The Cyclimids: Degron-inspired cereblon binders for targeted protein degradation

Abstract

Cereblon (CRBN) is an E3 ligase substrate adapter widely exploited for targeted protein degradation (TPD) strategies. However, achieving efficient and selective target degradation is a preeminent challenge with ligands that engage CRBN. Here, we report that the cyclimids, ligands derived from the C-terminal cyclic imide degrons of CRBN, exhibit distinct modes of interaction with CRBN and offer a facile approach for developing potent and selective bifunctional degraders. Quantitative TR-FRET-based characterization of 60 cyclimid degraders in binary and ternary complexes across different substrates revealed that ternary complex binding affinities correlated strongly with cellular degradation efficiency. Our studies establish the unique properties of the cyclimids as versatile warheads in TPD and a systematic biochemical approach for quantifying ternary complex formation to predict their cellular degradation activity, which together will accelerate the development of ligands that engage CRBN.

Main Text:

Targeted protein degradation (TPD) induced by small molecules is emerging as an attractive strategy for eliminating disease-relevant proteins from cells.^1–3 Small molecule degraders, such as molecular glues⁴ and proteolysis targeting chimeras (PROTACs),^1–3,5 can chemically induce the recruitment of target proteins to an E3 ligase complex and promote their subsequent ubiquitination and proteasomal degradation. Among the >600 E3 ligases encoded by the human genome, cereblon (CRBN), a substrate adapter in the CRL4^CRBN E3 ligase complex,^6–8 is one of the best-characterized substrate adapters for TPD applications. CRBN was originally discovered in association with intellectual disability⁹ and was later identified as the primary target of thalidomide.⁶ Thalidomide and its derivatives, lenalidomide and pomalidomide, are members of a class of pharmaceutical agents that have unique and pleiotropic clinical properties against multiple myeloma,^10,11 del(5q) myelodysplastic syndrome,¹² and other hemopoietic malignancies,¹³ as well as anti-inflammatory effects in the treatment of inflammatory diseases including erythema nodosum leprosum (ENL), rheumatoid arthritis, HIV-associated aphthous ulcers, and tuberculous meningitis.¹⁴ These effects are partially realized through binding to CRBN, thereby altering its substrate selection (e.g., IKZF1, IKZF3, CK1α, or GSPT1) for ubiquitination and degradation.^10–13 Engagement of CRBN using heterobifunctional degraders (e.g., PROTACs) by linking a thalidomide derivative to a ligand for a target protein has also spurred the development of many PROTACs for degradation of a myriad of target proteins and is a promising new modality in both chemical biology and drug discovery.^15–18

Despite the immense promise of degrader modalities, challenges in the development of efficient degrader molecules include (i) achieving high target selectivity, (ii) identifying and parsing through the diverse modes of target–degrader–E3 ligase ternary complexes, (iii) engineering precise control of the ternary complex orientation, and (iv) optimizing cellular degradation activity and characterizing these molecules in biochemical and cellular assays.¹⁷ Heterobifunctional degraders consist of two ligands, one for interacting with a target protein and the other for recruiting an E3 ligase, tethered by a linker. The design of heterobifunctional degraders through linker screening (termed “linkerology”)^17,19 offers one approach to tackle challenges in degrader discovery and optimization. The linker controls critical aspects of degrader design, such as the solution conformation of the degrader molecule,²⁰ as well as ternary complex engagement²¹ and orientation,²² thereby influencing target selectivity and degradation efficiency. As the formation of a productive ternary complex is crucial for efficient degradation events, various biophysical approaches have become powerful tools to study PROTAC-mediated ternary complexes. These approaches include fluorescence polarization (FP),²³ isothermal titration calorimetry (ITC),²⁴ surface plasmon resonance (SPR),²⁵ AlphaScreen/AlphaLISA,^24,26,27 TR-FRET,²² and other cell-based assays (e.g., separation of phase-based protein interaction reporter [SPPIER]²⁸ and bioluminescence resonance energy transfer [BRET]^29–31). Drawing on the assay principles of these pivotal prior works, a quantitative—and ideally high-throughput—method to systematically correlate ternary complex formation with the degradation efficacy of a heterobifunctional degrader would greatly expedite the development of these novel modalities.

A mechanism to systematically evaluate ternary complex formation by variation of the CRBN warhead, rather than the linker, may provide an orthogonal and tunable strategy to generate productive ternary complexes for TPD. Indeed, minor structural changes on the CRBN ligand can dramatically alter biological activities in cells,²² and the engagement of CRBN tends to be strongest with a 6-membered glutarimide in thalidomide derivatives and is significantly diminished when substituted with a 7-membered³² or 5-membered³³ cyclic imide. Efforts to identify new ligands for CRBN have revealed phenyl-glutarimides,³⁴ amino-glutarimides,³³ and dihydrouracils^35–37 as alternate 6-membered rings that can behave as ligands for CRBN. Recently, we discovered that C-terminal cyclic imides are previously overlooked post-translational modifications that arise from intramolecular cyclization of glutamine or asparagine, which afford a degron for the thalidomide-binding domain of CRBN (Fig. 1a).³⁸ CRBN flexibly recognizes a range of C-terminal cyclic imide dipeptides when conjugated to JQ1 to afford functional degraders for BRD4 (bromodomain-containing protein 4)³⁹ with intriguing differences in degradation efficiency that may be attributed to one of several factors, including differential engagement of CRBN, positive or negative ternary complex formation, or cellular permeability. Notably, both 6-membered cyclic imides and 5-membered cyclic imides served as effective degrons for CRBN, implying that the properties of these degron motifs may be in stark contrast to those of thalidomide derivatives when embedded in a bifunctional degrader and thus render new chemical space accessible. In this study, we designed and evaluated several classes of PROTACs bearing cyclic imide ligands, termed cyclimids, against a series of cellular targets. Employing a quantitative, high-throughput TR-FRET-based ternary complex assay platform, our results provide evidence that the cyclimids induce distinct modes of ternary complex formation between CRBN and target proteins in a linker-independent manner. The cyclimids show higher on-target selectivity via minimization of off-target CRBN-mediated molecular glue activity and maximization of ternary complex potency. Overall, our results demonstrate the potential for the cyclimids to yield potent and efficient PROTACs through systematic tuning of target–CRBN ternary complexes.

Fig. 1. The Cyclimids, degron-inspired CRBN ligands, exhibit distinct and diverse binding affinities against CRBN compared to thalidomide derivatives.

(a) CRBN recognizes the C-terminal cyclic imide degrons, which are mimicked by thalidomide, lenalidomide, and pomalidomide. Structures of thalidomide derivatives and the C-terminal cyclic imide degrons (XcQ and XcN, collectively called the cyclimids). The Cyclimids, a class of CRBN ligands inspired by the C-terminal cyclic imide degron of CRBN, may serve as unique CRBN warheads to develop heterobifunctional degraders. (b) Structures of thalidomide-based degraders (dBET6 and dBET6-cN) and cyclimid-based degraders (JQ1-XcQ and JQ1-XcN) for functional engagement of CRBN and BRD4 degradation in cells. (c) Schematic of the TR-FRET ligand-displacement assay for His₆-CRBN/DDB1. (d) Dose-titration of the indicated degraders in TR-FRET ligand displacement assays with His₆-CRBN/DDB1 and the determined equilibrium dissociation constants (K_D) of the indicated compounds against the His₆-CRBN/DDB1 complex. (e) Comparison between the pK_D values of cN/cQ cyclimid- and thalidomide-based degraders against CRBN/DDB1. In contrast to IMiD-based degraders, cN-cyclimid-based degraders generally bind to CRBN/DDB1 more tightly than their cQ counterparts. Data in (d) and (e) are presented as mean ± SD (n = 3 technical replicates). (f) Western blot of BRD4 after treatment of HEK293T cells for 4 h with dBET6, dBET6-cN, JQ1-FcQ, or JQ1-FcN over a 100 nM–10 μM dose-response range. Western blot data are representative of three independent replicates. For uncropped western blot images, see Figure S2 in the Supplementary Information.

The cyclimids exhibit distinct and diverse binding affinities against CRBN

To characterize the cyclimids as a unique class of CRBN ligands for TPD applications, we first constructed a library of 40 cyclimid-based degraders against BRD4. The cyclimid-based BRD4 degrader library is composed of a series of twenty 6-membered cyclic glutarimide (cQ)-bearing cyclimid BRD4 degraders (JQ1-XcQ),³⁸ and an additional set of twenty cyclimid degraders consisting of JQ1 linked to a dipeptide with a 5-membered cyclic aspartimide (cN) (JQ1-XcN; Fig. 1b). We installed all 20 canonical amino acids at the N–1 position (X) to the cyclic imide motif (cQ or cN) when constructing the library of cyclimids (XcQ or XcN). For direct comparison, we used the thalidomide-based degrader dBET6^15,16,22 and prepared its cN counterpart with no amino acid at the N–1 position (dBET6-cN; Fig. 1b). The same aliphatic linker was used across all cyclimid degraders to control for its contribution to degrader properties.

We first assessed the binding of the cyclimid-based degraders to CRBN by a time-resolved Förster resonance energy transfer (TR-FRET) ligand displacement assay using Thal-FITC as a tracer molecule, specific for the thalidomide-binding domain of CRBN (Fig. 1c, Supplementary Fig. 1a). Saturation binding of Thal-FITC against full-length His₆-CRBN/DDB1 detected by a CoraFluor-1-labeled anti-His₆ antibody⁴⁰ yielded an equilibrium dissociation constant (K_D) of 117 nM for His₆-CRBN/DDB1, which is comparable to the K_D values of thalidomide measured by fluorescence polarization (Supplementary Fig. 1b).⁴¹ We next determined the binding affinities of the 40 cyclimid-based degraders, along with the thalidomide-based degraders dBET6 and dBET6-cN, against CRBN/DDB1 by competition binding (Fig. 1d–e). The cyclimid degraders, JQ1-FcQ and JQ1-FcN, and dBET6 demonstrated similar levels of binding to CRBN (K_D values < 100 nM), whereas dBET6-cN exhibited more limited engagement of CRBN (K_D = 764 nM, Fig. 1d, Supplementary Table 1). All 40 cyclimid degraders functionally engaged CRBN and displayed K_D values ranging from 8 nM–1 μM, depending on the amino acid installed at the N–1 position (Fig. 1e, Supplementary Table 1). The degraders JQ1-cQ and JQ1-cN, which lack the amino acid at the N–1 position, also engaged CRBN at a moderate level independent of the ring size of the cyclic imide (Fig. 1e, Supplementary Fig. 1c). These data suggest that the additional amino acid at the N–1 position can tune binding affinities against CRBN in the cyclimid degraders. Notably, the cyclimid degraders with 5-membered cN generally exhibit higher affinities to CRBN than their 6-membered cQ counterparts (Fig. 1e).

The selected monofunctional cyclimids were also evaluated as ligands for CRBN, independent of the linker to JQ1 (Supplementary Fig. 1d–f). We observed a similar trend in which the binding affinity towards CRBN with cN-bearing dipeptides was generally superior to that of its cQ counterpart (Supplementary Fig. 1d). In contrast, pomalidomide (POM) exhibited much higher binding affinity than the 5-membered analog POM-cN³³ (~500-fold difference in K_D, Supplementary Fig. 1f). The inferior affinity of POM-cN compared to the parent cQ counterpart POM may be related to its inflexible projection of the isoindoline-1,3-dione backbone caused by the smaller ring size, which prevents favorable interactions with nearby residues at the thalidomide-binding domain of CRBN (Supplementary Fig. 1g–h). To further investigate the functional difference between 5-membered cN embedded in thalidomide derivatives or cyclimid scaffolds, the BRD4 degradation activity of dBET6-cN and JQ1-FcN, along with 6-membered cQ-based degraders dBET6 and JQ1-FcQ, were assessed in HEK293T cells (Fig. 1f). While dBET6 and JQ1-FcQ exhibited similar levels of BRD4 degradation, JQ1-FcN demonstrated efficient degradation of BRD4 at a concentration of 100 nM, whereas dBET6-cN was ineffective even at a concentration of 10 μM. The degradation efficiency directly correlates with the difference in binding affinity of cN against CRBN when incorporated within either thalidomide-based or cyclimid degraders.

Ternary complexes induced by cyclimid degraders are tunable and distinct from IMiD-based degraders

In addition to the distinct and diverse binding affinities of the cyclimids toward CRBN, we hypothesized that these cyclimid degraders may induce differential ternary complex modes relative to the thalidomide-based degraders and, therefore, screening the amino acid at the N–1 position, installed next to the cyclic imide motif, may afford rapid access to productive ternary complexes. We thus sought to evaluate the ternary complex formation between CRBN and each bromodomain of the target protein BRD4 (BD1 and BD2) induced by our complete panel of cyclimid or thalidomide derivative-based degraders. To circumvent the difficulty of extracting the absolute values of binding affinities for a three-body binding system,⁴² prior works introduced the concept of rendering a complex three-component system comprehensible by using a saturating excess of the target protein. This concept has been demonstrated through multiple biophysical assays, such as FP, ITC, SPR, and biolayer interferometry (BLI).^{22–25,29,30,42} We leveraged these concepts in the design of our TR-FRET assay platform to enhance throughput, quantification, material efficiency, and sensitivity, all to better facilitate the biochemical assessment of PROTAC-mediated ternary complexes.

To this end, we sought to modify our Thal-FITC/His₆-CRBN/DDB1 TR-FRET ligand displacement platform to enable quantitative ternary complex evaluation. Specifically, when [BRD4] ≫ K_{D, BRD4} and K_{D, CRBN} ≫ [CRBN], our TR-FRET ligand displacement platform can instead provide the equilibrium dissociation constant of the degrader–target binary complex for CRBN (K_{D, CRBN}[ternary], Fig. 2a). Saturation of the BRD4-targeting ligand (i.e., JQ1) can be achieved by using a high concentration of recombinant BD1 or BD2 in the assay solution, resulting in the formation of a binary complex which can compete with Thal-FITC in the TR-FRET ligand displacement assay against CRBN. To achieve near-saturation, the minimum concentration of GST-tagged BRD4 bromodomains BD1 or BD2 was first determined by the TR-FRET ligand displacement assays using JQ1-FITC⁴³ as a tracer (Supplementary Fig. 2a–d, Supplementary Table 1). All BRD4-targeting cyclimid degraders possess K_D values for BRD4(BD1) and BRD4(BD2) in the 20–150 nM range. As such, at a concentration of 1 μM BRD4(BD1) or BRD4(BD2) (approximately ~10–50-fold greater than the K_D value), this will result in essentially quantitative formation of cyclimid–BD1/2 binary complexes (corresponding to 92–98% formation of binary complex⁴⁴). Our TR-FRET assay methodology employs a simpler 2-site binding model with higher fidelity for regression analyses and is designed to be within the “binding regimen”⁴⁴ based on the K_D of the tracer Thal-FITC, resolving the limitations of previously used ligand displacement assays^22,23,30 to allow for the quantitative thermodynamic and kinetic characterization of ternary complexes. Furthermore, the cooperativity of binding can be directly calculated through these assays, which gives the cooperativity factor Alpha as a quantitative and comparable indicator to evaluate the ternary complex formation.^22,24,42 The cooperativity factor Alpha is defined by:

$$\text{Alpha}={\text{K}}_{\mathrm{D},\text{CRBN}}[\text{binary}]/{\text{K}}_{\mathrm{D},\text{CRBN}}[\text{ternary}]$$

where a positive cooperativity between two proteins results in log(Alpha) > 0, which suggests the direct interprotein interaction upon the ternary complex formation is favorable, while a negative cooperativity [log(Alpha) < 0] disfavors interprotein contacts (Fig. 2b). Because JQ1 binds to both BD1 and BD2 bromodomains with comparable affinities (K_D = 50 nM for BD1 and 90 nM for BD2),³⁹ we focused on whether the structurally diverse cyclimid-based degraders would interact with each bromodomain differently upon formation of a ternary complex with CRBN.

Fig. 2. Cyclimid degraders induce ternary complexes that are tunable and distinct from IMiD-based degraders.

(a) Schematic of the TR-FRET assay principle for determining K_D[binary] and K_D[ternary] against His₆-CRBN/DDB1. K_{D, CRBN}[ternary] values of the compounds were measured using the TR-FRET ligand displacement assay in the presence of a near-saturating concentration of either BD1 or BD2 domain of BRD4 (1 μM). (b) Schematic of the cooperativity factor Alpha. (c) Comparison of log(Alpha) values between cN and cQ cyclimid degraders. Log(Alpha) values for the cyclimids with the BD1 domain of BRD4 showed significant variance depending on the amino acid at the N–1 position. Data are presented as mean ± SD (n = 3 technical replicates). (d) Log(Alpha) for the selected cyclimid degraders with the BD1 domain of BRD4. Error bars represent mean ± SD.

Using this assay design, we measured K_{D, CRBN}[ternary, BD1 or BD2] and calculated the cooperativity factor Alpha for all 40 cyclimid degraders between His₆-CRBN/DDB1 and GST-BD1 or BD2 (Supplementary Table 1). The cyclimid degraders exhibited diverse cooperativity profiles compared to the thalidomide-based degrader dBET6, and the cooperativity of binding was dramatically altered by the amino acid installed at the N–1 position (Fig. 2c). The cyclimid degraders with neutral amino acids displayed positive cooperativity with BRD4(BD1) and negative cooperativity with BRD4(BD2). In contrast, dBET6 exhibited negative cooperativity for both bromodomains of BRD4, consistent with the previous report.²² Additionally, degraders without the amino acid at the N–1 position, JQ1-cQ and JQ1-cN, showed no significant cooperativity for either bromodomain, providing evidence that the N–1 amino acid is responsible for driving the differences in ternary complex formation. The log(Alpha) values of the cyclimid degraders varied significantly (e.g., an 80-fold difference in Alpha for BD1 between JQ1-YcQ and JQ1-PcQ; Fig. 2d, Supplementary Table 1). To further assess the selectivity of ternary complex formation mediated by the cyclimids with BD1 versus BD2 domains of BRD4, we calculated the BD1 and BD2 selectivity parameters derived from Alpha(BD1)/Alpha(BD2) for each degrader (Supplementary Fig. 2e). Among all cyclimid degraders tested, JQ1-YcQ showed the highest selectivity for the ternary complex formation with BRD4(BD1), with a 57-fold difference in Alpha for BD1 and BD2 (Supplementary Fig. 2f, Supplementary Table 1). Separately, we performed a modeling study of twenty cQ-based cyclimids to the thalidomide-binding domain of CRBN to investigate the impact of the N–1 position in the cyclimids on their interactions with CRBN (Supplementary Fig. 3a–b). While the N–1 residues do not appear to contribute differently to induce the closed conformation of CRBN, these models indicated that the amino acid at the N–1 position occupies a distinct molecular space relative to the phthalimide region of thalidomide or each other. Based on these models, the difference in cooperativity is likely due to the difference in the direction of each cyclimid that projects out from the thalidomide-binding domain of CRBN and their ability to induce a ternary complex that can yield efficient ubiquitination of a target protein. Collectively, these data demonstrate that the key to the diversity in ternary complex binding affinities of the cyclimids, as evidenced by K_{D, CRBN}[ternary] value and cooperativity factor Alpha, lies in the amino acid at the N–1 position adjacent to the cyclic imide motif, especially considering that all 40 cyclimid degraders possess the same linker motif.

K_{D, CRBN}[ternary] values are predictive of the cellular degradation activity of cyclimid degraders

In light of the significance of the amino acid at the N–1 position upon ternary complex formation, we next asked whether the ternary complex induced by the cyclimid degraders would be predictive of their cellular degradation activity. Prior work has demonstrated a generally strong correlation between cooperativity, ternary complex K_D or half-lives, and the degradation potency of the heterobifunctional degrader, although this trend has been only qualitatively demonstrated with select lead compounds across multiple reports.^{23–26,45–50} We therefore sought to extend these important precedents by conducting absolute quantitative measurements of the binary affinity, ternary affinity, and degradation on a scale that has not been previously performed. Collectively, we now provide a single report detailing the thorough correlation of biochemical and cellular activity for our library of 40 cyclimid-based BRD4 degraders and extend these observations to 20 additional cyclimids degraders for other targets.

First, to assess the selectivity in the degradation of each bromodomain of BRD4, we selected two pairs of cyclimid degraders, JQ1-YcQ/YcN and JQ1-AcQ/AcN, and the thalidomide-based analogs dBET6/dBET6-cN for evaluation in reporter HEK293T cells stably expressing either BRD4(BD1)-GFP or BRD4(BD2)-GFP.²² These two pairs of the cyclimid degraders were selected because JQ1-AcQ/AcN showed little to no ternary complex selectivity in our biochemical ternary complex assay, whereas JQ1-YcQ/YcN displayed 57- and 27-fold biochemical selectivity for ternary complexes with BD1. The induced degradation of BRD4(BD1) or BRD4(BD2) in these reporter cells after 4 h incubation with the selected cyclimid degraders or the thalidomide-based degraders was quantified directly in lysates adapting a TR-FRET-based strategy utilizing CoraFluor-labeled anti-GFP nanobodies⁵¹ (Fig. 3a, Supplementary Fig. 4a–b, Supplementary Table 1). Both JQ1-YcQ and JQ1-YcN exhibited higher potency for BD1 degradation with a ~5–10-fold difference in half-maximal degradation concentrations (DC₅₀) between BD1 and BD2, in good agreement with the BD1/BD2 selectivity parameter predicted from ternary complex studies [JQ1-YcQ: DC₅₀(BD1) = 103 nM; DC₅₀(BD2) = 513 nM, JQ1-YcN: DC₅₀(BD1) = 168 nM; DC₅₀(BD2) = 1.01 μM, Supplementary Table 1]. In contrast, the other pair of cyclimid degraders, JQ1-AcQ and JQ1-AcN, as well as dBET6, displayed comparable activity toward both bromodomains [JQ1-AcQ: DC₅₀(BD1) = 41 nM; DC₅₀(BD2) = 78 nM, JQ1-AcN: DC₅₀(BD1) = 34 nM; DC₅₀(BD2) = 36 nM, dBET6: DC₅₀(BD1) = 15 nM; DC₅₀(BD2) = 14 nM, Supplementary Table 1], consistent with a lack of selectivity between the individual recombinant bromodomains measured by in vitro ternary complex assays (Supplementary Fig. 3a).

Fig. 3. K_{D, CRBN}[ternary] values are predictive of the cellular degradation activity of cyclimid degraders.

(a) BRD4(BD1)-GFP or BRD4(BD2)-GFP levels in HEK293T cells stably expressing BRD4(BD1)-GFP or BRD4(BD2)-GFP after 4 h treatment with dBET6 and the selected cyclimid BRD4 degraders were measured by TR-FRET assay. For assay design, see Supplementary Fig. 4a. (b) Endogenous BRD4 protein levels in MDA-MB-231 cell lysate after 5 h treatment with dBET6 and the selected cyclimid BRD4 degraders were measured by TR-FRET assay. HSC, high signal control; LSC, low signal control. Data in (a) and (b) are presented as mean ± SD (n = 2 biologically independent samples). (c) Correlation (r_p = 0.82) between pDC₅₀ values of the selected cyclimid degraders for endogenous BRD4 degradation and their biochemical pK_{D, CRBN}[ternary] values. In the correlation graph, for each degrader, the most potent pK_{D, CRBN}[ternary] value between pK_{D, CRBN}[ternary, BD1] and pK_{D, CRBN}[ternary, BD2] was selected.

We next investigated the correlation between the degradation potency of endogenous BRD4 and K_{D, CRBN}[ternary] values, with the intent of characterizing the K_{D, CRBN}[ternary] values as a quantitative indicator for predicting the degradation potency profiles of the cyclimid degraders. To examine the degradation efficiencies of the cyclimid degraders for endogenous BRD4, MDA-MB-231 cells were treated for 5 h with a series of cyclimid degraders and controls in dose-response, followed by the quantification of endogenous BRD4 levels by TR-FRET⁴³ (Fig. 3b, Supplementary Fig. 5a). This subset of the cyclimid degraders (20 of 40 degraders) was chosen to reflect nonpolar aliphatic, aromatic, and polar amino acids at the N–1 position, along with control compounds (dBET6, dBET6-cN, JQ1-cQ, and JQ1-cN). The cyclimid degraders exhibited a 500-fold range of DC₅₀ values. The N–1 position of the cyclimids was essential for tuning the degradation properties, where in general cyclimid degraders with nonpolar or aromatic amino acid side chains at the N–1 position facilitated BRD4 degradation to a greater extent (Supplementary Table 1). This differential activity may be further amplified by decreased cellular permeability of analogs with polar side chains. The cellular degradation profile of the cyclimid degraders showed the strongest correlation with the pK_{D, CRBN}[ternary] values against their respective “most potent” BRD4(BD1) or BRD4(BD2) domain (p-value<0.0001, Fig. 3c). Cellular degradation also correlated significantly with binary CRBN target engagement (p-value=0.007, Supplementary Fig. 5b). In contrast, no correlation was observed with target engagement toward BRD4(BD1) and BRD4(BD2) (Supplementary Fig. 5c–d), most likely because all degraders possess the same BRD4 targeting ligand (JQ1). Further evaluation of the 20 cN-bearing cyclimid degraders in HEK293T cells showed dose-dependent degradation of BRD4 equivalent to or better than that observed in MDA-MB-231 cells (Supplementary Fig. 5e–h). The cN-based cyclimid degraders generally possess superior degradation abilities compared to their cQ counterparts, in alignment with the observed differences in binding affinities to CRBN both in binary and ternary systems (Supplementary Table 1). In sum, these results collected from a comprehensive library of cyclimid degraders demonstrate that the degradation efficacy of cyclimid degraders can be modulated through the formation of diverse ternary complex structures, enabling K_{D, CRBN}[ternary] values to serve as a quantitative indicator to predict the degradation activity.

Ligand-dependent dynamics of ternary complex dissociation and degradation with the cyclimids

The kinetic properties of PROTAC-mediated ternary complexes, such as their dissociative half-life (t_1/2), have also been shown to influence target protein degradation efficiency, with longer t_1/2 values shown to correlate with more efficient target degradation, at least for BRD4-targeted PROTACs.²⁵ Building on assay principles like those used in the NanoBRET-based drug residence time analysis,^29,30 we designed a TR-FRET-based ternary complex dissociation kinetics experiment. Using this platform, we measured the off-rates (k_off) of ternary complexes between His₆-CRBN/DDB1 and each bromodomain of BRD4 for a subset of the cyclimid degraders (Fig. 4a). Incubation of His₆-CRBN/DDB1 and GST-BRD4(BD1) or BRD4(BD2) (labeled with AF488-anti-His₆ antibody and CoraFluor-1-labeled GST antibody, respectively) with concentrations of bifunctional degraders equal or near to their respective K_{D, CRBN}[ternary] values (see Supplementary Table 1) resulted in the highest assay sensitivity. In the first set of experiments, BRD4(BD1) or BRD4(BD2) dissociation from the ternary complex was measured by adding excess JQ1, and dissociation kinetics of the ternary complex with JQ1-AcQ/AcN, JQ1-YcQ/YcN, JQ1-ScQ/ScN, JQ1-GcQ/GcN, or dBET6 were monitored using TR-FRET. These four pairs of cyclimid degraders were selected based on their biochemical selectivity for ternary complexes with BD1 versus BD2. For ternary complexes consisting of His₆-CRBN/DDB1 and BRD4(BD1), we observed dissociation half-lives (t_1/2) varying from ≤ 30 sec (the time from sample injection to the first measurement) to 270 sec across the degraders tested (Fig. 4b). JQ1-YcQ and JQ1-YcN yielded the most stable and long-lived complexes, consistent with their high positive cooperativity for BRD4(BD1) (Fig. 4c, Supplementary Table 2). With BRD4(BD2), however, all ternary complexes exhibited fast dissociation kinetics (t_1/2 ≤ 30 sec) regardless of the choice of degrader, which aligns with the negative cooperativity observed for most of the cyclimid degraders with this bromodomain (Fig. 4b, Supplementary Table 2). A reversed dissociation mechanism was also investigated, in which His₆-CRBN/DDB1 dissociation from the degrader-mediated ternary complex with each bromodomain of BRD4 was triggered by the addition of lenalidomide (Supplementary Fig. 6a). Similarly, ternary complexes with BRD4(BD1) displayed much slower dissociation kinetics relative to ternary complexes with BRD4(BD2) (Supplementary Fig. 6b–c).

Fig. 4. Ligand-dependent dynamics of ternary complex dissociation and degradation with the cyclimids.

(a) Schematic illustration of TR-FRET assay system employed to monitor BRD4 dissociation from the degrader-mediated ternary complex. (b) Kinetic measurements of BRD4 dissociation in the presence of the BD1 or BD2 domain of BRD4 from the ternary complex induced by the selected cyclimid degraders or dBET6. Data are presented as mean ± SD (n = 2 technical replicates) and fitted to a one-phase decay model in Prism 9. (c) Half-lives (t_1/2) of the His₆-CRBN/DDB1–BRD4(BD1) ternary complex mediated by the indicated cyclimid degraders and dBET6. Error bars represent mean ± SD (n = 2 technical replicates). (d) Kinetics of depletion of BRD4(BD1)-GFP in HEK293T cells stably expressing BRD4(BD1)-GFP after the treatment with JQ1-YcQ or dBET6. (e) Kinetics of depletion of BRD4(BD2)-GFP in HEK293T cells stably expressing BRD4(BD2)-GFP after the treatment with JQ1-YcQ or dBET6. Data in (d) and (e) are presented as mean ± SD (n = 2 biologically independent samples).

As a part of our investigation into the dissociation kinetics and ternary complex formation induced by these BRD4 degraders, the time- and concentration-dependent degradation kinetics of each bromodomain of BRD4 were determined for JQ1-YcQ and dBET6 (Fig. 4d–e). JQ1-YcQ achieved its maximum level of degradation (D_max) faster with BRD4(BD1) than with BRD4(BD2) (1.5 h for BD1; 3 h for BD2), which indicates that the more favorable ternary complex formation with BD1 than BD2, together with the longer off-rate for BD1 versus BD2, are likely driving the observed difference in degradation dynamics. In contrast, the time required to reach its maximum degradation efficiency with dBET6 was comparable for both bromodomains of BRD4, which reflects the slight difference in the cooperativity and the off-rates of dBET6-induced ternary complexes with BRD4(BD1) or BRD4(BD2). These data provide further evidence that the “occupancy” of cyclimid degraders in a ternary complex (i.e., how long the target protein remains tethered to CRBN by the degraders) is another useful parameter in evaluating the degradation activity of the cyclimid degraders, together with K_{D, CRBN}[ternary] values.

The cyclimid library is readily applicable to target other proteins for selective degradation

To further demonstrate the generalizability of the cyclimid platform and the robustness of our TR-FRET-based biochemical approach, we expanded the substrate scope of the cyclimid library beyond BRD4. We synthesized a set of cyclimid-based degraders targeting FKBP, a protein family belonging to the immunophilins, by substituting the thalidomide moiety of dFKBP-1¹⁵ with each of the ten cyclimids (Fig. 5a). The selection of cyclimids installed in FKBP degraders was based on our observations that nonpolar aliphatic and aromatic cyclimids generally demonstrated superior cellular degradation ability and binding affinity against CRBN compared to cyclimids with polar amino acids at the N–1 position. We additionally included proline because of its unique “bent” structure relative to the rest of the cyclimids. SLF,⁵² an analog of immunosuppressant FK506, is a high-affinity ligand for FKBP12, which has been exploited in thalidomide-based degraders like dFKBP-1. As with the evaluation of BRD4 degrader-mediated ternary complexes, using SLF-FITC as a tracer, TR-FRET assays were performed to determine K_{D, CRBN}[ternary, FKBP12] of dFKBP-cyclimid degraders and dFKBP-1 by measuring their binding towards CoraFluor-1-labeled His₆-CRBN/DDB1 in the presence of a near-saturating concentration of recombinant FKBP12 (Supplementary Fig. 7). The dFKBP-cyclimid degraders, like the cyclimid-based BRD4 degraders, exhibited a broad range of K_{D, CRBN}[ternary] values influenced by the amino acid choice at the N–1 position (Fig. 5b, Supplementary Table 3).

Fig. 5. The cyclimid library is readily applicable to target other proteins for selective degradation.

(a) Structure of dFKBP-1 and cyclimid-based FKBP degraders, dFKBP-cyclimid. (b) pK_{D, CRBN}[ternary] values of FKBP degraders against His₆-CRBN/DDB1 in the presence of a near-saturating concentration of FKBP12 (1 μM). Data are presented as mean ± SD (n = 2 technical replicates). (c) Western blot of FKBP12, FKBP51, and FKBP52 levels after treatment of HEK293T cells for 24 h with 10 μM dFKBP-1 or each dFKBP-cyclimid. (d) Correlation (r_p = 0.81) between pDC₅₀ values of FKBP12 degradation and biochemical pK_{D, CRBN}[ternary] values. pDC₅₀ values of FKBP12 degradation were obtained by quantification of FKBP12 levels relative to DMSO control. Corresponding blot images and quantifications are shown in Supplementary Fig. 8 and uncropped blot images are supplied in Supplementary Fig. 4. (e) Structure of dCDK-1 and cyclimid-based CDK degraders, dCDK-cyclimid. (f) Western blot of CDK4 and CDK6 levels after treatment of Jurkat cells for 4 h with 1 μM dCDK-1 or each dCDK-cyclimid. (g) Western blot of CDK4 and CDK6 levels after treatment of Jurkat cells for 4 h with 10 nM dCDK-1 or each dCDK-cyclimid. (h–j) Degradation of CDK4 (red) and CDK6 (blue) after treatment of Jurkat cells for 4 h with (h) dCDK-1, (i) dCDK-PcQ, or (j) dCDK-PcN over a dose-response range of 0.64 nM to 2 μM. Quantification of CDK4 and CDK6 levels were calculated relative to DMSO control. Data in (h)–(j) are presented as mean ± SD (n = 3 biologically independent samples). Corresponding blot images and quantifications are shown in Supplementary Fig. 9 and uncropped blot images are supplied in Supplementary Fig. 5. All western blot data are representative of at least three independent replicates. For uncropped western blot images, see Figure S2 in the Supplementary Information.

Due to the highly homologous binding domain of FKBPs, the FKBP-12 binder SLF has been shown to bind to other FKBP isoforms in the cell.⁵² The degradation activity of the dFKBP-cyclimid library was thus assessed against three members of the FKBP family – FKBP12, FKBP51, and FKBP52 – to determine the generality and selectivity of the cyclimid degraders. With the treatment of HEK293T cells at 1 μM, dFKBP-cyclimid degraders were as effective as the parent dFKBP-1 in promoting degradation of FKBP12, but did not induce degradation of FKBP51 and FKBP52 (Fig. 5c). Rescue of FKBP12 degradation was observed when cells were pre-treated with a proteasome inhibitor, a neddylation inhibitor, or excess lenalidomide to block CRBN engagement (Supplementary Fig. 8a). While the cyclimid library retains its unique ability to form various ternary complex configurations and to access the productive formation for degradation, each cyclimid revealed a differential potency when embedded in dFKBP structure in comparison with its counterpart BRD4 degrader. For instance, the cyclimid AcQ afforded one of the most potent BRD4 degraders (Fig. 3b, Supplementary Table 1), whereas LcQ was the most effective at FKBP12 degradation among the cQ series (Fig. 5c, Supplementary Table 3). To assess the relationship between degradation potency and the observed K_{D, CRBN}[ternary] values for dFKBP-cyclimids, we determined DC₅₀ values for a subset of dFKBP degraders (Supplementary Fig. 8b–d, Supplementary Table 3) and correlated these with pK_{D, CRBN}[ternary] values. A statistically significant correlation was also observed for the dFKBP-cyclimid series, further reinforcing the utility of K_{D, CRBN}[ternary] values as a quantitative indicator for predicting the degradation efficacy in cells (p-value=0.028, Fig. 5d).

We additionally constructed another set of cyclimid-based degraders for the Ser/Thr cell cycle kinases CDK4 and CDK6 to evaluate whether the cyclimids can provide a selective degrader by replacement of the pomalidomide moiety of dCDK-1 (Fig. 5e).^53,54 The same set of amino acids as employed for the cyclimid FKBP degraders was selected for constructing a series of CDK degraders. The dCDK-cyclimid library all facilitated complete degradation of CDK6 at 100 nM, with several degraders being functional at 10 nM (Fig. 5f–g, Supplementary Fig. 9a). Interestingly, the hook effect resulting in reduced degradation of CDK6 can be observed with the cQ-cyclimid degraders and the pomalidomide-based dCDK-1 at 10 μM, but not with the cN-cyclimid degraders, highlighting an advantageous feature of screening both cN- and cQ-cyclimids as warheads in TPD (Supplementary Fig. 9b). The addition of a proteasome inhibitor, a neddylation inhibitor, or excess lenalidomide completely inhibited CDK4 and CDK6 degradation by a cyclimid degrader, dCDK-PcN, or dCDK-1 (Supplementary Fig. 9c). While most dCDK-cyclimids conferred partial CDK4 depletion at 1 μM, dCDK-PcQ showed a suppressed level of CDK4 degradation relative to other CDK degraders (Fig. 5f). Deeper concentration-dependent profiling over a dose-response range of 0.64 nM to 2 μM with dCDK-PcQ, dCDK-PcN, and dCDK-1 revealed that all three degraders exhibited similar DC₅₀ values against CDK6 (DC₅₀ = 2–9 nM), whereas dCDK-PcN and dCDK-1, but not dCDK-PcQ, had a significant degradation activity against CDK4 (dCDK-1 and dCDK-PcN: DC₅₀ = 10–20 nM with D_max = 72 %, Fig. 5h, j; dCDK-PcQ: DC₅₀ = 89 nM with D_max = 25 %, Fig. 5i, Supplementary Table 4). Therefore, dCDK-PcQ is remarkably selective towards CDK6 in cellular degradation assays. Intriguingly, the cN-cyclimid series-derived degraders tended to show superior degradation efficacy to the cQ-cyclimid homologs, regardless of the choice of target protein (Supplementary Table 1, 3–4). These data indicate that the structurally flexible and versatile cyclimid library can readily tune target selectivity and offers an orthogonal, tunable, and efficient platform for optimizing target degradation and selectivity of heterobifunctional degraders through induction of distinct ternary complexes.

Cyclimid degraders do not recruit known neosubstrates of thalidomide and its derivatives

One limitation of thalidomide-based scaffolds in bifunctional degraders for TPD is their potential for off-target toxicity due to the inherent neosubstrate recruitment by the thalidomide-based warhead, including the teratogenic factor SALL4^55,56 or the essential translation termination factor GSPT1.¹³ As the cyclimids and thalidomide derivatives promote distinct binding profiles with CRBN, we expected to observe differentiation in their corresponding neosubstrate recruitment profiles (Fig. 6a). We thus evaluated whether the cyclimid-based bifunctional degraders mitigate the off-target degradation of known neosubstrates of thalidomide and its derivatives, adapting a previously reported target profiling platform.^57,58

Fig. 6. Cyclimid degraders do not recruit known thalidomide analog-sensitive neosubstrates.

(a) Schematic of the proposed reduced off-target degradation by the cyclimid-based degraders, in contrast to the thalidomide-based degraders. (b) Degradation of pomalidomide-sensitive zinc-finger (ZF) degrons in cells by the thalidomide- and cyclimid-based BRD4 degraders over a dose-response range of 32 nM to 20 μM. U2OS cells stably expressing 6 ZF degrons fused to GFP were treated with PROTACs followed by flow cytometry to assess ZF degradation. (c) Degradation of GSPT1 in cells using thalidomide- and cyclimid-based BRD4 degraders. HEK293T cells stably expressing GSPT1-GFP were treated with PROTACs followed by flow cytometry to assess GSPT1 degradation. (d) Degradation of pomalidomide-sensitive ZF degrons in cells by thalidomide- or cyclimid-based FKBP and CDK degraders over a dose-response range of 32 nM to 20 μM. U2OS cells stably expressing 2 ZF degrons fused to GFP were treated with degraders followed by flow cytometry to assess ZF degradation. (e) Degradation of GSPT1 in cells using thalidomide- and cyclimid-based FKBP and CDK degraders. HEK293T cells stably expressing GSPT1-GFP were treated with PROTACs followed by flow cytometry to assess GSPT1 degradation. Data in (b)–(e) are presented as mean (n = 3 biologically independent samples).

Specifically, to assess the off-target properties of the cyclimids, U2OS cells stably expressing several zinc-finger (ZF) degrons fused to GFP⁵⁷ were treated in dose-response with the selected cyclimid degraders, together with the thalidomide-based degraders dBET6 and ARV-825,^15,16 and the degradation of each ZF degron was monitored by flow cytometry (Fig. 6b, Supplementary Fig. 10a). Evaluation of known C2H2 zinc finger targets of thalidomide derivatives, including IKZF3 and ZFP91,^10,11,59 revealed their fast and rapid degradation by thalidomide-based degrader ARV-825. By contrast, almost no degradation of the same off-targets was observed with the cyclimid degraders (D_max ≤ 3%), confirming their high capability of reducing unwanted off-target interactions. A second thalidomide-based degrader, dBET6, also exhibited minimal off-target degradation comparable to the cyclimid degraders (D_max ≈ 6%). The off-target properties of the cyclimids were further explored with HEK293T cells stably expressing GSPT1-GFP. After treating the cell line with the same set of degraders, both the cyclimid and the thalidomide-based degraders considerably reduced the undesired degradation of GSPT1 (D_max ≤ 5% for the cyclimids; D_max ≤ 9% for dBET6; D_max ≤ 1% for ARV-825, Fig. 6c, Supplementary Fig. 10b). Some cyclimid degraders from the dFKBP and dCDK degrader library, dFKBP-LcQ/LcN and dCDK-PcQ/PcN, were additionally assessed for promiscuity. Using U2OS cells stably expressing ZF degrons of IKZF3 and ZFP91 fused with GFP and HEK293T cells stably expressing GSPT1-GFP, we tested for off-target degradation of these proteins with the dFKBP- and dCDK-cyclimids (Fig. 6d–e). The cyclimid degraders again did not affect the level of those neosubstrates commonly observed with thalidomide derivatives (D_max ≤ 1% for dFKBP-LcQ; D_max ≤ 4% for dFKBP-LcN; D_max ≤ 6% for dCDK-PcQ; D_max ≤ 8% for dCDK-PcN), demonstrating that the cyclimid ligands are unlikely to promote degradation of thalidomide analog-sensitive neosubstrates, thereby reducing unsought off-target degradation. As the use of a single ZF degron as a reporter may not fully translate to the degradation of the full-length protein, we performed orthogonal unbiased global proteomics to validate the target selectivity of the cyclimid degraders. Unbiased global proteomics provided further evidence for the high target selectivity of cyclimid degraders dCDK-PcQ and dCDK-PcN, which deplete CDK6 alone or both CDK6 and CDK4, respectively (Supplementary Fig. 10c–e).

Encouraged by the reduced off-target effects of the cyclimids, we characterized their pharmacokinetic properties to inform their translatability to in vivo systems. Other major limitations of the thalidomide analogs are their propensity for epimerization of the stereocenter in the glutarimide and their hydrolytic instabilities.⁶⁰ By contrast, the cyclimids offer advantages in reducing both destruction pathways. Several monofunctional and bifunctional cyclimid compounds were evaluated for metabolic stability against liver S9 fractions and in vitro permeability in a Caco-2 transwell assay, and compared to thalidomide, lenalidomide, or dBET6 (Supplementary Fig. 10f–g). The cyclimid-based degraders were found to have similar metabolic stability and in vitro cell permeability characteristics as the thalidomide-based degrader dBET6. Monofunctional, JQ1-independent cyclimids showed superior and more favorable metabolic stability than thalidomide derivatives. Thus, the cyclimids represent a unique class of CRBN warheads for use in TPD, which possess minimal thalidomide analog-sensitive off-target degradation and promising biochemical and cellular properties to accelerate degrader development for mechanistic studies and translational discovery programs.

Discussion

Here we report that the cyclimids, a new class of CRBN binders inspired by the C-terminal cyclic imide degron of CRBN, possess distinct modes of interaction with CRBN, thereby enabling efficient sampling of ternary complex space when employed as CRBN ligands in heterobifunctional degraders. A rich diversity in ternary complex binding affinities of the cyclimids, combined with their easily translatable properties to target different proteins, offers a facile approach for screening target degradation and selectivity in bifunctional degraders. The structural versatility of the cyclimids enables rapid and efficient screening and optimization to maximize the formation of productive ternary complexes^22,24,61 tailored for individual target proteins. An in-depth investigation of cyclimid degraders in binary and ternary complexes across two different substrates (BRD4 and FKBP12) using CoraFluor TR-FRET technology revealed that K_{D, CRBN}[ternary] values can serve as a quantitative indicator to predict degradation potency of the cyclimid degraders.

Our work establishes a novel concept for TPD in the ability to rapidly and systematically screen a series of CRBN warheads based on the C-terminal cyclic imide degron. This concept is readily enabled by the cyclimids and represents a complementary approach to the concept of “linkerology” that is widely utilized to discover effective degraders.^23,62,63 While we report here the use of the 20 canonical amino acids, a natural future extension is to utilize noncanonical amino acids (similar to the strategies employed for developing second-generation VHL-ligands⁶⁴), which will further expand the scope and impact of a warhead screening approach. Cumulatively, the systematic screening of the structure–activity relationship of the CRBN warhead, as we demonstrated with the comprehensive evaluation of over 60 bifunctional degraders, opens a new concept in screening the E3 ligase-binding warhead during degrader development.

The cyclimid library represents a unique class of CRBN-binding ligands for TPD applications, further expanding the current toolbox of emerging substitutes^33–37 for the thalidomide derivatives. Degradation of thalidomide analog-sensitive off-targets^56,65 was not observed with the cyclimid degraders evaluated here, providing evidence that the cyclimids may permit the design of clinically favorable CRBN-targeting therapeutics that avoid long-term side effects for patients. The peptidic properties of the cyclimids may be further harnessed by exploiting the ability of peptide-based molecules to interact with endogenous proteins, such as enzymes and receptors, that recognize specific amino acid sequences or post-translational modifications. These therapeutic strategies may realize desirable drug properties such as tissue-specific targeting and delivery. In light of the diverse suite of ternary complex structures accessible with the cyclimid-based bifunctional degraders, exploring non-peptidic structures will further diversify the cyclimid library and enhance desirable drug-like properties. CRBN-degron-inspired monofunctional degraders may also potentially emerge from these efforts in the future.

We additionally provide the first and most comprehensive evidence that the cyclimids preferentially engage CRBN as the 5-membered aspartimide ligand, which is inactive when installed as the thalidomide derivative. These properties from the cyclimids are in stark contrast to those observed with the corresponding thalidomide derivatives. Cyclization of asparagine is more prevalently observed than cyclization of glutamine under physiological conditions due to the formation of the 5-membered aspartimide being kinetically more favorable than the 6-membered glutarimide.⁶⁶ This observation implies that cyclic aspartimides may represent a more dominant degron motif for the thalidomide-binding domain of CRBN than cyclic glutarimides, and CRBN may be more receptive to 5-membered cyclic aspartimides. Focusing on functional differences between the two classes of the cyclimids bearing C-terminal cN or cQ, the cyclimid degraders with 5-membered cN generally exhibited higher affinities to CRBN than their 6-membered cQ counterparts. While one study found that cN and cQ-based peptides exhibited similar engagement with the human thalidomide-binding domain,⁶⁷ the discrepant results may be attributed to differences in the binding affinity measured by engagement of the thalidomide-binding domain alone versus the full-length CRBN/DDB1 employed herein. Regardless of the target protein chosen (e.g., BRD4, FKBP12, and CDK4/6), the cN-cyclimid degraders largely showed superior degradation efficacy than the cQ-cyclimid series. To further illuminate the biological relevance of the aspartimide and glutarimide modifications with CRBN biology, the mechanism of formation of C-terminal cyclic imides – when, where, and how these modifications are formed in the proteome – will need to be elucidated. With the maturation of studies mapping C-terminal cyclic imide sites and substrates across the proteome, the cyclimids may also provide an exciting avenue for designing small molecules that can further tap into native CRBN functions within the proteome.

STAR Methods

Resource availability

Lead contact

Further information and requests for resources and reagents generated in this study should be directed to the lead contacts, Christina M. Woo (cwoo@chemistry.harvard.edu) and Ralph Mazitschek (ralph@broadinstitute.org).

Materials availability

All cell lines and unique reagents generated in this study are available from the lead contacts with a completed Materials Transfer Agreement.

Experimental model and study participant details

Mammalian Cell Lines

HEK293T and Jurkat cells were obtained from American Type Culture Collection (ATCC). HEK293T cells stably expressing stably expressing BRD4(BD1)-GFP or BRD4(BD2)-GFP were kindly provided by the Fischer Lab (Dana-Farber Cancer Institute).²² U2OS cells stably expressing the GFP-fused zinc finger proteins were kindly provided by the Choudhary Lab (Broad Institute).⁵⁷ All cells were cultured in DMEM or RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1× penicillin-streptomycin. Cells were grown at 37 °C in a humidified atmosphere with 5% CO₂. Mycoplasma testing was performed regularly for all cell lines to check for contamination. For the collection of cell pellets, cells were dissociated and collected by two PBS washes and centrifugation at 500 × g, 24 °C, 3 min, followed by a final PBS wash of the pellets. The pellets were flash frozen with liquid nitrogen and stored at −80 °C until use.

Methods details

Western blotting procedures.

Unless otherwise noted, cells were lysed by probe sonication (5 sec on, 3 sec off, 15 sec in total, 11% amplitude) in 1–2% SDS in 1× PBS and cleared by centrifugation at 21,000 × g, 4 °C, 10 min. As needed, a BCA assay was performed to determine the protein concentration of lysates and the concentration was adjusted using lysis buffer. 5× SDS-PAGE loading buffer (5% (v/v) β-mercaptoethanol, 0.02% (w/v) bromophenol blue, 30% (v/v) glycerol, 10% (w/v) SDS/ 250 mM Tris pH 6.8) was added to the protein samples to a final concentration of 1× and the samples were heated at 95 °C for 5 min. Protein samples (8–15 μL per lane) were loaded on NuPAGE 3–8% Tris-Acetate precast gels or Criterion XT Tris-Acetate precast gels for high molecular weight proteins, 6/12% Tris-Glycine gels or 4–15% Criterion^™ TGX^™ precast gels for medium molecular weight proteins, or 16.5% Mini-PROTEAN^® Tris-Tricine gels or 15% Criterion Tris-HCl protein gels for low molecular weight proteins. Gels were transferred to membranes using the Invitrogen iBlot 2 dry blotting system and iBlot 2 nitrocellulose transfer stacks, using program P0 (1 min at 20 V, 4 min at 23 V, 2 min at 25 V) for most proteins and 8 min at 25 V for high molecular weight proteins. Membranes were stained with Ponceau S solution to visualize transfer and protein loading. After being blocked with 5% milk or BSA in TBST at 24 °C for 1 h, the membranes were incubated with primary antibodies at 24 °C for 1 h or at 4 °C for 1 to 24 h. Membranes were washed (3 × 5 min) with TBST and incubated with secondary antibodies at 24 °C for 1 h. Membranes were washed (3 × 5 min) with TBST and the results were obtained by chemiluminescence and/or IR imaging using Azure 600 or c600.

BRD4 degradation assay.

1.5×10⁶ HEK293T cells were seeded in DMEM supplemented with 10% FBS and 1× penicillin-streptomycin and incubated at 37 °C, 5% CO₂ for 30 min, then treated with compounds of interest and incubated at 37 °C, 5% CO₂ for 4 h prior to collection and lysis. If noted, cells were treated with 1 μM MLN4924 for 1 h before treatment with compounds following the initial 30 min incubation. All compounds were dissolved in DMSO, and the final DMSO concentration after addition of the compound to the cells did not exceed 0.2% v/v.

FKBP12, FKBP51, and FKBP52 degradation assay.

1.5×10⁶ HEK293T cells were seeded in DMEM supplemented with 10% FBS and 1× penicillin-streptomycin and incubated at 37 °C, 5% CO₂ for 30 min, then treated with compounds of interest and incubated at 37 °C, 5% CO₂ for 18 h prior to collection and lysis. If noted, cells were treated with 1 μM MLN4924 for 1 h before treatment with compounds following the initial 30 min incubation. All compounds were dissolved in DMSO and the final DMSO concentration after addition of the compound to the cells did not exceed did not exceed 0.2% v/v.

CDK4 and CDK6 degradation assay.

1.5×10⁶ Jurkat cells were seeded in RPMI supplemented with 10% FBS and 1× penicillin-streptomycin and incubated at 37 °C, 5% CO₂ for 30 min, then treated with compounds of interest and incubated at 37 °C, 5% CO₂ for 24 h prior to collection and lysis. If noted, cells were treated with 1 μM MLN4924 for 1 h before treatment with compounds following the initial 30 min incubation. All compounds were dissolved in DMSO and the final DMSO concentration after addition of the compound to the cells did not exceed did not exceed 0.2% (v/v).

Cloning, overexpression, and purification of His₆-FKBP12.

The cloning, overexpression, and purification of His₆-FKBP12 were adapted from Flaxman and co-workers.⁶⁸ In brief, Rosetta 2 (DE3) chemically competent cells transformed with the pGEX2T-His₆-FKBP12 plasmid were used to inoculate a culture of autoclaved Luria Bertani broth with 100 μg/mL ampicillin (50 mL). The culture was grown at 30 °C for 15–18 h with shaking at 200 rpm. The saturated culture was then diluted 1:100 into autoclaved Luria Bertani Broth (750 mL) with 100 μg/mL ampicillin in a 2 L baffled flask. The OD600 of the culture was monitored until the OD600 measured between 0.3–0.6, at which point protein expression was induced with 0.3 mM IPTG. After the addition of IPTG, the temperature of the incubator was reduced to 25 °C. The culture was incubated with shaking at 200 rpm at 25 °C for 5 h. Cells were harvested by centrifugation at 3,750 rpm (2,249 × g, 24 °C, 15 min). The pellet was re-suspended in lysis buffer (5 mL, 1 mM DTT, 0.5 mg/mL lysozyme/PBS), added protease inhibitor, and lysed by sonication (30 sec on, 10 sec off, 15% amplitude, 5 min total). DNAse (0.8 mg/mL) and RNAse (0.008 mg/mL) were added to the lysate and incubated for 30 min. After the clarification by centrifugation (10,000 x g, 4 °C, 10 min), followed by filtration through a 0.45 μm syringe filter, the clarified lysate was applied to the column with glutathione sepharose resin (1 mL) and shaken at 4 °C for 1–2 h. The unbound lysate was discarded, and the resin was sequentially washed with 1 mM DTT/PBS (20 mL) and thrombin cleavage buffer (5 mL, 0.3 mM CaCl2, 1 mM DTT /20 mM Tris, 100 mM NaCl, pH 6.8). The column was then treated with thrombin (100 units) in thrombin cleavage buffer (5 mL) and incubated with shaking at 4 °C for 15–18 h. The cleaved protein was collected from the column, which was further washed with 1 mM DTT/PBS (20 mL). The remaining GST was eluted with 10 mM reduced glutathione/50 mM Tris, pH 8.0 (5 mL). The thrombin cleavage and following washes were concentrated to 500 μL using a 3 kDa MWCO spin concentrator, with thrombin removal in conjunction with final buffer exchange. The final purification and buffer exchange were performed by SEC using a Superdex 75 GL 10/300 column, run with PBS at a flow rate of 0.5 mL/min. The desired protein was then collected from FPLC fractions and concentrated using a 3 kDa MWCO spin concentrator until the protein concentration was >1 mg/mL. Protein solution was aliquoted, flash frozen with liquid nitrogen, and stored at −80 °C until use.

TR-FRET measurements.

Unless otherwise noted, experiments were performed in white, 384-well microtiter plates (Corning 3572 or Greiner 781207) in 30 μL assay volume. TR-FRET measurements were acquired on a Tecan SPARK plate reader with SPARKCONTROL software version V2.1 (Tecan Group Ltd.), with the following settings: 340/50 nm excitation, 490/10 nm (Tb), and 520/10 nm (FITC, AF488, GFP) emission, 100 μs delay, 400 μs integration. The 490/10 nm and 520/10 nm emission channels were acquired with a 50% mirror and a dichroic 510 mirror, respectively, using independently optimized detector gain settings unless specified otherwise. The TR-FRET ratio was taken as the 520/490 nm intensity ratio on a per-well basis.

Protein labeling for TR-FRET measurements.

Anti-His₆ antibody (Abcam 18184), anti-GST antibody (Abcam 19256), anti-GST V_HH (ChromoTek ST-250), anti-GFP V_HH (ChromoTek GT-250), anti-RFP V_HH (ChromoTek RT-250), anti-rabbit Nano-Secondary (ChromoTek shurbGNHS-1), and His₆-CRBN/DDB1 complex were labeled with CoraFluor-1-Pfp ester or AF488-Tfp ester (ThermoFisher A37570) as previously described.^40,69 The following extinction coefficients were used to calculate protein concentration and degree-of-labeling (DOL): IgG E₂₈₀ = 210,000 M⁻¹cm⁻¹, anti-GST V_HH E₂₈₀ = 28,545 M⁻¹cm⁻¹, anti-GFP V_HH E₂₈₀ = 27,055 M⁻¹cm⁻¹, anti-RFP V_HH E₂₈₀ = 30,035 M⁻¹cm⁻¹, anti-rabbit Nano-Secondary E₂₈₀ = 24,075 M⁻¹cm⁻¹, His₆-CRBN-DDB1 E₂₈₀ = 167,000 M⁻¹cm⁻¹, CoraFluor-1-Pfp E₃₄₀ = 22,000 M⁻¹cm⁻¹, AF488-Tfp E₄₉₅ = 71,000 M⁻¹cm⁻¹. Protein conjugates were snap-frozen in liquid nitrogen, and stored at −80 °C.

Determination of equilibrium dissociation constants (K_D or K_D,app) for fluorescent tracers by TR-FRET.

The following conditions have been employed: (i) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2.5 nM His₆-CRBN/DDB1 complex in assay buffer (25 mM HEPES, 150 mM NaCl, 0.5 mg/mL BSA, 0.005% TWEEN-20, pH 7.5), (ii) 2.5 nM CoraFluor-1-labeled His₆-CRBN/DDB1 complex in assay buffer, (iii) 0.5 nM GST-BRD4(BD1) or GST-BRD4(BD2), 4 nM CoraFluor-1-labeled anti-GST V_HH in assay buffer, (iv) 2.5 nM CoraFluor-1-labeled anti-His₆ antibody, 5 nM His₆-FKBP12 in assay buffer, (v) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 0.5 nM His₆-GST-CDK4/GST-CD3 in assay buffer, (vi) 1 nM CoraFluor-1-labeled anti-6xHis antibody, 2 nM His₆-CDK6/GST-CD1 in assay buffer.

Fluorescent tracers (Thalidomide-FITC for CRBN-DDB1 assays, JQ1-FITC for BRD4 assays, SLF-FITC for FKBP12 assays, Palbociclib-FITC for CDK4/6 assays) were added in serial dilution using a HP D300 digital dispenser (Hewlett-Packard) and allowed to equilibrate for 1 h at room temperature before TR-FRET measurements were taken. For all assays using CoraFluor-1-labeled anti-epitope tag reagents (anti-6xHis antibody, anti-GST V_HH), nonspecific signal was determined from wells containing the CoraFluor-1-labeled conjugate alone in the absence of the target protein. For assays using CoraFluor-1-labeled His₆-CRBN/DDB1 complex, nonspecific signal was determined from wells containing 100 μM Lenalidomide-5′-NH₂. Data were fitted to a One Site – Specific Binding model in Prism 9.

For conditions (i), (iii), and (iv), the $K_{\mathrm{D}}$ determined from the one-site model was then used in Equation 1 to adjust for a two-site model due to the presence of dimeric GST protein or bivalent anti-6xHis antibody:

$$K_{D,2-site}=K_{D,1-site}\times \left(1+\sqrt{2}\right)$$ (1)

For conditions (v) and (vi), we did not attempt to determine the absolute $K_{\mathrm{D}}$ values for the Palbociclib-FITC tracer due to the potential presence of higher-order oligomeric complexes. For CDK4/6 experiments, $K_{\mathrm{D},\mathrm{a}\mathrm{p}\mathrm{p}}$ values for Palbociclib-FITC and test compounds are reported.

In subsequent TR-FRET ligand displacement assays, inhibitor $K_{\mathrm{D}}$ and $K_{\mathrm{D},\mathrm{a}\mathrm{p}\mathrm{p}}$ values have been calculated using Cheng-Prusoff principles, outlined in Equation 2 below:

$$K_{D}or{K}_{D,app}=\frac{{IC}_{50}}{1+\frac{\left[S\right]}{K_X}}$$ (2)

Where ${\mathrm{I}\mathrm{C}}_{50}$ is the measured ${\mathrm{I}\mathrm{C}}_{50}$ value, $\left[\mathrm{S}\right]$ is the concentration of fluorescent tracer, and $K_{\mathrm{X}}$ is the $K_{\mathrm{D}}$ or $K_{\mathrm{D},\mathrm{a}\mathrm{p}\mathrm{p}}$ of the fluorescent tracer for a given condition [4202581].

TR-FRET ligand displacement assays.

The following conditions have been employed: (i) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2.5 nM His₆-CRBN/DDB1 complex, and 50 nM Thalidomide-FITC in assay buffer, (ii) 2.5 nM CoraFluor-1-labeled His₆-CRBN/DDB1 complex, 200 nM Thalidomide-FITC in assay buffer, (iii) 4 nM GST-BRD4(BD1) or GST-BRD4(BD2), 4 nM CoraFluor-1-labeled anti-GST V_HH, and 20 nM JQ1-FITC in assay buffer, (iv) 2.5 nM CoraFluor-1-labeled anti-His₆ antibody, 5 nM His₆-FKBP12, and 60 nM SLF-FITC in assay buffer, (v) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2 nM His₆-GST-CDK4/GST-CD3, and 30 nM Palbociclib-FITC in assay buffer, (vi) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2 nM His₆-CDK6/GST-CD1, and 200 nM Palbociclib-FITC in assay buffer.

In all cases, test compounds were added in serial dilution (1:2 titration, 15-point, c_max = 10 μM) using a HP D300 digital dispenser and allowed to equilibrate for 1 h at room temperature before TR-FRET measurements were taken. For conditions (i-iii), the assay floor was determined from wells treated with 10 μM dBET6. For condition (iv), the assay floor was determined from wells treated with 10 μM dFKBP-1. For conditions (v-vi), the assay floor was determined from wells treated with 10 μM Palbociclib. The assay ceiling (top) was defined via a no-inhibitor control. Data were background-corrected, normalized and fitted to a four-parameter dose-response model [log(inhibitor vs. response – Variable slope (four parameters)] in Prism 9, with constraints of Top = 1 and Bottom = 0.

Determination of the cooperativity constant (α) for PROTAC-mediated ternary complexes of CRBN/DDB1 with recombinant BRD4 bromodomains and FKBP12.

Assays were performed in low-volume, white 384-well plates (ProxiPlate-384 Plus, PerkinElmer) in 10 μL assay volume. The following assay conditions have been employed: (i) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2.5 nM His₆-CRBN/DDB1 complex, 50 nM Thalidomide-FITC, and 1 μM GST-BRD4(BD1) in assay buffer, (ii) 1 nM CoraFluor-1-labeled anti-His₆ antibody, 2.5 nM His₆-CRBN/DDB1 complex, 50 nM Thalidomide-FITC, and 1 μM GST-BRD4(BD2) in assay buffer, (iii) 2.5 nM CoraFluor-1-labeled His₆-CRBN/DDB1 complex, 200 nM Thalidomide-FITC, and 1 μM His₆-FKBP12 in assay buffer.

In all cases, test compounds were added in serial dilution (1:4 titration, 7-point, c_max = 1 μM) using a HP D300 digital dispenser and allowed to equilibrate for 1 h at room temperature before TR-FRET measurements were taken. The assay floor was determined from wells treated with 10 μM dBET6, and the assay ceiling (top) was defined via a no-inhibitor control. Data were background-corrected, normalized and fitted to a four-parameter dose-response model [log(inhibitor vs. response – Variable slope (four parameters)] in Prism 9, with constraints of Top = 1 and Bottom = 0.

The cooperativity constant ($\mathrm{\alpha }$) was calculated via Equation 3 below:

$$\alpha =\frac{K_{D,binary}}{K_{D,ternary}}$$ (3)

Where $K_{\mathrm{D},\mathrm{b}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{r}\mathrm{y}}$ and $K_{\mathrm{D},\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{n}\mathrm{a}\mathrm{r}\mathrm{y}}$ are the respective inhibitor $K_{\mathrm{D}}$ values in the absence and presence of 1 μM recombinant target protein (e.g. GST-BRD4(BD1), GST-BRD4(BD2), or His₆-FKBP12).

TR-FRET assay to quantify endogenous BRD4 abundance in MDA-MB-231 cell extract after degrader treatment.

Assays were run as previously reported.⁴³ MDA-MB-231 cells were seeded into 96-well plates (Corning 3903 or Corning 3904) at a density of 20,000 cells/well in 100 μL cell culture medium and allowed to adhere overnight. A HP D300 digital dispenser was used to dispense serial dilutions of test compounds (1:2 titration, 12-point, c_max = 1 μM) normalized to 0.1% DMSO. Cells were incubated for 5 h at 37°C and 5% CO₂ then media was replaced with pre-warmed cell culture medium (150 μL/well) and residual test compound was washed out for 1 h at 37°C and 5% CO₂. After, media was aspirated and cells were washed with PBS (200 μL/well), followed by the addition of ice-cold lysis buffer (60 μL/well; 25 mM HEPES, 150 mM NaCl, 0.2% v/v Triton X-100, 0.02% TWEEN-20, pH 7.5 supplemented with 1 mM AEBSF hydrochloride (Combi-Blocks SS-7834), 250 U Benzonase (Sigma E1014), and 2 mM DTT). The plate was shaken at 1,000 rpm on an orbital shaker (Boekel Scientific Jitterbug, model 130000) for 10 min.

After, 10 μL of 7x control detection mix (7 nM CoraFluor-1-labeled anti-rabbit Nano-Secondary, 700 nM JQ1-FITC final concentrations, prepared in dilution buffer; 25 mM HEPES, 150 mM NaCl, 0.005% TWEEN-20, pH 7.5) was added to four of the eight DMSO-treated control wells, and 10 μL of complete detection mix (3.5 nM rabbit anti-BRD4 antibody, 7 nM CoraFluor-1-labeled anti-rabbit Nano-Secondary, 700 nM JQ1-FITC final concentrations, prepared in dilution buffer) was added to all other wells and allowed to equilibrate for 1 h. The plate was centrifuged at 2,000 × g for 1 min then lysate was transferred to a 384-well plate (30 μL × 2 TR-FRET replicates) using an adjustable electronic multichannel pipette (Matrix Equalizer, ThermoFisher 2231), and TR-FRET measurements were taken.

TR-FRET ratios were background-subtracted from the four (eight TR-FRET replicates) DMSO-treated wells containing control detection mix. The average TR-FRET intensity was then normalized to the four (eight TR-FRET replicates) DMSO-treated wells containing complete detection mix for each biological replicate, then data were fitted to a four-parameter dose-response model [log(inhibitor vs. response – Variable slope (four parameters)] in Prism 9, with constraints of Top = 1.

TR-FRET assay to quantify BRD4(BD1)-GFP and BRD4(BD2)-GFP in lysates after degrader treatment.

2.0×10⁵ HEK293T cells stably expressing either BRD4(BD1)-GFP or BRD4(BD2)-GFP cells²² were seeded in DMEM supplemented with 10% FBS and 1× penicillin-streptomycin and incubated at 37 °C, 5% CO₂ for 30 min, then treated with compounds of interest and incubated at 37 °C, 5% CO₂ for the indicated time prior to collection and lysis. All compounds were dissolved in DMSO, and the final DMSO concentration after addition of the compound to the cells did not exceed 0.2% v/v. After compound treatment, cells were collected, washed with PBS (500 μL), flash-frozen, and stored at −80 °C until further use.

Frozen cell pellets (~200,000 cells per pellet) were thawed on ice and to each pellet was added 200 uL ice-cold complete lysis buffer. The cell pellets were briefly vortexed and lysis was allowed to proceed at room temperature for 10 min. After, lysates were transferred to a 384-well plate (30 μL × 2 TR-FRET replicates), followed by the addition of 5 μL of 7x detection mix (28 nM CoraFluor-1-labeled anti-GFP V_HH in dilution buffer). The plate was allowed to equilibrate for 1 h at room temperature before TR-FRET measurements were taken.

TR-FRET ratios were background-subtracted from wells containing 0.5 mg/mL BSA + detection mix. TR-FRET ratios were then normalized to the DMSO-treated wells for each biological replicate. Data were fitted to a four-parameter dose-response model [log(inhibitor vs. response – Variable slope (four parameters)] in Prism 9, with constraints of Top = 1.

Determination of correlation between assays.

Pearson correlation coefficients (r_p) and p-values (two-tailed) were calculated using Prism 9 from measured pK_D and DC₅₀ values.

Off-rate measurements for degrader-mediated ternary complexes between His6-CRBN/DDB1 and individual GST-tagged bromodomains.

Assays were performed in low-volume, white 384-well plates (ProxiPlate-384 Plus, PerkinElmer) in 10 μL assay volume. A master solution of His₆-CRBN/DDB1 (10 nM), AF488-labeled anti-His₆ antibody (10 nM; Ab18184), CoraFluor-1-labeled anti-GST antibody (5 nM; Ab19256), and GST-BRD4(BD1) or GST-BRD4(BD2) (10 nM) was prepared and divided into 59 uL aliquots. After, 1 μL of 60x compound stock (in DMSO) was added to achieve the following concentrations: for GST-BRD4(BD1), 10 nM dBET6, 10 nM JQ1-YcQ, 10 nM JQ1-YcN, 20 nM JQ1-AcQ, 10 nM JQ1-AcN, 50 nM JQ1-ScQ, 10 nM JQ1-ScN, 120 nM JQ1-GcQ, 20 nM JQ1-GcN; for GST-BRD4(BD2), 20 nM dBET6, 275 nM JQ1-YcQ, 150 nM JQ1-YcN, 75 nM JQ1-AcQ, 20 nM JQ1-AcN, 150 nM JQ1-ScQ, 25 nM JQ1-ScN, 540 nM JQ1-GcQ, 90 nM JQ1-GcN. These concentrations were chosen explicitly as they are equal to or near the K_{D, CRBN}[ternary] values for each degrader (see Supplementary Table 1).

Solutions were allowed to equilibrate for 1 h at room temperature then 3 × 10 uL technical replicates were plated into wells of the 384-well plate before initial t = 0 TR-FRET measurements were taken. Following the addition of JQ1 (10 μM) or Lenalidomide-5′-NH₂ (50 μM), the time-dependent change in TR-FRET signal was recorded over 10 min in 30 s intervals using fixed detector gain settings. Data were background-corrected from wells containing no compound, normalized to the t = 0 TR-FRET intensity on a per-compound basis, and fitted to a one phase decay model using Prism 9.

Generation of off-target degradation reporter cell lines.

U2OS cells stably expressing the 23-aa zinc finger region of IKZF3, SALL4, ZFP91, ZNF276, ZNF654, and ZNF827 were kindly provided by the Choudhary Lab.⁵⁷ The procedures of molecular cloning and cell line generation were adapted from Sievers and coworkers.⁵⁸ The Cilantro2-GSPT1 construct was derived from the degradation reporter vector Cilantro2 and pcDNA3.1-GSPT1. The GSPT1 insert was amplified by overhang PCR using Primer 1 and Primer 2 to add the BsmBI restriction sites. The vector and insert were digested by BsmBI, ligated by Quick Ligation Kit, and transformed into NEB Stable cells. The sequence was validated by Sanger sequencing (Quintara Bio).

For virus production, HEK293T cells in a 10 cm plate were transfected by pMD2.G (6 μg), psPAX2 (4 μg) and Cilantro2-GSPT1 (10 μg) using TransIT-Pro reagent (20 μL) following the manufacturer’s guidelines. Lentiviruses were collected 48 h after transfection, spun down (500 × g, 5 min, 24 °C) and filtered with a 0.45 μm PES filter. For transduction, HEK293T cells in 1 mL DMEM+/− containing 10 μg/mL polybrene were seeded into 6-well plates containing the virus (500 μL) and incubated for 48 h and then selected by 2 μg/mL puromycin for 7 days. The GSPT1 reporter cell line was validated by a test degradation assay using the GSPT1 degrader CC-885.¹³

Off-target degradation assay by flow cytometry.

Cells were seeded in 96-well plates (3–4×10⁴ cells per well) 1 h before compound treatment. DMSO or each compound at each concentration was added into three separate wells (final DMSO concentration: 0.2%) and incubated at 37 °C, 5% CO₂ for 24 h. The cells were detached by trypsinization in 0.25% trypsin without phenol red (diluted 2× by PBS), followed by resuspension in DMEM +/− without phenol red. All plates were analyzed by FACSymphony (BD Biosciences) on the high throughput sampler.

Signal from at least 2,000 events per well was acquired, and the fluorescence signals of FITC and PE Texas Red were measured. GFP/mCherry ratios were determined by dividing the geometric mean of FITC signal by that of PE Texas Red signal on BD FACSDiva v9.1. For each condition, the arithmetic mean of ratios among triplicates was calculated and then normalized to that of the DMSO ratios to derive the GFP/mCherry (compound/DMSO) value.

Computational modeling.

Computational models were generated in MOE version 2022.02. The IMiD ligand from the indicated crystal structure was adapted to the indicated dipeptide and the adapted complex was subjected to energy minimization with Amber10:EHT force field followed by preparation with Protonate 3D.

Global quantitative proteomics sample preparation.

The proteomics sample preparation protocol was adapted from Donovan and coworkers⁵⁵ and off-line fractionation protocol was adapted from Batth and coworkers⁷⁰. Global proteomics samples were prepared in biological triplicate for each condition. 2.0×10⁶ Jurkat cells were seeded in 12-well plates and incubated at 37 °C, 5% CO₂ for 1 h. Small molecules of interest were then added, and cells were incubated at 37 °C, 5% CO₂ for 4 h. Cells were collected according to the described general procedure, lysed by probe sonication (5 sec on, 3 sec off, 15 sec in total, 11% amplitude) in lysis buffer (8 M urea, 50 mM NaCl, 50 mM HEPES, 1x protease/phosphatase inhibitor cocktail, pH 8.2), and cleared by centrifugation (21,000 × g, 4 °C, 10 min). After protein quantification by BCA protein assay, the lysates were diluted to 0.89 mg/mL with the lysis buffer. The diluted lysates (110 μL) were reduced by the addition of dithiothreitol (final concentration: 5 mM) at 24 °C for 30 min, then alkylated by addition of iodoacetamide (final concentration: 15 mM) and incubation in the dark at 24 °C for 30 min. The proteins in the samples were precipitated using methanol-chloroform precipitation. In brief, four volumes of chilled methanol, one volume of chilled chloroform, and three volumes of water were added sequentially to the lysates. The mixture was vortexed and centrifugated at 14,000 × g, 5 min, 4 °C, and the supernatant was aspirated. The protein pellet was washed with three volumes of chilled methanol and centrifugated at 14,000 × g, 5 min, 4 °C, and the resulting precipitated protein was air-dried.

The protein was dissolved in 25 μL 4M urea, 50 mM HEPES, pH 7.4, followed by the addition of 75 μL 200 mM HEPPS, pH 8.0. Lys-C (2.0 μg) was added to the mixture, and the digestion was allowed to proceed at 30 °C for 4 h without rotation. The samples were diluted by the addition of 200 mM HEPPS, pH 8.0 (100 μL), and further digested by trypsin (4.0 μg) at 37 °C for 18 h without rotation. Approximately 50 μg of peptides from each digested sample (107 μL) was taken for labeling with TMTpro 16-plex reagent (17 μL) at 24 °C for 1 h. TMT labeling was quenched by incubation with 5% hydroxylamine (7 μL) at 24 °C for 15 min. The TMT-labeled samples were combined and dried by a vacufuge. The combined, dried sample was then resuspended in 300 μL 0.1% formic acid and was subjected to desalting with Pierce Peptide desalting spin columns (Thermo Scientific, 89852). The desalted sample was combined, dried by a vacufuge, and then resuspended in 200 μL 5% acetonitrile in aqueous NH₄HCO₃ solution (final concentration: 10 mM). The sample was offline fractionated into 80 fractions by high pH reverse-phase HPLC (Agilent LC1260) through an AerisPpeptide XB-C18 column (Phenomenex) with mobile phase A containing 5% acetonitrile and aqueous NH₄HCO₃ solution (final concentration: 10 mM), and mobile phase B containing 90% acetonitrile and aqueous NH₄HCO₃ solution (final concentration: 1 mM). Both mobile phase solutions were adjusted to be at pH 8.0 with ammonium hydroxide and formic acid. Fractions were collected using a fraction collector (Agilent) in a 96-deep well plate (Agilent, 5043–9305). Samples were initially loaded onto the column at 1 mL/min for 4 min, after which the fractionation gradient commenced as follows: 1% B to 27% B in 50 min, 60% B in 4 min, and ramped to 70% B in 2 min. At this point, fraction collection was halted, and the gradient was held at 70% B for 5 min before being ramped back to 1% B, where the column was then washed and equilibrated. The 80 resulting fractions were transferred into low protein binding tubes and dried by a vacufuge. The dried fractions were resuspended in 20 μL 0.1% FA and pooled in a non-contiguous manner into 20 fractions (final volume: 80 μL), followed by subsequent mass spectrometry analysis.

Proteomics mass spectrometry acquisition procedures for protein level quantitation.

The resuspended samples (2 μL) were run on an Orbitrap Eclipse Tribrid Mass Spectrometer coupled with a Vanquish Neo HPLC system (Thermo) at the Harvard Center for Mass Spectrometry. The peptides were first trapped on a trapping cartridge (300 μm x 5 mm PepMap^™ Neo C18 Trap Cartridge, Thermo) prior to separation on a silica-chip-based micropillar column (μPAC, C18 pillar surface, 50 cm bed, Thermo). The column oven temperature was maintained at 35 °C. Peptides were eluted using a multi-step gradient at a flow rate of 300 nL/min over 180 min. The mobile phase A and weak wash liquid was water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid, and strong wash liquid was 80% acetonitrile with 0.1% formic acid. The mobile phase gradient consists of a linear 125 min gradient from 2% to 20% mobile phase B, followed by a 24 min increase to 35% B, a further 10 min increase to 95% B, a 20 min plateau phase at 95% B. The autosampler temperature was 7 °C. The column “Fast equilibration” was enabled. The Orbitrap Eclipse MS was operated in DDA mode with 3 s cycle time. The electrospray ionization was in positive mode with voltage of 2.1 kV and the capillary temperature at 275 °C. Dynamic exclusion was enabled with a mass tolerance of 10 ppm and exclusion duration of 60 s. Full scan was performed in the range of 400–1,600 m/z at a resolution of 120,000, RF lens 30%, normalized AGC target 200% and maximum injection time set to auto. Precursors were isolated in a window of 0.7 m/z and charge states of 2–6. MS2 fragmentation was performed by HCD using normalized collision energy of 38% at the resolution of 50,000. The normalized AGC target was set to 250%, and a maximum ion injection time set to 200 ms.

Mass spectrometry data analysis for protein level quantitation.

Analysis was performed in Thermo Scientific Proteome Discoverer version 2.4.1.15. The raw data were searched against SwissProt human (Homo sapiens) protein database (21 February 2019; 20,355 total entries) and contaminant proteins using the Sequest HT algorithm. Searches were performed with the following guidelines: spectra with a signal-to-noise ratio greater than 1.5; mass tolerance of 20 ppm for the precursor ions and 0.02 Da for fragment ions; full trypsin digestion; 2 missed cleavages; variable oxidation on methionine residues (+15.995 Da); static carboxyamidomethylation of cysteine residues (+57.021 Da); static TMTpro labeling (+304.207 Da) at lysine residues and N-termini. The TMT reporter ions were quantified using the Reporter Ions Quantifier node and normalized to the total peptide amount. Peptide spectral matches (PSMs) were filtered using a 1% false discovery rate (FDR) using Percolator. PSMs were filtered to PSMs in only one protein group with an isolation interference under 70%. For the obtained proteins, the data were filtered to include master and master candidate proteins with high protein FDR confidence, at least 2 unique peptides, and exclude all contaminant proteins. The abundance ratios and their associated p-values were obtained from Proteome Discoverer (p-values were calculated by one-way ANOVA with TukeyHSD post-hoc test).

Liver S9 fraction assay.

Assay was designed and performed by Pharmaron. In brief, a master solution containing 100 mM phosphate buffer, 5 mM MgCl₂ solution, 1 mg/mL Liver S9 fractions, and 50 μg/mL Alamethacin in ultra-pure water was prepared and pre-warmed at 37°C. To each well of a plate, 318 μL of the master mix solution, 40 μL of 10 mM nicotinamide adenine dinucleotide phosphate (NADPH), and 40 μL of 20 mM uridine diphosphate glucuronic acid (UDPGA) were added (final concentration: NADPH, 1 mM; UDPGA, 2 mM). For negative control, NADPH and UDPGA solutions were substituted with 80 μL of ultra-pure water. The reaction was initiated by adding 2 μL of 2 mM test compound or 2 μL of 400 μM Diclofenac to the wells (final concentration: test compound, 10 μM; Diclofenac, 2 μM), with each compound tested in duplicate. At designated time points (t = 0, 5, 15, 30, 60 min), 50 μL samples were collected from the reaction solution and quenched with cold acetonitrile containing internal standards. After centrifugation, the supernatant was subjected to LC-MS/MS analysis. Data was processed using Microsoft Excel, where peak area ratios were determined from extracted ion chromatograms. The in vitro half-life and intrinsic clearance were calculated based on the slope value ($k$) derived from linear regression of the natural logarithm of the remaining drug percentage against the incubation time curve.

The in vitro half-life (${\mathrm{t}}_{1/2}$) were determined from the slope value:

$$invitro{t}_{1/2}=-\frac{0.693}{k}$$

Conversion of the in vitro ${\mathrm{t}}_{1/2}$ into the in vitro intrinsic clearance (in vitro ${\mathrm{C}\mathrm{L}}_{\mathrm{i}\mathrm{n}\mathrm{t}}$, in μL/min/mg protein) was performed using the following equation (mean of duplicate determinations):

$$invitro{CL}_{int}=\frac{0.693}{t_{1/2}}\times \frac{volumeofincubation\left(\mathrm{\mu }\mathrm{L}\right)}{amountofproteins\left(mg\right)}$$

Caco-2 trans-well assay.

Assay was designed and performed by Pharmaron. In brief, Caco-2 cells seeded into a 96-well HTS Transwell plate were washed twice with pre-warmed HBSS (10 mM HEPES, pH 7.4), and incubated at 37 °C for 30 min. The stock solutions of the test compounds (1 mM in DMSO) were diluted with HBSS (10 mM HEPES, pH 7.4) to obtain 5 μM working solutions. The final concentration of DMSO in the incubation system was 0.5%. To determine the rate of drug transport in the apical (AP) to basolateral (BL) direction, 75 μL of 5 μM working solution of each compound was added to the Transwell insert (apical compartment), and the wells in the receiver plate (basolateral compartment) were filled with 235 μL of HBSS (10 mM HEPES, pH 7.4). To determine the rate of drug transport in the basolateral to apical direction, 235 μL of 5 μM working solution of each compound was added to the receiver plate wells (basolateral compartment), and then the Transwell inserts (apical compartment) were filled with 75 μL of HBSS (10 mM HEPES, pH 7.4). Samples at t = 0 were prepared by transferring 50 μL of 5 μM working solution to wells of the 96-deep well plate, followed by the addition of 200 μL cold methanol containing appropriate internal standards (100 nM alprazolam, 200 nM labetalol, 200nM caffeine, and 200 nM diclofenac). After the plates were incubated at 37 °C for 2 h, 50 μL samples from donor sides (apical compartment for AP→BL flux, and basolateral compartment for BL→AP) and receiver sides (basolateral compartment for AP→BL flux, and apical compartment for BL→AP) were transferred to wells of a new 96-well plate, followed by the addition of 4 volume of cold methanol containing appropriate internal standards (100 nM alprazolam, 200 nM labetalol, 200nM caffeine, and 200 nM diclofenac). Samples were then vortexed for 5 minutes and centrifuged at 3,220 × g for 40 min. An aliquot of 100 μL of the supernatant was mixed with an appropriate volume of ultra-pure water before LC-MS/MS analysis.

The apparent permeability coefficient (${\mathrm{P}}_{\mathrm{a}\mathrm{p}\mathrm{p}}$), in units of centimeter per second, were calculated for Caco-2 drug transport assays using the following equation:

$$P_{app}=\frac{(V_A\times {\left[drug\right]}_{acceptor})}{(Area\times Time\times {\left[drug\right]}_{initial,donor}}$$

Where ${\mathrm{V}}_{\mathrm{A}}$ is the volume (mL) in the acceptor well, Area is the surface area of the membrane (0.143 cm² for Transwell-96 Well Permeable Supports), and time is the total transport time in seconds.

The efflux ratio will be determined using the following equation:

$$Effluxratio=\frac{P_{app(B–A)}}{P_{app(A–B)}}$$

Where ${\mathrm{P}}_{\mathrm{a}\mathrm{p}\mathrm{p}(\mathrm{B}–\mathrm{A})}$ indicates the apparent permeability coefficient in basolateral to apical direction, and ${\mathrm{P}}_{\mathrm{a}\mathrm{p}\mathrm{p}(\mathrm{A}–\mathrm{B})}$ indicates the apparent permeability coefficient in apical to basolateral direction.

The recovery can be determined using the following equation:

$$Recovery\left(\mathrm{\%}\right)=\frac{(V_A\times {\left[drug\right]}_{acceptor}+V_D\times {\left[drug\right]}_{donor})}{(V_D\times {\left[drug\right]}_{initial,donor})}$$

Where ${\mathrm{V}}_{\mathrm{A}}$ is the volume (mL) in the acceptor well (0.235 mL for AP→BL flux, and 0.075 mL for BL→AP), ${\mathrm{V}}_{\mathrm{D}}$ is the volume (mL) in the donor well (0.075 mL for AP→BL flux, and 0.235 mL for BL→AP).

Data and code availability

• All data are available in the main text or the Supplementary Information. Proteomics data have been deposited to the PRIDE repository with the dataset identifier PXD041494. Source data are provided with this paper.

• This paper does not report any original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contacts upon request.

Supplementary Material

Supp Material

Supplementary Fig. 1. TR-FRET ligand displacement assay platform to quantitatively measure small molecule target engagement with His₆-CRBN/DDB1, BD1 and BD2 domains of BRD4. (a) Structure of Thal-FITC, a tracer used to assess the engagement with CRBN/DDB1. (b) Saturation binding of Thal-FITC to His₆-CRBN/DDB1 labeled with CoraFluor-1-anti-His₆ conjugates. Data are presented as mean ± SD (n = 3 technical replicates). (c) Structures of JQ1-cQ and JQ1-cN. (d) Comparison between monofunctional cQ- and cN-cyclimids regarding pK_D values (equilibrium dissociation constants; K_D) against His₆-CRBN/DDB1. The cN-cyclimids generally bind to CRBN/DDB1 more tightly than their cQ counterparts. (n = 3 technical replicates, error bars, which fall within symbols, represent mean ± SD). (e) Structures of pomalidomide (POM), POM-cN, Boc-FcQ, and Boc-FcN. (f) Dose-titration of the indicated monofunctional compounds in TR-FRET ligand displacement assays with His₆-CRBN/DDB1 and the determined K_D values of the indicated compounds against the His₆-CRBN/DDB1 complex. Data are presented as mean ± SD (n = 3 technical replicates). (g) Structural models of POM, POM-cN, Boc-FcQ, and Boc-FcN, built from the crystal structure of the CRBN/DDB1 complex bound to pomalidomide (PDB: 4CI3). Zoom-in of the thalidomide-binding pocket of CRBN with POM-cN (pink), and overlays of POM-cN with POM (orange), Boc-FcQ (green), or Boc-FcN (blue). (h) Sketch of POM, POM-cN, Boc-FcQ, and Boc-FcN and their interactions with residues in the thalidomide-binding pocket of CRBN. POM-cN is incapable of forming hydrogen bonds with Trp 382 of CRBN, in contrast to POM, Boc-FcQ, and Boc-FcN.

Supplementary Fig. 2. TR-FRET ligand displacement assay platform with GST-BRD4(BD1) and GST-BRD4(BD2). (a) Structure of JQ1-FITC, a tracer used to assess the engagement with BD1 and BD2 domains of BRD4. (b) Schematic of the TR-FRET ligand displacement assay using GST-BRD4(BD1) and GST-BRD4(BD2) with CoraFluor-1-labeled anti-GST nanobody. (c) Saturation binding of JQ1-FITC to GST-BRD4(BD1) labeled with CoraFluor-1-anti-GST nanobody. (d) Saturation binding of JQ1-FITC to GST-BRD4(BD2) labeled with CoraFluor-1-anti-GST nanobody. Data in (c) and (d) are presented as mean ± SD (n = 3 technical replicates). (e) Selectivity parameter between BD1 and BD2 degradation of thalidomide-based degraders (black), cQ-cyclimid (red), and cN-cyclimid (blue) BRD4 degraders. Cyclimid degraders display higher biochemical selectivity than dBET6. (f) Dose-titration of JQ1-YcQ in TR-FRET ligand displacement assays with His₆-CRBN/DDB1 in the absence of BRD4, or in the presence of a near-saturating concentration of either the BD1 or BD2 domain of BRD4 (1 μM). The obtained dissociation constants (K_{D, CRBN}[binary] or K_{D, CRBN}[ternary]) of JQ1-YcQ against the His₆-CRBN/DDB1 complex indicated a 60-fold difference between K_{D, CRBN}[ternary, BD1] and K_{D, CRBN}[ternary, BD2], i.e., Alpha(BD1) and Alpha(BD2). Data are presented as mean ± SD (n = 3 technical replicates).

Supplementary Fig. 3. Model of the cyclimids with CRBN. (a) Structural model of Ac-FcQ (yellow) and the other 19 cQ-based cyclimids (yellow), Ac-FcN (green), and thalidomide analog (orange) in the thalidomide-binding pocket of CRBN (PDB: 6BOY). (b) Structural model of JQ1-FcQ (red), dBET6 (gray), and iberdomide (green) in the ternary complex with CRBN (gray) and BRD4 (blue), built from the crystal structure of the CRBN/DDB1–BRD4(BD1) complex bound to dBET6 (PDB: 6BOY) and CRBN/DDB1 bound to iberdomide with DDB1 in the linear conformation (PDB: 8D7U). Left: Side view. Right: Top view.

Supplementary Fig. 4. TR-FRET assay platform to quantitatively measure BRD4(BD1)-GFP and BRD4(BD2)-GFP levels in whole cell extracts. (a) Schematic showing the TR-FRET assay design for quantification of BRD4(BD1) and BRD4(BD2) levels in cell lysates. (b) TR-FRET-based quantification of BRD4(BD1)-GFP and BRD4(BD2)-GFP in serially diluted cell lysates from HEK293T cells stably expressing BRD4(BD1)-GFP or BRD4(BD2)-GFP. (n = 2 technical replicates, error bars, which fall within symbols, represent mean ± SD).

Supplementary Fig. 5. Cellular degradation assays of cyclimid-based BRD4 degraders. (a) BRD4 levels in MDA-MB-231 cell lysate after 5 h treatment with selected cyclimid BRD4 degraders were measured by TR-FRET assay. HSC, high signal control; LSC, low signal control. Data are presented as mean ± SD (n = 2 biologically independent samples). (b–d) Correlation between pDC₅₀ values of the selected cyclimid degraders for endogenous BRD4 degradation and their pK_D[binary] values against (b) CRBN (r_p = 0.54), (c) BRD4(BD1) (r_p = −0.04), or (d) BRD4(BD2) (r_p = −0.02). (e–g) Western blots of BRD4 levels after treatment of HEK293T cells for 4 h with (e) 10 μM, (f) 1 μM, or (g) 100 nM dBET6 or the twenty cN-cyclimid BRD4 degraders. (h) Western blot of BRD4 levels after treatment of HEK293T cells with dBET6 and selected cN-cyclimid degraders for 4 h over a 1–100 nM dose-response range. All western blot data are representative of at least three independent replicates. For uncropped western blot images, see Supplementary Figure 3.

Supplementary Fig. 6. Ligand-dependent dynamics of CRBN/DDB1 dissociation from cyclimid degrader-induced ternary complexes with individual BRD4 bromodomains. (a) Schematic illustration of TR-FRET assay system employed to monitor His₆-CRBN/DDB1 dissociation from the degrader-mediated ternary complex. (b) Kinetic measurements of His₆-CRBN/DDB1 dissociation from the ternary complex induced by the selected cyclimids or dBET6. Data are presented as mean ± SD (n = 2 technical replicates) and fitted to a one-phase decay model in Prism 9. (c) Half-lives (t_1/2) of the His₆-CRBN/DDB1–BRD4(BD1) ternary complex mediated by the indicated cyclimid degraders and dBET6. Error bars represent mean ± SD (n = 2 technical replicates).

Supplementary Fig. 7. TR-FRET ligand displacement assay platform to measure small molecule target engagement with His₆-FKBP12 and His₆-CRBN/DDB1. (a) Structure of SLF-FITC, a tracer used to assess the engagement with FKBP12. (b) Saturation binding of SLF-FITC to His₆-FKBP12 labeled with CoraFluor-1-anti-His₆ conjugate. (c) Saturation binding of Thal-FITC to CoraFluor-1-labeled His₆-CRBN/DDB1. Data in (b) and (c) are presented as mean ± SD (n = 3 technical replicates). (d) Schematic of the TR-FRET assay principle for determining K_D[binary] and K_D[ternary] against His₆-CRBN/DDB1, in the absence or the presence of FKBP12. (e) Intact MS measurement of His₆-FKBP12.

Supplementary Fig. 8. Evaluation of dFKBP-cyclimids in targeted protein degradation of FKBP12. (a) FKBP12 levels in HEK293T cells treated with the indicated degrader in the absence or the presence of a proteasome inhibitor MG132 (10 μM), a neddylation inhibitor MLN4924 (1 μM), or excess lenalidomide (100 μM), and incubated for 24 h. Rescue of FKBP12 degradation was observed when cells were pre-treated with MG132, MLN4924, or lenalidomide. (b–c) FKBP12 levels in HEK293T cells treated with serially diluted degraders for 24 h. Quantification of FKBP12 levels were calculated relative to DMSO control, and the plots shown were used to calculate the tabulated half-maximal degradation concentration (DC₅₀) values for each degrader. Data in (b) and (c) are presented as mean ± SD (n = 3 biologically independent samples). (d) Corresponding western blots of FKBP12 levels after treatment of HEK293T cells for 24 h with dFKBP-1 or each of the selected dFKBP-cyclimids over a dose-response range of 1.28 nM and 20 μM. All western blot data are representative of at least three independent replicates. For uncropped western blot images, see Supplementary Figure 4.

Supplementary Fig. 9. Evaluation of dCDK-cyclimids in targeted protein degradation of CDK4 and CDK6. (a) Western blot of CDK4 and CDK6 levels after treatment of Jurkat cells for 4 h with 100 nM dCDK-1 or each dCDK-cyclimids. (b) Western blot of CDK4 and CDK6 levels after treatment of Jurkat cells for 4 h with 10 μM dCDK-1 or each dCDK-cyclimids. (c) CDK4 and CDK6 levels in Jurkat cells treated with the indicated degrader in the absence or the presence of a proteasome inhibitor MG132 (10 μM), a neddylation inhibitor MLN4924 (1 μM), or excess lenalidomide (100 μM), and incubated for 4 h. Rescue of CDK4 and CDK6 degradation was observed when cells were pre-treated with MG132, MLN4924, or lenalidomide. (d–e) Corresponding western blots of (d) CDK4 and (e) CDK6 after treatment for 4 h with dCDK-1, dCDK-PcQ, or dCDK-PcN over a dose-response range of 0.64 nM and 10 μM. Quantification of CDK4 and CDK6 levels were calculated relative to DMSO control to obtain the plots in Figure 5h–j. All western blot data are representative of at least three independent replicates. For uncropped western blot images, see Supplementary Figure 5.

Supplementary Fig. 10. The cyclimids have minimal thalidomide analog-sensitive off-target degradation and possess comparable metabolic stability and cell permeability properties to the thalidomide derivatives. (a) Degradation of IKZF3 ZF degrons in cells by pomalidomide, CC-885, dBET6, and JQ1-FcN over a dose-response range of 32 nM to 20 μM. U2OS cells stably expressing IKZF3 ZF degrons fused to GFP were treated with degraders followed by flow cytometry to assess IKZF3 ZF degradation. (b) Degradation of GSPT1 in cells using pomalidomide, CC-885, dBET6, and JQ1-FcN. HEK293T cells stably expressing GSPT1-GFP were treated with the indicated compounds followed by flow cytometry to assess GSPT1 degradation. Data in (a)–(b) are presented as mean (n = 3 biologically independent samples). (c–e) Quantitative proteomics of Jurkat cells after 4 h treatment with 0.1 μM of (c) dCDK-1, (d) dCDK-PcQ, or (e) dCDK-PcN. P-values for the abundance ratios were calculated by one-way ANOVA with TukeyHSD post-hoc test. (f) Metabolic stability of the selected monofunctional cyclimids, thalidomide, and lenalidomide, as well as heterobifunctional degraders bearing the cyclimids or the thalidomide analogs (dBET6), against liver S9 fractions. t_1/2, in vitro half-life; CL_int, in vitro intrinsic clearance. When the remaining compound (%) at 60 minutes is greater than 85%, the t_1/2 and CL_int were reported as “>255.85” and “<2.71”, respectively. (g) in vitro permeability of the selected monofunctional cyclimids and lenalidomide, as well as heterobifunctional degraders bearing the cyclimids or the thalidomide analogs (dBET6), in a Caco-2 trans-well assay. P_{app (A-B)}, the apparent permeability coefficient in apical (AP) to basolateral (BL) direction; P_{app (B-A)}, the apparent permeability coefficient in basolateral to apical direction. The efflux ratio was determined with the following equation: efflux ratio = P_app(B-A)/P_app(A-B). The recovery can be determined using the following equation, where V_A is the volume (mL) in the acceptor well (0.235 mL for AP→BL flux, and 0.075 mL for BL→AP flux) and V_D is the volume (mL) in the donor well (0.075 mL for AP→BL flux, and 0.235 mL for BL→AP): Recovery% = (V_A×[drug]_acceptor+V_D×[drug]_donor)/(V_D×[drug]_{initial, donor}).

Supplementary and Supplementary Information are available for this paper.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BRD4 (E2A7X)	Cell Signaling Technology	Cat# 13440S; RRID: AB_2687578
Vinculin (V284)	Bio-Rad	Cat# MCA465GA; RRID: AB_2214389
FKBP12	Abcam	Cat# ab24373; RRID: AB_732383
FKBP51	Abcam	Cat# ab126715; RRID: AB_11128515
FKBP52	Abcam	Cat# ab129097; RRID: AB_11150603
CDK4	Abcam	Cat# ab199728; RRID: AB_2920615
CDK6 (D4S8S)	Cell Signaling Technology	Cat# 13331; RRID:AB_2721897
β-actin (C4)	Santa Cruz Biotechnology	Cat# sc-47778; RRID: AB_626632
Anti-mouse-HRP	Rockland Immunochemicals	Cat# 610-1302; RRID:AB_219656
Anti-rabbit-HRP	Rockland Immunochemicals	Cat# 611-1302; RRID:AB_219720
Anti-mouse- IRDye® 800CW	LI-COR Biosciences	Cat# 925-32210; RRID:AB_2687825
Anti-His6	Abcam	Cat# ab18184; RRID:AB_444306
Anti-GST	Abcam	Cat# ab19256; RRID:AB_444809
Anti-rabbit Nano-Secondary (CTK0101)	ChromoTek	Cat# shurbGNHS-1
Anti-GST VHH	ChromoTek	Cat# ST-250; RRID:AB_2631368
Anti-GFP VHH	ChromoTek	Cat# GT-250; RRID:AB_2631361
Anti-RFP VHH	ChromoTek	Cat# RT-250; RRID:AB_2631365
Bacterial and virus strains
NEB Stable Competent E. coli	New England BioLabs	Cat# C3040H
Chemicals, peptides, and recombinant proteins
(R, S)-Lenalidomide	BioVision	Cat# 1862-25
Pomalidomide	TCI	Cat# P2074
MLN4924	Selleck Chemicals	Cat# S7109
Dulbecco’s Modified Eagle’s Medium (DMEM)	Genesee Scientific	Cat# 25-500
RPMI 1640	Gibco	Cat# 10-040-CV
Trypsin-EDTA	Fisher Scientific	Cat# 25200114
Fetal bovine serum (FBS)	Peak Serum	Cat# PS-FB2
Penicillin-streptomycin (100×)	Lonza	Cat# 17-602E
BCA solution (BCA Reagent A)	VWR	Cat# 786-847
Copper Solution (BCA Reagent B)	VWR	Cat# 76825-860
CoraFluor-1-Pfp	Payne, N. C. et al.40	N/A
AF488-Tfp	ThermoFisher	Cat# A37570
His6-CRBN/DDB1	Bristol Myers Squibb	a generous gift
Thal-FITC	This paper	N/A
GST-BRD4(BD1)	BPS	Cat# 31040
GST-BRD4(BD2)	Epicypher	Cat# 15-0013
JQ1-FITC	This paper	N/A
His6-FKBP12	This paper	N/A
SLF-FITC	This paper	N/A
His6-GST-CDK4/GST-CD3	BPS	Cat# 40104
His6-CDK6/GST-CD1	BPS	Cat# 40097
DMEM with 4.5 g/L Glucose, without phenol re	Lonza	Cat# 12-917F
Trypsin-EDTA (0.5%), no phenol red	ThermoFisher	Cat# 15400054
Opti-MEM I Reduced Serum Medium	ThermoFisher	Cat# 31985070
Polybrene	Santa Cruz Biotechnology	Cat# sc-134220
Puromycin	ThermoFisher	Cat# A1113803
TransIT-Pro Transfection Reagent	Mirus	Cat# MIR 5760
Protease/phosphatase inhibitor cocktail (100×)	CST	Cat# 5872
TMTpro 16-plex label reagent set	ThermoFisher	Cat# A44520
Critical commercial assays
Quick Ligation Kit	New England BioLabs	Cat# M2200S
Deposited data
Proteomics data (Supplementary Fig. 10c–e)	This paper	PRIDE: PXD041494
Experimental models: Cell lines
HEK293T	ATCC	Cat# CRL-3216
Jurkat	ATCC	Cat# TIB-152
HEK293T cells stably expressing stably expressing BRD4(BD1)-GFP	Nowak, R. P. et al.22	N/A
HEK293T cells stably expressing stably expressing BRD4(BD2)-GFP	Nowak, R. P. et al.22	N/A
U2OS cells stably expressing the GFP-fused zinc finger	Nguyen, T. M. et al.57	N/A
HEK293T cells stably expressing stably expressing GSPT1-GFP	This paper	N/A
Oligonucleotides
Primer 1: CGTCTCACCATGGAACTTTCAGAACC	This paper	N/A
Primer 2: CGTCTCTCGCCGTCTTTCTCTGGAACCAG	This paper	N/A
Recombinant DNA
pcDNA3.1-GSPT1	Genscript	Cat#: OHu24371
Cilantro2	Addgene	Cat# 74450
Cilantro2-GSPT1	This paper	N/A
pMD2.G	Addgene	Cat# 12259
psPAX2	Addgene	Cat# 12260
Software and algorithms
Microsoft Excel (v16.44)	Microsoft	N/A
GraphPad Prism (v8.4.3).	GraphPad Software	N/A
MOE (v2022.02)	Chemical Computing Group	N/A
Geneious (v11.0.3)	Geneious	N/A
ImageJ (v1.52q)	NIH	N/A
Adobe Photoshop (v21.1.1)	Adobe	N/A
Adobe Illustrator (v24.1)	Adobe	N/A
Other
NuPAGE 3–8% Tris-Acetate precast gels	ThermoFisher	Cat# EA0375BOX
Criterion XT Tris-Acetate precast gels	Bio-Rad	Cat# 3450131
4–15% Criterion TGX precast gels	Bio-Rad	Cat# 5671085
16.5% MiniPROTEAN ® Tris-Tricine gels	Bio-Rad	Cat# 4563066
15% Criterion Tris-HCl protein gels	Bio-Rad	Cat# 3450021
iBlot 2 nitrocellulose transfer stac	Invitrogen	Cat# IB23001, IB23002
Plates for TR-FRET	Corning	Cat# 3572
Plates for TR-FRET	Greiner	Cat# 781207
BsmBI	New England BioLabs	Cat# R0739S
Lys-C, Mass Spec Grade	Promega	Cat# VA1170
Trypsin, sequencing grade modified	Promega	Cat# V5111
Pierce Peptide Desalting Spin Column	Thermo	Cat# 89852
96-well plate, 1.0 mL, round wells	Agilent	Cat# 5043-9305

Significance:

The E3 ligase adapter Cereblon (CRBN) is one of the best-characterized targets for targeted protein degradation (TPD) applications, prompting numerous efforts to degrade a myriad of disease-related target proteins. However, achieving efficient and selective target degradation remains nontrivial with ligands that engage CRBN. In this study, we introduce a systematic framework for screening the E3 ligase binder to achieve high target selectivity and degradation efficiency, providing a robust, orthogonal strategy to varying the linker in heterobifunctional degraders (e.g., PROTACs). We demonstrate that the cyclimids, ligands derived from the C-terminal cyclic imide degrons for CRBN, fine-tune the binding and degradation properties of bifunctional degraders and offer a facile approach for developing potent and selective bifunctional degraders. By employing quantitative, high-throughput TR-FRET assays, we obtained a comprehensive evaluation of 60 cyclimid degraders in binary and ternary complexes with CRBN across different substrates (e.g., BRD4, FKBP12, and CDK4/6) to reveal strong correlation between the ternary complex binding affinities and cellular degradation efficiency. Notably, the cyclimid degraders with 5-membered cyclic asparagine (cN) generally exhibited higher affinities to CRBN and degradation potency than their 6-membered cyclic glutamine (cQ) counterparts. Our studies raise important implications for the physiological function of CRBN by demonstrating that the 5-membered cN preferentially engages CRBN as compared to the more traditional 6-membered cQ found in thalidomide and its derivatives. Additionally, the cyclimids possess minimal off-target degradation that are common neosubstrates of CRBN-based PROTACs and glues, and exhibited metabolic stability and permeability characteristics on par with the thalidomide-based degrader. Collectively, our results establish the utility of systematic screening of E3 ligase ligands to yield potent and efficient bifunctional degraders through tuning of target–CRBN ternary complexes, while raising interesting new directions relevant to CRBN biology.