Bifunctional Inhibitor Reveals NEK2 as a Therapeutic Target and Regulator of Oncogenic Pathways in Lymphoma

Diffuse large B-cell lymphoma (DLBCL) is the most diagnosed and aggressive lymphoma worldwide (1). Despite achieving long-term, disease-free survival in approximately 60% of patients with DLBCL, nearly half of patients experience refractory or relapsed disease that is often fatal (2–4). Over 10 years ago, two major DLBCL subgroups were identified on the basis of the genetic profiles of normal counterparts and presumed cell-of-origin (COO): Germinal center-like (GCB) and activated B-cell–like (ABC), which dichotomized patients according to outcome (1, 5–8). The IHC-based Hans algorithm was developed for clinical use but can only group tumors as GCB or non-GCB due to a limited scope of markers (9). Some clinics now use a small gene expression panel, Lymph2Cx, to classify tumors more definitively into GCB and ABC subgroups (10, 11). Recent genetic analyses of DLBCL proposing new molecular subtypes, up to six different subclasses, continue to reveal the complex nature of the disease and its varied clinical response (4, 12, 13).

The complexity of DLBCL is a major barrier to overcoming treatment failure with current immunochemotherapy (4, 12–14). Understanding the factors that drive aggressive disease in patients of diverse genetic backgrounds is critical to developing treatments that combat resistance and relapse. Strong evidence links the cell-cycle–associated Never in Mitosis Gene A (NIMA) related Kinase 2 (NEK2) to germinal center formation and growth (15), suggesting a role in DLBCL pathology. Although the cancer-enabling activities of NEK2 are well-documented in several cancers (16–21), studies of NEK2 in DLBCL are limited.

The NEK family of serine/threonine kinases (STK), discovered in Aspergillus, is essential for mitosis and cell-cycle regulation (22). The NEK family member NIMA is crucial for mitotic entry, highlighting its potential as a target for mitotic inhibition. The closest mammalian NIMA isoform is NEK2 (23). During G₂–M, NEK2 initiates centrosome separation; thus, deficiencies in NEK2 activity lead to G₂–M arrest, whereas NEK2 overactivity results in bypass of the G₂–M checkpoint (22, 24, 25). NEK2 engages in a myriad of phosphorylation events and protein–protein interactions within the intercentriolar linkage of the centrosome (26–28). Overexpression of NEK2 promotes premature chromosome separation, yielding daughter cells with frequent aneuploidy (29, 30), a common feature of many cancers, particularly DLBCL (4, 13). Though NEK2 expression was recently documented in DLBCL (31), it remains unclear whether NEK2 contributes to chemoresistance or if it is expressed in more aggressive DLBCL subtypes. Our previous work demonstrated that elevated NEK2 enhances AKT activity in myeloma and non–small cell lung carcinoma, and contributes to chemoresistance through NF-κB and PIM1 (21), a kinase that helps recently defined a more aggressive subset of DLBCL (4). When aberrantly expressed in indolent follicular lymphoma (FL), NEK2 drives the transformation into DLBCL (32, 33) and even de novo DLBCL exhibits elevated NEK2 mRNA compared with untransformed FL (33). These data implicate NEK2 in several key oncogenic pathways associated with DLBCL and suggest a role for NEK2 in more aggressive lymphomas, setting the foundation of NEK2 as a putative molecular target. However, no work thus far has investigated pharmacological NEK2 inhibition in lymphoma. In the first patient cohort of its size, we demonstrate that NEK2 expression is a predictor of poor outcome in DLBCL irrespective of COO subtype. We then developed a bifunctional inhibitor that inhibits NEK2 kinase activity and induces its proteasomal degradation. Our NEK2 inhibitor, NBI-961, shows potent antitumor effects by blocking cell-cycle progression through G₂–M and causing apoptosis, supporting an oncogenic function for NEK2 in DLBCL.

A total of 76 formalin-fixed, paraffin-embedded (FFPE) tissues from pre-treatment DLBCL biopsies were used in this study for NEK2 IHC and survival analyses under a waiver of written informed consent (IRB# 227939). Transformed and HIV-associated DLBCL were excluded from the cohort. For survival analyses, all patients received the R-CHOP standard treatment regimen that consists of rituximab, cyclophosphamide, doxorubicin (DOX), vincristine (VCR), and prednisone. Overall, there were 43 male and 33 female patients with a median age of 67 years at diagnosis (range, 20 – 98 years). COO subtype was assigned by the Hans algorithm (9) as 46 GCB (59%) and 30 non-GCB (41%). A patient-derived, lymphoma primary cell suspension from a 64-year-old male, 36206, was CD20⁺/CD10⁻/CD5⁻ lambda monotypic by flow cytometry, and CD79a⁺/PAX5⁺/BCL2⁺/BCL6⁻/MUM1⁻/MYC⁻ by immunostaining consistent with malignant B-cell lymphoma. The excess biopsy sample was taken from lymphoma in the scalp. The patient also presented with a sub-cranial infiltrate that extended into the dura but there was no detectable brain lesion. The primary cells, 36206, were obtained from the UAMS Tissue Biorepository and Procurement Core within a few hours of surgical removal, cultured, and bio-banked following written informed consent from patient before surgery (IRB# 228443). Sample 36206 was used in this study for cell-based assays to test efficacy of pharmacologic NEK2 inhibition. The use of human tissues and clinical data was approved by the University of Arkansas for Medical Sciences (UAMS) Institutional Review Board in accordance with the Declaration of Helsinki.

The DLBCL cell lines were purchased from the DSMZ [U2932 (DSMZ Cat# ACC-633, RRID:CVCL_1896) RIVA (DSMZ Cat# ACC-585, RRID:CVCL_1885)] or the ATCC [HT (ATCC Cat# CRL-2260, RRID:CVCL_1290), SUDHL4 (ATCC Cat# CRL-2957, RRID:CVCL_0539), SUDHL5 (ATCC Cat# CRL-2958, RRID:CVCL_1735), SUDHL6 (ATCC Cat # CRL-2959, RRID:CVCL_2206)] except for HBL1 and VAL, which were previously obtained from Dr. Rimsza (Mayo Clinic in Arizona, Scottsdale). The GM22761 (RRID: CVCL_1P76) and GM16113 (RRID: CVCL_7646) EBV-immortalized, benign peripheral B cells were obtained from the Coriell Institute. All cells were cultured at 37°C, 5% CO₂ in 10% FBS, 1% penicillin/streptomycin-supplemented RPMI-1640 media. Cell lines were tested for Mycoplasma every 6 months and authenticated by the University of Arizona Genetics Core using the PowerPlex 16 System (Promega), which consists of a forensic-style 15 autosomal STR loci, including 13 CODIS DNA markers (nine of the standard loci collected by the ATCC and DSMZ), Amelogenin, and a mouse-specific locus every 10–12 months (Supplementary Table S1).

NBI-961 (formerly CMP3a) was synthesized as previously described (34) and is outlined in Scheme 1 (Supplementary Fig. S1). Briefly, methyl thioglycolate (i) was cyclized with ethyl propiolate (ii) in methanol in the presence of sodium methoxide to afford methyl 3-hydroxythiophene-2-carboxylate. (iii) A Mitsunobu reaction in dry DCM between thiophene core and trifluoromethyl phenyl ethanol (iv) in presence of triphenyl phosphine (PPh₃) and diisopropyl azodicarboxylate (DIAD) yielded compound (v). Boronic acid derivative (vi) was generated via a Miyaura borylation reaction in dry THF in the presence of lithium diisopropylamide (LDA), which was used in next step without further purification. A Suzuki coupling reaction between boronic acid derivative (vii) and imidazo[1,2-a]pyridine (viii) generated compound 9. Then, compound 9 was heated in methanol with ammonia and ammonium hydroxide to convert ester to an amide (compound 10). The amide compound was then subjected to Suzuki coupling to afford compound NBI-961. ¹H NMR (400 MHz, DMSO) δ 8.41 (d, J = 7.2 Hz, 1H), 8.13 (s, 1H), 7.90 (s, 2H), 7.78 (d, J = 7.6 Hz, 4H), 7.55 (t, J = 7.7 Hz, 1H), 7.33 (d, J = 7.2 Hz, 1H), 7.15 (s, 1H), 7.09 (s, 1H), 6.02 (d, J = 6.7 Hz, 1H), 4.24 (t, J = 6.5 Hz, 2H), 2.73 (s, 2H), 2.21 (s, 6H), 2.09 (s, 1H), 1.76 (d, J = 6.1 Hz, 3H), m/z (ESI) calculated for C₂₈H₂₇F₃N₆O₂S [M+H]⁺: 567.19, found: 567.4. The purity of NBI-961 was 96% as determined by HPLC (Supplementary Fig. S2). INH154, the structure of which was previously described (35), was purchased from MedChemExpress (catalog no. HY-117154). Doxorubicin and vincristine were purchased from Cayman Chemical Company (item numbers 15007 and 11764). Sorafenib was purchased from LC Laboratories (Lot No: SFB-104). For in vitro experiments, the compounds were dissolved in DMSO. DMSO was used as the vehicle treatment control.

Paraffin slides were prepared from blocks of FFPE DLBCL tissues cut at 4 μm and stained using a VENTANA BenchMark XT instrument. The primary NEK2 antibody (Santa Cruz Biotechnology Cat# sc-55601, RRID:AB_1126558) was used at 1:50 dilution. The NEK2 antibody was used as supplied and incubated for 32 minutes at 37°C. Protein was detected with an OptiView DAB Detection Kit (760–700, VMSI). All slides were counterstained with hematoxylin for 4 minutes. Reactivity was considered in DLBCL cells with moderate to strong DAB brown staining for NEK2 in 5% increments. Two observers, S. Kendrick and V. Patel, independently evaluated the stained whole-tissue sections for NEK2 expression. A third observer, G.R. Post, resolved the discrepant cases (34%, 26/76). Discrepant scoring was defined as ≥20% difference in reactivity between the two observers. Both V. Patel and G.R. Post are formally trained hematopathologists. Discrepancies in scoring were due to the intratumor varied intensity (light to strong) of staining of DLBCL cells that frequently occurred within tumors. DLBCL cells were then dichotomized into low or high NEK2 expression groups based on whether the tumor was below or ≥ median reactivity, which was 80% cells, staining positive.

Cells were harvested by centrifugation, washed three times with cold PBS, and lysed on ice with RIPA buffer (Boston BioProducts) and 1x HALT protease/phosphatase inhibitor (Thermo Fisher Scientific) for 30 minutes. Lysates were centrifuged at 4°C for 10 minutes at 14,000 rpm, collected, and total protein concentration was quantified using a BCA assay according to the manufacturer's recommendations (Pierce). For each lysate, 100 μg of total protein was heated at 95°C for 10 minutes with 2X Laemmli loading buffer (Bio-Rad) supplemented with 5% β-mercaptoethanol. Lysates were resolved by SDS-PAGE at 220 V and transferred onto polyvinylidene difluoride membranes. Membranes were blocked for 1 hour at room temperature (RT) in 5% non-fat milk in PBS, rinsed with PBS, and incubated overnight at 4°C with primary antibody diluted with PBST. Membranes were then washed one time for 15 minutes followed by two, 10-minute washes in PBST while rocking at RT. Membranes were incubated with secondary antibody diluted with 1% non-fat milk in PBST for 1 hour while rocking at RT and rinsed three times with PBST. The following primary antibodies were used: Anti-NEK2 (Santa Cruz Biotechnology Cat# sc-55601, RRID:AB_1126558) anti-AKT (Cell Signaling Technology Cat# 9272, RRID:AB_329827), anti-phospho (Ser473) AKT (Proteintech Cat# 66444–1-Ig, RRID:AB_2782958), anti-TBP (Abcam Cat# ab818, RRID:AB_306337), and anti-ACTIN (Thermo Fisher Scientific Cat# PA1–183, RRID:AB_2539914). The following secondary antibodies were used: Goat anti-Rabbit IgG Dylight 800 (Thermo Fisher Scientific Cat# SA5–35571, RRID:AB_2556775) and Goat anti-Mouse Dylight 650 (Thermo Fisher Scientific Cat# SA5–10034, RRID:AB_2556614). The Precision Plus Protein Standard All Blue (Bio-Rad) was used as the molecular weight ladder. Membranes were imaged and converted to gray scale using the Bio-Rad ChemiDoc MP, except for the western blot analysis image of basal NEK2 levels in untreated cell lines, which required level adjustment using Creative Cloud Adobe Photoshop (from default input shadows, midtones, and highlights levels 0, 1.00, 255 to 0, 0.31, 255). Images were cropped and rotated as necessary using Creative Cloud Adobe Illustrator. Uncropped and unadjusted images of each membrane are provided in the Supplementary Data (Supplementary Figs. S3–S6).

RIVA, VAL, and SUDHL5 cell lines were treated with vehicle (0.1% DMSO) or NBI-961 and incubated for 96 hours at 37°C and 5% CO₂. Concentrations of NBI-961 were 240 nmol/L for VAL, 80 nmol/L for SUDHL5, and 25 nmol/L for RIVA, based on their different sensitivity and calculated GI₅₀ value. Samples were collected, reduced, and digested using filter-aided sample preparation (36). Tryptic peptides were labeled using tandem mass tagging, enriched for phosphopeptides, and submitted for mass spectrometric analysis on the Orbitrap Fusion Lumos and Eclipse Tribrid systems as described previously (37). To generate the heat maps and volcano plots, MaxQuant (version 1.6.5.0, Max Planck Institute) result files from searching the UniprotKB Homo sapiens database (April, 2019 for VAL and SUDHL5; October, 2019 for RIVA) were analyzed using ProteoViz as described previously (37).

The percentage of cell viability of the cell lines was determined by the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) colorimetric assay as per the manufacturer's protocol (Promega). Cells were seeded at a density of 20,000 cells per well and incubated with doxorubicin (DOX), vincristine (VCR), NBI-961, or sorafenib at 37°C and 5% CO₂ for 96 hours at a 2-fold serial dilution of 11 concentrations across a range of 0.001–40 μmol/L depending on the cell type. For the chemosensitization studies, cells were co-treated with NBI-961 and the chemotherapeutic drug DOX or VCR. The GI₅₀ values with the corresponding 95% confidence intervals were calculated after performing the comparison of fits for nonlinear regression models: Normalized response-variable slope, variable slope with four parameters, and variable slope with three parameters using GraphPad Prism software version 9. The best-fit regression model and predicted GI₅₀ value were selected on the basis of the sum-of-squares F Test. The therapeutic window was calculated on the basis of the ratio between the minimum effective concentration (MEC) and the minimum toxic concentration (MTC): Therapeutic window = MTC/MEC.

For the NBI-961 versus INH154 experiments, trypan blue exclusion was used to assess the number of live cells. The percentage of live cells was calculated in reference to the total number of cells counted by the Bio-Rad TC20-automated cell counter.

Cells were seeded at a density of 250,000 cells/mL and treated with vehicle (0.01% DMSO) or the indicated concentrations of NBI-961. For cell-cycle experiments, cells were harvested by centrifugation at the indicated time points and fixed for 2 – 4 hours at 4°C by resuspension in ice cold 70% ethanol. Cells were harvested by centrifugation and stained by resuspension in 0.75-mL propidium iodide (PI) solution (25 μg/mL PI, 20 μg/mL RNase, and 0.1% BSA in PBS) for 30 minutes at RT. The area under the first peak for PI (∼50,000 arbitrary units, AU) staining was gated as G₁, with the second peak (∼100,000 AU) gated as G₂–M. The area between peaks was gated as S phase, with cells below the first PI peak gated as sub-G₁. For apoptosis experiments, cells were harvested by centrifugation and washed with PBS. Cells were resuspended in 100 μL apoptosis staining solution [1 μg/mL PI, 2.5 μL Alexa Fluor 488 Annexin V, annexin-binding buffer (Thermo Fisher Scientific) to 100 μL] and stained for 15 minutes at RT. Then, 400 μL annexin-binding buffer was added to all samples, which were then transferred to ice and immediately analyzed. Cells staining positive for Annexin V were gated as early apoptotic, whereas cells staining double positive for Annexin V and PI were gated as late apoptotic and double negative cells gated as viable. For all flow cytometry experiments, cells were analyzed on the LSRFortessa (BD Biosciences), and doublets identified by plotting side scatter height over side scatter area were excluded from analysis.

Humanized NSG-TG(Hu-IL6), abbreviated as NSG-IL6, mice were housed and bred at the UAMS Animal Facility. They were created in Dr. Shultz's laboratory by microinjection of a BAC clone (GRCh37/hg19 chromosome7 22,724,723–22,964,038; BAC1) carrying a piece of chromosome 7 with the human IL6 gene and was microinjected into NSG mice (RRID: IMSR_JAX:005557) fertilized embryos. Mice were kept on a 12-hour light/12-hour dark cycle. RT was maintained at 24˚C – 26˚C. Food pellets and water were sterilized and provided ad libitum.

The U2932 DLBCL cell line was transfected with the luciferase gene (Lenti-UBC-RedFluc-T2A-EGFP) by using TransDux MAX Lentivirus Transduction Enhancer (System Biosciences; catalog no: LV860A-1) according to the manufacturer's protocol and maintained at 37°C, 5% CO₂ in 15% FBS, 1X penicillin/streptomycin, 1X Glutamax-supplemented RPMI-1640 media. Luciferase-expressing U2932 cells were washed with PBS three times and counted by Cellometer Mini (Nexcelom) using the trypan blue exclusion method. At 8–10 weeks of age, humanized NSG-IL6 mice, randomized for sex and cage mates, were injected via the tail vein with a total of 0.5 × 10⁶ U2932 cells (>90% viability) suspended in 100 μL PBS. U2932 cell engraftment was confirmed by early second week bioluminescence imaging after 10 minutes of D-luciferin (1.5 mg/mouse, PerkinElmer) intraperitoneal injection using an IVIS Imaging System 200 Series (PerkinElmer). After 10 days of U2932 cell transplantation, mice were randomized for sex, cage mate, and bioluminescence, and NBI-961 was administered via intraperitoneal injection at 5 mg/kg in PBS (n = 6) or PBS alone (n = 5) once a day. Mice were followed for 8-weeks post-treatment by weekly bioluminescence imaging and evaluating for morbidity (hair tufting, significant reductions of activity, and swollen body). This mouse study was approved by the UAMS IACUC (AUP #3987) and was conducted as per the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals.

Statistical analyses were performed with GraphPad Prism software version 9. Significance was set at a P value of <0.05 or an adjusted P value of <0.05. Data are represented as mean ± standard error of the mean from at least three independent experiments unless otherwise indicated. Patient survival was evaluated using Kaplan–Meier curves and the log-rank test. NEK2 was assessed as a covariate with COO subtype using the Cox proportional hazard model. The two-tailed Fisher's exact test was used to compare patient characteristics from NEK2-low and NEK2-high protein expression groups. Data collected and compared at a single time point were evaluated using a one-way ANOVA with the Dunnett's test for multiple comparisons except for the comparison between SUDHL5 lysate not treated or treated with phosphatase, which was evaluated using an unpaired two-tailed t test. Data collected and compared at multiple time points were evaluated using a two-way ANOVA with the Tukey test for multiple comparisons. Cell viability comparisons between NBI-961 and INH154 treatments were evaluated using the area under the receiver-operating characteristic curve.

The proteomics data generated and R scripts utilized in this study are publicly available at https://codeocean.com/capsule/4490151/tree/v1 or via ProteomeXchange with the identifier PXD015606. All raw mass spectrometry data from the proteomic studies are available at ftp://massive.ucsd.edu/MSV000092686/. All other data generated are available upon request from the corresponding authors.

NEK2 is gaining recognition as an oncogenic player in solid tumors, which prompted us to investigate NEK2 expression in de novo DLBCL. First, we re-analyzed DLBCL patient RNASeq data from the NCI Center for Cancer Research (13, 38), and found patients with NEK2 mRNA abundance at or above the cohort median tend to have worse overall survival (Fig. 1A). We then performed IHC staining for NEK2 within tumor tissue from our in-house DLBCL patient biopsies (Fig. 1B) to determine NEK2 protein levels. Most DLBCL tumors frequently expressed NEK2 protein with 62% of the cases (47/76) having at least 80% of DLBCL cells staining positive for NEK2 and an overall range of 20% – 100% malignant cells staining positive in all 76 cases (Fig. 1C). When NEK2 was expressed in 80% or more of the malignant cells (NEK2-high), patients had significantly inferior outcomes (Fig. 1D). To rule out the possibility that NEK2 may be expressed predominantly in the more aggressive COO subtype, we performed analysis of covariance and found that NEK2 protein remained a significant indicator of patient survival when considered as a covariate with the known prognostic indicator of the non-GCB subtype (ref. 9; Supplementary Table S2; Supplementary Fig. S7A). Although patient sex did not differ between NEK2-low and NEK2-high protein DLBCL tumors, patients were slightly older in the NEK2-high group although not significantly (Supplementary Table S3). Positivity of NEK2 protein did not correlate with age (Supplementary Fig. S7B), and thus age likely does not contribute to the inferior survival of the patients with NEK2-high tumors. Recognizing the better outcomes associated with GCB DLBCL, we examined whether NEK2 protein differed among subtypes. However, there was no difference in subtype distribution between the NEK2-low and NEK2-high tumors (Supplementary Table S3). Furthermore, there was no difference in NEK2 protein between the two subgroups (Supplementary Fig. S7C). Of note, our cohort displayed the expected trend toward the non-GCB COO subtype conferring poor survival (Supplementary Fig. S7D). However, low sample size and any misclassification from using the Hans algorithm (9, 10) likely account for not reaching statistical significance. Overall, the separation of survival based on NEK2 mRNA and protein expression suggests that NEK2 may contribute to aggressive disease in DLBCL.

To investigate the cellular role of NEK2 in DLBCL, we used our recently developed NEK2 inhibitor, NBI-961 (a.k.a. CMP3a; ref. 34; Fig. 2A). NBI-961 binds to the active protein kinase conformation (Fig. 2B) and efficiently inhibits NEK2 catalytic activity (34). NBI-961 exhibits high binding affinity for NEK2 relative to >90 other kinases (scanEDGE kinome panel, Eurofins DiscoverX; Fig. 2C; Supplementary Table S4). NBI-961 at 300 nmol/L displaced 98.2% of bound NEK2 and a few other kinases to a lesser extent. At 15 nmol/L, off-target effects were abolished, except for the Fms-Related Receptor Tyrosine Kinase 3 (FLT3). However, we previously observed NBI-961 catalytic inhibition of FLT3 at a less than 10-fold potency compared with NEK2 inhibition (34).

We then determined basal expression of NEK2 in a panel of established cells representing the molecular subtypes of DLBCL and two benign B-cell lines. At the mRNA level, the GCB cell lines SUDHL5 and HT display the highest NEK2 abundance compared with the benign B-cell line (Fig. 2D). However, this high expression persisted at the protein level only for SUDHL5 (Fig. 2E). We also observed a distinct upshifted band, especially in SUDHL5 lysates (Fig. 2E). Phosphatase significantly reduced the intensity of this band with a corresponding increase in total NEK2 (Fig. 2F), indicating that the upshifted band represents phosphorylated NEK2. Next, we tested how NEK2 inhibition with NBI-961 affects DLBCL cell viability. Overall, NBI-961 achieved growth inhibitory concentrations (GI₅₀) in the nanomolar range particularly in ABC-derived cells (Fig. 2G; Supplementary Table S5).

To identify the cause of compromised cell viability, we performed cell-cycle and apoptosis analyses in a subset of the NBI-961–sensitive (SUDHL5 and RIVA) and less-sensitive (VAL) DLBCL cell lines as well as the GM22671 benign B cells. Consistent with the established role of NEK2 coordinating progression through G₂–M (22), we observed a stark G₂–M arrest in SUDHL5 and RIVA and to a lesser extent in VAL cells (Fig. 3A). The G₂–M stalling occurred as early as 24 hours NBI-961 treatment, and for RIVA and VAL, the G₂–M arrest persisted through 48 hours. There was no detectable G₂–M arrest in the benign GM22671 cells. For SUDHL5 and RIVA, the G₂–M arrest at 24 hours coincided with significant apoptotic death that steadily increased over the 96 hours (Fig. 3B). In contrast, NBI-961 did not trigger apoptosis in VAL cells until 48 hours, and as with the G₂–M arrest, this effect occurred at 3- or 9.6-fold higher NBI-961 concentrations compared with SUDHL5 and RIVA. Similarly, NBI-961 induced apoptosis in GM22671 cells at 48 hours and required an approximately 3-fold higher concentration of NBI-961 relative to RIVA. As a positive control, a high concentration of DOX alone was capable of inducing apoptosis in all four cell lines (Supplementary Fig. S8).

Given the efficacy of NBI-961, we sought to determine the mechanism by which it inhibits NEK2 and compare its effects with the commercially available NEK2 inhibitor, INH154. Because NEK2 undergoes autophosphorylation (25), we used NEK2 phosphorylation as direct readout for NEK2 catalytic inhibition. After 24 hours treatment with NBI-961 at low nanomolar levels, we observed a more pronounced decrease in phosphorylated NEK2 (p-NEK2) in SUDHL5 cells compared with the less-sensitive VAL cell line (Fig. 4A and B). Unexpectedly, we found not only reduction in p-NEK2 but also total NEK2 protein (Fig. 4A and B) suggesting that NBI-961 can promote NEK2 proteasomal degradation. Consistent with this notion, the proteasome inhibitor bortezomib blocked NBI-961–induced loss of NEK2 without negating the loss of p-NEK2 (Fig. 4C). mRNA abundance remained constant with NBI-961, confirming that loss of NEK2 occurred at the protein level (Fig. 4D). These data demonstrate a bifunctional mechanism of action for NBI-961 in which NEK2 is both catalytically inhibited (loss of p-NEK2) and degraded by the proteasome (loss of total NEK2). The loss of NEK2 and p-NEK2 was also observed in RIVA and another NBI-961–sensitive ABC-DLBCL cell line, U2932 (Supplementary Fig. S9).

We then compared the effects of NBI-961 on cell survival to INH154, a previously identified NEK2 inhibitor. INH154 indirectly inhibits NEK2 by binding the kinetochore-associated protein HEC1, thus preventing HEC1/NEK2 interaction that subsequently signals for NEK2 proteasomal degradation in HeLa cells (35). Although NBI-961 reduced the percentage of live cells in both the SUDHL5 DLBCL cell line and primary patient-derived lymphoma cells, INH154 did not impact cell survival at 24 hours (Fig. 4E and F) or even 96 hours in the primary cells (Fig. 4F). Despite reports of NEK2 degradation in breast and cervical cancer cell lines (35), we did not observe loss of NEK2 in SUDHL5 or in primary patient-derived lymphoma cells following treatment with INH154 (Fig. 4G and H). The primary lymphoma cells were derived from an atypical extranodal biopsy in the scalp and thus follow-up studies with typical DLBCL are warranted. Because INH154 did not alter NEK2 activity or expression nor affect lymphoma cell viability, these findings reinforce that the lymphoma cell killing properties of NBI-961 are linked to the ability of NBI-961 to elicit NEK2 degradation. As further demonstration, NBI-961 does not degrade FLT3 nor alter phosphorylation of a downstream FLT3 substrate, mTOR, and is approximately 23 – 188x more cytotoxic than sorafenib, a tyrosine kinase receptor inhibitor that inhibits FLT3, VEGFR, and PDGFRβ (refs. 39, 40; Supplementary Fig. S10).

In showing that NEK2 inhibition disrupts the proliferation and survival of DLBCL cells in a subset of cell lines typifying the molecular heterogeneity in patient tumors, we proposed that the NBI-961–sensitive cells may predominantly express and therefore depend on NEK2 regulated pathways. To generate a network of NEK2-dependent pathways and substrates, we used an unbiased total- and phospho-proteomic approach in untreated SUDHL5 and VAL cells, which are matched for COO but differ in their sensitivity to NBI-961. Quantitative total proteomics revealed that SUDHL5 cells exhibit a cell-cycle and centrosome separation profile dominated by known NEK2-regulated proteins relative to VAL (Fig. 5A and B; Supplementary Table S6).

We then determined how the DLBCL proteome and phospho-proteome were altered by NBI-961. NBI-961–sensitive cells, SUDHL5 and RIVA, displayed significant changes to protein expression with NBI-961 treatment (Fig. 5C; Supplementary Tables S7 and S8). There was a greater overall change in RIVA cells (note the greater Log₂ fold change compared with SUDHL5), which may account for their greater sensitivity to NBI-961 (Figs. 2G and 3). We then grouped differentially expressed proteins following NBI-961 treatment for SUDHL5 and RIVA. Although few proteins overlapped, overlapping proteins as well as the 25 most significant proteins from each cell line were enriched for cell-cycle and centrosome-related pathways (Fig. 5D; Supplementary Table S9). In addition, the loss of NEK2 resulted in an altered global phosphorylation profile, including changes in phosphorylation status of known or predicted NEK2 substrates such as MAP2K2, KIF11, TOP2A, and EIF4ENIF1 (refs. 41, 42; Fig. 5E; Supplementary Tables S10 and S11). As expected for an STK, sequence motif analysis of phosphopeptides differentially detected in the NBI-961 versus vehicle-treated control group showed high probability for differential phosphorylation status of serine or threonine (Fig. 5F). The predominance of serine is consistent with previous work showing NEK2 exhibits preference for phospho-acceptor serines relative to threonines (43). A K3 kinase enrichment analysis (KEA3, v3) of proteins with significant loss of phosphorylation implicated cell-cycle kinases and a few cell death–associated kinases as likely drivers of the altered phosphorylation landscape following NBI-961 treatment (Fig. 5G; Supplementary Table S12). Despite KEA3 identifying NEK2 as a lower ranked kinase (18/42; P = 0.079), the top two kinases, cyclin-dependent kinase 2 (CDK2) and CDK4 are known to directly interact with NEK2 (30, 44) with evidence that expression of CDK4 and NEK2 are co-dependent (30). These phosphoproteomic analyses also identified proteins not yet associated with NEK2 that are potentially critical for NEK2-dependencies in DLBCL, including PKN1 and IKZF3, which were common to both cell line subtypes. Given that PKN1, IKZF3, and other known NEK2 substrates are implicated in regulation of AKT (21, 45–49), a well-recognized hub for B-cell pro-survival pathways (50, 51), we investigated whether NBI-961 inhibits AKT phosphorylation. NBI-961 resulted in a dose-dependent loss of phosphorylated AKT with a concomitant reduction in NEK2 (Fig. 5H). A modest decrease in total AKT reached a maximum over the three, 2-fold concentrations. Together, these studies indicate that by targeting NEK2 in DLBCL, downstream oncogenic signaling can also be inhibited.

Given the clinical interest in sensitizing resistant DLBCL to existing chemotherapy to improve patient response (52–54), we investigated the ability of NEK2 inactivation to potentiate DOX and VCR. Both agents can have severe toxic effects, including DOX-related cardiotoxicity (55–57), so reducing the effective concentration of these drugs would improve patient survivorship. In the NBI-961–sensitive DLBCL cell lines, SUDHL5 and RIVA, NBI-961 sensitized cells to previously noncytotoxic concentrations of DOX and VCR (Fig. 6A) that were 8 – 28-fold (DOX) and 5 – 7-fold (VCR) lower than their GI₅₀ value (Supplementary Table S5). The significant loss in cell viability by NBI-961 potentiation of DOX was due to an increase in apoptosis (Fig. 6B). In contrast, the less NBI-961–sensitive VAL DLBCL cells and the GM22671 benign cells showed no chemosensitization to DOX and VCR and no induction of apoptosis (Fig. 6A and B). Similarly, other DLBCL cell lines more sensitive to NBI-961 alone (Fig. 2G; Supplementary Table S5), U2932 and to a lesser extent HT and SUDHL6, were also sensitized to DOX and VCR by NBI-961 whereas HBL1, SUDHL4, and a second benign B-cell line, GM16113, were not (Supplementary Fig. S11). Considering 7.8 nmol/L as the minimal concentration for drug effect (MEC) and the GI₅₀ value for the benign B-cell lines as the minimal concentration for toxicity (MTC), we determined the therapeutic window of NBI-961 between the sensitive DLBCL cells and the benign B cells is 4.2, indicating that a 4.2x higher concentration of NBI-961 is required to cause toxicity to kill 50% of benign B cells than is needed to produce a therapeutic effect in the DLBCL cells.

Mouse models of DLBCL often fail to recapitulate human disease. For example, transgenic mice rely on expression of oncogenes that better reflect only specific subgroups of DLBCL (58). Xenograft models better reproduce late-stage disease but require implantation into a localized area, which does not reflect human disease distribution (59). We established a xenograft model in NSG-IL6 mice where, rather than implanting DLBCL cells in a localized region, cells are injected intravenously, and tumors grow in areas that better recapitulate the disease in humans (60). Using U2932 cells, a once-daily intraperitoneal injection of NBI-961 at 5 mg/kg for the duration of the experiment delayed disease onset (Fig. 6C and D) and prolonged mouse survival (Fig. 6E). NBI-961 was well tolerated in the mice with no visible signs of distress or toxicity, including no visible weight loss and no loss of appetite, activity, or tuft fur.

Although many patients with DLBCL achieve disease-free survival following the current R-CHOP regimen, about 40% experience relapsed disease that is often deadly (61). R-CHOP poses significant toxicity, including an increased risk for grade 3–4 adverse cardiovascular events (62). These outcomes, along with the heterogenous genetic background of DLBCL, emphasize the need for adjuvant therapies targeting new molecular DLBCL etiologies. Recent work found that the expression of the cell-cycle kinase NEK2 is elevated in DLBCL and at the mRNA level, confers poor patient outcome (31). Here, our study using larger cohorts supports these findings by showing both high NEK2 mRNA and protein expression is a significant predictor of patient outcome. We also demonstrate that this correlation is independent of the known influence of COO subtype on DLBCL prognosis and the potential to pharmacologically target NEK2.

Our NEK2 inhibitor NBI-961 is selective for NEK2, with only residual binding to one off-target kinase, FLT3. It is unclear whether non-malignant mature B cells or lymphomas rely on FLT3 signaling (63, 64), particularly in post-GC–derived malignancies such as ABC DLBCL (65). Together with our previous enzymatic inhibition assay that revealed an approximately 12-fold lower IC₅₀ value of NBI-961 for NEK2 relative to FLT3 (66), the unaltered FLT3 expression and downstream targeting we observed in DLBCL cells suggest NBI-961 is unlikely to affect FLT3 (34).

STK inhibition in cancer is highly susceptible to molecular relapse through multiple, distinct mechanisms, such as kinome rewiring associated with retained scaffolding functions after kinase inhibition, leading to complete response rates of as low as 4% following STK inhibition in lymphoma (67–70). Kinase degradation may offer more sustained reduction in signaling and at lower concentrations compared with kinase inhibition (71). NBI-961 exhibits an unprecedented bifunctional mechanism of action in which it both catalytically inhibits NEK2 and induces NEK2 proteasomal degradation, leading to lymphoma cell death. Because catalytic inhibition of NEK2 alone by INH154 was insufficient to induce lymphoma cell death, it is likely that noncatalytic activities of NEK2 are essential to drive proliferation of lymphomas. We suspect that NEK2 may serve as a scaffold protein to coordinate cell-cycle progression and prevent apoptosis, which may be a feature unique to certain subsets of DLBCL. Therefore, the added function of NEK2 degradation upon treatment with NBI-961 may enhance therapeutic benefit beyond traditional kinase inhibition by eliminating NEK2 scaffolding and protein–protein interactions, thus limiting the pathways by which resistance may occur (71). An important question to address in the future is how does disruption of NEK2 noncatalytic or scaffolding functions drive G₂–M arrest and cell death in DLBCL. NBI-961–induced NEK2 degradation can serve as a useful tool to identify these noncatalytic functions.

Sensitivity to NEK2 inhibition did not seem to correlate with any of the classic COO subtypes; although, two of the three ABC DLBCL cell lines displayed increased sensitivity to NBI-961 relative to all of the GCB cell lines. These two cell lines are also associated with the recently proposed molecular subtypes, BN2 (RIVA; ref. 12) and possibly A53 (U2932; ref. 60). BN2 cases tend to have upregulated NF-κB signaling (13), which is characteristic of ABC tumors and associated with increased NEK2 in other cancers (72, 73) A53 is characterized by high rates of aneuploidy, another condition associated with high NEK2 expression (20). Regardless, we show a differential response to NBI-961 treatment in terms of NEK2 kinase inhibition and degradation that is directly linked to effects on cell death and cell-cycle progression.

These experiments indicate a role for NEK2 in DLBCL cell viability and B-cell cycle progression. We demonstrate that the novel NEK2 inhibitor NBI-961 significantly attenuates lymphoma cell viability. Perhaps most strikingly, NBI-961 synergizes with the standard of care drugs DOX and VCR to induce apoptosis and significantly lower their effective doses in DLBCL cells but not in benign B cells, indicating the potential of NEK2 inhibition to reduce the amount and duration of existing treatment. Reducing the dose or frequency of chemotherapy could reduce the severity of dose-dependent neuro- and cardiotoxicities and lower the risk of secondary malignancies resulting from treatment. These studies establish a key role for NEK2 in DLBCL and set the foundation for development of anti-NEK2 therapeutic strategies.

M. McCrury reports grants from National Institutes of Health (NIGMS) during the conduct of the study. S.D. Byrum reports grants from NIH/NIGMS and NIH/NIGMS during the conduct of the study. A. Rodriguez reports grants from Bristol Myers Squibb outside the submitted work. L.D. Shultz reports Royalties from The Jackson Laboratory from their licensing NSG MHC class I/II knockout mice to For Profit Companies by L.D. Shultz. Note: This is stated in the article. B. Frett reports grants from National Institutes of Health during the conduct of the study; as well as reports a patent for WO2018081719A1 issued. S. Kendrick reports grants from National Institutes of Health (NIHGMS) during the conduct of the study. No disclosures were reported by the other authors.

M. McCrury: Formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K. Swafford: Formal analysis, investigation, visualization, methodology, writing–review and editing. S.L. Shuttleworth: Formal analysis, investigation, visualization, methodology, writing–review and editing. S.H. Mehdi: Investigation, visualization, methodology, writing–review and editing. B. Acharya: Investigation, methodology, writing–review and editing. D. Saha: Investigation, methodology, writing–review and editing. K. Naceanceno: Data curation, software, formal analysis, investigation, methodology, writing–review and editing. S.D. Byrum: Data curation, software, formal analysis, writing–review and editing. A.J. Storey: Data curation, software, formal analysis, investigation, writing–review and editing. Y.-Z. Xu: Investigation, writing–review and editing. C. Doshier: Investigation, writing–review and editing. V. Patel: Investigation, writing–review and editing. G.R. Post: Resources, investigation, writing–review and editing. A. De Loose: Resources, writing–review and editing. A. Rodriguez: Conceptualization, resources, writing–review and editing. L.D. Shultz: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing. F. Zhan: Formal analysis, writing–review and editing. D. Yoon: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing. B. Frett: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing. S. Kendrick: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing.

We thank Drs. S. Alam and A. Sharabura, who as medical students at the time of FFPET tissue collection assisted in collating the patient clinical data. The Kendrick and Frett laboratories are supported by the National Institutes of Health (NIH; P20GM121293, P20GM109005, and R24GM137786); and the UAMS Winthrop P. Rockefeller Cancer Institute Teams Science Award. The Yoon laboratory is supported by the UAMS Winthrop P. Rockefeller Cancer Center Institute Seeds of Science Award. The Shultz laboratory is supported by the NIH (CA034196). The Zhan laboratory is supported by NIH-NCI 1R01CA236814–01A1, 3R01-CA236814–03S1, U54-CA272691, US Department of Defense (DoD) CA180190, the Paula and Rodger Riney Foundation, and UAMS Winthrop P. Rockefeller Cancer Institute (WRCRI) Fund.