aire disease. Mice with the Aire-null mutation on the NOD background develop wasting disease, with a median survival time of 55 d (19) because of multiorgan lymphocytic infiltration that has particular predilection for the lung, exocrine pancreas, eye, and ovary/prostate. In contrast, NOD.Aire^−/− mice are protected from the insulitis that normally affects the NOD strain (19, 20). These observations were reproduced in our study (Fig. 2). Aire deficiency was indeed complemented by the transgenes, as iA mice in which the reporter was continuously on (iA^ON) survived to 100 d of age and beyond, developing the typical insulitis of NOD mice, rather than exocrine pancreatitis. Only rare isolated and mild infiltration of the lung or gonads was detected over the course of multiple experiments, comparable with that observed in similarly aged WT mice. Aire turnoff by continuous Dox exposure (iA^OFF) restored the full array of disease manifestations, confirming that Aire disease depends on regulated Aire expression. As expected, Dox treatment of either WT or KO mice had no effect.

The functional importance of Aire in stromal cells of peripheral lymphoid organs remains conjectural (17, 18), although the transferability of the autoimmune disease by grafts of Aire-deficient thymi (15) suggests that it is not relevant for Aire disease. Consistent with this notion, there was no correlation between peripheral Aire expression, which was largely insensitive to Dox (Fig. 1 H), and protection from disease, which was fully sensitive (Fig. 2). This peripheral expression is thus not relevant to the present study, although it may not be a coincidence that the same peripheral cells that normally express low levels of Aire are also those that show leaky expression of the transgenic reporter.

Thus, the inducible Aire system provided us with a means to vary the expression of and tolerance to Aire-controlled self-antigens. To dissect the influence of Aire expression on tolerance throughout life, we examined the incidence of wasting disease and autoimmune infiltration in iA mice in which Aire was expressed over different time intervals (Fig. 3). Dox treatment should result in clearance of peripheral tissue antigens within a matter of days, from their own molecular turnover and because Aire-positive cells turn over rapidly, with a half-life of 3–4 d (16). Turning Aire off before birth resulted in full-blown wasting disease and multiorgan autoimmunity (Fig. 3, C and D), but keeping it on for an increasing number of days after birth resulted in a progressively marked increase in the median survival age and in the protection from multiorgan infiltration (Fig. 3, E–H). Mice expressing Aire only until 21 d (iA^ON(0–21)) were indistinguishable from iA^ON controls, including the appearance of “normal” insulitis (Fig. 3 H). Several iA^ON(0–21) mice that were observed for >250 d remained similarly free of disease (unpublished data). These results demonstrate that Aire expression during the neonatal period is essential for self-tolerance but that it is dispensable beyond weaning age. This conclusion was confirmed by the reverse experiment, in which Aire was turned on only after birth; in spite of continued Aire expression for the remainder of their lives, these animals showed wasting and autoimmune pathology (Fig. 3 I). Thus, exposure to self-antigens during the neonatal period is both necessary and sufficient to prevent the devastating autoimmunity associated with Aire disease.

These data imply that escape of autoreactive lymphocytes into the periphery is more tolerable after weaning age than during the neonatal period, and they raise two connected but different questions: what is special about the neonatal period, and why don't the intolerant T cells emerging from the adult Aire^OFF thymus induce pathology? Neonatal T cells are peculiar in several respects, any of which could be relevant in the present context: their TCRs are largely devoid of N-region diversity and have a different intrinsic affinity for major histocompatibility complex molecules (21–23); they include relatively few CD4⁺Foxp3⁺ regulatory (T reg) cells (24), and they are able to access the parenchymal tissues (25). Finally, neonatal thymocytes emerge into peripheral T cell compartments that are largely empty and, thus, show lymphopenia-driven proliferation, a state which can provoke effector-like potentialities (26–30). Several experiments were set up to explore some of these explanations. First, iA^ON(0–21) mice were lethally irradiated and reconstituted with syngeneic bone marrow while Aire was turned (and maintained) off (Fig. 4 A), a setting which should recreate in an adult mouse the lymphopenic conditions of the neonate. In this case, iA^ON(0–21) mice did develop wasting disease and multiorgan lymphocytic infiltration (Fig. 4 A). Unmanipulated iA^ON(0–21) mice remained free of disease, and all other controls behaved as expected (Fig. S1). Similarly, partial T cell lymphopenia was induced in iA^ON(0–21) mice by injection of depleting anti-CD4/CD8 mAbs at 3 wk of age at the time of Aire turnoff. Multiorgan pathology ensued, albeit in the absence of wasting disease (Fig. 4 A; and Fig. S2 shows additional controls). Third, Aire^−/− neonates were infused with boluses of adult spleen and LN cells (10 × 10⁶ T-enriched cells at days 1 and 7), thus reducing their lymphopenic state. Distinct, albeit not complete, protection was achieved with this treatment, with less frequent and delayed wasting and a reduced intensity of histological lesions. Such protection could not be achieved by the transfer of 3 × 10⁵ FoxP3⁺ T reg cells alone, which is equivalent to the number of T reg cells present within the 10 × 10⁶ T cell–enriched cells' bolus (Fig. 4 B and Fig. S3). These results are not compatible with a major influence of the neonatal TCR repertoire or with an intrinsic “openness” of neonatal tissues. Neither are they compatible with an explanation based solely on the relative dearth of T reg cells at birth because after irradiation and reconstitution from bone marrow precursors, T cell populations are restored with largely normal proportions of T reg cells (31, 32). An explanation which is purely based on T reg cell deficiency would not account for the autoimmune pathology in reconstituted iA^ON(0–21) mice. Rather, they indicate that lymphopenia provides a permissive environment for the initiation of autoimmunity, because T cells spontaneously acquire effector functions and/or are able to enter tissues, in which escape of autoreactive cells from the thymus has deleterious consequences.

The autoimmune manifestations of Aire disease have striking similarities to those induced by neonatal thymectomy (d3Tx): widespread but slowly progressing autoimmunity, interindividual variability, and a clear parallel in the target organs preferentially affected on different genetic backgrounds (19, 33). The autoimmunity induced by d3Tx was initially thought to reflect defective T reg cells, but it has since been realized that the numbers and activity of T reg cells are normal in d3Tx mice (34, 35). A unifying interpretation is that the T cell compartment requires exposure to self-antigens in the thymus during the postnatal period. Eliminating this exposure by either thymectomy or shutdown of Aire expression results in incomplete tolerance. This tolerogenic exposure can be achieved only in the thymus and not in the periphery, a notion which is also supported by the classical studies of embryonic quail/chick embryonic grafts (36); pre/postnatal exposure to peripheral tissues, even as massive as an entire limb, cannot replace the contribution of thymic epithelium. In addition, the lymphopenic state inherent to a growing T cell pool exacerbates this dependency, which is consistent with the autoimmune manifestations often observed in lymphopenic situations, such as the “immune reconstitution inflammatory syndromes” observed after T cell depletion in HIV or other viral infections (37). In contrast, a fully established adult T cell pool can provide enough control, via regulatory cells or other mechanisms of peripheral tolerance, to buffer incoming self-reactive cells arising from an Aire-deficient thymus. But these buffers can be established only if thymic exposure to self has occurred previously in the formative stages.

The TOA construct contains Aire cDNA (positions 36–1879 from GenBank/EMBL/DDBJ under accession no. NM_009646.1) and an intron and polyadenylation site from the rabbit β-globin gene in the tetO-SB plasmid. The fragment was prepared for microinjection into NOD embryos by AatII and Eco47III digestion and purification as previously described (38). Progeny were screened by a PCR assay specific for transgenic Aire (primers: 5′-GCAACGTGCTGGTTGTTGTG-3′ and 5′-TGACTCCAAGTTGCCATCTG-3′). TA mice (14) were backcrossed more than seven times onto NOD, fixing NOD alleles at A^g7, Idd3, and Idd5.2. Aire^−/− mice on the NOD background have been previously described (19). Dox hydrochloride (Sigma-Aldrich) was provided in drinking water at 100 µg/ml, unless otherwise indicated. Equal sweetener was added to the water to mask Dox's bitter flavor, and water was changed three times per week. For prenatal or neonatal treatment, Dox-supplemented water was provided to pregnant or lactating mothers. To follow wasting disease, mice were weighed three times per week and sacrificed when they lost 15–20% body weight relative to littermates. For bone marrow chimeras, recipients were irradiated (9 Gy), reconstituted with 10 × 10⁶ bone marrow cells from a syngeneic WT mouse, and kept on with sulfatrim for the remainder of the experiment. For mAb depletion of CD4 and CD8 T cells, mice were treated i.p. with 0.5 mg each of anti-CD4 (clone GK1.5) and anti-CD8 (clone 2.43) every 3 d for a 2-wk period. All mice were maintained under specific pathogen-free conditions at the Harvard Medical School facility (protocol 02954 approved by Harvard Medical School's Institutional Animal Care and Use Committee). For lymphocyte transfer experiments, spleen and LN cells were depleted of non–T cells with the negative selection T cell isolation reagent (Miltenyi Biotec) and 10 × 10⁶ cells were injected i.p. into neonates. For T reg cell transfer experiments, T reg cells from spleen and LN of female Foxp3-GFP (NOD) transgenic mice were isolated by sorting (B220⁻CD4⁺GFP⁺) and 3 × 10⁵ T reg cells were injected i.p into neonates.

RNA was prepared from tissues or sorted cells with Trizol (Invitrogen), treated with DNase (Applied Biosystems) for 30 min at 37°C to remove contaminating DNA, and cDNA synthesized with random or gene-specific priming (GSP). For GSP, the 5′-TCATCTCTACCAGGTATAGTGAC-3′ primer was used. cDNA was used as a template for amplification by Taqman RT-PCR (Mx3000P; Agilent Technologies). Both WT and transgenic Aire transcripts were detected using the 5′-GTACAGCCGCCTGCATAGC-3′ and 5′-CCCTTTCCGGGACTGGTT-3′ primers and the Fam-5′-CTGGACGGCTTCCCAAAAGATGT-3′-Tamra probe. For detection of Aire transcripts only from the TAO transgene, the 5′-TGAGAACTTCAGGCTCC-3′ and 5′-GGAAAGGCACTGTCTATGGC-3′ primers and the Fam-5′-CGTCTGCTGAGGCTGCACCG-3′-Tamra probe were used. Whole thymus RNA from WT or iA^ON (depending on the experiment) was used as a reference. Data were visualized and analyzed with MxPro software (Agilent Technologies), using the standard curve method.

Suspensions of thymic stromal cells were prepared and stained for intracellular Aire as previously described (16). For stromal cells from peripheral LNs and spleen, small incisions were made in the tissues, which were agitated in RPMI supplemented with 5% FBS to release free lymphocytes. The remaining tissue was digested in RPMI 5% FBS with 0.5 mg/ml collagenase (Roche), 1 mg/ml dispase (Roche), and 1 mg/ml DNAase (Sigma-Aldrich), and the released cells were stained for FACS analysis.

Anti–mouse gp38–Alexa Fluor 647 and rat anti–mouse Aire (5H12 clone) were gifts from S. Turley (Dana Farber Cancer Institute, Boston, MA) and H. Scott (Walter and Elisa Hall Institute, Melbourne, Australia), respectively. Anti–keratin-5 and –8 antibodies were obtained from Covance. Flow cytometric analysis was performed on an LSRII, sorting on a FACSAria (BD). Data were analyzed with FlowJo software (Tree Star, Inc.).

Figs. S1–S3 show additional curves for control animals corresponding to Figs. 3 and 4.

We thank Hamish Scott, Shannon Turley, Nora Mauermann, Shohei Hori, Polly Matzinger, Shannon Turley, and the Aire Group for discussion and reagents, John Stockton for microinjections, Joyce LaVecchio and Giri Buruzala for flow cytometry, Chris Campbell and Angela Wilcox for genotyping, and Vanessa Tran and Kimie Hattori for help with mice.

This work was supported by R01 DK60027 and Young Chair funds to D. Mathis and C. Benoist, and by the Joslin's NIDDK-funded Diabetes and Endocrinology Research Center.

The authors have no conflicting financial interests.