Selective Macrocyclic WEE1 Kinase Inhibitors with Strong Efficacy against Patient-Derived Colorectal Cancer Organoids

Abstract

Macrocyclization can enhance the selectivity of acyclic compounds toward structurally similar biological targets such as kinases. WEE1 regulates cellular homeostasis and is a promising target in oncology. The clinical candidate AZD1775 (1) failed to progress past Phase II trials because of patient tolerability issues, likely due to off-target inhibition of polo-like kinase 1 (PLK1). Herein, a computer-aided drug design approach was conducted to develop a macrocycle based on the 1-WEE1 X-ray cocrystal structure. Significantly enhanced WEE1 inhibitory selectivity over PLK1 was determined for leading macrocycle 2, which also demonstrated broader kinome-wide selectivity. Patient-derived organoids from colorectal cancer (CRC) peritoneal and liver metastases, treated with 2, demonstrated comparably strong or enhanced anticancer efficacy compared to that of 1. Against patient-matched normal colon vs primary CRC organoids, 2 potently and selectively treated CRC, as well as enhanced DNA damage compared to 1. Finally, the X-ray cocrystal structure of 2 bound to WEE1 validated its computationally predicted bioactive binding mode.

Introduction

Macrocycles are cyclic molecules that typically contain a ring of 12 or more atoms. − The unique structural preorganization of macrocycles is advantageous in drug discovery, as it can enhance potency and selectivity for biological targets of similar catalytic domains, such as kinases. − In some cases, macrocyclization even resulted in mutant-selective kinase inhibitors that do not affect the wild-type enzyme.

Computer-aided drug design (CADD) has emerged as a transformative force in drug discovery, integrating computational chemistry, structural biology, and pharmacology. Comprehensive computational methods have been increasingly employed in drug development campaigns for rationally designing next-generation compounds, especially when drug–protein cocrystal structures are available. This is increasingly pertinent to the field of kinase inhibitor discovery due to the number of readily accessible cocrystal structures. A significant challenge in developing kinase inhibitor drugs is poor target selectivity due to kinase catalytic domains being highly concerved, particularly the ATP-binding pocket. A powerful strategy to address this challenge involves using CADD to derive macrocyclic kinase inhibitors from leading acyclic drug candidates, thereby preorganizing favorable bioactive conformations and ultimately enhancing kinome-wide selectivity. ,

The nuclear kinase, WEE1, is considered to be a master regulator of the cell division cycle by managing signaling pathways associated with the critical G2/M and intra-S DNA-damage repair checkpoints (DDRCs). WEE1 manages the activity of cyclin-dependent kinases, which are essential, along with their complementary cyclin substrates, for progressing cells through the S and G2 phases toward mitosis. −

In response to DNA damage, the ataxia-telangiectasia mutated and Rad3-related (ATR) signaling pathway triggers the G2/M DDRC (Figure ). Here, ATR kinase activates checkpoint kinase 1 (CHK1), which subsequently promotes WEE1 activity via phosphorylation. Active WEE1 then transfers an inhibitory phosphate group on the Y15 residue of CDK1, inactivating the crucial CDK1–cyclin B complex and consequently inducing cell cycle arrest at the G2 phase. This allows time for DNA repair processes to occur before the continuation of cell cycling. The activity of CDK1 is competitively regulated against WEE1 by the protein phosphatase CDC25, which removes the inhibitory phosphate on Y15 and consequently reactivates the CDK1–cyclin B complex. In addition, CHK1 indirectly regulates the activity of CDK1 through transferring an inhibitory phosphate group onto CDC25. ,

1.

Key components of the cell cycle DDRCs in response to DNA damage and replication stress, as well as a comprehensive summary of essential regulatory proteins of the intra-S and G2/M DDRCs, highlighting the critical role of WEE1. Circles with P indicate a phosphate group. Arrows indicate activation and T-shaped arrows indicate deactivation.

At the onset of mitosis, WEE1 activity is rapidly decreased to enhance CDK1–cyclin B levels. This occurs via phosphorylation of WEE1 on the S53 and S123 residues by polo-like kinase 1 (PLK1). However, during the completion of cell division, the active concentration of WEE1 increases to downregulate CDK1 activity and aid in the dramatic remodeling of the cytoskeleton. In the S phase, WEE1 is considered a key regulator of DNA replication homeostasis due to managing CDK2 activity, a component of the crucial CDK2–cyclin A complex. In response to replication stress, fork stalling can occur, which triggers the CDK2–cyclin A-regulated intra-S DDRC to initiate repair. Increasing literature precedence also suggests that WEE1 plays an important regulatory role in epigenetic regulation by managing histone synthesis, as well as participating in tumor cell innate immunity via indirectly triggering the interferon pathway. ,−

The critical role of WEE1 in the intra-S and G2/M DDRCs has deemed it an attractive target for cancer treatment, in the context of the development of ATP-competitive WEE1 inhibitors. Many human cancers possess a deficient G1/S DDRC as a result of possessing a mutated TP53 gene and consequently an ineffective tumor suppressor protein p53, which is regarded as a master regulator of this DDRC (Figure ). Therefore, malignant cells generally over-rely on the intra-S and G2/M DDRCs for DNA-damage reparatory processes. Accordingly, small-molecule inhibition of WEE1 can disrupt numerous key signaling pathways of the proliferative cancer cell division cycle, by instigating DNA damage and replication stress and ultimately warranting mitotic catastrophe. ,, In clinical settings, WEE1 inhibitors have been reported to be efficacious in monotherapy, but deliver significantly improved therapeutic outcomes in combination with additional anticancer agents, such as platinum-based chemotherapeutics or other intra-S and G2/M DDRC kinase inhibitors. ,,−

Currently, no small-molecule WEE1 inhibitor has been approved by regulatory bodies globally; however, a few have been clinically evaluated, such as AZD1775 (adavosertib, 1). ,, Phase II trials have assessed 1 for treating numerous cancer types, such as colorectal, renal, cervical, and lung, among others, as a monotherapy drug or in dual therapy along with other anticancer drugs such as carboplatin and paclitaxel, or DDRC checkpoint inhibitors, as well as with radiotherapy. ,,

In 2021, Zentalis disclosed a next-generation pyrimidinylpyrazolone WEE1 inhibitor clinical candidate, ZN-c3 (azenosertib), which was derived from the structure of 1 and demonstrated enhanced inhibitory selectivity for WEE1 over PLK1, as well as an improved kinome-wide selectivity profile. In 2026, Zentalis commenced a monotherapy Phase III randomized confirmatory study with ZN-c3 for treating patients with cyclin E1-positive platinum-resistant ovarian cancer, which has received Fast Track Designation for this indication. ZN-c3 has also been evaluated in a Phase I/II trial along with encorafenib and cetuximab for treating patients with BRAF-V600E-mutated mCRC; however, this trial has been terminated due to a change in therapeutic landscape. Structurally, ZN-c3 differs from 1 by its chiral cyclic alcohol functional group, compared to the isopropyl alcohol (IPA) functionality of 1 (structure shown in Figure A). In the clinic, two treatment-related grade 5 adverse events (death due to presumed sepsis) were reported in the DENALI trial, which led to a partial FDA clinical hold in 2024 on three studies. Despite this, the overall therapeutic index (TI) remains favorable, and the majority of patients tolerated therapy well. Furthermore, myelosuppressive clinical toxicities have also been reported in patients, albeit to a lesser extent than that for 1.

2.

Structures of pyrimidinylpyrazolone (A) and non-pyrimidinylpyrazolone (B) WEE1 inhibitor/degrader clinical candidates.

Cyclization of the allyl moiety with the pyridyl ring of 1 resulted in the macrocyclic clinical candidate SC0191 (structure shown in Figure A), which is currently being evaluated in a Phase II trial for treating mCRC; however, its toxicity profile has not been reported. ,

Additional small-molecule WEE1 inhibitor clinical candidates include those from the dichlorophenyl chemotype, such as Debio 0123 (zedoresertib, Phase II) and IMP-7065 (potrasertib, Phase I). Few patients treated with Debio 0123 were reported to suffer from hematological diseases, and overall, the drug was described to have a manageable safety profile. − Limited safety data has been reported for IMP-7065 (non-pyrimidinylpyrazolone WEE1 inhibitor/degrader clinical candidate structures shown in Figure B). Scaffold hopping from the allyl pyrimidinylpyrazolone of 1 to a cyclopropyl pyrrolopyrimidine resulted in another clinical candidate currently in Phase I trials, named SY-4835. However, no data on its safety profile has been reported to date. In 2025, BMS disclosed the first WEE1 molecular glue degrader clinical candidate, BMS-986463, which is currently in a Phase I trial for treating advanced malignant tumors. Other WEE1 inhibitor clinical candidates currently being evaluated in Phase I trials for treating cancer include APR-1051, SGR-3515 (dual WEE1/PKMYT1), and MRANK-106 (dual WEE1/YES1); however, their structures have not been disclosed and little information is reported on their safety profiles, besides early clinical results for APR-1051, which illustrated a manageable safety profile. −

The pyrimidinylpyrazolone scaffold of 1 bears a pyridyl ring substituted with an IPA group and an N-methyl piperazine substituted onto a phenyl ring (Figure A). The binding mode of 1 observed via analyzing the X-ray cocrystal structure of 1 (PDB ID: 5V5Y, 1.90 Å) bound within the ATP-binding pocket of WEE1 revealed a hydrogen bonding interaction between the hinge region C379 carbonyl and the secondary amine functionality of the phenyl ring, as well as between the amine of C379 and a pyrimidine N of the core (Figure B). Key π–π interactions between the pyrimidine and both F433 and Y378 were also observed, as well as a hydrogen bond interaction between the carbonyl of the pyrimidinylpyrazolone core and the amide of N376. The allyl group was positioned in a hydrophobic pocket (among residues V313, A326, and K328) and the pyridyl ring extended toward the G-loop region, with the dimethyl functionality of the IPA group occupying a hydrophobic surface. The hydroxy functionality of the IPA group was not observed to partake in an interaction with a nearby residue, and the N-methylpiperazine was exposed to the solvent.

3.

(A) Chemical structure of AZD1775 (1), a pyrimidinylpyrazolone WEE1 inhibitor drug candidate that had completed Phase II clinical trials for treating various types of cancers. (B) Summary of WEE1 amino acid residue binding mode of 1 via analyzing its X-ray cocrystal structure (PDB ID: 5V5Y, 1.90 Å) bound to WEE1 using ICM-Pro.

Kinase selectivity profiling of 1 demonstrated ≥90% inhibition at 1 μM concentration toward 13 out of 477 screened kinases, with off-target equipotent activity against PLK1 also reported. , Due to its critical role in the G2/M DDRC (vide supra), PLK1 is also an enticing drug target for cancer treatment, with inhibitors being clinically evaluated; however, patients treated with PLK1 inhibitors have commonly been reported to experience grade 3/4 hematological diseases such as neutropenia and thrombocytopenia. − Therefore, the view was formed that the tolerability issues experienced by cancer patients treated with 1 may be due to being insufficiently selective for WEE1, especially against PLK1, and this has driven interest in the discovery of selective WEE1 inhibitors that retain potent anticancer efficacy compared to their nonselective counterparts.

Our previous studies have demonstrated that patient-derived peritoneal and liver mCRC organoids are sensitive to treatment with 1 and other acyclic WEE1 inhibitors. − Building on these findings and a broader interest in developing selective WEE1 inhibitors with retained potent anticancer efficacy, here, an investigation was undertaken which involved an extensive CADD campaign for developing a macrocycle based on 1. This was achieved via computational molecular modeling on an X-ray cocrystal structure of 1 bound to WEE1. Thereafter, the macrocycles were tested for biochemical WEE1 inhibitory activity and selectivity over PLK1. Subsequently, the biochemically most potent macrocycles were tested in the context of three-dimensional (3D) anticancer activity using patient-derived organoids (PDOs) and compared to 1. Lastly, X-ray cocrystal structures were generated of our most advanced macrocycles to compare their WEE1 binding mode with that predicted computationally during the CADD efforts. Gratifyingly, the experimental models and computational predictions coincided, which suggested that the bioactive conformation of 1 was successfully fixed due to macrocyclization.

Results and Discussion

Rational CADD of a Macrocyclic WEE1 Inhibitor

Macrocycle 2 was rationally designed via analysis and computational molecular modeling on an X-ray cocrystal structure of 1 bound to WEE1 (Figure ). Opportunely, the geometrical conformation of 1 adopted within the active site of WEE1 demonstrated a U-shaped domain. This type of geometry is favorable for the design of macrocyclic derivatives from their bioactive acyclic counterparts. − ,,

4.

(A) Computational molecular modeling (using ICM-Pro) on an X-ray cocrystal structure of AZD1775 (1) bound to WEE1 (PDB ID: 5V5Y, 1.90 Å), highlighting key residue interactions and a calculated binding score. (B) CADD of ethereal macrocycle 2 via ligedit modification of the cocrystal structure of 1 bound to WEE1 (PDB ID: 5V5Y, 1.90 Å), highlighting theoretical residue interactions and a calculated binding score. Green spheres represent hydrogen bond interactions, and black spheres represent π–π interactions. RMSD determined via molecular dynamic (MD) simulation experiments (400 ns) of 1 (A) and 2 (B) bound to WEE1 using the Desmond software of Schrödinger Maestro.

Therefore, the structure of 1 was evaluated using the ligedit feature of ICM-Pro (version 3.9-4). The workflow involved preparing the drug–protein complex (PDB ID: 5V5Y, 1.90 Å) by using the ‘convert to PDB’ feature to produce an ICM object, which removed unbound water molecules, optimized hydrogen atoms, restored missing heavy atoms in amino acid side chains, and assigned atomic charges. Afterward, the ligand and receptor were set up at physiological pH (7.4), and a subsequent binding score was calculated. Here, the ligedit feature was utilized to apply interactive de novo design to incorporate a 3-atom alkyl linker with an oxygen atom at the 2-position of the linker and at the meta-position of both the pyridyl and phenyl rings of 1 to afford macrocycle 2. The linker length was chosen due to the distance between the rings, and an oxygen atom at the 2-position was chosen here due to greater synthetic ease in accessing macrocycles with ethereal linkers. The newly modified ligand was minimized within the active site of the protein, and then a binding score was calculated. It was noted that macrocycle 2 was also redocked to assess differences between minimization and redocking, and it was observed that the highest scoring conformer exhibited a similar binding mode and returned a similar binding score to those produced after minimization alone; therefore, results from minimization only will be discussed throughout. The WEE1 binding mode of 2 was consistent with 1 besides an additional π–π interaction between the pyridyl ring and residue F310. For a detailed description of the WEE1 binding mode of 1, see Figure .

The binding score analyzed from ICM-Pro is the Radial and Topological Convolution Neural Network (RTCNN) score, which measures native-like ligands versus decoys, according to molecular geometries of anticipated ligand–protein complexes. This score is a sum of values calculated for each atom of a ligand and their corresponding interactions with their protein target. The lower the RTCNN score is, the more likely a ligand is anticipated to bind tightly within the protein of interest. Hereafter, the RTCNN score will be referred to as ‘binding score’. As the binding score of 1 was −46.74, the rationally designed macrocycle 2 was also proposed to bind tightly to WEE1 due to returning a similar binding score (−47.47), and its conformational restriction was hypothesized to boost selectivity for WEE1 over other protein kinases, like PLK1. − ,,

The aforementioned structural models of 1 (experimentally) and 2 (ligedit modification of 1) bound to WEE1 produced via ICM-Pro were further subjected to MD simulations (400 ns) for refinement and stabilization in the solution system using the Desmond simulation package of Schrödinger LLC (Figure ). , As shown in Figure A the RMSD (root-mean-square deviation) of the original acyclic ligand 1 was already at the steady state but experienced fluctuation throughout the experiment between 1.5 and 2.5 Å. In Figure B, it was observed that the rationally designed macrocycle 2 was initially at an unstable state by assessing its RMSD, likely due to not being the original ligand of the complex, and fluctuated to approximately 3.0 Å; however, after 50 ns in the simulation, the complex reached a steady state with little fluctuation between 1.0 and 2.2 Å throughout the 400 ns experiment. Gratifyingly, it was obvious that the complex with 1 as a ligand exhibited a higher RMSD than the complex with 2 as a ligand, demonstrating more stabilization and less fluctuation brought by the semirigid macrocyclic domain of 2 when binding to the kinase. Overall, it was observed that 2 bound within the ATP-binding pocket of WEE1 formed a more stable complex with lower energy compared to 1, which suggested that it could be a more potent inhibitor of WEE1. With the promising binding score and RMSD results of 2 across the 400 ns simulation, synthetic efforts were then directed to accessing a truncated and simpler analogue of 2 to validate the hypothesis that a macrocycle based on 1 would not only retain WEE1 inhibitory potency but also improve selectivity over PLK1.

Validation of the Macrocyclization Strategy

Because of the molecular complexity of fully elaborated macrocycle 2, numerous synthetic steps would be required to synthesize this advanced compound. Therefore, an initial proof-of-concept validation strategy was devised that involved a comparative model experiment with simpler analogues, these being model macrocycle 3 as a surrogate for 2 and acyclic control compound 4 that represents the counterpart surrogate for 1. For a detailed description of the macrocyclization validation strategy, including compound structures and design rationale, see Figure S1 and Table S1 in Supporting Information.

Depending on the size of the ring being generated, macrocyclic structures can be somewhat strained. The higher the degree of ring strain that these compounds possess, the increasingly difficult their synthesis can be. , Therefore, building on the CADD campaign, a computational model was identified called StrainViz, which was utilized prior to commencing the synthesis of model macrocycle 3. This program functions to measure the theoretical localized strain energies of macrocycles and can infer the likelihood of successful synthesis. The localized strain analysis of 3 was conducted alongside the highly strained macrocyclic anticancer drug pacritinib for direct comparison, and the results are demonstrated in Figure . ,

5.

Determination of localized strain energies (kcal/mol) of the macrocyclic anticancer drug pacritinib and model macrocycle 3. All schematics are accompanied by a color bar. The bars display a color range that establishes the maximum and minimum localized strain energies experienced by the macrocycles. Red regions indicate greater localized areas of strain, while conversely, green areas indicate comparatively lower localized areas of strain. Yellow regions indicate moderate localized areas of strain relative to values displayed within the range. All strain energy values are displayed in kcal/mol.

Pacritinib was determined to experience extremely high localized strain (24.9 kcal/mol), particularly across the secondary amine functional group that bridges the southern pyrimidine and phenyl rings. Gratifyingly, model macrocycle 3 exhibited significantly lower localized strain (7.23 kcal/mol) compared to pacritinib, and the strain energies were observed to be more delocalized evenly throughout the 14-atom macrocyclic ring, particularly through the bonds that connect the pyrimidinylpyrazolone scaffold to the pyridine and phenyl rings, as well as across the ethereal linker. Overall, the results from StrainViz demonstrated that model macrocycle 3 was theoretically less strained than pacritinib; therefore, it was presumed that ring strain may not pose a significant challenge in accessing 3 synthetically.

Compounds 3 and 4 were successfully synthesized and then tested against WEE1 and PLK1 biochemically for IC₅₀ values, as well as WEE1/PLK1 IC₅₀ selectivity index (SI) values (see Scheme S1 and Table S2 in Supporting Information for synthetic procedures and assay parameters, respectively). In brief, the structural preorganization of 3 successfully prelocked a favored bioactive conformation for WEE1 over PLK1; as a result, a significant increase in the SI was determined for 3 (SI = 24), compared with the acyclic control compound (SI = 4). In addition, model macrocycle 3 retained WEE1 inhibitory potency in comparison to 4.

Taken together, the CADD campaign facilitated the rational design of macrocyclic WEE1 inhibitor 2, which was based on AZD1775 (1). This paved the way for a model macrocycle study to validate that macrocyclization can fix a preferred WEE1 bioactive conformation and enhance selectivity over that of PLK1. As the macrocyclization strategy was successful, efforts were then directed to identifying the best 3-atom linker of the macrocyclic scaffold, followed by the synthesis of fully elaborated macrocycles with the best performing linker.

Exploring Linker Variations within the Macrocycle

It was then of interest to explore heteroatom variations at the 2-position of the 3-atom linker of model macrocycle 3 and reincorporating the N-allyl moiety of 1. It has been observed in the literature that subtle changes in a macrocyclic linker, such as exchanging heteroatoms, can significantly impact potency and selectivity, especially against proteins with similar structural motifs. − Therefore, simple allyl macrocycles were designed with ether (2-O) 5, amine (2-NH) 6, methylene (2-C) 7, and thioether (2-S) 8 functionalities at the 2-position of the linker, and prior to initiating synthetic efforts, these compounds were subjected to a strain analysis by StrainViz to infer the likelihood of successful synthesis (see Figure S2 in Supporting Information for the strain analysis of macrocycles 6–8). After successful synthesis, these compounds were tested in a biochemical assay for IC₅₀ and SI values against the purified kinases, WEE1 and PLK1 (Table , and see Schemes S1 and S2 in Supporting Information for synthetic procedures).

1. Biochemical kinase Inhibitory Analysis of the Simple Allyl Macrocycles with Linker Heteroatom Variations, 2-O 5, 2-NH 6, 2-C 7, and 2-S 8 against WEE1 and PLK1 ^{, , ,}

^aIC₅₀ values were determined at Reaction Biology (Germany) via testing of the compound in a ³³PanQinase assay.

^bIC₅₀ value that is the mean of at least two experiments (n = 2) with ±SD (standard deviation).

^cThe SI refers to the PLK1/WEE1 IC₅₀ ratio.

^dSI calculations for comparison with AZD1775 (1) and propyl model macrocycle 3.

Surprisingly, 2-O-5 (IC₅₀ = 61 nM) not only retained potent WEE1 activity compared to clinical candidate 1 (IC₅₀ = 41 nM) but also exhibited profoundly enhanced inhibitory potency (148-fold) compared to its propyl counterpart 3 (IC₅₀ = 9000 nM). This reinforces that the allyl moiety is essential for potency within the macrocyclic system, which is likely positioned well within the hydrophobic pocket (among residues V313, A326, and K328, akin to 1) and stabilized by the semirigid macrocyclic scaffold when bound to WEE1. In addition, this highlights that subtle changes in functional groups (N-propyl to N-allyl) that extend from the macrocyclic core may drastically affect binding to biological targets due to the semirigid nature of the macrocycle, especially as an acyclic N-propyl analogue, 17 of 1 was reported to exhibit equipotent activity against WEE1 (see Figure S2 in Supporting Information for structure and biological data).

The structure of 5 was also confirmed by single-crystal X-ray diffraction (XRD), and its conformation was observed to closely match (RMSD = 0.81 Å) the theoretical binding mode predicted from the ligedit feature of ICM-Pro, which was based on the X-ray cocrystal structure of 1 bound to WEE1 (Figure , PDB ID: 5V5Y, 1.90 Å, see the Single-Crystal XRD section in Supporting Information for a detailed report). This similarity suggests that the simple macrocycle is preorganized in a low-energy conformation compatible with target binding, minimizing the entropic cost of reorganization upon complex formation. Consequently, 5 bypasses the requirement for elaborated functional groups such as the IPA and N-methyl piperazine in 1 and its acyclic analogues that are required in those cases for high WEE1 potency. Notably, key dihedral angles and the orientation of pharmacophoric groups are preserved between the solid-state and predicted bound structures, supporting the relevance of the computational model and the validity of the observed conformation for further CADD efforts. Furthermore, 5 retained a high binding score (−44.39) comparative with fully elaborated derivative 2 (−47.46), which was expected as 5 returned potent WEE1 inhibitory activity, and overall highlights that the macrocyclic scaffold itself, coupled with the N-allyl group, are important for tight binding within the active site of WEE1.

6.

(A) Single-crystal XRD structure diagram of simple allyl 2-O macrocycle 5. Ellipsoids are shown at the 50% probability level. (B) Theoretical WEE1 binding mode of 5 (gray) generated by ligedit (via ICM-Pro) modification of the X-ray cocrystal structure of 1 bound to WEE1 (PDB ID: 5V5Y, 1.90 Å). A binding score was calculated and the theoretical structure was superimposed with a conformer (pink) obtained via XRD with RMSD calculated. Green spheres represent hydrogen bonds, and purple spheres represent π–π interactions.

In addition, 2-NH 6 and 2-C 7 returned relatively potent activity against WEE1, with IC₅₀ values of 143 nM and 131 nM, respectively, while thioether 2-S 8 exhibited a significant loss in activity (IC₅₀ = 7300 nM).

With regard to PLK1 inhibition, 2-O 5 (IC₅₀ = 2100 nM) was significantly less active than 1 (IC₅₀ = 112 nM), returning SI values of 34 and 3, respectively. Moreover, in comparison to its N-propyl counterpart 3 (IC₅₀ = 213000 and SI = 24), the subtle modification to an N-allyl group in 5 not only profoundly enhanced WEE1 potency but also significantly increased selectivity over PLK1. A similar trend was observed with 2-NH 6 and 2-C 7, which returned IC₅₀ values of 2500 nM and 1200 nM, as well as SI values of 18 and 9, respectively. 2-S 8 was also significantly less active against PLK1 (IC₅₀ = 94000 nM), with an SI of 13, but was noted to be a weak WEE1 inhibitor.

Taken together, in this macrocyclic linker study, it was observed that subtle heteroatom modifications at the 2-position within the 3-atom alkyl linker significantly affected both WEE1 potency and selectivity over PLK1, akin to Lücking et al. Factors that contributed to the decision in evaluating the optimal linker for the development of novel elaborated macrocycles were good potency (WEE1 IC₅₀ < 150 nM) and a PLK1/WEE1 IC₅₀ SI value of approximately 10 or greater. Macrocycles 2-O 5, 2-NH 6, and 2-C 7 all satisfied the former parameters, but 5 emerged as the clear frontrunner in terms of both WEE1 potency and selectivity over PLK1 and was chosen as the candidate for further development. Indeed, 5 exhibited comparable potency to clinical candidate 1, although being a simpler compound, and was significantly more selective for WEE1 over PLK1 (>10-fold increase).

Synthetic Procedures of Elaborated Macrocycles

Having determined that the optimal macrocyclic linker contained an oxygen atom at the 2-position (2-O) of the linker, focus was then directed to elaborating 5 with at least one of the key functional groups also borne by 1, that being the N-methylpiperazine moiety of the phenyl ring for macrocycle 9 and the IPA group of the pyridyl ring for macrocycle 10 (see Figure S3 in Supporting Information for theoretical WEE1 binding mode and score of 9 and 10 determined via the ligedit feature of ICM-Pro).

Des-piperazine macrocycle 9 was synthesized according to Scheme and is described further in the Compound Characterization section in Supporting Information. Synthesis began with methylation of commercially available carboxylic acid 20 to 9–1, followed by reduction to primary alcohol 9–2. Next, silyl protection of 9–2 gave 9–3, which was then subjected to a lithiation reaction and quenching with acetone to afford 9–4. Thereafter, TBS-deprotection of 9–4 gave primary alcohol 9–5, which was then subjected to a Williamson ether synthesis with 19 to afford 9–6. A subsequent Ullmann-type reaction between 9–6 and 18 yielded 9–7, which was then reduced to 9–8. Finally, 9–8 was treated with Oxone, and an intramolecular S_NAr reaction afforded 9 in a total of 15 steps (17%). See Figure S5A in Supporting Information for a comparative RMSD analysis of the structure of 9 in its lowest energy conformation determined via single-crystal XRD, which was observed to closely match (RMSD = 0.91 Å) its theoretical binding mode predicted from the ligedit feature of ICM-Pro. This highlights that the structure’s lowest energy conformation is analogous to its bioactive conformation with WEE1.

1. Synthesis of Des-piperazine Macrocycle 9 .

^a Reagents and conditions: (a) SOCl₂, MeOH, 70 °C, 3 h; (b) LiAlH₄, Et₂O, 35 °C, 5 h; (c) TBSTf, Et₃N, DCM, 0 °C–RT, 18 h; (d) (1) n-BuLi, Et₂O, −78 °C, 1 h, (2) acetone or cyclobutanone, 0 °C for 30 min then 0 °C–RT for 1 h; (e) TBAF, THF, 0 °C–RT, 18 h; (f) Cs₂CO₃, MeCN, RT, 18 h; (g) N,N′-dimethylethylenediamine, CuI, K₂CO₃, 1,4-dioxane, 95 °C, 18 h; (h) Zn, NH₄Cl, acetone:H₂O, 0 °C for 1 min, RT for 18 h; and (i) Oxone, THF:H₂O, RT, 2 d (17%).

Des-IPA macrocycle 10 was synthesized according to Scheme and is described further in the Experimental Section in Supporting Information. Synthesis began with chlorination of previously synthesized intermediate 3–2 to 10–1. Commercially available aryl fluoride 21 was then subjected to an S_NAr reaction with 1-methylpiperazine to afford 10–2, which thereafter underwent a Williamson ether synthesis with 10–1 to produce 10–3. A subsequent Ullmann-type reaction between 18 and 10–3 gave 10–4, which was then reduced to 10–5. Finally, 10–5 was treated with Oxone, and a subsequent intramolecular S_NAr reaction afforded 10 in a total of 14 steps (13%).

2. Synthesis of Des-IPA Macrocycle 10 .

^a Reagents and conditions: (a) SOCl₂, DCM, RT, 3 h; (b) 1-methylpiperazine, 130 °C, 1 h; (c) NaOH, TBAHS, PhMe:H₂O, reflux, 24 h (d) N,N′-dimethylethylenediamine, CuI, K₂CO₃, 1,4-dioxane, 95 °C, 18 h; (e) Zn, NH₄Cl, acetone:H₂O, 0 °C for 1 min, RT for 18 h; (f) Oxone, THF:H₂O, RT, 2 d (13%)

Fully elaborated macrocycle 2, as well as related analogues 11–15, was synthesized according to Scheme and is described further in the Compound Characterization section in Supporting Information. See Figure S4 in Supporting Information for theoretical WEE1 binding modes of 11–15 determined via the ligedit feature of ICM-Pro, as well as a description on the design rationale of these advanced macrocycles.

3. Synthesis of Fully Elaborated Macrocycle 2 and Derivatives 11–15 .

^a Reagents and conditions: (a) (1) n-BuLi, Et₂O, −78 °C, 1 h, (2) acetone or cyclobutanone, 0 °C for 30 min then 0 °C–RT for 1 h; (b) N,N′-dimethylethylenediamine, CuI, K₂CO₃, 1,4-dioxane, 95 °C, 18 h; (c) TBAF, THF, 0 °C–RT, 18 h; (d) Cs₂CO₃, MeCN, RT, 2 d; (e) 1-methylpiperazine, methanol, dimethylamine, 1-methylhomopiperazine or 4-(pyrrolinin-1-yl)­piperidine, DMF or NMP, 60–100 °C, 1–3 d; (f) NH₄Cl, acetone:H₂O, 0 °C for 1 min, RT for 18 h; (g) H₂O₂, Mo catalyst, H₂O, RT, 1–2 d. Yield of final reactions (22–65%)

First, previously synthesized intermediate 9–3 was subjected to a lithiation reaction that was quenched with either anhydrous acetone or cyclobutanone to produce 9–4 and 15–1, respectively. A subsequent Ullmann-type reactions between either 9–4 or 15–1 and building block 18 afforded 2–1 and 15–2, respectively. Next, TBS-deprotection of 2–1 and 15–2 yielded 2–2 and 15–3, respectively. Williamson ether reactions between 2–2 or 15–3 and previously synthesized intermediate 22 gave key intermediates 2–3 and 15–4, respectively. Next, 2–3 and 15–4 were subjected to an S_NAr reaction with 1-methylpiperazine to afford 2–4 and 15–5, respectively, that were then reduced to produce respective penultimate intermediates 2–5 and 15–6. Finally, 2–5 and 15–6 were treated with hydrogen peroxide and a Molybdenum-micellar catalyst (synthesized via a literature protocol), and a subsequent intramolecular S_NAr reaction afforded 2 and 15 in a total of 17 steps. It is noted here that different reaction conditions were employed to mediate S-oxidation and a subsequent intramolecular S_NAr reaction to afford the advanced macrocycles. A milder S-oxidation procedure was found in recent literature that forms micelles in the solvent (H₂O). Here, the reagents can catalytically react with hydrogen peroxide in the Mo-based micelles and selectively form sulfoxides, instead of sulfones depending on the equivalents of hydrogen peroxide used. Utilizing this method resulted in improved yields and fewer byproducts, decreasing difficulties experienced with macrocycle purification, especially when Oxone was used as the reagent.

Moreover, 13-step key intermediate 2–3 was chosen to undergo late-stage diversification on the phenyl ring handle due to possessing a highly reactive para-fluoro nitrobenzene moiety. Therefore, 2–3 was subjected to an S_NAr reaction with the following nucleophiles: methanol, dimethylamine, 1-methylhomopiperazine, and 4-(pyrrolidin-1-yl)­piperidine, to produce intermediates 11–1–14–1, respectively. Subsequent reduction of 11–1–14–1 yielded the respective penultimate aniline intermediates 11–2–14–2, that were ultimately treated with hydrogen peroxide and the Mo-micellar catalyst, promoting an intramolecular S_NAr reaction to afford 11–14, respectively. Final compounds were obtained in 24–82% yields.

Furthermore, see Figure S5B in Supporting Information for a comparative RMSD analysis of the structure of 11 in its lowest energy conformation determined via single-crystal XRD, which was observed to closely match (RMSD = 0.85 Å) its theoretical binding mode predicted from the ligedit feature of ICM-Pro. This highlights that the structure’s lowest energy conformation is analogous to its bioactive conformation with WEE1, akin to macrocycles 5 and 9. For single-crystal XRD reports of macrocycles 5, 9, and 11, see the Single-Crystal XRD section in Supporting Information.

Elaborated Macrocycles Trend toward Similar Biochemical Potency to the Unelaborated Macrocycle Scaffold

With the elaborated macrocycles based on 5 successfully synthesized, it was of interest to test them against WEE1 and PLK1 in a biochemical assay to measure inhibitory potency and determine SI values (Table ).

2. Biochemical kinase inhibitory analysis of elaborated macrocycles 2 and 9–15 against WEE1 and PLK1, as Well as SI Values for Comparison against AZD1775 (1) and Frontrunner Simple Allyl Macrocycles 5–7 ^{, , , , , , ,}

^aElaborated macrocycles were synthesized based on simple macrocycle 5, which was derived from 1.

^bSI refers to the selectivity index; the PLK1/WEE1 activity ratio determined for comparison with 1.

^cPM003 is a PDO line established from the peritoneal metastatic CRC tissue.

^dPM025 is a PDO line established from the peritoneal metastatic CRC tissue.

^eQEH039LM is a PDO line established from the liver metastatic CRC tissue.

^fQEH042LM is a PDO line established from liver metastatic CRC tissue determined to be relatively resistant to treatment with WEE1 inhibitors compared with other PDO lines in this investigation.

^gPDOs were grown for 2 days and then treated with 7 doses of the WEE1 inhibitor in quadruplicate for 5 days, cell viability was assessed via Cell Titer-Glo assays, and IC₅₀ values were calculated via nonlinear regression. Cell viability experiments were conducted with 4 technical replicates, one biological experiment per PDO, except for ⁺mCRC PDO viability assay IC₅₀ values that represent the mean of two experiments each conducted in quadruplicate with ±SD. *Biochemical kinase inhibition assays conducted at Reaction Biology. IC₅₀ values that represent the mean of at least two experiments with ±SD (standard deviation) in one run, except for 16 which does not have biochemical data. Macrocycles 2 and 15 were tested against WEE1 and PLK1 (n = 2) during two independent runs, with different batches, and mean IC₅₀ values were reproducible (see Figure S16 in Supporting Information for WEE1 dose–response curves).

^hAnticancer efficacy of the best performing macrocyclic WEE1 inhibitors measured against peritoneal and liver mCRC PDOs for comparison against clinical candidates 1 and ZN-c3 (16).

First, des-piperazine macrocycle 9 (IC₅₀ = 40 nM) retained strong inhibitory potency against WEE1, akin to 1 (IC₅₀ = 41 nM). Next, des-IPA macrocycle 10 (IC₅₀ = 57 nM) also returned a strong potency against WEE1. Likewise, and gratifyingly, fully elaborated macrocycle 2 (IC₅₀ = 42 nM), which was the product of the CADD campaign (vide supra), also exhibited strong potency against WEE1. Thus, it was observed that the addition of either, or both, of the functional groups that are known to be important for WEE1 potency in the acyclic system (such as in 1) was not essential in the case of macrocycles, as these compounds did not recapitulate the expected potency boost in the biochemical assay. In this investigation, this highlights the dominating role of macrocyclization in conferring biochemical inhibitory potency, which has usurped that of the need for side chain extensions.

With respect to PLK1, des-piperazine macrocycle 9 returned an IC₅₀ value of 680 nM and an SI value of 17. This was almost half of what was calculated for simple allyl macrocycle 5 (SI = 34). Next, des-IPA macrocycle 10 returned a large decrease in potency against PLK1 (IC₅₀ = 7100 nM), and a significantly greater SI value of 125 was determined. Fully elaborated macrocycle 2 (IC₅₀ = 1600 nM and SI = 38) returned an almost 2-fold greater SI value than 9. Interestingly, it was observed that possessing an IPA group alone (such as in 9) extended from the pyridyl ring of the macrocyclic scaffold increased potency against PLK1. Conversely, possessing an N-methylpiperazine group alone (such as in 10) extended from the phenyl ring of the macrocyclic scaffold significantly increased selectivity for WEE1 over PLK1. Bearing both of said functional groups, in the case of 2, enhanced selectivity for WEE1 over PLK1 compared with unelaborated derivative 5, returning an SI value between those calculated for 9 and 10. Taken together, with the encouraging potency and selectivity results in regard to fully elaborated macrocycle 2, it was then of interest to conduct a brief SAR explorative analysis on the pyridyl and phenyl ring substituents in order to determine their effects on the WEE1 potency and selectivity over PLK1. It is noted here that even though des-IPA macrocycle 10 returned exceptional inhibitory selectivity for WEE1 against PLK1 (SI = 125), this compound returned weak anticancer efficacy (vide infra) and thus was not chosen as a candidate for further development.

The solvent-exposed N-methylpiperazine group substituted onto the phenyl ring of 2 was thereafter subjected to functional group diversification. Anisole 11 and dimethylamine 12 retained relatively potent WEE1 activity, returning IC₅₀ values of 61 nM and 50 nM, respectively. A 2-fold decrease in WEE1 activity was observed with homopiperazine 13 (IC₅₀ = 116 nM); however, relatively potent activity was exhibited by 4-(pyrrolidin-1-yl)­piperidine derivative 14 (IC₅₀ = 73 nM). Briefly focusing on the pyridyl ring of 2, rigidification of the branched alcohol moiety to a cyclobutanol in 15 (IC₅₀ = 57 nM) resulted in retained potent WEE1 inhibitory activity.

Regarding PLK1 inhibition and compared with macrocycle 2 (IC₅₀ = 1600 nM and SI = 38), anisole 11 (IC₅₀ = 1600 nM) exhibited relatively decreased WEE1 selectivity (SI = 23). However, dimethylamine 12 (IC₅₀ = 2500 nM) demonstrated an increase in WEE1 selectivity (SI = 50). WEE1 selectivity was observed to decrease with macrocycles that bear larger solvent-exposed groups: homopiperazine 13 (IC₅₀ = 1600 nM) and 4-(pyrrolidin-1-yl)­piperidine 14 (IC₅₀ = 760 nM), as they returned SI values of 12 and 10, respectively. Lastly, cyclobutanol 15 (IC₅₀ = 1300 nM), although relatively lower than that of 2, demonstrated good selectivity for WEE1 over PLK1, returning an SI value of 25. Overall, from this SAR study, it was observed that modifying the pyridyl and phenyl ring functional groups resulted in slight variations in WEE1 potency but more importantly significantly affected selectivity for WEE1 over PLK1.

Taken together, it was observed that functionalization of the biochemically active macrocyclic scaffold of 5 with the IPA group (on the pyridyl ring) and/or N-methylpiperazine group (on the phenyl ring), which are also borne by 1, significantly affected WEE1 inhibitory selectivity over PLK1, but only slightly affected WEE1 potency. This was surprising as it was expected that functional group elaboration onto the bioactive macrocyclic core, those of which are known to be important for acyclic AZD1775-like analogues, should in fact also be crucial in improving the WEE1 potency. However, in this investigation, it was clear that WEE1 potency was dominated by the macrocyclic core of 5, coupled with the presence of the N-allyl group, likely due to the success in prelocking a favored WEE1 bioactive conformation by means of extensive CADD efforts. These encouraging results emphasize the advantage of pursuing macrocyclization as a strategy for not only enhancing biological target selectivity over structurally similar counterparts but also improving or retaining potent binding affinity by limiting entropic loss of ligand–protein binding. As functionalization of the macrocycles did not confer the same biochemical potency increase that is typically observed in an acyclic system, rather unelaborated macrocycles such as 5 were found to exhibit equipotent activity against WEE1 compared with 1. It was therefore of interest to investigate how such structural modifications influenced the anticancer organoid-based potency.

Anticancer Efficacy Using Patient-Derived Colorectal Cancer Organoids

Colorectal cancer (CRC) is the second leading cause of cancer-related mortality worldwide. , CRC patients with peritoneal metastasis have the lowest survival rate of all mCRC types, affecting approximately 8–15% of patients with CRC. , In addition, almost 50% of CRC patients develop liver metastasis during disease progression. Unresectable mCRC to either site has limited treatment options; thus, the outlook for most advanced CRC patients has changed little in the last 15 years. , To improve outcomes, new and personalized treatment options are imperative.

PDOs are 3D communities of self-organizing cells that offer promise in evaluating clinical patient tumor responses to drug treatments since they retain the (epi)­genetic and molecular characteristics of the original cancer and can be grown indefinitely in vitro from resected malignant tissue (Figure ). −

7.

Primary tumor CRC and normal colon PDOs (RAH035) are representative of the human tissues from which they are derived. (A) Schematic of the pipeline from human CRC patients to drug testing. (B) Organoids are morphologically similar to the native tumor (H&E) and express intestinal epithelial markers, CDX2 and CK20, but not CK7. Scale bar: 20 μm (H&E and IHC). Far right are bright-field images of organoids. Scale bar: 100 μm (brightfield).

There is strong evidence that drug testing in PDOs can retrospectively mimic patient responses in the clinic. − PDO culture has remarkably high success rates for CRC, and we have recently established a pipeline for patient and tumor specific preclinical drug testing. This involved recruiting CRC patients, collecting normal healthy and tumor tissue for genetic analysis, and banking as patient PDOs, within a clinically actionable time frame.

It has been reported that CRC with mutated TP53 was sensitive to treatment with 1. In addition, Phase II clinical trials were completed using 1 in monotherapy for treating advanced CRC in the United Kingdom (FOCUS4-C) with patients that had mutated TP53 and RAS. In combination with irinotecan, 1 was also evaluated clinically in a Phase I trial as a second-line treatment for mCRC patients with mutated KRAS and BRAF.

Furthermore, we were the first to report the use of a PDO platform to evaluate the effects of AZD1775-like acyclic WEE1 inhibitor treatment on both normal healthy and tumor cell viability. Concurrently, this work was recognized as aligning well with the FDA’s recent initiative to phase out animal testing in favor of more human-relevant methods such as drug testing against PDOs for evaluating drug efficacy, safety, and toxicity.

In this investigation, the initial focus was on two TP53-mutated (as well as KRAS-WT and BRAF-WT) peritoneal mCRC PDO lines (PM003 and PM025) characterized in our previous study. PM003 and PM025 were determined to be sensitive to treatment with 1, and here, respective IC₅₀ values of 130 nM and 270 nM were obtained.

In addition, this study focused on two TP53-mutated liver mCRC PDO lines, namely, QEH039LM (previously characterized, with KRAS-WT and BRAF-WT), which was found to be sensitive to treatment with 1, returning an IC₅₀ value of 240 nM in this study, and QEH042LM, a novel mCRC PDO line (64 y/o female with mCRC adenocarcinoma, TP53-mutated, BRAF-V600E and KRAS-mutated) which was found to be more resistant to treatment with 1 (IC₅₀ = 937 nM). For histology of both liver mCRC PDO lines, see Figure S6 in Supporting Information. In addition, for more information about mutation sequencing of all (m)­CRC PDO lines of the PDO platform used in this investigation, see Figures S9 and S10 in Supporting Information.

To determine the anticancer efficacy of the potent macrocyclic WEE1 inhibitors, these compounds were evaluated against PM003 and PM025 in vitro, and cell viability was measured to calculate IC₅₀ values (Table ). Only the best-performing macrocycles that were determined to be most efficacious against PM003 and PM025 (peritoneal metastases) were also tested against QEH039LM and QEH042 liver metastases (LM).

First, simple potent macrocyclic WEE1 inhibitor 2-O 5 (WEE1 IC₅₀ = 61 nM and SI = 34) was tested against the peritoneal mCRC PDO lines. Weak efficacy against PM003 and PM025 was observed for 5, with IC₅₀ values of 9600 nM and 2100 nM, respectively. These results were surprising as 5 demonstrated potent activity against WEE1 in the biochemical assay. It was thus of interest to determine whether simple macrocycles 2-NH 6 (WEE1 IC₅₀ = 143 nM and SI = 18) and 2-C 7 (WEE1 IC₅₀ = 131 nM and SI = 9) were also inactive against the peritoneal metastases. Indeed, this was observed, as 6 returned an IC₅₀ value of 960 nM against PM003 and 1500 nM against PM025. Similarly, 7 returned an IC₅₀ value of 1100 nM against those of both PM003 and PM025.

Next, the elaborated derivatives of 5 were tested against peritoneal metastases. Comparatively, des-piperazine macrocycle 9 (WEE1 IC₅₀ = 40 nM and SI = 17) exhibited enhanced activity against PM003, returning an IC₅₀ value of 940 nM, and poor activity against PM025, returning an IC₅₀ value of 2100 nM. However, 9 was still significantly less active than 1 (IC₅₀ = 120 nM against PM003 and 290 nM against PM025). A decrease in anticancer efficacy was also observed with 10 (WEE1 IC₅₀ = 57 nM and SI = 125), returning an IC₅₀ value of 4000 nM against both PM003 and PM025. However, and gratifyingly, fully elaborated macrocycle 2 (WEE1 IC₅₀ = 42 nM and SI = 38), the product of the CADD campaign, demonstrated strong efficacy against PM003 and PM025, returning IC₅₀ values of 164 nM and 239 nM, respectively, and comparable to those of 1.

The macrocycles with phenyl ring modifications (exchanging the N-methylpiperazine group) were then tested against the peritoneal mCRC PDO lines. First, methoxy macrocycle 11 (WEE1 IC₅₀ = 61 nM and SI = 23) demonstrated modest activity against PM003 and PM025, returning IC₅₀ values of 850 nM and 1600 nM, respectively. An increase of efficacy against PM003 was observed by treatment with dimethylamine 12 (WEE1 IC₅₀ = 50 nM and SI = 50), returning an IC₅₀ value of 381 nM, and modest efficacy was observed against PM025 (IC₅₀ = 1600 nM). Homopiperazine 13 (WEE1 IC₅₀ = 116 nM and SI = 12) suffered a loss in activity against PM003 and PM025, returning IC₅₀ values of 2100 nM and 2300 nM, respectively. 4-(Pyrrolidin-1-yl)­piperidine analogue 14 (WEE1 IC₅₀ = 73 nM and SI = 10) exhibited modest efficacy against PM003 (IC₅₀ = 711 nM) and demonstrated enhanced efficacy against PM025 (IC₅₀ = 344 nM). Gratifyingly, with respect to the pyridyl ring modification (rigidifying the IPA group as a cyclobutanol), 15 (WEE1 IC₅₀ = 57 nM and SI = 25) exhibited strong potency against PM003 and PM025, returning IC₅₀ values of 143 nM and 273 nM, respectively. From this study of testing the biochemically active macrocyclic WEE1 inhibitors against the peritoneal metastases, the leading macrocycles that exhibited strong efficacy, which were also highly selective for WEE1 over PLK1, were fully elaborated macrocycle 2, which was the product of rational CADD efforts (vide supra), and its cyclobutanol derivative 15.

Thereafter, the leading macrocycles were tested against the liver mCRC PDO lines and compared with 1 (QEH039LM IC₅₀ = 240 nM and QEH042LM IC₅₀ = 940 nM). Macrocycle 2 returned an IC₅₀ value of 169 nM against QEH039LM and 740 nM against QEH042LM. In addition, 15 returned an IC₅₀ value of 147 nM against QEH039LM and 760 nM against QEH042LM, and taken together, both 2 and 15 demonstrated stronger efficacy than 1 against the LM (Figure ).

8.

Liver mCRC PDOs are sensitive to WEE1 inhibitors. (A) Representative images of liver mCRC organoids (QEH039LM and QEH042LM) treated with WEE1 inhibitors (2, 15, and 1). Scale bar: 500 μm. (B) Representative dose–response curves of QEH039LM and QEH042LM organoids treated with WEE1 inhibitors (2, 15,and 1). Organoids were treated in quadruplicate, and cell viability was assessed with Cell-Titer Glo assays in at least two independent experiments. Nonlinear regression curve fitting was used to calculate IC₅₀ values.

Furthermore, as the best-in-class clinical candidate, ZN-c3 (16) is the current frontrunner WEE1 inhibitor being evaluated clinically (Phase III clinical trial commenced); it was also of interest to determine the efficacy of this drug against the peritoneal mCRC PDO lines for comparison against the leading macrocycles 2 and 15 as well as clinical candidate 1. As expected, 2 retained strong efficacy against PM003 and PM025, returning IC₅₀ values of 127 nM and 112 nM, respectively. These anticancer activity results were comparable to AZD1775 (1), as well as macrocycles 2 and 15 (for dose–response curves and organoid images during treatment with 2 and 15, as well as 16, see Figure S7 in Supporting Information).

Through testing the biochemically potent macrocyclic WEE1 inhibitors in our unique CRC PDO platform, it was observed that some analogues were ineffective at reducing organoid viability. Numerous factors may have contributed the potent inhibition of the purified WEE1 enzyme in a biochemical assay but significantly decreased activity against CRC PDOs of some macrocycles, for example, challenges in maintaining activity due to nonspecific binding in cell media, transportation issues, modification by other enzymes in the cells, poor physicochemical properties resulting in inadequate cell permeability, and/or accessibility of the biochemical target, with the latter likely being the most relevant in this context, due to the 3D architecture and complexity of organoids. ,,, In addition, it was recognized that the flatness of drug candidates can negatively impact aqueous solubility and cell permeability due to molecular stacking, and introducing sp³-hybridized carbons onto aromatic rings is an effective strategy to disrupt this planarity. This could explain why simple planar macrocycles 5–7 demonstrated poor anticancer efficacy yet being potent WEE1 inhibitors, compared to those that have sp³ character, such as leading macrocycles 2 and 15, that possess an IPA and cyclobutanol group, respectively, substituted onto the pyridine ring, which are sp³-rich motifs.

Leading macrocycles 2 and 15 were subsequently tested against the patient-matched TP53-WT primary CRC tumor organoid line (RAH051T) and normal healthy colon organoid line (RAH051N) to measure a potential therapeutic window for treatment. Also, we determined whether these macrocyclic WEE1 inhibitors are efficacious against CRC with an intact G1/S DDRC (Figure ).

9.

CRC organoids are significantly more sensitive to WEE1 inhibitors than patient-matched normal healthy colon organoids. (A) Representative images of normal human colon (RAH051N) and matched tumor (RAH051T) organoids treated with WEE1 inhibitors (2 and 15). Scale bar: 500 μm. (B) Representative dose–response curves of normal human colon (RAH051N) and matched tumor (RAH051T) organoids treated with WEE1 inhibitors (2 and 15). Organoids were treated in quadruplicate, and cell viability was assessed with Cell-Titer Glo assays in at least two independent experiments. Nonlinear regression curve fitting was used to calculate IC₅₀ values.

Images were taken of these patient-matched PDO lines when treated with 2 and 15 at inhibition concentrations of 0.1 nM to 100 μM (DMSO as the vehicle), demonstrating concentration-dependent activity against both normal and tumor organoids. However, and gratifyingly, the dose–response curves of normal human colon (blue) and patient-matched tumor (red) RAH051 organoids indicated a therapeutic window where normal healthy organoids (RAH051N) were significantly less sensitive to treatment than tumor organoids (RAH051T). Nonlinear regression curve fitting was used to calculate IC₅₀ values; macrocycle 2 returned an IC₅₀ value of 1300 nM against RAH051N and 171 nM against RAH051T, returning a TI of 8. Macrocycle 15 returned an IC₅₀ value of 5700 nM against RAH051N and 118 nM against RAH051T, with an exceptional TI value of 48 determined. These results demonstrate that the leading macrocyclic WEE1 inhibitors were also strongly active against TP53-WT primary CRC with an intact G1/S DDRC. In addition, leading macrocycles 2 and 15 were tested against the novel patient-matched TP53-mutated primary CRC tumor organoid line (RAH035T) and normal healthy colon organoid line (RAH035N) to measure another potential therapeutic window for cancer treatment (histology of RAH035 represented in Figure ). In brief, 2 and 15 were also significantly more selective at potently treating CRC in RAH035. For a detailed description of this study, see Figure S8 in Supporting Information, and for more information on patient and genetic mutations, see Figure S9 in Supporting Information).

Furthermore, leading macrocycles 2 and 15 were found to abrogate CDK1 phosphorylation in TP53-mutated human triple-negative breast cancer cells (MDA-MB-231^HM) due to intracellular inhibition of WEE1. These results suggest that intracellular WEE1 inhibitory activity likely underlies the strong efficacy of 2 and 15 in (m)­CRC PDOs. For a detailed description of the study, see Figure S11 in Supporting Information.

Next, it was of interest to determine whether leading macrocycles 2 and 15 share a similar mechanism of action with 1 when treated against RAH051T by causing replication stress and DNA damage (Figure ).

10.

Lead macrocyclic WEE1 inhibitors have a similar mechanism of action to AZD1775 (1). (A) PDOs derived from the matched normal colon (RAH051N) and TP53-WT primary CRC tumor (RAH051T) were treated with 300 nM compound 1 and macrocyclic WEE1 inhibitors (compounds 2 and 15) or vehicle (DMSO) for 48 h before staining using the Apoptosis, DNA damage, and cell proliferation kit (BD Pharmingen) combined with flow cytometric analysis. Actively cycling cells were calculated as the percentage of cells that incorporated BrdU after a 3 h pulse with 20 μM BrdU. S phase cells were calculated with DAPI staining. DNA damage was calculated as the percentage of cells that had phosphorylated H2AX (γH2AX) using the antibody AlexaFluor 647 Mouse Anti H2AX (pS139). Results are from three independent (n = 3) experiments, and the t-test was used for statistical analysis. (B) PDOs derived from the matched normal colon (RAH051N) and TP53-WT primary CRC tumor (RAH051T) were treated with 300 nM compound 1 and macrocyclic WEE1 inhibitors (compounds 2 and 15) or vehicle (DMSO) for 48 h before analyzing by Western blot. α-Tubulin was used as a loading control. Bars represent the average of quantification of the relative pY15-CDK1/CDK1 (n = 3 biological replicates).

The number of actively cycling cells typically decreases during replication stress, and these cells were identified by incorporation of the BrdU label into treated CRC PDOs following a 3 h pulse of 20 μM BrdU. Concurrently, the percentage of S phase cells was calculated by the analysis of incorporated DAPI stain. An early response to the induction of dsDNA breaks is the phosphorylation of the histone variant H2AX on residue S139, forming γH2AX. γH2AX functions to recruit and localize DNA-damage repair proteins or cell cycle checkpoint factors to damaged DNA sites. Thus, DNA damage was calculated as the percentage of cells labeled with fluorescent antibodies specific for phosphorylated H2AX (pS139).

Lead macrocycles (2 and 15) broadly demonstrated a mechanism of action similar to that of clinical candidate 1 to induce DNA replication stress and were significantly more effective than 1 in inducing DNA damage in CRC cells. In addition, 2 and 15 significantly enhanced the percentage of CRC cells in the S phase compared to the vehicle, akin to 1. Lastly, and compared with 1, macrocycles 2 and 15 significantly decreased active cycling of CRC cells compared to that of the vehicle. The WEE1 inhibitor-induced effects on DNA replication and DNA damage were not observed in matched normal colon organoids, even though the Western blots demonstrated a significant intracellular decrease of pY15-CDK1 concentration due to WEE1 inhibition in both matched normal colon organoids and CRC organoids.

Although numerous simple macrocycles demonstrated potent inhibitory activity against WEE1 but exhibited weak effects on CRC PDO cell viability, a number of potent fully elaborated macrocyclic WEE1 inhibitors exhibited efficacy against the peritoneal mCRC PDO lines. Leading macrocycles 2 and 15 exhibited exceptional anticancer efficacy against all CRC PDOs utilized in this investigation, comparable to the best performing clinical candidates 1 and 16. It was observed that the strong anticancer efficacy of 2 and 15 was likely due to selectively inhibiting WEE1 intracellularly, which differs from 1, as its equipotent inhibition of both WEE1 and PLK1 may have driven its phenotype in CRC. , Therefore, the efficacy of compound 2 and 15 is not significantly different to AZD1775, but the specificity is improved. Differences in genetic backgrounds lead to the observed variability across organoid lines in response to the range of compounds. Identifying biomarkers to predict the response to these compounds will be conducted in the future. Taken together, these results emphasize the value in testing novel anticancer agents against PDOs for fast-tracking lead compound optimization and coincide with the FDA’s plan to prioritize PDO studies in drug development.

Binding Mode of 2 within WEE1

An X-ray cocrystal structure of fully elaborated macrocycle 2 bound within the ATP-binding pocket of WEE1 was generated and analyzed (Figure ).

11.

Computational molecular modeling (using ICM-Pro) on an (A) X-ray cocrystal structure of macrocycle 2 bound within the ATP-binding pocket of WEE1 (PDB ID: 9R55, 2.7 Å). (B) Overlay of the theoretical WEE1 binding mode of 2 (green, via ICM-Pro ligedit modification of 1 using PDB ID: 5V5Y, 1.85 Å) and its binding mode determined experimentally (PDB ID: 9R55, 2.7 Å), without residue interactions (left) and with residue interactions (right). RMSD was calculated using ICM-Pro. Green spheres represent hydrogen bond interactions, and black spheres represent π–π/cation interactions.

In Figure A, an X-ray cocrystal structure was generated of leading macrocycle 2 (PDB ID: 9R55, 2.7 Å) bound to WEE1, and all residue interactions were observed to be conserved when compared to the WEE1 complex with 1 (PDB ID: 5V5Y, 1.85 Å). The only differences were the addition of a π–cation interaction between the pyridine ring and K328, as well as a π–π interaction between the pyridine ring and F310. In Figure B, the binding modes of 2 that were determined computationally (see Figure ) and experimentally via an X-ray cocrystal structure were superimposed for comparison, and a low RMSD was determined (RMSD = 0.73 Å). The only differences were slight changes in the orientation of the 3-atom ethereal linker, cis-allyl group, IPA group, and N-methylpiperazine group, all of which were expected due to being flexible or rotatable functional groups. In addition to the residue interactions between WEE1 and 1 (see Figure ), the theoretical binding mode of 2 determined computationally via ICM-Pro ligedit modification of 1 in complex with WEE1 (PDB ID: 5V5Y, 1.85 Å, see Figure ) predicted the π–π interaction between the pyridine ring and F310, but not the π–cation interaction between the pyridine ring and K328, likely due to the residue being rotated in the opposite direction in the model, with the protein being mostly rigid. Furthermore, the experimental WEE1 binding modes of 1 and 2 were overlaid and compared, which also demonstrated low RMSD (0.65 Å), highlighting the success of the macrocyclization strategy in fixing the favored WEE1 bioactive conformation of 1 (see Figure S12 in Supporting Information). Nevertheless, these results showcase the advantageousness of using the ligedit feature of ICM-Pro for CADD of macrocyclic analogues when the X-ray cocrystal structures of acyclic drug candidates are available.

Additionally, an X-ray cocrystal structure of leading cyclobutanol macrocycle 15 bound within the ATP-binding pocket of WEE1 was generated and analyzed, which demonstrated a slightly different binding mode to 2 due to the rigid cyclobutane moiety being positioned in a hydrophobic pocket (PDB ID: 9R56, 3.1 Å). For a detailed description, see Figure S13 in Supporting Information. For complete crystallographic data of 2 and 15, see Table S3 in Supporting Information.

Kinome Selectivity

The stability of proteins decreases with increasing temperature, ultimately resulting in denaturation. The melting temperature (T _m) of a protein is the temperature at which the concentrations of the folded and unfolded forms are equal. ΔT _m tends to increase upon ligand binding, and this stability increase is a result of the free energy contribution. Therefore, ΔT _m values of ligand–protein complexes can be determined via differential scanning fluorimetry (DSF) in a thermal shift assay (TSA) in order to evaluate ligand–protein binding. Resulting ΔT _m data, in particular for kinases, correlate well with experimentally determined K _D and IC₅₀ values. ,, For an example comparison between ΔT _m and K _D values in regard to binding with the kinase AAK1 with known inhibitors, see Table S4 in Supporting Information. Briefly, a ΔT _m value of 4.4 °C resulted in a K _D value of 900 nM, a ΔT _m value of 7.2 °C resulted in a K _D value of 200 nM, a ΔT _m value of 9.1 °C resulted in a K _D value of 11 nM, and a ΔT _m value of 15.2 °C resulted in a K _D value of 1.2 nM.

Macrocyclization of acyclic drug candidates has been reported to increase selectivity for biological targets over proteins of similar structures, such as kinases, due to locking favored bioactive conformations. Therefore, it was of interest to profile leading macrocycle 2 (WEE1 IC₅₀ = 42 ± 3 nM, PLK1 IC₅₀ = 1600 ± 800 nM and SI = 34) against a panel of kinases to evaluate kinome-wide selectivity, beyond WEE1 selectivity over PLK1 alone. Here, 2 was tested over a panel of 100 proteins in a TSA using DSF to measure ligand–protein binding by calculating ΔT _m values (for completed data, see Figure S14 in Supporting Information).

Macrocycle 2 interacted with only 19 kinases (ΔT _m values of ≥3 °C); however, using the range discussed earlier in this section, 2 clearly demonstrated kinome-wide selectivity (Figure ). This was due to 2 returning a high-affinity thermal shift for WEE1 (ΔT _m = 15.2 °C) which notably exceeded the next highest kinase ABL1 (ΔT _m = 9.2 °C).

12.

Summary of the TSA kinome selectivity profile of leading macrocycle 2 against 100 proteins.

Specifically, 2 is expected to exhibit low single-digit nanomolar binding affinity with WEE1 due to returning a ΔT _m of 15.2 °C. Akin to WEE1, the only additional kinase from the panel that is expected to demonstrate tight binding with 2 is ABL1 (ΔT _m = 9.2 °C), with a predicted K _D value of 10–20 nM. Eleven additional kinases (ΔT _m values of >4.4 and <9.3 °C) are expected to return K _D values within 200–900 nM. However, taken together, 2 demonstrated significantly decreased binding (based on ΔT _m values) with all other proteins in the panel compared to WEE1, highlighting the positive impact on kinase selectivity conferred by macrocyclization. It was noted that clinical candidate AZD1775 (1) was reported to exhibit a similar high-affinity thermal signature for WEE1 (ΔT _m = 15.2 °C) compared to 2.

Building on these results, it was then of interest to further assess the broader kinome-wide selectivity of leading macrocycle 2, however, through direct WEE1 target engagement in cells. We first determined the target engagement of 2 in a cellular NanoBRET assay. Compound 2 potently bound to WEE1 with an EC₅₀ of 15 nM. Thereafter, 2 was subjected to nanoBRET cellular screening (K192 assay) for the percentage of occupation against 192 kinases at a concentration of 1 μM (Figure ).

14.

Summary of the cellular NanoBRET kinome selectivity profile of leading macrocycle 2 at 1 μM concentration against 192 kinases.

Apart from WEE1 (occupancy = 89%), there were only 6 other kinases with >50% occupancy, with the closest off-targets being MAP3K4 (occupancy = 82%) and TNNI3K (occupancy = 68%). For complete NanoBRET K192 assay data, see Figure S15 in Supporting Information. We then determined the potency of binding in the cellular NanoBRET assay of the closest off-target kinases MAP3K4 and TNNI3K as well as the closely related homologue WEE2 (Figure ) to further assess selectivity. Macrocycle 2 returned significantly decreased EC₅₀ values of 660, 500, and 415 nM against WEE2, MAP3K4, and TNNI3K, respectively, compared to the EC₅₀ value of 15 nM against WEE1. These results clearly demonstrated a good engagement window for 2 between WEE1 and the off-target kinases, approximately 28-fold or greater.

13.

NanoBRET cellular target engagement of WEE1, WEE2, MAP3K4, and TNNI3K via treatment with leading macrocycle 2 and corresponding dose–response curves.

Additionally, it was noted that the closest off-target protein determined in the TSA (see Figure ) was ABL1, and a cellular EC₅₀ value was not determined for 2 against this protein because it was tested against ABL2 for the percentage of occupancy. ABL2 and ABL1 are close homologues, which only differ by 2 solvent exposed amino acid residues in their ATP-binding domains. Macrocycle 2 only returned a percentage of occupancy of 30% for ABL2; therefore, similar results were expected for ABL1.

Taken together, through the DSF TSA kinase panel (100 proteins), cellular NanoBRET kinase panel (192 kinases), direct target engagement EC₅₀ values in cells (WEE1 vs its homologue WEE2 and off-targets MAP3K4 and TNNI3K), and biochemical inhibitory assays (WEE1, PLK1, and CHK1), 2 clearly demonstrated good kinase selectivity, which was the result of macrocyclization, establishing 2 as a selective WEE1 inhibitor chemical probe.

In Vitro Metabolic Evaluation of Leading Macrocycles

Building on the strong CRC PDO efficacy demonstrated by leading macrocycles 2 and 15, significantly decreased toxicity detected in patient-matched normal healthy colon organoids, and the exceptionally broader kinase selectivity profile of 2, it was then of interest to profile both 2 and 15 against human and mouse liver microsomes in vitro to assess their metabolic stability profiles (Table ).

3. In Vitro Metabolic Evaluation of Leading Macrocycles (2 and 15).

cmpd	T 1/2 (min)		CLint ((μL/min)/mg protein)		unbound CLint ((μL/min)/mg protein)
	mouse	human	mouse	human	mouse	human
2	16	7	90	205	143	326
15	16	5	85	287	137	464

^a T _1/2 refers to the degradation half-life in mouse and human microsomes.

^bCL_int refers to in vitro plasma-protein-bound intrinsic microsomal clearance.

^cUnbound CL_int refers to the in vitro plasma protein unbound intrinsic microsomal clearance.

It was determined that the degradation half-lives of 2 and 15 in both mouse and human liver microsomes (T _1/2 = 16 and 7 min, and 16 and 5 min, respectively) were very short. In addition, in both mouse and human liver microsomes, respective high intrinsic clearances (CL_int) are 90 and 205 μL/min/mg protein for 2 and 85 and 287 μL/min/mg protein for 15. In addition, respective unbound intrinsic clearances (unbound CL_int) are 143 and 326 μL/min/mg protein for 2 and 137 and 464 μL/min/mg protein for 15. Albeit being potent anticancer tool compounds, these results, overall, reflect pharmacokinetic characteristics of 2 and 15 that require future optimization if oral bioavailability is an end-point.

Conclusions

WEE1 is an appealing drug target for cancer treatment due to being a key modulator of cellular homeostasis. Being clinically efficacious but not well tolerated, the potent WEE1 inhibitor AZD1775 (1) was discontinued after several years of evaluation, purportedly due to selectivity limitations, with PLK1 reported as an equipotently inhibited off-target. Macrocyclization of acyclic drug candidates is a valuable method for prelocking favored bioactive conformations and can be utilized to enhance target selectivity and limit off-target interactions, especially if off-target proteins share similar catalytic domains. However, the synthetic preparation of macrocycles is typically difficult, as it often involves lengthy multistep routes and suffers from low yields. In this study, a CADD campaign was employed to rationally design a macrocycle 2 based on the structure of 1 to improve WEE1 selectivity, which involved computational molecular modeling and MD simulations on an X-ray cocrystal structure of 1 bound to WEE1. This rationale stemmed from the observation that 1 exhibits a U-shaped bioactive conformation in its WEE1-bound state. To capture this favorable geometry, macrocycle 2 was designed to include a 3-atom linker connecting the meta-positions of the pyridine and phenyl rings, with an oxygen incorporated at the 2-position to form an ether linkage, which is typically more synthetically accessible in macrocyclization pathways. To validate the macrocyclization strategy, we designed a simpler truncated model macrocycle 3 based on 2 and assessed its synthetic feasibility via a computational localized strain analysis. In a biochemical assay, 3 displayed markedly improved WEE1 selectivity over PLK1 compared with the corresponding acyclic control 4 derived from 1. Building on this success, an investigation on linker heteroatom modification was conducted on the simpler macrocyclic scaffold of 3, however, with the reinstatement of the key N-allyl group of 1, which ultimately resulted in frontrunning simple yet potent allyl macrocycle 5 (WEE1 IC₅₀ = 57 nM and SI = 25), which was highly selective over PLK1 and possessed an oxygen atom at the 2-position of its 3-atom alkyl linker. The simple yet potent macrocyclic scaffold of 5 was then subjected to elaboration with key functional groups also borne by 1, as well as an SAR explorative analysis on the pyridyl and phenyl rings to assess effects on WEE1 potency and selectivity over PLK1. Therefore, a focus set of elaborated macrocyclic WEE1 inhibitors (IC₅₀ values of <120 nM) with varying degrees of WEE1 inhibitory selectivity over PLK1 (SI = 10–125) were synthesized, including fully elaborated macrocycle 2 (WEE1 IC₅₀ = 57 nM and SI = 38) which was the product of the initial and extensive CADD campaign. All macrocycles from this investigation were found to be more selective for WEE1 over PLK1 than 1.

The macrocyclic WEE1 inhibitors were then tested against two peritoneal mCRC PDOs lines, PM003 and PM025, to determine anticancer efficacy, and the best performing macrocycles were indeed 2 (PM003 IC₅₀ = 164 nM and PM025 IC₅₀ = 239 nM) and cyclobutanol counterpart 15 (PM003 IC₅₀ = 143 nM and PM025 IC₅₀ = 273 nM). Moreover, the leading macrocycles were tested against two liver mCRC PDO lines, and 2 returned an IC₅₀ value of 169 nM against QEH039LM as well as 740 nM against QEH042LM. In addition, 15 returned an IC₅₀ value of 147 nM against QEH039LM and 760 nM against QEH042LM. Gratifyingly, leading macrocycles 2 and 15 demonstrated robust anticancer activity across all mCRC PDO lines in our proprietary PDO platform, showing efficacy comparable to and, in certain instances, exceeding that of the prior clinical frontrunner 1 and the current best-in-class agent ZN-c3 (15).

Furthermore, 2 and 15 were also tested against patient-matched normal healthy (RAH051N) and primary tumor CRC (RAH051T) PDOs, and a favorable therapeutic window was determined, i.e., increased tumor cell sensitivity to these macrocycles in comparison to normal matched colon cells, with 15 exhibiting an outstanding TI value of 49. This study also highlighted that the macrocyclic WEE1 inhibitors were also highly selective for and efficacious against CRC with TP53-WT compared to patient-matched normal colon organoids. In addition, when tested against RAH051T, 2 and 15 were determined to have caused significantly enhanced DNA damage and DNA replication compared to 1; however, induced effects on DNA replication and DNA damage were not observed in RAH051N, even though the Western blots demonstrated a significant intracellular decrease of pY15-CDK1 concentration due to WEE1 inhibition in both matched normal healthy colon and CRC organoids. Therefore, leading macrocycles 2 and 15, overall, broadly shared a similar yet superior mode of action.

Taken together, and gratifyingly, it was recognized that the strong anticancer efficacy of leading macrocycles 2 and 15 was likely due to the selective and potent inhibition of WEE1, consequently inducing enhanced DNA damage compared with 1. Accordingly, macrocycles 2 and 15 outperform 1 as anticancer agents, as the equipotent inhibition of WEE1 and PLK1 by 1 likely influenced its CRC phenotype, and the PLK1 inhibition may have contributed to the tolerability issues observed in patients, particularly the grade 3/4 hematological adverse events.

X-ray cocrystal structures of 2 and 15 bound within the active site of WEE1 were also generated, exhibiting similar experimental structural binding modes in agreement with those predicted computationally. Gratifyingly, the experimental binding mode of 2 was superimposed with the computationally predicted counterpart determined during the extensive CADD campaign for comparison, and a low RMSD was determined (RMSD = 0.73 Å), highlighting the accuracy of ICM-Pro in predicting macrocyclic binding modes from acyclic counterparts. In addition, these results demonstrate that the macrocyclization strategy was successful in fixing the favored bioactive conformation of 1.

In terms of broader kinome-wide selectivity, results from the DSF TSA kinase panel (100 proteins), cellular nanoBRET kinase panel (192 kinases), direct target engagement in cells via nanoBRET (WEE1 vs homologue WEE2 and close off-targets MAP3K4 and TNNI3K), and biochemical inhibitory assays (WEE1, PLK1, and CHK1) demonstrated that macrocycle 2 is highly selective for WEE1. It was noted that 2 potently inhibited WEE1 in cells (EC₅₀ = 15 nM) and returned significantly decreased EC₅₀ values of 660, 500, and 415 nM against WEE2, MAP3K4, and TNNI3K, respectively. Overall, these results demonstrated a good activity window for 2 between WEE1 and the off-target kinases, approximately 28-fold or greater. Macrocycle 2 clearly demonstrated kinase selectivity, which was conferred via macrocyclization.

Overall, this comprehensive acyclic drug to macrocyclic drug candidate CADD investigation provides critical insight into a new generation of selective macrocyclic WEE1 inhibitors that exhibit improved anticancer efficacy and reduced off-target activity and possess a large potential therapeutic window for treatment. In addition, this study demonstrates the value of CADD approaches, including interactive de novo design with the ligedit feature of ICM-Pro, for transforming nonselective drug candidates into macrocyclic derivatives with improved selectivity when X-ray cocrystal structures are available and off-targets exhibit similar structural motifs.

Experimental Section

Rational CADD of a Macrocyclic WEE1 Inhibitor

Computational molecular modeling was conducted using ICM-Pro (version 3.9–4). First, the X-ray cocrystal structure of the relevant ligand bound to WEE1 was analyzed (as the XR file) and was converted to a ‘PDB file’ in order to prepare the protein–ligand complex. Here, the boxes that were already ticked were left untouched, but the following boxes were ticked; “guess formal charges in ligands” and “auto fix backbone and side chain”. Next, on the ligedit tab, the “Setup Ligand” feature was conducted at a pH of 7.4. Afterward, the “Setup Receptor” feature was employed, with the receptor limited by a box encompassing the ligand bound within the active site of WEE1 and surrounding pocket residues. This resulted in an RTCNN binding score for the use of the ligedit feature for ligand modification. Novel macrocycles were designed using the ligedit feature by modifying AZD1775 (1, PDB ID: 5V5Y, 1.90 Å) to calculate binding scores after energy minimization and/or redocking for comparison against 1. X-ray cocrystal structures generated in this investigation for 2 (PDB ID: 9R55, 2.7 Å) and 15 (PDB ID: 9R56, 3.1 Å) were subjected to the aforementioned methods for analysis.

MD Simulations

The MD simulations of 1 (PDB ID: 5V5Y, 1.90 Å) and macrocycle 2 (generated using the ligedit feature of ICM-Pro on PDB ID: 5V5Y, 1.90 Å) were carried out using the Desmond simulation package of Schrödinger LLC. The NPT ensemble with a temperature of 300 K and a pressure of 1 bar was applied in all runs. The simulation length was 400 ns with a relaxation time 1 ps for ligands BT13 and BT18. The OPLS_4 force field parameters were used in all simulations. The long-range electrostatic interactions were calculated using the particle mesh Ewald method. The cutoff radius in Coulomb interactions was 9.0 Å. The H₂O molecules were explicitly described using the TIP3P model. The Martyna–Tuckerman–Klein chain coupling scheme with a coupling constant of 2.0 ps was used for the pressure control and the Nosé–Hoover chain coupling scheme for the temperature control. Nonbonded forces were calculated using an r-RESPA integrator where the short-range forces were updated every two steps and the long-range forces were updated every six steps. The trajectories were saved at 100 ps intervals for analysis. The behavior and interactions between the ligands and protein were analyzed using the Simulation Interaction Diagram tool implemented in the Desmond MD package. The stability of MD simulations was monitored by analyzing the root-mean-square deviation (RMSD) of the ligand and protein atom positions in time.

Strain Analysis via StrainViz

For a detailed description of StrainViz, see Colwell et al. In brief, geometry optimization of the structure of the relevant macrocycle was initially conducted using Gaussian-16. Individual fragments were then generated at appropriate disconnections of the structure using Avogadro (version 1.2.0). With the optimized geometry file of the relevant macrocycle and it is corresponding acyclic fragments in hand, StrainViz calculations were conducted using Becke, 3-parameter, Lee–Yang–Parr (B3LYP). Total and localized strain energies (kcal/mol) were obtained from StrainViz and visual representation of the relevant macrocycle conducted via Visual Molecular Dynamic version 1.9.4a48.

In Vitro Biochemical Kinase Assays

Kinase inhibitory assays were conducted at Reaction Biology (Germany). The IC₅₀ profiles of the compounds were determined using at least three protein kinases (WEE1, PLK1, and CHK1). IC₅₀ values were measured by testing 10 concentrations (variable concentration ranges) of each compound in duplicate or triplicate.

Sample Preparation

The compounds as powders were dissolved to 100× highest assay concentration (1 or 10 μM) in volumes of 100% DMSO. 100 μL of each of the stock solutions was transferred into column 2 of four 96-well master plates. Prior to testing, the 100× stock solutions in column 2 of the master plates were subjected to a serial, semilogarithmic dilution using 100% DMSO as a solvent. Subsequently, 2 × 10 μL from each well of the serial diluted copy plates were aliquoted with a 96-channel pipettor into x identical sets of compound dilution plates. In the process, 90 μL of H₂O was added to each well of each compound dilution plate. To minimize potential precipitation, H₂O was added to each plate only a few minutes before the transfer of the compound solutions into the assay plates. The plates were shaken thoroughly, resulting in compound dilution plates/10% DMSO. Several compounds were noted to have exhibited turbidity at the highest concentration(s) upon addition of water to the 100% DMSO stock solutions, indicating possible solubility issues. For the assays, 5 μL of solution from each well of the compound dilution plates/10% DMSO was transferred into the assay plates. The final volume of the assay was 50 μL. All compounds were tested at 10 final assay concentrations in variable ranges. The final DMSO concentration in the reaction cocktails was 1% in all cases.

Recombinant Protein Kinases

All protein kinases provided by RBE were expressed in Sf9 insect cells or in E. coli as recombinant GST-fusion proteins or His-tagged proteins, as either full-length or enzymatically active fragments. All kinases were produced from human cDNAs and purified by either GSH affinity chromatography or immobilized metal. Affinity tags were removed from several kinases during the purification. The purity of the protein kinases was examined by SDS-PAGE/Coomassie staining, and the identity was checked by mass spectroscopy.

Kinase Assays

A radiometric protein kinase assay (³³PanQinaseTM Activity Assay) was used for measuring the kinase activity of the protein kinases. All kinase assays were performed in 96-well ScintiPlatesTM from PerkinElmer (Boston, MA, USA) in a 50 μL reaction volume. The reaction cocktail was pipetted in four steps in the following order: 25 μL of assay buffer (standard buffer/[γ-33P]-ATP), 10 μL of ATP solution (in H₂O), 5 μL of the test compound (in 10% DMSO), and 10 μL of the enzyme/substrate mixture. The assay for all protein kinases contained 70 mM HEPES–NaOH pH 7.5, 3 mM MgCl₂, 3 μM Na-orthovanadate, 1.2 mM DTT, 50 μg/mL PEG20000, ATP (variable) concentrations, corresponding to the apparent ATP-Km of the respective kinase, [γ −33P]-ATP (approximately 9 × 1005 cpm per well), protein kinase, and substrate (variable amounts, see Table S1 in Supporting Information.).

PDOs

Human primary and metastatic CRC and normal colon tissue samples were obtained at the time of surgery for routine pathological analysis. All participants gave informed written consent, and research was conducted in accordance with the Declaration of Helsinki, the NHMRC Statement on Ethical Conduct in Human Research, and institutional approval (HREC/16/SAC/344 SSA/18/CALHN/71). Title of the review committee was Southern Adelaide Local Heath Network Human Research Ethics Committee (SALHN HREC). Tumor samples were propagated as described by Narasimhan et al. In brief, tissue samples were first minced and enzymatically digested in organoid digestion media containing advanced DMEM-F12 (Gibco; 12634-010), Antibiotic–Antimycotic (Gibco; 15240-062), 50 μg/mL Gentamicin (Gibco, 15750–060), 1% FCS, Collagenase IV 75 U/ml (Gibco; 17104019), Dispase 125 μg/mL (Gibco 17105-041), Hyaluronidase 20 μg/mL (Sigma; H3506), and Y27632 (Sigma; Y0503) in a water bath at 37 °C for 10 min, followed by a further 30 min shaking at 150 rpm in an orbital shaker at 37 °C. Samples were strained, washed, and plated in domes of GFR-Matrigel. Human CRC organoids were cultured in colorectal cancer media containing advanced DMEM/F12 (Gibco; 12634-010), Antibiotic–Antimycotic (Gibco; 15240-062), 50 μg/mL Gentamicin (Gibco, 15750-060), 10 mM HEPES (Sigma; H0887), 1x glutamax (Gibco; 35050–061), 2x B27 (Gibco; 17504044), 500 nM A83–01 (Tocris; 2939), 50 ng/mL hEGF (Sigma; SRP3027), 1 nM [Leu15]-Gastrin 1 (Sigma; G9145), 1 mM N-acetyl-L-cysteine (Sigma; A9165), 5 μM SB202190 (Sigma; S7067), 10 μM SB431542 (Sigma; S4317), and 10 μM Y27632 (Sigma; Y0503). Colorectal cancer media were changed twice weekly, with growth monitored until passaging was required.

Normal colorectal tissue samples were rocked in chelation buffer (3 mM EDTA, 50 μM DTT in PBS) for 1 h at room temperature. Crypts were released by shaking in PBS, washed, and then plated in GFR-Matrigel. Human colorectal organoids were cultured in normal colorectal media containing advanced DMEM/F12 (Gibco; 12634–010), Antibiotic–Antimycotic (Gibco; 15240–062), 50 μg/mL Gentamicin (Gibco, 15750–060), 50% HUB­(Wnt3a) conditioned media (as described in Lannagan et al.), 20% Rspo2 conditioned media (as described in Lannagan et al.), 10 mM HEPES (Sigma; H0887), 1× glutamax (Gibco; 35050–061), 1× B27 (Gibco; 17504044), 1× N2 (Gibco; 17502048), 2 μM A83–01 (Tocris; 2939), 10 mM Nicotinamide (Sigma; N0636), 100 ng/mL Noggin (Peprotech AF-250–38–250), 50 ng/mL mEGF (Peprotech; 315–09–500), 10 nM [Leu15]-Gastrin 1 (Sigma; G9145), 1 mM N-acetyl-L-cysteine (Sigma; A9165), 5 μM SB202190 (Sigma; S7067), and 10 μM Y27632 (Sigma; Y0503). Normal colorectal media were changed twice weekly, with growth monitored until passaging was required.

Immunohistochemistry (IHC)

Organoids were washed with cold PBS and fixed in 4% paraformaldehyde (Chem Cruz, sc-281692) for 1 h at RT. After washing extensively with PBS, organoids were suspended in warmed HistoGel (Thermo Fisher, HG-4000–012), then formalin-fixed, and paraffin-embedded.

Formalin-fixed, paraffin-embedded (FFPE) organoid and tissue sections (4 μm) were dewaxed and stained using hematoxylin and eosin (H&E). For IHC staining, slides were subjected to sodium citrate or EDTA antigen retrieval in a steamer for 30 min before quenching endogenous peroxidase with 3% hydrogen peroxide in PBS/0.05% Tween-20 (PBS-T) for 5 min. This was followed by incubation for 1 h at RT with the following primary antibodies diluted in 50% protein block (DAKO, X0909) and 50% PBS-T: rabbit anti-CDX2 (1:1000 Abcam, ab76541), rabbit anti-CK20 (1:500 Abcam, ab76126), or rabbit anti-CK7 (1:100 Abcam, ab181598). After incubating with the horse radish peroxidase (HRP)-conjugated secondary antibody (Goat antirabbit HRP, 1:1000 Abcam ab205718) for 30 min at RT, the samples were washed and labeled with the DAB kit (DAKO K3468) for 10 min. Slides were counterstained using hematoxylin.

Organoid Genetic Analysis

Organoid genomic DNA preparation and TP53 sequencing was conducted according to that previously described. The primers used to amplify KRAS exon 2 were KRAS F:5–GGTGGAGTATTTGATAGTGTATTAACC–3 and KRAS R: 5–AGAATGGTCCTGCACCAGTAA–3 (220-bp PCR product). The primers used to amplify BRAF exon 15 were BRAF F 5-TCATAATGCTTGCTCTGATAGGA-3 and BRAF R 5-GGCCAAAAATTTAATCAGTGGA-3 (223-bp PCR product). Sanger sequencing was conducted by the Australian Genome Research Facility.

Organoid Compound Activity Testing Assays

Organoids were dissociated into single cells by incubation in TrypLE (Gibco) with 10 μM Y27632 (Sigma) at 37 °C with mechanical dissociation. 1000 cells were seeded in a 5 μL GFR Matrigel (Corning) dome in white-walled 96-well plates (Corning). Solidified Matrigel domes were overlaid with 100 μL of the appropriate media. Once organoids had established (2 days post seeding), media were removed and replaced by 100 μL of drug-containing media. A 7-point 10-fold dilution series of each drug was assessed in quadruplicate. Drug concentrations ranged from 100 pM to 100 μM. Drugs were dissolved in DMSO to 1000-fold the final concentration of treatment (100 nM to 100 μM), and all treatment wells were normalized to 0.1% DMSO content. Drug-containing media were refreshed every 2–3 days for 5 days of drug treatment. Organoids were photographed (2 × lens on Olympus IX53 microscope), and then cell viability was quantitated by the Cell Titer Glo assay (Promega) using a GloMax microplate reader (Promega). Data analyses were performed using GraphPad Prism 7.02 software, applying nonlinear regression (curve fit) and the equation log­(inhibitor) versus response–variable slope (four parameters) to calculate IC₅₀ values. Most organoid/compound combinations were assessed in at least 2 independent assays, each in quadruplicate. Statistical analysis was conducted in GraphPad Prism using one-way ANOVA with Tukey’s multiple comparison test.

Apoptosis, DNA Damage, and Cell Proliferation Experiments

Human normal colon organoids and matched CRC organoids were grown for 7 days in 50 μL GFR-Matrigel domes with normal colon media or CRC media. 48 h before harvesting, media were replaced with 300 nM WEE1 inhibitor or vehicle (DMSO) in media. Extra domes of organoids were treated with the most effective macrocyclic WEE1 inhibitors (Cmpd 2 and 15) for staining with each individual antibody as controls for the instrument setup. 3 h prior to harvesting, organoids were pulse-labeled with 20 μM BrdU (BD Pharmingen). Organoids were dissociated into single cells by incubation in TrypLE (Gibco) with 10 μM Y27632, 2 mM EDTA, and 1.5 μg/mL DNase 1 at 37 °C with mechanical dissociation. Preparation of cells and analysis were conducted according to the manufacturer’s instruction for the Apoptosis, DNA Damage, and Cell Proliferation Kit (BD Pharmingen Cat # 562253). Controls for the instrument setup included organoids stained with each individual PerCP-Cy5.5 Anti-BrdU (Display in Log Mode); Alexa Fluor 647 Mouse Anti-H2AX (pS139) (Display in Log Mode); and DAPI (Display in Linear Mode). Stained cells were analyzed on a BD FACS Canto II flow cytometer at a rate no greater than 400 events/sec. Three biological replicate experiments were conducted.

Intracellular WEE1 Inhibition Western Blot Assays

Cell synchronization was performed as described by Bukhari et al. A highly metastatic variant of MDA-MB-231 cell lines (3 × 10⁵ cells, MDA-MB-231^HM) was seeded in a 6-well plate in DMEM supplemented with 10% FBS. The following day, the cells were then treated with 2 mM thymidine for 18 h. The cells were then incubated with fresh DMEM with 10% FBS for 8 h, followed by a second incubation with 2 mM thymidine for 18 h. The cells were subsequently treated with DMSO (0.2% v/v, vehicle control) or WEE1 inhibitors (2 and 15 at approximately 10 × IC₅₀ acquired from the WEE1 biochemical assay). After 2 h of treatment, the cells were collected in ice-cold PBS. The cell pellets were acquired after centrifugation, frozen, and stored at −20 °C until further analysis.

Protein expression was determined by using Western blot. The cell pellet was lysed in RIPA buffer (1.0% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 50 mM Tris) supplemented with Complete, Mini, EDTA-free Protease Inhibitor Cocktail (04693159001, Roche) for protein extraction. Protein concentration was determined via the Bradford assay. Protein samples were then separated by a 4–15% polyacrylamide gel (4568084, Biorad) and transferred on an Immun-Blot Low Fluorescence PVDF Membrane (1620264, Biorad) using a Trans-Blot Turbo Transfer System (Biorad). The membrane was blocked in 5% BSA in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 h. Following blocking, the membrane was incubated with primary antibodies against the protein of interest at 4 °C overnight. The membrane was then washed and subsequently incubated with Alexa Fluor-conjugated secondary antibodies against specific species. The membrane was then imaged on an Amersham Typhoon 5 Imager at a resolution of 100 μm. The following antibodies were used: rabbit anti-CDK1 p-Y15 (4539, Cell Signaling), rabbit anti-CDK1 (77055, Cell Signaling), mouse anti-β-actin (3700, Cell Signaling), Goat anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 647 (21245, Thermo Fisher), and F­(ab’)­2-Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (11017, Thermo Fisher). The molecular weight of both CDK1 and pY15-CDK1 is around 34 kDa.

The acquired image was analyzed on ImageJ. The image was first converted to an 8-bit format. A region of interest (ROI) was then created to include the protein band, and the mean gray value of the ROI was measured. An ROI of the same size was used to measure the protein band of all the samples, including the background. The mean gray value of the background was then subtracted from the value of the protein band. The background-subtracted mean gray value of pCDK1 was then normalized to that of CDK1 from the same sample. Values were then normalized to the DMSO control.

Human normal colon organoids and matched CRC organoids were grown for 7 days in 50 μL GFR-Matrigel domes with normal colon media or CRC media. 48 h before harvesting, media were replaced with 300 nM WEE1 inhibitor or vehicle (DMSO) in media. Organoids were collected in cold PBS and then immediately lysed in M-PER buffer (Thermo Fisher, 78501) supplemented with Complete, Mini, EDTA-free Protease Inhibitor Cocktail (Roche, 04693159001) and PhosSTOP Phosphatase inhibitors (Roche, 4906845001) for protein extraction. Protein concentration was determined via the BCA assay (Pierce, 23227). Protein samples were then separated by a 4–15% polyacrylamide gel (Biorad, 4561084) and transferred onto an Immuno-Blot Low Fluorescence PVDF Membrane (Biorad, 1620263). The membrane was blocked in 5% Skim milk in phosphate-buffered saline with 0.1% Tween-20 (TBST) for 1 h. Following blocking, the membrane was incubated with primary antibodies against the protein of interest at 4 °C overnight. The following antibodies were used: rabbit anti-CDK1 p-Y15 (Cell signaling, 4539), mouse anti-CDK1 (Cell Signaling 9116), rabbit anti-alpha-tubulin (Cell Signaling, 2144) Amersham ECL Rabbit IgG, HRP-linked F­(ab’)₂ fragment (Cytiva, NA9340), and Amersham ECL Mouse IgG, HRP-linked F­(ab’)₂ fragment (Cytiva, NA9310). Blots were imaged on a BioRad ChemiDoc MP system and quantified using ImageLab software.

TSA

The assay was performed as previously described. , Purified proteins were buffered in 25 mM HEPES (pH 7.5) and 500 mM NaCl and were assayed in a 384-well plate with a final protein concentration of 2 μM in a 10 μL assay volume. Macrocycle 2 was added to a final concentration of 20 μM, using an ECHO 550 acoustic dispenser (Labcyte). As a fluorescence probe, SYPRO-Orange (Molecular Probes) was added in a 1:5000 dilution. Filters for excitation and emission were set to 465 and 590 nm, respectively. The temperature was increased from 25 °C with 3 °C/min to a final temperature of 95 °C, while scanning, using the QuantStudio5 (Applied Biosystems). Data was analyzed through the Boltzmann equation in Protein Thermal Shift software (Applied Biosystems). Samples were measured in technical duplicates.

NanoBRET Assay

The assay was essentially performed as described. Briefly, HEK293T cells were transfected with the corresponding NanoLuc fusion construct for WEE1 (C-terminal NanoLuc), WEE2 (N-terminal NanoLuc), MAP3K4 (C-terminal NanoLuc), or TNNI3K (C-terminal NanoLuc) (Promega, NV2231, NV4501, NV1591, and NV4461, respectively) using FuGENE HD (Promega, cat no. E2312) as per the manufacturer’s recommended protocol. Proteins were allowed to express for 20 h at 37 °C with 5% CO₂. Transfected cells in T75 flasks were trypsinized, resuspended in OptiMEM without phenol red (Gibco, cat. no. 11058–021), and seeded into white, low volume 384-well plates (Greiner cat. No. 784075, 10 μL of the cell suspension per well at a density of 2 × 105 cells/mL). Serially diluted inhibitors and tracer K10 (Promega, TracerDB ID: T000008) at the recommended concentration taken from TracerDB (tracerdb.org) were transferred using an ECHO 650 acoustic dispenser (Labcyte). Treated cells were allowed to equilibrate for 2 h at 37 °C at 5% CO₂ prior to measurement. The NanoBRET NanoGlo Substrate and Extracellular NanoLuc Inhibitor were added as per the manufacturer’s protocol, and filtered luminescence was measured on a PHERAstar plate reader (BMG Labtech) equipped with a luminescence filter pair (450 nm BP filter (donor) and 610 nm LP filter (acceptor)). Data was analyzed and plotted in GraphPad Prism 10 using a normalized 3-parameter curve fit with the following equation: Y = Bottom + (Top-Bottom)/(1 + 10̂((X-LogIC₅₀)))X: log-transformed inhibitor concentration, and outlier detection was performed using the ROUT method (Q = 1) as implemented in GraphPad Prism. Data is shown in technical duplicates.

NanoBRET K192 Panel

To assess kinase inhibitor selectivity, the K192 Kinase Selectivity System (Promega, cat. no. NP4050) was performed as described in detail by Schwalm et al. In brief, for plate preparation, transfection mix was prepared in white 384-well small-volume plates (Greiner, cat. no. 784075) by preplating 3 μL of 20 μL/mL FuGene HD (Promega, cat no. E2311), diluted in optiMEM medium (Gibco, cat. no. 11058–021). 1μL of transfection-ready DNA from the K192 vector source plates was added using an ECHO 650 acoustic dispenser (Beckman Coulter). The mix was incubated for 30 min, and 6 μL of HEK293T cells at a concentration of 2.5 × 10⁵ cells/mL in optiMEM medium were added. The proteins were allowed to express for 20 h. After expression, Tracer K10 was added using the concentrations recommended in the K192 technical manual, and 1 μM inhibitor was added to every second well. After 2 h of equilibration, detection was carried out using substrate solution comprising optiMEM with a 1:166 dilution of the NanoBRET Nano-Glo Substrate and a 1:500 dilution of the Extracellular NanoLuc Inhibitor. 5 μL of the substrate solution was added to every well, and filtered luminescence was measured on a PHERAstar plate reader (BMG Labtech) equipped with a luminescence filter pair (450 nm BP filter (donor) and 610 nm LP filter (acceptor)). For every kinase, occupancy was calculated in Microsoft Excel and plotted using GraphPad Prism 10. Occupancy results are the mean of two separate experiments.

Co-Crystal Structures

Cloning and Expression

The gene encoding WEE1–292–575 was synthesized by GenScript Biotech and cloned into the pMal-TEV vector, providing cleavable N-terminal His- and MBP-tags. The plasmid was transformed into competent E. coli Rosetta (DE3) cells for protein expression. The cells were grown at 37 °C until OD₆₀₀ reached ∼ 1.0. The temperature was then lowered to 16 °C. The expression was induced using 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at OD₆₀₀ ∼ 2.5. The culture was grown for an additional 16 h, and then, the cells were harvested by centrifugation (Beckman centrifuge, 6000g at 4 °C).

Protein Purification

The cells were resuspended in lysis/binding buffer (30 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 7.5 @ 4 °C, 500 mM NaCl, 5% glycerol, 0.5 mM tris­(2-carboxyethyl)­phosphine (TCEP), and 10 mM imidazole) and lysed by sonication (SONICS vibra cell, 5 s on and 10 s off cycle using a total of 18 min) in the presence of the cOmplete EDTA-free protease inhibitor (Roche, Basel, CH). The protein was purified by Ni-NTA affinity chromatography, and elution was carried out using elution buffer (30 mM HEPES pH 7.5 @ 4 °C, 500 mM NaCl, 5% glycerol, 0.5 mM TCEP, and 300 mM imidazole). The eluate was dialyzed ON into gel filtration buffer (30 mM HEPES pH 7.5 @ 4 °C, 300 mM NaCl, 5% glycerol, 0.5 mM TCEP) to remove imidazole and simultaneously cleave the His-tag using 1 mg of tobacco etch virus (TEV) protease. The cleaved protein was passed through a Ni-NTA column to remove the TEV protease and subsequently injected into a size exclusion chromatography (SEC) column (GE Superdex 75 16/600 pg) pre-equilibrated with SEC buffer (30 mM HEPES pH 7.5 @ 4 °C, 100 mM NaCl, 5% glycerol, 0.5 mM TCEP). The purified protein was concentrated to 4 mg/mL and stored at −80 °C in SEC buffer after flash-freezing in liquid nitrogen.

Crystallization of WEE1-Compound Complexes

The compounds were mixed with the protein to a final concentration of 1 mM and incubated on ice for 1 h. Initial crystallization trials were performed using commercially available screens. A drop volume of 200 nL was equilibrated against 20 μL of reservoir solution in 96-well sitting drop SWISSCI plates. The crystallization plates were incubated at 20 °C. The crystals were obtained in 20% PEG6000, 10% ethylene glycol, 0.1 M MES pH 6.0, and 0.2 M NH₄Cl for the WEE1-MIPS-54861 (macrocycle 15) complex and 20% w/v PEG 8000, 0.2 M magnesium acetate tetrahydrate, and 0.1 M sodium cacodylate, pH 6.5 for the WEE1-MIPS-54859 (macrocycle 2) complex.

Data Collection and Processing

A single crystal was picked from the crystallization plate, treated with 25% ethylene glycol in reservoir solution as the cryoprotectant, and frozen in liquid nitrogen for each complex. The data were collected at the Diamond Light Source (DLS) I04 beamline. The temperature was maintained at 100 K during data collection. A total of 3600 images were collected, the data were processed, and the intensities were scaled and converted into structure factors using xia2-DIALS , as implemented at the DLS.

Structure Determination and Refinement

The structure of WEE1-MIPS-54859 (macrocycle 2) was solved by molecular replacement using the PHENIX.PHASER program of the PHENIX suite, and that of WEE1-MIPS-54861 (macrocycle 15) was solved by molecular replacement using the MOLREP program of the CCP4 suite, with the published WEE1 structure (PDB ID 1X8B) as the search template. The models were built manually using COOT, and the structures were refined by PHENIX.REFINE of the PHENIX suite (WEE1-MIPS-54859 (macrocycle 2)) and REFMAC , of the CCP4 suite (WEE1-MIPS-54861 (macrocycle 15)). The models were validated using the wwPDB Validation System (https://validate-rcsb-2.wwpdb.org/) prior to deposition. The coordinates were deposited in the Protein Data Bank with PDB IDs 9R55 (WEE1-MIPS-54859 (macrocycle 2)) and 9R56 (WEE1-MIPS-54861 (macrocycle 15)).

Chemistry

General Material and Methods

Small-molecule ¹H and ¹³C NMR spectra were acquired at 400.13 and 100.62 MHz, respectively, on an Avance III Nanobay 400 MHz Bruker spectrometer at 298 K, coupled to the BACS 60 automatic sample changer and equipped with a 5 mm PABBO BB-1H/D Z-GRD probe. Data acquisition and processing was managed using the XWINNMR (Bruker) software package version 3.5, and plotting was managed using MestReNova version 11.0.4–18998. All NMR chemical shifts (δ) were measured in parts per million (ppm) referenced to an internal standard of residual proteosolvent: chloroform-d (7.26 ppm for ¹H and 77.16 ppm for ¹³C), methanol-d ₄ (3.31 ppm for ¹H and 49.0 ppm for ¹³C), DMSO-d ₆ (2.50 ppm for ¹H and 39.52 ppm for ¹³C). Solvents used for NMR studies were purchased from Cambridge Isotope Laboratories. Each proton resonance was assigned according to the following convention: chemical shift (δ), multiplicity, coupling constant (J Hz), and number of protons. Each carbon resonance was assigned according to the following convention: chemical shift (δ), multiplicity (where no multiplicity is assigned a singlet peak was observed), and coupling constants (J Hz) for carbon–fluorine coupling. Multiplicity is quoted as br (broad), s (singlet), d (doublet), t (triplet), q (quartet), p (pentet), and m (multiplet).

Analysis by LRMS was performed in an Agilent 6120 Series Single Quadrupole LCMS coupled with an Agilent 1260 Infinity series ultra-HPLC, a 1200 series G1312B quaternary pump, a 1260 series G1367E HiP ALS autosampler, and a 1290 series G4212A photodiode array detector. Acquisition and analysis were performed using Chemstation software Rev.B.04.03 coupled with Mass Hunter Easy Access. The conditions for liquid chromatography were: column, Agilent Poroshell 120 EC-C18 2.7 μm (3 mm × 500 mm) 120 Å; column temperature: 35 °C; injection volume: 1 μL; solvent A: 0.1% formic acid in ultrapure H₂O; solvent B: 0.1% formic acid in MeCN; flow rate: 0.5 mL/min; gradient: 5–100% of solvent B over 2.5 min, maintained for a further 3.8 min; detection: 254 and 214 nm. The conditions for mass spectrometry were: quadrupole ion source; ion mode: API-ES; drying gas temp: 350 °C; vaporizer temperature: 200 °C; capillary voltage: 3000 V (positive and negative); scan range: 100–1000 m/z; step size: 0.1 s; acquisition time: 5 min.

HRMS analyses were acquired on an Agilent 6224 TOF LCMS Mass Spectrometer coupled to an Agilent 1290 Infinity series ultra HPLC. All data were acquired and reference mass-corrected via a dual-spray electrospray ionization (ESI) source. Each scan or data point on the Total Ion Chromatogram (TIC) is an average of 13,700 transients, producing a spectrum every second. Mass spectra were created by averaging the scans across each peak and background-subtracted against the first 10 s of the TIC. Acquisition was performed using Agilent Mass Hunter Data Acquisition software version B.05.00 Build 5.0.5042.2, and analysis was performed using Mass Hunter Qualitative Analysis Version B.05.00 Build 5.0.519.13. The conditions for mass spectrometry were ion mode: ESI; drying gas flow: 11 L/min; nebulizer: 45 psi; drying gas temperature: 325 °C; capillary voltage: 4000 V; fragmentor (160 V), skimmer (65 V); octupole RFV (750 V), scan range: 100–1500 m/z; internal reference ions: positive ion mode = m/z = 121.050873 and 922.009798. The conditions for liquid chromatography were reversed-phase HPLC analysis fitted with an Agilent Zorbax SB-C18 Rapid Resolution HT 1.8 μm (2.1 mm × 50 mm) 80 Å column; solvent A: 0.1% formic acid in ultrapure H₂O; solvent B: 0.1% formic acid in MeCN; flow rate: 0.5 mL/min; gradient: 5–100% of solvent B over 3.5 min.

All assayed compounds are >95% pure by analytical HPLC. Compound purity was determined using an Agilent 1260 Infinity Analytical HPLC (1260 Infinity G1322A Degasser, 1260 Infinity G1312B Binary pump, G1367E HiP ALS autosampler, 1260 Infinity G1316A Thermostated Column Compartment, and 1260 Infinity G4212B DAD detector). The conditions for liquid chromatography were: column, Agilent Zorbax Eclipse Plus C18 Rapid Resolution 3.5 μm (4.6 mm × 100 mm) 95 Å; column temperature: 35 °C; injection volume: 1 μL; solvent A: 0.1% trifluoroacetic acid (TFA) in ultrapure H₂O; solvent B: 0.1% TFA in MeCN; flow rate: 1.0 mL/min; gradient: 5–100% of solvent B over 9 min, maintained for a further 1 min; detection: 254 and 214 nm.

Glossary

Abbreviations

ATM

ataxia telangiectasia mutated kinase

ATR

ataxia telangiectasia and Rad-3-related kinase

CADD

computer-aided drug discovery

CDC25

cell division cycle 25

CDK1/2

cyclic-dependent kinases 1 and 2

CHK1

checkpoint kinase 1

cmpd

compound

DNA

DNA

DDRC

DNA-damage repair checkpoint

mCRC

metastatic colorectal cancer

mRNA

messenger ribonucleic acid

neg.

negative

N/A

not available

p

phosphorylated

PDOs

patient-derived organoids

PLK1

polo-like kinase 1

TPSA

topological polar surface area

SD

standard deviation

t _1/2

degradation half-life

XRD

X-ray diffraction

y/o

years old

2D

two-dimension

3D

three-dimensional.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.6c00018.

• Molecular formula strings (PDB)

• 3D PDB files of docking models (using PDB ID: 5V5Y) of AZD1775 (1), Cmpd 2, and Cmpd 15 (PDB)

• (PDB)

• Associated content sectionsvalidation of macrocyclization strategy, strain analysis of linker-modified simple allyl macrocycles, synthetic procedures of the simple macrocycles and acyclic control compound, CADD of elaborated macrocycles, histology of PDO lines derived from CRC LM, peritoneal mCRC organoid images and dose–response curves, measuring therapeutic windows using primary CRC PDOs with mutated TP53, patient summary and key gene mutations of the CRC PDO platform, leading macrocyclic WEE1 inhibitors abrogate CDK1 phosphorylation in cells, X-ray cocrystal structures of leading macrocycles 2 and 15 bound to WEE1, kinome profiling, and biochemical kinase inhibition assay reproducibility (CSV)

• General procedures, compound characterization, representative NMR spectra, representative HPLC chromatograms, and single-crystal XRD (PDF)

The authors declare no competing financial interest.