Maintenance of T Cell Specification and Differentiation Requires Recurrent Notch Receptor–Ligand Interactions

Mice.

C57BL/6 mice and timed-pregnant Swiss.NIH mice were obtained from the National Cancer Institute, Frederick Cancer Research and Development Center.

Cell Lines.

The OP9-control and OP9-DL1 cell lines were generated, as described previously (38), by infecting the bone marrow stromal cell line OP9 (39) with the empty MigR1 retroviral vector (1), or with the MigR1 retroviral vector engineered to express the Dll1 gene (40) 5′ of the internal ribosomal entry site, allowing the bicistronic expression of Dll1 and GFP. OP9 cells, OP9-control cells, and OP9-DL1 cells were cultured as a monolayer in OP9 media (αMEM supplemented with 20% FCS [HyClone] and 2.2 g/liter sodium bicarbonate), to which IL-6, IL-7, IL-15, and Flt3 ligand (all from PeproTech) were added to enhance the growth of T, B, and NK cells.

Isolation of Fetal Cells.

Day 14 fetal thymus and fetal liver were harvested, washed in 5 ml OP9 medium, and disrupted through 40-μm nylon mesh using a syringe plunger. CD24^low/Lin⁻ fetal liver cells were obtained by antibody/complement-mediated lysis, as described previously (16). In brief, 2 ml anti-CD24 (J11d.2) culture supernatant and a 1:10 dilution of Low-Tox rabbit complement (Cedar Lane) were added to single cell suspensions in 10 ml complete medium, and cells were incubated at 37°C for 30 min. After incubation, viable cells were recovered by discontinuous density gradient centrifugation over Lympholyte-Mammal (Cedar Lane) and washed before flow cytometric cell sorting.

Flow Cytometry and Cell Sorting.

Flow cytometry was performed using a FACSCalibur™ (BD Biosciences) instrument, as described previously (16). FITC-, PE-, biotin-, and APC-conjugated mAbs and streptavidin-APC were purchased from BD Biosciences. For analysis, live cells were gated based on forward and side scatter, and lack of propidium iodide uptake. Cells were sorted using a FACSDiVa™ (BD Biosciences). Bulk-sorted cells were ≥99% pure, as determined by postsort analysis.

OP9 Cell Cocultures.

Fetal thymus DN subsets were sorted based on the differential surface expression of CD117, CD44, and CD25, as illustrated in Fig. 1 a. Sorted DN subsets were seeded at 10³ cells/well into 24-well tissue culture plates containing either OP9-control or OP9-DL1 cells, to which 1 ng/ml IL-6, 1 ng/ml IL-7, 25 ng/ml IL-15, and 5 ng/ml Flt3 ligand were added to enhance the growth of T, B, and NK cells. The cells were cultured for 7 d, and then analyzed by flow cytometry. CD24^low/Lin⁻ fetal liver progenitor cells were sorted for CD117⁺ Sca-1^hi HPCs and cultured on either OP9-control or OP9-DL1 cells. Various concentrations of the presenilin Inhibitor X (Calbiochem) were added to the culture wells, as indicated (see Fig. 4). The culture medium containing equal volumes of DMSO alone or presenilin inhibitor (diluted in DMSO) was changed on the fourth day of culture.

Precursor Frequency Analysis.

Limiting dilution analysis was performed using the Clonecyte option of the FACSDiVa™ cell sorter. As mentioned above, IL-6, IL-7, IL-15, and Flt3 ligand were added to these assays to enhance the growth of T, B, and NK cells. Precisely 1, 3, 10, 20, 30, or 100 DN1 and DN2 thymocytes were deposited into 96-well plates containing OP9 cell monolayers, with 56 replicate wells for each sample group of DN1 cells and 53 replicate wells for each sample group of DN2 cells. Limiting dilution analysis was also performed using OP9-DL1 cells by sorting 1, 3, 10, or 30 DN1 and DN2 thymocytes into wells of 96-well plates containing OP9-DL1 cell monolayers, with 24 replicate wells for each sample group. After 12–14 d, the cells were harvested from individual wells and analyzed by flow cytometry. The presence of CD45⁺ NK1.1⁺ CD19⁻ NK cells, CD45⁺ NK1.1⁻ CD19⁺ B cells, and CD45⁺ NK1.1⁻ CD3⁺ T cells was scored and the progenitor frequency was determined by the method of maximum likelihood applied to the Poisson model (41). Clonal assays were performed by depositing single DN1 or DN2 thymocytes into individual wells of 96-well plates containing OP9-DL1 cells by flow cytometric cell sorting. The percentage of T cell progenitors, NK cell progenitors, and T/NK bipotent progenitors within the DN1 and DN2 thymocyte populations was determined flow cytometrically by analyzing the phenotype of cell populations derived from single cells after 11–14 d of culture (NK cell progenitors gave rise only to CD3⁻ NK1.1⁺ NK cells, T cell progenitors gave rise only to CD3⁺ T cells, and T/NK bipotent progenitors gave rise to both CD3⁻ NK1.1⁺ NK cells and CD3⁺ T cells).

PCR and RT-PCR.

Genomic DNA was purified using the EasyDNA kit (Invitrogen) from total day 14 fetal thymus, ex vivo DN2 thymocytes, T cells derived from DN2 thymocytes cultured on OP9-DL1, CD45⁺ NK1.1⁺ CD19⁻ NK cells sorted from DN2 cells cultured on OP9 cells for 14 d, and from embryonic fibroblasts. 100 ng of each DNA sample was amplified using a PTC-225 Peltier Thermal Cycler (MJ Research). Primers used for the TCR D-Jβ rearrangement analysis have been described previously (42). Products were separated by agarose gel electrophoresis and visualized by ethidium bromide staining. All PCR products shown correspond to expected molecular sizes.

Immunofluorescent Staining of Thymus Sections.

Transverse frozen sections of 5–6-wk-old C57BL/6 mouse thymus (5-μm thickness) were prepared by embedding the tissue in OCT (Fisher Scientific) followed by cryostatic sectioning onto glass slides. Sections were fixed in ice-cold acetone, dried, and rehydrated in PBS. Endogenous avidin binding activity was blocked using Avidin/Biotin Blocking Kit (Vector Laboratories). Nonspecific protein binding was blocked using PBS containing 5% FBS. This buffer was also used as diluent for all subsequent steps. Next, sections were stained with primary antibody, either goat polyclonal IgG recognizing Dll1 (catalog no. SC9932; Santa Cruz Biotechnology, Inc.), or a matching concentration of nonimmune goat IgG (Jackson ImmunoResearch Laboratories). Sections were washed extensively and stained with biotinylated donkey antiserum raised against goat IgG (Jackson ImmunoResearch Laboratories), followed by blocking of excess secondary antibody binding sites using goat IgG. Sections were then stained with a combination of streptavidin-conjugated Alexa 594 (Molecular Probes) and FITC-conjugated anti-pancytokeratin (catalog no. F3410; Sigma-Aldrich). After washing, nuclei were counterstained by mounting in Prolong solution (Molecular Probes) containing DAPI (Molecular Probes). Single color images (DAPI, FITC, and Alexa 594) were acquired on a fluorescent microscope. Exposure times for each fluorochrome were kept the same for both DL1 and control (nonspecific) primary antibodies. After acquisition, single color images were merged to generate RGB composites.

Online Supplemental Material.

Inhibition of Notch signaling does not affect B or NK cell development on OP9-control cells. See Fig. S1, which is available at http://www.jem.org/cgi/content/full/jem.20040394/DC1.

Requirement for Notch Receptor–Dll1 Interactions throughout Early T Cell Development.

To characterize the requirement for Notch receptor–ligand interactions during early thymocyte development, we sorted each of the three DN thymocyte subsets before β selection and cultured them (1,000 cells/well) on either OP9-control or OP9-DL1 cells for 7 d in conditions that effectively support T, B, and NK cell growth (Fig. 1). As described previously, DN1 cells contain multipotent lymphoid progenitors (27, 43), which, when cultured on OP9-control cells, gave rise to B and NK cells, but not T cells. In contrast, DN1 thymocytes cultured on OP9-DL1 cells gave rise to both T and NK cells, but not B cells (Fig. 1 b, left). DN1 cells proliferated on both stromal cell lines; however, in the presence of Dll1, this subset proliferated to a greater extent (148- vs. 1,150-fold expansion on OP9-control vs. OP9-DL1 cells, respectively; Fig. 1 b).

Consistent with the notion that DN2 progenitors are committed to the T cell lineage, this population failed to give rise to B cells when cultured on OP9-control cells, and T cells were observed almost exclusively when DN2 thymocytes were cultured on OP9-DL1 cells (Fig. 1 b, middle). However, DN2 progenitors gave rise to a substantial number of NK cells when cultured on OP9-control cells, whereas only a few NK cells were generated from DN2 cells cultured on OP9-DL1 cells (∼15-fold more NK cells on OP9-control vs. OP9-DL1 cells, respectively; Fig. 1 c). Therefore, our data supports the view that DN2 thymocytes may not be fully committed to the T cell lineage, and that these cells retain some plasticity toward the NK cell lineage when cultured in an environment that does not support further T cell commitment.

In contrast, DN3 cells, which bear irreversible hallmarks of T lineage commitment in the form of TCR-β gene rearrangements (34), were no longer capable of adopting an NK cell fate, failed to differentiate or expand on OP9-control cells (Fig. 1 b, right), and instead underwent cell death (not depicted). On the other hand, DN3 cells cultured on OP9-DL1 gave rise exclusively to T cells that progressed normally to the DP stage (Fig. 1 b, bottom right). These data show that a requirement for Notch signaling persists at the DN3 stage, even though these cells are already committed to the T cell lineage, implying a role for Notch signaling in functions other than lineage commitment (44).

Requirement for Notch Receptor–Dll1 Interactions during T Cell Lineage Commitment by DN2 Thymocytes.

The results in Fig. 1 b show that DN2 thymocytes do not appear to be fully committed to the T cell lineage, as this population retained the potential to give rise to an alternate cell fate under conditions that do not support T cell development. To determine the frequency of DN2 thymocytes that can differentiate into NK cells, we performed a limiting dilution analysis with DN2 and DN1 thymocytes cultured on OP9 cells. The results presented in Table I indicate that the NK cell progenitor frequency within the DN1 population (1 in 8.8 DN1 cells) is similar to that of the DN2 population (1 in 13.6 DN2 cells). This analysis also revealed that the B cell progenitor frequency within the DN1 subset appears to be surprisingly small (1 in 188), and as demonstrated in Fig. 1 b, DN2 cells do not contain any B cell lineage potential. Taken together, these data suggest that the majority of cells within the DN1 population are either committed T cell progenitors or bipotent T/NK cell progenitors, and that although the transition from the DN1 to DN2 stage results in a loss of B cell potential, it does not result in a substantial loss of NK cell lineage potential. This, in turn, suggests that continued Notch signals specifying T cell lineage fate are required at the DN2 stage.

To determine the relative frequency of committed T cell progenitors, NK cell progenitors, and bipotent T/NK cell progenitors within the early DN thymocyte subsets, we performed a clonal assay, in which single DN1 or DN2 thymocytes were placed into individual wells containing OP9-DL1 cells (Fig. 2). T cell lineage potential was determined by CD3 expression, and NK cell lineage potential was determined by the expression of NK1.1 in the absence of CD3 (Fig. 2 c). The frequency of T cell progenitors, NK cell progenitors, and bipotent T/NK cell progenitors within the DN1 and DN2 populations is shown in Fig. 2 a. These results indicate that the DN1 and DN2 subsets both contain a relatively high frequency of progenitors that give rise exclusively to T cells (27 out of 192 wells or ∼51% of the total number of reconstituted wells for DN1; 37 out of 192 wells or ∼90% of reconstituted wells for DN2), whereas the DN1 subset contained a greater percentage of NK cell lineage–only progenitors than the DN2 subset (11 out of 192 wells or ∼21% of reconstituted for DN1; 4 out of 192 wells or ∼10% of reconstituted for DN2). Notably, bipotent T/NK cell progenitors were only observed in wells seeded with DN1 thymocytes (15 out of 192 wells or ∼28% of reconstituted), whereas bipotent T/NK cell progenitors were not detected within the DN2 thymocyte subset (Fig. 2 a).

These results appear in contrast to the limiting dilution data presented in Table I, which clearly demonstrated a similar progenitor frequency for NK cells (either bipotent or NK cell lineage committed) within the first two DN thymocyte subsets. To determine whether the presence of Dll1 influences the ability of DN2 thymocytes to adopt the NK cell fate, an additional limiting dilution analysis was performed using OP9-DL1 cells to determine the progenitor frequency for T and NK cells within the DN1 and DN2 subsets (Table II).

The data shown in Fig. 2 b provides a comparison of the NK cell progenitor frequencies from the limiting dilution analysis of each population (from Tables I and II) when cultured on either control-OP9 cells (Fig. 2 b, left) or OP9-DL1 cells (Fig. 2 b, middle), as well as the T cell progenitor frequency for each population (Fig. 2 b, right). This analysis shows that the NK cell progenitor frequency was only slightly decreased for DN2 thymocytes as compared with DN1 thymocytes when cultured on OP9-control cells (11.4 to 7.4%, equivalent to only a ∼1.5-fold higher NK cell progenitor frequency in the DN1 subset), whereas there was a striking loss of NK cell progenitor frequency between the two thymocyte subsets when assayed on OP9-DL1 (4.3 to 0.3%, equivalent to a ∼15-fold higher NK cell progenitor frequency in the DN1 subset). As demonstrated in Fig. 2 a, the only NK cell potential detectable from single DN2 cells cultured on OP9-DL1 cells was derived from a small number of contaminating NK cell progenitors, suggesting that a proportion of committed NK cell progenitors in the thymus may express CD25, and thus cannot be distinguished from DN2 thymocytes. Therefore, the relatively high NK cell progenitor frequency observed from the DN2 population when cultured on control OP9 cells is likely derived from T cell progenitors, which can adopt an NK cell fate in the absence of continued Notch receptor–ligand interactions. In contrast, the progenitor frequency for T cells is not significantly different between the DN1 and DN2 subsets (66 and 64%, respectively) when cultured on OP9-DL1 cells (Fig. 2 b, right).

Notably, the NK cell progenitor frequency of the DN2 population is ∼25-fold higher when assayed on OP9-control versus OP9-DL1 cells. It is also important to note that the measured frequency of NK cell progenitors within the DN1 population is ∼2.5-fold higher when cultured on OP9-control versus OP9-DL1 cells, indicating that although T/NK cell bipotent progenitors are readily detected from DN1 cells cultured on OP9-DL1 (Fig. 2 a), Notch signaling acts at the DN1 stage to promote T lineage specification at the expense of NK cell development. Taken together, these data indicate that Notch receptor–ligand interactions direct T cell development throughout the DN1 and DN2 stages, and that the DN2 population contains progenitors that are specified to the T cell lineage, and thus require continuous Notch signaling both to maintain T lineage specification and to undergo final lineage commitment.

Status of the TCR-β Gene Rearrangement in NK Cells Derived from DN2 Cells.

Progenitor thymocytes begin the process of TCR gene rearrangement during the DN2 to DN3 transition (33, 34, 45, 46). Once thymocytes rearrange their TCR genes, they are considered irreversibly committed to the T cell lineage. The TCR-β gene locus of DN2 cells is in the germline configuration, although a small percentage of these cells have initiated the rearrangement process and display TCR D-Jβ rearrangements (33, 34, 45, 46). To determine which DN2 cells are capable of giving rise to NK cells when cultured on OP9 cells lacking Dll1 expression, DN2 thymocytes were cultured for 2 wk on OP9 cells, and the resulting NK cells were analyzed for D-Jβ rearrangement by PCR (Fig. 3). Multiple D-Jβ rearrangements were observed from fetal thymus–derived DNA, and from DNA derived from DN2 thymocytes cultured on OP9-DL1. In contrast, the genomic DNA from DN2-derived NK cells produced a single PCR product corresponding to the expected size of the germline locus, like that observed from the DNA of fibroblasts (Fig. 3). These results indicate that DN2 thymocytes that are removed from a Notch ligand–containing environment are unable to undergo TCR gene rearrangement and furthermore, DN2 thymocytes that have begun TCR gene rearrangement may no longer be capable of adopting an NK cell fate when transferred to an environment lacking Notch receptor–ligand interactions.

Different Thresholds of Notch Signaling Are Required during T Cell Lineage Commitment and Development.

To further elucidate the effect of Notch signaling during the initial stages of T cell lineage commitment, we cultured CD117⁺ Sca-1^hi fetal liver–derived HPCs on OP9-DL1 cells in the presence of increasing concentrations of a presenilin 1 and 2 inhibitor. A presenilin-dependent γ secretase activity is required for the cleavage of the intracellular transactivating domain of all four mammalian Notch receptors in response to engagement by ligand (47). Fig. 4 shows that HPCs cultured on OP9-DL1 cells in the absence of the presenilin 1/2 inhibitor (DMSO alone) failed to give rise to B lineage CD19⁺ cells, while efficiently generating primarily DN3 and pre-DP/DP T lineage cells and a small percentage of NK1.1⁺ NK cells after 7 d of culture. However, in the presence of increasing concentrations of the inhibitor there was a striking increase in the proportion of NK cells generated on OP9-DL1 cells (Fig. 4 a), which also correlated with a substantial increase in the absolute number of NK cells in the cultures (Fig. 4 c). Importantly, the effect on T versus NK cell development occurred at lower concentrations of inhibitor than was necessary to allow B cells to develop (Fig. 4 a). Furthermore, progression through the DN stages of T cell development was also impaired or delayed in the presence of low levels of inhibitor (Fig. 4 b). These results suggest that progenitors require relatively high levels of Notch signaling to initiate commitment to the T cell lineage and progress through the DN stages of development, such that at even low inhibitor concentrations, an increase in the generation of NK cells becomes apparent. On the other hand, it appears that the adoption of the B cell lineage by HPCs is sensitive to even low levels of Notch signaling, such that B cells only develop in the presence of the highest concentrations of the presenilin 1/2 inhibitor. The addition of presenilin 1/2 inhibitor to HPCs cultured on OP9-control cells had no effect on NK or B cell development (Fig. S1, available at http://www.jem.org/cgi/content/full/jem.20040394/DC1).

The finding that continuous Notch signaling is required for T cell commitment and during the initial DN stages of T cell development suggests that a Notch ligand must be expressed throughout the thymic cortex, where early T cell development occurs. Therefore, we analyzed transverse sections of adult thymus for the expression of Dll1 (Fig. 5). Cortical epithelial cells, which are present throughout the thymic cortex, were found to express Dll1, particularly near the cortico-medullary junction, as staining for Dll1 and cytokeratin colocalized on the same stromal cells. These findings are consistent with those reported by Harman et al. (48), in which transcripts for the different Notch ligands, including Dll1, were shown to be expressed by thymic stromal cells. Because cortical stromal cells provide the matrix for adhesion and migration of DN cells across the cortex (49), this pattern of Dll1 expression shows that Dll1 is in the position to provide the necessary Notch signals that induce and support the commitment and early development of T cells within the thymus.