Development of Thymic Microenvironments In Vitro Is Oxygen-dependent and Requires Permanent Presence of T-cell Progenitors

Abstract

Development of a mature T-cell repertoire in the thymus depends on lympho-stromal interaction between thymocytes and stromal cells. To facilitate intercellular contact, the epithelium in the thymus has differentiated into a unique three-dimensionally (3D)-oriented network. Here we analyze factors influencing induction and maintenance of the 3D configuration of the epithelial network in fetal thymic lobes in vitro. We show that the 3D configuration of the thymic stroma depends on (a) the oxygen pressure in vitro and (b) permanent physical contact between stromal cells and developing thymocytes. This latter feature is demonstrated by incubation of fetal thymic lobes with deoxyguanosine (d-Guo), inducing a 2D-organized thymic stroma, with thymic cysts appearing. Reconstitution of d-Guo-treated lobes with a limited number of flow-sorted T-cell progenitors restores the 3D configuration of the thymic epithelium, but only at high oxygen pressure. This study underlines the plasticity of thymic epithelium and shows that the unique organization of the thymic epithelium is dependent on both oxygen and crosstalk signals derived from developing thymocytes.

The thymus is concerned with the development of a broad repertoire of T-lymphocytes, reactive with foreign antigens but tolerant to self-determinants. During generation of this physiological T-cell repertoire, developing T-cells interact with various types of stromal cells located in different thymic microenvironments (van Ewijk 1991; Boyd et al. 1993; Anderson and Jenkinson 2001). Thus, thymic microenvironments induce expansion and maturation of T-cell progenitors while screening developing T-cells for reactivity and avidity with MHC-peptide complexes expressed by thymic stromal cells (Jameson and Bevan 1998; Derbinski et al. 2001).

Epithelial cells form the basis of these microenvironments. Importantly, the organization and positioning of epithelial cells is unique for the thymus. Whereas in other organs epithelial cells are usually placed on a basement membrane to form two-dimensionally (2D)-organized sheets of cells which line off functionally different domains, epithelial cells in the thymus are organized in a sponge-like three-dimensionally (3D)-oriented fashion. This typical configuration facilitates migration and lympho–stromal interaction during thymopoiesis (Kyewski et al. 1987; van Ewijk 1988; Lind et al. 2001; Petrie 2002; Petrie and van Ewijk 2002).

The uniqueness of the thymic stroma is already established early in ontogeny in the region of the third pharyngeal pouch. In this region, expression of Hoxa3 and Foxn1 regulates early steps in the development of the thymic epithelium (Boehm et al. 1995; Manley and Capecchi 1995; Nehls et al. 1996). In addition, Pax1 and Pax9 have been shown to contribute to the primary development of the thymic stroma (Wallin et al. 1996; Peters et al. 1998; Hetzer-Egger et al. 2002). In addition to specific gene expression, epithelial–mesenchymal interactions crucially determine the potential of the third pouch endoderm to develop into a thymic primordium (Bockman and Kirby 1984; Itoi and Amagai 1998; Suniara et al. 2000). Thus, a two-layered epithelial cyst arises at day 11 of gestation (Ed11), which at Ed12 results in an epithelial cell cluster (Itoi et al. 2001). At this stage in embryonic development, progenitor T-cells enter the thymic “anlage” to start proliferation and differentiation in close contact with epithelial cells.

Signals derived from invading T-cell progenitors have been shown to contribute to the organization, positioning, and maintenance of the differentiated network configuration of the thymic epithelium, a phenomenon earlier designated as “thymic crosstalk” (van Ewijk et al. 1994; Holländer et al. 1995). Experiments using immunodeficient mice harboring arrests at discrete different levels in T-cell development have revealed that during differentiation T-cells subsequently crosstalk to different subsets of thymic epithelial cells (Palmer et al. 1993; van Ewijk et al. 1999,2000).

In the present study we analyze requirements for proper development of thymic microenvironments in vitro. First, we show that maintenance of the typical thymic architecture requires a high oxygen pressure in the culture system. Lowering the pO² in the culture system results in reorganization of stromal cells and loss of one dimension in the tertiary organization of the thymic epithelium. Second, we show that maintenance of the tertiary organization of thymic microenvironments requires permanent physical contact between developing thymocytes and thymic epithelial cells. To demonstrate this latter feature, we have purified thymic epithelium using de-oxyguanosin (d-Guo). d-Guo treatment of fetal thymic lobes blocks lymphoid development and results in thymic lobes exclusively composed of stromal elements (Jenkinson et al. 1982). Similarly to culture of thymic lobes at low pO₂, the thymic epithelium loses one dimension in organization and develops many cysts lined with “classical” epithelial cells. We show that this induced 2D configuration of the thymic epithelium is not a static phenomenon. The 2D configuration can be remodeled back into 3D, provided that both T-cell progenitors and a local high pO² are simultaneously present in the thymic lobes.

This study underlines the impressive plasticity of the thymic stroma and shows that the typical architecture of the thymic microenvironments depends on oxygen, together with continuous crosstalk signaling by developing thymocytes. Physiologically, both factors guarantee induction and maintenance of proper T-lymphocyte development.

Materials and Methods

Mice

Time-mated pregnant C57Bl/6 (B6/Ly5.2) mice were obtained from SLC (Shizuoka, Japan), and B6Ly5.1 were maintained in the animal facility of the Kyoto University according to the institutional guidelines.

Organs and Cells

Fetal thymic lobes (TLs) of d14, d15 embryos (B6) as well as fetal livers (FLs) of d14 embryos (B6Ly5.1) were isolated using a stereomicroscope. FL and FT cell suspensions were prepared as described previously (Kawamoto et al. 1997). Progenitors for T-cells were isolated from d14 fetal liver as Lin⁻ Sca-1^lo c-kit⁺ CD45⁺ cells.

High Oxygen Submersion (HOS) and Low Oxygen Submersion (LOS) Organ Culture: Deoxyguanosine Treatment

FTs were obtained from d14 embryos, placed directly in wells of a 96-well V-bottom plate, and cultured for 5 or 9 days. To obtain hematopoietic cell-depleted FT lobes, thymuses were cultured on polycarbonate filters (pore size 8 μm) floating on culture medium containing 1.35 mM d-Guo for 6 days in humidified air containing 5% CO₂. The lobes were washed and placed individually in wells of a 96-well U-bottom plate. To these lobes, 100 D14FL Lin⁻ Sca-1^lo progenitors were added using a fine capillary tube under direct microscopic visualization, followed by low-speed centrifugation of the plate for 5 min (experimental details in Table 1). All organ cultures that were to grow under HOS conditions were placed and sealed in plastic bags (Ohmi Oder Air Service; Hikone, Japan). The air in the bags was replaced with a gas mixture of 70% O₂, 25% N₂, and 5% CO₂. These plates, as well as plates with FTs under LOS conditions (normal air containing 5% CO₂), were placed in a 37C incubator. The cultures were maintained in RPMI 1640 complete medium (Life Technologies; Tokyo, Japan). Supplements were 10% FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, 2 mg/ml sodium bicarbonate, 0.1 mM non-essential amino acid solution (Life Technologies), 5 × 10⁵ mM 2-mercaptoethanol, 100 ng/ml streptomycin, and 100 U/ml penicillin. Medium was changed on day 5 for 9-day cultures and a new gas mixture was infused into each bag. For 5-day cultures the medium was not changed. After culture, all lobes were harvested and either processed for FACS analysis or frozen in OCT compound for immunohistochemistry.

Table 1

Experimental protocols

			Progenitor cells (B6Ly5.1)
Group	Thymic lobes	Source	No.	Oxygen conditions	Culture days
1	D14 FTa	−	−	HOS	5
2	D14 FT	−	−	LOS	5
3	D15dGuoFTb	−	−	−	−
4	D15dGuoFT	−	−	HOS	9
5	D15dGuoFT	−	−	LOS	9
6	D15dGuoFT	D14FL Lin− Sca-1lo	100	HOS	9
7	D15dGuoFT	D14FL Lin− Sca-1lo	100	LOS	9

^a

Fetal thymus, day 14 of gestation.

^b

Fetal thymus, day 15 of gestation, cultured in deoxyguanosine for 6 days.

Monoclonal Antibodies and Flow Cytometric Analysis

After culture, cells were recovered and pooled in groups of three lobes, and were counted using trypan blue dye exclusion. Cells were incubated with antibodies and analyzed by flowcytometry (FACS Vantage; Becton Dickinson, Mountain View, CA) as previously described (Kawamoto et al. 1997). The following monoclonal antibodies were used in this study: PE-anti-Thy1.2 (5a-8; Caltag Laboratories, San Francisco, CA), FITC-anti-B220 (RA3-6B2; Caltag), FITC-anti-Mac1 (M1/70; Caltag), FITC-anti-Gr-1 (RA3−8C5; PharMingen, San Diego, CA), PE-anti-CD4 (GK1.5; Caltag), FITC-anti-CD8 (YTS169.4; Caltag), PE-anti-TCRαβ (H57597; Caltag), FITC-anti-TCRγδ (GL3; Caltag), FITC-anti-CD4 (GK1.5; Caltag), FITC-anti-CD3 (500A2; Caltag), PE-anti-CD44 (IM7; PharMingen), anti-Ly5.1 (a201.7; kindly donated by Dr. Y. Saga, Banyu Seiyaku, Tokyo, Japan), and anti-CD25 (PC61.5.3; American Type Culture Collection, Rockville, MD). Anti-Ly5.1 and anti-CD25 were labeled with Cyanine5 (Cy5) using a labeling kit (Biological Detection Systems; Pittsburgh, PA). Anti-FcRγII/III (FcR; 2.4G2, PharMingen) was used to block nonspecific binding of monoclonal antibodies to the FcR.

Monoclonal Antibodies, Single-chain Antibodies, and Immunohistochemistry

From all different experimental groups, 6-μm frozen sections were prepared using a Leitz cryostat and collected on 0.1% gelatin-coated microscope slides. After a brief acetone fixation, sections were incubated with the following monoclonal antibodies: T-cells, anti-Thy1 (clone 59AD2.2), anti-CD3 (clone KT3), anti-CD4 (clone H129.19), anti-CD8 (clone 536.7); macrophages, F4/80, ER-HR3, ER-TR9 and MOMA1; cortical epithelium, ER-TR4; medullary epithelium, ER-TR5; fibroblasts, ER-TR7, anti-MHC class I (clone M1/42), anti-MHC class II (clone M5/114); dendritic cells, N418. After washing in PBS/Tween, the slides were incubated with rabbit anti-rat IgG conjugated to horseradish peroxidase (Dako; Glostrup, Denmark) as secondary reagent or horseradish peroxidase conjugated to rat anti-hamster IgG (Dako), to detect N418.

For staining with monoclonal single-chain antibodies (scFv), sections were fixed with 1% paraformaldehyde for 5 min followed by a brief wash in PBS/Tween. Slides were incubated with the cortical epithelial marker TB4-4 (van Ewijk et al. 1997), the common epithelial marker TB4-20 (van Ewijk et al. 1997), and TB3-110, identifying basement membranes (our unpublished observations). After thorough washing with PBS/Tween, biotinylated anti-myc (clone 9E10; ATCC) was used as secondary reagent to react with the scFvs, followed by incubation with horseradish peroxidase-conjugated streptavidin (Jackson Immunoresearch Labs; West Grove, PA) as the tertiary reagent, all at proper titration. Finally, all sections were developed using H₂O₂ activated diaminobenzidine as a substrate. All steps were carried out at room temperature.

Results

This study is based on the use of a fetal thymic organ culture system in which thymic lobes are completely submerged in tissue culture medium (Watanabe and Katsura 1993; Kawamoto et al. 2000; Katsura 2002). The efficiency of this system equals the “classical” organ culture system in which fetal TLs are placed on a filter floating in tissue culture medium (Jenkinson et al. 1982). In addition, the stromal architecture of the thymus in submersion cultures is similar to that in the classical filter system (van Vliet et al. 1985; van Ewijk, unpublished observations). Moreover, the submersion culture system allows accurate regulation of the oxygen pressure during culture.

In the present study we cultured lobes at two different pO₂ levels: (a) oxygen concentration at 70%, resulting in a local high oxygen pressure of 400 mmHg in the thymic lobes (high oxygen submersion culture; HOS), and (b) oxygen concentration at 20% resulting in a low pO₂ of 175 mmHg (low oxygen submersion culture; LOS); (cf. Watanabe and Katsura 1993).

To study the influence of thymocytes on the development of the thymic epithelium, isolated T-cell progenitor cells were co-cultured with d-Guo-treated fetal TLs under either HOS or LOS conditions.

T-cell Development Occurs Normally at High pO₂ But Is Arrested at the Triple Negative CD44⁺CD25⁻ Stage at Low pO₂

In day 14 fetal TLs, cultured under HOS conditions (group 1 in Table 1; and Figure 1, upper row) T-cell development proceeded normally. After 5 days of culture, the majority of the cells in the TLs were Thy1⁺ cells (Figure 1A), with both CD4⁺CD8⁺ (DP, double positive) cells developing, as well as SP (single positive) CD4⁺ and SP CD8⁺ thymocytes (Figure 1B). Both αβ and γδ TCR-expressing T-cells were generated (Figure 1C; cf. Watanabe and Katsura 1993; Kawamoto et al. 2000). Analysis of subsets of TN cells in HOS cultures showed that subpopulations of T-cell progenitors could be detected that were either CD44⁻CD25⁻, CD44⁺CD25⁻, or CD44⁻CD25⁺ (Figure 1D).

In contrast, when TLs were placed under LOS conditions (group 2 in Table 1; and Figure 1, lower row), the cell recovery was much lower compared to the HOS cultures (see also legend to Figure 1). Although limited numbers of T-cells expressing CD4⁺CD8⁺ or CD4⁻CD8⁺ were still present (Figures 1E and 1F), as well as some T-cells expressing αβ TCR and γδ TCR (Figure 1G), most of the remaining thymocytes were arrested at the TN (CD3⁻CD4⁻CD8⁻) CD44⁺CD25⁻ stage (Figure 1H). Incubation with propidium iodide indicated that the large majority of cells in TLs cultured under LOS conditions remained alive after 5 days of culture (data not shown).

Figure 1

(A-H) FACS profiles of cells recovered from fetal TLs cultured under HOS and LOS conditions. Thymic lobes from d14 embryos (B6) were cultured under HOS conditions (group 1 in Table 1) or LOS conditions (group 2 in Table 1). Nine lobes per group were cultured. Cells were recovered and pooled in numbers of three. FACS profiles were similar among the three samples, and representative samples are shown. (D,H) Anti-B220, anti-Gr-1, and anti-Mac1 were also used to gate in the CD3 CD4 CD8 (TN) subset. Number of recovered cells (mean ± SD) in HOS and LOS cultures are 11.7 ± 3.1 × 10⁴ per lobe and 0.8 ± 0.15 × 10⁴ per lobe, respectively.

3D-oriented Thymic Epithelium Develops in Thymic Lobes Only at High pO₂

To study the architecture of thymic microenvironments under HOS and LOS conditions, d14 fetal TLs were placed in an organ culture under HOS (group 1 in Table 1) or LOS (group 2 in Table 1) conditions. After 5 days of culture, the lobes were frozen and prepared for immunohistochemistry. Microscopic analysis revealed two different areas in each TL cultured under HOS conditions (Figure 2). TB4-4, an scFv reactive with cortical epithelium, identified a peripheral area consisting of 3D-organized cortical epithelial cells and a central area with closely packed, undifferentiated epithelial cells (i.e., rounded cells with an euchromatic nucleus that lack long cytoplasm processes and show low cytokeratin expression; see Discussion) (Figure 2A, asterisk). An organized medulla was lacking but a few medullary epithelial cells, defined by the monoclonal antibody ER-TR5, were found scattered throughout the TL (Figure 2B). ER-TR7 identified the thymic capsule, septae, and individual fibroblasts in the center of the lobes (Figure 2D). Therefore, a welllocalized medulla defined by groups of ER-TR5⁺ epithelial cells remains absent.

In each lobe, T-cells were abundantly present (Figure 2C). Importantly, the large majority of T-cells colocalized with the fine reticular network of epithelial cells in the periphery of the TLs (compare Figures 2A and 2C), indicating that a 3D organization of thymic epithelium is required for T-cell development.

When d14 fetal TLs were cultured under LOS conditions (group 2 in Table 1), cellularity remained low. In addition, cortical epithelial cells did not create their typical network architecture. Instead, closely packed cuboidal cortical epithelial cells were observed (Figure 2E). A few small, rounded medullary epithelial cells occurred spread throughout the lobes (Figure 2F). Fibroblasts, dendritic cells, and macrophages were present in TLs grown either under HOS or LOS conditions; they did not differ in frequency or in morphology (data not shown).

The 3D Architecture of Thymic Epithelium Is Lost After d-Guo Treatment

The finding that HOS conditions were required for the organization of a normal reticular network of epithelial cells raises the question of whether oxygen has a direct influence on the differentiation of epithelial cells alone. To investigate this possibility, endogenous thymocytes were removed by culturing the thymic lobes in the presence of d-Guo.

To determine the direct effect of d-Guo on the thymic stroma, d15 lobes were incubated with d-Guo for 6 days (group 3 in Table 1) and directly frozen in OCT compound. Frozen sections stained with scFv TB4-4 showed that cortical epithelial cells, instead of creating a network, now formed strands of cuboidal cells (Figure 3A). Similarly, medullary epithelial cells had retracted their cytoplasmic extensions (Figure 3B). Furthermore, thymic cysts developed under the influence of d-Guo, as shown in Figures 3A and 3B. Staining with monoclonal antibody specific for the TCR-CD3 complex indicated that d-Guo treatment efficiently removed endogenous thymocytes (data not shown). Thus, d-Guo treatment of TLs not only causes depletion of endogenous thymocytes but also results in repositioning of both cortical and medullary epithelial cells.

Figure 2

Frozen sections of mouse fetal TLs cultured under HOS conditions for 5 days (group 1 in Table 1) and incubated with TB4-20 (all epithelial cells) (A), ER-TR5 (medullary epithelial cells) (B), KT3 (T cells) (C), ER-TR7 (fibroblasts) (D), or under LOS conditions and stained with TB4-20 (E) and ER-TR5 (F). Asterisks in A and C refer to the central (undifferentiated) epithelial area in the cultured thymic lobes. Arrows in E outline the undifferentiated epithelial area in the cultured thymic lobes. Arrows in F indicate medullary epithelial cells. Bar = 120 μm.

High pO₂ Per Se Has No Influence on the 3D Architecture of Cortical Epithelial Cells

To study whether the pO₂ directly influenced the differentiation of the thymic stroma we cultured d-Guo treated lobes for a period of 9 days at high pO₂ or at low pO₂ (HOS, c.q. LOS conditions; groups 4 and 5 in Table 1). As shown in Figure 3C, under high pO₂ cysts remained present, the characteristic 3D organized epithelial network structure remained absent, and only closely packed, small cortical epithelial cells were observed, which at the periphery of the thymic lobe were oriented parallel to the surrounding basement membrane. These findings suggest that oxygen per se does not promote the differentiation and positioning of the epithelial network in the thymus.

Figure 3

Frozen sections of mouse fetal TLs treated with d-Guo for 6 days without further culture (A,B; group 3 in Table 1), or after d-Guo treatment followed by culturing under HOS conditions (C,D; group 4 in Table 1), or under LOS conditions (E,F; group 5 in Table 1) for 5 days and incubated with TB4-20 (all epithelial cells) and ER-TR5 (medullary epithelial cells). Asterisks in A and C indicate thymic cysts. Bar = 120 μm.

Progenitor T-cells Induce Differentiation of the 3D Architecture of Thymic Epithelial Cells Only at High pO₂

To study the potential of developing thymocytes to induce differentiation of the thymic epithelium, we cocultured d-Guo-treated fetal TLs with T-cell progenitors under either high or low pO₂ conditions.

Endogenous T cells were first removed by culturing d14 fetal TLs in the presence of d-Guo for 6 days. Subsequently, these d-Guo-treated lobes were in vitro reconstituted with 100 sorted Lin⁻ c-kit⁺ Sca-1^lo d14 fetal liver cells, which contain T lineage-committed progenitors (Kawamoto et al. 1997,2000). Reconstituted lobes were then placed under HOS and LOS conditions (groups 6 and 7 in Table 1) and analyzed by flow cytometry for T-cell development. FACS data showed that under HOS conditions (group 6 in Table 1), T-cell differentiation in the reconstituted lobes progressed beyond the TN CD44⁺CD25⁻ stage to end-stage maturation (Figure 4). The majority of the TN cells had lost CD44 and upregulated CD25. In addition, DP and SP thymocytes were generated. In contrast, under LOS conditions (group 7 in Table 1), T-cell differentiation remained completely arrested at the TN CD44⁺CD25⁻ stage. Thus, T cell development in the FL reconstituted d-Guo-treated fetal TL is comparable to T-cell development in the normal unreconstituted fetal TL placed under HOS or LOS conditions.

To study the development of the d-Guo-treated thymic stroma after reconstitution with T-cell progenitors, we stained cryostat sections of TLs with antibodies defining subsets of thymic epithelial cells and compared the organization of the thymic stroma with normal untreated fetal TLs (group 1 in Table 1). Comparison of Figures 5A and 2A shows that the architecture of the thymic epithelium in both groups is quite similar. At the periphery of the reconstituted d-Guo-treated lobes, a well-differentiated MHC-expressing reticular network of cortical epithelial cells had developed (Figures 5A and 5D), with thymocytes positioned in between processes of the epithelial cells (Figure 5C). In the middle of the lobes epithelial cells were present, but these cells are mainly undifferentiated cells, while thymocytes remained absent in this region. Compared to normal HOS cultures, ER-TR5⁺ medullary epithelial cells were present at increased frequency (Figure 5B), but these cells were not positioned in discrete medullae. They remained intermingled with TB4-4⁺ cortical epithelial cells (compare with Figure 5A). T-cell differentiation appeared normal, as demonstrated in Figures 4 and 5C.

By contrast, in d-Guo-treated fetal liver cell reconstituted lobes placed under LOS conditions, T-cells did not progress beyond the CD44⁺CD25⁻ phenotype (Figure 4; and group 7 in Table 1). Moreover, in these lobes cortical epithelial cells were unable to form a 3D-oriented network and remained grouped together in strands of cuboidal epithelial cells (Figure 5E). Similarly, medullary epithelial cells did not expand and were unable to form medullae. Thymic cysts remained present in the lobes, as shown in Figure 5E. Thus, under low pO₂, the 3D network of the thymic stroma was not induced. Together, these experiments clearly demonstrate an impressive plasticity of the thymic stroma and indicate that T-cell progenitors and a high pO₂ are simultaneously required for the development and 3D orientation of thymic microenvironments.

Figure 4

FACS profiles of cells generated in d-Guo-treated lobes cultured with FL progenitors under HOS and LOS conditions. d-Guo-treated lobes (B6) were cultured for 9 days with 100 c-kit⁺Sca1^lo cells from B6Ly5.1 mice under HOS (group 6 in Table 1) or LOS conditions (group 7 in Table 1). Nine lobes per group were cultured. Cells were recovered and pooled in numbers of three. FACS profiles were similar among the three samples, and representative samples are shown. Virtually all recovered cells are Ly5.1⁺ (data not shown). Number of recovered cells (mean ± SD) in group 6 and 7 are 13.5 ± 4.5 × 10⁴ per lobe and 0.4 ± 0.08 × 10⁴ per lobe, respectively. In D and H, TN subset was gated as in Figure 1.

Figure 5

Frozen sections of mouse fetal TLs treated with d-Guo for 6 days followed by reconstitution with 100 sorted FL progenitors, cultured under HOS conditions for 9 days (group 6 in Table 1) and incubated with TB4-20 (all epithelial cells) (A), ER-TR5 (medullary epithelial cells) (B), anti-Thy-1 (C), and anti-MHC class II (D), or LOS conditions for 9 days (group 7 in Table 1) and incubated with TB4-20 (E) and ER-TR5 (F). Asterisks in A and C indicate thymic cysts. Bar = 120 μm.

Discussion

Early steps in the development of the mouse thymus can be easily accessed by in vitro culture of d14-derived thymic lobes. Within a period of 6 days all thymocyte subsets differentiate normally, as under normal in vivo conditions (Jenkinson et al. 1982). Likewise, thymic microenvironments develop in vitro, but their organization differs from the in vivo situation (van Vliet et al. 1985). Thymic medullae do not develop in cultured lobes, and cortical and medullary epithelial cells remain intermingled at the peripheral part of the thymic lobes (van Vliet et al. 1985; and this study). Absence of medullae can be explained by the absence of a developing thymic vasculature under tissue culture conditions, because Anderson et al. (2000) have shown that the intrathymic vasculature programs the positioning of medullary epithelial cells. However, this in vitro positioning defect appears to have no obvious consequences for the development of T-lymphocytes (Jenkinson et al. 1982; and our present data).

In the cultured fetal thymus, thymocytes develop at the peripheral side of the lobes, in areas where epithelial cells are found in the thymus-characteristic network configuration. They do not develop in the central part of the lobe where clustered epithelial cells are located.

Such clustered epithelial cells have an undifferentiated morphological phenotype. Electron microscopic studies (van Ewijk et al. unpublished) have revealed absence of the characteristic long cytoplasmic processes. Moreover, bundles of tonofilaments are absent in these cells, corresponding to extremely low levels of cytokeratin expression, while incubation of sections with the antibody MTS 24 (Gill et al. 2002) detects precursor epithelial cells in the central area of the thymic lobe (van Ewijk et al. unpublished). These observations indicate that differentiation of the thymic epithelial reticulum is under the control of developing thymocytes and that the network configuration of the thymic epithelium is a basic microenvironmental requirement for efficient thymopoiesis.

Organ culture of fetal thymic lobes requires oxygen. In the classical filter-based culture system, only a thin film of culture medium separates the developing thymocytes from the air in the incubator. In the submersion culture system, as we show here, thymopoiesis clearly depends on the pO₂ in the culture system. At high pO₂ the tertiary structure of thymic microenvironments is maintained. Lowering the pO₂ induces a dramatic redistribution of epithelial cells, leading to the formation of thymic cysts lined with cells normally not occurring in the thymus.

This 3D to 2D conversion induced by hypoxia does not lead to cell death in the cultured thymic epithelium and is, in our opinion, not an in vitro artifact. First, thymic cysts also occur in vivo, even in normal mice (Farr et al. 2002), but especially in mice with a defect in early T-cell development (van Ewijk et al. 2000). Second, our recently performed electron microscopic study revealed a vital ultra structure of the thymic epithelium at 2D conditions (unpublished observations). Third, we show here that the 2D epithelial phenotype is still able to support development of γδ T-cells. Finally, as shown by our present reconstitution experiments, the 2D nature can be remodeled back to 3D.

There is ample information in the literature on the requirement for oxygen during the development of the immune system. Interestingly, a recent in vivo study from Trubiani et al. (2002) employing rats maintained under hypoxic conditions shows that the outer thymic cortex in young rats is pO₂-sensitive. This area is known to comprise TN III (CD3⁻CD4⁻CD8⁻CD44⁻ CD25⁺) and TN IV (CD3⁻CD4⁻CD8⁻CD44⁻ CD25⁻) thymocytes (Lind et al. 2001; Petrie 2002), and hypoxic conditions lead to a cellular depletion of lymphoid cells within this domain. The authors also mention that the thymic microenvironment is affected by hypoxia, but this conclusion is not validated by clear-cut immunohistology.

Redistribution of the thymic epithelium under hypoxic conditions appears to be an indirect phenomenon, primarily caused by lack of crosstalk signals derived from developing thymocytes. We do not feel that the reported impressive changes in the architecture of the thymic reticulum are caused by the mere physical absence of developing thymocytes, because other experimental models in which thymocytes are removed from the thymic microenvironment, such as hydrocortisone treatment (van Vliet et al. 1986) or neutron irradiation (Huiskamp et al. 1985), do not induce the formation of thymic cysts. These and the present findings indicate that remodeling of the thymic stroma can be induced only during early development of the thymus. The presence of other “classical” epithelial cells normally not occurring in the hypoxic thymus indicates that the thymic epithelium has reverted its differentiation program which, in absence of developing thymocytes, results in a default pathway in which ciliated cells, absorptive cells, and goblet cells are generated. The presence of many clustered, rounded epithelial cells with an undifferentiated phenotype, together with the observed increased frequency of MTS 24⁺ cells, further indicates a change in the differentiation program of the thymic epithelial reticulum under hypoxic conditions. However, the effect is not directly dependent on oxygen because a similar stromal phenotype also occurs after chemical treatment of fetal thymic lobes with d-Guo. d-Guo is known to block T-cell development within a period of 6 days (Jenkinson et al. 1982). Subsequent culture of d-Guo-treated lobes at high pO₂ did not recover the 3D architecture of the thymic epithelium, indicating that lymphoidderived factors crucially influence the epithelial phenotype. Culture under low pO₂ primarily blocks thymocytes expressing the αβ TCR-expressing thymocytes at the TN I (CD3⁻CD4⁻CD8⁻CD44⁺ CD25⁻) stage, while γδ T-cells develop normally. Therefore, crosstalk signals from early αβ TCR-expressing thymocytes appear to be of fundamental importance in development and maintenance of the thymic epithelial architecture. These in vitro findings are supported by our previously published in vivo work in which we showed that a block in early T-cell development, as occurs in CD3∊ transgenic mice, results in a similar 3D to 2D conversion of the epithelial phenotype (Holländer et al. 1995), while reconstitution of CD3∊ transgenic mice with progenitor T-cells derived from RAG^null mice restores the 3D phenotype (van Ewijk et al. 2000). In agreement with these in vivo findings, we recently found that reconstitution of d-Guo-treated lobes with T-cell progenitors from RAG^null mice similarly restores the 3D phenotype of the thymic epithelium, supporting the notion that differentiation of TN I to TN III thymocytes is instrumental in the development of the cortical thymic stroma.

The question arises whether thymocytes at early stages in development induce this particular 3D configuration or whether these cells merely play a role in the maintenance of this architecture after other inductive events. In support of the latter notion, a recent study by Klug et al. (2002) argues that initial steps in the differentiation of the thymic epithelium, as defined by differential expression of cytokeratins, are independent of thymocytes, whereas later stages in epithelial development would require sustained crosstalk signaling from thymocytes to maintain proper thymic microenvironments.

The molecular nature of thymic crosstalk signaling is now slowly emerging. Several different factors have been shown to influence the typical architecture of the thymic stroma. (a) Tumor necrosis factor (TNF) appears to be involved, since TCR ligation in developing T-cells induces upregulation of TNFα in thymic stromal cells (Lerner et al. 1996). (b) Neurotrophins, secreted by developing thymocytes, have been suggested to regulate the tertiary structure of the thymic epithelium (García-Suárez et al. 2000). In their study on the thymic architecture of mice deficient for TrKA, a signal transducing receptor for nerve growth factor, thymic cysts were found, similar in structure to cysts reported in transgenic mice overexpressing the human CD3∊ gene (c.f. van Ewijk et al. 2000) and cysts described in the present paper. (c) A new immunoregulator gene Aire has been identified in subsets of thymic epithelial cells. Its expression also appears under control of developing T-cells and regulates the 3D organization of the thymic stroma (Zuklys et al. 2000). (d) Similarly, GKLF, a transcriptional regulator expressed in thymic epithelial cells, appears under control of TN cells undergoing TCR β gene rearrangement (Panigada et al. 1999). (e) Stat3 expression has been implicated in the maintenance of properly organized thymic microenvironments, because disruption of the Stat3 gene in epithelial cells would prevent crosstalk signaling by thymocytes, resulting in de-differentiation of the thymic epithelium (Sano et al. 2001). (f) Furthermore, Wnt-mediated signals have been implicated in the proliferation, differentiation, and cell fate of epithelial cells (Cadigan and Nusse 1997). In this line, a recent paper by Balciunaite et al. (2002) shows that thymic epithelial cell lines transfected by a reporter construct activated by co-culture with thymocytes (but not peripheral T-cells) leads to activation of the Wnt signal transduction pathway, which directly activates Foxn1 in epithelial cells.

In summary, our observations indicate that in the absence of crosstalk signaling, thymic epithelial cells retract their typical cytoplasmic extensions and convert to a “simple” epithelium. This type of epithelium is found in the primitive thymic anlage and also in other endoderm-derived organs such as the respiratory tract, the gastrointestinal tract, thyroid, and parathyroid. In the presence of developing thymocytes, thymic epithelial cells differentiate into proper organized microenvironments. Maintenance of these microenvironments requires a high pO₂ within the thymic lobes. The role of epithelial stem cells (Bennett et al. 2002; Blackburn et al. 2002; Gill et al. 2002) in thymic crosstalk is now under investigation.