Preclinical Evaluation of Proteolytic Targeting of LCK as a Therapeutic Approach in T-cell Acute Lymphoblastic Leukemia

Abstract

T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive hematological malignancy and there is an unmet need for targeted therapies, especially for patients with relapsed disease. We have recently identified pre-T cell receptor (preTCR) and lymphocyte-specific protein tyrosine kinase (LCK) signaling as a common therapeutic vulnerability in T-ALL. LCK inhibitor dasatinib showed efficacy against T-ALL in preclinical studies and also in patients with T-ALL, however, this is transient in most cases. Leveraging the Proteolysis Targeting Chimera (PROTAC) approach, we developed a series of LCK degraders using dasatinib as an LCK ligand and phenyl-glutarimide as a cereblon-directing moiety. Our lead compound SJ11646 exhibited marked efficiency in cereblon-mediated LCK degradation in T-ALL cells. Relative to dasatinib, SJ11646 showed up to three orders of magnitude higher cytotoxicity in LCK-activated T-ALL cell lines and primary leukemia samples in vitro, with drastically prolonged suppression of LCK signaling. In vivo pharmacokinetic and pharmacodynamic profiling indicated a 630% increase in the duration of LCK suppression by SJ11646 over dasatinib in patient-derived xenograft models of T-ALL, which translated into its extended leukemia-free survival over dasatinib in vivo. Lastly, SJ11646 retained a high binding affinity to 51 human kinases, particularly ABL1, KIT, and DDR1, all of which are known drug targets in other cancers. Taken together, our dasatinib-based phenyl-glutarimide PROTACs are promising therapeutic agents in T-ALL and also valuable tools for developing degradation-based therapeutics for other cancers.

One-sentence summary

Dasatinib-based proteolysis targeting chimera is effective in treating LCK-activated T-cell acute lymphoblastic leukemia.

Introduction

Acute lymphoblastic leukemia (ALL) is a common blood cancer in the United States. ALL can arise in both lymphoid lineages, with B-ALL and T-ALL accounting for 85% and 15% of cases(1). T-ALL is typically more aggressive and has worse survival than B-ALL(2, 3). In adults, only ~50% of patients with T-ALL survive more than five years from diagnosis(4). Relapsed and refractory T-ALL has an even more dismal prognosis, with the 5-year survival rate of 10%(5). Children with T-ALL fare relatively better with conventional cytotoxic chemotherapy, but at the cost of a range of debilitating long-term side effects(6). Therefore, the lack of innovative therapies for T-ALL is a barrier to improving the cure of this highly aggressive cancer.

Coupling systems biology analyses of T-ALL genomic data with ex vivo leukemia drug sensitivity profiling, we recently discovered that ~40% of T-ALL cases exhibited constitutive activation of the lymphocyte-specific protein tyrosine kinase (LCK) and thus were responsive to LCK inhibitor therapy(7). LCK dependency in T-ALL was driven by differentiation arrest at the CD4/CD8 double-negative DN3/DN4 stages, where preTCR-LCK signaling is triggered to promote T cell expansion prior to specification. This pro-proliferation signal is thought to be leukemogenic and required for T-ALL survival(7, 8). Our findings corroborate data from other groups’(7, 9–12), consistently pointing to LCK as a therapeutic target in T-ALL. LCK inhibitor dasatinib exhibited anti-leukemia efficacy in vivo in patient-derived xenografts of T-ALL, albeit only temporarily delaying leukemia growth(7). This is not unexpected because most small-molecule inhibitors exert reversible effects on their targets and sustained efficacy requires prolonged drug exposure which can be challenging to achieve pharmacokinetically. For this reason, it is attractive to develop proteolysis targeting chimeras (PROTACs) to degrade LCK and irreversibly suppress LCK signaling.

PROTACs are bifunctional molecules that target proteins of interest for degradation by hijacking the ubiquitin–proteasome system. A PROTAC molecule typically consists of a ligand for the target protein and an E3 binding moiety that are connected by a linker. Through its dual binding capability, the PROTAC brings the target protein within close proximity of an E3 ligase, which in turn results in its ubiquitination and subsequent proteasomal degradation(13, 14). Once the target protein is degraded, the PROTAC is released and can participate in additional rounds of induced proteolysis, which enables the elimination of therapeutic targets in super-stoichiometric quantities(15). PROTACs have been shown to effectively induce degradation of a wide range of therapeutic targets, including enzymes, transcription factors, and scaffolding proteins(16–22). In particular, tyrosine kinases are highly amendable as PROTAC targets because of their known importance in a variety of diseases and the availability of small molecule inhibitors as potential ligands(23, 24).

Targeting cereblon (encoded by the CRBN gene), a substrate recognition domain of CRL4^CRBN E3 ubiquitin ligase, is currently one of the most frequently reported approaches in PROTAC discovery. This is mostly due to drug-like properties of cereblon ligands, immunomodulatory imide drugs (IMiDs) such as lenalidomide and thalidomide.(25, 26). However, IMiD-based PROTACs readily undergo hydrolysis, even in commonly utilized cell culture media, which can affect their cell efficacy. To address this, we have recently identified phenyl-glutarimide as an alternative cereblon binder with several advantages over classical IMiDs in PROTAC design, such as chemical stability, smaller size, lower polarity, greater ligand efficiency, and synthetic feasibility(27).

Here, we describe the design, synthesis, and preclinical evaluation of a series of LCK-targeting PROTACs. Using dasatinib as the LCK ligand, we developed ten LCK degraders with a phenyl glutarimide-based cereblon binder. Our lead PROTAC molecule showed efficiency in LCK proteolysis and cytotoxicity in T-ALL, more potent than the dasatinib(22), with minimal cytotoxic effects on normal hematopoietic cells. Our in vivo pharmacokinetic and pharmacodynamic studies revealed a markedly improved exposure-to-efficacy relationship and anti-leukemic efficacy of PROTAC molecules relative to dasatinib. Systematically screening the human kinome, we also showed that our dasatinib-derived PROTAC broadly targets ABL and SRC family kinases and therefore is a powerful tool for developing a unique class of drugs against these targets.

Results

Development of LCK PROTACs and their in vitro characterization

To evaluate the feasibility of developing LCK degraders, we synthesized a series of PROTAC molecules that are comprised of three components: (1) dasatinib as the LCK ligand with a reported dissociation constant (Kd) of 7 nM(28); (2) a phenyl-glutarimide moiety as the cereblon-directing moiety(27); and (3) a linker with varied composition to modulate activity and physicochemical properties (Fig. 1A). Based on structural analysis of the dasatinib-ABL complex, the linker attachment site to dasatinib was hypothesized to permit PROTAC binding to LCK(29). In parallel, we also synthesized a set of IMiD-based dasatinib PROTAC described previously(22) as references (compounds 1a, 1b, and 1c, Fig. 1A and fig. S1). LCK PROTACs were then evaluated for their physicochemical properties related to adsorption, disposition, metabolism, and excretion, including stability in aqueous solution and plasma, metabolic stability using mouse liver microsome, and aqueous solubility (Table S1). We measured the influx and efflux of LCK PROTACs across the intestinal cell CACO-2 monolayers. All PROTACs showed relatively modest permeability, with an apparent permeability coefficient (P_app A/B) ranging from 0.06 to 7.61 nm/s.

Figure 1. Dasatinib-based PROTAC molecules induce LCK degradation with cytotoxic effects in T-ALL.

(A) Structures of a previously reported dasatinib-IMiD PROTAC (Compound 1a(22),) and phenyl-glutarimide-based LCK targeting PROTACs developed herein (compounds 2a-j). (B) LCK degradation in T-ALL induced by PROTAC molecules. LCK-activated T-ALL cell line KOPT-K1 was treated with Compound 2a-j, IMiD-PROTAC, or dasatinib (100nM for 24 hours) and was harvested for protein extraction and western blotting. The lower panel shows the image for western blots, and the upper panel depicts quantification of LCK normalized to GAPDH. Bars are shown as the mean and standard error of measurement (SEM, n=3). (C) Cytotoxicity of LCK PROTACs in T-ALL. KOPT-K1 cells were treated with each phenyl-glutarimide PROTAC, IMiD-PROTAC or dasatinib for 72 hours and cell viability was determined by CTG assay. Cell viability is shown as mean and standard deviation (SD, n=3).

Next, we sought to evaluate LCK PROTACs for their efficiency in target protein degradation and cytotoxic effects in T-ALL. For this, we chose the T-ALL cell line KOPT-K1 as the model system because of its constitutive activation of LCK and in vitro sensitivity to dasatinib(7). We also performed a genome-wide CRISPR screen to identify genes essential for leukemia viability in this model. Using this assay, we confirmed LCK-dependency as a therapeutic vulnerability in this type of T-ALL whereas none of the other SRC family kinases exhibited any effects on cell survival (fig. S2). At 100nM, all PROTACs induced LCK degradation after 24-hour exposure, whereas dasatinib did not affect the amount of LCK protein (Fig. 1B). IMiD-based compounds (1a, 1b, and 1c, Fig. 1B and fig. S3) showed an average of 86.7% (ranging from 82.9%-94.2%) LCK degradation, compared to an average of 96.6% (ranging from 86.1%-99.1%) with the phenyl glutarimide-based PROTACs (P=0.0059, t-test, Fig. 1B). The degree of LCK degradation by a given PROTAC was positively correlated with its cytotoxicity in KOPT-K1 measured as LC₅₀, the concentration to inhibit leukemia growth by 50% (P=0.0127, Pearson correlation). Among phenyl-glutarimide PROTACs, Compound 2a was the most cytotoxic with an LC₅₀ of 0.083pM, which was 1,561-fold lower than that of dasatinib with an LC₅₀ of 0.13nM (P<0.001, t-test, Fig. 1C); six exhibited cytotoxicity comparable to dasatinib, and two were less cytotoxic. By contrast, the IMiD-based PROTACs displayed lower cytotoxicity than dasatinib with LC₅₀ values of 9.37, 2.22, and 12.71nM (Fig. 1C and fig. S1). Owing to its high potency, we selected Compound 2a for further evaluations, henceforth referred to as SJ11646.

SJ11646 efficiently and rapidly induces LCK proteolysis in T-ALL with minimal effects on normal hematopoietic cells

Tested at six concentrations between 10⁻⁷ and 100nM, SJ11646 was shown to induce LCK degradation in KOPT-K1 T-ALL cells in a dose-dependent manner, with an estimated DC₅₀ of 0.00838pM, the concentration required to degrade target protein by 50% (Fig. 2A–B). By contrast, the IMiD-based PROTACs 1a, 1b, and 1c had a DC₅₀ of 9.97nM, 4.57nM, and 7.59nM, respectively, and there was no LCK degradation by dasatinib across all concentrations (Fig. 2C, fig. S3, and fig. S4). As expected, PROTAC treatment also led to a loss of LCK phosphorylation (pLCK) (Fig. 2A, fig. S3, and fig. S4).

Figure 2. SJ11646 efficiently induces LCK degradation in vitro in a dose- and time-dependent fashion.

(A-C) LCK abundance after various dosages of SJ11646, with dasatinib as the control (ctrl). KOPT-K1 cells were treated with increasing concentrations of SJ11646 (Panel A shows Western blot with quantification shown in Panel B) or dasatinib (C) for 24 hours. Proteins were extracted for western blotting for LCK and pLCK (Y394). (D-E) LCK degradation overtime after treatment with SJ11646, with dasatinib as control. KOPT-K1 cells were treated with SJ11646 (D) or dasatinib (E) at 100nM and proteins were harvested at 0, 1, 3, 6, 24 hours after treatment. LCK and pLCK (Y394) amounts are determined by using western blotting. (F) Ex vivo sensitivity of human primary T-ALL cells to SJ11646. Human leukemia cells from six patient-derived T-ALL xenografts (PDX) were treated with various concentrations of SJ11646 or dasatinib or for 96 hours using a stromal cell co-culture system. Cell viability was determined by high-content imaging analysis. Data are shown in mean and SD (n=2). (G) In vitro cytotoxicity of SJ11646 in human CD34⁺ and peripheral mononucleate cells (PBMCs). Cells were treated with SJ11646 or dasatinib for 72 hours and cell viability was measured by CTG assay. Data are shown as mean and SD (n=2).

Next, we examined the time course of LCK degradation by treating KOPT-K1 cells with SJ11646, IMiD PROTACs 1a, 1b, 1c, or dasatinib at 100nM for 1, 3, 6, and 24 hours. With SJ11646, we observed 92.6% LCK degradation and complete loss of pLCK within 3 hours (Fig. 2D, fig. S4), with an estimated t_1/2, the time required for LCK or pLCK to reduce to half of their initial amounts, of 1.37 and 1.04 hours, respectively. IMiD PROTACs also exhibited a rapid loss of LCK and pLCK (fig. S3). Compared to PROTACs, dasatinib showed a fast-acting effect with a complete loss of pLCK by 1 hour, with no impact on LCK abundance (Fig. 2E, fig. S4).

Because dasatinib has already been used clinically as an LCK inhibitor in patients with T-ALL, we focused on comparing SJ11646 against dasatinib as the benchmark compound for the remainder of this study. To further profile the anti-leukemia effects of SJ11646, we tested it against a panel of T-ALL blasts collected from 10 patient-derived xenografts (PDX). In all six PDX samples that were LCK-dependent(7), SJ11646 showed superior cytotoxicity relative to dasatinib across concentrations (Fig. 2F). In the remaining four PDXs and T-ALL cell line Jurkat that were LCK-independent, our PROTAC showed no cytotoxicity at concentrations tested (fig. S5), suggesting LCK-specific effects as we described previously(7).

Because LCK activation is essential in normal T cell functions, we also examined the cytotoxicity of SJ11646 in hematopoietic cells from healthy donors. We tested CD34⁺ cells isolated from cord blood as well as peripheral blood mononuclear cells for the degree of apoptosis induced by SJ11646 or dasatinib. SJ11646 showed lower toxicity than dasatinib in both CD34⁺ cells (LC₅₀ of 726.42 and 92.88nM, respectively, Fig. 2G), with similar results in peripheral blood mononuclear cells (LC₅₀ of 13.84 and 2.81nM for, respectively, Fig. 2G). We also tested resting and activated normal T cells and observed similar cytotoxic effects of SJ11646 as compared to dasatinib in these cell populations (fig. S5). Taken together, our results suggest that SJ11646 is a potent and fast-acting LCK degrader with cytotoxic effects specific to LCK-dependent T-ALL cells.

SJ11646 is a potent and selective degrader of ABL and SRC family kinases

As a multi-target tyrosine kinase inhibitor, dasatinib binds to a relatively wide spectrum of kinases with a high affinity, particularly ABL1 and SRC family kinases, including LCK(28, 30). Hence, we sought to comprehensively profile the kinase selectivity of SJ11646 in comparison with dasatinib. Using a competition binding kinase assay(31), we first measured the dissociation constant Kd of SJ11646 against LCK, ABL1, and SRC and observed binding kinetics similar to dasatinib: LCK, Kd=0.14nM for SJ11646 and 0.09 for dasatinib (Fig. 3A); ABL1, Kd= 0.054nM and 0.022nM (Fig. 3B); SRC, Kd=0.17nM and 0.19nM (Fig. 3C). Furthermore, we performed scanMAX assay to profile the affinity of SJ11646 and dasatinib for 468 human kinases(31) (Table S2). Overall, these two compounds showed a rather similar selectivity profile (Fig. 3D–E, fig. S6), with CSK, LCK, and ABL1 as the top-ranking targets. In fact, there was a strong positive correlation of relative binding between dasatinib and SJ11646 across kinases (R²=0.9233, P<0.0001, Pearson correlation, Fig. 3F), indicating that our phenyl-glutarimide LCK PROTAC retained dasatinib’s targeting selectivity without inducing new interactions with other kinases.

Figure 3. Kinase selectivity of SJ11646 in comparison to dasatinib.

A-C. Kd analysis of SJ11646 and dasatinib against LCK (A), ABL1 (B) and SRC (C). A competition assay using DNA-tagged kinase and immobilized ligand on beads was performed to evaluate the binding affinity of dasatinib or SJ11646 for each target kinase. The binding assay was done with 11 concentrations, data are shown in mean and SD (n=2). (D-E) Kinome scan assay indicates similar selectivity profile of SJ11646 (D) and dasatinib (E). A competition assay was performed with a fixed dose of 100nM for dasatinib and SJ11646 against 468 kinases (including mutant proteins, see Supplementary Fig. 6A–B). In the Treespot visualization, the size of the circle indicates the affinity between the compound and a target kinase. (F) Comparison of kinase selectivity profile of dasatinib and SJ11646. Binding efficiency was determined by Kinome scan assay and x- and y-axes represent %binding by dasatinib and SJ11646, respectively. (G-I) Proteomic profiling to identify changes in protein expression with dasatinib treatment (G), SJ11646 (H), or those specifically affected by SJ11646 relative to dasatinib (I). KOPT-K1 cells were treated with vehicle control, dasatinib, or SJ11646 at 100 nM for six hours before being subjected to protein collection and TMT-based proteomic profiling. In the volcano plots, x-axes are log₂ transformed foldchange and y-axes are −log₁₀ transformed P-value. (J) Western blot of CSK and other members of the SRC family kinases in KOPT-K1 cells. KOPT-K1 cells were treated with dasatinib or SJ11646 at 100nM for 24 hours and immunoblotting was performed to quantify SRC family kinases and CSK. (K) In vitro sensitivity of BCR-ABL ALL cell line SUP-B15 to SJ11646 or dasatinib. SUP-B15 cells were treated with various concentrations of each compound for 72 hours and cell viability was determined by CTG assay. (L) Western blot of BCR-ABL protein in SUP-B15 cells. SUP-B15 cells were treated with dasatinib or SJ11646 at 100nM for 24 hours and immunoblotting was performed to quantify BCR-ABL.

Furthermore, we performed unbiased proteomic profiling of KOPT-K1 cells to determine the global changes caused by dasatinib and SJ11646 treatment. In this experiment, we treated KOPT-K1 cells with dasatinib or SJ11646 at 100 nM for six hours and collected proteins for Tandem Mass Tag proteomics assay (Table S3). We identified 10,513 unique proteins, of which only DUSP6 and CISH were commonly downregulated in both dasatinib-treated and SJ11646-treated cells (P<10⁻⁵, t-test, log₂[fold change]>1.5), and both are LCK downstream signal proteins(32, 33) (Figure 3G–I). By contrast, there were only three proteins, namely CSK, LCK, and SRC that were uniquely affected by SJ11646.

SJ11646-mediated degradation of SRC family kinases was also confirmed by Western blotting in KOPT-K1 cells (Fig. 3J). In B-ALL cell line SUP-B15 which harbors the BCR-ABL fusion gene, SJ11646 exhibited cytotoxicity with an estimated LC₅₀ of 0.0123 pM, 55,114-fold higher than dasatinib (Fig. 3K). As expected, SJ11646 induced degradation of BCR-ABL fusion protein, whereas dasatinib did not (Fig. 3L). These results suggest that the improved therapeutic effects of SJ11646 over dasatinib may extend to other cancers where dasatinib target genes are implicated.

SJ11646-induced LCK degradation is cereblon-dependent

To confirm SJ11646-induced LCK degradation was mediated by cereblon, we performed a competition assay using lenalidomide, which has a high affinity for cereblon but has no effect on T-ALL survival and growth (fig. S7). In KOPT-K1 cells, the addition of lenalidomide (10 μM) blunted the cytotoxic effects of SJ11646, with an increase of its LC₅₀ by 21,978-fold (Fig. 4A), whereas lenalidomide did not alter dasatinib LC₅₀ (Fig. 4B). Lenalidomide also reduced the degree of LCK proteolysis by SJ11646 in vitro, although the inhibition of LCK phosphorylation appeared to be affected to a lesser extent (Fig. 4C). As expected, blocking E3 ligase binding by lenalidomide had no effects on dasatinib inhibition of LCK (Fig. 4D).

Figure 4. SJ11646-induced LCK degradation requires cereblon.

(A-B) Lenalidomide competition assay. KOPT-K1 cells were treated with various concentrations of SJ11646 (A) or dasatinib (B) in the presence or absence of lenalidomide (10μM) for 72 hours and CTG assay was performed to measure cell viability. (C-D) Western blot of LCK in KOPT-K1 cells treated with two concentrations of SJ11646 (C) or dasatinib (D) in media with and without 10μM lenalidomide. (E) KOPT-K1 cells with CRBN^KO were treated with dasatinib or SJ11646 at 100nM for 24 hours and proteins were collected for immunoblotting. (F-G) Sensitivity of CRBN^KO cells to SJ11646 and dasatinib. CRBN^WT and CRBN^KO KOPT-K1 cells were treated with multiple concentrations of SJ11646 (F) or dasatinib. (G) for 72 hours and cell viability was quantified by CTG assay. (H) Structure of negative control for SJ11646. An analog of SJ11646 (SJ11646_neg.) was synthesized with the PG moiety modified (marked with arrow). (I) Drug response curve of SJ11646_neg. KOPT-K1 cells were treated with multiple doses of negative control (SJ11646_neg.) for 72 hours and CTG assay was used to measure cell viability. (J) KOPT-K1 cells were incubated with serial concentrations of SJ11646_neg. for 24 hours and LCK and pLCK (Y394) concentrations were determined by western blotting. (K) An in vitro AlphaLISA assay was performed to evaluate the stability of LCK-PROTAC-cereblon ternary complex. Stability is noted by increased AlphaLISA signal.

Furthermore, CRBN deletion using CRISPR editing (CRBN^KO) blunted the effects of SJ11646 on LCK and pLCK degradation and its cytotoxicity in KOPT-K1 cells (Fig. 4E–F, fig. S7), but did not alter cell sensitivity to dasatinib(Fig. 4G). Finally, we synthesized an N-methylated analog of SJ11646 in which the N-methyl group is positioned to abrogate cereblon binding affinity (SJ11646_neg., Fig. 4H). This negative control molecule exhibited reduced cytotoxicity in T-ALL cells, comparable to that observed with SJ11646 in the presence of lenalidomide and in CRBN^KO cells (Fig. 4I). SJ11646_neg. did not induce LCK degradation, with only residual suppression of pLCK (Fig. 4J). Finally, SJ11646 showed a high amplitude signal in AlphaLISA assay confirming ternary complex formation with LCK and cereblon, as indicated by the characteristic bell-shaped response (Fig. 4K). Taken together, SJ11646 induced LCK degradation and T-ALL apoptosis in a cereblon-dependent fashion.

SJ11646 shows improved pharmacodynamic profile relative to dasatinib

PROTACs are often reported to have prolonged therapeutic effects due to the irreversible proteolytic elimination of their protein targets, as compared to transitory blockade of signal transduction by small-molecule inhibitors(24). To evaluate SJ11646 in this context, we performed an in vitro wash-out experiment where KOPT-K1 cells were treated with PROTAC or dasatinib at 100nM for 18 hours before the drugs were removed. Cells were then cultured in drug-free media, and cell growth was monitored daily for over 10 days (Fig. 5A). The vehicle-treated cells expanded exponentially with time whereas both dasatinib and SJ11646 caused growth inhibition (Fig. 5B–C). Dasatinib-treated cells began to recover at 96 hours post drug removal but the growth of cells in the SJ11646 group was suppressed for up to 240 hours (Fig. 5B–C). In parallel, the amount of pLCK gradually rose after dasatinib removal and returned to baseline at 120 hours, whereas pLCK remained undetectable in the SJ11646-treated group for at least 240 hrs (fig. S8).

Figure 5. SJ11646 induces prolonged suppression of LCK signaling in vitro and in vivo.

(A) Study design of the in vitro drug wash-out assay. KOPT-K1 cells were treated with dasatinib or SJ11646 at 100nM for 18 hours before drugs were removed. Cells were then incubated in drug-free media with cell growth monitored daily. (B-C) Viable cell number (B) and cell viability (C) were measured by automatic cell counter. Arrows indicate the time when drugs were washed out. (D) Schema for in vivo pharmacokinetic and pharmacodynamic profiling of SJ11646 and dasatinib. Primary T-ALL cells were injected into NSG mice. Once leukemia burden in peripheral blood reached 50%, a single dose of dasatinib or SJ11646 was given by intraperitoneal (i.p.) injection. Bone marrow cells were harvested for LCK and pLCK (Y394) quantification by Western blotting. (E-F) Western blotting for LCK and pLCK (Y394) in bone marrow leukemia cells post treatment with dasatinib (E) or SJ11646 (F). Cells were collected 0, 2, 8, 24, 48 and 72 hours after drug administration. (G-H) Quantification of LCK and pLCK (Y394) normalized to GAPDH based on Western blots in Panels E and F, respectively. Data are shown as mean and SEM (n=3). I-J. Pharmacokinetic and pharmacodynamic modeling of SJ11646 and dasatinib in T-ALL xenografts. Drug concentration in plasma was determined using LC-MS (right y-axis). pLCK was quantified using Western blotting and is labeled on left y-axes. Pharmacokinetic analysis was performed using one-compartment model with first-order absorption and linear elimination. Blue curve and shaded regions represent median, 10^th and 90^th percentile of model-estimated pLCK amount. Black circles are measured plasma drug concentrations. Black solid curves show the median model estimated dasatinib (I) or SJ11646 (J) concentrations.

Next, we characterized the in vivo pharmacokinetic and pharmacodynamic properties of SJ11646 using PDX models of T-ALL and included dasatinib for comparison. In this experiment, we treated T-ALL bearing NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice with a single intraperitoneal injection of dasatinib or SJ11646 at the same molar dosage (10 mg/kg for dasatinib and 15 mg/kg for SJ11646). Plasma samples were collected to measure drug clearance, and leukemia cells were harvested from bone marrow to measure LCK degradation and phosphorylation as the pharmacodynamic endpoints (Fig. 5D). Both dasatinib and SJ11646 exhibited relatively rapid elimination in vivo, with a t_1/2 of 0.92 (half life time, relative standard error (RSE)(34): 2%) and 1.3 (RSE: 5%) hours estimated using a one-compartment model. The peak concentration (C_max) was also comparable between the two, with the area under curve (AUC) at 1,402 (RSE: 5%) and 1,382 (RSE: 6%) ng·hour/mL, respectively (Table S4). Upon injection, both compounds induced rapid and complete loss of pLCK, as early as 3 hours (Fig. 5E–F). However, LCK phosphorylation started to recover at ~8 hours in dasatinib-treated mice and returned to the pre-treatment amount at 48 hours. On the other hand, SJ11646 produced a prolonged and complete suppression of pLCK lasting at least 24 hours post-injection, with concomitant degradation of LCK (Fig. 5G–5H).

Integrating the pharmacokinetic and pharmacodynamic data, we modeled the exposure-to-response relationship for both dasatinib and SJ11646. That is, we used pLCK as a proxy marker for the net therapeutic effect, and this was then evaluated as a function of plasma drug concentration (Table S4). As Shown in Fig. 5I–J, SJ11646 produced deeper and longer effects on pLCK than dasatinib given at the same molar dosage, despite its slightly faster clearance. The maximal pLCK depletion was 97% (ranging from 95% to 99% for the 10^th and 90^th percentile of simulation) and 87% (85%, 88%) for SJ11646 and dasatinib, respectively, with an EC₅₀, half maximal effective concentration, of 0.07 and 14.5 ng/mL. The duration of >50% pLCK depletion was 52 hours (40, >72) for SJ11646 and 8 (7.5, 8.9) hours for dasatinib. Therefore, a single injection of SJ11646 led to a degree of LCK signaling suppression that was 630% longer than dasatinib.

SJ11646 has superior in vivo anti-leukemic efficacy compared to dasatinib

Because SJ11646 produced deeper and longer-lasting LCK suppression than dasatinib, we reasoned that this would translate into a superior anti-leukemic effect in vivo. Using two independent T-ALL PDX models, we tested both compounds as single-agent therapy and monitored leukemia response over time. Three days after leukemia inoculation, mice were treated daily with vehicle control or equal molar doses of dasatinib (10 mg/kg) or SJ11646 (15 mg/kg) for eight weeks by intraperitoneal injection, whereas the amount of human leukemia in peripheral blood was measured weekly by using flow cytometry (Fig. 6A). In both PDX models, dasatinib and SJ11646 repressed T-ALL growth with clear efficacy measured as leukemia-free survival compared to vehicle-treated mice (Fig. 6B–E). However, SJ11646 showed significantly greater anti-leukemic efficacy than dasatinib (P<0.0001, t-test, for both PDX cases): by the time dasatinib-treated mice reached humane endpoint (75% blast), animals on the SJ11646 arm showed only 12.9% and 0.9% blast in PDX_1 and PDX_2, respectively, with marked increases in survival relative to dasatinib therapy (Fig. 6B–E). We did not observe any significant changes in bodyweight or gross toxicities during SJ11646 treatment (fig. S9).

Figure 6. In vivo anti-leukemic efficacy of SJ11646 in T-ALL PDX models.

(A) Schema of in vivo efficacy study. Primary T-ALL cells from two independent PDX models were injected into NSG mice through the tail vein, and mice were given vehicle control, dasatinib or SJ11646 by i.p. injection daily starting three days post T-ALL inoculation. Leukemia burden in peripheral blood was monitored weekly three weeks after implantation. (B-E) Anti-leukemia efficacy of SJ11646 and dasatinib in two T-ALL PDX models. (B, D) Leukemia growth curve of PDX_1 (B) and PDX_2 (D). Leukemia growth was measured by weekly flow-cytometry and shown as human leukemia blast %, with each curve representing an individual animal (n=5). (C, E) Survival analysis of SJ11646- and dasatinib-treated mice. The probability of survival was estimated for both PDX_1 (C) and PDX_2 (E) after treatment of SJ11646 or dasatinib. P-value was estimated with Cox-regression (n=5).

For both PDX models, T-ALL eventually returned in the majority of mice with clearly acquired resistance to SJ11646 (fig. S10). Relapsed leukemia in PDX_2 showed a substantial downregulation of LCK compared to cells prior to PROTAC treatment, whereas this was not true in PDX_1 (fig. S10). We also performed whole genome seq and identified a list of genomic aberrations unique to relapsed leukemia (fig. S10, Table S5), although no mutations were found in CRBN or LCK. Taken together, the phenyl-glutarimide PROTAC SJ11646 has superior in vivo anti-leukemia efficacy than dasatinib although leukemia can also develop resistance to PROTAC therapy.

Discussion

Targeted protein degradation by hijacking the ubiquitin proteasome system has emerged as a unique paradigm in drug discovery across therapeutic areas(13). Focusing on T-ALL, we developed a series of PROTACs to target oncoprotein LCK using dasatinib as the bait and phenyl-glutarimide as the cereblon binder. Lead PROTAC SJ11646 elicited efficient proteolysis of LCK in T-ALL cell lines and PDX models, with markedly improved anti-leukemic effects in vitro and in vivo relative to dasatinib. Our preclinical pharmacokinetic and pharmacodynamic studies indicate that SJ11646 has a stronger and more prolonged effect on LCK suppression in vitro and in vivo. These LCK-degrading PROTACs are not only potentially exciting therapeutic agents in T-ALL, but also valuable tools for studying the chemical biology of oncoprotein LCK in benign and malignant hematopoiesis.

Generally, the physicochemical properties of PROTACs fall outside the traditional small molecule drug space, which is often reflected in their less favorable pharmacokinetic profiles(16, 35). However, following intraperitoneal administration at an equimolar dosage, SJ11646 and dasatinib presented almost identical systemic exposure, with an AUC of 1402 and 1382 ng·hour/ml, respectively. However, unlike dasatinib, SJ11646 sustained the suppression of pLCK in mice long after it was cleared from systemic circulation. This prolonged pharmacodynamic effect translated into a superior leukemia-free survival but also raised concerns about potentially increased toxicity. Encouragingly, we found that SJ11646 was not more cytotoxic than dasatinib in CD34⁺ cells or normal mononuclear cells in the blood. Because LCK signaling is essential for T cell activation, we hypothesized that activated T cells would be highly susceptible to SJ11646, relative to dasatinib. However, this was not the case, and we reason that T-ALL must have a much higher dependency on LCK than that seen during normal T cell activation.

SJ11646 almost completely retained the kinase selectivity of dasatinib, despite the addition of linker and the phenyl-glutarimide moieties. Prior reports of the use of a promiscuous kinase inhibitor showed increased selectivity from parent inhibitor to PROTAC(23) The structural determinants of the substrate specificity of dasatinib-based PROTAC may be multifactorial, such as ternary complex stability, lysine availability, and future studies are warranted to carefully define these features. On the other hand, this also suggests that our PROTAC may be effective in targeting a great number of kinases beyond LCK, particularly notable is ABL, an important therapeutic target in chronic myeloid leukemia and Ph⁺ ALL. There is growing interest in developing dasatinib-based PROTACs for BCR-ABL degradation as PROTACs can abolish the potential scaffolding function of this fusion protein(22, 36–39). These previously reported molecules were IMiD-based(22, 23), and optimized mostly by focusing on the BCR-ABL degradation, but they rarely exhibited superior cytotoxicity to dasatinib(36, 37, 39). Our phenyl glutarimide-based PROTAC showed enhanced cytotoxicity in BCR-ABL positive SUP-B15 cells, pointing to its potential value in treating BCR-ABL positive cancers. Beyond BCR-ABL, other oncogenic proteins such as KIT, for gastrointestinal stromal tumor(40), and DDR1, for several types of solid tumors,(41–44) could also be targeted by SJ11646 (Table S2) and can potentially benefit from PROTAC-based therapy. Further studies will be required to confirm the therapeutic effects of SJ11646 in these cancers, and it remains unclear whether it would produce better efficacy than dasatinib as seen in T-ALL.

There are a number of limitations of this study. First, systematic in vivo toxicity evaluation of SJ11646 is needed, beyond the limited in vitro assays performed here. In particular, dasatinib has been linked to T cell inactivation during immunotherapy(45) and this should be examined carefully with LCK PROTACs. A potential strategy to overcome PROTAC toxicity is to take advantage of tissue-specific expression of different E3 ligases. By switching to a ligand of a tumor-specific E3 ligase, we can improve therapeutic effects in malignant cells and also minimize the toxicity to normal tissues. Second, even though our phenyl-glutarimide PROTAC SJ11646 showed improved potency over IMiD-based molecules, the mechanistic basis of these differences is unclear. Analogs of the lead compound with additional structural variations should be tested to identify features contributing to dasatinib PROTAC activity. Lastly, our pharmacokinetic and pharmacodynamic studies of SJ11646 were performed with the compound administered by intraperitoneal injection which bypasses drug metabolism in the liver. If the oral bioavailability of SJ11646 is poor, its advantage over dasatinib would be limited because oral formulation of dasatinib is already highly effective.

In conclusion, we have developed SJ11646, a phenyl glutarimide-based PROTAC, as a unique LCK targeting agent in T-ALL. The superior in vitro and in vivo potency of SJ11646 compared to dasatinib points to the value of targeted protein degradation in cancer drug discovery. Further optimization of this lead compound may improve its pharmacological properties for clinical development and strategies to rationally combine LCK targeting agents with other chemotherapeutics should also be evaluated carefully.

Materials and methods

Study Design

The objective of this study is to develop proteolytic agents targeting LCK in T-ALL. We synthesized a series of PROTAC molecules and evaluated their biochemical properties and therapeutic effects using NMR, western blot, proteomic profiling, kinase binding screening, in vitro drug sensitivity profiling, pharmacokinetic and pharmacodynamic profiling, in vivo drug efficacy test, etc. We obtained primary T-ALL samples and developed PDX models, with informed consent from all participants as approved by the Institutional Review Board of St. Jude Children’s Research Hospital (reference number: XPD-17-089). All in vivo experiments are performed following the guidance of the Institutional Animal Care and Use Committee. Mice were randomized before each experiment but investigators were not blinded to group allocation. The health status of each mouse was monitored daily, and mice that met the predefined study endpoints were humanely euthanized. For other assays, at least two independent experiments were performed, with biological replicates or triplicates.

LCK PROTAC synthesis

Unless specified, reagents and solvents were obtained from commercial suppliers and used without purification. Reactions were typically set up under air and carried out under nitrogen atmosphere. Thin-layer chromatography was performed using either Merck Millipore silica gel 60G F₂₅₄ glass plates or Biotage KP-NH plates and visualized with a 254 nm UV lamp for detection. Automated flash column chromatography was carried out on a Biotage SP1 flash chromatography system using Sfar silica gel, Sfar amino, or SNAP C18 columns. Evaporation was conducted using a Büchi Rotovapor R-205. NMR spectra were acquired on either a Bruker 400 MHz or Bruker 500 MHz spectrometer in the solvents indicated, and the spectra were processed using MestReNova (12.0) with chemical shifts (ppm) referenced to the solvent peak. Signals are designated as: s, singlet; br s, broad singlet; d, doublet; dd, doublet of doublets; ddd, doublet of doublet of doublets; ddt, doublet of doublets of triplets; dtd, doublet of triplets of doublets; t, triplet; td, triplet of doublets; tdt, triplet of doublets of triplets; q, quartet; qd, quartet of doublets; qt, quartet of triplets; p, pentet; m, multiplet. Coupling constants (J) are in hertz (Hz). Final compound purity (>95% by UV/ELSD unless specified otherwise) was assessed using UPLC-MS (Acquity PDA detector, Acquity SQ detector and Acquity UPLC BEH-C18 column 1.7 μm, 2.1 × 50 mm [Waters Corp]) with mobile phase of 0.1% formic acid in H₂O and acetonitrile. High-resolution mass spectral data was obtained using a Waters Xevo G2 QTof mass spectrometer.

Details for the synthesis of all PROTACs are described in the Supplementary Methods.

Liver microsomes stability assay

Mouse liver microsomal degradation is determined using multiple time points to monitor the rate of disappearance of the parent compound during incubation. NADPH regenerating agent solutions A and B and mouse liver microsomes (CD-1) were obtained from fisher Scientific (Woburn, MA). Pooled human liver microsomes were purchased from XenoTech. Ninety-six deep well plates were obtained from Midsci (# P-DW-11-C). Ninety-six analytical plates were obtained from Corning Incorporated (#3363). Sample preparation for microsomal stability was modified from prior publications(46). A set of incubation times of 0, 15, 30, 60, 120, and 240 min were used. DMSO stock solutions of test compounds and verapamil (system control) were prepared at the concentration of 10 mM. Concentrated mouse liver microsomes (20mg/mL protein concentration) and 0.5 M EDTA were diluted into 0.1 M potassium phosphate buffer (pH7.4) and mixed well, followed by compound solutions being spiked. This solution was mixed and 90 μL was transferred to 6-time points plates (each in triplicate wells). For the time 0 h plate, 3 fold (v:v) cold acetonitrile with internal standard (40 ng/ml warfarin) was added to each well, followed by the addition of NADPH regenerating agent (mixing NADPH solutions A and B in PBS, pH7.4) and no incubation. For other five-time points’ plate, NADPH regenerating agent was added to each well to initiate the reaction, the plate was incubated at 37° C for the designed time point, followed by quenching of the reaction by adding a 3-fold volume of cold acetonitrile with internal standard to each well. The final concentration of each component applied in this reaction was liver microsome protein at 0.5 mg/mL, EDTA at 1 mM, compound at 10 μM, NADPH A at 1.3 mM, and NADPH B at 0.4 U/of the plates were sealed and mixed well and were centrifuged at 4000 rpm for 20 min. The supernatants were transferred to analytical plates and diluted by millipore water appropriately for analysis by LC–MS/MS. Conditions for SCIEX Qtrap UHPLC-MS/MS system was described separately. The metabolic stability is evaluated via the half-life from least-squares fit of the multiple time points based on first-order kinetics.

Caco-2 permeability assay

High throughput Caco-2 permeability was performed using the 96-well Transwell 0.4 μm polycarbonate membrane 96-well system with a modified method(47). Caco-2 cells were maintained at 37°C in a humidified incubator with an atmosphere of 5% CO₂. The cells were cultured in 75 cm² flasks with Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% fetal bovine serum (FBS), 1% non-essential amino acids (NEAA), 100 units/ml of penicillin, and 100 μg/ml of streptomycin. The Caco-2 cells were seeded onto inserts at a density of 0.165×10⁵ cells/insert. The medium in the wells were exchanged every other day, and the trans epithelial electrical resistance (TEER) value was measured using an automated TEER measurement system (World Precision Instruments Inc.). Caco-2 cells were grown for 7 days to reach consistent TEER values.

For transport experiments, each cultured monolayer on the 96-well plate was washed twice with a transport buffer (10 mM HBSS/25mM HEPES, pH 7.4). The permeability assay was initiated by the addition of each compound solution (10 μmol/L) into inserts (apical side, A) or receivers (basolateral side, B). Caco-2 cell monolayers were incubated for 2 h at 37°C. Fractions were collected from receivers (if apical to basal permeability; A→B) or inserts (if basal to apical permeability; B→A), and concentrations were assessed by UPLC/MS (Waters). All compounds were tested in triplicates.

The A→B (or B→A) apparent permeability coefficients (Pappa, cm/s) of each compound were calculated using the equation, Papp=dQ/dt×1/AC₀. The flux of a drug across the monolayer is dQ/dt (μmol/s). The initial drug concentration on the apical side is C₀ (μmol/L). The surface area of the monolayer is A (cm²). The efflux ratio was determined by dividing the Papp in the B→A direction by the Papp in the A→B direction. An efflux ratio >2 suggested that a given substrate was actively transported across the membrane.

In vitro sensitivity profiling of LCK PROTACs using CTG assay

For ALL cell lines (KOPT-K1, Jurkat, and SUP-B15), human CD34⁺ cells, peripheral blood mononuclear cells (PBMCs), resting and CD3/28 activated T-cells, CTG (CellTiter-Glo) assay was used to determine their sensitivity to dasatinib and PROTACs. CD34⁺ cells were thawed and cultured for 7 days prior to cell viability assay. On day 0, cells of ALL cell lines or T-cells were collected and resuspended in RPMI1640 supplemented by 10% FBS at the density of 62,500 cells/ml. CD34⁺ cells were resuspended in DMEM (Gibco, #12430–062) supplemented with 10% FBS at the density of 250,000 cells/ml and PBMCs were resuspended in DMEM (10% FBS) at 416,667 cells/ml. For ALL cell lines, T-cells, and PBMCs, 24μl cell suspension was plated on 384-well plates (1,500 cells/well, 5,000 and 10,000 cells/well, respectively) and 80ul of CD34⁺ cell suspension was plated on 96-well plates (20,000 cells/ml). Drug stock was thawed at room temperature, and working solution was made by serial dilution with the corresponding medium and 6 or 20μl drug solution was added to cell suspension in 384- or 96-well plates. Vehicle control and blank (medium control) were set appropriately. After three days of incubation at 37°C/5% CO₂, 30/100μl of CTG solution (Promega, #G9241) was added to each well of 384-/96-well plates. Cells were then incubated at room temperature on a shaker for five minutes before measuring the luminescence.

Measurement of PROTAC-mediated LCK degradation

Untreated and treated cells were harvested and washed once with ice-cold phosphate-buffered saline (Gibco, 10010–023). Cell pellets were lysed with RIPA lysis and extraction buffer (Thermo Scientific, #89901) supplemented with protease and phosphatase inhibitor cocktail (Thermo Scientific, #78440). Protein lysates were incubated on ice with gentle shaking for 15 minutes before being centrifuged at 4°C/15,000 rpm for 15 minutes. Supernatants were transferred into new centrifuge tubes, and an equal volume of 2-mercaptoethanol (Bio-Rad, #1610710) supplemented 2x laemmli sample buffer (Bio-Rad, #1610737) was added to protein lysate. Protein samples were heated at 99°C before being stored at −20°C freezer or western blotting. Equal amounts of protein samples were run on precast 4–15% Tris-glycine Mini-PROTEAN TGX gels (Bio-Rad, # 4561086). Resolved proteins were transferred onto Immobilon-FL PVDF membranes (Millipore, #IPFL00010). The membranes were blocked with Intercept (TBS) blocking buffer (LI-COR, #927–60001) for one hour at room temperature. The membranes were then probed with primary antibodies (Cell Signaling, #2657, 6943, 5714) at an optimal concentration (dilution factor of 1:1000 for LCK and pLCK, 1:10000 for GAPDH) in the same blocking buffer supplemented with 0.2% Tween 20 (Fisher BioReagents, #BP337–500) overnight at 4°C. The membranes were washed with TBST three times (10 minutes each time on a shaker) and incubated with the IRDye 800CW goat anti-rabbit and goat anti-mouse IgG secondary antibodies (LI-COR, #926–32211 and 926–32210) at room temperature for 2 hours. Excessive antibodies were washed with TBST, and the membranes were exposed in LI-COR Odyssey imaging system. Immunoblotting data were analyzed with Image Studio software (lite version 5.2). For the detection of other proteins, the same procedure was followed with substitutions of the primary antibodies to the following: SRC family kinases (CST, #9320T); ABL1 (CST, #2862S); Cereblon (CST, #71810S).

Proteomic profiling and data analysis

KOPT-K1 cells were lysed, and the protein concentrations were quantified as previously described(48). Lysis buffer (50 mM HEPES, pH 8.5, 8 M urea, and 0.5% sodium deoxycholate, 1:10 v/v) were added into the frozen samples with 1x PhosSTOP phosphatase inhibitor cocktail (Sigma-Aldrich, #4906845001). Protein concentration was measured by the bicinchoninic acid assay (BCA) assay (Thermo Fisher, # 23225) and then confirmed by Coomassie-stained short sodium dodecyl sulfate (SDS) gels.

Approximately 1mg quantified protein samples in the lysis buffer with 8 M urea for each TMT channel were digested first with Lys-C (Wako, #125–05061, 1:100 w/w) at 21°C for 3 h, then diluted 4-fold to reduce urea to 2 M for trypsin digestion (Promega, #VA1060, 1:50 w/w) at 21°C overnight. Reduction and alkylation were carried out by adding DTT and IAA. Trifluoroacetic acid was added to 1.5% to acidify the solution and precipitate lipids. After centrifugation, the supernatant was desalted with the Ultra-micro spin c18 column (Harvard apparatus, # 74–7242), and then dried by Speedvac. Each sample was resuspended in 50 mM HEPES (pH 8.5) for TMT labeling, and then mixed equally. The reaction was then quenched by addition of 5% hydroxylamine and desalted using the Sep-Pak C18 cartridge (Waters, # WAT036945).

The TMT labeled samples were fractionated by offline basic pH reverse phase liquid chromatography (LC), and each of these fractions was analyzed by the acidic pH reverse phase Liquid chromatography–mass spectrometry (LC-MS/MS)(49). 80 fractions were collected by offline basic LC (3 hr gradient on a XBridge C18 column, 3.5μm particle size, 4.6 mm × 25 cm, Waters; buffer A: 10 mM ammonium formate, pH 8.0; buffer B: 95% acetonitrile, 10 mM ammonium formate, pH 8.0), and then concatenated into 40 fractions(48).

For acidic pH LC-MS/MS analysis, each fraction was run sequentially on a ConAnn C18 column (75μm × 20 cm, 1.7μm particle size, 65°C to reduce backpressure) interfaced with an Q Exactive HF Orbitrap MS (Thermo Fisher). Peptides were eluted by a 90 min gradient (buffer A: 0.2% formic acid, 5% DMSO; buffer B: 67% acetonitrile, 0.2% formic acid, 5% DMSO). MS settings included the MS1 scan (450–1600 m/z, 60,000 resolution, 1 × 106 AGC and 50ms maximal ion time) and 20 data-dependent MS2 scans (fixed first mass of 120 m/z, 60,000 resolution, 1 × 105 AGC, 105ms maximal ion time, HCD, 32% normalized collision energy, 1.0 m/z isolation window with 0.2 m/z offset, and ~15 s dynamic exclusion).

The identification was performed with the in-house JUMP search engine for improved sensitivity and specificity(50). Major parameters included precursor and product ion mass tolerance (±15 ppm), full trypticity, static mass shift for the TMT tags (+304.20714) and carbamidomethyl modification of 57.02146 on cysteine, dynamic mass shift for Met oxidation (+15.99491), maximal missed cleavage n = 2, and maximal modification sites n = 3. Putative PSMs were filtered by mass accuracy and then grouped by precursor ion charge state and filtered by JUMP-based matching scores (Jscore and ΔJn) to reduce FDR below 1% for proteins during the whole proteome analysis.

AlphaLISA Assay (LCK-CRBN/DDB1)

His-tagged CRBN-DDB1 protein was prepared following the procedure reported by Matyskiela et al(51) and GST-tagged LCK (1–509) protein was acquired from GeneTex. The assay mixture contained 240 nM His-tagged CRBN-DDB1, 60 nM GST-tagged LCK (1–509), 20 μg/mL nickel chelate AlphaLISA acceptor, and glutathione donor beads (PerkinElmer, # 6765300) and serially diluted test compounds in a buffer comprising 25 mM HEPES, pH 7.4, 100 mM NaCl, 0.1% BSA, 0.05% tween20, and 0.005% proclin-300. A 96-well plate containing 5× test compound, 5× His-tagged CRBN/DDB1, and 5× GST-tagged LCK was incubated at rt for 1 h. After the incubation, 5 μL of the solution was transferred to a 384-well white AlphaPlate (PerkinElmer) in duplicate followed by 10 μL nickel chelate AlphaLISA acceptor (1:100) and 10 μL glutathione donor beads (1:100). The plate was sealed and mixed on a MixMate (Eppendorf) for 60 min at room temperature and then luminescence detection was collected on an Envision plate reader (PerkinElmer).

CRISPR screen for essential genes in KOPT-K1 cell line

Genome-wide CRISPR/Ca9 screen was performed to identify genes essential for T-ALL survival, using KOPT-K1 cell line as the model system. Experimental procedure was based on previously published protocol(52) with minor modifications. Briefly, Cas9-expressing KOPT-K1 cells were first generated by lentiviral transduction, and single cell dilution was performed to develop and select clones with high level of Cas9 for genome-wide CRISPR screen. Cas9-expressing KOPT-K1 cells were transfected with the GeCKO V2 sgRNA library (Addgene, #1000000048) at low multiplicity of infection (0.3). Transfected cells were selected with puromycin for two days and then cultured for another 6 days to allow proliferation. Genomic DNA was extracted from surviving cells and sgRNAs were amplified by PCR followed by Illumina sequencing(53). The sgRNA enrichment was quantified based on read counts, and fold change from baseline (day 0) to day 6 was used for estimate gene dependency score(53).

Pharmacokinetic (PK) and pharmacodynamic (PD) modeling

For pharmacokinetic modeling, non-linear mixed effects modeling analysis was performed with Monolix (version 5.1.0) using the Stochastic Approximation Expectation-Maximization (SAEM) method. A one-compartment PK model with first-order absorption and linear elimination was used to model dasatinib/SJ11646. The pharmacokinetic parameters estimated included: ka (1/hrs), the absorption rate constant; ke (1/hrs), the elimination rate constant; and V/f (ml/kg), apparent volume. The parameter “f” is the unknown bioavailability. The inter-individual variability of the parameters was assumed to be log-normally distributed. A proportional residual error model was used with assumed normal distribution of the residuals. The median and 90^th percent prediction interval for AUC_0->24hr, AUC_0->infinity, C_max, T_max (time to peak drug concentration), t_1/2, and clearance were estimated using n=100 sets of pharmacokinetic parameters randomly sampled from the population pharmacokinetic parameter distribution.

The dynamics of pLCK in the presence of dasatinib or SJ11646 are modeled with an indirect response model where the drug de-phosphorylates LCK. Specifically, the rate constant describing de-phosphorylation in the following model (k_off) is an increasing function of the drug concentration. The model is defined as follows:

Where:

$$\begin{gathered} \frac{\mathit{\text{dpLCK}}}{dt}=k_{on}-k_{off}\cdot \left(1+\frac{E_{MAX}\cdot C^n}{E{C}_{50}^n+C^n}\right)\cdot \mathit{\text{dLCK}} \\ {k}_{on}=pLC{K}_{\mathit{\text{baseline}}}\cdot k_{off} \end{gathered}$$

In the above equation k_on is the rate constant describing phosphorylation of LCK, k_off is the rate constant describing de-phoshorylation of pLCK. The equation relating k_on to k_off maintains steady-state concentrations of pLCK (pLCK baseline) in the absence of drug. The effect of dasatinib or SJ11646 concentration (C: determined using the above PK model) is described by the saturable Hill function where E_MAX is the maximum effect, EC₅₀ is the concentration that causes a 50% of maximum effect, and n is the Hill coefficient that quantifies the steepness of the Hill function.

The PD model parameters were estimating with maximum likelihood estimation using a naive pooled data approach by first estimating the PK model parameters, then fixing these PK parameters and then estimating the PD parameters (MATLAB).

In vivo efficacy evaluation

For in vivo efficacy evaluation of SJ11646, primary human T-ALL cells from each case were injected into female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice between 8–12 weeks of age through tail vein (2 million cells/mouse, resuspended in 200μl sterile PBS). Health status of all injected mice were monitored every day. SJ11646 or dasatinib was given to T-ALL bearing NSG mice starting at three days post injection. SJ11646 was dissolved in 5% NMP + 40% PEG400 + 5% Solutol HS-15 + 0.5% polyvinyl alcohol (50%) + 49.5% saline; Dasatinib (LC Laboratories, #D-3307) was dissolved in 4% DMSO + 30% PEG300 + 5% Tween 80 + 61%. SJ11646 and dasatinib were administrated once daily through intraperitoneal injection at 15 mg/kg and 10 mg/kg, respectively. Starting from two weeks after injection, peripheral blood was obtained by retro-orbital bleeding and subjected to flow cytometry to determine the level of human leukemia (humanCD45 and humanCD7 double positive) weekly. Mice with human blast percentage over 75% in blood or sign of moribund were considered lethal in survival analysis and euthanized.

Genomic profiling

For both T-ALL PDX_1 and PDX_2, we performed whole-genome seq (WGS) of baseline and relapse leukemia (following SJ11646 treatment). Libraries were constructed using Kapa Hyperprep kit (Roche) and sequenced for 2 × 151 bp pair-end reads. WGS analyses were performed following procedures established previously(54). Sequencing reads were mapped to GRCh38 reference using bwa(55) and sorted by coordinates using samtools(56). Duplicated reads were removed by Picard and excluded from downstream variant calling. Single nucleotide variants and short indels were called by Mutect2(57), SomaticSniper(58), Varscan2(59), MuSE(60) and Strelka2(61). Coding variations were manually reviewed, and those present in relapse leukemia but absent in their corresponding baseline samples (prior to SJ11646 treatment) were summarized using Oncoplot.

Statistical analysis

All statistical tests were two-sided and were chosen according to data distribution as described wherever appropriate; the threshold for statistical significance was defined as P < 0.05.

Details for all other experiments are described in the Supplementary Methods.

Data and materials availability

All data associated with this study are present in the paper or supplementary materials. Requests for materials should be sent to the corresponding authors and material transfer agreements are required.