Targeted Degradation of CDK9 Potently Disrupts the MYC-Regulated Network

Summary

CDK9 coordinates signaling events that regulate transcription and is implicated in oncogenic pathways, making it an actionable target for drug development. While numerous CDK9 inhibitors have been developed, success in the clinic has been limited. Targeted degradation offers a promising alternative. A comprehensive evaluation of degradation versus inhibition is needed to assess when degradation might offer superior therapeutic outcomes. We report a selective and potent CDK9 degrader with rapid kinetics, comparing its downstream effects to those of a conventional inhibitor. We validated that CDK9 inhibition triggers a compensatory feedback mechanism that dampens its anticipated effect on MYC expression and found that this was absent when degraded. Importantly, degradation is more effective at disrupting MYC transcriptional regulation and subsequently destabilizing nucleolar homeostasis, likely by abrogation of both enzymatic and scaffolding functions of CDK9. These findings suggest that CDK9 degradation offers a more robust strategy to overcome limitations associated with its inhibition.

Graphical Abstract

Significance

CDK9 plays an important role in the regulation of MYC expression and its downstream activity, both of which are frequently altered in cancers. However, prolong inhibition of CDK9 can lead to a compensatory increase in MYC levels, posing a challenge for therapeutic development. We developed a selective CDK9 PROTAC degrader with rapid kinetics, allowing us to differentiate the effects of CDK9 degradation from its inhibition in the regulation of MYC. We show that CDK9 degradation leads to a rapid downregulation of MYC and prevents the paradoxical increase observed with inhibition. The mechanistic difference between the two modalities was also evident in their differential effect on nucleolar homeostasis, a MYC-regulated cellular process. This offers a possible synthetic lethality avenue to exploit in the development of CDK9-targeted therapies.

eTOC blurb

Toure et al. investigate a known challenge in CDK9 inhibition – the compensatory increase in MYC levels. They demonstrate that a selective CDK9 degrader with rapid kinetics circumvents this effect, providing mechanistic insights into degradation versus inhibition and highlighting its potential therapeutic advantages.

Introduction

MYC gene expression is an important hallmark of stimulated signaling pathways that promote cell proliferation^1,2. Dysregulation of MYC expression resulting from genomic amplification or increased gene copy number, among a variety of other genomic alterations, is a key driver in cancer development and progression³. Thus, the suppression of MYC transcription and its downstream programs has been a long-standing goal in the development of cancer therapeutics.

CDK9 is the catalytic subunit of the positive transcription elongation factor b (P-TEFb), and is crucial for the regulation of MYC. MYC gene expression depends highly on CDK9 which is tethered to the MYC promoter through interaction of the P-TEFb complex with BRD4^2,4 and in some cellular context the scaffolding activity of the ERK protein¹. CDK9’s influence on MYC dynamics extends to the protein level as well. CDK9 mediated phosphorylation of MYC at serine 62 protects the oncoprotein from degradation⁵. Consequently, MYC genomic amplification induces an increased dependence on CDK9 which in turn becomes critical for the maintenance of MYC addicted tumor state⁶. More than twenty CDK9 inhibitors have been evaluated in clinical trials involving both hematologic and solid tumors. Unfortunately, a combination of off- and on-target toxicity and the lack of objective response has restricted progress to FDA approval for these agents⁷. Sustained inhibition of CDK9 can induce a compensatory increase in MYC expression². This mechanism of resistance is thought to be driven by activation of inactive cellular CDK9 through the bromodomain protein BRD4 and its channeling of available CDK9 to the MYC promoter².

The sustained interest in advancing molecules with desirable pharmacological attributes against MYC and CDK9 has been bolstered by advances in targeted protein degradation⁸. It is likely that an acute and potent degradation of CDK9 would circumvent the resistance mechanism described above and lead to a more robust attenuation of MYC activity. While the pharmacodynamics of inhibitors are driven by drug concentration, those of degraders appear to be driven by the target resynthesis rate⁹. Moreover, degraders have been shown to act sub-stoichiometrically with rapid kinetics⁸. These attributes may confer PROTACs and other degraders an advantage in the form of greater resilience to drug resistance induced by long-term exposure to high concentration of small molecule inhibitors^8,10.

Moreover, targeted protein degradation may offer unique advantages over other modalities to study the transient and temporal changes in cellular signaling networks resulting from the acute depletion of proteins. Accordingly, several CDK9 degraders have been reported over the last few years^11-18 that effectively degrade the kinase.

Here, we undertook our own efforts to design PROTAC molecules using the ultra-selective CDK9 inhibitor KI-ARv-03¹⁹, which was previously discovered in our group and the starting point for the clinical candidate KB-0742²⁰ (istisociclib) that was recently evaluated in Phase I/II clinical trials for patients with relapsed or refractory solid tumors or Non-Hodgkin Lymphoma (NCT04718675)^21-23. We integrated transcriptional, proteomics, phosphoproteomics, and imaging approaches to comprehensively characterize the downstream effects and cellular adaptations that result from CDK9 degradation.

We report a selective and potent targeted CDK9 degrader that rapidly downregulates MYC levels. In addition, we explore cellular adaptations to CDK9 degradation and gain new insights into potential biomarkers indicative of contexts in which degradation might work better than inhibiting the kinase activity of CDK9. Our findings suggest that the selective degradation of CDK9 presents an attractive strategy for a robust attenuation of dysregulated MYC transcription and could provide an optimal therapeutic path for patients with aggressive and metastatic MYC-driven cancers.

Results

KI-CDK9d-32 is a potent CDK9 degrader with rapid kinetics

We designed and synthesized a first set of CDK9 PROTACs using KI-ARv-03 as the CDK9 recruitment moiety. The Small Molecule Microarray (SMM) platform used to discover KI-ARv-03 offers a suitable and intuitive starting point for the development of PROTAC molecules. Specifically, the original SMM screen involved compounds captured on isocyanate-coated glass slides, enabling rapid nomination of potential exit vectors for the attachment of linkers for PROTAC generation²⁴. The intrinsic primary amine of the KI-ARv-03 molecule was used for linker connection. To construct the initial library of bivalent degraders, we used commercially available linkers to gain rough estimates of linker length and rigidity needed for degradation of CDK9 using KI-ARv-03 and Pomalidomide as recruiting element for the ubiquitin ligase cereblon (CRBN)^25,26.

The potential of these compounds to reduce CDK9 protein levels in MOLT-4 cells, a T-lymphoblast cell line with MYC dependence and where CDK9 is a known therapeutic target, was evaluated. The cells were treated with 1 μM of each compound for six hours. CDK9 protein levels were assessed using Western blot.

Compound KI-CDK9d-08, a compound with a biphenyl linker (Figure 1A and 1B(i)), stood out as the most effective PROTAC from our preliminary library. We attributed this result to the rigidity of the biphenyl linker leading to favorable ternary complex formation with CDK9 and CRBN³⁰.

Figure 1: KI-CDK9d-32 is a potent CDK9 degrader with rapid kinetics.

A) Summary of KI-CDK9d-32 development. KI-ARv-03 and pomalidomide were used as the CDK9 binding and CRBN binding moieties, respectively. Linker optimization followed prevailing standards for achieving better properties. B) Luminescence evaluation of endogenous HiBiT-tagged CDK9. HiBiT is a CRISPR-mediated system that enables the development of luminescence-based assays to quickly assess the levels of endogenous proteins²⁷. Using this assay, CDK9 level was evaluated in MOLT-4 cells after 4-hours of treatment with KI-CDK9d-08 (i) or KI-CDK9d-32 (ii). C) Kinetics evaluation of KI-CDK9d-32 at 1, 2, 4, 6 and 12 hours using luminescence readout of HiBiT-tagged CDK9 in MOLT-4. The curves were fitted to the one-parameter exponential decay model²⁸ and degradation rates at each concentration estimated. The kinetics profile follows that of a classical hook-effect²⁹ model wherein there is a goldilocks effect, the maximal rate of degradation is bounded by a low and a high concentration. The rate at 16 nM was 0.83 Hr⁻¹, the fastest rate of degradation (1.04 Hr⁻¹) for our experimental data occurs at 126 nM, and drops to 0.80 Hr⁻¹ at 355 nM when the hook-effect starts to dominate.

Considering the insights gained from our initial degrader library, subsequent optimization efforts were directed toward the KI-CDK9d-08 scaffold (Figure 1A). We employed emerging design strategies from the targeted degradation space, such as linker optimization and IMiD substitutions, to enhance degradation efficiency and minimize unintended polypharmacology effects resulting from the neo-substrates targeted by CRBN²⁶. This allows for a precise assessment of the biological effects driven primarily by CDK9 degradation. The field has seen significant advancements in the design and optimization principles for PROTAC molecules, especially after identifying a suitable starting point^26,31. The incorporation of saturated rings, like piperidine and piperazine, into the linker has demonstrated improvements in potency and physicochemical characteristics, including solubility and metabolic stability³¹. Moreover, altering the attachment point of the linker to the phthalimide ring of the IMiD moiety and replacing pomalidomide with lenalidomide have proven to be effective for diminishing the known off-target effects, mainly CRBN neo-substrates, of pomalidomide-based PROTACs³¹.

These strategies enabled us to successfully design a second-generation collection of degraders with enhanced properties from the KI-CDK9d-08 scaffold. We evaluated the degradation efficiency of our optimized compounds in cell-based assays and selected KI-CDK9d-32 (Figure 1A), a sub-nanomolar degrader, as the leading candidate for generating biological insights. KI-CDK9d-32 achieves near 100% degradation and a DC50 of 0.89 nM following 4 hours of treatment in MOLT-4 cells (Figure 1B(ii)). Consistent with the “hook-effect”, which is often observed with PROTACs and reflects the saturation of both protein members of the ternary complex with their respective binding elements and rendering the formation of a ternary complex unfavorable above a certain concentration threshold³², we observed limited degradation efficiency starting at 355 nM for the early time-points. Further assessment degradation kinetics showed that this effect became increasingly negligeable as treatment time elapsed. CDK9 levels remained below 10% of the DMSO baseline for at least 12 hours after treatment (Figure 1C). We believed this convergence to complete CDK9 degradation over the course of our study to be the result of slower CDK9 resynthesis rates²⁹ and possibly a positive feedback mechanism in the CDK9-MYC axis.

KI-CDK9d-32 is a highly selective and potent CDK9 degrader that induces rapid reduction of MYC protein levels

KI-CDK9d-32 effectively reduced CDK9 levels through the ubiquitin-proteasome pathway, as confirmed by co-treatment with the proteasome inhibitor MG132 (Figure 2A). Additionally, the effect of the degrader on CDK9 levels was rescued through co-treatments with excess KB-0742, pomalidomide, or KI-CDK9d-32N, a negative control PROTAC with impaired CRBN binding. The extent of CDK9 rescue correlated with downstream functional effects on MYC levels and POLR2A C-terminal domain serine-2 phosphorylation, particularly with pomalidomide and MG132 co-treatments. In contrast, co-treatment with CDK9 binders resulted only in partial rescue, likely because high levels of inhibition can mirror the effects of degradation on these functional markers. The observed compensatory increase in MYC following KB-0742 treatment is consistent with previous studies².

Figure 2: KI-CDK9d-32 is a highly selective and potent CDK9 degrader that induces rapid reduction of the levels of MYC and MYC-dependent regulatory kinases.

A) Western Blots showing degradation and rescue assessments in MOLT-4 after 4 hours of treatment. Concentrations used: KI-CDK9d-32 (50 nM (++), or 15 nM (+) for co-treatments); KB-0742 (5 μM); Pomalidomide (5 μM); KI-CDK9d-32N (5 μM); MG132 (5 μM). The Cofilin protein was used as internal reference. A set of parallel experiments in PSN-1, RH-4, and HeLa are reported in supplementary Figure 1C. B-C) Quantitative mass-spectrometry assessment of the protein-level effects of KI-CDK9d-32. KI-CDK9d-32 has a robust effect on MYC-driven processes based on proteomics assessments in MOLT-4 cells. Cells were treated with DMSO or 50 nM of KI-CDK9d-32 in four biological replicates. Protein lysates were harvested after 4 hours of compound exposure. (B) Volcano plot representation of the 4 hour time-point. CDK9, MYC, and MYC target genes from one of the molecular signatures database (MSigDB) Hallmark collection are shown. (C) Top 10 up and down regulated proteins from A. D) Enrichment analysis of the top 10% of genes that were differentially impacted – (i) enriched MSigDB Hallmark pathways and (ii) enriched KEGG pathways. The “Enrichment Ratio” is the ratio of observed significant proteins in a cluster belonging to a given pathway to the total number of proteins that could have been identified in that category in the background universe. In this case, we used all proteins identified in the experiment as the cell-specific background. E) Evaluation of differential peptide phosphorylation after 4 hours of MOLT-4 exposure to 50 nM of KI-CDK9d-32. The log2[Normalized Ratio] is the ratio of DMSO-normalized peptide to protein abundance in MOLT-4 cells treated with 50nM KI-CDK9d-32 for 4 hours. Analysis utilized four biological replicates and the Limma package for differential analysis post-median normalization. F) Kinase-substrate enrichment analysis of the significantly enriched or depleted phosphopetides (adjusted p-value < 0.05, and absolute value log2[Normalized Ratio] > .3) using the PhosphoSitePlus kinase prediction web interface (phosphosite.org/kinaseLibraryAction). CDK9 was predicted as the sole significantly inhibited kinase after correcting the p-values for multiple hypotheses.

As selectivity is a critical consideration when targeting kinases, we employed quantitative mass spectrometry to perform an assessment of the selectivity of degradation and specific impacts of KI-CDK9d-32 on proteins in MOLT-4 cells. CDK9 and MYC exhibited the most pronounced reduction in protein levels. We observed a 5-fold reduction in CDK9, which coincided with a 3-fold decrease in MYC protein levels (Figures 2B and 2C). Additionally, the degradation of CDK9 and MYC was followed by reduction in the levels of kinases in the MYC regulatory network (i.e., AURKA, PLK1, CDK11B, and WEE1). The expression of these kinases were previously shown to depend on MYC in double-hit lymphomas³³. Moreover, AURKA and PLK1 interact with PTEN and PI3K/AKT to form a regulatory axis that drives MYC overexpression in aggressive lymphomas, establishing the likelihood of a positive feedback loop between these kinases and MYC^34-36. We further validated this effect at the transcript level through a RT-qPCR quantification of the mRNA transcripts of MYC, AURKA, and PLK1 (Supplementary Figure S3A).

Given that PROTACs can function as inhibitors and potentially have off-targets even if they don’t degrade them, we carried out a global kinase inhibition assay against wild kinases at Km[ATP] and .1 and 10 μM of the degrader. We selected these concentrations to assess selectivity at the maximal concentration used in our studies (i.e., 100 nM) as well as match the concentration used in profiling the parent inhibitor to ensure comparability and identify a subset of kinases for biochemical IC50 determination. At 100nM, the degrader inhibits about 90% of CDK9 activity. No other kinase is impacted at that concentration (Supplementary Table S3).

At 10 μM of KI-CDK9d-32, we identify a subset of 14 kinases whose activities were inhibited (>75% inhibition, Supplementary Figure 1A and Table S2). We then followed up with determining the biochemical affinities (IC₅₀s) of these kinases and compare them to data previously collected on the parent molecule, KI-ARv-03¹⁹. Developing KI-ARv-03 into a PROTAC resulted in a more potent CDK9 binder with CDK9/Cyclin T1 IC₅₀ = 3 nM at 10 μM ATP (Supplementary Figure 1B). The compound was found to be selective for CDK9/Cyclin T1 with more than 140X fold selectivity over other kinases profiled except for CDK19/Cyclin C which was at 40X. This was an exciting result given the challenge in achieving selectivity against this class of proteins.

We conducted pathway over-representation analyses on the top 10% of proteins that were enriched (“Up”) or depleted (“Down”) in the quantitative mass spectrometry data. We performed these analyses using gene sets from the Molecular Signatures Database (MSigDB)³⁷ and the Kyoto Encyclopedia of Genes and Genomes (KEGG)³⁸ to identify pathways that were significantly enriched in the two clusters. The results demonstrated that proteins involved in cell cycle regulation, response to cellular stress and growth signals were depleted, with G2M checkpoint and E2F targets being the most significant negatively enriched pathways on treatment with the degrader, KI-CDK9d-32 (Figure 2D(i)). Depletion of proteins in these pathways indicates a strong inhibitory effect on cell cycle progression suggesting a coordinated cellular effort to halt proliferation and growth, in response to CDK9 degradation. Moreover, ribosome biogenesis is the most significantly KEGG enriched pathway for proteins in the “Up” cluster after the 4-hour treatment with 50 nM KI-CDK9d-32 (Figure 2D(ii)). This positive enrichment of ribosomal proteins in the proteomic dataset was in alignment with nucleolar stress condition during which release of nucleolar proteins to the nucleoplasm and cytoplasm occurs thus increasing their solubility³⁹. This is further substantiated by the depletion of MYC and stabilization of TP53 observed following compound treatment (Figure 2B-C)³⁹. These findings substantiate the degrader’s induction of a considerable disruption of the regulatory network associated with MYC.

We carried out global quantitative phosphoproteomics (Ser/Thr) evaluation to further assess the selectivity of the kinase inhibitory activity of KI-CDK9d-32 as well as identify key early signaling adaptations. To deconvolve changes in protein abundance from those of the phosphorylation levels, we normalized the fold changes in the abundance of phosphopeptides to those at the protein level. The normalized ratios isolate the changes in the phosphorylation status. Depleted sites were enriched in CDK9 motifs as predicted by kinase-substrate enrichment analysis (Figure 2E and 2F)⁴⁰, validating the selectivity of KI-CDK9d-32. The top proteins with differentially altered phosphosites were predominantly enriched in mRNA metabolism, ribosomal maturation, and protein synthesis processes. For instance, we observe significant reduction in phosphorylated 4EBP1, the protein encoded by the EIF4EPB1 gene. This reduction in phosphorylation at specific residues, including Threonine-37/Threonine-46, is known to favor the interaction between 4EBP1 and eIF4E, a factor that is essential for the recruitment of the 40s ribosomal subunits to the 5’ end of mRNA⁴¹. Thus, 4EBP1 binding to eIF4E induces an inhibitory effect on translation^35,41,42. The phosphorylation state of nucleolar proteins plays an important role in the assembly of the nucleolus as well as its structural and functional integrity⁴³.

CDK9 degradation disrupts nucleolar homeostasis, a MYC-regulated process

Ribosome biogenesis, the process that governs the assembly of the ribosomal subunits in the nucleolus, essential for protein synthesis, is tightly regulated by MYC. The suppression of MYC network in cancer can lead to a collapse of ribosome biogenesis and induce widespread suppression of protein synthesis^44,45. This can be especially disruptive to cancer cells given their elevated reliance on protein synthesis for aberrant proliferation. Numerous studies have demonstrated that CDK9 inhibition or silencing can have downstream effects on ribosomal biogenesis processes^11,46,47. Consistent with these results, our proteomics analyses found disruption in ribosome biogenesis as one of the most significantly impacted cellular process following CDK9 degradation using KI-CDK9d-32. As seen with earlier studies¹¹, we find the effect from CDK9 degradation on ribosome biogenesis to be stronger than that induced by inhibition (Figure 2D(ii)). Given the clear biological phenotype emerging from our data, we carried out high resolution immunofluorescence microscopy to explore this further.

The Nucleophosmin protein (NPM1) and the RNA-binding protein DExD-box helicase 21 (DDX21) are key markers of nucleolar dynamics. NPM1 is a scaffold protein that plays a critical role in the assembly of the nucleolus and is localized at the nucleolar rim, the outer region of the nucleolus^48,49. DDX21 is localized at the core of the granular compartment and is known to engage in several protein-protein interactions that drive ribosomal RNA metabolism⁵⁰ (Figure 3B). Thus, we directly monitored the impact of the compounds on the nucleolar structural stability by staining HeLa cells for NPM1 and DDX21. Ribosome biogenesis has been extensively studied in HeLa cells^51,52, making them a well-suited model for these immunofluorescence experiments.

Figure 3: KI-CDK9d-32 potently destabilizes nucleolar homeostasis.

A) Fluorescence imaging of HeLa cells following treatment with the inhibitor, degrader, and relevant controls (DMSO, Negative Degrader, and Actinomycin D, a well-known inducer of nucleolar stress). Images were processed using ImageJ. KI-CDK9d-32N is a negative control degrader, lacking CRBN binding. Supplementary figure S4 provides additional details on the dose-response effects from both KB-0742 and KI-CDK9d-32. B) Illustration of the nucleolar compartments.

KI-CDK9d-32 treatment destabilized nucleolar homeostasis as early as 2 hours, indicating that potent CDK9 degradation destabilizes nucleolar homeostasis by targeting, either directly or indirectly, the nucleolar rim, the outer layer of the nucleolus defined by NPM1 (Figure 3A). Interestingly, this phenotype diverges from the classical model of nucleolar stress. Unlike Actinomycin D, a known inducer of nucleolar stress, CDK9 degradation had little to no impact on the core granular compartment of the nucleolus. Additional brightfield imaging of HeLa cells treated with KI-CDK9d-32 confirmed that this compound induces significant nucleolar disruption, even in the absence of specific staining, reinforcing the conclusion that CDK9 degradation selectively impacts nucleolar architecture (Supplemental Figure S4C). This is further substantiated by a RT-qPCR analysis that revealed that KI-CDK9d-32, like Actinomycin D, significantly reduced pre-rRNA levels, indicating nucleolar disruption (Figure S4D). However, the effect of Actinomycin D was more pronounced, which aligns with its established role in inhibiting RNA polymerase I, the regulator rRNA synthesis³⁹.

These results demonstrate that KI-CDK9d-32 mediated depletion of CDK9 and MYC levels leads to a robust disruption of MYC regulated processes. While MYC is known to regulate ribosome biogenesis at the transcriptional level, we show here that disrupting the MYC-CDK9 transcriptional network causes the collapse of the nucleolus, the site of ribosome biogenesis.

CDK9 degradation induces sustained disruption of MYC, and bypasses a known compensatory mechanism

The degradation of CDK9 led to rapid and potent downregulation of MYC at the protein level, which is significant given MYC’s inherently fast turnover. This observation suggests extremely rapid CDK9 degradation kinetics that effectively shuts down MYC transcription. In contrast, previous studies have reported a compensatory increase in MYC levels following CDK9 pharmacological inhibition². CDK9 degradation, on the contrary, seems to robustly bypass this resistance mechanism⁵³ and as a result likely to have a pronounced impact on MYC-driven processes.

To substantiate these results, we conducted qPCR assays to compare the effects of KI-CDK9d-32 and KB-0742 on CDK9 and MYC transcripts levels. Consistent with the findings of Lu et al.², we observed an increase in MYC mRNA levels within 2 hours of CDK9 inhibition in a dose- and time-dependent manner (Figure 4A and Supplementary Figure S3A). In contrast, treatment with KI-CDK9d-32 significantly suppressed transcription of MYC at all concentrations tested, also in a dose- and time-dependent manner. This suggests that CDK9 may play a scaffolding role at the MYC promoter, in line with a recent study by Agudo-Ibañez et al. that demonstrated the role of ERK in “anchoring CDK9 to the MYC promoter”¹. Interestingly, the effects on CDK9 transcript levels did not provide a clear distinction between degradation and inhibition during the early time points tested. However, on aggregate, the data suggest a positive feedback loop between MYC and CDK9. Other factors, such as transcriptional kinetics and assay sensitivity, may obscure the observed effect on CDK9.

Figure 4: CDK9 degradation induces sustained disruption of MYC, and bypasses a known compensatory mechanism.

A) RT-qPCR averaged across 3 biological replicates, and at least 3 technical replicates for all conditions. Degrader KI-CDK9d-32 repression of MYC mRNA levels is consistent in a time and dose-dependent manner (15 nM – 125 nM). Inhibition with KB-0742 shows a paradoxical increase in MYC transcripts at earlier time-points. The relative expression was determined using the abundant POLR2A as reference, and DMSO-treated samples as control. B) RNA sequencing evaluation of transcript levels following 4 hours of treatment with 1.2 μM of KB-0742 and 15 nM of KI-CDK9d-32. LFC is the log2 of the ratio of Inhibitor/Degrader. The transcripts of genes with LFC > 0 are differentially downregulated by the degrader. C - E) Enrichment analysis of the top 20% of genes that were differentially downregulated in degrader versus inhibitor at multiple time points. As in B, LFC is the log2 of the ratio of Inhibitor/Degrader.

To validate the above results and identify transcripts that display differential responses to the two pharmacological approaches, we carried out RNA sequencing experiments using synthetic RNA spike-ins to control for changes in the overall transcriptional state⁵⁴. MOLT-4 cells were treated with 15 nM and 1.2 μM of KI-CDK9d-32 and KB-0742, respectively, for 2, 4, and 8 hours across four biological replicates. Consistent with our qPCR data, CDK9 degradation using 15 nM KI-CDK9d-32 induces a rapid and sustained downregulation of MYC mRNA expression more than 6-fold relative to CDK9 inhibition (Figure 4B). Moreover, this reduction in MYC expression triggers a repression of CDK9 transcript levels in the RNA sequencing data, indicative of a potential positive feedback mechanism. This positive feedback mechanism becomes very apparent at the 8-hour mark, where CDK9 expression is reduced by about 5-fold by the degrader.

To further analyze the transcriptional signature of each compound across the time-points tested, we performed over-representation enrichment analyses. Recognizing the limitations of relying solely on applying fixed fold-change thresholds, which may overlook biologically significant changes at earlier time-points, we employed a percentile-based significance filter. Specifically, included genes in the enrichment analyses were required to fall within the top 20% of absolute fold changes within a given condition tested. This allowed us to capture genes that show similar trends in their differential expression across timepoints even if their fold changes at earlier timepoints are modest.

We observed subtle temporal differences between CDK9 degradation and inhibition. Degradation had a potent effect on early growth signaling and innate inflammation as evidenced by the negative enrichment of TNF-α Signaling Via NF-κB, and the Unfolded Protein Response pathways⁵⁵ (Figure 4C). It appears that CDK9 degradation rapidly downregulated BRD4-mediated processes, which was expected given the structural and functional relationship between the P-TEFb and BRD4^4,55,56. Moreover, the MYC Targets gene sets were differentially impacted. Comparing the effects of the two pharmacological approaches, we found that degradation had the strongest effect on the Hallmark “MYC Targets V2”, the set most correlated with ribosome biogenesis, with the tryptophan-aspartate repeat domain 43 (WDR43) one of the topmost downregulated. The effects of CDK9 Inhibition, on the other hand, was most pronounced against the DNA damage pathway (Figure 4C-E, Supplementary Figure S3C).

In parallel to the above experiment in MOLT-4, we also evaluated the effects of KI-CDK9d-32 and KB-0742 treatment on transcriptional programs in two additional cell lines, the pancreatic adenocarcinoma cell line PSN-1 and the rhabdomyosarcoma line RH-4 to determine the presence of persistent signals across different cell lines. These experiments were carried out at the same concentrations as those described above, for durations of 4 and 8 hours.

The TNF-α Signaling Via NF-κB hallmark pathway was the most persistently differentially downregulated across all cell lines in the degrader treated samples relative to the samples treated with the inhibitor. This pathway was consistently the earliest downregulated pathway in the samples treated with the degrader across all three cell lines. At the other end, the inhibitor appears most effective against genes involved in the maintenance of genomic stability. The DNA-repair pathway was significantly downregulated by the inhibitor in 2 of the three cell lines (Figure 4C-E). These observations are reflective of important mechanistic differences driven in part by complete abrogation of CKD9 kinase and scaffolding functions by the degrader. The molecular consequence of which is a breakdown of the BRD4-P-TEFb complex, an insight that aligns with the rapid and sustained degradation of critical early response genes regulated by BRD4 (e.g., MYC and IFRD1)^2,4,55. Moreover, the kinetics of action of degradation versus inhibition of CDK9 is likely to be an important determinant. This underscores the significance of a probe with the rapid kinetics to enable a temporal assessment of the primary responses to inhibition versus degradation as observed in this study.

Overall, the degrader KI-CDK9d-32 appears to have more pronounced effects than the inhibitor KB-0742 on transcriptional repression of genes important for proliferation especially in MOLT-4 and PSN-1. This trend is reversed for most genes in the context of RH-4, where the inhibitor’s effects on MYC transcription appeared to be stronger. The subtle differences exhibited might be due to differences in transcriptional programs driven by lineage-specific core transcriptional regulatory circuitries^57,58 (Supplementary Figures S5).

Response to KI-CDK9d-32 depends on CRBN and the activity of ABC transporters

We assessed the effect of KI-CDK9d-32 and its negative control KI-CDK9d-32N on the viability of an initial panel of three cell lines, MOLT-4, PSN-1, and RH-4. The viability effects were compared to KI-ARv-03, KB-0742, and a known CDK9 degrader, Thal-SNS-32. KI-CDK9d-32 induced the most pronounced cytotoxic effects relative to the inhibitors and degrader in the comparison set (Figure 5A and 5B).

Figure 5 : KI-CDK9d-32 demonstrates strong sensitivity in MOLT-4, PSN-1, and RH-4.

A) Dose-response curves from three cell lines (MOLT-4, PSN-1, and RH-4) measured 120 hours post-treatment with a panel of compounds: KI-CDK9d-32, KI-CDK9d-32N, KI-ARv-03, KB-0742, and Thal-SNS-32. X-axis is log transformed with concentration shown in anti-log. Error bars indicate the mean ± standard deviation from n=3 technical replicates. B) Tabulated IC₅₀ values for the corresponding conditions at 72 and 120 hours post-treatment. Concentration ranges for degraders were 0 – 500 nM, while inhibitors and KI-CDK9d-32N were tested from 0 – 10 μM.

As CDK9 perturbation continues to be pursued as a therapeutic strategy for a variety of cancers, we were interested in determining which cellular models, if any, are likely to be most impacted by CDK9 degradation. We carried out broad cell line sensitivity profiling of ~800 cancer cell lines through the Broad Institute’s PRISM sensitivity profiling platform⁵⁹.

Given the pan-essential nature of CDK9⁶⁰, we anticipated broad cytotoxic activity with both agents. Consistent with this assessment, treatment with KB-0742 and KI-CDK9d-32 had strong cytotoxic effects. The mean IC₅₀ values were 926 nM and 82.6 nM for KB-0742 and KI-CDK9d-32, respectively. KB-0742 had a widespread cytotoxicity starting at 3.3 μM. The degrader exhibited a more selective cytotoxicity profile up to the maximal applied dose of 1.5 μM (Figure 6A and supplementary Figure S6A), about 100X the concentration used for the transcriptional profiling.

Figure 6: KI-CDK9d-32 has strong cytotoxic effects relative to the inhibitors used, but activity is limited in cells with high ABCB1 level.

A) Distribution of the IC₅₀ values resulting from a pooled screen of ~800 cell lines through the Broad Institute’s PRISM platform. PRISM offers a high throughput approach for compound screening in cancer cells derived from several lineages. KB-0742 and KI-CDK9d-32 were evaluated in 9-point three-fold dilutions series with top concentrations of 30 μM and 1.5 μM, respectively. B) Top significant Depmap gene expression biomarkers (q values < .05) of sensitivity to KI-CDK9d-32 (log2(IC₅₀)). CRBN and ABCB1 were top markers of sensitivity and resistance, respectively (supplement S6B provides an expanded figure). C) Correlation between the Z-scores of AUC values from the PRISM pooled screen and a secondary individual cell line screen. D) UMAP dimensionality reduction followed by HDBSCAN clustering of cell lines in both the PRISM pooled screen and the secondary individual cell line screen. The Z-score of IC₅₀ and AUC values, relative metrics of compound sensitivity in the respective platforms, were used as inputs into UMAP to obtain a reduced dimensional data that used in HDBSCAN to obtain unbiased clustering of cell lines based on sensitivity patterns to the degrader and inhibitor. E, F) Dose-response plots of representative sets of cell lines from the clusters obtained in D. Viability values are relative to DMSO. Error bars represent the mean +/− sd from three and two replicates from the PRISM screen and secondary screen, respectively. The cell lines plotted in each cluster are listed in Supplementary Figure S6D. G, H) Comparison of IC₅₀ values of the full cluster constituents from the PRISM screen. I, J) Show comparison to the expression profile of the strongest biomarkers from (B). For G through J, statistical comparisons were done using the pair-wise Wilcoxon rank-sum test between cluster 1 and the rest of the clusters, respectively. Statistical significance symbols as follows: ns: p > .05, *: p <= 0.05, **: p <= 0.01, ***: p <= 0.001, ****: p <= 0.0001

To determine the strongest biomarkers driving response to CDK9 degradation using KI-CDK9d-32, we examined the correlations between area-under the dose-response curve (AUC) values and the multi-omics features available through the DepMap portal following the standard PRISM analysis workflow⁵⁹. CRBN expression, protein level, and copy number alteration consistently emerged as the most significant driver of sensitivity for degrader KI-CDK9d-32 (Figure 6B, Supplementary S6B). These observations are likely attributable to KI-CDK9d-32 as a lenalidomide-based degrader that induces the proximity of CDK9 and CRBN. Moreover, at the other end of the response spectrum, ABC-transporter mediated efflux activity surfaced as the top driver of resistance to KI-CDK9d-32: the gene expression and protein levels for ABCB1 were significantly correlated with higher AUC values (Figure 6B). This finding aligns with an earlier report on the role of ABCB1 in promoting resistance to PROTAC-based degraders⁶¹, and was consistent with results from a Madin-Darby canine kidney (MDCK) permeability assessment, an assay employed to determine whether compounds of interest are substrates for efflux transporters (Data not shown).

In line with the proteomic and transcriptomic datasets discussed above, we also observed the ribosomal protein L15 (RPL15) as one of the top 10 biomarkers that correlated with response (low IC₅₀ values) to KI-CDK9d-32 (S6B). Although we observed robust degradation of CDK9 and MYC in biochemical assays, their expression levels were not among the top correlates with cell viability in our cell line profiles, highlighting the complex interplay of these genes in cancer biology.

To validate the above insights from the PRISM screen, we evaluated viability effects from the degrader against a set of 300 cell lines from a previously described cell repository at the MGH Cancer Center⁶². Cells were treated with either KI-CDK9d-32, or KI-CDK9d-32N, the negative control analogue of the degrader that differs only in the addition of methyl group to the glutarimide ring of lenalidomide and thereby preserving CDK9 binding while preventing recruitment CRBN. There was strong correlation between AUC values from the PRISM pooled and the non-pooled viability screening approaches (Figure 6C).

The impact of transporter activity complicated our efforts to sufficiently delineate cell lines differentially responsive to KI-CDK9d-32 and KB-0742 based on underlying mechanism. Despite this limitation of our data, we were able to identify clusters of cell lines that showed stronger sensitivity to one treatment over the other (Figure 6D,E,F). Interestingly, high expression of ABCB1 had similar effects on the response to both drugs, allowing a deeper dive into other biomarkers (Figure 6G-I). Supplementary Table S1 provides a full list of the cells used in this analysis as well as their response patterns. Clusters 1, 2, and 4 exhibited similar sensitivity levels to both compounds, while cluster 3 was responsive to degradation but resistant to inhibition. Pathway enrichment analysis of the top gene expression correlates identified genes involved in the structural regulation of the ribosome (enriched in cluster 3) as key determinants of sensitivity to KI-CDK9d-32. Further analysis of the clusters revealed that KRAS activation was correlated with resistance to inhibitor KB-0742 (Supplementary Figure S6C). KRAS activation triggers the MAPK/ERK and PI3K/AKT pathways that are upstream of MYC and may compensate for the inhibition of CDK9^1,5. CDK9 degradation not only prevent a compensatory increase in MYC but may also be resilient to other mechanisms of resistance linked to inhibition. Additional studies may be fruitful in characterizing the scaffolding functionalities of CDK9 in the regulation of MYC transcription.

Discussion

Targeting MYC dependencies has been a significant therapeutic goal, promising to develop cures for various cancers. Approximately 15% of the global transcriptome is found to be dependent on the MYC regulatory network⁶³. Dysregulation of MYC is estimated to occur in a vast number of cancers, with suggestions that up to 70% of cancers may exhibit MYC alterations⁶⁴. The P-TEFb protein complex, which consists of CDK9 and CCNT1, plays a crucial role in the MYC network, affecting MYC’s transcription, activity, and stability^6,64-66. As a result, numerous CDK9 inhibitors have entered clinical trials⁷.

The initial drug candidates, like flavopiridol, encountered significant challenges in clinical trials due to their broad targeting within the Cyclin-dependent Kinase protein family. The newer generation of CDK9 inhibitors, such as KB-0742, AZD4573, and VIP152, have demonstrated improved selectivity, and clinical trials are ongoing. However, the efficacy of CDK9 inhibition as a standalone treatment remains a subject of debate, particularly due to resistance mechanisms like the compensatory increase in MYC, or BCL-xL levels^2,67. Combination regimens to address these limitations continue to be explored.

The emerging field of targeted protein degradation offers an avenue for probing CDK9 biology further and generating therapeutic hypotheses. Its appeal lies in the ability to inhibit both the enzymatic and scaffolding functions of target proteins, potentially preventing compensation or kinome rewiring^2,68. Several groups have reported CDK9 degraders that exhibit greater degradation selectivity than their precursor inhibitors. Here, we present the development a novel CDK9 degrader. We take this further by carrying out a comprehensive characterization of the effects of CDK9 degradation on transcription, global protein stability, and post-translation phosphorylation state. We found that CKD9 degradation induced a robust silencing of MYC transcription network. We see no signs of a compensatory increase in MYC mRNA expression over the study time points as previously documented with inhibition². This suggests that CDK9 degradation could potentially offer additional therapeutic advantages over its inhibition.

Although certain cell lines with high transporter activity exhibited reduced sensitivity to KI-CDK9d-32, the overall sensitivity profile remained robust over inhibition. This is in line with the catalytic mechanism of action of degraders and underscores their potency at low intracellular doses. Further studies on KI-CDK9d-32 are necessary to confirm these findings in vivo. Additionally, the compound’s improved potency and physico-chemical properties highlight the potential for exploring novel delivery and targeting strategies. The findings presented here underscore the importance of continued exploration into the nuances of targeted degradation, not just as a means of drug development but as a fundamental lens through which we can better understand cellular regulation and disease progression.

Limitations of the study

In this study, we provide evidence of mechanistic differences between CDK9 inhibition and degradation. While we observed a previously reported compensatory increase in MYC levels following CDK9 inhibition² – an effect that is absent with degradation – our findings are limited in generalizability. We only compared KI-CDK9d-32 and KB-0742, which share a common parent molecule. Thus, the observed differences may stem partly from variations in kinetics and potency rather than intrinsic mechanistic differences alone. A broader evaluation of multiple CDK9 degrader and inhibitor pairs would be necessary to validate and extend these findings. Additionally, our study focused on early cellular responses (<12 hours), potentially missing later protein turnover and adaptive mechanisms that could emerge with prolonged exposure.

Our analysis also lacks detailed structural insights into ternary complex formation and the role of linker composition on the enhanced binding affinity of KI-CDK9d-32 following PROTAC conversion. Further computational modeling or X-ray crystallography could provide deeper mechanistic understanding.

Lastly, this study primarily relies on in vitro biochemical assays and cell-line models, and does not capture the in vivo pharmacokinetics, efficacy, or toxicity profiles of the degrader. Future studies would include in vivo validation and explore potential limitations of the dependence on CRBN levels and ABCB1 resistance, which may influence the therapeutic applicability of this PROTAC.

STAR Methods text

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines

Human acute lymphoblastic leukemia cells MOLT-4 (ATCC, CRL-1582, sex of cell: male), rhabdomyosarcoma cells RH-4 (Cellosaurus, CVL_5916, sex of cell: female), pancreatic adenocarcinoma cells PSN-1 (ATCC, CRL-3211, sex of cell: male), and adenocarcinoma HeLa cells (ATCC, CCL-2, sex of cell: female) were obtained from the Koch Institute’s Preclinical Modeling (ES Cell and Transgenics Facility) Core. Cell lines were maintained at 5% CO₂ at 37°C. Cells were cultured in RPMI 1640 (Thermo Fisher Scientific, 11875093) + 10% fatal bovine serum (FBS). Cell lines were tested intermittently throughout studies using MycoAlert Mycoplasma Detection Kit (Lonza, Cat#LT07-418), generally a few days after thawing and immediately before significant studies (e.g. Proteomics, RNA-seq experiments).

METHOD DETAILS

HiBiT-tagged CDK9 in MOLT-4 cells

To generate MOLT-4 cells with an endogenous HiBiT-tagged CDK9, we utilized the CRISPR-Cas9 system to insert the HiBiT tag at the N-terminus of CDK9 immediately following the start codon^27,28. In summary, Alt-R sgRNA and Cas9 Nuclease were combined to form a ribonucleoprotein complex (RNP), as per the guidelines provided by Integrated DNA Technologies (IDT). The transfection mixture was prepared by combining 5μL of the RNP complex, 2.4μL of 100μM Alt-R HDR donor oligonucleotides, 1.2μL of Alt-R Cas9 electroporation enhancer, 20μL of MOLT-4 cell suspension (containing 2x10^5 cells), and 1.4μL of PBS, reaching a total volume of 30μL per transfection. Electroporation was carried out using the 4D Nucleofector System (Lonza) and Nucleocuvette^™ Strips, following the CA-137 program recommended by the manufacturer for MOLT-4 cells (Lonza, V4XC-2032). Post-electroporation, cells were incubated at room temperature for 5 minutes before being transferred to a six-well plate for cultivation. The electroporated cells were allowed a recovery period of one week. For the generation of monoclonal cell populations, cells were seeded in conditioned growth media at a limiting dilution to achieve approximately 0.5 cell per well in 96-well plates. After about two weeks of expansion, clones were screened for the insertion.

CDK9 degradation assay

MOLT-4 cells expressing HiBiT-tagged CDK9 were cultured in RPMI 1640 medium (Thermo Fisher Scientific, 11875093) supplemented with 10% fetal bovine serum (FBS) and maintained in a 5% CO2, 37°C incubator. For degrader screening, dose-response, and kinetic studies, cells were seeded at a density of 10x10^3 cells per well in 96-well plates. The cells were treated with various compounds and a DMSO vehicle control in 100μL of growth medium. Following each treatment time point, replicate plates were equilibrated to room temperature, and an equal volume of 2X Nano-Glo HiBiT Lytic Reagent (Promega N3030)—comprising Nano-Glo HiBiT Lytic Buffer, Nano-Glo HiBiT Lytic Substrate, and LgBiT Protein—was added. The mixture was then agitated on an orbital shaker for 10 minutes to ensure thorough mixing, followed by a 10-minute incubation at room temperature to facilitate equilibration between LgBiT and HiBiT in the lysate. Luminescence was measured using the Tecan Infinite M200 plate reader.

Immunofluorescence assay

Cells were seeded at 3x10⁴ cells per well in 24-well plates with glass coverslips (Fisher Scientific, 1254580) and incubated overnight for attachment. If applicable, drug treatments were administered for the specified durations. Cells were fixed with 4% paraformaldehyde directly in the media for 15 minutes at room temperature, followed by washes with phosphate-buffered saline (PBS) and stored at 4°C. Overnight blocking was performed at 4°C using PBSA buffer (PBS with 1% BSA, 0.1% Triton X-100, 0.05% sodium azide). Primary antibodies, DDX21 (Novus Biologicals, NB100-1718) and NPM1 (Abcam ab180607), were diluted to 1:200 in PBSA and applied to cells for 2 hours at room temperature. Secondary antibodies, Alexa Fluor 488 (Thermo Fisher Scientific, A-11008) and Alexa Fluor 647 (Thermo Fisher Scientific, A-27040), were diluted to 1:1,000 in PBSA and incubated for 1 hour at room temperature. Afterward, cells underwent three 5-minute washes with PBSA and two 5-minute washes with PBS, a brief water rinse, and were mounted on glass slides using ProLong Diamond antifade mountant (Invitrogen, P36961). Nuclei were stained with Hoechst 33342 (Thermo Fisher Scientific, 62249) at a 1:1000 dilution. Images were captured on a DeltaVision 2 TIRF microscope (60X objective) and processed accordingly.

Western blot assay:

Cells were seeded at a density of 50-60% in RPMI-1640 supplemented with 10% FBS, approximately 12 hours prior to treatment. The cells were treated with either compound or DMSO. After 4 hours of treatment, cells were harvested and lysed using RIPA buffer (Thermo Fisher Scientific, cat. 89900), which was supplemented with a cocktail of protease inhibitors (Thermo Fisher Scientific, cat. 87786), and phosphatase inhibitors (PhosSTOP, Millipore SIGMA, cat. 4906845001). Lysates were vortexed and spun at 20,000 g for 10 min at 4° C. The resulting supernatants were quantified using the bicinchoninic acid (BCA) assay (Thermo Fisher Scientific, cat. 23225) to ensure uniformity and added to loading buffer (2x Laemmli Sample Buffer, Bio-Rad, cat. 1610737; and 2-Mercaptoethanol, Bio-Rad cat. 161070). The samples were boiled at 90° C for 7 min. Samples were electrophoresed at 200 V using a 4-20% Criterion Protein Gel (Bio-Rad, cat. 5671094) and transferred to nitrocellulose membrane (Bio-Rad, cat. 1620112) using the trans-blot turbo transfer system (Bio-Rad, cat. 1704150EDU) according to manufacturer’s instructions. Relative protein levels were determined via immunoblotting (CDK9, c-MYC, CFL1, p-Serine 2 CTD POLR2A, CDK7, Histone 3).

Quantitative Mass-spectrometry

Protein extraction

MOLT-4 cells were seeded in 6-well plates at a density of 2.4 × 10^6 cells per well, in 4 mL of RPMI-1640 supplemented with 10% FBS, approximately 12 hours prior to treatment. The cells were treated with either KI-CDK9d-32 or DMSO. A total of four biological replicates were prepared. At 1 and 4 hours post-treatment, the cells were lysed using RIPA buffer (Thermo Fisher Scientific, cat. 89900), which was supplemented with a cocktail of protease inhibitors (Thermo Fisher Scientific, cat. 87786), and phosphatase inhibitors (PhosSTOP, Millipore SIGMA, cat. 4906845001). The proteins in the lysates were precipitated using cold acetone (−20°C), following a standard protocol (Thermo Fisher Scientific TECH TIP #49), employing a single cycle of precipitation. The samples were then stored at −80°C until further analysis by mass spectrometry, as described below.

Reduction, Alkylation, and Tryptic Digestion

Proteins were first reduced with 10 mM dithiothreitol (Sigma) for 1 hour at 56°C and then alkylated with 55 mM iodoacetamide (Sigma) for 1 hour at 25°C in the dark. The samples were diluted with 100 mM ammonium bicarbonate to decrease urea concentration to 1M. Proteins were digested overnight at 25°C with modified trypsin (Promega) using an enzyme-to-substrate ratio of 1:50 in 100 mM ammonium bicarbonate, pH 8.9. The reaction was stopped by adding formic acid (99.9%, Sigma) to achieve a final concentration of 5%. Samples were then desalted using Pierce Peptide Desalting Spin Columns (cat. # 89852).

TMT Labeling

Desalted samples underwent TMTpro^™ 16plex (Thermo Fisher Scientific) labeling according to Thermo Fisher Scientific’s protocol. Samples were resuspended in 100 mM TEAB, vortexed, and briefly centrifuged. Anhydrous acetonitrile (20 μL) was added to the TMT label reagents, followed by vortexing, brief centrifugation, and a 5-minute dissolution period. Then, 20 μL of TMT reagents were added to each of the 100 μL samples, which were then vortexed and briefly centrifuged. The samples were incubated for 1 hour at room temperature. To quench the reaction, 5 μL of 5% hydroxylamine was added to each sample and incubated for 15 minutes. Equal volumes of each channel were combined and concentrated to dryness via speed-vac.

Phosphopeptide enrichment

A 10% aliquot of the combined TMT-labeled sample was reserved for HPLC fractionation to obtain the unmodified protein data. The remainder underwent phosphopeptide enrichment using the High-Select Fe-NTA Phosphopeptide Enrichment Kit (Thermo Fisher Scientific) as per the manufacturer's instructions.

HPLC fractionation of non-phosphopeptides

To acquire the unmodified protein data, 10% of the sample was fractionated using HPLC. The dried TMT-labeled peptides were resuspended in 10 mM TEAB and fractionated using an Agilent 1100 HPLC system with high pH buffers. Buffer A consisted of 10 mM triethylammonium bicarbonate (TEAB), and Buffer B was 10 mM TEAB in 99% acetonitrile. Fractions were collected every minute from the 10th to the 90th minute, resulting in 80 fractions. Every 15^th fraction was pooled resulting in 15 total fractions for further analysis. Specifically, the 15 fractions were dried, resuspended in 10 μL of 0.2% formic acid, and 10% of this mixture was injected into the LC-MS.

The peptides were resolved over the following gradient run:

Time	% Buffer B	Flow rate uL
0	1	200
2	1	200
10	5	200
72	35	200
87	70	200
92	70	200
93	1	200
100	1	200

High pH fractionation of Phosphopeptides

Phosphopeptides were fractionated by Pierce High pH Reversed-Phase Peptide Fractionation Kit (Cat# 84868) per manufacturer’s instructions. Dried fractions were re-suspended in 3 μL of 0.2% formic acid, and 33% of this solution was injected into the LC-MS.

LC-MS/MS measurement

The fractionated TMT-labeled tryptic peptides were separated by reverse-phase HPLC (Thermo Ultimate 3000) using a Thermo Fisher Scientific PepMap RSLC C18 column (2 μm tip, 75 μm x 50 cm PN# ES903) over a gradient before nano-electrospray ionization in an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific). The solvents used were 0.1% formic acid in water (Solvent A) and 0.1% formic acid in acetonitrile (Solvent B). The gradient conditions were 1% B (0 – 10 min at 300 nL / min); 1% B (10 – 15 min, 300 nL / min to 200 nL / min); 1 – 5% B (15 – 20 min, 200 nL / min); 5 – 25% B (20 – 120 min, 200 nL / min); 25 – 35% B (120 – 140 min, 200nL / min); 35 – 80% B (140 – 155 min, 200 nL / min); 80% B (155 – 160 min, 200 nL / min); 80 – 1% B (160 – 160.1 min, 200 nL / min); 1% B (160.1 – 180 min, 200 nL / min). The mass spectrometer was operated in a data-dependent mode. The parameters for the full scan MS consisted of resolution of 60,000 across 450 – 1600 m/z and maximum IT 50 ms. The full MS scan was followed by MS/MS for as many precursor ions in a 3 s cycle with a NCE of 32, dynamic exclusion of 30 s and resolution of 45,000.

Protein ID & data prep

The raw mass spectrometry data (.raw files) were analyzed using Sequest HT within Proteome Discoverer (Thermo Fisher Scientific). The search parameters included a 10 ppm mass tolerance for precursor ions and a 0.05 Da tolerance for fragment ion mass. Up to two missed cleavages by trypsin were allowed. Fixed modifications included carbamidomethylation of cysteine and TMT labeling of lysines and peptide N-termini. Variable modifications included oxidation of methionine, loss of methionine at the protein N-terminus, acetylation of the protein N-terminus, and combined Met-loss plus acetylation at the N-terminus. Searches were conducted against both a human protein database (homo sapiens UniProt FASTA files) and an in-house contaminant database.

Peptide spectrum matches (PSMs) were exported from Proteome Discoverer (PD) and subjected to addition filtering and processing using a proteomic analysis workflow (PAW) described previously^71,72. In summary, the PD exported PSMs were filtered based on the following default pipeline criteria: a q-value cutoff of 0.05, a mass deviation within ±20 ppm, and precursor isolation interference below 50%. PSMs were further filtered by charge state (2 to 4), peptide length (6 to 28 amino acids), and rank (considering only top-ranked hits). This filtered PSMs file was then used for high confident peptide and protein inference. At least 2 distinct peptides had to be present support a protein identification within a sample, and the minimum number of tryptic termini was 2, ensuring that peptides are fully tryptic. In a final processing step, proteins with insufficient evidence to be considered unique were grouped. Intensities are then added to the peptide and protein-level summaries and exported to R for quantitation and analysis.

Protein ID & data prep for phosphopeptides

Phosphopeptide raw data were processed with Proteome Discoverer version 3.0 (Thermo Fisher Scientific), peptide and protein identification were performed with Mascot version 2.4 (Matrix Science), and phosphorylation site localization probabilities were assessed using ptmRS⁷³ within Proteome Discoverer. Spectra were searched against the UniProt homo sapiens database. Search parameters were set to trypsin digestion with a maximum of two missed cleavages. Mass tolerance was set to 10 ppm at the MS1 level and 20 mmu at the MS2 level. Precursor ions and TMT reporter ions were removed from MS/MS spectra prior to searching using the non-fragment filter node in Proteome Discoverer. Cysteine carbamidomethylation, TMT-labeled lysine, and TMT-labeled peptide N-termini were set as static modifications; phosphorylation of tyrosine, serine, and threonine, and oxidation of methionine were set as dynamic modifications.

PSMs were exported from Proteome Discoverer and filtered to keep those with high quality: search engine rank of 1, delta mass within −10 to 10 ppm, expectation value <0.05, and ions score >15. PSMs missing values in more than 25% of TMT channels were discarded, remaining missing values were imputed with the minimum intensity observed in the dataset. Only peptides with an average abundance of 1000 or more across TMT channels, and with a ptmRS probability greater than 50 for at least one site were maintained. TMT reporter ion intensities were summed across PSMs sharing a common phosphopeptide sequence. To address the issue of redundancy, phosphopeptides derived from identical proteins with matching phosphorylation patterns were aggregated to streamline the dataset, ensuring each unique phosphosite was represented without unnecessary repetition. This aggregation was carried out by matching peptides based on their annotated sequences, modifications in master proteins, and positions in master proteins. Furthermore, motif analysis, leveraging the PhosphoSitePlus database⁷⁴, was carried out to map phosphorylation sites to specific known motifs.

Differential analysis for non-phosphopeptides

The pre-processed global quantitative mass spectrometry data were loaded into R. For each unique protein, abundance values were summed. Normalization involved alignment of the datasets by applying a median-centering normalization across all samples. This process adjusted each sample’s abundance values by the ratio of the overall median to the sample’s median, thereby aligning median values across the dataset. Differential analysis was conducted using the Limma package in R, which facilitated the identification of differentially enriched or depleted proteins relative to the DMSO-treated control samples within each experimental condition. Gene set over-representation analysis was carried out on 10% of the most significant enriched or depleted proteins (adjusted p values < .05, and the absolute value of log 2 fold change > .5) to identify the cellular processes and functions most affected by the degrader. The clusterProfiler⁷⁵ package from R was used. We used all proteins identified in the experiment as the universe for the over-representation analysis.

Differential analysis for phosphopeptides

The pre-processed phosphopeptide data were loaded into R. For each unique peptide, abundance values were summed. Normalization involved two main steps: First, for drug-treated samples, abundance values were normalized against control samples within each experimental condition. Next, to align the datasets, a median-centering normalization was applied across all samples. This process adjusted each sample's abundance values by the ratio of the overall median to the sample's median, thereby aligning median values across the dataset. These normalized phosphopeptide data were then integrated with protein data, which underwent a similar normalization process, ensuring only peptides present in both datasets were retained. Differential analysis was conducted using the Limma package in R, which facilitated the identification of differentially phosphorylated peptides in relation to their overall protein levels. Gene over-representation analysis was carried out on the most significant enriched or depleted phosphopeptides (adjusted p values < .05, and the absolute value of log 2 fold change > .5) to identify the cellular compartments most impacted. The clusterprofiler package from R was used. The background universe used was all proteins identified in the global quantitative mass spectrometry.

RT-qPCR

RNA extraction

MOLT-4 cells were cultured in RPMI-1640 medium supplemented with 10% FBS. Cells were seeded at a density of 2x10^6 cells/well in 6-well plates. Twelve hours after seeding, the cells in the assay wells were treated with KI-CDK9d-32 at concentrations of 15nM, 50nM, and 125nM; KB-0742 at concentrations of 125nM, 1.2μM, and 5μM; or DMSO (as a control) for durations of 2, 4, and 8 hours. At the end of the specified assay periods, RNA was extracted using the PureLink^™ RNA Mini Kit (Invitrogen) and homogenized using the QIAshredder (Qiagen). Equal amounts of RNA were then used for cDNA synthesis with the iScript^™ Advanced cDNA Synthesis Kit (Bio-Rad Laboratories, catalog number 1725038). RT-qPCR analysis was performed on the CFX384 Touch^™ Real-Time PCR Detection System (Bio-Rad), following the manufacturer's specified protocols and using SYBR Green (Roche/Kapa). The following primer pairs were used: MYC (5'-tcctcggattctctgctctc-3'/ 5'-tcttcctcatcttcttgttcctc-3' ), CDK9 (5'-ctgaagaaggtgctgatggaa-3' / 5'--agttgaccacattctcgtgtt-3'), POLR2A (5'- ggatgatctggaatgctcag-3' /5'- attccttgactccctccac-3').

Data Analysis

The raw quantification cycle (Cq) values from RT-qPCR were exported from the CFX384 Real-Time PCR Detection System and imported into R software for statistical analysis. The median Cq values were computed from a minimum of three technical replicates for each unique combination of treatment, time point, specific gene, and concentration. To normalize gene expression, the delta Cq (ΔCq) for each condition was calculated by subtracting the Cq value of the POLR2A reference gene from the Cq value of the target gene. The delta delta Cq (ΔΔCq) was then determined by subtracting the ΔCq of the DMSO control (the normalization control) from the ΔCq of each experimental condition. This step quantifies relative gene expression levels. The fold change in expression for each condition was defined as 2^(-ΔΔCq). The average fold change in expression from three biological replicates was used to compare the effects of the compounds on gene expression.

RNA Sequencing

RNA extraction, synthetic RNA spike-in

Four biological replicates were used for these experiments. For each biological replicate, 10³ cells were seeded and treated with treated with KB-0742, KI-CDK9d-32, or DMSO. At the appropriate endpoint (2, 4, or 8 hours after treatment), cells were immediately lysed in lysis buffer (from PerkinElmer’s chemagic RNA Tissue 360 H96, VD200615) supplemented with BME flash frozen and stored at −80c. After all biological replicates were collected, samples were thawed in deep 96-well plates. An equal amount (~30pg) of Lexogen’s ERCC spike-in mix (SIRV-3) were added at this point following manufacturer’s recommendations. The PerkinElmer’s Chemagic 360 was used for bead-based RNA extraction following manufacturer’s standard protocol.

RNA was quantified using an Agilent Fragment Analyzer, purified using 2X RNA SPRI beads, and 1ng of total RNA was used for Illumina library preparation. Libraries were prepared using a reduced volume version of SMART-seq2⁷⁶ performed on an SPT Mosquito HV using 10 cycles of amplification for cDNA generation and 16 cycles for library generation using NexteraXT (Illumina). Libraries were sequenced on an Illumina NextSeq500 using paired end 38nt reads.

Data processing and differential analysis

RNA-seq data was used to quantify transcripts from the hg38 human assembly with the gencode version 43 annotation using the nf-core/rnaseq workflow revision 3.12.0⁷⁷. The genomic target included the LexogenSIRVData spike-in reference. Gene level summaries were prepared from the star_salmon quantitation using tximport version 1.28.0⁷⁸ running under R version 4.3.0 (R Core Team 2021) with tidyverse version 2.0.0⁷⁹. Differential expression analysis was done with DESeq2 version 1.40.1^80,81 using apeglm log fold change shrinkage⁸². The R image used for this work is available (docker://bumproo/rnaseqclass23).

For each of the three cell lines, MOLT-4, PSN-1, and RH-4, protein coding genes and genes with sufficient variance and expressions across all conditions (variance > 0.1, and mean > 0.1) were used for downstream analyses. Enrichment analysis was conducted on the top 20% of genes that were differentially expressed when comparing the degrader directly with the inhibitor. The clusterprofiler R package⁷⁵ was used for the over-representation analysis against the Molecular Signature Database Hallmark gene set. All protein coding genes in each dataset was used as background for that specific cell line to preserve any cell-specific expression. Preranked Gene Set Enrichment Analysis⁸³ was done using javaGSEA version 4.3.2 with msigDb v2023.2 human gene sets⁸⁴.The ranking metric used for GSEA was the Wald statistic produced by DESeq2 from differential expression tests run using data normalized to SIRV spike-ins.

Cell viability assessment

After 12 hours post-seeding, cells in assay wells were treated with vehicle (DMSO) or compound stocks dissolved in DMSO. The cells were seeded at the following densities to accommodate for growth over the experimental period: MOLT-4 at 6.5x10³ and 3.0x10³, PSN-1 at 2.0x10³ and 1.0x10³, and RH-4 at 3x10³ and 1.5x10³ cells per well for 72-hour and 120-hour endpoint readings, respectively. Treatments included degraders (e.g., KI-CDK9d-32, Thal-SNS-32) at concentrations up to 500nM, inhibitors (e.g., KI-CDK9d-32N, KI-ARv-03, and KB-0742) up to 10μM, and the DMSO control. The percentage of DMSO by volume was kept below 0.4%. Cell viability was determined using the CellTiter-Glo assay (Promega), with luminescence measured by a Tecan Infinite M200 plate reader at the specified time points. Viability was determined as the normalized luminescence to the DMSO control for each cell line and replicate (n=3 replicates). The dose-response curves were fitted to a four-parameter log-logistic model to estimate IC50 values for both the 72-hour and 120-hour endpoints using the drc package in R.

High throughput cell sensitivity profiling

PRISM Pooled Cell Sensitivity Screen:

The PRISM cell sensitivity screen was carried out as previously described⁵⁹. In summary, approximately 800 barcoded cell lines in pools of 20-25 were thawed and plated into 384-well plates (1250 cells/well for adherent cells, 2000 cells/well for suspension or mixed suspension/adherent pools). Cells were treated in triplicate with threefold dilutions starting at 30 μM and 1.5 μM of KB-0742 and KI-CDK9d-32, respectively. Cells were incubated for 120 hours, then lysed. Each cell’s barcode was read out by mRNA based Luminex detection as described previously and input to a standardized R pipeline (https://github.com/broadinstitute/prism_data_processing) to generate viability estimates relative to vehicle treatment and fit dose-response curves. The IC₅₀ and area under the dose-response-curve (AUC) values were used as metrics of therapeutic potency in cell lines and correlated with gene expression (transcripts/million) and proteomics data as annotated in the Cancer Cell Line Encyclopedia (CCLE).

MGH Cancer Center Sensitivity Screen:

Cell lines are maintained as part of the MGH Center for Molecular Therapeutics for drug sensitivity profiling⁶² (Garnett et al, Nature 2012). All lines were tested for mycoplasma and grown in either RPMI or DMEM/F12 media with 10% fetal bovine serum. Cells were seeded into 384-well plates one day prior to drug addition. Cells were treated with vehicle (DMSO) or compound stocks dissolved in DMSO with a PerkinElmer JANUS workstation. Following drug addition, cells were incubated for 5 days. Cell viability was then determined using the CellTiter-Glo assay (Promega), with luminescence measured by a PerkinElmer Envision plate reader. Viability was determined as the normalized luminescence to DMSO control for each cell line (n=2 replicates). Dose-response curves were fitted to a four-parameter log-logistic model using SciPy scipy.optimize.curve_fit to estimate IC₅₀ and AUC values.

Sensitivity Analysis and Clustering:

To generate cell sensitivity clusters for KI-CDK9d-32 and KB-0742, AUC and IC50 values for cells in both the pooled and Individual cell line screens were independently standardized using z-scores (i.e., by scaling the mean to 0 and the standard deviation to 1). This led to relative AUC and IC50 values, controlling for systematic differences attributable to varying dose ranges. The scaled values were subjected to UMAP-based dimensionality reduction followed by hdbscan clustering. The UMAP parameters used were n_neighbors = 5, min_dist = .01, metric = Euclidean. For hdbscan, the default minPts=10 was used.

Chemical synthesis and characterization

KI-CDK9d-08

4'-(((2-(2,6-dioxopiperidin-3-yl)-1,3-dioxoisoindolin-4-yl)amino)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)-[1,1'-biphenyl]-4-carboxamide

tert-butyl 4'-(hydroxymethyl)-[1,1'-biphenyl]-4-carboxylate (3):

A mixture of 4-(tert-Butoxycarbonyl)phenylboronic acid, pinacol ester (1, 1.00 g, 3.29 mmol), (4-Bromophenyl)methanol (2, 615 mg, 3.29 mmol), Pd(PPh₃)₂Cl₂ (115 mg, 0.164 mmol), and Cs₂CO₃ (2.14 g, 6.57 mmol) in 1,4-dioxane (8.00 mL) and H₂O (2.00 mL) was stirred at 100 °C. After 3.5 hours, the mixture was filtered through a pad of Celite and rinsed with ethyl acetate. The filtrate was concentrated to give a crude material, which was purified by silica gel column chromatography (hexane:ethyl acetate = 90:10 to 60:40) to yield the title compound as a white solid (997 mg, quant). ¹H NMR (500 MHz, CDCl₃) δ 8.05 (d, J = 8.5 Hz, 2H), 7.62 (dd, J = 8.3, 6.6 Hz, 4H), 7.46 (d, J = 8.0 Hz, 2H), 4.76 (s, 2H), 1.62 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 165.82, 144.87, 140.89, 139.68, 130.96, 130.09, 127.65, 127.58, 126.95, 81.21, 65.13, 28.37. QToF HRMS m/z: calcd for C18H20NaO3⁺ [M+Na⁺] = 307.1305; Found 307.1311.

2-(2,6-dioxopiperidin-3-yl)-4-nitroisoindoline-1,3-dione (6):

A mixture of 4-nitroisobenzofuran-1,3-dione (4, 2.00 g, 10.4 mmol) and 3-aminopiperidine-2,6-dione hydrochloride (5, 1.88 g, 11.4 mmol) and KOAc (3.15 g, 32.1 mmol) in AcOH (20.8 mL) was stirred at 90 °C overnight. The mixture was concentrated, and the resulting solid material was washed with methanol. The title compound was obtained as a gray solid (3.32 g, quant). ¹H NMR (500 MHz, DMSO) δ 11.17 (s, 1H), 8.35 (d, J = 8.0 Hz, 1H), 8.24 (d, J = 7.6 Hz, 1H), 8.12 (t, J = 7.8 Hz, 1H), 5.20 (dd, J = 12.9, 5.3 Hz, 1H), 2.89 (ddd, J = 17.2, 13.9, 5.4 Hz, 1H), 2.66 – 2.57 (m, 1H), 2.56 – 2.45 (m, 1H), 2.12 – 2.03 (m, 1H). ¹³C NMR (126 MHz, DMSO) δ 172.74, 169.52, 165.19, 162.54, 144.44, 136.84, 133.02, 128.89, 127.32, 122.57, 49.44, 30.88, 21.74. QToF HRMS m/z: calcd for C13H9KN3O6⁺ [M+K⁺] = 342.0123; Found 342.0127.

2-(2,6-dioxo-1-((2-(trimethylsilyl)ethoxy)methyl)piperidin-3-yl)-4-nitroisoindoline-1,3-dione (7):

DBU (2.22 mL, 14.9 mmol) and SEMCl (1.98 mL, 11.2 mmol) were added to a stirred solution of 6 (2.26 g, 7.45 mmol) in DMF (24.8 mL) at room temperature. After 2 hours, the mixture was quenched by adding saturated aq. NH4Cl, and extracted with ethyl acetate. The combined organic phase was dried over anhydrous Na2SO4. Filtration and concentration gave the crude material, which was purified by silica gel column chromatography (hexane:ethyl acetate = 90:10 to 50:50) to yield the title compound as a white solid (1.31 g, 41%). ¹H NMR (500 MHz, DMSO) δ 8.36 (d, J = 8.1 Hz, 1H), 8.24 (d, J = 7.4 Hz, 1H), 8.13 (t, J = 7.8 Hz, 1H), 5.34 (dd, J = 13.1, 5.4 Hz, 1H), 5.08 (s, 2H), 3.52 (dtd, J = 30.6, 9.7, 6.4 Hz, 2H), 3.03 (ddd, J = 17.3, 14.0, 5.4 Hz, 1H), 2.85 – 2.77 (m, 1H), 2.60 – 2.51 (m, 1H), 2.15 – 2.07 (m, 1H), 0.91 – 0.77 (m, 2H), −0.02 (s, 9H). ¹³C NMR (126 MHz, DMSO) δ 171.56, 169.47, 165.11, 162.45, 144.46, 136.86, 132.98, 128.91, 127.30, 122.53, 68.35, 65.99, 49.98, 31.09, 20.73, 17.46, −1.38. QToF HRMS m/z: calcd for C19H27N4O7Si⁺ [M+NH₄⁺] = 451.1644; Found 451.1648.

N-(2-(2,6-dioxo-1-((2-(trimethylsilyl)ethoxy)methyl)piperidin-3-yl)-1,3-dioxoisoindolin-4-yl)-2-nitrobenzenesulfonamide (9):

A mixture of 7 (660 mg, 1.52 mmol) and Pd/C (10%, 81.0 mg, 0.076 mmol) in ethanol (7.60 mL) was stirred under hydrogen atmosphere at room temperature overnight. The mixture was filtered through a pad of Celite, then concentrated. The resulting crude material was used in the next step without further purification. To a stirred solution of the crude material in pyridine (5.10 mL), 2-nitrobenzenesulfonyl chloride (1.01 g, 4.58 mmol) was added, then the mixture was stirred at 40 °C overnight. After cooling to room temperature, the mixture was quenched by adding 10 drops of H₂O. Concentration and purification by silica gel column chromatography (hexane:ethyl acetate = 80:20 to 40:60 including 1% Et₃N, then 100% ethyl acetate) gave the title compound as a yellow solid (521 mg, 58%, 2 steps). ¹H NMR (500 MHz, CDCl₃) δ 9.66 (d, J = 4.1 Hz, 1H), 8.19 (dt, J = 7.3, 1.8 Hz, 1H), 8.09 (dd, J = 8.6, 3.5 Hz, 1H), 7.92 (dt, J = 7.7, 1.9 Hz, 1H), 7.80 – 7.72 (m, 2H), 7.70 (t, J = 7.9 Hz, 1H), 7.53 (dd, J = 7.3, 2.2 Hz, 1H), 5.24 (s, 2H), 4.99 – 4.91 (m, 1H), 3.67 – 3.53 (m, 2H), 3.05 – 2.92 (m, 1H), 2.85 – 2.72 (m, 2H), 2.16 – 2.05 (m, 1H), 0.98 – 0.88 (m, 2H), −0.01 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 168.64, 168.08, 166.43, 148.18, 136.54, 135.93, 134.86, 133.08, 132.78, 132.23, 131.14, 126.15, 122.68, 119.13, 116.86, 69.38, 67.59, 50.22, 32.08, 21.84, 18.17, −1.32. AccuTOF DART HRMS m/z: calcd for C25H32N5O9SiS [M+NH₄⁺] = 606.1685; Found 606.1721.

tert-butyl 4'-(((N-(2-(2,6-dioxo-1-((2-(trimethylsilyl)ethoxy)methyl)piperidin-3-yl)-1,3-dioxoisoindolin-4-yl)-2-nitrophenyl)sulfonamido)methyl)-[1,1'-biphenyl]-4-carboxylate (10):

Triphenylphosphine (80.2 mg, 0.306 mmol) and diisopropyl azodicarboxylate (60.2 mL, 0.306 mmol) were added to a stirred solution of 9 (150 mg, 0.255 mmol) and 3 (87.0 mg, 0.306 mmol) in THF (2.55 mL) at room temperature. After 22 hours, the mixture was quenched with H₂O, and extracted with ethyl acetate. The combined organic layer was dried over anhydrous Na2SO4. Filtration and concentration gave a crude material, which was purified by silica gel column chromatography (hexane:ethyl acetate = 90:10 to 50:50) followed by prep TLC (hexane:ethyl acetate = 50:50) to yield the title compound as a white solid (116 mg, 53%). ¹H NMR (500 MHz, CDCl₃) δ 8.02 (d, J = 8.5 Hz, 2H), 7.85 – 7.80 (m, 1H), 7.72 – 7.61 (m, 5H), 7.56 (d, J = 8.4 Hz, 2H), 7.53 – 7.46 (m, 3H), 7.31 (d, J = 8.2 Hz, 2H), 5.55 – 5.31 (m, 1H), 5.22 (s, 2H), 4.85 – 4.78 (m, 2H), 3.62 (pd, J = 9.4, 7.3 Hz, 2H), 3.00 – 2.92 (m, 1H), 2.79 – 2.46 (m, 2H), 2.03 – 1.94 (m, 1H), 1.60 (s, 9H), 0.95 (t, J = 8.2 Hz, 2H), 0.00 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 170.79, 168.21, 166.38, 165.71, 165.26, 148.10, 144.35, 140.07, 135.45, 135.07, 134.78, 134.02, 133.28, 132.31, 131.74, 131.13, 131.02, 130.07, 129.66, 128.31, 127.61, 126.91, 126.88, 124.36, 124.18, 81.23, 69.39, 67.64, 55.27, 50.15, 32.08, 28.35, 21.72, 18.27, −1.26. QToF HRMS m/z: calcd for C43H46N4NaO11SSi⁺ [M+Na⁺] = 877.2545; Found 877.2547.

tert-butyl 4'-(((2-(2,6-dioxo-1-((2-(trimethylsilyl)ethoxy)methyl)piperidin-3-yl)-1,3-dioxoisoindolin-4-yl)amino)methyl)-[1,1'-biphenyl]-4-carboxylate (11):

Cs₂CO₃ (133 mg, 0.407 mmol) and 4-bromothiophenol (51.3 mg, 0.271 mmol) were added to a stirred solution of 10 (116 mg, 0.136 mmol) in DMF (1.36 mL) at room temperature. After 1 hour, the mixture was poured into H₂O, and extracted with CH₂Cl₂. The combined organic layer was dried over anhydrous Na2SO4 and concentrated. Purification by silica gel column chromatography (hexane:ethyl acetate = 90:10 to 70:30) gave the title compound as a yellow amorphous (81.2 mg, 89%).¹H NMR (500 MHz, CDCl₃) δ 8.05 (d, J = 8.4 Hz, 2H), 7.61 (dd, J = 8.2, 6.0 Hz, 4H), 7.49 – 7.41 (m, 3H), 7.13 (d, J = 7.1 Hz, 1H), 6.85 (d, J = 8.5 Hz, 1H), 6.74 (t, J = 5.9 Hz, 1H), 5.28 (s, 2H), 4.99 – 4.92 (m, 1H), 4.56 (d, J = 5.7 Hz, 2H), 3.63 (dtd, J = 31.5, 9.8, 6.7 Hz, 2H), 3.03 – 2.94 (m, 1H), 2.86 – 2.73 (m, 2H), 2.17 – 2.07 (m, 1H), 1.61 (s, 9H), 0.95 (ddd, J = 9.8, 6.7, 2.7 Hz, 2H), −0.01 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 171.11, 169.66, 169.23, 167.70, 165.74, 146.70, 144.61, 139.71, 137.74, 136.28, 132.66, 131.06, 130.12, 127.88, 127.64, 126.90, 117.18, 112.21, 110.86, 81.20, 69.30, 67.53, 49.78, 46.61, 32.20, 28.36, 22.15, 18.22, −1.31. QToF HRMS m/z: calcd for C37H43N3NaO7Si⁺ [M+Na⁺] = 692.2762; Found 692.2765.

4'-(((2-(2,6-dioxopiperidin-3-yl)-1,3-dioxoisoindolin-4-yl)amino)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)-[1,1'-biphenyl]-4-carboxamide (KI-CDK9d-08):

A mixture of 11 (81.2 mg, 0.121 mmol) and 4 N HCl in 1,4-dioxane (1.21 mL) was stirred at 40 °C overnight. Concentration gave a crude mixture including 4'-(((2-(1-(hydroxymethyl)-2,6-dioxopiperidin-3-yl)-1,3-dioxoisoindolin-4-yl)amino)methyl)-[1,1'-biphenyl]-4-carboxylic acid (12), which was used in the next step without further purification. To a stirred solution of the crude material and KI-ARv3 hydrochloride (61.6 mg, 0.208 mmol) in DMF (1.04 mL), iPr₂NEt (54.4 mL, 0.313 mmol) and HATU (79.2 mg, 0.208 mmol) were added at room temperature. After stirring overnight, the mixture was poured into H₂O, and extracted with CH₂Cl₂. The combined organic layer was dried over anhydrous Na2SO4. Filtration and concentration gave a crude mixture, which was purified by silica gel column chromatography (CH₂Cl₂:MeOH = 99:1 to 95:5) to yield a mixture of the title compound and 4'-(((2-(1-(hydroxymethyl)-2,6-dioxopiperidin-3-yl)-1,3-dioxoisoindolin-4-yl)amino)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)-[1,1'-biphenyl]-4-carboxamide. To a stirred solution of the mixture in DMF (1 mL) was added N1,N2-Dimethylethane-1,2-diamine (9.20 mL, 0.0853 mmol) at 0 ℃. After 1.5 hours, the mixture was poured into H₂O, and extracted with 10% MeOH in CH₂Cl₂. The combined organic layer was dried over anhydrous Na₂SO₄. Filtration and concentration gave a crude material, which was purified on HPLC (MeCN:H₂O = 10:90 to 90:10, including 0.1% HCO₂H) to yield the title compound as a yellow solid (32.2 mg, 52%, 3 steps). ¹H NMR (500 MHz, DMSO) δ 11.12 (s, 1H), 8.44 (d, J = 7.2 Hz, 1H), 8.02 (d, J = 2.2 Hz, 1H), 7.94 (d, J = 8.5 Hz, 2H), 7.75 (d, J = 8.1 Hz, 2H), 7.71 (d, J = 7.5 Hz, 2H), 7.55 – 7.46 (m, 3H), 7.30 (t, J = 6.3 Hz, 1H), 7.03 (d, J = 7.1 Hz, 1H), 6.99 (d, J = 8.6 Hz, 1H), 6.30 (d, J = 2.1 Hz, 1H), 6.06 (s, 1H), 5.09 (dd, J = 12.9, 5.2 Hz, 1H), 4.62 (d, J = 6.3 Hz, 2H), 4.51 (q, J = 7.2 Hz, 1H), 4.28 (q, J = 7.1 Hz, 1H), 2.90 (ddd, J = 17.9, 13.8, 5.4 Hz, 1H), 2.65 – 2.52 (m, 5H), 2.30 – 2.21 (m, 1H), 2.20 – 2.01 (m, 3H), 1.85 – 1.59 (m, 4H), 0.92 (t, J = 7.4 Hz, 3H). ¹³C NMR (126 MHz, DMSO) δ 172.89, 170.16, 168.81, 167.33, 165.66, 162.20, 148.65, 146.13, 146.06, 143.17, 142.30, 138.98, 137.96, 136.17, 133.47, 132.27, 128.02, 127.68, 127.06, 126.30, 117.71, 110.84, 109.67, 93.64, 85.03, 54.94, 51.54, 49.28, 48.62, 45.13, 38.42, 31.08, 31.03, 30.63, 22.20, 21.96, 13.79. QToF HRMS m/z: calcd for C41H41N8O5⁺ [M+H⁺] = 725.3194; Found 725.3206.

KI-CDK9d-32

4-((4-(2-(2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)benzamide

tert-butyl 4-((4-(2-(2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)benzoate (15):

Tert-butyl 4-(bromomethyl)benzoate (14, 7.7 mg, 0.0285 mmol) and K₂CO₃ (11.2 mg, 0.0813 mmol) were added to a stirred mixture of 3-(1-oxo-5-(piperazin-1-yl)isoindolin-2-yl)piperidine-2,6-dione⁸⁵ (13, 9.9 mg, 0.0271 mmol) in DMF (0.500 mL), and the mixture was warmed to 40 °C. After stirring overnight, the mixture was poured into H₂O, and extracted with ethyl acetate. The combined organic layer was dried over anhydrous Na₂SO₄. Filtration and concentration gave a crude mixture, which was purified by prepTLC (5% methanol in CH₂Cl₂) to yield the title compound as a white solid (10.1 mg, 72%). ¹H NMR (500 MHz, CDCl₃) δ 8.11 – 8.02 (m, 1H), 8.00 – 7.93 (m, 2H), 7.72 (d, J = 8.6 Hz, 1H), 7.41 (d, J = 8.2 Hz, 2H), 6.98 (dd, J = 8.6, 2.2 Hz, 1H), 6.86 (d, J = 2.2 Hz, 1H), 5.19 (dd, J = 13.2, 5.0 Hz, 1H), 4.40 (d, J = 15.7 Hz, 1H), 4.25 (d, J = 15.6 Hz, 1H), 3.61 (s, 2H), 3.32 (t, J = 5.0 Hz, 4H), 2.97 – 2.76 (m, 2H), 2.60 (t, J = 5.1 Hz, 4H), 2.38 – 2.25 (m, 1H), 2.19 (dtd, J = 13.0, 5.3, 2.5 Hz, 1H), 1.60 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 171.28, 169.88, 169.66, 165.82, 154.56, 143.77, 142.81, 131.24, 129.64, 128.96, 125.15, 121.89, 115.62, 108.39, 81.12, 62.71, 52.93, 51.84, 48.45, 47.03, 31.75, 28.35, 23.63. QToF HRMS m/z: calcd for C29H35N4O5⁺ [M+H⁺] = 519.2602; Found 519.2609.

4-((4-(2-(2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)benzamide (KI-CDK9d-32):

A mixture of 15 (23.9 mg, 0.0461 mmol) and 4 N HCl in 1,4-dioxane (1.00 mL) was stirred at 40 °C overnight. Concentration gave a crude mixture, which was used in the next step without further purification. iPr₂NEt (24.1 mL, 0.138 mmol) and HATU (35.1 mg, 0.0922 mmol) were added to a stirred solution of the crude material and KI-ARv3 hydrochloride (27.3 mg, 0.0922 mmol) in DMF (1.00 mL) at room temperature. After stirring overnight, the mixture was poured into H₂O, and extracted with CH₂Cl₂. The combined organic layer was dried over anhydrous Na₂SO₄. Filtration and concentration gave a crude mixture, which was purified on HPLC (MeCN:H₂O = 10:90 to 90:10, including 0.1% HCO₂H) to yield a formic acid salt of the title compound as a white solid (11.2 mg, 35% in 2 steps). ¹H NMR (500 MHz, DMSO) δ 10.94 (s, 1H), 8.37 (d, J = 7.3 Hz, 1H), 8.15 (s, 1H), 8.01 (d, J = 2.3 Hz, 1H), 7.83 (d, J = 8.0 Hz, 2H), 7.67 (d, J = 7.7 Hz, 1H), 7.52 (d, J = 8.4 Hz, 1H), 7.43 (d, J = 8.0 Hz, 2H), 7.08 – 7.02 (m, 2H), 6.30 (d, J = 2.2 Hz, 1H), 6.05 (s, 1H), 5.04 (dd, J = 13.3, 5.1 Hz, 1H), 4.54 – 4.43 (m, 1H), 4.32 (d, J = 16.9 Hz, 1H), 4.29 – 4.24 (m, 1H), 4.20 (d, J = 17.0 Hz, 1H), 3.59 (s, 2H), 3.33 – 3.27 (m, 4H), 2.95 – 2.84 (m, 1H), 2.65 – 2.58 (m, 3H), 2.57 – 2.46 (m, 5H), 2.41 – 2.29 (m, 1H), 2.29 – 2.20 (m, 1H), 2.19 – 1.99 (m, 2H), 1.99 – 1.91 (m, 1H), 1.83 – 1.58 (m, 4H), 0.92 (t, J = 7.4 Hz, 3H). ¹³C NMR (126 MHz, DMSO) δ 172.95, 171.31, 168.35, 165.87, 163.22, 162.27, 153.74, 148.83, 146.05, 144.04, 143.10, 141.22, 133.60, 128.64, 127.34, 123.76, 121.53, 114.74, 108.41, 93.68, 84.97, 61.53, 52.34, 51.47, 51.41, 49.19, 47.60, 46.99, 40.43, 38.40, 31.27, 31.06, 30.60, 22.59, 21.95, 13.79. QToF HRMS m/z: calcd for C39H46N9O4⁺ [M+H⁺] = 704.3667; Found 704.3682.

KI-CDK9d-D32N

4-((4-(2-(1-methyl-2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)benzamide

tert-butyl 4-(2-(1-methyl-2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazine-1-carboxylate (19):

A mixture of tert-butyl 4-(3-(bromomethyl)-4-(methoxycarbonyl)phenyl)piperazine-1-carboxylate⁸⁵ (17, 77.9 mg, 0.188 mmol), 3-amino-1-methylpiperidine-2,6-dione hydrochloride⁸⁵ (18, 35.3 mg, 0.198 mmol) and iPr₂NEt (24.1 mL, 0.138 mmol) in acetonitrile (0.627 mL) was stirred at 80 °C for 24 hours. The mixture was then concentrated and purified by silica gel column chromatography (hexane:ethyl acetate = 50:50 to 0:100) to yield the title compound as a purple solid (40.7 mg, 49%). ¹H NMR (500 MHz, CDCl₃) δ 7.75 (d, J = 8.7 Hz, 1H), 6.98 (d, J = 8.7 Hz, 1H), 6.87 (s, 1H), 5.15 (dd, J = 13.5, 5.0 Hz, 1H), 4.38 (d, J = 15.4 Hz, 1H), 4.25 (d, J = 15.6 Hz, 1H), 3.59 (t, J = 5.2 Hz, 4H), 3.27 (t, J = 5.2 Hz, 4H), 3.17 (s, 3H), 3.02 – 2.93 (m, 1H), 2.90 – 2.79 (m, 1H), 2.35 – 2.22 (m, 1H), 2.20 – 2.11 (m, 1H), 1.48 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 171.50, 170.48, 169.61, 154.78, 154.42, 143.82, 125.22, 122.72, 116.06, 108.91, 80.29, 76.91, 52.52, 48.62, 47.20, 32.24, 28.55, 27.28, 22.96. QToF HRMS m/z: calcd for C23H31N4O5⁺ [M+H⁺] = 443.2289; Found 443.2296.

tert-butyl 4-((4-(2-(1-methyl-2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)benzoate (20):

A mixture of 19 (40.7 mg, 0.0920 mmol) and 4 N HCl in 1,4-dioxane (1.00 mL) was stirred at room temperature. After 3 hours, the mixture was concentrated to give a crude mixture, which was used in the next reaction without further purification. Tert-butyl 4-(bromomethyl)benzoate (14, 26.2 mg, 0.0966 mmol) and K₂CO₃ (38.1 mg, 0.276 mmol) were added to a stirred solution of the crude mixture in DMF (1.00 mL), then the mixture was warmed to 40 °C. After stirring overnight, the mixture was poured into H₂O, and extracted with ethyl acetate. The combined organic layer was dried over anhydrous Na₂SO₄. Filtration and concentration gave a crude mixture, which was purified by silica gel column chromatography (hexane:ethyl acetate = 50:50 to 0:100) to yield the title compound as a white foam (41.8 mg, 85% in 2 steps). ¹H NMR (500 MHz, CDCl₃) δ 7.96 (d, J = 7.8 Hz, 2H), 7.73 (d, J = 8.6 Hz, 1H), 7.41 (d, J = 8.0 Hz, 2H), 6.98 (d, J = 8.5 Hz, 1H), 6.86 (s, 1H), 5.15 (dd, J = 13.4, 5.0 Hz, 1H), 4.37 (d, J = 15.6 Hz, 1H), 4.24 (d, J = 15.6 Hz, 1H), 3.61 (s, 2H), 3.31 (t, J = 5.0 Hz, 4H), 3.17 (s, 3H), 3.02 – 2.93 (m, 1H), 2.90 – 2.79 (m, 1H), 2.60 (t, J = 5.0 Hz, 4H), 2.27 (tt, J = 13.3, 6.7 Hz, 1H), 2.19 – 2.10 (m, 1H), 1.59 (s, 9H). ¹³C NMR (126 MHz, CDCl₃) δ 171.54, 170.52, 169.73, 165.81, 154.52, 143.80, 142.85, 131.21, 129.63, 128.94, 125.11, 122.15, 115.62, 108.41, 81.10, 62.72, 52.95, 52.50, 48.50, 47.20, 32.26, 28.35, 27.27, 22.97. QToF HRMS m/z: calcd for C30H37N4O5⁺ [M+H⁺] = 533.2758; Found 533.2762.

4-((4-(2-(1-methyl-2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)benzoic acid (21):

A mixture of 20 (41.8 mg, 0.0785 mmol) and 4 N HCl in 1,4-dioxane (2.00 mL) was warmed to 40 °. After stirring overnight, the mixture was heated to 70 °C and stirred for 2 hours. The mixture was then concentrated and purified on HPLC (MeCN:H₂O = 10:90 to 90:10, including 0.1% HCO₂H) to yield the title compound as a white solid (23.8 mg, 64%). ¹H NMR (500 MHz, Pyr) δ 8.49 (d, J = 7.9 Hz, 2H), 7.98 (d, J = 8.5 Hz, 1H), 7.66 – 7.55 (m, 2H), 7.08 (dd, J = 8.5, 2.2 Hz, 1H), 7.00 (d, J = 2.2 Hz, 1H), 5.59 (dd, J = 13.7, 5.0 Hz, 1H), 4.51 (d, J = 16.1 Hz, 1H), 4.39 (d, J = 16.0 Hz, 1H), 3.58 (s, 2H), 3.36 – 3.30 (m, 4H), 3.12 (s, 3H), 2.96 – 2.89 (m, 2H), 2.57 (t, J = 5.0 Hz, 4H), 2.44 – 2.33 (m, 1H), 2.08 – 1.99 (m, 1H). ¹³C NMR (126 MHz, Pyr) δ 172.19, 171.63, 170.02, 169.41, 155.01, 145.01, 144.11, 131.90, 130.71, 129.65, 125.15, 123.16, 115.91, 109.31, 62.90, 53.44, 53.29, 48.80, 47.88, 32.71, 27.26, 23.27. QToF HRMS m/z: calcd for C26H29N4O5⁺ [M+H⁺] = 477.2132; Found 477.2142.

4-((4-(2-(1-methyl-2,6-dioxopiperidin-3-yl)-1-oxoisoindolin-5-yl)piperazin-1-yl)methyl)-N-((1R,3R)-3-((5-propylpyrazolo[1,5-a]pyrimidin-7-yl)amino)cyclopentyl)benzamide (KI-CDK9d-32N):

iPr2NEt (26.1 mL, 0.150 mmol) and HATU (57.0 mg, 0.150 mmol) were added to a stirred solution of 21 (23.8 mg, 0.0499 mmol) and KI-ARv3 hydrochloride (29.5 mg, 0.0999 mmol) in DMF (1.00 mL) at room temperature. After stirring overnight, the mixture was poured into H₂O, and extracted with 10% methanol in CH₂Cl₂. The combined organic layer was dried over anhydrous Na₂SO₄. Filtration and concentration gave a crude mixture, which was purified on HPLC (MeCN:H₂O = 10:90 to 90:10, including 0.1% HCO₂H) and prepTLC (8% methanol in CH₂Cl₂) to yield the title compound as a white solid (16.9 mg, 47%). ¹H NMR (500 MHz, DMSO) δ 8.32 (d, J = 7.3 Hz, 1H), 7.97 (d, J = 2.3 Hz, 1H), 7.79 (d, J = 7.9 Hz, 2H), 7.63 (d, J = 7.6 Hz, 1H), 7.49 (d, J = 8.8 Hz, 1H), 7.39 (d, J = 8.0 Hz, 2H), 7.04 – 6.98 (m, 2H), 6.26 (d, J = 2.2 Hz, 1H), 6.00 (s, 1H), 5.07 (dd, J = 13.5, 5.0 Hz, 1H), 4.45 (q, J = 7.1 Hz, 1H), 4.28 (d, J = 16.9 Hz, 1H), 4.23 (q, J = 7.1 Hz, 1H), 4.15 (d, J = 17.0 Hz, 1H), 3.54 (s, 2H), 3.26 (t, J = 4.8 Hz, 3H), 2.95 (s, 3H), 2.98 – 2.88 (m, 1H), 2.75 – 2.66 (m, 1H), 2.61 – 2.54 (m, 2H), 2.49 – 2.44 (m, 4H), 2.37 – 2.26 (m, 1H), 2.25 – 2.16 (m, 1H), 2.15 – 1.98 (m, 3H), 1.98 – 1.88 (m, 1H), 1.79 – 1.55 (m, 5H), 0.88 (t, J = 7.4 Hz, 3H). ¹³C NMR (126 MHz, DMSO) δ 171.97, 170.94, 168.37, 165.86, 162.26, 153.76, 148.83, 146.05, 144.05, 143.09, 141.24, 133.59, 128.62, 127.33, 123.79, 121.51, 114.73, 108.39, 93.67, 84.96, 61.53, 52.34, 51.95, 51.47, 49.19, 47.60, 47.01, 40.43, 38.39, 31.42, 31.05, 30.60, 26.56, 21.94, 21.83, 13.79. QToF HRMS m/z: calcd for C40H48N9O4⁺ [M+H⁺] = 718.3824; Found 718.3834.

QUANTIFICATION AND STATISTICAL ANALYSIS

Details of statistical approaches and software used to analyze specific data are provided in the relevant method section and/or image caption. Significance was set at p < 0.05.

Supplementary Material

2

Table S2. Excel file containing kinase profile data at 10 uM KI-CDK9d-32, related to Figure 2

3

Table S3. Excel file containing kinase profile data at 100 nM KI-CDK9d-32, related to Figure 2

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-Vinculin (E1E9V)	Cell Signaling Technology	Cat#13901S
Mouse monoclonal anti-Rpb1 CTD (4H8)	Cell Signaling Technology	Cat#2629S
Rabbit monoclonal anti-cMYC (D84C12)	Cell Signaling Technology	Cat#5605S
Rabbit monoclonal anti-P-Rpb1 S2 (E1Z3G)	Cell Signaling Technology	Cat#13499S
Rabbit anti-Histone H3	Cell Signaling Technology	Cat#9715L
Rabbit monoclonal anti-GAPDH (14C10)	Cell Signaling Technology	Cat#2118S
Mouse monoclonal anti-CDK7 (MO1)	Cell Signaling Technology	Cat#2916S
Rabbit monoclonal anti-CDK9 (C12F7)	Cell Signaling Technology	Cat#2316S
Rabbit monoclonal anti-CDK10 (D8Z7Z)	Cell Signaling Technology	Cat#36106S
Ant-mouse IgG HRP-Linked	Cell Signaling Technology	Cat#7076S
Anti-rabbit IgG HRP-Linked	Cell Signaling Technology	Cat#7074S
Rabbit polyclonal anti-DDX21	Novus Biologicals	Cat#NB100-1718
Rabbit monoclonal anti-NPM1	Abcam	Cat#ab180607
Alexa Fluor 488	Thermo Fisher Scientific	Cat#A-11008
Alexa Fluor 647	Thermo Fisher Scientific	Cat#A-27040
Chemicals, peptides, and recombinant proteins
KI-CDK9d-32	This manuscript	N/A
KI-CDK9d-32N	This manuscript	N/A
KI-ARv-03	Koehler Lab	Richters et al.
KB-0742 dihydrochloride	Medchem Express	Cat#HY-137478A
Protease inhibitors	Thermo FIsher Scientific	Cat#87786
Phosphatase inhibitors (PhosSTOP)	Millipore SIGMA	Cat#4906845001
Critical commercial assays
HotSpot Wildtype Kinase Panel	Reaction Biology	N/A
Nano-Glo HiBiT Lytic Detection System	Promega	Cat#N3030
PRISM Screen	Broad Institute	N/A
Deposited data
RNA Sequencing data of MOLT-4 Cells treated with KI-CDK9d-32 and KB-0742	This manuscript	GEO: GSE263153
Mass Spectrometry Proteomics data of MOLT-4 Cells treated with KI-CDK9d-32	This manuscript	ProteomeXchange: PXD052300; 10.6019/PXD052300
Experimental models: Cell lines
Human: MOLT-4	ATCC	cat#CRL-1582
Human: RH-4	Cellosaurus	cat#CVL_5916
Human: PSN-1	ATCC	cat#CRL-3211
Human: HeLa	ATCC	cat#CCL-2
Oligonucleotides
MYC Forward: 5'-tcctcggattctctgctctc-3'	This manuscript	N/A
MYC Reverse: 5'-tcttcctcatcttcttgttcctc-3'	This manuscript	N/A
CDK9 Forward: 5'-ctgaagaaggtgctgatggaa-3'	This manuscript	N/A
CDK9 Reverse: 5'-agttgaccacattctcgtgtt-3'	This manuscript	N/A
POLR2A Forward: 5'-ggatgatctggaatgctcag-3’	This manuscript	N/A
POLR2A Reverse: 5'-attccttgactccctccac-3'	This manuscript	N/A
Software and algorithms
clusterProfiler	Bioconductor	10.18129/B9.bioc.clusterProfiler
nf-core/rnaseq workflow revision 3.12.0	nf-core.re	N/A

Highlights.

• KI-CDK9d-32 is a highly potent and selective CDK9 degrader.

• KI-CDK9d-32 leads to rapid downregulation of MYC protein and mRNA transcripts levels.

• KI-CDK9d-32 represses MYC pathways and destabilizes nucleolar homeostasis.

• Multidrug resistance ABCB1 gene is a top resistance marker for KI-CDK9d-32.