Systems biology-enabled targeting of NF-κΒ and BCL2 overcomes microenvironment-mediated BH3-mimetic resistance in DLBCL

Abstract

In Diffuse Large B-cell Lymphoma (DLBCL), elevated anti-apoptotic BCL2-family proteins (e.g., MCL1, BCL2, BCLXL) and NF-κB subunits (RelA, RelB, cRel) confer poor prognosis. Heterogeneous expression, regulatory complexity, and redundancy offsetting the inhibition of individual proteins, complicate the assignment of targeted therapy. We combined flow cytometry 'fingerprinting', immunofluorescence imaging, and computational modeling to identify therapeutic vulnerabilities in DLBCL. The combined workflow predicted selective responses to BCL2 inhibition (venetoclax) and non-canonical NF-κB inhibition (Amgen16). Within the U2932 cell line we identified distinct resistance mechanisms to BCL2 inhibition in cellular sub-populations recapitulating intratumoral heterogeneity. Co-cultures with CD40L-expressing stromal cells, mimicking the tumor microenvironment (TME), induced resistance to BCL2 and BCLXL targeting BH3-mimetics via cell-type specific upregulation of BCLXL or MCL1. Computational models, validated experimentally, showed that basal NF-κB activation determined whether CD40 activation drove BH3-mimetic resistance through upregulation of RelB and BCLXL, or cRel and MCL1. High basal NF-κB activity could be overcome by inhibiting BTK to resensitize cells to BH3-mimetics in CD40L co-culture. Importantly, non-canonical NF-κB inhibition overcame heterogeneous compensatory BCL2 upregulation, restoring sensitivity to both BCL2- and BCLXL-targeting BH3-mimetics. Combined molecular fingerprinting and computational modelling provides a strategy for the precision use of BH3-mimetics and NF-κB inhibitors in DLBCL.

Fig. 1. Summary of existing knowledge of links between the Tumor Microenvironment (TME), NF-κB dimers, and BCL2-family proteins in lymphoma.

The canonical NF-κB pathway (green) is mediated in part by B-cell receptor (BCR) signaling. BCR signaling transduces signals through Bruton’s tyrosine kinase (BTK), and NF-κB essential modulator (NEMO), to primarily activate RelA- and cRel-containing NF-κB dimers. Single cell RNA sequencing has revealed that the primary impact of the TME on DLBCL cells is through the non-canonical NF-κB pathway (purple). Macrophages (Mϕ), dendritic cells (DC), and cancer associated fibroblasts (CAFS), activate non-canonical signaling through B cell-activating factor receptor (BAFFR). T cells in the TME secrete CD40 ligand. Both BAFFR and CD40 activation lead to stabilization of NF-κB-inducing kinase (NIK) and activation of RelB-containing NF-κB dimers. While RelB activity has been shown to upregulate BCLXL in chronic lymphocytic leukemia, and there are frequently reported links between NF-κB and BCL2, it is unknown which NF-κB dimers induce which BCL2-family proteins (indicated by gray arrows), and how this drives response to therapies in the context of the DLBCL TME.

Fig. 2. NF-κB and BCL2 ‘fingerprinting’ predicts sensitivities to inhibition of non-canonical NF-κΒ and BCL2 in Diffuse Large B cell Lymphoma (DLBCL).

A Representative dot plots of RelA and RelB Median Fluorescence Intensity (MFI) values for the DLBCL cell lines SUDHL8, SUDHL10, U2932, and RIVA. Workflow pipeline for the productions of NF-κΒ fingerprints. B Relative RelA and RelB expression levels measured with flow cytometry, in the cell lines RIVA, U2932, SUDHL8 and SUDHL10 as fingerprints, with dots representing the median in a single experiment, also shown as probability density curves (right and top). C Bar graphs of RelA and RelB median fluorescence intensity (MFI) values as mean ± standard deviation of two independent experiments. (*P < 0.05, (**P < 0.01; one-way ANOVA with Tukey’s comparisons test). D Cell viability in response to 0.125-100μΜ of Amgen16 in the indicated DLBCL cell lines post a 24-h treatment, error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control. E LC₅₀ values for the indicated cell lines in response to Amgen16 with mean indicated, three independent experiments (**P < 0.01, ***P < 0.001 unpaired Ttest). F Relative MCL1, and BCL2, and BCLXL expression levels measured with flow cytometry, in the cell lines RIVA, U2932, SUDHL8, and SUDHL10 as fingerprints, with dots representing the median in a single experiment, also shown as probability density curves (right and top). G Bar graphs of MCL1, and BCL2, and BCLXL median fluorescence intensity (MFI) values as mean ± standard deviation of three independent experiments. (**P < 0.01, ***P < 0.001, ****P < 0.0001; one-way ANOVA with Tukey’s comparisons test). H Cell viability in response to 0.0001–100 μΜ of ABT199 (venetoclax) in the indicated DLBCL cell lines, post a 24-h treatment, error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control. I LC₅₀ values for the indicated cell lines in response to ABT199 with mean indicated, of three independent experiments (**P < 0.01, ***P < 0.001 unpaired Ttest).

Fig. 3. Differential BCL2 family abundance in the sub-clonal cell line U2932 predicts differential responses to ABT199 (venetoclax), TME mimicking hCD40L-3T3 co-culture confers resistance to BCL2 inhibition.

A Flow cytometry dot plot showing the distinct levels of CD20 Median Fluorescence Intensity (MFI) in the U2932 subclones R1 and R2. B Cell viability in response to 0.0001–100 μΜ of ABT199in the indicated U2932 subclones R1 and R2, post 24-h treatment, error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control. C Bar graphs of MCL1, and BCL2, and BCLXL median fluorescence intensity (MFI) values as mean ± standard deviation of three independent experiments (****P < 0.0001; unpaired t-test). D Schematic of the co-culture system used. U2932 cells were co-cultured with hCD40L-3T3 cells prior to treatment with increasing doses of ABT199. E Viability of U2932 cells cultured for 24 h with or without CD40-Ligand NIH3T3 in the presence of 0.0001–100μΜ of ABT199. Error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (**P < 0.01, ***P < 0.001 multiple unpaired T tests using the Holm-Šídák corrections test). F LC50 values for the U2932 cell line in monoculture vs co-culture with hCD40L-3T3 in response to ABT199, with error bars representing the mean ± standard deviation of three independent experiments(**P < 0.01, unpairedTtest).G MCL1, BCL2, and BCLXL levels, shown as fold changes of 24-h stimulation with hCD40L-3T3 cells to unstimulated control in U2932 cells with error bars representing the mean ± standard deviation of five independent experiments (*P < 0.05, **P < 0.01 one-way ANOVA with Tukey’s comparisons test).

Fig. 4. Tumor microenvironment-mimicking co-culture induced resistance to BH3-mimetics in Diffuse Large B Cell Lymphoma (DLBCL) cell lines.

A Viability of RIVA cells cultured for 24 h with or without hCD40L-3T3 cells in the presence of 0.0001–100μΜ of the BCL-2 inhibitor ABT199 (venetoclax). Error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control. B LC₅₀ values for RIVA in each condition with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (*P < 0.05, unpaired Ttest). C MCL1, BCL2, and BCLXL levels, shown as fold changes of 24-h stimulation with hCD40L-3T3 cells to unstimulated control in RIVA with error bars representing the mean ± standard deviation of five independent experiments (**P < 0.01, ***P < 0.001 one-way ANOVA with Tukey’s comparisons test. D Viability of SUDHL8 cells cultured for 24 h with or without hCD40L-3T3 cells in the presence of 0.0001–100 μΜ of the BCLXL inhibitor A1331852. Error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control. E LC₅₀ values for SUDHL8 in each condition with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (***P < 0.001, unpaired Ttest). F MCL1, BCL2, and BCLXL levels, shown as fold changes of 24-h stimulation with hCD40L-3T3 cells to unstimulated control in SUDHL8 with error bars representing the mean ± standard deviation of five independent experiments (*P < 0.05, one-way ANOVA with Tukey’s comparisons test).

Fig. 5. Immunofluorescence microscopy reveals differential NF-κB basal RelA activity in Diffuse Large B-cell Lymphoma (DLBCL) cell lines.

A Schematic of the canonical and non-canonical NF-κB signaling pathway illustrating how non-canonical NF-κB activation results in the induction of BCLXL and how elevated basal canonical signaling induces multiple negative regulators of NF-κB which may contribute to cell line specific induction of MCL1. B Representative immunofluorescence microscopy following staining in RIVA cells for RelA (NF-κB), DAPI (nucleus), actin (cytoplasm), to measure nuclear and cytoplasmic subcellular localization. C Quantification immunofluorescence microscopy showing nuclear to cytoplasmic ratio of cRel and RelA in single cells. Data shows a representative replicate in the indicated cell lines. Violin width indicates data density. D Quantification of the nuclear to cytoplasmic ratio of cRel (top) and RelA (bottom) as bar graphs with error bars of the mean ± standard deviation of three independent experiments in RIVA, U2932 and SUDHL8 (*P < 0.05; One-Way ANOVA).

Fig. 6. Co-culture with hCD40L-3T3 cells upregulates NF-κB RelB and BCLXL in DLBCL, and selectively upregulates NF-κB cRel and MCL1 through signaling crosstalk when NF-κB RelA is chronically active.

A Abundances of the heterodimers RelA:p50 (left) and cRel:p50 (right) bound to inhibiting NF-κΒ proteins (IκB), as simulated by computational modeling for basal canonical NF-κΒ activation state in RIVA, and high basal activation state in SUDHL8. See supplementary modeling description. B Computational modeling results showing the nuclear abundance of the indicated NF-κB dimers in a simulation with low basal nuclear RelA (RIVA, pink), and a simulation high basal nuclear RelA (SUDHL8, teal). The mean (line) and standard deviation (shaded region) of 25 cells is indicated. C Schematic demonstrating the mechanism by which NF-κB contributes to the induction of BCLXL in a cell line with normal basal signaling (RIVA) upon stimulation with CD40L (right). hCD40L-3T3 mediated activation of NIK results in p100 processing into p52, leading to nuclear translocation of RelB:p52 and increased expression of BCLXL. Gray = inactive pathways and low abundance proteins, color = active pathways and predominant complexes. D Schematic demonstrating how crosstalk emerges between CD40, NIK, cRel, and MCL1. Increase basal RelA results in increase p100, and IκBδ (left). hCD40L-3T3 mediated activation of NIK results in processing of IκBδ and release of cRel:p50, which translocates to the nucleus and potentially upregulates expression of MCL1 (right). Gray = inactive pathways and low abundance proteins, color = active pathways and predominant complexes. Quantification of immunofluorescence microscopy in RIVA (E) and SUDHL8 (F) cells showing nuclear to cytoplasmic ratio of RelA, RelB and cRel. Data post 24 h of monoculture and hCD40L-3T3co-culture is shown side by side. Single cells of a representative replicate are shown (left), with violin width indicating data density. Quantification of the nuclear to cytoplasmic ratio of RelA, RelB, and cRel is shown as bar graphs (right) with error bars displaying the mean ± standard deviation of three independent experiments (*P < 0.05, **P < 0.01; unpaired ttest). G Schematic of the co-culture system used, following 24 h of treatment with 0.0001–100 μΜ of the BCLXL inhibitor A1331852 ± 0.1 μM of the Bruton’s Kinase (BTK) inhibitor Ibrutinib. H Cell viability of SUDHL8 in response to 0.0001–100 μΜ of the BCLXL inhibitor A1331852 post a 24-h treatment in monoculture, post a 24-h hCD40L-3T3co-culture or post a 24-h hCD40L-3T3 co-culture with the addition of 0.1 μM of the BTK inhibitor, Ibrutinib. LC₅₀ values for each condition is shown (right) with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (***P < 0.001 one-way ANOVA with Tukey’s comparisons test).

Fig. 7. IκBε knock-out mouse model reveals cRel-dependent induction of MCL1.

A NF-κB ChIP-seq data from Zhao et al. 2014 [53] showing binding of the indicated NF-κB subunits around the transcription start site (TSS) of the indicated anti-apoptotic genes in lymphoblastoid B cell line. A 1 kb window around the TSS is indicated. B Model-based Analysis for ChIP-Seq 2 (MACS2) scores for the indicated NF-κB subunits within 1 kb of the MCL1 promoter across all ChIP-Seq datasets reported in ChIP Atlas Zou et al. 2024 [53], after outlier removal using ROUT method (q = 1%, ****P < 0.0001). C Schematic of normal basal NF-κΒ signaling (left), schematic of NF-κB signaling predicting the unknown effect of IκΒε^–/– on cRel and MCL-1 induction. D Computational modeling simulation of RelA and cRel abundance in IκBε^–/– and IκB^ακB/κB demonstrated as fold changes normalized to wild-type. Error bars represent the mean and ± standard deviation of 25 cells (***P < 0.001, unpaired ttest). E RelA and cRel abundances in IκBε^–/– and IκBα^κB/κB cells derived from primary splenocytes, demonstrated as fold changes to wild-type, error bars representing the mean ± standard deviation. F Experimental pipeline showing the workflow of isolating and purifying primary B cells from wild type, IκBε^–/^– and IκBα^κB/κB mouse genotypes. G Abundances of MCL1 and BCLXL acquired with flow cytometry in primary B cells isolated and purified from IκBε^–/^– (left) and IκBα^κB/κB (right) mouse genotypes, demonstrated as fold changes normalized to wild-type. Error bars represent the mean ± standard deviation of three independent experiments (*P < 0.05, unpaired ttest).

Fig. 8. Targeting the NF-κB inducing kinase (NIK) to overcome tumor microenvironment (TME) resistance in Diffuse large B cell lymphoma (DLBCL).

A Overview of signaling between TME, NF-κB, and BCL2-family proteins indicating interactions revealed here with dashed lines, and NIK-inhibitor Amgen16 displayed in red. B Schematic of the co-culture system used, following 24 h of treatment with 0.0001–100 μΜ of venetoclax (ABT199) or A1331852 ± 50 μM of the NIK inhibitor Amgen16. C Western blot analysis of p100 and p52 levels in RIVA cells under monoculture, post a 24-h hCD40-3T3 co-culture or post a 24-h hCD40L-3T3 co-culture with the addition of 50 = μM of Amgen16. Quantification of p52:p100 ratio from western blot data, normalized to total protein across conditions. Bar graphs depict the p52:p100 ratio of two independent experiments. Uncropped blots are presented in the supplementary material. D Fold change to monoculture in MCL1 and BCLXL protein levels in RIVA cells under the indicated conditions. Data points represent paired measurements from individual biological replicates. Statistical analysis was performed using paired T tests with significance indicated as *P < 0.05, ns = not significant. E Cell viability of RIVA cells in response to 0.0001–100μΜ of the BCL-2 inhibitor ABT199 following a 24-h treatment under three conditions: monoculture, co-culture with hCD40L-3T3 cells, and co-culture with the addition of 50 μM of the NIK inhibitor Amgen16. LC₅₀ values for each condition are shown (right) with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control under four conditions: monoculture, co-culture with hCD40L-3T3 cells, co-culture with the addition of 50 μM of the NIK inhibitor Amgen16, and co-culture with the addition of 0.001 μM of the NIK inhibitor CW15337 (*P < 0.05, **P < 0.01, one-way ANOVA with Tukey’s comparisons test). F Cell viability of U2932 cells in response to 0.0001–100 μΜ of ABT199 following a 24-h treatment under three conditions: monoculture, co-culture with hCD40L-3T3 cells, and co-culture with the addition of 50 μM of the NIK inhibitor Amgen16. LC₅₀ values for each condition are shown (right) with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (*P < 0.05, **P < 0.01, unpaired Ttest). G Schematic representation of the proposed mechanism by which NIK inhibition via Amgen16 disrupts CD40-mediated NF-κB crosstalk, thereby restoring sensitivity to A1331852 (BCLXL inhibitor) by reducing BCLXL and MCL1 levels in the presence of TME signals. H Fold change to monoculture in MCL1 and BCLXL protein levels in SUDHL8 cells cultured under co-culture and co-culture with Amgen16 conditions. Data points represent paired measurements from individual biological replicates. Statistical analysis was performed using paired T tests (*P < 0.05, **P < 0.01). I Cell viability of SUDHL8 cells in response to 0.0001–100μΜ of the BCLXL inhibitor A1331852 following a 24-h treatment under three conditions: monoculture, co-culture with hCD40L-3T3 cells, and co-culture with the addition of 50 μM of the NIK inhibitor Amgen16. LC₅₀ values for each condition are shown (right) with error bars representing the mean ± standard deviation of three independent experiments, normalized to the untreated control (***P < 0.001; one-way ANOVA with Tukey’s comparisons test).