Previous analysis on the thymus of erythropoietin-producing hepatocyte kinases (Eph) B knockout mice and chimeras revealed that Eph-Eph receptor–interacting proteins (ephrins) are expressed both on T cells and thymic epithelial cells (TECs) and play a role in defining the thymus microenvironments. In the current study, we have used the Cre-LoxP system to selectively delete ephrin-B1 and/or ephrin-B2 in either thymocytes (EfnB1thy/thy, EfnB2thy/thy, and EfnB1thy/thyEfnB2thy/thy mice) or TECs (EfnB1tec/tec, EfnB2tec/tec, and EfnB1tec/tecEfnB2tec/tec mice) and determine the relevance of these Eph ligands in T cell differentiation and thymus histology. Our results indicate that ephrin-B1 and ephrin-B2 expressed on thymocytes play an autonomous role in T cell development and, expressed on TECs, their nonautonomous roles are partially overlapping. The effects of the lack of ephrin-B1 and/or ephrin-B2 on either thymocytes or TECs are more severe and specific on thymic epithelium, contribute to the cell intermingling necessary for thymus organization, and affect cortical TEC subpopulation phenotype and location. Moreover, ephrin-B1 and ephrin-B2 seem to be involved in the temporal appearance of distinct cortical TECs subsets defined by different Ly51 levels of expression on the ontogeny.
The thymus is an epithelial-based organ that provides a unique microenvironment for the generation of functional T lymphocytes capable of efficiently reacting to foreign Ags and, at the same time, tolerating self-Ags (1). During their development, thymic epithelial cells (TECs) and developing thymocytes establish a mutual cross talk that is necessary for the functional maturation of both thymic cell components (2, 3). The unique three-dimensional architecture of the thymic epithelial network facilitates the TEC–thymocyte (T-TEC) interactions (4). The adult thymus is, therefore, the sum of different microenvironments each composed of specific cell types with different functions (5). Lymphoid progenitors mature while they are guided orderly through these niches by different signaling systems (6). Although we know that the cortical compartment is responsible for expansion of the double-negative (DN) subpopulation as well as for positive selection (7) and that in the medullary compartment negative selection takes place (8, 9), we are still far from having a complete characterization of more specific niches contained in these two main thymic compartments.
The tyrosine kinase receptor family erythropoietin-producing hepatocyte kinases (Eph) and their ligand Eph receptor–interacting proteins (ephrins) provide positional information for cells and regulate cell-to-cell contacts, cell migration, attraction/repulsion phenomena, cell survival, and differentiation, participating in multiple morphogenetic processes (10, 11). Previous studies have demonstrated a relevant role for some members of this large family of molecules in different aspects of thymus biology. EphB2 and EphB3, and their ligands ephrin-B1 and -B2, are expressed on both thymocytes and TECs (12), and Eph B2 and/or EphB3 knockout mice showed an abnormal thymic development that affects mainly the epithelial component, including the cortex/medulla distribution, TEC morphology, and different epithelial-specific marker expression (13). They also showed decreased numbers of thymocytes and altered T differentiation particularly affecting the DN cell stage (12). Studies based on chimeras revealed that these receptors play an autonomous role both in T cell differentiation and TEC development, but they also affect nonautonomously the other thymic component (14, 15).
In the current study, we have used a Cre-LoxP model for the deletion of genes encoding for ephrin-B1 or -B2, ligands of EphB2 and EphB3, specifically either in TECs or in thymocytes, to evaluate in vivo the relevance of their expression in each thymic component on the biology of the organ for a better understanding of crosstalk. The obtained results confirm the importance of Eph–ephrins in T-TEC interactions and shed light on the characterization and development of thymic cortex microenvironments.
Materials and Methods
The C57/Bl6-EfnB1LoxP/LoxP mice were kindly donated by D.J. Anderson (California Institute of Technology, Pasadena, CA) and those of the C57/Bl6-EfnB2LoxP/LoxP strain by P. Soriano (Fred Hutchinson Cancer Research Center, Seattle, WA). The C57/Bl6-Tg(K5-Cre)Jt strain was developed by J.J. Takeda and donated by the Center for Animal Resources and Development of the Kumamoto University. C57/Bl6-Tg(lck-Cre) mice (Taconic Europe, Ry, Denmark) and CD1 mice (Harlan, Ibérica, Spain) were commercially purchased. All animals were bred and maintained under pathogen-free conditions in the facilities of the Complutense University of Madrid.
For mouse genotyping, specific sequences of the different alleles or transgenes were amplified by RT-PCR. We also performed RT-PCR to determine the efficiency of recombination, amplifying the intact or recombined allele from DNA isolated from either thymocytes or TEC suspensions. We have used the following oligonucleotides synthesized by Sigma-Genosys: EfnB1 A, 5′-AGCAGTGGGGTAGTAGTGACTACC-3′; B, 5′-TGGCCTTACACCCGCTTAAGA-3′; and C, 5′-CCTAACCACAGATGGTGGCC-3′; A + B, 54°C temperature of annealing (Ta), wild-type (wt) 250 bp, Loxp 500 bp; B + C, 58°C Ta, recombined allele detection 700 bp; EfnB2 wt forward, 5′-GCTGCCCGCGGCCGGTCCCAACG-3′ and wt reverse, 5′-CCGTTAGTGGCAACGTCCTCCGTCCTCG-3′; 61°C Ta, wt 580 bp. LoxP forward, 5′-GAGCCCCAGGTTCTAGAATAACTTCG-3′ and LoxP reverse, 5′-AAGTTATAAGCTTCAACGCGTCG-3′; 61°C Ta, LoxP 320 bp; C, 5′-CGGCCGGTCCATAACTTCGTATAGCA-3′, wt reverse + C, 65°C Ta, deleted 309 bp; Cre-K5 forward, 5′-ATGCCAATGCCCCCTCAGTTCCT-3′ and reverse, 5′-TGCCCCTTTTTATCCCTTCCAGA-3′; 62°C Ta, 300 bp. Cre-Lck forward, 5′-CGATGCAACGAGTGATGAGG-3′ and reverse, 5′-GCATTGCTGTCACTTGGTCGT-3′; 61°C Ta, 287 bp.
For light microscopy analysis, 8-μm cryosections were stained with 0.5% toluidine blue during 5 min and dehydrated in graded ethanol (70–100%) for 5 min and xylene (5 min) (Panreac). Sections were mounted in DPEX (Panreac).
Ulex europaeus agglutinin-1 (UEA-1) biotinylated (Vector Laboratories) (UEA). After three washes for 5 min in PBS, sections were incubated with the secondary Abs during 45 min at room temperature. Primary Abs were detected using anti-rabbit IgG Alexa Fluor 488, anti-rat IgG-biotinylated, and anti-rat Ig G/IgM Texas Red Abs (Molecular Probes, Invitrogen). After washing in cold PBS three times during 5 min, sections were mounted with Antifade Prolong Gold (Molecular Probes, Invitrogen) and analyzed using a Confocal Leica SP2 microscope (Leica Microsystems) at the Centre of Cytometry and Fluorescence Microscopy (Complutense University, Madrid, Spain).
Thymocytes were obtained by mechanical disaggregation of the thymuses. TECs were obtained by digestion with 2.5% trypsin. Embryonic day (E)18.5 or adult thymuses were carefully cut up with scissors into very small pieces and then incubated at 37°C during 15 min with trypsin. Fine needles were used to obtain a unicellular suspension. Trypsin activity was neutralized with 10% FCS, and the resultant cell suspension washed repeatedly in PBS 5 mM EDTA. Both TEC and thymocyte total number on each fetal lobule were determined by flow cytometry with the help of a suspension of a known concentration of beads (AlignFlow Plus flow cytometry alignment beads, 6 μm; Invitrogen) for calculating the exact volume analyzed. To detect different surface markers, cells were incubated at 4°C during 20 min with saturating concentrations of fluorochrome-conjugated Abs. The following Abs were used: CD4 (RM4-5), CD8 (53-6.7), TCRab (H57-597), CD44 (IM7), CD25 (PC61), lineage mixture (CD3ε, CD11b, CD45R/B220, Ly-76, Ly-6G, and Ly-6C), CD117 (2B8), epithelial cell adhesion molecule (EpCAM; G8.8), CD45 (30-F11), Delta-like ligand 4 (DLL4; HMD4-1), MHC class II (MHC II; M5/114.15.2), and anti-MTS20 (kindly donated by Richard Boyd, Monash University, Melbourne, Australia). For cell-cycle analysis, cells were fixed overnight in Cellfix (BD Biosciences) after surface marker staining and incubated afterward with Hoechst 33342 (Molecular Probes, Invitrogen) in EtOH 30% in PBS/1% BSA for 30 min at room temperature. Cells were analyzed in a BD LSR (BD Biosciences) using CellQuest software at the Centre of Cytometry and Fluorescence Microscopy (Complutense University). For cell-death analysis, cell suspensions were stained with Annexin-V–FITC (Roche Diagnostics) in HEPES buffer 1% FCS for 20 min at 4°C. To discard necrotic cells, the cell suspension was afterward incubated with 7-aminoactinomycin D (7-AAD; BD Biosciences) for 10 min at room temperature and resuspended for analysis. Annexin-V+/7-AAD− cells were considered to be apoptotic cells. Cells were analyzed in an FACSCalibur (BD Biosciences) and with CellQuest software. A total of 20,000 cells/sample at least were analyzed, and nonviable cells were excluded by forward-side scatter in all cases. Ten wt adult mice, five EfnB1tec/tec mice, nine EfnB2 tec/tec mice, and seven EfnB1tec/tecEfnB2tec/tec mice were used for thymocyte analysis.
Characteristics of mutant mice
We performed conventional crossing between C57/Bl6-EfnB1LoxP/LoxP or C57/Bl6-EfnB2LoxP/LoxP mice strains and the C57/Bl6-Tg(K5-Cre)Jt strain, in which recombinase is expressed under the control of keratin 5 gene promoter, to get the new strains: EfnB1loxP/loxP-Tg(K5-Cre)Jt (EfnB1tec/tec) and EfnB2loxP/loxP-Tg(K5-Cre)Jt (EfnB2tec/tec), respectively. We obtained the double-mutant strain (EfnB1tec/tecEfnB2tec/tec) by crossing these two single-mutant strains.
K5 is expressed in the basal and mitotically active layers of most stratified epithelia (16) and downregulated throughout epithelial differentiation (17). At E11.5, most or all of the epithelial cells of the thymic anlage express K5 (18, 19). In addition, it has been previously demonstrated that thymuses from mice coming from crossing the C57/Bl6-Tg(K5-Cre)Jt strain, used in our study, and mice carrying the transgene STOPflox-GFP (CAG-loxP-STOP-loxP-GFP) (20) showed GFP+ TECs all over the organ (21). So, we assume that in the EfnB1tec/tec, EfnB2tec/tec, and EfnB1tec/tecEfnB2tec/tec mice strains generated, Efnb1 and Efnb2 genes are deleted approximately from E11.5. The specific deletion of the genes in TECs and not in thymocytes was confirmed by RT-PCR both on single and double adult and fetal mutants (Fig. 1Aa).
The three mice genotypes generated (EfnB1tec/tec, EfnB2tec/tec, and EfnB1tec/tecEfnB2tec/tec) presented a Mendelian frequency at birth. EfnB1tec/tec mice showed no fertility or viability alterations; however, both EfnB2tec/tec and double mutants presented alterations in other epithelia showing shorter hair (not shown) and abnormally open eyelids (Fig. 1Ab). In mouse embryos, eyelid closing is a consequence of the active proliferation of epithelium on both sides of the corneal surface to the opposite side. Once epithelial fusion has taken place, dermis coming from neural crest cells keeps developing underneath it. At E16.5, wt mouse fetuses have closed eyelids; they start to open them at 3 to 4 d after birth and have them totally opened at 12 d old (22). All EfnB1tec/tecEfnB2tec/tec mice and most of the EfnB2tec/tec were born with open eyelids (Fig. 1Ab), an alteration that was observable, at least from E17.5. This same phenotype has been previously described in EfnB1+/−EfnB2+/− mice (23). Our observations indicate that the lack of these ephrins in the epithelium is enough to prevent eyelid closure.
We also performed conventional crossing between C57/Bl6-EfnB1LoxP/LoxP or C57/Bl6-EfnB2LoxP/LoxP mice strains and the C57/Bl6-Tg(lck-Cre)Jt strain in which recombinase is expressed under control of the lck gene proximal promoter, which is more active in immature thymocytes (24), generating the new strains: EfnB1loxP/loxP-TgN(Lck-Cre) and EfnB2loxP/loxP-TgN(Lck-Cre), EfnB1thy/thy and EfnB2thy/thy, respectively. We obtained the double-mutant strain (EfnB1thy/thyEfnB2 thy/thy) by crossing these two single-mutant strains.
Lck is an src family protein that associates to cytoplasmic tails of CD4 and CD8 coreceptors (25). It was shown that, when GFP is expressed under the proximal lck promoter, the most immature CD44+CD25− thymocytes are the first ones to express GFP. More than 70% of CD44+CD25+ cells (DN2) and most CD44−CD25+ thymocytes (DN3) were GFP+ in 4-wk-old animals (26). In our study, we confirmed by RT-PCR that allele deletion had taken place in thymocytes but not in other nonlymphoid tissues (Fig. 1Ac) that, accordingly, did not exhibit significant alterations.
For all of the parameters analyzed, no differences between EfnB1loxP/loxP, EfnB2loxP/loxP, and wt mice were found (data not shown), and thus, these genotypes were considered as wt.
Ephrin-B1 and/or ephrin-B2 expression on TECs is necessary for correct development of the thymus and a normal T cell differentiation
Mice with ephrin-B1– and/or ephrin-B2–deficient TECs, particularly the EfnB1tec/tec EfnB2tec/tec mice, presented smaller thymuses than wt ones (Fig. 1B). The thymic sections obtained from single-mutant mice and stained with toluidine blue revealed a normal cortex–medulla distribution, although there were variations between individuals, and a stellate-shaped medulla as observed in wt mice. In the EfnB1tec/tec EfnB2tec/tec mice, however, we observed an extremely thin and scarcely dense cortex that surrounded an inner, little branched medulla filled up with thymocytes (Fig. 1B).
The K5 and K8 expression pattern analysis showed an increase of the proportion of K5+K8+ cells in the three mutant mice thymuses (Fig. 1Ca–d). We have observed larger and more abundant medullary cysts in both EfnB1tec/tec and EfnB1tec/tec EfnB2tec/tec mice (Fig. 1Ca–d). Vascular and trabecular component distribution were also evaluated. Blood vessels, fibroblasts, and extracellular matrix components occupied a larger proportion of the organ in the mutant mice than in wt ones, particularly in the EfnB1tec/tecEfnB2tec/tec mice (Fig. 1Ce–h, 1Ci, 1Cl). This finding did not seem to be the result of a higher development of these elements but the consequence of a lower number of thymocytes and/or TECs in the mutant organs.
We analyzed by immunohistofluorescence the thymocyte arrangement in relation to the epithelial network. EfnB1tec/tec EfnB2tec/tec and approximately one half of the EfnB2tec/tec analyzed mice showed a densely packed epithelial network (Fig. 1Cc, 1Cd), depleted of CD45+ lymphoid cells (not shown). The other half of EfnB2tec/tec mice analyzed and all of the EfnB1tec/tec mice showed thymuses with a heterogeneous organization, containing areas with numerous thymocytes and few TECs that formed an outstretched epithelial network, and regions consisting of highly packed TECs with few or no thymocytes (Fig. 1D).
In contrast, the three mutant genotypes showed lower thymocyte numbers than the wt genotype (Fig. 2A). The flow cytometry study demonstrated that most EfnB1tec/tec thymuses showed no differences in the proportions of distinct thymocyte subpopulations with respect to wt values; from five EfnB1tec/tec mice analyzed, only two showed decreased proportions of single-positive (SP) mature cell (Fig. 2B). From nine EfnB2tec/tec thymuses analyzed, three presented decreased proportions of TCRαβ+ cells (not shown) and a slightly increased proportion of DN4 (CD44−CD25−) cells, plus a reduction of the TCRαβhiCD69hi subset within the double-positive (DP) subpopulation (Fig. 2B). Occasionally, (1 out of 9) EfnB2tec/tec thymuses exhibited increased DN3 (CD44−CD25+) cell proportions, with an inherent decline in the percentage of DP cells (not shown).
Although all EfnB1tec/tecEfnB2tec/tec mice showed reduced numbers of TCRαβ+ cells and most of them also a reduction in TCRαβhiCD69hi, there were notable differences in the thymocyte phenotypes among individuals. From a total of seven mice, we observed percentual accumulations in different thymocyte subpopulations, including DN1, DN3, CD8inm, and/or DP cells (Fig. 2C). These changes in the lymphoid content of EfnB1tec/tecEfnB2tec/tec thymuses were reflected in the thymic histology related to the degree of development of the medullary or cortical compartments (Fig. 2D). These results suggested a role of ephrin-B2 in the transitions from DN cells to the DP cell compartment and of both ephrins from this last one to SP cells, which correlated with the underdeveloped cortex observed in the EfnB2tec/tec and the smaller size of medullary areas of both EfnB2tec/tec and EfnB1tec/tec thymuses. The impairment of early T cell development was more frequent in EfnB2tec/tecEfnB1tec/tec thymuses than in EfnB2tec/tec thymuses, suggesting the occurrence of a cooperative role of both ephrin-B1 and -B2 in the DN-to-DP cell transition.
We also evaluated whether decreased numbers of thymocytes observed in mutant thymuses were related to changes in thymocyte survival or proliferation. Whereas no significant variations occurred in EfnB1tec/tec or EfnB2tec/tec mice, the EfnB2tec/tecEfnB1tec/tec mutants exhibited an increase in the apoptotic rate without changes in thymocyte proliferation (data not shown).
Ephrin-B1 and/or ephrin-B2 expression in thymocytes is necessary for a normal T cell differentiation and a correct histological organization of the organ
Decreased numbers of thymocytes were observed in EfnB1thy/thy, EfnB2thy/thy, and EfnB1thy/thyEfnB2thy/thy adult mice respect to control values, without significant differences between genotypes (Fig. 3A). The study of T cell differentiation revealed an increase in the percentage of DN thymocytes in the three thymocyte-conditioned mutant mice; this increase correlated, in some cases, with a decrease in the DP cell percentage (Fig. 3B) and, in other ones, with decreased numbers of both DP and SP cells (not shown). Changes in the DN cell subset included a lower percentage and number of DN1 cells and also an increase in the DN3 cell subset in correlation with a decline in the later stage, DN4 (Fig. 3C), apparently reflecting alterations in the DN3-to-DN4 cell progression. These alterations in the subpopulation proportions were accompanied by a higher percentage of apoptotic cells, especially in the DN cell subpopulation of the three mutant genotypes (Fig. 3D). Interestingly, the proportion of proliferating cells did not decrease but underwent a significant increase, particularly in the DP cell subset and some other stages (Fig. 3E).
It is known that the phenotype induced by mutations in distinct members of the Eph/ephrin family is dependent on the murine genetic background (27). We had already found differences between EphA4−/− mice generated in the C57/Bl6 and C57/Bl6-CD1 strains. When we analyzed both ephrin-B mutations conditioned to T cell lineage in a mixed genetic background C57/Bl6-CD1, we observed a similar phenotype than that described in C57/Bl6 mice, finding higher variability between individuals in the phenotype severity and individuals with a more severe phenotype (Fig. 3F). They showed a smaller thymus, fewer thymocytes (EfnB1thy/thy, 69–82 × 106; EfnB2thy/thy, 22–100 × 106), and higher proportions of DN cells (Fig. 3G) than mutant mice generated in C57/Bl6 background.
The histological study of thymocyte-conditioned ephrin-B mutant thymuses demonstrated that in both C57/BL6 and C57/BL6-CD1 mixed-background mice, the medullary areas were smaller, the proportion of K5+K8+ TECs was slightly increased (Fig. 3Ha), and a more densely packed epithelial network occurred in both cortex and medulla (Fig. 3Hb). The C57/BL6-CD1/EfnB1thy/thy mutants showed more and larger epithelial cysts (Fig. 3Hc), whereas C57/BL6-CD1/ EfnB2thy/thy mice exhibited important alterations in the structure of the organ. Many of them contained large areas devoid of epithelium but populated by CD45+ cells, next to regions without thymocytes and a strongly packed epithelium (Fig. 3Hc). In other thymuses of these mice, the T-TEC distribution was similar to that of control mice but showed a denser epithelial network (not shown).
Thus, all mutations affect thymocyte development, although in a different way. Although both thymocyte-conditioned ephrin-B1 and ephrin-B2 mutations affect T cell development autonomously, the TEC-conditioned mutation of EfnB2 has a greater nonautonomous impact on thymocyte development than the TEC-conditioned EfnB1 mutation. All of the TEC-conditioned and thymocyte-conditioned mutations of EfnB1 and/or EfnB2 affect epithelial organization and phenotype. To go further into this issue we focused on the analysis of cortical epithelium.
Ephrin-B expression on both TECs and thymocytes is necessary to define the phenotype and correct arrangement of adult cortical TEC subpopulations
We studied the cortical compartment of the mutant thymuses by analyzing the expression of Ly51, previously described as a cortical epithelial marker (28, 29). In agreement, MTS10, a medullary TEC (mTEC)-specific marker, and Ly51 expression were almost mutually exclusive in wt adult thymus (Fig. 4c). We also found that most Ly51+ cells of the corticomedullary junction also expressed K5 (Fig. 4c, 4e, 4f) and, therefore, corresponded to K8+K5+ cells (Fig. 4b).
The thymus of EfnB1tec/tec, EfnB2tec/tec, and EfnB1tec/tecEfnB2tec/tec mice presented increased numbers of Ly51+K5+ TECs (Fig. 4e, 4f, 4h, 4i, 4k, 4l, 4n, 4o), a finding also observed in the thymus of C57/Bl6-CD1 EfnB1thy/thy (Fig. 4q, 4r) and EfnB1thy/thy (Fig. 4t, 4u) mice and in that of C57/Bl6 EfnB1thy/thyEfnB2thy/thy mice (not shown). These results implied that at least part of the K5+K8+ TECs for which proportion had increased in these mice (Fig. 1C) were also Ly51+, cortical TECs (cTECs).
In the wt thymuses, we could distinguish by immunohistofluorescence two cTEC subpopulations based on the levels of Ly51 expression (Fig. 4a, 4d). Ly51hi TECs appeared in the subcapsulary cortex, forming a discontinuous layer of TECs in the cortex–medulla limit. The rest of the cortex was constituted by some Ly51hi TECs arranged longitudinally and radially around the medulla and, between them, a Ly51lo round-shaped cell subpopulation (Fig. 4a, 4d). We confirmed by flow cytometry these two levels of expression of Ly51 in adult thymic TECs (Supplemental Fig. 1A). As it has been previously described (28, 30), we observed that Ly51hi TECs are MHC IIhi, and the Ly51lo subset correlates to the MHC IIlo one (Supplemental Fig. 1B). The combined staining of Ly51 and of the Notch ligand DLL4 revealed that the Ly51hi TEC subset was also DLL4+, whereas most of the Ly51lo subset was DLL4− (Supplemental Fig. 1C). The association of the expression of DLL4, specifically to the DN microenvironment, has been reported (31–33), in that the different levels of expression of Ly51 seem to distinguish between two functionally different cTEC subsets. Mutant mice showed alterations in Ly51 expression and the distribution of these cTEC subpopulations. In EfnB1tec/tec thymuses, the difference in Ly51 expression was less evident, and cTECs were retracted, forming a more homogeneous and dense compartment (Fig. 4g, 4h). In the EfnB2tec/tec thymuses, the Ly51 expression clearly defined two different regions: in the outer cortex, most cells were Ly51lo, whereas in the inner cortex, they were Ly51hi (Fig. 4j, 4k). EfnB1tec/tecEfnB2tec/tec thymuses exhibited a mixed phenotype: they contained stronger stained Ly51 cells in the inner region and Ly51lo cells in the outer one, as in EfnB2tec/tec thymuses (Fig. 4m, 4n) but, as in the EfnB1tec/tec thymuses, the general appearance of the cortical compartment was more homogeneous on the basis of Ly51 expression.
In the thymuses with mutations conditioned to thymocytes, especially in those of mixed genetic background, the cortical phenotype was more similar to that of EfnB1tec/tec thymuses with a more homogeneous Ly51 staining (Fig. 4p, 4q, 4s, 4t).
We observed no differences between none of the mutant mice thymuses and wt ones in the MHC II expression analyzed by immunofluorescence (data not shown).
The absence of EfnB1 and/or B2 from TECs induced a delay in the maturation of thymic cortical epithelium
Because the EfnB1tec/tecEfnB2tec/tec thymuses showed the more severe alterations, we analyzed in these mutants possible changes in number, survival, and proliferation of TECs, as well as thymic cortex development during ontogeny.
We evaluated TEC number by flow cytometry, observing a reduction in TEC number in EfnB1tec/tecEfnB2tec/tec thymus at E14.5 and also at E18.5 (Fig. 5A). These lower numbers of mutant TECs correlated with an increased proportion of apoptotic TECs observed in EfnB1tec/tecEfnB2tec/tec thymuses at E18.5 and 25 postnatal days (PN) (Fig. 5B). However, no changes were detected in the proliferation rate (S + G2 + M phase TEC proportion) of mutant TECs (Fig. 5C).
At E13.5, UEA+ TECs progeny are exclusively mTECs (34), and we and other authors (28) have observed that, in adult thymus, Ly51 and UEA expression are mutually exclusive: Ly51+UEA− and Ly51−UEA+ subsets correspond to cTECs and mTECs, respectively. Thus, by using these two markers, we analyzed by flow cytometry evolution of the two populations during thymus ontogeny (Fig. 5A). TECs were identified as CD45−EpCAM+ cells in total thymus suspensions, and UEA− TECs were considered to be mostly cortical cells and UEA+ ones, mTECs.
As previously described (28), in wt thymuses, at earlier stages of thymus development, cTECs represented the majority of thymic epithelium; mTECs increased gradually—at the perinatal period, the cTEC/mTEC ratio was ∼1, and, after birth, the mTEC number exceeded that of cTECs (Fig. 5Aa). Ly51 expression gradually increased from E12.5 onward, finally giving rise to different postnatal cortical subpopulations on the basis of Ly51 and/or UEA expression (Fig. 5Aa, 5Ba). Remarkably, when we analyzed the absolute number of TECs within each level of Ly51 expression along development, the number of each fetal subpopulation either remained unchanged or increased, but never decreased with age (Fig. 5Ca), suggesting the sequential appearance of distinct epithelial subpopulations with increasing expression of Ly51 rather than the occurrence of a single subpopulation that upregulates Ly51 expression.
In EfnB1tec/tecEfnB2tec/tec thymuses, mTECs were less represented than in wt ones at postnatal stages but not during fetal life (Fig. 5A), confirming that EfnB1 and EfnB2 must also play a role in the postnatal expansion of medulla. We also found differences in the UEA− Ly51+ TEC cell populations from E14.5 onward (Fig. 5B). At each stage, the proportion of Ly51lo TECs was higher in mutant than in wt thymuses. Absolute numbers revealed that this percentual accumulation of Ly51lo cells corresponded to a delayed appearance of the Ly51hi populations (Fig. 5C). At E12.5, the absolute number of Ly51− TECs was slightly lower in mutant thymuses than in wt ones. However, at E14.5, this subpopulation recovered normal numbers that remained at E18.5. At E14.5, both Ly51lo and Ly51med subsets of mutant mice are lower than in wt. At E18.5, the numbers of both Ly51med and Ly51hi cells remained significantly lower in mutant mice. Altogether, these results suggest a delay of the cortical TEC differentiation in EfnB1tec/tecEfnB2tec/tec mutant mice.
Immunofluorescence microscopy analysis of Ly51 staining revealed very low levels of expression at fetal stages, and we were unable to visualize these differences between wt and EfnB1tec/tecEfnB2tec/tec fetal thymuses. After birth, mutant thymuses already showed visibly smaller cortical Ly51+ regions than wt ones (not shown).
As MTS20 is a marker associated with early TEC progenitors, we analyzed the combined expression of UEA and MTS20 in the CD45−EpCAM+ cell population. Considering that, from E14.5 onward, UEA− TECs are cortical Ly51+ cells, we identified UEA+MTS20+ cells as immature mTECs and UEA−MTS20+ as immature cortical epithelial cells (Fig. 5D). At E12.5, all TECs were MTS20+ (35) (Fig. 5E). In both UEA− and UEA+ TECs, the MTS20 expression was progressively downregulated, although in the medullary compartment of UEA+, the percentage of MTS20+ cells remained higher than in cortical UEA− TECs. In contrast, we did not observe significant differences in the proportions of MTS20 cells between mutant and wt thymuses (Fig. 5E). Considering that delayed appearance of cTEC populations with increasing levels of Ly51 in EfnB1tec/tec EfnB2tec/tec thymuses did not correlate with accumulation of MTS20+ cells, we analyzed in wt mice the combined staining of these two markers during thymus development. We found that, in all studied stages of wt thymuses, part of each cell subpopulation defined by the level of Ly51 expression also expressed MTS20 (Fig. 6B). Gradually, there was a reduction in the proportion of MTS20+ TECs within each Ly51+ subpopulation (Fig. 6B), but the first cells appearing in each of them were MTS20+ (i.e., 100% Ly51med TECs were MTS20+ at E13.5) (Fig. 6B, 6C).
Therefore, two different processes occurred during the development of thymic cortical epithelium: the sequential appearance of populations with an increasing level of expression of the Ly51 cortical marker and the downregulation of MTS20 in each of them, EfnB1 and/or EfnB2 expression on TECs being required for the former.
Our results demonstrate that EfnB1 and/or EfnB2 deletion conditioned either to thymocytes or TECs has an impact on T cell development but also on the histological organization of thymus. These results, therefore, confirm and extend previous data (13, 14, 36, 37) that suggested both autonomous and nonautonomous effects of Eph/Ephrin-B in distinct aspects of thymus biology.
The deletion of EfnB1 and/or EfnB2 conditioned to immature thymocytes has an autonomous effect on T cell differentiation that results in hypocellularity and a partial blockade of T cell development at the DN3 stage, as recently reported by other authors who studied similarly conditioned mice (36, 38). Although previous reports on the contribution of each ephrin to T cell development seem contradictory (36, 38), our results indicate that the lack of EfnB1 or EfnB2 in thymocytes is sufficient to induce a partial blockade of T cell differentiation, affecting largely the DN cell compartment. Although some redundancy cannot be completely discarded, we have found no significant differences in the thymocyte phenotypes of single- and double-mutant mice. Thus, ephrin-B1 and ephrin-B2 expressed on thymocytes must have cooperative rather than redundant roles in thymocyte development.
Few data are available about the mechanism driving this autonomous role of ephrin-B signaling in T cell development. The increased proportions of DN3 cells observed could be related to a role of ephrin-B in modulation of the signaling of molecules involved in β-selection or DN cell progression, such as Notch or IL-7. In this way, Luo and colleagues (39) have recently reported reduced IL-7Rα expression in thymocytes of EfnB1thy/thyEfnB2thy/thy. In contrast, increased cell death observed in the DN cell subset and, in some mice, in DP and SP cells of the three studied mutants contribute to the decrease in thymic cellularity. In addition, we reported decreased lymphoid progenitor colonization of Eph B2−/− thymic stroma (37). Nevertheless, in thymocyte-conditioned ephrin-B1/-B2 mutants, there is also a significant increase in the proportions of proliferating cells, especially in the DP cell compartment, that could partially recover the cell content, a finding previously reported by others (40).
One important factor to explain the variability in the severity of mutant phenotypes are the differences in the genetic background of the mice analyzed. More severe and variable phenotypes occur in mutant mice of C57/B16-CD1 mixed background than in the C57/B16 strain. We found a similar variability in the thymocyte phenotypes of EphA4-deficient mice of different genetic backgrounds (41) in agreement with other studies on various Ephs and Ephrins in different systems (27, 42). Ephs and Ephrins act in coordination with many signaling networks that are globally affected by the mouse genetic background (43).
The lack of ephrin-B1 and/or ephrin-B2 in TECs also affects T cell development. In this case, the effects of ephrin-B2 absence on TECs seem to be more important, but ephrin-B1 may also be relevant as EfnB1tec/tecEfnB2tec/tec mice show an earlier and more severe blockade on DN cells than single TEC-conditioned mutants. However, it has been described that ephrin-B2 deletion conditioned to IL-7 expression does not interfere with T cell differentiation (44). Differences between these results and ours could be either due to the high variability observed between individuals and/or to differences in the time and pool of cells affected by ephrin-B2 deletion. This would argue in favor of a main role of ephrin-B2 in TEC development that would indirectly affect adult T cell development. However, several data support a direct role of both ephrin-B1 and -B2 expressed by TECs on T cell development. Both ephrin-Bs are expressed by mature TECs (12), and published results reveal that EphB2 and/or EphB3 forward signaling on thymocytes that depend on signals provided by TECs modulate DP T-TEC adhesion, T cell development, and TCR stimulation (14, 45). Therefore, our current results on the effects of TEC-conditioned mutations on adult T cell development can be interpreted as the sum of a direct role plus the indirect consequence of an altered TEC maturation.
Both thymocyte- and TEC-conditioned deletion of Efnb1 and/or Efnb2 have effects not only on T cell development but also on the epithelial organization of thymus. In all studied mutants, particularly in the EfnB1tec/tecEfnB2tec/tec mice, the thymuses are smaller and show a higher proportion of K5+K8+ cells and epithelial cysts than wt thymuses, exhibiting less developed cortex and medulla. This implies that ephrin-Bs need to be expressed in both thymocytes and TECs, probably affecting their cross-talk mechanisms. Accordingly, all mutant thymuses show alterations in the organization of epithelial network and T-TEC interconnections. The role of Ephs and ephrins in the regulation of attraction–repulsion phenomena, the establishment of tissue boundaries and exclusive borders, and the intermingling of different cell populations is widely known (46). In a particular case as DP–TEC interaction, previous results from our laboratory had demonstrated that Eph–ephrins modulate both kinetic of conjugates and functional immunologic synapses formation (14, 45). In accordance to this, our current results show that the deficiency of ephrins B1 and/or B2 on TECs causes alterations on positive selection that are reflected on a decline of TCRαβhiCD69hi percentages within the DP subset.
With respect to the cortical epithelium, different levels of Ly51 expression allow us to distinguish two cTEC subsets in adult thymus. Cortical thymic epithelial heterogeneity has been broadly described, and some recent data identify functional cortical niches in association with the regulated expression of several markers, such as DLL4, MHC II, or IL-7 in cTECs (32, 47). Our results demonstrate that Ly51hi, but not Ly51lo, expresses DLL4, a Notch ligand, that drives T cell lineage commitment (48, 49). During early phases of thymic organogenesis, DLL4 is expressed by the majority of TECs, and its expression diminishes with age, being confined mainly to a fraction of cTECs at the corticomedullary junction in the adult thymus (49, 50). Moreover, it has been recently described that DN3 thymocytes require a DLL4-induced Notch signal to differentiate into DP cells (31–33). In contrast, in RAG−/− mice, which have a blockade on T cell development at DN3, most TECs are MHC IIhiLy51hi, and just 15% of the TECs are MHC IIloLy51lo, this latter subpopulation being higher (30%) in TCRα−/− mice, in which T cell differentiation is blocked in the DP to SP transition (28). Taken together, these data suggest that Ly51hi TECs would correspond to the niche of DN thymocytes and that of DP cells would be composed largely by Ly51lo TECs. These also suggest that the sequential appearance of the different T cell subpopulations could contribute to the differentiation of TEC subsets (51).
Our immunofluorescence study is consistent with this hypothesis as, in wt thymus, Ly51hi TECs are at the corticomedullary junction, arranged longitudinally throughout the cortex and in the subcapsule. Ly51lo TECs occupy the interradial spaces of cortex, and their morphology resembles the thymic nurse cells, tightly embracing DP thymocytes (52). This arrangement fits well with the described DN and DP niches and the journey followed by developing thymocytes throughout the cortex: DN cells would go up from the corticomedullary junction toward the subcapsule through the Ly51hi niche and then come back as DP cells through the Ly51lo TECs, where they suffer positive selection (Fig. 7).
Some of the studied mutant mice (EfnB1tec/tec, EfnB1thy/thy, and EfnB2thy/thy) show a more homogeneous Ly51 expression, so Ly51hi and Ly51lo subpopulations are not so clearly distinguishable, and most TECs are Ly51med/hi cells. Also, TEC morphology has changed in that they tend to form round groups with thymocytes instead of being radially arranged. In EfnB2tec/tec mice, there are two subpopulations with a different level of Ly51 expression segregated into two regions: one formed by several layers of Ly51hi TECs next to the medulla and the other one containing Ly51lo TECs in the outer cortex (Fig. 7). These results indicate that thymocytes must express both ephrin-B1 and -B2, and TECs express ephrin-B1 for the normal differentiation and/or expansion of these cTEC subpopulations. The similarities observed among EfnB1thy/thy, EfnB2thy/thy, and EfnB1tec/tec mice, but with the latter showing no important alterations in T cell development, suggest that the phenotypes may be due to difficulties in T-TEC cross talk rather than in the thymocyte depletion. This idea also supports T-TEC cross talk, implying direct cell–cell contacts dependent on Eph–ephrin-mediated signaling rather than on soluble molecules. In addition, ephrin-B2 expressed by TECs is necessary for topological organization of the distinct cTEC subpopulations, once they have differentiated. Taking into account that TECs also express the ephrin-B2–specific receptors, Eph B4 and Eph B6 (data not shown), it is worth speculating that the regulated expression of these molecules could contribute to defining more specific intracortical niches.
The cortical phenotype of ephrin-B–conditioned deficient mutant thymuses could be due to alterations in TEC generation, maintenance, or both processes. In wt mice, our results reveal that during ontogeny subpopulations of TECs with increasing Ly51 expression arise sequentially, according to the following pattern: a pool of MTS20+ cells expressing a particular level of Ly51 would give rise to MTS20+ cells with higher Ly51 expression, some of which will later mature, losing MTS20 expression (Fig. 6). This model suggests that MTS20 expression is associated with an early, presumably proliferating stage within each Ly51-expressing subset rather than to cortical early TEC stage. In contrast, these data contain an apparent contradiction: whereas in adult thymus, Ly51hi cTECs harbor DN thymocytes (28), in the early fetal thymus, cTECs are Ly51lo. This suggests that we cannot compare fetal and adult epithelium and that after their generation, the cTEC subpopulations keep developing, and as part of this maturation process, they would acquire higher Ly51 levels.
The analysis of the development of EfnB1tec/tecEfnB2tec/tec thymuses revealed a delay in this sequential appearance of the cTEC subpopulations, not observing changes in the MTS20+ cTECs proportions within each Ly51+ subset in comparison with wt values (data not shown). These results indicate that ephrin-B1 or/and ephrin-B2 expression on TECs is necessary for correct appearance of the different cortical epithelial cell subsets during ontogeny but not for their maturation marked by downregulation of MTS20 expression.
In contrast, most studied mutants (EfnB1thy/thy, EfnB2thy/thy, EfnB1tec/tec, and EfnB1tec/tecEfnB2tec/tec) show increased proportions of K5+K8+ TECs, a part of which are Ly51hiK5+ cells. In fact, our immunohistochemical analysis demonstrates that most K5+K8+ TECs of the corticomedullary junction are Ly51hi cells. In agreement, Gray and colleagues (28) described that ∼25% of isolated Ly51+ TECs were K5+K8+. It can be concluded that the K5+K8+ cell population adjacent to the MTS10+ medullary region, classically considered to contain adult common epithelial cell progenitors, could be part of the cortical compartment. Moreover, it has been previously proposed that this cell population is, or largely contains, cortical-specific progenitors based on the fact that thymuses of Krm−/−, a negative regulator of the Wnt canonic signaling pathway, knockout mice show increased proportions of K5+K8+ cells that accumulate in the thymic cortex (53), and the overexpression of DKK1, an antagonist of Wnt pathway, in the thymic epithelium of young mice results in a rapid thymic degeneration that courses with an almost total loss of K5+K8+ cells and a hypoplasic thymic cortex (54). Although we cannot rule out the existence of potential common epithelial progenitors and/or committed medullary epithelial precursors within the K5+K8+ cell compartment of the corticomedullary junction, our results suggest that they could be a minority population in adult thymus.
According to these results, ephrin-B1 and ephrin-B2 expressed on thymic epithelium seem to be necessary for a correct temporary appearance of the cTEC subpopulations but not for their maturation through downregulation of MTS20 expression. We could speculate that a similar process of an altered generation of cTEC subsets could occur during postnatal life in these mice, which could explain the increased numbers of K5+K8+Ly51hiMTS10− TECs observed later in the adult mutant thymuses.
The authors have no financial conflicts of interest.
We thank Dr. D.J. Anderson (California Institute of Technology, Pasadena, CA) for the donation of C57/Bl6-EfnB1LoxP/LoxP mice and Dr. J.J. Takeda and the Center for Animal Resources and Development of the Kumamoto University (Kumamoto, Japan) for the gift of C57/Bl6-Tg(K5-Cre)Jt strain. We also thank the Cytometry and Fluorescence Microscopy Center at Complutense University of Madrid for the use of the facilities and technical assistance, the Developmental Studies Hybridoma Bank of Iowa University for supplying the anti-K8 keratin Ab, and Dr. Richard Boyd (Monash University, Melbourne, VIC, Australia) for the donation of the MTS10 Ab.
This work was supported by Grants BFU 2007-65520 and BFU2010-18250 from the Spanish Ministry of Science and Innovation, RD06/0010/0003 from the Spanish Ministry of Health, Social Services and Equality, and S-BIO/0204/2006 from the Regional Government of Madrid.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- 7-aminoactinomycin D
- cortical thymic epithelial cell
- Delta-like ligand 4
- embryonic day
- epithelial cell adhesion molecule
- erythropoietin-producing hepatocyte kinase
- erythropoietin-producing hepatocyte kinase receptor–interacting protein
- MHC II
- MHC class II
- medullary thymic epithelial cell
- mouse thymic stroma
- postnatal day
- temperature of annealing
- thymic epithelial cell
- thymocyte–thymic epithelial cell
- Ulex europaeus agglutinin-1
- Received July 12, 2012.
- Accepted January 8, 2013.
- Copyright © 2013 by The American Association of Immunologists, Inc.