Research ArticleCancer

Safe targeting of T cell acute lymphoblastic leukemia by pathology-specific NOTCH inhibition

See allHide authors and affiliations

Science Translational Medicine  29 May 2019:
Vol. 11, Issue 494, eaau6246
DOI: 10.1126/scitranslmed.aau6246

The right presenilin for the job

The NOTCH signaling pathway is frequently mutated in T cell acute lymphoblastic leukemia and therefore presents a potential therapeutic target. Previous researchers tried to inhibit γ-secretase, a protease that cleaves NOTCH and thereby activates it. Unfortunately, nonspecific inhibition of γ-secretase proved to be too toxic for clinical use because of on-target side effects in a variety of healthy tissues. Habets et al. found that presenilin-1, a component of some γ-secretase complexes, is highly expressed in T cell leukemia relative to healthy T cells. The authors then demonstrated that targeting presenilin-1 is effective and safe in mouse models, suggesting this strategy’s potential for translation.


Given the high frequency of activating NOTCH1 mutations in T cell acute lymphoblastic leukemia (T-ALL), inhibition of the γ-secretase complex remains an attractive target to prevent ligand-independent release of the cytoplasmic tail and oncogenic NOTCH1 signaling. However, four different γ-secretase complexes exist, and available inhibitors block all complexes equally. As a result, these cause severe “on-target” gastrointestinal tract, skin, and thymus toxicity, limiting their therapeutic application. Here, we demonstrate that genetic deletion or pharmacologic inhibition of the presenilin-1 (PSEN1) subclass of γ-secretase complexes is highly effective in decreasing leukemia while avoiding dose-limiting toxicities. Clinically, T-ALL samples were found to selectively express only PSEN1-containing γ-secretase complexes. The conditional knockout of Psen1 in developing T cells attenuated the development of a mutant NOTCH1-driven leukemia in mice in vivo but did not abrogate normal T cell development. Treatment of T-ALL cell lines with the selective PSEN1 inhibitor MRK-560 effectively decreased mutant NOTCH1 processing and led to cell cycle arrest. These observations were extended to T-ALL patient-derived xenografts in vivo, demonstrating that MRK-560 treatment decreases leukemia burden and increased overall survival without any associated gut toxicity. Therefore, PSEN1-selective compounds provide a potential therapeutic strategy for safe and effective targeting of T-ALL and possibly also for other diseases in which NOTCH signaling plays a role.


γ-Secretases are a group of widely expressed intramembrane-cleaving proteases. The enzymes process clinically relevant substrates such as amyloid precursor protein (APP) and NOTCH and have been explored as drug targets in Alzheimer’s disease, cancer, and other disorders (1). However, the clinical use of γ-secretase inhibitors (GSIs) has been hampered by severe mechanism-based dose-limiting toxicity, predominantly intestinal goblet cell hyperplasia and concomitant severe diarrhea (2), thymus atrophy and ablated T cell development (3, 4), splenic marginal zone atrophy (5), and skin lesions (6). This toxicity is due to systemic NOTCH inhibition because NOTCH receptors require γ-secretase complex–mediated processing for the release and nuclear translocation of the NOTCH intracellular domain (NICD) to activate gene expression (7).

It is, however, important to realize that the enzymatic activity of the γ-secretase complex reflects the combined activity of at least four different complexes. Clinical trials have been performed with broad-spectrum, nonselective GSIs that target all four complexes equally. The question remains whether inhibition of some but not all γ-secretase complexes, by subunit-selective targeting, could provide a way forward to safe targeting.

The four different γ-secretase complexes each contain one nicastrin (NCSTN) and one presenilin enhancer 2 (PEN-2) subunits. In addition, two different APH-1 proteins, APH-1A or APH-1B, and two different presenilin (PSEN) proteins, PSEN1 or PSEN2, exist. One of the APH-1 and one of the PSEN proteins combine with the two stable subunits to generate four subcomplexes, which differ in their PSEN subunit and/or APH-1 subunit (8, 9). Different γ-secretase complexes can be expressed simultaneously by the same cells and at the same time (1, 10) but differ in their subcellular distribution and their biological function (1113). Gene targeting of the different complexes revealed a variety of different outcomes, ranging from very severe embryonic lethal phenotypes (7, 1416), to subtle behavioral phenotypes (17, 18), to normal mice (19).

The γ-secretase complexes have been mainly investigated in the context of Alzheimer’s disease because they process the APP to generate different amyloid-β profiles (12). In addition, GSIs have been investigated as targeted therapeutics for T cell acute lymphoblastic leukemia (T-ALL) (20), an aggressive hematologic malignancy resulting from the transformation of immature T cell progenitors (21, 22). In T-ALL, activating NOTCH1 mutations are the most common mutations observed, present in about 60% of all cases (2325). These mutations result in increased and ligand-independent oncogenic NOTCH1 signaling, promoting T cell transformation through the physiologic functions of NOTCH1 in the thymus (24, 26). However, like physiologic NOTCH1 signaling, oncogenic mutant NOTCH1 signaling still requires γ-secretase processing for activation, providing the rationale for GSIs as a therapeutic approach for T-ALL (20, 2729).

Beyond T-ALL, increasing evidence is supporting a pathogenic role for NOTCH gain-of-function mutations in solid tumors. These cancers often display inappropriate NOTCH signaling due to overexpression of NOTCH receptors and/or ligands or loss of negative regulation of NOTCH signaling (30). Because of this, GSIs were also investigated for a number of hematological and solid cancers, including breast cancer, pancreatic cancer, glioma, non–small cell lung cancer, and colorectal cancer (30). However, despite extensive research to develop several classes of GSIs, none of them sufficiently spared physiologic NOTCH signaling, and none of these drugs was successful in clinical trials (6). This has resulted in a halt to the therapeutic development of GSIs in these areas. The hypothesis that selective inhibition of one of the γ-secretase complexes alone, targeting the enzyme most active in tumors while sparing other γ-secretase complexes in the hope of preserving physiological NOTCH signaling in the healthy tissues, however, has not been tested.

Here, we wanted to address whether selective γ-secretase inhibition could be a valid therapeutic approach. Given the high frequency of NOTCH1 mutations that result in ligand-independent but γ-secretase–dependent signaling in T-ALL, this cancer provides a perfect model to evaluate this hypothesis. We selectively targeted the PSEN1-containing γ-secretase complexes while leaving the PSEN2-containing complexes untargeted. By doing so, we observed therapeutic efficacy against T-ALL without the toxicity inherent to broad-spectrum GSIs.


PSEN1 is highly expressed in T-ALL and regulates NOTCH1 cleavage

The PSEN subunits PSEN1 or PSEN2 provide the catalytic center of the different γ-secretase complexes (Fig. 1A). Microarray analysis revealed that both PSEN1 and PSEN2 are expressed during human T cell development (31), although PSEN1 has about fourfold higher expression compared to PSEN2 (Fig. 1B). However, PSEN1 expression is >30-fold higher than PSEN2 expression in T-ALL cell lines and primary T-ALL patient samples (Fig. 1C) (32, 33). NOTCH1 signaling is essential for early T cell development in both mice and humans (26, 34); therefore, we examined whether Psen1 loss would abrogate normal T cell development in mice. Previously, a conditional Psen1 knockout mouse was generated by targeting exon 2 and exon 3 (35). We developed a conditional Psen1 knockout mouse targeting exon 1, using a DNA construct that also contained a tag (Psen1f/f) (fig. S1A) and then crossed it with a CD2Cre transgenic mouse (36) to inactivate Psen1 in developing T cells (CD2CrePsen1Δ/Δ mice). The introduction of loxP sites and a tag into the Psen1 gene locus alone caused a 50% decrease in PSEN1 protein expression in thymocytes but did not result in major T cell defects (Fig. 1, D to F, and fig. S1, B to D). The complete loss of PSEN1 expression in thymocytes of CD2CrePsen1Δ/Δ mice did not alter the frequency of major T cell populations in the thymus (Fig. 1, D to F). Although a small compensatory increase of PSEN2 expression was seen after loss of PSEN1, this was insufficient to overcome the overall loss in γ-secretase complex formation assessed by the strong reduction in NCSTN maturation that otherwise only occurs when a functional complex is present (fig. S1, B and C). To study the possible decrease in T cell proliferation, we generated ex vivo cultures of mouse pro-T cells derived from C57BL/6 wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ mice (fig. S2A). These pro-T cells are cultured on DLL4 (NOTCH1 ligand)–coated plates and are strictly dependent on DLL4-induced NOTCH signaling (37). Pro-T cells derived from CD2CrePsen1Δ/Δ mice showed comparable proliferation to controls (Fig. 1G), demonstrating that PSEN1 is dispensable for normal T cell development. Apparently, the presence of PSEN2 is sufficient to maintain adequate NOTCH signaling to support normal development.

Fig. 1 PSEN1 deletion specifically reduces oncogenic mutant NOTCH1 signaling in T cells.

(A) Schematic representation of the four different γ-secretase complexes that exist in humans. All complexes contain nicastrin (NCSTN), PEN-2, and either APH-1A or APH-1B and PSEN1 or PSEN2. (B) PSEN1 (probe X203460_s_at) and PSEN2 (probe X211373_s_at) expression in sorted subsets of human thymocytes. Average expression and SD in samples obtained from two different donors are shown. (C) RNA sequencing (RNA-seq) fragments per kilobase of transcript per million (FPKM) PSEN1 and PSEN2 expression in T-ALL cell lines (ALL-SIL, CCRF-CEM, DND-41, HPB-ALL, HSB2, Jurkat, Karpas45, KE37, LOUCY, MOLT14, MOLT4, P12-Ichikawa, PEER, PF382, RPMI-8402, RM2, SUPT1, SUPT13, and T-ALL1) and T-ALL patient samples. (D) Representative flow cytometry plots of thymocyte populations stained with antibodies against CD4 and CD8 or CD44 and CD25 in C57BL/6 wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ mice. (E) Quantification of mature thymic T cell populations in relative numbers (wild type, n = 5; Psen1f/f, n = 4; CD2CrePsen1Δ/Δ, n = 4). DP, double-positive; DN, double-negative. (F) Quantification of immature thymic T cell populations in relative numbers (wild type, n = 5; Psen1f/f, n = 4; CD2CrePsen1Δ/Δ, n = 4). (G) Relative cell numbers in ex vivo cultures of mouse pro-T cells derived from C57BL/6 wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ mice and grown for 7 days (n = 3 for all). (H) Relative cell numbers in ex vivo cultures of mouse pro-T cells derived from C57BL/6 wild-type (n = 2), Psen1f/f (n = 3), or CD2CrePsen1Δ/Δ (n = 3) mice transduced with MSCV-NOTCH1-L1601P-ΔP-IRES-GFP and grown for 7 days without DLL4. (I) Western blot analysis of C57BL/6 wild-type (n = 4), Psen1f/f (n = 4), or CD2CrePsen1Δ/Δ (n = 3) NOTCH1-L1601P-ΔP pro-T cells for NICD1, PSEN1, PSEN2, NCSTN, and β-actin cultured in the absence of DLL4 for 72 hours. Quantification of NICD1 expression is shown on the right. NICD1, NOTCH1 intracellular domain; Mat, mature; Immat, immature; CTF, C-terminal fragment. All graphs show the mean values, and error bars represent SD. P values for (C) were calculated using two-tailed Student’s t test. ****P ≤ 0.0001. P values in (E) to (I) were calculated using one-way analysis of variance (ANOVA). *P ≤ 0.05, ***P ≤ 0.001, and ****P ≤ 0.0001.

We next investigated whether Psen1 gene inactivation could affect processing of mutant forms of the NOTCH1 receptor found in T-ALL. We transduced the mouse pro-T cells established from C57BL/6 wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ mice with a constitutively active mutant NOTCH1 receptor, containing a heterodimerization and PEST domain mutation (NOTCH1-L1601P-ΔP). This mutant NOTCH1 receptor provides a substrate for γ-secretase, which is independent of ligand stimulation (24). C57BL/6 wild-type and Psen1f/f-derived pro-T cells transduced with mutant NOTCH1 receptors were able to survive and proliferate without DLL4, confirming that the NOTCH1-L1601P-ΔP mutant can signal in the absence of DLL4. PSEN1 deletion significantly (P ≤ 0.05) reduced pro-T cell proliferation and survival of NOTCH1-L1601P-ΔP–expressing cells (Fig. 1H) and decreased mutant NOTCH1 receptor processing and NICD1 formation (Fig. 1I). Together, these data indicate that loss of PSEN1 specifically reduces oncogenic NOTCH1 signaling in T cells while having limited effect on physiologic DLL4-mediated NOTCH signaling.

Genetic deletion of PSEN1 prolongs survival in NOTCH1-induced T cell leukemia in vivo

Given that loss of PSEN1 perturbs the processing and downstream signaling of mutant NOTCH1 receptors involved in T-ALL, we next assessed whether Psen1 deletion could also impair mutant NOTCH1-induced T-ALL development in vivo. To this end, we transplanted wild-type C57BL/6 mice with syngeneic wild-type C57BL/6, Psen1f/f, or CD2CrePsen1Δ/Δ hematopoietic progenitors, expressing equivalent amounts of ΔEGF-NOTCH1-L1601P-ΔP, inferred from the extent of green fluorescent protein (GFP) expression, after retroviral transduction. (Fig. 2, A and B, and fig. S2B). Mice transplanted with wild-type or Psen1f/f cells transduced with ΔEGF-NOTCH1-L1601P-ΔP showed circulating GFP-positive (GFP+) CD4+CD8+ T cells at 6 weeks after transplantation, indicative of leukemia development (Fig. 2, C and D). Mice transplanted with CD2CrePsen1Δ/Δ cells transduced with ΔEGF-NOTCH1-L1601P-ΔP were largely devoid of circulating GFP+ T cells at 6 weeks after transplantation. At 9 weeks, mice transplanted with wild-type and Psen1f/f cells displayed splenomegaly and thymic enlargement due to CD4+CD8+ leukemic cell infiltration, which was still absent in mice transplanted with CD2CrePsen1Δ/Δ cells (Fig. 2, E and F, and fig. S2C). These differences were not due to impaired homing and engraftment after transplantation of the CD2CrePsen1Δ/Δ hematopoietic progenitors (fig. S2D). Eventually, only 9 of 17 mice transplanted with CD2CrePsen1Δ/Δ progenitors developed CD4+CD8+ leukemia with a median overall survival of 120 days compared to a median survival of 73 and 72 days in wild-type and Psen1f/f transplanted mice, respectively (P < 0.0001; Fig. 2, G and H). These data show that deficiency of PSEN1 alone is sufficient to majorly affect murine T-ALL induction.

Fig. 2 PSEN1 deletion impairs mutant NOTCH1-induced T-ALL development.

(A) Schematic of primary bone marrow transplant using hematopoietic stem and progenitor cells from C57BL/6 wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ donor mice transduced with MSCV-ΔEGF-NOTCH1-L1601P-ΔP-IRES-GFP transplanted into sublethally irradiated wild-type C57BL/6 recipient mice. (B) Representative flow cytometry plots for the numbers of GFP+-transduced lineage-negative cells before injection into recipient mice. (C) Representative flow cytometry plots for the numbers of circulating GFP+ cells in peripheral blood from mice transplanted with wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP 6 weeks after transplantation. Quantification of the GFP+ population is shown on the right (n = 9, 14, and 14 mice, respectively). SSC, side scatter. (D) Representative flow cytometry plots of peripheral blood from mice transplanted with wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP stained with antibodies to CD4 and CD8 6 weeks after transplantation. Quantification of the CD4+CD8+ double-positive population is shown on the right (n = 9, 14, and 14 mice, respectively). APC, allophycocyanin; PE, phycoerythrin. (E) Size and weight of spleens from mice transplanted with wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP analyzed 9 weeks after transplantation (n = 3). (F) Quantification of GFP+ cells in spleens from mice transplanted with wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP analyzed 9 weeks after transplantation (n = 3). (G) Kaplan-Meier survival curves of mice transplanted with wild-type, Psen1f/f, or CD2CrePsen1Δ/Δ progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP. All graphs show the mean values, and error bars represent SD. (H) Representative flow cytometry of end-stage leukemia in the different cohorts of mice. P values in (C) to (F) were calculated using one-way ANOVA. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001. P value in (F) was calculated using the log-rank test. ****P ≤ 0.0001.

Next, we set out to elucidate whether Psen1 loss in an already established leukemia could still impair disease progression. To this end, we generated inducible Psen1 conditional knockouts by crossing Psen1f/f to Rosa26Cre-ERT2 mice, producing R26Cre-ERT2Psen1f/f mice. C57BL/6 mice were then transplanted with R26Cre-ERT2Psen1f/f hematopoietic progenitors transduced with ΔEGF-NOTCH1-L1601P-ΔP. After successful engraftment, Psen1 was specifically deleted in transplanted donor cells through tamoxifen treatment (100 mg/kg for 5 days). These experiments were performed in primary recipient mice and were also repeated in tertiary transplanted recipient mice to assess whether loss of PSEN1 in an aggressive leukemia setting still hampers disease progression (Fig. 3A). Analysis of R26Cre-ERT2Psen1f/f leukemic cells recovered after tamoxifen treatment showed complete ablation of PSEN1 (fig. S3A). Specific Psen1 deletion in the donor cells reduced leukemia burden, assessed by the fraction of circulating GFP+ cells in the peripheral blood, by twofold at 5 weeks after transplantation and by more than sixfold at 9 weeks after transplantation in the primary recipient mice, compared to vehicle-treated mice (Fig. 3B). One week after tamoxifen treatment, analysis of age-matched mice showed that Psen1 deletion reduced splenomegaly by 50% compared to vehicle-treated mice (Fig. 3C). Furthermore, Psen1 deletion by tamoxifen treatment in an established leukemia increased the median overall survival to 142 days compared to 84.5 days for vehicle-treated mice (P < 0.0001; Fig. 3D). The tamoxifen-treated mice that did develop leukemia had lost PSEN1, excluding the possibility of having escaped recombination but did show robust PSEN2 expression (fig. S3B). Tamoxifen treatment had no effect on leukemia progression and overall survival in mice transplanted with wild-type C57BL/6 tumor cells expressing Psen1 endogenously compared to vehicle controls (fig. S3C).

Fig. 3 Genetic targeting of PSEN1 impairs mutant NOTCH1 leukemia maintenance.

(A) Schematic of primary bone marrow transplant using hematopoietic stem and progenitor cells from R26Cre-ERT2Psen1f/f donor mice transduced with MSCV-ΔEGF-NOTCH1-L1601P-ΔP-IRES-GFP before transplantation into sublethally irradiated wild-type C57BL/6 recipient mice. Once engraftment was confirmed (3 weeks after transplantation), primary transplant mice were either treated with tamoxifen (100 mg/kg per day) (or vehicle) by intraperitoneal (IP) injection for five consecutive days to delete PSEN1 specifically in transplanted donor cells or used for secondary/tertiary transplants before treatment with tamoxifen (100 mg/kg per day) by intraperitoneal injection for five consecutive days to delete PSEN1. (B) The numbers of circulating GFP+ cells in peripheral blood from mice transplanted with R26Cre-ERT2Psen1f/f progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP after tamoxifen or vehicle treatment 5 weeks (n = 13 and n = 14 mice, respectively) and 9 weeks after transplant (n = 9 and n = 11 mice, respectively). (C) Spleen weight and quantification of GFP+ cells in spleens from mice transplanted with R26Cre-ERT2Psen1f/f progenitors expressing ΔEGF-NOTCH1-L1601P-ΔP 5 weeks after transplant, treated with either vehicle or tamoxifen (n = 3 mice). (D) Kaplan-Meier survival curves of primary R26Cre-ERT2Psen1f/f transplant mice expressing ΔEGF-NOTCH1-L1601P-ΔP treated with either vehicle or tamoxifen. (E) Kaplan-Meier survival curves of tertiary R26Cre-ERT2Psen1f/f mice expressing ΔEGF-NOTCH1-L1601P-ΔP treated with either vehicle or tamoxifen. Gray boxes represent the treatment period. All graphs show the mean values, and error bars represent SD. P values in (B) and (C) were calculated using two-tailed Student’s t test. P values in (D) and (E) were calculated using the log-rank test. **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001. Vehicle: corn oil.

Last, we tested whether Psen1 was required for leukemia maintenance and progression by targeting Psen1 in a more aggressive and developed leukemia. Tertiary recipients were transplanted with leukemic cells from mice that suffered from full-blown leukemia, and these tertiary recipients were treated with vehicle or tamoxifen. Even in this aggressive leukemia model, Psen1 deletion significantly increased survival (P < 0.001; Fig. 3E). Together, these data show the importance of PSEN1 in leukemia development and maintenance and validate PSEN1 as a potential target for therapy in T-ALL with NOTCH1 mutations.

Pharmacologic inhibition of PSEN1 impairs leukemia progression and prolongs survival in vivo

The decrease in leukemic burden identified through genetic loss of Psen1 prompted us to investigate whether selective pharmacological PSEN1 inhibition would be a viable strategy for T-ALL treatment. To this end, we tested whether the PSEN1-selective inhibitor MRK-560, which has ~100-fold selectivity for PSEN1 over PSEN2 (fig. S4) (38), could block mutant NOTCH1 receptor signaling in human T-ALL cell lines. MRK-560 treatment reduced NICD1 generation in HPB-ALL, DND-41, and Jurkat cell lines and resulted in a dose-dependent decrease of proliferation in HPB-ALL and DND-41, which depend on NOTCH signaling for their survival (Fig. 4, A and B). Jurkat T-ALL cells harbor a PTEN deletion and are not dependent on NOTCH signaling for their proliferation, which explains why these cells do not show a decrease in survival upon MRK-560 treatment (Fig. 4B and table S1) (28, 39, 40). The effect of the PSEN1-selective GSI MRK-560 on proliferation could be attributed to a block in cell cycle (Fig. 4, C and D).

Fig. 4 PSEN1-selective inhibition impairs T-ALL cell proliferation.

(A) Western blot analysis for NICD1 in HPB-ALL, DND-41, and Jurkat cells in response to MRK-560 and DAPT (N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester) treatment. (B) Proliferation of NOTCH-dependent HPB-ALL and DND-41 cells and NOTCH-independent Jurkat cells in response to MRK-560 (n = 3). (C) Cell cycle analysis of HPB-ALL cells treated with increasing doses of MRK-560. Edu, 5-ethynyl-2′-deoxyuridine. (D) Quantification of the percentages of HPB-ALL and DND-41 cells in G1 phase upon treatment with increasing doses of MRK-560. (E) Kaplan-Meier survival curves for mice with ΔEGF-NOTCH1-L1601P-ΔP–induced leukemia treated with vehicle or MRK-560 (30 μmol/kg per day) for 14 days. Gray boxes represent the treatment period. Graphs show the mean values, and error bars represent SD. P values in (D) were calculated using two-way ANOVA. P values in (E) were calculated using the log-rank test. *P ≤ 0.05, ***P ≤ 0.001. DMSO, dimethyl sulfoxide.

Next, we determined whether pharmacological PSEN1 inhibition impaired T-ALL in vivo. ΔEGF-NOTCH1-L1601P-ΔP T-ALL lymphoblasts were injected into secondary recipients and treated with MRK-560 (30 μmol/kg) or vehicle for 14 days. MRK-560 treatment resulted in strong antileukemic effects and improved median survival to 30 days compared to 18 days in vehicle-treated mice (P = 0.0009; Fig. 4E). These data show marked therapeutic effects for pharmacological PSEN1 inhibition in T-ALL models in vitro and in vivo.

Having validated selective PSEN1 inhibition in a NOTCH1-driven mouse leukemia model, we next investigated the efficacy of PSEN1 targeting in human patient-derived xenograft (PDX) in vivo models. Immunodeficient NOD.Cg-Prkdscid Il2rgtm1WjlH2-Ab1tm1Gru Tg(HLA-DRB1)31Dmz/Szj (NSG) mice were injected with four genetically different PDX T-ALL samples with different NOTCH1 mutations (table S1). Mice were randomized into vehicle and MRK-560 treatment arms and were treated with MRK-560 (30 μmol/kg) or vehicle for 14 days by subcutaneous injection. MRK-560 treatment significantly reduced leukemia burden compared to vehicle-treated mice, as assessed by peripheral blood counts of human CD45+ cells and in vivo bioluminescence (P ≤ 0.01) (Fig. 5, A to D, and fig. S5, A and B). The various degrees of response to MRK-560 treatment were not explained by differences in PSEN1 expression in the PDX models (fig. S5C).

Fig. 5 Pharmacological PSEN1 inhibition attenuates leukemia in PDX.

Percentages of human CD45+ cells in peripheral blood from mice transplanted with patient samples (A) 389E, (B) XC63, and (C) XC65 treated with vehicle or MRK-560 (30 μmol/kg per day) for 14 days. (D) Bioluminescence images depicting mice with leukemia burden closest to the mean among animals transplanted with patient sample XC65 and treated with vehicle or MRK-560 for 14 days. Quantification is shown in the right panel. (E to G) Quantification of human CD45+ cells in spleen and overall spleen weight from mice transplanted with patient samples (E) 389E, (F) XC63, and (G) XC65 after 14 days of treatment with vehicle or MRK-560 (n = 3 for 389E and n = 4 for XC63 and XC65). (H to J) Kaplan-Meier survival curves of mice transplanted with patient samples (H) 389E, (I) XC63, and (J) XC65 and treated with vehicle or MRK-560 for 14 days. Treatment of XC65 with MRK-560 and survival analysis began once leukemia burden was shown to be high in the blood (~50% hCD45+ cells). (K) Quantitative polymerase chain reaction (qPCR) analysis of DTX1, NOTCH3, and MYC expression in splenocytes from mice transplanted with patient sample XC63 or XC65 after 14 days of vehicle or MRK-560 treatment (n = 4). Graphs show the mean values, and error bars represent SD. Gray boxes represent the treatment period. P values in (A) to (D) were calculated using two-way ANOVA and two-tailed Student’s t test. P values in (E), (F), (G), and (K) were calculated using two-tailed Student’s t test. P values in (H) to (J) were calculated using the log-rank test. **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001. Vehicle: 20% hydroxypropyl-β-cyclodextrin (HPβCD) in 0.1 M meglumine. CNRQ, calibrated normalized relative quantity.

Analysis of age-matched vehicle- versus MRK-560–treated mice at the end of a 2-week treatment period showed up to 60% reduction in splenomegaly and up to 40% reduction in leukemic cell infiltration in the spleens of MRK-560–treated mice (Fig. 5, E to G). Human leukocyte antigen staining revealed a marked reduction of human leukemia cells in the spleen of MRK-560–treated animals compared to vehicle-treated mice, associated with reduced Ki67 staining, in line with reduced leukemia burden and inhibition of proliferation (fig. S5, D and E). MRK-560 treatment significantly prolonged survival compared to vehicle-treated mice in all three NOTCH1 mutant T-ALL patient samples tested here (1.1-fold, P = 0.0001 for 389E; 1.5-fold, P = 0.0011 for XC63; and 1.6-fold, P = 0.0027 for XC65; Fig. 5, H to J). Single agent MRK-560 treatment was still effective even when treatment was initiated in mice with a high leukemia burden (“curative setting”), which resembles the clinical setting where disease is only detected when blood counts start to change (Fig. 5, C and J). In these conditions, we observed a 1.5-fold increased survival compared to vehicle-treated mice (P = 0.0011; Fig. 5J). Moreover, relapsing mice also remained sensitive to a second round of MRK-560 treatment, indicating no short-term outgrowth of an overtly resistant leukemia clone (fig. S5F). Analysis of leukemic cells after MRK-560 treatment showed that canonical NOTCH target genes MYC, DTX1, and NOTCH3 were down-regulated in T-ALL cells from MRK-560–treated mice compared to vehicle-treated mice (Fig. 5K), confirming on-target effects for MRK-560.

Pharmacological PSEN1-selective targeting does not cause gastrointestinal toxicity or T cell developmental defect

The major hurdle in adopting GSIs clinically has been the on-target NOTCH-related toxicity, resulting in severe gastrointestinal goblet cell hyperplasia or defective T cell development. Treatment with the PSEN1-selective GSI MRK-560 did not cause any pathological changes in the gastrointestinal architecture or increased numbers of secretory goblet cells, assessed by periodic acid–Schiff (PAS) staining (Fig. 6A). In contrast, treatment with the classical broad-spectrum GSI dibenzazepine (DBZ; 10 μmol/kg) resulted in a fourfold increase in the number of secretory goblet cells, characteristic of gastrointestinal toxicity due to systemic NOTCH inhibition, as previously reported (2). Moreover, 4-week treatment of wild-type mice with intermittent dosing of DBZ, as described previously (41), was still more toxic compared to MRK-560 (fig. S6). The expression of PSEN1 and PSEN2 within the human small intestine is nearly equivalent (Fig. 6B), suggesting that, in the absence of PSEN1, the activity of PSEN2 may be sufficient to maintain normal gastrointestinal physiology. Furthermore, immunophenotyping revealed no major defects in thymic T cell development in healthy C57BL/6 mice treated with MRK-560 for 14 days, with cells progressing normally into CD4+CD8+ T cells (Fig. 6, C to E), in line with the genetics experiments shown in Fig. 1. In contrast, mice treated with DBZ showed defective T cell development, evidenced by a marked reduction in CD4+CD8+ T cells and an increase in DN1 (double-negative) stage T cells, as previously reported (26) for classical GSIs. Together, these data provide evidence for a clear therapeutic window for selective PSEN1 targeting in T-ALL.

Fig. 6 Pharmacological selective PSEN1 targeting does not cause gastrointestinal toxicity or T cell developmental defects.

(A) Representative images of PAS staining of intestines from mice treated with vehicle, MRK-560, or the broad-spectrum GSI DBZ for 14 days to assess the number of secretory goblet cells. Scale bars, 100 μm. Quantification of the number of goblet cells per centimeter of villus is shown on the right (vehicle, n = 4; MRK-560, n = 4; and DBZ, n = 3). (B) Human PSEN1 and PSEN2 gene expression in small intestine. Data for this analysis were obtained from the genotype-tissue expression (GTEx) portal and database of genotypes and phenotypes (dbGaP) accession number phs000424.vN.pN. (C) Representative flow cytometry plots of thymocyte populations stained with antibodies to CD4 and CD8 or CD44 and CD25 in C57BL/6 mice treated with vehicle, MRK-560, or DBZ for 14 days. TPM, transcripts per million. (D) Quantification of intrathymic T cell populations in relative numbers (vehicle, n = 5; MRK-560, n = 5; and DBZ, n = 3). All graphs show the mean values, and error bars represent SD. The P values in (A) and (D) were calculated using one-way ANOVA. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.


Activating NOTCH1 mutations are found in about 60% of patients with T-ALL and across all T-ALL subtypes (24, 32). Therefore, targeting NOTCH1 by using GSIs that block NOTCH processing and activation has been of continuing clinical interest in the treatment of T-ALL. The gastrointestinal toxicity inherent to general GSIs is partially ameliorated with intermittent dosing regimens; nonetheless, the most recent clinical trials were still halted prematurely because of severe adverse effects (2, 6, 42). Here, we deliver proof of concept that selective targeting of PSEN1 γ-secretase complexes provides a viable and attractive alternative therapeutic approach. Both genetic and pharmacological experiments showed consistently that PSEN1 targeting alone is sufficient to strongly mitigate leukemia development, both in mutant NOTCH1-driven leukemia mouse models and in human PDX models. Using a tamoxifen-inducible model for Psen1 deletion, we showed that these effects were also observed when Psen1 was targeted after leukemia had developed, which more closely resembles the clinical situation. The beneficial effects of pharmacological PSEN1 targeting were also observed when treatment was initiated at high blast count, which is classified clinically as higher risk with worse prognosis (43, 44). These data show that selective targeting of PSEN1 γ-secretase complexes provides a potent approach with high antileukemic activity, comparable to complete γ-secretase inhibition (45). These data are supported by an earlier observation that PSEN1 is required for the development of a DLL4-driven T cell lymphoma on a PSEN2-null background (46). Although that work demonstrated that PSEN1 is necessary, it did not show that it is sufficient to inhibit PSEN1 alone to block the lymphoma. Some mice in our experiments, with confirmed PSEN1 deletion, still developed leukemia, likely due to the observed compensatory high PSEN2 expression in their cells. Therefore, although PSEN2 is not expressed in human T-ALL, reactivation of PSEN2 expression might be a potential resistance mechanism. Similarly, γ-secretase inhibition can reduce the frequency of leukemic stem cells because mutant NOTCH1 signaling plays an important role in their maintenance (4750). Although not tested here, our findings suggest that selective PSEN1 targeting might also have the promise to target these quiescent and therapy-resistant cells believed to be responsible for T-ALL relapse.

The selective pharmacological inhibition of PSEN1 did not result in gastrointestinal toxicity or T cell development defects in mice, unlike the adverse effects observed with complete γ-secretase inhibition by DBZ. Our work extends previous studies in Alzheimer’s disease models, where selective γ-secretase inhibition was efficacious and tolerable, although treatment periods in these studies were much shorter than in the current work (38, 51). This lack of toxicity in vivo may be partially explained by the observation that PSEN2 and PSEN1 expression is more equivalent within the gastrointestinal tract and developing T cells compared to T-ALL cells. Psen2 knockout mice exposed to MRK-560 did display gastrointestinal and thymus toxicity, comparable to full γ-secretase inhibition, indicating that PSEN2 is responsible for the protective effects (38). Thus, although PSEN2 is insufficient to compensate for Psen1 deletion during development (15), sparing PSEN2 activity maintains physiologically relevant signaling in the gut and hematopoietic system during adult life. To confirm that PSEN2 activity can compensate for the lack of PSEN1 activity, future studies will require a complete selective genetic knockout of Psen1 in the gut, for example, using a villin-Cre mouse model. However, for all practical means, it is clear that PSEN1-selective inhibition is highly preferable to the broad-spectrum inhibition that is induced by the previously tested GSIs. As shown above, in human T-ALL cells, PSEN1 expression is much higher than PSEN2 expression (Fig. 1), and PSEN2 expression is likely too low to provide sufficient support for leukemia development. Measuring PSEN1 and PSEN2 expression in circulating T-ALL cells might be a helpful parameter to decide whether selective PSEN1 targeting might be beneficial. In addition, PSEN1 and PSEN2 complexes have different subcellular profiles, resulting in substrate selectivity due to differential compartmentalization (11). Therefore, the differential sensitivity for leukemic cells versus normal tissue observed with the PSEN1 inhibitor might also be partially explained by enzyme-substrate specificity due to different subcellular localization. Further work should determine whether mutant NOTCH1 is processed at the cell surface, where PSEN1 complexes are predominantly expressed, whereas wild-type NOTCH1 is processed additionally in endocytic compartments, where PSEN2 complexes reside (11).

Our study was conducted using preclinical experimental models that may not completely recapitulate their human counterparts. The murine leukemia models used were solely driven by mutant NOTCH1 signaling, whereas patients diagnosed with leukemia often display multiple different genetic lesions. However, the four human PDX samples used to determine the potency of pharmacological PSEN1-selective targeting displayed various mutational backgrounds, strengthening our preclinical findings. In addition, human PDX models have shown concordance between preclinical results and corresponding available clinical data (52), making them a good model to study potential therapeutic approaches. Moreover, the overall increased safety profile of selective PSEN1 γ-secretase inhibition warrants optimism for further clinical development. Careful monitoring of the thymus, gastrointestinal tract, and also skin, which is susceptible to tumorigenesis when Psen1 is deleted (6, 53), will be necessary when further addressing the clinical feasibility of PSEN1-selective targeting in patients.

Last, if further proof of concept can be established in patients with T-ALL, then even more applications for selective γ-secretase inhibition in the clinic could be envisaged. Safe-selective γ-secretase inhibition might be useful as additional therapy in a variety of other cancers with deregulated NOTCH signaling, such as B cell lymphoma, breast cancer, and glioma (5456). Furthermore, clinical interest in γ-secretase inhibition might become revived in other therapeutic areas where development was stopped because of gastrointestinal side effects, such as acute hearing loss (57, 58), peritoneal fibrosis as a complication of dialysis in renal disease (59), and atherosclerosis (60, 61).


Study design

We hypothesized that selective targeting of specific γ-secretase subunits is a safe strategy to target NOTCH1 mutant T-ALL. Cell culture experiments using cell lines were performed at least three times. Ex vivo T cell cultures were performed using at least three different mice to generate pro-T cells, unless otherwise noted. For treatment studies involving mice, disease burden was determined at treatment initiation, and animals were rank-ordered and divided into treatment arms after assuring that mean disease burden was comparable among groups. For toxicity studies in mice, animals were randomly assigned to treatment groups. Sample sizes for in vivo experiments were chosen on the basis of previous experience and power calculations of expected differences with this type of experiments. Initial bone marrow transplants were replicated to test for variation in leukemia progression among different experiments. After this was ruled out, bone marrow transplantation experiments were no longer replicated, but sample size was large, and results were reproducible among animals. For testing the effect of MRK-560 on human PDX samples in mice, every experiment was performed only once with five mice per treatment intervention. However, four independent PDX samples were chosen to ensure reproducibility among different patient samples to test for a broader applicability of the approach. Once conditions for an experiment were optimized, all data were included in the absence of a specific technical or procedural reason that confounded the interpretation of a finding. For bioluminescence imaging, the value from one vehicle-treated mouse on day 21 was excluded from the analysis for technical reasons due to an incorrect intraperitoneal injection. Bioluminescence values for this mouse before and after day 21 are present in the analysis. For histological analysis of intestinal toxicity, investigators were blinded during goblet cell counting. During data collection and analysis, investigators were blinded to group allocation.

Cell culture, expression plasmids, and retrovirus production

HPB-ALL, DND-41, and Jurkat cell lines were cultured in RPMI 1640 medium supplemented with 20% fetal bovine serum (FBS). Human embryonic kidney (HEK) 293T cells were cultured in RPMI 1640 medium supplemented with 10% FBS. Mouse embryonic fibroblasts (MEFs) knocked out for Psen and Aph1 and rescued with PSEN1 and APH1A or PSEN2 and APH1A expression were described previously (12). MEFs were transduced with pMSCV-NOTCH1ΔE-RFP-puromycin viral vectors, and after puromycin selection, red fluorescence protein (RFP)–positive cells were selected through fluorescence-activated cell sorting (FACS) and cultured in Dulbecco’s minimum essential medium (DMEM)/F12 medium supplemented with 10% FBS. All other constructs used were cloned into the pMSCV-IRES-GFP vector. Viral vectors were produced in HEK293T cells using an EcoPack packaging plasmid and GeneJuice transfection reagent (Merck-Millipore), and virus was harvested 48 hours after transfection. The stromal cell-free culture system used to generate mouse pro-T cells was described previously (62) and is summarized in Supplementary Materials and Methods.


DAPT was a gift from Janssen Pharmaceutica, and DBZ [(S)-2-(2-(3,5-difluorophenyl)acetamido)-N-((S)-5-methyl-6-oxo-6,7-dihydro-5H-dibenzo[b,d]azepin-7-yl) propanamide] was purchased from Selleckchem. Dimethyl sulfoxide was purchased from VWR, and (2-hydroxypropyl)-β-cyclodextrin, meglumine, corn oil, and tamoxifen were purchased from Sigma-Aldrich. MRK-560 (N-[4-(4-chlorophenyl)sulfonyl-4-(2,5-difluorophenyl)cyclohexyl]-1,1,1-trifluoromethanesulfonamide) was provided by Janssen Pharmaceutica and synthesized as described in Supplementary Materials and Methods (63).

Gene expression profiling

Human thymocytes were extracted from thymus tissue from children undergoing cardiac surgery and were obtained and used according to the guidelines of the Medical Ethical Commission of the Ghent University Hospital, Belgium (approval B670201319452). FACS-mediated cell sorting was used to isolate the different thymocyte subsets, representing the different stages of normal T cell development, as described previously (31, 64).

Mice and animal procedures

All experiments were approved by the Ethical Committee on Animal Experimenting of the University of Leuven. Psen1 conditional knockout mice (Psen1tm2.1Bdes) were generated by homologous recombination in E14 embryonic stem cell line by introducing two loxP sites flanking exon 1. The first loxP site was introduced in intron 1 and the second together with a Frt-flanked hygromycin B selection marker cassette in intron 2. In addition, a double tag encoding calmodulin-binding protein and 3×FLAG was inserted immediately after the ATG start codon. F1 offspring were crossed with Gt(ROSA)26Sortm1(FLP1)Dym strain, and deletion of the selection cassette was confirmed by Southern blotting and PCR analysis. To generate mice carrying a Psen1 deletion in all committed B and T cell progenitors, we crossed the mice harboring the conditional Psen1 targeted allele with a specific Cre deleter line B6.Cg-Tg(CD2-icre)4Kio/J (the Jackson Laboratory, 008520), generating CD2CrePsen1Δ/Δ mice. To generate conditional inducible Psen1 knockout mice, we bred animals harboring the conditional Psen1 allele with B6.129-Gt(ROSA)26Sortm1(cre/ERT2)Tyj/J mice, which express a tamoxifen-inducible form of the Cre recombinase, generating Rosa26Cre-ERt2Psen1f/f mice. All colonies were kept on an inbred C57Bl/6J background, which was also used for wild-type control animals and recipient animals for primary and secondary transplants. NSG mice used for PDX experiments were purchased from Harlan Laboratories. During experiments, mice were housed in individually ventilated cages enriched with wood-wool and shavings as bedding, given access to water and food ad libitum, and monitored daily.

Murine bone marrow transplantation

Bone marrow transplantation to generate the mouse model for ΔEGF-NOTCH1-L1601P-ΔP was performed as described (65). Briefly, 6- to 12-week-old male mice were euthanized, and bone marrow cells were harvested from femur and tibia. Lineage-negative cells were enriched by negative selection using biotinylated antibodies directed against nonhematopoietic stem cells and nonprogenitor cells [CD5, CD11b, CD19, CD45R/B220, Ly6G/C(Gr-1), TER119, and 7-4] and streptavidin-coated magnetic particles (RapidSpheres, STEMCELL Technologies) and cultured overnight in RPMI 1640 with 20% FBS with interleukin-3 (IL-3) (10 ng/ml; PeproTech), IL-6 (10 ng/ml; PeproTech), stem cell factor (50 ng/ml; PeproTech), and penicillin-streptomycin. The following day, 1 × 106 cells were transduced by spinoculation (90 min at 2500 rpm) with viral supernatant (ΔEGF-NOTCH1-L1601P-ΔP) and polybrene (8 μg/ml). The following day, the cells were washed in phosphate-buffered saline (PBS) and injected (1 × 106 cells/0.3 ml) into the lateral tail vein of sublethally irradiated (5 Gy) syngeneic 8- to 12-week-old female C57BL/6 recipient mice. For secondary transplants, leukemic T cells were obtained from splenic tissue derived from mice transplanted with ΔEGF-NOTCH1-L1601P-ΔP wild-type cells. After isolation, 1 million cells were transplanted intravenously into sublethally irradiated (2.5 Gy) 8- to 10-week-old wild-type (C57BL/6) female recipient mice. To delete the Psen1 gene in NOTCH1-induced mouse leukemia, animals were treated with tamoxifen (100 mg/kg) by intraperitoneal injection for five consecutive days.

Human primary leukemia samples

Clinical leukemia samples were obtained with informed consent at local institutions, and all experiments were conducted on protocols approved by the Ethical Committee of the University of Leuven.

In vivo imaging

For in vivo bioimaging, cells from NOTCH1 mutant T-ALL sample XC63 were injected into the tail vein of 6- to 12-week-old NSG mice, and human leukemic cell expansion was monitored through hCD45 staining on peripheral blood samples. After successful engraftment and leukemic disease development, mice were euthanized, and single cells were isolated from the spleen, containing >90% hCD45+ cells. Splenocytes were transduced overnight with lentivirus pCH-SFFV-eGFP-P2A-fLuc, and GFP+ cells were sorted using a S3 Sorter (Bio-Rad) before being transplanted back into NSG mice via tail vein injection. Upon confirmation that leukemic cells were >95% GFP+, leukemic cells were isolated from the spleen and retransplanted into a larger cohort of NSG mice for treatment studies. For in vivo bioluminescence imaging, anesthesia was induced in an induction chamber with 2% isoflurane in 100% oxygen at a flow rate of 2 liters/min and maintained in the IVIS Spectrum (Caliper Life Sciences) with a 1.5% mixture at 0.5 liters/min. Before each imaging session, the mice were injected subcutaneously with d-luciferin (126 mg/kg; Promega, Leiden, the Netherlands) dissolved in PBS (15 mg/ml). Next, they were positioned in the IVIS, and consecutive 2-min frames were acquired until the maximum signal was reached. Data are reported as the total flux per second from the whole mouse.

Toxicity studies, immunohistochemistry, immunofluorescence, and flow cytometry

Healthy 12-week-old C57BL/6 mice were treated for 14 days with vehicle (20% HPβCD in 0.1 M meglumine) or MRK-560 (30 μmol/kg) by subcutaneous injection or DBZ (10 μmol/kg) or vehicle (0.5% methylcellulose, 0.1% Tween 80) by intraperitoneal injection. For the intermittent 4-week treatment, C57BL/6 mice were treated with the same concentrations of DBZ and MRK-560 but at a 5-day on/2-day off schedule. After treatment, thymus was isolated, and single cells were prepared and stained for CD4, CD8, CD44, and CD25 to assess T cell development. To assess goblet cell hyperplasia, intestines from treated NSG mice were harvested and flushed with PBS and 10% neutral buffered formalin before fixation. For every mouse, a minimum of 10 villi were assessed for the number of secretory goblet cells. In addition, total body weight from all treated mice was assessed during and after treatment.

Patient-derived xenografts

Human leukemic bone marrow cells from NOTCH1 mutant T-ALL patients 389E, XC63, XC65, and XB47 (table S1) were injected into the tail vein of 6- to 12-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. After successful engraftment, splenocytes were harvested and reinjected at a concentration of 1 × 106 cells into secondary recipient NSG mice to create secondary transplants. Human leukemic cells were identified in peripheral blood samples by anti-hCD45 (APC, eBioscience) staining by flow cytometry or by luciferase in vivo bioimaging using the IVIS Spectrum (Caliper Life Sciences). After the leukemic clone was detectable in the blood or by bioimaging, mice were segregated randomly into treatment groups and treated daily for 14 days with vehicle or MRK-560 (30 μmol/kg) dissolved in 20% HPβCD in 0.1 M meglumine by subcutaneous injection.

Statistical analyses

All analyses were performed using GraphPad Prism. Comparisons between two groups were performed by the Student’s unpaired two-tailed t test. One-way ANOVA was used to examine differences when comparing effects in three groups, namely, comparing wild-type, Psen1f/f, and CD2CrePsen1Δ/Δ for leukemia progression, T cell development, and in ex vivo T cell cultures. For comparison of cell growth of T-ALL cells or leukemia progression in PDX models over time, differences were assessed by two-way ANOVA. Tukey post hoc analysis was performed to correct for multiple comparisons. Survival in mouse experiments was represented with Kaplan-Meier survival curves, and statistical significance was calculated using the log-rank test.


Materials and Methods

Fig. S1. PSEN1 deletion does not affect T cell development.

Fig. S2. PSEN1 deletion does not affect engraftment in bone marrow transplants.

Fig. S3. Tamoxifen does not affect leukemia progression.

Fig. S4. MRK-560 shows selectivity for PSEN1 over PSEN2.

Fig. S5. PDX samples were treated with MRK-560.

Fig. S6. Long-term treatment with MRK-560 has less gastrointestinal toxicity compared to DBZ.

Table S1. T-ALL cell lines and PDX models.

Table S2. Primers used for qPCR.

References (6668)


Acknowledgments: Synthesis of MRK-560 took place at Janssen Pharmaceutica Neuroscience Medicinal Chemistry by R.N., as a visiting scientist. Funding: This work was supported by the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen (FWO), the KU Leuven and VIB, a Methusalem grant from the KU Leuven/Flemish Government and Stichting Alzheimer Onderzoek to B.d.S. B.d.S. is supported by the Bax-Vanluffelen Chair for Alzheimer’s Disease and “Opening the Future” of the Leuven Universiteit Fonds. J.C. is supported by an ERC-consolidator grant (617340) and “Kom op tegen Kanker” (stand up to cancer), the Flemish Cancer Society. R.A.H. is supported by a FWO fellowship (12I2317N). Author contributions: R.A.H., C.E.d.B., J.C., and B.d.S. conceived the study, designed and analyzed experiments, and prepared the manuscript. R.A.H. and C.E.d.B. performed most of the experiments with the help of D.N., I.L., D.V., L.S., J.D., and A.L. S.D. analyzed the RNA-seq data, and T.T. analyzed the microarray data. R.N. performed MRK-560 synthesis. All authors reviewed the manuscript and agreed with the final submission. Competing interests: B.d.S. and R.N. received a grant from Janssen Pharmaceutica for the development of GSIs. All other authors declare that they have no competing interests. Data and materials availability: Data for PSEN1 and PSEN2 expression in human small intestine were obtained from the GTEx portal and dbGaP accession number phs000424.vN.pN. RNA-seq data on T-ALL cell lines are available in the EGA database with accession number EGAD00001000849. RNA-seq data for human T-ALL can be accessed at

Stay Connected to Science Translational Medicine

Navigate This Article