pocketz.doi.bio/2026-06-13_163349_nature_s41467-024-50642-0

Dual-site molecular glues for enhancing protein-protein interactions of the CDK12-DDB1 complex | Nature Communications

URL: https://www.nature.com/articles/s41467-024-50642-0

Saved: 2026-06-13T23:33:55.062Z

Download PDF Article Open access Published: 01 August 2024 Dual-site molecular glues for enhancing protein-protein interactions of the CDK12-DDB1 complex Zemin Zhang, Yuanqing Li, Jie Yang, Jiacheng Li, Xiongqiang Lin, Ting Liu, Shiling Yang, Jin Lin, Shengyu Xue, Jiamin Yu, Cailing Tang, Ziteng Li, Liping Liu, Zhengzheng Ye, Yanan Deng, Zhihai Li, Kaixian Chen, Hong Ding, Cheng Luo & Hua Lin 

Nature Communications volume  15, Article number: 6477 (2024) Cite this article

17k Accesses

21 Citations

5 Altmetric

Metrics details

Abstract

Protein-protein interactions (PPIs) stabilization with molecular glues plays a crucial role in drug discovery, albeit with significant challenges. In this study, we propose a dual-site approach, targeting the PPI region and its dynamic surroundings. We conduct molecular dynamics simulations to identify critical sites on the PPI that stabilize the cyclin-dependent kinase 12 - DNA damage-binding protein 1 (CDK12-DDB1) complex, resulting in further cyclin K degradation. This exploration leads to the creation of LL-K12-18, a dual-site molecular glue, which enhances the glue properties to augment degradation kinetics and efficiency. Notably, LL-K12-18 demonstrates strong inhibition of gene transcription and anti-proliferative effects in tumor cells, showing significant potency improvements in MDA-MB-231 (88-fold) and MDA-MB-468 cells (307-fold) when compared to its precursor compound SR-4835. These findings underscore the potential of dual-site approaches in disrupting CDK12 function and offer a structural insight-based framework for the design of cyclin K molecular glues.

Similar content being viewed by others Design principles for cyclin K molecular glue degraders Article Open access 07 September 2023 The CDK12 inhibitor SR-4835 functions as a molecular glue that promotes cyclin K degradation in melanoma Article Open access 16 December 2023 Discovery of molecular glues that bind FKBP12 and structurally distinct targets using DNA-encoded libraries Article Open access 22 April 2026 Introduction

Numerous protein−protein interactions (PPIs) are essential to human cells and their molecular recognition is central to biological processes1,2,3. Molecular glues, inducing or stabilizing PPIs, are changing the paradigm of clinically regulated protein interactions. In the past two decades, molecular glue drugs, such as thalidomide, tacrolimus, and rapamycin, have treated countless patients and generated tens of billions in revenue4,5. At the same time, molecular glues and proteolysis targeting chimeras6 together inspired the two main strategies of targeted protein degradation and the former has the advantages of targeting undruggable proteins, lower molecular weight and drug-like physical and chemical properties7,8,9,10,11,12. These advantages highlight the promise of molecular glues in regulating PPIs.

Over the past decades, the discovery of molecular glues has relied mainly on serendipity, phenotypic screening, mechanistic validations or post-hoc rationalization13,14,15,16,17,18,19,20,21,22. In recent years, there has been an increased use of systematic screens to identify molecular glues. For example, CR823 and NCT0224 were identified to be cyclin-dependent kinase 12 - DNA damage-binding protein 1 (CDK12-DDB1) molecular glues by an external screening strategy or by a genetic screening strategy. They induced cyclin K degradation by linking CDK12–cyclin K to the DDB1– Cullin 4 (CUL4) – RING box protein-1 (RBX1) E3 ligase, without the cereblon (CRBN) substrate receptor. Interestingly, the CDK12 inhibitor SR-4835 was discovered to be a superior CDK12-DDB1 molecular glue compared to CR823,24. SR-4835, which is highly homologous to CR8, shares similar chemical characteristics with CR8, not only significantly stabilizing the interaction between CDK12 and DDB1, but also possessing enhanced selectivity and drug-like properties, making SR-4835 a more promising candidate for the optimization of molecular glues compared to other cyclin K degraders such as CR8, NCT02 and HQ46124,25,26,27. Since the optimization of molecular glues was relatively ambiguous, we envisioned using SR-4835 to explore general laws further.

The investigation of the dynamic regulation process of a protein or a target is a powerful means for drug discovery. By capturing the dynamic features, targetable allosteric sites, allosteric Post-translational modifications (PTM) sites, reshaped pockets or even additional binding pockets can be referred and applied to complex drug design28,29,30,31,32. However, the research on how molecular glues dynamically regulate the stability of the entire complex is limited. To investigate this, we focused on the CDK12-DDB1 complex, which mediates the ubiquitination and degradation of cyclin K. According to the reported structures resolved by crystallography, molecular glue ligand CR8 occupies the ATP pocket and induces PPI through the solvent-exposed part23. Simultaneously, CDK12 is also regulated by the interaction, further stabilizing the complex by allosteric means. These clues suggest an added rational direction in molecular glue design and optimization from a dynamic perspective.

In the preparation of this manuscript, the Thomä group reported the evaluation of 91 cyclin K candidate degraders and their mechanisms33. Structure-activity relationship (SAR) analysis were presented, utilizing resolved structures to elucidate the critical binding site with DDB1 for optimal activity. The most potent cyclin K degrader identified was DS18, exhibiting an EC50 of 9 ± 0.2 nM (TR-FRET). DS18 is a derivative of SR-4835, demonstrating a 1.8-fold increase in potency compared to SR-483533. Despite advancements, significantly enhancing the activity of molecular glues through rational design of the PPI component remains a formidable challenge34.

In this work, our objective is to delve deeper into the SAR of molecular glues, leveraging SR-4835. Drawing inspiration from the crystal structure of CDK12-DDB1, the allosteric stabilization process and our discoveries relate to the CDK-cyclin complex degraders35,36,37, we introduce the concept of ‘dual-site molecular glues’. This innovative approach targets the unoccupied interface between complexes, which is crucial for the overall stability. Dual-site molecular glues, exemplified by LL-K12-18, demonstrated amplified effects by fostering CDK12-DDB1 complex formation, leading to increased protein degradation and anti-proliferation activity. This paradigm relies on heightened interactions, dynamic stability regulation and an expanded binding interface for PPIs.

Results Emerging pockets may allosterically regulate protein complex stability

The characteristic degradation of cyclin K by CDK12-DDB1 molecular glues in cells indicates that the inhibition mechanisms of CDK12 not only involve occupying the kinase substrate pocket, but also result from the formation of a degradation complex. Recent structural research has elucidated the interaction details of molecular glues with CDK12 and the endogenous E3 ligase (DDB1). This emphasizes that the study of molecular glues should be based on the complex structure rather than individual kinases. For instance, due to the presence of DDB1, the ATP pocket is reshaped in the degradation complex, partly blocking the solvent-exposed region in CDK12. Concurrently, the C-terminal extension peptide (CTE) within the kinase domain of CDK12 inserts into the groove between the BPA and BPC domains of DDB1. Mutations within CTE can significantly attenuate the DDB1-CDK12 interaction induced by molecular glues23. These structural differences in CTE suggest that the formation of the degradation complex facilitated by molecular glues, undergoes a series of dynamic stabilizations and the mechanism of molecular glues to regulate the entire complex allosterically is a key clue for uncovering degraders (Supplementary Fig. 1a).

To investigate the dynamic regulatory mechanism of CDK12 molecular glue in facilitating the stability of the CDK12 and DDB1 complex, we performed molecular dynamics (MD) simulations on the ternary complex including DDB1, CDK12, and cyclin K based on crystal structures. In parallel, we docked the selective CDK12 regulator SR-4835, which has been reported as a molecular glue and cyclin K degrader24, to the complex to explore the allosteric effects of molecular glues and optimization strategies (Fig. 1a). We simulated the complex with and without SR-4835 for 120 ns, respectively and acquired two trajectories, ‘SR-4835’ and ‘No SR-4835’ trajectories. The stability of the complex was enhanced by SR-4835, as indicated by the lower RMSD value throughout the simulation (Supplementary Fig. 1b). Additionally, the estimated RMSF value of residues in the ‘SR-4835’ trajectory was generally lower than in the ‘No SR-4835’ trajectory (Supplementary Fig. 1c). Surprisingly, the delta RMSF analysis revealed that the CTE was significantly stabilized by SR-4835 binding, consistent with structural research (Supplementary Fig. 1d, e)23. These results suggest that SR-4835 enhances the interaction between CDK12 and DDB1 not only through the direct binding of the hydrophobic group on DDB1, but also by regulating the binding interface of CTE allosterically.

Fig. 1: Design of a dual-site molecular glue strategy based on an additional allosteric stabilization region. The alternative text for this image may have been generated using AI. Full size image

a Scheme for structure preparation and simulation process for MD. b Pocket prediction of SR-4835 binding site (magentas) and the emerged outer pocket (green) in the DDB1-CDK12 binding interface. c Illustration of residue communication paths starting from DDB1-Pro951. d Derivation direction for dual-site molecular glues. a–d DDB1 is shown in orange, CDK12 in blue and cyclin K in green. d SR-4835 is shown in yellow and distance in silver dashed line.

Inspired by the simulation results, we examined residues in the region of the binding pockets that could potentially affect the stability of the CTE remotely. This information can guide the optimization of molecular glues. As a result of protein pocket prediction, an additional pocket was identified in the outer region near the SR-4835 binding site, primarily composed of residues from DDB1, which fused with the substrate binding site of CDK12 to form a larger pocket (Fig. 1b). Protein structure network (PSN) model was applied to the entire complex to investigate the residue interactions between the additional identified pocket and the CTE of CDK1238. The PSN analysis revealed crosslinks between DDB1 and CDK12, along with critical nodes and edges emerging in the allosteric regulation process (Supplementary Fig. 2a). The residue communication paths with the highest frequency, calculated from Protein Structure Network Path analyses (PSNPATH), indicated that the binding site of SR-4835 could regulate CTE through a relatively short path involving CDK12-His818, CDK12-Leu1033, CDK12-Leu828 and CDK12-Pro1034 (Supplementary Fig. 2b, c). Additionally, residues within DDB1 connecting the dichloro-benzimidazole group on SR-4835 formed a regulation path to CTE, including DDB1-Ile909 and DDB1-Arg928, highlighting the strong correlation between the dichloro-benzimidazole surrounding pocket and complex stability. Notably, DDB1-Pro951, located relatively distant from SR-4835 and positioned at the outer pocket, also formed paths through dichloro-benzimidazole surrounding residues to CDK12-His1040 and CDK12-Pro1034 (Fig. 1c, and Supplementary Fig. 2b). Communication paths originating from DDB1-Pro951 revealed in MD simulation suggested that the allosteric effects could be further enhanced by residues around the outer pocket. Furthermore, the PSN model revealed that different regions of this additional formed fusion pocket can synergistically regulate the stability of the whole protein complex through allosteric action.

Design of dual-site molecular glues

The dynamic simulation and the PSN model provided insight into an additional targetable pocket to enhance the interaction. This pocket involves DDB1-Pro951 located near the binding site of SR-4835 (Fig. 1d). Previous studies have shown that modifying the solvent-exposed region of CDK12 is essential to produce more effective derivatives, but such modifications can be hit-or-miss or achieve very little, thus rational solvent region design remains a major challenge23,33,34,39. We assumed that the added pocket in the CDK12-DDB1 complex could be a secondary gluing site to form dual-site molecular glues and explore the optimization space. Considering the appropriate orientation of morpholine groups at the DDB1-CDK12 interface (Supplementary Fig. 3a), we chose this direction of extension for the development of cyclin K degrader. Based on the interaction of the native gluing moiety with DDB1 (Supplementary Fig. 3b), our objective was to augment CTE modulation efficacy by an additional gluing group. Initially, a range of piperazine derivatives were synthesized by changing morpholine to an easily modifiable piperazine. Then, we evaluated the effects of a series of compounds on the degradation of cyclin K and anti-proliferation effects in MDA-MB-231 and MDA-MB-468 cells.

Discovery and exploration of dual-site molecular glues

Using MDA-MB-231 cells as an example, the anti-proliferation activities of LL-K12-1 and LL-K12-2 were observed to be lower than that of SR-4835 (Table 1). The observed effect is attributed to unfavorable steric collisions of the methylpyrazole group (Supplementary Fig. 4a, b), consistent with recent findings33. Enhancing the flexibility of molecules proved to be an effective modification strategy, as the overall rigidity hindered the adjustment of molecular conformation28,40,41. Therefore, we replaced the larger methylpyrazole with smaller methyl (LL-K12-3), ethyl (LL-K12-4) and isopropyl groups (LL-K12-5) in subsequent experiments. Among these modifications, LL-K12-4 exhibited a more effective degradation effect, which was 11.1-fold enhanced compared with LL-K12-2. It successfully degraded cyclin K protein at a concentration of 5 nM and resulted in an EC50 of 27.42 nM in cells (Table 1). The size of the substituent in the hydrophobic pocket had a direct impact on activity and the appropriate substituent volume was essential for achieving optimal degradation activity. Excessive substituent size, as seen in LL-K12-5 or the opposite, as in LL-K12-3, resulted in reduced potency. Docking results suggested that LL-K12-4 not only forms hydrogen-bonding interactions with Glu735 of CDK12, but also has appropriately sized substituents. (Supplementary Fig. 5a–c).

Table 1 Structures of compounds SR-4835 and LL-K12-1–5 Full size table

LL-K12-4 was found to be superior to its counterparts (Table 2), where we attempted substitutions with methanesulfonyl piperazine (LL-K12-6), acetamide piperidine (LL-K12-7), butoxy (LL-K12-8) and triazole derivative (LL-K12-9). In our endeavor to target potential dynamic regulatory sites and an additional gluing site within the CDK12-DDB1 complex, we explored two approaches involving N-aromatic rings and aliphatic amines (Table 2). This exploration included the addition of pyridine substituents at the ortho- (LL-K12-10), meta- (LL-K12-11) and para-positions (LL-K12-12). Similarly, introducing polar aromatic rings around DDB1-Arg928 resulted in modest increases in activities, such as a 2.1-fold increase in the activity of LL-K12-11 than LL-K12-4. Whereas introduction of the benzene ring (LL-K12-13) or differently linked pyridine rings (LL-K12-14 and LL-K12-15) did not. The change of N-aromatic rings in activity was somewhat limited, likely due to competition with benzimidazole for Pi-cation binding sites.

Table 2 Structure–activity relationship of compounds LL-K12-4, 6–18 Full size table

Interestingly, we finally found LL-K12-18, which emerged as the remarkably potent degrader of cyclin K currently (EC50 = 0.37 nM). utilizing a diethylglycine derivative as the additional gluing moiety. When diethyl glycine derivative (LL-K12-18) was replaced by acetyl glycine derivative (LL-K12-16) or acetyl alanine derivative (LL-K12-17), the activities decreased more than 100-fold (Table 2). The acetylation of the amino group in these derivatives primarily blocked the hydrogen bond with DDB1-Phe949. While western blot results demonstrated that LL-K12-18 significantly degraded cyclin K but less affecting CDK12 at 5 nM, The EC50 of LL-K12-18 was determined to be 0.37 nM in MDA-MB-231 cells, resulting in 80-fold increase in relative activity compared to SR-4835 and a 74-fold increase compared to LL-K12-4, which was consistent with the general trend of the findings in MDA-MB-468 cells. In addition, the intriguing discrepancy between reduced kinase inhibition and enhanced cellular suppression in LL-K12-18 was reminiscent of NCT0226. Furthermore, to validate whether LL-K12-18’s selectivity was affected, we assessed other kinases within the CDK family and found their selectivity remained intact (Supplementary Fig. 8). This indicates a reduction of the inhibitory efficacy of LL-K12-18 and a remarkable enhancement of its degradative capability.

The binding model of LL-K12-18 was consistent with the SR-4835 binding site (Fig. 2a). Through the piperazine linker, the diethylamine acetyl moiety was directed towards an additional gluing site and the carbonyl and amine groups formed hydrogen bonds with CDK12-Glu735 and DDB1-Phe949, respectively, reestablishing the gluing effect and forming a dual-site molecular glue (Fig. 2b, and Supplementary Fig. 9a–c). Due to the formation of a key hydrogen bond with DDB1 (Fig. 2c), the gluing effect in the additional gluing region was 70-fold more active than of other compounds without a dual-site gluing effect, such as LL-K12-4.

Fig. 2: LL-K12-18 binding model and enhancement of CDK12-DDB1 interaction. The alternative text for this image may have been generated using AI. Full size image

a The binding pocket of CDK12-DDB1 for LL-K12-18. b Residues involved in LL-K12-18 degrader recognition. c Schematic interactions in CDK12-DDB1 and LL-K12-18. a–c CDK12 is shown in blue and DDB1 in orange. Interactions are represented by dashed lines. Hydrogen bonds are shown in yellow and π-cation interactions in green. d Binding affinity induced by SR-4835 or LL-K12-18 detected by SPR assay. e Recombinant Strep-DDB1 pulled-down by overexpressed CDK12-cyclin K in the presence of 1 μM SR-4835 or LL-K12-18. All blots in one subfigure came from an identical sample and were processed in parallel. Similar results were found in two repeated assays.

Lastly, we explored the replacement of dichloro-benzimidazole in the original gluing site with 4,5-difluoro-benzimidazole (LL-K12-19) and 5,6-difluoro-benzimidazole (LL-K12-20). As anticipated, this led to a decrease in cyclin K degradation, possibly due to the stronger electron-withdrawing groups reducing the strength of cation-π interactions, thereby lowering the gluing effect (Table S1)33. These findings underscore the potential of LL-K12-18 for further investigation, as it represents a potent dual-site molecular glue with picomolar-level activity.

Dual-site molecular glue (LL-K12-18) enhances CDK12-DDB1 protein interaction

To evaluate whether the glue-like effect could be enhanced after structural optimization by the dual-site approach, surface plasma resonance (SPR) assays were applied to investigate the interaction between DDB1 and CDK12. In this assay, a marked increase in binding signal was noted when the CDK12-cyclin K complex was incubated with SR-4835 or LL-K12-18, but not with the apo CDK12-cyclin K complex or the compounds alone (Fig. 2d, and Supplementary Fig. 10a, b), which aligns with the molecular glue effects of SR-4835 and LL-K12-18 on the DDB1-CDK12 interface. Specifically, fitted kinetic parameters indicated that LL-K12-18 promotes a higher-affinity interaction between DDB1 and CDK12 compared to SR-4835, largely due to a slower dissociation rate (Fig. 2d). Further, an in vitro pull-down assay demonstrated enhanced capture of DDB1 protein by LL-K12-18 in the presence of overexpressed CDK12 protein (Fig. 2e). Therefore, the dual-site approach could further improve the molecular-glue like effect to adhere the partners within the degradation complex.

LL-K12-18 facilitates rapid cyclin K degradation via the ubiquitin-proteasome system

DC50 curves of LL-K12-18 and SR-4835 on cyclin K were generated to investigate changes in degradation capacity. Western blot results revealed that LL-K12-18 could almost completely degrade cyclin K at 10 nM within 4 h. The DC50 value of LL-K12-18 was 0.38 nM, a 50-fold boost than SR-4835, closely matching the enhancement observed in anti-proliferation EC50 values (Fig. 3a–c). Unlike cyclin K, the effects of SR-4835 and LL-K12-18 on CDK12 levels were significantly less pronounced, which can be attributed to CDK12’s role as an adapter within the degradation complex. In terms of degradation kinetics, 50 nM LL-K12-18 could nearly entirely degrade cyclin K within 2 h, a significantly faster degradation rate compared to SR-4835 (Fig. 3d, e).

Fig. 3: Degradation characteristics of LL-K12-18. The alternative text for this image may have been generated using AI. Full size image

Protein levels of CDK12 and cyclin K after 4 h treatment with LL-K12-18 (a) and SR-4835 (b) in MDA-MB-231 cells. c DC50 curves quantified by western blot assay (n = 3, Supplementary Fig. 11). Data are presented as mean values +/- SD. Time course of cyclin K degradation by 50 nM LL-K12-18 (d)and SR-4835 (e). Cyclin K levels rescued by 1 h pre-treatment with 10 μM MLN-4924 (f), 1 μM MG132 or reduced levels of DDB1 knockdown by shRNA(g). All blots in one subfigure came from an identical sample and were processed in parallel. Similar results were found in two repeated assays.

To further elucidate the degradation mechanism, chemical probes MG132 and MLN-4924 were employed to interfere with the ubiquitin-proteasome system. Inhibition of the 26 S proteasome by MG132 or deactivation of the E3 ligase by NAE inhibitor MLN-4924 led to the rescue of cyclin K levels (Fig. 3f). Additionally, cyclin K levels were rescued upon DDB1 reduction through shRNA (Fig. 3g). These findings indicate that the degradation process of cyclin K by LL-K12-18 is dependent on the DDB1-mediated ubiquitin-proteasome system.

LL-K12-18 selectively degrades cyclin K and down-regulates DDR genes in MDA-MB-231 cells

Considering the enhanced molecular glue-like properties of LL-K12-18, we assessed its degradation selectivity using quantitative proteomics. Within 2 h of treatment, LL-K12-18 significantly decreased cyclin K levels more than SR-4835 (Fig. 4a, and Supplementary Fig. 12). Notably, cyclin K degradation was both effective and selective by LL-K12-18. Concurrently, LL-K12-18, but not SR-4835, reduced CDK13 levels, aligning with recent findings33,42. These studies suggest that molecular glue-based degraders of cyclin K preferentially decrease CDK13 levels, likely due to CDK13’s stability being contingent upon cyclin K. Further, we extended the treatment duration to 6 h with a relatively higher dose of 2 μM LL-K12-18 and minimal changes in protein levels were observed for cyclins and CDKs, except for cyclin K (Fig. 4b). Confirming this selectivity, LL-K12-18 reduced the phosphorylation of Ser2 on RNA polymerase II C-terminal domain (RNA POL II CTD) but weakly affected on Ser5, indicating its inhibition of downstream phosphorylation sites specific to CDK12 rather than other transcriptional CDKs like CDK7 (Fig. 4c)43.

Fig. 4: LL-K12-18 leads to selective degradation of cyclin K (CCNK), reduction of phosphorylation on Ser2 and inhibition of DDR genes in MDA-MB-231 cells. The alternative text for this image may have been generated using AI. Full size image

a Comparison of proteomics in LL-K12-18 treated MDA-MB-231 cells versus DMSO control. p value was calculated by one-way analysis of variance (ANOVA) and Tukey’s honestly significant difference (HSD) test. b Protein levels of cyclins and CDKs after 6 h treatment of LL-K12-18. c Phosphorylation levels of Ser2 and Ser5 and total RNA POL II were detected after 6 h treatment of LL-K12-18. d Transcription levels of DDR genes inhibited by SR-4835 and LL-K12-18 in 6 h (n = 3). Data are presented as mean values +/- SD and p value are illustrated as * < 0.05, ** < 0.01, *** < 0.001 by two-tailed multiple t test. e Influences of protein markers in MDA-MB-231 cells by 24 h treatment of LL-K12-18. All blots in one subfigure came from an identical sample and were processed in parallel. Similar results were found in two repeated assays.

CDK12 has been linked to DDR genes in triple negative breast cancer (TNBC)25,44. In line with its rapid degradation of cyclin K, LL-K12-18 efficiently reduced the mRNA levels of DDR-related genes, even more effectively than SR-4835 at a dose of 50 nM (Fig. 4d). The protein levels of ATM and RAD51 were also reduced by LL-K12-18 after 12 or 24 h of treatment, accompanied by increased levels of γ-H2AX and cleaved-PARP, markers of DNA damage and apoptosis (Fig. 4e, and Supplementary Fig. 13). These results demonstrated that LL-K12-18 could efficiently and selectively interfere with CDK12 function at the cellular level.

Discussion

In this study, we proposed a molecular glue design approach centered on the dynamic regulatory features of the complex to identify critical sites that mediate complex stability and amplify the areas of PPIs through dual-site gluing. By deciphering the principles of the original gluing region, we identified an additional adhesive region through MD simulation and PSN model and the dual-site compound LL-K12-18 was developed using a dual-site approach. Initially, we modulated the flexibility of the parental inhibitor, resulting in LL-K12-4 binding to Glu735 of CDK12, whose activity enhanced and preliminarily validated through molecular docking and cell experiments. Subsequently, we explored the SAR through a series of molecular glues and leveraged Phe949 of DDB1 to induce additional PPIs, ultimately yielding dual-site molecular glue LL-K12-18 (EC50 = 0.37 nM) with an 80-fold enhanced potency than SR-4835 in MDA-MB-231 cells and with a 307-fold boost potency (EC50 = 0.03 nM) in MDA-MB-468 cells, while the degradation efficiency (DC50 = 0.38 nM) increased 50-fold. SPR and in vitro pull-down assay confirmed the dual-site gluing could enhance the affinity between DDB1 and CDK12 with a slower dissociation rate, corroborating molecular docking results of LL-K12-18. These results indicated that the dual-site approach bolstered CDK12 and DDB1 adhesion in cells.

The dual-site effect extended beyond enhanced interaction to faster degradation of cyclin K, improved gene transcription regulation and downstream signaling. Cyclin K degradation by 50 nM LL-K12-18 reached nearly 100% within 2 h, significantly outpacing the degradation rate of the single-site molecular glue SR-4835. LL-K12-18 demonstrated superior protease-dependent degradation of cyclin K and its effect could be rescued by proteasome inhibitors. Furthermore, it displayed exceptional degradation selectivity, downregulated CDK12 downstream phosphorylation sites and modulated DNA damage and apoptosis markers. These results highlight LL-K12-18’s potent, rapid and selective interference with CDK12 function and suggest that molecular glue can be effectively enhanced by targeting the unoccupied interface of CDK12-DDB1. In conclusion, this study advances our understanding of molecular glues by providing practical insights into cyclin K molecular glues design and optimization, thereby paving the way for cyclin K-based drug discovery.

Methods Chemistry

Intermediates 8, 10a ~ c, 12 and 13a, b were prepared according to the literatures25,45. Intermediate 8 reacted with morpholine, piperazine or acetylpiperazine to give rise to SR-4835 or LL-K12-1 ~ 2 by SNAr reaction, respectively. Intermediates 10a ~ c were alkylation with iodomethane, iodoethane or 2-iodopropane to afford 11a ~ c respectively and then reacted with piperazine or acetylpiperazine to provide intermediate 12 or LL-K12-3 ~ 5 (Fig. 5).

Fig. 5: Synthesis of compounds SR-4835 and LL-K12-1 ~ 20. The alternative text for this image may have been generated using AI. Full size image

Reagents and conditions: (i) Boc-Glycine, HOBT, EDCI, DCM, DMF, rt., 2 h; (ii) AcOH, 70 °C, 2 h; (iii) 1,4-dioxane, HCl (conc.), rt., overnight; (iv) TfOH, DCM, rt., 5 min; m-CPBA, Mesitylene, DCM, 60 °C, 30 min; (v) 2,6-dichloropurine, CuBr, Et3N, CHCl3, 60 °C, 4 h; (vi) 4c, i-PrOH, DIPEA, 90 °C, 2 h; (vii) Corresponding amines, n-BuOH, DIPEA, 130 °C, 16 h for SR-4835, LL-K12-1 ~ 2; (viii) Iodoalkanes, Et3N, DMF, 60 °C, 16 h; (ix) 4c, i-PrOH, DIPEA, 55 °C, 8 h; (x)1-Acetylpiperazine, n-BuOH, DIPEA, 120 °C, 16 h for LL-K12-3 ~ 5; (xi) piperazine, n-BuOH, DIPEA, 130 °C, 16 h; (xii) Corresponding acids, HATU, DIPEA, DMF, rt. or Acetyl-L-alanine, TCFH, NMI, MeCN, rt., overnight for LL-K12-10 ~ 13, 16 ~ 18; 4-Pyridinecarboxaldehyde DCE, NaBH(OAc)3, 0 °C-rt., overnight for K12-15; (xiii) Corresponding amines, n-BuOH, DIPEA, 16 h 130 °C for LL-K12-6 ~ 7, 14; n-BuOH as solvent, Cs2CO3 instead of DIPEA for LL-K12-8; 170 °C instead of 130 °C for LL-K12-9; (xiv) 10b, i-PrOH, DIPEA, 50 °C, 2 h; (xv) 5,6,7,8- Tetrahydro-[1,2,4]triazolo[1,5-a] pyrazine, n-BuOH, DIPEA, 170 °C, 16 h for LL-K12-19 ~ 20.

Intermediate 12 coupled with corresponding acids in the presence of HATU or TCFH/NMI to provide LL-K12-10 ~ 13 or LL-K12-16 ~ 18. The corresponding amines or alcohol were reacted with 11b by DIPEA or Cs2CO3 under different heating conditions to obtain LL-K12-13 ~ 14, 16 ~ 18. LL-K12-15 was obtained by reductive amination of 4-pyridylaldehyde and 12 (Fig. 5).

Similar to the synthesis of intermediate 11b, dichlorophenyl was replaced by difluorophenyl. Intermediate 4a or b, corresponding purine and corresponding amine were sequentially reacted to provide LL-K12-19 and LL-K12-20 respectively (Fig. 5).

Detailed synthesis of intermediates and target compounds, as well as, 1H NMR of intermediates, 1H and 13C NMR or HPLC of target compounds, images of these spectra are available in the Supplementary Information.

Molecular dynamics simulation

The initial structure for molecular dynamic simulation came from DDB1-CDK12-Cyclin K complex with PDB code 6TD3 while the CR8 was excluded or replaced by SR-4835 by induced-fit docking. The optimized pose of SR-4835 and the corresponding RSEP charge was calculated by gaussian-09 software at the basis set of 6-311+ +g(d,p). The force field parameters for compounds and proteins were integrated in AMBER14 software and the system was further placed in TIP3P water box. The process of pre-equilibrium and dynamic simulation was carried out using GROMACS with time step of 2 fs. Complex of BPC domain in DDB1 and CDK12 were extracted for RMSD and RMSF analysis and PSN modeling.

Protein pocket prediction

Protein pocket prediction was carried out using Fpocket. The input structure was the pre-balanced pose of BPC-CDK12 complex. Features of pockets were further analyzed by PyMOL software. Residues around pockets were selected with a distance cut-off of 3.5 Å around the assumed pocket molecule generated by Fpocket46.

Correlation analysis, protein structure network and PSNPATH searching

The correlation analysis (CORR), PSN and PSNPATH searching were carried out using wordom 0.22-rc147. The correlation analysis was calculated at the level of residue using algorithm of DCC. PSNs were generated at each frame of trajectory, which represents the structure of proteins as a network composed of nodes (side chains of amino acids) and edges (non-covalent interactions). The interaction strengths were calculated as Eq. (1):

$${I}{{ij}}=\frac{{n}{{ij}}}{{N}_{i}}\times 100$$ (1)

Where Iij represents the interaction between residues i and j, nij is the number of atom-atom pairs within a distance cut-off of 4.5 Å between residues i and j, Ni was the normalization factor for residue types i. PSNPATH were searched between the residues around the pocket selected in protein pocket prediction section and the 1034–1047 amino acids of CDK12. Paths were included with the CORR cut-off of 0.5 and Imin cut-off of 3.5. Frequency of paths was calculated as Eq. (2):

$${F}{{path}}=\frac{{n}{{path}}}{{n}_{f}}\times 100$$ (2)

where Fpath represents the frequency of the path, npath is the number of frames where the path could be searched in completely and uninterruptedly, nf is the number of frames in trajectory. Also, interaction frequency of edges between residues was calculated as Eq. (3):

$${F}{{ij}}=\frac{{e}{{ij}}}{{n}_{f}}\times 100$$ (3)

Where Fij represents the interaction frequency of edges between residues i and j, eij represents the number of frames where the interaction existed in at least one of the paths that could be searched. Paths with top frequency and a cut-off of 0.2 were selected for further analysis. The representative PSNPATH network plot was generated by Cytoscape 3.6.0 software and the edges were filtered by interaction frequency cut-off of 0.05.

Protein expression and purification of DDB1

The gene of full-length human DDB1(GenBank ID: Q16531) was cloned into the pFastBac1 vector containing N-terminal Strep II tag and the construct was expressed in SF9 insect cells utilizing a baculovirus expression system. The SF9 cells expressing the full-length DDB1 protein were harvested and resuspended in a buffer containing 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 2 mM MgCl2, 2 mM tris (2-carboxyethyl)phosphine (TCEP), 1×phenylmethylsulfonylfluoride (PMSF) and 1×protease inhibitor cocktail (Sigma). Cells were lysed with a high-pressure homogenizer operating at 200 bar for 5 min. Then the supernatant was isolated by centrifugation at 42,000 × g at 4 °C for 1 h, followed by incubation with Strep Beads for 3 h. After washing the beads with 20 column volumes of wash buffer of 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 2 mM TCEP, the DDB1 protein was eluted with a buffer containing 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 2 mM TCEP and 5 mM D-biotin. The purified protein was then concentrated and subjected to gel filtration chromatography by Superdex 200 Increase 10/300 GL using a buffer composed of 20 mM HEPES (pH 7.5), 150 mM NaCl and 2 mM TCEP.

Surface plasma resonance (SPR) assay

All SPR assays were conducted with Biacore T200 instrument. Recombinant human DDB1 protein was coupled on CM5 chip (Cytiva). For the detection of protein-protein interaction, CDK12(714-1063 amino acids)/cyclin K(1-267 amino acids) complex (purchased from Biortus Biosciences, Wuxi, Jiangsu, China) were incubated with SR-4835 or LL-K12-18 at molar ratio of 1:10 and protein complex without compounds was set for negative control. Then, the mixes of protein and compounds were diluted with SPR buffer (20 mM HEPES, pH = 7.5, 150 mM NaCl, 0.05% DMSO). The diluted mixes were flowed through DDB1 coupled chip and the association, dissociation and binding dissociation constants were calculated in kinetic analysis mode using Biacore evaluation software. For the detection of protein-compound interaction, SR-4835 and LL-K12-18 were diluted with SPR buffer and flowed through DDB1 coupled chip without incubation with CDK12/cyclin K complex. All figures were generated with GraphPad 8.0 software.

Plasmids

For in vitro pull-down assay, gene of full-lengthed cyclin K with HA tag and kinase domain of CDK12(714-1063 amino acids) with 3×Flag tag were subcloned to pcDNA-3.1 vector respectively. All tags were cloned to the N-terminal of each gene. For knock-down assay, sequences for stem-loop formed shRNA transcription were inserted into GV248 vector. The targeting sequences were as follows:

sh-DDB1-1#: CCTATCACAATGGTGACAAAT

sh-DDB1-2#: CGACCGTAAGAAGGTGACTTT

sh-Scramble: TTCTCCGAACGTGTCACGT

Western blot and antibodies

Samples were harvested by 1×laemmli buffer and loaded on 6%–15% SDS-PAGE gel. Proteins were separated by electrophoresis and transferred to nitrocellulose membrane. All membranes were blocked by 1×TBST buffer (10 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween-20) with 5% Non-fat milk at room temperature for 1 h follow by the incubation of primary antibody at 4 °C overnight. Immunoblots of each protein were detected by EZ ECL Pico chemiluminescent reagent (Life iBio, cat#: AP34L024) after incubation of secondary antibody for 1 h at room temperature and three times washing by 1×TBST. Relative protein levels were calculated by ImageJ software. Primary and secondary antibodies applied in this article were as follows:

Anti-cyclin K (Santa Cruz Biotechnology, cat#: sc-376371), anti-CDK1 (Proteintech, cat#: 19532-1-AP), anti-CDK2 (Cell Signaling Technology, cat#: 2546 S), anti-CDK4 (Proteintech, cat#: 11026-1-AP), anti-CDK6 (abcam cat#: ab124821), anti-CDK7 (Cell Signaling Technology, cat#: 2090 S), anti-CDK9 (Cell Signaling Technology, cat#: 2316 S), anti-CDK12 (Cell Signaling Technology, cat#: 11973 S), anti-cyclin A2 (Cell Signaling Technology, cat#: 4656 S), anti-cyclin D1 (Cell Signaling Technology, cat#: 2922 S), anti-cyclin E1 (Cell Signaling Technology, cat#: 4129 T), anti-cyclin H (Cell Signaling Technology, cat#: 2927 S), anti-cyclin T1 (abcam, cat#: ab184703), α-Tubulin (proteintech, cat#: 66031-1-Ig), anti-GAPDH (abclonal, cat#: AC002), anti-ATM (Cell Signaling Technology, cat#: 2873 T), anti-RAD51 (Cell Signaling Technology, cat#: 8875 S), anti-Cleaved-PARP (Cell Signaling Technology, cat#: 9541 T), anti-γ-H2AX (Cell Signaling Technology, cat#: 9718 S), anti-RNA polymerase II CTD pSer2 (Millipore cat#: 04-1571), anti-RNA polymerase II CTD pSer5 (Millipore cat#: 04-1572), anti-RNA polymerase II (active motif cat#: 101307), anti-DDB1 (Santa Cruz Biotechnology cat#: sc-376860), anti-Flag-tag (Proteintech, cat#: 20543-1-Ap), anti-Strep II-Tag (abclonal, cat#: AE066), anti-HA-tag (PTM Bio, cat#: PTM-5389), anti-rabbit IgG antibodies (BBI, cat#: D110058-0100), anti-mouse IgG antibodies (BBI, cat#: D110087-0100).

Pull-down assay

HEK-293T cells were cultured in 10 cm dish in DMEM complete medium (DMEM medium with 10% FBS and 1×Penicillin-Streptomycin Solution) in 37 °C. After grown to 80% of moiety, 6 μg of pcDNA-3.1-3×Flag-CDK12 (714-1063 amino acids) and pcDNA-3.1-HA-cyclin K were transfected with EZ Trans Transfection Reagent (Life iBio, cat#: AC04L099) according to the manufacturer’s instructions. After 24 h of transfection, cells were harvested in pre-cooled PBS with 1×protease inhibitor cocktail (MedChemExpress, cat#: HY-K0010), then the pellet was collected by centrifugation and lysed on ice in pre-cooled IP buffer (50 mM HEPES, pH 7.5, 300 mM NaCl, 0.1%Tween-20, 1×protease inhibitor cocktail). Cell lysate was frozen in liquid nitrogen and thawed in room temperature for 3 times repeats and centrifuged for 15 min at 13,000 × g. The supernatant was divided equally and 15 μL of anti-Flag magnetic beads (Sigma Aldrich, cat#: M8823) were added to each division. Parallelly, 15 μL of anti-Flag magnetic beads were added to IP buffer with equal volume. After rotary incubation for 2 h at 4 °C, beads in each sample tubes were washed by IP-buffer for 3 times and harvested by a magnetic stand. Subsequently, 500 nM recombinant Strep-DDB1 and 1 μM SR-4835, LL-K12-18 or DMSO were added to each tube followed with rotary incubation for 1 h at 4 °C. Then, beads were harvested by a magnetic stand and washed 3 times by IP buffer supplemented with SR-4835, LL-K12-18 or DMSO. All captured proteins were eluted with 1 mg/mL of 3×FLAG peptide (SigmaAldrich, cat#: F4799) with rotation at 4 °C for 40 min and protein levels were detected by western blot assay.

Cell proliferation assay

MDA-MB-231 and MDA-MB-468 cells were seeded on 96-well plates (1500 cells per well) and cultured in DMEM complete medium at 37 °C overnight. Compounds with concentration of gradient were treated for 120 h in 37 °C incubator. 3 repeats for each concentration. The viability of cells was detected by CellCounting-Lite (Vazyme, cat#: DD1101-02). Values of each well were read by Multimode Plate Reader (EnVision, PerkinElmer) and EC50 values were fitted by GraphPad Prism 8 software.

qRT-PCR

MDA-MB-231 cells were seeded on 6-well plates (5 × 105 cells per well) and cultured overnight until completely adhere to the plate, followed by treatment of compounds for 6 h. Total RNA were isolated with RNA isolator Total RNA Extraction Reagent (Vazyme, cat#: R401-01-AA) and cDNA library were obtained by reverse transcription using HiScript III RT SuperMix for qPCR (Vazyme, cat#: R323-01). PCR reaction was carried out in ChamQ SYBR qPCR Master Mix (Vazyme, cat#: Q331-02) on QuantStudio 6 Flex (Thermo). Relative transcription levels of each gene were standardized by GAPDH and calculated according to the ΔΔCt method. Primers applied in this article were as follows:

GAPDH-F: GGAGCGAGATCCCTCCAAAAT

GAPDH-R: GGCTGTTGTCATACTTCTCATGG

BRCA1-F: GAAACCGTGCCAAAAGACTTC

BRCA1-R: CCAAGGTTAGAGAGTTGGACAC

FANCI-F: CCACCTTTGGTCTATCAGCTTC

FANCI-R: CAACATCCAATAGCTCGTCACC

FANCD2-F: AAAACGGGAGAGAGTCAGAATCA

FANCD2-R: ACGCTCACAAGACAAAAGGCA

RAD51-F: CAACCCATTTCACGGTTAGAGC

RAD51-R: TTCTTTGGCGCATAGGCAACA

SMARCC1-F: AGCTGTTTATCGACGGAAGGA

SMARCC1-R: GCATCCGCATGAACATACTTCTT

ATM-F: ATCTGCTGCCGTCAACTAGAA

ATM-R: GATCTCGAATCAGGCGCTTAAA

ATR-F: GGCCAAAGGCAGTTGTATTGA

ATR-R: GTGAGTACCCCAAAAATAGCAGG

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The mass spectrometry proteomics data generated in this study have been deposited to the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repository under accession code PXD05346948,49. Other raw data and processed data essential to this work are provided in the Supplementary Information and Source Data file. Source data are provided with this paper.

References

Zhang, Q. C. et al. Structure-based prediction of protein–protein interactions on a genome-wide scale. Nature 490, 556–560 (2012).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar

Kim, M. et al. A protein interaction landscape of breast cancer. Science 374, eabf3066 (2021).

Article   CAS   PubMed   PubMed Central   Google Scholar

Garlick, J. M. & Mapp, A. K. Selective modulation of dynamic protein complexes. Cell Chem. Biol. 27, 986–997 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar

Dewey, J. A. et al. Molecular glue discovery: current and future approaches. J. Med. Chem. 66, 9278–9296 (2023).

Article   CAS   PubMed   PubMed Central   Google Scholar

Xue, Y., Bolinger, A. A. & Zhou, J. Novel approaches to targeted protein degradation technologies in drug discovery. Expert Opin. Drug Discov. 18, 467–483 (2023).

Article   CAS   PubMed   PubMed Central   Google Scholar

Kathleen et al. Protacs: chimeric molecules that target proteins to the Skp1–Cullin–F box complex for ubiquitination and degradation. Proc. Natl Acad. Sci. 98, 8554–8559 (2001).

Article   Google Scholar

Békés, M., Langley, D. R. & Crews, C. M. PROTAC targeted protein degraders: the past is prologue. Nat. Rev. Drug Discov. 21, 181–200 (2022).

Article   PubMed   PubMed Central   Google Scholar

Burslem, G. M. & Crews, C. M. Proteolysis-targeting chimeras as therapeutics and tools for biological discovery. Cell 181, 102–114 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar

Schreiber, S. L. The rise of molecular glues. Cell 184, 3–9 (2021).

Article   CAS   PubMed   Google Scholar

Tinworth, C. P. & Young, R. J. Facts, patterns, and principles in drug discovery: appraising the rule of 5 with measured physicochemical data. J. Med. Chem. 63, 10091–10108 (2020).

Article   CAS   PubMed   Google Scholar

Kozicka, Z. & Thomä, N. H. Haven’t got a glue: protein surface variation for the design of molecular glue degraders. Cell Chem. Biol. 28, 1032–1047 (2021).

Article   CAS   PubMed   Google Scholar

Luh, L. M. et al. Prey for the proteasome: targeted protein degradation—a medicinal chemist’s perspective. Angew. Chem. Int. Ed. 59, 15448–15466 (2020).

Article   CAS   Google Scholar

Kong, N. R. & Jones, L. H. Clinical translation of targeted protein degraders. Clin. Pharm. Ther. 114, 558–568 (2023).

Article   CAS   Google Scholar

Wu, H. et al. Molecular glues modulate protein functions by inducing protein aggregation: a promising therapeutic strategy of small molecules for disease treatment. Acta Pharm. Sin. B 12, 3548–3566 (2022).

Article   MathSciNet   CAS   PubMed   PubMed Central   Google Scholar

Dong, G., Ding, Y., He, S. & Sheng, C. Molecular glues for targeted protein degradation: from serendipity to rational discovery. J. Med. Chem. 64, 10606–10620 (2021).

Article   CAS   PubMed   Google Scholar

Ito, T. et al. Identification of a primary target of thalidomide teratogenicity. Science 327, 1345–1350 (2010).

Article   ADS   CAS   PubMed   Google Scholar

Narla, A. E. et al. Lenalidomide causes selective degradation of IKZF1 and IKZF3 in multiple myeloma cells. Science 343, 301–305 (2014).

Article   ADS   PubMed   Google Scholar

Lu, G. et al. The myeloma drug lenalidomide promotes the cereblon-dependent destruction of Ikaros proteins. Science 343, 305–309 (2014).

Article   ADS   CAS   PubMed   Google Scholar

Fischer, E. S. et al. Structure of the DDB1–CRBN E3 ubiquitin ligase in complex with thalidomide. Nature 512, 49–53 (2014).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar

Krönke, J. et al. Lenalidomide induces ubiquitination and degradation of CK1α in del (5q) MDS. Nature 523, 183–188 (2015).

Article   ADS   PubMed   PubMed Central   Google Scholar

Matyskiela, M. E. et al. A novel cereblon modulator recruits GSPT1 to the CRL4CRBN ubiquitin ligase. Nature 535, 252–257 (2016).

Article   ADS   CAS   PubMed   Google Scholar

Han, T. et al. Anticancer sulfonamides target splicing by inducing RBM39 degradation via recruitment to DCAF15. Science 356, eaal3755 (2017).

Article   PubMed   Google Scholar

Slabicki, M. et al. The CDK inhibitor CR8 acts as a molecular glue degrader that depletes cyclin K. Nature 585, 293–297 (2020).

Article   CAS   PubMed   PubMed Central   Google Scholar

Dieter, S. M. et al. Degradation of CCNK/CDK12 is a druggable vulnerability of colorectal cancer. Cell Rep. 36, 109394 (2021).

Article   CAS   PubMed   Google Scholar

Quereda, V. et al. Therapeutic targeting of CDK12/CDK13 in triple-negative breast cancer. Cancer Cell 36, 545–558 (2019).

Article   CAS   PubMed   Google Scholar

Lv, L. et al. Discovery of a molecular glue promoting CDK12-DDB1 interaction to trigger cyclin K degradation. Elife 9, e59994 (2020).

Article   MathSciNet   CAS   PubMed   PubMed Central   Google Scholar

Jorda, R. et al. 3,5,7-Substituted pyrazolo[4,3-d]pyrimidine inhibitors of cyclin-dependent kinases and cyclin K degraders. J. Med. Chem. 65, 8881–8896 (2022).

Article   CAS   PubMed   Google Scholar

Rehman, A. U. et al. Computational approaches for the design of modulators targeting protein-protein interactions. Expert Opin. Drug Discov. 18, 315–333 (2023).

Article   PubMed   PubMed Central   Google Scholar

Zhang, H. et al. Dynamics of post-translational modification inspires drug design in the kinase family. J. Med. Chem. 64, 15111–15125 (2021).

Article   ADS   CAS   PubMed   Google Scholar

Zhang, Y. et al. Structural basis of ketamine action on human NMDA receptors. Nature 596, 301–305 (2021).

Article   ADS   CAS   PubMed   Google Scholar

Sun, Z. et al. Covalent inhibitors allosterically block the activation of Rho family proteins and suppress cancer cell invasion. Adv. Sci. 7, 2000098 (2020).

Article   CAS   Google Scholar

Rui, H., Ashton, K. S., Min, J., Wang, C. & Potts, P. R. Protein–protein interfaces in molecular glue-induced ternary complexes: classification, characterization, and prediction. RSC Chem. Biol. 4, 192–215 (2023).

Article   CAS   PubMed   PubMed Central   Google Scholar

Kozicka, Z. et al. Design principles for cyclin K molecular glue degraders. Nat. Chem. Biol. 20, 93–102 (2024).

Article   CAS   PubMed   Google Scholar

Jiang, W., Jiang, Y., Luo, Y., Qiao, W. & Yang, T. Facilitating the development of molecular glues: opportunities from serendipity and rational design. Eur. J. Med. Chem. 263, 115950 (2024).

Article   CAS   PubMed   Google Scholar

Li, J. et al. Discovery of small-molecule degraders of the CDK9-cyclin T1 complex for targeting transcriptional addiction in prostate cancer. J. Med. Chem. 65, 11034–11057 (2022).

Article   CAS   PubMed   Google Scholar

Wang, M. et al. Discovery of LL-K8-22: a selective, durable, and small-molecule degrader of the CDK8-cyclin C complex. J. Med. Chem. 66, 4932–4951 (2023).

Article   CAS   PubMed   Google Scholar

Lin, R. et al. Discovery of HyT‐based degraders of CDK9‐cyclin T1 complex. Chem. Biodivers. 20, e202300769 (2023).

Article   ADS   CAS   PubMed   Google Scholar

Brinda, K. V. & Vishveshwara, S. A network representation of protein structures: implications for protein stability. Biophys. J. 89, 4159–4170 (2005).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar

Slabicki, M. et al. Small-molecule-induced polymerization triggers degradation of BCL6. Nature 588, 164–168 (2020).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar

Sinko, W., Lindert, S. & McCammon, J. A. Accounting for receptor flexibility and enhanced sampling methods in computer‐aided drug design. Chem. Biol. Drug Des. 81, 41–49 (2013).

Article   CAS   PubMed   Google Scholar

Antunes, D. A., Devaurs, D. & Kavraki, L. E. Understanding the challenges of protein flexibility in drug design. Expert Opin. Drug Discov. 10, 1301–1313 (2015).

Article   CAS   PubMed   Google Scholar

Thomas, K. L. et al. Degradation by design: new cyclin k degraders from old CDK inhibitors. ACS Chem. Biol. 19, 173–184 (2024).

Article   CAS   PubMed   PubMed Central   Google Scholar

Akhtar, M. S. et al. TFIIH kinase places bivalent marks on the carboxy-terminal domain of RNA polymerase II. Mol. Cell 34, 387–393 (2009).

Article   CAS   PubMed   PubMed Central   Google Scholar

Blazek, D. et al. The cyclin K/CDK12 complex maintains genomic stability via regulation of expression of DNA damage response genes. Genes Dev. 25, 2158–2172 (2011).

Article   CAS   PubMed   PubMed Central   Google Scholar

Johannes, J. W. et al. Structure-based design of selective noncovalent CDK12 inhibitors. ChemMedChem 13, 231–235 (2018).

Article   CAS   PubMed   Google Scholar

Le Guilloux, V., Schmidtke, P. & Tuffery, P. Fpocket: an open source platform for ligand pocket detection. BMC Bioinform. 10, 168 (2009).

Article   Google Scholar

Seeber, M., Cecchini, M., Rao, F., Settanni, G. & Caflisch, A. Wordom: a program for efficient analysis of molecular dynamics simulations. Bioinformatics 23, 2625–2627 (2007).

Article   CAS   PubMed   Google Scholar

Ma, J. et al. iProX: an integrated proteome resource. Nucleic Acids Res. 47, D1211–D1217 (2019).

Article   PubMed   Google Scholar

Chen, T. et al. iProX in 2021: connecting proteomics data sharing with big data. Nucleic Acids Res. 50, D1522–D1527 (2022).

Article   CAS   PubMed   Google Scholar

Download references

Acknowledgements

We gratefully acknowledge financial support from the National Key R&D Program of China (2022YFC3400500 to C.L.; 2023YFD1800102 and 2022YFC2804800 to Z.L.), National Natural Science Foundation of China (22377013 to H.L., 81821005, U23A20108 and 92253303 to C.L.), the Fujian Provincial Natural Science Foundation (2021J01203 to H.L.), the State Key Laboratory of Drug Research (SKLDR-2023-KF-07 to H.L.), the Science and Technology Commission of Shanghai Municipality (YDZX20233100004032 to C.L.), the project of National Multidisciplinary Innovation Team of Traditional Chinese Medicine supported by National Administration of Traditional Chinese Medicine (ZYYCXTD202004 to C.L.), Major Program of Guangzhou National Laboratory (GZNL2023A02012 to C.L. and H.L.), High-level New R&D Institute (2019B090904008 to C.L.), High-level Innovative Research Institute (2021B0909050003 to C.L.), the Department of Science and Technology of Guangdong Province, the Zhejiang Provincial Natural Science Foundation of China (LY23H190002 to Z.L.); Research Funds from Hangzhou Institute for Advanced Study (2023HIAS-Y023 to Z.L.) and the Institutional Technology Service Center of Shanghai Institute of Materia Medica, Chinese Academy of Sciences (Shanghai, China) for technical assistance in mass spectrometry experiments and analysis. ChatGPT (chat.openai.com) was used to check grammar and make a few minor grammatical improvements to the text.

Author information Author notes

These authors contributed equally: Zemin Zhang, Yuanqing Li, Jie Yang, Jiacheng Li, Xiongqiang Lin, Ting Liu.

Authors and Affiliations

The School of Pharmacy, Fujian Medical University, Fuzhou, China

Zemin Zhang, Xiongqiang Lin, Ting Liu, Jin Lin & Cheng Luo

School of Chinese Materia Medica, Nanjing University of Chinese Medicine, Nanjing, China

Yuanqing Li, Shengyu Xue, Jiamin Yu, Cailing Tang & Kaixian Chen

State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China

Yuanqing Li, Jiacheng Li, Shengyu Xue, Jiamin Yu, Cailing Tang, Ziteng Li, Kaixian Chen, Hong Ding, Cheng Luo & Hua Lin

Key Laboratory of Microbial Pathogenesis and Interventions of Fujian Province University, the Key Laboratory of Innate Immune Biology of Fujian Province, Biomedical Research Center of South China, College of Life Sciences, Fujian Normal University, Fuzhou, China

Jie Yang, Shiling Yang & Hua Lin

University of Chinese Academy of Sciences, Beijing, China

Ziteng Li & Kaixian Chen

Zhongshan Institute for Drug Discovery, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Zhongshan, China

Liping Liu, Cheng Luo & Hua Lin

School of Pharmaceutical Science and Technology, Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences, Hangzhou, China

Zhengzheng Ye, Yanan Deng, Zhihai Li & Cheng Luo

State Key Laboratory of Functions and Applications of Medicinal Plants, School of Pharmacy, Guizhou Medical University, Guiyang, China

Hong Ding & Cheng Luo

Contributions

C.L. initialed the project, H.L. designed the compounds. Z.Z., J.Yang, X.L., and T.L. synthesized and characterized the compounds using HPLC, LC-MS, and NMR. J. Lin, J.Yang, and S.Y. performed the HPLC assay. Y.L., J.Li., S.X., J.Yu., C.T., Zi.L., and L.L., designed and performed kinase assay and all cell experiments. Z.Y., Y.D., and Zh.L. provided DDB1 protein, Z.Z. and Y.L. wrote the manuscript and completed computer-aided drug design. H.L. revised the manuscript. K.C., H.D., C.L., and H.L. supervised the whole research. All authors discussed the results and commented on the manuscript.

Corresponding authors

Correspondence to Hong Ding, Cheng Luo or Hua Lin.

Ethics declarations Competing interests

The authors declare no competing interests.

Peer review Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information Supplementary Information (download PDF ) Peer Review File (download PDF ) Reporting Summary (download PDF ) Source data Source Data (download ZIP ) Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article Cite this article

Zhang, Z., Li, Y., Yang, J. et al. Dual-site molecular glues for enhancing protein-protein interactions of the CDK12-DDB1 complex. Nat Commun 15, 6477 (2024). https://doi.org/10.1038/s41467-024-50642-0

Download citation

Received 06 January 2024

Accepted 18 July 2024

Published 01 August 2024

Version of record 01 August 2024

DOI https://doi.org/10.1038/s41467-024-50642-0

Share this article

Anyone you share the following link with will be able to read this content:

Get shareable link

Provided by the Springer Nature SharedIt content-sharing initiative

Subjects Chemical modification Proteins Small molecules This article is cited by CDK12 and CDK13 in oncology: from RNA regulation to therapeutic targeting Julia Dudkiewicz-GarbiczPaweł K. Włodarski

Cellular Oncology (2026)

DDB1 engagement defines the selectivity of S656 analogs for cyclin K degradation over CDK inhibition Céline MoisonRodrigo Mendoza-SanchezGuy Sauvageau

EMBO Reports (2025)