Experiments with animals complied with NIH’s Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at Michigan State University. Pgr^Cre/+ mice (7) and Pgr^f/f (8) were a gift from Dr. John Lydon (Baylor University) and Dr. Jay Ko (University of Illinois, Urbana Champaign). Breeding pairs of Rosa^mTmG/mTmG (9), Foxn1 ^Cre/+ (10), and wild type (WT) C57BL/6J and Balb/cJ were purchased from Jackson Laboratories (Bar Harbor, ME). All animals were on the C57BL/6J background and kept in conventional caging with a 12-hour light/dark cycle and fed ad libitum. Female transgenic animals (8-10 weeks old) were mated with 11-20-week-old wild type (WT) C57BL/6J or Balb/c males for fertility studies. Females were sacrificed prior to pregnancy and on day of gestation (GD) 6.5, 14.5, 16.5, and 18.5. The presence of the mucus plug was designated GD0.5.

Total RNA was extracted from thymus obtained from C57BL/6J non-pregnant and pregnant (GD16.5) animals to measure transcript levels of genes. After euthanization, thymi were harvested, frozen in liquid nitrogen, and stored at -80°C. RNA was extracted using TRIzol reagent (ThermoFisher, Waltham, MA, USA). Whole thymi were submerged in 1mL of TRIzol and dissociated using Omni Bead Homogenizer (Bead Rupter 12, catalogue number: 19-050A, Omni International Inc., GA, USA). Homogenate was transferred to Phase Lock Gel-Heavy tubes (1.5mL), 0.2mL of chloroform was added per 1mL of TRIzol Reagent, then centrifuged for 10 minutes at 12,000g, with centrifuge set at 4°C. The clear, aqueous phase was transferred to a 1.5mL sterile Eppendorf tube, and RNA was precipitated by adding 0,5mL of isopropyl alcohol per 1mL of TRIzol Reagent. Samples were mixed by inversion, then incubated at room temperature for 10 min, and centrifuged at 12,000g for 10 minutes at 4°C. The RNA pellet was washed with 1mL of 75% ethanol, centrifuged at 7,500g for 5 minutes at 4°C. The resultant RNA pellet was dried at room temperature and dissolved using sterile RNAse free water. RNA quality was assessed using a Nanodrop Lite spectrophotometer (ThermoFisher, Waltham, MA, USA). Quantitect Reverse Transcription Kit (Qiagen, Hilden, Germany) was used for cDNA synthesis from the RNA template followed by real-time quantitative polymerase chain reaction (RT-qPCR) using Taqman Probes ( Table 1 ) on a QuantStudio 5 instrument for 35 cycles (Applied Biosystems, Waltham, MA, USA). The Taqman probes are validated by the manufacturer to have amplification efficiencies close to 100%. Fold change was calculated using the delta-delta cycle threshold (ddCT) method normalized to Gapdh using a custom-made RStudio package, tidyQ (https://soohyuna.github.io/tidyQ/)

For immunofluorescence, tissues were fixed in 4% PFA at 4°C, transferred to 30% sucrose overnight, embedded in OCT medium (Sakura Finetek, Torrance, CA), and frozen in a bath of 2-methylbutane chilled in liquid nitrogen. Tissue cryosections (5µm) were prepared and stained using the following antibodies: rabbit monoclonal anti-progesterone receptor (PGR) (SP2, catalogue number: SAB5500165-100UL, tested concentration at 1:400, Sigma-Aldrich, St. Louis, MO); cytokeratin 8 (cTEC marker) (Troma-1, 1mL supernatant, stock concentration: 50ng/mL; tested concentration: 1ng/mL, Developmental Studies Hybridoma Bank, IA, USA), and cytokeratin 5 (mTEC marker) (stock concentration: 1mg/mL; tested concentration: 2ug/mL, Biolegend, San Diego, CA). Goat anti-rabbit IgG-Alexa Fluor 546 and goat anti-rat IgG-Alexa Fluor 488 (stock concentration: 2mg/mL, tested concentration: 0.01mg/mL, ThermoFisher, Waltham, MA, USA) were used as secondary antibodies, and PGR+ areas from stained sections were quantified using CellProfiler (11).

To image the entire section of the non-pregnant and pregnant thymus, 5µm thymic sections were first stained with anti-cytokeratin 5 (medulla), anti-cytokeratin 8 (cortex), counterstained with DAPI (Vectashield), and stitched using Leica TCS SP8 X Confocal Laser Scanning Microscope System at 10x magnification. The resulting LIF files were imported into Imaris v9.2.1 (Bitplane). Under surpass mode, surface was created for areas positive for cytokeratin 5, cytokeratin 8 and DAPI using the surface module. Summary statistics were exported from each file and compiled to calculate percent coverage by cytokeratin 5 positive areas in comparison to DAPI ( Figure 1E ).

Heterozygous Foxn1^cre/+ females were crossed to homozygous Pgr^f/f males. Resulting F1 females were crossed to Pgr^f/f males to create experimental Foxn1^Cre/+;Pgr^f/f (Pgr^d/d ) and control Foxn1^+/+ ;Pgr^f/f (Pgr^f/f or Pgr^f/+ ) F2 females. To create reporter lines, female Foxn1^Cre/+ or Pgr^Cre/+ were crossed with dual-fluorescence Rosa26^mTmG/mTmG males to generate Foxn1^Cre/+ ; Rosa26^mTmG/+ and Pgr^Cre/+;Rosa26^mTmG/+ offspring. DNA was extracted from the tail, spleen, ovary, and thymus to genotype for the presence/absence of the floxed Pgr gene. Primer sequences to genotype Pgr^f/f animals were provided by the group of Dr. Jay Ko (University of Illinois at Urbana-Champaign). Three primers were used to generate WT band at 226bp, flox band at 276bp and a KO band at 370bp in length:

Polymerase Chain Reaction (PCR) protocol is as follows: Step 1: 5 minutes at 94°C, Step2: 1 minute at 94°C, Step 3: 1 minute at 57°C, Step 4: 2 minutes at 72°C. Steps 2 to 4 were repeated for 35 cycles, followed by a hold for 10 minutes at 72°C. PCR products were run on a 2% agarose gel and exposed under UV light using a MyECL imager (ThermoFisher, Waltham, MA, USA). As per the description from Jackson Laboratory where Foxn1^cre strain was purchased (stock number: 018448), Foxn1 is also expressed by the skin keratinocytes. The F1 genotype results show the presence of WT, KO and flox alleles due to the tail digest containing mixture of Foxn1+ cells, which possess alleles of WT/KO, and the rest, which remain WT/FL. We leveraged this information to identify Pgr^d/d animals using tail snips, genotype of which show both a KO and flox bands from the tail snip, indicating that Foxn1^cre+ cells are KO/KO, while rest are FL/KO. This is also true in the ovary, uterus, spleen, and thymus as shown below. Thymus also shows FL/KO because of the presence of thymocytes and other cell types in the thymus that do not express Foxn1, and thus remain FL/KO ( Supplementary Figure 1 ).

Female control (Pgr ^f/f) and Pgr^d/d mice (6-8 weeks old) were bred to major histocompatibility complex (MHC)-matched (C57BL/6J) or mismatched (Balb/c) males for 4-6 cycles of continuous pregnancy. Pups were counted and weighed at weaning (21 days postpartum), sexed using ano-genital distance, then euthanized.

Sera were collected from nonpregnant (NP) and pregnant females at GD16.5 via cardiac puncture and assayed for progesterone and estradiol by enzyme-linked immunosorbent assay (ELISA) at the University of Virginia Center for Research in Reproduction Ligand Assay and Analysis Core (Charlottesville, VA, USA). Progesterone ELISA (Catalog no. IB79105, IBL, Minneapolis, MN) had an assay range of 0.3-40 ng/ml, and intra-/inter-assay coefficients of variation of 6.0% and 10.1%, respectively. Estradiol (catalog no. ES-190S-100; Calbiotech, Spring Valley, CA) was measured with an assay range of 3-300 pg/ml and intra-/inter-assay CVs of 5.7% and 10.2%. Further details of assays are available (https://med.virginia.edu/research-in-reproduction/ligand-assay-analysis-core, last accessed June 10, 2021).

To prepare a single cell suspension of thymocytes, thymus was dissected from the mouse, placed on top of 70 µm mesh, pressed through the filter using a back of 1mL syringe, collected into a 15mL tube, and centrifuged at 400g for 5 min. Single cell suspensions were filtered through 70um mesh, resuspended in flow staining buffer (FSB, 5% fetal bovine serum in PBS) and incubated with ACK (Ammonium-Chloride-Phosphorus) buffer to lyse red blood cells. Cell number was determined using a Countess II (ThermoFisher, Waltham, MA, USA). For viability, all cells were stained with Live/Dead™ Fixable Blue Dead Cell Stain kit (Catalog number: L23105, Thermofisher, Waltham, MA, USA) on ice for 30 min followed by Fc receptor block (Biolegend, San Diego, CA, USA) on ice for 20 min. Prior to cell surface staining, cells were stained with anti-CCR7 antibody at 37°C for 30 min. Post washing in FSB twice, cells were stained using a cocktail of cell-surface antibodies for 30 min on ice. All samples were analyzed using spectral flow cytometry, Cytek^(R)Aurora (Cytek Biosciences, CA, USA), with 5 laser capability, at Michigan State University Flow Cytometry Core. Flow cytometric analysis software, Kaluza v1.3 (Beckman Coulter, IN, USA) was used for analysis and generation of gating strategy. A full list of the antibodies is available in Table 2 . The following antibodies were purchased from Biolegend: CD4 (GK1.5), CD8 (53-6.7), CD25 (PC61), CD44 (IM7), CD69 (H1.2F3), and CCR7 (4B12).

Power analysis using G*Power was conducted to determine the sample size for all experiments. RStudio software was used to conduct statistical analysis and to graph all data. The Shapiro test was used to assess normality, which was followed by either parametric (normally distributed) or non-parametric (not normally distributed) test. For normally distributed data, Student’s t-test was used to compare the means of two independent samples; one-way ANOVA was used for more three or more independent samples. For non-parametric data, Wilcoxon and Kruskal-Wallis tests were used. In all incidences, P values less than 0.05 were deemed statistically significant.

We first assessed changes in thymic architecture, mass, and cellularity across pregnancy. C57BL/6J females were paired with C57BL/6J males and sacrificed on GD6.5, 14.5, or 18.5. We confirmed that thymic mass and cellularity decreased progressively across murine pregnancy (Mass: 69.5 ± 6.9 mg vs. 14.7 ± 4.0mg in NP and GD18.5, respectively; Cellularity: 125x10⁶ ± 64x10⁶ vs. 23x10⁶ ± 4.7x10⁶ cells in NP and GD18.5, respectively) ( Figures 1A–C , p<0.05). To compare the changes in the thymic architecture between virgin and pregnant female mice, we performed immunofluorescence on thymic sections with anti-cytokeratin 5 (K5) and anti-cytokeratin 8 (K8) antibodies, which define medullary and cortical thymic epithelial cells, and the areas they occupy, respectively. Using IMARIS software to quantify the areas of the cortex and medulla, we observed shrinkage of cortical area together with relative expansion of medullary area with pregnancy ( Figure 1E ).

Because estrogen and progesterone play an important role in pregnancy-associated thymic involution, we analyzed the gene expression of steroid hormone receptors in the thymus. We used RT-qPCR to quantify mRNA expression of Pgr, Esr1, Esr2, Pgrmc1, and Pgrmc2 in the thymus of NP and GD16.5 C57BL/6J females. Pgr mRNA significantly rose in WT thymus at GD16.5 as compared to NP (p = 0.0041), whereas transcripts for Esr1, Esr2, Pgrmc1, Pgrmc2 remained unchanged ( Figure 2A , p > 0.05). We also conducted immunofluorescence to understand the spatial distribution and dynamics of PGR protein expression throughout gestation. PGR expression was sparse in NP thymus but increased dramatically by GD14.5 ( Figure 2B , p < 0.0001, Supplementary Figure 2 ). Additionally, PGR immunofluorescence using anti-PGR antibody showed nuclear localization in large reticular cells suggestive of thymic epithelial cells (TECs), found mostly in the cortex ( Figure 2C ).

We confirmed the rise in cellular expression and location of Pgr using a dual reporter mouse model, Pgr^cre/+;Rosa26^mTmG/+ , in which cells that have expressed Pgr gene are identified by membrane-associated GFP reporter ( Figure 3A ). Macroscopic visualization showed a striking increase in Pgr-driven GFP expression by GD14.5 as compared to NP thymus ( Figure 3B ). This increase could also be seen at the cellular level, which confirmed the lack of expression in thymocytes, with GFP reporter expression consistent with Pgr expression by thymic epithelial cells ( Figures 3C, D ). No GFP was observed in control Pgr^+/+;Rosa26^mTmG/+ GD14.5 thymi ( Supplementary Figure 3 ).

To test the significance of TEC-PGR in pregnancy-associated thymic involution, we created a conditional deletion model in which Pgr gene is deleted specifically from TEC. To demonstrate that Foxn1+ TECs expressed PGR, we crossed Rosa26^mTmG/+ mice (9) with Foxn1^cre/cre mice, in which Cre expression is driven by the TEC-specific promoter for Foxn1 (10, 12). Combining this model with PGR immunofluorescence confirmed the expression of PGR predominantly by Foxn1+ TECs in pregnancy ( Figure 4A ).

To ablate Pgr expression from Foxn1+ TECs, we crossed Foxn1^cre/cre mice with Pgr^f/f mice, and then backcrossed F1 females to Pgr^f/f males. This cross abolished thymic PGR protein expression in Pgr-deficient (Pgr^d/d ) mice but not in flox controls (Pgr^{f/f or f/+} ); uterine expression of PGR protein remained unaffected ( Figure 4B ). The serum concentration of estradiol and progesterone were also comparable between controls and Pgr^d/d animals ( Supplementary Figure 4A ). We then established timed mating of Pgr^f/f and Pgr^f/+ control and sacrificed animals at GD16.5. In pregnancy, the thymus weight of GD14.5-16.5 WT and floxed control animals both decreased significantly (p < 0.0001), by approximately 2.8-fold and 2.1-fold, respectively. Thymi of Pgr^d/d animals also decreased (p < 0.001), but to a lesser extent (1.1-fold). Similarly, the cellularity of the thymus dropped dramatically in both WT and floxed control animals during pregnancy (5.6-fold and 5-fold, respectively; p < 0.00001). Cellularity of Pgr^d/d females also decreased during pregnancy, but to a lesser extent (1.5-fold) ( Figure 4C ). We also compared the thymic weight between genotypes. While the thymic weight was comparable between genotypes in non-pregnant state, in pregnancy, we found that thymic weight of Pgr-deficient females was significantly higher compared to our pregnant controls ( Figure 4D , p < 0.00001). Thymic cellularity corresponded with thymic weight as well. Overall, there was approximately a 5.6-fold change in thymic cellularity between NP and GD14.5-GD16.5 for C57BL6WT, 5-fold change for flox controls, and only 1.5-fold change for our Pgr-deficient females. We did not observe significant difference between GD14.5 and GD16.5 thymic cellularity in any of the genotypes ( Supplementary Figure 5 , p > 0.05).

To determine whether the thymic architectural changes were also apparent in our transgenic mouse model in pregnancy, we stained the thymus of Pgr^d/d and flox controls at GD16.5 and compared the distribution of the cortex to the medulla ( Supplementary Figure 4B ). Our flox control thymus at GD16.5 resembled that of the wild-type thymus at similar gestation of pregnancy ( Figure 1 ). Interestingly, the architecture of the Pgr^d/d thymus at GD16.5 resembled that of a NP wild type, with medullary islands distributed throughout the cortex, instead of coalescing in the middle. This outcome suggests a relationship between thymic involution and changes in thymic architecture with pregnancy. In the absence of thymic involution, the architectural changes do not seem to occur.

In order to determine if thymocyte development was impaired in our Pgr-deficient females, we conducted flow cytometry from the thymus and measured the proportions of single positive (SP) CD4, CD8 T cells as well as double negative (DN) and double positive (DP) thymocytes from C57BL6WT, flox controls and our Pgr-deficient females at NP and pregnant state. Despite the lack of thymic involution in our Pgr-deficient females, proportions of CD4, CD8, DN, and DP cells were comparable between NP and late gestation of pregnancy in all genotypes ( Figures 5A–C , p > 0.05). These results demonstrated that the pregnancy-associated thymic involution is dependent on the expression of Pgr by Foxn1+ TEC, but the lack of involution does not impact the overall proportions of T cells in the thymus.

We next explored whether progesterone regulates expression of genes that could cause changes in thymic function and architecture. We were particularly interested in whether ablation of Pgr impacts gene expression by mTECs because Pgr-deficient thymus at pregnant state resembled that of the non-pregnant thymus. Therefore, we compared thymi of Pgr^f/f and Pgr^d/d NP or GD16.5 mice for expression of genes that regulate self-antigens (Aire, Fezf2) as well as several tissue-restricted antigens that are found in gestational tissues. We also included Pgr to confirm our conditional knock out phenotype, and Gad1, an AIRE-independent gene to determine whether its expression changed with pregnancy. In our flox controls, both Aire and Fezf2 increased significantly, which appeared to be partially (Aire) and completely (Fezf2) dependent on progesterone ( Figure 6 , p < 0.001). In addition, genes regulated by Aire, including the fetal/placental genes Hbby (p < 0.01) and Trap1a, (p < 0.001) and the decidua-associated gene Prl8a2 (p < 0.01), were all significantly upregulated during pregnancy; this effect was ablated for Hbby and Prl8a2 in pregnant Pgr^d/d mice ( Figure 6 , p > 0.05). Finally, genes that regulate thymocyte trafficking (Ccl19, Ccl21a) and medullary size (Tnfsf11/RANKL, Tnfrsf11a/RANK) were investigated. In our controls, Ccl19 (p < 0.01), Ccl21a (p < 0.001), and Tnfrsf11a (p < 0.01) mRNA rose significantly during pregnancy, while Tnrsf11 mRNA remained unchanged ( Figure 6 ). These changes were not affected in Pgr^d/d females, however, mRNA for Tnfrsf11a (p < 0.05) and Tnfrsf11 (p < 0.01) significantly dropped in these mice between NP and GD16.5 of pregnancy.

Previous work showed impaired pregnancy when thymectomized female mice were transplanted with a thymus lacking Pgr (5). We asked whether pregnancy is similarly impaired in our TEC-specific Pgr conditional KO model. We saw that the mean litter size across syngeneic pregnancies was slightly diminished at GD16.5 in our Pgr^d/d animals; however, the difference was not statistically significant [ Supplementary Figure 6 , 8.42 ± 1.60 (Pgr^f/f and Pgr^f/wt ) VS. 7.54 ± 2.62 (Pgr^d/d ), p = 0.81]. Similarly, the proportion of resorbed fetuses were not different between controls (9.32%, or 11/118 implantation sites, of 14 control dams) and Pgr^d/d (14.45%, or 12/83 implantation sites, of 11 Pgr^d/d dams, p>0.05, chi-square analysis). We also assessed overall fertility by pairing control or Pgr^d/d females with either C57BL/6 males (syngeneic pregnancy) or Balb/c males (allogeneic pregnancy) for long term fertility trial. The average litter size from Pgr^d/d dams was significantly smaller than that of control dams when mated to allogeneic, but not to syngeneic sires ( Figure 7A ). Correspondingly, the mean pup weight at weaning was higher in these pregnancies; this was true regardless of sex ( Figure 7B ). In addition, cumulative numbers of pups across three pregnancies were smaller in these pregnancies (Controls, 32.76 ± 0.88 vs. 25 ± 1 pups; n=3 dams; p < 0.005) ( Figure 7C ). Collectively, the data indicate that fertility in allogeneically-bred Pgr^d/d females is reduced in comparison to allogeneically-bred control females.