C57BL6/J mice were purchased from Jackson Laboratory. Foxn1^lacZ/lacZ (Z/Z) and Foxn1^Cre mice were generated by Dr. Manley’s laboratory as described previously and are maintained on a C57Bl6/J background (37, 38). Lin28a^fl/flb^fl/fl mice (Lin28b^tm2.1GqdaLin28b^tm2.1Gqda/J, stock No. 02395) were purchased from the Jackson Laboratory (39). R26iLin28a inducible transgenic mice were originally provided by Eric Moss (Department of Molecular Biology, Rowan University, Stratford, NJ 08084 USA) and transferred from Dr. Jianfu Chen (Department of Genetics, University of Georgia. Current address: Herman Ostrow School of Dentistry of USC) (40). The Lin28a^{fl / fl}b^fl/fl;Foxn1^Cre TEC-specific deletion of Lin28a and b and R26iLin28a;Foxn1Cre TEC-specific overexpression of Lin28a mice were generated by crossing the Lin28a^fl/flb^fl/fl strain or the R26iLin28a strain with the Foxn1^Cre mouse strain, respectively. All analyses were performed using littermate animals whenever possible. Data from both male and female mice were combined because no difference was detected based on sex. All mice were maintained in a pathogen-free facility at the University of Georgia. The experiments were approved by the University of Georgia Institutional Animal Care and Use Committee.

For thymocyte analysis, freshly isolated thymocytes in suspension (1×10⁶) were used for each sample. Cells were blocked with an anti-CD16/32 (Clone: 93) antibody before staining. For a phenotypic profile of thymocyte subsets, the thymocytes were incubated with anti-CD4 APC-Cy7 (GK1.5), anti-CD8 PE-cy7 (53-6.7), anti-CD44 allophycocyanin (APC) (IM7), and anti-CD25 PerCP (3C7) on ice for 20 min.

For TEC isolation, one thymic lobe (fetal thymi were mixed) was cut into approximately 1 mm³ sections and gently washed in 2% FBS RPMI 1640 medium to remove thymocytes. The thymic pieces were digested in 5 ml of collagenase/dispose (1 mg/ml, Roche) plus DNase I (20 ng/mL, Sigma) in 2% FBS RPMI 1640 medium, placed in a 37°C water bath for 60 minutes, and agitated by passing through an 18G needle 8 times and a 25G needle twice. Cells were then filtered by passing through a 70 μm cell strainer.

For TEC phenotypical analysis, the digested cells (1-2 ×10⁶) were incubated with anti-CD45-PE-Cy7 (30-F11), MHCII-PerCp (M5/114.15.2), EpCAM-APC (G8.8), UEA-1-FITC (Vector, CA), and Ly51-PE (6C3) for 20 min. Cells were then placed in a 1% PFA PBS solution for analysis. For LIN28a antibody staining, we followed the Cell Signaling protocol for intracellular antibody staining. Briefly, digested thymic cells (1-2 ×10⁷/ml) were fixed in 4% PFA-PBS and placed in a 37°C water bath for 10 min. The cells were permeabilized by slowly adding 1 ml of ice-cold 90% methanol and incubating for 30 min on ice. The cells were then washed in 2% FBS-PBS buffer and incubated with anti-LIN28a (1:200) and UEA-1-Biotin at RT for 1 hr. The cells were then washed and incubated with the following antibodies: anti-CD45-PE-Cy7, MHCII-PerCp, EpCAM-APC, Ly51-PE, Strepavidin-APC-Cy7, and donkey anti-Rabbit-Alexa-488 (1:800) at RT for 30 min. For BrdU and LIN28a staining together, anti-CD45-PE-Cy7, MHCII-PerCp, EpCAM-PE, UEA-1-biotin following streptavidin-APC-Cy7, BrdU-FITC and LIN28a following donkey anti-Rabbit-Alexa-647 were used.

For TEC sorting, the digested cells were incubated with anti-CD45 APC, EpCAM PE, and MHCII FITC. The CD45^-Epcan⁺MHCII⁺ cells were sorted as TECs and separated into MHCII^hi and MHCII^lo subpopulations by MoFlo-DXP cytometry (Dako Cytomation). All antibodies were purchased from Biolegend if not noted (San Diego, CA). Phenotypical analysis was performed with a Cyan ADP Flow Cytometer (Beckman Coulter, Miami, FL). The data were analyzed by Flowjo™ Software (Tree Star, Ashland, OR).

Primary antibodies included specific polyclonal rabbit anti-LIN28a (A117, Cell Signaling Technology, Beverly, MA; 1:100); specific polyclonal rabbit anti-LIN28b (ProteinTech Group, Inc; Chicago, IL; 1:50) (25, 41); goat anti-mouse FOXN1 (WHN G-20, Santa Cruz Biotechnology; Santa Cruz, CA. 1:200); and rat anti-IA/IE (M5/114.15.2, Biolegend, San Diego, CA 1:200). Secondary antibodies included donkey anti-rabbit IgG-FITC (1:200), donkey anti-goat IgG Alexa-555 (1:800), donkey anti-rat IgG Alexa-555 (1:800); UEA-1-biotin (Vector Laboratories, Brulingame, CA 1:200); and streptavidin Alexa-647 (1:800) (Jackson Immunoresearch Laboratories, West Grove, PA). For FOXN1 staining, the embryos and adult thymi were fixed with 4% paraformaldehyde (PFA) in PBS at 4°C overnight. The samples were dehydrated in gradient ethanol solutions (70, 80, 90, 96, and 100%), embedded in paraffin, and then sectioned at 10 μm. The slides were rinsed in xylene before rehydration through a reversed ethanol series. Antigen retrieval was performed by boiling slides in pH 6, 10 mM sodium citrate buffer for 30 mins. For MHCII staining, embryos and adult thymi embedded in OCT were snap-frozen in liquid nitrogen and then stored at -80°C until sectioning. These samples were cut at 10 μm and fixed in ice-cold acetone for 2 min. All antibodies were diluted in 0.1% BSA+PBS, the primary antibodies were then added to each sample after blockage with 10% donkey serum overnight, and the secondary antibodies or streptavidin were stained for 40 min at RT. Images were obtained using a Zeiss Axioplan2 imaging microscope with an AxioCam HRM and AxioVision Rel 4.5 software (Jena, Germany). Quantitation of fluorescent cells was measured using ImageJ free software.

Each mouse was given a single intraperitoneal injection. with 1 mg of BrdU (Sigma−Aldrich) and hydrated with BrdU-containing water (0.8 mg/ml) for 5 days. One thymic lobe was digested for TEC, and one was ground for thymocyte analysis. For BrdU staining, freshly isolated thymocytes were incubated with anti-CD4-PE-Cy7, CD8-APC-Cy7, CD25-PerCp, and CD44-APC antibodies. The digested cells were incubated with anti-CD45 PE-Cy7, EpCAM APC, MHCII PE, and UEA-1-Biotin following Avidin-APC-Cy7. The surface-stained cells were then fixed and permeabilized in PBS containing 1% paraformaldehyde plus 0.01% Tween 20 for 48 h at 4°C and then incubated with 2 mg/ml DNase I for 15 min in a 37°C water bath. FITC-anti-BrdU Ab (clone 3D4; BD Pharmingen) was used for BrdU staining according to the manufacturer’s instructions. For Annexin V staining, freshly isolated thymocytes (5 X 10⁵) or total digested thymic cells (1 X 10⁶) were incubated with surface antibodies and then incubated with Annexin V FITC and PI in Annexin V binding buffer. The samples were analyzed within 1 hour following the protocol of the Annexin V kit (Biolegend, San Diego, CA).

The fetal thymi and sorted total TECs or MHCII^lo and MHCII^hi TEC subsets were extracted for mRNA by the RNeasy micro kit (QIAGEN). Reverse transcription PCR was performed with the superscript III system (Invitrogen). Gene expression levels of Lin28a (Mm00524077_m1), Lin28b (Mm01190673_m1), Let-7g (Mm04231484_s1), and H2-IAb (00439216_m1) in total TECs or MHCII^lo and MHCII^hi subsets of TECs were measured by Q-PCR, with a GAPDH FAM (Mm99999915_g1) primer/probe used as an endogenous control. All primers/probes were ordered from Applied Biosystems. Q-PCR was performed following the manufacturer’s instructions in a 10 μl volume using the AB 7500 Sequence Detector.

Mice were anesthetized using isoflurane anesthesia and then placed in dorsal recumbency. A 0.5-1 cm ventral midline incision was made in the scrotum, and the skin retracted to expose the tunica. The tunica was pierced, and the testes were pushed out sequentially. The testes were raised to expose the underlying blood vessels and tubules. Mouse testes were removed using forceps; cauterization was not needed. All deferential vessels and ducts were replaced back into the tunica. Skin incisions were closed with stainless steel wound clips. All of the excised paired tissues from each animal were retained to verify completeness of removal. Sham castration was performed as described above except for the removal of the testes.

All data were collected in a Microsoft Excel file and analyzed using Prism software (GraphPad Software, Boston, MA 02110) by one-way analysis of variance (ANOVA)-Bonferroni test or Student’s t test, P value in two-tailed. For statistical analysis, the significance was set at P ≤ 0.05.

To assess LIN28 in the fetal thymus, we first performed immunofluorescence (IF) on E13.5 and E18.5 thymi from C57BL6/J WT mice using LIN28a and LIN28bspecific antibodies. No LIN28a signal was detected in the E13.5 and E18.5 thymi ( Supplementary Figures 1A, B ). At E13.5, LIN28b was detected in the cytoplasm of both FOXN1^-negative cells (FOXN1^-) ( Figure 1A , white arrows) and in a minor subset of FOXN1-positive (FOXN1⁺) TECs ( Figure 1A , pink arrows). A small number of MHCII⁺ K14⁺ early medullary TECs (mTECs) were also detected in the center of the thymus but were found to be LIN28b negative ( Figure 1B , Supplementary Figures 1C, D ). At E18.5, LIN28b was found to be expressed in a subset of TECs in both the cortex and medulla, colocalizing with FOXN1 and MHCII ( Figures 1C, D ; pink arrows). Quantitative analysis showed the frequency of LIN28b⁺ cells increased in total FOXN1⁺ cells ( Figure 1E ). Some LIN28b^lo FOXN1^- cells were also detected in the outer area but were absent from the central area of the thymi, which might represent early developing thymocytes ( Figures 1C, D ; white arrows).

To further investigate expression in fetal TECs, we analyzed the gene expression of Lin28a and b and Let-7g in fetal TEC subsets sorted based on EpCAM and MHCII expression at E13.5 and E18.5. At E13.5, all TECs are MHCII negative or low ( Supplementary Figures 2A-D ). Lin28a was not detected or was present at very low levels in all fetal TEC subsets. Lin28b was correlated with the expression of MHCII, increasing significantly in MHCII^lo compared to MHCII^- TEC at E13.5, increasing further at E18.5 in MHCII^lo TEC, and increasing two-fold higher in MHCII^hi than in MHCII^lo at E18.5 ( Figure 1F ). Let-7g expression also correlated with MHCII expression and was higher in MHCII^lo TECs than MHCII^- TECs at E13.5 and higher in MHCII^hi than MHCII^lo TECs at E18.5. However, at E18.5, expression declined in MHCII^lo TECs relative to that at E13.5 ( Figure 1F ), suggesting more complex regulation. These results indicated that Lin28b and Let7g but not Lin28a may be involved in TEC differentiation during fetal thymic development.

To understand Lin28 and Let-7 expression in the postnatal thymus, we first assessed LIN28a and LIN28b at the protein level in postnatal TECs. Sections from C57BL6/J WT thymi obtained at 2 months of age were stained along with FOXN1 and MHCII. Both LIN28a and LIN28b in the medulla colocalized with both FOXN1 ( Figures 2A, B ) and MHCII ( Figures 2C, D ), indicating that they are each expressed in a subset of mTECs at this stage. Quantitative analysis showed the frequency of LIN28b⁺ cells in total FOXN1⁺ cells was higher than LIN28a⁺ cells ( Figure 2E ). We further investigated the gene expression of Lin28a, Lin28b, and Let-7g in sorted MHCII^lo and MHCII^hi TEC subsets from 2-month-old WT mice ( Figure 2F ). Both Lin28a and Lin28b were preferentially expressed in MHCII^hi TECs with little or no expression in MHCII^lo TECs. Lin28b expression was 2.8-fold higher than that of Lin28a. Conversely, Let-7g was almost exclusively expressed in MHCII^lo TECs ( Figure 2F ). This opposing expression profile is consistent with the known reciprocal regulation pattern for Lin28 and Let-7 in other tissues. These patterns for Lin28a and Lin28b are at least broadly consistent with Let-7g expression. However, the Let-7g pattern is the opposite of that observed at E18.5, where Let-7g was in both MHCII subsets but higher in MHCII^hi ( Figure 1F ), which suggests complex regulation of this miRNA gene.

We next traced the temporal dynamics of Lin28a, Lin28b, and Let-7g expression in postnatal MHCII^lo and MHCII^hi TEC subsets. The expression of both Lin28a and Lin28b gradually increased from 1 week, peaked at 1 month, and then gradually declined at 2 and 6 months in MHCII^hi TECs; a parallel pattern was found at a much lower level for Lin28b in MHCII^lo TECs ( Figures 2G, H ). Let-7g showed a similar temporal profile but in MHCII^lo TECs, with very low levels in MHCII^hi TECs ( Figure 2I ). Consistent with these gene expression kinetics, IF staining showed that LIN28b was present at a low level in a subset of FOXN1⁺ mTECs at day 20 ( Figure 2J ), increasing in cell number and intensity per cell at 2 months ( Figure 2K ), then dramatically declined or was undetectable at 6 months ( Figure 2L ). These data indicate that the expression of Lin28a, Lin28b, and Let-7g is correlated with both thymic growth and the early stages of age-associated involution.

Lin28a and Lin28b single and double null mutants have a variety of defects that could confound analysis of their function in TECs (20, 21); therefore, we generated TEC-specific deletion of Lin28a and Lin28b using Foxn1^+/Cre (38). Since similar genetic and phenotypic analysis results were shown at multiple time points, we show here representative results from 2-month-old mice. Q-PCR analysis showed that at 2 months, both Lin28a and Lin28b transcripts were undetectable in Lin28a^fl/flb^fl/fl;Foxn1^Cre/+ double mutant TECs and that only Lin28a was detected in the MHCII^hi TECs from Lin28a^+/+b^fl/fl;Foxn1^Cre/+ single mutants ( Figures 3A, B ). Specificity of LIN28a and LIN28b antibody staining on the thymic sections from Lin28a^fl/flb^fl/fl ;Foxn1^+/+ control and Lin28a^fl/flb^fl/fl ;Foxn1^Cre/+ double knockout adult mice confirmed the specific deletion of Lin28a and b in TECs ( Supplementary Figures 3A–D ). Deletion of Lin28b alone or of both Lin28a and Lin28b was associated with increased Let-7g expression in MHCII^hi TECs and even showed a 1.3-fold increase in MHCII^lo TECs ( Figure 3C ), despite overall low expression of both Lin28a and Lin28b in these cells. These data support a negative regulatory relationship between Lin28 and Let-7g expression in MHCII^hi TECs.

We next assessed the role of Lin28 and Let-7g in TEC differentiation and MHCII expression in 2-month-old double mutant mice. TEC-specific deletion of both Lin28a and Lin28b caused a significant reduction in both the percentage of TECs within total CD45^- stromal cells and in total TEC number ( Figures 3D–F ). Reductions were similar in both MHCII^lo and MHCII^hi TEC numbers, with a slight decline in the frequency of MHCII^hi TECs ( Figures 3G–K ). Analysis of cortical (cTEC) and mTEC populations showed that this reduction was entirely in mTECs in the Lin28a^fl/flb^fl/fl Foxn1^Cre/+ double mutants ( Figures 3L–Q ). These TEC defects caused a significant reduction in thymus size ( Figure 3R ) and in total thymocyte number ( Figure 3S ) starting at 6 weeks of age. The overall decline in thymocyte numbers was reflected in small but significant declines in DP and CD4SP numbers ( Figures 3T–V ). These results suggest that Lin28a and Lin28b impact thymocyte differentiation indirectly, primarily by regulating the differentiation and/or proliferation of MHCII^hi mTECs in the postnatal thymus.

To further test LIN28 activity in TECs, we overexpressed (OE) Lin28a by activating the iLin28a transgene (iLin28a-Tg) (40) specifically in TECs with Foxn1Cre (38). Activation of the transgene resulted in a 34-fold increase in Lin28a expression in MHCII^lo TECs and a 14-fold increase in MHCII^hi TECs in 6-week-old iLin28a^+/Tg ;Foxn1^Cre/+ (iLin28a OE) transgenic mice ( Figure 4A ). Flow cytometry analysis was consistent with these data, with a low level of LIN28a signal detected in MHCII⁺ TECs in iLin28a^+/+ controls and a strong LIN28a signal in the iLin28a OE mice in both MHCII^lo and MHCII^hi TECs, although LIN28a overexpression was detected only in a subset of TECs ( Figures 4B, C ). Consistent with the mRNA levels and flow cytometry, LIN28a was difficult to detect by IF in the iLin28a^+/+ control thymus at 3 weeks of age ( Figures 4D, F ), but LIN28a dramatically increased in iLin28a OE mice ( Figures 4E, G ). LIN28a protein in transgenics was found primarily in the cytoplasm but was also present in the nucleus of TECs ( Figure 4E , inset). Interestingly, Lin28b expression in MHCII^hi subsets was reduced in iLin28a OE mice compared to controls ( Figure 4H ), which suggests cross-regulation of these two related genes. Expression of the transgene was associated with a 60% reduction in Let-7g in MHCII^lo TECs but no significant change in Let-7g in MHCII^hi TECs ( Figure 4I ), possibly because of the reduction in Lin28b expression in these cells.

Overexpression of iLin28a in TECs caused an increase in thymus weight and the ratio of thymus weight to body weight in iLin28a OE mice ( Supplementary Figures 4A-D ), although it did not change the frequencies of cTECs and mTECs ( Figures 5A–D ). Increased thymus size was associated with significant increases in the number of both cTECs and mTECs ( Figures 5E–G ). To further analyze TEC phenotypes, we assessed MHCII expression. The frequency of MHCII^lo cTECs was increased, and the frequency of MHCII^hi mTECs was reduced in iLin28a OE mice ( Figures 5H–L ). Consistent with the overall increased size, the numbers of all MHCII subsets were increased, except for MHC^hi cTECs ( Supplementary Figures 4E, F ). Overall, the ratio of MHCII^lo to MHCII^hi cells increased in iLin28a OE mice ( Supplementary Figure 4G ).

Consistent with increased thymus size and TEC number, total thymocyte number was significantly increased in iLin28a OE mice ( Figure 5M ) with increases in the number of all subsets except for CD4SP cells, which declined in percentage ( Figures 5N, O ; Supplementary Figure 4H ). The decline in CD4SP cell frequency is consistent with the decline in the frequency of MHCII^hi mTECs in iLin28a OE mice ( Figure 5L ).

The above results showed that TEC-specific deletion of both Lin28a and Lin28b caused a selective reduction in MHCII^hi mTECs, while overexpression of Lin28a expanded all TEC subsets but preferentially expanded MHCII^lo TECs. Because Lin28a and Lin28b are typically expressed in stem/progenitor cells associated with cell proliferation, we hypothesized that Lin28a and Lin28b promote TEC proliferation in the postnatal thymus, especially in MHCII^lo TECs. To test this hypothesis, we measured TEC proliferation in iLin28a OE mice at 6 weeks of age. Compared to control mice, in iLin28a OE mice, the total number of BrdU⁺ TECs was significantly increased in total TECs ( Figures 6A, B ) but was selectively higher in MHCII^lo TECs ( Figures 6C–E ). LIN28a⁺ TECs incorporated a significantly higher level of BrdU than LIN28a^- TECs, indicating that this is a cell-autonomous effect ( Figures 6F, G ). In addition, the frequency of BrdU⁺ cells did not change in thymocyte subsets ( Supplementary Figure 5A ), and thymocytes and TECs from both mutant and control mice did not show a significant change in apoptosis ( Supplementary Figures 5B-D ). These data suggest that overexpressing Lin28a primarily affects TEC subsets by directly or indirectly increasing the proliferation of MHCII^lo TECs.

TEC differentiation and proliferation are also controlled by the transcription factor FOXN1, and Foxn1 downregulation is considered an initiating event for age-associated thymic involution (37). The dynamic patterns of Lin28a, Lin28b, and Let-7g across age, the loss of MHCII^hi mTECs in Lin28a and Lin28b loss-of-function mutants and the selective effect on proliferation of MHCII^lo TECs with Lin28a OE all mirror aspects of Foxn1 expression and function (37). To test the relationship between FOXN1 and Lin28b and Let-7g in TECs, we generated an allelic series using two alleles of Foxn1: the hypomorphic allele Foxn1^Lacz/LacZ (Z/Z) (37) and the null allele Foxn1^nu (42). In Z/Z mutants, Foxn1 expression is downregulated in TECs beginning at P7, resulting in a relative reduction in MHCII^hi and an increase in MHCII^lo TEC subsets. Combining the Z allele with the nu allele results in a progressive decline in thymus size and TEC phenotypes (37). We first analyzed Lin28b and Let-7g gene expression in total TECs from +/+, +/Z, Z/Z and Z/N (Z/nude) mice at 3 weeks of age. We found that with the progressive reduction in Foxn1 expression, both Lin28b and Let-7g in total TECs were downregulated in a Foxn1 dose-specific manner ( Figures 7A, B ). Consistent with their wild-type expression patterns, Lin28b was reduced in MHCII^hi TECs, and Let-7g was reduced in MHCII^lo TECs in the Z/Z mutants compared to +/Z controls ( Figures 7C, D ). These data indicated that the differential expression of both genes is controlled directly or indirectly by FOXN1.

To further investigate the regulation of Lin28 and Let-7 relative to thymic involution, we performed sex steroid ablation (SSA)-induced thymic rebound in 2.5- and 6-month-old Wt mice and analyzed gene expression early at day 3.5 after castration. SSA significantly increased Lin28b but not Lin28a expression specifically in MHCII^hi TECs at both 2.5 and 6 months, while expression in MHCII^lo TECs remained undetected ( Figures 8A, B ). SSA also caused a reduction in Let-7g expression in both MHCII^lo and MHCII^hi TECs at 2.5 months but no change in either subset at 6 months ( Figure 8C ). Based on microRNA search results, we found that the mouse class II transactivator-encoding gene Ciita (a critical regulator of MHCII gene expression) and several H2 genes had the predicted target sites for Let-7g and some other Let-7s ( Table 1 ). To test the effect of Let-7g, we further measured the changes of these two genes after SSA, and consistent with down-regulation of Let-7g, both Ciita and H2-IAb mRNA were increased in MHCII^lo and MHCII^hi TECs at 2.5 months, but still increased in MHCII^hi TECs at 6 months even without Let-7g change ( Figures 8D, E ). These changes in gene expression were associated with an increase in the total number of TECs ( Figure 8F ), including MHCII^lo and MHCII^hi TECs ( Figure 8G ). Hiweverthe increase was much greater in MHCII^hi TECs ( Figure 8H ), causing an increase in the ratio of MHCII^hi : MHCII^lo ( Figure 8I ). These data suggest that SSA-induced thymic rebound is associated with an increase in Lin28b in MHCII^hi TECs, with higher frequency of MHCII^hi TECs, and increased MHCII gene expression in these cells. The increase in MHCII gene expression in these cells may be caused by reducing the inhibitory effect of Let-7g to Ciita and H2 genes.