Introduction

Gastric cancer (GC) is the leading cause of cancer-related death worldwide [1, 2]. Despite the accelerated progress of the diagnosis and traditional treatments of GC, however, the biological and molecular mechanisms underlying GC development are still lacking. Therefore, identifying novel biomarkers and exploring therapeutic targets may shed light on individualized therapy strategies for GC patients.

Oxidative stress is the cellular pathogenic outcome resulting from an imbalance between reactive oxygen species (ROS) generation and the antioxidant capacity [3]. Aberrant oxidative stress occurs in most human cancers and contributes to cancer progression [3, 4]. The transcription factor nuclear factor (erythroid-derived 2)-like 2 (NRF2), as a prime target of cellular defense against oxidative stress, has long been identified to be a cytoprotective factor for its anti-inflammation and anti-cancer activities [4, 5]. However, its paradoxical activity has been recently received considerable controversies for its ‘‘dark side’’ profile in tumor progression [6,7,8,9,10,11,12,13]. Accumulating evidence described an oncogenic activity of aberrant NRF2 in cancer progression [7,8,9,10] and the suppressive role of NRF2 inhibitors in various human cancers [11,12,13]. However, the detailed mechanism underlying aberrant NRF2 activation possesses the “dark side” profile in cancer progression remains largely unknown.

Gene associated with retinoid-IFN-induced mortality 19 (GRIM-19), also named NDUFA13, as one of IFN/RA-inducible GRIM products, was originally identified as a potential tumor suppressor in various human cancers [14,15,16]. Subsequently, GRIM-19 was identified as an essential subunit of the mitochondrial respiratory chain (MRC) complex I [17, 18] and was reported to repress STAT3 activation via function interaction [19,20,21]. More recently, aberrant GRIM-19 expression was further observed in numerous human cancers [22,23,24,25,26,27,28,29,30]. Our previous study revealed that GRIM-19 was frequently downregulated in chronic atrophic gastritis (CAG) and its downregulation contributed to GC tumorigenesis via STAT3 [21]. However, the mechanism underlying GRIM-19 dysregulation and its role in GC metastasis have not yet been completely elucidated. Therefore, improved understanding of the role of GRIM-19 deficiency in GC metastasis should open new avenues and will develop novel therapeutic strategies for GC. In this study, we first identified dysregulation mechanism and the role of GRIM-19 in human GC metastasis. Then, we further investigated the possible relationship between GRIM-19 deficiency and aberrant NRF2 activation.

Materials and methods

Please see online supplemental data [10, 13, 21, 29,30,31,32, 34].

Results

Downregulated GRIM-19 predicts poor survival prognosis of human GC

To investigate the GRIM-19 expression profile in human GC, we evaluated GRIM-19 mRNA levels from NCBI GEO and TCGA STAD online database and found that GRIM-19 mRNA was significantly decreased in the primary GC tissues compared with non-tumorous tissues (Figure. S1A), suggesting deficient GRIM-19 expressions in human GC. However, we did not observe a significant difference of GRIM-19 mRNA level between intestinal-type and diffuse-type GC (Figure. S1B), indicating that GRIM-19 deficiency is independent of Lauren’s histologic type in GC [21]. To investigate the potential prognostic and predictive value of GRIM-19 in GC progression, we next analyzed the correlation of GRIM-19 expression with a serial of survival analyses in GC patients. We found that GRIM-19 downregulation was significantly correlated with shorter OS (n = 876, P = 3.3e-07) and PFS (n = 174, P = 0.038) (Figure. S1C). Furthermore, low GRIM-19 was related to shorter FPS (n = 641, P = 5.8e-07) and poor PPS (n = 499, P = 2.7e-06) (Figure. S1D). Notably, low GRIM-19 mRNA levels could distinguish a subset of GC patients with increased risk of poor DFS (n = 422, P = 0.014) and worse DMFS (n = 444, P = 0.017) (Figure. S1E), suggesting a clinical significance of GRIM-19 deficiency for the outcomes of GC patients. These data indicate that GRIM-19 may serve as a potential prognosis biomarker for the outcome of GC.

Genomic hypermethylation contributes to GRIM-19 downregulation in human GC

To ascertain the possible mechanism underlying GRIM-19 downregulation in human GC, we investigated genetic alterations of the GRIM-19 genome in a TCGA GC cohort with 34 normal cases and 415 GC cases. However, genetic analysis showed that only 5% primary GC cases (20/395) harbored the genetic mutations including truncating mutation, amplification, deep deletion and mRNA upregulation in the GRIM-19 genome (Fig. 1a), indicating that alternative mechanisms may be involved in GRIM-19 downregulation in human GC. To better understand whether DNA methylation is involved in aberrant GRIM-19 in GC, we analyzed the methylation status of GRIM-19 genome from this TCGA GC cohort, which was detected by Illumina Human Methylation 450 k Arrays with 21 methylation-specific probes for 21 CpG islands in four major regions of GRIM-19 gene (Fig. 1b). In this GC cohort, we observed that GRIM-19 mRNA expression was significantly downregulated in human GC (Fig. 1c) while total GRIM-19 methylation level of GC tissues was significantly higher than that of normal tissues (Fig. 1d). Furthermore, GRIM-19 mRNA expression was significantly correlated with its total methylation level in the GRIM-19 gene (Fig. 1e). Thus, these data suggesting that DNA hypermethylation is involved in GRIM-19 downregulation in GC.

Fig. 1
figure 1

Genomic hypermethylation contributes to GRIM-19 deficiency in human GC. a Genetic alteration analysis of GRIM-19 gene in human GC. Genetic alterations including truncating mutation, amplification, deep deletion and mRNA upregulation in the GRIM-19 genome were analyzed from primary GC cases (n = 395) according to the online TCGA database. b Schematic diagrams of CpG sites and methylation-specific probes in the GRIM-19 genome. The methylation of GRIM-19 gene was analyzed by the Illumina Human Methylation 450 k Array in 326 GC samples based on the online TCGA database. Four major regions harbouring 21 CpG islands were identified at TSS 1500 ( – 1500 bp–200 bp ahead of the transcription start site), TSS 200 ( – 200 bp–0 bp ahead of the TSS), Exon 1 (E1) and gene body regions in GRIM-19 gene and were detected by 21 methylation-specific probes. c–e Downregulation of GRIM-19 level is associated with hypermethylation of GRIM-19 gene in human GC. The difference of GRIM-19 mRNA (c), total methylation level (d) and their correlation (e) were analyzed in human GC cohorts. f Relative methylation difference of different CpG regions. Methylation difference between promoter region (TSS1500 &TSS200) and E1 & gene body region and methylation difference among TSS1500, TSS200, E1 and gene body regions were analyzed. g–h Correlation analysis of GRIM-19 mRNA level with methylation level of different CpG regions. GRIM-19 mRNA level and methylation level of different CpG islands were analyzed in the promoter region (TSS1500 &TSS200), TSS1500 region, TSS200 region (g), and E1 & gene body region, E1 region and gene body region (h), respectively. Relative methylation levels of different CpG islands in the GRIM-19 gene were assessed by a relative methylation ratio between the GRIM-19 gene and the keep-housing β-ACTB gene from the same sample. Data are presented as mean ± SD. *p < 0.05, **p < 0.01 and ***p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To further characterize the details of GRIM-19 methylation, we investigated relative methylation levels of four major CpG islands in the GRIM-19 genome including TSS1500, TSS200, Exon 1 and gene body region. Interestingly, we found relative methylation levels in exon 1 and gene body region were significantly higher than that of promoter region (TSS1500 & TSS200) while TSS1500 and gene body regions were significantly higher than that of TSS200 & exon 1 regions (Fig. 1f). Furthermore, GRIM-19 expression was significantly correlated with relative methylation levels of TSS1500 (Fig. 1g) and gene body region (Fig. 1h), indicating that hypermethylation of TSS1500 and gene body is involved in the regulation of GRIM-19 expression in human GC. Moreover, although the methylation data of one probe in gene body was absent in original data, we still observed that near 50% (6/12) methylation-specific probes in TSS1500 (5/8 probes, Figure. S3) and gene body region (1/4 probes, Figure. S4) presented significantly negative correlations between relative methylation levels and GRIM-19 expression, indicating that these CpG islands should be responsible for the hypermethylation of TSS1500 and gene body. Thus, these data suggest that hypermethylation-mediated GRIM-19 inactivation may contribute to its downregulation in human GC.

GRIM-19 deficiency facilitates GC metastasis in vitro and in vivo

To elucidate the functional role of GRIM-19 expression in GC metastasis, loss-of-function strategy by CRISP/Cas9 sgRNA-mediated GRIM-19 gene editing was conducted to knockdown GRIM-19 expression in SGC-7901 and MKN-45 cell lines (Figure. S2A–C), which have moderate GRIM-19 expression compared with normal GES-1 cells [21, 29]. We found that CRISP/Cas9-mediated sgRNA-134 presented a significant inhibition to GRIM-19 mRNA compared with negative control (NC) in three sgRNA sites (Figure. S2D) and generated a stable heteroplasmic deletion in GRIM-19 genome (Figure. S2E). Thus, SGC-7901 and MKN-45 cells expressing sgRNA-134 and NC were used for further analysis. Next, we investigated the impact of GRIM-19 deficiency on GC cell migration and invasion in vitro. We observed that GRIM-19 deficiency significantly enhanced the abilities of GC cell migration and cell invasion (Fig. 2a–b), suggesting that GRIM-19 deficiency promotes GC metastasis in vitro.

Fig. 2
figure 2

GRIM-19 deficiency facilitates GC metastasis in vitro and in vivo. a–b Deficiency of GRIM-19 facilitated GC cell migration and invasion in vitro. GRIM-19-deficient GC cells (SGC-7901–134, MKN-45–134) and corresponding NC cells were subject to cell migration assay (a) and cell invasion assay (b) using 24-well non-coated or Matrigel-coated transwell chambers (8-μm pore size). Cell numbers were counted in 5–10 random fields. c–e GRIM-19 deficiency enhanced GC distant lung and liver metastasis in vivo. GC cells (2 × 106 cells /100 µl PBS buffer/mice) were tail intravenously (T.V.) injected into the nu/nu mice (n = 5–8 mice/group). After the indicated time point, lungs and livers were collected, and visible metastatic tumor (indicated by black arrow) were counted and confirmed by H&E staining (c–d). Micro-metastasis marker vimentin was detected by IHC staining in liver and lung metastasis tissues (e). f–g GRIM-19 deficiency promoted GC peritoneal dissemination metastasis in vivo. MKN-45–134 cells or NC cells (1 × 106 cells /100 µl PBS buffer/mice) were peritoneally injected (i.p.) into the nu/nu mice (n = 5–8 mice/group). After indicated times, visible metastatic tumor (indicated by black arrow) were counted, and the liver metastasis tumor was confirmed by H&E staining (f) and vimentin expression was detected by IHC staining in liver metastasis tissues (g). Data are presented as mean ± SD of three independent experiments. Representative images are shown. Scale bar: 50 μm. *p < 0.05, **p < 0.01 and ***p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

In complementary in vivo distant metastasis mice model, we found that GRIM-19 deficiency significantly increased distant lung and liver metastasis of GC cells (Fig. 2c,d,e), suggesting GRIM-19 deficiency accelerates GC distant metastasis. Meanwhile, we further established a peritoneal dissemination metastasis mice model using GRIM-19-deficient MKN-45 cells, which presents a liver metastasis organotropism in mice [21, 29]. We found that GRIM-19 deficiency significantly enhanced peritoneal metastasis of MKN-45 cells in vivo (Fig. 2f), and markedly increased vimentin expression (a micrometastasis marker) in liver micrometastasis nodes (Fig. 2g). Thus, these data indicate that GRIM-19 deficiency increases GC migration and invasion, thereby facilitating metastasis in vivo.

ROS scavenger abrogates GRIM-19 deficiency-driven GC metastasis in vitro and in vivo

Since GRIM-19 is an essential subunit of the MRC complex I, we sought to know whether aberrant oxidative stress is involved in GRIM-19 deficiency-induced GC metastasis. As shown in Fig. 3a,b, we observed that GRIM-19 deficiency significantly enhanced intracellular ROS and mROS productions, as well as 8-oxo-deoxyguanosine (8-OHdG) expression, a ROS-induced DNA damage marker [10]. Furthermore, we also found that GRIM-19 deficiency significantly decreased NADP + /NADPH ratios but increased the GSH /GSSG ratio in GC cells (Fig. 3c). These data suggest that GRIM-19 deficiency triggers aberrant oxidative stress by ROS generation in GC cells.

Fig.3
figure 3

ROS scavenger reverses GRIM-19 deficiency-driven metastasis in vitro and in vivo. a–c GRIM-19 deficiency causes aberrant oxidative stress in GC cells. Intracellular ROS and mROS generations were detected by FCM (a) and DNA damage marker 8-oxo-deoxyguanosine (8-OHdG) was detected by IF staining (b). NADP + /NADPH and GSH/GSSG ratios were analyzed in GRIM-19 deficient GC cells (c). (d–e) ROS scavenger abrogates GRIM-19 deficiency-driven cell migration and invasion in vitro. SGC-7901–134 (1 × 105) and MKN-45–134 (2 × 105) GC cells were subject to cell migration (d) and cell invasion (e) assays upon NAC treatment (0 mM, 5 mM and 10 mM) for 24 h. Cell numbers were counted in 5–10 random fields. f–g ROS scavenger attenuates GRIM-19 deficiency-driven GC distant metastasis in vivo. Indicated cells were pre-treated with or without NAC (10 mM) and then T.V.-injected into nu/nu mice (3 × 106 /mice in 100 µl PBS, n = 5–8 mice/group). After 24 h, experimental mice were peritoneally injected (i.p.) with NAC (100 mg/kg) or 0.9% NaCl twice per week. At the indicated time point, visible lung and liver metastasis tumors were counted and confirmed by H&E staining (f). Vimentin expression was detected by IHC staining in lung and liver metastasis tissues (g). h–i ROS scavenger abrogates GRIM-19 deficiency-driven GC peritoneal metastasis in vivo. MKN-45–134 cells pre-treated by NAC (10 mM) were peritoneally injected (i.p.) into nu/nu mice (2 × 106 /mice in 100 µl PBS, n = 5–8 mice/group). 24 h after injection, experimental mice were i.p. injected with NAC (100 mg/kg) twice per week, as mentioned above. After the indicated time point, liver metastasis (indicated by black arrow) was confirmed by H&E staining (h), and vimentin expression was detected by IHC staining (i). Representative images are shown. Scale bars: 50 μm. Data are presented as mean ± SD. *p < 0.05, **p < 0.01 and ***p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To demonstrate the essential role of ROS in GRIM-19 deficiency-driven GC metastasis, we applied the ROS scavenger-NAC intervention to assess whether ROS scavenger could suppress GRIM-19 deficiency-driven GC metastasis. In vitro, we found that NAC treatment significantly abrogated GRIM-19 deficiency-driven cell migration and invasion with a dose-dependent manner (Fig. 3d, e). In distant organ metastasis mice, we further found that the lung and liver metastasis of GRIM-19-deficient SGC-7901 cells was significantly attenuated after peritoneal NAC administration (Fig. 3f). Furthermore, vimentin was markedly decreased in micrometastasis nodes of lung or liver tissues (Fig. 3g). In peritoneal metastasis model, peritoneal NAC administration also markedly attenuated mesenteric dissemination and liver metastasis of GRIM-19-deficient MKN-45 cells (Fig. 3h) and liver micrometastasis nodes were further confirmed by vimentin staining (Fig. 3i). Therefore, these results suggest that aberrant ROS play an essential role in GRIM-19 deficiency-driven GC metastasis.

GRIM-19 deficiency induces aberrant NRF2 activation via ROS-dependent manner

ROS has attracted many concerns not only for its role in oxidative stress but also for its vital second messenger in signal transduction like NRF2 pathway [10]. Therefore, we investigated whether GRIM-19 deficiency could induce aberrant NRF2 via ROS. In GRIM-19-deficient GC cells, GRIM-19 deficiency significantly increased NRF2 and p-NRF2 expression (Fig. 4a,b) and antioxidant response element (ARE)-driven NRF2 transcriptional activation (Fig. 4c). Furthermore, upregulated NRF2 expression was confirmed in lung and liver metastasis tissues (Fig. 4d), suggesting that GRIM-19 deficiency causes aberrant NRF2 activation in human GC. Moreover, the qRT-PCR analysis showed that GRIM-19 deficiency not only significantly upregulated NRF2-responsive gene expressions, but also significantly enhanced a serial of metastasis-associated gene levels including HIF-1α, COX-2, VEGF-A, VEGF-C, MMP-9 and BACH-1 in GC cells (Fig. 4e). Together, these results suggest that GRIM-19 deficiency-induced aberrant NRF2 activation is associated with GC metastasis.

Fig.4
figure 4

GRIM-19 deficiency induces aberrant NRF2 via a ROS-dependent manner. a–e GRIM-19 deficiency enhances NRF2 activation of GC cells in vitro and in vivo. NRF2 and p-NRF2 expressions in GRIM-19-deficient SGC-7901 and MKN-45 cells were detected by western blotting (a) and IF staining (b), respectively. NRF2/ARE transcriptional activation was analyzed by luciferase reporter gene assay (c). NRF2 proteins in metastasis tissues were analyzed by IHC staining (d), and NRF2-responsive genes and metastasis-associated genes levels were detected by Real-time PCR (e). f–i ROS scavenger attenuates GRIM-19 deficiency-driven NRF2 activation in vitro and in vivo. GRIM-19 deficient GC cells were treated with or without ROS scavenger-NAC. NRF2 and p-NRF2 expressions were detected by western blotting (f) while NRF2/ARE transcriptional activation was analyzed by luciferase reporter gene assay (g). NRF2 proteins in lung and liver metastasis tissues were analyzed by IHC staining (h). NRF2-responsive genes and metastasis-associated gene levels were detected by Real-time PCR (i). β-actin was as an internal control. Representative images are shown. Scale bars: 50 μm. Data are presented as mean ± SD. *p < 0.05, **p < 0.01 and ***p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To further clarify the critical role of ROS in GRIM-19 deficiency-driven NRF2 activation, we investigated whether ROS inhibition could reverse GRIM-19 deficiency-driven NRF2 activation. By NAC intervention, we found that NAC treatment significantly attenuated NRF2 and p-NRF2 levels (Fig. 4f) as well as NRF2/ARE luciferase activation in GRIM-19-deficient GC cells (Fig. 4g). Furthermore, NRF2 protein was markedly reduced upon NAC treatment in lung and liver metastasis tissues (Fig. 4h), and NRF2-responsive and metastasis-associated gene levels were significantly attenuated after NAC treatment in GRIM-19-deficient GC cells (Fig. 4i). Collectively, these data indicate that GRIM-19 deficiency induces aberrant NRF2 activation via ROS-dependent manner.

NRF2 activation plays an oncogenic role in GC metastasis

To investigate the functional role of NRF2 activation in GRIM-19 deficiency-driven GC metastasis, we applied ML-385, a specific NRF2 inhibitor to explore whether pharmaceutical NRF2 inhibition could abrogate GRIM-19 deficiency-driven GC metastases in vitro and in vivo. In vitro assays, we found that ML-385 treatment significantly abrogated GRIM-19 deficiency-driven GC cell migration and invasion with a dose-dependent manner (Fig. 5a,b). Meanwhile, qRT-PCR showed that ML-385 treatment significantly attenuated NRF2-responsive genes and metastasis-associated gene levels in GRIM-19-deficient GC cells (Fig. 5c,d), indicating that NRF2 inhibition reverses GRIM-19 deficiency-driven GC metastases in vitro. In distant metastasis mice model in vivo, we observed that the treatment of nontoxic low-dose ML-385 (30 mg/kg, once per week) significantly attenuated lung and liver metastasis (Fig. 5e) and vimentin expression in metastasis tissues (Fig. 5f). Moreover, in peritoneal metastasis mice, we found that ML-385 intervention markedly reduced mesenteric dissemination and liver metastasis of GRIM-19-deficient MKN-45 cells (Fig. 5g,h). Thus, these data indicate an oncogenic role of NRF2 in GRIM-19 deficiency-driven GC metastasis.

Fig. 5
figure 5

Pharmaceutical NRF2 blockage reverses GRIM-19 deficiency-driven metastasis in vitro and in vivo. (a,b) NRF2 blockage attenuates GRIM-19 deficiency-driven cell migration and invasion in vitro. GRIM-19-deficient GC cells were subject to cell migration (a) and cell invasion assay (b) upon treatment with NRF2 specific inhibitor-ML-385 (0, 5 µM, 10 µM). Migration or invasion cells were counted in 5–10 random fields. c,d NRF2 inhibition reverses GRIM-19 deficiency-driven NRF2 downstream genes and metastasis genes levels. GRIM-19 deficient GC cells were treated with ML-385 (0, 5 µM, 10 µM). NRF2-responsive genes and metastasis-associated gene levels were detected by real-time PCR. β-Actin gene was as an internal control. e,f Pharmaceutical NRF2 inhibition attenuates GRIM-19 deficiency-driven distant metastasis in vivo. GRIM-19 deficient SGC-7901 cells were pre-treated with or without ML-385 and then were tail intravenously (T.V.) injected into nu/nu mice (n = 5–8 mice/group). 24 h after injection, experimental mice were peritoneally injected (i.p.) with or without ML-385 (30 mg/kg) once per week as described in M&M. At the indicated time point, the visible metastatic tumors from lung and liver (indicated by black arrow) were counted and confirmed by H&E staining (e). Vimentin expression was detected by IHC staining (f). g,h Pharmaceutical NRF2 inhibition abrogates GRIM-19 deficiency-driven GC peritoneal metastasis in vivo. GRIM-19-deficient MKN-45 cells pre-treated with or without ML-385 (10 µM) for 24 h were peritoneally injected (i.p.) into nu/nu mice (n = 5–8 mice/group). After injection, experimental mice were treated with or without ML-385 (30 mg/kg) once per week as described above. After indicated time point, the visible metastatic tumors on the liver surfaces (indicated by black arrow) were counted and confirmed by H&E staining (g), and vimentin protein was detected by IHC staining (h). Data are presented as mean ± SD. Representative images are shown. Scale bars: 50 μm. *p < 0.05, ** < 0.01 and ***p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To better define the exact role of NRF2 in GC metastasis, we directly investigated the effect of NRF2 activation or inactivation on cell migration and invasion in parental GC cells. By ML-385 treatment, we observed that pharmacological NRF2 inhibition significantly suppressed GC cell migration and cell invasion with a dose-dependent manner (Figure. S5A-B). Meanwhile, NRF2-responsive targets and metastasis-associated gene levels were significantly reduced upon ML-385 treatment in these parental GC cells (Figure. S5C-D), suggesting that NRF2 inhibition can suppress GC metastasis. Next, using α-lipoic acid (ALA), a well-characterized NRF2 activator, we found that pharmacological NRF2 activation significantly promoted cell migration and cell invasion with a dose-dependent manner in parental GC cells (Figure. S6A-B). Meanwhile, we also found that ALA treatment markedly enhanced NRF2-responsive targets as well as those metastasis-associated genes expressions in parental GC cells (Figure. S6C-D), indicating that NRF2 activation can promote GC metastasis. Taking together, our results strongly suggest that NRF2 activation plays an oncogenic role in GC metastasis.

Heme oxygenase 1 (HO-1) is a critical player of oncogenic ROS-NRF2 axis-driven GC metastasis in GRIM-19-deficient GC

Although stress-responsive GCLM, NQC-1 and HO-1 proteins are well-known NRF2-responsive targets, whether these targets are involved in NRF2-mediated GC metastasis remains largely unknown. Recent studies reported aberrant HO-1 activation in many different types of human cancers [35,36,37,38,39], promoting us to focus on the role of HO-1 in GC metastasis. Using western blot and IHC analyses, we found that GRIM-19 deficiency significantly increased HO-1 expression in vitro (Fig. 6a) and metastatic lung and liver tissues in vivo (Fig. 6b) while these aberrant HO-1 expressions were significantly abrogated upon NAC treatment (Figure. S7A-B) or ML-385 intervention (Figure. S8A-B) in vitro and in vivo, indicating that HO-1 activation is dependent on ROS-NRF2 axis in GC cells.

Fig. 6
figure 6

Pharmaceutical HO-1 inhibition abrogates GRIM-19 deficiency-driven GC metastasis in vitro. a,b GRIM-19 deficiency enhances HO-1 levels in GC cells in vitro and in vivo. The expressions of HO-1 protein were detected by western blotting in GRIM-19-deficient GC cells (a) and IHC staining in lung and liver metastasis tissues (b), respectively. c HO-1 inhibition attenuates GRIM-19 deficiency-driven metastasis-associated genes levels. GRIM-19 deficient GC cells were treated with HO-1 specific inhibitor (0, 5 and 10 µM) for 6 h and metastasis-associated genes were detected by real-time PCR. β-Actin was an internal control. d,e HO-1 blockage reverses GRIM-19 deficiency-driven cell migration and cell invasion in GC cells. GRIM-19 deficient GC cells were subject to cell migration (c) and cell invasion assay (d) upon treatment with specific HO-1 inhibitor (0, 5 µM, 10 µM). Migration or invasion cells were counted in 5–10 random fields. Data are presented as mean ± SD of three independent experiments. Scale bar: 50 μm. Representative images are shown. *p < 0.05, ** p < 0.01 and *** p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To further test the role of HO-1 in GC metastasis, we investigated whether HO-1 inhibition could suppress parental GC metastasis, even reverse GRIM-19 deficiency-driven GC metastasis. By treatment of HO-1-IN-1 hydrochloride, a specific HO-1 inhibitor, we found that HO-1 inhibition significantly decreased GC cell migration and invasion (Figure. S9A-B) in parental GC cells, indicating an oncogenic role of HO-1 activation in GC metastasis. Moreover, we also found that HO-1 inhibition significantly downregulated HO-1 mRNA level and metastasis-associated gene expressions in GRIM-19-deficient GC cells (Fig. 6c). Notably, we observed that HO-1 inhibition significantly abrogated GRIM-19 deficiency-driven cell migration and invasion in vitro (Fig. 6d,e), suggesting that HO-1 inhibition could reverse GRIM-19 deficiency-driven GC metastasis. Collectively, these data suggest that oncogenic HO-1 activation is required to ROS-NRF2 axis-driven GC metastasis in GRIM-19-deficient GC cells.

HO-1 promotes GC metastasis by activating NRF2 via a positive-feedback loop

Given the oncogenic role of NRF2 and HO-1 in GRIM-19 deficiency-driven GC metastasis, we further defined the possible mechanism underlying how NRF2-responsive HO-1 activation promotes GC metastasis. In parental GC cells, we observed that HO-1 inhibitor significantly inhibited both HO-1 and NRF2 expressions and NRF2/ARE transcriptional activation (Figure. S10A-B). Meanwhile, HO-1 inhibitor treatment significantly abrogated NRF2-responsive targets and metastasis-associated gene expressions (Figure. S10C), indicating a feedback regulation of HO-1 to NRF2 activation in GC cells. Furthermore, we observed that the HO-1 inhibitor not only decreased HO-1 expression but also significantly attenuated the GRIM-19 deficiency-driven NRF2 level in GRIM-19-deficient GC cells (Fig. 7a,b), indicating that HO-1 inhibition can directly abrogate GRIM-19 deficiency-driven NRF2 activation. These data indicate that HO-1 may regulate NRF2 activation via a positive feedback in GC cells.

Fig. 7
figure 7

HO-1 inhibition antagonizes NRF2 activation via a positive-feedback loop. a,b HO-1 blockage abrogates GRIM-19 deficiency-driven NRF2 activation in vitro. GRIM-19 deficient GC cells were treated with HO-1 specific inhibitor (0, 5 µM,10 µM). HO-1 and NRF2 expression were analyzed by western blotting (a) and NRF2/ARE transcriptional activation was detected by luciferase reporter gene assay (b). c,d HO-1 inhibition attenuates ALA-induced GC metastasis in vitro. Parental GC cells were co-treated with or without α-Lipoic Acid (ALA) (0 mM or 0.2 mM) and HO-1 specific inhibitor (0 µM, 5 µM,10 µM). Cell migration (c) and cell invasion assay (d) were performed. Migration or invasion cells were counted in 5–10 random fields. e,f HO-1 blockage reverses ALA-induced NRF2 activation. Parental GC cells were co-treated with or without ALA and HO-1 inhibitor, as mentioned above. HO-1 and NRF2 expression were analyzed by western blotting (E), and NRF2/ARE transcriptional activation was detected by luciferase reporter gene assay (f). g HO-1 inhibition abrogates ALA-induced NRF2 target genes and metastasis-associated gene levels. Parental GC cells were co-treated with or without ALA and HO-1 inhibitor as mentioned above. NRF2 target genes and metastasis-associated gene levels were detected by real-time PCR. β-actin acted as an internal control. Representative images are shown. Data are presented as mean ± SD of three independent experiments. *p < 0.05, ** p < 0.01 and *** p < 0.001 between the indicated two groups determined by unpaired student’s t-test

Full size image

To investigate whether HO-1 inhibition could attenuate NRF2-driven GC metastasis, we first conducted a double intervention assay using HO-1 inhibitor and NRF2 activator in parental GC cells. As shown in Fig. 7c,d, ALA-induced cell migration and invasion were significantly reversed by HO-1 inhibitor in parental GC cells, suggesting HO-1 inhibition can directly antagonize NRF2-induced GC metastasis in vitro. Furthermore, we observed that HO-1 inhibition significantly abrogated ALA-induced NRF2 and HO-1 activations (Fig. 7e) as well as NRF2/ARE transcriptional activation (Fig. 7f). Moreover, we also found that HO-1 inactivation significantly decreased ALA-induced NRF2 downstream targets and metastasis-associated gene levels in GC cells (Fig. 7g), indicating a positive-feedback loop between HO-1 and NRF2. Therefore, these results indicate that HO-1 activation maintains oncogenic NRF2 activation via a positive-feedback loop, promoting a serial of metastasis-associated gene expressions, thereby leading to GC metastasis.

Discussion

Elucidating the critical role of oxidative stress in cancer metastasis may shed light on a novelty diagnostic and therapeutic target for GC treatment. Here, we identified that genomic hypermethylation contributed to the frequent downregulation of GRIM-19 in GC tissues, and its downregulation was correlated with poor survival in GC patients. Functionally, we revealed that GRIM-19 deficiency promoted GC metastasis in vitro and in vivo, which can be abrogated by the pharmacological inhibition of ROS, NRF2 and HO-1 respectively. Mechanistically, GRIM-19 deficiency-induced excessive oxidative stress by triggering ROS production, resulting in oncogenic NRF2 activation and NRF2-responsive HO-1 expression. Moreover, aberrant HO-1 activation promoted oncogenic NRF2 activation via a positive feedback loop, promoting a series of metastasis-associated gene expression. Our study provides a causal relationship of GRIM-19 deficiency and aberrant NRF2 activation in GC metastasis.

Our finding reveals that DNA hypermethylation rather than genetic alteration contributes to GRIM-19 downregulation and its downregulation predicts poor survival prognosis in human GC, offering a potential prognostic biomarker for human GC. Our previous finding revealed that GRIM-19 protein was downregulated in GC tissues and precancerous CAG lesions [21]. Herein, we used online database with expanded GC cohorts to explore the correlation of GRIM-19 mRNA and survival prognosis. As expected, we also found that GRIM-19 mRNA was significantly decreased in GC tissues but was independent of intestinal-type and diffuse-type in GC. These results are consistent with our previous finding that GRIM-19 downregulation was correlated with clinical stage and lymph node metastasis but was independent of Lauren’s histologic type [21], further indicating that GRIM-19 a potential prognostic biomarker for human GC. A recent finding suggested that the promoter hypermethylation (± 1 kb from TSS) is involved in GRIM-19 downregulation in head and neck squamous cell carcinoma [28]. In this study, we also revealed that GRIM-19 hypermethylation rather than genetic alteration is correlated with GRIM-19 downregulation. Furthermore, we identified two core hypermethylation regions (promoter TSS1500 and gene body region) are correlated with GRIM-19 downregulation. Therefore, DNA hypermethylation may provide mechanism insights of GRIM-19 downregulation in human GC.

Our finding demonstrates that GRIM-19 deficiency-induced oxidative stress promotes GC metastasis via triggering ROS generation, improving our understanding of the role of aberrant oxidative stress in the tumor progression. Although our previous work revealed a suppressive aspect of GRIM-19 overexpression to GC metastasis [29], whether GRIM-19 deficiency should be responsible for the frequent metastasis in advanced GC has not yet been defined. ROS, a normal byproduct of cellular metabolism, has attracted many concerns for its vital cell signaling messenger in the regulation of signal transduction pathways. GRIM-19 has been identified to be a nuclear-encoded essential subunit of the MRC complex I, which was shown to be the main source of ROS generation in intact mammalian mitochondria in vitro [25]. Here, we demonstrated that GRIM-19 deficiency promoted GC metastasis in vitro and in vivo by triggering aberrant oxidative stress via ROS production while ROS scavenger significantly attenuated GRIM-19 deficiency-induced GC metastasis in vitro and in vivo. Our data strongly suggest that ROS-mediated aberrant oxidative stress plays an essential role in GRIM-19 deficiency-driven GC metastasis, further improving our understanding of the role of aberrant oxidative stress in the GC progression. However, recent contradicting results have raised the possibility that ROS limits distant metastasis [6]. There is a possibility that specific characteristics of antioxidants and cell lines with different gene mutations may contribute to the contradictory results regarding the role of ROS and antioxidant supplementation in tumor metastasis [10]. Further investigations will be necessary to dissect the role of oxidative stress in cancer metastasis in different tumor types and models.

Our study reveals an oncogenic NRF2 activation in GRIM-19 deficiency-associated GC metastasis, improving our understanding of the “dark side” of NRF2 activation in tumor progression. NRF2 has long been identified as a well-known anti-cancer molecular and a serial of antioxidants suppresses tumor progression [4, 5]. However, recent evidence indicates the “dark side” profile of NRF2 activation in tumor progression [6,7,8,9,10,11,12,13] and shows that aberrant NRF2 accelerates cancer progression while antioxidants accelerate migration and invasion of cancer cells [9, 10, 35]. In addition, our recent finding also revealed that NRF2-mediated anti-oxidative stress deteriorated tumor metastasis while NRF2 knockdown decreased HCC metastasis [10]. Here, our data demonstrated that GRIM-19 deficiency promoted GC metastasis by ROS-dependent NRF2 activation while specific ROS or NRF2 inhibition abrogated GRIM-19 deficiency-driven GC metastasis. Moreover, NRF2 activator ALA promoted GC metastasis, whereas NRF2 specific inhibitor ML-385 suppressed GC metastasis, which is consistent with recent findings that NRF2 inhibitors could antagonize human cancers [11,12,13]. Interestingly, our data also showed that GRIM-19 deficiency markedly increased hemangiogenesis and lymphangiogenesis in metastatic tissues, which can be significantly abrogated by ROS or NRF2 inhibition (Figure. S11), indicating hemangiogenesis and lymphangiogenesis are involved in ROS-NRF2 axis-driven metastasis in GRIM-19-deficient GC cells. Therefore, our results strongly suggest that NRF2 plays an oncogenic role in GRIM-19 deficiency-driven GC metastasis.

Our finding describes that NRF2-responsive HO-1 contributes to oncogenic NRF2 activation in GC metastasis via a positive-feedback NRF2-HO-1 loop, offering new avenues to develop novel therapeutic strategies for counteracting oncogenic NRF2 by targeting HO-1. Our results demonstrate that NAC suppresses GC metastasis, whereas NRF2 activation promotes GC metastasis, raising a topic concerning clarification of the paradoxical role of antioxidants in tumor metastasis. HO-1, a well-known NRF2-responsive target, plays a pivotal role in the maintenance of cellular redox homeostasis [35]. However, recent evidence revealed that aberrant HO-1 in many different types of human tumors plays an oncogenic role in tumor metastasis [36,37,38,39, 41]. Here, our data showed that the pharmacological inhibition of ROS or NRF2 attenuated HO-1 expression while HO-1 inhibition abrogated GC metastasis in GRIM-19-deficient or parental GC cells, suggesting that NRF2-responsive HO-1 activation is required to oncogenic ROS-NRF2 axis-mediated GC metastasis. Notably, our findings further showed that HO-1 inhibition not only antagonized ALA-promoted GC metastasis but also blocked NRF2 activation, indicating a positive-feedback NRF2-HO-1 loop. Moreover, recent evidence showed that oncogenic HO-1 activation is associated with a serial of metastasis-associated proteins, including MMP-9, VEGF-A, HIF-1α and BACH-1 [36,37,38,39, 42,43,44,45]. The MMP-9 and VEGF-A were upregulated in HO-1 overexpressing human GC cells but decreased in HO-1-inhibited cells [38]. Our data showed that HO-1 inhibition also blocked these metastasis-associated gene expressions in parental or GRIM-19-deficient GC cells. Thus, HO-1, as a critical downstream player of the ROS-NRF2 axis, feedback promotes oncogenic NRF2 activation, promoting metastasis-associated genes expression, thereby facilitating GC metastasis. Our finding provides a clinical caution to administrate NRF2 antioxidants in GC patients, and more comprehensive preclinical and clinical studies should be performed to ensure the safety of antioxidants in cancer patients.

In summary, our data suggest that GRIM-19 deficiency accelerates GC metastasis through the oncogenic ROS-NRF2-HO-1 axis via a positive-feedback NRF2-HO-1 loop (Figure. S12). This finding not only offers a mechanistic insight into the oncogenic NRF2 activation in tumor progression, but also provides new avenues to counteract the “dark side” of NRF2 by targeting HO-1.