Two Temporally Distinct Specification Events of TECs.

The expression of Foxn1 is considered to specify thymic identity of pharyngeal epithelial cells. Foxn1 expression can first be detected in the prospective thymus domain of the common thymus/parathyroid anlage at approximately embryonic day 10.5 (E10.5) (2, 3). It has been suggested that cervical thymi develop later than the orthotopic thymus (7), possibly as a result of a second specification event (8) (Fig. 1A). Indeed, rudiments of cervical thymi are first observed at E15.5 as small patches of Foxn1-expressing epithelial cells, most often located next to the developing parathyroid (Fig. 1B). At this time, the orthotopic thymus has already separated from the common parathyroid/thymic primordium and descended to a location above the heart. This observation suggests that the onset of Foxn1 expression in cervical thymi occurs ≈5 d later than in the thoracic thymus. However, it is possible that the rudiments of the cervical thymus are initially so small that they escape detection before E15.5. To examine this possibility, we attempted to ablate Foxn1-expressing cells during embryogenesis before the presumed second specification event. To this end, we created transgenic mice expressing the human diphtheria toxin receptor (DTR) gene (17) under the control of the Foxn1 promoter. The Foxn1 promoter faithfully directs expression of transgenes into TECs (8, 18) of both cervical and orthotopic thymi (7). Female mice transgenic for the Foxn1:eEGFP reporter (7) were mated with males transgenic for the Foxn1:DTR construct; in this mating scheme, the female is spared the cytotoxic effects of diphtheria toxin (DT), eliminating possible systemic effects on the developing offspring. DT was injected three times, at E10.5, E11.5, and E12.5 (i.e., beginning at the time of onset of Foxn1 expression in the thoracic thymus, but well before the first detectable signs of a developing cervical thymus). Foxn1:eGFP single-transgenic embryos and Foxn1:eGFP/Foxn1:DTR double-transgenic embryos were analyzed at various time points thereafter. As expected, the degree of TEC ablation was dose dependent (Fig. 1C); with higher doses, complete destruction of the thoracic thymic anlage (as determined by GFP fluorescence) was achieved. Siblings lacking the Foxn1:DTR transgene had a normal thymus, indicating that the observed phenotype was the result of a transgene-specific effect. The orthotopic thymus did not recover in treated double-transgenic embryos, suggesting that it was completely destroyed and that TECs of the embryonic thoracic thymus (including the presumptive progenitor cells) express Foxn1. However, when embryos treated between E10.5 and E12.5 were examined shortly before birth, at E18.5, many exhibited two (rarely only one) small thymic masses in the neck region (Fig. 1D), reminiscent of cervical thymi in both size and location (7). Given the effectiveness of Foxn1-directed cytoablation, this experiment therefore directly shows that the onset of Foxn1 expression (including the Foxn1 promoter-driven transgenes) in the anlagen of the cervical thymi indeed occurs later than in the orthotopic thymus.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Varying times of onset of Foxn1 expression in the epithelial cells of thoracic and cervical thymi. (A) Hypothetical scheme of the development of organ anlagen from the epithelium of the third pharyngeal pouch in mouse embryos. A common anlage (yellow) differentiates into separate parathyroid (green) and thymic (red) domains, but a small undifferentiated primordium remains; at a later stage, this gives rise to the anlage of the cervical thymus (Upper). Foxn1-directed cytoablation during early stages of embryogenesis destroys the primordium of the thoracic thymus but does not affect the subsequent development of the cervical thymus (Lower). (B) RNA in situ hybridization analysis of adjacent transverse sections of an E15.5 mouse embryo. The cervical thymus with Foxn1-positive epithelium is located directly adjacent to the parathyroid (see enlargements underneath). Hybridization with Ttf1, which is expressed in the primordium of the thyroid and the epithelium of the larynx, is shown for orientation (it is expressed at low levels in parathyroid and cervical thymus). Note that at this stage of development, the thoracic thymus has already descended to the mediastinum. This section is representative of a group of four embryos. (Scale bars: Upper, 100 μm; Lower, 50 μm.) (C) Macroscopic view of the mediastinal region of Foxn1:eGFP single- and Foxn1:eGFP, Foxn1:DTR double-transgenic embryos treated with the indicated doses of diphtheria toxin (DT). Pregnant female mice were injected with DT three times at E10.5, E11.5, and E12.5; photographs were taken under UV illumination at E14.5. The pictures shown are representative of groups of two, three, and five embryos dissected for the three treatment/genotype combinations, respectively. (Scale bar: 100 μm.) (D) Macroscopic view of thoracic and neck regions of Foxn1:eGFP single- and Foxn1:eGFP, Foxn1:DTR double-transgenic embryos treated with DT as in C. Photographs were taken under UV (Left and Right) or bright-light (Center) illumination. Anatomical landmarks are indicated, and the two cervical thymi are marked with arrows. The embryos shown are representative of groups of >100 (Left) and 20 (Center and Right). (Scale bar: 500 μm.)

Thymic Epithelial Progenitor Cells Express Foxn1.

Our previous findings (9) and the above experiment indicated that the bipotent thymic epithelial progenitor expresses Foxn1. We verified this conclusion using cytoablation based on a different cytotoxic principle. To this end, we created mice transgenic for a fusion gene between the fluorescent protein CFP and the bacterial nitroreductase (NTR) (19), obviating the need to combine two transgenes in the same animal. Nitroreductase is capable of converting prodrugs (such as CB1954) into effective cytotoxic agents, leading to cell death. In the transgenic mouse strains developed here, treatment with either DT or CB1954 cannot be extended beyond the E15.5 time point, owing to the occurrence of severe defects in the skin of newborn mice, because Foxn1 is also expressed in this tissue (20). To be able to examine whether TEC progenitors in the postnatal thymus still express Foxn1, intact orthotopic (thoracic) thymi from transgenic and nontransgenic E15.5 embryos were transplanted under the kidney capsule of nude mice (lacking the Foxn1:CFP-NTR transgene). One week after transplantation, corresponding to postnatal day 4, mice were treated with CB1954. The transplants were analyzed either 4 or 9–10 wk after the last treatment, providing enough time for any remaining progenitors to give rise to a thymopoietic epithelial microenvironment. Thymic tissue of transgenic transplants was completely resorbed in CB1954-treated mice as indicated by the lack of fluorescent cells (Fig. 2A); correspondingly, few, if any, CD4/CD8 double-positive thymocytes (DPs) were recovered from the entire kidney tissue (Fig. 2B). Hence, it appears that the epithelial progenitor cells did not survive transplantation and subsequent treatment with CB1954, compatible with their expressing Foxn1. This finding suggests that, at least in embryonic and early postnatal periods, epithelial progenitors are Foxn1 positive; otherwise, thymopoiesis should have recovered after cytoablative treatment of Foxn1-expressing cells.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Thymic epithelial progenitors express Foxn1. (A) Thymic lobes from E15.5 Foxn1:CFP-NTR transgenic mice were transplanted under the kidney capsule of nude mice. After 1 wk, animals were treated with vehicle or vehicle plus CB1954 at 500 μg four times per week for 3 wk. After a recovery period, kidneys were removed and photographed under bright-light and UV illumination. (Scale bar: 2.5 mm.) (B) Number of CD4/CD8 double-positive T cells isolated from thymus transplants (such as those shown in A) 4 (Left) or 9–10 (Right) weeks after the last treatment. The number of thymic lobes (n) analyzed, the presence (+) or absence (−) of the Foxn1:CFP-NTR transgene (tg), and treatment with vehicle (−) or CB1954 (+) for the respective groups are noted. (C) Macroscopic views of thoracic regions of sibling wild-type or Foxn1:CFP-NTR transgenic mice at 2 mo of age photographed under bright-light (Upper) and UV illumination (Lower). During the embryonic period, pregnant wild-type mothers received i.p. injections of CB1954 at 500 μg per mouse at E12.5, E13.5, and E14.5. Thymic tissue is outlined with solid lines; the heart is indicated with broken lines. Note that no thymic tissue could be detected for the animal shown in the center panel. (Scale bar: 2.5 mm.) (D) Macroscopic appearance of a small thoracic thymus (about 1/8 the size of that in wild-type sibling) observed at the age of 2 wk in a Foxn1:CFP-NTR animal treated as in C. In the photograph taken under UV illumination, note several clusters of strongly fluorescent epithelial cells, corresponding to individual medullary islets. (Scale bar: 500 μm.) (E) Number of CD4/CD8 double-positive thymocytes recovered from adolescent and adult mice exposed in utero to CB1954 as in C. The number of animals in each group is shown underneath the graph. Animals from group 1 were exposed to CB1954 in utero at E12.5, E13.5, and E14.5; animals from group 2 at E13.5, E.14.5, and E15.5.

The conclusion that embryonic thymic epithelial progenitor cells express Foxn1 was confirmed in an additional set of experiments. In this scheme, initial in utero treatment with CB1954 was done either between E12.5 and E14.5 or between E13.5 and E15.5, to achieve a selective near-total reduction of TECs, as suggested by prior dose-finding experiments. We then monitored thymopoietic activity of wild-type and Foxn1:CFP-NTR transgenic littermates at various time points after birth. We expected complete recovery of thymopoietic activity only in cases where TECs in the orthotopic thymus recovered to approximately normal numbers (note that cervical thymi are at least two orders of magnitude smaller than orthotopic thymi) (6, 7). As in the previous experiment, nontransgenic siblings were not affected by the treatment during the embryonic period; the size of the thymus and the number of thymocytes was indistinguishable from age-matched untreated control animals. By contrast, in all transgenic animals, the thymus was either smaller or entirely absent. In animals in which small thymic lobes were found (Fig. 2C), the characteristic pattern of densely packed medullae and less-populated cortical regions could be readily discerned by their distinctly different fluorescence (Fig. 2D). In some cases, the location of small masses of thymic tissue was again reminiscent of cervical thymi (as seen in the experiments with Foxn1:DTR transgene); in others, they clearly corresponded to small orthotopic thymi in the mediastinum. Correspondingly, when analyzed at various time points after birth, the number of DP thymocytes recovered per animal from all thymic tissues was highly variable, but always significantly reduced or nil in transgenic animals compared with wild-type controls (Fig. 2E). The low number of athymic mice in our cohort was expected, because the frequency of cervical thymi is relatively high in the FVB strain used here for the generation of transgenic mice. When residual thymic tissue was present, the relative distributions of thymocyte subsets were essentially normal, suggesting that T-cell development was quantitatively impaired but qualitatively normal. Importantly, at each time point, mice were identified that showed no evidence of ongoing thymopoiesis. Mice with low numbers of DPs also had fewer T cells in the periphery; these cells had the phenotype of activated peripheral T cells (CD44^high, CD62L^low, CD25^high), compatible with homeostatic proliferation in a lymphopenic host. Collectively, these experiments suggest that prenatal or early postnatal loss of TECs cannot be compensated for at later time points.

Foxn1 Expression in TECs During Ontogeny.

In contrast with embryonic and early postnatal thymi, the adult orthotopic thymus contains Foxn1-negative TECs, and it has been suggested that these cells correspond to a Foxn1-independent, but functionally competent, cell lineage (15). We addressed this possibility in two ways. First, we used an analytical approach to establish the developmental relationship between Foxn1-positive and Foxn1-negative TECs. To this end, we confirmed that Foxn1-negative TECs are readily detected in Foxn1:eGFP transgenic mice several weeks after birth (Fig. S1A). To determine whether these cells had previously expressed Foxn1, we marked them for their history of Foxn1 expression using a Foxn1:Cre transgene (21) on the Rosa26-YFP background, in which YFP is expressed only after Cre-mediated excision of a stop cassette (22). In these thymi, essentially all TECs express the fluorescent marker (Fig. S1B). Though this type of analysis does not rule out the presence of a small subset of Foxn1-negative TECs without a history of Foxn1 expression, it suggests that the majority of Foxn1-negative TECs arises from Foxn1-positive TECs (Fig. 3). We confirmed that the activity of the Foxn1:eGFP reporter transgene correlates with endogenous Foxn1 gene expression in TECs (Fig. S2). In a second experiment, our aim was to address the functional capacity of Foxn1-negative TECs. Initially, we attempted to ablate Foxn1-expressing cells in adult transgenic mice to create a thymic microenvironment solely composed of Foxn1-negative TECs. Unfortunately, as was the case for treatment in late stages of gestation, DT and CB1954 were toxic (most likely owing to the effects on the skin) and caused high mortality rates. Therefore, this strategy was abandoned and instead a model of subacute intoxication of TECs was developed. Polyglutamine stretches in various proteins are associated with neurodegenerative disorders (23), and it has been shown that such cellular pathology can also be achieved in other cell types (24). Proteins carrying only a short stretch of glutamines are considered nonpathogenic, whereas stretches longer than 50 residues are regarded as toxic (23). We reasoned that we could exploit this property to examine whether the postnatal TEC compartment contains a functionally relevant Foxn1-negative lineage, which would be unaffected by the polyglutamine-associated toxicity. To this end, we developed two additional transgenic strains of mice expressing variants of EGFP modified by the N-terminal addition of either 19 or 82 glutamine residues (24), under the control of the Foxn1 promoter. In specific pathogen-free conditions, mice expressing these transgenes do not have obvious skin defects, are fertile, and have a normal life span. As expected, thymopoiesis proceeded unimpaired in Foxn1:eGFP and Foxn1:Q19-eGFP mice (Fig. 4A). By contrast, thymopoiesis was severely compromised in the Foxn1:Q82-eGFP strain, as judged by the reduced numbers of DPs (Fig. 4A Upper) and the decline of the thymopoietic index (ratio of DP thymocytes to TECs; Fig. 4A Lower). These findings suggest that toxicity in TECs is subacute/chronic rather than acute, and that the phenotype is due to functional deterioration of TECs rather than to their acute loss. Hence, in Foxn1:Q82-eGFP transgenic mice, the thymic microenvironment is formed normally but deteriorates functionally after birth. Interestingly, CD4 and CD8 single-positive T-cell subsets were equally affected (as evidenced by the unchanged CD4/CD8 ratio), although we had originally expected that CD8 cells would be more affected owing to the presumed impairment of the proteasome-dependent protein degradation pathway and the resulting reduction of MHC class I ligands. Hence, the cytopathic effect of polyglutamine expression in TECs appears to affect general aspects of TEC physiology.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Foxn1-negative cells in the adult thymus are descendants of Foxn1-positive cells. Flow cytometric determination of the percentage of fluorescent TECs (Epcam⁺CD45⁻) for the indicated genotypes at two different time points. For representative flow cytometric profiles, see Fig. S1.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Expression of polyglutamine-containing proteins causes thymic malfunction. (A) Thymopoietic activity in age- and sex-matched mice transgenic for the indicated constructs, expressed relative to control. Values for CD4/CD8 double-positive thymocytes (Upper) and thymopoietic index [number of CD4/CD8 double-positive thymocytes divided by number of thymic epithelial cells (Lower)] are shown. The number of mice per data point (mean ± SD were applicable) is indicated and applies to both panels. (B) Abnormal architecture of thymic microenvironment in Foxn1:Q82-eGFP transgenic mice at 2 mo of age. Note the impaired separation of cortex (marked by K8 staining) and medulla (distinguished by strong GFP staining) (Upper). The altered shapes of medullary and cortical TECs are particularly evident in higher-power views of corticomedullary junctions (Lower). The sections are derived from the 2-mo-old animals also shown in C. The tissue architecture of Foxn1:Q19-eGFP transgenic mice is normal and comparable to Foxn1:eGFP mice. (Scale bars: Upper, 200 μm; Lower, 100 μm.) m, medulla; c, cortex. (C) Distribution of thymocyte subsets in sex-matched mice transgenic for the indicated constructs, analyzed at 2 and 5 mo of age. The numbers of total thymocytes are given above the panels; the relative proportions of thymocyte subsets are indicated in percentages (red numbers). For this experiment, flow cytometry profiles are representative of four animals in each group. Results from wild-type, Foxn1:eGFP, and Foxn1:Q19-eGFP transgenic mice are similar. (D) Macroscopic appearance of thymic tissue in aged mice. Note the size of the thymus and the widespread fluorescence of thymic epithelial cells in Foxn1:eGFP mice (50 wk of age) and the involuted thymus (outlined by dashed line) of a Foxn1:Q82-eGFP mouse that contains only few fluorescent cell clusters (42 wk of age). (Scale bar: 100 μm.)

The functional impairment correlates with structural abnormalities. Foxn1:eGFP mice exhibit a normal architecture with well-separated cortex and medulla, strong GFP staining in the medulla, and reticular GFP/K8 double-positive network in the cortex. In Foxn1:Q82-eGFP mice, the demarcation of cortex and medulla is less clear, and the reticular network in the cortex is much less elaborate (Fig. 4B). By 5 mo of age, the majority of Foxn1:Q82-eGFP mice lack appreciable thymopoietic activity, as evidenced by the number of CD4/CD8 double-positive thymocytes; the relative increase of DN thymocytes in Foxn1:Q82-eGFP mice suggests that T-cell development is affected at an early stage (Fig. 4C). At this age, Foxn1-negative cells comprise a substantial fraction of TECs in Foxn1:Q82-eGFP mice; for instance, at 5 mo of age, 25.1 ± 9.8% (n = 5) do not express the transgene. Lack of Foxn1:Q82-eGFP transgene expression correlates with diminished endogenous Foxn1 activity (Fig. S2). Nonetheless, the thymopoietic index is disproportionally (∼1,000×) lower than in wild-type or Foxn1:Q19-eGFP and Foxn1:eGFP mice. This strongly indicates that Foxn1-negative cells do not have any appreciable thymopoietic function. The most likely explanation for this finding is that, even after the cessation of transgene expression, the functional capacity of TECs does not recover. However, we observed that the loss of thymopoietic activity does not affect the entire cohort of Foxn1:Q82-eGFP mice, as some mice sustained minimal thymopoiesis up to the age of 2 y. These thymi contained Foxn1-positive TECs, albeit in much lower numbers than those of Foxn1:eGFP and Foxn1:Q19-eGFP strains (Fig. 4D). In contrast with peripheral T cells in Foxn1:eGFP and Foxn1:Q19-eGFP mice, Foxn1:Q82-eGFP mice had fewer peripheral T cells, exhibiting the phenotype of activated T cells, as expected for a lymphopenic environment.