C57BL/6J, MRL/MpJ-Fas^lpr/J, NOD mice (The Jackson Laboratory, Bar Harbor, ME, USA), and Foxp3^RFP reporter mice (30), kindly provided by S. Huber, were housed and maintained under specific pathogen-free conditions. Mice were mated and the presence of a vaginal plug was considered day 0.5 of pregnancy (E0.5). On day 18.5 (E18.5) mice were treated by i.p. injection of 0.1 mg betamethasone (Sigma-Aldrich, Germany) in PBS or vehicle control (PBS). When indicated, 4- to 5-week-old animals were given 0.1 mg betamethasone i.p. or vehicle control (PBS) 24 h prior to organ harvesting. All animal experiments were performed in accordance with national and institutional guidelines on animal care and ethics.

Thymi, lymph nodes and spleens were harvested and single-cell suspensions prepared by mechanical disruption and filtering through a 70 µm nylon mesh. Red blood cells were lysed when necessary and Fc receptors were blocked using anti-CD16/32 Abs prior to staining. Anti-mouse Abs used in this study were: anti-CD3e eFluor 450 (500A2), anti-CD4 APC-eFluor 780 (RM4-5) and anti-CD8α PE-Cy7 (53-6.7) from eBioscience (San Diego, CA, USA); anti-CD4 FITC (RM4-5), anti-CD4 BV421 (GK1.5), anti-CD8α PerCP-Cy5.5 (53–6.7), anti-CD25 PE (3C7), and anti-TCRβ chain PerCPCy5.5 (H57-597) from BioLegend (San Diego, CA, USA), and anti-CD3ε FITC (145-2C11) from BD Biosciences (San Jose, CA, USA).

Analysis of TCR Vβ usage was performed using 15 FITC-conjugated monoclonal antibodies recognizing the TCR chains Vβ2, 3, 4, 5.1/5.2, 6, 7, 8.1/8.2, 8.3, 9, 10^b, 11, 12, 13, 14, and 17^a TCRs (BD Biosciences, San Jose, CA, USA). The percentage of Vβ⁺ cells was determined in each subset after gating on CD4⁺, CD8⁺, and DN cells.

Data were collected on a Flow Cytometer (FacsCanto II, BD Biosciences, San Jose, CA, USA) and analyzed using FlowJo Software (Tree Star, Ashland, OR, USA). Cell sorting of thymocyte populations was performed on a FACS Aria IIIu (BD Biosciences, San Jose, CA, USA) carried out by the FACS sorting Core Unit of the University Medical Center Hamburg-Eppendorf.

One million thymocytes from 5-weeks-old untreated C57B1/6J mice were cultured in RPMI-1640 (Gibco, CA, USA) in a 96-well round bottom plate (Thermo Fisher Scientific, Germany) and incubated with increasing concentrations of betamethasone (0.1–100 nM) with or without the addition of 1 µg/ml mifepristone (RU486, Sigma-Aldrich). After 16 h at 37°C, cells were harvested, washed with 1 × Annexin V Binding Buffer (Exbio, Czech Republic) and subsequently stained for linage markers (CD3, CD4, CD8, CD25). For detection of dead and apoptotic cells, Dead Cell Stain-Pacific Orange (Invitrogen, CA, USA) and Annexin V FITC (BD Biosciences, San Jose, CA, USA) were used according to the manufacturer’s instructions.

Pooled single-cell suspensions from inguinal, axillary, brachial, lumbar, and superficial cervical lymph nodes were stained with the cell proliferation dye eFluor 670^® (eBioscience, San Diego, CA, USA) in order to differentiate proliferating and non-proliferating cells. Shortly, lymph node cells were washed with PBS and resuspended in 2 µM eFluor 670^® dye in PBS for 10 min at 37°C in the dark. Labeling was stopped by adding 4–5 volumes of cold complete medium (RPMI 1640, 10% FCS, 1% penicillin/streptomycin, 2 mM l-glutamine, 50 µM 2-Mercaptoethanol) and incubated on ice for 5 min. Labeled cells were then cultured in complete medium in the presence or absence of IL-2 (100 U/ml). Cells were plated at 1 × 10⁵/well in a 96-well plate (Thermo Fisher Scientific, Germany) for up to 5 days at 37°C in 5% CO₂. After 3, 4, and 5 days of culture, lymph node cells were collected, washed with PBS and stained with Dead Cell Stain-Pacific Orange and anti-TCRβ chain PerCPCy5.5, anti-CD4 APC-eFluor 780, and anti-CD8α PE-Cy7 antibodies. The proliferation of CD4⁺, CD8⁺, and DN cells was evaluated according to eFluor dilution measured by flow cytometry (FacsCanto II, BD Biosciences).

RNA was extracted from sorted T cell subsets or from thymocytes after in vivo or in vitro treatment using the RNeasy Mini Plus kit (QIAGEN, Hilden, Germany) and cDNA was synthesized with the M-MLV Reverse Transkriptase kit (Invitrogen). TaqMan gene expression assay (LifeTechnologies, CA, USA) was used to detect GAPDH (Hs02758991_g1) expression. 18S and FoxP3 expression were determined using SYBR^® green and following primers: 18S forward: 5′-CGGCTACCACATCCAAGGAA-3′ 18S reverse: 5′-GCTGGAATTACCGCGGCT-3′; FoxP3 forward: 5′-GGCCCTTCTCCAGGACAGA-3′ FoxP3 reverse: 5′-GCTGATCATGGCTGGGTTGT-3′.

Statistical analysis of TCR Vβ chain usage was performed with Matlab R2016b (The Mathworks). The fractions of positive cells for each Vβ chain, as well as the remaining fraction of cells that was not positive for any of the measured Vβ chains (other Vβ), were log or square-root transformed to obtain normally distributed data. Using N-way ANOVA, the Vβ chain fractions for each cell type were correlated to treatment and possible interaction terms, with correction for litter size and subject nested within treatment. Upon reaching statistical significance, pairwise comparison with Fisher’s least significant difference correction was performed to identify which Vβ chain(s) were differently expressed upon treatment. For calling statistical significance, alpha of 0.05 was applied in all analyses.

Additional statistical analysis, including unpaired Student’s t-test, Gehan–Breslow–Wilcoxon test, and Mantel–Cox test were performed with GraphPad Prism Software (La Jolla, CA, USA) and are indicated in the corresponding figure legends. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

This study was carried out in accordance with the recommendations of the Declaration of Helsinki for animal experimental investigation and the Principles of Laboratory Animal Care (NIH pub.85–23 revised 1985). The protocol was approved by the local animal ethics committees (ethical approval 119/13 and 122/12 obtained from the state authority of Hamburg, and DMAH8948 obtained from the Generalitat de Catalunya).

We have previously shown in C57BL/6J mice that prenatal betamethasone treatment causes a profound reduction in the thymic volume and cell numbers in the offspring (16). The question that arises is if the massive death and subsequent accelerated replenishment of the thymus has consequences upon T cell repertoire selection and immunity later in life. To explore this, we took advantage of the MRL/MpJ-Fas^lpr (hereafter referred to as MRL/lpr) autoimmunity-prone mouse strain, which spontaneously develops lupus-like glomerulonephritis and vasculitis as result of autoantibody production and immune complex deposition (32). In this strain, we first sought to confirm the effects of prenatal glucocorticoid treatment on the thymus. After treating the pregnant dams (E18.5) with betamethasone (Figure 1A), at postnatal day 1 (PND1) we did not observe any difference in the weight of the pups (Figure 1B), but a drastic reduction in the number of living thymocytes (Figure 1C). Not surprisingly, thymocyte loss was caused by a massive reduction in the CD4⁺CD8⁺ DP compartment and, as a consequence, a compensational increase in the frequency of DN cells (Figures 1D,E) could be observed. This effect was transient, since in the adult offspring the percentage of DP thymocytes was similar in both groups (not shown). Figure 1E shows a direct comparison of the composition of the thymocyte compartment in a sham- (upper row) vs. a betamethasone-treated (lower row) animal. The density plot in the right panels demonstrates the shift from maximal representation of DP cells in the untreated animals to a maximum of DN cells in the animals treated with betamethasone. Importantly, the range of DP cell loss within a litter was highly variable, with some animals displaying marginal effects while others have nearly lost the DP compartment (Figure 1D). This variation is likely the result of different exposure of each individual fetus to betamethasone (16). The frequencies of CD4SP and CD8SP cells remained similar, although we could notice a reduction in absolute cell counts (not shown).

We next asked if the massive perinatal thymocyte death had long lasting consequences on the peripheral T cell compartment. While the MRL strain is prone to autoreactivity, the defect in fas (lpr, lymphoproliferation) leads to uncontrolled expansion and accumulation of autoreactive cells in the peripheral lymphoid organs, and accelerates, rather than initiates, disease (33). These autoreactive cells harbor the phenotype CD3⁺B220⁺CD4⁻CD8⁻ (DN T cells), and steadily increase with disease progression to become the most abundant subset in spleen and lymph nodes. Importantly, the TCR Vβ usage of public clones in the enlarged lymph nodes and in T cells infiltrating the kidneys of diseased animals is limited to a few families, including Vβ2, Vβ6, Vβ8.2, Vβ8.3, and Vβ10, underlining their pathogenic relevance (34–36). To assess if prenatal glucocorticoid treatment had an effect on the autoimmune repertoire, we performed flow cytometric analysis of TCR Vβ chain usage in peripheral CD4⁺, CD8⁺ and DN T cells of young (5- to 7-week old) and aged (15- to 17-week old) offspring of mothers treated with vehicle or betamethasone. While the frequencies of CD4⁺, CD8⁺, and DN T cells at 5–7 weeks of age, before appearance of disease symptoms, were similar in both groups, accumulation of DN T cells in the older animals was clearly more prominent in the group that had received betamethasone prenatally (Figures 2A,B). Moreover, analysis of Vβ chain usage on the different T cell subsets indicated a bias in TCR receptor Vβ expression in these mice, with higher representation of the potentially autoreactive Vβ2- (7.97 vs 8.95%, P = 0.0564), Vβ8.1/8.2- (10.91 vs 12.85%, P = 0.0379), and Vβ10^b- (1.97 vs 2.85%, P = 0.0113) expressing DN T-cells in young animals (5–7 weeks) (Figures 2C,E) and further differences in the frequency of Vβ5.1 CD4⁺ and CD8⁺ cells, Vβ11 in CD4⁺ T cells, and Vβ12 in CD4⁺ T and DN T cells (Table S1 in Supplementary Material). Similarly, the TCR Vβ usage in MRL/lpr animals with advanced disease (at 15–17 weeks of age) was also biased, with differences in the frequency of Vβ14 (7.46 vs 8.46%, P = 0.0114) CD8⁺ T cells and, as in young animals, in the disease-relevant Vβ10^b (4.78 vs 6.18%, P = 0.0389) bearing DN cells (Figures 2D,F; Table S2 in Supplementary Material).

We next sought to determine whether the changes observed in the TCR Vβ repertoire resulted in a higher proliferation, which would reflect the presence of more autoreactive cells. For this, we performed AMLR to measure proliferation of autoreactive T cells in response to endogenous antigens. While no external TCR stimulus was given, IL-2 was added to the cultures to support the incipient proliferation of the few autoreactive precursors. Although not significant, our data show higher proliferation rates of both CD4⁺ and CD8⁺ cells in the betamethasone-exposed animals (Figure 2G). As expected, DN cells showed maximal proliferation, and no differences were found between the treated and non-treated groups, since the accumulating DN cells are chronically activated in both groups.

The lpr defect in this mouse strain leads to a progressive enlargement of the lymphoid organs, enhancing the disease phenotype of the MRL strain (33). Therefore, we would expect that a T cell repertoire biased toward more autoreactivity would result in larger lymphoid organs. In agreement with increased amounts of pathogenic TCR Vβ families and enhanced T cell proliferation, the spleens and lymph nodes were considerably larger in the animals whose mothers had been treated with betamethasone (Figures 2H,I), supporting the concept of a more autoreactive T cell repertoire. However, a higher frequency of autoreactive T cells and enlarged lymphoid organs did not translate in accelerated course of disease, as assessed by the levels of dsDNA autoantibodies (Figure 2J) and proteinuria (Figure S1 in Supplementary Material) from 6 to 16 weeks of age, which were similar in both groups. Of note, prenatal betamethasone treatment resulted in a transiently lower body weight (Figure 2K) and higher production of ds-DNA specific IgG autoantibodies (Figure 2J) at 6 weeks of age, but this difference disappeared soon after. Thus, prenatal treatment of the pregnant females with betamethasone elicits drastic thymocyte death in the postnatal thymus, which results in persistent changes in the TCR repertoire promoting the expansion of autoreactive Vβ families.

Despite the critical role of T cells in the initiation of autoimmunity in MRL/lpr mice, other mechanisms such as the production of pathogenic autoantibodies and renal immune-complex deposition leading to complement activation, contribute fundamentally to disease progression. Therefore, we turned to a second model of autoimmunity with a more direct involvement of T cells in the immune pathogenesis, the NOD mouse model for T1D to assess persisting effects of prenatal betamethasone treatment. This spontaneous model of autoimmune diabetes is characterized by the activation of autoreactive T cells and the destruction of insulin producing beta-cells in the pancreas. After prenatal administration of vehicle or betamethasone to non-diabetic females (E 18.5), glucose levels were measured in the female offspring for the occurrence of T1D until 25 weeks of age (Figure 3A). Surprisingly, we found that fetal glucocorticoid exposure was protective against the development of T1D in females (Figure 3B). At 25 weeks of age, only 25% of the NOD females in the control group were disease-free, while in the betamethasone treatment group it was 78%. Moreover, T1D onset was delayed in the treated group (from 14 weeks of age) when compared to control mice (from 11 weeks of age), consistently with the protective effect of the drug. There were no differences between the two groups in the frequencies of CD4⁺, Treg cells, CD8⁺, or CD3⁺ DN spleen cells at 6 weeks of age (Figure 3C). We next investigated whether the different incidence of disease could be traced to changes in the TCR Vβ repertoire. In the NOD mice, the most commonly found TCR Vβ chains in T cells infiltrating the pancreas are Vβ4 (37), and Vβ12 (38, 39), followed by Vβ8.1, Vβ2, and Vβ6 (40, 41). Our TCR Vβ usage analysis revealed that Vβ12 (6.71 vs. 5.45%, P = 0.0077) CD4⁺ T cells, and Vβ4 (2.66 vs. 2.14%, P = 0.0683) and Vβ6 (9.28 vs. 7.39%, P = 0.0124) CD8⁺ T cells were underrepresented in the spleen of the female NOD mice whose mothers received betamethasone (Figures 3D,E; Table S3 in Supplementary Material). This lower frequency of pathogenic TCRs was reflected in the lower degree of immune cell infiltration in the pancreas of animals that had not yet developed T1D at 25 weeks of age, determined in a second mouse cohort (Figures 3F–H). Here, the incidence was somewhat lower in both groups, but still 46% of sham-treated females were disease-free at 25 weeks, while 62% of the betamethasone-treated had normal glucose levels (P = 0.0847, Figure S2 in Supplementary Material). Therefore, we conclude that a bias in the T cell repertoire caused by prenatal glucocorticoid treatment contributes to protection from T1D in the NOD mice.

In addition to changes in TCR Vβ usage, another possibility that could affect development of autoimmunity is the quantity or quality of Treg cells. Treg cells are less sensitive to glucocorticoid challenge than conventional CD4⁺ cells (42, 43). However, it is not known if prenatal treatment with betamethasone results in enrichment of Treg cells in the newborn and, more importantly, if there are consequences for the peripheral Treg cell compartment later in life. In a first set of experiments we exposed thymocytes from 6 weeks old (not prenatally treated) C57BL/6J mice to increasing concentrations (0.1–100 nM) of betamethasone in vitro and measured the survival of Treg cells compared to conventional CD4⁺ T cells. We used mifepristone (RU486), a synthetic steroid with potent anti-glucocorticoid properties, to confirm the specificity of the betamethasone effects. We found that the relative frequency of surviving CD4⁺CD25⁺ SP thymocytes (which in the naïve mice are considered Treg cells) increased in response to betamethasone in a dose-dependent manner, and this effect was completely inhibited by the addition of RU486 (Figure 4A). More than a threefold increase in the frequency of Treg cells compared to the untreated controls was achieved at 10 nM (Figure 4A). In contrast, conventional CD4⁺ single-positive thymocytes did not significantly change in proportion with increasing concentrations of betamethasone (Figure 4B). To confirm that it is indeed the Treg cell subset the one increasing, we performed RT-PCR on the cultured cells and found increased Foxp3 expression at higher betamethasone concentration (Figure 4C). We obtained similar results when thymocytes derived from untreated Foxp3^RFP reporter mice were treated with betamethasone and Foxp3 expression could be monitored by flow cytometry (Figure S3 in Supplementary Material). To test the effects of glucocorticoids on the frequency of Treg cells in vivo, we treated 6 weeks old mice i.p. with 0.1 mg betamethasone. Analysis of the thymus 24 h post-treatment revealed a net increase in CD4⁺CD25⁺ in the surviving cells (Figure 4D). Finally, to find out if prenatal betamethasone exposure has an effect on the frequency of neonatal Treg cells, we treated pregnant females and analyzed Treg precursor cells in the DP thymocytes and in the CD4SP subpopulations in the offspring at postnatal day 0. As shown in Figure 4E, we could detect an increased frequency of DP CD25⁺ Treg precursor cells and a slightly elevated proportion of CD4⁺CD25⁺ thymic-derived Treg cells 24 h post-treatment (Figure 4F). This neonatal bias, however, did not persist until adulthood, since the percentage of Treg cells in the spleen of animals at 5–7 weeks of age was similar in both groups (Figure 4G). Moreover, no differences were observed in the expression of neuropilin-1 (Figure 4H) or CD62L (not shown), indicating similar percentages of thymic and peripherally induced Treg cells (44), and similar levels of activation in both groups. Proliferation and inflammatory cytokine production of spleen T cells at this age was also comparable in both groups (not shown). Our data demonstrate that Treg thymocyte precursor cells are less sensitive than conventional CD4⁺ precursors to prenatal administration of betamethasone, resulting in a transient increase in the frequency of Treg precursor cells after birth. However, these differences are not persisting, and do not elicit changes in the future Treg cell compartment.

In the two models of autoimmune disease that we have analyzed, disease develops spontaneously, and the TCR repertoire of the pathogenic cells is oligoclonal. In both cases, we have shown that antenatal betamethasone treatment elicits changes in the TCR Vβ usage affecting the autoreactive T cell repertoire. To address if indeed glucocorticoid-induced changes in the TCR repertoire are responsible for this effect, we took advantage of a model in which only few TCRs are engaged in the pathogenic response, namely EAE induced by MOG_35–55 peptide. In contrast to the more polyclonal response in the lupus-prone MRL/lpr mouse and in the NOD mouse model, in this induced model of autoimmunity, half of the CNS infiltrating cells bear Vβ8.1 or Vβ8.2 (45–47). We therefore treated pregnant C57BL/6J females with betamethasone at E18.5, and let the offspring develop normally, until injection of the MOG peptide at 6 weeks of age (Figure 5A). After disease induction, mice were examined daily and scored for clinical symptoms of disease for 30 days. No differences could be seen in disease onset, clinical score or survival between the animals whose mothers had been treated with betamethasone or with vehicle (Figures 5B,C). Analysis of the TCR repertoire revealed significant changes in Vβ chain usage between the two groups, namely affecting Vβ5.1/5.2 (2.73 vs. 3.27%, P = 0.0004) and Vβ7 (2.24 vs. 1.99%, P = 0.0315) in CD4⁺ and Vβ7 (5.79 vs. 5.22%, P = 0.0040) in CD8⁺ T cells. However, the frequency of the disease-relevant Vβ8.1/8.2 chains remained unchanged (Figures 5D,E; Table S4 in Supplementary Material). Our data show that prenatal betamethasone is not altering the highly restricted pathogenic TCR repertoire in EAE, and—consequently—the disease course is similar in both groups of mice.