- 1 Prolongevity hormone FGF21 protects against immune senescence by delaying age-related thymic involution
- 1.1 RESULTS
- 1.2 DISCUSSION
- 1.3 MATERIALS AND METHODS
- 1.4 SI MATERIALS AND METHODS
- 1.5 ACKNOWLEDGMENTS
- 1.6 FOOTNOTES
- 1.7 REFERENCES
The degenerative changes in thymus precede age-related loss of function in other organs (1–4). As human lifespan continues to increase, it has been hypothesized that the ability to retain a functional level of thymic lymphopoiesis beyond the time limit set by evolutionary pressures may be an important strategy to extend healthspan (3, 4). Therefore, the ability to enhance thymic lymphopoiesis is thought to be central to the rejuvenation of T-cell–mediated immune surveillance in the elderly (1–7). Aging is associated with marked perturbations in the stromal cell microenvironment of the thymus (8, 9). This includes a reduction in thymopoiesis-supporting thymic epithelial cells (TECs) (10), an increase in fibroblasts (11, 12), and emergence of adipocytes (4, 13) of unknown origin and function. Accordingly, recent efforts have focused on targeting TECs for the rejuvenation of the aging thymus (12, 14). Emerging evidence indicates that immune–metabolic interactions control several aspects of the thymic involution process and age-related inflammation (13). We have shown that byproducts of thymic fatty acids and lipids result in accumulation of “lipotoxic DAMPs” (damage-associated molecular patterns), which triggers innate immune-sensing mechanisms such as inflammasome activation that link aging to thymic demise (15). Immune–metabolic interactions within the aging thymus produce a local proinflammatory state that directly compromises the thymic stromal microenvironment, thymic lymphopoiesis, and serves as a precursor of systemic immune dysregulation in the elderly (5, 8). Despite progress in the field, the thymic growth factors that regulate thymic involution are incompletely understood.
The fibroblast growth factors (FGFs) constitute a family of 22 proteins that regulate diverse biological processes such as growth, development, differentiation, and wound repair (16). Prior studies showed that FGF7/keratinocyte growth factor (KGF) administration in aged mice partially reversed thymic involution (17–19). Notably, unlike most FGFs, FGF21 lacks affinity for heparan sulfate in the extracellular matrix and thus can be secreted to act in an endocrine fashion (20). FGF21 is predominantly secreted from liver but is also expressed in thymus (21). FGF21 is a prolongevity hormone that elicits it biological effects by binding to βKlotho in complex with FGF receptor (FGFR) 1c, 2c, or 3c, but not FGFR4 (16, 22, 23). FGF21 supports host survival during states of energy deficit by increasing ketogenesis and fuel utilization through mitochondrial fatty acid oxidation (16, 23, 24). Interestingly, energy deficit induced by the prolongevity intervention of caloric restriction (CR) reduces ectopic thymic lipid and maintains thymopoiesis in aged mice (13). This raises the question of whether signals that stimulate mobilization of ectopic lipid mediate the salutary effects of CR on immune function. Here we present evidence that FGF21 and βKlotho are coexpressed in TECs and maintain T-cell diversity in models of aging and hematopoietic stem cell transplantation (HSCT) by enhancing thymic function.
FGF21 and βKlotho Are Expressed in Thymic Stromal Cells.
Our initial microarray profiling studies revealed that thymic Fgf21 expression declines with age. To confirm these findings, real-time PCR analysis showed that aging is associated with a reduction in thymic FGF21 mRNA expression, whereas CR significantly protected against loss of Fgf21 expression in thymus (Fig. 1A). Consistent with prior studies (17, 21), Fgf21 and FGF receptors are expressed in thymus along with βKlotho (Klb) (Fig. 1B). Interestingly, although thymic FGF21 is reduced with age (Fig. 1C), Klb and Fgfr1 showed a reciprocal increase in expression (Fig. 1 D and E), whereas no age-dependent regulation of Fgfr2, Fgfr3, or Fgfr4was found (Fig. 1 F–H). In analyses of hematopoietic and stromal cells from young and old mice, we found that Fgf21, Klb, and Fgfr1 mRNA are predominantly expressed in thymic stromal cells (TSCs) and regulated with aging (Fig. 1 I–K).
To further characterize FGF21 expression in thymus, we sorted TECs (CD45−Epcam+) and fibroblasts (CD45−PDGFRα+) from young mice. FGF21 mRNA expression was highest in TECs (Fig. 1L), where it was present at greater than threefold higher levels than in liver, the primary source of circulating FGF21. Immunostaining of thymic cryosections revealed that βKlotho is expressed in a subpopulation of Keratin8+cortical TECs (Fig. 1M), some of which seem to be thymic nurse cells (25). In complementary studies, we also examined whether βKlotho is expressed in TECs expressing FoxN1, a transcription factor that is critical for thymopoiesis (26). To do this, we used transgenic mice harboring a fluorescent membrane dTomato/membrane EGFP (mT/mG) Cre reporter construct (27) that marks FoxN1 Cre excision by a heritable switch from membrane-targeted tdTomato expression to membrane-targeted EGFP expression. Examination of Foxn1-Cre:mT/mG mice thymi revealed that βKlotho is colocalized with Foxn1+ TECs (Fig. 1N). In addition, βKlotho was expressed in endothelial cells of double-walled postcapillary venules (PCVs) in the corticomedullary junction of the thymus (Fig. 1O). PCVs are critical for import of hematopoietic stem cells into thymus and export of mature CD4 and CD8 cells. These data suggest that FGF21 may regulate thymic function by acting on both TECs and PCVs.
FGF21 Overexpression Prevents Age-Related Thymic Involution.
Given that Fgf21 expression in thymus decreases with aging, we next investigated thymic status in a line of Fgf21-transgenic (tg) mice that compared with WT animals show 50–100 times higher circulating FGF21 concentrations (28). Therefore, the Fgf21tg and WT littermates were aged up to 18 mo to examine the impact of FGF21 overexpression on age-related thymic involution. Consistent with an overall reduction in body weight and size (29), the thymi and spleens of middle-aged Fgf21tg mice were significantly smaller than those of WT littermates (Fig. 2 A and B). When normalized for total body weight, the thymic size as well as cellularity of Fgf21tg mice were significantly higher than those of the control littermates (Fig. 2 A and B). The male and female Fgf21Tg mice do not display a difference in body weight (see Fig. S2C). Overexpression of FGF21 did not alter the T-cell development stages (Fig. S1 A and B), but when normalized to body weight, FGF21 gain of function significantly (P < 0.05) increased the total CD4 single-positive (CDSP), CD8 single-positive (CD8SP), CD4+CD8+ double-positive (DP), and CD4−CD8− double-negative (DN) thymocyte subpopulations (Fig. 2C). In addition, compared with WT controls, the middle-aged Fgf21tgmice displayed a significant reduction in Lin−Sca1+Kit+ (LSK) in bone marrow (Fig. S1 C and D). However, the reduction in LSKs in Fgf21tg mice was not associated with thymic involution and could represent increased exit of these progenitors from bone marrow. Hallmark features of thymic aging include loss of corticomedullary junctions and emergence of ectopic adipocytes (1–4). Examination of thymic architecture revealed that in comparison with age-matched WT littermates, 14-mo-old Fgf21tg mice displayed preservation of cortical and medullary cellularity (Fig. 2D and Fig. S2A). Interestingly, overexpression of FGF21 was associated with a reduction in ectopic adipocytes in the subcapsular zone of thymus (Fig. 2E). Furthermore, instead of the typical accumulation of white adipocytes in the perithymic region of middle-aged WT animals, the Fgf21tg mice had an increase in brown adipose tissue adjacent to thymus (Fig. 2 D and E and Fig. S2 A and B). These data agree with the prior finding that FGF21 causes browning of white adipose tissue (30).
The channelling of ectopic lipid into nonoxidative pathways can lead to the generation of ceramides, which causes NLRP3 inflammasome-dependent thymic macrophage activation and inflammation (3, 15). Interestingly, electron microscopy analysis revealed that macrophages in the aging WT thymi contained spiculate crystalline material reminiscent of Charcot–Leyden crystals (Fig. 2 F and G) (31, 32), phagocytosed lipid droplets, and large protein aggregates, suggesting defective autophagy. There were also enlarged lysosomes with electrodense material (Fig. 2F and Fig. S2 D and E), which are associated with NLRP3 inflammasome activation and thymic involution (15). Although βKlotho is not expressed in macrophages (Fig. 1J), consistent with reduced thymic damage, thymi from Fgf21tg mice had significantly reduced macrophages with large crystals (Fig. S2D). These data suggest that enhanced FGF21 signaling in TSCs reduces the overall burden of DAMP clearance by macrophages, which may indirectly participate in lowering age-related thymic inflammation.
Stromal cells in thymus, including cortical (c) and medullary (m)TECs, are essential for T-cell development (10–12, 25). Aging is associated with reduced proliferation and survival of TECs (8, 9, 11). We found an increase in the number of cTECs (Fig. 2H) without any change in mTECs in Fgf21tg mice (Fig. S2F). Although mTECs predominate in thymus, the relative increase in the cTEC:mTEC ratio in middle-aged mice is consistent with prior studies (11). Aging is also associated with changes in the composition of TSCs, with a typical increase in thymic fibroblasts (8, 9, 11). The overexpression of FGF21 prevented the age-related increase in thymic fibroblasts (Fig. 2I) and maintained the cTEC architecture (Fig. 2J). To determine the mechanism of FGF21’s effects on thymic function, we evaluated whether FGF21 can act on thymic stroma. Consistent with our finding that βKlotho and FGFRs are expressed in TSCs, FGF21 treatment induced the phosphorylation of ERK (Fig. 2K), suggesting that FGF21 acts directly on TSCs. Furthermore, the preservation of cTEC function was reflected by increased expression of TEC-specific genes, early V antigen (Eva), and growth factors such as Il7 and Fgf7 (Fig. 2L). No significant changes in the expression of Aire, Beta5t, Dll4, or Rank were observed between the thymi of 14-mo-old WT and Fgf21tg mice (Fig. S2G). Together, these data suggest that FGF21 maintains the thymic microenvironment during aging by lowering thymic lipotoxicity and promoting TEC function.
T-cell development is dependent on lympho-stromal interactions that control progression of the earliest thymocyte progenitors (ETPs) into mature T cells (5–7). Age-related thymic involution is also linked to reduction in frequency of ETPs (6). Interestingly, overexpression of FGF21 significantly increased the frequency of ETPs (Fig. 3A and Fig. S3A). Given that decline in T-cell diversity is one of the major mechanisms that contributes to immune senescence and reduced immune surveillance in aging (33, 34), we next investigated the impact of FGF21 on peripheral T-cell diversity in middle-aged mice. Interestingly, compared with 14-mo-old WT mice, age-matched Fgf21tg mice had a significant increase in frequency of CD4 and CD8 naïve (CD62L+CD44lo) cells and a reduction in age-induced expansion of effector memory (E/M) cells (CD62L−CD44hi) (Fig. 3 B and C and Fig. S3 F and G). Furthermore, examination of an additional cohort of 18-mo-old Fgf21tg mice confirmed that FGF21 overexpression protects against age-related loss of naïve and E/M T-cell expansion (Fig. 3 D and E and Fig. S3 H and I). Given that spleen size and total splenocyte counts are lower (Fig. S3J) and proportional to lower body weight in Fgf21tg mice, the total naïve and E/M T-cell counts were normalized to body weight to represent the impact of FGF21 on T-cell diversity (Fig. S3 G and I). Together, data from two cohorts (14 and 18 mo) aged independently in two separate mouse facilities (University of Texas Southwestern Medical Center and Yale School of Medicine) demonstrate robust protective effects of FGF21 on T-cell senescence that are not influenced by husbandry conditions that may influence microbiota.
It is known that preexisting naïve T cells in the periphery can also undergo proliferation to compensate for age-related reduction in thymic T-cell export (35, 36). Importantly, Klb is not expressed in T cells, suggesting that FGF21 does not act directly on peripheral T cells. To further evaluate thymic function, we also quantified signal-joint T-cell receptor (TCR) excision circles (sjTRECs) as a surrogate marker for recent thymic emigrants (37, 38). Consistent with an increase in naïve T-cell frequency in middle-aged Fgf21tg mice, the sjTREC content in splenic T cells was also significantly higher than in control WT littermates (Fig. 3F), suggesting increased thymopoiesis. TCR diversity is conferred by VJ and VDJ recombinations in complementarity determining region 3 (CDR3) of newly generated T cells in thymus (33, 39). Hence, each Vβ–Jβ combination is represented as a Gaussian distribution of 6–10 CDR3 lengths with consecutive addition of 3 bp representing in-frame rearrangement (40). The CDR3 polymorphism analysis through TCR spectratyping revealed that 14-mo-old Fgf21tg mice do not display significant perturbations of TCR repertoire (Fig. S4A). Given that these mice are middle-aged, no perturbations in other Vβ subtypes were observed (Fig. S4B). Taken together, these data show that FGF21 prevents age-related deterioration of peripheral T-cell diversity indirectly by increasing thymic T-cell production.
Loss of FGF21 Compromises Thymic Reconstitution.
We next investigated whether loss of FGF21 function affects thymic aging. The global FGF21-deficient mice do not display any changes in total thymic cellularity or T-cell development and ETPs at 2–3 mo of age (Fig. 4A and Figs. S3 B–E and S5 A and B) in young Fgf21−/− mice, suggesting that FGF21 is not required for thymic development. Interestingly, by 1 y age of age, ablation of FGF21 was manifested in greater loss of total thymic cellularity (Fig. 4A). Compared with 12-mo-old littermate controls, Fgf21−/− mice displayed a trend toward reduction in cTECs and mTECs that did not reach statistical significance (Fig. 4B and Fig. S5E). The middle-aged Fgf21−/−animals displayed significantly higher loss of naïve T cells and greater frequencies and numbers of E/M cells compared with age-matched littermate control animals (Fig. 4C and Fig. S5D). No changes in CD4 naïve and E/M subsets were observed in young Fgf21−/− mice (Fig. S5C). These data suggest that FGF21 deficiency with age accelerates thymic involution and loss of naïve T cells.
Age-related thymic degeneration is a significant impediment in cancer patients undergoing HSCT because the conditioning regimens damage already reduced stromal cell niches in recipient thymi (41–44). In elderly patients, the impaired T-cell reconstitution due to thymic damage after HSCT results in prolonged posttransplant T-cell deficiency and significant mortality and morbidity (43). We found that compared with WT mice there was increased mortality in Fgf21−/− mice that underwent lethal irradiation and HSCT (Fig. 4D). This was not due to reduced chimerism in bone marrow (Fig. S5F). Consistent with an important role for FGF21 in thymic function, we found that compared with WT mice the ablation of FGF21 significantly reduced thymic reconstitution (Fig. 4E and Fig. S5G). Lack of FGF21 in the host stromal compartment was associated with a significant reduction in donor DPs without changes in the CD4SP and CD8SPs (Fig. 4E and Fig. S5 H and I). These data suggest that loss of FGF21-mediated immune–metabolic interactions impairs thymic reconstitution following HSCT.
Thymic involution likely occurs as a consequence of both intrinsic defects in thymocyte progenitors and a failure to maintain a functional TEC compartment (5–7, 41). Aging reduces the number of TECs with a concomitant increase in lipid-laden cells, fibroblasts, and adipocytes (8, 9, 11). Among the FGF family, FGF7/KGF promotes thymic lymphopoiesis by acting on TECs (18, 19, 45). Unlike FGF7, FGF21 lacks the conventional FGF heparin-binding domain and hence can diffuse away from its cellular source of production to act in a paracrine or endocrine manner (20). FGF21 requires βKlotho for its action and is known to increase energy expenditure and exert antidiabetic and prolongevity effects (16, 22). Consistent with prior studies (29) and similar to CR mice (13), overexpression of FGF21 increases brown adipose tissue in the perithymic region and reduces ectopic lipid within the thymic space, suggesting a reduction in thymic lipotoxicity. Our data demonstrate a previously unidentified function of FGF21 as a prothymic molecule that is highly expressed in TECs and may diffuse in thymus to signal via discrete subpopulations of βKlotho-expressing Ker8+ cTECs, FoxN1+ TECs, and PCVs.
In addition, given that FGF21 increases lipid utilization, including enhanced adipose tissue browning, it is likely that FGF21 reduces overall lipid-derived DAMPs in aging thymus. Crystals are seldom formed spontaneously in mammalian tissues. Crystal-storing histiocytosis is a rare disease associated with the accumulation of crystalline material in macrophages and excessive inflammation (31). Surprisingly, in thymi of aged mice, spiculate crystal-containing macrophages were located in thymic medulla. The crystalline material is reminiscent of Ym1-like Charcot–Leyden crystals, which are linked to higher IL-1β secretion and exuberant innate immune response (31, 32). Importantly, the FGF21 coreceptor βKlotho is not expressed in macrophages (Fig. 1J), suggesting that the reduction in ectopic lipid and crystals in macrophages from Fgf21tg mice is secondary to an overall reduction in thymic involution rather than a direct effect on macrophage NLRP3. Understanding precisely how aging and FGF21 overexpression regulates the generation of crystalline material in thymic macrophages will require additional studies.
Prior studies showed that overexpression of FGF21 in mice increases serum adiponectin levels, improves insulin sensitivity, and extends lifespan by ∼40% (16). Given that aging of thymus precedes the development of systemic metabolic abnormalities, our data suggest that FGF21 acts directly on thymus. We have previously published that Fgf21tg mice eat the same or more than WT mice (16, 24). Thus, the effects on thymic biology observed in Fgf21tg mice are not due to caloric restriction. Despite FGF21’s robust effects on longevity and metabolism, Fgf21−/− mice do not display overt changes in metabolism or lifespan, suggesting alternate compensatory mechanisms.
With regard to thymic biology, the FGF21 gain and loss of function studies revealed that FGF21 plays a role on maintaining thymic microenvironment during aging, when the thymus undergoes lipoatrophy. In addition, the increased lethality of FGF21-deficient mice during the conditioning regimen of radiation suggests that FGF21 is required for immune–metabolic interactions that maintain homeostasis and protect against tissue damage. Loss of FGF21 was accordingly associated with reduced T-cell reconstitution in a clinical model of HSCT. However, global deletion of Fgf21 in a knockout mouse model may not mimic the much more discrete and gradual loss of thymic Fgf21 expression over time in aging. Thus, future studies using TEC specific and inducible down-regulation of FGF21 signaling may provide definitive insights on role of this pathway in thymic aging and reconstitution.
FGF21 is currently being pursued for the treatment of obesity and type 2 diabetes (46); our findings suggest that FGF21 also exerts positive immunoregulatory effects. Taken together, our data demonstrate that FGF21 links metabolic and immune systems and regulates peripheral T-cell homeostasis by preventing age-related thymic degeneration.
MATERIALS AND METHODS
Mice and Animal Care.
The C57BL/6-Tg(Apoe-Fgf21)1Sakl/J mice C57BL/6J Fgf21−/− and control littermates on C57BL6 background were obtained from University of Texas Southwestern Medical Center. Mice were housed in a pathogen-free facility with a 12-h light/12-h dark cycle with free access to food and water. All mice were fed a standard chow diet consisting of 4.5% fat (5002; LabDiet). The WT and 40% caloric-restricted mice were obtained from the National Institute on Aging Rodent Colony.
The lethal irradiation to ablate hematopoietic cells was performed using an X-Rad300, X-ray small animal irradiator. One week before irradiation, the recipient mice were be given acidified, antibiotic water. The lineage-depleted bone marrow cells from CD45.1+ (B6.SJLPtprca Pep3b/BoyJ) were transplanted to irradiated (750 cGy) syngeneic WT, Fgf21tg, and Fgf21−/− mice via tail vein injection. The mice were killed 2 wk after the HSCT for analysis of T-cell reconstitution. All experiments were in compliance with ref. 47 and were approved by the Institutional Animal Care and Use Committee at Pennington Biomedical Research Center and Yale University.
To identify ETPs, thymocytes were labeled for lineage-positive cell by using PE-conjugated anti-CD11b, Gr-1, CD45R, CD3, CD8, αβTCR, γδTCR, pan-NK, NK1.1, CD11c, CD19, Ter119, and CD127 antibodies but no CD4 (eBioscience), followed by staining with APC-conjugated anti-CD25 and FITC-conjugated anti–c-kit (eBioscience). The PE-labeled lineage-negative cells lacking CD25 and expressing c-kit were designated as ETPs, as previously described. For lymphocytes analysis after bone marrow transplantation, thymocytes are stained for CD4, CD8, CD45.1, and CD45.2 cells followed by staining with FITC-, PE-, PerCP-, and APC-conjugated antibodies (eBioscience). To identify naïve and effecter or memory T cells, splenocytes were incubated with PerCP-conjugated anti-CD4, APC-conjugated anti-CD8, PE-conjugated anti-CD62L, and FITC-conjugated anti-CD44 antibodies. Anti-MTS15 antibody for fibroblast analyses was a generous gift from Richard Boyd, Monash University, Melbourne. All of the FACS data were analyzed by postcollection compensation using FlowJo (Tree Star, Inc.) software.
Western Blot Analysis.
We conducted the immunoblot analysis for phosphor-ERK1, 2 in CD45- and TSCs as described previously (48). The protein immune complexes were detected using specific fluorescent secondary antibodies conjugated with IRDye 800CW (Rockland) and membranes were imaged using an Odyssey infrared imaging system (LI-COR).
Immunohistochemistry and Electron Microscopy.
The thymi were collected from mice and fixed in 4% (vol/vol) buffered paraformaldehyde and embedded in paraffin and optimal cutting temperature compound then cut into 5- to 7-μm-thick sections. Tissue sections were stained with H&E, UEA1/Troma1, KLB/keratin 8, and KLB/MECA32. The images were acquired using Axiovert 40 microscope and Leica SP5 confocal microscope. The animals were perfused with paraformaldehyde fixative and ultrathin thymus sections were cut on a Leica ultramicrotome into 70-nm-thick sections, collected on Formvar-coated single-slot grids, analyzed with Tecnai 12 Biotwin EM (FEI), and evaluated and photographed in a JEM 1010 electron microscope (JEOL) equipped with a Multiscan 792 digital camera (Gatan).
The total RNA from thymus tissue in different age time point was extracted using RNeasy Lipid Tissue Mini Kit (Qiagen). Total RNA was digested by DNase (Invitrogen). The cDNA synthesis and real-time RT-PCR was performed as described previously (Bio-Rad). Quantitative real-time RT-PCR analyses were done in duplicate on the ABI PRISM 7900 Sequence Detector TaqMan system with the SYBR Green PCR kit as instructed by the manufacturer (Applied Biosystems). GAPDH was used for normalization human and mouse genes accordingly. Primers were designed using NCBI Primer software based on GenBank sequence data. Primer sequences are listed in Table S1. sjTREC real-time PCR and TCR spectratyping details are provided in SI Materials and Methods.
SI MATERIALS AND METHODS
Quantification of sjTRECs.
CD4+ T subsets were isolated from splenocytes using mouse CD4+ T cells positive section kit (Invitrogen). The sorted cells were lysed in 100 mg/L proteinase K (Sigma) for 1 h at 56 °C followed by 10 min at 95 °C. The amount of TRECs in 5 × 106 cells was determined by real-time quantitative PCR using the ABI PRISM 7900 Sequence Detector TaqMan system as described previously (15, 38) (Applied Biosystems). The PCR was performed with mδRec and ψJα specific primers and mδRec-ψJα fluorescent probe as described previously (38). The standard curves for murine TRECs were generated by using δRec ψJα TREC PCR product cloned into a pCR-XL-TOPO plasmid that was generously provided by Gregory Sempowski, Duke University School of Medicine, Durham, NC.
Vβ TCR Spectratyping Analysis.
The analysis of hypervariable CDR3 of β chain offers a practical approach for the global 26 qualitative assessment of diversity of TCR repertoire. For TCR spectratyping and CDR3 length analysis PCR, a FAM-labeled nested constant β-region primer is used in combination with 24 multiplexed forward murine Vβ-specific primers. PCR was performed for 35 cycles with denaturation at 94 °C for 30 s, annealing for 55 °C for 30 s, and 1 min extension at 72 °C and the PCR products were analyzed on an ABI3130 genetic analyzer as described previously (13, 15). Each Vβ–Jβ rearrangement is visualized by six to eight peaks and each peak represents one or a set of T-cell clones bearing the same CDR3 length. Each peak was analyzed and quantified with ABI PRISM GeneScan analysis software (Applied Biosystems), based on size and density. Data were used to calculate the area under the curve for each Vβ family. Each peak, representing a distinct CDR3 of a certain length, was quantified with statistical software (BioMed Immunotech).
A two-tailed Student’s t test was used to test for differences between genotypes or treatments (*P < 0.05 and P < 0.01). The results are expressed as the mean ± SEM. The differences between means and the effects of treatments were determined by one-way ANOVA using Tukey’s test (Sigma Stat), which protects the significance (P < 0.05) of all pair combinations.
We thank Kim Nguyen, Angie Bookout, and Yuan Zhang for assistance with animal experiments; Klara Szigeti-Buck for electron microscopy analyses of thymus; and Emily L. Goldberg for reading the manuscript. This work is supported in part by NIH Grants AG043608, AI105097, and DK090556 (to V.D.D.) and R01DK067158 (to S.A.K. and D.J.M.) and Robert A. Welch Foundation Grants I-1558 (to S.A.K.) and I-1275 (to D.J.M.). D.J.M. was supported by the Howard Hughes Medical Institute.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1514511113/-/DCSupplemental.