Neuroblastoma is the second most common tumor in children. The cause of neuroblastoma is thought to lie in aberrant development of embryonic neural crest cells and is accompanied by low MHC-1 expression and suppression of the NF-κB transcription factor, thereby gearing cells toward escape from immunosurveillance. Here, we assess regulation of the MHC-1 gene in neuroblastoma to enhance its immunogenic potential for therapeutic T-cell targeting. A genome-wide CRISPR screen identified N4BP1 and TNIP1 as inhibitory factors of NF-κB-mediated MHC-1 expression in neuroblastoma. Patients with advanced stage neuroblastoma who expressed high levels of TNIP1 and N4BP1 exhibited worse overall survival. Depletion of N4BP1 or TNIP1 increased NF-κB and MHC-1 expression and stimulated recognition by antigen-specific CD8+ T cells. We confirmed that TNIP1 inhibited canonical NF-κB member RelA by preventing activation of the RelA/p50 NF-κB dimer. Furthermore, N4BP1 inhibited both canonical and noncanonical NF-κB through binding of deubiquitinating enzyme CEZANNE, resulting in stabilization of TRAF3 and degradation of NF-κB-inducing kinase NIK. These data suggest that N4BP1/CEZANNE or TNIP1 may be candidate targets for immunotherapy in neuroblastoma tumors and should lift NF-κB suppression, thereby triggering increased peptide/MHC1-mediated tumor reactivity to enhance therapeutic T-cell targeting.
Aberrant regulation of NF-κB and MHC-1 in neuroblastoma tumors provides new targets for immunotherapeutic approaches against neuroblastoma.
Neuroblastoma accounts for 15% of pediatric cancer mortality (1–3). It is the most common solid tumor in children with survival rates of merely 40% in advanced stage (stage 4) disease (1). Neuroblastoma is an embryonic cancer that originates from neural crest cells. The mutational burden of neuroblastoma tumors is low (4, 5), hence so far, no neo-antigens have been reported for neuroblastoma. Instead, expression of cancer/testis antigens have been described in neuroblastoma (6–8) reflecting the embryonically undifferentiated status of this type of tumor.
Immunotherapy for stage 4 neuroblastoma gains traction. Currently, antibody therapy targeting the neuroblastoma-specific surface disialoganglioside GD2 is standard protocol succeeding radiotherapy, chemotherapy, surgery, and autologous stem cell transplantation. In addition, various immunotherapies are in trial stages (9, 10). However, due to its low immunogenicity (11, 12), neuroblastoma tumors continue to pose a challenge for immunotherapeutic approaches. Neuroblastoma typically presents with extremely low levels of MHC class 1 (MHC-1), especially in stage 4 patients (7), preventing recognition and lysis of neuroblastoma cells by cytotoxic T lymphocytes (CTL).
Unique for neuroblastoma is the abnormally low levels of NF-κB protein that have been reported (13, 14), which sets neuroblastoma aside from all other solid tumors in which NF-κB is usually constitutively active (15). MHC-1 levels in neuroblastoma are, however, controlled by NF-κB (16). Activation of NF-κB induces transient MHC-1 membrane expression (13, 14, 17) that facilitates recognition by CTLs in vitro (13, 18–20), underscoring that reconstitution of MHC-1 by activation of the NF-κB pathway in these tumors might increase their immunogenicity.
In this study, we aimed to identify target proteins that control peptide/MHC-1 display in neuroblastoma by performing a genome-wide CRISPR knockout screen combining NF-κB activity and cell-surface MHC-1 membrane expression. We show that both canonical and noncanonical NF-κB may contribute to neuroblastoma immunogenic display and give mechanistic insight into the inhibitory mechanisms of these target proteins.
Materials and Methods
Cells and culture conditions
Neuroblastoma cell lines GIMEN, Sk-N-SH, SH-Sy5y Sk-N-BE and IMR32 were obtained via the Academic Medical Center of Amsterdam. Melanoma cell line BLM was obtained via the Radboud University Medical Center of Nijmegen. All cells were maintained in DMEM GlutaMAX supplemented with 10% FCS (Biowest), 50 U/mL penicillin (Life Technologies) and 50 μg/mL streptomycin (Life Technologies) and cultured at 37°C at 5% CO2. Cell authentication was conducted through microarray analysis, performed at AMC (NCBI GEO Accession No. GSE 16476; http://www.ncbi.nlm.nih.gov/geo/) confirming neuroblastoma expression profiles, which were confirmed for selected neuroblastoma markers (LIN28B, MYCN, GD2, CD56, and CD81) by qPCR and/or FACS analysis when cells arrived at UMC Utrecht (2012). Cells were frozen in individual aliquots after less than 10 passages and tested regularly for Mycoplasma contamination.
Reagents and antibodies
Human recombinant TNF was purchased from Miltenyi. Decitabine, LPS and poly(I:C) were purchased from Sigma-Aldrich. Hoechst was purchased from Thermo Fisher Scientific. A list of all antibodies used can be found in the Supplementary Material.
Constructs and plasmids
For generating NF-κB reporter cells, a pTRH1 plasmid was used containing an NF-κB transcriptional response element (TRE) and a minimum CMV (mCMV) promoter followed by eGFP gene and upstream of the blasticidin S resistance gene (BSR) from Bacillus cereus (21). Clonal cell lines GIMEN ΔN4BP1 and ΔTNIP1, used in all follow-up experiments, were analyzed by TA-cloning (Strataclone PCR cloning kit, agilent #240205). At least 20 colonies were picked and subjected to sequencing to ensure no allele would be unannotated. For non-clonal cell lines, PCR product was subjected to sanger sequencing to obtain mutation profiles based on TIDE assay (22), using nested sanger sequencing primers Fw_N4BP1_seq: 5′-CTACAGCTCTCCAAGTGAAACATTTGAG-3′ and Fw_TNIP1_seq: 5′-CATTCTCCCACACTCGAGTGG- 3′. Information about gRNA sequences can be found in the Supplementary Material.
Cells were detached with trypsin (Life Technologies) and washed twice in PBS containing 2% FCS (Biowest) and 0.1% sodium azide (NaN3, Sigma-Aldrich). Next, antigen nonspecific binding was prevented by prior incubation of cells with 10% mouse serum (Fitzgerald). Cells were incubated with indicated anti-human antibodies for 30 minutes on ice and washed twice. In case of intracellular stainings, cells were fixed and permeabilized (Fix/Perm kit BD Biosciences) after surface staining and subsequently stained with indicated antibodies for 30 minutes on ice and washed twice. Cells were taken up in PBS (2% FCS/0,1% NaN3) to be used for measurement using either FACSCanto II or Fortessa with FACS Diva Version 6.13 (BD Biosciences). Analysis was done using FlowJo Version 10 software.
CRISPR screen analysis
An overview and detailed documentation of all commands and code used for the analysis can be found in the following Github repository: https://github.com/ElmerS/spel_et_al_2018. A summary of the procedure is written below. Sequence reads were uploaded to NIH Sequence Read Archive with access number: SRP147989.
Sequence read analysis
Raw sequence reads were extracted from the FASTQ files using AWK and subsequently, using regular expression in SED, sequences of 19-21b flanked by ACCG and GTT were distilled. Bowtie 1 (23) was used to align the reads to the lentivirus library in two consecutive rounds. In the first round, a single mismatch was allowed while demanding alignment to only a single guide. In the second round, the reads that aligned to multiple guides in the first round where re-aligned with a more stringent setting of zero mismatches to see if they could be aligned unambiguously.
Scoring sgRNA using differential expression analysis
Readcount files were generated from Bowtie output using a Bash script and differential expression analysis was carried out using the R package DESeq2 (24). The NF-κBlowMHClow population from both screens were used as replicates for the low population and the NF-κBhighMHChigh populations as replicates for the high population. Differences in guide abundance between the low and high populations were tested using a Walt test with default size factor normalization of DESeq2. Shrinkage of Log2FoldChanges was applied. Guides with an FDR-corrected (Benjamini-Hochberg) P value that met the threshold of 0.05 were designated significant.
Candidate gene identification
Using a custom Python script, the number of guides per gene with a FDR corrected P value below or equal to the threshold was determined. Genes having at least three guides that met this threshold and which also showed identical directionality were selected as candidates. In total 9 genes fulfilled these criteria. To include the nontargeting control guides in this analysis, 167 hypothetical genes were randomly assigned 6 nontargeting control guides. For follow-up, those genes were selected that showed at least 2 significant gRNA hits above 500 reads.
Cell were cultured in LabTek 8 well chambers in culture medium. Cells were subsequently washed and fixed for 10 minutes at RT using PBS supplemented with 4% formaldehyde. Cells were permeabilized and blocked for 60 minutes with PBS supplemented with 5% BSA and 0,3% Triton-X100. Next, slides were incubated with primary antibodies diluted in PBS supplemented with 1% BSA and 0.3% Triton-X100 for 16 hours. Cells were washed with PBS and incubated with secondary antibodies for 1 hour. After washing, slides mounted on a microscope slide using Mowiol 4-88 (Carl Roth, Germany). Samples were imaged using a Zeiss LSM710 confocal microscope equipped with a Plan-Apochromat 63 × 1.40 oil differential interference contrast M27 objective (Zeiss). Fluorescent signals were detected with PMTs set at the appropriate bandwidth using 633 nm helium neon laser for Alexa647. Images were processed using the Zeiss Enhanced Navigation (ZEN) 2009 software.
Quantitative real-time PCR
Total RNA was isolated using RNA isolation columns (Qiagen) according to the manufacturer's instructions. cDNA was synthesized from 1 μg of total RNA using the iScript cDNA synthesis kit (Bio-Rad). Real-time PCR was performed using IQ SYBR Green PCR Supermix (Bio-Rad) and the CFX96 Touch Real-Time PCR Detection System (Bio-Rad), according to the manufacturer's instructions. PCR assays were done in triplicate. Data were calculated as values relative to GAPDH and further analyzed using GraphPad Prism 7.02.
T-cell activation assay
A HCMV pp65-specific CD8+ T-cell clone was prepared previously. In brief, T cells from an HLA-A*0201+ donor were stained with HLA-A2/pp65495-503 tetramers, and subsequently single-cell sorted in a 96-well plate (Thermo) containing irradiated B-LCL feeder cells (1 × 105 cells/mL, irradiated with 70 Gy) and PBMCs from 3 healthy donors (1 × 106 cells/mL, irradiated with 30 Gy). One μg/mL leucoagglutinin PHA-L (Sigma-Aldrich) and 120 U/mL of recombinant IL-2 (Immunotools) were added. T-cell clones specific to pp65495-503 were selected using tetramer staining. Positive clones were restimulated and expanded during several stimulation cycles and frozen in aliquots that were freshly thawed before each use in an assay. Neuroblastoma cells were cocultured with HCMV pp65-specific CD8+ T cells 4 to 5 hours in the presence of Golgistop (1/1,500; BD Biosciences). Cells were subsequently stained for surface markers and presence of intracellular IFNγ and TNF, followed by flow cytometry–based analysis.
Cells were lysed in sample buffer, separated by SDS-PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore) by wet immunoblotting. Membranes were blocked with 4% nonfat milk powder (Campina) in PBS 0.1% Tween-20 (Bio-Rad) for 1 hour and subsequently incubated with a primary antibody for 16 hours at 4°C. After washing, membranes were incubated with HRP-coupled secondary antibody for 1 hour followed by analysis on chemidoc imaging system (Bio-Rad).
Cells were lysed in PBS(1% CHAPS, 30 mmol/L Tris-HCL pH 8.0, 150 mmol/L NaCl) for 30 minutes on ice and spun down at 13,000 × g for 15 minutes. Supernatants were incubated with uncoupled or antibody-coupled beads (Protein A-agarose beads, Roche) for 16 hours under constant agitation. Beads were separated from lysates by centrifugation, washed twice and resuspended in sample buffer. Protein interactions were analyzed using immunoblotting (see above).
Immunoprecipitation beads were heated in sample loading buffer for 5 minutes. at 95°C, eluates were run into the stacking of a 4% to 12% Bis-Tris gel and coomassie-stained bands were excised. Proteins were reduced with 6.5 mmol/L DTT, alkylated with 54 mmol/L iodoacetamide and digested in-gel with trypsin (Gold, mass spectrometry grade, Promega, 3 ng/μL) overnight at 37°C. Extracted peptides were vacuum dried, reconstituded in 10% formic acid and analyzed by nanoLC-MS/MS on an Orbitrap Fusion Tribrid mass spectrometer equipped with a Proxeon nLC1000 system (Thermo Scientific). Peptides were loaded directly on the analytical column (Agilent Poroshell EC-C18 120 2.7 μm, 50 μm × 500 mm, packed in-house) and separated in a 140-minutes gradient containing a 124-minutes linear increase from 6% to 30% solvent B (0.1% formic acid/80% acetonitrile), with 0.1% formic acid/water as Solvent A. Further settings were as described previously (25).
Raw data were analyzed by MaxQuant (version 188.8.131.52; ref. 26) using the label-free quantitation (LFQ) algorithm and the “match between runs” tool, with default settings. MS/MS data were searched against the human Swissprot database (20,197 entries, release 2016_09) complemented with a list of common contaminants and concatenated with the reversed version of all sequences. Trypsin/P was chosen as cleavage specificity allowing two missed cleavages. Carbamidomethylation (C) and oxidation (M) were set as fixed and variable modification, respectively. LFQ intensities were log2-transformed in Perseus (version 184.108.40.206; ref. 27), after which, proteins were filtered for at least two valid values (out of 3 total). Missing values were replaced by imputation based on a normal distribution using a width of 0.3 and a downshift of 1.8. P values of all quantified proteins were calculated in a Student t test and plotted against the log2 (bait/control) LFQ abundance difference in a Volcano plot. Proteins were considered enriched by immunoprecipitation when they met the criteria P < 0.05 and log2 LFQ[bait/control] ≥ 2.
RNA seq dataset (GSE62564) was selected from the Gene Expression Omnibus and analyzed through the Genomics Analysis and Visualization Platform. Kaplan–Meier estimates of overall survival were determined using automatically generated expression cutoff values based on statistical significant differences using log-rank test.
All statistical analyses were performed using Graphpad Prism 7.02.
Low expression of MHC-1 and NF-κB in neuroblastoma
Children of merely 3-months-old can present with neuroblastoma tumors, which advanced the idea that this type of tumor arises from erroneous embryonic development rather than external mutagenic factors. Therefore, considering the early developmental origins of these tumors, neuroblastoma might have originated from precursors that are yet to develop immunological features such as MHC-1, which is conceptionally different from tumors that downregulate MHC-1 as immune evasion strategy. To gain insight into MHC-1 gene expression and its regulation in neuroblastoma, we analyzed expression datasets of neuroblastoma tumors, healthy neural crest cells and healthy mature nerve tissue for MHC-1 and its transcription factors. Although MHC-1 expression in neuroblastoma was indeed lower than healthy mature nerve tissue, we found no differences in MHC-1 gene expression between neuroblastoma tumors and healthy neural crest cells (Fig. 1A). Neuroblastoma appears to reflect the underdeveloped state of the neural crest in which MHC-1 expression is still low, but can be evoked.
Transcriptional control of MHC-1 is well understood as various transcription factors have been identified that establish MHC-1 in human cells, including NF-κB, NLRC5 and IRF1. Overall, these MHC-1 transcription factors were not highly expressed in neuroblastoma (Fig. 1B; Supplementary Fig. S1). NLRC5 and IRF1 were not highly expressed in neural crest cells either, as we found similar expression levels between neuroblastoma tumors and neural crest cells (Supplementary Fig. S1). In contrast, gene expression of multiple NF-κB family members (RelA, RelB, NFKB1, NFKB2, and Rel) was higher in neural crest cells than in neuroblastoma tumors. In particular, expression of RelA, RelB, and NFKB1 was significantly decreased in neuroblastoma tumors (Fig. 1B). Also compared with mature nerve tissue, RelA and NFKB1 expression was reduced. As neural crest cells differentiate toward mature nerve cells, the MHC-1 locus becomes available for transcription causing the developmental upregulation of MHC-1. Although in healthy cells NF-κB expression stimulates de novo MHC-1 transcription, in neuroblastoma NF-κB expression is suppressed causing a lack in the normal developmental MHC-1 upregulation. Gene expression of MHC-1 alleles correlated with NFKB gene expression in stage-4 neuroblastoma tumors (Fig. 1C; Supplementary Fig. S1), indicating that MHC-1 expression can be increased in neuroblastoma by NF-κB. Of note, NF-κB family member Rel was upregulated in neuroblastoma, which is in agreement with its role in antiapoptotic signaling (28, 29).
Thus, the abnormal development from healthy neural crest cell to malignant neuroblastoma coincides with downregulation of NF-κB family members. In line with this observation, several studies have shown that transcriptional activity of NF-κB in neuroblastoma cells is indeed repressed (13, 14, 17). These data together support that reversal of repressed NF-κB might steer neuroblastoma away from tumorigenic development and possibly unlock MHC-1 expression for immunogenic display.
Identification of NF-κB–dependent MHC-1 suppressors in neuroblastoma
NF-κB is repressed in neuroblastoma on both gene expression level (Fig. 1B) as well as activation level (13, 14, 17). Activation of NF-κB can temporarily restore MHC-1 expression in neuroblastoma, suggesting that NF-κB signaling is repressed by unknown mechanisms, leading to low immunogenicity of neuroblastoma tumors. To establish stable MHC-1 expression in neuroblastoma, we set out to identify the factors that negatively regulate NF-κB–mediated MHC-1 expression in neuroblastoma.
To this end, we generated a clonal cell line bearing an NF-κB reporter, which upon activation of the NF-κB pathway expresses eGFP (Fig. 1D). Indeed, elevated MHC-1 levels were observed upon stimulation of the NF-κB pathway using TNF or TLR ligands. These cells were mutagenized using a genome-wide CRISPR/Cas9 knockout library containing 118.000 individual gRNAs with an average of 6 gRNAs per gene (30). The mutagenized cells were expanded and stained for cell surface expression of MHC-1. NF-κBnegMHCneg and NF-κBposMHCpos populations were isolated using FACSsorting. Next-generation sequencing was used to identify gRNAs enriched for each population (Fig. 1E).
Each individual gRNA was traced back and scored for its enrichment in either the NF-κBnegMHCneg or the NF-κBposMHCpos population (Fig. 1F; Supplementary Fig. S2). As 6 different gRNAs were used per gene, the scores of all 6 gene-specific gRNAs combined were used to calculate significant gene hits. This resulted in a total of 9 genes that were identified as candidate regulators of NF-κB activity and MHC-1 expression in neuroblastoma cells (Supplementary Table S1), including 4 positive regulators and 5 suppressors. We selected Nedd4-binding protein 1 (N4BP1), TNFAIP3-interacting protein 1 (TNIP1), cylindromatosis (CYLD), and TNF-receptor–associated factor 2 (TRAF2) as candidate suppressors of NF-κB and MHC-1 in neuroblastoma cells shown in Fig. 1F (See Materials and Methods for selection criteria). The expression of N4BP1, TNIP1, and TRAF2, but not CYLD, correlated with worse survival in patients with stage-4 neuroblastoma (Fig. 1G), suggesting an important role for NF-κB suppression in malicious outcome of neuroblastoma.
CYLD and TRAF2 are well-established regulators of NF-κB signaling. Whereas CYLD inhibits NF-κB upon receptor signaling (31, 32), TRAF2 maintains an inactive NF-κB in steady-state conditions (33), but not upon signaling. Furthermore, the inhibitory effect of TNIP1 on NF-κB is shown to impede canonical NF-κB (34, 35). Moreover, N4BP1 has been implicated in NF-κB inhibition (36); however, the mechanism remains elusive.
These results suggest that upon depletion of N4BP1 or TNIP1, NF-κBnegMHCneg neuroblastoma cells transform into cells endowed with stable expression of MHC-1 and an active NF-κB transcription factor.
Depletion of NF-κB suppressors enhances MHC-1 in neuroblastoma
Using CRISPR/Cas9, we generated N4BP1- and TNIP1-depleted neuroblastoma reporter clones and empty vector-transduced wild type (WT) reporter cells as controls (Supplementary Fig. S3). NF-κB activity and MHC-1 expression were measured by flow cytometry. Depletion of N4BP1 or TNIP1 resulted in activation of the NF-κB reporter (Fig. 2A). Activation of NF-κB was also shown by confocal microscopy of WT and ΔN4BP1 and ΔTNIP1 neuroblastoma cells (Fig. 2B). Total RelA protein localized to the cytosol in WT cells, but spontaneously translocated to the nucleus in ΔN4BP1 and ΔTNIP1 cells. Nuclear translocation requires phosphorylation of RelA protein, indeed, phospho-RelA was only detected in ΔN4BP1 and ΔTNIP1 cells and not WT neuroblastoma cells. In addition, ΔN4BP1 and ΔTNIP1 cells showed increased surface levels of MHC-1 compared with WT cells (Fig. 2C), corroborating the findings of the CRISPR screen. Moreover, inversed correlation between N4BP1 and MHC-1 gene expression could be observed in 46 patients (Supplementary Fig. S4) and associated with a significant difference in survival probability, with high expression of N4BP1 relating to worse survival (Fig. 2D). For 42 patients an inversed correlation between TNIP1 and MHC-1 gene expression was observed, which also resulted in worse survival for TNIP1high-expressing tumors.
To investigate whether the inhibitory effects of N4BP1 and TNIP1 on NF-κB and MHC-1 are specific for neuroblastoma, we generated a panel of N4BP1 and TNIP1 mutants in a range of human cell lines (Supplementary Fig. S5). MHC-1 expression levels were measured in WT and knockout cells using flow cytometry (Fig. 2C). Depletion of N4BP1 or TNIP1 generally resulted in increased MHC-1 membrane expression across five neuroblastoma cell lines. Differences in gene expression of either HLA-A or HLA-B genes were measured using qPCR (Fig. 2E). Interestingly, in non-neuroblastoma cell lines HEK293T and BLM, MHC-1 upregulation was not clearly different between WT and knockout cells, suggesting that the high steady-state expression levels of MHC-1 in HEK293T and BLM make them less prone for increased MHC-1 expression. However, immunoblot analysis of ΔN4BP1 and ΔTNIP1 cells of HEK293T and BLM shows increased levels of phospho-RelA, indicating that NF-κB is activated (Supplementary Fig. S6). These results suggest that NF-κB inhibition by N4BP1 and TNIP1 is conserved across multiple cell types, but its effect on MHC-1 levels is cell-type specific.
Antitumor T-cell reactivity against NF-κB–active neuroblastoma cells
Expression of peptide/MHC-1 complexes on the neuroblastoma cell surface is essential to T-cell recognition and possible T-cell–based immunotherapy (13, 19, 20). We next addressed whether increased MHC-1 expression in ΔN4BP1 and ΔTNIP1 neuroblastoma cells is sufficient to trigger functional antigen presentation and T-cell reactivity. results in functional antigen presentation and T-cell reactivity, and therefore performed co-culture experiments using neuroblastoma cells and MHC-matched antigen-specific CTLs. To test this, we generated WT and knockout neuroblastoma cells that stably express the model antigen pp65, and exposed these cells to MHC-1–restricted pp65-specific CTLs in coculture experiments. We found that although CTLs were activated upon coculture with ΔN4BP1 or ΔTNIP1 cells, WT cells failed to do so (Fig. 3A). Indicative of T-cell activation and granule release, the CTLs produced IFNγ and showed LAMP1 surface staining, respectively.
Demethylating agent decitabine (DAC) has been shown to increase MHC-1 expression and is currently being scrutinized in several clinical trials for patients with neuroblastoma (37–39). Treatment with decitabine indeed increased MHC-1 levels on neuroblastoma cells to a similar extent as ΔN4BP1/TNIP1 depletion (Fig. 3B). The combination of DNA demethylation with active NF-κB transcription shows a 2-fold higher induction of MHC-1 surface expression. Such a synergistic effect is also observed when testing the T-cell reactivity against decitabine-treated knockout cells. T-cell recognition of ΔN4BP1 and ΔTNIP1 cells increased by 7-fold and 5-fold, respectively, when treated with decitabine (Fig. 3A). Thus, active NF-κB makes neuroblastoma cells immunogenic for T-cell reactivity. These results highlight the significance of NF-κB modulation for anti-neuroblastoma immunotherapies, alone or as adjuvant treatment to currently tested protocols.
Differential NF-κB activation in ΔN4BP1 and ΔTNIP1 cells
TNIP1 modulates NF-κB activity by destabilizing the IKK complex (40) with a consequential sequestering of RelA by IκB, less nuclear translocation and diminished transcription of target genes including MHC-1. How N4BP1 modulates NF-κB activity remains unclear. To investigate which NF-κB members play a role in ΔN4BP1 cells, we looked at the expression levels of the five transcriptionally active subunits (phospho-RelA, RelB, p50, p52, and c-Rel) of NF-κB by immunoblot. Compared with WT, ΔN4BP1 cells showed increased levels of phosphorylated RelA, of RelB and p52 (Fig. 4A). This is different from ΔTNIP1 cells that only showed increases in p-RelA and RelB, but no effects on the other subunits. Next, we tested whether these subunits also contribute to NF-κB-mediated gene expression by disrupting each subunit in the ΔN4BP1 or ΔTNIP1 neuroblastoma reporter cells using CRISPR/Cas9. We designed gRNAs against the NF-κB genes RelA, RelB, NFKB1 and NFKB2 and used these to transduce WT, ΔN4BP1 and ΔTNIP1 cells. NF-κB reporter activity was measured seven days after transduction. In ΔN4BP1 cells, NF-κB reporter activity was decreased when RelA or NFKB2 (= p52) was depleted (Fig. 4B). In ΔTNIP1 cells, NF-κB reporter activity was decreased only when RelA was depleted. In contrast, RelB depletion did not affect reporter activity in either ΔN4BP1 or ΔTNIP1 cells, despite increased RelB expression.
Thus, although N4BP1 and TNIP1 both inhibit NF-κB activity, they appear to do so using different mechanisms. In ΔN4BP1 cells both RelA and p52 contribute to NF-κB-mediated gene transcription, indicating that N4BP1 inhibition broadly affects both canonical and noncanoncial NF-κB subunits. In ΔTNIP1 cells on the other hand, only canonical RelA appears to be the driver of NF-κB-mediated gene activation.
N4BP1 coordinates degradation of NF-κB–initiating kinase NIK
To gain more insight into the mechanism by which N4BP1 inhibits both p-RelA and p52, we decided to seek out the binding partners of N4BP1. To this end, a protein-protein interaction experiment was performed combining antibody-mediated pulldown with mass spectrometry. WT neuroblastoma cells were transfected with His-tagged N4BP1-expressing constructs and after 48 hours lysed for coimmunoprecipitation experiments using anti-His beads. Mass spectrometry was used to identify proteins bound to N4BP1-His.
A variety of binding partners of N4BP1 were identified and among them, one in particular caught our attention. The de-ubiquinating enzyme CEZANNE was significantly enriched in N4BP1-His coimmunoprecipitates (Fig. 5A). As observed for N4BP1, also CEZANNE expression levels correlated with worse survival in patients with stage-4 neuroblastoma (Fig. 5B). Mechanistically, CEZANNE stabilizes TRAF3 by removing TRAF3-bound ubiquitin modifications, thereby controlling NF-κB activation (41). Stabilized TRAF3 takes part in a protein complex that retains NIK and facilitates ubiquitination and degradation of NIK. Conversely, the degradation of TRAF3 causes NIK to accumulate and phosphorylate down-stream targets thereby initiating p100 cleavage into p52 and activation of noncanonical NF-κB. In addition, TRAF3 degradation was shown to also activate the canonical pathway involving RelA (42).
To confirm that N4BP1 is involved in stabilizing TRAF3 and break-down of NIK, the protein levels of TRAF3 and NIK were measured in WT and ΔN4BP1 cells using immunoblot (Fig. 5C and D). In ΔN4BP1 cells, we detected decreased levels of TRAF3 protein compared with WT cells, suggesting that TRAF3 is degraded when N4BP1 is absent. In addition, higher levels of NIK kinase are detected in ΔN4BP1 cells compared with WT cells, indicating that NIK accumulates in absence of N4BP1. Given that TRAF3 degradation and subsequent NIK accumulation leads to processing of p100 into p52, these results could explain the spontaneous activation of noncanonical NF-κB in ΔN4BP1 neuroblastoma cells.
TRAF3 was shown to also control the canonical pathway through attenuating phosphorylation and degradation of the Inhibitor of NF-κB (IκB; ref. 42) and keeping RelA protein sequestered. Therefore, to determine a possible role of IκB in ΔN4BP1 cells, we measured the levels of IκB bound to RelA using immunoprecipitation followed by immunoblot (Fig. 5E). We found that RelA was in reduced amounts complexed with IκB in ΔN4BP1 cells, rendering RelA free for activation. In contrast, in WT cells, IκB was clearly bound to RelA, indicating sequestering and inhibition of RelA and canonical NF-κB.
Thus, N4BP1 functions as a negative regulator of the NF-κB pathway in neuroblastoma, as contributed by both the canonical and noncanonical branches, where N4BP1 uses CEZANNE to stabilize TRAF3, resulting in degradation of NIK and sequestering of RelA by IκB. Upon depletion of N4BP1, TRAF3 is degraded, NIK levels accumulate and IκB detaches from RelA for ensued signaling to target genes, including MHC-1.
In this study, we investigated the repressed NF-κB–mediated MHC-1 expression in neuroblastoma. We show that gene expression levels of NF-κB members are actively repressed during tumorigenic development of neural crest cells toward neuroblastoma tumor cells. Repression of NF-κB proteins in neuroblastoma has been reported previously (13, 17). Van 't Veer and colleagues (14) use neuroblastoma rat cell lines to show decreased expression of p105/p50 protein and regulation thereof by the oncogene NMYC. In contrast, a study in human neuroblastoma cell lines shows low transcriptional availability of NF-κB in neuroblastoma in an NMYC-independent manner (17). In support of the latter study, the performed CRISPR/Cas9 knockout screen we describe here also did not identify NMYC as a negative regulator of NF-κB. However, our screen results were not saturated due to technical restraints therefore a role for NMYC in NF-κB inhibition in neuroblastoma remains plausible. Of note, the p105/p50 subunit did not contribute to NF-κB activity in ΔN4BP1 or ΔTNIP1 neuroblastoma cells, indicating that N4BP1 or TNIP1 do not use NMYC-p50 regulation for NF-κB inhibition.
Nevertheless, restoration of canonical NF-κB can increase MHC-1 expression in neuroblastoma (13, 14, 17). Transfection of NMYC-amplified cells with a p50-expressing construct showed MHC-1 expression (14). Moreover, transfection of RelA increased MHC-1 membrane levels in neuroblastoma (13, 17). Combined activation of NF-κB and the IRF1 transcription factor further increased MHC-1 expression (13), however we have no indication that IRF1 is involved in MHC-1 upregulation observed in ΔN4BP1 or ΔTNIP1 neuroblastoma cells (Supplementary Fig. S7). In ΔN4BP1 and ΔTNIP1 cells we clearly observe activation of canonical NF-κB as RelA is dissociated from IκB protein and highly phosphorylated. In addition, in ΔN4BP1 cells we also observe activation of noncanonical NF-κB involving the RelB and p100/p52 proteins.
N4BP1 was first described as an inhibitor of the E3 ligase Itch thereby stabilizing cell death regulators such as p73α and c-Jun (43). In 2009, N4BP1 was implicated in NF-κB regulation by Fenner and colleagues (44). In a coimmunoprecipitation experiment using polyubiquitin, N4BP1 and other proteins were isolated. A number of these ubiquitin-binding proteins were revealed as negative regulators of NF-κB. Although N4BP1 was not followed-up in this study, its ubiquitin-binding capacity implicated the possibility for regulating NF-κB. Then, in 2011 a direct role of N4BP1 in controlling NF-κB activity was first observed (36). Through overexpression in NF-κB reporter cells, N4BP1 was found as the most effective negative regulator of NF-κB. We here extend the knowledge about N4BP1 as an NF-κB suppressor by showing its interaction with deubiquitinating enzyme CEZANNE. CEZANNE regulates TRAF3 levels (41), by which it controls both canonical and noncanonical NF-κB (42). As such, N4BP1 prevents NF-κB activation by orchestrating CEZANNE-mediated deubiquitination of TRAF3 thereby stabilizing TRAF3 protein. Intact TRAF3 results in constant degradation of NIK and thus obstruction of noncanonical NF-κB activation. Also, RelA protein was sequestered by IκB when TRAF3 was present, leading to inhibition of canonical NF-κB.
The NF-κB suppressors N4BP1, CEZANNE, and TNIP1 independently correlate with worse survival in patients with stage-4 neuroblastoma. Despite the clear role shown here of NF-κB suppressors preventing T-cell recognition, there is yet no evidence for an immunological reason underpinning this difference in survival. Since low expression of NF-κB suppressors did not prevent neuroblastoma development, a survival difference might be explained by predisposition of the tumor for post-treatment immune responses. One possibility is that after stem cell transplantation of patients with neuroblastoma, newly developing immune cells can eliminate residing tumor cells that express low levels of NF-κB suppressors better than tumor cells that express high levels of NF-κB suppressors, which would result in better clearance of NF-κB–active neuroblastomas. Alternatively, low expression of NF-κB suppressors may render neuroblastoma tumors more sensitive for current treatment. For example, NF-κB activation is required for certain cytotoxic agents to kill neuroblastoma cells (45, 46). Also, efficacy of retinoic acid-induced differentiation of neuroblastoma cells requires an active NF-κB pathway (47–49).
Taken together, we have uncovered that aberrant NF-κB functioning in neuroblastoma results in decreased MHC-1 expression, escape from T-cell immune surveillance and poor survival in stage-4 patients. The strategy chosen here, combining an NF-κB reporter and MHC-1 cell surface staining in combination with a genome-wide CRISPR/Cas9 library, has to our knowledge not been performed before. The identification of novel and previously identified members of the NF-κB signaling pathway illustrates that this strategy has been successful. A limitation to our study is that several described NF-κB and MHC-1 regulators were not retrieved using this method. A possible explanation for this is the quality of this early generation CRIPSR/Cas9 library requires further optimization. A library containing more, and better validated, gRNAs targeting each gene could lead to a better separation of mutants. Nevertheless, we have identified NF-κB suppressors that may serve as drug targets to increase neuroblastoma immunogenicity, which is necessary for immunotherapies to translate into clinical benefits for patients with stage-4 neuroblastoma.
Disclosure of Potential Conflicts of Interest
T.R. Brummelkamp has ownership interest (including stock, patents, etc.) in Scenic Biotech and Haplogen. No potential conflicts of interest were disclosed by the other authors.
Conception and design: L. Spel, J. Nieuwenhuis, S. Nierkens, M. Boes
Development of methodology: L. Spel, J. Nieuwenhuis, S. Nierkens, M. Boes
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Spel, J. Nieuwenhuis, R. Haarsma, O.B. Bleijerveld, M. Altelaar, T.R. Brummelkamp, S. Nierkens
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Spel, J. Nieuwenhuis, R. Haarsma, E. Stickel, O.B. Bleijerveld, M. Altelaar, T.R. Brummelkamp, S. Nierkens, M. Boes
Writing, review, and/or revision of the manuscript: L. Spel, J. Nieuwenhuis, R. Haarsma, E. Stickel, J.J. Boelens, S. Nierkens, M. Boes
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Spel, S. Nierkens
Study supervision: J.J. Boelens, S. Nierkens, M. Boes
We thank J.E. Carette for kindly providing us with the NF-κB reporter plasmid. This work was supported by the Villa Joep Foundation and the Netherlands Organization for Scientific Research (NWO), as part of the National Roadmap Large-scale Research Facilities of the Netherlands, Proteins@Work (project number 184.032.201). S. Nierkens and J.J. Boelens received funding from The Villa Joep Foundation and AMMODO. O.B. Bleijerveld and M. Altelaar received funding from Dutch Scientific Research (NWO), as part of the National Roadmap Large-scale Research Facilities of the Netherlands.