A Bivalent Molecular Glue Linking Lysine Acetyltransferases to Oncogene-induced Cell Death

SUMMARY:

Developing cancer therapies that induce specific death of malignant cells is critical for preventing relapse. Highly effective strategies, such as immunotherapy, exemplify this principle. Here we provide the mechanistic basis for a small-molecule approach that leverages chemically induced proximity (CIP) to kill diffuse large B cell lymphoma, the most common non-Hodgkin lymphoma. We developed lysine acetyltransferase (KAT)-based TCIPs (Transcriptional/epigenetic Chemical Inducers of Proximity), or KAT-TCIPs, that redirect p300/CBP to activate cell death networks repressed by the oncogenic driver, BCL6. Our lead KAT-TCIP reprograms the epigenome to initiate apoptosis. A crystal structure of the chemically induced p300-BCL6 complex reveals how chance protein-protein interactions may be exploited to confer the potency and selectivity of KAT-TCIPs. Thus, oncogenic drivers can be co-opted to activate robust cell death. Consistent with their gain-of-function mechanism, TCIPs recruiting different transcriptional activators – p300, BRD4, or CDK9 – produce distinct genomic responses, suggesting specialized therapeutic uses.

IN BRIEF:

Chemically induced proximity of lysine acetyltransferases (KATs) with BCL6 reprograms epigenetic signaling to eliminate lymphoma tumors. Structural and mechanistic studies demonstrate that fortuitous protein-protein contacts convert proximity-induction into targeted changes on chromatin, revealing a key mechanism by which small molecules can co-opt oncogenic transcriptional regulators to elicit malignant cell death.

Graphical Abstract

INTRODUCTION

The paralogous lysine acetyltransferases (KATs) p300 (E1A-associated protein p300, encoded by EP300) and CBP (CREB-binding protein, encoded by CREBBP) are essential regulators of gene expression in healthy and malignant cells¹. p300 and CBP (hereafter: p300/CBP) catalyze histone acetylation, interact with transcription factors (TFs), and activate transcription at promoters and enhancers^2,3. In diverse malignancies, p300/CBP sustain oncogenic transcriptional networks by activating clusters of gene-regulatory regions called enhancers and super-enhancers^4–9. Given their central role in transcriptional regulation, we sought to redirect p300/CBP toward pro-apoptotic transcription in malignant cells using small molecule chemical inducers of proximity (CIPs).

CIPs originally served as essential tools for revealing induced proximity governs diverse biological processes¹⁰ including signal transduction¹¹, post-translational-modification^12,13, affinity modulation of protein surfaces¹⁴, protein degradation¹⁵, chromatin regulation^16–19, and transcription initiation and elongation^20–23. Collectively, these studies established a foundation for the therapeutic development of CIPs, including Proteolysis-targeting chimeras (PROTACs)²⁴ and affinity-modulating RAS inhibitors^25–28. Structural studies showed that CIPs can facilitate protein complex formation by creating a composite binding surface containing the small molecule and the two binding proteins²⁹. Small molecules that assemble cooperative complexes between unrelated proteins are called molecular glues³⁰.

While CIPs are emerging as preclinical tools to degrade p300/CBP in hematological malignancies and solid cancers^31–34, the development of p300/CBP-redirecting therapeutic CIPs remains limited. Prior efforts have largely relied on transgene overexpression³⁵, protein tags^36–38, or large, DNA-binding pyrrole-imidazole polyamide moieties³⁹, technologies not readily translatable to the clinic. Two exceptions are CIPs that harness KAT activity to inactivate a mutant form of p53 via targeted acetylation^40,41.

We recently introduced a class of Transcriptional/epigenetic CIPs (TCIPs) that redirect transcriptional activators to chromatin bound by DNA sequence-specific TFs^22,23. These bivalent molecules redirected the RNA Polymerase II elongation-associated factors Bromodomain and extra-terminal containing protein 4 (BRD4) and Cyclin-dependent kinase 9 (CDK9) to B cell lymphoma 6 (BCL6), a TF that represses pro-apoptotic and growth arrest genes^42–44 and is deregulated in 40 to 60% of diffuse large B cell lymphomas (DLBCLs)⁴⁵. TCIPs rewired BCL6 to activate transcriptionally silent proapoptotic pathways and potently killed cancer cells. Their gain-of-function mechanism, distinct from BCL6 degraders and inhibitors^46–50, produced unique preclinical activity²³. Expanding this approach hinges on identifying diverse transcriptional activators and developing cooperative molecular glues, necessitating structural insight into ternary complexes to drive chemical optimization.

We postulated that CIPs recruiting different classes of transcriptional activators would elicit distinct, though partially convergent, transcriptional outputs. Accordingly, we developed lysine acetyltransferase TCIPs (KAT-TCIPs) that redirect p300/CBP to BCL6-bound genomic loci. We used a systematic chemical design cascade to generate molecules that induced selective p300/CBP-BCL6 complexes in DLBCL cells, driving potent cell killing (IC₅₀ = 0.80 nM) and in vivo activity in lymphoma xenograft models. KAT-TCIPs unidirectionally redistributed p300/CBP to BCL6 loci, driving localized acetylation and target-gene activation. The crystal structure of a KAT-TCIP in complex with p300 and BCL6 revealed selectivity and potency arise from the chance formation of complementary protein-protein interactions on the composite drug-protein interface. KAT-TCIPs demonstrated partially overlapping but distinct genomic and biological activities relative to BRD4- and CDK9-based BCL6 TCIPs, illustrating how transcriptional activators can be modularly deployed to initiate tailored programmed cell death across B cell malignancies.

RESULTS

KAT-TCIPs induce structurally defined biochemical ternary complexes

To explore the potential of harnessing co-activating lysine acetyltransferase activity for transcriptional activation, we synthesized a library of bivalent compounds designed to recruit p300/CBP to BCL6-bound genes (Figure 1A). These bivalent compounds link the BCL6 Broad-Complex, Tramtrack, and Bric à brac (BTB) domain ligand BI-3812⁴⁹ to the p300/CBP-specific bromodomain (BD) inhibitors GNE-781⁵¹ and Cmpd33⁵² (Figures 1B, S1A, and S1B) via conjugatable handles identified from X-ray co-crystal structures. We employed a library of alkyl and PEG linkers to sample chemical space and identify cell-permeable KAT-TCIPs that induce productive ternary complexes (Figures S1A and S1B).

Figure 1. The Crystal Structure of a KAT-TCIP-induced Ternary Complex.

(A) KAT-TCIPs targeting p300/CBP were designed to activate BCL6-controlled cell death and cell-cycle arrest in DLCBL cells.

(B) Domain structures of full-length p300/CBP and BCL6.

(C) Co-crystal structure of MNN-02-155, BCL6-BTB, and p300-BD. A BCL6-BTB homodimer binds two molecules of MNN-02-155, each of which engages one protomer of p300-BD; inset highlights neo-hydrogen bonds.

(D) Binary TR-FRET tracer displacement assay measuring binding of p300-BD-WT, -Q1082A, or -Q1082R to MNN-02-155 in the absence (binary) or presence (ternary) of BCL6 at saturating concentrations, using the fluorescent p300-BD tracer MNN-06-112 (Figure S3D). α = K_{d, app} binary/K_{d, app} ternary. n = 3 independent experiments, mean.

To evaluate the KAT-TCIP library, we used a BCL6-controlled GFP reporter in the DLBCL cell line KARPAS422 (K422)²². MNN-02-155 produced the highest GFP signal, ~5-times greater than DMSO treatment (Figure S1C). We observed a hook effect in GFP activation, caused by high compound concentrations saturating both protein partners and preventing complex formation^10,20. This result suggested MNN-02-155 promotes a ternary complex between p300/CBP-BD and BCL6-BTB that mediates transcriptional activation.

To understand the mode of compound binding, we solved the crystal structure of the BCL6-MNN-02-155-p300 complex at a 2.1 Å resolution (Table S1). The asymmetric unit contained a BCL6-BTB homodimer and two copies of p300-BD molecules, each tethered to BCL6 by MNN-02-155 (Figures 1C and S2A–C, Video S1). MNN-02-155 engaged p300 and BCL6 without significant rearrangements of the parental monovalent protein-compound structures, CBP-GNE-781⁵¹ (RMSD = 0.35 Å, Figure S2D) and BCL6-BI-3802⁴⁹ (a degrader analog of BI-3812, RMSD = 0.43 Å, Figure S2E). Compound binding induced several fortuitous interactions that constituted a p300-BCL6 neo-interface, including hydrogen bonds between the carbonyl of p300-G1085 and the amide nitrogen of the flexible MNN-02-155 linker (3.4 Å), and between the terminal guanidinium group of BCL6-R24 and the backbone carbonyl of p300-Q1082 (3.0 Å, Figure 1C). Additionally, p300-Q1082 engaged BCL6-R24 through shape complementation. Overlaying the CBP-GNE-781 structure indicated an identical set of CBP-BCL6 interactions was possible (Figure S2F). These contacts buried an average surface area of 164.5 Ų, less than 3% of each domain subunit’s total surface area (Figure S2G). Thus, a few opportunistic contacts induced by compound-mediated dimerization are sufficient to drive thermodynamically favorable ternary complex assembly.

The identification of compound-induced contacts raised the possibility of cooperativity in ternary complex formation⁵³, a hallmark of molecular glues^30,54. We measured cooperativity (α)^55,56 for p300 using recombinant p300-BD and BCL6-BTB domains and a time-resolved fluorescence resonance energy transfer (TR-FRET) tracer displacement assay, defining α as the ratio of the K_{d, app} measured for binary p300-MNN-02-155 binding to ternary p300-[BCL6-MNN-02-155] binding. For p300-WT, α was 6.9 (Figure 1D), indicating strong ternary complex cooperativity and enhanced MNN-02-155 binding to p300 in the presence of BCL6. To assess the role of the observed side chain interactions for KAT-TCIP binding and cooperativity, we introduced p300-Q1082A and p300-Q1082R mutations, designed to reduce residue bulk or to electrostatically clash with BCL6-R24, respectively. Both mutations diminished ternary complex cooperativity (p300-Q1082A α ~ 2.3; p300-Q1082R α ~ 4.6; Figure 1D). As expected, BCL6 did not affect the affinity of the monovalent inhibitor GNE-781 to p300-BD (p300 WT α ~ 1.1, p300-Q1082A α ~ 0.88; p300-Q1082R α ~ 1.0; Figures S2H and S2I). We conclude that neo-protein-protein interactions mediated by KAT-TCIP geometry, rather than ligand site plasticity, drive cooperative binding, a principle that may apply to other ternary complexes induced through bivalent small molecules.

Optimization of KAT-TCIP molecules with antiproliferative activity

We hypothesized that rigidifying the KAT-TCIP linker would lower the entropic cost of productive complex formation and accordingly generated a focused library of rigidified KAT-TCIPs (Figure S1A). By docking GNE-781-based compounds into the co-crystal structure, performing energy minimization, and a 300 ns molecular dynamics (MD) simulation, we identified a new KAT-TCIP, TCIP3, that exhibited the lowest linker energy strain (Figures 2A and S3A). TCIP3 also activated BCL6 expression in the K422 GFP reporter assay to a greater degree than MNN-02-155 (Figures 2B and S3B) and exhibited a hook effect. We thus used TCIP3 for follow-up mechanistic studies.

Figure 2. Structure and Activity of TCIP3.

(A) Structure of TCIP3, NEG1, and NEG2.

(B) Activation of a BCL6-controlled GFP reporter construct in K422 cells after treatment for 24 h; 3 biological replicates, mean ± s.e.m.

(C) Cell viability after 72 h of compound treatment in SUDHL5 cells; 3–5 biological replicates, mean ± s.e.m.

(D) Cell viability IC₅₀ values (nM) after 7-day treatment in DLBCL and leukemia cells plotted against relative BCL6 protein expression (Figure S4C); 3–4 biological replicates, mean ± s.d.; R of BCL6 expression versus area under viability curve (AUC) computed by Pearson’s correlation; P-value computed by two-sided Student’s t-test.

(E) Number of catalytically-functional CREBBP/EP300 alleles in DLBCL and leukemia cell lines and IC₅₀ (nM) of cell viability corresponding to Figure 2D. GCB: Germinal center B-like; BL: Burkitt Lymphoma; CML: chronic myelogenous leukemia. *viability > 50% at highest dose; Figure S4E.

(F) Cell viability after 72 h treatment of TCIP3 or GNE-781 in RL cells harboring catalytically inactive CBP mutations (Y1503C, Y1482N, or R1446C) and a monoallelic p300 deletion (WT/KO); mean ± s.e.m. of two clones.

To distinguish the effects of ternary complex formation from p300/CBP-BD or BCL6-BTB inhibition while controlling for compound size and cell permeability, we synthesized negative controls containing the TCIP3 linker that disrupt binding to either BCL6-BTB (NEG1) or p300/CBP-BD (NEG2) (Figure 2A). NanoBRET⁵⁷ intracellular probe-displacement assays in HEK293T cells (Figures S3C and S3D) confirmed NEG1 and NEG2 retained membrane permeability and binding to p300/CBP-BD or BCL6-BTB, respectively. Neither NEG1 nor NEG2 activated GFP expression (Figure 2B). Throughout our mechanistic studies, we used these negative controls to determine the requirement for ternary complex formation.

Having established that KAT-TCIPs activate BCL6-dependent transcription, we examined their effects on the viability of DLBCL cells. In a 72-hour cell viability assay, TCIP3 inhibited the proliferation of SUDHL5 DLBCL cells with an IC₅₀ of 0.80 nM (Figure 1C). TCIP3 was over 1,000 times more cytotoxic than either NEG1 or NEG2, and nearly all GNE-781-based KAT-TCIPs were 10 – 38,000-times more potent than the inhibition, degradation, or co-inhibition of p300/CBP and BCL6 (Figures 2C and S1A). TCIP3 was at least an order of magnitude more potent than p300/CBP KAT inhibition (A-485⁵⁸), p300/CBP degradation (dCBP-1³⁴), or BCL6 degradation (ARV-393⁵⁹) (Figures 2C, S3E, and S3F).

The enhanced potency of KAT-TCIPs relative to BCL6 loss-of-function modalities prompted us to quantify the dependence of TCIP3 activity on cellular BCL6 engagement. Fractional occupancy calculations based on NanoBRET studies in HEK293T cells (Figure S3C) revealed that TCIP3 binds to < 10% of BCL6 at 10 nM, a concentration that kills 90% of cells (Figure 2C) and activates the BCL6 reporter 5-fold (Figure 2B). In contrast, BI-3812 binds > 90% of BCL6 at concentrations required for antiproliferative effects (1 μM). The modest BCL6 engagement of TCIP3 at concentrations where it is active suggests that TCIP3 kills DLBCL cells via a gain-of-function model.

We next evaluated the context-specificity of TCIP3-mediated cytotoxicity. In non-malignant B and T cell populations isolated from primary tonsillar lymphocytes, enriched for BCL6-expressing germinal center B cells and T follicular helper cells⁶⁰, and in primary human fibroblasts, which have little BCL6, TCIP3 exhibited minimal cytotoxicity relative to SUDHL5 cells (Figure S4A) and was less toxic than GNE-781, dCBP-1, and A-485. Across eight lymphoma lines with variable levels of BCL6 expression and one leukemia cell line (K562) with negligible BCL6 (Figures 2D and S4C), there was an inverse correlation between BCL6 levels and the area under the dose-response TCIP3 cell viability curve (AUC) (Pearson’s R = −0.83, P = 0.0057). TCIP3 inhibited the viability of high-BCL6 lines RL, SUDHL5, SUDHL4, K422, and DB at comparable potencies, 200 to 5500 times greater than the co-treatment of the inhibitors GNE-781 and BI-3812 (Figures 2D, S4E, and S4F). In cells with low BCL6 levels, TCIP3 performed comparably to the co-treatment of GNE-781 and BI-3812 (Figures 2D, S4E, and S4F). We conclude that TCIP3 exerts its cytotoxicity in a malignant- and BCL6-dependent manner.

We investigated how KAT dosage affects TCIP3 sensitivity by analyzing whole-genome sequencing data, focusing on the frequently-mutated^61,62 alleles of CREBBP and EP300. (Figures 2E and S4D; Table S2). Despite having only one wild-type CREBBP/EP300 allele, the high-BCL6 line K422 was comparably sensitive to TCIP3 as SUDHL5, SUDHL4, RL, and DB (Figures 2E and S4E). To measure the impact of EP300/CREBBP dosage on compound sensitivity in an isogenic context, we biallelically mutated residues to inactivate the CBP KAT domain (R1446C, Y1482N, Y1503C) in RL cells, which have only one functional EP300 allele (WT/KO)⁶³. TCIP3 retained equivalent cytotoxicity in cells with only one WT KAT as in cells with all four WT alleles (Figure 2F). In contrast, GNE-781 was more potent in KAT-domain-mutant cells (Figure 2F), consistent with the reported increased susceptibility of cell lines with loss-of-function mutations in CREBBP to p300/CBP bromodomain inhibition⁶⁴. Neither NEG1 nor NEG2 were cytotoxic in these cells (Figures S4G, S4H). Comparable TCIP3 sensitivity in the presence of fewer copies of catalytically active p300/CBP indicates that TCIP3 relies on a fraction of catalytically-functional p300/CBP to kill cells.

We next asked if TCIP3 suppresses malignant cell growth through BCL6-dependent sequestration of p300/CBP away from its normal chromosomal substrates. We overexpressed the BCL6 BTB domain tagged with a nuclear localization signal (NLS) in BCL6-low K562 cells (Figure S4I). This construct binds TCIP3 but lacks the DNA-binding zinc finger domains required to bind death genes on chromatin. K562 cells are otherwise sensitive to p300/CBP acetyltransferase inhibition⁶⁵. Overexpressing BCL6-BTB did not increase sensitivity to TCIP3 (Figure S4I). Our results indicate that TCIP3 activity requires not only high levels of BCL6, but also genomic-binding and active repression of cell death programs. Collectively, cellular profiling indicates that TCIP3-mediated CIP between p300/CBP and BCL6 drives a potent, gain-of-function effect requiring only fractional engagement of either protein, distinct from the effects of protein inhibition, sequestration, or degradation.

Selective ternary complex formation in cells is required for activity

To examine if KAT-TCIP cytotoxicity depends on concomitant cellular engagement of both proteins, we co-treated SUDHL5 cells with 1 nM of TCIP3 and either GNE-781⁵¹ (Figure 3A) or one of three BCL6-BTB inhibitors: BI-3812⁴⁹, GSK137⁵⁰, and an analog of CCT373566 (CCT373566a, File S1)⁴⁷ (Figure 3B). Co-treatment with each inhibitor buffered the effects of TCIP3 on cell viability (Figure 3A, 3B). Therefore, chemically-induced cellular complex formation between p300/CBP and BCL6 is required for cell killing.

Figure 3. TCIP3 is a Molecular Glue that Kills Cells Via Chemical Induced Proximity.

(A) SUDHL5 cell viability after competitive titration of constant 1 nM TCIP3 with p300/CBP bromodomain or (B) BCL6-BTB domain inhibitors (72 h; 3 biological replicates, mean ± s.e.m.).

(C) Isothermal calorimetry (ITC) traces of ternary complex formation; representative of 3 independent experiments.

(D) K_d and α (α = K_d binary/K_d ternary) values for each binding event, calculated from n = 2–3 independent experiments, mean ± s.e.m. Corresponds to Figure S5F.

(E) p300 immunoprecipitation-mass spectrometry (IP-MS) from SUDHL5 cells and (F) FLAG IP-MS from genomic knock-in FLAG-tagged BCL6 SUDHL5 cells treated with 1 nM TCIP3 for 2 h. Proteins containing peptides only after TCIP3 treatment were imputed. For (E) and (F), dashed lines indicate cutoffs of |log₂(fold change)| ≥ 2 and P ≤ 0.01; 3 biological replicates; P-values computed by a moderated t-test.

(G) ChIP-seq analysis of p300 binding and (H) BCL6 binding at high-confidence BCL6 binding sites at indicated treatments in SUDHL5 cells. For (G) and (H), signal profiles are averaged from three (p300 ChIP-seq) or two (FLAG-BCL6-ChIP-seq) sequence-depth-normalized biological replicates with inputs subtracted.

While almost all KAT-TCIPs induced p300/CBP-BCL6 complexes as measured by TR-FRET (Figures S5A and S5B), TCIP3 exhibited one of the highest increases in p300/CBP-BD-BCL6-BTB TR-FRET signal (Figures S5B–S5D). We compared cooperativity across KAT-TCIPs of varying potency and found that TCIP3 and MNN-02-155 produced similar α values, whereas less cytotoxic compounds (MNN-02-156 and MNN-02-162) exhibited minimal cooperativity (Figure S5E). These results indicate many linker architectures can support ternary-complex-driven activity, consistent with our structural data suggesting fortuitous protein-protein contacts are sufficient for compound potency.

We next used isothermal calorimetry (ITC) to quantify the thermodynamic cycle of ternary complex formation^53,66. TCIP3 exhibited enhanced affinity for the ternary complex compared to binary binding with either p300-BD or BCL6-BTB (α^{p300 into BCL6:TCIP3} = 3.5, α^{BCL6 into p300:TCIP3} = 13, Figures 3C, 3D, and S5F), consistent with TR-FRET measurements (Figure S5E). The large entropic penalty observed for ternary binding may reflect protein ordering upon complex formation, which is offset by favorable enthalpic interactions that stabilize the complex (Figure S5F). Consistent with the path-independence of free energy, there was only a 4% deviation in the ΔG of complex formation between the p300- and BCL6-nucleated binding mechanisms (Figure S5F). Our biophysical data confirms that TCIP3 acts as a molecular glue by inducing the formation of a cooperative ternary complex containing new, stabilizing interactions between p300/CBP and BCL6.

The human genome encodes 62 structurally homologous bromodomains⁶⁷ and 183 proteins with BTB domains⁶⁸. We assessed the selectivity of TCIP3 using immunoprecipitation-mass spectrometry (IP-MS), treating SUDHL5 cells with 1 nM TCIP3 for 2 hours (Table S3). Following p300 immunoprecipitation, the only significantly enriched protein upon TCIP3 cellular treatment was BCL6 (P = 1.8 × 10⁻⁹, log₂(fold change) = 4.0) (Figure 3E and S6A). We next performed IP–MS in SUDHL5 cells harboring a genomically integrated C-terminal FLAG tag at the BCL6 locus. The only proteins that showed statistically significant co-enrichment after immunoprecipitation with an anti-FLAG antibody were CBP (P < 0.0001, log₂(fold change) = 3.7) and p300 (imputed log₂(fold change) = 3.7), which had no detected peptides in the DMSO condition (Figures 3F, S6B, and S6C; Table S3). Other bromodomains only bound TCIP3 at high concentrations in a biochemical assay (10 μM, Figure S6D). To validate these results in a genetically unmodified cell, we immunoprecipitated BCL6 using an antibody raised against an epitope in its N-terminus and observed 10- to 40-fold enrichment of both p300 and CBP (Figure S6E). Our profiling indicates that TCIP3 selectively induces an assembly of BCL6-p300/CBP complexes in cells. While cooperativity between the BTB and BD drives thermodynamically favored complex formation, interactions between BCL6 and p300/CBP outside of the BTB and bromodomains nucleated by TCIP3 may also contribute to complex selectivity.

Enrichment of p300 at BCL6-bound genomic loci

To define the specific effects of TCIP3-mediated ternary complex formation on the epigenetic landscape, we conducted chromatin immunoprecipitation-sequencing (ChIP-seq) for p300 and FLAG-BCL6 upon compound addition to SUDHL5 cells. 10,059 and 3,611 peaks were reconstructed for p300 ChIP-seq and FLAG-BCL6 ChIP-seq, respectively. Most of the variance between conditions was attributable to compound treatment (Figures S7A and S7B). TCIP3 (1 nM, 2 hours) increased p300 binding 4-fold at FLAG-BCL6 summits genome-wide (Figure 3G). 80% of peaks that increased in p300 binding overlapped with BCL6-bound peaks (Figures S7C and S7D). BCL6 occupancy at these loci increased modestly (1.5-fold, on average, Figure 3H) and 76% of peaks that increased in BCL6 binding overlapped with the peaks that gained p300 binding (Figures S7E and S7F). This likely reflects compound-induced stabilization, as has been previously observed²³. There was minimal reduction of p300 binding (29, 0.3% of all peaks decreased, Figure S7C) or BCL6 binding (144, 4.0% of all peaks decreased, Figure S7E) after treatment.

We also identified 7,379 BCL6 binding sites in untreated SUDHL5 cells by Cleavage Under Targets and Release Using Nuclease (CUT&RUN⁶⁹) using an endogenous BCL6 antibody. Endogenous BCL6 CUT&RUN signal correlated highly with FLAG-BCL6 ChIP-seq signal in both DMSO and TCIP3-treated conditions (Figure S7G), and both methods recovered the annotated BCL6 binding-site motif (Figure S7H). Consistent with ChIP-seq analyses, TCIP3 induced p300 binding at BCL6 peaks defined by CUT&RUN and stabilized BCL6 binding at these same regions (Figure S7I). Collectively, our genomic data indicates that TCIP3 rapidly and unidirectionally recruits p300 to BCL6-bound chromatin.

KAT-TCIPs were designed to harness the lysine acetyltransferase activity of p300/CBP. To assess whether ternary-complex geometry supports enzymatic activity, we overlaid the p300-MNN-02-155-BCL6-BTB structure onto the p300 catalytic core (PDB: 6GYR), which contains the catalytic KAT domain^70,71, the p300/CBP auto-inhibitory loop (AIL)⁷², and the really interesting new gene (RING) domain⁷⁰. This model positions active p300 protomers in an orientation compatible with acetylation of BCL6 and proximal chromatin (Figure S7J, teal). Because BCL6 is an obligate homodimer ^73,74, each chromatin-bound BCL6 site could recruit four molecules of p300 without steric clash (Figure S7J, gray). p300 activation occurs through trans auto-acetylation across homodimers assembled on dimeric binding partners⁷¹. Structural modeling suggests BCL6 may facilitate KAT trans-acetylation, reinforcing local chromatin acetylation and higher-order p300 oligomerization on chromatin.

Rapid acetylation of BCL6 and BCL6-proximal chromatin

Selective BCL6-p300/CBP ternary complex formation in cells and on chromatin implicated a proximity-dependent molecular mechanism of cell death. Because p300/CBP acetylates a wide range of proximal targets⁷⁵, we assessed total histone acetylation after treating SUDHL5 cells with 1 nM TCIP3, the approximate antiproliferative IC₅₀ (Figure 2C). Of the potential histone lysines that undergo acetylation, we focused on histone H3-K27 and histone H2B-K20 (H3K27ac and H2BK20ac, respectively) since they co-localize with and broadly mark active enhancers and promoters^76,77. TCIP3 treatment did not significantly alter global H3K27ac or H2BK20ac levels (Figure 4A) despite significantly inducing other molecular changes, including repression of the c-MYC protein and induction of the cell cycle inhibitor p27 (Figure S8A). Thus, TCIP3 does not globally inhibit p300/CBP acetyltransferase activity, and its effects arise from specific activity modulation.

Figure 4. TCIP3 Redistributes p300/CBP Activity to BCL6 and Proximal Chromatin.

(A) Representative immunoblot of SUDHL5 cells treated with 1 nM TCIP3 for the indicated time (left) and quantification of H3K20ac and H3K27ac levels (right). 3 biological replicates, mean ± s.e.m. P-values computed by Fisher’s LSD test after ANOVA; No comparisons to DMSO were significant (P > 0.05).

(B) Acetylated lysine (K-ac)-IP and representative BCL6 immunoblot after 1 h of TCIP3 in SUDHL5 cells (left); quantification of BCL6 K-ac levels (right). 3 biological replicates, mean ± s.e.m. P-values computed by Fisher’s LSD test after ANOVA; *P < 0.05 vs DMSO.

(C) Changes in H3K27ac and H2BK20ac by ChIP-seq.

(D) Enrichment of predicted TF binding in gained H3K27ac and H2BK20ac peaks calculated by overlap with public blood-lineage ChIP-seq datasets; full enrichment data in Table S4; P-values computed by two-sided Fisher’s exact test and adjusted by Benjamini-Hochberg.

(E) Induction of H2BK20ac, H3K27ac, p300, and FLAG-BCL6 with time at the promoter of the BCL6 target gene ARID3B; track signals are averaged from three (p300 ChIP-seq) or two (H3K27ac, H2BK20ac, and FLAG-BCL6 ChIP-seq) sequence-depth normalized biological replicates with inputs subtracted. For BCL6 CUT&RUN, signal averaged from two sequence-depth normalized biological replicates.

(F) Overlap of gained and lost H3K27ac peaks after 1 h of 1 nM TCIP3 with annotated enhancers and super-enhancers in SUDHL5 cells.

(G) H3K27ac changes at annotated enhancers and super-enhancers after 2 h of 1 nM TCIP3. For (C) and (G): significant peaks: adj. P ≤ 0.05 and |log₂(fold change)| ≥ 0.5; 2 biological replicates, P-values computed by two-sided Wald test and adjusted for multiple comparisons by Benjamini-Hochberg.

(H) Gene set enrichment analysis (GSEA) of ranked log₂(fold change) in gene expression after 1 nM TCIP3 in SUDHL5 cells across 4 timepoints, and in SUDHL4, RL, K422, and OCILY19 cells at 4 h; only all gene sets adj. P ≤ 0.05 at all timepoints displayed; positive and negative normalized enrichment scores (NES) indicate gene sets enriched in TCIP3-induced and decreased genes, respectively; P-values computed by permutation and adjusted for multiple comparisons by Benjamini-Hochberg. Expanded gene set names are in Figure S9B.

(I) Changes in p300 at all enhancers and super-enhancers, classified by change in H3K27ac from (C); P-values adjusted by Tukey’s test after ANOVA; ****adj. P < 0.0001, ns: not significant.

(J) Model of TCIP3-induced recruitment of p300/CBP to BCL6 loci.

p300/CBP-mediated acetylation of BCL6 in its disordered repression domain 2 derepresses BCL6 target genes by blocking co-repressor binding without altering DNA-binding capacity^78,79. Immunoprecipitation of all acetylated proteins from SUDHL5 nuclear extracts with a pan-acetyl-lysine antibody showed TCIP3 induced a dose-dependent increase in acetylated BCL6 at 1 hour (Figure 4B), potentially contributing to its gained activation ability.

To examine the effects of TCIP3 on histone tail lysines near chromosomal BCL6 binding sites, we conducted ChIP-seq for H2BK20ac and H3K27ac after addition of 1 nM TCIP3 to SUDHL5 cells for 15 minutes, 1 hour, and 2 hours. 69,719 and 66,995 peaks across all timepoints were reconstructed for H2BK20ac and H3K27ac, respectively. Most of the variance between conditions was attributable to compound treatment (Figure S8B). Differential peak analysis detected large gains and losses of acetylation: after 1 hour, 936 H2BK20ac peaks were gained and 4,346 were lost, while 533 H3K27ac peaks were gained and 5,216 were lost (Figure 4C).

Peaks that gained acetylation were enriched for BCL6 binding sites in human B-cell and blood cancer cell lines (Figure 4D; Table S4). Enrichment for BCL6-binding sites was more pronounced in regions that gained H2BK20ac than in regions that gained H3K27ac, consistent with the reported propensity of p300/CBP to catalyze H2BK20ac⁷⁶. Loci that gained p300 and acetylation included pro-apoptotic BCL6-targets such as ARID3B⁸⁰ (Figure 4E) and cell cycle inhibitors such as CDKN1B (See discussion below, Figure S14B). 18% (H3K27ac) to 26% (H2BK20ac) of gained acetylation regions in TCIP3-treated cells overlapped with BCL6 binding sites (Figures 3H and S8C). This is consistent with BCL6 occupancy calculations (Figure S3C) showing TCIP3 engages only a fraction of BCL6, limiting acetylation increases and transcriptional activation to a subset of BCL6-regulated programs. The remaining gains could reflect secondary effects, off-target activity, or transient BCL6 binding.

Reduced activity of oncogenic regulatory regions independent of p300 occupancy

In contrast to gains in histone acetylation, losses occurred at annotated enhancers and super-enhancers in SUDHL5 cells (Figures 4F and S8D). Losses of H3K27ac and H2BK20ac were modest in magnitude across enhancers and super-enhancers, including at those bound by BCL6, but statistically significant (adj. P ≤ 0.05) after 15 min, 1, and 2 hours of TCIP3 treatment (Figure S8E). The greatest losses in H3K27ac were concentrated in super-enhancers, broad regions of elevated histone acetylation that regulate the expression of cell-identity and proliferation genes^6,7,81 (Figure 4G). These losses were observed at several super-enhancers proximal to master regulators of the germinal center B cell and oncogenic drivers, including the BCL6 gene itself⁸² (Figure S8F) as well as preferentially in the most highly acetylated regions (Figures S8G and S8H).

Since enhancer and super-enhancer acetylation promotes transcription⁸³, we hypothesized that TCIP3 should produce decreases in enhancer- and super-enhancer-target gene expression concurrent with increases in BCL6-regulated genes. RNA-sequencing (RNA-seq) after 30 min, 1, 2, and 4 hours of 1 nM TCIP3 addition in SUDHL5 cells revealed large numbers of genes induced (1,510) and decreased (2,126) (Figure S9A). Gene set enrichment analysis showed that genes associated with BCL6 mutation or chemical perturbation increased, and genes associated with p300/CBP knockdown decreased (Figures 4H and S9B). Three additional high-BCL6 DLBCL cell lines (Figures 2D and S4C) with different mutations in EP300 and CREBBP (Figure 2E) showed similar transcriptional responses at 4 hours (Figure 4H). As predicted, most genes near (super)-enhancers that lost H3K27ac decreased in expression (Figures S9C and S9D). Consistent with ternary-complex-dependent activity (Figure 3), OCILY19, which has negligible BCL6 protein (Figures 2D and S4C), did not exhibit significant changes in gene expression (Figure 4H). Overall H3K27ac and gene expression changes exhibited a weak but statistically significant positive correlation (Figure S9E), indicating a modest direct relationship between histone acetylation and transcription of nearby genes that is consistent with published data^76,84.

Despite decreases in enhancer- and super-enhancer acetylation and associated gene expression, we did not observe generalized, statistically significant decreases in p300 binding at enhancers or super-enhancers (Figures 4I, S9F, and S9G). However, increased p300 binding at a subset of enhancers correlated with gained acetylation (Figure 4I). Enhancers and super-enhancers are known to be preferentially sensitive to small changes in co-activator concentrations due to limiting quantities of co-activators in the nucleus, competition with 10-fold more abundant histone deacetylases⁸⁵, and the presence of highly cooperative networks of interacting proteins^6,7. Our data suggest fractional redirection of p300/CBP activity to BCL6 – both to acetylate nearby histones as well as BCL6 protein – is sufficient to perturb sensitive DLBCL enhancer networks and reduce the transcription of enhancer and super-enhancer target genes. Meanwhile, increased transcription of BCL6 target genes arises from p300/CBP recruitment to BCL6 loci and the consequent acetylation of histones and BCL6 itself (Figure 4J).

TCIP3 partially phenocopies proteomic response of p300/CBP degradation

Given that TCIP3 perturbed p300/CBP-regulated networks, we compared the proteomic effects of 10 nM TCIP3 and 250 nM of the p300/CBP degrader dCBP-1 at 6 (Figure S10A) and 24 hours (Figure S10B). TCIP3-induced global proteomic changes correlated (Pearson’s R = 0.77, P < 0.0001) with dCBP-1-induced changes (Figure 5A). This was striking given that TCIP3 was dosed at a 25-fold lower concentration and neither decreased p300 binding at regulatory regions nor depleted p300/CBP (Figures 4I, 5A, and S10A–C). Instead, p300/CBP protein levels increased upon TCIP3 (Figures 5A and S10A–C) treatment in contrast to 250 nM dCBP-1, which induced statistically significant losses in p300/CBP (Figures 5A and S10A–C). This increase is consistent with observed compound-induced protein stabilization (Figures 3G, 3H) and increased transcripts of the BCL6-target gene EP300⁸⁶ (Figure 5C). Both TCIP3 and dCBP-1 induced significant decreases of germinal center B-cell-specific TFs including MEF2B, IRF8, SPIB (PU.1-related)^3,42,87, and modestly of BCL6 itself (Figures 5A and S10C). This is consistent with the decreases in transcription observed at the genes encoding these TFs and the decreases in acetylation observed at their super-enhancers (Figures 4F, 4G, S8F, S9C, and S9D). Thus, minimal redirection of p300/CBP to BCL6 is sufficient to decommission p300/CBP-regulated signaling.

Figure 5. Activation of Apoptotic Signaling by TCIP3.

(A) Whole-proteome profiling of indicated treatments in SUDHL5 cells (24h); labeled TCIP3-regulated proteins are statistically significantly (adj. P ≤ 0.05); P-values computed using a moderated t-test and adjusted by Benjamini-Hochberg.

(B) MSigDB Hallmark 2020 pathway enrichment among significantly increased proteins in SUDHL5 (adj. P < 0.05, log₂(foldchange) > 1) after TCIP3 treatment (24 h, 10 nM).

(C) Correlation of proteomic (10 nM) and bulk transcriptomic changes (1 nM) in SUDHL5 cells treated with TCIP3 at indicated timepoints; only genes whose transcripts changed significantly (adj. P ≤ 0.05) are shown; RNA-seq P-value computed by two-sided Wald test and adjusted by Benjamini-Hochberg; for (A) and (C), 3 biological replicates; R computed by Pearson’s correlation and P-value computed by two-sided Student’s t-test.

(D) Representative immunoblot of pro-apoptotic proteins and (E) quantification of BBC3/PUMA after indicated treatments in SUDHL5 cells for 48 h. 3 biological replicates, mean ± s.e.m. P-values computed by Fisher’s LSD test after ANOVA; **P < 0.01, *P < 0.05 vs. DMSO.

(F) Annexin V-positive SUDHL5 cells with indicated treatments at 24, 48, or 72 h. For (F) and (G) 3–12 biological replicates, mean ± s.e.m. P-values adjusted by Tukey’s test after ANOVA; ****adj. P < 0.0001, ***adj. P < 0.001, **adj. P < 0.01, *adj. P < 0.05. The following comparisons were significant: 24h – ****TCIP1 vs TCIP3 and DMSO, *TCIP3 vs DMSO; 48h – ****TCIP1 and TCIP3 vs DMSO; 72h – ****TCIP1 and TCIP3 vs NEG1, NEG2, and DMSO, ****dCBP-1 vs DMSO, ***dCBP-1 vs NEG2, and **dCBP-1 vs NEG1.

(G) Annexin V-positive SUDHL5, RL, and K422 cells with indicated treatments for 72 h.

(H) TUNEL and total DNA content co-staining of SUDHL5 cells treated with indicated compounds for 24, 48, or 72 h. P-values computed by Fisher’s LSD test after ANOVA; *P < 0.05 vs. DMSO; 4 biological replicates, mean ± s.e.m.

(I) Cell cycle distribution of fixed SUDHL5, RL, or K422 cells after 24 h treatment with indicated compounds; 3 biological replicates, mean ± s.e.m. P-values adjusted by Dunnett’s T3 test after ANOVA; ****adj. P < 0.0001, ***adj. P < 0.001, **adj. P < 0.01, *adj. P < 0.05 vs. TCIP3.

Despite the correlation in global proteomics between TCIP3 and dCBP-1, several proteins uniquely increased in abundance after TCIP3 treatment, including BCL6-regulated and/or p53-target genes that play critical roles in apoptosis such as BBC3/PUMA, ARID3A, and ARID3B (Figures 5A and S10A–C). Proteins significantly increased by TCIP3 but not dCBP1 were enriched with high confidence for p53-target and apoptosis proteins (Figures 5B and S10D). Proteomic changes induced by TCIP3 correlated with transcriptomic effects (Pearson’s R = 0.31, P < 0.0001) (Figure 5C). Notably, the BCL6 degrader ARV-393⁵⁹ failed to induce p27 expression or MYC depletion to the same degree as TCIP3 at equivalent doses, despite efficiently degrading BCL6 (Figure S10E). These data indicate TCIP3 distinctively activates apoptotic and cell cycle arrest protein signaling relative to p300/CBP and BCL6 degraders.

Induction of apoptosis and G1 arrest

We quantified the kinetics of TCIP3-induced apoptosis by analyzing the apoptotic signaling cascade. Annexin V staining for the externalization of phosphatidylserine⁸⁸, one of the first events in a cell undergoing apoptosis⁸⁹, showed that significant increases in Annexin V-positive cells began after 48 hours of 10 nM TCIP3 treatment in SUDHL5 cells (Figure 5F), coinciding with increased PUMA/BBC3 (Figures 5D, 5E, and S11A). RL, but not K422 cells, displayed a significant increase in Annexin V positivity and PUMA (Figures 5G, S11B, and S11C), indicating heterogeneous death mechanisms across DLBCL lines. In SUDHL5, levels of cleaved caspase-3, the terminal executioner protease in the apoptotic cascade⁹⁰, increased coincident with the increase in Annexin V-positive cells (Figure S11D). DNA fragmentation by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining (Figure 5H) and loss of membrane integrity by Trypan blue staining (Figure S11E) both began after 48 hours of compound treatment. dCBP-1 induced caspase-3 cleavage after 4 hours (Figure S11D) and apoptosis after 72 hours (Figure 5F) but did not activate PUMA (Figures 5D and 5E), indicating a different mechanism of cell death. A 24-hour exposure to TCIP3 was sufficient to trigger apoptosis comparable to a 72-hour continuous treatment (Figure S11F), suggesting activation of one or more irreversible apoptotic signals within 24 hours, resulting in apoptosis over the subsequent 24–48 hours.

In addition to elevating apoptotic proteins, TCIP3 induced cell-cycle inhibitors such as CDKN1B/p27 concomitant with decreased levels of master regulators of germinal center B cell proliferation^91–93 (Figure 5C). Accordingly, we observed significant enrichment in cells arrested in G0/G1 after TCIP3 treatment and a corresponding significant reduction in S phase populations in SUDHL5, RL, and K422 cells (Figure 5I). The measurements of apoptosis and cell cycle analysis across DLBCL cell lines indicate that while TCIP3 induces arrest of cell cycle progression uniformly across BCL6-high cell lines, the timing and commitment to apoptosis is variably penetrant.

Mechanistic diversity of BCL6-targeting TCIPs

The kinetics of apoptosis induction and cell cycle arrest differed strikingly from those of previously described BCL6-targeting TCIPs, TCIP1²² and CDK-TCIP1²³, which recruit elongation factors associated with RNA Polymerase II to BCL6. TCIP3 treatment exhibited slower kinetics of apoptosis induction relative to TCIP1 (Figure 5F), which recruits BRD4 to activate elongation at BCL6-target pro-apoptotic BH3-only genes such as BIM/BCL2L11 and PMAIP1²². Concurrent analysis of cell cycle and apoptotic effects showed that TCIP3, unlike TCIP1, arrests cells in G1 before inducing apoptosis (Figures 5H and S12A). Unbiased clustering of transcriptomic changes demonstrated that TCIP3 produced a different gene expression program that clustered independently from TCIP1 and CDK-TCIP1 (Figure S12B). Only TCIP3 induced changes in chromatin acetylation (Figure S12C). These mechanistic differences corresponded to distinct sensitivity profiles among 859 cell lines profiled in the PRISM^65,94 collection (Figure S12D). Nevertheless, all TCIPs were highly toxic to BCL6-driven lymphomas (Figures S12D and S12E). Other sensitivities remain to be explored.

Rapid and potent c-MYC repression

c-MYC, the most repressed protein following TCIP3 treatment, is a critical oncogenic driver of germinal center proliferation^95,96,97, prompting us to examine the contribution of its loss (Figures 6A and S8A) to the antiproliferative mechanism. c-MYC transcripts were rapidly and significantly depleted upon 1 nM TCIP3 treatment across the BCL6-high cell lines SUDHL5, SUDHL4, RL, and K422 (Figure S13A). In SUDHL5, c-MYC mRNA decreased with a t_1/2 of 33 minutes (Figure 6B). Protein repression at 4 hours (IC₅₀ ~ 1.04 nM, Figure S13B) led to significant loss of c-MYC-target gene expression at the same dose (Figure S13C) and exhibited a t_1/2 of 2.5 hours (Figure S13D).

Figure 6. TCIP3 Induces Rapid and Potent Reduction in c-MYC.

(A) c-MYC peptides detected by global proteomics after indicated treatments in SUDHL5 cells for 24 h; intensities are means of 3 biological replicates. Lines represent median and interquartile range.

(B) Time-course of c-MYC transcripts in SUDHL5 cells or (C) in DLBCL and leukemia cells with varying BCL6 expression¹¹⁶ (Figure S4C) treated with indicated compounds (RT-qPCR, normalized to GAPDH and DMSO treatment). SUDHL5 data is the same in 6B and 6C.

(D) Doxycycline-inducible c-MYC overexpression in SUDHL5 and RL cells.

(E) S phase SUDHL5^{TRE-3xFLAG-MYC} and RL^{TRE-3xFLAG-MYC} cells treated 24 h with doxycycline or vehicle, then 24 h with 10 nM TCIP3 or DMSO; 3 biological replicates, P-values computed by two-tailed ratio paired Students’ t-test.

(F) Cell viability of SUDHL5^{TRE-3xFLAG-MYC} and RL^{TRE-3xFLAG-MYC} cells treated 24 h with doxycycline or vehicle, followed by 72 h treatment with TCIP3. For (B), (C), and (F): mean ± s.e.m., 3 biological replicates.

Multiple lines of evidence indicate that TCIP3-mediated repression of c-MYC is mechanistically distinct from p300/CBP inhibition. Repression of c-MYC transcripts correlated with BCL6 levels across six lymphoma cell lines (Figure 6C). Competition of TCIP3 with BI-3812 or GNE-781 buffered the loss of c-MYC mRNA (Figure S13E). Neither BI-3812, GNE-781, NEG1, NEG2, nor dCBP-1 significantly repressed c-MYC compared to TCIP3 (Figures 6B and S13F). Only modest decreases in histone acetylation were observed at the c-MYC promoter and its known regulatory enhancers^98,99 at timepoints following transcript repression, likely as a consequence of lost transcription (Figures S13G and S13H). Finally, p300 binding at the c-MYC promoter or its enhancers did not change or slightly increased (Figure S13G), in contrast to p300/CBP bromodomain inhibitors and degraders, which repress c-MYC transcription by reducing p300/CBP binding to chromatin at these regions^34,51.

TCIP3 engages multiple effectors to produce a gain-of-function program

To evaluate whether c-MYC repression drives the cellular response to TCIP3, we overexpressed doxycycline(dox)-inducible 3x-FLAG-MYC in SUDHL5, RL, and K422 cells (TRE-3x-FLAG-MYC) (Figures 6D and S13I). Analysis of the cell cycle at 24 hours indicated that overexpression of c-MYC partially reversed the S-phase block (Figure 6E) and afforded a modest (3.8-fold) proliferative benefit in SUDHL5 cells (Figure 6F). However, these effects were not observed in RL or K422 cells (Figures 6E, F, and S13I), suggesting that c-MYC repression, while rapid and conserved across BCL6-dependent DLBCL cell lines, contributes to G1 arrest in some DLBCL cell lines but is insufficient for the cumulative antiproliferative effect of TCIP3.

We next investigated the contribution of the cell cycle inhibitor p27/CDKN1B, a BCL6-target gene¹⁰⁰ that was the most significantly induced protein by TCIP3 (Figure 5A). TCIP3 induced expression of CDKN1B across four BCL6-high DLBCL lines (SUDHL5, SUDHL4, K422, and RL; Figure S14A). In SUDHL5 cells, TCIP3 rapidly recruited p300 and induced histone acetylation at the CDKN1B locus (Figure S14B), suggesting it was a direct target of TCIP3-induced p300/CBP-BCL6 complexes. Deletion of p27 by CRISPR-Cas9 (Figure S14C) partially rescued the anti-proliferative effect of 1 nM TCIP3 treatment relative to a cell line infected with a non-targeting (NT) guide RNA (Figure S14D). Rescue of anti-proliferation in p27-KO cells was due to a failure to induce p27 (Figure S14E) and arrest cells at the G1/S transition (Figure S14F). Consistent with this observation, inhibition of the p27-target cyclin-dependent kinases 4 and 6¹⁰¹ via Palbociclib¹⁰² synergized with TCIP3 (Figure S14G) and cancer cell lines dependent on CDK4 were significantly (adj. P < 0.01) more sensitive to TCIP3 (Figure S14H). However, since neither MYC overexpression (Figures 6D and 6E) nor p27 knockout (Figure S14) fully rescued TCIP3-mediated cell death, despite representing the most highly repressed and induced genes, respectively, we conclude that TCIP3 acts through a coordinated gain-of-function program involving multiple genes rather than a single effector.

Specific ablation of BCL6-driven cells in vivo

We evaluated the metabolic stability of TCIP3 in mouse liver microsomes and observed an intrinsic clearance of 50 μM/min/mg and a t_1/2 of 13.9 minutes (Figure S15A). In vivo pharmacokinetic analysis following a single intraperitoneal dose (5 mg/kg) in mice showed favorable properties, including a half-life of 3.33 h and a C_max of 3.67 μM (Figures S15B, S15C). In plasma protein binding studies, 4.04% and 2.28% of TCIP3 remained unbound in mouse and human plasma, respectively (Figure S15D). The maximum free drug concentrations exceeded the cell killing IC₅₀ in SUDHL5 cells by over 100-fold.

We assessed the ability of TCIP3 to ablate cycling GC B cells in mice immunized with sheep red blood (Figure S16A). These cells are highly enriched for BCL6 expression and serve as a model for lymphomas of GC origin¹⁰³. TCIP3 induced a dose-dependent depletion of GC B cells relative to vehicle treatment, with 5 mg/kg twice daily (bid) intraperitoneal administration nearly eliminating this population (Figures 7A, 7B, S16B, and S16C). Peanut agglutinin (PNA) staining of spleens confirmed a loss of detectable GC B cells in mice treated with 5 mg/kg bid TCIP3 and marked reductions at other TCIP3 doses (Figures 7C and S16D). Total B cell numbers and weights remained unchanged (Figures 7D, 7E, S16E, and S16F). Examination (by H.V.) of heart, lungs, liver, spleen, and kidney after hematoxylin and eosin (H&E) staining revealed no overt organ toxicity or infiltration of inflammatory lymphocytes (Figure S16G). These data contrast with the phenotype of BCL6-knockout mice, which die of toxic hyperinflammation stemming from global BCL6 de-repression¹⁰⁴.

Figure 7. TCIP3 Ablates GC B Cells in Immunized Mice and Eliminates DLBCL Tumors in CDX Models.

(A) Representative gating and (B) quantification of GC B cells (GL7+, Fas+) from splenic B cells (B220+) in immunized mice treated intraperitoneally with the indicated doses of TCIP3 or vehicle. qd = once daily, bid = twice daily. P-values adjusted by Dunnett’s T3 multiple comparison’s test following ANOVA; ****adj. P < 0.0001, ***adj. P < 0.001, **adj. P < 0.01, *adj. P < 0.05 vs vehicle; n = 10 mice, mean ± s.d.

(C) Representative PNA staining of spleens from immunized mice corresponding to Figure S16D.

(D) Total B cells as percentage of live splenic B cells in immunized mice. n = 10 mice, mean ± s.d.

(E) Body weight change (%) over time of immunized mice. P-values adjusted using Sidak’s multiple-comparisons test after two-way repeated-measures ANOVA with Geisser-Greenhouse correction comparing treatment to vehicle within each timepoint; *P < 0.05 for only 5 mg/kg qd and bid TCIP3 treatment relative to vehicle at day 10. n = 10 mice, mean ± s.d.

(F) 5 representative bioluminescent images and (G) average tumor bioluminescence of SUDHL5 CDX mice treated with vehicle or 5 mg/kg TCIP3 bid. Lines represent the mean total flux at each time point; shaded regions represent ± s.e.m. n = 10 mice (including mice euthanized due to weight loss or to harvest organs over the course of the study). P-value computed by two-sided Welch’s unpaired t-test of area under the curve measurements of total flux through day 11 of the study.

To test the efficacy of TCIP3 treatment on killing BCL6-driven lymphomas in vivo, cell-derived xenograft (CDX) NOD scid gamma (NSG) mice bearing bioluminescent SUDHL5 tumors were administered 5 mg/kg TCIP3 or vehicle bid intraperitoneally over the course of 21 days. Treatment with TCIP3 resulted in complete or near-complete tumor clearance by day 11 of treatment (Figures 7F, 7G, and S17A). Two mice exhibited excessive weight loss and were removed from the study after tumor clearance on day 12, an effect likely addressable through dose optimization. Otherwise, TCIP3 exhibited no obvious signs of toxicity (Figures S17B and S17C).

Because BCL6-knockout mice die of a complex inflammatory reaction^104–106, we examined serum cytokine levels on day 11. Cytokine profiling revealed no significant differences in levels of inflammatory cytokines, including IL-1α, IL-1ß, IL-6, and the T helper 2 (Th2) cell lymphokines IL-4, IL-5, and IL-13 that have been suggested to contribute to the inflammatory phenotype in BCL6-knockout mice¹⁰⁶ (Figure S17D). These in vivo findings suggest that TCIP3 can selectively eliminate BCL6-positive lymphomas with minimal off-target toxicity.

DISCUSSION

Structural studies of KAT-TCIPs provide a general strategy to exploit chance interactions for the discovery of potent and selective CIPs. Induced proximity between unrelated proteins enables conformational sampling of opposing surfaces, creating opportunities for complementary contacts to emerge. The X-ray crystal structure of a KAT-TCIP in complex with p300 and BCL6 revealed fortuitous amino acid contacts comprising a composite surface¹⁴ independent of conformational plasticity¹⁰⁷ (Figure 1). Guided by these insights, we designed a library of KAT-TCIPs bearing rigid linkers to minimize entropic penalties and optimize favorable contacts, resulting in the identification of TCIP3, which demonstrated sub-nanomolar cell killing in lymphoma cells (Figure 2). Biophysical analyses revealed that TCIP3 is a bivalent molecular glue that induces cooperative binding between p300 and BCL6. This feature may contribute to its beneficial pharmacological profile^66,108–110 and confer in vivo stability that enabled potent ablation of germinal centers in immunized mice and the elimination of DLBCL tumors in CDX lymphoma models (Figure 7). These data suggest AI-driven docking strategies leveraging existing structures could accelerate the optimization of CIP-enabling compounds for therapeutic purposes.

The combinatorial nature of transcription led us to hypothesize that recruiting acetyltransferases to BCL6 would produce distinct global transcriptional outputs and cell fates relative to other activators. Indeed, TCIP3 induced unique gene expression changes and cell arrest and death kinetics relative to the previously developed TCIPs that recruit the elongation factors BRD4 and CDK9, consistent with the complex multistep nature of transcriptional activation. These findings highlight the modular pharmacology of TCIPs, suggesting compounds targeting the same TF but recruiting distinct transcriptional/epigenetic modifiers could be used sequentially to overcome mechanism-based resistance or confer context-dependent therapeutic advantages in different etiologic subtypes of B cell malignancies^{64,78,111,62,112}.

Our studies demonstrate that TCIP3 functions through a dominant gain-of-function mechanism to produce cell death. TCIP3 unidirectionally redirects only a fraction of p300/CBP to select BCL6 loci, and sub-stoichiometric protein engagement is sufficient to drive sub-nanomolar cell killing and rapid DLBCL tumor clearance in CDX models (Figure 7). KAT-TCIPs exploit the amplifying nature of transcriptional networks, whereby modest increases in pro-death transcripts propagate into irreversible and deterministic cell fate decisions. Consistent with this gain-of-function mechanism, the pharmacology of TCIP3 could not be phenocopied by either sequestration of p300/CBP or degradation of BCL6. Because TCIP3 does not meaningfully inhibit KATs, it may enable a superior therapeutic window. Reprogramming, rather than eliminating, p300/CBP, enables TCIP3 and KAT-TCIPs to engage multiple parallel cell death and proliferative-arrest networks. This strategy mimics natural cell death mechanisms that employ redundant pathways and operate for a human lifetime without the development of resistance^90,113,114.

LIMITATIONS OF THE STUDY

KAT-TCIPs recruit the acetyltransferases p300/CBP to the master transcriptional repressor BCL6 to activate cell death in DLBCL cells. We primarily assessed our lead KAT-CIP molecule TCIP3 in DLBCL cell lines with variable BCL6 expression, and the extent to which TCIP3’s pharmacology can be generalized across other DLBCL subtypes and other cancers remains to be explored. Additionally, we only incorporated two p300/CBP bromodomain-containing ligands and one BCL6-binding ligand into our KAT-TCIP designs. Other ligands may demonstrate different potencies and pharmacology. Although we investigated KAT-TCIP mediated acetylation of BCL6 and histone proteins, we did not comprehensively assess the possibility of other protein substrates. TCIP3 should be considered a tool molecule to manipulate cancer cell signaling and not a therapeutic candidate.

RESOURCE AVAILABILITY

LEAD CONTACT

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Gerald R. Crabtree (crabtree@stanford.edu).

MATERIALS AVAILABILITY

All compounds and unique reagents generated in this study are available upon request from the lead contact with a completed materials transfer agreement.

DATA AND CODE AVAILABILITY

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE¹¹⁵ partner repository with the dataset identifier PXD059919, which can be accessed using the Username: “reviewer_pxd059919@ebi.ac.uk” and Password: “6SYm8M9vSt59”. Genomic sequencing data has been deposited to GSE287542, GSE287543, and GSE313235. The X-ray co-crystal structure of MNN-02-155 in complex with p300-BD and BCL6-BTB has been deposited to the Protein Data Bank (PDB: 9MZA). Whole genome sequencing data for the cell line SUDHL5 has been deposited under the BioProject accession number PRJNA1392158 and can be accessed via the following link: https://urldefense.com/v3/__https://dataview.ncbi.nlm.nih.gov/object/PRJNA1392158?reviewer=5r488c897c6hlchcf6m4sn2okp__;!!G92We9drHetJ8EofZw!d6Qb-DbzJ_Ruoi07Zr76t8B6OAIjiFgXkLTC16ug0sWjCbMhyYi7PaysHSY5C8u6kto22Br1jjQE761Xpqu-JL0$. All other materials are available from the authors upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell Culture

Lymphoma and leukemia cells were grown in RPMI-1640 medium (ATCC 30–2001) supplemented with 10% fetal bovine serum (FBS) and 1% 100X Penicillin-Streptomycin (Gibco, 15140122) in incubators at 37°C with 5% carbon dioxide. OCILY19 cells were grown in MEM-α (with nucleosides, Thermo Fisher A1049001) supplemented with 10% fetal bovine serum (FBS) and 1% 100X Penicillin-Streptomycin (Gibco, 15140122) in incubators at 37°C with 5% carbon dioxide. K422 cells were obtained from Sigma (06101702). DB, SUDHL5, and RL cells were obtained from the American Tissue Culture Collection (ATCC). Raji cells, originally from ATCC, were a gift from J. Cochran's laboratory at Stanford University. OCILY19 cells, originally from DSMZ, and SUDHL4 cells, originally from ATTC, were a gift from D. Felsher’s laboratory at Stanford University. TOLEDO and K562 cells were originally obtained from ATCC and were kindly shared by A. Alizadeh's laboratory at Stanford University. Primary human tonsillar lymphocytes were isolated from two separate donors (male, 6 years old; male, 44 years old) from the laboratory of M. M. Davis under IRB protocol numbers IRB-60741 (adult tonsils) and IRB-30837 (pediatric tonsils) according to published procedures⁶⁰. BJ CRL-2522 human fibroblasts were obtained from ATCC and grown in DMEM media (ThermoFisher 11965118) supplemented with 10% fetal bovine serum (FBS) and 1% 100X Penicillin-Streptomycin (Gibco, 15140122) in incubators at 37°C with 5% carbon dioxide. Cells were routinely checked for mycoplasma and immediately checked upon suspicion. No cultures tested positive.

Mice

For the in vivo pharmacokinetic (PK) study, C57BL/6 male mice were housed individually in ventilated JAG 75 cages (IVC) with micro-isolator lids. HEPA-filtered air was supplied into each cage at a rate of 60 air exchanges per hour for the mice. The dark/light cycle was set for 8:00pm on-8:00pm off. The temperature was set for 72°F and was maintained within ± 2°F. The humidity was set to low/Hi 30–70%. There was a computerized system in place to control and/or monitor temperatures within the Animal Holding Room. Each animal room was equipped with a thermos-hygrometer that was monitored and recorded daily on the room log. All procedures related to PK studies were approved by the Scripps Florida Institutional Animal Care and Use Committee (IACUC) and the Scripps Vivarium is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. For the mouse GC experiments, 8-week old male C57BL/6J (Jackson Laboratory) mice were used. All mouse GC experiments were approved by Institutional Animal Care and Use Committee (IACUC) at MD Anderson Cancer Center with the IACUC number #00002290-RN00. For the SUDHL5 xenograft experiments, eight-week-old female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice were purchased from the Jackson Laboratory. All animal procedures related to this study were conducted in accordance with institutional guidelines and approved by the Institutional Animal Care and Use Committee (IACUC #00002290-RN01).

METHOD DETAILS

Cell Viability Measurements

Compound treatment for 72 hours:

Thirty thousand cells were seeded in 100 μl of media per well of a 96-well plate and treated with drug for the indicated times and doses. A resazurin-based indicator of cell health (PrestoBlue; P50200, Thermo Fisher) was added for 1.5 h at 37°C, at which point the fluorescence ratio at 560/590 nm was recorded (Tecan Spark). The background fluorescence was subtracted, and the signal was normalized to DMSO-treated cells. IC₅₀ measurements on cell lines were calculated using at least three biological replicates (separate cell passages). Fit of dose–response curves to data and statistical analysis was performed using GraphPad PRISM using the four-parameter log(inhibitor) vs response function.

Compound treatment for 7 days:

Fifteen thousand cells were seeded in 100 μl of media per well of a U-bottom 96-well plate and treated with digitally dispensed drugs (Tecan D300e) for the indicated times and doses. After 96 hours, plates were centrifuged for 750 rpm for 3 minutes, and old media was exchanged for 100 μl of fresh media. Drug treatment was then repeated at the indicated doses. 72 hours following media change, cells were transferred to 96 well flat-bottom opaque plates. Viability was measured by adding 25 μL Cell Titer-Glo reagent (Promega G7570) to each well and luminescence was measured (BMG Labtech Pherastar FS). The background luminescence was subtracted, and the signal was normalized to DMSO-treated cells. IC₅₀ measurements on cell lines were calculated using at least three biological replicates by separate cell passages. Fit of dose–response curves to data and statistical analysis was performed using GraphPad PRISM using the four-parameter log(inhibitor) vs response function.

Compound synergy treatment and Chou-Talalay analysis:

One thousand cells were seeded in 50 μL of media per well of a 384-well plate and treated with digitally dispensed drugs (Tecan D300e). Compounds were tested in a 10-point dose-response matrix using two-fold concentration steps, with a maximum concentration of 10 nM for TCIP3 and 25,000 nM for Palbociclib. Viability was measured as described above, “Compound treatment for 72 hours,” and background- and DMSO-normalized fluorescence values extracted. Synergy analysis proceeded as suggested in¹¹⁷: first, the individual compound data was fit to the 2-parameter median-effect model log(effect/(1-effect)) vs log(concentration) using the R linear regression fit function “lm” and the median dose, D_m, extracted (D_m = e^{(-intercept/slope)}); for TCIP3 and Palbociclib, the fit coefficient of determination R² was 0.95 and 0.93, respectively. Then, for synergy analysis, the Combination Index, CI, was calculated for the indicated combined drug effect size x (e.g., 50%, or ED₅₀) as CI = [TCIP3]/D_{x, TCIP3} + [Palbociclib]/D_{x, Palbociclib}, where [TCIP3] or [Palbociclib] were the empirical compound concentrations added in combination which produced the effect size; and D_{x, TCIP3} or D_{x, Palbociclib} are the concentrations required to achieve the effect size (e.g., 50%) if the compound was added alone, modeled from the individual compound model fits as D_x = D_m x (effect/(1-effect))^(1/slope). Uncertainty for the effective concentrations D_x was estimated by resampling fitted model parameters from their estimated covariance to generate a distribution of D_x at the specified effect level, and propagated to Combination Index calculations to derive a 95% confidence interval. The normalized isobologram (concentrations of compound as a fraction of effective dose, i.e., [TCIP3]/D_{x, TCIP3} vs. [Palbociclib]/D_{x, Palbociclib}) was plotted at the indicated effect size (e.g., 50% or ED₅₀) with combination points indicated that produce the viability effect within a tolerance of 0.05, colored by their CI: CI < 1 indicates synergy; CI > 1 indicates antagonism; CI = 1 indicates additivity.

Trypan Blue Cell Counting Assay

Thirty thousand cells were seeded into a 96-well plate and treated with digitally dispensed drugs (Tecan D300e) for the indicated times and doses. At either 24, 48, 72, or 96 hours, cells were transferred to a U-bottom well, centrifuged for 4 minutes at 500g, then aspirated. 10 μL phosphate-buffered saline (PBS) pH 7.4 and 10 μL Trypan Blue (Invitrogen T10282) were added to the well and mixed, at which point 10 μL of the mixture was transferred to a cell counting slide and the percentage of cells alive was recorded.

CDKN1B Knockout Proliferation Assay

100,000 cells were plated in 6 mL media and treated with either DMSO (0.1%) or TCIP3 (1 nM or 10 nM, at 0.1% DMSO final concentration). Growth was monitored for 9 days by live cell counting using Trypan Blue (Invitrogen T10282). Media containing 0.1% DMSO was supplemented on day 7 for only DMSO-treated cells to avoid overgrowth.

PRISM Cell Proliferation Assay

The PRISM cell proliferation assay was carried out as previously described¹¹⁸. Briefly, up to 859 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 with an 8-point dose curve starting at 10 μM with threefold dilutions in triplicate and incubated for 120 hours, then lysed. Each cell’s barcode was read out by mRNA-based Luminex detection as described previously⁶⁵ 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 area under the dose-response-curve (AUC), which is correlated with drug potency, was used as a metric of drug potency in a cell line, and correlated (Pearson’s) with dependency of the cell line to gene knockout⁹⁴.

Protein Expression and Purification

The bacterial expression vectors for 6xHis-p300-BD and 6xHis-CBP-BD used for isothermal calorimetry, crystallography, and TR-FRET assays were generous gifts from Nicola Burgess-Brown (p300: Addgene plasmid 74658; http://n2t.net/addgene:74658; RRID:Addgene_74658; CBP: Addgene plasmid 38977; http://n2t.net/addgene:38977; RRID:Addgene_38977). Mutations in p300 (Uniprot: Q09472) were introduced by site-directed mutagenesis (NEB E0554). The bacterial expression vector for BCL6-BTB used for isothermal calorimetry and crystallography (renamed pSG219C) was created as follows: codon-optimized coding sequences for human BCL6 (residues 5–129; Uniprot: P41182) were synthesized and cloned into pET-48b(+) (Novagen). The open reading frame codes for N-terminal Trx and 6xHis tags as well as a 3C cleavage site. The bacterial expression vector for BCL6-BTB used for TR-FRET assays (renamed pSG233) included a C-terminal GS linker and AviTag (GLNDIFEAQKIEWHE) used for biotinylation. Both vectors coded for the following BCL6 mutations: C8Q, C67R, C84N¹¹⁹. These enhance stability but do not affect the affinity for BI3812 or for SMRT.

Protein expression was carried out for 18 hours in Rosetta(DE3) cells (Novagen 70954) at 18°C before pelleting cells by centrifugation. Cell pellets were resuspended in ~2 mL per liter buffer D800 (20 mM HEPES, pH 7.5; 800 mM NaCl; 10 mM imidazole, pH 8.0; 2 mM beta-mercaptoethanol; 10 % glycerol (v:v)) supplemented with protease inhibitors (1 mM PMSF, 1 mM benzamidine, ~20 μg/ml pepstatin, aprotinin, and leupeptin), and stored at −80 °C.

To purify 6xHis-CBP-BD and 6xHis-p300-BD, cell pellets were thawed in warm water. All subsequent steps were carried out at 4 °C or on ice. Cells were lysed by sonication before centrifugation at 3,214 g for 1 hour. Clarified lysate was mixed with ~0.5 mL/L of culture cobalt resin (GoldBio) for one hour. Beads were washed by low-speed centrifugation and subsequently by gravity flow with ~25 column volumes of buffer D800 (followed by ~10 column volumes of buffer B50 (D800 but with 50 mM NaCl). Protein was eluted with 50 mL C50 (B50 with 400 mM imidazole), and the eluate was applied to a 5 mL anion exchange column (Q HP, Cytiva) and eluted via salt gradient (8 CV; B50 to D800). Peak fractions were concentrated by ultrafiltration before application to a 24 ml gel filtration column (S200 increase, Cytiva) charged with GF150 buffer (20 mM Tris-HCl, pH 8.5, 150 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine (TCEP)). Peak fractions were again concentrated by ultrafiltration, supplemented with 5% glycerol by volume (final), and aliquoted and frozen at −80 °C. Individual aliquots were thawed and stored for no more than 24 hours at 4 °C or on ice before use or disposal.

Purification of BCL6-BTB for crystallography and isothermal calorimetry followed the same procedure as 6xHis-CBP-BD and 6xHis-p300-BD with the following modifications: 6xHis-3C (homemade, produced with pET-NT*-HRV3CP, a kind gift from Gottfried Otting (Addgene plasmid 162795; http://n2t.net/addgene:162795; RRID:Addgene_162795)) was added to the eluate from the anion exchange column for 18 hours at 4 °C under slow rotation. Imidazole concentration was adjusted to 50 mM and the mixture was applied to a 5 mL nickel column (HisTrap FF Crude, Cytiva) and the flow-through captured. Flow-through was concentrated by ultrafiltration before application to a gel filtration column and peak fractions were concentrated and stored as written above.

Purification of biotinylated BCL6-BTB-AviTag for TR-FRET assays followed the same procedure as BCL6-BTB with the following modification: following cleavage of the tag with 6xHis-3C, the mixture was charged with 14xHis-BirA (homemade; produced with pTP264, a kind gift from Dirk Görlich (Addgene plasmid 149334; http://n2t.net/addgene:149334; RRID:Addgene_149334)) with 0.01 mM D-biotin (Sigma 2031), 10 mM MgCl₂, and 10 mM ATP and nutated for 1 hour 30 °C. The sample was applied to a 5 mL nickel column (HisTrap FF Crude, Cytiva) and the flow-through captured. Flow-through was concentrated by ultrafiltration before application to a gel filtration column and peak fractions were concentrated and stored as written above. Biotinylation efficiency was confirmed to be almost 100% by incubation of a small sample with streptavidin beads and monitoring of the flowthrough by SDS-PAGE.

TR-FRET Ternary Assay

10 μL reactions containing 10 nM 6x-His-p300-BD or 6x-His-CBP-BD, 200 nM Biotinylated-Avi-BCL6-BTB, 20 nM streptavidin-FITC (SA1001, Thermo), and 1:400 anti-6x-His-terbium (PerkinElmer 61HI2TLF) in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.1% BSA, 0.1% NP-40, and 1 mM TCEP were plated in low-volume 384 well plates. Drugs were digitally dispensed (Tecan D300e) into protein-containing wells, and the plate was allowed to incubate in the dark for 1 hour at room temperature. Emission at 490 nm (terbium) and 520 nm (FITC) was measured on a PHERAstar FS plate reader (BMG Labtech) upon excitation with 337 nm. The ratio of signal at 520 nm to 490 nm was calculated and normalized to DMSO-treated protein.

TR-FRET Binary Assay

10 μL reactions containing 25 nM 6x-His-p300-BD or 6x-His-CBP-BD, 50 nM MNN-06-112, 1:400 anti-6x-His-terbium (PerkinElmer 61HI2TLF), ± 10 μM BCL6-BTB in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.1% BSA, 0.1% NP-40, and 1 mM TCEP were plated in low-volume 384 well plates. Drugs were digitally dispensed (Tecan D300e) into protein-containing wells, and the plate was allowed to incubate in the dark for 1 hour at room temperature. Emission at 490 nm (terbium) and 520 nm (FITC) was measured on a PHERAstar FS plate reader (BMG Labtech) upon excitation with 337 nm. The ratio of signal at 520 nm to 490 nm was calculated and normalized to DMSO-treated protein. The K_{d, app} for each compound was calculated using the Cheng-Prusoff correction (K_{d, app} = IC₅₀/ (1 + [S]/K_{d, tracer}), where [S] is the concentration of protein (50 nM) and K_{d, tracer} is the binding affinity of MNN-06-112 to p300-BD-WT, determined through a titration of MNN-06-112 into p300-BD-WT and anti-His-Tb using the conditions described in this assay. The resulting curve was fitted using a one-site specific binding model in Prism 9, and adjusted for a two-site model due to the presence of anti-His-Tb using the equation K_d-two-site = K_d-one-site x (1 + √2), yielding 1849 nM.

Crystallography

In GF150 buffer, 1 mg of BCL6-BTB protein was incubated for 1 hour on ice with 2.5 equivalents of MNN-02-155 that was diluted 10-fold from a 10 mM DMSO stock in GF150 prior to mixing. The mixture was then purified by size-exclusion chromatography as above. The centermost peaks were collected, concentrated by ultrafiltration, and then incubated with 1.2 equivalents of 6x-His-p300-BD on ice for 1 hour. This mixture was purified by size-exclusion chromatography, which resulted in two distinct peaks. Fractions from the early-eluting peak were collected and concentrated to approximately 4.5 mg/mL in a 3k Amicon Ultra-4 Centrifugal filter. The sample was used immediately for crystallization by the sitting drop vapor diffusion method. Rectangular plate-like crystals grew within 48 hours in multiple conditions. Crystals was cryoprotected in a solution containing the crystallization solution supplemented with 25% glycerol before cryo-cooling by dipping the crystal in liquid nitrogen. Data was collected at Stanford SSRL experimental beamline 9–2. The best diffraction dataset came from crystals grown in 0.15 M DL-Malic acid pH 7.0, 20 % PEG 3,350. Data were integrated and scaled in the P212121 space group using Aimless and data quality analyzed using Ctruncate^120,121. We used data to a minimum Bragg spacing of 2.1 Å according to the CC1/2 cutoff suggested by Ctruncate. The L-test indicated twinning was not present in the crystal.

Initial phases were determined by molecular replacement using MOLREP¹²² as implemented in CCP4¹²³. We used as search models crystal structures of the BCL6 BTB domain (6CQ1) and the p300 bromodomain (5BT3) with the small-molecule ligands removed¹²⁴. The resolution cutoff for molecular placement was 3 Å, and we specified a multicopy search for hetero-multimers. This operation successfully placed individual copies of BCL6 and p300. Initial refinement using REFMAC¹²⁵ and analysis of the overall Matthews Coefficient indicated the likely presence of two copies each for BCL6 and p300. Thus, these initial coordinates were used as an input for PHASER¹²⁶, which placed two copies of the starting model in the unit cell. Refinement of this model using REFMAC converged rapidly and produced unambiguous extra density corresponding to the compound included in crystallography experiments. In later rounds of refinement, we added this compound, generating restraints using AceDRG¹²⁷. We also added water molecules in later refinement rounds. Manual model adjustments were done in Coot¹²⁸.

We note that, while the L-test indicated no twinning, the presence of multiple screw axes in the space group, as well as the existence of non-crystallographic symmetry within the biologically relevant protomer (2 BCL6:2 p300:2 KAT-TCIP) could potentially complicate data reduction and analysis. Specifically, we explored the possibility of translational non-crystallographic symmetry resulting in the averaging of non-equivalent protomers along a screw axis, which would result in higher-than-expected intensity statistics at low resolutions and would thus complicate L-tests for twinning. Unaccounted-for tNCS may also explain the relatively high Rfree value in the final refined model. However, the Phaser-MR solution was superior given the P212121 space group versus other orthorhombic possibilities (PHASER Refined LLG = 2749.7 versus 701.5 for the next best space group, P21221). Additionally, refinement produced features of the small molecule, which was not present in the molecular replacement models, indicative of high quality and unbiased maps. We note the presence of similar cases of potential unresolved tNCS in the literature with similar statistics. See¹²⁹ and references therein. The final refined model has Rwork/Rfree values of .217/.281. There are no Ramachandran outliers, and the overall clashscore is 9. Analysis of the protein-protein interface was conducted using PISA¹³⁰.

Isothermal Calorimetry

12 hours prior to performing ITC, frozen stocks of 6xHis-p300-BD and BCL6-BTB were dialyzed for two hours in ITC buffer at 4°C (50 mM HEPES pH 7.4, 100 μM TCEP, 150 mM NaCl), followed by dialysis in fresh ITC buffer overnight at 4°C to remove glycerol and equalize protein mixtures. Samples were centrifuged at 10,000g for 10 min to remove any precipitate. In all titrations, DMSO was added to protein mixtures to match the DMSO concentration of ligand dissolved in ITC buffer. For binary assays with TCIP3, dialyzed 6xHis-p300-BD or BCL6-BTB were titrated from the syringe into a cell containing TCIP3 dissolved in ITC buffer. For p300 titrations, 50 μM or 25 μM protein was titrated into 5 μM or 2.5 μM, respectively, of TCIP3. For BCL6 titrations, 70 μM or 40 μM protein was titrated into 7 μM or 5 μM, respectively, of TCIP3. For ternary titrations, 20× 6xHis-p300-BD was incubated with 1x TCIP3 in the cell to drive saturation of the binary complex, followed by a titration of 10x BCL6-BTB from the syringe, at 310 rpm stirring at 25°C. The following concentrations were used in each run: 15 μM BCL6-BTB, 30 μM p300, and 1.5 μM TCIP3; 18 μM BCL6-BTB, 36 μM 6xHis-p300-BD, and 1.8 μM TCIP3; or 10 μM BCL6-BTB, 20 μM 6xHis-p30-BD, and 1 μM TCIP3. Alternatively, 20x BCL6-BTB was incubated with 1x TCIP3 in the cell to drive saturation of the binary complex, followed by a titration of 10× 6xHis-p300-BD from the syringe, at 310 rpm stirring at 25°C. The following concentrations were used in each run: 30 μM BCL6, 15 μM 6xHis-p300-BD, and 1.5 μM TCIP3; 20 μM BCL6-BTB, 10 μM 6xHis-p300-BD, and 1 μM TCIP3; or 40 μM BCL6-BTB, 20 μM 6xHis-p300-BD, and 2 μM TCIP3. The first one or two injections and outliers from instrument noise were routinely excluded. Data were fit to a one-site model using MicroCal LLC Origin software. The following injection parameters were used for each run: (Total injection number: 35; Cell temp: 25 °C; Ref power: 10; Initial delay: 250 s; stirring speed: 310; Feedback mode: High; volume: 8 μL; Duration: 13.7 s; Filter period: 2; initial injection volume: 2 μL).

Percent Deviation in Free Energy Calculations:

Gibbs free energy changes (ΔG) for each binding permutation were calculated using the equation ΔG = -RTlnK, where R is the gas constant (1.987 cal/mol·K), T is the temperature in Kelvin (298.15 K), and K is the binding constant (1/M) obtained from the ITC fit. Percent deviation between the ΔG of the pathway that involves binding to p300 first (ΔG_p300) and the pathway that involves binding to BCL6 first (ΔG_BCL6) was calculated according to the equation (100*|ΔG_p300 - ΔG_BCL6|/((ΔG_p300 + ΔG_BCL6)/2)), where ΔG_p300 = ΔG of [p300-WT into TCIP3] + ΔG of [BCL6-WT into [TCIP3 + p300-WT]], and ΔG_BCL6 = ΔG of BCL6-WT + ΔG of [p300-WT into [TCIP3 + BCL6-WT]].

NanoBRET

BCL6-BTB NanoBRET:

The construct for NanoLuciferase (NanoLuc)-tagged BCL6-BTB was created as follows in a lentiviral construct: an N-terminal NanoLuc (subcloned from pNLSF-1, Promega N1351) was fused to the BTB domain of human BCL6 (Uniprot: P41182; aa1–129) with an internal GSG linker followed by a V5 tag. HEK293T cells were plated at a density of 6×10⁵ cells/mL in 2 mL of DMEM/well in a tissue culture treated 6-well plate and were allowed to incubate overnight at 37 °C. The next day, each well of cells was transfected with 2 ug of NanoLuc-BCL6-BTB (renamed pNSG218) using Lipofectamine 2000. The transfected cells were allowed to incubate overnight at 37 °C. The following day, the transfected cells were trypsizined, washed with PBS, counted with Trypan Blue, and brought to a final concentration of 1.25×10⁵ cells/mL in Fluorobrite DMEM (Gibco A1896701) supplemented with 10% FBS. 1 uM TNL-15 probe was added to the cells before plating 5000 cells per well in a volume of 40 uL per well in a 384-well plate. Control wells without the inclusion of TNL-15 probe were plated before its addition to the cells to be used as a negative control for data analysis. The plated cells were allowed to incubate overnight at 37 °C. The next day, the cells were treated with test compounds using a Tecan D300e Digital drug dispenser, normalized with DMSO, and allowed to incubate at 37 °C for 1 h. After 1 h, 5 uL of NanoBRET NanoGlo Substrate plus Extracellular NanoLuc Inhibitor was added to each well (Promega N1661) and mixed on an orbital shaker at 200 g for 30 seconds. Data was then obtained using a PheraStar FS plate reader (BMG Labtech) measuring luminescence with 520-BP and 450-BP filters. Wells were normalized to treatment with DMSO.

p300-BD and CBP-BD NanoBRET:

The constructs for NanoLuc-tagged p300-BD (SG234) and CBP-BD (SG235) were subcloned in pNLF-1 (Promega N1351) and created as follows: an N-terminal NanoLuc was fused to the bromodomains (BD) of human p300 (Uniprot: Q09472; aa1040–1161) or human CBP (Uniprot: Q92793; aa1081–1197) with an internal GSSG linker. HEK293T cells were plated at a density of 6×10⁵ cells/mL in 2 mL of DMEM/well in a tissue culture treated 6-well plate and were allowed to incubate overnight at 37 °C. The next day, each well of cells was transfected with 2 ug of NanoLuc-p300-BD or NanoLuc-CBP-BD using Lipofectamine 2000. The transfected cells were allowed to incubate overnight at 37 °C. The following day, the transfected cells were trypsizined, washed with PBS, counted with Trypan Blue, and brought to a final concentration of 1.25×10⁵ cells/mL in Fluorobrite DMEM supplemented with 10% FBS. 100 nM of MNN-05-112 probe was added to the cells before plating 5000 cells per well in a volume of 40 uL per well in a 384-well plate. Control wells without the inclusion of MNN-05-112 probe were plated before its addition to the cells as a negative control for data analysis. The plated cells were allowed to incubate overnight at 37 °C. The next day, the cells were treated with test compounds using a Tecan D300e Digital drug dispenser, normalized with DMSO, and allowed to incubate at 37 °C for 1 h. After 1 h, 5 uL of NanoBRET NanoGlo Substrate plus Extracellular NanoLuc Inhibitor was added to each well and mixed on an orbital shaker at 200 g for 30 seconds. Data was then obtained using a PheraStarFS plate reader (BMG Labtech) measuring luminescence with 520-BP and 450-BP filters. Wells were normalized to treatment with DMSO.

For all NanoBRET assays, a dose-response curve with 3 technical replicates was constructed for each compound, and corrected BRET ratios were calculated according to manufacturer assay protocol (Promega TM439). Data was fit using a standard four parameter log-logistic function using the R package drc or using GraphPad Prism.

Determination of compound occupancy

Calculations following a previously described procedure¹³¹ were performed to convert the displacement of the TNL-15 probe (measured by loss of NanoBRET signal) to percentage occupancy. The probe concentration was chosen to minimize shifting of apparent IC₅₀ but to also provide an adequate assay window; in general, it was slightly less than the EC₅₀ of the BRET response to the probe titrated alone. Occupancy percentage (%) was calculated according to Eq. 1 below:

$$Occupancy(\mathrm{\%})=\left[1-\frac{X-Z}{Y-Z}\right]*100$$

where $X$ is the BRET signal in the presence of both compound and probe; $Y$ is the BRET signal in the presence of only the probe; and $Z$ is the BRET signal background in the absence of probe or compound but only matched DMSO % (v:v). Occupancy values were plotted as a function of compound concentration and fit to a standard four-parameter log-logistic function using GraphPad Prism.

Annexin V staining

25,000 cells were plated in 200 μL of media in 96 well U-bottom plates and treated with compound. 72 hours later, the plate was spun at 500g for 4 minutes at 4°C, at which point cells were washed in 100 μL cold 2.5% FBS in PBS. Centrifugation and washing was repeated once more. Cells were then resuspended in 50 μL cold binding buffer (10 mM HEPES pH 7.4, 140 mM NaCl, 2.5 mM CaCl₂) and 2.5 μL FITC-Annexin-V (Biolegend 640922). A no-stain control was also included to draw gates. The plate was allowed to incubate for 15 minutes at room temperature, at which point 100 μL of 2.5% FBS in PBS was added to each well. The plate was gently agitated through shaking, and samples were analyzed by flow cytometry on a BD Accuri. At least 50k events were collected per sample.

Cell Cycle Analysis

2 million cells were plated in 4 mL of media in 6-well plates and treated with 10 nM of drug. After 22 hours, cells were dosed with 10 μM 5-ethynyl-2’-deoxyuridine (EdU) and allowed to incubate for an additional 2 hours, at which point cells were harvested and collected by centrifuging at 300g for 4 minutes at 4°C. Cells were washed once with 1% BSA in PBS. 1.1 million cells in 1% BSA/PBS were then transferred to flow cytometry polypropylene tubes and centrifuged at 300g for 4 minutes at 4°C. Supernatant was removed, and cells were fixed for 15 minutes in the dark at room temperature with 100 μL of 4% PFA in PBS, in which gentle flicking was used to resuspend cells. Cells were then washed with 1 mL of 1% BSA/PBS and permeabilized with 100 μL of 0.1% Triton-X for 30 minutes at room temperature in the dark. Cells were washed with 3 mL 1% BSA/PBS, then incubated in the dark for 30 minutes with a cocktail of AlexaFluor-488-Azide (438 μL PBS, 10 μL 100 mM CuSO₄, 2 μL 488-azide, and 50 μL of 1x EdU Reaction Buffer dissolved in water) (Invitrogen C10337). Cells were washed with 1 mL 1% BSA in PBS, then incubated with a reaction cocktail of 7-AAD/RNAseA for 30 minutes in the dark (500 μL 1% BSA/PBS, 2 μL 7-AAD (BD 559925), 5 μL RNAseA (Invitrogen 12091021)). Cells were washed with 1 mL 1% BSA/PBS centrifuged a final time, then resuspended in 500 μL 1% BSA/PBS. Cells were gently agitated by shaking, then analyzed through flow cytometry on a BD Accuri. At least 100k events were collected per sample. Gates were drawn from single-stain and no-stain controls.

TUNEL Analysis of DNA Fragmentation

1 million cells were plated in 6 mL of media in 6-well plates and treated with compound for indicated timepoints and doses. 1.2 million total cells were counted, washed in PBS, fixed in 4% paraformaldehyde/PBS at a concentration of 10M/mL for 15 minutes at room temperature in the dark, washed in PBS, and stored in 70% ethanol at −20°C until ready for processing (at least 24 hours). DNA breaks in fixed cells were labeled with bromolated deoxyuridine (Br-dUTP or BrdU) by incubation with deoxynucleotidyl transferase (TdT) for 60 min at 37°C, washed, and stained with FITC-labeled anti-BrdU antibodies for 30 min at room temperature in the dark (BD 556405). Cells were resuspended in 1% BSA/PBS with 1:100 RNaseA (Invitrogen 12091021) and co-stained with 7-AAD for 30 min at room temperature in the dark (BD 559925). The suspension was gently agitated by shaking and analyzed by flow cytometry within 1 hour (BD Accuri).

BCL6 Reporter Assay

K422 cells were lentivirally transduced with a construct containing the reporter. Description of the reporter construct has been published previously²². After selection, cells were plated and treated with indicated amount of compound for 24 hours. Cells were washed in 2.5% FBS/PBS, stained with 1:250 v/v of 7-AAD (BD 559925) to distinguish live from dead cells, and harvested for flow cytometry on a BD Accuri. Given the polyclonal population after transduction, the area under the curve of the histogram representing FITC signal across all live cells was calculated as an integrative measure of total GFP signal. A GFP-positive gate two standard deviations from the mean was drawn from non-transduced cells and the area past a constant threshold was calculated and normalized to the signal from cells treated with DMSO.

Generation of Overexpression and Knockout Cell Lines

NanoLuc-BCL6-BTB overexpression:

This construct (SG247) is the same as the one used for BCL6 NanoBRET studies above, with the addition of a 2x-SV40-NLS-V5 tag.

c-MYC overexpression:

An N-terminal fusion of 3xFLAG followed by a GGSGS linker fused to the coding sequence of human c-MYC (NM_002467.6) was subcloned into pCW57-MCS1-2A-MCS2 (a gift from Adam Karpf (Addgene plasmid 71782; http://n2t.net/addgene:71782; RRID:Addgene_71782)).

CDKN1B knockout:

CRISPR knockout constructs were generated using the lentiCRISPR v2 plasmid (Addgene plasmid # 52961; http://n2t.net/addgene:52961, a gift from Feng Zhang). 4 gRNAs designed using CRISPick¹³² targeting isoform-conserved exons in human CDKN1B were used to make 4 lentiCRISPR v2 plasmids. Target sequences for CDKN1B were: (1) AGTTCTACTACAGACCCCCG; (2) TTAGCCGGAGCCCCAATTAA; (3) GCCACTCGTACTTGCCCTCT; and (4) TGGACCACGA AGAGTTAACC. Two gRNAs were pooled for the non-targeting control; their sequences were (1) CAACGTCGCGAACGTCGTAT and (2) ACCACCGTTCGTACCGGTCG.

Lentivirus Production and Cell Infection:

Lentivirus was produced from lenti-x HEK293T cells (Clontech #632180) grown in a 15 cm plate via polyethylenimine transfection. Briefly, cells were transfected using 2^nd-generation packaging plasmids (psPAX2 and pMD2.G) and the respective constructs above, and media was replaced after 24 hours. For CDKN1B knockout cells, the 4 lentiCRISPR v2 plasmids were added at equimolar ratios to generate a pooled virus. The 2 non-targeting lentiCRISPR v2 plasmids were also added at equimolar ratios to generate a pooled non-targeting virus. Media containing virus 72 h after transfection was harvested, filtered, and used immediately or concentrated in PBS by ultracentrifugation (2 hours, 20,000 rpm, Beckman-Coulter Optima XE), flash-frozen, and stored at −80°C. Cells were lentivirally transduced with the overexpression construct by spinfection of virus (1000 g for 1 hour at 30°C) and selected using puromycin (1 μg/uL).

RNA Extraction, qPCR, and Sequencing Library Preparation

Cells were plated at 1M cells/mL and harvested in TRIsure (Bioline 38033). RNA was extracted using Direct-zol RNA MicroPrep columns (Zymo R2062) treated with DNAseI. cDNA was prepared (Meridian Bioscience BIO-65054) and used for qPCR (Meridian Bioscience BIO-94050) using an AppliedBiosystems QuantStudio 6Pro. Primer sequences were the following:

c-MYC fwd: CCTTCTCTCCGTCCTCGGAT;

c-MYC rev: CTTCTTGTTCCTCCTCAGAGTCG;

GAPDH fwd: GCCAGCCGAGCCACAT;

GAPDH rev: CTTTACCAGAGTTAAAAGCAGCCC.

For sequencing library preparation of SUDHL5 and K422 cells, polyA-tailed mRNA was enriched (NEB E7490) and prepared into paired-end libraries (NEB E7765) and indexed (NEB E7335). Library size distributions were confirmed using an Agilent Bioanalyzer and High Sensitivity DNA reagents (Agilent 5067) and concentrations determined by qPCR. Equimolar pooled libraries were sequenced on an Illumina NovaSeq with 2 × 150 bp cycles.

For RNA sequencing for RL, SUDHL4, and OCILY19 cells, 500,000 cells were harvested in 50 μL Zymo DNA/RNA Shield (R1100). For sequencing library preparation, 3’ end counting was used (Plasmidsaurus): polyA mRNA was captured, reverse transcribed, and barcoded with a unique molecular identifier; then, double-stranded cDNA was generated and tagmented; finally, libraries were amplified using Illumina index primers for sequencing.

Acetyl-lysine and BCL6 Immunoprecipitation for Immunoblotting

Cells were plated at 1M cells/mL and treated with compound or DMSO at indicated timepoints and doses. Cells were harvested on ice, counted, normalized by count, and washed 1X in PBS containing compound or DMSO. For only BCL6 immunoprecipitation, compound or DMSO was maintained in nuclear preparation buffers at identical concentration to cell treatment throughout. Nuclei were prepared by incubation in Buffer A (25 mM HEPES pH 7.5, 25 mM KCl, 0.05 mM EDTA, 1mM MgCl₂, 10% glycerol, 0.1% NP-40) supplemented with protease inhibitors (1 mM PMSF, ~20 mg/ml pepstatin, aprotinin, and leupeptin) at 4 °C for 7 minutes. Nuclear preparation was confirmed by Trypan blue staining and nuclei were pelleted by centrifugation for 5 min at 500g at 4 °C. Pelleted nuclei were resuspended in IP Buffer (25 mM HEPES pH 7.5, 150mM KCl, 0.05 mM EDTA, 1 mM MgCl₂, 10% glycerol, 0.1% NP-40) supplemented with protease inhibitors and 1 μL/250 units benzonase (Sigma E1014). Chromatin was removed by incubation with rotation for 30 minutes and nuclei sheared using a 27-gauge needle exactly five times. Insoluble material was pelleted by centrifugation at 21,000g for 10 min at 4 °C and the supernatant containing soluble nuclear protein preserved. Extracts were normalized by total protein concentration (Bradford) and identical amounts of total protein and concentrations were used for immunoprecipitation. 1 μg acetylated-lysine antibodies (Cell Signaling 9441), anti-BCL6 antibodies (Cell Signaling D65C10), or normal rabbit IgG (Cell Signaling 2729) and paramagnetic beads conjugated to Protein G (Thermo 10003D) were added to samples and incubated with rotation at 4 °C for 18 hours. Samples were washed five times with 1 mL IP Buffer supplemented with protease inhibitors and eluted by denaturation in 1X NuPage LDS/RIPA sample buffer (Thermo NP0008) supplemented with beta-mercaptoethanol by incubation at 95 °C for 5 minutes. For quantification, total signal was normalized to total BCL6 signal, and subsequently each immunoblot was normalized to DMSO treatment.

Western Blots

Cells were plated at 1M cells/mL and treated with drug at indicated timepoints and doses. Cells were harvested on ice in RIPA buffer (50mM Tris-HCl pH 8, 150mM NaCl, 1% NP-40, 0.1% DOC, 1% SDS, protease inhibitor cocktail (~20 mg/ml pepstatin, aprotinin, and leupeptin), 1mM DTT) and 1:200 benzonase (Sigma E1014) was added and incubated for 20 minutes. After 10 min centrifugation at 14,000g and 4 °C, the supernatant was collected and protein concentration was measured by Bradford. SDS-PAGE analysis was carried out in either 4–12% or 12% Bis-Tris PAGE gels (Thermo NW04120 and NW04127). Antibodies used for immunoblots were: BCL6 (1:1000 v:v, Cell Signaling D65C10), p300 (1:250, Santa Cruz F-4 sc-48343), CBP (1:1000, Cell Signaling D6C5), c-MYC (1:1000, Cell Signaling D84C12), Caspase-3 (1:1000, Cell Signaling 9662), GAPDH (1:2000, Santa Cruz 6C5 sc-32233), BBC3/PUMA (1:1000, Cell Signaling E2P7G), FOXO3 (1:1000, Cell Signaling 75D8), H3K27ac (1:1000, abcam ab4729), H2BK20ac (1:1000, Cell Signaling D709W), H3 (1:10,000, Cell Signaling 1B1B2), H2B (1:10,000, Cell Signaling D2H6), p27/CDKN1B (1:1000, Cell Signaling D69C12), FLAG (1:1000, Sigma F1804) and p53 (1:1000, Santa Cruz DO-1). ImageStudio (Licor) was used for blot imaging and quantification.

Quantitative Global Proteome Profiling via LC-MS/MS

Cell Treatment.

SUDHL5 cells (6M cells in 6 mL RPMI-1640 media supplemented with 10% FBS and 1x pen-strep) were treated with 0.1% DMSO, 10 nM TCIP3, 10 nM NEG1, 10 nM NEG2, or 250 nM dCBP1 (1000x stocks in DMSO for final 0.1% DMSO concentration) in 3 biologically independent replicates. The cells were washed twice with TBS (50 mM Tris, pH 8.5, and 150 mM NaCl) and then stored at −80 °C until use.

Lysis & Digestion.

Lysis was performed by first thermally denaturing the samples in residual wash buffer for 5 min at 95 °C. Fresh Lysis Buffer (8 M urea, 150 mM NaCl, and 100 mM HEPES, pH 8.0, in MS-grade water) was then added, and the lysates were homogenized using needle ultrasonication. Protein concentrations were then measured via a Bradford protein concentration assay (Bio-Rad 5000006) and then normalized to 30 μg of 2 mg/mL by diluting with Lysis Buffer. The samples were then reduced and alkylated simultaneously by adding final concentrations of freshly prepared 10 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP; Sigma Aldrich C4706) and 40 mM chloroacetamide (CAM; Sigma Aldrich C0267) in 100 mM HEPES, pH 8.0, and shaking for 30–60 min. The samples were then diluted with 120 μL of 20 mM HEPES, pH 8.0, and 1.29 mM CaCl₂ (Sigma Aldrich C4901) to a final concentration 1 mM. Digestion using 1:100 (w/w) protease to protein lysate with MS-grade Trypsin/Lys-C Protease Mix (Thermo A40009) was then performed by mixing overnight at 37 °C. The samples were then acidified by adding trifluoroacetic acid (TFA) until the sample reached pH ≤ 3, as confirmed by pH paper.

SPE Desalting.

Samples were then desalted using SOLAμ SPE HRP Peptide Desalting columns (Thermo 60209–001) with a Positive Pressure Manifold and all MS-grade reagents. The desalting cartridges were activated with 200 μL of acetonitrile and then equilibrated with two washes of 200 μL Wash Buffer (2% ACN + 0.2% TFA in water). After loading samples into the cartridges, the samples were washed three times with 200 μL Wash Buffer and then eluted using 100 μL 50% acetonitrile with 0.1% formic acid (FA) in water. The desalted peptides were then immediately dried on a centrifugal vacuum concentrator (Thermo SPD120-115). The samples were then reconstituted in 0.1% FA in water and then analyzed by LC-MS/MS (see below).

Generation of FLAG-BCL6 knock-in SUDHL5 cell lines

The procedure for the construction of FLAG-BCL6 knock-in cell lines was adapted from previously described procedures (Savic et al. 2015, Meadows et al. 2020). Plasmids expressing wildtype Cas9 under the control of chicken β-actin (CBA) and a human U6 promoter-driven were obtained from Addgene (plasmids #104046 and #104047; http://n2t.net/addgene:104046 and http://n2t.net/addgene:104047; RRID:Addgene_104046 and 104047, respectively; PX458_BCL6_iso1_1 and PX458_BCL6_iso1_2 were a gift from Eric Mendenhall & Richard M. Myers). gRNAs were designed to direct Cas9 nuclease activity near the stop codon of BCL6. The donor plasmid contained a 3x FLAG epitope tag, P2A linker, and a Neomycin resistance gene, flanked by regions homologous to the C-terminus of the BCL6 gene.

SUDHL5 cells were electroporated with donor and gRNA plasmids in a cuvette using the Amaxa^™ Cell Line Nucleofector^™ Kit V (Lonza, VVCA-1003) and Amaxa^™ Nucleofector^™ II transfection machine. 2 million cells were resuspended in Nucleofector solution containing supplement with 10 μg pooled plasmid (5 μg donor plasmid and 2.5 μg of each gRNA). This was done in biological triplicate (i.e. three separate electroporations performed), with cells from each electroporation maintained independently thereafter. Immediately after electroporation, cells were transferred to culture flask containing pre-warmed culture media. 24 hours post-transfection, cells were selected with G418 (Geneticin) at a concentration of 800 μg/mL. Selection was performed for 7 days. Cells were maintained under selection as a polyclonal pool for the generation of cell stocks. Subsequently, single clones were isolated by limiting dilution. To validate homologous recombination, genomic DNA from FLAG-BCL6 SUDHL5 cells was isolated using DNeasy Blood and Tissue kit (Qiagen, 69504). The C-terminus of the BCL6 gene was amplified using Q5 Hot Start High Fidelity 2X Master Mix (NEB M0494) with primers that bind internal to the 3x Flag tag and external to the homologous arms. The PCR product was purified using QIAquick PCR Purification Kit (Qiagen) and submitted for Sanger sequencing.

FLAG and p300 Immunoprecipitation

50 million FLAG-BCL6 knock-in SUDHL5 cells were treated compound or DMSO (0.1%) at indicated doses for 2 hours. Cells were lysed in NE10 buffer (20 mM HEPES (pH 7.5), 10 mM KCl, 1 mM MgCl2, 0.1% Triton X-100 (v/v), protease inhibitors (Roche), 15 mM ß-mercaptoethanol), dounced 15 times and pelleted 5 min at 500 g. Nuclei were washed in NE10 buffer and then digested with 250 units benzonase (Millipore) for 30 min rotating at 25°C. Nuclei were resuspended in NE150 buffer (NE10 supplemented with 150mM NaCl) and incubated for 20 min. Lysates were pelleted at 16,000 g for 20 min at 4°C and supernatants were immunoprecipitated by incubating with 2.5 μg FLAG M2 antibody (F1804, Millipore-Sigma) or 2.5 μg p300 (A300–358A, Bethyl Laboratories) antibody with Dynabeads Protein G (Thermo Fisher) overnight at 4°C. The IP fraction was recovered by magnetic separation followed by three washes with NE10 buffer containing 150mM–300mM NaCl but without Triton-X. For mass spectrometry analysis, the IP was then eluted from the beads with Ammonium Hydroxide, pH 11–12; 3% v/v.

Immunoprecipitation-Mass Spectrometry (IP-MS)

Immunoprecipitation was performed as described above for FLAG-BCL6 and p300 in three biologically independent replicates. Enriched proteins were eluted three times with 150 μL of 3% v/v ammonium hydroxide (pH 11–12) for 15 min with gentle agitation. The collected eluate was then dried using a centrifugal vacuum concentrator (Thermo SPD120–115). The samples were reconstituted in 50 μL of reduction-alkylation buffer containing 8 M urea, 10 mM TCEP, 40 mM chloroacetamide, 4.4 mM CaCl₂, and 100 mM HEPES, pH 8.0 and then incubated for at least 30 min while shaking. The samples were then diluted with 160 μL of 20 mM HEPES, pH 8.0, and then digested overnight at 37 °C with 500 ng of MS-grade Trypsin/Lys-C Protease Mix (50 ng/μL). The samples were then acidified by adding trifluoroacetic acid (TFA) until the sample reached pH ≤ 3, as confirmed by pH paper. Samples were then desalted using SPE desalting (described above).

LC-MS/MS diaPASEF Acquisition & Statistical Analysis

diaPASEF LC-MS/MS Analysis.

The reconstituted desalted peptides were then analyzed using a nanoElute 2 UHPLC (Bruker Daltonics, Bremen, Germany) coupled to a timsTOF HT (Bruker Daltonics, Bremen, Germany) via a CaptiveSpray nano-electrospray source. The peptides were separated in the UHPLC using an Aurora Ultimate nanoflow UHPLC column with CSI fitting (25 cm × 75 μm ID, 1.7 μm C18; IonOptics AUR3-25075C18-CSI) at a flow rate of 400 nL/min with column temperature maintained at 50 °C using Mobile Phase A (MPA; 3% acetonitrile + 0.1% FA in water) and Mobile Phase B (MPB; 0.1% FA in ACN). The 70 min gradient started at 6% MPB while increasing to 17% MPB at 40 min, 25% MPB at 55 min, 34% MPB at 64 min, 85% MPB at 65 min, and finishing at 92% MPB at 70 min.

The TIMS elution voltages were calibrated linearly with three points (Agilent ESI-L Tuning Mix Ions; 622, 922, 1,222 m/z) to determine the reduced ion mobility coefficients (1/K₀). diaPASEF was performed using the MS settings 100 m/z for Scan Begin and 1700 m/z for Scan End in positive mode, the TIMS settings 0.70 V·s/cm² for 1/K₀ start, 1.30 V·s/cm² for 1/K₀ end, ramp time of 120.0 ms, 100% duty cycle, ramp rate of 7.93 Hz, and the capillary voltage set to 1600 V. diaPASEF windows from mass range 226.8 Da to 1226.8 Da and mobility range 0.70 1/K₀ to 1.30 1/K₀ were designed to provide 25 Da windows covering doubly and triply charged peptides as confirmed by DDA-PASEF scans, whereas singly charged peptides were excluded from the acquisition due to their position in the m/z-ion mobility plane.

Raw data processing.

The raw diaPASEF files were processed using library-free analysis in FragPipe 22.0¹³³. DIA spectrum deconvolution was performed using diaTracer 1.1.5 with the following default settings: (i) “Delta Apex IM” to 0.01, (ii) “Delta Apex RT” to 3, (iii) “RF max” to 500, (iv) “Corr threshold” to 0.3, and (v) mass defect filter enabled with offset set to 0.1. The reviewed Homo sapiens protein sequence database was obtained from UniProt (07/13/2024; 20,468 entities) with decoys and common contaminants. In the MS Fragger 4.1 database search, the following settings were used: (i) initial precursor and fragment mass tolerances of 10 ppm and 20 ppm, respectively, (ii) enabled spectrum deisotoping, mass calibration, and parameter optimization, and (iii) isotope error set to “0/1/2”. For protein digestion, “stricttrypsin” for fully tryptic peptides was enabled with up to 1 missed cleavage, peptide length from 7 to 50, and peptide mass range from 500 to 5,000 Da. For modifications, methionine oxidation (2 max occurrences) and N-terminal acetylation (1 max occurrence) were set as variable modifications (maximum up to 3), while cysteine carbamidomethylation was set as a fixed modification. For validation, MSBooster (DIA-NN model) and Percolator were used for RT and MS/MS spectra prediction and PSM rescoring, while ProteinProphet (--maxppmdiff 2000000) was used for protein inference with FDR filtering (--picked --prot 0.01). The spectral library was then generated by EasyPQP 0.1.49 using default settings. Peptides were then quantified using DIA-NN¹³⁴ 1.9.1 with 0.1% FDR and QuantUMS high accuracy settings. The DIA-NN parquet report containing peptide quantification and scoring was then analyzed in R.

Statistical Analysis.

DIA-NN quantified peptides were then further filtered including (i) common contaminants and reverse sequences, (ii) 1% FDR filtering at the global.q.value (precursor Q value across all samples) and pg.q.value (protein group Q value in single injection), (iii) quantification of a given peptide in at least two replicates in one condition, and (iv) removal of singly charged peptides and non-proteotypic (unique) peptides. Protein intensities were then re-calculated using the MaxLFQ method¹³⁴ provided in the DIA-NN R package. Differential statistics was then performed using the DEqMS¹³⁵ R package, which performs a LIMMA-moderated t-test with an adjustment for number of detected peptides per protein, to determine the p-value, fold change, and Benjamini-Hochberg adjusted p-value. For the mean peptide intensity plots, the arithmetic mean was calculated for unique peptide precursors across conditions imputing a value of 1 unit for replicates where a peptide was not detected. For protein-level imputation in IP-MS, proteins were selected for imputation if (i) a fold change using DEqMS differential statistics could not be calculated and (ii) the protein was not detected in DMSO condition at all but detected in at least all but one replicate in the compound-treated condition. Selected proteins had imputed 15^th percentile MaxLFQ protein intensities for the DMSO condition and the subsequent fold changes between DMSO and compound conditions were recalculated.

RNA-seq Analysis

Raw reads were checked for quality using fastqc (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and trimmed from adapters using cutadapt¹³⁶ using parameters cutadapt -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA -b AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT --nextseq-trim=20 --minimum-length 1. Transcripts were quantified using kallisto¹³⁷ against the human Gencode v33 indexed transcriptome and annotations. Transcript isoforms were collapsed to genes and differential gene analysis was performed using DESeq2¹³⁸ using apeglm¹³⁹ to shrink fold changes. Pathway analyses were performed using Enrichr¹⁴⁰ on genes defined by significance cutoffs as detailed in figure legends. GSEA analysis was performed using fgsea¹⁴¹ on differential genes ranked by log₂(fold change) with an in-house dataset consisting of LymphDB (L. Staudt, NIH, https://lymphochip.nih.gov/signaturedb/), MSigDB pathways, and an internal dataset generated at MD Anderson of lymphoma-specific signaling. Correlation analyses with ChIP-seq used only genes with normalized (using DESeq2¹³⁸, relative log expression (RLE)) mean expression ≥ 32; this number was chosen by automatic independent filtering for outlier and low mean counts¹³⁸. Unbiased clustering analyses with TCIP1²² and CDK-TCIP1²³ were performed using pheatmap¹⁴² using parameters cutree_cols = 6, cutree_rows=2 including only significantly changed genes in any of the treatment conditions (|log₂(fold change)| ≥ 0.5 and adj. P ≤ 0.05; P-values computed by two-sided Wald test and adjusted by Benjamini-Hochberg). Reprocessed published datasets from the Sequence Read Archive (SRA) were the following: TCIP1 – SRX20228454, SRX20228455, SRX202284546, SRX20228457, SRX20228458, SRX20228459, SRX20228448, SRX20228449, SRX20228450; CDK-TCIP1 – SRX22117221, SRX22117222, SRX22117223, SRX22117224, SRX22117225, SRX22117226, SRX22117227, SRX22117228, SRX22117229).

ChIP-seq Experiment and Library Preparation

25–30 million cells were treated with TCIP3 or DMSO for indicated timepoints. Cells were washed in PBS containing TCIP3 or DMSO and crosslinked for 11 min in CiA Fix Buffer (50 mM HEPES pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 100 mM NaCl) with addition of formaldehyde to a final concentration of 1%. The crosslinking reaction was quenched by glycine added at 0.125 M final concentration. Crosslinked cells were centrifuged at 1,000 × g for 5 min. Nuclei were prepared by 10 min incubation of resuspended pellet in CiA NP-Rinse 1 buffer (50 mM HEPES pH 8.0, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% IPEGAL CA-630, 0.25% Triton X100) followed by wash in CiA NP-Rinse 2 buffer (10 mM Tris pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 200 mM NaCl). The pellet was resuspended in CiA Covaris Shearing Buffer (0.1% SDS,1 mM EDTA pH 8.0, 10 mM Tris HCl pH 8.0) with protease inhibitors and sonicated for 20 min with a Covaris E220 sonicator (Peak Power 140, Duty Factor 5.0, Cycles/Burst 200). The distribution of fragments was confirmed with D1000 Tapestation or by agarose gel electrophoresis. 20 μg of chromatin per ChIP was used with anti-H3K27ac antibodies (abcam ab4729) and anti-H2Bk20ac antibodies (Cell Signaling D709W), with 40 ng Drosophila chromatin (53083, ActiveMotif) spiked in. 15 ug of chromatin per ChIP was used with anti-p300 antibodies (Bethyl Labs, cat#: A300-358A) with 12 ng Drosophila chromatin (53083, ActiveMotif) spiked in. 40 μg of chromatin per ChIP on SUDHL5-FLAG-BCL6 cells was used with anti-FLAG antibodies (Sigma, cat#: F1804), with 32 ng Drosophila chromatin (53083, ActiveMotif) spiked in. After overnight incubation at 4 °C in IP buffer (50 mM HEPES pH 7.5, 300mM NaCl, 1mM EDTA, 1% Triton X100, 0.1% sodium deoxycholic acid salt (DOC), 0.1% SDS), IPs were washed twice with IP buffer, once with DOC buffer (10 mM Tris pH 8, 0.25 M LiCl, 0.5% IPEGAL CA-630, 0.5% sodium deoxycholic acid salt (DOC), 1mM EDTA), and once with 10 mM Tris/1 mM EDTA buffer (TE) pH 8. IPs and inputs were reverse-crosslinked in TE/0.5% SDS/0.5 μg/μL proteinase K for 55 °C /3 hours then 65 °C /18 hours, then DNA was purified using a PCR cleanup spin column (Takara #74609). Paired-end sequencing libraries were constructed using an NEBNext Ultra II DNA kit (E7645S) and indexed (NEB E7335). Library size distributions were confirmed using an Agilent Bioanalyzer and High Sensitivity DNA reagents (Agilent 5067) and concentrations determined by qPCR. Equimolar pooled libraries were sequenced on an Illumina NovaSeq XPlus with 2 × 150 bp cycles.

CUT&RUN Experiment and Library Preparation

CUT&RUN was performed as previously described^143,144 with minor modifications. Briefly, cells were collected and washed by PBS. After final wash, nuclei were isolated using nuclei extraction buffer (20mM HEPES pH 7.9, 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 1x cOmplete protease inhibitors (Roche 11836153001), 0.5mM spermidine) and then washed twice with wash buffer (20mM HEPES pH 7.5, 150mM NaCl, 0.5mM Spermidine, 1x cOmplete protease inhibitor). Five-hundred thousand nuclei were counted and incubated with activated Concanavalin A coated beads (EpiCypher 21–1411) and subsequently mixed with BCL6 antibodies (Cell Signaling D65C10) in antibody buffer (wash buffer + 0.01% digitonin + 2 mM EDTA) for overnight incubation on a nutator in a cold room. IgG controls (normal rabbit IgG, Cell Signaling 2729) were set in parallel for antibody enrichment and specificity validation. Nuclei were then washed three times with digitonin buffer (wash buffer + 0.01% digitonin) and incubated with pAG-MNase (EpiCypher 15–1116) for 10 minutes at room temperature. After 10 minutes of incubation, nuclei were washed three times again with digitonin buffer and resuspended with prechilled low-salt, high-Ca2+ buffer (20 mM HEPES pH 7.5, 0.5mM spermidine, 1x cOmplete protease inhibitor, 10mM CaCl₂) to initiate 1-hour digestion at 0°C using an ice block. The reaction was terminated by switching into stop buffer (340 mM NaCl, 20mM EDTA, 4 mM EGTA, 50 μg/mL RNaseA, 50 μg/mL glycogen) and incubating at 37°C for 10 minutes to release digested DNA from the nuclei. DNA was then purified with Ampure beads (Beckman A63882), subjected to two-sided size selection (0.5X to remove large fragment and 2X to recover desired nucleosomal DNA fragments). Sequencing libraries were generated with KAPA Hyper Prep Kits (Roche KK8502) using 14 cycles of PCR amplification. Libraries were validated on a Tapestation 4200 (Agilent G2991BA), quantified by Qubit High Sensitivity dsDNA Kit (Life Technologies Q32854), multiplexed and sequenced on a NovaSeq6000 using 2× 100 bp cycles.

ChIP-seq and CUT&RUN Analysis

The data quality was checked using fastq (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The raw reads were trimmed from adapters with cutadapt (parameters: -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA -AAGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT) and raw reads were aligned to hg38 human genome assembly using bowtie2 (parameters: --local --maxins 1000). Low quality reads, duplicated reads and reads with multiple alignments were removed using samtools¹⁴⁵ and Picard (https://broadinstitute.github.io/picard/). macs2¹⁴⁶ was used to reconstruct peaks with FDR cutoff of 0.05. Bedtools¹⁴⁷ was used to find a consensus set of peaks by merging peaks across multiple conditions (bedtools merge); peaks within 1000 bp (-d 1000) were merged for all ChIP-seq datasets. deepTools¹⁴⁸ was used to generate coverage densities across multiple experimental conditions (deeptools computeMatrix and deeptools plotProfile) and to generate bigwig files (deeptools bamCoverage), where reads mapping to ENCODE blacklist regions were excluded¹⁴⁹. All browser tracks and metaprofiles shown were calculated with sequence-depth-normalized, replicate-averaged (deeptools bigWigAverage) and, for ChIP-seq, input-subtracted data (deeptools bigWigCompare). Tracks were plotted using trackplot¹⁵⁰. Peak differential analysis and PCA analysis was performed using DESeq2¹³⁸ using apeglm¹³⁹ to shrink fold changes across relevant peaksets. Overlap analyses were performed using valr¹⁵¹. Enhancers were annotated using ROSE⁷ by stitching together H3K27ac peaks in untreated cells within 12.5 kb but excluding regions within 2 kb of a transcription start site unless within a larger H3K27ac domain. For analyses of differential changes at enhancers, only read counts across annotated enhancers were considered in the peakset input to DESeq2. Peaks and enhancers were assigned first to their nearest gene by simple linear distance and manually annotated as needed from literature. TF motif enrichment analysis around the summits of peaks reconstructed from FLAG-BCL6 ChIP-seq and BCL6 CUT&RUN datasets was performed on non-redundant summits using findMotifsGenome.pl from HOMER¹⁵² using parameters hg38 -size 200.

Cell Line Somatic Variant Analysis

SUDHL5 whole genome sequencing was performed (Novogene) on genomic DNA extracted from 2 million cells (MACHERY-NAGEL #740952.50). Whole genome sequencing data for K422, DB, and OCILY19 was obtained from the NCBI Sequence Read Archive (SRA) under the following accession numbers: K422 - SRP020237, OCILY19 - SRP020237, DB - SRP020237, and RAJI - SRP189367. Whole genome sequencing data for K562 was obtained from ENCODE, accession number #ENCBS806UYV. All genomes were sequenced to a minimum median depth of 22x. Somatic variants were called using the nf-core/sarek pipeline (v3.5.1)¹⁵³. Reads were aligned to GRCh38 using bwa-mem (0.7.18)¹⁵⁴ and duplicates were removed with GATK4 MarkDuplicates (4.5.0.0). Variants were called using Mutect2 (4.5.0.0) and annotated using SnpEff (GRCh38.105)¹⁵⁵ and ensemble VEP (113.0)¹⁵⁶. Synonymous variants were removed and analysis was restricted to variants with the PASS filter within the exonic regions of TP53, EP300, and CREBBP. Benchmarking of variant callsets against public data was performed using the Omics Somatic Mutations dataset from the 25Q3 public release of DepMap. For cell lines RL, SUDHL4, and TOLEDO, variant calls were taken from DepMap.

Linker Strain Energy Calculations:

Docking of KAT-TCIPs:

The crystal structure 9MZA containing BCL6, p300 and MNN-02-155 was processed by removing solvent and crystallographic molecules. A docking grid was defined using the center of mass of MNN-02-155 as the smallest box encompassing the entire ligand, extended up to 13 Å in all directions. The box dimensions along each axis were limited to a maximum of 35 Å. Docking of MNN-02-187, MNN-02-197, MNN-02-161, MNN-03-039, MNN-03-040, MNN-03-041, MNN-03-050, MNN-02-162, MNN-02-156, RCS-IJD-001, MNN-02-195, MNN-02-196, MNN-03-049, MNN-02-155, TCIP3, MNN-02-160, and MNN-03-037 was carried out using Deep Origin docking software. During docking, ligand warheads were constrained to the binding geometry observed in the reference ligand MNN-02-155, ensuring that each warhead adopted the correct orientation toward BCL6 and p300.

Molecular Dynamics (MD) simulations of ternary complex:

Each structure with a docked ligand was processed using the Deep Origin SystemPrep package that performed tautomers generation, kekulization, and protonation of the ligand and protein. The resulting complex was solvated in a rectangular periodic box with a 1 nm buffer of explicit water molecules on all sides. Solvation employed the TIP3P water model and a physiological ionic strength of 0.15 M was maintained by adding Na+ and Cl− ions to achieve both electroneutrality and the desired salt concentration. Simulations were carried out using the Deep Origin Simulation Suite. The protein was parameterized with the AMBER FF14SB force field, while the ligand was assigned parameters using the GAFF2 force field, with atomic charges derived via the AM1-BCC method. All production simulations were conducted in the NPT ensemble at 298.15 K and 1 bar, using a BAOAB Langevin integrator and a Monte Carlo barostat. The protocol included energy minimization, followed by a 2 ns NVT equilibration and a 5 ns NPT relaxation. One 300 ns production trajectory was generated per ligand-protein system, during which positional restraint was applied to the center of mass of the entire system to maintain it near the center of the simulation box. Additionally, rotational constraints were imposed on the BCL6 chains to preserve their orientation during the simulations. Convergence analyses indicated stable ligand poses throughout the simulation. Each trajectory was processed using CPPTRAJ (AmberTools¹⁵⁷) to center the ligand coordinates at the origin and apply autoimage to account for periodic boundary conditions. For conformational analysis, we retained only the ligand coordinates and selected the initial conformation obtained after energy minimization and representative 10 conformations from the last 10 ns of the production run. The representative conformations were selected by performing clustering with hierarchical Ward linkage method as implemented in TTClust¹⁵⁸ using root-mean-square deviation (RMSD) of heavy atoms.

Calculation of linker strain energy:

The minimized structure and the centroid structures from each cluster were, then, used for subsequent strain energy calculations. We performed strain energy calculations for the linker atoms of each ligand conformation, which was defined as ΔE_strain = E_bound - E_unbound, where E_bound and E_unbound represent the protein-bound and solution-phase global minimum conformations, respectively. The E_bound and E_unbound were obtained using quantum mechanical calculations using PySCF¹⁵⁹ at the ωB97M-V/def2-TZVPD level with the SMD implicit solvation model¹⁶⁰. For analysis, linker regions were extracted from complete ligand structures, retaining an acetyl capping group present in all studied ligands. Bound conformations from MD trajectories were subjected to single-point calculations without further optimization to preserve protein-induced distortions. Internal coordinate refinement is performed automatically within the Deep Origin software. Unbound conformations were identified through extensive conformational search on complete ligand structures using distance geometry methods followed by semi-empirical optimization and hierarchical filtering based on energy and structural similarity criteria. From the lowest-energy conformer, the linker region with acetyl cap was extracted to serve as the unbound reference state for ωB97M-V/def2-TZVPD/SMD calculations. Strain energies were evaluated for: (i) the energy-minimized conformation and (ii) 10 cluster representatives from the MD analysis. A population-weighted average was calculated as ⟨ΔE_strain⟩ = Σ(w_i × ΔE_strain,i). Finally, the difference between strain energies obtained for minimized and MD-averaged conformations from the end of the production run were taken, and then for each system, averaged over the two bound ligands to assess the strain that protein conformations impose on the ligand of interest. Because the crystallographic ternary complex contains two copies of the protein–ligand assembly, each MD system included two bound ligand warheads, which were analyzed and averaged.

Pharmacokinetic Measurements

Plasma protein binding:

Plasma protein binding was assessed using equilibrium dialysis. Samples were tested in triplicate using the RED Rapid Equilibrium Dialysis Device (Thermo Fisher Scientific 90006). Plasma samples containing the compound at an initial concentration of 2 μM were loaded into the plasma chamber, while phosphate-buffered saline (PBS) was added to the receiver chamber. The plate was sealed and incubated at 37°C with gentle shaking for 6 hours. A 25 μl sample was taken from the plasma and PBS chambers, which was then diluted with either blank PBS or plasma to achieve a 1:1 ratio of plasma: PBS for all samples. The concentration of the drug in the plasma and PBS chambers was determined by LC-MS/MS. The fraction of drug bound to plasma protein was calculated as ([plasma] – [PBS]) / [plasma].

In vivo pharmacokinetics (PK) study:

The PK study was performed in the Drug Metabolism and Pharmacokinetics (DMPK) core facility at Scripps Florida (https://www.scripps.edu/science-and-medicine/cores-and-services/dmpk-core/index.html). 5 mg/kg of TCIP3 was injected intraperitoneally into C57BL/6 male mice (n=3 in treatment and n=3 in vehicle condition) using a 25–29 gauge needle to deliver 10 μL/g body weight of a formulation of 0.5 mg/mL TCIP3 in 5% DMSO, 5% Tween-80, and 90% saline. The vehicle was the same formulation (5/5/90 DMSO/Tween-80/Saline). The formulation was checked to be a clear solution and after administration the animal was returned to its cage. Plasma levels of compound were measured at 0, 5, 10, 30, 60, 120, 240, 360, 480, and 1440 minutes after compound administration, processed by precipitation using acetonitrile, and analyzed by LC-MS/MS. PK parameters were calculated using the noncompartmental analysis tool of WinNonlin Enterprise software (version 6.3).

Microsomal Stability

1 μM of compound was incubated with 1 mg/mL hepatic mouse microsomes in 100 mM potassium phosphate buffer, pH 7.4. The reaction was initiated by adding NADPH (1 mM final concentration). Aliquots were removed at 0, 5, 10, 20, 40, and 60 minutes and added to acetonitrile (3X v:v) to quench the reaction and facilitate protein precipitation. NADPH dependence of the reaction was evaluated with -NADPH samples. At the completion of the assay, the samples were centrifuged through a Millipore Multiscreen Solvinert 0.45-micron low binding PTFE hydrophilic filter plate (MSRPN0450) and analyzed by LC-MS/MS.

Mouse Immunization and Germinal Center Response Analysis

8-week old male C57BL/6J mice were injected with 0.5 mL of a 2% sheep red blood cell suspension (Cocalico Biologicals) in PBS intraperitoneally to initiate a germinal center response. 10 mice were randomly grouped to be treated with either vehicle (8% DMSO, 5% Tween-80 and 87% saline) or TCIP3 at one of three dose administrations (2.5 mg/kg qd, 5 mg/kg qd, or 5 mg/kg bid, where qd = once per day, and bid = twice per day). All treatments were administered intraperitoneally and initiated three days following injection of sheep blood. Mice were monitored daily and weighed every other day. After 8 days of treatment, mice were euthanized. Spleens were homogenized by filtering through a 40 μm cell strainer followed by red blood cell lysis using ACK Lysing Buffer (A1049201, Thermo Fisher Scientific). Cells were resuspended in PBS with 0.5% BSA, performed with Fc blocking (422302, BioLegend), and labeled with anti-CD138 (142508, BioLegend), anti-GL7 (144614, BioLegend), anti-IgD (405723, BioLegend), anti-CD86 (105037, BioLegend), anti-IgM (406539, BioLegend), anti-B220 (103246, BioLegend), anti-Fas (557653, BD Biosciences), anti-CD184 (742171, BD Biosciences), anti-IgG (102742, BioLegend), and antiCD38 (747479, BD Biosciences) conjugated with fluorochromes for 30 minutes at room temperature. Samples were acquired on the Cytek Aurora flow cytometer and data was analyzed using FlowJo software.

SUDHL5 Xenograft Model

The SUDHL5 human DLBCL cell line was obtained from ATCC and genetically modified using a lentiviral vector to stably express luciferase and iRFP720. iRFP720-positive cells were maintained in RPMI-1640 supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37°C with 5% CO₂ and were routinely tested for mycoplasma. For implantation, cells were harvested during exponential growth, counted by trypan blue exclusion, and resuspended in sterile PBS or a 1:1 PBS:Matrigel (Corning) mixture at a concentration of 25 million cells per mL. For subcutaneous xenografts, 10 million cells in 400 μL of the PBS: Matrigel mixture were injected into the left flank of eight-week-old female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (Jackson Laboratory) using a 27-gauge needle. Tumor burden was monitored by in vivo imaging system (IVIS) bioluminescence imaging twice weekly, and body weight was recorded every Monday, Wednesday, and Friday. TCIP3 (5 mg/kg, twice daily) or vehicle control was administered according to the designated treatment schedule.

Serum Immunoprofiling

Serum samples were isolated at day 11 from three representative SUDHL5 CDX mice treated with either 5 mg/kg TCIP3 or vehicle bid. Immuno-assays were performed by the Immunoassay Team at the Stanford Human Immune Monitoring Center. 40 μL PBS was added to 20 μL of sample to dilute samples 3-fold. From this diluted sample, 25 μL was used for the mouse 48-plex kit (Life Technologies Corporation, Carlsbad, CA). Prepared samples were mixed with antibody-linked magnetic beads in a 96-well plate and incubated overnight with shaking at 4°C. Incubation steps (cold and room temperature) were performed on an orbital shaker at 500–600 rpm. Plates were washed twice with wash buffer in a BioTek ELx405 washer (BioTek Instruments, Winooski, VT), incubated for 1 hour at room temperature with biotinylated detection antibody, and washed again. Streptavidin-PE was added for 30 minutes with shaking. Plates were washed again and Reading Buffer was added to wells for detection in the Luminex FlexMap3D Instrument with a lower bound of 50 beads per sample per cytokine. Each sample was measured in duplicated wells. Custom quality control beads (Assay Chex; Radix BioSolutions, Georgetown, TX) were added to all wells. Wells with a bead count <50 were flagged, and data with a bead count <20 were analyzed with caution.

Chemical Synthesis: See Methods S1 and Data S1 (Document S1)

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using Graphpad Prism 10 or RStudio. Statistical significance was assessed through two-sided t-tests, Fisher’s LSD tests after ANOVA, two-sided Fisher’s exact tests, Dunnett’s T3 tests after ANOVA, two-sided Wald tests, Tukey’s tests after ANOVA, or Sidak’s multiple-comparisons tests after two-way repeated-measures ANOVA with a Geisser-Greenhouse correction. Where appropriate, P-values were adjusted for multiple comparisons by Benjamini-Hochberg. Statistical significance is denoted as follows: ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. Unless otherwise noted, all biological samples (n) were ≥ 3. Error bars represent standard deviation (s.d.) or standard error of the mean (s.e.m.).

Supplementary Material

1

Table S2. Whole Genome Sequencing Data with Variants in Cell Lines, related to Figure 2. Contains variants detected in cell lines with publicly available datasets or in cell lines sequenced in-house. Legend is on Sheet 1 of the Table.

2

Document S1: Figures S1–S17; Table S1; Data S1–S3; and Methods S1

3

Table S3. Enriched proteins after p300 IP-MS in SUDHL5 cells, FLAG IP-MS in FLAG-tagged BCL6 SUDHL5 cells, or after global proteome profiling after TCIP3, dCBP-1, NEG1, and NEG2 treatment in SUDHL5 cells; related to Figures 3 and 5.

4

Table S4. Full results of enrichment of histone acetylation peaks in public transcription factor ChIP-seq datasets in blood-lineage cells (ChIP-atlas), related to Figure 4. Contains all significantly enriched TFs in H3K27ac and H2BK20ac differential peaks. Legend is on Sheet 1 of the Table.

5

Video S1: Co-Crystal Structure of MNN-02-155 in complex with p300-BD and BCL6-BTB, related to Figure 1.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-p300 (F-4)	Santa Cruz Biotechnology	Cat# sc-48343; RRID:AB_628075
Rabbit anti-CBP (D6C5)	Cell Signaling Technology	Cat# 7389; RRID:AB_2616020
Rabbit anti-c-Myc (D84C12)	Cell Signaling Technology	Cat# 5605; RRID:AB_2798629
Rabbit anti-BCL6 (D65C10)	Cell Signaling Technology	Cat# 5650; RRID:AB_10949970
Rabbit anti-Caspase-3 (9662)	Cell Signaling Technology	Cat# 9662; RRID:AB_331439
Rabbit anti-PUMA (E2P7G)	Cell Signaling Technology	Cat# 98672; RRID:AB_3096180
Rabbit anti-FoxO3a (75D8)	Cell Signaling Technology	Cat# 2497; RRID:AB_2106669
Mouse anti-GAPDH (C65)	Santa Cruz Biotechnology	Cat# sc-32233; RRID:AB_627679
Mouse anti-Histone H3 (1B1B2)	Cell Signaling Technology	Cat# 14269; RRID:AB_2756816
Rabbit anti-Histone H2B (D2H6)	Cell Signaling Technology	Cat# 12364; RRID:AB_2714167
Rabbit anti-p27 Kip1 (D69C12)	Cell Signaling Technology	Cat# 3686; RRID:AB_2077850
Rabbit anti-H3K27ac (ab4729)	Abcam	Cat# ab4729; RRID:AB_2118291
Rabbit anti-H2BK20ac (D709W)	Cell Signaling Technology	Cat# 34156; RRID:AB_2799047
Mouse anti-FLAG M2	Millipore Sigma	Cat# F1804; RRID:AB_262044
Mouse anti-p53 (DO-1)	Santa Cruz Biotechnology	Cat# sc-126; RRID:AB_628082
Rabbit anti-acetylated lysine (9441)	Cell Signaling Technology	Cat# 9441; RRID:AB_331805
Rabbit anti-IgG (2729)	Cell Signaling Technology	Cat# 2729; RRID:AB_1031062
Rabbit anti-p300 (A300-358A)	Bethyl Laboratories	Cat# A300-358A; RRID:AB_185565
FITC-labeled anti-BrdU antibodies (BD Pharmingen™ APO-BRDU™ Kit)	BD Biosciences	Cat# 556405; RRID:AB_2869068
Brilliant Violet 421 ™ anti-mouse CD138 (Syndecan-1) Antibody	Biolegend	Cat# 142508; RRID:AB_11203544
Pacific Blue™ anti-mouse/human GL7 Antigen (T and B cell Activation Marker) Antibody	Biolegend	Cat# 144614; RRID:AB_2563292
Brilliant Violet 510™ anti-mouse IgD Antibody	Biolegend	Cat# 405723; RRID:AB_2562742
Brilliant Violet 605™ anti-mouse CD86 Antibody	Biolegend	Cat# 105037; RRID:AB_11204429
Brilliant Violet 711™ anti-mouse IgM Antibody	Biolegend	Cat# 406539; RRID:AB_2814386
Brilliant Violet 785™ anti-mouse/human CD45R/B220 Antibody	Biolegend	Cat# 103246; RRID:AB_2563256
Anti-Fas (BD Pharmingen™ PE-Cy™7 Hamster Anti-Mouse CD95)	BD Biosciences	Cat# 557653; RRID:AB_396768
BD OptiBuild™ BB700 Rat Anti-Mouse CD184	BD Biosciences	Cat# 742171; RRID:AB_2871409
Brilliant Violet 750 anti-mouse IgG1	BioLegend	Cat# 102742; N/A
Alexa Fluor 700 anti-mouse CD38	BD Biosciences	Cat# 747479; RRID: AB_2872155
Bacterial and virus strains
Rosetta(DE3) cells	Novagen	Cat# 70954
Chemicals, peptides, and recombinant proteins
Dynabeads Protein G	Thermo Fisher	Cat# 10003D
Benzonase Nuclease	Millipore Sigma	Cat# E1014-25KU
TCIP3	This paper	N/A
MNN-02-155	This paper	N/A
TNL-15	This paper	N/A
MNN-06-112	This paper	N/A
TRIsure	Bioline	Cat# BIO-38033
Chymostatin	Millipore Sigma	Cat# 230790
Leupeptin	Millipore Sigma	Cat# 108975
Pepstatin A	Millipore Sigma	Cat# 516481
Protease Inhibitor	Millipore Sigma	Cat# 4693132001
PhosSTOP phosphatase inhibitor	Millipore Sigma	Cat# 4906845001
Digitonin	Millipore Sigma	Cat# 300410-250MG
RPMI-1640 medium	ATCC	Cat# 30-2001
Fetal bovine serum	Thermo Fisher	Cat# A5669201
Penicillin-Streptomycin	Gibco	Cat# 15140122
DMEM Media	Thermo Fisher	Cat# 11965118
PrestoBlue	Thermo Fisher	Cat# P50200
Cell Titer-Glo Reagent	Promega	Cat# G7570
Palbociclib	Sigma Aldrich	Cat# PZ0383; Cas# 571190-30-2
pAG-MNase	EpiCypher	Cat# 15-116
Trypan Blue	Invitrogen	Cat# T10282
DMSO	ATCC	Cat# 4-X
Cobalt resin	GoldBio	Cat# H-310
Streptavidin-FITC	Thermo Fisher	Cat# SA1001
6x-His-terbium	PerkinElmner	Cat# 61HI2TLF
Lipofectamine 2000	Thermo Fisher	Cat# 11668027
Fluorobrite DMEM	Gibco	Cat# A1896701
Extracellular NanoLuc Inhibitor	Promega	Cat# N1661
FITC-Annexin-V	Biolegend	Cat# 640922
5-ethynyl-2’-deoxyuridine (EdU)	Invitrogen	Cat# C10337
Triton-X	Thermo Fisher	Cat# A16046.AE
AlexaFluor-488-Azide	Invitrogen	Cat# C10337
Copper Sulfate	Invitrogen	Cat# C10337
1x EdU Reaction Buffer	Invitrogen	Cat# C10337
7-AAD	BD Biosciences	Cat# 559925
RNAseA	Invitrogen	Cat# 12091021
BrdU	BD Biosciences	Cat# 556405
TdT	BD Biosciences	Cat# 556405
puromycin	Gibco	Cat# 11668027
DNA/RNA Shield	Zymo	Cat# R1100
1x NuPage LDS/RIPA sample buffer	Thermo	Cat# NP0008
dCBP-1	MedChemExpress	Cat# HY-134582; Cas# 2484739-25-3
A-485	MedChemExpress	Cat# HY-107455; Cas# 1889279-16-6
ARV-393	MedChemExpress	Cat# HY-158105; Cas# 2851885-95-3
Calcium Chloride	Sigma Aldrich	Cat# C4901
MS-grade Trypsin/Lys-C Protease Mix	Thermo Fisher	Cat# A40009
Cncanavalin A coated beads	EpiCypher	Cat# 21-1411s
Ampure beads	Beckman	Cat# A63882
Sheep red blood cell suspension	Cocalico Biologicals	N/A
Chloroacetamide	Sigma Aldrich	Cat# C0267
ACK Lysing Buffer	Thermo Fisher	Cat# A1049201
Human TruStain FcX™ (Fc Receptor Blocking Solution)	BioLegend	Cat# 422302
Bio-Rad Protein Assay Dye (Bradford)	Bio-Rad	Cat# 500-0006
MEM-α (with nucleosides)	Thermo Fisher	Cat# A1049001
D-Biotin	Sigma Aldrich	Cat# 2031
Drosophila Spike-in Chromatin	Active Motif	Cat# 53083
Matrigel	Corning	Cat# 356234
Critical commercial assays
Direct-zol RNA Microprep Kits	Zymo Research	Cat# R2060
Q5 Hot Start High Fidelity 2X Master Mix	New England Biolabs	Cat# M0494
Amaxa Cell Line Nucleofector Kit V	Lonza	Cat# VVCA-1003
Q5-Site Directed Mutagenesis Kit	New England Biolabs	Cat# E0554S
NEBNext® Poly(A) mRNA Magnetic Isolation Module	New England Biolabs	Cat# E7490
NEBNext® Ultra™ II Directional RNA Library Prep with Sample Purification Beads	New England Biolabs	Cat# E7765
NEBNext® Multiplex Oligos for Illumina® (Index Primers Set 1)	New England Biolabs	Cat# E7335
NEBNext® Ultra™ II DNA Library Prep Kit for Illumina®	New England Biolabs	Cat# E7645S
NEBNext® Multiplex Oligos for Illumina® (Index Primers Set 1)	New England Biolabs	Cat# E7335
Agilent DNA 1000 Kit	Agilent	Cat# 5067-1504
KAPA Hyper Prep Kits	Roche	Cat# KK8502
Qubit High Sensitivity dsDNA Kit	Life Technologies	Cat# Q32854
NucleoSpin Tissue, Mini kit for DNA from cells and tissue	MACHEREY-NAGEL	Cat# 740952.50
ProcartaPlex™ Mouse Immune Monitoring Panel, 48plex	Life Technologies	Cat# EPX480-20834-901
Deposited data
Genomic sequencing data	This paper	GEO: GSE287542, GSE287543, and GSE313235.
Mass spectrometry data	This paper	PRIDE: PXD046031
SUDHL5 Whole Genome Sequencing data	This paper	BioProject accession number PRJNA1392158
X-ray co-crystal structure of MNN-02-155 in complex with p300-BD and BCL6-BTB	This paper	PDB: 9MZA
Experimental models: Cell lines
OCILY19	DSMZ, Felsher Lab	Cat# ACC 528
K422	Millipore Sigma	Cat# 06101702
DB	ATCC	Cat# CRL-2289
SUDHL5	ATCC	Cat# CRL-2958
RL	ATCC	Cat# CRL-2261
RAJI	ATCC	Cat# CCL-86
SUDHL4	ATCC	Cat# CRL-2957
TOLEDO	ATCC	Cat# CRL-2631
K562	ATCC	Cat# CCL-243
BJ CRL-2522 Human Fibroblasts	ATCC	Cat# CRL-2522
Primary human tonsillar lymphocytes	Davis Lab	N/A
HEK293T cells	Millipore Sigma	Cat# 85120602
Lenti-x HEK293T	Clontech	Cat# 632180
Experimental models: Organisms/strains
C57BL/6J male mice	The Jackson Laboratory	Cat# 000664; RRID:IMSR_JAX:000664
Eight-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice	The Jackson Laboratory	Cat# 005557; RRID:IMSR_JAX:005557
Oligonucleotides
CDKN1B sgRNA 1: AGTTCTACTACAGACCCCCG	This paper	N/A
CDKN1B sgRNA 2: TTAGCCGGAGCCCCAATTAA	This paper	N/A
CDKN1B sgRNA 3: GCCACTCGTACTTGCCCTCT	This paper	N/A
CDKN1B sgRNA 4: TGGACCACGAAGAGTTAACC	This paper	N/A
Non-targeting sgRNA 1: CAACGTCGCGAACGTCGTAT	This paper	N/A
Non-targeting sgRNA 2: ACCACCGTTCGTACCGGTCG	This paper	N/A
c-MYC fwd: CCTTCTCTCCGTCCTCGGAT	This paper	N/A
c-MYC rev: CTTCTTGTTCCTCCTCAGAGTCG	This paper	N/A
GAPDH fwd: GCCAGCCGAGCCACAT	This paper	N/A
GAPDH rev: CTTTACCAGAGTTAAAAGCAGCCC	This paper	N/A
Recombinant DNA
6xHis-p300-BD	Gift from Nicola Burgess-Brown	Addgene# 74658
6xHis-CBP-BD	Gift from Nicola Burgess-Brown	Addgene# 38977
6xHis-p300-BD-Q1082A	This paper	N/A
6xHis-p300-BD-Q1082R	This paper	N/A
pSG219C	This paper	N/A
pSG233	This paper	N/A
NT*-HRV3CP	Abdelkader et al.161	Addgene# 162795
pTP264 (14xHis-BirA)	Pleiner et al.162	Addgene# 149334
pNSG218	This paper	N/A
SG234	This paper	N/A
SG235	This paper	N/A
BCL6 Reporter Construct	Gourisankar et al.22	N/A
SG247	This paper	N/A
3xFLAG-c-MYC	This paper	N/A
pCW57-MCS1-2A-MCS2	Barger et al.163	Addgene# 71782
lentiCRISPR v2 plasmid	Sanjana et al.164	Addgene# 52961
psPAX2	Gift from Didier Trono Lab	Addgene# 12260
pMD2.G	Gift from Didier Trono Lab	Addgene# 12259
PX458_BCL6_iso1_1	Savic et al.165	Addgene#104046
PX458_BCL6_iso1_2	Savic et al.165	Addgene# 104047
pNLSF-1	Promega	Cat# N1351
pET-48b(+)	Sigma - Novagen	Cat# 71462-3
Software and algorithms
Adobe Creative Cloud	Adobe	https://www.adobe.com/creativecloud.html
Rstudio	RStudio	https://www.rstudio.com/
Python 3.12	Python Programming Language	https://www.python.org/
Image Studio 6.0	LICORbio	https://www.licorbio.com/image-studio
Geneious Prime	Dotmatics	https://www.geneious.com/
GraphPad Prism 10	Dotmatics	https://www.graphpad.com/
fastQC	The Babraham Institute	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
deeptools	Ramirez et al.148	https://deeptools.readthedocs.io/en/latest/
DEseq2	Love et al.138	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Kallisto	Bray et al.137	https://kallisto.readthedocs.io/en/latest/
Enrichr	Chen et al.140	https://maayanlab.doud/Enrichr/
Cutadapt	Martin et al.136	https://cutadapt.readthedocs.io/en/stable/
Picard	Broad Institute	https://broadinstitute.github.io/picard/
macs2	Zhang et al.146	https://github.com/macs3-project/MACS
HOMER	Heinz et al.166	http://homer.ucsd.edu/homer/homer.ucsd.edu/homer/
bedtools	Quinlan et al.147	https://bedtools.readthedocs.io/en/latest/
samtools	Li et al.145	https://www.htslib.org/
CCP4	Collaborative Computational Project No. 4	https://www.ccp4.ac.uk/
REFMAC	MRC Laboratory of Molecular Biology	https://www2.mrc-lmb.cam.ac.uk/groups/murshudov/content/refmac/refmac.html
PHASER	CCP4	https://phenix-online.org/documentation/reference/phaser.html
Coot	MRC Laboratory of Molecular Biology	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
FragPipe 22.0	Nesvizhskii Lab	https://fragpipe.nesvilab.org/
diaTracer 1.1.5	Nesvizhskii Lab	https://diatracer.nesvilab.org/
bwa-mem (0.7.18)	Li et al.154	https://bio-bwa.sourceforge.net/
GATK4	Broad Institute	https://gatk.broadinstitute.org/hc/en-us
SnpEff (GRCh38.105)	Cingolani et al.155	https://pcingola.github.io/SnpEff/
VEP (113.0)	McLaren et al.156	https://useast.ensembl.org/info/docs/tools/vep/index.html
Nf-core/sarek pipeline (v3.5.1)	Ewels et al.153	https://nf-co.re/sarek/3.8.1/
bowtie2	Langmead and Salzberg167	https://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Apeglm	Bioconductor	https://bioconductor.org/packages/release/bioc/html/apeglm.html
Valr	Riemondy et al.168	https://rnabioco.github.io/valr/
ROSE	Loven et al.6; Whyte et al.7	https://bitbucket.org/young_computation/rose/src/master/
CPPTRAJ (AmberTools)	Cheatham Lab	https://amberhub.chpc.utah.edu/cpptraj/
TTClust	Tubiana et al.158	https://github.com/tubiana/TTClust
PySCF	Sun et al.169	https://pyscf.org/
Other
SOLAμ SPE HRP Peptide Desalting columns	Thermo Fisher	Cat# 60209-001
Multiscreen® 96 well Plate, PTFE membrane	Millipore Sigma	Cat# MSRPN0410
PCR cleanup spin column	Takara	Cat# 74609
Custom quality control beads Assay Chex	Radix BioSolutions	N/A
Centrifugal Vacuum Concentrator	Thermo Fisher	Cat# SPD120-115

HIGHLIGHTS:

• Small molecule-induced proximity between KATs and BCL6 elicits malignant cell death

• Fortuitous protein-protein contacts at complex interfaces enhance TCIP3 activity

• Fractional redirection of p300/CBP to BCL6 reprograms lymphoma epigenetic signaling

• Chemically induced proximity kills BCL6-driven lymphoma cells and tumors in vivo