Cutting Edge: Three-Dimensional Architecture of the Thymus Is Required to Maintain Delta-Like Expression Necessary for Inducing T Cell Development¹

The thymus plays a central role in inducing and supporting the differentiation of hemopoietic progenitor cells (HPCs)³ into T cells. An important feature of the thymic microenvironment is its three-dimensional (3-D) organization, consisting of an ordered architecture of thymic stromal cells (TSCs), which may be epithelial or mesenchymal in origin, through which the developing thymocytes migrate and mature (1). This 3-D configuration maximizes the interaction of developing thymocytes with the supporting stromal cells, allowing intercellular cross-talk integral to the development of both T cells and TSCs (2). With this in mind, in vitro models for the generation of T cells were thought to require an intact 3-D thymic architecture, such as that maintained by fetal thymic organ cultures (FTOCs) (3) or reaggregate thymic organ cultures (RTOCs) (4). Both FTOCs and RTOCs support the generation of all T cell subsets. In contrast to RTOCs, in which TSCs and HPCs are assembled to form a 3-D structure, two-dimensional monolayers composed of freshly isolated TSCs or cell lines derived from TSCs were shown to be incapable of replicating a functional thymic microenvironment (5).

The requirement for a 3-D-supporting stroma appears to be unique to T cell development, as the in vitro development of other hemopoietic lineages, including B cells and NK cells, does not require a 3-D structure (6). However, this paradigm had to be revised with the advent of the OP9-DL1 coculture system (7, 8), in which T cell development progresses unimpeded on heterologous bone marrow stromal cell monolayers. Nevertheless, the reason behind the seemingly paradoxical finding that fresh ex vivo thymic stromal cell monolayer cultures (TSMCs) fail to support T lymphopoiesis remained unresolved.

To elucidate the mechanism behind the failure of TSMCs to support T cell development, we examined the expression levels of Notch ligands in TSCs in both monolayer culture and in FTOCs. Strikingly, expression of Delta-like (Dll) 1 and its closely related homologue, Dll4, were extinguished in TSMCs, while both transcripts were maintained in FTOCs and RTOCs. In contrast, Jagged 1 (Jag1) expression remained similar in both TSMCs and FTOCs, whereas Jag2 expression was down-regulated in TSMCs. Importantly, TSCs transduced to re-express Dll1 or Dll4 in monolayer culture regained the ability to support T lineage commitment and differentiation. This indicates that the loss of Dll expression (Dll1 and Dll4) by TSCs in monolayer culture is a causal mechanism behind the erosion of their ability to support T lymphopoiesis.

HPCs were isolated as previously described (7). Briefly, day 14–15 fetal livers obtained from timed-pregnant CD-1 mice (Charles River Breeding Laboratories) were depleted of mature cells using anti-CD24 mAb (J11d.2) plus complement (9), and viable cells were subsequently flow cytometrically sorted for CD117⁺Sca-1⁺ cells. NIH3T3 cells were provided by Dr. M. Julius (Sunnybrook & Women’s Research Institute). GP+E.86-GFP, GP+E.86-DL1, GP+E.86-DL4 viral packaging cell lines and OP9-DL1 cells were generated in our laboratory by T. M. Schmitt (7). All animal procedures were approved by Sunnybrook & Women’s Research Institute’s Animal Care Committee.

Fetal thymic lobes isolated on days 14 and15 of gestation were cultured as previously described (3, 10). FTOCs were either treated with 1.1 mM 2-deoxyguanosine (dG) (Sigma-Aldrich) or left untreated for 5 days (3). dG-FTOCs were reconstituted with 2000 Sca-1⁺CD117⁺ HPCs and incubated as previously described (10).

To establish RTOCs and TSMCs, dG-FTOCs were incubated with serum-free medium containing 0.05% trypsin and 2 mM EDTA for 20–25 min at 37°C. Lobes were then disaggregated into a single-cell suspension and filtered through 70-μm nylon mesh. For RTOCs,∼10⁶ TSCs were resuspended in 20 μl of FTOC medium and placed on FTOC rafts as previously described (10). For TSMCs, ∼60,000 TSCs were added to each well of a 96-well flat-bottom plate to form monolayers. After overnight incubation, TSMCs were retrovirally transduced by incubating the cultures with fresh retroviral supernatants from GP+E.86 packaging lines each day for 4 days, then cultured for 2 days in medium. Two hundred HPCs were added to each confluent well of TSMCs supplemented with 5 ng/ml Flt-3 ligand (PeproTech) and 1 ng/ml IL-7 (PeproTech).

Similar protocols were used to transduce NIH3T3 cells. After transduction, NIH3T3 cells were sorted for GFP expression. Before coculture, transduced NIH3T3 cells were gamma-irradiated (∼55 Gy) and plated at 40,000 cells/well in 96-well flat-bottom plates overnight. Cultures were seeded with 200 HPCs the next day as above. Hemopoietic cells were transferred to a fresh plate of gamma-irradiated NIH3T3 cells every 5 days and supplemented with Flt-3 ligand and IL-7 as above, with 25 ng/ml stem cell factor (PeproTech) included only in the first 5-day culture.

Total RNA was isolated from TSMCs and FTOCs using TRIzol (Invitrogen Life Technologies). cDNA was prepared using Superscript III (Invitrogen Life Technologies). Semiquantitative PCR was performed using serially diluted cDNA, normalized based on β-actin signal. PCR primers are as follows: β-actin (11); Jag1, Jag2, and Dll1 (12); keratin 8 (13) and keratin-5F (5′-CGACCCCACCATCCAGCG), keratin-5R (5′-TCTCCCCCAGAACCCCGT); Dll4F (5′-GGTGACCTGGCGAACAGACGAGCAAAAT) and Dll4R (5′-GGATCCGGGGAAGCTGGGTGGCAA); Foxn1F (5′-TTTGGCTTTGAGGAGGGC) and Foxn1R (5′-GTTGCAAGGGAGGCTGGTA). All PCR products were resolved by agarose gel electrophoresis and corresponded to the predicted molecular mass.

All mAbs were purchased from BD Biosciences: PE-conjugated Sca-1, CD8α, CD44, and allophycocyanin-conjugated CD117, CD4, CD25, and CD19. Cells were stained and sorted as previously described (9). All sorts yielded ≥99% cell purity, as determined by post-sort analysis.

To analyze the role of the thymic architecture in supporting T cell development, we first depleted FTOCs of endogenous thymocytes by treatment with dG (3). The resulting dG-FTOCs were either left intact or trypsinized and plated as a monolayer culture (TSMC). Both intact dG-FTOCs and TSMCs were cocultured with CD117⁺Sca-1⁺ HPCs obtained from day 14 fetal livers (Fig. 1). Flow cytometric analysis at days 7–13 of coculture showed that TSMCs failed to support differentiation of HPCs past the initial CD4⁻CD8⁻ (double-negative, DN) stage of T cell development (DN1, CD44⁺ CD25⁻). In contrast, intact FTOCs promoted T lymphopoiesis through all DN stages and yielded CD4⁺CD8⁺ double-positive (DP) and single-positive (SP) cells (Fig. 1). These results indicate that interactions between TSCs and HPCs necessary for the induction of T cell development are only maintained in FTOCs, but not in TSMCs.

To elucidate the intrinsic deficiency between TSMCs and FTOCs, we examined expression of characteristic thymic stromal genes under various culture conditions by RT-PCR. RNA samples were obtained from unmanipulated FTOCs, dG-FTOCs, and dG-FTOCs cultured for an additional 5 days either intact or dissociated as TSMCs. Fig. 2 A shows that expression of keratin-8 transcripts was similar under all conditions analyzed, demonstrating that the abundance of thymic epithelial cells (TECs) was equivalent in TSMC vs intact FTOCs. In contrast, keratin-5 expression was diminished in TSMCs compared with intact FTOCs. Since keratin-8 is normally expressed in all developing TECs, while keratin-5 is present only in more mature TECs (14), this suggests that the conditions used for TSMCs were not fully conducive to the normal developmental maturation of TECs.

Because OP9-DL1 cells support T lymphopoiesis in monolayer conditions, but native OP9 cocultures fail to do so, we focused our attention on expression of various Notch ligands in TSMCs. As revealed in Fig. 2 A, Jag1 expression in TSMCs remained similar to that in dG-FTOCs. In marked contrast, however, the expression level of Jag2 was dramatically reduced in TSMCs, as were Dll1 and Dll4 transcripts that were beyond detection level, while all three transcripts were detected in intact FTOCs. To pinpoint elements responsible for the loss of both Dll molecules, we cocultured TSMCs with day 16 fetal thymocytes, which would provide cross-talk stimulation known to be necessary for proper thymic organization during development (15). However, the presence or absence of thymocytes resulted in no difference in the expression levels of Dll1 or Dll4 in TSMCs (data not shown).

To address whether Dll expression by TSCs is critically dependent on a 3-D stromal structure, TSCs were used in RTOCs in the absence of thymocytes (Fig. 2,B). In contrast to TSMCs, TSCs placed in RTOCs maintain expression of both Dll1 and Dll4. Additionally, we also observed that the expression of Foxn1 (2) is down-regulated in TMSCs, but not in FTOCs or RTOCs (Fig. 2 B). Although there is no evidence for direct regulation of Dll transcription by Foxn1, mutations in foxn1 responsible for the nude phenotype in mice result in the loss of both Dll1 and Dll4 expression in the fetal thymic anlage (16).

We explored whether restoration of either Dll1 or Dll4 expression in TSMCs would be sufficient to re-establish T lymphopoietic function. To this end, TSMCs were retrovirally transduced to express either Dll1 or Dll4 plus GFP, or GFP alone as control. Following a transduction period of 4 and 6 days in normal medium, we consistently achieved 60–80% transduction efficiency, as indicated by GFP expression (Fig. 3,A). We next verified that Dll1 and Dll4 were expressed in the transduced TSMCs. As shown in Fig. 3,B, comparison of Dll expression levels in TSMCs relative to freshly dissociated TSCs reveals that TSMCs transduced to express Dll1 or Dll4 regained expression of each of the respective Notch ligands. In contrast, expression levels of Dll transcripts in TSMCs transduced with GFP alone were undetectable after 10 days of culture, consistent with our previous results (see Fig. 2).

To ascertain whether TSMCs with forced Dll1 or Dll4 expression would regain their ability to support T cell development, HPCs were cocultured with the transduced TSMCs. As expected, TSMC-GFP cocultures failed to generate a significant population of DN CD25⁺ T lineage cells, and DP cells were not present at later time points (Fig. 4,A). In striking contrast, >20% of hemopoietic cells expressed CD25 in TSMC-DL1 and TSMC-DL4 cocultures after 10 days. Furthermore, nearly 50% of the cells had advanced to the DP stage and ∼10% reached the CD8 SP stage (Fig. 4 A). These results demonstrate that forced expression of either Dll1 or Dll4 alone in TSMC cells is sufficient to confer the ability to induce and support T cell development. Moreover, at least one Dll ligand is necessary for T lymphopoiesis in TSMC culture, but each Dll ligand is functionally redundant and can compensate for the loss of the other. In addition, these findings indicate that the lack of T lymphopoietic potential in TSMCs is due to a loss of Dll expression following disruption of the 3-D stromal architecture. This suggests that 3-D interactions between TSCs are necessary to maintain normal Notch ligand expression in the thymus.

We also observed a reproducible and complete block in B lymphopoiesis from HPCs in TSMC-DL cocultures, as indicated by the absence of CD19⁺ cells (Fig. 4,B). This result is similar to that observed for OP9-DL1. In contrast, TSMC-GFP cocultures gave rise to B lineage cells, with ∼70% of the cells expressing CD19 (Fig. 4 B). These findings in TSMCs support the notion that forced expression of either Notch ligand, Dll1 or Dll4, is sufficient to direct the commitment of HPCs toward a T cell fate, at the expense of a B cell fate that would otherwise predominate.

To address whether expression of specific Dll family members is the only critical requirement for HPCs to generate T lineage cells in vitro, we investigated whether expression of Dll1 or Dll4 by NIH3T3 fibroblast cells is sufficient to support T cell development. To this end, retrovirally transduced NIH3T3 cells expressing Dll1, Dll4, or GFP alone were cocultured with HPCs, under similar conditions to TSMC, and analyzed on days 10 and 17 for evidence of lymphopoiesis. Notably, none of the NIH3T3 cocultures was capable of generating T lineage cells, suggesting that Dll-Notch signals alone are not sufficient to support T lymphopoiesis (Fig. 5). Importantly, although NIH3T3-GFP cells gave rise to B lineage cells under the conditions tested, NIH3T3-DL1 or NIH3T3-DL4 cells did not support the generation of B lineage cells from HPCs (Fig. 5). Thus, in addition to the requirement for appropriate Dll expression in stromal cells, T lymphopoiesis also requires critical interactions provided in TSMCs that are lacking in NIH3T3 cells. Thus, Dll-Notch signaling alone may be sufficient to suppress B lymphopoiesis without inducing T lymphopoiesis.

The development of the OP9-DL1 culture system has helped to re-imagine T cell development in vitro and overturned the notion that it required a 3-D thymic stromal structure. However, the inability of freshly isolated TSC monolayers to induce and support T cell development form HPCs has since remained a paradoxical finding in light of the OP9-DL1 system. In this study, we demonstrate that TSCs appear to require an intact 3-D structure to maintain expression of two critical Notch ligands, Dll1 and Dll4, both of which are normally expressed by undisturbed TSCs. Transduced TSMCs re-expressing Dll1 or Dll4 regain the ability to support T cell development, clearly establishing that a critical missing feature of TSMCs is the expression of a necessary Dll family member. It also demonstrates that Dll4 has similar properties as Dll1 in inducing HPCs to adopt a T cell fate and to differentiate into DP and SP T cells.

Both Dll1 and Dll4 are expressed in fetal thymus and appear to have redundant functions in their ability to induce T cell generation. This redundancy can explain results from a recent report showing that conditional deletion of Dll1 in vivo was dispensable for T cell development (17). We show here that the dramatic down-regulation of both Dll1 and Dll4 expression in TSMC, to an extent that it may mimic the effect of a double Dll1/Dll4 deletion in the thymus, results in a complete block in T cell development. This block is not compensated for by the residual Jag1 and Jag2 expression that remain unchanged or reduced, respectively. Further, other reports demonstrated that Jag1 and Jag2 overexpression in bone marrow derived cell lines failed to induce T-lymphopoiesis from HPCs (18, 19, 20). Although Jag2 has been shown to influence NK cell development (20), the evidence so far suggests that it is unlikely that the loss of Jag2 expression in TSMCs is responsible for the loss in T lymphopoiesis.

The renewed ability of TSMCs with forced expression of Dll1 or Dll4 to support T cell development clearly establishes that a critical missing feature of cultured TSCs is the expression of the required Dll family members. Nevertheless, overexpression of Dll1 or Dll4 does not de facto transform any stromal cell monolayer into one capable of supporting T lymphopoiesis. Several possibilities exist for the failure of NIH3T3-Dll cocultures to generate T cells. These include possible expression of inhibitory factors (e.g., TGF-β) (21), insufficient expression of other regulatory factors (e.g., Hedgehog or WNTs) (22, 23), or an inability of NIH3T3 cells to support lymphocyte lineage commitment to a T cell fate (despite supporting B cell development in the presence of exogenous IL-7). In an attempt to uncover possible inhibitory signals, we performed cocultures using mixed OP9-DL1 and NIH3T3 cells; however, these conditions could efficiently generate DP T cells (data not shown), indicating that NIH3T3 cells did not possess any intrinsic inhibitory properties. In any case, our results point to a requirement for additional factors necessary for the generation of T cells from HPCs.

Our results imply that normal expression levels of Dll1 and Dll4 are maintained through close intercellular interactions between TSCs in a 3-D arrangement. In this study, we show that FTOC and RTOC matrices devoid of lymphocytes are still able to maintain Dll expression for at least 5 days despite thymocyte depletion (Fig. 2); this suggests that the key stimulatory intercellular interactions missing in TSMCs may involve TECs and mesenchymal cells (24). Candidate molecules for these interactions include fibroblast growth factors derived from mesenchymal cells interacting with TECs expressing fibroblast growth factor R2-IIIb (25) and homotypic interactions of E-cadherin between neighboring TECs (26). Although the downstream effects of these specific signals on Dll expression are unknown, concomitant loss of that Foxn1 and Dll transcripts in TSMCs (Fig. 2 B) merits examining whether Foxn1 mediates an event regulating the expression of Dll molecules (16). Identifying factors involved in stromal cell interactions will be an important additional step in outlining the signaling pathways responsible for maintaining Dll expression in an intact thymic microenvironment, thus in turn conferring the unique ability of the thymus to generate T cells.

We thank H. T. Petrie and J. R. Carlyle for critical review of this manuscript and G. Knowles for expert assistance in cell sorting.

The authors have no financial conflict of interest.