Property-Based Optimization of Cereblon-Based Molecular Glue Degraders

Abstract

Molecular glue degraders that bind the E3 ligase adaptor cereblon are clinically precedented drugs used to treat cancer. Highly potent next-generation degraders are in development that target difficult-to-drug neosubstrates, such as transcription factors, for several different indications. This Microperspective covers the medicinal chemistry strategies being used to advance the property-based optimization of cereblon-based molecular glue degraders through consideration of their physicochemical characteristics.

The immunomodulatory imide drugs (IMiDs) thalidomide, lenalidomide and pomalidomide (Figure ) are molecular glues that bind cereblon (CRBN, a substrate receptor of the CRL4^CRBN E3 ligase complex) and remodel its surface to mediate interactions with neosubstrates. The induced proximity of the neosubstrate with CRL4^CRBN triggers its polyubiquitination and proteasomal degradation. Canonical neosubstrates possess a structural motif called a G-loop degron consisting of a β-hairpin with a critical glycine residue that mediates ternary complex formation – other amino acids at this position would clash and prevent binding to the neofunctionalized glue/CRBN surface. Neosubstrates are often difficult-to-drug targets, such as zinc-finger transcription factors IKZF1 (Ikaros) and IKZF3 (Aiolos), the depletion of which underpins the efficacy of the IMiDs in treating multiple myeloma. The very polar glutarimide pharmacophore present within the IMiDs binds a tritryptophan cage in the thalidomide binding domain (TBD) at the CRBN surface. Chemically tethering glutarimide analogs to ligands for proteins-of-interest (POIs) creates heterobifunctional proteolysis targeting chimeras (PROTACs) that provides a modular approach to mediate target degradation through induced proximity pharmacology.

1.

Physicochemical properties of IKZF1/3 degraders. Top row: first generation IMiDs. Second row: next-generation CELMoDs and cemsidomide. Electronic structure calculations for thalidomide, lenalidomide, and pomalidomide, and the highest occupied molecular orbitals are shown as isosurfaces. High electron density of the amino and carbonyl motifs of lenalidomide is circled. The increase in size and lipophilicity of the next-generation degraders compared to the IMiDs is noticeable.

CRBN-based molecular glues have generated considerable excitement in the drug discovery community because of the potential for targeted protein degradation (TPD) to expand the druggable proteome. Several articles and reviews have described the pharmacology of CRBN-based degraders and the pharmacokinetic (PK) features of PROTACs, but relatively less attention has been given to molecular glue physicochemistry and property-based optimization, which is the focus of this Microperspective.

The IMiDs were developed as racemic mixtures due to the rapid racemization of the chiral center at the 3-position of the glutarimide. This issue can be addressed through replacement of the exchangeable hydrogen atom with deuterium. As a result, the deuterated analogs are stabilized, and the enantiomers can be separated and assayed individually. The technique was applied to avadomide (Figure ), another IKZF1/3 degrader that was in trials for treating multiple myeloma and lymphoma, and deuterated (S)-avadomide (or SP-3164) is now in clinical development. Unexpectedly, 15 mg/kg SP-3164 dosed orally in mice had considerably higher C _max and AUC than the detected (S)-isomer of a 30 mg/kg avadomide dose, illustrating potential pharmacokinetic (PK) advantages of deuteration.

Recently, we assessed the physicochemistry of the IMiDs and how their distinct properties affect in vivo PK and distribution. It might be expected that the presence of an additional carbonyl motif in pomalidomide compared to lenalidomide would increase polarity, but pomalidomide is approximately 10-times more lipophilic (Figure ). There are two important molecular features of pomalidomide that significantly reduce its polarity compared to lenalidomide. First, the sp² hybridization of the carbon atom of the additional carbonyl group enables extended π-delocalization across the phthalimide ring system, increasing lipophilicity. Conversely, the isolated benzamide motif of lenalidomide increases polarity, an effect reflected in the electronic structures (Figure ). Second, the pomalidomide amino group undergoes an intramolecular hydrogen bond with the additional carbonyl oxygen, but this is not possible for lenalidomide where it is unshielded. Another interesting physicochemical feature of pomalidomide is that the push–pull π electronics impart on the drug (and related analogs) intrinsic fluorescent properties which have even been used clinically as a sensitive marker of exposure in patient plasma. ,

As a consequence of its high polarity, lenalidomide is renally cleared, unlike pomalidomide which is oxidatively metabolized, and so patients with impaired renal function require dose adjustments to avoid excessive lenalidomide exposure and potential toxicities. Avadomide also undergoes significant renal clearance (43% of the administered dose in patients) likely due to the polarity of its amino quinazolinone core and so must be dose adjusted in renally impaired patients too. Additionally, due to its hydrophilicity, lenalidomide has low permeability that translates to impaired CNS penetration and the IMiD thus lacks the sedation and neuropathy side effects seen for thalidomide.

We reasoned that these relative properties of lenalidomide and pomalidomide may translate more broadly to other molecular glue degrader analogs. Based on pairwise comparisons of phthalimide and isoindolinone degraders in our lab we find that phthalimides are 6-fold more lipophilic than their isoindolinone congeners, which translates to a 4-fold improvement in permeability (Caco-2 flux) and a 2–3-fold increase in metabolic rate (HLM CL_int). Therefore, to improve the permeability of an isoindolinone degrader, the scaffold could be switched to a phthalimide. Conversely, replacing the phthalimide core of a degrader with an isoindolinone may improve metabolic stability. These broad principles may not always hold of course, but we have been able to successfully apply the learnings to the property-based optimization of degraders.

Both phthalimide and isoindolinone scaffolds have been used to develop next-generation CRBN E3 ligase modulatory compounds (CELMoDs) that degrade IKZF1/3 much more effectively than the original IMiDs. These agents, iberdomide (CC-220), , golcadomide (CC-99282) and mezigdomide (CC-92480) are currently in clinical trials for the treatment of multiple myelomas and lymphomas (Figure ), and their highly potent and efficient degradation activity helps address resistance that occurs during IMiD treatment through CRBN downregulation. The rate of racemization of these CELMoDs is slow compared to the IMiDs, enabling enantiomer separation and testing, and their active (S)-enantiomers are being advanced in trials.

A recent report described the use of efficiency metrics tailored to degraders that take into account both potency (DC₅₀) and depth (D _max) of degradation, and a retrospective analysis of the development of golcadomide was provided. Lipophilic efficiency (LipE), which normalizes potency for Log D, has become a useful metric used by medicinal chemists through hit-to-lead and lead optimization (LO): ,

$$\mathrm{LipE}={\mathrm{pIC}}_{50}-\mathrm{Log}D$$

Essentially, the higher the LipE number the better because simply improving potency by making the molecule more lipophilic (to drive hydrophobic interactions for example) will usually be detrimental to drug-like properties, including metabolic stability, selectivity and solubility. In the case of degraders, a new value called composite degradation potency (CDP) was defined:

$$\mathrm{CDP}={(\mathrm{p}\mathrm{D}\mathrm{C}}_{50}\times D_{\max })/\mathrm{S}\mathrm{F}$$

SF is a scaling factor, set to 100 in this case. Lipophilic efficiency for degraders (LipED) then becomes

$$\mathrm{LipED}=\mathrm{CDP}-\mathrm{Log}D$$

Lenalidomide and pomalidomide are so small and polar (Figure ) it would be somewhat inevitable that more effective IKZF1/3 degraders would be larger and more lipophilic to optimize ternary complex formation. Indeed, the LogD of iberdomide is 1.9,⁷ and the cLogPs of golcadomide and mezigdomide are 1.4 and 3.0 respectively (Figure ). Over the course of the golcadomide LO program, LipED clearly improved significantly, from <2 to ∼5, and early in the campaign, highly efficient degraders were identified (high CDP) but with high LogD and thus poor properties, so were not progressed. LipED will no doubt help guide the medicinal chemistry optimization of degraders in the future. We have previously utilized an alternative metric to describe the effectiveness of degraders, the area under the curve (AUC) of the plot of protein level versus drug concentration, and so presumably a similar efficiency metric could be generated using degradation AUC.

Cemsidomide (CFT7455, Figure ) was the result of a structure-based design strategy to increase the degradation efficiency of the IMiDs. Merging benzoimidazolone and pomalidomide scaffolds yielded a new benzoisoindolinone core. A tail group similar to that in iberdomide was incorporated into the scaffold to provide cemsidomide that had subnanomolar IKZF1 DC₅₀ and 48% bioavailability in mice. Cemsidomide possessed dose proportional PK when given at submilligram doses to humans, with a half-life of 2 days. Noticeably, the CELMoDs and cemsidomide all possess either morpholine or piperazine groups which are privileged motifs incorporated into medicinal chemistry design strategies to improve drug-like properties, solubility in particular. ,

CK1α is another neosubstrate of lenalidomide, and its degradation provides an effective treatment of myelodysplastic syndrome (MDS) patients with deletion of chromosome 5q (del­(5q)). With the objective of developing a selective CK1α degrader, a recent report described the potency and property-based optimization of cereblon-based molecular glues. Compound 13 was found to be a potent CK1α degrader, but the benzamide was degraded in plasma, presumably through the action of amidases (Figure ). Cyclization of the amide into a heterocyclic bioisostere furnished SJ3149 (Figure ), a potent and highly selective degrader with improved metabolic stability. SJ3149 also possessed excellent stability in hepatocytes, but a high clearance and rather modest oral bioavailability in mice of 9%, suggesting an in vitro/in vivo disconnect that may be due to alternative clearance mechanisms.

2.

Potencies and pharmacokinetic properties of CRBN-based molecular glue degraders targeting CK1α, IKZF2, GSPT1, and VAV1 neosubstrates.

Functional genomics, proteomics and structural bioinformatics studies of the IMiDs and related molecules have revealed many additional CRBN neosubstrates of therapeutic relevance, that have become the targets of drug discovery campaigns. IKZF2 is a target linked to a stable Treg phenotype in the tumor microenvironment, suggesting a degrader of the zinc-finger transcription factor may have immunotherapeutic potential. The G-loop degron of IKZF2 contains a key histidine residue, compared to a glutamine in IKZF1/3, which hinders IMiD-mediated degradation. A more flexible glutarimide-containing anilinomaleimide scaffold was developed previously that enabled ternary complex formation with IKZF2 that established proof-of-principle pharmacological degradation of the target, although the compounds also degraded IKZF1/3. In a subsequent report, elaboration of the isoindolinone scaffold provided DKY709 (Figure ) that was also shown to mediate degradation of IKZF2 selectively and was advanced to clinical trials for solid tumor indications. The piperidine motif was incorporated to improve solubility, and the benzyl group enhanced permeability (over the derivative possessing a free NH piperidine). DKY709 possessed rapid absorption, moderate clearance and good oral bioavailability in both mice and monkeys. A related isoindolinone selective degrader of IKZF2 was reported recently, PVTX-405 (Figure ), bearing a spiro-benzylpiperidine, that possessed superior efficacy to DKY709 and good PK. PVTX-405 also had weaker hERG activity than DKY709 that might derive from a lower pK _a of the basic piperidine nitrogen and hERG binding SAR related to the addition of the methyl pyrazole unit.

Phenotypic screening of libraries of thalidomide analogs in cancer cell lines by several groups led to the identification of the translation termination factor GSPT1 as a degradable target and its pharmacological downregulation mediated potent antiproliferative effects. ,− CC-885 was identified as the prototype GSPT1 degrader, and others such as SJ6986 and MRT-2359 were subsequently developed and profiled in PK experiments, the latter being progressed to clinical trials (Figure ). GSPT1 degraders possess lipophilic termini, which can be explained by interactions with hydrophobic residues in this region of the GSPT1 interface. , Equally, hydrogen bonding potential in the ‘linker’ enables interactions within the ternary complex, which balances the overall physicochemistry of the molecules, thus avoiding high compound lipophilicity. Consequently, SJ6986 and MRT-2359 clearly illustrate that these orally bioavailable GSPT1 CRBN-based molecular glue degraders possess drug-like properties.

Another recently described neosubstrate is the guanine nucleotide exchange factor VAV1, a target linked to autoimmune and chronic inflammatory diseases. MRT-6160 (Figure ), which utilizes a slightly different phenyl glutarimide scaffold to reported CELMoDs, is a highly efficient and selective VAV1 degrader that has demonstrated good oral bioavailability and long half-life in humans. MRT-6160 may be suitable for neurological autoimmune indications due to its potential for high CNS exposure resulting from its excellent physicochemical properties (MW 392, Log D 1.5, 1 hydrogen bond donor, 3 hydrogen bond acceptors, 3 rotatable bonds, cf. median values for CNS drug candidates: MW 360, cLogD 2.2, 1 hydrogen bond donor, 4.5 rotatable bonds).

Other CRBN-based molecular glue degraders have entered clinical trials that degrade new neosubstrates, including WIZ, that enables activation of fetal hemoglobin for the treatment of sickle cell disease, and WEE1 for solid tumors. Although the structures of these degraders have yet to be published, they must of course possess optimized physicochemical properties to achieve the desired PKPD profiles necessary for advancement to human experiments.

Recently, we described the application of a new technique called covalent anchoring using sulfonyl exchange (CASE) to explore novel structure–activity and structure–property relationships of new CRBN scaffolds. The incorporation of a sulfonyl fluoride warhead into the 6-position of the isoindolinone molecular glue degrader EM12 to deliver EM12-SF (Figure ), mediated a significant increase in binding potency through covalent engagement of His353 in the sensor loop of CRBN. The electrophile also rescued the potency of isoindoline EM364, (EM364-SF, CRBN IC₅₀ = 71 nM) suggesting optimal equilibrium binding motifs may “buy-back” potency, enabling removal of the warhead. Indeed, appending the IKZF1/3 optimized tail of iberdomide yielded CPD-2743 which retained the CRBN binding affinity and degradation potency of EM12 itself. Importantly, CPD-2743 lacked SALL4 degradation, a target linked to the teratogenicity of EM12 and the IMiDs. , Unsurprisingly, the removal of the polar carbonyl motif slightly increased lipophilicity compared to iberdomide which translated to a considerable improvement in permeability, with no evidence of efflux. Therefore, CASE exemplifies a new medicinal chemistry technique that facilitates the exploration of new chemical property space and could be used more broadly in drug discovery.

3.

Covalent anchoring using sulfonyl exchange (CASE) leading to the development of isoindoline CPD-2743.

Interestingly, when we switched the electrophile to a fluorosulfate, the compound EM12-FS (Figure ) still site-specifically modified His353, but this time CRL4^CRBN was synthetically neofunctionalized to degrade a novel noncanonical neosubstrate called NTAQ1, which performs the first step in the Arg/N-degron pathway. The fluorosulfate warhead is intrinsically less reactive than sulfonyl fluoride due to the additional oxygen donating electron density into the sulfur atom, reducing its electrophilicity. As a result, EM12-FS has considerably higher plasma stability than EM12-SF, which is a liability for sulfonyl exchange warheads more generally (Figure ). Sulfonyl triazoles EM12-SO₂Tr1 and EM12-SO₂Tr2 were developed to efficiently label His353, but these were also hydrolyzed readily in plasma (Figure ). We decided to explore the labeling efficiency and plasma stability of a series of new sulfonyl diazole warheads attached to the EM12 scaffold that were designed to attenuate the intrinsic electrophilicity of the warhead. By simply plotting extent of CRBN labeling versus plasma half-life we were able to identify a broad correlation, as might be expected, but the sulfonyl imidazole EM12-SO₂Im (Figure ) retained higher plasma stability than expected based on its efficient labeling of CRBN. Our studies have revealed sulfonyl imidazole and fluorosulfate warheads are clearly suitable for drug development due to their latent site-specific reactivity with nucleophilic amino acid side chains while retaining excellent chemical and metabolic stability. These findings could be useful for the property-based optimization of sulfonyl exchange-bearing covalent modulators of CRBN and other proteins in the future.

4.

Potency and plasma stability of a series of covalent CRBN binders.

Targeted protein degraders, and covalent E3 modulators in particular, possess pharmacodynamic advantages over traditional inhibitors since only a small fraction of CRBN needs to be neofunctionalized to affect efficient catalytic degradation of the target. Consequently, PKPD mismatches potentially allow for intermittent dosing regimens of relatively high clearance degraders, which may lower the AUC needed to drive efficacy, and thus enhance therapeutic indices. However, high unbound C _max-driven liabilities such as hERG toxicity would need to be understood for such strategies, especially for lipophilic and basic degraders, which appears to be a common physicochemical feature of clinical candidates, as described above.

CRBN-based molecular glue degraders continue to be a considerable focus of drug discovery research because the modality enables difficult-to-drug targets to be addressed for the first time. As for all small molecule drugs, careful consideration of physicochemical properties will be necessary to ensure balanced PKPD and safety profiles are achieved. Key learnings from the development of IMiDs and degraders currently in clinical trials will clearly facilitate medicinal chemistry optimization of the next-generation of CRBN modulators.

NOTE: All cereblon modulators should be treated as teratogenicity risks.

Glossary

Abbreviations

AUC

area under the curve

CASE

covalent anchoring using sulfonyl exchange

CDP

composite degradation potency

CELMoD

CRBN E3 ligase modulatory compound

Chrom LogD

chromatographic assessment of the logarithm of the distribution coefficient at pH 7.4 (Log D)

C _max,

highest concentration of a compound dosed in vivo

CNS

central nervous system

CRBN

cereblon

DC₅₀

concentration of drug required to reduce protein signal by 50%

D _max

maximum amount of protein degraded

F

oral bioavailability

HLM

human liver microsomes

IMiD

immunomodulatory drug

LipE

lipophilic efficiency

MW

molecular weight

LipED

lipophilic efficiency for degraders

PD

pharmacodynamics

PK

pharmacokinetics

PROTAC

proteolysis-targeting chimera

SAR

structure–activity relationship.

The author declares the following competing financial interest(s): LHJ is a co-founder of Anvia, serves as an advisor to ONO Pharmaceuticals, Merck KGaA, Matchpoint Therapeutics, Lime Therapeutics, Immunovalence, Belharra Therapeutics, and holds equity in Rapafusyn Pharmaceuticals and Hyku Biosciences.