SOX11 Inhibitors Are Cytotoxic in Mantle Cell Lymphoma

Mantle cell lymphoma (MCL) is a lethal subtype of B-cell lymphoma and represents approximately 6% of all non–Hodgkin lymphoma (NHL), the most common hematologic malignancy worldwide (1). Despite advances in chemotherapy and immunotherapy, patients with MCL have a median survival of 7–8 years, exhibit a continuous pattern of relapse and in clinical trials their survival curves have not yet reached a plateau. Upon relapse, targeted agents such as bortezomib, lenalidomide, and Bruton tyrosine kinase inhibitor (BTKi) ibrutinib have shown excellent clinical activity in patients with MCL but are not curative (2–4). Thus, advances in therapeutics for this disease are urgently needed (5). The pathogenesis of MCL is characterized by cell-cycle dysregulation and cyclin D1 (CCND1) overexpression caused by the t(11;14) (6, 7) translocation, leading to clonal expansion of malignant B lymphocytes. A subset of patients with CCND1-negative MCL have been identified with more than half of them harboring translocations that lead to overexpression of other CCND2/3 due to rearrangements with immunoglobulin genes (8). However, the CCND1/2/3 overexpression gene signatures in patients with MCL are indistinguishable from each other and mouse models overexpressing CCND1 do not recapitulate the MCL disease phenotype (9) suggesting their overexpression alone is not sufficient to drive the disease and that additional genetics lesions are necessary for MCL pathogenesis (10). Indeed, whole-exome sequencing of primary MCL patient samples has identified recurrent mutations in TP53, CDKN2A, CDKN2C, ATM, KMT2D, KMT2B, KMT2C, SMARCA4, and NOTCH isoforms among other genes in subsets of patients (11–15). Despite these findings, early transformation events that precipitate these genetic events are not well understood, and mouse models with these individual mutations do not recapitulate disease phenotype. The current management of MCL involves anti-CD20 antibodies and chemotherapy. The BTKi ibrutinib (16) and the BCL2 inhibitor (BCL2i) venetoclax (17, 18) are emerging MCL treatment options that produce high response rates but modest durations of response of 17.5 and 14 months, respectively, in this patient population. However, resistance to one or both of these agents frequently develops in MCL and new treatments, especially for chemorefractory MCL resistant to BTKi or BCL2i are the most urgent unmet need in clinical practice (19).

SOX11 belongs to the SOXC family of high mobility group (HMG) transcription factors (TF), which consists of SOX4, SOX11, and SOX12 (20). HMG TFs bind to DNA and facilitate conformational changes that allow binding of other TFs, usually leading to activation or repression of downstream genes. In early B cells, SOX4, a closely related SOX11 family member with identical target binding motifs, can inhibit differentiation via direct inhibition of WNT signaling (21). SOXC TFs have overlapping roles in central and peripheral nerve development, and it is likely that these factors play a role in B-cell development, as SOX11-null mice do not form spleens, and SOX4-deficient mice do not form B lymphocytes (22).

SOX11 protein is expressed in the majority (78%–93%) of patients with MCL and is specific for MCL as compared with other types of NHL (23–25). SOX11 is expressed even in the 5%–10% of patients with CCND1-negative MCL (6). In serial biopsies obtained from patients with MCL, SOX11 is found to be expressed specifically in premalignant lymph nodes, suggesting SOX11 upregulation is an early event in the malignant transformation of B lymphocytes in MCL (26). Microarray studies performed on primary MCL samples in our lab confirm SOX11 expression to be significantly higher in CD19⁺ tumor cells as compared with naïve B cells from healthy controls (27). Furthermore, SOX11 expression in MCL is associated with poor prognosis (23).

SOX11 depletion by RNAi in human MCL cell lines leads to reduced tumor growth in xenograft models (28). Additional studies corroborate the finding that SOX11 contributes to MCL pathogenesis in part via regulation of BLIMP1 downstream targets (29). Our group published a chromatin immunoprecipitation (ChIP-seq) analysis in the MCL cell line Granta-519 that identified direct binding targets of SOX11 within critical pathways involved in cell proliferation and cell-cycle control, including the WNT, protein kinase A, and TGFβ receptor signaling cascades (27). We developed the first murine models of SOX11 overexpression (Eμ-SOX11) which consistently develops an aberrant oligoclonal CD5⁺CD19⁺CD23⁻ B-cell proliferation that is identical to clinical MCL (30). In addition, RNA sequencing of CD5⁺CD19⁺ splenocytes from SOX11-overexpressing mice show increased B-cell receptor (BCR) signaling compared with wild-type controls. This is particularly relevant as constitutively active BCR signaling has been recently implicated in human MCL (31–33). Ibrutinib, a BCR signaling inhibitor (BTKi), shows significant therapeutic activity with a 60% overall response rate in patients with relapsed MCL (16). Still, the majority of patients relapse after ibrutinib treatment, and relapses are associated with acquired mutations in the binding pocket for ibrutinib (33). Because SOX11 is upstream of BTK, identification of SOX11 targets that are key intermediates for MCL pathogenesis may reveal new therapeutic targets for patients that relapse on ibrutinib and/or venetoclax (BCL2i), which is emerging as a promising therapy for MCL (17, 18).

SOX11 presents a potential new target in MCL, but TFs in general are considered “undruggable” because most have interaction surfaces that are devoid of pockets and grooves for the binding of small molecules. Nonetheless, because SOX proteins bind to the minor groove of DNA and derive part of their specificity through a severe distortion of the DNA, we hypothesized that this might lend to cavities at the protein–DNA interface that are amenable to targeting by small molecules. Indeed, putative binding pockets for small molecules could be identified at the protein–DNA interface of a SOX11–DNA homology model we built using the crystal structure of the SOX4–DNA-binding domain (DBD) as a template (34). Using a combination of in silico screening predictions and experimental validations, we report here first-in-class small-molecule inhibitors of SOX11 with on-target anti-MCL activity in BTKi-resistant cellular models in vitro and in patient-derived ex vivo models of ibrutinib resistance.

A total of 10 μg/mL (His)₆-SOX11-DBD was captured on the gold-coated sensor chip (SCR NiHC200M; XanTec bioanalytics) by injecting at 10 μL/minute for 180 seconds. Compounds were dissolved in 25 mmol/L Tris, 150 mmol/L NaCl, 5 mmol/L MgCl₂, 0.005% Tween-20, 2% DMSO, pH 7.4 and injected at 25 μL/minute for 425 seconds. The signals were measured at 25°C using Reichert 2SPR (Reichert Technologies) The association and dissociation times were set at 120 and 300 seconds, respectively. Three wash steps of 180 seconds at 25°C were included between each experimental run to ensure a steady baseline. The k_a, k_d, and K_d values were determined using Reichert TraceDrawer software. Average and SD were calculated from two or more separate experiments.

50 nmol/L (His)₆-mNeonGreen-SOX11-DBD was incubated with compounds at room temperature for 20 minutes in 96-well plates followed by 37.5 nmol/L AlexaFluor 568–conjugated double-stranded oligonucleotide (prepared as FAM-ds-oligo) for 20 minutes. Plates were placed in a Safire2 (Tecan Group) spectroscope with excitation at 488 (mNeonGreen; λ_ex/em = 506/517 nm) and emission was recorded from 512 to 620 nm (AlexaFluor-568; λ_ex/em = 578/603 nm). Control wells included buffer alone, compound alone, mNeonGreen-SOX11-DBD (donor) alone and AlexaFluor 568–dsDNA oligo (acceptor) alone and were used for background subtraction. Noninterference from compound intrinsic fluorescence signals was confirmed because <2% signal in excess of buffer alone control was observed for each compound. Fluorescence resonance energy transfer (FRET) ratios were calculated according to the TechNote TNPJ100.04 from ProZyme (https://prozyme.com/pages/tech-notes).

Statistical comparisons were made using two-tailed Student t test with Prism software (GraphPad Prism, RRID:SCR_002798) to calculate P values: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. The average and mean SD were calculated after experiments were done three times unless indicated.

All studies involving human samples were performed under Ohio State University (OSU) Institutional Review Board committee–approved protocols, through which informed written consent was obtained, and deidentified samples were utilized. Primary MCL cells were obtained from the peripheral blood of 4 patients with acquired ibrutinib-resistant MCL. Patient characteristics are presented in Supplementary Table S4. Cells were thawed and placed in RPMI1640 medium supplemented with 10% FBS and 1 mmol/L sodium pyruvate for 3 hours. Cells were then treated with either compound or DMSO control. All animal studies were carried out in accordance with the guidelines of the Institute for Animal Studies at OSU (Columbus, OH). The MCL patient-derived xenograft (PDX) mouse model was generated by engrafting peripheral blood mononuclear cells from Patient # 2 (Pt. 2) with acquired ibrutinib-resistant MCL intravenously into NSG mice and subsequently passaged five times. CD5⁺CD19⁺ MCL cells used for the described experiments were purified from the spleen of mice once removal criteria were met.

MCL cell lines JeKo-1, Mino, Z-138, and JVM-2 were purchased from ATCC. All cell lines were authenticated by short tandem repeat DNA profiling and routinely tested for Mycoplasma infection as per standard practice (MycoAlert, Lonza Bioscience). All MCL cell lines were maintained in RPMI1640 medium (Mod.) 1× with l-Glutamine (Corning) supplemented with 10% FBS (Gemini Bio-Products) and 1% penicillin-streptomycin 100× solution (Corning). The cells were grown at 37°C in a humidified incubator containing 5% CO₂.

Cell viability was determined by a fluorometric resazurin reduction method (CellTiter-Blue; Promega) following the manufacturer's instructions. A total of 100,000 cells in 100 μL of RPMI1640 medium were plated in 96-well flat bottom Falcon Polystyrene Microplates (Corning) and treated with compounds (8 replicates per condition). Cells were incubated for 72 hours. After incubation, 20 μL CellTiter-Blue reagent was added to each well and incubated for another 2 hours. Plates were put into a fluorescence plate reader that records fluorescence at 560/590 nm to get optical density (OD) values. The number of viable cells in each treated well was calculated, based on the linear least squares regression of the standard curve (OD vs. cell concentration). The viability of cells treated with compounds was normalized to the viability of cells treated with 0.2% DMSO. Cell counts were confirmed with trypan blue exclusion assay on the countess automated cell counter (Invitrogen) according to the manufacturer's specifications. The viability of cells treated with compounds for 48 hours was normalized to the viability of cells treated with 0.2% DMSO.

Gene expression profiles of tumor samples from patients newly diagnosed with MCL before any treatment and naïve B cells (NBC) from healthy donor specimens were used to evaluate mRNA expression levels of SOX C TFs SOX4, SOX11, and SOX12 (Gene Expression Omnibus accession number GSE70910; refs. 16, 35). Sample collection and laboratory studies were in compliance with Institutional Review Board and Helsinki protocols. CD19⁺ cells from 26 patients with MCL treated at the NIH were purified by magnetic bead sorting from peripheral blood or lymph node products before freezing to ensure greater than 90% purity. For controls, purified normal immunoglobulin D positive (IgD⁺) NBCs were obtained with the use of magnetic bead sorting from specimens from five healthy donors undergoing routine tonsillectomy (for non-neoplastic indications) at the Children's Hospital at Montefiore (Bronx, NY). Briefly, total RNA was extracted using the RNeasy kit (Qiagen) and profiled by Affymetrix Human Genome U133 Plus 2.0 arrays according to the manufacturer's instructions. CEL files were processed using Affymetrix Expression Console software and normalized by the robust multi-averaging method.

Cells were treated with compounds and harvested after 24 hours. Subsequent to harvesting, cells were prepped for analysis using the eBioscience Annexin V Apoptosis Detection Kit (Thermo Fisher Scientific) and following the manufacturer's instructions. Cells were analyzed using the BD LSRFortessa flow cytometer. Data analysis was done using Cytobank (15). For combination studies, isobolograms were made using the values for affected fraction of cells and combination index (CI) values were determined using CompuSyn software (http://www.biosoft.com/w/calcusyn.htm). The CI value correlates with the effect of combination treatment. A CI of <0.9 is considered synergistic, a CI of ≥ 0.9 or ≤ 1.1 is considered additive, and a CI of >1.1 is considered antagonistic.

After treatment, cells were stained with 5 μL FITC-Annexin V antibody (BD Biosciences) and 5 μL propidium iodide solution (BD Biosciences) for 15 minutes at room temperature in the provided binding buffer before they were analyzed on a BD LSRFortessa flow cytometer.

Cells were treated with compounds and harvested after 24 hours. Cells were washed and fixed by using the eBioscience Foxp3/Transcription Factor Fixation/Permeabilization Concentrate and Diluent kit (Thermo Fisher Scientific) and following the manufacturer's instructions. Fixed cells were stained with Alexa Fluor 647 Mouse Anti-BTK (pY223)/Itk (pY180) (BD Biosciences) and analyzed using the BD LSRFortessa flow cytometer. Data analysis was done using Cytobank (15).

A total of 1 million cells/mL were treated with compounds and incubated for 24 hours. RNA was extracted using the RNeasy Plus Mini Kit (Qiagen) according to the manufacturer's protocols. cDNA was prepared using SuperScript VILO Master Mix (Thermo Fisher Scientific) and detected by SsoFast EvaGreen Supermix (Bio-Rad Laboratories) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Gene expression was normalized to hypoxanthine phosphoribosyltransferase (HPRT) and expressed relative to cells treated with 0.2% DMSO using the 2^–ΔΔCT formula. Thermal cycler conditions were: initial step of 30 seconds at 95°C followed by 40 cycles of 5 seconds at 95°C (denature) and 5 seconds at 60°C (anneal/extend) followed by 5 minutes at 55°C to 95°C in increments of 0.5°C.

• HPRT Forward Primer: 5′-AAAGGACCCCACGAAGTGTT-3′

• HPRT Reverse Primer: 5′-TCAAGGGCATATCCTACAACAA-3′

• SMAD3 Forward Primer: 5′-TGGACGCAGGTTCTCCAAAC-3′

• SMAD3 Reverse Primer: 5′-CCGGCTCGCAGTAGGTAAC-3′

• PAX5 Forward Primer: 5′-GAGCGGGTGTGTGACAATGA-3′

• PAX5 Reverse Primer: 5′-GCACCGGAGACTCCTGAATAC-3′

• PPP3CA Forward Primer: 5′-CCAAGTCACCGGCTTACAG-3′

• PPP3CA Reverse Primer: 5′-CCTCCTTCATAAGATGCGCCTT-3′

For whole-cell extracts, cells were lysed in cold RIPA buffer containing cOmplete, Mini Protease Inhibitor Cocktail and phosphatase inhibitors (Sigma-Aldrich) and cell lysates were clarified by centrifugation. Proteins were analyzed by immunoblot using standard procedures. Primary antibodies to the following proteins were used: Actin Antibody (Santa Cruz Biotechnology, catalog no. sc-1615, RRID:AB_630835) and Anti-SOX11 antibody (Atlas Antibodies, catalog no. AMAb90502, RRID:AB_2665568). For primary cells, antibodies to the following proteins were used: SOX11 (Abcam); BTK, phospho BTK (Tyr223), and GAPDH (Cell Signaling Technology). Blot patterns were analyzed using ImageJ software (16), providing a quantitative measure of protein expression (ImageJ, RRID:SCR_003070).

For original data, please contact [email protected].

The SOXC family consists of three members, SOX4, SOX11, and SOX12, with high sequence and structural homology. To validate SOX11 as a target for development of small-molecule inhibitors, we used microarray analysis to examine the expression levels of SOX4, SOX11, and SOX12 in naïve B cells from five healthy donors and 26 patients with MCL (Fig. 1). While we observed similar baseline expression levels of SOX4 and SOX12 in both healthy and MCL-derived B cells, the mRNA expression of SOX11 was specifically increased approximately 6-fold in MCL patient samples. These findings are consistent with previous IHC studies in patients with MCL (23, 36–38).

Previous ChIP-seq experiments suggest that SOX11 mediates the overexpression of several oncogenes by binding directly to their DNA regulatory elements (27). SOX proteins, including SOX11 recognize and bind to the cognate DNA sequence [∼4 (TTGT)] (34) via a small positively charged DBD. Amino acids from the SOX-DBD insert into the minor groove of the DNA recognition sequence, resulting in sharp bending of the DNA, while the conformation of the protein remains largely unchanged (39). We reasoned that small molecules that bind to the DNA interacting surface of SOX11 could have the potential to preclude DNA binding.

To enable in silico screening of small molecules that bind to the SOX11-DBD, we derived a homology model based on the closest TF with an available crystal structure, that is, the SOX4-DBD (ref. 34; see details in the Materials and Methods section). There are only four different residues between SOX4 and SOX11 in this domain: Q71, L99, D104, and Q110 in murine SOX4 correspond to K61, M89, E94, and R100 in human SOX11, respectively.

The first in silico screening at this SOX11-DBD homology model using the NCGC pharmaceutical collection (http://tripod.nih.gov/npc/) dataset (40) prioritized 67 compounds for experimental testing (see Materials and Methods section for details). The fluorescence anisotropy (FA) assay used to assess inhibition of SOX11–DNA interaction in the presence of 20 μmol/L of each compound identified Flavitan as a potential inhibitor (Supplementary Fig. S1A) because of its > 50% inhibition of DNA binding (Supplementary Fig. S1B). However, this molecule was not studied further because of its high polarity, high molecular weight, and inactivity in MCL cell lines. Instead, its binding pose within the SOX11-DBD was used as a reference to prioritize additional hits from virtual screening of approximately 10 million ready-to-dock molecules from the ZINC version 12 Drugs Now subset (see virtual screening workflow in Supplementary Fig. S2; refer to Materials and Methods for details). Specifically, results of a structural interaction fingerprint (SIFt; ref. 41) assessment of the binding pose of Flavitan were compared with SIFt results obtained for the identified 112 top-scoring ligands from the aforementioned virtual screening using SIFt Tanimoto similarity. A total of 31 ligands with a SIFt Tanimoto similarity ≥ 0.4 with the SIFt assessment of Flavitan were selected for experimental analysis, but only five of them (Cpd A–E) were purchased.

Next, we tested the effect of compounds A–E on the viability of SOX11-expressing Z-138 MCL cell line versus the SOX11-negative (SOX11⁻) JVM-2 MCL cell line (Supplementary Fig. S3). Of the five compounds assayed, compound E displayed selective growth inhibition of Z-138 indicating that compound E may be exerting its cellular activity by binding to SOX11. On the basis of these results, we identified structural analogs with at least 80% similarity to compound E by querying the ZINC database. Of these, we procured two compounds that were available for purchase which we refer to as compound R and compound T (Fig. 2).

We used surface plasmon resonance (SPR) to determine whether compound E, compound R, and compound T (referred to as SOX11i) bind directly to (His)₆-SOX11-DBD immobilized on a Ni-NTA gold chip. As can be seen in the sensograms in Fig. 3A, all three compounds were found to bind SOX11-DBD with micromolar affinities (Supplementary Table S1). Compound N, an inactive compound from our screen, was also assayed and did not associate with SOX11-DBD while Sm4, a pan-SOXi small molecule (42) and a duplex DNA containing the SOX11 binding motif displayed expected affinities (Supplementary Table S1). To evaluate whether the binding of these compounds to SOX11-DBD deterred the binding of DNA, we sought to employ the FA assay (described above and Materials and Methods) which measures FA at fluorescein excitation and emission. Unexpectedly, the intrinsic fluorescence of the compounds complicated data analysis. These compounds have spectral properties over a broad emission range and hence were unsuitable to be evaluated by the FA assay. As an alternative, we developed a FRET-based assay in which a duplex DNA carrying the SOX11 binding sequence is labeled with the fluorophore Alexa Fluor 568 on the 5′-end of one strand and assessed against the SOX11-DBD fused to an N-terminal (His)₆-mNeonGreen (mNG) fluorescent protein. Binding of SOX11-DBD to the DNA brings the fluorophores closer due to DNA bending, leading to a concomitant increase in FRET (Fig. 3B). Small molecules that disrupt SOX11-DBD and DNA binding are expected to reduce FRET efficiency. We observed biochemical inhibition of the SOX11-DNA interaction by compounds E, R, and T (Fig. 3C). Compound E at 20 μmol/L caused the most inhibition of the SOX11-DBD:DNA interaction with unlabeled DNA used as a competitive positive control. To confirm the inhibition of SOX11-DBD:DNA interaction by SOX11i, we employed an orthogonal electrophoretic mobility shift assay (EMSA). For all three SOX11i, we observed a dose-dependent reduction in the mobility shift of DNA mediated by incubation with SOX11-DBD (Supplementary Fig. S4) with compound E showing the lowest IC₅₀ value.

To ensure the cellular growth inhibition and biochemical activity, we observed was not the result of SOX11i binding directly to DNA we employed a topoisomerase-based DNA unwinding assay that detects small-molecule intercalators of DNA. Ibrutinib was used a negative control and the established DNA intercalator, ethidium bromide as a positive control. We were able to confirm that all three SOX11i and ibrutinib do not directly bind DNA at 20 μmol/L; that is, they did not interfere with the DNA unwinding activity of topoisomerase I whereas ethidium bromide completely inhibited its ability to relax supercoiled plasmid DNA (Supplementary Fig. S5).

We next expanded our single-dose observation of SOX11-specific growth inhibition of Z-138 cells by compound E by conducting two different dose–response assays (0–40 μmol/L) in three SOX11-positive (SOX⁺) MCL cell lines (JeKo-1, Mino, and Z-138) versus the SOX11⁻ JVM-2 MCL cell line (see Fig. 4A for SOX11 status).

We used the metabolic CellTiter-Blue assay to determine viability of these MCL cell lines upon treatment with SOX11i for 72 hours. Both compound E and compound R had robust growth inhibitory activity in SOX11⁺ cells (IC₅₀: 12–16 μmol/L) as compared with SOX11 JVM-2 cells (IC_50: 30–32 μmol/L) with a 2.5- to 2.7-fold selectivity. Compound T had less effect in SOX11⁺ cells (IC₅₀: 14–34 μmol/L); however, it was still selective as IC₅₀ was not reached in JVM-2 cells (Fig. 4B and C; Supplementary Table S2).

We further tested the SOX11-dependent growth inhibitory activity of SOX11i by siRNA-mediated knockdown of SOX11 in SOX11⁺ JeKo-1 cells. An increase in survival of si-SOX11 cells compared with control cells treated with 5 μmol/L compound R confirmed its SOX11 dependency and on-target cytotoxicity (Supplementary Fig. S6). In addition, we treated PBMCs from two healthy donors with compound R for 24, 48, and 72 hours and found viability profiles matching those for SOX11⁻ JVM-2 cells (Supplementary Fig. S7).

The SOX11-specific growth inhibition exerted by these SOX11i could be the result of cytotoxic cell death. We performed flow cytometric analysis by staining SOX11i-treated MCL cells with Annexin V and 7-AAD to directly assay and quantify early and late apoptotic cell death. As can be seen in Fig. 4D and E and Supplementary Table S3, compound E had an apoptotic IC₅₀ value of 15 μmol/L in JeKo-1, 12 μmol/L in Z-138, and 12 μmol/L in Mino cells whereas in JVM-2 IC₅₀ was not reached. Compound T had an IC₅₀ of 14 μmol/L in JeKo-1, 19 μmol/L in Z-138, and 33 μmol/L in Mino cells while IC₅₀ was not reached in JVM-2 cells. Compound R was relatively more active with IC₅₀ values of 10 μmol/L in JeKo-1, 12 μmol/L in Z-138, and 11 μmol/L in Mino cells. The IC₅₀ value in JVM-2 was 35 μmol/L, showing SOX11 selectivity of similar magnitude to growth inhibitory assays.

One of the characteristics of MCL is increased BCR signaling. Eμ-SOX11 transgenic mice overexpressing SOX11 specifically in B cells have increased activating tyrosine phosphorylation of Bruton tyrosine kinase (pBTK-Y223) and phospholipase C γ (pPLCγ-Y759; ref. 30). Upon observing SOX11-specific anti-MCL cytotoxicity, we tested whether the SOX11i affect pBTK-Y223 levels in SOX11-expressing MCL cell lines, JeKo-1, Z-138, and Mino, using JVM-2 cells as a comparator. Cells were treated with compounds E, R, and T (10–40 μmol/L range) for 24 hours and flow cytometry was used to measure pBTK -Y223 levels. Compound E inhibited pBTK levels in all three SOX11⁺ cell lines with an IC₅₀ of 28–43 μmol/L (Fig. 5A). Compound R was relatively more active with IC₅₀ ranging from 23 to 34 μmol/L. Compound T was the least active, reaching IC₅₀ for only one of the three cell lines (JeKo-1). Reassuringly, pBTK level in SOX11⁻ JVM-2 cells was unaffected by all three SOX11i up to 40 μmol/L, further confirming a SOX11-specific mechanism of action. Positive control BTKi ibrutinib on the other hand, reduced pBTK level in all four cell lines irrespective of SOX11 expression status (Fig. 5B).

For further mechanistic validation of the inhibition of the transcriptional activity of SOX11 by SOX11i, we investigated the expression levels of SOX11 target genes. Mothers against decapentaplegic homolog 3 (SMAD3) and Serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform encoded by the PPP3CA gene, are MCL-specific target genes of SOX11 and mediators for the TGFβ pathway. Both these genes are repressed by SOX11 overexpression in MCL cell lines Granta-519 and Z-138 and their modulation by SOX11 has been validated by ChIP-qPCR and reporter assays (27). Conversely, paired box 5 (PAX5), a regulator of terminal B-cell differentiation is a validated MCL-specific gene activated by SOX11 that promotes a more aggressive disease phenotype (28).

Treatment of three SOX11⁺ cell lines JeKo-1, Mino, and Z-138 with 20 μmol/L SOX11i E, R, and T for 24 hours caused a significant de-repression of both SMAD3 and PPP3CA target genes and downregulation of PAX5. These transcriptional effects were virtually absent in SOX11⁻ JVM-2 cells (Fig. 5C). These results provide further evidence that these SOX11i interfere with SOX11-DNA interaction and thus inhibit both activating and repressing functions of SOX11.

Because compound R showed the strongest in vitro SOX11-specific anti-MCL activity and was efficacious in the majority of other assays, it became our SOX11i of choice for testing in ibrutinib-resistant primary MCL. SOX11 expression was compared in primary samples from 4 patients with ibrutinib-resistant MCL (see Supplementary Table S4 for patient characteristics) and two healthy volunteers. As expected, the two healthy volunteers were SOX11⁻, however, 3 of the 4 patients with ibrutinib-resistant MCL were positive for SOX11 expression (Fig. 6A). Treatment with compound R induced cytotoxicity specifically in the three SOX11⁺ ibrutinib-resistant (Pt. 1, 2, 3) but not in the SOX11⁻ ibrutinib-resistant (Pt. 4) MCL patient samples (Fig. 6B). In a PDX mouse model of one of these SOX11⁺ subjects (Pt. 2), compound R treatment led to complete loss of BTK phosphorylation confirming its on-target mechanism of action (Fig. 6C). We also tested compound R in combination with ibrutinib in SOX11⁺ JeKo-1, Z-138, and Mino cells and found strong synergy between the two agents (Fig. 6D) as depicted in the normalized isobolograms (Fig. 6E–G). A similar experiment with SOX11⁻ JVM-2 cells showed antagonism between compound R and ibrutinib (Fig. 6H) further confirming the on-target activity of compound R.

TF SOX11 is expressed in the majority of patients with MCL and is an important contributor to MCL pathogenesis. It is developmentally regulated, and is not expressed in adult tissues. As such, SOX11 represents a potential new drug target in MCL, but as a TF, it would be considered “undruggable.” Although, substantial success has been achieved in targeting protein–protein interactions mediated by TFs (43) or by targeting a ligand-binding domain outside of the DBD (44), targeting protein–DNA interactions still constitutes a major challenge. The overall convexity of DBDs (to bind the DNA major groove, for example) and the lack of pockets usually found in other targets such as kinases and GPCRs make it difficult to target DBDs with small molecules. Still, a small molecule (FDI-6) targeting the TF FOXM1-DBD (45) has been developed, as well as another one (Sm4) targeting the SOX18-DBD, which also inhibits SOX11-DBD albeit at sub-millimolar concentrations (42). Our in silico screening using a SOX11-DBD homology model led to the successful identification of three structurally related compounds (SOX11i; E, R, and T) that bind the SOX11-DBD and perturb its interaction with DNA. Specifically, our binding data from cell-free assays (SPR) confirm that SOX11i (E, R, and T) physically bind to SOX11-DBD with micromolar affinities and FRET and EMSA assays show that the compounds interfere with SOX11:DNA binding in vitro. During the preparation of our article, Dodonova and colleagues (46) published the X-ray crystal structure of the SOX11-DBD in complex with DNA (PDB code 6T78). Importantly, our three-dimensional homology model of SOX11-DBD utilized for virtual screening is confirmed by their experimentally determined structure with an RMSD of 1 Å (Supplementary Fig. S8).

TP53-mutated MCL is associated with the worst prognosis (12, 47, 48). Our study utilized four different MCL cell lines. Z-138 and JVM-2 have wild-type TP53, whereas MINO (missense V147G mutant), JeKo-1 (deleted) have defective TP53. Our data demonstrate that sensitivity to SOX11i (e.g., compound R) is SOX11 specific and independent of TP53 status because Z-138, MINO, and JeKo-1 are all sensitive but JVM-2 is not. Hence, SOX11i could provide a much-needed avenue for treatment of TP53-mutated MCL.

Our SOX11 inhibitors represent a proof of concept that DNA–TF interaction can be targeted to abrogate aberrant gene regulation and cellular growth in cancer. Our current molecules are early tool compounds for probing downstream effects of TF inhibition in vitro. The SOX11i we have discovered have potencies in the micromolar range and they also have limited aqueous solubility which has prevented us from testing them in our transgenic and PDX mouse models. We are actively pursuing chemical modifications to make inhibitors that are more potent, selective, and with desirable pharmacologic properties for in vivo application.

Cellular resistance to chemotherapy is a major obstacle to cancer treatment in MCL. The most significant advancement in the treatment of this challenging disease is the introduction of molecularly targeted therapies such as BTKi and BCL2i. However, patients develop resistance to BTKi due to a number of mechanisms including C481S mutations in the BTK-binding pocket preventing ibrutinib binding and median survival is less than 4 months after BTKi resistance (49). Similarly, resistance to BCL2i has been described because of upregulation of the AKT pathway (50). The development of resistance to therapeutics in BTKi and BCL2i setting represents the most urgent unmet need in MCL. Our data using primary MCL cells and PDX models demonstrate that our lead SOX11i, compound R, can overcome ibrutinib resistance. The approval of anti-CD19 CAR-T cells (51) and ongoing trials of CD20 × CD3 bispecific antibodies (52, 53) represent promising new approaches for patients with relapsed MCL, but are not currently curative. Given the room for improvement in MCL therapeutics, our results lay the foundation for future optimization of the chemical probes identified thus far toward inhibitors for therapeutic application in treating naïve and relapsed/refractory patients with MCL.

These novel small-molecule inhibitors will also be useful for understanding the pathogenesis of other SOX11⁺ malignancies such as epithelial ovarian tumors (54, 55), medulloblastoma (56), gliomas (35), and basal-like breast cancer (57), and ultimately expand therapeutic options for SOX11-expressing human cancers.

S.S. Jatiani reports grants from NIH, Icahn School of Medicine at Mt Sinai, and Celgene Consortium during the conduct of the study; in addition, S.S. Jatiani has a patent for SOX11 INHIBITORS FOR THE TREATMENT OF MANTLE CELL LYMPHOMA Serial Number 63/010,408 issued pending. S. Christie reports grants from NIGMS-funded Integrated Pharmacological Sciences Training Program T32 GM062754 during the conduct of the study. A. Kapoor reports grants from Celgene during the conduct of the study; in addition, A. Kapoor has a patent for SOX11 INHIBITOR FOR MANTLE CELL LYMPHOMA pending. C. Lee reports grants from Celgene during the conduct of the study; in addition, C. Lee has a patent for Targeting the SOX11 Transcription Factor in Mantle Cell Lymphoma pending. A. Wiestner reports grants from Acerta Pharma, Merck, Nurix, Genmab, Verastem, and Pharmacyclics outside the submitted work. J. Jin reports grants from Celgene during the conduct of the study; in addition, J. Jin has a patent for SOX11 Inhibitors pending. M. Filizola reports grants from Celgene during the conduct of the study, as well as grants from Celgene outside the submitted work; in addition, M. Filizola has a patent for 63/039,704 pending. A.K. Aggarwal reports grants from Celgene during the conduct of the study, as well as grants from Celgene outside the submitted work; in addition, A. Aggarwal has a patent 37100591 pending. S. Parekh reports grants from NCI R01 CA244899 and Celgene Consortium during the conduct of the study, as well as grants from Amgen, Karyopharm, and BMS and personal fees from Foundation Medicine outside the submitted work; in addition, S. Parekh has a patent for SOX11 INHIBITORS FOR THE TREATMENT OF MANTLE CELL LYMPHOMA Serial Number 63/010,408 issued. No disclosures were reported by the other authors.

S.S. Jatiani: Conceptualization, data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S. Christie: Data curation, formal analysis, investigation, visualization, writing–original draft, writing–review and editing. V.V. Leshchenko: resources, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. R. Jain: Data curation, formal analysis, supervision, investigation, visualization, methodology, writing–review and editing. A. Kapoor: Data curation, software, investigation, visualization, writing–review and editing. P. Bisignano: Data curation, software, formal analysis, investigation. C. Lee: Software, formal analysis, investigation, writing–original draft. H.U. Kaniskan: Resources, investigation, methodology, writing–original draft. D. Edwards: Data curation, formal analysis, investigation. F. Meng: Investigation, methodology, writing–original draft. A. Laganà: Software, formal analysis, validation, writing–review and editing. Y. Youssef: Formal analysis, investigation, writing–original draft. A. Wiestner: Conceptualization, formal analysis, investigation, writing–original draft. L. Alinari: Conceptualization, resources, formal analysis, supervision, methodology, writing–review and editing. J. Jin: Resources, supervision, writing–review and editing. M. Filizola: Conceptualization, resources, data curation, software, formal analysis, supervision, writing–review and editing. A.K. Aggarwal: Conceptualization, resources, formal analysis, supervision, visualization, writing–original draft, writing–review and editing. S. Parekh: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing.

We are grateful for support from Celgene Consortium Funding, NIH R01 CA244899, NIGMS-funded Integrated Pharmacological Sciences Training Program T32 GM062754, and Tisch Cancer Center Development Grant. Computations were run on resources available through the Scientific Computing Facility at the Icahn School of Medicine at Mount Sinai supported by the Office of Research Infrastructure of the NIH under award numbers S10OD018522 and S10OD026880.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.