Hepatic NOD2 promotes hepatocarcinogenesis via a RIP2-mediated proinflammatory response and a novel nuclear autophagy-mediated DNA damage mechanism

HCC specimens and paired adjacent non-tumor liver tissues were collected from 141 patients (Zhongshan Hospital HCC1, i.e., ZHH1 cohort, 2005–2010) receiving surgical resection at Zhongshan Hospital, Fudan University. Paired tumor and adjacent liver tissues were collected from 80 patients (ZHH2 cohort, 2012–2015) for mRNA analysis, 30 of which were randomly selected for western blot and immunohistochemistry (IHC) analyses. HCC specimens and normal liver tissues were collected from patients receiving surgical resection or liver transplantation at Zhongshan Hospital, Fudan University (n = 10). Postoperative follow-ups were performed as previously reported [17]. Overall survival (OS) was defined as the interval between the surgery and death or contact failure or the last investigation point [17]. Time to recurrence (TTR) is defined as the interval between surgery and the diagnosis of recurrence (intrahepatic recurrence or extrahepatic metastasis) [17]. This project was performed in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Zhongshan Hospital, Fudan University. Informed consents were obtained from all patients.

Reagents used in this study include diethylnitrosamine (DEN, Sigma-Aldrich, St Louis, MO, USA), muramyl dipeptide (MDP, Invitrogen, San Diego, CA, USA), 3-methyladenine (3-MA, Selleck Chemicals, Shanghai, China), MG-132 (MedChemExpress, NJ, USA), Lyso-Tracker Red (Beyotime, Shanghai, China) and microsocial nuclease (New England Biolabs, Ipswich, MA, USA). The primary antibodies and dilutions used are listed in Additional file 1: Table S1. Biotinylated secondary antibodies were purchased from Santa Cruz; TRITC, Alexa Fluor 488 or Alexa Fluor 647-conjugated secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA, USA). HEK293T cell line was originally purchased from the American Type Culture Collection (ATCC) and cultured in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin in an incubator with 5% CO₂ at 37 °C.

TMAs of the ZHH1 cohort were constructed as described previously [17]. To quantify protein expression levels based on TMAs, a scoring system was used based on the IHC staining signal intensity (0, negative; 1, weak; 2, moderate; 3, strong) and the percentage of positively stained cells (0, 0%; 1, 1–25%; 2, 26–50%; 3, 51–75%; 4, 76–100%) [17]. Composite expression scores were calculated by multiplying the scores of staining intensity and positive cell percentage and classified as negative (score 0), weak (scores 1–4), moderate (scores 5–8) and strong (scores 9–12) [17]. The IHC staining scores were independently evaluated by two pathologists in a double-blind manner. Receiver operating characteristic (ROC) curve analysis showed that the optimal cutoff value of NOD2 was 8, with the maximum area under the curve (AUC) of 0.691. Thus, NOD2 IHC score ≥ 8 was considered as high NOD2 expression, whereas score < 8 was defined as low.

RNA-Seq data of 371 HCC and 50 normal liver tissues were obtained from The Cancer Genome Atlas (TCGA) database (http://​cancergenome.​nih.​gov) for analysis of NOD2 mRNA expression. OS was determined using Kaplan–Meier survival analysis; Gene Set Enrichment Analysis (GSEA) was performed using GSEA software (http://​software.​broadinstitute.​org/​gsea/​downloads.​jsp). For assessing the association of NOD2 expression with genetic instability in TCGA HCC cohort, 104 patients with TP53 mutation and 255 patients without TP53 mutations were retrieved for analysis.

Mice with homozygous conditional-null alleles of Nod2 (Nod2^f/f) were generated with the help of the Model Animal Research Center, Nanjing University (Nanjing, China). Briefly, single-guide RNA (sgRNA), Cas9 mRNA and the donor vector targeting the exon 3 and 4 sites were microinjected into mouse zygotes to create a Nod2-flox allele (Nod2^f/f) with two LoxP sites flanking exons 3 and 4. Nod2^f/f mice were bred to Albumin-Cre mice (Jackson Laboratory, Bar Harbor, ME; 003574) to generate hepatocyte-specific Nod2-knockout mice (Nod2^△hep). Wild-type (WT) mice were purchased from Jackson Laboratory. Rip2^f/f (T005428) and Lmna^f/f mice (T007824) were purchased from GemPharmatech (Nanjing, China) and bred to Albumin-Cre mice to generate Rip2^△hep, Lmna^△hep and Lmna/Rip2^△hep mice. Primers used for genotyping are listed in Additional file 1: Table S2. All animals are in the C57BL/6 background. All mice were provided with food and water ad libitum and maintained in a 12-h light–dark cycle under standard conditions at the Fudan University (Shanghai, China).

For chemically induced HCC, male mice aged 14–16 days were given a single intraperitoneal (i.p.) injection of DEN (25 mg/kg) and 8 subsequent biweekly injections of CCl₄ (1.2 ml/kg, 1:4 diluted with olive oil, i.p., starting 4 weeks after DEN). Mice were killed at the indicated time points after DEN injection.

All experiments preformed were approved by the Research Ethics Committee of Zhongshan Hospital and conducted according to the institutional guidelines.

Data were analyzed with SPSS 20.0 (SPSS Inc., Chicago, IL, USA). Unpaired Student’s t test or the Mann–Whitney U test was used for comparison between two groups. Paired t test was used to compare paired data. Analysis of variance (ANOVA) was used to compare 3 or more groups. Ordinary one-way or two-way ANOVA with a Sidak test was used to make multiple comparisons between different groups. Chi-square test was used for comparing categorical variables. Cox proportional hazard model was used for univariate and multivariate analyses. ROC curve analysis was used to determine the optimal cutoff value of the parameters by SPSS. Survival was analyzed by the Kaplan–Meier (log-rank) test. Pearson correlation analysis was used to assess the association between the expression of NOD2 and lamin A/C or p-RIP2. All data are expressed as mean ± standard deviation (SD). Difference between groups is considered statistically significant if P < 0.05.

A complete description of the methods, including histological analyses, isolation and culture of primary mouse hepatocytes, cell transfection and infection, reactive oxygen species (ROS) detection, ALT and AST analyses, quantitative polymerase chain reaction, western blot analysis, TaqMan Copy Number Analysis, MDP Concentration measurement, co-immunoprecipitation (co-IP) and mass spectrometry (MS) analysis, confocal microscopy, transmission electron microscope (TEM) and live-cell imaging, GST pull-down assay, subcellular fractionation, comet assay and non-homologous end joining (NHEJ) assay, is available in Additional file 2: Online materials and methods.

To determine the role of NOD2 in HCC development, we first examined its expression pattern in HCC samples (ZHH2, n = 80) using qPCR. Compared to the matched adjacent non-tumor (ANT) liver tissues, NOD2 expression was increased in 67.5% (54/80) of HCC samples (Fig. 1a), which was confirmed in the analysis of an HCC cohort using TCGA database and in the analysis of two cohorts (GSE124535, GSE121248) from Gene Expression Omnibus (GEO, https://​www.​ncbi.​nlm.​nih.​gov/​gds) datasets (Fig. 1b). Western blots of 30 randomly selected HCC samples from the ZHH2 cohort demonstrated increased NOD2 protein in HCC (Fig. 1c). In addition, IHC analyses of TMAs from the ZHH1 cohort (n = 141) indicated that HCCs have increased NOD2 protein expression compared to ANT tissues (Fig. 1d, e1), with an average score of 6.11 ± 3.69 for HCC and 4.19 ± 2.88 for ANT tissues (P < 0.001). Notably, phosphorylated RIP2 (p-RIP2), a downstream indicator of NOD2 activation, was also higher in HCCs than in ANT tissues (average score 7.97 ± 3.49 vs. 5.53 ± 3.25, P < 0.001; Fig. 1e2). Furthermore, NOD2 level was positively correlated with p-RIP2 in HCC (Fig. 1h). Consistently, both western blot and IHC analyses demonstrated that NOD2 levels increase gradually during the progression of DEN/CCl₄-induced HCC, which is also accompanied by an increase in RIP2 phosphorylation (Additional file 3: Fig. S1), indicating NOD2 activation in HCC.

To determine the clinical significance of NOD2 in HCC, we examined the association between NOD2 expression and clinicopathological parameters in the ZHH1 cohort (n = 141). HCC tumors were divided into NOD2-high and NOD2-low group based on the cutoff value of their IHC scores (Additional file 3: Fig. S2). We found that the NOD2 expression level positively correlated with tumor aggressive phenotypes, including severe cirrhosis (P = 0.012), frequent vascular invasion (P = 0.007), multiple tumor number (P = 0.003), large tumor size (P = 0.032), high alpha fetoprotein (AFP, P = 0.014), advanced Edmondson grade (P = 0.002) and BCLC stage (P = 0.029) (Additional file 1: Table S3). Kaplan–Meier plot showed that NOD2-high patients have a significantly shorter median OS (33.72 vs. 74.17 months, P < 0.001) and TTR (18.84 vs. 58.20 months, P < 0.001) than NOD2-low patients (Fig. 1f). The shorter median OS in NOD2-high patients was confirmed in the analysis of the TCGA cohort (46.57 vs. 81.87 months, P = 0.011, Fig. 1g, n = 371). Univariate and multivariate analysis demonstrated that, apart from vascular invasion and tumor number, NOD2 expression remains as an independent prognostic factor of both OS (hazard ratio = 3.447, 95% CI: 1.931–6.154, P < 0.001) and TTR (hazard ratio = 3.010, 95% CI: 1.861–4.868, P < 0.001) (Additional file 1: Table S4).

These results established the clinical relevance of NOD2 in HCC and suggested that NOD2 may play a role in the development of HCC.

Next, we constructed hepatocyte-specific Nod2-knockout (Nod2^△hep) mice (Additional file 3: Fig. S3) and studied the role of hepatic NOD2 in a DEN/CCl₄ HCC model of Nod2^f/f and Nod2^△hep mice. Eight months after DEN injection, all Nod2^f/f mice (15/15) developed liver tumors, whereas only 73.3% (11/15) of Nod2^△hep mice did. The number of tumor nodules was nearly 1.5-fold lower in Nod2^△hep mice than that in Nod2^f/f mice (17.36 ± 4.93 vs. 27.40 ± 6.98, P < 0.01). The maximal tumor diameters in Nod2^△hep mice were almost threefold lower than those in Nod2^f/f mice (3.41 ± 2.10 vs. 9.77 ± 2.42, P < 0.01) (Fig. 2a, b). Histopathological analysis revealed that liver tumors in Nod2^△hep and Nod2^f/f mice were basophilic and showed similar pathological characteristics (Fig. 2c). DEN/CCl₄-treated Nod2^△hep mice had a higher survival rate than DEN/CCl₄-treated Nod2^f/f mice (Fig. 2d). These results showed that loss of hepatic NOD2 attenuates the tumorigenesis of DEN/CCl₄-induced HCC.

To determine whether NOD2 deletion affects HCC development at the cellular level, we measured cell death and proliferation in DEN/CCl₄-treated livers. Compared to Nod2^f/f mice, Nod2^△hep mice displayed more apoptotic hepatocytes (Fig. 2e) and attenuated Ki67⁺ cells (Fig. 2f). The hepatocyte injury indicator serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were lower in the serum of Nod2^△hep mice (Fig. 2g). The degree of liver fibrosis was also decreased in Nod2^△hep mice, indicated by Sirius Red and Masson staining 4 months after DEN/CCl₄ treatment (Fig. 2h).

Chronic inflammatory response, DNA damage and subsequent genomic instability are hallmarks of carcinogens-induced HCC [18]. We performed RNA-seq analyses of livers from DEN/CCl₄-treated Nod2^f/f and Nod2^△hep mice (n = 3, male) (SRA accession: PRJNA685328). Hierarchical clustering showed that multiple inflammatory cytokines (IL-6, TNF-α, IL-1, IFN-γ, etc.) were downregulated in Nod2^△hep livers, while genes involved in DNA damage repair upregulated, compared to Nod2^f/f livers (Fig. 3a and Additional file 1: Table S5). KEGG pathway enrichment analysis showed that genes downregulated in Nod2^△hep were enriched for the pro-inflammatory pathways (NF-κB, JAK-STAT and MAPK), while genes upregulated were enriched for DNA damage repair pathways (NHEJ and base excision repair), compared to Nod2^f/f livers (Fig. 3b).

Consistent with RNA-seq data, inflammatory cytokine genes (IL-6, TNF-α, IL-1β and IFN-γ), pro-inflammatory pathways (NF-κB (p65), JAK2/STAT3 and MAPK (ERK, JNK, P38), as well as ROS biosynthesis genes were positively co-enriched with NOD2 (Additional file 3: Fig. S4a, b) in GSEA analyses of the TCGA HCC dataset (n = 371) with the gene ontology (GO) and Hallmark gene sets. Reduced pro-inflammatory cytokine levels (IL-6, TNF-α, IL-1β and IFN-γ) and attenuated activation of pro-inflammatory pathways (NF-κB (p65), JAK2/STAT3 and MAPK (JNK, ERK, P38)) were further confirmed in DEN/CCl₄-treated Nod2^△hep livers using qRT-PCR, ELISA and western blot analyses, respectively (Fig. 3c, d and Additional file 3: Fig. S5). Consistently, liver macrophages infiltration, as quantified with F4/80 staining, was diminished in Nod2^△hep compared to Nod2^f/f mice (Fig. 3e).

In line with RNA-seq data, DNA damage recognition and repair pathways were negatively co-enriched with NOD2 in GSEA analyses of the TCGA HCC dataset (Additional file 3: Fig. S4c). Decreased 8-OHdG expression (a DNA damage marker) (Fig. 4a), γ-H2AX⁺ (a DNA double-strand break marker) cell numbers (Fig. 4b) and ROS levels (Fig. 4c) were observed in DEN/CCl₄-treated Nod2^△hep livers, compared to those in DEN/CCl₄-treated Nod2^f/f livers. Attenuated DDR pathways (ATM/CHK2, ATR/CHK1) activation and reduced γ-H2AX level were demonstrated with western blot analysis of DEN/CCl₄-treated Nod2^△hep livers (Fig. 4d). Remarkably, DEN/CCl₄-treated Nod2^△hep livers presented decreased genetic instability (14.81%) compared to DEN/CCl₄-treated Nod2^f/f livers (44.44%) (Fig. 4e), as measured by allelic imbalances at chromosomal common fragile sites (CFS) using the TaqMan copy number assay [19].

These results collectively demonstrate that hepatic NOD2 knockout attenuates the inflammatory response and DNA damage in DEN/CCl₄-induced HCC model.

Increased intestinal permeability may allow the translocation of gut microbial metabolites such as MDP, into the liver via gut-liver axis in HCC patients. In line with this hypothesis, we observed an increased MDP level in HCC tissues, compared to that in normal liver tissue (Additional file 3: Fig. S6). To determine the role of MDP in hepatocarcinogenesis in vivo, we treated WT, Nod2^△hep and Rip2^△hep mice with MDP [20] in DEN/CCl₄-induced HCC model (modeling procedures detailed in Additional file 3: Fig. S7a). Six months after DEN injection, compared with untreated group, MDP treatment significantly increased the tumor incidence (40.0% (6/15) vs. 80% (12/15)) and tumor burden (tumor number: 4.67 ± 0.82 vs. 10.83 ± 1.85, P < 0.01; tumor size: 7.12 ± 2.04 vs. 11.58 ± 2.13, P < 0.01), in the WT, but not in Nod2^△hep (tumor incidence: 27.8% (5/18) vs. 33.3% (5/15); tumor number: 2.2 ± 0.84 vs. 2.40 ± 1.14, P > 0.05; tumor size: 2.00 ± 0.71 vs. 3.60 ± 1.14, P > 0.05), indicating that MDP promotes tumorigenesis in an NOD2-dependent manner. Unexpectedly, MDP-induced increase of tumor incidence and burden was still observed in Rip2^△hep mice (tumor incidence: 46.7% (7/15) vs. 66.7% (10/15); tumor number: 4.00 ± 0.82 vs. 7.71 ± 1.25, P < 0.01; tumor size: 4.64 ± 1.11 vs. 8.00 ± 0.87, P < 0.01) (Additional file 3: Fig. S7b, c), although this increase was moderate, indicating that a RIP2-independent signal may be also involved in MDP-promoted hepatocarcinogenesis. Kaplan–Meier plot also showed that compared with untreated group, MDP treatment significantly decreased the survival rate in the WT (median survival time: 329 vs. 268 days) and Rip^∆hep (median survival time: 359 vs. 302 days), but only slightly decreased the survival rate in Nod2^△hep (median survival time: 398 vs. 378 days) in DEN/CCl₄ HCC model.

It is well established that NOD2 activation drives inflammatory response via RIP2 binding and phosphorylation [9]. To determine whether the pro-inflammatory effect of NOD2 activation is mediated by RIP2 in the liver, we measured the expression of pro-inflammatory cytokines and macrophages infiltration in livers of mice from the MDP-treated HCC model. One month after DEN injection, MDP treatment significantly increased pro-inflammatory responses (expression of IL-6, TNF-α, IL-1β and IFN-γ, and the number of infiltrating macrophages) and DNA damage (8-OHdG and H2AX expression) in WT mice. NOD2 deficiency blocked MDP-exaggerated pro-inflammatory responses and DNA damage. In contrast, RIP2 deficiency only blocked MDP-potentiated inflammatory responses, but had no effect on MDP-increased DNA damage (Additional file 3: Fig. S7e–g), indicating a RIP2-dependent and a RIP2-independent mechanism in promoting pro-inflammatory responses and exaggerating DNA damage, respectively.

To determine the effect of NOD2 activation on DEN-induced inflammatory response in vitro, we treated primary hepatocytes with DEN alone or with DEN plus MDP (DEN/MDP). Consistent with our findings in vivo (Fig. 3a–d), MDP aggravated DEN-induced inflammatory response, evidenced by increased expression of IL-6, TNF-α, IL-1β and IFN-γ in hepatocytes (Additional file 3: Fig. S8a) and potentiated DEN-induced activation of NF-κB (p65), JAK2/STAT3 and MAPK (JNK, ERK,P38) (Additional file 3: Fig. S8b), which are well-known pro-inflammatory pathways downstream of RIP2 [9, 21]. In contrast, MDP did not enhance DEN-induced inflammatory response in hepatocytes from Rip2^△hep mice. Together, these findings indicated that, in a Rip2-dependent manner, NOD2 activation turns on the NF-κB, JAK2/STAT3 and MAPK signaling pathways to promote a pro-inflammatory response in the hepatocytes (Additional file 3: Fig. S8c).

We have found that Rip2 knock-out had no effect on MDP-aggravated DNA damage in the DEN/CCl₄ HCC model (Additional file 3: Fig. S7g). Similarly, when exposed to DEN in vitro, hepatocytes from Rip2^△hep mice still had aggravated DNA damage after MDP treatment, as determined by the activation of DDR pathways (ATM/CHK2, ATR/CHK1) and elevation of γ-H2AX level via western blots (Additional file 3: Fig. S9), further indicating that NOD2 activation in hepatocytes aggravates DEN-induced DNA damage via a RIP2-independent mechanism.

We have shown that NOD2 activation RIP2 independently aggravates DNA damage. To gain further insights into the mechanistic interplay between NOD2 and DNA damage, we used co-immunoprecipitation and liquid chromatography-tandem mass spectrometry (LC–MS/MS) to search for NOD2-interacting proteins and identified 28 potential candidates (Additional file 1: Table S6). Among them, nuclear matrix protein lamin A/C (Lmna) was identified by five unique peptides (Additional file 3: Fig. S10a, b). Interaction network analysis using RNA-seq data (Nod2^△hep and Nod2^f/f transcriptomes), String database (https://​string-db.​org/​) and Cytoscape software suggested that Lmna might play a central role in the NOD2-aggravated DNA damage (Additional file 3: Fig. S10c). Bidirectional co-IP assays verified the physical interaction between endogenous NOD2 and lamin A/C in primary hepatocytes (Fig. 5a). GST pull-down assays showed that purified GST-NOD2 could directly interact with lamin A/C via NOD2 C‐terminal LRR domain (amino acids 619–1040) (Fig. 5b). Confocal immunofluorescence showed that, upon MDP treatment, cytoplasmic NOD2 underwent nuclear translocation and co-localized with lamin A/C (Fig. 5c).

Using the cNLS Mapper program (http://​nls-mapper.​iab.​keio.​ac.​jp/​cgi-bin/​NLS_​Mapper_​form.​cgi) [22], a bipartite nuclear localization sequences (NLS), ₄₀₇LRKFVRTECQLKGFSEEGIQLYLRKHHREP₄₃₆, were identified on NOD2 (Additional file 3: Fig. S11a). Immunofluorescence and western blot studies showed that deletion of this NLS completely abolished the ability of NOD2 to translocate into the nucleus (Additional file 3: Fig. S11b, c).

To determine how NOD2 regulates lamin A/C, we measured the effect of NOD2 activation on lamin A/C protein level and found that MDP treatment concentration- and time-dependently decreased lamin A/C protein in primary hepatocytes, and that this effect is NOD2-dependent, but RIP2-independent, because NOD2 deficiency, but not RIP2 deficiency, could abolish this effect (Fig. 5d and Additional file 3: Fig. S12a). MDP treatment did not affect the mRNA level (Additional file 3: Fig. S12b), but decreased the protein stability of lamin A/C (Additional file 3: Fig. S12c). MDP-induced lamin A/C degradation was blocked by 3-MA (an autophagy inhibitor) (Additional file 3: Fig. S12d), but not by MG-132 (a proteasome inhibitor) (Additional file 3: Fig. S12e), indicating an autophagy-dependent pathway. In support of these results, we found that MDP treatment induced nuclear autophagy, evidenced by increased LC3-II and decreased p62 in the nucleus (Fig. 5e and Additional file 3: Fig. S12f). MDP-induced nuclear autophagy could be abolished by NOD2 deficiency, but not RIP2 deficiency (Fig. 5f and Additional file 3: Fig. S12g), indicating a NOD2-dependent, but RIP2-independent manner.

TEM demonstrated more nuclear autophagosomes (Fig. 5g and Additional file 3: Fig. S12h), increased nuclear membrane collapse and less perinuclear heterochromatin in MDP-treated primary hepatocytes, compared to vehicle (Fig. 5g), indicating that MDP triggers nuclear autophagy events. Consistently, we found that NOD2 interacted with ATG16L1, a protein governing autophagosome formation, in hepatocytes following MDP treatment (Additional file 3: Fig. S13). Confocal immunofluorescence analysis revealed that lamin A/C co-localized with GFP-LC3 in nucleus after 15 min of MDP stimulation (Fig. 5h, middle panel), while nuclear lamin A/C, HA-NOD2 and DAPI bodies appeared in cytoplasm and co-localized with GFP-LC3 after 150 min of MDP stimulation (Fig. 5h, bottom panel), indicating that lamin A/C is an autophagy substrate upon MDP stimulation.

Live-cell imaging on mCherry-GFP-lamin A/C-expressing hepatocytes showed that upon MDP stimulation, lamin A/C underwent a nucleus-to-cytoplasm transport process at 0–120 min, accompanied by nuclear membrane blebbing at 45 min. Lamin A/C gradually entered the lysosome at 135–165 min, as lamin A/C displayed cytoplasmic red-only bodies at 165 min in cells, and GFP signal was quenched by the acidic pH environment inside the lysosome [23] (Additional file 3: Fig. S14a). Super-resolution microscopy analysis further confirmed that lamin A/C was degraded by lysosome pathway, evidenced by co-localization of lamin A/C and Lyso-Tracker Red in cytoplasm of MDP-stimulated hepatocytes (Additional file 3: Fig. S14b).

Together, these findings show that NOD2 activation triggers a novel nuclear autophagy pathway to degrade lamin A/C protein (Additional file 3: Fig. S14c).

Lamin A/C plays important roles in DNA repair and genomic stability [24, 25], by promoting recruitment of 53BP1 to the damage foci and enrichment of SIRT6 to the chromatin upon DNA damage, thus facilitating DNA damage repair through NHEJ pathways [26‐28]. Consistently, our Comet assay and double-labeled immunofluorescence of γ-H2AX/53BP1 showed that MDP treatment aggravated DEN-induced DNA damage (Additional file 3: Fig. S15) and reduced the number of γ-H2AX/53BP1-colocalized foci (Fig. 6a, b), respectively, in hepatocytes. Western blots of chromatin-bound fractionation and NHEJ assay showed that MDP treatment decreased the amount of chromatin-bound SITR6 (Fig. 6e) and diminished the activity of NHEJ pathway (Fig. 6f), respectively, in DEN-treated hepatocytes. All these effects were NOD2-dependent and could be restored by lamin A/C overexpression (Additional file 3: Fig. S15, Fig. 6c–f). Together, our results suggest that bacterial MDP reduces DNA repair and increases DNA damage through NOD2-lamin A/C axis.

To investigate whether lamin A/C is decreased in HCC, we next measured the lamin A/C protein level in the livers of DEN/CCl₄-treated mice. Lamin A/C protein gradually decreased in WT mice, but not Nod2^△hep mice, during the progression of DEN/CCl₄-induced HCC (Additional file 3: Fig. S16a). This decrease appeared to be NOD-dependent, because Nod2^△hep mice had higher lamin A/C level than Nod2^f/f mice (Additional file 3: Fig. S16b).

To determine the role of lamin A/C in NOD2-promoted hepatocarcinogenesis in vivo, we treated Lamn^△hep and Rip2/Lamn^△hep mice with MDP in DEN/CCl₄-induced HCC model. Compared with untreated group, MDP treatment cannot affect DNA damage (8-OHdG and H2AX expression) (Fig. 7a), but increased pro-inflammatory responses (expression of IL-6, TNF-α, IL-1β and IFN-γ, and the number of infiltrating macrophages) (Additional file 3: Fig. S17a, b), in lamin A/C^△hep mice. In addition, MDP treatment still moderately increased the tumor burden in lamin A/C^△hep mice (tumor number: 16.50 ± 2.43 vs. 22.75 ± 3.92, P < 0.01; tumor size: 6.58 ± 1.53 vs. 9.38 ± 1.51, P < 0.01) (Fig. 7b). Notably, hepatic double-knockout of Lamin A/C and Rip2 abolished MDP-promoted hepatocarcinogenesis, evidenced by unchanged tumor burden (tumor number: 11.80 ± 2.17 vs. 12.40 ± 2.30, P > 0.05; tumor size: 3.10 ± 1.19 vs. 3.80 ± 1.04, P > 0.05), DNA damage and pro-inflammatory responses in Rip2/Lamn^△hep mice (Fig. 7a, b, Additional file 3: Fig. S17a, b).

Finally, to validate our findings in human samples, we analyzed the association between NOD2 and lamin A/C in the ZHH1 cohort (n = 141) with IHC. We found that, in contrast to the increased NOD2 level (Fig. 1c, 1), lamin A/C level was decreased in HCCs, compared to their matched ANTs (average score 8.84 ± 3.582 vs. 9.76 ± 2.957, P = 0.001; Fig. 7c, d). Notably, NOD2-high tumors showed low lamin A/C expression and vice versa. Pearson correlation analysis demonstrated a negative correlation between NOD2 and lamin A/C protein level in HCC (Fig. 7e).

In line with our findings that NOD2 deficiency decreases genetic instability in DEN/CCl₄-treated HCC mice model (Fig. 4e), we found that NOD2-high HCCs had higher genetic instability than NOD2-low HCCs, as evidenced by a higher level of allelic imbalances (AI, 81.48% vs. 69.44%) at chromosomal common fragile sites (CFS) (Fig. 7f). Results were further confirmed in TCGA HCC cohort, in which we found a higher NOD2 expression in HCCs with mutated TP53 (a biomarker of genomic instability [29]) (Fig. 7g). Moreover, we also examined the mutation status of LMNA and TP53 in ZHH1 and TCGA HCC cohort. We found that LMNA presents a low mutation status in HCC [(2/141, 1.42%) in ZHH1; (5/359, 1.39%) in TCGA cohort]. LMNA-Mut groups showed increased TP53 mutation (1/2, 50% in ZHH1; 3/5, 60% in TCGA), compared with LMNA-Wt groups (24/139, 17.3%; 101/354, 28.53%), indicated that LMNA mutation could be associated with increased genomic instability. (Additional file 3: Fig. S18).

Our study still has limitation that the mechanism for a higher expression of NOD2 in HCC is still unclear. Previous studies have demonstrated that NOD2 promotor contains two NF-κB binding sites that enable the transcriptional activation of NOD2, following stimulation with TNF-α, which resulted in increased NOD2 mRNA and protein level [44‐46]. Moreover, this up-regulation is augmented by IFN-γ [44]. We estimated that up-regulation of NOD2 expression might be part of a positive regulatory loop induced by inflammatory cytokines or bacterial components, which needs the further investigation.