Harwood Academic Publishers imprint, part of the Gordon and Breach Publishing Group Printed in Malaysia An Adult Thymic Stromal-Cell Suspension Model for In Vitro Positive Selection

Presented here is a cell-suspension model for positive selection using thymocytes from αβ-TCR (H-2Db-restricted) transgenic mice specific to the lymphocytic choriomeningitis virus (LCMV) on a nonselecting MHC background (H-2d or TAP-1 –/–), cocultured with freshly isolated adult thymus stromal cells of the selecting MHC type. The thymic stromal cells alone induced positive selection of functional CD4-CD8+ cells whose kinetics and efficiency were enhanced by nominal peptide. Fibroblasts expressing the selecting MHC alone did not induce positive selection; however, together with nonselecting stroma and nominal peptide, there was inefficient positive. These results suggest multiple signaling in positive selection with selection events able to occur on multiple-cell types. The ease with which this model can be manipulated should greatly facilitate the resolution of the mechanisms of positive selection in normal and pathological states.


INTRODUCTION
Engagement of the TCR is a pivotal step in thymocyte development, ultimately resulting in the survival (positive selection) or loss (negative selection) of developing T cells. The molecular and cellular interactions necessary for these selection events, however, are still poorly understood. MHC transfected fibroblast lines (Pawlowski et al., 1993) and cloned thymic epithelial-cell lines Vukmanovic et al., 1992) were able to induce positive *Corresponding author. selection after intrathymic injection. However, the possibility for multicellular involvement--fibroblasts providing the selecting MHC combined with appropriate thymic epithelial cells delivering inducing signals--has not been addressed.
In conjunction with TCR transgenic mice, fetal thymic organ culture (FTOC) has been a useful in vitro model for investigating the effects of peptide in thymic selection processes (Ashton-Rickardt et al., 1994;Hogquist et al., 1994;Sebzda et al., 1994).
Reaggregate FTOC have also been developed to include defined populations of T-cell precursors and embryonic stroma (Anderson et al., 1994;Merkenschlager and Fisher, 1994). However, in these FTOC models, embryonic thymi are the source of Tcell precursors and stroma, yet the nature of the embryonic thymus differs markedly from that of the adult . Hence, they may be inappropriate to study the mechanisms and cellular involvement in adult thymic positive selection and abnormalities therein that are manifest postnatally. Furthermore, the downregulation of RAG-1 expression in immature CD4+CD8 thymocytes (Brandle et al., 1992;Kouskoff et al., 1995), a common source of precursor cells for some current assays for positive selection, indicates that cells at this stage may have already received selection signals in vivo, prior to culture. We thus sought to develop an in vitro model for positive selection that uses adult stroma and requires minimal pretreatment of T-cell precursors yet eliminates the possibility that selection signals had already occurred in vivo.
This cell-suspension model is more easily manipulated and should enable a more detailed dissection of the thymic stromal elements involved in adult thymic selection and hence allow investigations into possible thymic selection detects present in autoimmune diseases.

An Adult Thymic Stromal-Cell-Suspension Model for In Vitro Positive Selection
The basic method is outlined in Figure 1. P14 transgenic mice expressing a TCR (V/38.1, Vce2) specific for the LCMV glycoprotein peptide (amino acid sequence 33-41; KAVYNFATM), restricted to H-2Db, were bred onto a nonselecting MHC (H-2d) background. Rag-1 expression is not downregulated in TCR transgenic CD4+CD8 thymocytes on a nonselecting MHC background, suggesting that positive selection signals had not occurred (Kouskoff et al., 1995). Day 17-18 gestation was optimal for target cells since we found this was the earliest age with significant levels of CD4+CD8 cells and minimal rearrangement of the endogenous Vce chain that would have occurred. Selecting (C57B 16; H2b) stromal cells were freshly purified with minimal enzyme treatment to reduce potential loss of relevant surface molecules. The stromal cells were comprised essentially of cortical thymic nurse cells (---30%) and other epithelial cells (--30%), fibroblasts (--5%), macrophages and dendritic cells (--20%), endothelial cells (--5%), and lymphocytes (lymphocyte:stromal cell ratio of--2:1). In these studies, multiple cell types were deliberately kept to more closely mimic the normal thymus. Current work involves separating individual cell types.
The transgenic thymocyte to adult stromal cell ratio was 5:1. A three-dimensional structure necessary to support the development of T cells was maintained by culturing the cell suspension as hanging drops. This also allows for maximum gaseous exchange and for the contact of T-cell precursors with stromal cells at the meniscus. Furthermore, as a cell suspension in 28-30/xl, the diffusion of cytokines important for cell viability and thymic selection is not hampered or too dilute. Although this system allows some threedimensional structure to be retained, the cells were still essentially a suspension at the end of the 5-day culture. The nominal LCMV peptide (amino acids 33-41; KAVYNFATM) was usually added once at the beginning of culture at concentrations ranging from 10 -4 to 10 -14 M. As peptides degrade in the presence of fetal calf serum (FCS), in some experiments, the LCMV peptides were also added on a daily basis at concentrations ranging from 10 -5 to 10 -14 M. The P14 transgenic precursors were distinguished from the stroma-associated lymphocytes by use of Ly 5 congenic mice (for H-2 b stroma) or by MHC class I haplotype (for H-2 k stroma).
De Novo Positive Selection of CD4-CD8 Cells When Cocultured with Selecting MHC Stroma The El7 transgenic (H-2d) cells remained at the CD4+CD8 stage by day 5 when cultured alone or with nonselecting MHC stroma (H-2k) in the absence Figure 2A  Essentially, the co-culture system involves gently digesting adult (3to 5-week-old) H-2b thymi with enzymes then enriching for stromal cells by zonal unit-gravity elutriation. These freshly prepared stromal cells were then co-cultured in hanging drops with El7 Tcell precursors from P14 TCR transgenic mice specific for the lymphocytic choriomeningitis virus (on a nonselecting background), over a period of 4 to 5 days in the absence or presence of nominal LCMV peptide. defined by the CD41CD81 phenotype (Swat et al., 1991) evident. Similarily, when cultured only with H-2 b lymphocytes, the transgenic cells remained at the double-positive stage, but there was substantial apop-tosis in the presence of 10 -5 M peptide, consistent with previous work showing that thymocytes can selftolerize . When cultured with H-2b-expressing fibroblasts (MC57), they also remained El7 transgenic cells were presorted for CD4+CD8 thymocytes to eliminate the possibility that immature CD8 intermediate cells at El7 were merely expanding in culture. These CD4+CD8 cells were cocultured with H-2 stroma in the absence of peptide as described and harvested on day 5. CD4 vs. CD8 profiles of recovered transgenic cells are shown. To illustrate that these cocultures can also support the differentiation of double-negative cells, CD4-CD8cells were cocultured with H-2 stroma and harvested on day 5. CD4 vs. CD8 profiles of recovered transgenic cells are shown. CD4+CD8+, suggesting that MHC alone cannot induce positive selection ( Figure 3). As a positive control, the LCMV-pulsed MC57 cells were able to activate mature T cells derived from the thymus or periphery of LCMV-transgenic H-2 b mice (data not shown).
When the transgenic thymocytes were cocultured with selecting (H-2b) stroma, however, significant levels of CD4-CD8 cells developed (--15% of total cells) by day 5 (Figure 2A). This compares favorably with the in vivo situation; in the H-2 b transgenic thymus (selecting MHC environment), approximately 18% of cells are single-positive for CD8. Given that peptides rescue positive selection in FTOC using lobes deficient in MHC class I/peptide complexes (Ashton-Rickardt et al., 1993;Hogquist et al., 1993;Sebzda et al., 1994), we examined the influence of the nominal LCMV peptide added at the onset of coculture. Positive selection was enhanced to approximately 40% in the presence of LCMV peptide at 10 -5 M ( Figure 2A). The reduced level of CD4+CD8 cells in the presence of nonselecting H-2 k stroma, relative to the other control cultures, is unlikely to be due to increased cell death, but rather to the presence of macrophages in the stromal preparation that we have shown avidly phagocytose presumably apoptotic CD4+CD8 cells within 48 to 72 hr (Izon et al., 1989).
Absolute transgenic cell numbers recovered from cocultures with selecting MHC thymic stroma demonstrated at least a tenfold increase relative to cocultures with nonselecting stroma in the absence of   peptide, and an approximately twentyfold increase in the presence of the nominal LCMV peptide (Table 1). Only approximately 1% of transgenic cells (predominantly CD4+CD8 cells) were recovered after 5 days in culture in the absence of selecting stroma, compared to 6-7% in the presence of selecting stroma. Clearly, a large amount of cell loss is evident regardless of the single specificity of the transgenic TCR, most likely due to the sensitivity of immature thymocytes in culture. Despite this, the efficiency of positive selection in this model is approximately 5-10%, which compares favorably with intrathymic injections of target thymocytes (Lundberg and Shortman, 1994). Compared to transgenic cells cocultured with nonselecting stroma, there was a 200-fold increase in CD4-CD8 cells recovered when transgenic cells were cocultured with selecting stroma. A similar increase was evident in the presence of 10 -5 M nominal LCMV peptide; many transgenic cells would have been initially deleted due to the high concentration of peptide. In the presence of 10 -7 M LCMV, there was a 300-fold increase in CD4-CD8 cells, and at 10 -l M LCMV, there was 500-fold increase ( Table 2). Upregulation of the TCR as a consequence of positive selection is clearly demonstrated in Figure 4. Prior to culture, the El7 transgenic thymocytes contained virtually no TCR hi CD4-CD8 cells, as shown by their Vc2 expression (Figure 4). Identical profiles were obtained for V/38 expression (data not shown). The El7 CD4+CD8 cells and CD4-CD8 intermediates in particular showed low TCR expression, in contrast to transgenic CD4-CD8 cells induced when cocultured with selecting MHC stroma in the presence or absence of the nominal LCMV peptide (Figure 4).

CD4-CD8 Cells Developed De Novo from Presorted CD4+CD8 Cells
To further establish that the CD4-CD8 cells developed de novo and were not due to proliferation of preexisting mature cells, the El7 transgenic cells (H-2d) were presorted into CD4+CD8 and CD4-CDSpopulations prior to coculture with H-2 b stroma. In both cases, significant levels (16-23%) of CD4-CD8 cells developed by day 4 of culture even in the absence of nominal peptide ( Figure 2B). We also examined the phenotypic and functional status of the minor subset of El7 CD4-CD8 cells (4-8%) present in the nonselecting MHC thymus. In accordance with their low TCR status, these cells were immature intermediates since they were uniformly TSA-1 hi  and HSA hi ( Figure 5) and rapidly lost in culturemby 48 hr, 94% of the cells were CD4+CD8 (Figure 6). This clearly demonstrates that CD4-CD8 cells do not originate from preexisting single-positive cells and that the co-Tg+H-2b stroma Tg+H-2b stroma + 10-7M LCMV 71% 82% Vc 2 expression of CD4-CD8 + cells culture system can support positive selection from double-negative cells presumably via CD4+CD8 intermediates since these develop within 24 to 48 hr when the CD4-CD8cells are cultured alone (data not shown). Finally, if the CD4-CD8 cells at the end of culture are simply preexisting mature cells, they would be expected to preferentially survive in cultures of transgenic thymocytes alone and even more so to proliferate in the presence of peptide-pulsed antigen-presenting cells. In neither case was this observed ( with irradiated H-2 b splenocytes as a source of antigen-presenting cells and 10 -5 or 10 -8 M LCMV peptide, conditions that vigorously stimulate thymocytes and lymph nodes from LCMV transgenic H-2 b mice (data not shown).
As an alternative source of transgenic thymocytes that have not received prior positive selection signals in vivo, we made use of P14 TCR transgenic mice on a TAP-1-deficient (H-2b) background, where development of CD4-8 but not CD+CD8 cells is arrested (Ashton-Rickardt et al., 1993). Neonatal transgenic thymocytes from these mice were further depleted of CD4-CD8 cells by sorting for CD4+T cells (leaving only CD4+8 and CD4+8 cells) prior to coculturing with freshly purified thymic stroma from H-2 b or H-2 k mice. By day 4, TCR hi CD4-CD8 cells were evident, but only with the H-2 b thymic stroma, clearly demonstrating de novo positive selection in the CD4+CD8 precursors, driven by endogenous or FCSderived peptide. In these cultures, CD4-CD8 cell numbers increased from 0 to 2-3 104 cells per 10 6 original transgenic cells by day 4 (data not shown).

Positively Selected CD8 Cells Are Functionally Mature
The positively selected CD4-CD8 cells were functionally mature both in terms of proliferation and cytotoxicity. Transgenic El7 thymocytes (H-2d) were cultured alone or with selecting (H-2b) or nonselecting (H-2 d or H-2k) thymic stroma in the absence or presence of 10-7M nominal LCMV peptide for 5 days. The stroma had been irradiated to prevent any contribution by the stroma-associated lymphocytes. After the primary coculture, the cells were harvested and further cultured for 2 days with fresh irradiated H-2 b spleen cells as a source of antigen-presenting cells that had been pulsed with LCMV peptide. Transgenic cells when initially cultured alone or with nonselecting stroma for 5 days, in the absence and presence of specific peptide, showed no proliferation when challenged with fresh antigen-presenting cells loaded with LCMV peptide (10-6M), nor were they able to specifically lyse LCMV-(10-SM) loaded EL4 target cells. Transgenic cells cultured with selecting stroma in the absence and presence of peptide for 5 days, however, proliferated in response to LCMV-(10-6M) loaded antigen presenting cells (Figure 7). Phenotypically, these cells remained CD4-CD8 hi and were able to specifically lyse LCMV-(10-SM) loaded EL4 target cells (Table 3).

CD69 Expression
The early T-cell activation marker, CD69, is transiently expressed during thymocyte selection, both positive and negative (Swat et al., 1993;Brandle et al., 1994) LCMV peptide, CD69 increased from <5% at day 0, to 20-25% by day 2, reaching a maximum of approximately 50% at day 4. With 10 -5 M LCMV peptide, there was a biphasic expression of CD69 (data not shown); the initial increase (day 1) coincided with the negative selection of transgenic cells with high concentration of peptide, consistent with recent findings that CD69 is also expressed on the surface of apoptosing cells (Kersh and Hedrick, 1995;Kishimoto et al., 1995). The second increase in CD69 (day 3) would be a consequence of positive selection. CD5 is another cell-surface molecule that is upregulated during positive selection (Fowlkes et al., 1985;Lanier et al., 1986). We also found CD5 was upregulated on the selected single-positive CD8 cells in the co-cultures (data not shown), providing further evidence for de novo positive selection.

Positive Selection Can Involve Multiple-Cell Types
Positive selection was not induced in the presence of nonselecting MHC stroma. To determine whether MHC alone can positively select, El7 transgenic thymocytes were cocultured with H-2D b expressing MC57 cells for 5 days. No positive selection was evident in cocultures in the absence or presence of the LCMV peptide. However, when nonselecting stroma was included with 10 -5 M LCMV peptide and MC57 cells, 11% of total transgenic cells were of the CD4-CD8 phenotype at day 5 ( Figure 3).

DISCUSSION
To date, thymocyte selection has been primarily studied in the embryonic thymus, but how well this mimics that which occurs in the adult thymus is questionable. Here we present a cell-suspension model for in vitro positive selection that utilizes adult stroma, and hence will provide a good comparison with current fetal thymic models. The ease with which this system can be manipulated should enable a detailed dissection of the basis to positive selection, particularly the early kinetics and gene-activation events. Furthermore, by using adult stroma, it should allow investigations into the aetiological significance of thymic stromal-cell disorders that are manifest in the postembryonic period. Essentially, this system involves coculturing transgenic thymocyte precursors of known TCR specificity, but in a nonselecting environment, with purified adult stroma of the selecting and nonselecting MHC type. This system has the added advantage of being able to separate out distinct stromal-cell subsets, a technique that we are currently investigating. To avoid the possibility that positive selection signals had occurred prior to culture but were not yet apparent, and to minimize the preculture handling of precursor T cells, we used El7 thymocytes from TCR transgenic mice specific to LCMV/H-2b, which were maintained on a nonselecting MHC (H-2) or a TAP-1-deficient background. This is the first stage where CD4+CD8 cells are evident; these and a low percentage of CD3-CD4-CD8 were TCR-/l, HSAhi, and TSA-1 hi, confirming their immature status. After culture alone or in the presence of H-2 bbearing nonthymic epithelial cells, the latter cells developed to CD4+CD8 cells and there was no progression of any cells to the mature CD4-CD8 cells, confirming that these are a consequence of de novo positive selection in the presence of thymic epithelium. Selecting stromal cells (H-2b) were Proliferation 30000 25000 20000 15000 10000 5000 :::::::::::::::::::::::::::::::::::::::::::: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Tg alone Tg+ H-2b Str Tg+ H-2 b Str +10-7M  apositively selected CD4-CD8 cells from cocultures containing selecting H-2 stroma, both in the absence and presence of the nominal LCMV peptide (10-7 M), were able to specifically lyse target cells that were loaded with the same nominal LCMV peptide at a higher concentration (10 -5 M). Transgenic cells cultured alone under the same conditions were unable to specifically lyse LCMV-loaded target cells.
freshly purified to reduce the likelihood that important surface molecules had been downregulated in culture.
It has been argued that MHC alone is all that is required to induce positive selection, with the main body of evidence being from intrathymic injections of cell lines expressing the selecting MHC into a nonselecting thymus Vukmanovic et al., 1992). These experiments, however, did not exclude the possibility that accessory signals that may be required for positive selection were received from the nonselecting thymic epithelial cells. We found that positive selectio/a was only observed with thymic stromal cells and not when transgenic precursors were cocultured with H-2b-expressing thymocytes, irradiated spleen cells, or fibroblasts (MC57) in the presence or absence of peptide. Positive selection was evident by the downregulation of the CD4 coreceptor, upregulation of TCR, and upregulation of CD69 and CD5. Similarly, Anderson and colleagues have found in reaggregate thymic organ cultures that thymic epithelium was essential for positive selection (Anderson et al., 1996). When the precursors were cocultured with H-2b-expressing fibroblasts and thymic stroma from a nonselecting MHC background in the presence of peptide, we did observe some positive selection, albeit inefficient. Our data are consistent with the possibility that TCR-peptide/MHC interactions, an essential requirement for positive selection, work in conjunction with other stromal accessory signals to induce efficient complete positive selection. Certainly, although positive selection can be multicellular, it is more efficient if the correct MHC is coexpressed with appropriate differentiation molecules on dedicated thymic stroma, presumably epithelium.
The issue of peptide involvement in positive selection was briefly investigated using the nominal agonist LCMV peptide at various concentrations. Whereas positive selection did occur in the absence of specific nominal peptide, presumably due to the presence of endogenous peptide, it was enhanced in the presence of the LCMV peptide, as indicated by the increase in CD4-CD8 cell numbers recovered after culture. This was most evident in cocultures containing 10 -l M LCMV. In cocultures containing higher concentrations of peptide such as 10 -5 M, doublepositive cells initially apoptosed, evident in kinetics studies (data not shown). By day 1, however, the double-negative cells that had converted to doublepositive cells were then likely to have been selected at a concentration of peptide somewhat lower than the original level, due to degradation of the peptide in the presence of FCS. The CD4-CD8 cells selected either on the endogenous peptide or in the presence of the nominal LCMV peptide were functionally mature, both in their ability to proliferate in response to LCMV-peptide-loaded antigen-presenting cells and in their capacity to specifically lyse LCMV-loaded EL4 target cells. This is significant in the light of recent publications suggesting that only antagonist peptides can positively select to produce phenotypically and functionally mature CD8 T cells.
In summary, this in vitro model demonstrates all the salient features of positive selection, and highlights the importance of peptide in the efficiency of selection. This system should enable a more precise dissection of the stromal cells and the role of naturally occurring self-peptides and other molecules involved in the thymic selection processes, and allow early activation events to be monitored. Furthermore, the ease with which this model can be manipulated should greatly facilitate resolution of the aetiological significance of thymic stromal cells, particularly epithelium, abnormalities that are manifest postnatally.

MATERIALS AND METHODS
Mice P14 TCR transgenic mice, H-2Db-restricted, were bred onto a nonselecting H-2 d background and maintained at the Monash University animal house.
Ly 5 congenic C57B16 mice and CBA mice were maintained at the Monash University animal house. The P14 TCR transgenic mice on a TAP-l-deficient background were a kind gift from S. Tonegawa, the Massachusetts Institute of Technology.

Peptides
The nominal LCMV agonist peptide (amino acids 33-41; KAVYNFATM), the original cysteine at anchor position 41 in the wildtype LCMV peptide having been replaced by methionine to prevent dimer formation, were a kind gift from Hanspeter Pircher (University of Freiburg).

T-Cell Precursors
Transgenic embryonic thymi (E 17) from P14 a/3 TCR transgenic mice (specific to the lymphocytic choriomeningitis virus; H-2b-restricted) on a H-2 d (Pircher et al., 1989) or Tap-l-deficient background were the source of precursor T cells. Cell suspensions were prepared at a concentration of 3.4 10 6 cells/ml. To remove any CD8 single-positive cells, thymocytes were presorted with CD4-magnetic beads (MACS Magnetic Microbeads, Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) using a VarioMACS (VarioMACS, Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). Enriched cells were mostly of the CD4+CD8 phenotype with a small population of CD4+CD8 intermediate cells. In some experiments, pre-sorted CD4-CD8or CD4-CD8 int cell populations were also used as target cells.

Co-culture
Transgenic El7 thymocytes were mixed with the stromal cells at a ratio of 5:1 and co-cultured as hanging drops in inverted Terasaki plates at 37C, 5% CO2 generally over a 5-day period. Ly5.2 expression or MHC class I (S17.71 anti H-2k) distinguished P14 transgenic thymocytes from stromal-associated lymphocytes. Postculture analysis involved staining cells with Ly 5.2-FITC (PharMingen, San Diego), CDS-Biotin/Tricolour (PharMingen), and CD4-PE (PharMingen), or CD4 APC and Va2 PE (PharMingen). Dead cells were gated out by a lymphocyte viability gate or by using propidium iodide where possible. Cells were analyzed by flow cytometry using a FACScan (Becton-Dickinson, Mountain View, CA). Cell population gates were determined using the nontransgenic stromal-associated lymphocyte component of the coculture. supported by grants from the National Health and Medical Research Council of Australia.

Functional Assays
Cocultures of P14 H-2 d transgenic thymocytes and freshly purified H-2 b stroma were prepared and analyzed as described before with the addition that purified stroma was irradiated at 3000 fads prior to co-culture to prevent proliferation of stroma-associated lymphocytes. Cells were harvested at day 5 and placed into round-bottom wells of a 96-well plate (5 l04 cells/well) with IL-2 (25 units) and 5 105 irradiated (3000 rads) H-2 b splenocytes that had been pulsed with 10 -6 M LCMV peptide, for 48 hr. For the proliferation assay, (3H) thymidine was added for 16 hours and incorporated radioactivity measured on a/3 counter. For the cytotoxicity assay, target cells (EL4 cell line) were incubated with 51 Cr, with or without 10 -5 M LCMV peptide, for 45 min. Effector cells were then incubated with target cells for 4 hr in tripling dilutions, the supernatant removed, and chromium release measured on a 3/counter. Percentage-specific lysis of target cells was calculated by: cpm test cpm minimum release 100 cpm maximum release cpm minimum release Statistical Analysis Viable cells were counted under a fluorescence microscope using ethidium bromide/acridine orange and standardized as the number of recovered cells per 106 original El7 put into culture at day 0.