The requirement for Notch signaling at the β-selection checkpoint in vivo is absolute and independent of the pre–T cell receptor

To inhibit signaling from all Notch receptors (Notch1–4), we generated DNMAML^f/+ mice (14) (Fig. 1 A). These mice harbor a cassette encoding the GFP-tagged Notch inhibitor DNMAML downstream of a floxed sequence that prevents transcription of DNMAML. Cre recombinase results in DNMAML expression from the ubiquitously active ROSA26 promoter. Notch-deprived cells can be tracked through GFP detection. DNMAML^f/+ mice were crossed to LC tg mice to generate LCD mice and compared with control LC littermates. GFP expression was detected in a small percentage of CD44⁺CD25⁺ DN2 cells (Fig. 1 B). In subsequent stages, DNMAML was induced in >50% of CD44⁻CD25⁺ DN3 cells and >95% of CD44⁻CD25⁻ DN4, TCRβ⁻CD8⁺ immature single-positive and CD4⁺CD8⁺ DP cells.

LCD thymi had impaired αβ T cell development, as shown by a decreased percentage of DP and an increased percentage of DN cells (Fig. 1 C), translating into a four to fivefold decrease in thymic cellularity (Fig. 1 D). The absolute number of DN cells was preserved, whereas numbers of immature single-positive, DP, CD4⁺ single positive (SP) and mature CD8⁺ SP cells were decreased, which was consistent with a block at the DN-DP transition. Among CD3⁺ thymocytes, the percentage of TCRγ⁺ cells was increased in LCD mice (Fig. 1 C), an effect secondary to decreased TCRβ⁺ cells because the absolute number of TCRγ⁺ cells was maintained (Fig. 1 D). This was observed despite expression of DNMAML in ∼80% of LCD γδ thymocytes (Fig. 1 B). Regarding αβ T cell development, our observations in LCD mice were consistent with findings reported after LC-mediated inactivation of Notch1 or CSL/RBP-J (7, 9). In contrast, the preservation of γδ thymocytes in LCD mice was similar to LC x Notch1^f/f mice (7) but differed from the increase in absolute γδ cell numbers in LC x CSL/RBP-J^f/f mice (9). This cannot be explained by Notch2–4 activity after Notch1 deletion because signaling from all four Notch receptors is inhibited both by DNMAML (14) and the absence of CSL/RBP-J. A potential explanation is that Notch-independent CSL/RBP-J–mediated transcriptional repression plays a role during γδ development. This repressor function would be lost in the absence of CSL/RBP-J but is unaffected by DNMAML. Alternatively, subtle differences in genetic background or timing of Notch inactivation may account for the discrepant results.

We studied the phenotype of Lin⁻ DN thymocytes in LC and LCD mice (Fig. 2), using GFP to better define the effects of Notch deprivation. The proportions of DN2, DN3, and DN4 cells were similar in LC and LCD thymi (Fig. 2 A). LCD DN3 cells expressed higher levels of CD25 than LC DN3 cells (Fig. 2 A). Although the difference was modest, it was statistically significant (Fig. S1). In addition to DN3 cells, abnormal CD25^hi DN2 cells were also present in LCD mice. These findings were reminiscent of the CD25^hi DN2-DN3 population observed after LC-mediated inactivation of Notch1 or CSL/RBP-J (7, 9). This population was hypothesized to represent Notch-deprived DN2-DN3 cells that failed to undergo β-selection, as in mice lacking pre-TCR components (7). However, all CD25^hi DN2-DN3 cells in LCD mice were GFP⁻, whereas GFP⁺ DN3 cells expressed lower levels of CD25 than the GFP⁻ DN3 cohort (Fig. 2 A). These findings indicated that CD25^hi DN2-DN3 cells emerged in LCD mice not as a direct consequence of Notch deprivation but because of a non–cell autonomous effect; e.g., an abnormality in intrathymic niches occupied by DNMAML⁺ DN cells.

We next assessed LC/LCD DN3 and DN4 cells for i.c. TCRβ and TCRγ (Fig. 2 B and Fig. S2). The percentage of i.c. TCRβ⁺ DN3 and DN4 cells showed a modest, though statistically significant, decrease in LCD as compared with LC mice. The percentage of i.c. TCRγ⁺ thymocytes was not significantly different.

We then examined LC/LCD DN3 cells for phenotypic changes associated with β-selection (Fig. 2, C–E). DN3 cells exposed to pre-TCR signals exhibit active cell cycling, increased cell size, and CD27 up-regulation (DN3b population) (15). FSC^hiCD27^hi or CD44^loCD27^hi DN3b cells were reduced in LCD as compared with LC thymi (Fig. 2 C). When assessed for DNA content, a smaller proportion of DN3 cells were in the S-G2M phases of the cell cycle (Fig. 2 D) and fewer cells incorporated BrdU in LCD as compared with LC DN3 cells (Fig. 2 E). In contrast, no significant proliferation defect was detected in DN4 cells. The proportion of i.c. TCRβ⁺ cells was reduced among BrdU⁺ LCD as compared with LC DN3 cells (LC, 68 ± 3% vs. LCD, 43 ± 4%; mean ± SEM; P < 0.01), a finding that was consistent with reports of impaired proliferation, predominantly in αβ lineage cells in vitro (16, 17). Collectively, these findings indicate that Notch-deprived DN3 cells, although detected phenotypically as CD25^lo cells at the DN3-DN4 transition, failed to undergo the typical changes associated with β-selection. We recently reported that Notch directly up-regulates c-myc transcription in primary DN3 cells and T cell leukemia cell lines, suggesting that abrogation of the Notch–c-myc axis contributes to the β-selection defects in LCD mice (18).

In vitro studies using Rag-deficient DN3 cells on OP9-DL1 stroma found that Notch signaling had important roles in cell metabolism and survival. In contrast, we did not detect consistent abnormalities in Annexin V staining and cellular bioenergetics, as measured by tetramethylrhodamine ethyl ester labeling in LCD thymi (unpublished data). This does not rule out Notch-related changes in these parameters because compromised/dying thymocytes are rapidly eliminated and have been notoriously difficult to detect in vivo (19).

The in vivo DNMAML-related defects were reminiscent of in vitro findings using OP9 cells in which the main consequences of Notch deprivation were growth arrest, decreased cell size, and eventual cell death (10, 11). To better understand changes induced by the loss of Notch signaling, we assessed cell size and CD25 expression in cultures of LC DN3 or LCD GFP⁺ DN3 thymocytes with OP9-DL1 cells (Fig. 3 A). After 24 h, LCD GFP⁺ DN3 cells exhibited decreased cell size and CD25 expression when compared with LC DN3 cells. Therefore, these changes resulted from a transcriptional effect of Notch signaling mediated by the ICN-CSL-RBP-J–MAML complex. The rapid modulation of CD25 expression suggested that Cd25 (Il2ra) is a direct transcriptional Notch target in thymocytes (20). A chromatin precipitation (ChIP) assay showed that Notch1 associated with two conserved CSL/RBP-J binding sites in the Il2ra locus; one is located immediately upstream of the transcription start site and the other in intron 3 (Fig. 3 B) These findings confirm that Il2ra is a Notch transcriptional target in developing thymocytes.

Collectively, these observations suggest a scenario in which Notch-deprived DN3 cells down-regulate CD25 and fail to proliferate. Thus, some CD44⁻CD25^lo DN3-DN4 thymocytes in LCD mice might represent Notch-deprived DN3 cells that have aberrantly down-regulated CD25. However, these cells did not accumulate to a large extent in LCD mice, as shown by the nearly normal frequency of i.c. TCRβ⁺ cells among LCD DN4 thymocytes (Fig. 2 B). The minimal accumulation of Notch-deprived DN cells in LCD mice might be related to their growth arrest (Fig. 2 D) and/or rapid elimination by apoptosis, which was consistent with in vitro work showing poor survival of cells lacking both Notch and pre-TCR signaling (16). In comparison to LCD mice, the accumulation of aberrant CD25^lo DN3-DN4 cells appeared more prominent in LC x Notch1^f/f and, to a lesser extent, in LC x CSL/RBP-J^f/f mice, even if overall impairment of T cell development was similar (7, 9). We hypothesize that residual Notch signaling through Notch2–4 receptors in LC x Notch1^f/f mice, or preformed CSL/RBP-J protein in LC x CSL/RBP-J^f/f mice, contributed to enhanced accumulation of aberrant DN3-DN4 cells without allowing normal progression through β-selection.

As in the absence of Notch1 and CSL/RBP-J (7, 9), i.c. TCRβ expression was decreased in LCD DN3 cells (Fig. 2 B). To assess if this was functionally relevant in vivo, we crossed LCD mice to Tcrb tg mice (Fig. 4, A and B). DNMAML did not affect transgene expression (Fig. S3). The Tcrb transgene failed to restore cellularity of LCD thymi back to normal numbers (Fig. 4 A), despite a twofold increase in LCD/Tcrb tg mice. The percentage of LCD DP cells was decreased with and without the transgene (Fig. 4 B), although the decrease was slightly less pronounced with Tcrb. However, we consistently observed a higher percentage of GFP⁻ cells among DP thymocytes of LCD/Tcrb tg than in non-tg LCD mice. This suggested that the transgene accelerated the transition through DN2-DN4 stages of development, allowing more DP cells to arise without being exposed to DNMAML. Together, these results indicated that a Tcrb transgene did not rescue Notch inactivation at the β-selection checkpoint in vivo.

In addition to its effect on TCRβ, Notch might regulate pTα expression (12). To assess if this played a limiting role in vivo, we crossed LCD mice to TcrAND tg mice (Fig. 4, C and D). These mice express Tcra/b transgenes that can substitute for pre-TCR function (21), though with reduced efficiency (22). Expression of the Tcra/b transgenes did not restore thymic cellularity (Fig. 4 C). Accordingly, the percentage of DP cells was reduced in LCD/TcrAND as compared with TcrAND mice (Fig. 4 D). Similar results were observed with DO11.10 tg mice (unpublished data).

These results show that Notch deprivation in vivo cannot be rescued by restoring pre-TCR function. Instead, they suggest that Notch and the pre-TCR act in parallel pathways. An important future task will be to characterize the interactions between Notch, pre-TCR signals, and other partners that are active during β-selection, such as E proteins. Of note, recent work indicates that Notch and E proteins cooperate during T lineage commitment (23).

In vitro experiments suggested an absolute requirement for Notch at the DN-DP transition (10, 11). However, the generation of LCD DP cells was reduced but not abolished in vivo. This could be explained by a less stringent requirement for Notch signaling in vivo or by the precise kinetics of LC-mediated excision. To differentiate between these possibilities, we purified GFP⁻ and GFP⁺ DN3 cells from LCD mice and performed intrathymic injections (Fig. 5). Control LC DN3 cells gave rise to donor-derived DP/SP T cells 10 d after injection (Fig. 5 A, top). In contrast, Notch-deprived GFP⁺ LCD DN3 cells gave rise to no or barely detectable progeny (Fig. 5 A, middle). There was at least a 3-log reduction in donor-derived cells in the absence of Notch signaling (Fig. 5 B). When GFP⁻ LCD DN3 cells were injected, significant numbers of donor-derived DP/SP cells were observed at day 10 (Fig. 5 A, bottom), and >50% of these cells were GFP⁺ as a result of DNMAML induction in vivo between injection and analysis. These results indicate that the requirement for Notch during β-selection is as stringent in vivo as in vitro. Furthermore, the apparently partial differentiation block observed in vivo results from late Notch inactivation in a fraction of DN3-DN4 cells.

In conclusion, our observations bring new insights into the role of Notch at the β-selection checkpoint in vivo. By tracking Notch-deprived cells, we demonstrated an absolute requirement for Notch signaling during β-selection that was not rescued by restoring pre-TCR function. Our results are consistent with a predominant and limiting effect of Notch on proliferation and cell survival, although we cannot rule out additional effects on differentiation. Because the effect was observed on interference with the Notch transcriptional activation complex, future work to identify relevant downstream mediators of Notch will have to focus on transcriptional targets of Notch signaling.

DNMAML^f/+ mice were generated as previously described (14). The DNMAML-GFP construct encodes amino acids 13–74 of MAML1 fused to GFP (13, 14). DNMAML^f/+ mice were crossed to LC tg mice (Taconic). LCD mice were compared with LC littermates. B6.CD45.1 and Tcrb tg mice expressing a TCRβ chain from the DO11.10 hybridoma were obtained from Taconic. Tg(TcrAND) and Tg(DO11.10) Tcra/Tcrb tg mice were purchased from the Jackson Laboratory. Experimental protocols were approved by the University of Pennsylvania Office of Regulatory Affairs.

The following antibodies were from obtained from BD Biosciences or eBioscience: PE anti-CD25 (PC61), CD27 (LG.7F9), TCRβ (H57-597), TCRγ (GL3), CD4 (RM4-5), and CD3 (145-2C11); APC anti-CD4, TCRβ, CD44 (IM7), and BrdU; biotinylated anti-CD45.2 (104), CD8 (53-6.7), TCRβ, TCRγ, CD4 (GK1.5), CD3, NK1.1 (PK136), B220 (RA3-6B2), CD19 (1D3), CD11b (M1/70), Gr1 (RB6-8C5), and CD11c (HL3); APC-Cy7 anti-CD25; and PE-Cy5.5 anti-CD44 and PE-Cy7 anti-CD45.1 (A20). Biotinylated antibodies were revealed with streptavidin-PerCP (BD Biosciences), Pacific Blue (Invitrogen), or PE–Texas red (Caltag). Lineage⁺ cells were defined with anti-CD8, TCRβ, TCRγ, NK1.1, CD3, B220, CD19, CD11b, Gr1, and CD11c.

Cells were stained in PBS/2% FCS. i.c. staining was performed with fixation/permeabilization or BrdU labeling kits (Becton Dickinson). 0.5 mg BrdU was administered i.p. 3 and 1 h before death. Cells were sorted on a FACS DiVa (Beckton Dickinson) or a MoFlo (DakoCytomation). Analysis was performed on a FACS Calibur or LSR II (Becton Dickinson). DAPI was used to exclude dead cells or assess DNA content in fixed cells. Files were analyzed with software (FlowJo; Tree Star, Inc.).

1–2 × 10⁵ sorted DN3 cells were injected intrathymically in anesthetized B6.CD45.1 recipients given 500 rad 2–6 h before injection.

OP9-DL1 cells were provided by J.C. Zuniga-Pflucker (University of Toronto, Toronto, Canada) and used as previously described (24). Progenitors were seeded into 24-well plates containing a stromal monolayer with 1 ng/ml mIL-7 (PeproTech).

ChIP was performed from Rag-2^−/− DN3 cells using Notch1 TAD domain–specific antiserum (25), anti-acetylated histone 4 (Upstate Biotechnology), or rabbit IgG (Santa Cruz Biotechnology, Inc.), as previously described (18). Quantitative PCR was performed with SYBR green (Applied Biosystems) and the following Il2ra-specific primers: 25K, 5′ CAGTCATTGGTTGGCCACTCT 3′ and 5′ GGACCTCCATGCAGACATCA 3′; promoter, 5′ TGTTGAGTCTTCTGGGGGAGAA 3′ and 5′ CTAGGAGGTGTGGGCAGTGTTT 3′; and intron 3, 5′ TGCAGCATGGGTCAAATGAA 3′ and 5′ AGGTCTCCCCAGGAAAAGTCAC 5′.

Fig. S1 depicts CD25 median fluorescence intensity in Lin⁻ CD44 DN3/DN4 LCD as compared with LC thymocytes. Fig. S2 shows expression of i.c. TCRβ and i.c. TCRγ in Lin⁻ DN3-DN4 LC and LCD thymocytes (a representative example is presented). Fig. S3 shows expression of i.c. TCRβ in the presence or absence of a Tcrb transgene in LC and LCD DN thymocyte subsets.

We thank Juan-Carlos Zuniga-Pflucker and Julian Lum for reagents and help.

This work was supported by grants from the National Institutes of Health (NIH) to A. Bhandoola and W.S. Pear. Individual support was provided by a grant from the Damon Runyon Cancer Research Foundation to I. Maillard (DRG-102-05), NIH training grants to L. Tu (T32HL007439-27) and J. Millholland (T32CA09140-31-35), and a Leukemia & Lymphoma Society Fellow Award to K. Keeshan and a Scholar Award to A. Bhandoola.

The authors have no conflicting financial interests.