Foxn1 overexpression promotes thymic epithelial progenitor cell proliferation and mTEC maintenance, but does not prevent thymic involution

The thymus provides the essential microenvironment for lineage commitment and development of T cells. Thymic epithelial cells (TECs) include medullary (mTEC) and cortical (cTEC) cells that mediate the homing of lymphoid progenitors and the proliferation, survival and differentiation of developing T cells. TECs derive from the third pharyngeal pouch endoderm during fetal development, which proliferates to form the thymic rudiment at about embryonic day (E) 11.5 in mouse embryonic development (Alawam et al., 2020; Gordon and Manley, 2011). In both humans and mice, the thymic rudiment becomes functional only after transcriptional activation of the Foxn1 gene in the thymic epithelium (Brissette et al., 1996; Oh et al., 2020; Romano et al., 2013).

FOXN1 is a key transcription factor that controls most aspects of TEC proliferation and differentiation. Foxn1 is selectively expressed in thymic and skin epithelia, where it regulates the expression of downstream molecular targets to control both growth and differentiation (Brissette et al., 1996). In mice, rats (Nehls et al., 1994) and humans (Romano et al., 2012), null mutations in the Foxn1 gene lead to a hairless phenotype (hence the name ‘nude’ mouse) and alymphoid cystic thymic dysgenesis due to defective TEC differentiation (Brissette et al., 1996; Kreins et al., 2021; Nehls et al., 1996, 1994; Rota et al., 2021). In the skin, FOXN1 promotes keratinocyte proliferation and suppresses differentiation, and must be downregulated for differentiation to proceed (Bukowska et al., 2018; Li et al., 2007). In contrast, FOXN1 is required in fetal and postnatal TECs both for proliferation and for progression of differentiation at multiple stages in cTEC and mTEC sub-lineage development (Nowell et al., 2011; Oh et al., 2020; Su et al., 2003; Vaidya et al., 2016). TECs are exquisitely sensitive to FOXN1 levels, and even small changes in FOXN1 dose can affect phenotypes (Chen et al., 2009; Nowell et al., 2011). Furthermore, maintenance of FOXN1 levels is required for postnatal thymus homeostasis. In Foxn1^lacZ mutant mice, premature postnatal downregulation of Foxn1 causes disorganization and atrophy of the thymus similar to premature aging-related thymic involution (Chen et al., 2009), and two different models of transgenic overexpression of Foxn1 feature prolonged maintenance of thymic size, structure and function, delaying involution (Bredenkamp et al., 2014a; Zook et al., 2011).

The role of FOXN1 is in TEC progenitor/stem cells (TEPCs/TESCs) is not yet clear. This uncertainty is in part due to lack of clarity on the unique phenotypes and functional capacity of both fetal and postnatal TEPCs. In the fetal thymus, the EpCam⁺Plet1⁺ TEC population includes a common thymic epithelial precursor (TEPC), from which both cTECs and mTECs will be subsequently generated (Alawam et al., 2020; Bennett et al., 2002; Gray et al., 2006). As these cells were originally identified based on their phenotype in Foxn1 null nude mice (Blackburn et al., 1996), it is possible that TESCs are Foxn1 negative, at least initially. Furthermore, TEPCs/TESCs can persist for long periods in the absence of or under very low Foxn1 levels, and can be activated by increasing Foxn1 expression (Bleul et al., 2006; Jin et al., 2014). However, there is also evidence that all TECs express Foxn1 at some point early in their differentiation (Corbeaux et al., 2010; O'Neill et al., 2016). Furthermore, it is clear that during differentiation both cTEC and mTEC lineages modulate Foxn1 levels over a wide range (Chen et al., 2009; Hirakawa et al., 2018; Nowell et al., 2011). FOXN1 may have varying effects at different levels in these distinct populations; for example, there is evidence that lower FOXN1 levels preferentially promote proliferation in less mature major histocompatibility complex (MHC) Class II low (MHCII^lo) TECs, whereas higher levels promote differentiation into MHCII^hi populations (Bredenkamp et al., 2014a; Chen et al., 2009; Nowell et al., 2011). Taken together, these and other data support a model in which TESCs/TEPCs express, at most, very low Foxn1 levels, and that its differential upregulation controls TEC proliferation and differentiation at multiple stages of both mTEC and cTEC differentiation. Although understanding these dynamics is crucial, neither how Foxn1 levels are regulated in these different subpopulations nor the specific role(s) of FOXN1 across TEC subsets is yet well understood.

In this study, we investigated the quantitative and TEC subset-specific roles of FOXN1 using K5.Foxn1 transgenic (Weiner et al., 2007), Foxn1^lacZ (Chen et al., 2009) and nude (Foxn1 null; Nehls et al., 1994) mouse strains. We show that the K5.Foxn1 transgene drives Foxn1 expression broadly in TECs, resulting in FOXN1 levels 4- to 6-fold higher than normal. However, unlike previously published overexpression models, this transgene did not result in increased thymus size or delay thymus size reduction with involution in either wild-type or Foxn1^lacZ/lacZ mice. However, K5.Foxn1 did improve TEC differentiation and thymus function. Importantly, this transgene drove significantly enhanced Foxn1 expression in Plet1⁺ cells, which are putative TEC progenitors. This higher expression resulted in both an expanded Plet1⁺ population and their increased proliferation, as well as ectopic expression of the mTEC progenitor markers claudin 3 and 4 (Hamazaki et al., 2007) and the mTEC differentiation marker UEA-1 (Ulex europaeus agglutinin 1) within Plet1⁺ cells. In the absence of endogenous Foxn1, the K5.Foxn1 transgene was sufficient to drive formation of a small thymus, with a bias toward mTEC development and an expanded claudin 3- and 4-positive (Cld3,4⁺) population. These results show that moderate upregulation of Foxn1 in TECs broadly biases them toward differentiation without increasing proliferation. Furthermore, Foxn1 upregulation within a TESC/TEPC-containing population causes both increased proliferation and misexpression of mTEC markers. Thus, the effects of increasing Foxn1 levels are context dependent, a finding that has important implications for efforts to delay thymic involution or rejuvenate the involuted thymus through manipulation of Foxn1 levels.

The K5.Foxn1 transgenic mouse line was developed in the Brissette lab to investigate the role of Foxn1 in skin and hair development by driving its overexpression using a keratin 5 (K5; KRT5) promoter (Weiner et al., 2007). As K5 is also expressed in multiple populations of TECs, we evaluated thymus phenotypes across aging. We measured the relative increase in Foxn1 expression at the mRNA level in mTECs [defined as CD45 (PTPRC)⁻EpCam⁺MHC⁺UEA-1⁺ cells] and cTECs (defined as CD45⁻EpCam⁺MHC⁺UEA-1⁻ cells) from K5.Foxn1Tg and wild-type mouse thymi. We used primers specific for the transgene mRNA product, as well as common primers that amplify both the transgenic and endogenous Foxn1 transcripts. Total expression of Foxn1 mRNA was increased in TECs from K5.Foxn1Tg mice compared with wild-type controls (Fig. 1A). Although expression of the endogenous K5 gene in the thymus is primarily restricted to mTECs, the K5.Foxn1 transgene drives expression in both cTECs (UEA-1⁻) and mTECs (UEA-1⁺). Foxn1 expression in UEA-1⁺ mTECs increased more than 6-fold, whereas expression in cTECs increased approximately 4-fold (Fig. 1B). Overall FOXN1 protein levels as detected by immunofluorescence were also clearly increased in both cortex and medulla in transgenic mice (Fig. 1C). Analysis of fluorescence intensity showed increases in both average fluorescence intensity (Fig. 1D) and the proportion of cells with high FOXN1 levels (Fig. 1E) compared with wild type.

This level of overexpression was considerably more modest than those observed in previously reported Foxn1 overexpression models (≥20-fold in Bredenkamp et al., 2014b; Zook et al., 2011), both of which displayed increased thymus size and delayed involution. To investigate the effect of this more moderate level of Foxn1 overexpression effects on the thymus, we first measured thymus size and weight at 1 month and 6 months of age. To our surprise, thymus size was not changed by the presence of the transgene (Fig. 1F,G). The overall architecture of the transgenic thymus, as examined by either Hematoxylin & Eosin staining at 1 and 6 months (Fig. 1G), or by keratin 5 (K5) and keratin 8 (K8; KRT8) staining at 1 month (Fig. 1H), was also not obviously affected. Thymocytes from 1- or 6-month-old transgenic mice were similar to wild-type controls in the relative frequencies of CD4⁺ or CD8⁺ single-positive (SP), CD4⁺8⁺ double-positive (DP), CD4⁻8⁻ double-negative (DN) thymocytes, or CD25 (IL2RA)- and CD44-labeled DN subsets (Fig. 1I; Fig. S1). These data indicated that the modest level of overexpression in this model was insufficient to affect steady-state phenotypes, or to prevent or delay involution in general. However, analysis of UEA-1 staining in 1-, 6- and 10 month-old mice showed an increase in the frequency and intensity of UEA-1⁺ mTECs, both by immunofluorescence (IF) (Fig. 1J-L) and by flow cytometry (Fig. 1M), with a selective increase in the frequency of UEA-1⁺MHCII^hi cells (Fig. 1N). These data suggest that although initial thymus development, steady-state phenotypes, and involution as measured by thymus size were unaffected, mTEC differentiation was maintained better with aging compared with wild-type mice.

We have previously reported an allele of Foxn1, Foxn1^lacZ, in which insertion of the lacZ gene into the 3′UTR results in postnatal premature downregulation of Foxn1 to levels about 35% of wild type (Chen et al., 2009) (referred to hereafter as Foxn1^Z). This early downregulation results in premature thymic involution that is phenotypically similar to, but much more rapid than, normal aging-related involution, including loss of mTEC markers, reduced MHC Class II (MHCII) expression, and disorganization of cortical-medullary structure (Chen et al., 2009). We tested whether increasing Foxn1 levels using the K5.Foxn1 transgene in Foxn1^Z/Z mice would prevent or modulate premature involution in this model. As in the K5.Foxn1 transgenics on a Foxn1 wild-type background, thymus size was not changed by the presence of the transgene at either 1 or 6 months of age in either Foxn1^+/Z (+/Z) or Foxn1^Z/Z (Z/Z) mice (Fig. 2A,B). Both thymocyte and TEC cellularity at 4 weeks of age were similar regardless of presence of the transgene (Figs 2A, 3A). Analysis of the degree to which Foxn1 expression was rescued in Foxn1^+/Z;K5.Foxn1 or Foxn1^Z/Z;K5.Foxn1 mice showed that overall Foxn1 expression was higher in all TEC subsets in Foxn1^Z/Z;K5.Foxn1 mice compared with Foxn1^+/Z or Foxn1^Z/Z genotypes, resulting in a modest overexpression of 4- to 5-fold (Fig. 2C). This degree of expression was similar to that of Foxn1^+/Z;K5.Foxn1 TECs. We note that at this stage Foxn1 expression in Foxn1^Z/Z mice does not differ from controls when sorted into TEC subsets by phenotype, despite our previously documenting that the lacZ allele is reduced in expression at this time point (Chen et al., 2009). We have consistently observed this phenomenon and conclude that this is a consequence of sorting by phenotypes that are themselves determined by Foxn1 expression levels.

To investigate further how TEC differentiation in the Z/Z mutants was affected by the K5.Foxn1 transgene, we examined the expression of region-specific markers at 1 and 6 months of age by immunohistochemistry (IHC). For cTECs, we used a cTEC-specific catalytic subunit of thymoproteasome-b5-thymus (β5t), which is a downstream target of FOXN1 (Uddin et al., 2017; Žuklys et al., 2016), and CD205 (LY75), an early marker of cTEC differentiation (Baik et al., 2013; Jiang et al., 1995). We used both keratin 14 (K14; KRT14) and UEA-1 to evaluate mTEC differentiation, which mark largely non-overlapping mTEC subsets by IHC (Klug et al., 1998).

CD205 was relatively unaffected by changes in Foxn1 levels in these models at either age (Fig. 2D,E). At 1 month of age, cortical staining for β5t protein was similar between +/Z and Z/Z thymus in the presence and absence of the K5.Foxn1 transgene, suggesting that the levels of FOXN1 protein in all of these genotypes was still sufficient for β5t expression (Fig. 2D). However, by 6 months of age β5t was present at very low levels in Z/Z thymus compared with +/Z and was upregulated in K5.Foxn1;Z/Z thymi to levels similar to controls (Fig. 2F).

We previously showed that mTECs in general and UEA-1⁺ mTECs in particular are decreased in Z/Z mutants (Chen et al., 2009). We confirmed this result by both IHC and flow cytometry; the fluorescence intensity and percentage of UEA-1⁺ mTECs were lower in Z/Z compared with +/Z mice at 1 month of age (Fig. 2G,H; Fig. S2B,C). Addition of the K5.Foxn1 transgene increased UEA-1 levels, in both +/Z and Z/Z thymi even at 1 month (Fig. 2G). The frequency of UEA-1⁺ cells in both +/Z and Z/Z thymi was also increased at 1 month with the addition of the transgene, and the frequency in K5.Foxn1;Z/Z was rescued to a level similar to +/Z mice (Fig. 2H).

Given the early phenotypes in mTECs, we further investigated mTEC differentiation at 1 month by measuring MHCII and AIRE levels. In both +/Z and Z/Z mice, presence of the transgene did not affect the frequency of UEA-1⁺MHCII^hi mTECs (Fig. 3A,B). There were, however, shifts in MHCII^lo cells specifically in K5.Foxn1;Z/Z mice, with more MHCII^lo mTECs and fewer MHCII^lo cTECs compared with Z/Z (Fig. 3B). As the total number of TECs was not changed in these mice (Fig. S3A), this shift in the frequency of MHCII^lo mTECs reflected an overall increase in the mTEC:cTEC ratio (Fig. 3C). We then evaluated AIRE, a crucial mTEC maturation marker, by IHC. AIRE levels decreased more than 80% in Z/Z mutants on a per cell basis compared with +/Z in 1-month-old mice (Fig. 3D,E), and the frequency of AIRE⁺ cells declined more than 50% (Fig. 3F,G). Presence of the transgene did not affect AIRE levels in +/Z mTECs, but in K5.Foxn1;Z/Z mice both AIRE levels (Fig. 3D,E) and frequency (Fig. 3F,G) were rescued, similar to +/Z mice. These data are consistent with other reports that Aire expression declines with either aging or with declines in Foxn1 expression (Coder et al., 2015; Xia et al., 2012).

These data suggest that maintaining and/or increasing Foxn1 expression using the K5.Foxn1 transgene prevented the declines in both cTEC and mTEC differentiation in Foxn1^Z/Z mice, with a more substantial positive impact on mTECs.

In Foxn1^Z/Z mutant mice, the TEC defect results in rapid, non-cell-autonomous defects in thymocyte development, including reduced total cellularity, selective reduction in CD4⁺ SP cells, and decreased DN1a,b/early T-cell precursor (ETP) cells (Chen et al., 2009). Because presence of the transgene did impact TEC differentiation, we tested whether the presence of the transgene in +/Z and Z/Z mice improved thymocyte differentiation at 1 and 6 months of age. Similar to wild-type mice (Fig. 1H), in +/Z heterozygotes presence of the transgene did not significantly change total thymocyte numbers or the numbers or relative frequencies of CD4⁻8⁻ DN, CD4⁺8⁺ DP, or CD4⁺ or CD8⁺ SP thymocytes (Fig. 4A,B; Fig. S3A-C,E,F), or of the DN subsets defined by CD44 and CD25 (DN1-4) (Fig. 4C,D; Fig. S3G,H). In K5.Foxn1;Z/Z mice, the numbers and frequencies of the DN and DP subsets were also unaffected, although those of both CD4 and CD8 SP thymocytes were increased slightly, but significantly, relative to Z/Z alone at both ages (Fig. 4A,B; Fig. S3C).

We have previously reported that within the DN population, Z/Z mice have increased frequency of CD44⁺CD25⁻ DN1 cells relative to wild type and heterozygotes (Fig. 4C,D) (Chen et al., 2009) due to an increase in a specific CD44⁺CD25^lo subset (Fig. 4C, red box). This subset likely represents a partial block in DN1-DN2 differentiation, as CD44⁺CD25⁺ (DN2) and CD44⁻CD25⁺ (DN3) cells were decreased (Fig. 4C,D) (Chen et al., 2009; Xiao et al., 2018). Notably, this aberrant DN1 population was absent in K5.Foxn1;Z/Z mice (Fig. 4C). Both the DN1 and DN3 frequencies improved in the K5.Foxn1;Z/Z mice, although they did not reach control levels, and DN2 frequencies did not change (Fig. 4C,D; Fig. S3G,H). As we did not include lineage markers, the DN1 subset in this analysis includes both c-Kit⁺ (ETP) T-lineage progenitors as well as B and NK lineage cells, all of which can be separated by their expression of HSA (CD24) and c-Kit (Porritt et al., 2004). Analysis of these subsets revealed that the T lineage-specific CD24⁺c-Kit⁺ DN1a,b/ETP cell population, which is decreased in Z/Z mutants, was partially rescued in K5.Foxn1;Z/Z mice at 1 month (Fig. S3I,J), and was similar to +/Z controls by 6 months (Fig. 4E,F). This rescue could have been due to changes in DLL4 levels, which were rescued to levels similar to that in +/Z controls in K5.Foxn1;Z/Z mice (Fig. 4G).

We also assessed the generation of FOXP3⁺ regulatory T (T_reg) cells at 6 months. T_reg frequencies were similar in +/Z, Z/Z and K5.Foxn1;Z/Z thymi (Fig. 5). Surprisingly, K5.Foxn1;+/Z thymi exhibited markedly increased frequencies and absolute numbers of FOXP3⁺ T_reg cells (Fig. 5; Fig. S3D). These T_reg cells expressed normal levels of CD4, TCRβ and Foxp3 (Fig. S4A-D). Increases in FOXP3⁺ cells were also seen in K5.Foxn1 transgenics on a wild-type background, as FOXP3⁺ cells declined between 1 and 6 months in wild type, but were maintained in K5.Foxn1 transgenics at 6 months (Fig. S4E,F). This result suggests that higher FOXN1 levels create a microenvironment that biases T cell development toward T_reg cell generation, and that the better maintenance of mTEC phenotypes in transgenics sustained T_reg cell generation with aging.

Taken together, these data indicate that presence of the transgene did not affect the major categories of normal thymocyte development in wild-type or +/Z mice, with the exception of increased T_reg cell production. In addition, the transgene improved thymocyte differentiation defects in Z/Z mice, especially increasing ETPs to levels similar to controls. This result is consistent with improved, but not normal, TEC phenotypes that better support thymocyte development.

Several reports have identified Plet1 as a marker for a population of cells with both cortical and medullary progenitor activity, both at fetal and postnatal stages (Bennett et al., 2002; Gray et al., 2006; Ulyanchenko et al., 2016). In addition, Plet1⁻Cld3,4^hi UEA-1⁺ TECs represent progenitors for postnatal mTECs, including AIRE⁺ mTECs (Hamazaki et al., 2007). Given the improved maintenance of mTEC populations in the presence of the K5.Foxn1 transgene, we evaluated the impact of the K5.Foxn1 transgene on these progenitor populations in both wild-type mice and in the context of the Foxn1^Z allele by IHC. The frequency of Plet1⁺ cells was significantly increased in K5.Foxn1 transgenic mice compared with wild-type controls (Fig. 6A,B; Fig. S5A, B). When all three markers were evaluated on +/Z and Z/Z mice with and without the K5.Foxn1 transgene, Plet1⁻Cld3^hiUEA-1^hi mTEC-committed progenitors were present in all genotypes (Fig. 6C, pink arrows). In the +/Z thymus, most Plet1⁺ cells had a range of claudin 3 expression levels and were UEA-1⁻ (white arrows), although a minority of these cells were also UEA-1^lo (yellow arrows). Similar results were seen in the Z/Z mutants. In contrast, in both the K5.Foxn1;+/Z and K5.Foxn1;Z/Z thymi nearly all Plet1⁺ cells were both claudin 3⁺ and UEA-1⁺ (yellow arrows).

To determine whether this co-expression in the presence of the transgene correlated with presence of FOXN1, we performed IHC for FOXN1 and Plet1. In +/Z and Z/Z mice, 90% of the Plet1^hi cells were negative for FOXN1 (Fig. 6D, white arrows; Fig. S5C). In contrast, in both K5.Foxn1;Z/+ and K4.Foxn1;Z/Z mice, ∼50% of Plet1⁺ cells were FOXN1⁺ (Fig. 6D, yellow arrows; Fig. S5C), although with a range of levels, indicating that the transgene was driving ectopic Foxn1 expression in this population that contains TEC progenitors. Given that Foxn1 has been shown to be involved in the regulation of TEC proliferation (Chen et al., 2009; Itoi et al., 2007; O'Neill et al., 2016), we used bromodeoxyuridine (BrdU) incorporation to test whether increased Foxn1 expression in these cells affected their proliferative status. We found that only 20% of Plet1⁺ cells in +/Z and Z/Z thymi incorporated BrdU during a 5-day period of exposure, whereas the majority of Plet1⁺ TECs in K5.Foxn1;+/Z and K5.Foxn1;Z/Z thymi were proliferating during this time window (Fig. 6E; Fig. S5D). Additional analysis showed that there was no overall change in TEC proliferation in either MHC^hi or MHC^lo TECs with the presence of Foxn1 transgene, including in Z/Z mice that have reduced MHCII^lo proliferation (Chen et al., 2009) (Fig. S5E,F). Thus, the increase in Plet1⁺ TEC proliferation was specific to that population.

The Foxn1 null thymus phenotype (nude) results from an early block in thymus development, with the resulting rudiment composed of both Plet1⁺ TEPCs and Cld3⁺UEA-1⁺ medullary progenitors (Depreter et al., 2008; Hamazaki et al., 2007; Nehls et al., 1996; Nowell et al., 2011). To test whether the K5.Foxn1 transgene can rescue this phenotype, we crossed it into the nude background. In K5.Foxn1;nu/nu mice, the transgene was the only source of Foxn1 because endogenous Foxn1 is eliminated by the nude mutation (Nehls et al., 1996). Addition of the transgene to +/nu heterozygotes did not significantly impact thymus weight (P=0.073), the number of FOXN1⁺ cells, or overall cortico-medullary organization (Fig. 7A,B,E-G). IHC for K8 (cTECs), K5 and UEA-1 (mTECs) showed minor disruption of the cortical and medullary organization and regions of K8^lo TECs in K5.Foxn1Tg;+/nu mice (Fig. 7C). Analysis of UEA-1 staining showed that addition of the transgene to +/nu increased the frequency of UEA-1⁺ mTECs but did not impact claudin 3 staining (Fig. 7D), consistent with the results in wild-type and +/Z mice (Figs 3, 5).

Addition of the transgene in the nu/nu thymus caused a dramatic improvement in thymus development. K5.Foxn1;nu/nu thymi were clearly larger than in nu/nu (Fig. 7A,B), and the number of FOXN1⁺ cells increased to nearly that of +/nu and K5.Foxn1Tg;+/nu mice (Fig. 7E,F), although the density of FOXN1⁺ cells was increased 2-fold (Fig. 7E,G). FOXN1 expression in TECs is required to recruit endothelial cells and to generate the cellular and molecular environment needed for normal thymic vascularization (Bryson et al., 2013; Mori et al., 2010). In the nude thymus anlagen, no CD31 (PECAM1)⁺ cells were detected in the epithelial region (Fig. 7H). All other genotypes showed the presence of vasculature, indicating that transgene-driven TEC differentiation is capable of recruiting vasculature into the thymic rudiment. CD31⁺ blood vessels also had a higher density in K5.Foxn1Tg;nu/nu thymus compared with controls (Fig. 7H,I), consistent with the increased density of FOXN1⁺ TECs. The restoration of TECs in K5.Foxn1Tg;nu/nu mice also supports normal development of CD4⁺, CD8⁺ T cells in the thymus (Fig. S6A-H). The percentage of CD4⁺, CD8⁺ T cells from K5.Foxn1Tg;nu/nu spleen was significantly lower than that of controls in 2-week-old mice, but reached similar levels by four weeks of age (Fig. S6I-L).

The K5.Foxn1;nu/nu thymus exhibited restoration of cortical and medullary compartments, with the medulla relatively expanded (Fig. 7B). Consistent with previous reports (Chen et al., 2009), nu/nu thymi showed broad expression of K8 with central small numbers of K5⁺ or UEA-1^loCld3⁺ cells (Fig. 7C,D). K5.Foxn1Tg;nu/nu thymus displayed K8 expression that was similar to that of K5.Foxn1Tg;+/nu thymus, with K8^lo regions in the cortex and scattered K8⁺ cells in the medullary regions (Fig. 7C). K5 expression was observed throughout the medulla, but also in scattered cells in the cortex that were primarily K8⁺ (Fig. 7C), consistent with expansion of a progenitor-containing K8⁺K5⁺ population (Klug et al., 2002). In contrast to K5 expression, UEA-1⁺ mTECs were localized to clearly delineated medullary areas of the K5.Foxn1Tg;nu/nu thymus.

Claudin 3⁺ cells are present in the nude mouse thymic rudiment, consistent with a previous report that some mTEC lineage divergence occurs in the absence of FOXN1 (Nowell et al., 2011). Our analysis confirmed that claudin 3⁺ cells are present in the nu/nu thymus and are mostly UEA-1 negative or low, similar to +/nu controls (Fig. 7D). The number and frequency of claudin 3⁺ cells was dramatically expanded in K5.Foxn1Tg;nu/nu mice, compared with all other genotypes, the majority of which were UEA-1⁺. This analysis suggested that the majority of mTECs in the K5.Foxn1Tg;nu/nu thymus have phenotypes consistent with an mTEC progenitor (Fig. 7D).

In summary, Foxn1 expression from the transgene is capable of driving substantial thymus differentiation and growth from the nude thymus anlage. The resulting TECs have expanded compartments of cells previously shown to contain TEC progenitors (K8⁺K5⁺ and claudin 3⁺ cells), consistent with the transgene driving both proliferation and differentiation of these progenitors.

Given previous data showing that Foxn1 overexpression in TECs attenuated and delayed thymic involution (Bredenkamp et al., 2014a; Zook et al., 2011), the fact that Foxn1 overexpression from this K5.Foxn1 transgene neither causes thymic overgrowth nor impacts the timing or degree of thymus size reduction during involution is surprising. However, this result is consistent with our data demonstrating that the thymus is highly sensitive to Foxn1 dosage, as reductions in Foxn1 mRNA levels of as little as 15% has measurable corresponding reductions in thymus size and TEC phenotypes (Chen et al., 2009). Foxn1 overexpression from the K5.Foxn1 transgene does have specific impacts on TEC and overall thymus phenotypes with improved mTEC maintenance and expanded TEC progenitor populations on a wild-type background. In addition, this transgene improves proliferation, differentiation and organization of both cTECs and mTECs in Foxn1^lacZ homozygous mice, including upregulation of MHCII expression in all TECs. These changes particularly impact mTEC numbers, differentiation and maintenance. Furthermore, when expressed in Foxn1 null nude mice the transgene drives sufficient Foxn1 expression to promote substantial thymus development biased toward mTECs.

So why would expression of the K5.Foxn1 transgene result in a different phenotype compared with the other published accounts of Foxn1 overexpression in TECs? There are two main differences between this transgene and the previously published accounts that could underlie this seemingly paradoxical result. First, this transgene drives Foxn1 expression at a level that is higher than normal endogenous levels, but lower than that reported for the other two transgenes. As Foxn1 is dosage sensitive in loss-of-function models, it is certainly possible that very high levels of expression are required to cause the thymic overgrowth. Second, the promoter used in this transgene drives Foxn1 overexpression in progenitors that are usually Foxn1 low or negative. The two prior reports used either a K14 promoter (Zook et al., 2011) or Foxn1^Cre (Bredenkamp et al., 2014a) to drive Foxn1 expression, neither of which would drive expression in the earliest TEC progenitors. K5, in contrast, is a known marker of TECs with progenitor activity, as well as most or all mTECs (Klug et al., 2002; Ulyanchenko et al., 2016), and our data clearly show that this transgene drives Foxn1 expression in Plet1⁺ cells, which have been implicated as progenitors and are normally Foxn1 negative (or below the level of detection), and results in their proliferation and relative expansion. Both features of this transgene could contribute to observed phenotypes; the expression in both bipotent Plet1⁺ progenitors and claudin 3⁺ mTEC progenitors (both of which are present in the absence of Foxn1) could bias them towards proliferation, whereas the modest overexpression in immature TECs (likely contained within the MHCII^lo compartment) is insufficient to cause the expansion phenotypes seen under conditions of significantly higher Foxn1 overexpression driven by the K14 or Foxn1 promoters in other studies. This combination of effects in different TEC populations could result in the failure to observe thymus overgrowth while driving the expansion of progenitors, both in the K5.Foxn1 transgenics alone, and in combination with the Foxn1^Z/Z mutation.

Other than its effects on progenitors, increased FOXN1 levels via the K5.Foxn1 transgene had the most obvious effects on the mTEC sublineage in the postnatal thymus. In addition to causing expansion of Cldn3⁺ mTEC progenitors, the transgene appears to bias TEC differentiation directly toward the mTEC lineage, with increased expression of mTEC markers in the Plet1⁺ compartment. In addition, both UEA-1 and AIRE mTEC differentiation markers are upregulated more broadly in mTECs. The upregulation of AIRE, both in cell numbers and on a per cell basis, is particularly important, as it could indicate an impact on self-tolerance. The efficiency of negative selection is dependent on the presentation of peripheral tissue-specific self-antigens (Klein et al., 2014), the expression of which is in part regulated by the Aire gene in mTECs (Koh et al., 2018; Peterson et al., 2008). Postnatal reduction of Foxn1 levels in the Z/Z mice induces loss of mTECs and reduces both the numbers of AIRE⁺ mTECs and its expression level, both of which are increased with Foxn1 overexpression. Although there is no evidence that Aire is itself a direct target of FOXN1 (Žuklys et al., 2016), these data support the idea that Aire expression is dependent on, and correlated with, FOXN1 levels.

An effect on negative selection is also indicated by the dramatic increase in the production of FOXP3⁺ T_reg cells, which are an important component of peripheral tolerance, and are generated in the thymus by diverting cells from apoptosis during the process of negative selection. Other than this phenotype, the main effects of the transgene on thymocyte differentiation were relatively mild, although consistent with known roles for FOXN1-dependent processes. No significant differences in the main thymocyte stages defined by CD4 and CD8 were seen between +/Z controls and K5.Foxn1;+/Z transgenics; this is consistent with the relatively minor effects on TEC differentiation and the lack of any change in thymus size. However, the increased Foxn1 expression from the transgene did fully or partially rescue the thymocyte differentiation phenotypes seen in the Z/Z mice. In particular, the frequency of DN1a,b/ETP progenitors was recovered consistent with its dependence on the expression of Dll4, a known Foxn1 direct target, on cTECs (Žuklys et al., 2016). These results are consistent with the idea that most aspects of thymocyte differentiation are sensitive to reduced Foxn1 dose, but may be relatively insensitive to increased Foxn1 levels.

Efforts to generate TECs via either directed differentiation or reprogramming continue to be of interest, as do targeting the in vivo thymus for rejuvenation via a variety of approaches towards the therapeutic goal of improving immune function after damage or with aging. The results from this study highlight the importance of carefully controlling both target cell type and Foxn1 dosage in these attempts. Whereas high levels of overexpression cause thymic overgrowth that does increase output but can be detrimental, the lower levels of expression and targeting to progenitors shown here did not result in overgrowth, despite the expansion of progenitor populations. This overexpression maintained and improved thymic phenotypes, especially in the medulla, without overgrowth. These results demonstrate that the effects of Foxn1 overexpression are pleiotropic and differ depending on the cell type and level of expression.

The Foxn1^lacZ allele was generated by the Manley lab (Chen et al., 2009), and may be obtained by contacting Dr Manley upon completion of a simple MTA agreement for academic users. Nude mice (Foxn1^nu) were purchased from The Jackson Laboratory Animal Resource Unit (Bar Harbor, ME, USA). Genotyping for both alleles was as described (Chen et al., 2009). The K5.Foxn1 transgenic line was provided by Dr Janice L. Brissette (to whom inquiries should be addressed; Cell Biology, Suny Downstate Medical Center, NY, USA), and was genotyped by PCR as previously reported (Weiner et al., 2007). All experiments using animals received prior review and approval by the UGA institutional IACUC in accordance with AALAC accreditation guidelines.

The following antibodies were used in this study: anti-CD4 (GK1.5, BioLegend, 248, lot100406, 1:150), anti-CD8 (53-6.7, BioLegend, 155, lot100708, 1:150), anti-CD25 (PC61, BioLegend, 4262, lot102030, 1:150), anti-CD44 (IM7, BioLegend, 312, lot103012, 1:150), anti-K5 (AF138, Covance, 1:500), anti-K8 (provided by Dr Ellen Richie, The University of Texas MD Anderson Cancer Center, TX, USA), anti-Foxn1 (G20, Santa Cruz Biotechnology, 1:200), anti-β5t (PD021, MLB International, 1:100), biotinylated UEA-1 (B-1065-2, Vector Laboratories, 1:400), anti-CD205 (dp200, Abcam, EPR5233, lot ab124897, 1:100), anti-K14 (AF64, Covance, 1:500), anti-CD45 (30-F11, BioLegend, 97, lot 103112, 1:150), anti-EpCAM (G8.8, BioLegend, 4726, lot 118206, 1:150), anti-I-A/I-E (M5/114.15.2, BioLegend, 366, lot 107606, 1:150), anti-Aire (M-300, Santa Cruz Biotechnology, sc-33189, 1:200), anti-CD24 (M1/69, BioLegend, 343, lot 101808, 1:150), anti-CD117 (2B8, BioLegend, 1945, lot 105812, 1:150), anti-CD19 (6D5, BioLegend, 1530, lot 115508, 1:150), anti-Foxp3 (FJK-16 s, eBioscience, 1:200), anti-Plet1 (provided by Dr Ellen Richie), anti-claudin 3 (PA5-32353, Invitrogen, 1:100), anti-BrdU (3D4, RUO, BD Pharmingen, 1:100), anti-CD31 (MEC 13.3, RUO, BD Pharmingen, 1:100). Fluorochrome-conjugated anti-Ig second step reagents were purchased from Jackson ImmunoResearch (705-585-003, 1:400; 715-545-150, 1:400; 711-605-152, 1:400; 715-585-151, 1:400). Binding of biotinylated antibodies was detected by fluorochrome-conjugated streptavidin (Invitrogen, S11223, 1:800).

Thymic stromal cells were isolated as described previously (Chen et al., 2009). Briefly, adult thymi were minced, digested in collagenase (Roche), then collagenase/dispase (Roche) and passed through a 100-μm mesh to remove debris.

Serial frozen sections (10 µm) from mouse thymus were air-dried for 30 min before acetone fixation. Thin sections were blocked with normal donkey serum and subsequently incubated with optimal dilutions of primary antibodies for at least 1 h at room temperature before washing and incubation with appropriate fluorochrome-conjugated secondary reagents. Controls included slides incubated with nonimmune species-matched Ig or isotype-matched mouse Ig. For multiple antibody staining, the sections were incubated simultaneously with primary antibodies from different species. Microscopic analysis was performed with a Zeiss Axioplan2 microscope or Zeiss LSM 710.

Cells in PBS containing 2% fetal bovine serum and 0.1% sodium azide were incubated with directly conjugated or biotinylated antibodies on ice for 30 min followed by two washes. Binding of biotinylated antibodies was detected with PerCp-SA. Cells were analyzed with a CyAn flow cytometer (Beckman Coulter Life Sciences) equipped with an argon laser (488 nm) for FITC and PE excitation and a helium-neon laser (633 nm) for APC and PerCp-SA excitation. Data were collected on using a four-decade log amplifier and were stored in list mode for subsequent analysis using FlowJo software.

Incorporation was initiated by intraperitoneal injection of BrdU (1 mg in PBS; Sigma-Aldrich) and maintained in drinking water for 5 days (0.8 mg/ml BrdU). Thymic stromal cells were isolated and surface labeled, then fixed in 1% paraformaldehyde, 0.01% Tween 20 (BDH Laboratory Supplies) in PBS overnight at 4°C. Cells were washed in PBS, recovered by centrifugation (8 min at 600 g), and incubated in DNase I (50 Kunitz; Roche) for 30 min at 37°C. After washing, cells were stained with FITC-conjugated anti-BrdU (BD Pharmingen) for 1 h at room temperature. For IF staining, to detect BrdU incorporation, thymic sections were incubated in 2 N HCl for 20 min at room temperature. After washing in PBS, the sections were incubated with mouse anti-BrdU (BD Pharmingen) for 1 h at room temperature followed by incubation with FITC anti-mouse IgG (Jackson ImmunoResearch).

Total RNA from sorted CD45⁻EpCam⁺UEA-1⁺ and CD45⁻EpCam⁺UEA-1⁻ thymic stromal cells from 4-week-old mice (Fig. S7) were isolated using the RNeasy kit (QIAGEN). cDNA was synthesized from the total RNA and used as a template for real-time relative quantification of Foxn1 and Dll4 using commercially available probes and reagents (Mm01298125_g1, Hs01117333_m1; Invitrogen) on an Applied Biosystems 7500 Real Time PCR System.

Fluorescence intensity and cell density for IHC experiments were analyzed using freely available ImageJ software (Fig. S8).

Data are presented as the mean and s.e.m. Comparisons between two groups were made using unpaired, one-tailed Student's t-test, or ANOVA for multiple group comparisons. P<0.05 was considered significant.

We acknowledge Ashley S.K. Simon for performing early experiments on K5.Foxn1 transgenic mice that led to the current manuscript. We thank Ellen Richie for helpful advice and comments on the manuscript. We thank J. Nelson in the Center for Tropical and Emerging Global Diseases Flow Cytometry Facility at the University of Georgia for flow cytometry and cell sorting technical support. This work was supported by the Biomedical Microscopy Core at UGA.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200995.reviewer-comments.pdf