Ryanodine receptor 2 promotes colorectal cancer metastasis by the ROS/BACH1 axis

There is no targeted therapy for KRAS proto‐oncogene, GTPase (KRAS)‐mutant metastatic colorectal cancer (mCRC) because the underlying mechanism remains obscure. Based on bioinformatic analysis, this study aims to elucidate a potential gene target for which an approved drug is available, and to reveal the function as well as the underlying mechanism of the candidate gene. Here, we identified that ryanodine receptor 2 (RyR2) expression was upregulated in KRAS‐mutant mCRC, and that this promoted cancer cell metastasis. S107, an approved drug to inhibit calcium release from RyR2 in the clinic, inhibited cancer cell metastasis both in vitro and in vivo. High expression of RyR2 predicts poor survival in our patient cohort. CRC patients with serosa invasion and vascular tumor thrombus are characterized by high RyR2 expression. Analysis of expression profiles upon RyR2 knockdown and inhibition, revealed a set of metastasis‐related molecules, and identified BTB domain and CNC homolog 1 (BACH1) as the main transcription factor regulated by RyR2. RyR2 regulates cellular reactive oxygen species (ROS) levels, which activates nuclear factor erythroid 2‐related factor 2 (Nrf2; also known as NFE2L2) and HMOX1 expression, and thus BACH1 accumulation. Collectively, this study provides evidence that the RyR2/ROS/BACH1 axis may be a potential intervention target for CRC metastasis.


Introduction
Colorectal cancer (CRC) represents one of the most common digestive tract malignancies worldwide [1]. Half the diagnosed cases will develop into metastatic colorectal cancer (mCRC), which is the final step of cancer development and the main cause of CRC-related death [2]. Given the pervasive activation of the EGF receptor (EGFR) signaling pathway in CRC and mCRC, the EGFR antibodies cetuximab and panitumumab are approved to treat mCRC together with the traditional chemotherapy including folinic acid, fluorouracil and irinotecan (FOL-FOX) and folinic acid, fluorouracil and oxaliplatin (FOL-FIRI) [3]. However, mCRC patients with KRAS activation mutation (generally at codons 12 and 13) are excluded from this treatment modality, resulting in a poor prognosis [4]. An appealing discovery of AMG510 as a specific inhibitor of KRASG12C in 2019 is encouraging, although this type of mutation represents only 3% of CRC patients [5]. Thus proper intervention and treatment are still both urgently needed for this subset of mCRC patients. However, this is hindered by a lack of knowledge of the mechanisms underlying metastasis formation of KRAS mutant patients.
Calcium regulates diverse biological processes including cancer metastasis [5]. An early study showed that STIM/ORAI, which mediates extracellular calcium fluxing into endoplasmic reticulum (ER) to refill the intracellular calcium reservoir, is critical for breast cancer metastasis [6]. The next decade witnessed numerous studies demonstrating the crucial role of calcium in cancer progression, which involves a great number of molecules, including calcium channels, transporters as well as calcium binding proteins, and diverse pathways regulated by intracellular calcium levels [7]. However, the cellular outputs triggered by calcium changes are complex due to both the pattern of calcium stimulation and the cellular calcium toolbox. To elicit a calcium-related signal, extracellular stimuli converge on calcium channels residing in ER [8]. ER-resident calcium channels consist of two families, the inositol triphosphate receptors 1-3 (IP3R1-3) and the ryanodine receptors 1-3 (RYR1-3) [9]. IP3R bind intracellular inositol 1,4,5 triphosphate (IP3), the end product of phospholipase C (PLC), and release calcium from ER to cytosol [10]. RyR show a tendency for expression in the brain, heart and muscle, where they sense membrane potential changes [11]. As such, IP3R are considered functional in non-excitable cells, whereas RyR are mainly involved in excitable cells, including nerve and muscle cells.
In this study, we found that RyR2 was upregulated in KRAS mutant mCRC patients, and RyR2 high expression conferred poor survival. RyR2 gene silencing by knocking down or pharmacological inhibition by small molecule S107 decreased cancer cell metastasis both in vitro and in vivo. A mechanism study uncovered genes downstream of RyR2, and enriched BACH1 as the main transcription factor (TF) affected by RyR2 inhibition. Further research revealed that RyR2 regulated BACH1 levels via modulating cellular reactive oxygen species (ROS) level. Collectively, the results of this study provide the therapeutic possibility that RyR2 targeting by S107 may be repurposed to intervene with metastasis of CRC patients.

Cell culture
HEK293T and SW480, HCT116, HT29 and CT26 were purchased from Cell Bank of Type Culture Collection of Chinese Academy of Sciences, Shanghai Institute of Cell Biology, Chinese Academy of Sciences. All cells were grown in DMEM (Gibco, Carlsbad, CA, USA) supplemented with 10% FBS (Anlite, Shanghai, China) and 1% penicillin/streptomycin (Sangon Biotech, Shanghai, China), except for CT26 cells, which was cultured with RPMI-1640 (Gibco). Cell lines were incubated at 37°C in a humidified 5% CO 2 atmosphere.

CRC sample and tissue microarray
A total of 195 patients diagnosed pathologically with colorectal cancer were enrolled in this study. Samples were collected at The First People's Hospital of Huzhou from 2018 to 2019 in compliance with the protocol for tissue collection approved by the Ethics Committee of First Affiliated Hospital, Huzhou University (approval number: 2020KYLL002). All methodologies conformed to the standards set by the Declaration of Helsinki. Informed consent was signed by all patients and all experiments were approved by the Ethical Committee. Detailed pathological examination parameters were recorded and available for analysis. Tissue microarray (TMA) was constructed from CRC samples embedded in paraffin, and stained with anti-RYR2 (diluted 1 : 100), followed by scanning, photographing and scoring using the Vectra2 system (PerkinElmer, Waltham, MA, USA).

RNA isolation and real time PCR
RNA extraction and cDNA preparation have been described earlier [12]. Briefly, total RNA was isolated

Immunohistochemistry and immunofluorescence
The procedures of these two assays have been described earlier [12]. Anti-RYR2 (Proteintech) was diluted 1 : 100 to stain CRC TMA slides. Anti-phospho-CREB(Ser133) (Affinity, Changzhou, China) was diluted 1 : 100 in an immunofluorescence (IF) assay. Briefly, for immunohistochemistry (IHC), sections of clinical specimens were deparaffinized with xylene and rehydrated with ethanol, followed by staining with anti-RYR2 antibody and horseradish peroxidase (HRP)-linked anti-rabbit IgG, and further developed with 3,3 0 -diaminobenzidine (DAB). TMA sections were further photographed and analyzed by Vecture 2 (PerkinElmer). The same algorithm was used to score every core. For IF, cells on slides were fixed in 4% formaldehyde, followed by staining with indicated primary antibodies and fluorescent secondary antibody (Alexa Fluor 488-conjugated donkey anti-rabbit IgG and Alexa Fluor 555-conjugated goat anti-rabbit IgG, 1 : 1000) and were photographed with a confocal microscope (Zeiss, Cambridge, UK; LSM 880NLO FILM).

Western blot and immunoprecipitation
Western blot analysis was performed as described previously [12]. 3xFlag-BACH1 immunoprecipitation was performed according to the anti-Flag M2 manual. Briefly, 2 days post transfection, HEK293T cells were washed twice in ice-cold PBS, and lysed in lysis buffer (50 mM Tris-HCl, pH = 7.4 with 150 mM NaCl, 1 mM EDTA and 1% Triton X-100) with protease inhibitors for 10 min on ice, after centrifugation at 20 000 g for 15 min at 4°C. Supernatants were incubated with anti-Flag M2 beads on a rotator overnight in cold room. After incubation, the beads were pelleted and washed five times in TBS (50 mM Tris-HCl, 150 mM NaCl, pH = 7.4), then elution with 3xFlag peptides for 1 h. The eluate was resolved by SDS/PAGE western blot.

In vitro migration assay
The migration assay was performed as described previously [12]. In vitro migration assay was conducted in 24-well inserts (Corning Inc., Corning, NY, USA). Briefly, when cells reached a confluence of 70-90%, they were trypsinized, and 1 9 10 6 cells were seeded on the top and complete medium was added to the bottom. After incubation for 48 h, non-invasive cells were removed from the upper end of the insert with a cotton swab. The bottom cells (invasive cells) were fixed with 4% paraformaldehyde for 20 min, stained with a 0.1% crystal violet solution for 30 min, and photographed using a microscope. The number of cells was counted, and data were presented as the means of three randomly selected fields.

RNA-seq analysis of transcriptome
A total of 3 9 10 6 SW480 cells were seeded into 10-cm dishes at two replicates; for the S107 treatment group, the final S107 concentration was 10 lM. Two days later, cells were washed with ice-cold PBS three times, followed by dissociation in 1 mL TRIzol reagent. Samples were further subjected to RNA extraction, library construction and high-throughput sequencing (Illumina NovaSeq; Illumina Inc., San Diego, CA, USA). Raw data were quality controlled and mapped to the genome by HISAT2, followed by transformation to FPKM. DESeq was used to calculate deregulated genes (DEG) with¦log 2 FoldChange¦ > 1 and P < 0.05. To obtain DEG between RYR2 low and RYR2 high in the Cancer Genome Atlas (TCGA) data, we first scored the mRNA expression data of TCGA-COAD and TCGA-READ using the R package ESTIMATE to acquire immune and stroma score for every patient. We chose the relative pure tumor tissues by selecting total score < À500 and immune score < À100. Respectively, 22 and 12 patients were grouped into RYR2 low and RYR2 high . DEG were obtained using the R package LIMMA. All heatmap pictures were drawn by the R package PHEATMAP. TF prediction was performed using the DAVID website.

Dual luciferase reporter assay
Reporter sequences for BACH1, CHX10, FREAC7, HFH3 and SOX9 were synthesized and inserted into backbone plasmid; sequences for each TF can be found in the Supplemental Table S2. To perform dual luciferase reporter assay, 0.1 lg reporter plasmid and 0.02 Renilla luciferase vector were transformed into cells using lipo3000 reagent for three replicates. Two days after transfection, the measurements were complied using the instruction manual.

Animal studies
BALB/c mice (Shanghai SLAC Laboratory Animal Co. Ltd., Shanghai, China) were housed according to a 12-h light (7 a.m.)/dark (7 p.m.) cycle at 25°C, with ad libitum access to water and rodent standard chow diet and maintenance under pathogen-free conditions. All animal studies were approved by and performed in accordance with Institutional Animal Care and Use Committee of Shanghai Institutes for Nutritional Sciences, Chinese Academy of Sciences (SIBS-2018-XD-3). Metastatic murine colorectal cancer cell line CT26 was used in this study. For the weekly luciferase reporter assay, mice were injected intraperitoneally with D-luciferin. They were anesthetized by isoflurane and photographed in the IVIS imaging system (Xenogen, Alameda, CA, USA). For lung metastasis, 4 9 10 5 cells were injected into the tail vein of each nude mouse. Two weeks post injection, all mice were sacrificed and lungs were subjected to fixation in formalin followed by embedding in paraffin and then H&E staining.
For splenic injection, 5 9 10 5 cells were injected into mouse spleen. S107 was used 2 days before the surgery and was injected intraperitoneally every other day at a concentration of 30 mgÁkg À1 . Ten days post injection, the mice were sacrificed and their livers excised.

Heme detection
Cell-free heme was determined by the colorimetric method using a commercially available kit (MAK316-1KT; Sigma-Aldrich, St. Louis, MO, USA). Briefly, cells in a 10-cm dish with a confluence of about 80% were washed with ice-cold PBS, followed by collection of cells with a scrapper. Cells were pelleted and resuspended in lysis buffer (PBS with protease inhibitor), followed by cellular disruption by sonication. The remaining steps followed those in the manufacturer's manual.

Statistical analysis
All data are presented as the mean AE standard error of the mean. Student's t-test was used for the comparison of measurable variants of two groups. All experiments were performed with at least three biological duplicates (n = 3) for each group. Survival curves were calculated using the Kaplan-Meier method, and differences were assessed by a log-rank test. The criterion for significance was P < 0.05 for all comparisons.

RyR2 upregulation is related to colorectal cancer metastasis behavior
To identify differentially expressed genes in KRAS mutant metastatic CRC patients, we analyzed TCGA CRC data using the following strategy. Patients were first divided into two groups according to their metastasis status. For each group, patients were further separated by KRAS status. Candidate genes should be upregulated in KRAS mutant mCRC patients compared with KRAS WT mCRC patients, but showed no significant changes in the non-metastasis group. These criteria yielded 13 genes. The candidate genes were further screened by shRNA in the SW480 cell line which showed KRAS G12V mutation, followed by examination of migration ability in vitro. We then examined whether there was a targeted drug available for the candidate genes which led to RyR2 as the final target (Fig. 1A). RyR2 showed upregulation in KRAS mutant mCRC patients compared with KRAS WT patients, but its expression was not changed in non-metastasis patients (Fig. 1B). To gain insights into the clinical significance of RyR2 upregulation, we analyzed TCGA CRC data and demonstrated that high expression of RyR2 in CRC patients resulted in a shorter survival time (hazard ratio [HR] = 1.8, P = 0.011; Fig. 1C) as well as shorter disease-free survival (HR = 1.6, P = 0.042; Fig. 1D). Consistently, in another cohort comprising 575 patients, we also observed a shorter overall survival time associated with high RYR2 expression (Fig. S1A). Kamal and colleagues reported the largest transcriptome data of mCRC and primary CRC tissues [12]. By analyzing this dataset, we found that RyR2 expression was significantly higher in mCRC tissues than in primary CRC tissues (Fig. 1E).
To further characterize the correlation between RyR2 and CRC metastasis behavior, we stained a TMA consisting of 195 primary CRC tissues. Scoring for each tissue core was accomplished by Vectra 2 and was further manually curated (Fig. 1F). Statistical analysis of IHC score with clinicopathologic parameters showed that RyR2 expression was not correlated with tumor size (P = 0.114), P53 status (P = 0.279) or Ki67 status (P = 0.731), indicating that RyR2 expression did not affect tumor growth (Table S1). Four metastasis-associated parameters were included in our cohort: serosa invasion, vascular tumor thrombus, perineural invasion and lymph node invasion. Perineural invasion and lymph node invasion were not correlated with RyR2 expression. Importantly, RyR2 expression was significantly upregulated in CRC patients with tumor serosa invasion (P = 0.0049) and intratumoral vascular tumor thrombus (P = 0.0155) (Fig. 1G). Furthermore, in the two patients enrolled in our cohort who manifested liver metastases, high RyR2 expression could be observed in the primary tissues from both patients (Fig. 1H). Taken together, high RyR2 expression in primary CRC tissue was correlated with serosa invasion and vascular tumor thrombus formation, and predicted a poor prognosis. The mCRC tissues showed elevated RyR2 expression compared with the primary CRC tissues. For mCRC patients, RyR2 was overexpressed in KRAS mutant subtype compared with KRAS WT patients.

RyR2 knockdown or inhibition decreased CRC cell metastasis
To demonstrate the cellular function of RyR2, we first knocked down RyR2 expression using two independent shRNA sequences in KRAS mutant CRC cell line DLD-1 (KRASG13V), SW480 (KRASG12V) and HCT116 (KRASG13D) ( Fig. 2A; Fig. S1B). RyR2 is a critical molecule in the calcium-induced calcium release (CICR) process, and ATP has been shown to increase cellular calcium. We tested whether ATP-induced intracellular calcium increase was inhibited by RyR2 knockdown (KD). ATP 5 lM substantially increased intracellular free calcium [Ca 2+ ] i in DLD-1 control cells, while this [Ca 2+ ] i increase was hindered after RyR2 KD, although RyR2 KD did not affect the responsive time point and maintenance time window (Fig. 2B). Transwell assay demonstrated significant inhibition of cellular migration ability after RyR2 KD (Fig. 2C,D). While we found that RyR2 KD still inhibited cellular migration in a KRAS wildtype cell line RKO, the inhibitory extent was limited to 10%, compared with more 50% in SW480 and DLD-1 (Fig. S1C). S107 is a small molecule that enhances calstabin2 binding to RyR2 (R2474S) and is orally available to prevent cardiac arrhythmias and raise the seizure threshold [13]. We found that S107 treatment in SW480 cells dampened [Ca 2+ ] i enhancement triggered by 5 lM ATP (Fig. 2E). Consistently, S107 treatment inhibited SW480 and DLD-1 cell migration (Fig. 2F,  G). Further examination of cellular growth revealed that RyR2 KD or its inhibition by S107 did not affect CRC cell growth (Fig. S1D-G).
We next asked whether RyR2 KD or inhibition could decrease CRC cell metastasis in vivo. To this end, we used the highly metastatic CRC cell line CT26, which carried KRASG12D mutation. We first knocked down the expression of RyR2 in CT26-luciferase cells (Fig. 3A). Delivery of cancer cells into mice vasculature via tail vein injection developed a strong signal in lungs at 14 days post injection. However, RyR2 KD resulted in a significant decrease in signal magnitude (Fig. 3B). Consistently, whereas hundreds of metastatic foci were observed on the surfaces of lungs from control mice, lungs from the RyR2 KD groups presented with much fewer metastatic foci (Fig. 3C); statistical analysis demonstrated an average of 126 AE 8 metastatic foci in  the control group, and 27 AE 10 and 21 AE 8 metastatic foci for the two shRyR2 groups, respectively (Fig. 3D). Whole-mount HE staining of lungs revealed a higher tumor burden in the control group compared with the shRyR2 groups (Fig. 3E,F). We next sought to examine the therapeutic effect of S107 in impeding CRC metastasis. CT26-luci cells were pretreated with S107 for 2 days before being injected into mice tail veins, followed by S107 treatment at 30 mgÁkg À1 via intraperitoneal injection every other day (Fig. 4A). Bioluminescent images showed significantly decreased signals in the S107 treatment group compared with the control group (Fig. 4B). Furthermore, S107 treatment reduced the number of metastasis foci when compared with the control group (Fig. 4C,D), and also attenuated tumor burden (Fig. 4E,F). Moreover, using splenic injection of CT26 cells to model CRC liver metastasis, we observed decreased metastasis formation following S107 treatment (Fig. S2). Taken together, RyR2 inhibition decreased CRC cell metastasis both in vitro and in vivo.

RyR2 regulates a set of metastasisassociated genes
To gain insights into the molecular changes after RyR2 KD or S107 treatment, we performed RNA-seq analysis.
The result demonstrated 112 codownregulated genes and 174 co-upregulated genes following RyR2 KD or S107 treatment (Fig. 5A). A manual check of these 286 deregulated genes (RyR2 associated DEG) revealed enrichment in cancer metastasis (68 genes), brain-or heart-biased expression (35 genes), transcription regulation (23 genes), ROS related function (21 genes), cytoskeleton or extracellular matrix (ECM) (18 genes), ion channel or transporters (13 genes) (Fig. 5B). To further identify functional genes during colorectal cancer metastasis, we analyzed gene expression of these 286 DEG in primary and metastatic CRC tissues by in silico analysis of previously reported data (GSE131418). To this end, we identified 37 genes downregulated by RyR2 KD and S107 treatment that showed a higher expression in metastatic CRC tissues, and 71 genes upregulated by RyR2 KD or S107 treatment that showed a lower expression in metastatic CRC tissues (Fig. 5C).
We then validated the RNA-seq results by RT-PCR; the results showed that expressions of the representative molecules such as ROS-related CKMT1A, CKMT2 and MAOA as well as metastasis-associated KISS1, KISS1R, MLF2 and EGLN3 were consistent with our RNA-seq results (Fig. 5D,E). Moreover, Western blot (WB) result also confirmed the RNA-seq  results (Fig. 5F,G). These data revealed a set of metastasis-associated genes downstream of RyR2.

Transcription factor BACH1 was regulated by RyR2
As RyR2 regulated a large number of metastasisrelated genes, we proposed that RyR2 might reshape cell fate. TF controlled cell fate by regulating the expression of multiple genes. We thus asked whether there existed any TF whose activities were modulated by RyR2. To this end, we took advantage of three lines of evidence: (i) TF enriched by DEG following RyR2 KD; (ii) TF enriched by DEG following S107 treatment; (iii) TF enriched by DEG derived from TCGA data, by comparing low and high RyR2 expression samples (Fig. 6A). We thus enriched 55, 45 and 34 TF for these three groups, respectively. Five TF emerged as shared by all three groups (Fig. 6B,C). To further verify this result, we first constructed luciferase reporter plasmids for the five TF according to their binding sites (Fig. 6D). Dual luciferase reporter assay showed high basal level transcriptional activity of BACH1 and low activities of the other four TF. Moreover, RyR2 KD significantly decreased BACH1 reporter activity (Fig. 6E). RyR2 KD induced a decrease in BACH1 reporter activity that could also be observed in HCT116 cells (Fig. 6F). In an attempt to explore the underlying mechanism, we found that RyR2 KD or inhibition by S107 downregulated the protein level of BACH1 (Fig. 6G) in SW480 cells. The results could also be reproduced in HCT116 cells (Fig. 6H) as well as in HT29 cells and CT26 cells (Fig. 6I,J). Thus, BACH1 was regulated by RyR2.

RyR2-induced ROS production and cellular heme level
We next asked how RyR2 KD decreased BACH1 expression. As RyR2 was an ER-resident calcium channel, we first checked whether classical pathways regulated by calcium were changed by RyR2 KD. Excitation-transcription coupling described transcriptional activity regulated by calcium-pCREB axis in cardiac myocytes. However, pCREB(Ser133) was not changed following RyR2 KD in SW480 cells (Fig. 7A). Another important cellular aspect regulated by calcium was PKC signaling pathway. In SW480 cells, RyR2 KD did not affect either pPKC(Thr497) or pPYK2 (Tyr402) (Fig. 7B). In addition, BACH1 did not interact with known molecules downstream of calcium (calmodulin, CaMKIV, CaMKII) (Fig. S3A). We identified an array of proteins interacting with BACH1  using proximity proteomics, and selected some proteins whose functions were closely related to calcium; however, none of these proteins interacted with BACH1 (Fig. S3B). Thus, we inferred that RyR2 influenced BACH1 expression in a more indirect manner. Previous studies on BACH1 revealed its regulation by antioxidants and Nrf2 [14,15] which were related to cellular ROS. This and our RNA-seq results led us to determine cellular ROS changes following RyR2 KD or inhibition. RyR2 KD did indeed significantly decrease cellular ROS level in both SW480 and HCT116 cells (Fig. 7C,D). Consistently, S107 treatment also reduced ROS levels in both cell lines (Fig. 7E,F). Cellular ROS was mainly derived from superoxide species, which were inevitably produced by oxidative phosphorylation in mitochondria. RyR2 KD led to a significant decrease in superoxide species in SW480 and HCT116 cells, as measured by mitoSOX signal (Fig. 7G,H). S107 treatment also inhibited superoxide species production in SW480 and HCT116   cells (Fig. 7I,J). As a result, RyR2 KD or inhibition by S107 downregulated the Nrf2 level (Fig. 5F,G). Consequently, antioxidant response element (ARE) reporter activity was significantly dampened by RyR2 KD or inhibition (Fig. 7K,L). As previously reported, transcriptional upregulation of heme oxygenase 1 (HMOX1) by Nrf2-mediated BACH1 degradation. We also found downregulation of HMOX1 expression after RyR2 KD or inhibition (Fig. 5F,G). Since BACH1 directly regulated the cellular heme level, we measured free heme levels in SW480 and DLD-1 cells with RyR2 KD. The results showed that RyR2 KD significantly enhanced cellular heme levels (Fig. S4). Taken together, RyR2 KD or inhibition decreased cellular ROS levels, which in turn inactivated Nrf2 activity, and thus HMOX1 downregulation and heme accumulation followed by BACH1 degradation.
3.6. Genes downstream of RyR2/BACH1 affected cellular motility BACH1 is well established as a pro-metastasis TF. However, different BACH1 downstream targets have been reported in certain kinds of cancer types. Combining RyR2 DEG and putative targets predicted by JAS-PAR, we identified a gene list of BACH1 downstream genes, which were further exemplified by its putative targets STARD8 and TIAM2. Both RyR2 KD and inhibition by S107 downregulated mRNA levels of STARD8 and TIAM2 (Fig. 8A,B). BACH1 KD in HCT116 cells also hindered mRNA expression of STARD8 and TIAM2 (Fig. 8C). Furthermore, BACH1 expression in 293T cells enhanced promoter activity of STARD8. However, this regulation was not observed for the promoter of TIAM2 (Fig. 8D). Instead, the BACH1 binding site was detected at the enhancer region of TIAM2, as indicated in GeneCards by Hi-C sequencing data. We next overexpressed TIAM2 in SW480 and HCT116 cells (Fig. 8E). The results demonstrated that overexpression of TIAM2 enhanced cellular motility (Fig. 8E). Moreover, mRNA expression of STARD8 and TIAM2 was highly correlated with that of RyR2 in TCGA-COAD dataset (Fig. 8F,G). Thus, the RyR2/BACH1 axis regulated STARD8 and TIAM2 in order to regulate CRC cell motility.

Discussion
There is still an urgent need for appropriate therapeutic modality for mCRC patients with KRAS mutation.
To identify druggable genes that can mediate metastasis of this group of patients, we combined bioinformatic analysis, molecular and cell biology and found that RyR2 and its inhibitor S107 possessed the potential to be repurposed to intervene with KRAS mutant CRC metastasis. RyR2 was upregulated in mCRC patients with KRAS mutation and its expression was associated with poor prognosis of CRC patients. Moreover, its high expression was associated with serosa invasion and tumor vascular thrombus. This is evidence that RyR2 expression in tumor cells was closely related to metastasis behavior of CRC. Previous studies on RyR2 were mainly focused on the physiological and pathological roles in heart and brain, where RyR2 was highly expressed; few studies have reported its role in cancer metastasis [16,17]. Multiple congenital RyR2 mutations cause calcium leakage from ER and lead to catecholaminergic polymorphic ventricular tachycardia (CPVT) in humans [18]. S107 was first reported to inhibit calcium leakage from mutant RyR2 through enhancement of the binding affinity between RyR2 and calstabin2 [13]. We found that S107 actually inhibited calcium leakage in SW480 and DLD-1 cells with WT RyR2. The ability of S107 to inhibit cancer cell metastasis and ROS production was consistent with the results of RyR2 KD. Furthermore, mice treated with S107 at 30 mgÁkg À1 via intraperitoneal injection for 2 weeks presented with a healthy appearance. As S107 was commercially available, we wondered whether this drug could be repurposed to intervene with CRC metastasis.
Calcium has long been linked to cancer metastasis [19]. One of the early studies shows that Orai1 and STIM1, which mediate extracellular calcium influx into ER via store-operated calcium entry (SOCE), were critical for breast cancer cell metastasis [6]. That study highlighted the essential role of ER-resident calcium in cancer metastasis. Calcium release from ER to cytosol was mainly governed by six members (RyR1-3 and IP3R1-3) [9]. Although these six channels share a similar function, it is intriguing to find that only high RyR2 expression predicts a poor prognosis in CRC (Fig. S5). High RyR2 expression in cardiomyocytes leads to spontaneous Ca 2+ leakage from ER [20], and subsequent diminished systolic Ca 2+ transients which are not necessary for cancer cells. Effectors downstream of RyR2 include NF-jB/cyclin D1 in pulmonary arterial smooth muscle cells [21], glucose oxidative phosphorylation [22] and ROS production by calcium entry into mitochondria matrix in cardiomyocytes [23], Interferon b (IFNb) signaling in 293T and monocytes among others [24]. The discovery here supports the model that RyR2 overexpression increases superoxide species production in mitochondria and subsequent ROS levels in cytosol, followed by KEAP1 inactivation, and Nrf2 nuclear translocation and BACH1 accumulation (Fig. 8H).
BACH1 is well established as a promoter of cancer metastasis, especially in non-small cell lung cancers with NFE2L2 mutation [14,15]. Evidence also exists showing that BACH1 promotes CRC cell metastasis [25][26][27]. Generally, BACH1 is a member of the Cap 'n' Collar and basic region leucine zipper family of TF and it acts as a transcription repressor of target genes [28]. We found that RyR2 KD decreases the cellular BACH1 level, and thus dysregulation of BACH1regulated metabolism and metastasis genes. BACH1 does not interact with canonical calcium-related molecules, including calmodulin, CaMKII and CaMKIV, indicating that BACH1 is not a first-line effector of cellular calcium change. The crosstalk between calcium and BACH1 sheds further light on the regulation of cancer metastasis by calcium.
Intimate relationships between cancer metastasis and mitochondrial activity, ROS and Nrf2 are frequently reported [29][30][31][32][33]. Targeting calcium is also recognized as a potential option in cancer therapy [19]. However, considering the huge amounts of molecules involved in these pathways, it is not easy to identify a specific druggable gene in a specific cancer type.

Conclusions
Through bioinformatics analysis and experimental assays, our results provide evidence that RyR2 is a potential therapeutic target for intervention of CRC metastasis.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. RyR2 inhibition did not affect cellular growth. Fig. S2. RyR2 inhibition decreased CRC liver metastasis in vivo. Fig. S3. RyR2 did not interact with calcium-related molecules. Fig. S4. RyR2 inhibition increased cellular heme level. Fig. S5. Overall survival and disease-free survival analysis of ITPR and RYR. Table S1. Relationship between expression of RYR2 in CRC and clinicopathologic features. Table S2. Materials and primers used in this study.