Leukemia-specific delivery of mutant NOTCH1 targeted therapy

The successful perturbation of folate metabolism as an approach to treat patients with cancer was first described in a landmark paper in the New England Journal of Medicine in 1948. Sidney Farber described the results of the clinical testing of the folate antagonist aminopterin in five children with acute lymphoblastic leukemia (ALL; Farber and Diamond, 1948). That study, for the first time, demonstrated that leukemia cells are highly dependent on folate metabolism while establishing the first reported clinical responses of childhood ALL to drug therapy. Subsequently, the targeting of folic acid metabolism became the foundation of successful ALL treatment.

Folic acid (FA) is a water-soluble vitamin (B9) used as a one-carbon donor in the biosynthesis of the essential purines and thymidylate necessary for the production of DNA and RNA (Fig. 1 A). Folate enters cells by two mechanisms: (1) the reduced folate carrier, a ubiquitously expressed protein with low affinity for folate (Whetstine et al., 2002; Matherly et al., 2007), or (2) folate receptor (FR), which is virtually absent in normal cells but has high affinity for FA (Shen et al., 1997; Kelemen, 2006). The FR family consists of four different proteins: FR1–4 or FRα, β, γ, and δ (Antony, 1992, 1996). Several lines of evidence suggest that FRs are aberrantly expressed in rapidly dividing cells, including cancer cells (Ross et al., 1999; Wang et al., 2000; Lynn et al., 2015). The most extensively characterized FRs in cancer are FR1 and FR2, encoded by the FOLR genes located on the long arm of chromosome 11 (q11.3–q13.5). FR1, for example, is overexpressed in several tumors: adenocarcinomas of the ovary, uterus, and pituitary gland and mesothelioma (Garin-Chesa et al., 1993; Parker et al., 2005). Indeed, FR1 expression is 10–100-fold higher in non-mucinous epithelial ovarian tumors than in normal kidney, lung, or breast epithelial cells (Parker et al., 2005; Kalli et al., 2008). FR2, on the other hand, is constitutively expressed in activated macrophages and acute myeloid leukemia (AML; Ross et al., 1999; Wang et al., 2000; Pan et al., 2002; Paulos et al., 2004b; Lynn et al., 2015).

Because of the high dependence of leukemia cells on folate metabolism, we speculated that T-ALL cells might express FRs on their cellular surface. Therefore, T-ALL became an ideal system to further evaluate our drug delivery strategy with the development of a folate-drug conjugate that leverages the expression of FR on ALL cells for tumor cell specificity, as well as the activity of a natural product with specificity for mutant NOTCH1, a protein recurrently mutated in this disease.

Folate conjugation has been conceptually advanced as an anticancer strategy by Low and colleagues, who have worked to deliver imaging agents and cytotoxins in a tumor-specific manner for many years (Leamon and Low, 1991, 1994; Leamon et al., 2002; Low and Kularatne, 2009). This concept is borne from the effective clinical use of tumor-localizing strategies to increase the effective molarity of cytotoxic agents on the plasma membrane of tumor cells, notably also including antibody–drug conjugates and nanomaterials (Lee and Low, 1994; Kularatne and Low, 2010; Ab et al., 2015). Despite these advances, to date, folate-conjugated agents have not demonstrated evident clinical utility, which necessitates the ongoing investigation of this strategy. Here, we work to expand the tumor-specific therapeutic index by targeting a folate-addicted cancer with a pathway-specific inhibitor.

In T-ALL, transcription factors play a fundamental role in controlling tumorigenesis, and Notch receptors (NOTCH1–4) function as ligand-activated transcription factors that mediate a conserved signaling pathway critical in controlling cell fate determination, stem cell quiescence, and differentiation (Tanigaki and Honjo, 2007). Gain- and loss-of-function Notch mutations have been characterized in a constellation of diseases and are frequently observed in cancer (Penton et al., 2012). Moreover, Notch signaling is thought to contribute to chemotherapy resistance (Wang et al., 2014; Takebe et al., 2015). Hence, several studies strongly support the development of Notch inhibition for targeted cancer therapy. This idea is most compelling for T-ALL, where activating mutations of NOTCH1 are present in 55–60% of cases (Ellisen et al., 1991; Weng et al., 2004), and cancer dependence has been well established (Girard et al., 1996; Capobianco et al., 1997; Aster et al., 2000; Yanagawa et al., 2000; Weng et al., 2004; Beverly et al., 2005; Armstrong et al., 2009; Dail et al., 2010).

Recently, we used gene expression signature, cell-based screens to discover the SERCA inhibitor thapsigargin (Fig. 1 B) as a pathway-specific modulator of mutated NOTCH1 signaling in T-ALL (Roti et al., 2013). This compound had on-target activity in mouse models of human T-ALL, although with efficacy limitations attributable to a narrow therapeutic index. Still, we identified that at thapsigargin concentrations sufficient to inhibit mutant NOTCH1 in vivo, wild-type NOTCH1 and NOTCH2 receptors are properly processed (Roti et al., 2013). This selectivity provides a therapeutic window not observed before with other Notch inhibitors, such as γ-secretase inhibitors or antibody-based approaches, which showed equivalent inhibitory activity against wild-type Notch.

Thapsigargin is a sesquiterpene-γ-lactone isolated from the plant Thapsia garganica. Thapsigargin selectively inhibits SERCA, leading to a depletion of ER calcium storage and sustained elevation of cytosolic calcium (Lytton et al., 1991). SERCA channels are critical to maintain intracellular calcium homeostasis in all cell types. Thus, the direct delivery of thapsigargin to animals or humans might be expected to incur cardiac toxicity secondary to calcium ion shifts. We envisioned that the nonspecific toxicity of thapsigargin could be avoided by tagging an active derivative to FA and using a folate-derived drug delivery strategy to take advantage of the specificity of thapsigargin toward mutant NOTCH1. We hypothesized that the compound would bind with FR specifically on T-ALL cells compared with normal cells (Fig. 1 C). The inhibitory motif could be delivered into the cell and cleaved from the folate tag (Fig. 1 C). The released free inhibitory motif could then bind to SERCA in the leukemia cell to block the maturation of mutant NOTCH1 (Fig. 1 C). This strategy could direct the natural product specifically to the cancer cell to further enhance leukemia-specific mutant NOTCH1 inhibition.

We designed a folate-thapsigargin derivative based on the following principles: the dual-function molecule should actively bind to FR; the connection between FA and thapsigargin should be stable in serum but cleaved in the intracellular compartment; and the resulting thapsigargin derivative should potently inhibit SERCA activity. We investigated potential anchor sites for installing FA on thapsigargin. Structure–function inferences from the literature suggested that core modifications to appending esters may retain antiproliferative activity (Nielsen et al., 1995; Christensen et al., 1999; Jakobsen et al., 2001; Andrews et al., 2007). The butyl ester bond at C8 was readily, selectively cleaved from the isolated natural product under basic conditions to produce the secondary alcohol, 8-O-debutanoylthapsigargin (Thap-OH; Fig. 1 D and Fig. S1). We then connected the carboxylate of FA to the C8-alcohol of Thap-OH to generate the folate conjugate JQ-FT (Fig. 1 E and Fig. S1; see Synthesis of small molecule inhibitors and probes section for full synthetic procedures).

To establish the expression of folate receptor alleles in human T-ALL, we analyzed the mRNA transcript levels of FOLR1 and FOLR2 in 17 T-ALL cell lines and three primary leukemia samples by quantitative RT-PCR. We observed that FOLR2 was expressed in all leukemia samples, whereas FOLR1 expression was measurable in only 3/20 cases tested (Fig. 2 A). To confirm stable expression of surface polypeptides, we developed methods for FR1 and FR2 flow cytometry. Because FR isoforms are polypeptides of 220–237 amino acids that share 68–79% sequence identity (Antony, 1996), we first evaluated the specificity of FR antibodies against FR1 and FR2 using a stably transduced NOTCH1-mutated T-ALL cell line (RPMI 8402) overexpressing FR1 or FR2 and established the lack of antibody cross-reactivity by flow cytometry (Fig. 2 B). Western blotting of lysates from nine T-ALL cell lines with the isoform-specific FR2 antibody confirmed strong expression of FR2 across all of the samples (Fig. 2 C). We did not observe a significant difference in FR2 levels among NOTCH1-mutated (ALL/SIL, DND41, HPB-ALL, KOPTK1, PF382, RPMI 8402) versus wild-type (LOUCY, MOLT16, SUPT11) T-ALL cell lines. These results establish strong expression of FR2.

To assess functional engagement of the folate receptor on T-ALL cells, we generated fluorescence-tagged FA probes, FL-TAMRA and FL-FITC, as tool compounds (Fig. 2 D and Fig. S2). With FL-TAMRA treatment, all tested T-ALL cell lines showed a concentration-dependent increase in fluorescence signal by flow cytometry, notably independent of NOTCH1 mutational status (Fig. 2 E). T-ALL lines demonstrated stronger FL-FITC labeling compared with T cells isolated from peripheral blood mononuclear cells (PBMCs), providing support for leukemia-specific targeting (Fig. 2 F). Collectively, these observations indicate that functional FR2 expression is increased in T-ALL compared with normal cells, further supporting a rationale for folate-mediated delivery in leukemia.

To explore the mechanism of binding and delivery of folate conjugates in T-ALL, we performed competition and temperature-sensitivity studies of the FL-FITC probe. The T-ALL cell line RPMI 8402 was engineered to overexpress either FR1 or FR2 (Fig. 2 B) and then was treated with increasing concentrations of FL-FITC. We observed a concentration-dependent increase in the FITC signal with overexpression of either FR1 or FR2 (Fig. 3 A). We next addressed whether FA could compete with the FL-FITC probe for uptake. T-ALL cells were grown in medium depleted of FA for 48 h, treated with FL-FITC (1 or 10 µM), and assessed by flow cytometry. T-ALL cells grown in folate-depleted medium exhibited increased fluorescence compared with control cells incubated in standard folate-replete culture conditions (Fig. 3 B) or folate-depleted conditions supplemented with free folate (10 µM; Fig. 3 C). Acidic washing of FL-FITC–incubated cells did not eliminate fluorescence intensity, supporting internalization of the folate conjugate fluorescence probe, as opposed to nonspecific binding to the cell surface (Fig. 3 D).

FL-FITC uptake also showed an energy dependence. T-ALL cells with FR2 overexpression cultured at 4°C were unable to take up FL-FITC. This observation supports an active, endocytic mechanism of uptake, as low temperature blocks endocytosis at 4°C because of altered membrane fluidity (Mamdouh et al., 1996; Punnonen et al., 1998; de Figueiredo and Soares, 2000; Fig. 3 E), in keeping with prior folate-conjugate studies (Leamon and Low, 1991).

FRs are glycosylphosphatidylinositol-anchored membrane proteins associated with caveolar structures (Rothberg et al., 1990; Rijnboutt et al., 1996). To determine whether the endocytic process is mediated by caveolae, we tested FL-FITC uptake in T-ALL cells that were pretreated with filipin, a transient inhibitor of caveolin-mediated endocytosis. The filipin-pretreated T-ALL cells produced significantly reduced FITC fluorescence signal (mean reduction 38.9 ± 2.5%), confirming that the folate conjugate uptake into the cell is mediated by caveolar transport (Fig. 3 F). Together, these results demonstrate that folate-conjugated probes are internalized into T-ALL cells by FR-dependent, caveolae-mediated endocytosis.

To test the hypothesis that Thap-OH targets SERCA in T-ALL cells, we first performed a competitive pull-down assay in which protein lysates were treated with a novel biotinylated derivative of thapsigargin (Thap-biotin in Fig. S3), alone or in the presence of increasing concentrations of free, competitive Thap-OH. Binding to Thap-biotin in a complex mixture was confirmed by immunoblot, as free Thap-OH competed off biotinylated thapsigargin from SERCA2 and SERCA3 (Fig. 4 A).

We previously described that the SERCA inhibitors thapsigargin and cyclopiazonic acid impair NOTCH1 maturation, leading to an accumulation of full-length, unprocessed polypeptides in the ER/Golgi subcellular compartment (Roti et al., 2013). An immediate consequence of SERCA inhibition is a decrement in NOTCH1 protein display on the surface of T-ALL cells. The treatment of T-ALL cells with Thap-OH resulted in a concentration-dependent reduction in NOTCH1 expression on the cell surface by flow cytometry, as was observed with control thapsigargin treatment. Thap-OH was again found to be less potent than the natural product, as predicted from prior studies (Christensen et al., 1993; Jakobsen et al., 2001; Fig. 4 B). To further support the hypothesis that Thap-OH impairs mutant NOTCH1 maturation, we evaluated levels of full-length, transmembrane and activated NOTCH1 (ICN1) by Western blotting. Lysates from T-ALL cell lines treated with 1 µM Thap-OH were immunoblotted with an antibody specific for the cytoplasmic portion of NOTCH1 that recognizes both unprocessed NOTCH1 (FL-N1; ∼270 kD) and the furin-processed transmembrane subunit (TM-N1; ∼110 kD). Consistent with the flow cytometry data, Thap-OH reduced the levels of the furin-processed transmembrane NOTCH1 subunit, but not the unprocessed full-length NOTCH1 precursor, in multiple T-ALL cell lines (Fig. 4 C). Moreover, Thap-OH decreased ICN1 levels in T-ALL cells, suggesting that the cleavage product retains the potent anti-NOTCH1 properties observed with SERCA inhibition.

As expected, treatment with Thap-OH was associated with a decrease in T-ALL cell viability, as measured by dose-ranging ATP content assays (Fig. 4 D). Furthermore, the selectivity for mutant (ALL/SIL, DND41, PF382, RPMI 8402) compared with wild-type (LOUCY, MOLT16, SUPT11) NOTCH1 was retained with Thap-OH; mutant T-ALL cells were more sensitive to the effects of Thap-OH than NOTCH1 wild-type cells (Fig. 4 D), and there was no effect on NOTCH1 maturation in the wild-type cell lines at the concentrations tested (Fig. 4, E–G). Thus, cell lines carrying NOTCH1 alleles with heterodimerization domain mutations were more sensitive to Thap-OH than cells with wild-type NOTCH1 alleles.

Thapsigargin is a known inducer of the unfolded protein response (UPR). As such, its derivatives may trigger a cellular response that affects the stable expression or trafficking/recycling of folate receptors. FR recycling has variably been reported to take 5.7–20 h in tumor cells, ∼8 h in normal organs (Paulos et al., 2004a), and ∼10–20 min in macrophages in inflamed tissues (Varghese et al., 2014). Thus, to exclude the possibility that FR2 is a target of UPR, we first treated T-ALL cell lines with increasing concentrations of Thap-OH for 24 h. No change in FR2 expression was observed by Western blotting (Fig. 5 A). Similarly, in time course studies (0–48 h), no change in FR2 expression was observed in T-ALL treated with JQ-FT or Thap-OH (Fig. 5 B). This result indicates that FR2 is not affected by Thap-OH at concentrations targeting NOTCH1.

Together, these data demonstrate that Thap-OH preferentially inhibits mutant NOTCH1 receptors while sparing wild-type NOTCH1 and FR2 expression, supporting Thap-OH as a suitable and targeted payload for folate-conjugation.

We next studied the effects of JQ-FT in a panel of T-ALL cell lines that contain activating mutations in the heterodimerization domain of NOTCH1 and/or protein-stabilizing deletions within the PEST degradation domain. In all T-ALL cell lines studied, JQ-FT impaired cell growth, leading to G1 arrest and rapid induction of apoptosis (Fig. 6, A–D). As expected based on our FL-FITC uptake studies, a greater effect on cell viability was observed in cell lines overexpressing FR1 or FR2 (Fig. 6 E). To demonstrate that JQ-FT enters the cells through an FR-based mechanism, we treated T-ALL in the presence or absence of FA (10 and 100 µM; Paulos et al., 2004a). Cells co-treated with FA are rescued from JQ-FT treatment, whereas no differential response was observed in cells treated with Thap-OH (Fig. 6 F), supporting the mechanism of internalization and subsequent cleavage of JQ-FT through an FR mediated process and consistent with the lack of spontaneous degradation to Thap-OH in culture media conditions (Fig. S4).

As observed with thapsigargin and Thap-OH, treatment of T-ALL cell lines with JQ-FT led to accumulation of full-length NOTCH1 (Fig. 7 A). A decrement of transmembrane NOTCH1 (TM-N1) was confirmed by Western blotting and flow cytometry analysis (Fig. 7, A and B). Loss of ICN1 (Fig. 7 A) caused the suppression of NOTCH1 target genes as measured by RT-PCR (Fig. 7 C). Consistent with a folate receptor–mediated process, FA competition rescued the effects of JQ-FT on transmembrane and activated NOTCH1 as measured by Western blotting and flow cytometry (Fig. 7, D and E). FA did not rescue the effects of Thap-OH on NOTCH1 protein levels (Fig. 7, D and E). To establish whether the effects of JQ-FT on cell viability were caused by impaired NOTCH1 activation, the NOTCH1-dependent T-ALL cell line RPMI 8402 was transduced with MigR1-ICN1, to rescue effects on full-length NOTCH1 processing, versus an empty MigR1 vector control. Exogenous expression of ICN1 attenuated the growth-inhibitory effects of JQ-FT, in keeping with the function of ICN1 downstream of ER processing and surface γ-secretase cleavage in Notch pathway activation (Fig. 7 F).

To begin to assess the translational significance of these findings, we studied leukemia cells from T-ALL patient-derived xenografts (PDXs). JQ-FT treatment inhibited the viability of NOTCH1-mutated PDX cells in vitro (Fig. 8 A). In addition, JQ-FT treatment resulted in loss of transmembrane NOTCH1, leading to the depletion of detectable ICN1 (Fig. 8, B and C). In contrast, no effect was observed in PDX T-ALL cells with wild-type NOTCH1 (Fig. 8 B). Consistent with these results, no transcriptional changes were observed in NOTCH1 target genes in wild-type PDX samples, whereas expression of canonical NOTCH1 target genes, DTX and MYC, was decreased in the NOTCH1 mutant samples (Fig. 8 D). These results provide strong support for the mechanistic thesis that mutated NOTCH1 receptors are more sensitive to JQ-FT treatment in human T-ALL, prompting proof-of-concept studies in T-ALL models in vivo.

In vitro studies provide valuable mechanistic insights but may not recapitulate tumor microenvironment and metabolic conditions in vivo. Indeed, a significant limitation of research on folate-conjugate drugs is the inconsistency between free folate concentrations (and other middle metabolites) in culture conditions in vitro versus in vivo. High levels of free FA in the serum may block the binding and uptake of JQ-FT in FR2-positive T-ALL, compromising the anti-NOTCH1 leukemia effect observed in vitro. To explore the therapeutic efficacy of JQ-FT in vivo, we studied effects on a syngeneic T-ALL mouse model carrying NOTCH1 L1601P ΔPEST, a common mutation observed in the human disease. First, leukemia cells obtained from this model were treated with JQ-FT for 24 h in vitro. Consistent with the results observed in cell lines and PDX cells, ex vivo treatment with JQ-FT diminished the proliferation (Fig. 9 A), ICN1 expression (Fig. 9, B and C), and transcription of canonical NOTCH1 targets Hes1 and Dtx1 (Fig. 9, D and E).

Although the JQ-FT conjugate has not yet been optimized for pharmacologic properties, we were eager to explore the utility of the compound as an in vivo chemical probe, while assessing putative tolerability of folate-conjugated thapsigargin. We first established the feasibility of administering JQ-FT (60 mg/kg per day) in mice, administered by daily i.p. injection. Notably, this tolerated dose is 150-fold improved over our prior established maximum tolerated dose of unconjugated thapsigargin (Roti et al., 2013). Next, we performed pharmacokinetics and tissue distribution of JQ-FT at 60 mg/kg per day in male CD1 mice, which revealed that JQ-FT exhibits broad biodistribution, a low basal rate of hydrolysis to the active thapsigargin, and excellent pharmacologic stability. After a single i.p. injection, JQ-FT achieved a maximum serum concentration of 1,272 ng/ml in bone marrow, 67,952 ng/ml in liver, 2,244 ng/ml in plasma, and 6,416 ng/ml in spleen, with a basal rate of hydrolysis to the active thapsigargin (ratio JQ-FT/Thap-OH; Fig. 9 F). Because these concentrations reached an active concentration range as established in vitro (1–10 µM), we initiated treatment studies of JQ-FT (60 mg/kg i.p. daily) in mice with established T-ALL. After 5 d of treatment, a decrease in tumor growth was observed, confirmed pathologically by a decrease in leukemic infiltration in spleen and liver (Fig. 9 G) and clinically by a reduction in spleen weight (Fig. 9 H). Bone marrow infiltration by T-ALL, the primary site of human disease, was markedly inhibited by JQ-FT, as confirmed by flow cytometric analysis for GFP⁺ leukemia cells (Fig. 9 I). Pharmacodynamic modulation of the Notch pathway was importantly validated by measurement of reduced ICN1 expression in T-ALL cells from JQ-FT–treated animals, compared with vehicle-treated controls (Fig. 9, J and K).

In the last decade, NOTCH1 has been identified as one of the most frequently mutated genes across all cancers (Lawrence et al., 2014). In hematologic malignancies, activating NOTCH1 mutations are observed in chronic lymphocytic leukemia (CLL; Di Ianni et al., 2009; Puente et al., 2011) and mantle cell lymphoma (Kridel et al., 2012) and at an exceptionally high rate in T-ALL (Weng et al., 2004), where NOTCH1 mutations represent the most common actionable genetic abnormality. Targeted NOTCH1 therapies, such as γ-secretase inhibitors and receptor-blocking antibodies, have entered early-stage clinical trials. However, these modalities have the liability of inhibiting normal NOTCH1 and NOTCH2. In addition to the known potential for gut toxicity (van Es et al., 2005; Deangelo and Silverman, 2006), there is also a significant concern for secondary malignancies, as Notch receptors are established context-specific tumor suppressor genes (Dotto, 2008; Lobry et al., 2011). Thus, the development of tumor-directed inhibitors with selective activity against mutated proteins is highly desirable.

Although inhibition of SERCA proteins to selectively target mutated NOTCH1 with free thapsigargin is promising, thapsigargin is poorly tolerated. Here, we provide a solution involving folate conjugation. We recognize efforts from other groups to create thapsigargin derivatives (Hua et al., 1995; Procida et al., 1998), some of which have been advanced to tumor models (Christensen et al., 2009). In one derivative, mipsagargin (G202; Doan et al., 2015), thapsigargin is conjugated to a peptidyl substrate of the proteolytic enzyme prostate-specific membrane antigen for localized delivery. A phase I clinical study has demonstrated the safety and tolerability of G202 in patients with advanced cancer (Doan et al., 2015). Recognizing the limitations of peptide delivery mechanisms, we have here pursued an all-chemical strategy.

Direct FR targeting has been used in therapeutics for FR overexpressed cancers via the development of anti-FR antibodies in ovarian and lung cancer (Armstrong et al., 2013; Thomas et al., 2013). For example, MORAb-003, farletuzumab, is a humanized antibody targeting FR1 with efficacy demonstrated in preclinical xenograft models of ovarian cancer (Armstrong et al., 2013). Lynn et al. (2015) reported the development of T cells expressing a chimeric antigen receptor (CAR) specific for human FR2 (m909). M909 CAR T demonstrated activity against human AML cell lines in vitro and prevented AML tumor growth in vivo (Lynn et al., 2015). Interestingly, m909 CARs do not inhibit healthy CD34⁺ hematopoietic stem cells further supporting the development of a FR-targeting approach in leukemia (Lynn et al., 2015). A second approach uses the high affinity of FA to FR to deliver a folate-conjugated chemotherapeutic agent. An example is the development of vintafolide, a water-soluble derivative of FA and the vinca alkaloid desacetylvinblastine hydrazide (Vergote and Leamon, 2015). In preclinical studies, vintafolide has been shown to bind to FR1-positive tumors and exert direct FR anticancer activity (Reddy et al., 2007). A subsequent phase II clinical trial demonstrated that the combination of vintafolide and liposomal doxorubicin significantly improved the median progression-free survival in women with platinum-resistant ovarian cancer (Vergote and Leamon, 2015). However, recent studies of vintafolide failed to confirm these results in a phase III trial (PROCEED trial EC-FV-06), and the study was suspended. Importantly, the decision was not based on safety concerns for the patients enrolled in the trial. Although these results are disappointing, they were partially predictable based on previous experiences that demonstrated the modest clinical benefit in patients treated with vinca alkaloids in this subset of highly aggressive tumors (Rothenberg et al., 2004; Katsaros et al., 2005; Herzog et al., 2008). We hypothesize that simply tagging a generic chemotherapeutic agent to FA may not be sufficient: it is likely important that the selected agent independently demonstrate a robust tumor-specific inhibitory effect. To this end, the folic conjugation serves to simply improve the agent’s delivery, uptake, and specificity.

In this study, we have proposed, for the first time, the use of small-molecule folate-mediated delivery of thapsigargin to enable selective and target-specific drug delivery to T-ALL cells. The inhibitor, thapsigargin, was connected to FA with a cleavable bond and was transferred into the cell after binding to FR on the cell surface. The expression of FR in T-ALL enabled the selective recognition of the designed molecule, JQ-FT, by cancer cells. The cleavable bond feature of the molecule facilitated direct delivery of the inhibitory motif to the target (SERCA) and subsequently blocked mutant NOTCH1 maturation. This strategy avoided complicated manufacturing processes, such as drug–antibody conjugation, but still allowed selective delivery to the cancer cell. Indeed, the dose of JQ-FT tolerated in mice was >150-fold that of thapsigargin, supporting the more selective uptake of the derivatized product. Importantly, in our in vivo experiments, mice were not restricted to a low folate chow to demonstrate FR-mediated antitumor effect in vivo. Such a strategy had been reported in previously published preclinical studies testing folate-drug conjugates (Leamon et al., 2008). JQ-FT treatment with a low folate diet is expected to have even greater efficacy in vivo. Although we provide here a prototype molecule, JQ-FT is not yet a drug molecule. Additional medicinal chemistry will be needed to optimize the compound for clinical translation, including optimization of the probe, delivery strategy, and schedule of administration. For example, the potency of the molecule will likely require improvement, potentially by the addition of a cleavable spacer between FA and thapsigargin, avoiding the need for Thap-OH, which is less potent than thapsigargin.

In conclusion, the described approach enhanced the therapeutic window of thapsigargin as a NOTCH1 inhibitor providing dual selectivity: leukemia over normal cells and NOTCH1 mutated over wild-type receptors. Given the important role of mutations in NOTCH1 in many cancers, JQ-FT offers a potential strategy in treating other tumors with NOTCH1 mutations, such as CLL and non–small cell lung cancer. Furthermore, our study demonstrated both in vitro and in vivo that the folate-assisted, pathway-specific drug delivery strategy could be an efficient method to solve a common drug delivery problem in the present era of targeted cancer therapy.

Human cell lines MOLT16, LOUCY, DND41, and HPB-ALL were purchased from Leibniz-Institut DSMZ-German collection of microorganisms and cell cultures and ATCC; identity of ALL/SIL, KOPTK1, PF382, SUPT11, and RPMI 8402 was confirmed by PCR sequencing for known NOTCH1 mutations and short tandem repeat loci profiling. Cells were cultured in RPMI 1640 (Cellgro) with 10% FBS (Sigma-Aldrich) and 1% penicillin-streptomycin and incubated at 37°C with 5% CO₂. Primary patient cells were collected from peripheral blood or bone marrow aspirates, after obtaining informed consent under Dana-Farber Cancer Institute Internal Review Board (DFCI IRB)–approved protocols. Patient-derived xenografted cells were provided by David Weinstock at the Dana-Farber Cancer Institute under DFCI IRB–approved protocols. Cells were cultured short-term in RPMI 1640 with 10% FBS and 1% penicillin-streptomycin and incubated at 37°C with 5% CO₂ supplemented with IL-2 (0.1 ng/ml) and IL-7 (0.1 ng/ml; Peprotec).

Separation of T cells from PBMCs was obtained by depletion of non-T cells using the Pan T Cell Isolation kit (130–096-535), an LS Column, and a MidiMACS Separator (Miltenyi Biotec). PBMCs were first isolated from human healthy donors by using a density gradient medium (Ficoll-Paque PLUS; GE Healthcare). All primary human samples were de-identified and obtained after informed consent through DFCI IRB–approved protocol 04-430.

Isolated PBMCs were immediately incubated with a pan-T cell microbeads cocktail for 15 min at 4°C and separated using a magnetic column according to the manufacturer’s protocol. RPMI medium was used for the incubation of isolated T cells. Cells were labeled with CD3-FITC (Beckman Coulter) and analyzed by flow cytometry with an Accuri C6 (BD).

Oligonucleotides encoding cDNA-ORF (FOLR1 and FOLR2) were cloned into pLK317 obtained from the Broad Institute of Harvard and MIT. For virus production, 500,000 293T cells plated in 10-cm plates were transfected with 1 µg pLK317 and packaging plasmids according to the FuGENE 6 protocol (Roche Diagnostics). Medium was changed to RPMI 1640 24 h posttransfection, and viral supernatant was harvested and filtered 48 h posttransfection. The virus was concentrated five times using Peg-it virus precipitation solution (System Bioscences). Cells were infected for 2 h at 37°C with 2 ml lentivirus and 8 µg/ml polybrene (Sigma-Aldrich). Cells were selected 48 h later with 1 µg/ml puromycin (Sigma-Aldrich). Viral supernatant production and retroviral infections were performed as previously described for MigR1 retroviral vectors (Aster et al., 2000). Transduction efficiency for MigR1 was monitored by assessing GFP expression with a FACScan flow cytometer (BD). After viral infection, GFP-positive cells were sorted by flow cytometry with a FACSAria II (BD) and cultured under compound E (0.5 µM) negative selection for 10 d. Experiments were conducted 3 or more days after removal of compound E.

Cell growth was assessed using the Cell-TiterGlo ATP-based assay (Promega). Luminescence was measured using a Fluostar Omega instrument (BMG-Labtech). Apoptosis was measured using a Caspase-Glo 3/7 assay (Promega), or by annexin V and propidium iodide staining by flow cytometry (eBioscience). Cellular DNA content was assessed by staining with propidium iodide (50 g/ml). Cells were analyzed by flow cytometry with a FACScan flow cytometer (BD) and FlowJo V10 (Tree Star) analytical software. At least 20,000 events were acquired, and all determinations were replicated at least twice.

To assess cell viability upon JQ-FT treatment, in vitro murine leukemia cells expressing NOTCH1 L1601P ΔPEST were cultured in Optimem medium supplemented with 10% FBS, 1% penicillin/streptomycin, and murine IL-7 (10 ng/ml) on a feeder layer of OP9 stromal cells. Viability was examined using the cell viability kit with counting beads from BD, gating out dead cells and particles. Results were analyzed using FlowJo software.

Primers and probes for real-time RT-PCR were obtained from Applied Biosystems (GAPDH #402869, RPL13A #Hs01926559_g1, MYC #Hs00153401_m1, HES1 #Hs00172878_m1, DTX1 #Hs00269995_m1, FOLR1 #Hs01124177_m1, FOLR2 #Hs01044732_g1). The data were analyzed using the ΔΔCT method and plotted as percentage of transcript compared with vehicle.

Western blots were stained with antibodies specific for γ-secretase–cleaved NOTCH1 (Val1744; #4147 or #2421; Cell Signaling Technology), the intracellular transcriptional activation domain of NOTCH1 (Hasserjian et al., 1996), or the C terminus of NOTCH1 (#SC-6014 [C-20]; Santa Cruz Biotechnology). Control stains were performed with antibodies specific for actin (#ACTN05; Thermo Fisher Scientific), vinculin (#2907; Abcam), or GAPDH (#137179; Santa Cruz Biotechnology). Blots were developed with anti-rabbit HRP (#NA9340V; Amersham) or anti-mouse-HRP (#NA9310V; Amersham).

Expression of FR1 and FR2 was determined using isoform-specific antibodies (#MAB5646, R&D Systems; #103988, Abcam). Immunofluorescence staining was performed on permeabilized cells using a murine monoclonal antibody against NOTCH1 (3294, Abcam), and species-specific secondary antibodies linked to Alexa Fluor 488. Slides were mounted with Prolong Gold anti-fade reagents and counterstained with DAPI (Invitrogen). Images were acquired using a Zeiss LSM510 confocal microscope at 100× power. Cell surface NOTCH1 was evaluated by staining nonpermeabilized cells with monoclonal anti-human NOTCH1 antibody (#FAB5317P; R&D Systems) as previously described (Roti et al., 2013).

The fluorescent folate analogues folate-FITC (FL-FITC) and folate-TAMRA (FL-TAMRA) were synthesized as described in the Supplemental Material. T-ALL cells were grown in the presence or absence of FL-FITC, and after treatment, cells were washed in cold PBS containing 2 mmol/L EDTA. Background fluorescence on the cell surface due to nonspecific FA fluorescence was determined in each experiment by incubating cells with free FA. For internalization studies, T-ALL cells were incubated for 6 h at 37°C and subsequently either washed as described above or subjected to a short acidic cold wash with PBS, 50 mM glycine pH 4.0 to eliminate cell surface-bound fluorescence. Fluorescent signal was assessed using flow cytometry with an Accuri C6 (BD). A minimum of 10,000 events was collected for each biological sample. TAMRA fluorescence was assessed using a Fluostar Omega instrument (BMG-labtech) with excitation set at 541 wavelength and emission at 568 nM.

Pharmacokinetics and biodistribution studies were performed at ChemPartner in Shanghai, China. In brief, three male CD1 mice were i.p. injected per time point with 60 mg/kg JQ-FT dissolved in 65% D5W + 30% PEG-400 + 5% Tween-80 and sampling of plasma, bone marrow, liver and spleen were collected at multiple time points. JQ-FT concentration was assessed by liquid chromatography (LC)/tandem mass spectrometry (MS-MS) method and pharmacokinetics parameters (T_max, C_max, T_1/2, AUC, etc.) calculated with WinNonlin V 6.2 statistics software (Pharsight Corporation) using a noncompartmental model.

Animals for efficacy studies were maintained in specific pathogen–free facilities at the Irving Cancer Research Center at Columbia University Medical Campus. Animal procedures were approved by the Columbia University IACUC. To generate NOTCH1-induced T-ALL tumors in mice, retroviral transduction of bone marrow cells enriched in lineage negative cells (#130-090-858; Miltenyi) according to the manufacturer’s guidelines was performed with an activated form of the NOTCH1 oncogene (NOTCH1 L1601P ΔPEST). Cells were then transplanted via intravenous injection into lethally irradiated recipients. To evaluate the activity of the JQ-FT compound in vivo, leukemia cells expressing the mutant form of NOTCH1 and GFP were reinjected into sub-lethally irradiated secondary recipients (4 Gy) 6–8 wk old C57BL6 female mice (Taconic Farms). Secondary tumor development was monitored via GFP detection in peripheral blood. Once disease was established, animals with homogeneous tumor burden were divided into two treatment groups of four mice each: vehicle (65% D5W + 30% PEG-400 + 5% Tween-80 only) or 60 mg/kg of JQ-FT. Animals were treated with either vehicle or intraperitoneal injection of compound at 60 mg/kg daily for 5 d. Antileukemic activity of JQ-FT was assessed by measuring spleen weight and percentage of GFP positive (GFP⁺) leukemia cells in the bone marrow.

Compounds were obtained from the following sources: compound E (#ALX-270-415-M001, ENZO Life Sciences), filipin (#F9765, Sigma-Aldrich), FA (#F8758, Sigma-Aldrich). Thapsigargin was directly purchased from Adipogene (#AG-CN2-0003). The structure and purity of compounds were further confirmed by nuclear magnetic resistance (NMR) and liquid chromatography and mass spectrometry (LCMS).

Reactions were run as described in the individual procedures using standard double manifold and syringe techniques; glassware was dried by baking in an oven at 130°C for 12 h before use. Solvents for reactions were purchased anhydrous from Sigma-Aldrich and used as received. HPLC grade solvents were used for aqueous work ups and chromatography. Reagents were used as received. Reactions were monitored by thin-layer chromatography using EMD silica gel 60 F254 (250-micrometer) glass-backed plates (visualized by UV fluorescence quenching and staining with KMnO₄) and by LC-MS using a Waters Acquity BEH C18 2 × 50-mm 1.7-µm particle column (50°C) eluting at 1 ml/min with H₂O/acetonitrile [0.2% vol/vol added formic acid; 95:5(0min)→5:95(3.60min)] using alternating positive/negative electrospray ionization (125–1000 amu) and UV detection (210–350 nm). Flash column chromatography was performed using Merck grade 9385 silica gel 60 Å pore size (230–400 mesh; Sigma-Aldrich, St. Louis, MO, USA). Melting points were obtained using a capillary melting point apparatus and are uncorrected. 1H NMR spectra were recorded at 400 MHz on a Bruker spectrometer and are reported in ppm using the residual solvent signal (dimethylsulfoxide-d6 = 2.50 ppm; chloroform-d = 7.27 ppm; methanol-d4 = 3.31 ppm; dichloromethane-d2 = 5.32 ppm) as an internal standard. Data are reported as: {(δ shift), [(s = singlet, d = doublet, dd, doublet of doublets, ddd = doublet of a dd, t = triplet, quin = quintet, sept = septet, br = broad, ap = apparent), (J = coupling constant in Hz) and (integration)]}. Proton-decoupled ¹³C NMR spectra were recorded at 100 MHz on a Bruker spectrometer (Bruker Daltonics Inc.) and are reported in ppm using the residual solvent signal (chloroform-d = 77.0 ppm; dimethylsulfoxide-d6 = 39.51 ppm; methanol-d4 = 49.15 ppm) as an internal standard. Infrared spectra were recorded using an ATR-FTIR instrument. High-resolution mass spectra were acquired by flow injection on a qTOF Premiere Mass Spectrometer operating in ES+ ionization with resolution ∼15,000.

To a solution of thapsigargin (200 mg, 0.31 mmol) in methanol (4 ml), triethylamine (0.5 ml) was added at 23°C. The resulting clear solution was stirred at 23°C for 6 h. The solvent was removed in vacuo. The crude reaction was purified directly using column chromatography (MeOH-CH₂Cl₂, 0 to 15% gradient), and produced Thap-OH as white foam (170 mg, 94% yield). MS: m/z (M+1)⁺: 581.3.

To a solution of Thap-OH (17 mg) in DMSO (1.6 ml) was added FA (27 mg, 0.06 mmol), N, N’-dicyclohexylcarbodiimide (DCC, 9.7 mg, 0.077 mmol), and 4-dimethylaminopyridine (DMAP, 3.8 mg, 0.031 mmol). The reaction was stirred at 23°C for 16 h. The reaction mixture was further diluted with methanol and was directly purified by HPLC to afford JQ-FT as yellow powder (13 mg, 45% yield). ¹H NMR (500 MHz, DMSO-d₆) d ppm 0.79–0.89 (m, 5 H), 0.98–1.02 (m, 1 H), 1.11–1.17 (m, 4 H), 1.19–1.33 (m, 19 H), 1.44–1.55 (m, 3 H), 1.64–1.75 (m, 5 H), 1.78–1.82 (m, 8 H), 1.85–1.90 (m, 5 H), 1.95–2.02 (m, 2 H), 2.18–2.35 (m, 6 H), 3.07 (d, J = 11.60 Hz, 1 H), 4.26–4.32 (m, 1 H), 4.35 (br. s., 1 H), 4.45–4.52 (m, 3 H), 5.24–5.27 (m, 1 H), 5.39 (t, J = 3.51 Hz, 1 H), 5.52 (br. s., 1 H), 5.59 (br. s., 1 H), 6.03–6.19 (m, 5 H), 6.55–6.71 (m, 3 H), 7.61 (d, J = 8.85 Hz, 3 H), 8.25 (d, J = 7.63 Hz, 1 H), and 8.60–8.71 (m, 2 H). MS: m/z (M+1)⁺: 1004.4.

FA (502.5 mg, 1.14 mmol) and DMAP (692.5 mg, 5.67 mmol) were suspended in DMSO (5 ml). N,N′-diisopropylcarbodiimide (DIC; 350.8 µl, 2.27 mmol) was added to the reaction followed by a solution of SI-1 (204.9 mg, 1.17 mmol) in DMSO (2 ml). The reaction was stirred at room temperature, and an additional 2 ml of DMSO was added to help solubilize the reactants. The reaction was stirred at room temperature overnight. Water was added to the solution, and a yellow-orange solid precipitated out. The reaction was filtered, and the filtrate was collected and lyophilized. The residue was redissolved in methanol and purified via HPLC to give the desired product, SI-2, as a yellow-orange solid. MS: m/z 599.5 (M+1)⁺.

SI-2 was suspended in 4 M HCl in dioxane (6 ml) and stirred at room temperature for 3.5 h. The reaction was concentrated in vacuo, redissolved in methanol, and purified via HPLC to give the desired product SI-3. MS: m/z 499.4 (M+1)⁺.

Free amine, SI-3 (37.5 mg, 0.08 mmol) and DIPEA (133.4 µl, 0.77 mmol) were suspended in THF (5 ml). MeCN (1 ml) was added to the reaction to help the solubility of the reaction. FITC (26.7 mg, 0.07 mmol) was then added, and the reaction was stirred at room temperature for 3 h. The reaction was purified directly via HPLC to afford the desired product, FL-FITC as a yellow solid. ¹H NMR (500 MHz, DMSO-d₆) d ppm 1.18–1.30 (m, 2 H), 1.56–1.64 (m, 1 H), 1.73–1.82 (m, 2 H), 1.85–2.13 (m, 5 H), 2.31–2.45 (m, 4 H), 3.15–3.21 (m, 6 H), 4.03–4.16 (m, 3 H), 4.32 (s, 1 H), 4.49 (s, 2 H), 6.50–6.72 (m, 17 H), 6.98 (d, J = 7.63 Hz, 1 H), 7.09 (d, J = 8.24 Hz, 1 H), 7.17 (d, J = 8.54 Hz, 1 H), 7.57–7.69 (m, 4 H), 7.70–7.77 (m, 1 H), 8.08–8.25 (m, 3 H), 8.63–8.69 (m, 1 H), 9.07 (s, 1 H), and 9.97 (s, 1 H). MS: m/z 887 (M+1)⁺.

Free amine, SI-3 (13.8 mg, 0.028 mmol) and DMAP (35.5 mg, 0.289 mmol) were dissolved in DMSO (2 ml). A solution of 5,6-TAMRA succinimidyl ester (10.7 mg, 0.020 mmol) in DMSO (1 ml) was added, and the reaction was stirred at room temperature overnight. The reaction was purified via HPLC to afford the desired product, FL-TAMRA as a dark purple solid. ¹H NMR (500 MHz, DMSO-d₆) d ppm 1.13–1.20 (m, 1 H), 1.23 (s, 1 H), 1.77–1.84 (m, 1 H), 1.85–1.93 (m, 2 H), 1.93–2.01 (m, 1 H), 2.04–2.18 (m, 1 H), 2.36 (s, 1 H), 2.42 (t, J = 7.32 Hz, 2 H), 3.26 (s, 17 H), 4.10 (t, J = 5.95 Hz, 3 H), 4.29–4.42 (m, 1 H), 4.48 (s, 2 H), 5.73–5.77 (m, 1 H), 5.93–5.97 (m, 1 H), 6.52–6.58 (m, 1 H), 6.60–6.68 (m, 2 H), 6.95 (s, 2 H), 7.00–7.11 (m, 6 H), 7.54–7.69 (m, 4 H), 7.85–7.96 (m, 1 H), 8.11–8.15 (m, 1 H), 8.17 (d, J = 7.93 Hz, 1 H), 8.20–8.25 (m, 1 H), 8.29 (dd, J = 7.93, 1.83 Hz, 2 H), 8.62–8.72 (m, 2 H), 8.75–8.82 (m, 1 H), 8.92 (s, 1 H), 9.06 (s, 1 H), and 9.97 (s, 1 H). MS: m/z 911.8 (M+1)⁺.

To a solution of Thap-OH (17 mg) in DMSO (1.6 ml) was added biotin-PEG₂-COOH acid (27 mg, 0.06 mmol), DCC (9.7 mg, 0.077 mmol), and DMAP (3.8 mg, 0.031 mmol). The reaction was stirred at 23°C for 16 h. The reaction mixture was further diluted with methanol and was directly purified by HPLC to afford Thap-biotin as colorless oil (13 mg, 42% yield). MS: m/z (M+1)⁺: 1144.6.

The compound stability of JQ-FT was assessed by LCMS. Next, we collected cell culture media (supernatant) and protein lysates of ALL/SIL cells treated with 20 µM of JQ-FT or 20 µM Thap-OH. In protein lysates of JQ-FT–treated cells, we detected the presence of a Thap-OH peak and the disappearance of the JQ-FT peak, indicating intracellular cleavage of the probe (Fig. S4, A–C). In contrast, in cell culture media of treated cells, we observed the presence of a JQ-FT peak and the lack of a Thap-OH peak, indicating the absence of spontaneous cleavage of the probe in in vitro conditions (Fig. S4, D and E).

The detailed syntheses of compound Thap-OH, JQ-FT, biotin-Thap, FL-TAMRA, and FL-FITC are described in the Supplemental material. Fig. S1 describes the synthesis of JQ-FT. Fig. S2 summarizes the synthesis of fluorescence FA derivatives FL-FITC and FL-TAMRA. Fig. S3 describes the synthesis of Thap-biotin. Fig. S4 depicts the LC-MS traces of T-ALL cell line (ALL/SIL) treated with JQ-FT and Thap-OH.

This work was supported by a SCOR Award from the Leukemia and Lymphoma Society (J.E. Bradner, K. Stegmaier, and J. Qi), the William Lawrence and Blanche Hughes Foundation (K. Stegmaier), the Children’s Leukemia Research Association (K. Stegmaier), the National Cancer Institute (R35 CA210030), the Claudia Adams Barr Program in Cancer Research (G. Roti and K. Stegmaier), Fondazione Umberto Veronesi (G. Roti), and Associazione Italiana per la Ricerca sul Cancro (n. 17107; G. Roti).

J.E. Bradner is now the President of the Novartis Institutes for Biomedical Research. The other authors declare no competing financial interests.

Author contributions: G. Roti, J. Qi, J.E. Bradner, and K. Stegmaier conceived the project and cowrote the manuscript with input from all coauthors. G. Roti and J. Qi performed the data analysis and interpreted the results. G. Roti, S. Kitara, and A. Su performed the cell-based work. J. Qi, A.C. Varca, and L. Wu designed and synthesized all the probes. M. Sanchez-Martin, A. Saur Conway, A.L. Kung, and A.A. Ferrando performed the in vivo studies and interpreted the results. J.E. Bradner and K. Stegmaier supervised the research.