RelA NF-κB subunit activation as a therapeutic target in diffuse large B-cell lymphoma

It has been well established that nuclear factor kappa-B (NF-κB) activation is important for tumor cell growth and survival. RelA/p65 and p50 are the most common NF-κB subunits and involved in the classical NF-κB pathway. However, the prognostic and biological significance of RelA/p65 is equivocal in the field. In this study, we assessed RelA/p65 nuclear expression by immunohistochemistry in 487 patients with de novo diffuse large B-cell lymphoma (DLBCL), and studied the effects of molecular and pharmacological inhibition of NF-κB on cell viability. We found RelA/p65 nuclear expression, without associations with other apparent genetic or phenotypic abnormalities, had unfavorable prognostic impact in patients with stage I/II DLBCL. Gene expressionprofiling analysis suggested immune dysregulation and antiapoptosis may be relevant for the poorer prognosis associated with p65 hyperactivation in germinal center B-cell–like (GCB) DLBCL and in activated B-cell–like (ABC) DLBCL, respectively. We knocked down individual NF-κB subunits in representative DLBCL cells in vitro, and found targeting p65 was more effective than targeting other NF-κB subunits in inhibiting cell growth and survival. In summary, RelA/p65 nuclear overexpression correlates with significant poor survival in early-stage DLBCL patients, and therapeutic targeting RelA/p65 is effective in inhibiting proliferation and survival of DLBCL with NF-κB hyperactivation.


INTRODUCTION
Diffuse large B-cell lymphoma (DLBCL), the most common form of aggressive non-Hodgkin lymphoma, accounts for nearly 40% of non-Hodgkin lymphomas [1]. Although most cases of DLBCL are curable with the standard immunochemotherapy regimen, rituximab plus cyclophosphamide, hydroxydaunomycin, vincristine, and prednisone (R-CHOP), 30-40% of patients have drug-resistant disease or recurrence [2]. DLBCL is a highly heterogeneous disease. Based on gene expression profiling (GEP), DLBCL can be classified into two molecular subtypes: germinal center B-celllike (GCB) and activated B-cell-like (ABC) DLBCL [3]. The ABC subtype of DLBCL typically exhibits constitutive activation of the nuclear factor-kappaB (NF-κB) pathway [4,5] and patients have inferior clinical outcomes compared with patients with GCB-DLBCL [6,7]. Recent studies have shown that NF-κB expression is not limited to ABC-DLBCL but also can occur in GCB-DLBCL [8][9][10].
Despite the well-established role of NF-κB signaling in lymphoma pathogenesis and treatment resistance, conflicting results on the prognostic significance of NF-κB and RelA/p65 expression (as a surrogate marker of NF-κB activation) in DLBCL have been reported by previous clinical studies [8,9,[34][35][36]. To help clarify the prognostic effect of RelA/p65 nuclear expression, in this study we evaluated nuclear expression of RelA/p65 by immunohistochemistry (IHC) in a large cohort of DLBCL treated with R-CHOP, and studied the prognostic effects and gene expression profiles associated B-cell lymphoma (DLBCL), and studied the effects of molecular and pharmacological inhibition of NF-κB on cell viability. We found RelA/p65 nuclear expression, without associations with other apparent genetic or phenotypic abnormalities, had unfavorable prognostic impact in patients with stage I/II DLBCL. Gene expressionprofiling analysis suggested immune dysregulation and antiapoptosis may be relevant for the poorer prognosis associated with p65 hyperactivation in germinal center B-cell-like (GCB) DLBCL and in activated Bcell-like (ABC) DLBCL, respectively. We knocked down individual NF-κB subunits in representative DLBCL cells in vitro, and found targeting p65 was more effective than targeting other NF-κB subunits in inhibiting cell growth and survival. In summary, RelA/p65 nuclear overexpression correlates with significant poor survival in earlystage DLBCL patients, and therapeutic targeting RelA/p65 is effective in inhibiting proliferation and survival of DLBCL with NF-κB hyperactivation.
with p65 nuclear expression. Moreover, we inactivated individual NF-κB subunits in vitro and investigated their differential effects on proliferation and apoptosis of DLBCL cells which highlighted the important therapeutic value of RelA/p65.

RESULTS
p65 hyperactivation has significant adverse impact in early-stage DLBCL p65 expression was evaluable in 487 DLBCL patients, including 287 men and 200 women. GCB/ABC ratio was close to 1 (243 GCB and 239 ABC). The median age of the patients in the study group was 63 years, and 58% of the study cohort had elderly age (≥60). Immunohistochemical results showed that 58% of DLBCL samples had p65 nuclear expression indicative of p65 activation at various levels ( Fig. 1A) with a mean level of 14%. p65 nuclear expression was not specific for ABC-DLBCL. In fact, the GCB-DLBCL group had a slightly higher mean level of p65 nuclear expression (16.1%) than the ABC-DLBC group (12.6%) (Fig. 1A). Table 1 showed the clinical and pathological features of the study cohort.
Low levels (10-40%) of p65 nuclear expression did not have significant prognostic impact in DLBCL (Fig. 1B). However, high p65 nuclear expression (p65 high , ≥50% tumor cells with p65 positive nuclei) correlated with significantly shorter PFS and OS durations in patients with stage I/II DLBCL and in patients with an International Prognostic Index score (IPI) ≤2 (Fig. 1B,  Fig. 2A). In contrast, in patients with stage III/IV DLBCL or an IPI >2, p65 expression was not prognostic. p65 high patients with stage I/II DLBCL had similar survival rates compared with p65 high patients with stage III/IV DLBCL (Fig. 2B).

p65 nuclear expression correlates with p50 nuclear expression in DLBCL
We found high p65 nuclear expression was significantly associated with p50 + and p50 high nuclear expression in overall DLBCL, GCB-DLBCL, and ABC-DLBCL (Table 1), suggesting the predominance of p65/p50 dimer activation via the canonical NF-κB pathway [9]. Significant association with c-Rel + nuclear expression was also found in overall DLBCL and GCB-DLBCL (p50/c-Rel is another dimer activated via the canonical pathway [37,38]). No significant association was observed between p65 high and RelB + . p65 high showed significant association with p52 + in overall DLBCL but not in either GCB or ABC subset.
Nuclear expression of p50, p52, and c-Rel did not show further prognostic effects among the p65 high patients. We did not observe associations of p65 high with any other adverse biomarkers such as TP53 mutations, MYC/BCL2 translocation, and Myc/Bcl-2 overexpression which may confound the prognostic effects [39][40][41][42]. In contrast, in the GCB but not the ABC subgroup, p65 high compared with p65 low patients less frequently had Bcl-2 overexpression (18% vs. 40%, P = 0.036).

p65 hyperactivation has significant adverse impact in patients with wild-type TP53
Cases of DLBCL with wild-type TP53 (WT-TP53) had significantly lower levels of RELA mRNA (P = 0.018, Fig. 1D) and a trend toward lower nuclear p65 levels (P = 0.11) than those with mutated TP53 (MUT-TP53), suggesting that wild-type p53 suppressed RELA NF-κB expression. Conversely, p65 antagonized p53 function as suggested by survival analysis: in WT-TP53 DLBCL, patients with p65 high expression correlated with significantly decreased PFS (P = 0.0076, Fig. 2C) and OS (P = 0.0082) rates than patients with p65 low tumors, independent of GCB and ABC cell-of-origin. However, when subdivided cohorts by disease stages, we found the prognostic impact was only significant in patients with stage I/II disease (P < 0.0001 for PFS, P = 0.0004 for OS). Also in MUT-TP53 patients with stage I/II DLBCL, positive p65 nuclear expression was associated with significant poorer survival; in contrast, opposite trends were observed in MUT-TP53 patients with stage III/IV DLBCL (Fig. 1E).

Multivariate survival analysis
Multivariate survival analysis (Cox regression) for high p65 nuclear expression with adjustments for clinical variables confirmed that p65 high was an independent adverse prognostic factor in patients with GCB-DLBCL and in patients with WT-TP53 DLBCL, but not in the overall study group, the ABC-DLBCL subgroup, or the MUT-TP53 DLBCL subgroup (Table 2).  Abbreviations: p65 high , high levels of nuclear p65; p65 low , low levels of p65 nuclear expression; LDH, lactate dehydrogenase; IPI, international prognostic index; CR, complete remission; PR, partial response; GCB, germinal center B-cell-like; ABC, activated B-cell-like; WT-TP53, wildtype TP53; MUT-TP53, mutated TP53. Some of the clinicopathologic data were not available. Percentages are calculated among cases with specific data available. Significant P values in bold.

GEP analysis suggests different signaling pathways activated in GCB-and ABC-DLBCL
To gain insight into the molecular mechanisms underlying the prognostic effects of p65 hyperactivation in DLBCL, we compared gene expression profiles of p65 high and p65 low tumors. p65 high patients showed GEP signatures compared with other DLBCL including p65 − DLBCL patients (IHC <10%), stronger in GCB-DLBCL than in ABC-DLBCL subset (Fig. 3A, Fig. 4, Tables 3-4).
In line with the unfavorable prognosis of patients with p65 high DLBCL, GEP analysis found that JUN and PTPRD (involved in cell cycle progression) were upregulated (1.43-fold and 1.31-fold respectively) whereas pro-apoptotic NOXA/PMAIP1 and BTG3 which negatively regulates proliferation and cell cycle progression were downregulated (1.62-fold and 1.45fold, respectively) in p65 high DLBCL compared with p65 low DLBCL. RBMS1 which transactivates MYC was upregulated (1.48-fold) in p65 high compared with p65 low GCB-DLBCL (Table 3). Paradoxically, antiapoptotic

AGING (Albany NY)
BIRC6, MCM10 (involved in the initiation of eukaryotic genome replication), CARS (cysteinyl-tRNA synthetase) and PA2G4 (involved in growth regulation) were downregulated in p65 high patients, and TENC1 (which inhibits AKT1 signaling) was upregulated in p65 high DLBCL.
When analyzed in GCB and ABC subtypes individually, we found such paradoxical association was limited in the GCB subset. RNF130 involved in apoptosis showed 1.81-fold upregulation in p65 high GCB-DLBCL patients. In contrast, in p65 high ABC-DLBCL, antiapoptotic BIRC5 and BCL2L2 were significantly upregulated whereas pro-apoptotic NOXA/PMAIP1 was significantly downregulated (Fig. 3B-C), in addition to the proliferative signatures (such as upregulation of genes involved in replication, transcription, translation, and metabolism) in ABC-DLBCL (Table 3). Signaling, ion channels Cell cycle, DNA metabolism, transcription and translation regulation Transport, trafficking, protein folding, chaperone  In the GEP comparison in overall DLBCL, a few immune-related genes were also found up-or down-regulated in p65 high DLBCL compared with p65 low DLBCL, including upregulation of LCP2 (lymphocyte cytosolic protein 2, involved in T cell receptor signaling, 1.27-fold) and TEK (antiinflammatory, 1.21-fold), and downregulation of UHRF1 (an epigenetic regulator promoting proliferation of immunosuppressive Treg cell, 1.48-fold downregulation) [43] in p65 high DLBCL (Table 3).

AGING (Albany NY)
These data indicated that the antiapoptotic and proproliferation function of p65 was primarily activated in ABC-DLBCL, whereas immune dysregulation might be more relevant for the significantly adverse impact of p65 hyperactivation in GCB-DLBCL. We further analyzed the expression of p65-activating upstream signals in GCB-and ABC-DLBCL. TNFRSF1A encoding a TNF-α receptor which can activate NF-κB by degrading inhibitory IkBα (canonical activation), was upregulated in the overall p65 high group than in the overall p65 low group (P = 0.0001). LPAR3 which encodes a receptor for lyso-phosphatidic acid/LPA was downregulated in p65 high DLBCL. In the ABC-DLBCL subset only, PSMB1, which encodes a 20S core beta subunit of the proteasome B-type family, was upregulated in p65 high compared with p65 low patients (suggesting canonical activation of NF-kB). Bruton tyrosine kinase gene BTK which plays an important role in BCR signaling activation (canonical activation), and TNFRSF13B which encodes the tumor necrosis factor receptor for APRIL and BAFF were significantly upregulated in p65 high ABC-DLBCL but not in p65 high GCB-DLBCL ( Fig. 3D-E). In comparison, in GCB-DLBCL but not in ABC-DLBCL, TNFRSF13C which encodes the receptor specific for BAFF (non-canonical activation), MYD88 which encodes an adapter protein essential for the Toll-like receptor (TLR) and interleukin-1 receptor signaling pathways, and MAP3K14/NIK which is involved in non-canonical activation of NF-κB were significantly upregulated in the p65 high compared with p65 low group (Fig. 3F-H).

Molecular inhibition of constitutive NF-κB activation in DLBCL cell lines
First, we examined whether specific inhibition of NF-κB was sufficient to block cell survival by overexpressing a super repressor mutant form of IκBα Signaling, ion channels Cell cycle, DNA metabolism, transcription and translation regulation Cell adhesion, cytoskeleton, collagen, extracellular matrix IncRNA genes, other function Abbreviations: p65 high , high p65 nuclear expression (immunohistochemistry results: ≥50%); p65 − , negative p65 nuclear expression (immunohistochemistry results: <10%); FDR, false discovery rate. *Upregulated genes in bold.

AGING (Albany NY)
(pCMV-IκBαM) in a representative DLBCL cell line, MS, that has been previously shown to have constitutive NF-κB activation [44]. IκBαM binds to NF-κB subunits but cannot be phosphorylated on the basis of alanine substitution for serines 32 and 36, acting as a dominant negative (DN) and thereby preventing the NF-κB subunits from translocating into the nucleus. Transient transfection (70-80 efficiency and 75% viability) of a DLBCL-MS cell line with the DN-IκBαM leads to the induction of IκBα protein level while suppressing constitutive NF-κB activation (Fig. 5A). In addition, cells over-expressing the DN-IκBαM are prone to apoptosis as demonstrated by Annexin V binding assays (Fig. 5B). This result suggests that constitutive NF-κB

AGING (Albany NY)
activation is required for the survival of this cell line.
To determine the functional significance of each NF-κB subunit on growth and survival regulation in DLBCL, we used specific validated shRNAs to selectively silence each NF-κB component individually in four representative DLBCL cell lines (two GCB-DLBCLs, two ABC-DLBCLs). These validated shRNAs inhibited endogenous NF-κB by more than 70% (Fig. 5C). Except for p52, downregulation of p65, c-Rel, and RelB protein expression with individual shRNAs inhibited DLBCL cell growth (thymidine incorporation assay), and inhibition of p65 was most effective (Fig. 5D), particularly in cell lines with mutated p53 (MS, LP and HB).

Pharmacological targeting of constitutive NF-κB activation in DLBCL cells
To evaluate the effects of pharmacological inhibition of NF-κB activation on transcription activities of NF-κB subunits and DLBCL cell growth and survival, we select-ed the proteasome inhibitor bortezomib (BZ), and the small molecule NF-κB inhibitor BAY 11-7082 (BAY-11) that selectively inhibits the phosphorylation and degradation of IκBα [45][46][47] in MS (GCB-DLBCL) cells.
To ascertain whether BZ or BAY-11 has an effect on constitutive NF-κB activation in DLBCL cells, we performed EMSA with nuclear extracts purified from BZ-or BAY-11-treated GCB-DLBCL cell line (MS). After BZ or BAY-11 treatment, NF-κB DNA-binding activity (Fig. 6A) and the level of phosphorylated-IκBα (Fig. 6B) gradually declined in a dose-dependent manner in the MS DLBCL cell line. Confocal microscopy analysis also demonstrated that BZ or BAY-11 treatment inhibits the nuclear accumulation of p50 and p65 NF-κB subunits, leading to DNA fragmentation, indicative of cells undergoing apoptosis (Fig. 6C). Next, we evaluated the effects of BZ or BAY-11 on DLBCL cell viability using in vitro proliferation assays in four representative DLBCL cell AGING (Albany NY) lines (two ABC, two GCB). Both BZ and BAY-11 inhibited cell proliferation in the representative DLBC cell lines in a dose-dependent manner (Fig. 7A). BZ and BAY-11 inhibit NF-κB by different mechanisms, as we analyzed the cell lysates from BZ-treated and BAY-11treated DLBCL-MS cells to a 20S proteasome proteolysis assay, and found proteasome activity was substantially inhibited (>50%) after 3 hours of BZ treatment, whereas BAY-11 treatment did not affect proteasome activity in a similar time points (Fig. 7B).
To determine whether the cell growth inhibition effects of BZ and BAY-11 involve their activity in the cell cycle regulation, we analyzed the cell cycle profile. As shown in Fig. 7C, in a representative DLBCL cell line (MS), both BZ-and BAY-11-treated DLBCL cells accumulated in the G0/G1 phase of the cell cycle, while cells in G2M and S phases were decreased. In addition, BZ or BAY-11 treatments in MS DLBCL cell line resulting in cells undergoing apoptosis in a timedependent manner (Fig. 7D). To verify that these cells had actually undergone apoptosis, we measured the generation of caspase 3 activities. DLBCL cells treated with BZ or BAY-11 activated caspase 3 activity after 12 hours of treatment, which can be block with a caspase 3 inhibitor (DVED) but not with a caspase 1 inhibitor (VAD) (Fig. 7E). In addition, a known caspase substrate, poly-(ADP-ribose) polymerase (PARP), was cleaved after BZ or BAY-11 treatment (Fig. 7F). These experiments provide additional and interesting insights into the putative role of NF-κB in DLBCL cell proliferation and viability maintenance.

DISCUSSION
In this study we studied the significance of RelA/p65 NF-κB in DLBCL by two ways. In the first part of this study, we evaluated the prognostic significance of RelA/p65 nuclear expression in a large cohort of de novo DLBCL treated with R-CHOP (n = 487). Although p65 nuclear expression may not be a strong prognostic marker in overall DLBCL, we found p65 hyperactivation (IHC ≥50%) had significant adverse impact on survival of patients with stage I/II DLBCL independent of cell-oforigin and TP53 mutation status, even though it was not associated with apparent genetic or phenotypic abnormalities, such as TP53 mutations, MYC/BCL2 translocation, and Myc/Bcl-2 overexpression.
The adverse prognostic significance of p65 hyperactivation was also seen in the GCB-DLBCL subtype overall and in the subset with wild-type TP53, but not in the subsets with strong unfavorable factors including advanced stages, TP53 mutations, and ABC cell-of-origin. In addition, lower levels (10-40%) of p65 nuclear expression did not have significant prognostic impact in DLBCL. This limited significance of p65 expression in DLBCL may reflect different signaling transductions pathways activating p65, different p65 NF-κB dimers, and complicated p65 functions influenced by other factors including p53 in different stimulatory signals. For example, NF-κB p65 activation induced by cytotoxic stimuli promotes apoptosis in mouse embryo fibroblasts, which contrasts with the prosurvival function of p65 induced by inflammatory cytokines [30]. In various cancer cell lines, p65 and p53 formed p65/p53 complex and bound to DNA targets; the function of p65 and fate of tumor cells are significantly affected by p53 and stress levels [32].
Others have shown that there is transcriptional and functional crosstalk between NF-κB and p53. p53 can negatively regulate NF-κB activation by regulating IKK1 expression [29] and suppressing glycolysis [28]; NF-κB and p53 antagonize each other's function in apoptosis, proliferation and tumor invasion that appears to depend on cellular context. Overall the results in this study suggested that p65 and wild-type p53 counteracted each other in DLBCL, and that the inhibition of p53 tumor suppressor function by p65 hyperactivation had a significant adverse impact on clinical outcomes.
GEP suggested that BCR, TNF, TLR, and mitogenactivated protein kinase signaling pathways were all implicated in p65 hyperactivation in DLBCL. These upstream pathways were activated preferably in ABC-DLBCL or GCB-DLBCL, and correspondingly, resulting in different downstream pathways in ABC and GCB subtypes. In ABC-DLBCL, p65 high GEP signatures were featured by proliferation and antiapoptosis, whereas in GCB-DLBCL in which subgroup p65 hyperactivation showed significant adverse prognostic impact, p65 high expression was accompanied with upregulation of some pro-apoptosis genes as well as many immune genes. Upregulation of LILRB1 and CD163 in p65 high patients suggested immune suppression and dysregulation, which may contribute to the associated poor prognosis.
In the second part of this study, we tested whether NF-κB p65 subunit in particular is a potential molecular target by the proteasome inhibitor bortezomib and the small molecule NF-κB inhibitor BAY-11 in vitro.
Previous studies have shown that proteasome inhibitors have better antitumor efficacy in patients with ABC-DLBCL than in patients with GCB-DLBCL, probably due to higher p65 expression in the ABC subtype [45,48,52]. Consistently, our GEP analysis also found the proteasome gene PSMB1 was upregulated in ABC-DLBCL. Our in vitro experiments found that these anti-NF-κB agents can effectively inhibit p65 protein expression and DNA binding activity, leading to cell cycle arrest, decreased cell proliferation, and apoptosis induction in both GCB and ABC types of p65overexpressing DLBCL cell lines. Intriguingly, representative DLBCL cell lines with mutated p53 are more sensitive to p65 shRNA targeting approach as compared to a cell line with wild-type p53, opposite the AGING (Albany NY) prognostic effects observed in the DLBCL study cohort. These findings may suggested that although p65 subunit only manifested prognostic significance in certain DLBCL subsets due to the complexity of NF-κB dimers and activating mechanisms, in vitro experiments nonetheless demonstrated that NF-κB overexpressing DLBCL cells were addictive to NF-κB and vulnerable for NF-κB inhibitors. This vulnerability of DLBCL cells was also apparent in the context of mutated p53; p65 may have an important role in the oncogenic activities of mutated p53 in DLBCL. Importantly, in vitro p65 subunit stood out as a critical factor in controlling cell growth and survival and showed the most sensitivity to molecular and pharmacological inhibition of NF-κB activation. Therefore, our current study in both patient samples and DLBCL cell lines provided additional insights into the putative roles of NF-κB p65 in immune regulation, DLBCL cell proliferation, and viability maintenance, and the utility of p65 as a biomarker to stratify DLBCL patients to receive alternative therapeutic regimens including agents targeting NF-κB [52]. However, these findings warrant further investigation and validation in more representative DLBCL cell lines as well as primary DLBCL cells.
In summary, we provide clinical and experimental data that RelA/p65 NF-κB has prognostic and therapeutic value in DLBCL. High p65 nuclear expression is a significant adverse biomarker in patients with earlystage (I/II) DLBCL. Pharmacological p65 inactivation effectively inhibited cell growth and survival in both GCB-DLBCL and ABC-DLBCL cell lines with p65 hyperactivation.

Immunohistochemical staining
Immunohistochemistry for p65 and other NF-κB subunits using specific antibodies (Abcam, Cambridge, MA) was performed on tissue microarrays of formalinfixed, paraffin-embedded lymphoma samples using methods described previously [10,49]. The immunohistochemical stains were assessed in 10% increments by three pathologists blinded to the clinical outcomes. Disagreements about the percentage of positive cells were resolved by joint review at a multi-headed microscope.

Gene expression profiling
GEP was performed using the Affymetrix GeneChip Human Genome U133 Plus 2.0 array (Santa Clara, CA) and CEL files were deposited in the NCBI Gene Expression Omnibus repository (GSE#31312) [49]. GEP were available for 444 DLBCL patients of this study cohort with high or low levels of p65 nuclear expression. The P values for differential expression obtained via multiple t-tests were corrected for false discovery rates using the beta-uniform mixture method.

Transfection
Transfection experiments in DLBCL cells with validated green fluorescent protein (GFP)-shRNAs were performed in vitro in representative transfectable DLBCL cells, using the Neon transfection system from Invitrogen (Life Technologies Corporation, Grand Island, NY) as described previously [44], and were repeated at least twice to verify reproducible experimental results. Twenty-four hours after transfection, GFP + cells were sorted by a fluorescenceactivated cell sorter (FACS) flow cytometer and plated. Cell proliferation was measured 96 hours after sorting by thymidine incorporation assays, while some cells were lysed for Western blot analysis for NF-κB subunit inhibition. A set of four shRNA plasmids for each NF-κB subunit was tested and the optimal (>75%) gene knock-down shRNA plasmid was selected.
Therapeutic NF-κB inhibition experiment DLBCL cells were treated with increasing doses of bortezomib (0-50 nM), or organic compounds BAY 11-7082/BAY-11 (1µM) for 6-48 hours and subjected to cell proliferation assays, electromobility gel shift assay (EMSA), immunofluorescence, apoptosis detection assay, and cell cycle analysis according to the manufacturer's instructions or procedures as previously described [51]. Data are representative of three independent experiments.

Thymidine incorporation assays
In vitro thymidine incorporation proliferation assays were performed as described previously 1