Valosin-containing protein-regulated endoplasmic reticulum stress causes NOD2-dependent inflammatory responses

NOD2 polymorphisms may affect sensing of the bacterial muramyl dipeptide (MDP) and trigger perturbed inflammatory responses. Genetic screening of a patient with immunodeficiency and enteropathy revealed a rare homozygous missense mutation in the first CARD domain of NOD2 (ENST00000300589; c.160G > A, p.E54K). Biochemical assays confirmed impaired NOD2-dependent signaling and proinflammatory cytokine production in patient’s cells and heterologous cellular models with overexpression of the NOD2 mutant. Immunoprecipitation-coupled mass spectrometry unveiled the ATPase valosin-containing protein (VCP) as novel interaction partner of wildtype NOD2, while the binding to the NOD2 variant p.E54K was abrogated. Knockdown of VCP in coloncarcinoma cells led to impaired NF-κB activity and IL8 expression upon MDP stimulation. In contrast, tunicamycin-induced ER stress resulted in increased IL8, CXCL1, and CXCL2 production in cells with knockdown of VCP, while enhanced expression of these proinflammatory molecules was abolished upon knockout of NOD2. Taken together, these data suggest that VCP-mediated inflammatory responses upon ER stress are NOD2-dependent.

Here, we report a rare missense mutation affecting the first CARD domain of NOD2 that has been identified in a child with enteropathy and was associated with defective MDP-dependent signaling, abrogated interaction with RIPK2, as well as impaired cytokine responses. The characterization of this rare mutation unveiled Valosin-containing protein (VCP) as novel interaction partner of NOD2. VCP is a ubiquitously expressed ATPase with pleiotropic functions in controlling Ubiquitin-proteasome system (UPS)-mediated protein degradation in endoplasmic reticulum (ER)-associated protein degradation (ERAD), apoptosis, and autophagy [11][12][13] . Our study highlighted that VCP-mediated proinflammatory responses during ER stress are NOD2-dependent.

Results
Defective NOD2 signaling caused by a biallelic germline mutation affecting the first CARD domain. Whole exome sequencing was conducted to elucidate the genetic etiology in a one-year-old female patient presenting with intractable diarrhea, recurrent perianal candida dermatitis, hemophagocytic lymphohystiocytosis (HLH), and prolonged Norovirus infection. Genetic screening revealed rare homozygous missense mutations in NOD2 (ENST00000300589; c.160G > A, p.E54K) and STXBP2 (ENST00000441779; c.949C > G, p.L317V). Even though the variant in STXBP2 was predicted to be benign, this mutation likely has caused HLH associated with impaired NK cell degranulation ( Supplementary Fig. 1). By contrast, the NOD2 missense mutation has been proposed as deleterious based on PolyPhen and SIFT and thus, may contribute to the gastrointestinal phenotypes (intractable diarrhea, recurrent perianal dermatitis). Our study was not specifically designed to address whether the NOD2 variant was the causal or risk factor of disease in our patient, but the distinct location of the variant triggered assessment of NOD2-mediated signaling and interacting networks, since previously reported CD-associated NOD2 variants are mainly localized in or near the LRR domain ( Fig. 1A) [14][15][16] .
To elucidate the effects of the identified NOD2 mutation on the canonical NOD2-mediated signaling, we assessed intracytoplasmic TNF production in patient's peripheral blood mononuclear cells (PBMC)-derived monocytes upon stimulation with the ligand L18-MDP. While patient-derived monocytes showed a normal response to lipopolysaccharide (LPS), TNF expression was reduced in L18-MDP-treated cells as compared with healthy donors (Fig. 1B). In addition, patient-derived neutrophils showed reduced CD62L shedding upon MDP stimulation (Fig. 1C), confirming defective NOD2 signaling in patient innate immune cells. Correspondingly, patient's PBMCs revealed reduced phosphorylation of NF-κB p65, ERK1/2, and/or p38 upon stimulation with L18-MDP (Fig. 1D). Furthermore, luciferase reporter assays on HEK293T cells ectopically expressing the NOD2 variant p.E54K showed impaired NF-κB activity in response to L18-MDP comparable with the NOD2 variant p.L1007fsX1008 that has previously been associated with CD 15 (Fig. 2A).
To gain insights into the pathomechanisms of the mutation in the context of intestinal inflammation, we engineered coloncarcinoma-derived HCT116 cells with a CRISPR/Cas9-mediated NOD2 knockout (KO) and subsequent lentiviral overexpression of wild-type (WT) or mutant (p.E54K and p.L1007fsX1008) NOD2 variants. In contrast to WT reconstitution, HCT116 cells with NOD2 mutants showed reduced expression and secretion of IL-8 upon treatment with L18-MDP (Fig. 2B).
Emerging evidence highlights that NOD2 has critical functions apart from peptidoglycan (PGN) sensing. Notably, NOD2 has been implicated in mediating proinflammatory responses triggered by ER stress 17 . To assess the PGN-independent functions of NOD2, we evaluated the expression of the proinflammatory cytokine IL8 in engineered HCT116 cells upon ER stress. Treatment with tunicamycin resulted in impaired IL8 production in cells with expression of the NOD2 variant p.E54K as compared with WT NOD2 reconstituted cells. Interestingly, we could detect normal expression of IL8 for the NOD2 variant p.L1007fsX1008 suggesting genotype-specific mechanisms of ER stress-induced proinflammation in the context of NOD2 deficiency (Fig. 2C). Impaired RIPK2 binding and ubiquitination by the NOD2 mutant p.E54K. NOD2 is critical in mediating inflammatory signaling pathways in response to invading pathogens via the interaction of RIPK2 with its CARD domain 6 . Previously reported NOD2 polymorphisms associated with CD are mainly localized in the LRR domain 14,[18][19][20][21][22][23][24][25] , however two heterozygous variants (p.R38M and p.R138Q) affecting the first and second CARD domain have been suggested to alter RIPK2 recruitment and NF-κB signaling 26 . To assess whether the NOD2 variant p.E54K affects interaction with RIPK2, we conducted co-immunoprecipitation experiments using anti-FLAG beads in HEK293T cells ectopically expressing FLAG-tagged NOD2 WT or mutants along with WT RIPK2. While the NOD2 variant p.L1007fsX1008 showed normal binding to RIPK2, the interaction of the mutant p.E54K and RIPK2 was significantly reduced (Fig. 2D).
Previous studies have demonstrated that polyubiquitination and autophosphorylation of RIPK2 is triggered upon NOD2 activation 4,7,27,28 . Despite the different abilities of the NOD2 mutants p.E54K and p.L1007fsX1008 to interact with RIPK2, we could detect reduced phosphorylation of RIPK2 at position S176 by immunoblotting as well as impaired ubiquitination of RIPK2 by employing tandem ubiquitin-binding entities upon MDP stimulation in both NOD2 variants (Fig. 2E). These data suggest that defective polyubiquitination and/or autophosphorylation of RIPK2 may represent an underlying common mechanism for reduced MDP-dependent signaling associated with NOD2 polymorphisms.  (Fig. 3A). Using co-immunoprecipitation and nano liquid chromatography tandem mass spectrometry (LC-MS/MS), we identified VCP as novel NOD2 interacting protein enriched in cells reconstituted with WT NOD2. In contrast, cells with expression of the variants p.E54K or p.L1007fsX1008 showed abrogated interaction of VCP with NOD2 ( Fig. 3B and C). Notably, we also identified vimentin (VIM) and carbamoyl phosphate synthetase/aspartate transcarbamylase/ dihydroorotase (CAD) among the list of known NOD2 interacting proteins; thus increasing the confide in our screening approach. While vimentin was enriched in both NOD2 WT and p.E54K expressing cells as compared with RFP controls, CAD was found to be enriched only in cells overexpressing WT NOD2 (Fig. 3B). Finally, the interaction of endogenous VCP/NOD2 proteins in HCT116 cells was confirmed by immunoprecipitation with an antibody binding to VCP and co-precipitation of NOD2 ( Supplementary Fig. 2). VCP is an evolutionarily conserved AAA + ATPase governing diverse biological functions, in particular in the UPS and ER-associated degradation (ERAD) 29 . VCP mutations have been associated with several diseases such as myopathy, Paget's disease, dementia, amyotrophic lateral sclerosis and Huntington's disease 29,30 . However, the exact role of VCP in health and disease remains elusive. Previously, VCP has been listed as potential NOD2 interaction partner in one of three yeast two hybrid (Y2H) screens 31 and proteomic studies have revealed VCP in the group of proteins that are differentially expressed in HEK293T cells overexpressing the NOD2 variant www.nature.com/scientificreports/ p.L1007fsX1008 32 . These studies support our findings that VCP is an interaction partner of NOD2 but did not provide any functional links for the regulatory role of VCP in NOD2 signaling.

VCP controlled ER stress causes inflammatory responses in a NOD2-dependent manner.
To study VCP function in the context of NOD2 signaling, we engineered heterologous HCT116 cells with siRNAmediated knockdown of VCP and evaluated MDP-induced NF-κB activity and IL8 expression. Luciferase reporter assays showed impaired NF-κB activation in cells with knockdown of VCP in response to L18-MDP treatment (Fig. 4A). Correspondingly, we could detect decreased expression of IL8 in MDP-treated cells upon VCP knockdown, as compared with cells transfected with non-targeting siRNA (Fig. 4B). However, we could not observe a direct effect of VCP knockdown on phosphorylation of RIPK2 (S176) or binding of RIPK2 to NOD2 ( Supplementary Fig. 3). VCP plays a critical role in controlling UPR and inhibition of VCP resulted in increased ER stress 33 . Moreover, enhanced ER stress and activated UPR in intestinal epithelial cells have been reported in patients with CD and ulcerative colitis (UC) 34 . Recently, NOD1 and NOD2 have been proposed as molecular bridges linking ER stress to proinflammatory responses 17 . To examine the influence of NOD2 on VCP-mediated ER stress functions, we transfected HCT116 cells with NOD2 KO or lentiviral reconstitution of NOD2 WT with siRNAs targeting VCP. Knockdown efficiency was assessed by qPCR measurement of VCP mRNA as well as immunoblotting of VCP protein (Supplementary Figs. 4 and 5). Using this heterologous cellular model, we could confirm previous findings that VCP knockdown induces increased UPR, as demonstrated by enhanced expression of C/EBP homologous protein (CHOP) (Fig. 4C) as well as increased phosphorylation of PERK and eIF2α, while splicing of XBP1 appeared unaffected ( Supplementary Fig. 5). Whereas additional knockout of NOD2 resulted in slight increase of the activated PERK-eIF2α axis ( Supplementary Fig. 5), we could not observe a significant upregulation of the transcription of CHOP, which is a downstream target modulated by all three signaling branches of UPR (Fig. 4C). Moreover, treatment with tunicamcyin in NOD2 WT reconstituted cells induced enhanced IL8, CXCL1, and CXCL2 expression upon knockdown of VCP, while increased proinflammatory cytokine and   www.nature.com/scientificreports/ chemokine expression could not be observed in NOD2 KO cells ( Fig. 4D and E). Differences in IL8, CXCL1, and CXCL2 expression between NOD2 knockout and WT reconstituted cells were not associated with increased cell death (Supplementary Fig. 6). Taken together, these findings suggest that inflammatory responses caused by VCP-regulated ER stress are NOD2-dependent.

Discussion
NOD2 is a key receptor of innate immunity and the first genetic locus that has been associated with inflammatory bowel disease (IBD) 14,15 . Since most CD-associated NOD2 variants are located in the LRR domain, the identification of a biallelic missense mutation affecting the CARD domain of NOD2 in a patient with enteropathy prompted us to investigate the signaling and interaction network of the mutant NOD2 protein in greater detail. Even though the HLH-associated phenotype in our patient is likely caused by the STXBP2 mutation, the NOD2 sequence variant may be a risk factor contributing to the gastrointestinal phenotypes (intractable diarrhea, recurrent perianal dermatitis). The function of NOD2 in pathogen recognition or peptidoglycan sensing has been previously acknowledged 35,36 , however the role of NOD2 during ER stress remains still elusive. Our biochemical study showed impaired NOD2-governed PGN-dependent and independent signaling in primary patient cells as well as cellular models and unveiled VCP as novel interaction partner of NOD2 that regulates ER stress-mediated inflammatory responses. Several studies have suggested the implication of dysregulated UPR in different inflammatory conditions such as neurodegenerative diseases and IBD [37][38][39][40] . A growing body of evidence indicate reciprocal relationships between inflammation and ER stress 41 . While inflammatory stimuli like pattern-recognition receptor (PRR) ligands or ROS can induce UPR, activation of the three main UPR pathways can trigger NF-κB-and MAPKdependent inflammatory responses leading to the expression of proinflammatory cytokines such as IL-6 and TNF 17,[42][43][44] . In the context of intestinal inflammation, mice with KO of Ire1 and Xbp have been shown to have increased sensitivity to dextran sodium sulfate (DSS)-induced colitis 37,38 .
NOD2 has been previously implicated in regulating ER stress 17 . For example, Laccase domain containing-1 (LACC1)-dependent induction of ER stress has been documented in macrophages upon MDP stimulation 45 . Furthermore, previous studies have shown that NOD1/NOD2/RIPK2-dependent inflammation can be triggered by ER stress in mouse bone-marrow-derived macrophages (BMDMs) via the IRE1α/TRAF2 pathway 17   www.nature.com/scientificreports/ that NOD2 mediates proinflammatory responses upon different types of UPR via interaction of its nucleotide binding domains with sphingosine-1-phosphate 47 . Corresponding to their finding that ER stress activated NOD2 independent of the LRR domain, we could observe comparable transcriptional level of IL8 in cells overexpressing the NOD2 p.L1007fsX1008 variant and WT NOD2. Thus, our investigation on a rare sequence indicated altered ER stress as possible mechanism how NOD2 polymorphisms may contribute to disease development and behavior. The CARD domain of NOD2 is known to be important for the interaction with the adaptor protein RIPK2 48 . Our study revealed impaired RIPK2 interaction for the identified mutation p.E54K affecting the CARD domain but not the LRR domain variant p.L1007fsX1008. However, we could observe reduced phosphorylation and abrogated ubiquitination of RIPK2 as a potential common pathomechanism for the impaired MDP-triggered NF-κB activity in NOD2-deficient cells. Consistently, X-linked Inhibitor of Apoptosis (XIAP) E3 ubiquitin ligase activity mediating ubiquitination of RIPK2 has been previously reported to be indispensable for NF-κB activation initiated by NOD2 stimulation 49 . The relevance of this signaling axis for human disease has been demonstrated by loss-of-function XIAP mutations causing a severe immunodeficiency disorder 50,51 . In routine diagnostics, analysis of defective MDP-dependent NOD2 signaling is used to determine XIAP deficiency 52 . Since both mutations exhibited abrogated RIPK2 ubiquitination, future studies investigating the recruitment of XIAP to the NOD2 complex might provide further insights on the pathomechanisms of the NOD2 variants.
In view of the impaired RIPK2 interaction and posttranslational modification, we sought to decipher the interaction network of the NOD2 variant p.E54K. Using an immunoprecipitation-coupled mass spectrometry screen, we identified VCP as a novel NOD2 interaction partner that was associated with wild-type NOD2 protein but not with the NOD2 variants p.E54K and p.L1007fsX1008. VCP is an abundant ubiquitin-dependent ATPase that implicates in myriad of cellular processes such as ERAD, autophagy, DNA damage response, apoptosis and   www.nature.com/scientificreports/ ubiquitin-proteasome-dependent protein degradation [11][12][13]53,54 . Heterozygous germline mutations in VCP have been previously associated with Paget disease of bone and frontotemporal dementia, amyotrophic lateral sclerosis (ALS) and type 2 Charcot-Marie-Tooth disease [55][56][57] . Furthermore, increased level of proinflammatory cytokines have been observed in the plasma and myoblasts of patients with VCP mutations 58 . The function of VCP in the ERAD pathway has been reported to be regulated through interaction with the deubiquitinase ATAXIN3 59 . Interestingly phosphorylation of ATAXIN3 by NOD2 and TLR2 in myeloid cells has been recently shown to mediate mitochondrial reactive oxygen species production and bacterial clearance 60 . To evaluate plausible functions of VCP in NOD2 signaling, we used VCP-silenced cellular models that were stimulated with the NOD2 canonical stimuli MDP. Consistent with previous studies demonstrating impaired TNF-and IL-1β-triggered NF-κB signaling in VCP-deficient cells 61 , we observed impaired NF-κB activity and proinflammatory cytokine responses in cells with knockdown of VCP upon MDP treatment comparable to NOD2-deficient cells. Strikingly, our data unveiled VCP as a negative regulator of NOD2 activity during tunicamycin-induced ER stress, as VCP silencing resulted in NOD2-dependent hyperinflammatory responses. While VCP-dependent CHOP transcription was not affected by knockout of NOD2, expression of the members of the CXC family of chemokines IL8 (CXCL8), CXCL1, and CXCL2 was increased in a NOD2-dependent manner. These chemokines have been shown to be important in the regulation of neutrophil activation and migration as well as the induction of exaggerated angiogenesis at sites of inflammation 62 . Notably, the expression and activity of these molecules were positively correlated with the grade of inflammation in IBD patients [63][64][65][66] . Thus, altered NOD2-dependent proinflammatory cytokine responses upon ER stress may present a new link in the context of intestinal inflammation, however the exact mechanisms how VCP acts on NOD2 signaling remains elusive. Our data suggested that knockdown of VCP alters UPR but does not directly affect phosphorylation of RIPK2 or binding of RIPK2 to NOD2. Therefore, we propose that VCP-regulated UPR can be sensed by NOD2 and can trigger inflammatory responses in a NOD2-dependent manner. Previous studies have suggested that activation of NF-κB signaling is mediating ER stress-derived inflammation, however we could not observe substantial alteration of NF-κB p65 phosphorylation upon tunicamycin stimulation in VCP-silenced cells (data not shown). Further studies are required to profile ER stress-induced NOD2-dependent proinflammatory responses in greater detail and to decipher the underlying molecular mechanisms. Taken together, molecular characterization of a rare germline mutation affecting the first CARD domain of NOD2 unveils VCP as novel interaction partner. Functional studies show that VCP controlled ER stress induces inflammatory responses in a NOD2-dependent fashion; thus, providing a new potential mechanistic link and therapeutic target in NOD2-related intestinal inflammation.

Methods
Patient. Written informed consent was obtained from the patient, first-degree relatives, and healthy donors for the collection of peripheral blood. The investigation was approved by the respective institutional review boards of the LMU Munich and conducted in accordance with current ethical and legal frameworks. DNA sequencing. Next-generation sequencing was performed at the Dr. von Hauner Children's Hospital NGS facility. Genomic DNA was isolated from whole blood (Qiagen) for the generation of whole-exome libraries using the SureSelect XT Human All Exon V6 + UTR kit (Agilent Technologies). Barcoded libraries were sequenced with a NextSeq 500 platform (Illumina) to an average coverage depth of 90x. Bioinformatics analysis used Burrows-Wheeler Aligner (BWA 0.7.15), Genome Analysis ToolKit (GATK 3.6) and Variant Effect Predictor (VEP 89). The frequency filtering used allele frequencies from public (e.g. ExAC, GnomAD and GME) and in house databases. The potentially causative variants were confirmed by Sanger sequencing for the patient and informative family members.
Cell culture and stimulation. Peripheral  Immunoblotting and silver staining. Cells were lysed in 1× cell lysis buffer (Cell Signaling Technologies) supplemented with 1 mM phenylmethylsulfonyl fluoride and 1× protease inhibitors. Normalization of protein concentration was performed by Bradford assay and equal amount of proteins were subjected to 10-12% SDS-PAGE followed by immunoblotting using different antibodies. Chemiluminescence signals were detected using the SuperSignal West Dura detection kit (Thermo Fisher Scientific) on the ChemiDoc TM XRS + System (Bio-Rad) and analyzed with the ImageLab TM software (Bio-Rad). SDS-PAGE silver staining was performed using silverQuest TM (Invitrogen) according to the manufacturer's protocols.
Intracellular flow cytometry. Cells were washed with PBS and then fixed and permeabilized using the Cytofix/Cytoperm Kit (BD Bioscience) and stained with CD14, CD3 and TNF antibodies. Flow cytometry was performed on the LSRFortessa TM flow cytometer (BD Biosciences) and analyzed with the Flowjo V9 software (TreeStar) 52 .
Engineering of NOD2 knockout cell lines using CRISPR/Cas9 genome editing. Analysis of cell death in HCT116 coloncarcinoma cell lines. NOD2 knockout and lentiviral reconstituted NOD2 WT HCT116 cells treated with siRNAs targeting human VCP and non-targeting (NT) siRNA oligonucleotides were stimulated with 5 µg/ml tunicamycin (Sigma-Aldrich) for 24 h. To measure cell death, HCT116 cells were stained with Annexin V and DAPI (Thermo Fisher Scientific) and analyzed by flow cytometry.
Gels were Coomassie stained (Simply Blue, Expedeon) and the area containing proteins was excised. Gel slices were destained (50% acetonitril, 50 mM NH4HCO3) and subjected to in-gel digestion using the following steps: For protein reduction and alkylation, gel slices were first incubated in 45 mM DTE/50 mM NH4HCO3 for 30 min at 55 °C and then incubated for 30 min in 100 mM iodoacaetamide/50 mM NH4HCO3. In-gel digestion was done using 0.7 µg Trypsin at 37 °C overnight. Samples were analyzed by nano-LC MS/MS using an Ultimate 3000 nano liquid chromatography system (ThermoFisher Scientific) coupled to a TripleTOF 5600 + instrument (Sciex). As solvent A 0.1% formic acid and as solvent B acetonitrile with 0.1% formic acid was used. Peptides were separated at a flow rate of 200 nL/min on an Acclaim PepMap RSLC C18 column (75 μm × 50 cm, Thermo Fisher Scientific) with the following gradient: from 2% B to 25% B in 120 min followed by 25% B to 50% B in 10 min. For mass spectrometry, the ion source was operated at a needle voltage of 2.3 kV. Mass spectra were acquired in cycles of one MS scan from 400 m/z to 1250 m/z and up to 40 data dependent MS/MS scans of the most intensive peptide signals. For protein identification (FDR < 1%) and label free quantification, the Max-Quant software platform 68 was used in combination with the Human subset of the UniProt database. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD031539 69 .
Statistical analysis. Prism version 6 (GraphPad Software, USA) was used for statistical analysis of experimental data. Probability (P) values were calculated using two-way repeated-measures ANOVA and P values < 0.05 were considered to be statistically significant. Statistical details of experiments can be found in the figures and figure legends. Biologically independent experiments are referred to as n.

Data availability
The data generated in this study is available upon request, please contact the corresponding author.