Discovery of Selective and Orally Bioavailable Heterobifunctional Degraders of Cyclin-Dependent Kinase 2

Abstract

Cyclin-dependent kinase 2 (CDK2) plays an important role in cell cycle regulation and has emerged as a compelling target for the treatment of cancer, largely because of its potential to overcome the resistance associated with CDK4/6 inhibition. Efforts to develop CDK2 inhibitors have historically proven challenging due to undesirable safety profiles associated with inhibiting off-target CDK isoforms. Herein, we describe the structure-guided discovery of a series of orally bioavailable and selective degraders of CDK2. Degrader 37 demonstrated improved phenotypic selectivity compared to a clinical CDK2 inhibitor, with greater specificity for disease-relevant cyclin E1 (CCNE1)-amplified cancer cells vs nonamplified cohort. The antitumor activity of 37 in mice bearing CCNE1-amplified HCC1569 tumors correlated with sustained >90% degradation of CDK2 and sustained 90% inhibition of Rb phosphorylation.

Introduction

The CDKs consist of a large family of serine/threonine kinases that regulate key cellular processes such as RNA transcription and cell cycle progression via the binding of specific regulatory subunits, known as cyclins. , CDKs are less active in their monomeric form compared with their cyclin-CDK heterodimer complex, and their activation also requires post-translational modifications involving an intricate sequence of phosphorylation reactions. Dysregulation and overactivation of cell cycle CDKs can cause uncontrolled cell proliferation linked to cancer initiation and progression. Pharmacological CDK inhibition through cell cycle arrest and apoptosis has therefore emerged as a promising approach in cancer therapy. Several pan-CDK inhibitors have entered clinical trials; however, it was recognized that the efficacy and safety balance of these would need to be improved upon via more selective CDK targeting. Recent FDA approvals of dual CDK4/6 inhibitors such as palbociclib, ribociclib, and abemaciclib for the treatment of HR+/HER2– breast cancer have rekindled interest in the potential of CDK selective inhibition. CDK2 is of particular interest since a compensatory pathway involving this enzyme has been identified both preclinically and in patients as a resistance mechanism to CDK4/6 inhibition. , Therefore, combined inhibition of CDK2 with CDK4/6 may provide a more durable therapeutic response. However, exquisite CDK2 selectivity, particularly over CDK1, will likely be needed to avoid the exacerbation of CDK4/6 SMi-related heme and GI toxicities.

CDK2 is an important cell cycle regulator that is active during G1 phase and throughout S phase, where its cyclin partner switches from cyclin E to cyclin A. CDK2 is activated by the binding of cyclins E and A, phosphorylation by the CDK-activating kinase (CAK) complex, and dephosphorylation of tyrosine-15 by cell division cycle 25 A (CDC25A). − In complex with cyclin E, CDK2 phosphorylates and inactivates Rb, leading to the release of the E2F transcription factor, expression of G1/S transition related genes, and transition from G1 to S phase. There has been additional interest in CDK2 inhibition due to the finding that CDK2 hyperactivity or amplification of CCNE1, the gene that encodes the cyclin partner to CDK2, cyclin E, drives certain cancers such as ovarian, endometrial, and breast. − The CDK2 knockout mouse is viable with few phenotypes, suggesting a CDK2 inhibitor would be well tolerated, with the potential to safely treat both CCNE1-amplified cancers alone, or in combination with CDK4/6 inhibitors in HR+/HER2- metastatic breast cancers. −

Consequently, there has been much interest in the development of selective CDK2 inhibitors. This is a challenging task due to the highly conserved nature of the ATP binding site between CDK isoforms, and poor selectivity, particularly versus CDK1, has limited the clinical utility of CDK2 inhibitors. However, advances have been made toward this goal, and several CDK2 inhibitors have now progressed into clinical trials, including PF-07104091, BLU-222, INCB123667, INX-315, ARTS-021, BG-68501, and AZD-8421; however, it remains unclear if improvements in CDK1 selectivity are sufficient to enhance the clinical therapeutic index. −

Targeted protein degradation utilizes bifunctional molecules that recruit a protein of interest in proximity to an E3 ubiquitin ligase to trigger protein degradation. This approach has the advantage of eliminating both the enzymatic activity and scaffolding functions of therapeutic proteins of interest and the potential to overcome resistance that emerges from treatment with small molecule inhibitors. CDK2 degraders have recently appeared in the scientific and patent literature, and NXT3964 has entered clinical trials. −

Results and Discussion

At the onset of our work, we envisioned that this alternative modality might enhance selectivity over CDK1 through more productive ternary complex formation. In order to maximize selectivity over CDK1, we decided not to rely solely on preferential ternary complex formation but to design binders of CDK2 that were as selective as possible. We envisioned that it may be possible to dial out the CDK4,6 inhibition of PF-06873600 and cpd 5, by replacement of their amino-piperidine-sulfonamides with the methyl-anilino-sulfonamide unit present in CDK1,2,5 inhibitor TMX-3010 (Figure ).

1.

Reported nonselective inhibitors of CDK2.

The basis of this selectivity strategy was that the tolyl substituent in TMX-3010 would interact favorably with Phe82 in CDK2, which is a smaller, more polar histidine in CDK4 and CDK6 (Figure ). Additionally, a water molecule is present in CDK4 and 6 that interacts favorably with the histidine residue, which the tolyl substituent would interact unfavorably with.

2.

Rationale for CDK4 and 6 sparing activity of TMX-3010 in a model of TMX-3010 and cpd 5 bound to CDK2 (PDB code: 1JSV). As seen in the binding pose, the tolyl methyl group makes a favorable van der Waals interaction with Phe82 of CDK2, which is a smaller, more polar histidine in CDK4 and 6.

Consistent with our design hypothesis, we were delighted to find that hybrid compound 1 displayed >44× and >124-fold selectivity over CDK4 and CDK6, respectively, while possessing sub nM biochemical IC₅₀ (Figure ).

3.

Scaffold hopping from reported unselective CDK2 inhibitors to a selective CDK2 inhibitor 1, and identification of the CDK2 degrader 2.

Several degraders were synthesized based on CDK2 inhibitor 1. Degrader 2 showed a biochemical inhibition profile similar to 1, and as described in a recent Kymera publication, showed several advantages over traditional small molecule inhibitors in vitro, which translated in vivo with greater ease in driving antitumor activity via requisite pathway inhibition. Despite their biological advantages over small molecule inhibitors, oral bioavailability in this series was a challenge. Thus, we sought to design a CDK2 ligand with physicochemical properties better positioned to enable access to an orally bioavailable degrader. In our design strategy, we increased the three-dimensionality of the ligand (Fsp3) and removed the strong hydrogen bond acceptor (HBA) in the pyrido-pyrimidinone core. We also hypothesized that incorporation of the tertiary alcohol featured in PF-06873600 may improve solubility and reduce intrinsic clearance (Figure ).

4.

Rational design of CDK2 ligands with improved physicochemical properties.

Installation of a trifluoromethyl group at the 5-position of the pyrimidine core was incorporated to lessen the potential for undesired metabolism on the electron-rich diaminopyrimidine. The structural similarity of recently described CDK2 inhibitors (Figure , example 10) provided confidence in maintaining CDK2 activity on scaffold hopping. Although less potent than 1, identifying cyclohexanes 3 and 4 as double-digit nM inhibitors of CDK2 was encouraging (Table ). Cyclohexane isomer 3, with the hydroxy group in a cis conformation relative to the pyrimidine substituent, displayed a better CDK family selectivity than trans isomer-4 (e.g., 47 and >357-fold CDK1 and 9 sparing for 3 vs 16 and 278-fold for 4), so was further profiled.

1. Profiles of Compounds 1 and 3–7 .

*Units of 10^–6 cm/s. **Units of mL/min/kg. ^#IV (2 mg/kg)/PO (10 mg/kg), Cl in units of mL/min/kg, V_ss in units of L/kg, T _1/2 in units of h.

A cocrystal structure of compound 3 in CDK/CCNE1 was obtained (see Figure a). The molecule forms two hydrogen bonds to Leu83 of the hinge region of the ATP binding site, with a possible weak CH···O hydrogen bond to Glu81. Like 1, the cycloalkyl ring occupies the ribose pocket. The sulfonamide makes three additional hydrogen bonds to Asp86 and Lys89, and the reduced potency of cyclohexanol 3 vs 1 can be explained by reduced hydrophobic contact in the back part of the pocket, particularly to Phe80 and Ala144 (Figure b). The CDK2 potency was improved by transitioning from 4-hydroxy-cyclohexane 3 to hydroxypiperidines 5 and 6. In the cocrystal structure of compound 3, further validated with molecular dynamics, the hydroxyl group in the 4-position maintained a weak hydrogen bond to the backbone carbonyl of Gln131. Modeling suggested that its repositioning to the 3-position would decrease the distance to the carbonyl, therefore improving the interaction. The added potency with the transition to piperidine could be explained by the difference in energy barrier of axial versus equatorial substitutions on piperidines vs cyclohexanes. For the hydroxyl to be in the preferred position to interact with Gln131 (Figure a), the ring needs to be in axial–axial conformation, which will be more favorable with a piperidine. Racemate 6 and more potent enantiomer 7 were single-digit nanomolar CDK2 inhibitors with good selectivity against key CDK family isoforms (e.g., 42 and >5000-fold selective vs CDK1 and 9, respectively, for 7). Compounds 3 and 7 showed increased selectivity over compounds from the earlier series when tested against a panel of nonmutant kinases (Eurofins DiscoverX). At a concentration of 1 μM, only 2/403 and 4/403 kinases tested showed >90% inhibition, respectively, versus control, compared to 18/403 kinases for 1.

5.

(a) Crystal structure of compound 3 (green)/CDK2 (PDB code: 9NYQ); (b) overlay of 3 with 1 (pink) in the crystal structure of 3/CDK2, with wire mesh representing the surface of the binding site (PDB code: 9NYQ).

The new CDK2 ligands were profiled in vitro and in vivo and showed significantly improved physicochemical properties and PK profile, as detailed in Table . In contrast to compound 1, cyclohexanol 3 was metabolically stable in human and rat microsomes, almost 10-fold more soluble, and showed improved apparent permeability (7.6 × 10^–6 cm/s) in wild-type MDCK cell monolayers. These improved in vitro ADME properties resulted in a low clearance and high oral bioavailability (86%) in rat, in contrast to earlier CDK2 ligand 1 (rat F = 3%). Piperidine 7 also showed good in vitro ADME and PK properties, albeit with higher microsome intrinsic clearance. Overall, new selective and drug-like CDK2 ligands 3 and 7 fulfilled our design criteria, providing the CDK2 ligand profiles we desired for subsequent degrader design.

Initial degraders were designed and synthesized based on achiral CDK2 ligand 3, due to its overall well-balanced properties. By incorporating the linker from our initial series of CDK2 degraders 2, we were pleased to identify potent single-digit nanomolar degraders 8 and 9 by AlphaLISA analysis of CDK2 degradation following treatment with compounds in MKN1 CCNE1^amp cells. However, compounds 8 and 9 displayed a low rat oral bioavailability. One strategy we explored to improve oral bioavailability was shielding of the sulfonamide NH by employing sulfonamides bearing nonprimary alkyl groups. We hypothesized that this modification might also lower human microsome intrinsic clearance since it is known that sulfonamides stabilize α-carbon radicals, increasing the rate of oxidation at the α-carbon. Indeed, metabolite identification studies in this subseries confirmed alkyl sulfonamide cleavage as a major metabolic pathway (data not shown). A substantial increase in oral bioavailability was realized on incorporation of a chiral methyl group adjacent to the sulfonamide NH (Table , compounds 10 and 11). Modeling analysis indicated a clear trend in decreasing 3D-PSA for compounds 9–11 compared to 8, thereby impacting permeability and consequently oral absorption. Despite gains in rat oral bioavailability, human microsome intrinsic clearance remained high, and a deuterium atom was incorporated at the α-carbon to further reduce the rate of alkyl sulfonamide cleavage. However, this modification had little impact on intrinsic clearance (compare 12 with 10 and 11). This was also found to be the case in degraders containing more potent piperidine-based ligands (e.g., 13).

2. Early Degrader Optimization.

compound	R 1	R 2	X	MKN1 CDK2 DC50 /D max	MDCK-WT Papp/MDR1 ER	logD7.4/FaSSIF	calc. minE 3D-PSA	HLM/RLM clint	rat PK (Cl, Vss, T 1/2, %F)
8	H	H	H	4.9/90%	1.0/>30	4.1, 153	286	79/<21	59/15/5.7/1%
9	H	H	F	12/86%	2.0/>54	4.6, 173	233	102/26	45/11/4.9/5%
10	(R)-Me	H	F	13/85%	1.5/>530	5.7, 110	219	181/<21	4/1/2.8/10%
11	(S)-Me	H	F	200/63%	3.1/>230	5.6, 115	218	96/<21	13/2/3.5/33%
12	Rac-Me	D	F	39/72%	2.4/>26	4.9, 131		118/<21	6/2/3.1/29%
13				1/95%	2.3/91	4.6, 62		143/25

^aUnits of nM.

^bUnits of 10^–6 cm/s.

^cUnits of μM.

^dSee the Supporting Information for details.

^eUnits of mL/min/kg.

^fIV (2 mg/kg)/PO (10 mg/kg), Cl in units of mL/min/kg, V_ss in units of L/kg, T _1/2 in units of h.

An alternative strategy to lower human intrinsic clearance was exploredreplacement of the sulfonamide alkyl ether side chain with cyclic alternatives, while conserving the general structure of the remainder of the linker (Table ). Gratifyingly, cyclohexyl-piperazine 14 and cyclobutyl-piperazines 15 and 16 demonstrated significantly improved metabolic stability. To increase the oral absorption of compounds, such as 15 and 16, several medicinal chemistry strategies were explored. Sterically shielding the sulfonamide acidic proton (Table , compound 16 vs 17 and 18), modulation of linker lipophilicity and basicity (compare compound 15 to 19), or incorporating a monobasic rather than dibasic linker (exemplified by 20) all failed to improve oral bioavailability. Finally, transitioning away from benzimidazolone CRBN binders increased passive permeability and oral bioavailability, albeit with a drop in degradation potency (Table , 15 vs 21). This increase in passive permeability correlated with a decrease in polar surface area (calc minimum energy 3D-PSA of 292 vs 254 for 15 and 21, respectively). Compound 21 also presented low to moderate clearance, moderate volume of distribution, and moderate half-life.

3. Improvement in PK Parameters via Linker and CRBN Ligand Modification.

*Racemic at 3-hydroxypiperidine. **Units of nM. ^@Units of mL/min/kg. ^$Units of 10^–6 cm/s. ^#IV (2 mg/kg)/PO (10 mg/kg), Cl in units of mL/min/kg, V_ss in units of L/kg, T _1/2 in units of h.

Having achieved improved PK with compound 21, we shifted our focus toward improving its degradation potency. We hypothesized that reducing PSA and MW via linker shortening would maintain or improve the PK properties and potentially improve degradation potency through more productive ternary complex formation. Thus, we designed a discrete set of degraders with linker lengths between 4 and 10 atoms (shortest path), incorporating non-benzimidazolone-based CRBN ligands phenylglutarimide and anilino glutarimide, to increase permeability and oral bioavailability. Noteworthy compounds are characterized in Table . Compounds with very short linkers, exemplified by 22 and 23, showed suboptimal degradation potency. Transitioning to a neutral linker resulted in a shallower degradation (compound 25). The para-position of a phenylglutarimide or meta-position of an anilino glutarimide appeared to be the most productive exit vectors to the linker (compare compounds 23 vs 26 and 27 vs 28). Compared with 14, 15, and 17, compounds possessing shorter linkers between CDK2 and CRBN ligands showed a higher normalized area under the curve (AUC) in a CDK2/Cyclin E1-CRBN ternary complex AlphaLISA assay (Table ). Compound 24 creates an optimal positive cooperative system that favors stable ternary complex formation by inducing novel interactions between CDK2 and CRBN. This results in a higher and wider AlphaLISA bell-shaped curve, increasing the AUC, for example, compared to 23, a compound with a slightly shorter linker. Compound 28 was used as the 100% control, from which normalized AUCs for other compounds were calculated.

4. Multiparameter Optimization of CRBN Ligand and Linker.

*Units of nM. ^@Units of mL/min/kg. ^$Units of 10^–6 cm/s. ^#IV (2 mg/kg)/PO (10 mg/kg), Cl in units of mL/min/kg, V_ss in units of L/kg, T _1/2 in units of h.

5. CDK2/CRBN/Degrader Ternary Complex AlphaLISA Assay.

compound	normalized AUC
14	10
15	9.5
17	13
19	27
20	7.9
21	4.0
22	0
23	4.8
24	92
26	0
27	35
28	100

A cryo-EM structure was obtained of CDK2/Cyclin E1-compound 24-CRBN/DDB1 at 3.9 Å resolution. CRBN, together with its adapter protein DDB1, engages CUL4 and catalyzes polyubiquitination and thus controls the degradation of CDK2/Cyclin E1. Compound 24 induces a novel, non-native protein–protein interaction surface, where the CDK2 N-lobe and the C-lobe engage the LON- and TBD-domains of CRBN, respectively. His-353 of CRBN engages the backbone of CDK2 Ile-10 and the piperazine linker of compound 24. Additionally, CRBN Lon domain residue Phe150, residing in a β-hairpin loop, sandwiches between CDK2 Tyr77 and Tyr19. Lastly, CDK2 Arg200 favorably interacts with the backbone of CRBN Lys392 and Ile393. All these interactions help stabilize the ternary complex (Figure ).

6.

(a) Cryo-EM structure of CDK2/Cyclin E1 – compound 24 – CRBN/DDB1 at 3.9 Å resolution (PDB code: 9NYR). Compound 24 induced non-native CDK2-CRBN interactions as highlighted in insets (b, c, and d). (b) Zoomed-in view of compound 24 (sticks) and density (transparent surface). His-353 of CRBN engages the backbone of CDK2 Ile-10 and the piperazine linker of compound 24. (c) CRBN Lon domain residue Phe150, residing in a β-hairpin loop, sandwiches between CDK2 Tyr77 and Tyr19, and (d) CDK2 Arg200 interacts favorably with the backbone of CRBN Lys392 and Ile393.

Compounds 24 and 28 provided an optimal balance of degradation potency and favorable PK. The latter compound and related anilino glutarimide analogues were subsequently deprioritized due to reduced degradation selectivity over CDK1 and other optimization challenges. Further characterization of 24 identified time- and NADPH-dependent inhibition of the CYP3A4 enzyme. We hypothesized that the generation of a quinone methide reactive metabolite from the ortho-methyl aniline motif on the CDK2 binder may have been responsible. Strategies such as fluorination or deuteration of the ortho-methyl group, fluorination of the aniline ring, or replacement of the ortho-methyl group with alternative alkyl groups did not resolve the time-dependent inhibition (TDI) (data not shown). Direct replacements of the ortho-methyl group were assessed, and the results are shown in Table . Gratifyingly, removal or replacement of the methyl group with halogen atoms (Table , compounds 30–32) eliminated the TDI liability, consistent with the quinone methide reactive metabolite hypothesis.

6. Removal of CYP3A4 TDI while Maintaining Good Kinome Selectivity.

compound	X	CYP3A4 TDI (±NADPH) μM	kinomescan S(10) score
29	Me	0.6/>10	0.007
30	H	>10/>10	0.04
31	F	>10/>10	0
32	Cl	>10/>10	0

^aSelectivity Score or S-score is a quantitative measure of compound selectivity. It is calculated by dividing the number of kinases that compounds bind to by the total number of distinct kinases tested, excluding mutant variants. S(10) = (number of nonmutant kinases with % Ctrl <10)/(number of nonmutant kinases tested).

Kinome selectivity was closely monitored due to the methyl substituent interacting with a less conserved residue. Fluorine and chlorine analogues 31 and 32 displayed excellent selectivity across the kinome in contrast to 30, where reduced selectivity was observed. Fluorine analogue 31, a more potent degrader than 32 (data not shown), was chosen for final optimization. Compound 31 was a low-clearance, orally bioavailable single-digit nanomolar degrader of CDK2. Surprisingly, during routine profiling, we observed an erosion of the CDK1 degradation selectivity in degrader 31. This finding was further confirmed via isomer 33 (Table ).

7. Late-Stage Multiparameter Optimization and Discovery of 37 .

*Units of nM. ^@Units of mL/min/kg. ^#IV (2 mg/kg)/PO (10 mg/kg), Cl in units of mL/min/kg, V_ss in units of L/kg, T _1/2 in units of h.

Mining our SAR, we found that increasing steric bulk in the linker drove higher levels of selectivity over CDK1, as exemplified by compounds 34 and 35. Overall, 35 was a well-balanced compound; however, basic amines of this subclass were subsequently deprioritized due to higher human microsomal intrinsic clearance and poor oral bioavailability in higher species. Replacement of the chiral methyl group with a fluorine atom at the 3-position of the piperidine gave diastereomers 36 and 37, and of the four possible stereoisomers (data for two shown), the (3S,4R)-3-fluoro-4-amino-piperidine stereoisomer 37 was the most potent degrader with the lowest human microsomal intrinsic clearance.

Based on its potency, selectivity over CDK1, and in vivo pharmacokinetic (PK) properties in rats, 37 was characterized further. The selectivity for degradation of CDK2 over other CDK family members was assessed in the CCNE1^nonamp TOV21G cell line. Importantly, 37 showed >100-fold selectivity over CDK1, and >500-fold selectivity over CDK9, the CDK off targets of greatest safety concern (Tables and ).

8. Further In Vitro Biological Characterization of 37 .

assay	results
MKN1 pRB IC50/IC90/Imax	23 nM/110 nM/91%
TOV21G CDK1/2/4/6/9 DC50	2130/17/>10,000/1276/>10,000 nM
Cell cycle (max G1 arrest to G2/M accumulation)	32×
CCNE1amp NC IC50 (MKN1)	9 nM
CCNE1nonamp NC IC50 (TOV21G)	5760 nM

^asee the Supporting Information for full experimental details.

Our group has previously shown phenotypic differentiation of a selective CDK2 degrader from a CDK2 SMi driven by enhanced CDK2-to-CDK1 degradation selectivity through differential ternary complex formation. In this scaffold class, we similarly observed an improved phenotypic selectivity window in a cell cycle functional analysis. With 37, maximum on-target effect (max G1 arrest) was observed at 78 nM; while the emergence of off-target activity (increased G2/M accumulation), a phenotype consistent with CDK1 inhibition was observed at 2500 nM (32-fold window) (Table and Supporting Information). Given the essentiality of CDK1 in proliferating cells, we envision that enhancement of the CDK2-to-CDK1 selectivity ratio will engender 37 with greater specificity for CDK2-dependent CCNE1^amp cancer cells. Indeed, 37 showed a superior selectivity window for CCNE1^amp to CCNE1^nonamp (32-fold) compared to PF-07104091 (5-fold). Consistent with improved cellular CDK1 and phenotypic selectivity, 37 also shows potent single-digit nanomolar antiproliferative activity in MKN1 CCNE1^amp cells compared to weak micromolar activity in TOV21G CCNE1^nonamp cells (Table ).

Following single intravenous bolus injection and oral dosing in rat, dog, and monkey, 37 was characterized by low to moderate clearance in plasma, moderate to high volume of distribution at steady state, and moderate terminal half-lives (Table ). The oral bioavailability ranged between 12 and 21%.

9. PK Parameters of 37 in Preclinical Species.

IV PK parameter	rat	monkey	dog
IV Dose (mg/kg)	2	0.07	0.1
CL (mL/min/kg)	10	9.1	3.9
Vss (L/kg)	5	5.7	4.2
T 1/2 (h)	6.7	8.6	14.4
PO PK parameters
PO Dose (mg/kg)	10	1	1
T max (h)	2.0	4.0	4.0
C max (μM)	0.261	0.014	0.026
AUClast (μM × h)	3.63	0.250	0.557
F	21%	12%	12%

The selectivity of 37 for CDK2 was investigated using tandem mass tag-based deep proteome profiling of over 8000 proteins. Treatment of human PBMC cells with 800 nM 37 (10-fold DC₉₀) for 8 h revealed a high level of selectivity, demonstrating CDK2 as the only significantly downregulated protein of over 8000 identified proteins (Figure ).

7.

Volcano plot showing deep tandem mass tag proteomics analysis in human PBMC cells treated with 37 for 8 h. No off-target protein degradation was observed.

Based on its promising profile, we examined the PD effect and antitumor efficacy of 37 in the CCNE1-amplified invasive ductal breast carcinoma HCC1569 tumor xenograft model in mice. Following oral dosing of 37 at 10, 25, and 50 mg/kg BID or 50 mg/kg QD for 3 days, peripheral blood plasma concentrations and the percent of CDK2 and pRB protein remaining in tumor tissue relative to tumor from vehicle-treated animals were determined. 37 demonstrated greater than dose proportional exposure between 10 and 50 mg/kg BID dosing (Table ).

10. PK Parameters of 37 in a HCC1569 Xenograft Model .

		dose and schedule
PK parameters	unit	10 mg/kg BID × 3 days	25 mg/kg BID × 3 days	50 mg/kg QD × 3 days	50 mg/kg BID × 3 days
T 1/2	h	2.82	2.64	2.51	2.39
C max	ng/mL	791	2233	6990	5343
AUC0–24h	h × ng/mL	12,487	35,467	68,868	94,865
AUC0–24h/D	h × mg/mL	624	709	1377	949
AUC0–24h free	h × ng/mL	12.5	14.2	27.5	19.0

^a 37 mouse plasma protein binding = 98.02%.

The plasma protein binding adjusted plasma free concentration of 37 exceeded the HCC1569 in vitro free CDK2 DC₉₀ and pRb IC₉₀ (FBS binding adjusted) for 6 h at 10 mg/kg, while doses of 25–50 mg/kg BID maintained free plasma concentration above the in vitro DC₉₀ for the dosing interval (12 h). Sustained ≥90% CDK2 degradation through C_trough was observed when plasma concentrations of 37 were greater than the adjusted in vitro DC₉₀, which was observed at doses of 25 mg/kg and higher. Sustained ≥90% degradation of CDK2 was correlated with >90% inhibition of pRB and free plasma AUC values ≥404 h × ng/mL (Figure a), which is consistent with prior findings. To determine the antitumor activity, HCC1569 tumor bearing mice were treated with 37 at 25 or 50 mg/kg PO BID or 50 mg/kg PO QD for 28 days (Figure b). Tumor growth inhibition was reported as the Day 28 mean percent change in tumor volume (%dTV) (Table ). Robust antitumor activity resulting in stable disease approaching tumor stasis was observed in response to 37 treatment at 25 mg/kg PO BID, 50 mg/kg PO BID, and 50 mg/kg PO QD. Clinical CDK2 inhibitor PF-07104091 at 150 mg/kg PO BID demonstrated similar activity. These findings demonstrate that tumor stasis is correlated with sustained >90% degradation of CDK2, leading to sustained 90% inhibition of pRb, consistent with the cell cycle arrest induced by CDK2 degradation in vitro.

8.

(a) PK/PD and (b) efficacy of 37 in an HCC1569 xenograft model.

11. Antitumor Activity of 37 in a HCC1569 Xenograft Model.

compound	dose (mg/kg)	schedule	day 28%dTV (mean ± SEM)	antitumor activity
37	25	BID	–26 ± 9	stable disease
37	50	QD	20 + 18	stable disease
37	50	BID	–24 ± 10	stable disease
PF-07104091	150	BID	4 ± 15	stable disease

^aDosed every 12 h.

^b%dTV = % change in tumor volume.

^c%dTV > 30% is progressive disease, 30 to −30% is stable disease (0% = Tumor Stasis), −30 to −99% is partial response, and −100% is complete response.

Compound 37 contains a CRBN binding glutarimide ring, which is, in principle, prone to stereoisomer interconversion under physiological conditions. Therefore, we initiated studies to assess how labile the imide stereocenter was, to inform whether compound 37 should be developed as a racemate or single diastereomer. 37 was separated by chiral SFC to give single diastereomers 37-a and 37-b in >95% purity. Both compounds were independently incubated with rat, dog, NHP, and human plasma, where the half-lives for chiral interconversion of 37-a to 37-b were 1.3, 5.5, 6.2, and 5.0 h, respectively. The corresponding half-lives for the chiral interconversion from 37-b to 37-a were 1.0, 1.2, 5.5, and 5.0 h, respectively. Next, we examined the in vivo pharmacokinetic profiles of compounds 37-a and 37-b in rats and dogs after IV and oral dosing. Following single IV and oral dose administration of 37-a in rats, the bioavailability of 37-a was 12.5% and 37-b was formed in vivo as a result of interconversion. IV and oral dosing of 37-a in rats led to 28 and 33% of 37-b, respectively, while dosing of 37-b led to 29 and 32% of 37-a, respectively. Those data clearly indicated that moderate interconversion between 37 diastereomers occurs in vivo in rats. The same experiments were conducted in dogs, and similar findings were observed. For example, in dogs, IV and oral dosing of 37-a led to 50 and 55% of 37-b, respectively, while dosing of 37-b led to 33 and 36% of 37-a, respectively. Taken together, these data established that the interconversion observed in rat, dog, and NHP plasma was recapitulated in vivo, and our findings supported the potential development of 37 as a mixture of diastereomers at the imide chiral center.

Synthesis

Amine 44 required for the synthesis of degrader 37 was prepared via a six-step procedure (Scheme ). Palladium-catalyzed thioetherification was followed by two nucleophilic aromatic substitution reactions on a trisubstituted pyrimidine. Subsequent oxidation with NCS, sulfonamide formation, and deprotection gave intermediate 44. In parallel, a chemoselective Suzuki coupling on iodobromofluorobenzene 45 followed by Buchwald coupling and two deprotections gave aldehyde 49 (Scheme ). A combination of intermediates 49 and 44 in a reductive amination reaction gave 37.

1. Synthesis of Amine 44 Needed for the Preparation of 37 .

^a Reagents and conditions: (a) BnSH, Pd₂(dba)₃, xantphos, DIEA, dioxane, 80 °C, 81%; (b) 2,4-dichloro-5-(trifluoromethyl)­pyrimidine, ZnCl₂, NEt₃, DCE, t-BuOH, 0 °C then RT, 56%; (c) (3S)-3-methylpiperidin-3-ol hydrochloride, NEt₃, MeCN, RT, 49%; (d) NCS, AcOH, MeCN, water, RT, 88%; (e) tert-butyl (3S,4R)-4-amino-3-fluoro-piperidine-1-carboxylate, DCM, RT, 95%; (f) TFA, DCM, RT, 96%.

2. Synthesis of 37 .

^a Reagents and conditions: (a) 2,6-dibenzyloxy-3-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)­pyridine, Pd­(dppf)­Cl₂.CH₂Cl₂, K₂CO₃, dioxane, water, 100 °C, 55%; (b) 4-(dimethoxymethyl)­piperidine, Pd-PEPPSI-IHeptCl, Cs₂CO₃, dioxane, 100 °C, 84%; (c) H₂ (40 psi), 10% Pd/C, THF, 40 °C, 93%; (d) HCOOH, 80 °C, 94%; (e) 44, NaBH­(OAc)₃, NEt₃, HOAc, RT, 78%.

Conclusions

Herein, we describe the discovery of molecules that induce the potent and selective degradation of CDK2. Achieving oral exposure is traditionally a challenge in the bRo5 heterobifunctional degrader space. Our optimization process therefore included the design of novel CDK2 ligands with excellent ADME and PK properties to increase the likelihood of success.

Highlighting the unique challenge associated with the multiparameter optimization of heterobifunctional degraders, our discovery of lead compound 37 included several innovations: (1) reduction of microsome intrinsic clearance by employing sulfonamido-substituted-cycloalkane/piperidine bearing linkers; (2) increasing oral bioavailability by minimizing 3D-PSA through CRBN ligand modification; (3) achievement of deeper CDK2 degradation (D_max) by minimization of linker length. The optimal linker induced novel CDK2-CRBN interactions as observed in the Cryo-EM ternary complex structure; (4) elimination of a degrader CYP3A4 TDI liability via CDK2 ligand modification; and (5) late-stage modulation of degradation selectivity in a highly conserved enzyme family class via linker structural modification.

Biological characterization of lead compound 37 showed CDK2 as the only protein downregulated in proteomic analysis. It displayed phenotypic differentiation from a CDK2 SMi in a cell cycle functional analysis, presumably driven by enhanced CDK2-to-CDK1 selectivity and antiproliferative activity preferentially in CCNE1-amplified cancer cell lines. Evaluation of 37 in a mouse xenograft model revealed robust antitumor activity. Based on its encouraging profile, it was advanced into preclinical safety studies, and results will be reported in due course. Recent disclosures of CDK2 inhibitor clinical safety data indicate some potential for heme and GI-related toxicities. , It will be critical to understand whether a selectivity profile exemplified by 37 can mitigate these liabilities.

Experimental Section

In Vivo Pharmacology

All of the procedures related to animal handling, care, and treatment in this study were performed according to guidelines approved by the Institutional Animal Care and Use Committee (IACUC) of Pharmaron (Ningbo, P.R. China) following the guidance of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). In addition, all portions of this study were performed at Pharmaron and adhered to the study protocol provided or approved by the sponsor and applicable standard operating procedures (SOPs). The approval ID for the Animal Use Protocol (AUP) is that of ON-CELL-XEN-06012023. Animals were housed according to general procedures for animal care and housing in accordance with the standard, Commission on Life Sciences, National Research Council, and Standard Operating Procedures (SOPs) of Pharmaron. Female NOG mice (Vital River Laboratory Animal Technology Co., Ltd. strain no. 408) were acclimated to the facility for 7 days prior to enrollment. HCC1569 human breast carcinoma cells were maintained in RPMI 1640 + 10% Fetal bovine serum. Mice were subcutaneously implanted on the central right flank with HCC1569 tumor cells (6 × 10⁶ + Matrigel) in 0.2 mL of RPMI 1640 for tumor development. Mice were assigned to groups (n = 9 for efficacy, n = 3 per time point for PKPD) such that the mean tumor volume was the same for each treatment group. The measurement of tumor size was conducted twice a week for efficacy, as well as once on sample collection days. Animals were dosed orally (PO) as indicated at 10 mL/kg dosing volume.

General Synthesis Methods

All compounds are >95% pure by HPLC analysis. All commercial reagents and catalysts were used as provided by the commercial supplier without purification. Unless otherwise indicated, all reactions were mechanically stirred and run under inert (N2 or Ar) conditions. Reaction progress was monitored by thin-layer chromatography (TLC) and/or LC-MS. Solvents for synthesis, extraction, and chromatography were of reagent grade and used as received. Moisture-sensitive reactions were carried out under an atmosphere of argon, and anhydrous solvents were used as provided by the commercial supplier. Reaction progress was monitored by HPLC, LC–MS, or thin-layer chromatography. Crude products were immediately purified using preparative reversed phase HPLC methodology with UV detection (instrument: Shimadzu LC-MS-2020; column: Phenomenex luna C18 150 × 25 × 10 μm and Waters Xbridge BEH C18 150 × 25 × 10 μm; eluent A: 0.225% FA in H₂O and 10 mM NH4HCO₃ in H₂O; eluent B: acetonitrile; UV detection: 210 nm) or flash chromatography on silica gel. The fractions obtained were concentrated under reduced pressure to remove organic volatiles. Unless otherwise indicated, all compounds have greater than 95% purity. A detailed description of LC-MS methods is included in the Supporting Information. 1H NMR and 13C NMR spectra were recorded in solvents indicated below at RT with Bruker AVANCE NEO 400 MHz operating at 400 MHz for 1H NMR, at 377 MHz for 19F NMR, and at 101 MHz for 13C NMR. Chemical shifts are reported in ppm relative to tetramethylsilane (TMS) as an internal standard. The descriptions of the coupling patterns of 1H NMR signals are based on the optical appearance of the signals and do not necessarily reflect the physically correct interpretation. In general, the chemical shift information refers to the center of the signal. In the case of multiplets, intervals are given. Spin multiplicities are reported as s = singlet, br s = broad singlet, d = doublet, dd = doublet of doublets, t = triplet, q = quartet, and m = multiplet.

Synthesis of 3-fluoro-N-[(3S,4R)-3-fluoro-4-piperidyl]-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­benzenesulfonamide (44)

Step 1. 4-Benzylsulfanyl-2-fluoro-aniline (39)

To a solution of 4-bromo-2-fluoro-aniline (20.0 g, 105 mmol), DIEA (55.0 mL, 316 mmol), xantphos (6.09 g, 10.5 mmol), and Pd₂(dba)₃ (9.64 g, 10.5 mmol) in dioxane (200 mL) was added BnSH (16.9 mL, 144 mmol), and then the mixture was stirred at 80 °C for 15 h. On completion, the mixture was diluted with water (150 mL) and extracted with EA (200 mL × 3). The water layers were quenched with 5% aq NaOCl (200 mL). The combined organic layers were dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure to give a residue, which was purified by reverse phase (aqueous formic acid/acetonitrile) to give the title compound (20.0 g, 81% yield) as a brown oil. ¹H NMR (400 MHz, DMSO-d ₆) δ 7.30–7.16 (m, 5H), 6.98 (dd, J = 2.0, 11.6 Hz, 1H), 6.87 (dd, J = 2.0, 8.4 Hz, 1H), 6.70–6.65 (m, 1H), 5.27 (s, 2H), 4.01 (s, 2H). LC-MS (ESI⁺) m/z 234.1 (M + H)⁺.

Step 2. N-(4-Benzylsulfanyl-2-fluoro-phenyl)-4-chloro-5-(trifluoromethyl)­pyrimidin-2-amine (40)

To a solution of 2,4-dichloro-5-(trifluoromethyl)­pyrimidine (18.6 g, 85.7 mmol) in t-BuOH (200 mL) and DCE (200 mL) was added ZnCl₂ (1 M, 103 mL) at 0 °C, and the mixture was stirred at 0 °C for 1 h. 4-Benzylsulfanyl-2-fluoro-aniline 39 (20 g, 85.7 mmol) in t-BuOH (200 mL), DCE (200 mL), and TEA (13.1 mL, 94.3 mmol) were added at 0 °C, and then the mixture was stirred at 25 °C for 14 h. On completion, the mixture was diluted with water (150 mL) and extracted with EA (200 mL × 3). The combined organic layers were dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure to give the residue. The residue was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (20.0 g, 56% yield) as a yellow solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 10.32 (s, 1H), 8.73 (s, 1H), 7.47–7.45 (m, 1H), 7.39 (d, J = 7.2 Hz, 2H), 7.30 (s, 3H), 7.27–7.22 (m, 1H), 7.17 (dd, J = 1.6, 8.4 Hz, 1H), 4.30 (s, 2H). LC-MS (ESI+) m/z 413.8 (M + H)⁺.

Step 3. (3S)-1-[2-(4-Benzylsulfanyl-2-fluoro-anilino)-5-(trifluoromethyl)­pyrimidin-4-yl]-3-methyl-piperidin-3-ol (41)

To a solution N-(4-benzylsulfanyl-2-fluoro-phenyl)-4-chloro-5-(trifluoromethyl)­pyrimidin-2-amine (22.0 g, 53.2 mmol) and (3S)-3-methylpiperidin-3-ol hydrochloride (12.1 g, 79.7 mmol) in MeCN (500 mL) was added TEA (22.2 mL, 159 mmol); then the mixture was stirred at 25 °C for 4 h. On completion, the mixture was concentrated under reduced pressure to give a residue, which was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (13.0 g, 49% yield) as a white solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 9.18 (s, 1H), 8.30 (s, 1H), 7.61 (t, J = 8.4 Hz, 1H), 7.40–7.19 (m, 6H), 7.12–7.10 (m, 1H), 4.43 (s, 1H), 4.25 (s, 2H), 3.62–3.49 (m, 1H), 3.38–3.33 (m, 1H), 3.31–3.23 (m, 2H), 1.76–1.72 (m, 1H), 1.60–1.51 (m, 2H), 1.47–1.38 (m, 1H), 1.02 (s, 3H). LC-MS (ESI+) m/z 493.7 (M + H)⁺.

Step 4. 3-Fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl] Amino]­benzenesulfonyl Chloride (42)

To a solution of (3S)-1-[2-(4-benzylsulfanyl-2-fluoro-anilino)-5-(trifluoromethyl)­pyrimidin-4-yl]-3-methyl-piperidin-3-ol (9 g, 18.3 mmol) in MeCN (90 mL), HOAc (9 mL) and H₂O (1 mL) was added NCS (7.32 g, 54.8 mmol), and then the mixture was stirred at 25 °C for 0.5 h. On completion, the mixture was diluted with water (200 mL) and extracted with DCM (100 mL × 4), and the combined organic layers were dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure to give a residue, which was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (7.53 g, 88% yield) as a yellow solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 10.34 (s, 1H), 8.59 (s, 1H), 7.80 (t, J = 8.0 Hz, 1H), 7.51–7.39 (m, 2H), 3.93 (d, J = 13.2 Hz, 1H), 3.67 (d, J = 12.8 Hz, 1H), 3.40 (d, J = 13.2 Hz, 1H), 3.30–3.25 (m, 1H), 1.83–1.78 (m, 1H), 1.64–1.56 (m, 2H), 1.54–1.47 (m, 1H), 1.08 (s, 3H). LC-MS (ESI⁺) m/z 468.9 (M + H)⁺.

Step 5. tert-Butyl (3S,4R)-3-fluoro-4-[[3-fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­phenyl]­sulfonylamino]­piperidine-1-carboxylate (43)

A solution of 3-fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl] amino]­benzenesulfonyl chloride (11.3 g, 24.1 mmol) and tert-butyl (3S,4R)-4-amino-3-fluoro-piperidine-1-carboxylate (10.5 g, 48.2 mmol) in DCM (120 mL) was stirred at 25 °C for 10 h. On completion, the mixture was filtered, the filter cake was diluted with water (100 mL) and extracted with DCM (50 mL × 3), and the combined organic layers and the filtrate were concentrated under reduced pressure. The crude product was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (14.9 g, 95% yield) as a yellow solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 9.59 (s, 1H), 8.39 (s, 1H), 8.16–8.06 (m, 2H), 7.70–7.61 (m, 2H), 4.63–4.45 (m, 2H), 4.09–4.07 (m, 1H), 3.83 (s, 1H), 3.67–3.64 (m, 1H), 3.54–3.38 (m, 3H), 3.33–3.25 (m, 2H), 1.85–1.73 (m, 1H), 1.63–1.44 (m, 5H), 1.35 (s, 9H), 1.28–1.22 (m, 1H), 1.04 (s, 3H). LC-MS (ESI⁺) m/z 651.2 (M + H)⁺.

Step 6. 3-Fluoro-N-[(3S,4R)-3-fluoro-4-piperidyl]-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­benzenesulfonamide trifluoroacetate (44)

A solution of tert-butyl (3S,4R)-3-fluoro-4-[[3-fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­phenyl]­sulfonylamino]­piperidine-1-carboxylate (13.0 g, 19.9 mmol) in DCM (50 mL) and TFA (20 mL) was stirred at 25 °C for 0.5 h. On completion, the mixture was concentrated under reduced pressure to give the title compound (12.8 g, 96% yield) as a yellow gum. LC-MS (ESI⁺) m/z 551.2 (M + H)⁺.

Synthesis of N-[(3S,4R)-1-[[1-[4-(2,6-Dioxo-3-piperidyl)-2-fluoro-phenyl]-4-piperidyl]­methyl]-3-fluoro-4-piperidyl]-3-fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­benzenesulfonamide (37)

Step 1. 2,6-Dibenzyloxy-3-(4-bromo-3-fluoro-phenyl)­pyridine (46)

To a solution of 1-bromo-2-fluoro-4-iodo-benzene 45 (39.7 g, 132 mmol) and 2,6-dibenzyloxy-3-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)­pyridine (50.0 g, 120 mmol) in dioxane (500 mL) and water (100 mL) were added K₂CO₃ (33.1 g, 239 mmol) and Pd­(dppf)­Cl₂ (8.77 g, 12.0 mmol), and the mixture was stirred at 100 °C for 10 h under a nitrogen atmosphere. On completion, the mixture was filtered, and the filtrate was diluted with water (1 L) and extracted with ethyl acetate (400 mL × 5). The combined organic layers were concentrated under reduced pressure to give the crude product, which was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (30.5 g, 55% yield) as a yellow solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 7.82 (d, J = 8.0 Hz, 1H), 7.69 (t, J = 8.0 Hz, 1H), 7.57 (dd, J = 1.6, 10.8 Hz, 1H), 7.44–7.31 (m, 11H), 6.57 (d, J = 8.0 Hz, 1H), 5.42 (s, 2H), 5.38 (s, 2H). LC-MS (ESI⁺) m/z 465.9 (M + H)⁺.

Step 2. 2,6-Dibenzyloxy-3-[4-[4-(dimethoxymethyl)-1-piperidyl]-3-fluoro-phenyl]­pyridine (47)

To a solution of 2,6-dibenzyloxy-3-(4-bromo-3-fluoro-phenyl)­pyridine 46 (20.0 g, 43.1 mmol) and 4-(dimethoxymethyl)­piperidine (13.7 g, 86.1 mmol) in dioxane (200 mL) were added Cs₂CO₃ (42.1 g, 129 mmol) and Pd-PEPPSI-IHeptCl (2.10 g, 2.15 mmol), and then the mixture was stirred at 100 °C for 10 h. On completion, the mixture was filtered, and the filtrate was concentrated under reduced pressure to give the residue. The residue was purified by column chromatography (petroleum ether/ethyl acetate) to give the title compound (19.6 g, 84% yield) as a yellow solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 7.74 (d, J = 8.0 Hz, 1H), 7.44–7.26 (m, 12H), 7.01 (t, J = 8.8 Hz, 1H), 6.53 (d, J = 8.0 Hz, 1H), 5.39 (d, J = 18.4 Hz, 4H), 4.11 (d, J = 6.4 Hz, 1H), 3.38 (d, J = 11.6 Hz, 2H), 3.28 (s, 6H), 2.62 (t, J = 11.6 Hz, 2H), 1.77–1.64 (m, 3H), 1.40–1.37 (m, 2H). LC-MS (ESI⁺) m/z 543.5 (M + H)⁺.

Step 3. 3-[4-[4-(Dimethoxymethyl)-1-piperidyl]-3-fluoro-phenyl]­piperidine-2,6-dione (48)

To a mixture of 10% Pd/C (25.0 g, 53.4 mmol) in THF (300 mL) was added 2,6-dibenzyloxy-3-[4-[4-(dimethoxymethyl)-1-piperidyl]-3-fluoro-phenyl]­pyridine (29.0 g, 53.4 mmol), and then the mixture was degassed and charged with hydrogen three times and then stirred at 40 °C under 40 psi hydrogen for 48 h. On completion, the mixture was filtered, and the filtrate was concentrated under reduced pressure to give the title compound (18.2 g, 93% yield) as a gray solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 10.82 (s, 1H), 7.04–6.90 (m, 3H), 4.11 (d, J = 6.4 Hz, 1H), 3.79 (dd, J = 4.8, 11.6 Hz, 1H), 3.32–3.31 (m, 2H), 3.27 (s, 6H), 2.66–2.55 (m, 3H), 2.47–2.46 (m, 1H), 2.24–2.13 (m, 1H), 2.02–1.93 (m, 1H), 1.75–1.64 (m, 3H), 1.43–1.34 (m, 2H). LC-MS (ESI⁺) m/z 387.2 (M + Na)⁺.

Step 4. 1-[4-(2,6-Dioxo-3-piperidyl)-2-fluoro-phenyl]­piperidine-4-carbaldehyde (49)

A solution of 3-[4-[4-(dimethoxymethyl)-1-piperidyl]-3-fluoro-phenyl]­piperidine-2,6-dione 48 (11.7 g, 32.1 mmol) in formic acid (50 mL) was stirred at 80 °C for 1 h. On completion, the mixture was concentrated under reduced pressure to give the title compound (11.0 g, 94% yield) as a yellow solid. LC-MS (ESI⁺) m/z 319.1 (M + H)⁺.

Step 5. N-[(3S,4R)-1-[[1-[4-(2,6-Dioxo-3-piperidyl)-2-fluoro-phenyl]-4-piperidyl]­methyl]-3-fluoro-4-piperidyl]-3-fluoro-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­benzenesulfonamide (37)

To a solution of 3-fluoro-N-[(3S,4R)-3-fluoro-4-piperidyl]-4-[[4-[(3S)-3-hydroxy-3-methyl-1-piperidyl]-5-(trifluoromethyl)­pyrimidin-2-yl]­amino]­benzenesulfonamide 44 (12.8 g, 19.3 mmol, TFA) in THF (150 mL) was added triethylamine (8 mL, 57.8 mmol) and HOAc (2.2 mL, 38.5 mmol) to adjust the pH to a range of 6 to 7; then, 1-[4-(2,6-dioxo-3-piperidyl)-2-fluoro-phenyl]­piperidine-4-carbaldehyde 49 (10.5 g, 28.9 mmol, FA) was added, and the mixture was stirred at 25 °C for 0.5 h. NaBH­(OAc)₃ (6.12 g, 28.9 mmol) was added, and the mixture was stirred at 25 °C for 0.5 h. On completion, the mixture was quenched with water (5 mL), diluted with NaHCO₃ (150 mL), and extracted with EA (80 mL × 5). The combined organic layers were dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure to give the residue. The residue was purified by prep-HPLC (column: Phenomenex Luna c18 250 mm × 100 mm × 10 μm; mobile phase: [water (NH₄HCO₃)-ACN]; gradient: 43–73% B over 24 min) to give the title compound (12.8 g, 78% yield) as a white solid. ¹H NMR (400 MHz, DMSO-d ₆) δ 10.80 (s, 1H), 9.56 (s, 1H), 8.39 (s, 1H), 8.10 (t, J = 8.0 Hz, 1H), 8.00 (s, 1H), 7.71–7.59 (m, 2H), 7.04–6.88 (m, 3H), 4.57–4.40 (m, 2H), 3.78 (dd, J = 4.8, 11.6 Hz, 1H), 3.65 (dd, J = 5.6, 8.8 Hz, 1H), 3.48–3.34 (m, 2H), 3.30–3.20 (m, 4H), 3.01–2.91 (m, 1H), 2.72–2.54 (m, 4H), 2.50–2.38 (m, 2H), 2.20–2.10 (m, 3H), 2.03–1.90 (m, 2H), 1.83–1.64 (m, 4H), 1.62–1.51 (m, 3H), 1.49–1.40 (m, 1H), 1.32–1.15 (m, 3H), 1.04 (s, 3H). LC-MS (ESI⁺) m/z 853.1 (M + H)⁺. ¹³C NMR (101 MHz, DMSO-d ₆ ) δ = 174.1, 173.4, 160.3, 159.6, 158.5, 155.9, 154.5, 153.4, 152.0, 139.3 (d, J = 32 Hz), 137.0 (d, J = 24 Hz), 133.1 (d, J = 32 Hz), 131.2 (d, J = 44 Hz), 124.8, 124.3, 122.5, 119.1, 116.0 (d, J = 84 Hz), 113.8 (d, J = 88 Hz), 100.1, 99.8, 89.3, 87.5, 67.6, 63.0, 58.3, 55.3 (d, J = 80 Hz), 52.4 (d, J = 76 Hz), 51.3, 50.6, 47.5, 46.4, 37.0, 32.7, 31.5, 30.4 (d, J = 52 Hz), 26.8, 25.6, 20.9. ¹⁹F NMR (377 MHz, DMSO-d₆) δ −54.3, −120.2, −122.9, −198.6; MS (ESI): m/z 853.2 (M + H) ⁺.

Glossary

Abbreviations Used

ABT

1-aminobenzotriazole

ADME

absorption, distribution, metabolism, and excretion

ATP

adenosine triphosphate

AUC

area under the curve

BDC

bile duct-cannulated

BID

bis in die (twice a day)

CRBN

cereblon

CYP

cytochrome P450

DCE

dichloroethane

DCM

dichloromethane

DDB1

DNA damage-binding protein 1

DIEA

diethylamine

DMF

N,N,-dimethylformamide

DMSO

dimethyl sulfoxide

ESP

electrostatic potential

FA

formic acid

FaSSIF

fasted state simulated intestinal fluid

FDA

U.S. Food and Drug Administration

GI

gastrointestinal

GPCR

G-protein-coupled receptor

HCC

hepatocellular carcinoma

HER2

human epidermal growth factor receptor 2

hERG

human ether-a-go-go-related gene

HLM

human liver microsomes

IC50

half-maximal inhibition concentration

IV

intravenous

HR+

hormone receptor positive

LON

Lon protease-like

LC-MS

liquid chromatography–mass spectrometry

MDCK

Madin–Darby canine kidney

MDR1

multidrug resistance 1

MeOH

methanol

MW

molecular weight

NADPH

nicotinamide adenine dinucleotide phosphate

NCS

N-chlorosuccinimide

NHP

nonhuman primate

PBMC

peripheral blood mononuclear cell

PD

pharmacodynamic

PDB

protein database

PBS

phosphate-buffered saline

P-gp

P-glycoprotein

PK

pharmacokinetic

PO

per os (by mouth)

PSA

polar surface area

QD

quaque die (once a day)

Rb

retinoblastoma protein

RLM

rat liver microsomes

RNA

ribonucleic acid

RT

room temperature

SAR

structure–activity relationship

SFC

supercritical fluid chromatography

SMi

small molecule inhibitor

TBD

thalidomide binding domain

Tbk1/Ikki

binding domain

TDI

time-dependent inhibition

TEA

triethylamine

TFA

trifluoroacetic acid

THF

tetrahydrofuran

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.5c01160.

• Description of biochemical and cellular assays, proteomics, and in vivo pharmacology protocols, in vitro ADME and computational chemistry methods, kinomescan data for compounds 3, 7, 29, 30, 31, and 32, in vivo diastereomer interconversion summary for compound 37, crystallographic and cryo-EM methods, chemical synthesis of compounds 3–37, ¹H NMR, ¹⁹F NMR, and ¹³C NMR spectra of compound 37, HPLC traces of compounds 3, 5, 10, 11, 19, 20, 28, 32, 34, 36, and 37 (PDF)

• Molecular formula strings (CSV)

All authors have given approval to the final version of the manuscript.

The authors declare the following competing financial interest(s): All authors are or were employees of Kymera Therapeutics and may hold company stocks or stock options with Kymera Therapeutics.