Notch and Presenilin Regulate Cellular Expansion and Cytokine Secretion but Cannot Instruct Th1/Th2 Fate Acquisition

Recent reports suggested that Delta1, 4 and Jagged1, 2 possessed the ability to instruct CD4+ T cell into selection of Th1 or Th2 fates, respectively, although the underlying mechanism endowing the cleaved Notch receptor with memory of ligand involved in its activation remains elusive. To examine this, we prepared artificial antigen-presenting cells expressing either DLL1 or Jag1. Although both ligands were efficient in inducing Notch2 cleavage and activation in CD4+ T or reporter cells, the presence of Lunatic Fringe in CD4+ T cells inhibited Jag1 activation of Notch1 receptor. Neither ligand could induce Th1 or Th2 fate choice independently of cytokines or redirect cytokine-driven Th1 or Th2 development. Instead, we find that Notch ligands only augment cytokine production during T cell differentiation in the presence of polarizing IL-12 and IL-4. Moreover, the differentiation choices of naïve CD4+ T cells lacking γ-secretase, RBP-J, or both in response to polarizing cytokines revealed that neither presenilin proteins nor RBP-J were required for cytokine-induced Th1/Th2 fate selection. However, presenilins facilitate cellular proliferation and cytokine secretion in an RBP-J (and thus, Notch) independent manner. The controversies surrounding the role of Notch and presenilins in Th1/Th2 polarization may reflect their role as genetic modifiers of T-helper cells differentiation.

Recent studies have suggested that distinct Notch ligands expressed on APCs might regulate Th1 and Th2 fate choice [29,30,31,32]. Notch proteins are membrane-bound receptors that regulate diverse cell fate decisions in multi-cellular organisms [33].
Notch signaling regulates developmental processes during hematopoiesis and lymphopoiesis, and is essential for differentiation of single-positive T-cells from the common lymphoid progenitor [34,35,36,37,38,39,40]. Despite the assumption that manipulation of this linear pathway by different strategies should lead to a similar set of observations, the role of Notch signaling in Th1 and Th2 development has been controversial (Table S1A & B) [29]. Particularly provocative were reports that DLL1 biased naïve CD4 + T cells towards the Th1 fate [31], whereas Jag1 biased toward Th2 [30], suggesting that pathogens drive distinct T helper fate choices through the induction of alternative Notch ligands on antigen presenting cells [30].
While the induction of Th1 or Th2 development by distinct Notch ligands might resemble the mechanism of fate induction mediated by cytokines such as IL-12 and IL-4, it is difficult to explain how activation of Notch receptors by its ligands could instruct divergent fates. Understanding this difficulty requires an appreciation of how Notch is activated: binding of DLL1 or Jag1 ligands to the Notch extra-cellular domain triggers a conformational change that exposes a b-strand of Notch to cleavage by ADAM family metalloproteases [41,42]. This cleavage results in shedding of the ectodomain [43,44], generating an intermediate that is recognized by Nicastrin [45], a component of the enzyme csecretase. Nicastrin then transfers truncated Notch into the active site of c-secretase, which cleaves the Notch transmembrane domain near the inner leaflet [46,47]. Following this cleavage, the Notch intracellular domain (NICD) translocates to the nucleus where it regulates gene expression [47,48]. The four mammalian Notch receptors regulate transcription through a common DNA binding protein, RBP-J [49], and require the recruitment of mastermind-like (MAML) proteins [50,51,52,53] and additional co-activators [54,55] to initiate transcription on target promoters. Given this activation mechanism, it is unclear how NICD could retain the memory of which ligand induced ectodomain shedding and translate this memory into distinct transcription profiles. Therefore, it is immensely interesting to explore the basis of such ''ligand memory'' and to explain how DLL1 could induce Th1 through T-bet upregulation [31], whereas Jag1 could initiate Th2 development by inducing GATA-3/IL-4 expression [30,56,57], when both ligands should lead to essentially the same intracellular signal within the T cells.
We examined the activity of Notch ligands in directing Th1/ Th2 differentiation. In contrast to previous reports [30,31], we demonstrate that DLL1 and Jag1 are insufficient for instructing specification of either Th1 or Th2 fates in the absence of polarizing cytokines, and can mildly enhance cytokine-induced Th1/Th2 responses. In addition, we show for the first time that Jag1 is incapable of activating Notch1 signaling in naïve CD4 + T cells, which express Lunatic Fringe. We also examined the requirement for Notch signaling on CD4 + T cell fate specification by removal of Presenilin1 (PS-1) and PS-2 genes, which encode the c-secretase catalytic subunits, and by removal of the nuclear Notch co-activator RBP-J, to separate Notch-independent from Notch dependent activities of c-secretase.
First, we find that Notch signaling is not necessary for cytokineinduced Th1/Th2 fate selection, consistent with some previous studies that identified a co-stimulatory role for Notch [58,59,60]. This analysis, however, uncovered two novel, RBP-J-independent functions of presenilin, one contributing to the proliferative response and the other to secretion of cytokines in T-helper cells. Taken together, our data suggest that intact Notch signaling and Presenilins function permit optimal peripheral T-helper cell responses, rather than exerting direct influences on Th1/Th2 differentiation choices.
The primers 59-CGGGATCCGCCAATGCGGTCCCCACG-GACGCGC-39 and 59-GTTCTCGAGCTATACGATGTATTC-CATCCGGTT-3 were used to amplify Jagged1 cDNA. The primers 59-GAGGATCCGCCATGGGCCGTCGGAGCGCGC-TAG-39 and 59-TTACTCGAGTTACACCTCAGTCGCTATAA-CAC-39 were used to amplify DLL1 cDNA. The PCR amplified fragments were first blunt cloned into PCS2+ vectors and then replaced with original cDNA sequences by restriction digest. PCS2+DLL1 and PCS2+Jagged1 were then completely sequenced, digested with BamHI and XhoI, and the ligands subcloned into the Bgl II and XhoI site on the IRES-GFP-RV vector.

Cell lines
CHO cell lines stably expressing either full-length Notch2 (CHOfNotch2) or full-length Delta-like-1 (fD1-CHO) were a generous gift from Drs. S. Chiba and H. Hirai (University of Tokyo, Japan). Parental CHO cells expressing MHC class II molecules (I-A d haplotype) and B7-1 were a generous gift from Dr. A. Sharpe (Harvard University, Boston, MA) and were used to generate artificial APC lines that express Notch ligands. Retroviral infection of CHO cells was performed with a protocol modified from Dr. G. Nolan's lab (Stanford University, CA). Briefly, retrovirus was packaged by transfection of 293T cells with constructs pYITG, pCGP and the viral vectors using the calcium phosphate method (in BES buffered saline with chloroquine at a final concentration of 25 mM). After 9 h, transfected 293T cells were washed once and replenished with new media. Virioncontaining supernatant (10 ml) was harvested 48 h post-transfection, filtered (0.45 mM) and transferred to CHO cells (4.5 ml of 293T supernatant in 6.5 ml of media supplemented with polybrene at a final concentration of 5 mg/ml). After 48 h, 6 to 25% of cells were infected. Three different GFP hi APC lines (IRES-GFP-RV infected or CHO-B7; DLL1-IRES-GFP-RV infected or CHO-DLL1; Jagged1-IRES-GFP-RV infected or CHO-Jag1) were FACS sorted to .95% purity, stained with I-A d and B7-1-phycoerythrin and analyzed by flow cytometry to confirm a comparable level of staining for MHC and B7 ligands (80-90% of the GFP hi population; Figure S1A). All APC cells were maintained in Iscove's DMEM, supplemented with 10% heatinactivated FBS, nonessential amino acids, sodium pyruvate, penicillin/streptomycin and b-mercaptoethanol. 293 T cells were maintained in DMEM media according to ATCC protocol.
Co-culturing Experiment 1.2610 5 of CHOfNotch2 cells was seeded on 24-well plate 24 h prior to transfection. PCS2+bgal (control for transfection), TP1-luc and PCS2+ (as DNA carrier) plasmids were transfected by Lipofectamine TM 2000 according to manufacturer protocol. Artifical APC lines or fD1-CHO (positive control, at 0.1610 6 ) were added 24 hr after the transfection and luciferase assay was carried out 48 h later.

Luciferase Assays
Cells were harvested after 48 h in co-culture and washed once with phosphate-buffered saline (PBS). Cells were lysed in 100 ml of lysis buffer (100 mM KPO 4 buffer, pH 7.8; 0.2% Triton; 1 mM dithiothreitol (DTT); protease inhibitors) at room temperature for 10 min. 5 ml of lysate was used to determine b-galactosidase concentration (to normalize for transfection efficiency) according to the Tropix Galacton chemiluminescent substrate instructions. 50 ml of lysate incubated with luciferin assay buffer (30 mM Tricine, pH 7.8; 3 mM ATP; 15 mM MgSO 4 ; 10 mM DTT; 0.2 mM CoA; 1 mM luciferin) was used to determine luciferase activity using a Tropix TR717 luminometer.

T cell purification and in vitro differentiation Naïve CD4 + T cells Purification
Two different methods were used in this study. The first method, used to generate Figure 3B, employed MACS column and anti-CD4 magnetic beads (Miltenyi Biotech, Auburn, CA) for rapid separation of CD4 + T cells. For other experiments in this study, a two-step purification protocol was used. The CD19+ fraction was first removed with anti-CD19 magnetic beads on MACS column (Miltenyi Biotech, Auburn, CA). Subsequently, the CD19-negative fraction was FACS sorted (MoFlo TM , Dako) using CD4-FITC (Caltag, RM2501) and CD62L-phycoerythrin antibodies (Caltag, RM4304) to purify CD4 hi , CD62L hi population.
Priming by APC 2610 6 artificial APC cell lines were seeded on a 60-mm culture plate one day prior to the experiment. On the day of the experiment, APC cells were treated with 100 mg/ml of mitomycin C (Sigma) at 37uC for 1 h, washed twice in PBS and lifted with 0.2 M EDTA. APCs were then seeded (at 0.25610 6 ) on 48-well plates (for subsequent activation of T cells) or added (at 0.1610 6 ) to transfected CHOfNotch2 (for reporter assay). To control for the activity of natural APCs, 5610 6 irradiated (2000 Rad) splenocytes from BALB/c mice (spl) were used. 0.5610 6 naïve CD4+ T cells isolated to .98% purity from DO 11.10 mice were added to the 0.25610 6 APCs in the presence of 0.3 mM OVA peptide and under one of the 4 polarizing conditions. Th1: 10 U/ml of IL-12 and 10 mg/ml of anti-IL-4 (11B11). Th2: 100 U/ml of IL-4, 3 mg/ml of anti-IL-12 (TOSH) and 10 mg/ml of anti-IFNc (H22). Drift: Only media. Neutral: 10 mg/ml of anti-IL-4 (11B11), 3 mg/ml of anti-IL-12 (TOSH) and 10 mg/ml of anti-IFNc (H22). Polarization experiments were carried out in a 7-day cycle, starting with activation on day 0. On day 2, cells were expanded into fresh media containing 40 U/ml of IL-2 (see Table S4A for additional details). T cells were collected on day 7 and counted. 0.5610 6 T cells were then re-stimulated either with plate-bound anti-CD3 for 24 h and its supernatant collected for ELISA, or PMA (50 ng/ml) and Ionomycin (1 mM) in the presence of Brefeldin A (1 mg/ml, Epicenter Technology, Epicenter Technologies, Madison, WI) for the 4 h and subjected to intracellular cytokine staining.

Priming by anti-CD3/anti-CD28
48 well plates were coated with anti-CD3 (500A2: 1 mg/ml) and anti-CD28 (0.8 mg/ml) overnight at 4uC. 0.5610 6 of purified naïve CD4 + T cells (.95%) from different genotypes (Het, PSdko, Rko & PSRtko) were stimulated in 6-day cycles, starting with activation on day 0 in either Th1 or Th2 polarizing conditions (see below). On day 2, cells were expanded into fresh media plus 40 U/ml of IL-2 (see Table S4B and C for details) and resting cells were collected on day 6/7 and counted. 0.5610 6 of cells from each genotype were re-stimulated overnight with plate-bound anti-CD3 or PMA (50 ng/ml) and Ionomycin (1 mM). Brefeldin A (1 mg/ml; Epicentre Technology) was added for the final 4 h of each stimulation. Cells were subjected to intracellular cytokine staining and supernatant collected for ELISA.

Tabulation of cytokine production
Ten independent in vitro polarization experiments were conducted for Figures 4-6, comprising of at least one control (Het) and two test genotypes (PSdko, Rko or PSRtko) per experiment. The level of secreted cytokines was measured empirically. In each experiment, the highest secreted cytokine level was set as 100% and used to normalize the values from other genotypes. Results from the independent experiments were compiled with these percentiles and presented as relative cytokine level. The normality of distribution and significance were calculated by Kolmogorov-Smirnov and two-tailed paired t-test respectively.

Notch ligands expressed by APCs are functional and do not alter T cell proliferation
Based on in vitro differentiation assays of naïve CD4 + T cells, an inductive role for Notch in promoting Th1 or Th2 development was recently suggested [30]. To test how Notch signals could Results are mean6s.d. from three independent experiments. (B) DLL1 and Jag1 ligands do not cause significant differences in the level of IL-4 production under ''drift'' and ''neutral'' conditions. They only marginally enhance IL-4 cytokine secretion under Th2 polarizing conditions. Activated T cells were re-stimulated on Day 7 with anti-CD3 antibody for 24 h and before supernatant collection and ELISA. Results are mean6s.d. of three independent experiments. The P value was determined by student two-tailed t test. (C) Only CHO-DLL1 APC line triggers Notch1 cleavage in activated CD4 + T cells. Naïve CD4 + T cells from DO10.11 mice were purified to .99% purity by the two-step protocol and primed with various APC lines and 0.3 mM OVA peptide in 3 different polarizing conditions. They were isolated 24 h later and probed with V1774 and actin antibodies. WB: western blot. (D) Detection of Lunatic Fringe in CD4 + T cells. Naïve CD4 + T cells purified from DO11.10 mice were activated with anti-CD3/CD28 antibodies under the 4 polarizing conditions for 24 h and harvested for western with anti-Lfng antibody. Lysate from newborn pup (P1) was used as positive control whereas negative control was lysate from NIH3T3 cells. WB: western blot. doi:10.1371/journal.pone.0002823.g002 Activated Notch1 is detected in activated control CD4 + T cells under both Th1 and Th2 polarizing conditions but not detected in T cells that have targeted ablation of PS1/PS2 alleles. Non transfected HEK293T cells were used as negative control, while cells transfected with PCS2+N1DE was used as a positive control. CD4 + T cells were isolated with anti-CD4 magnetic beads on MACS column to .95% purity from the spleens of 2 months old littermates. This protocol allows co-purification of natural APCs that provide the Notch ligands (compare with Figure 2C where no Notch1 activation was observed when a two-step purification method was used). T cells were activated with anti-CD3/CD28 in Th1 or Th2 polarizing conditions for 24 h. T cells were then FACS sorted for CD4 + population and probed with V1744 antibody. WB: western blot. (C) Proliferation capacity was measured by 3 [H]-thymidine incorporation. Reduced proliferation was observed in PSdko and PSRtko cells under both Th1 and Th2 polarizing conditions. Rko cells were not significantly different than controls. Results are presented as mean6S.D. of five wells and representative of at least three independent experiments. (D, E) Reduction in the final number of viable PSdko and PSRtko T cells 6 days after activation. Naïve CD4 + T cells from different genotypes were stimulated with anti-CD3/CD28 antibodies for 2 days in Th1 and Th2 conditions before they were expanded in fresh media containing IL-2 cytokine. After 6 days of culture, T cells were harvested and viable cells were counted. Each circle/diamond indicates data of individual mouse. In Figures 1-4, circle (#) indicates data point in which CD4 + T cells were expanded by regimen 2 (Supplemental Table 4B), whereas diamond (e) indicates result in which T cells were expanded according to regimen 3 (Supplemental Table 4C). The P value was determined by student two- induce different T-helper cell differentiation programs, we created Notch-ligand expressing APCs by modifying Chinese hamster ovary (CHO) cells that stably express B7-1 and MHC class II (I-A d ) to also express either DLL1 or Jag1 ligands ( Figure 1A and Figure S1A). To confirm the effectiveness of Notch ligands expressed by these APCs, we co-cultured our panel of APC lines with CHOfNotch2 reporter cells that express a full-length Notch2 receptor [66] and the TP-1 luciferase reporter cassette [65]. We found that the APCs that express DLL1 or Jag1 could robustly stimulate luciferase activity in CHOfNotch2 cells. In contrast, the APC line CHO-B7, which contains GFP-expressing retroviral vector and lacks Notch ligands, failed to induce luciferase activity in CHOfNotch2 cells ( Figure 1B). These results are an important positive control demonstrating the functional integrity of the DLL1 and Jag1 ligands expressed by our APCs.
Given that attaining optimal T-helper differentiation requires proliferation [67,68,69,70], we tested the proliferative responses of naïve CD4 + T cells activated by APCs with or without Notch ligands. To inhibit APC proliferation, APC were exposed to 1 hr of 100 mg/ml mitomycin C treatment. This regimen effectively blocked APC proliferation; allowed Notch ligands on the APC lines to induce luciferase activity in CHOfNotch2reporter lines ( Figure 1B); and triggered proliferative responses in naïve  DO11.10 CD4 + T exposed to three different concentrations of OVA peptide ( Figure 1C). These results indicated that activation of Notch by APCs in the presence of CD28 co-stimulation did not enhance antigen-driven T cell proliferation. More importantly, this system provides us with the ability to test whether functional Notch ligands on APCs can instruct Th1 or Th2 fate selection.

Functional Notch ligands expressed by APCs do not instruct T cell differentiation
To examine the role of Notch ligands on T cell differentiation, naïve DO11.10 CD4 + T cells were activated with these three APC cell lines under 4 conditions ( Figure 1A). We included both Th1and Th2-polarizing conditions to test if either ligand could augment or inhibit cytokine-driven differentiation. We also included two kinds of non-polarizing conditions to test if Notch ligands themselves were sufficient to bias/instruct differentiation: in one, cytokines are neither added nor neutralized to test whether Notch ligands can bias toward Th1 or Th2 differentiation. If they could induce IL4, for example, this condition will allow for autostimulation. In the second type of non-polarizing condition, polarizing cytokines are neutralized to test whether Notch ligands can induce Th1 or Th2 fate acquisition on their own. Finally, we used irradiated BALB/c splenocytes as APCs under all conditions as a positive control for normal cytokine-induced Th1 and Th2 differentiation.
Activated T cells were passage 7 days (detailed in Table S4A), harvested and counted. Equal number of viable cells was restimulated with PMA/ionomycin (4 h in the presence of BFA for intracellular staining) or anti-CD3 (for ELISA of the supernatant). Artificial APC lines induced T cells expansion 2 to 3.6 fold greater than irradiated splenocytes under all conditions ( Figure S1C). These results indicated that our APC lines are comparable to (if not better than) natural APCs in priming naïve CD4 + T cells.
Under Th1 polarizing conditions, we found high levels of intracellular IFN-c production in CD4 + T cells stimulated with any APC line, regardless of Notch ligand expression (Figure 2A; individual data points were depicted in Figure S2A-C & Table  S2A). CD4 + T cells primed with splenocytes as APCs produced the highest percentage of IFNc-positive cells detected by intracellular staining (ICS) (78%), closely followed by T cells primed with CHO-DLL1 (64%; P = 0.006) and by CHO-B7 or CHO-Jag1 (47-48%). Similar results were observed when secreted IFN-c was measured by ELISA of the supernatant (detailed in Table S2B). Likewise, under Th2 polarizing conditions, high levels of IL-4 were produced by CD4 + T cells stimulated with any APC line, regardless of Notch ligand expression ( Figure 2B, detailed in Table  S3B). CD4 + T cells primed with CHO-DLL1 and CHO-Jag1 secreted slightly higher amounts of IL-4 compared to the CHO-B7 APC line or splenocytes. These results indicate that Notch ligands can augment cytokine-induced Th1 or Th2 differentiation but cannot interfere with this specification process. In particular, DLL1, suggested to induce Th1 development [30,31], did not redirect T cells toward a Th1 fate under Th2-polarizing conditions. Likewise, Jag1, suggested to induce Th2 development [30], did not re-direct T cells toward a Th2 fate under Th1-polarizing conditions.
Next we examined non-polarizing conditions of activation. Surprisingly, neither DLL1 nor Jag1 had a significant effect on Th1 or Th2 development under either of the non-polarizing conditions compared to the CHO-B7 APC line (Figure 2A, B, and Table S2, 3). When cytokines were neither added nor neutralized, DLL1 expression by APCs did not cause a significant increase in IFN-c production, as would have been predicted from a previous study [30,31]. Similarly, Jag1 expression on APCs did not increase IL-4 production, apparently excluding a role for Notch ligands in biasing Th1/Th2 fate choice. Finally, under conditions where polarizing cytokines were neutralized, neither DLL1 nor Jag1 led to significant changes in IFN-c or IL-4 production (Figure 2A, B), indicating that these ligands are not sufficient for driving Th1 or Th2 fate choice. These data are inconsistent with the results reported for DLL1 and Jag1 expression on DCEK hi7 fibroblasts as artificial APCs [30], which claimed DLL1 and Jag1 to induce Th1 and Th2 development, respectively. Thus, our results would appear to exclude an instructive role for Notch ligands in inducing actual Th1/Th2 development from naïve CD4 + T cells.
The inability of Notch ligands expressed by our APC lines to induce Th1 or Th2 differentiation could have resulted from their inability to activate Notch signaling in naïve CD4 + T cells, despite their demonstrated activity in reporter cells ( Figure 1B). Thus, we tested if Notch1 activation occurred in naïve CD4 + T cells co- cultured with CHO-B7 APC, CHO-DLL1 or CHO-Jag1 by directly measuring the production of NICD1. We found that robust Notch1 activation was induced in CD4 + T cells by CHO-DLL1, but not by the CHO-B7 cell line, as expected ( Figure 2C). However, we found no evidence of Notch1 activation in CD4 + T cells induced by CHO-Jag1 cells ( Figure 2C). Lunatic Fringe (Lfng) can modify the glycosylation pattern of Notch1 receptors (but not Notch2) in a manner that potentiates DLL1-mediated signaling but inhibits Jag1-mediated signaling [71,72,73]. The inability of Jag1 to activate Notch1 in CD4 + T cells but to activate Notch2 signaling in a reporter line ( Figure 1B) would be consistent with Lfng activity rendering their Notch1 receptor insensitive to activation by Jag1. To test this, we measured Lfng protein expression by Western analysis in activated naïve CD4 + T cells ( Figure 2D). Notably, Lfng protein was easily detected in CD4 + T cells under all assay conditions as well as naïve CD4 + T cells from different genetic backgrounds (data not shown). This result could explain the observed lack of effect of Jag1 on Th1 differentiation. In summary, these results confirms that Notch signaling is activated by DLL1 in CD4 + T cells, yet this is insufficient to instruct either Th1 or Th2 fate specification, nor can NICD act to redirect CD4 + T cell differentiation under these conditions ( Figure 2).

Conditional removal of presenilin and RBP-J in CD4 + T cells
The results above demonstrated that Notch activation was not sufficient to instruct T-helper cell fate selection. To test for a genetic requirement for Notch signaling in CD4 + T differentiation along Th1/Th2 lineages, we used a conditional deletion approach to eliminate all presenilin (PS) activity in CD4 + T cells. This is similar to the approach taken in a recent study [74], except that different targeted PS-1 and PS-2 alleles [62] were used. Due to strain or allele differences that may have restricted CD4-Cre expression to the DP stage, we were able to obtain normal numbers of CD4 + T cells from all genotypes, in contrast to the observations reported by Laky and Fowlkes [74]. For our analysis, we used naïve CD4 + T cells isolated from mice of the genotypes CD4-Cre Tg/+ , PS1-1 C/+ , PS2 2/2 , RBP-J c/+ (Het); CD4-Cre Tg/+ , PS1-1 C/C , PS2 2/2 , RBP-J c/+ (PSdko); CD4-Cre Tg/+ , PS1-1 C/+ , PS2 2/2 , RBP-J c/c (Rko) or CD4-Cre tg/+ , and PS1-1 C/C , PS2 2/2 , RBP-J c/c (PSRtko) (See Materials and Methods). We confirmed that complete deletion of targeted alleles occurred in naïve CD4 + T cells purified from PSdko, Rko or PSRtko, lacking expression of either PS-1 or RBP-J as expected ( Figure 3A), whereas normal levels of PS-1 or RBP-J were expressed in heterozygote littermate controls. Next, we asked if Notch signaling is activated during the process of T cell activation in our system. In the presence of natural APCs, control CD4 + T cells showed evidence of Notch1 activation 24 h after stimulation, whereas PSdko CD4 T cells showed no accumulation of NICD1, indicating an absence of Notch activation ( Figure 3B). These controls confirmed that Notch signaling is active during normal T cell activation and that our genetic manipulations have successfully eliminated Notch signaling as intended prior to activation of CD4 + T cell.
An RBP-J independent activity of presenilin contributes to CD4 + T cell expansion We next examined the proliferative capacity of activated CD4 + T cells under both Th1 and Th2 conditions at 48 h by H 3thymidine incorporation assay ( Figure 3C). Over several independent experiments, Notch ligands did not enhance proliferation in the presence of B7 co-stimulation ( Figure 1C), whereas presenilin-deficient (PSdko) and presenlin-and RBP-J-deficient (PSRtko) T cells consistently exhibited lower proliferation ( Figure 3C). Consistent with the proliferation results, cellular expansion, measured by counting the number of viable cells 6 days after activation, was also significantly impaired in PSdko and PSRtko populations ( Figure 3D & E). Specifically, the heterozygote Th1 cultures showed a mean cell number of 8610 6 cells, which was reduced to 4610 6 cells in PSdko cells. The heterozygote Th2 cultures, which showed a mean cell number of 14610 6 cells, was also reduced to 4610 6 cells in PSdko cells. A similar reduction in cellular expansion was observed when stimulated T cells were treated with csecretase inhibitors [58,59]. In contrast, the deletion of RBP-J did not significantly alter cellular expansion in T cells ( Figure 3D, E), in agreement with earlier observations [75]. Interestingly, we saw a significant reduction in cellular expansion of triple mutant (PSRtko) T cells lacking PS-1, PS-2 and RBP-J when compared to control T cells from heterozygous littermates ( Figure 3D and E). This result revealed a requirement for presenilin, but not RBP-J, in T cell expansion, consistent with a role for either a Notch-independent function of c-secretase, or an RBP-J-independent function of Notch. Given that Notch ligand did not alter the level of proliferation induced by OVA peptide (Figure 1C), a Notchindependent function of c-secretase seems most likely to be required for optimal T cell proliferation.
We also examined the expression of the transcription factors Tbet and GATA-3 in these cells 6 days after activation ( Figure 3F and G). Notably, we found that T-bet was expressed by all cells, regardless of genotype ( Figure 3F). In contrast, GATA-3 was present only in T cells exposed to Th2 polarizing conditions, regardless of their genotype ( Figure 3G). The persistence of T-bet and GATA-3 expression under Th1 or Th2 conditions indicated that cytokine stimulation was not dependent on intact Notch signaling. Since other methods of manipulating Notch signaling have been reported to influence these processes in vitro [30,31,56,57,76,77], we next turned our attention to determining the effects of presenilin and RBP-J deficiency on Th1/Th2 fate selection under polarizing conditions. An RBP-J independent presenilin activity regulates the levels of Th1 and Th2 cytokine secretion If the Notch pathway is linear, then deletion of any one of its components should cause the same effect as deletion of any other component. In particular, a linear Notch pathway would predict that deletion of RBP-J [30,75] would result in similar changes as would Notch blockade by c-secretase inhibitors, or even by overexpression of dominant-negative (DN) MAML [78]. However, evidence has been accumulated that this may not be the case. In particular, removal of RBP-J and over-expression of DN-MAML both inhibited Th2 differentiation, but c-secretase inhibitors did not block Th2 differentiation, and instead inhibited Th1 differentiation [77]. Because RBP-J is associated with co-repressors in the absence of a Notch signal [79], this discrepancy was suggested to reflect de-repression of a critical target (e.g. T-bet; [77]). Likewise, activation of the Notch pathway has not led to the same result. For example, over-expression of NICD1 in CD4 + T cells induced T-bet in some studies but not in others [31,77]. In contrast, NICD1 induced IL-4 and GATA-3 in other studies [30,56,57,76]. Thus, there is evidence that the effects of Notch inhibition or activation may be context dependent, perhaps due to the existence of a bifurcation in the pathway.
For these reasons, we compared Th1 and Th2 development in CD4 + T cells that lacked presenilin or RBP-J proteins (Figure 4, 5, 6). We examined the ability of these T cells to generate IFN-c by ICS following differentiation in vitro after activation under both conditions in response to either overnight anti-CD3 treatment ( Figure 4A) or PMA/Ionomycin treatment ( Figure 4B). The percent of IFN-c expressing cells from individual mice is shown for each experiment as a circle or diamond, and the mean of all mice in one group is presented as a bar. First, IFN-c is produced by T cells only under Th1 conditions, and not under Th2 conditions, for littermates from all genotypes. Second, the absence of presenilin or RBP-J did not compromise the acquisition of Th1 fate (assessed as the ability to express IFN-c) under Th1 conditions, measured following re-stimulation with anti-CD3 or with PMA/Ionomycin. In agreement with previous reports, we see an increase in the percent of RBP-J-deficient T cells producing IFN-c under Th1 conditions [30,75]. Thus, RBP-J-deficient CD4 + T cells are capable of Th1 differentiation.
Although we observed a reduction in the percentage of IFN-c positive, presenilin deficient (PSdko) T cells, this difference was not statistically significant when compared to CD4 + T cells isolated from heterozygous littermates. Thus, commitment to Th1 development was still functional in CD4 + T cells lacking presenilin activity (Figure 4), inconsistent with the observations based on pharmacologic inhibition of c-secretase [77]. Furthermore, PSdko CD4 + T cells differentiated under Th1 conditions committed to IFN-c production at a much higher frequency than when differentiated under Th2 conditions, demonstrating their ability to functionally respond to Th1-inducing stimuli. No significant difference was detected between PSRtko and heterozygous T cells, or between PSdko and PSRtko T cells ( Figure 4A, B). In summary, as measured by ICS, cytokine-induced Th1 fate commitment does not require the canonical Notch pathway [78].
Next we examined the ability of these T cells to secrete IFN-c under Th1 and Th2 conditions after re-stimulating an equal number of T cells by either anti-CD3 treatment ( Figure 5A) or PMA/Ionomycin treatment ( Figure 5B). The amount of IFN-c secretion is shown for individual mice and the mean is presented as a horizontal bar. Generally, the results were similar to those obtained by intracellular cytokine staining, although quantitative differences are now evident. First, for all genotypes tested, IFN-c secretion was always significantly higher in Th1 conditions compared to Th2 conditions, indicating that the commitment to Th1 development was generally intact in PSdko T cells and RBP-Jdeficient (Rko) T cells. However, the amount of IFN-c that was secreted from PSdko T cells was reduced by about half, compared to control T cells from heterozygous littermates ( Figure 5A). In contrast, despite an increase in the number of IFN-c positive cells ( Figure 4A), IFN-c secretion from RBP-J-deficient T cells was not statistically different from control T cells ( Figure 5A). To explain the increase in Th1 commitment seen in RBP-J null cells, it was proposed that de-repression of the T-bet gene occurred when RBP-J was removed [77]. If true, removing RBP-J in presenilindeficient T cells (PSRtko) would de-repress T-bet and thus restore IFN-c secretion. Instead, PSRtko T cells also exhibited a similar reduction in IFN-c secretion to PSdko cells when compared to control T cells, even though the level was still far higher than when they were activated under Th2 conditions. In summary, the overall level of Th1 commitment, as assessed by ICS, was not significantly reduced in PSdko or PSRtko cells; however, we observed a variable reduction (50-70%) in the magnitude of IFN-c secretion from T cells deficient in presenilin compared to all other T cells. These results suggest that an RBP-J independent action of c-secretase is necessary to achieve maximal secretion of IFNc from Th1 cells, independent of whether anti-CD3 or PMA/ Ionomycin were used for re-stimulation ( Figure 5).
We also analyzed IL-4 secretion for these genotypes activated under both Th1 and Th2 conditions ( Figure 6). IL-4 was produced by all genotypes of T cells selectively under Th2 conditions, indicating that Th2 fate specification was intact in both presenilindeficient and RBP-J-deficient T cells. Further, RBP-J-deficient T cells showed no significant differences in IL-4 production compared to controls T cells from heterozygous littermates ( Figure 6). Similar to what we saw with IFN-c secretion by Th1 cells, IL-4 secretion from PSdko T cells were somewhat reduced compared to controls. Although this reduction was not statistically significant in comparison to heterozygous littermates, it was statistically significant in comparison with RBP-J deficient T cells ( Figure 6A, B). Similar reduction (,20-40%) was also seen in triple mutant PSRtko T cells. Thus, we uncovered a general, RBP-J independent action of presenilin or c-secretase that contributes to the magnitude of cytokine secretion by differentiated Th1 and Th2 cells.

Discussion
Examples of Notch-dependent decisions in the hematopoietic system include the development of Marginal Zone B cells [80,81,82], multiple steps during T-cell development [63,83,84,85,86,87,88,89,90,91], and vascular development (reviewed in [92]). In the examples above, the dependence of the developmental decision on Notch is similarly revealed independently of the experimental design or genetic background. Due to the linearity of the Notch signaling pathway involved in these decisions, the outcome of Notch inhibition is not influenced by the step at which Notch inhibition occurs (e.g., ligand binding, c-secretase cleavage, association with RBP-J, or assembly of the activation complex). In contrast, the contribution of Notch signaling to peripheral T cell differentiation remains highly controversial because different experimental approaches have resulted in strikingly different outcomes (Table S1) [29]. In particular, the finding most difficult to reconcile with known mechanisms of Notch activation was that specific Notch ligands were able to instruct distinct T-helper fates [30]. These results were provocative because they implied a mechanism of ''ligand memory'' by which different ligands could produce distinct NICD activities in the nucleus.
This study tested the proposal of ''ligand memory'' by recreating artificial APCs that express functional Notch ligands DLL1 and Jag1 that were previously suggested to instruct Th1 or Th2 differentiation, respectively [30]. In our system, we provide clear evidence verifying the ability of Notch ligands to activate Notch-dependent transcription from one or more receptors. Moreover, our controlled experiments were capable of detecting a bias towards Th1 and Th2 differentiation when natural APCs were used (Figure 2A, B). Despite that, neither DLL1 nor Jag1 could instruct or redirect T-helper fate specification in this assay. Specifically, Notch1 and Notch2 activation occurred in response to DLL1, yet DLL1 could not induce Th1 nor inhibit Th2 differentiation ( Figure 2C). Surprisingly, we discovered that Jag1 was incapable of activating Notch1 in naïve CD4 + T cells, most likely due to the presence of Lfng in naïve CD4 + T cells from several genetic backgrounds ( Figure 2D and unpublished data). This finding further diminishes the likelihood that Jag1 plays a significant role in Th2 differentiation through Notch1 activation [71,93]. Although Jag1 could still activate Notch2 in the presence of Lfng, previous studies indicated that NICD1, and not NICD2, triggered robust Th2 responses (Amsen et al., 2003, Amsen et al., 2007Fang et al., 2007).
Some previous studies have induced Notch activation with immobilized Notch ligands [31]. One caveat is that we cannot directly compare the ''level'' of Notch activation produced by APC-expressed ligands with those produced by immobilized DLL1-Fc molecules [31]. Conceivably, higher levels of NICD were achieved in the study by Maekawa [31], which titrated the level of immobilized (i.e., non-physiological) ligand presented to the CD4 + T cells. Regarding the discrepancy between APCs used in different studies, it will be necessary to perform side-by-side comparisons to resolve this apparent paradox.
Instead of identifying a mechanism of ''ligand memory'' capable of instructing T-helper fate, we conclude that Notch is not capable of inducing the initial steps towards fate acquisition. Rather, our data suggest that Notch can, at most, cooperate with cytokines to optimize T-helper differentiation ( Figure 2C & D), consistent with a costimulatory role that has been suggested previously [59,60,78]. The distinct transcriptional activities of NICD during Th1 or Th2 differentiation in vitro could be explained if accessibility to target gene loci was influenced by a previous exposure to cytokine-induced remodeling, independent of Notch ligands.
We find that removal of intact Notch signaling by targeted deletion of Presenilins and/or RBP-J genes does not prevent cytokine-induced Th1/Th2 fate specification (Figure 4, 5, 6). This finding would be inconsistent with a genetic requirement for Notch signaling during the initial fate selection. Since the T-helper differentiation program is poised for rapid execution in response to multiple cues, the choice of experimental system and the ''strength'' of the stimulus will impact the amplitude of the contribution Notch makes. Our results suggest that intact Notch signaling may act to allow cells to attain the maximum level of commitment when insufficiently stimulated, highlighted by the physiological requirement of RBP-J for optimal Th2 response in T cells primed by parasite-exposed APC [56]. Although these reports indicate that Notch signaling may augment GATA-3 transcription in vivo, they did not demonstrate a specific role for Jagged in this process [56,57], nor did they ask if stronger T-cell activation would bypass the need for Notch, as would be predicted by a co-stimulatory function of Notch in this process.
The present study is one of a few studies that have compared the effects of disrupting the Notch pathway during T-helper differentiation at two independent positions in the pathway. Importantly however, this is the only study that has performed epistatic analysis of c-secretase and RBP-J deficiencies in this system. As defective T cell expansion and cytokine production were not observed in the absence of RBP-J alone, this epistatic analysis has revealed unexpected RBP-J-independent (and perhaps Notch-independent) functions of presenilins in regulating these processes. One function, regulating optimal T cell expansion, was reported by others [58,59]. Given that proliferation is necessary to attain optimal T-helper differentiation [67,68,69,70], a defect in proliferation may have led to the misinterpretation of c-secretase function in T-helper cell specification. The second function, observed under all the experimental conditions we deployed in this study, contributes to the production or secretion of cytokines from committed, differentiated T-helper cells. Since we have inactivated c-secretase by deleting the Presenilin genes, one or both of the above functions may reflect a protease-independent activity of presenilin [94,95]. Distinguishing between these two presenilin activities will require epistatic analyses with the other c-secretase components, but that is beyond the scope of this study.
In summary, the data presented here are inconsistent with the instructive role of Notch in Th1/Th2 fate specification. Instead, separable functions of Notch and presenilin as genetic modifiers of T-helper cell differentiation pathway(s) can account for the dependence of published conclusions on genetic background and experimental systems [29]. Despite clues pointing to TCR activation as the pathway modified by Notch and/or presenilin [74], this remains an important open question that will have to be addressed experimentally. Table S1 Proposed roles of Notch signaling in peripheral T cell development. Table 1 summarized the proposed regulatory roles of Notch signaling in T cell activation/proliferation (A) and Th1/ Th2 differentiation (B). The references are labeled with either Arabic or Roman numeral to highlight their conflicting conclusions regarding the functions of Notch in these processes. Found at: doi:10.1371/journal.pone.0002823.s003 (0.08 MB DOC)

Table S2
Summary of the level of IFN-c produced (A) Percent of CD4+ T cells stained positive for intracellular IFN-c from three independent APC-primed experiments described in Fig. 1 & 2.
The ICS values were presented as flow cytometry plot in Supplemental Fig. 2A-C. The mean and standard deviation was calculated using the percent of IFN-c positive cells and presented graphically in Fig. 2A. (B) The level of IFN-c secreted by T cells activated with various APCs lines under different polarizing conditions. Equal numbers of T cells were re-stimulated on Day 7 with anti-CD3 for 24 hr. The supernatant was harvested and the level of secreted cytokines was measured with ELISA. Found at: doi:10.1371/journal.pone.0002823.s004 (0.08 MB DOC)

Table S3
Summary of the level of IL-4 produced (A) Percent of CD4+ T cells stained positive for intracellular IL-4 from three independent APC primed experiments described in Fig. 1 & 2.
The ICS values were presented as flow cytometry plots in Supplemental Fig. 2A-C. Note that the values of IL-4 ICS are highly variable and are dependent on the types of conjugated antibodies used. (B) The level of IL-4 secreted by T cells activated with various APCs lines under different polarizing conditions. ELISA of IL-4 was carried out as described in Supplemental Table  2B. The mean and standard deviation was calculated and presented graphically in Fig. 2B.