Heterogeneity of thymic stromal cells and thymocyte differentiation: a cell culture approach

Functions of the immune system are regulated by a coordinate series of events including a cascade of differentiation and proliferation phases in which lymphoid cell precursors acquire the ability to recognize and react to antigenic stimuli. For T lymphocytes, many of these events occur in the thymus, which is the major site of T-cell development. The thymic microenvironment provides soluble factors (i.e. cytokines, thymic hormones and growth factors) known to be involved in T-cell proliferation and differentiation (reviewed by Goldstein, 1984 and by Shortman, 1992), MHC Class I and II structures needed to shape the T-cell repertoire and extracellular matrix (ECM) components, which contribute to development-related processes by mediating direct cell to cell interactions. Such a functionally complex microenvironment frames the intrinsic processes of lymphoid cell maturation and the acquisition of antigenic reactivity (i.e. positive or negative selection of T-cells), which is mainly driven by the T cell antigen receptor (TCR) and accessory molecule engagement.

The complex lympho-stromal structure of the thymus and the close association of thymocytes with stromal cells high-light a role for specific intrathymic cell to cell interactions in T-cell development. For this reason in recent years several researchers have established different experimental models of thymic cell culture, which have allowed them to obtain a number of items of information on the mechanisms involved in T-cell proliferation, differentiation and selection.

After summarizing the main phases of thymic ontogeny of both lymphoid and stromal cell compartments, we will briefly review recent evidence on the role of the heteroge-neous non-lymphoid component of the thymus in T-cell development and selection in vitro.

The epithelial thymic anlage originates from the fusion of the endoderm of the third pharyngeal pouch and the ectoderm of the third branchial cleft (reviewed by Lobach and Haynes, 1987). Cells derived from the corresponding region of the neural crest also participate in this fusion (Le Lièvre and Le Douarin, 1975). Furthermore, bone marrow-derived stromal cells participate in the final organization of the thymus (Jordan et al., 1979; Kraal et al., 1990). Such a het-erogeneity in ontogeny accounts for the heterogeneous composition of the thymic stroma, in fact, fibroblasts, macrophages, dendritic and different types of epithelial cells can all be found in the adult thymus.

Human and murine thymic epithelial cells have been the subject of several studies in vivo that demonstrated their morphological and phenotypical heterogeneity (Van de Wijngaert et al., 1984; Rouse et al., 1988). These studies also showed a specific correlation between morphology, phenotype and localization of the different epithelial cell types (Van Vliet et al., 1985). All these observations resulted in the definition, by monoclonal antibodies (mAbs) of at least five types of Clusters of Thymic Epithelial Staining (CTES) patterns (Kampinga et al., 1989).

With respect to the architectural organization, it has been shown previously that, although T-cell progenitors enter the murine thymus by day 11, the organ is not divided into cortex and medulla regions before the 13th day of gestation (Van Vliet et al., 1985). Moreover, at this time the initial expression of MHC class II antigens is observed. These are initially localized in the stromal cells of the medulla, and become diffuse by day 16 (Jenkinson et al., 1981). In contrast class I MHC antigens on stromal thymic cells are not expressed before day 16 (Jenkinson et al., 1981). Interestingly, it has been shown that the onset of proliferation in fetal thymus occurs on day 13 (Van Vliet et al., 1985), when cortex and medulla are distinct and MHC class II expression begins.

Stromal cells of hematogenous origin appear later: macrophages are not found before day 14, while dendritic interdigitating cells appear even later (Kraal et al., 1990). This progressive arrangement of the thymic non-lymphoid architecture suggests that thymic stroma adjusts itself to different functions during ontogeny. In addition to providing a microenvironment suitable for homing and sustaining precursors and progressively maturing lymphoid cells, thymic stromal cells have also been suggested to play a specific functional role in positive or negative selection of thymocytes. The cortical thymic epithelial cells are the main cell type believed to mediate positive selection, whereas medullary macrophages, dendritic cells and medullary epithelial cells would appear to be responsible for negative selection, which refers to the induction of tolerance by the clonal elimination of autoreactive cells (reviewed by Ferrick et al., 1989). The best model for investigating the role of different stromal cells in the establishment of tolerance has been provided by recent in vivo studies, which utilized MHC transgenic mice (reviewed by Benoist and Mathis, 1989) and bone marrow chimeras (reviewed by Sprent et al., 1988).

Contribution of neural crest migrants to thymic development has been reported to be essential. Cell elements migrate from the cranial portion (anterior to the fifth somite) of the neural crest to provide mesenchyme around the epithelial primordia of the thymus (Le Lièvre and Le Douarin, 1975). Interaction between the neural crestderived mesenchyme and epithelium is necessary for the organ morphogenesis (Auerbach, 1960). In fact, ablation of the chicken embryo cephalic neural crest results in thymic aplasia/hypoplasia (Bockman and Kirby, 1984). The further fate and functions played by neural crest-derived cells in thymic stroma and in the series of events throughout the thymocyte differentiation pathway remain unclear. Nakamura and Ayer-Le Lièvre (1986) suggested that thymic myoid cells, which are predominantly situated in the medulla and involved in the autoimmune disease myasthenia gravis, are derived from neural crest cells. Expression of a number of antigens found in neural and neuroendocrine cells (such as the ganglioside A2B5, the secretory granuleassociated chromogranin-A, neuron-specific enolase, the surface glycoproteins HISL-5, -9, -14 and ID-4) has been described in thymic epithelium (reviewed by Geenen et al., 1989). In particular, the lymphoepithelial complexes, known as thymic nurse cells express most of these neuroendocrine markers, and also synthesize several neuropeptides (oxytocin, vasopressin and neurophysin) (Geenen et al., 1986). Somatostatin, neurotensin, Met-Enkephalin, neuropeptide-Y, the pre-protachykinin-A-derived substance P and neurokinin-A have also been described in thymic epithelium (Fuller and Verity, 1989; Ericson et al., 1990). The ability of thymic stromal cells to produce neuropeptides, together with the presence of receptors for neurotransmitters and neuropeptides in T cells and their ability to respond to the latter substances (Richman and Arnason, 1979; Singh et al., 1979; Calvo et al., 1992), suggest that the neural-derived component of the thymus plays a role in T-cell development.

In recent years, the combination of different approaches (study of fetal ontogeny, establishment of lines of transgenic mice and its complementary ‘gene knockout’ technique) for the study of intrathymic T-cell differentiation has allowed us to gain insights into the major pathways of thymocyte differentiation (Petrie et al., 1990; reviewed by Von Boehmer, 1990). Moreover, the wide production of monoclonal antibodies and antisera specific for both the lymphocyte and stromal cell surface antigens has facilitated the analysis of developmentally related stages of thymocyte differentiation and the characterization of stromal cells directly associated with differentiating thymocytes.

Most researchers agree that hemopoietic-(fetal liveror bone marrow-) derived precursors of mature T cells colonize the murine thymic anlage at about day 11 of fetal life and continue to migrate to it throughout adult life (Owen and Ritter, 1969).

Penit and Vasseur (1988) showed that adult pro-T cells initially are localized to the thymic cortico-medullary junction or to the subcapsule as CD4⁻ CD8⁻ (double negative, DN) thymocytes. In the adult thymus these DN cells, which display the highest mitotic rate, accumulate as DN, CD3⁻ thymocytes in the subcapsular region, where the thymic stroma is characterized mainly by baskets of elongated epithelial cells and thymic nurse cells (Brekelmans and Van Ewjik, 1990). The subcapsular epithelial cells express antigens detected by mAbs of type I and II of CTES (Brekelmans and Van Ewjik, 1990).

In the murine fetal thymus the first CD3⁺ thymocytes are detected on day 14 of development (Bluestone et al., 1987). Moreover, on day 14 to 15 of gestation the first TCR-γδexpressing thymocytes can be observed (Hawran and Allison, 1988). TCR-αβ⁺ thymocytes are not found in the thymus until day 17 of gestation (Roehm et al., 1984).

The expression of CD4 and CD8 is first observed on day 15 to 16 of development (Ceredig et al., 1983). It has also been observed that there is a short period (on day 16) in which only CD8 is expressed (Ceredig et al., 1983). By day 17 all thymocytes coexpress CD4 and CD8. In the adult thymus, this CD4⁺,CD8⁺ population (double positive, DP), which expresses little or no CD3, is located in the cortex. These cells are in close association with epithelial cells characterized by long, branching cytoplasmic processes (spider-shaped) defined by mAbs of type I and III of CTES (Brekelmans and Van Ewjik, 1990).

Finally the mature CD4⁺,CD8⁻ and CD4⁻,CD8⁺, both CD3⁺ and TCR-αβ⁺, (single positive, SP) thymocytes appear in murine fetal thymus on day 18 and 19, respectively. In the adult thymus SP thymocytes are located in the medulla, which is characterized by a loose network of round epithelial cells (positive for type I, II and IV of CTES) and dendritic interdigitating cells. This latter cell type is mainly localized to the cortico-medullary junction (Duyvestyn et al., 1982).

These observations, together with those of Penit (1986) on the proliferative status and localization of thymocyte subsets, support the proposals of many authors that thymocyte differentiation proceeds from immature subcapsular DN CD3⁻ thymocytes, through the intermediate stages of cortical DP thymocytes with increasing expression of CD3 to medullary mature SP, CD3^high, TCRαβ⁺ thymocytes. Even though this is supposed to be the main pathway of thymocyte differentiation, more recently, different additional and/or alternative pathways, both intraand extra-thymic, of T-cell differentiation, have been reported (Matsumoto et al., 1991; Hattori et al., 1989; Rocha et al., 1992; and reviewed by Shortman, 1992). The pathways of thymocyte differentiation overlap an intrathymic selection process in which clonal cell subsets are either deleted (by apoptosis) or rescued. A central mechanism in determining this negative versus positive selection process, and in defining the fate of differentiating thymocytes and their extrathymic functions, involves variable expression and affinity of thymocyte TCR(s) for MHC class I and II structures and different antigens presented by thymic stromal cells, as well as the involvement of the accessory molecules CD4 and CD8 (reviewed by Boyd and Hugo, 1991; by Von Boehmer, 1992; by Pardoll and Carrera, 1992).

Although major advances have been made in the understanding of both cellular and molecular aspects of thymocyte differentiation with respect to the role of thymocyte subsets, little is known about their relationships with thymic stromal components. The distribution of different lymphoid and stromal cells within the thymus is suggestive of specific relationships between different subsets of either cell compartments, which may be necessary for the various stages of T-cell differentiation.

The first in vitro approach to the study of T-cell/thymic stroma interaction and its relevance to T-cell differentiation has been provided by the use of fetal thymus organ cultures (Jenkinson et al., 1982; Born et al., 1987; Ceredig, 1988). These, while representing a good model to analyze the stages of T-cell development in vitro, like the intact thymus, constitute a black box in which it is not possible to distinguish the individual roles of different stromal cells. More recently, advances in cell culture technologies have allowed the separation of different thymic stromal components. This has allowed a better identification and characterization of the sources of soluble factors involved in thymocyte differentiation, and helped to point out a specific role for different stromal cells and the underlying mechanisms involved in sustaining T-cell proliferation, differentiation and/or selection.

In this regard, by isolating structures representing different types of direct cell to cell lympho-stromal interactions (thymocyte/macrophage rosettes; thymocyte/dendritic interdigitating cells rosettes; lymphoepithelial complexes, designated as thymic nurse cells) from fresh thymuses, Kyewski and coworkers (1982) showed that the thymocyte population associated with all three types of stromal cells were enriched in actively dividing cells. In a subsequent study these authors elegantly showed that one of the proliferation-activating stimuli was the consequence of antigen presentation (Kyewski et al., 1984), since they observed that I-A⁺dendritic interdigitating cells were capable of presenting antigen to thymocytes in vitro.

Cell culture techniques have now been established for each thymic stromal cell type. Moreover co-culture assays have been performed with all stromal cells and lymphoid precursors or thymocytes at different stages of differentiation (Palacios et al., 1989; Nishimura et al. 1990; Tatsumi et al., 1990; Sen-Majundar et al., 1992). Data generated from co-culture studies are often conflicting, suggesting the complexity of the thymocyte differentiation steps, in keeping with the multiple pathways of T-cell intrathymic and extrathymic development (reviewed by Shortman, 1992).

By using the in vitro approach, several studies have suggested that the interactions between different stromal cells and thymocytes are sustained by an abundance of adhesion molecules present (e.g. CD2/LFA-3 (Denning et al., 1987), LFA-1/ICAM-1 (Fine and Kruisbeek, 1991), Mac-1 (Papiernik and El Rouby, 1988) and fibronectin and its receptors (Utsumi et al., 1991; Sawada et al., 1992)). This plethora of possible adhesion mechanisms involved in thymocyte/stromal thymic cell interactions is likely to provide coordinate and/or alternative mechanisms of cell to cell contact, sustaining stromal cell-driven T-cell differentiation. For instance, some recent results obtained by utilizing co-culture assays, suggest a critical role for ECM components in sustaining thymocyte differentiation, since blocking the interaction between fibronectin and its receptor(s) impaired the adhesion of thymocytes to the stromal cells and their subsequent differentiation (Utsumi et al., 1991; Sawada et al., 1992). We recently confirmed these data and also observed that the modulation of fibronectin production by thymic stromal cell lines may represent a means by which agents known to influence cell differentiation (e.g. retinoic acid) impair the in vitro thymocyte developmental program (our unpublished data).

Several reports have attempted to elucidate whether pure populations of individual thymic epithelial cell types are capable of sustaining a specific step in thymocyte subset differentiation, in addition to providing information on the soluble factors involved in proliferation and differentiation processes. In this regard, Gutierrez and Palacios (1991) demonstrated that phenotypically different epithelial cell lines obtained from newborn or fetal mouse thymus were involved in the differentiation of T-cell progenitors into distinct thymocyte subsets. The cell lines also displayed a distinct cell line-specific pattern of cytokine production. It was suggested that multiple types of epithelial thymic cells are required to sustain the development of the whole spectrum of thymocyte subsets.

Major contributions to the understanding of the role of soluble factors, produced by thymic epithelial cells, in the induction of proliferation and differentiation of immature thymocytes, have been provided by several authors. Both human and murine thymic cell lines constitutively produce cytokines that play a role in sustaining thymocyte or precursor cells proliferation and/or differentiation. These include IL-1α and β, IL-3, IL-6, IL-7, GM-CSF, TSTGF and LIF (Sano et al., 1989; Le et al., 1990a; Gutierrez and Palacios, 1991; Sakata et al., 1992; Screpanti et al., 1992). Cytokine production can be regulated by different growth factors, potentially acting upon stromal cells in an autocrine fashion, (e.g. EGF and TGFα) (Le et al., 1991) as well as by the triggering of adhesion-involved structures (e.g. LFA-3) resulting from an interaction with thymocytes (Le et al., 1990b). The production of IL6 and LIF is particularly intriguing. It has been shown that these cytokines belong to the neuropoietic cytokine family of factors (also including CNTF and oncostatin-M). All share the IL-6 signal transduction co-receptor gp130 (Ip et al., 1992) and are able to regulate both hemo-lymphopoietic maturation and growth, survival and differentiation of neurons (reviewed by Hall and Rao, 1992). Due to the heterogeneous ontogeny of thymic stromal cells, a neurotrophic microenvironment might be important for sustaining neural crest derivatives in thymic stroma. We recently proved this prediction by the identification of neural cells in several murine thymic stromal cell primary cultures. These cells carry a number of neuron-specific markers (neurofilament cytoskeletal proteins and synapsin I), in addition to other neuroendocrine cell markers (Screpanti et al., 1992). Interestingly, these thymic stromal cell lines provide a complex neurotrophic microenvironment, as they synthesize several members of both the trk-receptor-interacting neurotrophin (e.g. NGF and BDNF) and neuropoietic cytokine (e.g. IL-6 and LIF) families (Sano et al., 1989; Le et al., 1990a; Sakata et al., 1992; Screpanti et al., 1992; and our unpublished data). NGF and IL6 are able to enhance the neural phenotype of thymic stromal cells in an autocrine/paracrine way. Interestingly, NGF also enhanced IL6 gene expression, suggesting an intrathymic cross-talk between the two neurokine families (Screpanti et al., 1992).

With respect to the role of thymic stromal cells in sustaining thymocyte selection, Papiernik and coworkers (Papiernik et al., 1987; Papiernik and El Rouby, 1988), using a lympho/stromal cell co-culture system, demonstrated that bone marrow-derived thymic stromal cells cocultured with thymocytes were able to preferentially support the CD4 SP subset proliferation. The addition of anti-I-A or anti-I-E antibodies to the syngeneic co-culture blocked this proliferation. Furthermore, they observed inhibition of CD4 SP proliferation by antibodies directed against adhesion-involved structures, such as LFA-1 and Mac-1, showing that direct contact between lymphoid and stromal cells is required to enhance the proliferation and positive selection processes. Recent evidence that cultured thymic epithelial cells are also involved in positive selection is given by Vukmanovic et al. (1992), who showed that a thymic epithelial cell line, capable of antigen processing and presentation in vitro, can induce thymocyte positive selection in vivo after intrathymic injection.

With respect to the thymocyte negative selection process, the lympho-stromal co-culture model has been utilized by several groups in order to further understand the role of different thymic stromal cells. Kosaka et al. (1989) demonstrated that a murine thymic stromal cell line had the potential to eliminate an antigen-specific T-cell clone that was restricted by the class II MHC antigens of stromal cells in vitro. They also suggested that such a negative selection process was TCR/MHC engagement-dependent, as it could be prevented by anti-class II MHC or anti-CD3 antibodies. Furthermore, the same authors described a mechanism by which the immunosuppressant drug cyclosporin A could induce autoimmunity. They demonstrated that the addition of cyclosporin A to the lympho/stromal co-culture blocked the deletion of a T clone with a given specificity, that was otherwise induced by thymic stromal monolayer (Kosaka et al., 1990). This co-culture model has also been used to determine the minimum amount of peptide required for negative thymocyte selection by a thymic nurse cell line (Iwabuchi et al., 1992). In addition to TCR/MHC engagement, accessory adhesion molecules, such as the LFA1α/ICAM.1 interaction, participate in the negative selection process in vitro (Carlow et al., 1992). Thus, stromal cell-directed thymocyte differentiation in vitro allowed the observation that molecules which mediate cell to cell contact (i.e. LFA-1/ICAM-1 or TCR/MHC) are able to play a bifunctional role in either enhancing thymocyte development or inducing T-cell deletion. Likewise, the lympho/stromal culture provides an interesting model to dissect the events that determine either choice following the interaction of the thymocyte with the stromal cell.

The complex series of events determining the pathways of intrathymic T-cell generation are accompanied by an equally complex network of possible interactions in different thymic stromal cell populations. The development of appropriate thymic stromal cell cultures has recently provided major advances in the dissection into single steps of the overall process of thymocyte development. As summarized in this review, these events are mediated by several adhesion molecules, ECM components and their receptors, specific cytokines, growth factors and by a spectrum of MHC/TCR interactions, which lead to the thymocyte selection process.

The results generated by the cell culture system are often conflicting, as they only represent fragments of the overall differentiation process in which the roles played by the single components are imbalanced by the in vitro analysis of the single steps. More contradictory pictures are generated by the comparison of in vitro with in vivo models. For instance, the role of some soluble factors in thymic ontogeny (i.e. IL-2 and IL-4) is questioned by the normal thymocyte development observed in either IL-2 or IL-4 gene-deficient mice obtained by targeted recombination (Schorle et al., 1991; Kuhn et al., 1991). Likewise, although the results reviewed above suggest a main role for the expression of class II MHC by thymic stromal cells in thymic development, mice lacking MHC class II molecules do not have major modifications in thymocyte maturation, as one would expect, whereas they present significant alterations in splenic lymphocytes (Cosgrove et al., 1991). These surprising data obtained in transgenic mice might be accounted for by the redundancy of the events supporting thymocyte development, which provides alternative pathways and/or overlapping functions for those gene products. In this regard, the cell culture system might help to dissect the role of each molecular event. The study of the in vitro developmental pattern, using thymocyte precursors or thymic stroma from single gene disrupted mice, could help to elucidate this problem.

In conclusion the cell culture system, by providing the opportunity to directly manipulate both stromal cells and/or thymocytes at the functional and genetical levels, could shed light on the molecular mechanisms involved in specific cell to cell interactions in each stage of thymopoiesis and its alteration in disease. Particularly, by using thymic stromal cell culture, it is possible to address the T-cell maturation and/or education study at the level of the single clonal cell (thymocyte or stromal cell). This is of critical importance to allow the assembly of all the steps necessary to build up the puzzle of the intrathymic T-cell development.

We are grateful to Drs A. MacKay and R. Testi for helpful discussions and critical reading of the manuscript. This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC), by the AIDS program and by the Consiglio Nazionale delle Ricerche (CNR).