ABL001

Global PROTAC Toolbox for Degrading BCR−ABL Overcomes Drug- Resistant Mutants and Adverse Effects

ABSTRACT: The BCR−ABL fusion oncoprotein causes chronic myeloid leukemia or acute lymphoblastic leukemia in Ph+ patients because the ABL kinase is constitutively activated. However, current clinical treatment with ABL inhibitors is seriously limited by drug resistance and adverse effects. Although the emerging proteolysis-targeting chimeras (PROTACs) have been introduced to degrade BCR−ABL, most of them showed limited activity and could not overcome the common drug-resistant mutants, especially for T315I mutant. Herein, we systematically designed a set of unique PROTACs by globally targeting all the three binding sites of BCR−ABL, including dasatinib-, ponatinib-, and asciminib- based PROTACs. Our ponatinib-based PROTACs showed practical activity as dasatinib-based PROTACs, while no reported ponatinib-based PROTACs could degrade BCR−ABL before. As a proof of concept, some additional dasatinib-based PROTACs were then designed to degrade T315I mutant too. We provided a global PROTAC toolbox for degrading both wild-type and T315I-mutated BCR−ABL from each binding site. More importantly, these PROTACs showed better selectivity and less adverse effects than the inhibitors, indicating that PROTACs had great potential for overcoming clinical drug resistance and safety issues.

INTRODUCTION
The Philadelphia (Ph) translocation between human chromo-some 9 and chromosome 22 generates BCR−ABL fusion oncoprotein, which leads to the constitutive activation of ABL tyrosine kinase and thus causes unregulated proliferation of cancer cells in all chronic myeloid leukemia (CML) and some Ph chromosome-positive (Ph+) acute lymphoblastic leukemia (ALL).1 As the first tyrosine kinase inhibitor (TKI) and first- generation ABL inhibitor, imatinib has significantly improvedthe clinical outcomes and is utilized as the first-line treatment for patients with CML.2 Although it becomes the paradigm for targeted cancer therapy, imatinib cannot help approximately 40% of patients because of the intolerance and drug resistance.3 The BCR−ABL mutations, especially for T315I, and overexpression seriously contribute to the drug resistance.4 To some extent, the advent of second-generation (nilotinib, dasatinib, and bosutinib) and third-generation (ponatinib) ABL inhibitors with reportedly lower mutations provides multiple options for patients.3 However, the new TKIs could not inhibit all the resistant mutants; on the other hand, theadverse effects, such as vascular disease for ponatinib and pulmonary hypertension for dasatinib heavily limit their utilization.3 In particular, ponatinib, the only targeted drug for T315I mutant of BCR−ABL, was transiently removed from the market in 2013 due to its serious vascular adverse events (VAEs) and later reauthorized with a revised indication statement and a black box warning.5 So far, there is no newapproved drug for targeted therapy of T315I mutant. Therefore, new advances are expected to meet the clinical needs.With potential to address the challenges in current drug development programs, the rapidly emerging proteolysis- targeting chimera (PROTAC, Figure 1A) technology utilizes heterobifunctional small molecules to recruit an E3 ubiquitin ligase and target protein.

Upon ternary complex formation, the E3 ligase-mediated ubiquitination on target protein led to subsequent degradation by the proteasome.6 In most cases, inhibitors could only interfere with single function of the multiple functional proteins, while PROTACs induce the protein degradation and interfere with all protein functions. On the other hand, PROTACs need further chemical modifications on original ligands and the participation of multiple cellular machineries in several steps which may provide the opportunity for selectively degrading homology proteins and avoiding off-target effects. These advantages expand the drug space and probably reduce adverse effects sothat efforts have been made towards the development of PROTAC-targeting BCR−ABL. According to their binding sites, ABL inhibitors can be divided into five types (type I−V),among which type I, type II, and type IV have been paid more attention (Figure 1B).7 Although BCR−ABL-targeting PRO- TACs based on these inhibitors have been investigated (Figure1C),6,8−11 only type I-based PROTACs give meaningful potency for wild-type K562 CML cells but they do not work for T315I mutant. The degrading efficiency of other type PROTACs is at the submicromolar or micromolar level. Especially, for type II-based PROTACs, only imatinib was used as the moiety for recognizing BCR−ABL,11,12 whereas degradation could only be observed over 30 μM on wild- type cells. Ponatinib was also prepared as a PROTAC,13 but any degradation of BCR−ABL by it was never reported. Herein, we believe that it is likely to develop potent PROTACs to overcome the drug-resistant mutants by quickly exploring the global recognizing sites on BCR−ABL with our new probes from click chemistry. What’s more, these PROTACs have potential for reducing some serious adverse effects of the inhibitors.

RESULTS AND DISCUSSION
Global Design, Synthesis, and Evaluation of aToolbox for Degrading BCR−ABL. To outperform state of the art of BCR−ABL-targeting PROTACs, we envisioned that the opportunities might be seized by globally and systematically exploring the binding sites of BCR−ABL. The alkyne tags were designed to be installed on heterocycles ofimatinib (type II inhibitor), dasatinib (type I inhibitor), asciminib (type IV inhibitors), and ponatinib (type II inhibitor) that were exposed to solvent (Figure 1C, Schemes 1-4). Then, click chemistry with cereblon (CRBN) ligands, varied pomalidomide derivatives, was performed to synthesize the target PROTACs quickly (Figure 2A, Schemes 1-4). PEGylated linkers were chosen due to their good compatibility with all three binding pockets.6−11 On the other hand, PEGylated linkers usually have good biocompatibility andimprove water solubility of compounds. Some key inter- mediates were first prepared before the final click chemistry. For the imatinib-based PROTACs (Scheme 1, compounds 1− 5), intermediate 21 was synthesized with aniline and acyl chloride building blocks (Scheme 1), which was then substituted by piperazine to generate intermediate 22. After amide condensation, the alkyne-tagged intermediate 23 was prepared. For the dasatinib-based PROTACs (Scheme 2, compounds 6−10), intermediate 29 was generated via an aromatic nucleophilic substitution reaction on which the alkyne tag was installed via amide condensation too. Intermediate 30 was then used in the click chemistry. For the asciminib-based PROTACs (Scheme 3, compounds 11− 15), compound 31 was first prepared with acyl chloride, followed by the generation of compound 32. After an aromatic nucleophilic substitution reaction, the secondary amine was installed on the pyridine ring. The −Boc group of compound 32 was then removed by trifluoroacetic acid (TFA), and a Suzuki cross-coupling was conducted to generate compound 34, which was then modified with alkyne.

After the deprotection of compound 35 under acidic condition, compound 36 was obtained and used for the synthesis of final PROTACs. To prepare the ponatinib-based PROTACs (Scheme 4, compounds 16−20), amide condensation was performed first to prepare compound 38, followed by reduction of the nitro group. Another amide condensation was conducted to generate intermediate 41, which was then used for the click chemistry for preparing the final PROTACs. After evaluation via western blots on wild-type K562 cells, the imatinib-based PROTACs did not induce any degradation of BCR−ABL even at 10 μM, as reported before (Figures S1 and 2B, compounds 1−5),6,11 while the dasatinib-basedPROTACs efficiently reduced the BCR−ABL protein levels (Figures S2 and 2B, compounds 6−10 and Figure 2C). Themost potent P22D (compound 10) had a similar degradingpotency with other dasatinib-based PROTACs such as DA-6- 2-2-6-CRBN and SIAIS178.6,10 However, all these dasatinib- based PROTACs should be incapable of degrading T315I mutant due to the loss of binding affinity, only asciminib- and ponatinib-based PROTACs were capable of degrading it.14 In contrast to the moderate degrading activity of asciminib-based PROTACs among which P19As (compound 14) with a half degrading concentration (DC50) of 200 nM was the best (Figures S3 and 2B, compounds 11−16 and Figure 2C), the ponatinib-based PROTACs (Figures S4 and 2B, compounds 17−20 and Figure 2C) showed more practical potency that were close to dasatinib-based PROTACs. The protein level of BCR−ABL was decreased by 50% after treatment with P19P (compound 19) for 32 h at a concentration of 20 nM.

Although these PROTACs exhibited less antiproliferative activities than the corresponding inhibitors (Figure 2D), they were good and practical enough in comparison with the first- line therapeutic imatinib. The more obvious increasing IC50 of P19As (compound 14) could be led by its weaker degrading potency.We observed some valuable points by systematically exploring this global toolbox. First, we believed that thebinding affinity of ligands with BCR−ABL was important for the formation of the ternary complex, which could be concluded by comparing imatinib-based PROTACs with ponatinib-based PROTACs (Figure 2B). They had similar connecting positions with CRBN ligands in type II site, but ponatinib-based PROTACs significantly possessed strongerbinding affinity over imatinib-based PROTACs that were reflected by the corresponding inhibitors’ cellular activities (Figure 2D). Although weak binders meant to be compatible in PROTACs, the well-matched protein−protein interactions between target proteins and E3 ligases might contribute to more binding affinities to the degrading complex in these cases. To mimic the formation of the ternary complex, we conducted protein−protein docking in silico between BCR−ABL and CRBN in which the distances from ABL ligands to CRBN ligands were constrained to less than the linear length of linkers in PROTACs. These constraints played roles like binding forces from PROTACs that allowed target proteins to bind E3 ligases reasonably. The generated models with low- energy states were searched with which the PROTACs were finally docked to obtain the whole ternary complex models (Figure 3). P19P (compound 19) and P19As (compound 14) should induce CRBN to recognize inactive BCR−ABL inwhich conformation of ABL was autoinhibited by interactions of the SH2/3 domains with the kinase domain, while these intramolecular interactions should not be involved in a disassembled state of active conformation.

In consideration of near cellular activities of ponatinib, dasatinib, and asciminib,14 we believed that the degrading difference of their corresponding PROTACs was mainly created by protein−protein interactions in degrading complexes. CRBN interacted with the kinase domain in the models of both P19P(compound 19) and P22D (compound 10) but generating differential interacting surfaces. Its area for P19P (compound 19) was obviously less than P22D (compound 10) (Figure 3, 1949.4 Å2 vs 3032.9 Å2), which might explain the slight weaker degrading potency of P19P (compound 19) than P22D (compound 10). Although the interacting surface area in the model of P19As (compound 14) was as large as P22D (compound 10), the major interactions existed between SH2/3 domains and CRBN (Figure 3C). The SH2/3 domains regulated the activity of the kinase domain by dynamically docking and undocking to it.16 What’s more, there was another domain of BCR−ABL, F-actin binding domain (FABD), close to the interacting surface that might impair the function of the degrading complex by possible protein−protein crashes.17 TheScheme 5. Representative Synthetic Route of Additional Dasatinib Analogues; Compounds 45−49 were Synthesized as Compound 43moderate potency of P19As (compound 14) and other previous asciminib-based PROTACs was likely to be influenced by these factors. On the other hand, ponatinib- based PROTAC with a linker of 9-atom length was prepared in a previous study, but the degradation of BCR−ABL by ponatinib-based PROTAC has never been reported.13 We observed that all of the PROTACs were sensitive to the length of linkers, especially for ponatinib-based PROTACs (Figure 2B). The degrading efficiency of P10P (compound 16) with a linker of 10-atom length and the previous ponatinib-based PROTACs obviously weakened due to the possible protein steric clashes under a short-linker distance.Ponatinib- and Asciminib-Based PROTACs DegradedDrug-Resistant Mutants. We then evaluated their degrading abilities for representative drug-resistant mutants, including T315I at the gatekeeper, E255K at the p-loop, H396R at the activation loop, and V468F at the c-helix (Figure 4A). The most important T315I mutant could directly impair the binding of imatinib, dasatinib, and their PROTACs with ABL, and thus, P19P (compound 19) and P19As (compound 14) were first studied for it.

Both P19P (compound 19) and P19As (compound 14) could decrease the protein level of T315I mutant in BCR−ABL-transformed BaF3 murine cells, but theirDC50 values were obviously larger than those in wild-type human K562 cells (Figures 2B and 4B), which might result from the reduced binding affinity with mutants and species difference of CRBN.14,18 To our delight, P19P (compound 19) still exhibited practical antiproliferative activity (EC50 = 28.5 nM) for T315I mutant (Figure 4C). Although ponatinib showed stronger activity, the obvious inhibitory effects on parental BaF3 cells indicated its serious cytotoxicity for normal cells (Figure 4C). Our asciminib-based PROTACs, P19As (compound 14), also showed better activity than the previous type IV ligand-based PROTACs for T315I mutant (Figure 4C).9 Clinical resistant mutants against asciminib like V468F have been reported so that its utilization was limited.14 Besides T315I, P19P (compound 19) could degrade V468F and other mutants such as E255K and H396R in transiently transfectedHeLa cells that constantly expressed BCR−ABL (Figure 4D). Because more than 100 mutants have been identified for thedrug resistance of ABL inhibitors to date and new mutants emerged continually,8 P19P (compound 19) had great opportunity for overcoming them as ponatinib (Figure 4C).New Dasatinib-Based PROTACs Degraded T315I Mutant. To build a toolbox for degrading BCR−ABL mutants from all three binding pockets, especially for achieving thedegradability for T315I, we decided to develop new dasatinib- based PROTACs. In the binding complexes (Figure 5A), dasatinib bound the active conformation of ABL in which the activation loop oriented “in” toward the ATP-binding pocket, while ponatinib induced inactive conformation where the activation loop oriented “out” from the ATP pocket. However, they shared some common features in binding with the ATP pocket that could guide us to design new analogues of dasatinib to overcome T315I mutant. The unique alkyne group in ponatinib allowed it to insert into the ATP pocket from the opposite side of dasatinib and avoid the steric hindrance from T315I mutant (Figure 1B).

In contrast, the gatekeeper T315I prevented dasatinib from getting inserted into a hydrophobic pocket and reduced its binding affinity a lot (Figure 5A). Inspired by this difference between dasatinib and ponatinib, the resistance to dasatinib in T315I mutant could be overcome by replacing the corresponding amide bond in dasatinib with some hydrophobic small functional groups as the alkyne in ponatinib (Figure 5B). A truncated compound (42) was first synthesized but showed no activity, indicating that theremoved hydrophobic group was important for binding. We then synthesized several analogues of dasatinib (Figure 5B, compounds 43−49 and Scheme 5) by installing the hydro- phobic benzene ring via an alkene group, which could decrease the steric hindrance and increase the hydrophobic interactions with T315I mutant. To prepare compound 42, two successive aromatic nucleophilic substitution reactions were conducted to generate compound 52 at first. After removing the −Boc group, compound 42 was obtained by installing the alkyne group on compound 53 through amide condensation. For compound 43 and 44, Witting reaction was performed first to get compound 57 as a Z/E mixture. The −Boc group was then removed by TFA, followed by two successive aromatic nucleophilic substitution reactions for the synthesis of compound 60. Compound 60 containing Z/E isomers was treated with iodine so that only E isomer remained, which was then treated with TFA and BBr3 to generate compound 61 and 62, respectively. After installing the alkyne group via amide condensation, compounds 43 and 44 were produced. Compounds 45−49 were synthesized as compound 43.Compound 45 was identified as an effective ligand of T315I mutant which inhibited half of the BaF3−BCR−ABL (T315I) cells at 0.57 μM. After coupling with the CRBN ligand via click chemistry, P22D1 (compound 50), a new dasatinib-based PROTAC (Figure 5C), was obtained and tested for degrading T315I mutant.

As expected, both P22D (compound 10) and the previously reported DA-6-2-2-6-CRBN could not induce any degradation of T315I mutant even at 30 μM, but P22D1 (compound 50) obviously decreased the protein level of T315I at 1 μM (Figure 5D). In the antiproliferative assays, no good activities were observed for dasatinib, P22D (compound 10)and DA-6-2-2-6-CRBN, but P22D1 (compound 50) showed better inhibition on BCR−ABL (T315I) stable BaF3 cells than the parental BaF3 cells (Figure 5E, EC50 = 6.1 μM vs EC50 > 10 μM). It means that P22D1 (compound 50) worked by mainly targeting T315I mutant other than some off-targets.PROTAC Decreased Adverse Effects. Many promiscuous kinase inhibitors such as ponatinib have off-target effects by binding to a similar ATP pocket,13 which resulted in adverse effects.19,20 The inhibition of ponatinib on parental BaF3 cells (Figure 4C) has demonstrated its toxicity and off-target effects. The serious cardiovascular toxicity was also evaluated on cardiovascular cells as before.20−22 As shown in Figure 6A,B, ponatinib induced cytotoxicity on rat cardiac myocyte (H9C2) and human umbilical vein endothelial cell (HUVEC) under EC50 values of 379 and 515 nM, respectively. By contrast, P19P (compound 19) did not inhibit these cells even at 10μM. To elucidate the better selectivity of PROTAC over the inhibitor, we tested inhibition to kinases that were closely related with its activity or adverse effects. P19P (compound 19) still showed strong activity on the desired ABL (T315I) kinase, though it was 13-fold weaker than ponatinib (Figure 6C, IC50 = 13 nM vs IC50 = 1.0 nM). It has been hypothesized that ponatinib harmfully interferes with the platelet function through inhibitions of LYN and SRC kinases.23 In our tests, their inhibitions by P19P (compound 19) drastically decreased compared with ponatinib (Figure 6D,E), especially for SRC that was less inhibited by 257-fold (IC50 = 10 nM vs IC50 = 2575 nM).

There was also a report regarding the pathogenesis of ponatinib-associated VAEs, which was resulted from VEGFR2 inhibition.21 The activity of VEGFR2 decreased by 50% in the presence of P19P (compound 19) at 441 nM, while ponatinib inhibited 50% of its activity at 26 nM (Figure 6F). Therefore, ponatinib dramatically inhibited VEGFR2-related HUVEC tube formation at 200 nM in the matrigel-based assay, whereas P19P (compound 19) had no such adverse effect at 1 μM (Figure 6G). Because there are other off-targets of ponatinib, including FLT3, c-KIT, and PDGFR, we then evaluated their degradation by P19P (compound 19, Figure S5). Only FLT3 could be moderately degraded by P19P (compound 19, DC50 is larger than 300 nM), while there was very slight or no degradations for c-KIT, SRC, and PDGFRα (Figure S5). All the above-mentioned data suggested that our PROTACs exhibited better selectivity and safety than inhibitors. Another way to reducing the adverse effects is combination of several drugs with the synergistic effect. We performed a combination study of P19P (compound 19) with other anticancer molecules, including the BET inhibitor ABBV-075, PI3K inhibitor copanlisib, hedgehog pathway inhibitor GDC-0449, and HDAC inhibitor vorinostat,according to the Chou−Talalay method (Figure 6H).24 Thecombination index (CI) was less than 1 around 50% of the fraction affected by the dose (Fa) when P19P (compound 19)and GDC-049 were combined, indicating that they had synergistic effect. What’s more, it was predicted that the synergistic effect remained at higher Fa regions (Figure 6H). There was great potential for further reducing the dosages and adverse effects via synergistic combinations.

CONCLUSIONS
Although PROTACs provide a new strategy for attacking target proteins by easily tethering the ligands of interested proteins with ligands of E3 ligases, their rational designs are challenging due to the complexity of the degrading complex. The ubiquitin of lysine on the target protein’s surface requires precisely matched components in the degrading machinery, and thus, case by case degradability is usually observed. The current development of PROTACs for clinical usage, especially for oral administration, needs extensive optimization as the ligand-based drug development in traditional medicinal chemistry. In the black box of developing drug-like PROTACs, quickly evaluating their degradability and exhibiting their advantages over original inhibitors are keys for further investment in them. Therefore, we prepare a toolbox for degrading BCR−ABL from different binding pockets via click chemistry, which allows us to globally study the degradability
and advantages of PROTACs. The dasatinib- and asciminib- based PROTACs P22D (compound 10) and P19As (compound 14) are equally good as those previously reported,6,9 while ponatinib- and additional dasatinib-based PROTACs P19P (compound 19) and P22D1 (compound 50) are far better than those, especially for degrading T315I mutant. Besides the advances over previous PROTACs, our toolbox also exhibited obvious advantages over inhibitors in some respects.

For instance, dasatinib and asciminib are limited by drug-resistant mutants,14 whereas ponatinib leads to serious adverse effects. P19P (compound 19) overcomes dasatinib-resistant T315I and asciminib-resistant V468F mutants14 and reduces the adverse effects of ponatinib. PROTACs in this toolbox degrade both wild-type and drug- resistant mutants from every binding site, which can be particularly meaningful for cross-resistance and synergistic combinations from different pockets. In addition, something about designing PROTACs of BCR− ABL learnt by their systematic comparison should be emphasized again. To achieve efficient degradation, strong binders are possibly necessary for BCR−ABL because degradation is rarely observed when the ligand is replaced from ponatinib to imatinib (Figure 2B). Because protein− protein interactions induced by PROTACs are crucial for the formation of the degrading complex, neighboring protein structures can influence the degradability positively or negatively when distinct E3 ligases or recognized sites on target proteins are involved.9 The dasatinib- and ponatinib- based PROTACs possess better degradability than asciminib- based PROTACs (Figure 2B), indicating that the ATP-binding pocket of BCR−ABL should be preferred for CRBN-based PROTACs. The advances in this work pave a way to develop more drug-like PROTACs for degrading BCR−ABL in the future. The great potential and opportunity for ABL001 overcoming drug resistance and adverse effects by PROTACs are spreading out.