LncRNA CRCMSL interferes in phospholipid unsaturation to suppress colorectal cancer progression via reducing membrane fluidity.

Introduction
Colorectal cancer (CRC) is currently the third most common cancer worldwide [1], causing nearly 900,000 cancer-related deaths each year [2], and its incidence has been increasing around the world [3]. Despite several treatment options for CRC, the prognosis of patients with metastatic CRC is not promising [1], urging us to search for effective markers or intervention targets for CRC metastasis.
Long noncoding RNAs (lncRNAs) are transcripts between 200–10,000 nt that do not encode proteins and play various roles in biological processes [4]. They are thought to be involved in cancer progression by regulating proliferation, survival, migration or genomic stability [5]. In recent years, several studies have reported that lncRNAs participate in metabolism regulation by affecting enzymes [6], [7], [8] and transcription factors [9], [10] of glycolysis as well as reprogramming lipid metabolism [11], [12]. However, beyond the energy supply related ideas mentioned above, few studies have explored other phenomena caused by lncRNA-associated metabolism, such as whether lipid metabolism reprogramming affects the properties or function of phospholipid membranes, which requires further investigation.
Reprogrammed metabolism is a hallmark of CRC and considered an important basis for CRC metastasis [13]. Fatty acid metabolism contributes to tumor invasion and migration [14], and rate-limiting enzyme for de novo fatty acid synthesis is acetyl-CoA carboxylase 1 (ACC1), whose abnormal expression and activity have been reported in several human malignant tumors [15], [16]. Interestingly, silencing of ACC1 expression resulted in inhibition of proliferation and caspase-mediated apoptosis in prostate cancer cells, but presented no cytotoxic effects in nonmalignant cells with relatively inactive lipid metabolism [17]. Although ACC1 has been regarded as a promising target for inhibiting lipid metabolism in tumors [16], the clinical application of its inhibitors is still limited to the treatment of lipid metabolism diseases. As of the latest reported phase II trial, patients with NAFLD treated with the ACC inhibitor ND630 (also known as firsocostat) tolerated the treatment well, with no adverse effects requiring discontinuation, and symptoms of hyperlipidemia, obesity, and hepatocellular steatosis were effectively improved [18], [19]. Nevertheless, there are no reports on the use of ND630 in tumors, leaving a possibility and challenge for metastatic CRC treatment strategies.
Recent studies have shown that induction of ferroptosis is a promising therapeutic strategy for CRC [20], [21], [22]. Ferroptosis is a non-apoptotic form of cell death that is caused by the accumulation of reactive oxygen species (ROS) associated with lipid peroxidation, which is regulated by three primary mechanisms including the following: labile iron pool (LIP), phospholipids containing polyunsaturated fatty acid chains (PUFA-PLs) and inhibition of glutathione (GSH) antioxidant pathway. Ferritin, acyl-CoA synthetase long-chain family member 4 (ACSL4) and glutathione peroxidase 4 (GPX4) are involved in each of these three mechanisms and were therefore used as protein markers to assess ferroptosis process in this study.
Fatty acids that form the hydrophobic portion of phospholipid bilayers, can be classified into saturated fatty acids (SFAs), monounsaturated fatty acids (MUFAs) and polyunsaturated fatty acids (PUFAs) according to the degree of saturation. SFAs are straight, leading to ordered and compact phospholipid bilayers with lower membrane fluidity, whereas MUFAs and PUFAs are bent by carbon–carbon double bonds, forming disordered and loose phospholipid bilayers with higher membrane fluidity. High membrane fluidity in malignant tumors have been reported as early as the last century [23], but has been given poor attention. In recent years, several studies have reported and concerned about this phenomenon through various experiments, suggesting that high membrane fluidity contributes to the progression of CRC and other tumors [24], [25], [26], [27], [28], which remains to be further studied. In this study, membrane fluidity of live cells was assessed by Laurdan generalized polarization (GP) detection and fluorescence recovery after photobleaching (FRAP) assay. Higher fluidity of phospholipid bilayers corresponds to smaller polarity and faster fluorescence recovery.

Results

Screening, function prediction and preliminary validation of metastasis and prognosis-related genes in CRC
To explore intervention targets for CRC metastasis, sixteen moderately differentiated adenocarcinoma cases (eight with metastasis and eight without metastasis, Table S1) were collected and made into microarray for high-throughput transcriptome sequencing (GSE113296). Differentially expressed genes between metastatic and non-metastatic samples were screened (limma P-value <0.01, fold change ≥1.8 or ≤0.556) and annotated as Ensemble gene ID, becoming candidates for subsequent prognostic analysis.
Three large sample datasets were selected for prognosis-related secondary screening: two from the GSE database (GSE39582 and GSE17536) and one from the TCGA database (TCGA-COAD). The expression matrix of candidate genes and survival information of patients from the three datasets were obtained, as described in the Methods section. Applying univariate Cox regression to preliminarily assess the prognostic performance of candidate genes in the three datasets, we noticed that ENSG00000258701, ENSG00000267194, and ENSG00000168393 appeared in the log-rank test P-value top 10 of three databases (Table S2), while only ENSG00000267194 acted as a protective factor in each dataset (Table S3), which was further presented in Kaplan-Meier curves (Fig. 1B–D).
The potential functions of ENSG00000267194 in the three datasets were predicted by GSEA, and the results suggested that ENSG00000267194 may affect fatty acid and phospholipid metabolism (Fig. S1A-C), which were reported in detail by enrichment plots (Fig. 1E–J). ENSG00000267194 was previously named Colorectal Cancer Metastasis-Suppressed LncRNA (CRCMSL), and a part of its anti-metastasis mechanism has been identified [29].
To explore the relationship between CRCMSL and lipid metabolism in CRC, CRCMSL was knocked down in the CRC cell line SW620 (Fig. 1K), which naturally expresses high level of CRCMSL [29]. CRCMSL-knockdown and vector-control SW620 cells were examined for lipidomics, and differential lipid metabolites were screened by P-value <0.05, and VIP ≥ 1. The results suggested that lipid metabolites were generally elevated by CRCMSL-knockdown in SW620 cells (Fig. 1L), the major class of which was phospholipids (Fig. 1M), preliminarily confirming that CRCMSL regulates lipid metabolism in CRC.

Ferroptosis contributes to CRCMSL-related suppression of CRC metastasis
Further analysis of fatty acyl carbon chains of differential phospholipids caused by CRCMSL knockdown revealed that CRCMSL-knockdown not only upregulated the lipid synthesis but also remodeled fatty acyl chains of phospholipids (PLs) and glycerolipids (GLs) to contain a higher proportion of monounsaturated fatty acids (MUFAs) and a lower proportion of polyunsaturated fatty acids (PUFAs) (Fig. 2A-B& Fig. S2A-B). Compared to PUFAs, MUFAs with only one carbon–carbon double bond contribute to lower lipid peroxidation, which means that CRCMSL may affect lipid peroxidation-dependent ferroptosis through fatty acid metabolism. Stearoyl-coenzyme A desaturase (SCD), as a key enzyme that catalyses the synthesis of MUFAs from SFAs [30], is considered a recovery target for CRCMSL-related fatty acid metabolism.
The relationship between CRCMSL-related fatty acid metabolism and ferroptosis was explored. Classical protein markers were detected to assess ferroptosis process, which is positively correlated with ferritin and ACSL4 whereas negatively correlated with GPX4. Lentiviruses were used to overexpress CRCMSL in the CRC cell line HCT116 (Fig. 2C), which naturally expresses low level of CRCMSL [29]. Increased ACSL4 and ferritin and decreased GPX4 were detected in CRCMSL-overexpressed cells, and these trends were restored by SCD (Fig. 2D&S2C-D). In contrast, decreased ACSL4 and ferritin and increased GPX4 were detected in CRCMSL-knockdown cells and were returned by SCD inhibitor PluriSln1 (Fig. 2E & Fig. S2E). The results suggested that CRCMSL may promote ferroptosis by restricting fatty acid metabolism. Lipid peroxidation signals (Fig. 2F–G), cellular (Fig. 2H–I) and mitochondrial ROS (Fig. 2J–K) signals enhanced by CRCMSL overexpression and reduced by CRCMSL-knockdown could also be restored by SCD and PluriSln1, respectively, suggesting that CRCMSL causes lipid-related oxidative stress on CRC cells. Alteration of GSH levels indicated that CRCMSL could be resistant to the GPX4-associated antioxidant effects (Fig. 2L–M). In addition, CRCMSL overexpression-induced mitochondrial morphology changes in HCT116 cells (becoming smaller and rounder, with deeper electron density) were restored to some extent by SCD (Fig. 2N), and mitochondrial aberrations prevented by CRCMSL-knockdown were returned through PluriSln1 (Fig. 2O). These results demonstrated that CRCMSL promotes ferroptosis by restricting fatty acid metabolism.
Ferroptosis is commonly thought to be associated with the proliferative activity of tumors, whereas CRCMSL was screened as a metastasis suppressor; thus, Transwell assays were required to detect the ferroptosis-associated invasion and migration capacity of CRC cells, corresponding to the number of cells crossing the Matrigel and migrating to the high serum concentration side, respectively. The classic ferroptosis inhibitor Fer1 rescued invaded cells reduced by CRCMSL overexpression (Fig. 2P&Fig. S3), whereas the ferroptosis inducer Erastin partly reversed invaded cells increased by CRCMSL-knockdown (Fig. 2Q & Fig. S3). Changes in the number of migrated cells showed a pattern similar to that described above (Fig. 2R–S & Fig. S3). These suggested that CRCMSL-induced ferroptosis contributes to the suppression of CRC invasion and migration.

CRCMSL reduces membrane fluidity to suppress CRC metastasis
Compared to SFAs, MUFAs with one carbon–carbon double bond lead to morphological heterogeneity of phospholipid molecules and disordered arrangement of phospholipid bilayers, which should result in higher membrane fluidity. As a key enzyme that catalyzes the synthesis of MUFAs from SFAs, SCD is considered a possible recovery target for CRCMSL-related membrane fluidity.
Laurdan GP detection was performed to assess the order of phospholipid bilayers. The GP value was positively correlated with the order degree of phospholipid bilayers and negatively correlated with the membrane fluidity (Fig. 3A). We detected elevated GP values in CRCMSL-overexpressed cells and diminished GP values in CRCMSL-knockdown cells, which could be restored by SCD and PluriSln1, respectively (Fig. 3B–C), suggesting that CRCMSL leads to more ordered phospholipid bilayers in CRC cells by interfering in phospholipid unsaturation. FRAP assays were performed to measure the rate of membrane flow by fluorescence half recovery time (Thalf) in the bleached region, which is also negatively correlated with membrane fluidity (Fig. 3D). Fluorescence recovery of the membrane was slowed down in CRCMSL-overexpressed cells and could be restored by SCD (Fig. 3E). CRCMSL knockdown accelerated fluorescence recovery in SW620 cells, which was retarded by PluriSln1 (Fig. 3F). These results further verified that CRCMSL reduces membrane fluidity in CRC cells by interfering in phospholipid unsaturation and that SCD is also capable as a recovery target for CRCMSL-related membrane fluidity.
Transwell assays were performed to assess whether membrane fluidity contributes to invasion and migration capacity of CRC cells. The number of invaded and migrated cells decreased by CRCMSL was rescued by SCD (Fig. 3G, I & Fig. S4), and its inhibitor PluriSln1 could normalized the prosperous invaded and migrated cells caused by CRCMSL-knockdown (Fig. 3H, J & Fig. S4). These data suggested that CRCMSL suppresses invasion and migration capability of CRC cells by reducing their membrane fluidity.

CRCMSL restricts ACC activity to promote ferroptosis and reduce membrane fluidity
To investigate the molecular mechanisms by which CRCMSL affects lipid metabolism, the interaction of CRCMSL with key enzymes involved in fatty acid metabolism was performed utilizing RNA pull-down assay. The target proteins of CRCMSL were enriched in the pull-down products, and the latter were identified by western blot assays. The results showed that CRCMSL interacts with ACC1 instead of with SCD (Fig. 4A). ACC1 is the key enzyme for de novo synthesis of SFAs, and the latter is catalyzed into MUFAs by SCD. Given that ACC1 is a carboxylase capable of binding biotin, which may contribute to non-specific binding to streptavidin magnetic beads in RNA pull-down assays, RNA immunoprecipitation (RIP) assays were performed to validate the interaction between ACC1 and CRCMSL. The results showed that ACC1 enriches CRCMSL compared to the IgG control (Fig. 4B–C).
The effect of CRCMSL on the expression of ACC1 was detected using qPCR and western blot assays. In HCT116 and SW480 cells, overexpression of CRCMSL resulted in about 15–20 % decrease in the transcriptional level of ACC1 (Fig. 4D–E), whereas in SW620, HT29 and LS174T cells, knockdown of CRCMSL resulted in about 8–30 % increase in the transcriptional level of ACC1 (Fig. 4F & Fig. S5A). Meanwhile, inactivation of ACC1 through phosphorylation at Ser79 was promoted by CRCMSL overexpression in HCT116 and SW480 cells (Fig. 4G–H) and suppressed by CRCMSL knockdown in SW620 and HT29 cells (Fig. 4I). The result of transcriptome was consistent with the experimental data of western blot (Fig. S5B). These results suggested that CRCMSL promotes phosphorylation-related inactivation of ACC1, which may trigger transcriptional compensation of ACC1. ACC activity detection showed that CRCMSL is a negative regulator of ACC activity in CRC cells (Fig. 4J-K), and the ACC oral inhibitor ND630, which has passed Phase II clinical trials, inhibited ACC activity combined with CRCMSL in HCT116 and SW480 cells (Fig. 4J) and restored the increased ACC activity caused by CRCMSL knockdown in SW620 and HT29 cells (Fig. 4K).
To confirm whether ACC1 inhibition contributes to ferroptosis, we attempted to assess ND630 in ferroptosis related assays. Western blot showed that ND630 differentially rescued CRCMSL-induced alterations in ACSL4, ferritin and GPX4 (Fig. 5A). Cellular ROS detection resulted that ND630 could partly restored oxidative stress relieved by CRCMSL-knockdown (Fig. 5B). GSH levels elevated in CRCMSL-knockdown cells were normalized by ND630 (Fig. 5C). Flow cytometry further verified that CRCMSL-knockdown related reduction of lipid peroxidation can be recovered by ND630 (Fig. 5D). These results demonstrated that CRCMSL promotes ferroptosis by restricting ACC activity.
Laurdan and FRAP assays were conducted to analyze whether ACC1 inhibition leads to membrane fluidity alternation associated with CRCMSL. We found that ND630 could further promoted CRCMSL-induced GP values increase (Fig. 5E) and rescued GP values diminished by CRCMSL-knockdown (Fig. 5F), which suggested that CRCMSL leads to more ordered phospholipid bilayers in CRC cells by restricting ACC activity. Besides, fluorescence recovery of the membrane slowed down in CRCMSL-overexpressed cells could be further retarded by ND630 (Fig. 5G), and fluorescence recovery acceleration dependent on CRCMSL-knockdown was also decelerated by ND630 (Fig. 5H). These results indicated that CRCMSL reduces membrane fluidity in CRC cells by restricting ACC activity, which could better effect in combination with ND630.
Before in vivo experiments, the role of ND630 in CRC metastasis needed to be determined by Transwell assays. When combination with ND630, CRCMSL-related decrease of invaded and migrated cells were further declined (Fig. 5I-K&S6), and increase of invaded and migrated cells caused by CRCMSL-knockdown were restored (Fig. 5J–L & Fig. S6). These results suggested that CRCMSL suppresses the invasion and migration capability of CRC cells by restricting ACC activity, and that the combination with ND630 may better suppress CRC progression.

CRCMSL suppresses CRC progression in vivo by restricting fatty acid metabolism
Orthotopic and subcutaneous xenografts in nude mice were used to assess CRC progression in vivo (Fig. 6A). For orthotopic xenografts, CRCMSL-overexpressed and vector-control HCT116 cells were injected into the cecum mesentery of 5-week-old nude mice after transfection with SCD or the corresponding vector control. Four weeks later, tumor-bearing mice were sacrificed by cervical dislocation under deep anesthesia. The number of metastatic foci in the liver demonstrated that CRCMSL suppresses CRC metastasis in vivo by interfering in phospholipid unsaturation (Fig. 6B). For subcutaneous xenografts, HCT116 cells, treated as described above, were injected into the axilla of 5-week-old nude mice. Nine days later, tumor-bearing mice were measured every three days for five measurements. The volume and weight of the xenografts suggested that CRCMSL suppresses CRC proliferation in vivo by interfering in phospholipid unsaturation (Fig. 6C). Consistent with the western blot results, IHC staining of subcutaneous xenografts also demonstrated ferroptosis contributes to CRCMSL-related suppression of CRC (Fig. 6D).
To observe the efficacy of ND630 on CRC in vivo, CRCMSL-overexpression and vector-control HCT116 cells were injected into the cecum mesentery or axilla of 5-week-old nude mice to construct orthotopic or subcutaneous xenografts, respectively (Fig. 7A). Four weeks after orthotopic xenograft injection, or nine days after subcutaneous xenograft injection, tumor-bearing mice were administered 5 mg/kg ND630 or an equivalent volume of solvent daily for two weeks. The results demonstrated that CRCMSL suppresses CRC progression in vivo by restricting ACC activity, and its combination with ND630 resulted in better suppression of CRC progression (Fig. 7B–C). IHC staining results demonstrated ND630 contributes to CRCMSL-related lipid peroxidation of in vivo (Fig. 7D).

Discussion
Membrane fluidity of cells mainly depends on the fatty acyl chains of phospholipid bilayers, the raw material of which is fatty acids. ACC1 initiates and rate-limits the de novo synthesis of SFAs, and the latter is catalyzed to MUFAs by SCD [31]. As a key enzyme in this step, SCD is considered a regulatory target for membrane fluidity [25], [32], [33]. There are two isoforms of SCD (also known as hSCD1) and SCD5 found in human tissues, while the former is the predominant isoform in the colon [34], [35], making SCD the only regulation target of membrane fluidity in this study. Similar to ACC1 inhibitors, the clinical use of SCD inhibitors is currently limited to lipid metabolism disorders, and no clinical trials have reported a SCD inhibitor that is not only safe but also effective [36]. In this study, we found that CRCMSL suppresses CRC metastasis by reducing membrane fluidity in CRC cells, and the SCD inhibitor PluriSln1 was demonstrated to be a potential drug that could restore the enhanced metastatic capacity and membrane fluidity caused by CRCMSL knockdown.
Ferroptosis is a kind of cell death associated with phospholipid peroxidation. Compared to MUFAs, PUFAs also contribute to the high membrane fluidity of phospholipid bilayers but are more likely to trigger lipid peroxidation-dependent ferroptosis due to their excessive carbon–carbon double bonds. This study revealed how CRC cells remodel lipid metabolism through the de novo synthesis pathway, striking a favorable balance between membrane fluidity requirement and the risk of lipid peroxidation.
Human CRC cell lines with high expression of CRCMSL, such as SW620, HT29, LS174T, and LOVO [29], were unable to stably construct orthotopic xenografts sufficient for analysis, and the expression levels of CRCMSL in other CRC cell lines were too low to be knocked down. In addition, CRCMSL knockdown on the basis of CRCMSL overexpression in HCT116 cells was also attempted, but the knockdown effect was poor, which may be due to the excessive initiation ability of the CMV promoter used for exogenous overexpression. Therefore, experiments for assessing CRC progression related to CRCMSL knockdown were finally aborted, while orthotopic xenografts using HCT116 cells still presented convincing results, demonstrating the effect of CRCMSL and ND630 on CRC progression in vivo.
In this study, we discovered a novel mechanism for metastasis-suppressing potency of CRCMSL that promotes ferroptosis and reduces membrane fluidity by restricting ACC activity, and demonstrated the effect of CRCMSL and ND630 on CRC progression in vivo, providing laboratory evidence for the development of CRC therapeutic strategies (Fig. 8).

Materials and methods

Cell culture, gene knockdown and overexpression
Human CRC cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences and cultured in RPMI-1640 medium (KeyGEN, Cat# KGM31800N) supplemented with 10 % fetal bovine serum (Gibco, Cat# A3160802) at 37 °C and 5 % CO2. All cell lines used in this study were authenticated by performing short tandem repeat (STR) profiling within two years, and experiments were performed in cells less than six months after resuscitation, during which cells were tested every two months to ensure that they were mycoplasma-free.
Molecular drugs were stored in DMSO at −80 °C and added to the culture medium when needed in vitro. 10 mM Fer1 (Selleck, Cat# S7243) and 10 mM Erastin (Selleck, Cat# S7242) were used at 2.5 μM and 10 μM, respectively. 2 mM ND630 (DC Chemicals, Cat# DC10173) was diluted to 0.2 μM and 40 nM for HCT116 and SW620 cells, respectively. 100 mM PluriSln1 (MedChemExpress, Cat# HY-15700) was diluted to 50 μM in both HCT116 and SW620 cells.
Two pSuper-puro shRNA vectors for CRCMSL were constructed in our laboratory based on the following sequences: shCRCMSL#1 5′-GCTAGTTAGGAGCTTGCTA-3′ and shCRCMSL#2 5′-GGTCTCCAGATGTGTGTAT-3′. The CRCMSL-overexpression lentivirus was constructed and synthesized by GenePharma (Suzhou, China). Plasmid for SCD-overexpressing pEnCMV-SCD (human)-3 × FLAG (Cat# P20373) was purchased from Miaolingbio (Wuhan, China).
CRC cells were transfected with plasmids using Lipofectamine 3000 Reagents (Invitrogen, Cat# L3000075) or INVI DNA RNA Transfection Reagents (Invigentech, Cat# IV1216100) according to the manufacturer’s protocol. Stable cell lines were screened using eukaryote-specific antibiotics after transfection.

Lipidomics
Lipidomics of CRC cells was performed using LC-MS according to previous studies [37], [38], [39]. After washing three times with cold PBS, CRC cells were resuspended in chloroform–methanol solution (2:1, −20 °C) and lysed twice with liquid nitrogen and vortex mixer. Lipid metabolites in the lysate were extracted twice with chloroform–methanol solution (2:1, −20 °C), concentrated by vacuum centrifugation, dissolved in isopropanol, and filtered through a 0.22 μm membrane to obtain the samples to be tested [38] of each sample to be tested were used as quality control samples, and the rest were detected by LC-MS (Thermo Fisher Scientific, UltiMate 3000 and Q Exactive Focus) as previously described [39]. Raw data were annotated and aligned (R. T. Tolerance = 0.25, m-Score Threshold = 5) using LipidSearch software (v4), and then internally normalized by the peak response values. Unit-variance scaling (UV) conversion was performed prior to multivariate statistical analysis, and the latter was performed as previous publication [40]. Differential lipid metabolites were screened by P-value < 0.05 and VIP ≥ 1 (OPLS-DA) [41].

Lipid peroxidation detection
After washing twice with PBS, 5 × 106 CRC cells were resuspended in 1 mL of RPMI-1640 containing 5 μM lipid peroxidation sensor BODIPY@ 581/591C-11 (Thermo Fisher, Cat# D3861) for 30 min at 37 °C with 5 % CO2, and the sample tubes were flicked gently every 5 min. The fluorescence intensity of the oxidation phase (excited by a 488 nm laser) was recorded by BD LSRFortessa X-20 multidimensional high-definition flow cytometry analyzer, and the results were analyzed using FlowJo software (v10).

Cellular and mitochondrial ROS assays
Cellular ROS was detected using Reactive Oxygen Species Assay Kit (Beyotime, Cat# S0033S) according to manufacturer's instructions. Mitochondrial ROS was detected using Mitochondrial Superoxide Assay Kit with MitoSOX Red (Beyotime, Cat# S0061S) according to manufacturer's instructions.

GSH assays
GSH levels were detected using ROS Human Reduced Glutathione (GSH) Elisa Kit (JSBOSSEN, Cat# BS-E4274H2) according to manufacturer's instructions.

Measurement of membrane fluidity
The membrane fluidity of CRC cells was measured using Laurdan generalized polarization (GP) detection and fluorescence recovery after photobleaching (FRAP) assays. CRC cells were attached to glass-bottom culture dishes (Nest, Cat# 801002) and observed using an LSM880 confocal microscope (live cell workstation set to 37 °C and 5 % CO2) and the ZEN software (Zeiss).
For Laurdan GP detection, cells were washed twice with PBS and stained with Laurdan (Glpbio, Cat# GC18338) in PBS at 5 μg/mL for 30 min at 37 °C. With 16 bits depth, 1024 × 1024 resolution, 6 scan speed and 1 airy unit, images of ordered phase (excited by 405 nm laser and recorded between 400–460 nm), disordered phase (excited by 405 nm laser and recorded between 470–530 nm) and live cells (recorded in transmitted light) were simultaneously acquired at 100 × objective and zoom 1.5. From each sample, ten fields of view were randomly obtained, and the images were analyzed using ImageJ software, as previously described [42].
For FRAP assays, cells were washed twice with PBS and stained with BODIPY@ 500/510 C1, C12 (Thermo Fisher Scientific, Cat# D3823) at 2 μg/mL in PBS for 15 min at 37 °C. With 8 bits depth, 1024 × 1024 resolution, 1 airy unit and definite focus on, three 5 µm × 5 µm frames were drawn to track the fluorescence intensity (excited by 488 nm laser and recorded between 475–650 nm) of target cell, neighbor control cell and background at 63 × objective and zoom 2. After three scans, the frame of the target cell was bleached with 50 % laser power, whereas the other two were not; all of them were tracked 25 times after bleaching. From each sample, ten target cells were randomly selected, and their fluorescence recovery curve and half recovery time (Thalf) were analyzed using the FRAP function of ZEN software.

Acetyl-CoA carboxylase (ACC) activity detection
Total proteins containing ACC (crude enzyme solution) were extracted using RIPA lysis buffer with protease inhibitor cocktail and quantified using BCA Protein Assay Kit (KeyGEN BioTECH, Cat# KGP902). ACC activity of the crude enzyme solution was detected using ACC Activity Detection Kit (Solarbio, Cat# BC0410) according to the manufacturer's instructions, with the following formula: ACC activity (U/mg protein) = phosphorus yield × 20/protein concentration. Reactions for phosphorus measurement were performed in dark 96-well culture plates with OD660nm measured by microplate spectrophotometer (Epoch). Phosphorus yield (μmol/mL) was calculated from ΔOD660 according to the manufacturer's instructions. The unit of protein concentration is μg/mL.

Orthotopic and subcutaneous xenografts in nude mice
Four-week-old female BALB/c athymic nude mice were obtained from Guangdong Provincial Animal Center (Guangzhou, China) before the CRC cell lines were ready and raised under specific pathogen-free conditions for a week. The experiments were performed according to the ethical guidelines approved by Southern Medical University Animal Care and Use Committee. For orthotopic xenografts, 1 × 106 HCT116 cells suspended in 100 μL PBS were injected into the cecum mesentery of 5-week-old nude mice (eight mice per group). Four weeks later, tumor-bearing mice were sacrificed by cervical dislocation under deep anesthesia, and the livers were quickly harvested into ice-cold saline and then fixed for H&E staining.
For subcutaneous xenografts, 2 × 106 HCT116 cells suspended in 100 μL PBS were injected into the axilla of 5-week-old nude mice (six mice per group). Nine days later, the tumor volume was measured every 3 days according to the following formula: Volume (mm3) = width2 (mm2) × length (mm) × 0.5. After five measurements, tumor-bearing mice were sacrificed by cervical dislocation under deep anesthesia, and xenografts were quickly harvested and weighed, and then fixed for H&E and IHC staining.
For ND630 treatment assays, ND630 was dissolved in 1 % DMSO and 99 % saline. Four weeks after orthotopic xenograft injection or nine days after subcutaneous xenograft injection, tumor-bearing mice were administered 5 mg/kg ND630 or an equivalent volume of solvent daily for two weeks.

Ethical Approval
All experiments involving patients were approved by the Ethics Committee of Shunde Hospital, Southern Medical University (No. 20210813) and complied with the Declaration of Helsinki. Informed consent was not required because the data were analyzed anonymously. All animal experiments were performed according to the ethical guidelines approved by Southern Medical University Animal Care and Use Committee (No. SMUL202310033).

Bioinformatics and statistical analysis
Differentially expressed genes between metastatic and non-metastatic CRC tissues were screened using the R package limma (P-value <0.01, fold change ≥1.8 or ≤0.556) and annotated as Ensemble gene ID according to probe information (GPL18180). Two large sample CRC datasets, GSE39582 (585 cases) and GSE17536 (177 cases), were obtained from the GEO website (https://www.ncbi.nlm.nih.gov/geo) and the COAD dataset (539 cases) was obtained from the GDC Data Portal (https://portal.gdc.cancer.gov). Cases with complete gene expression matrix and prognosis information (579 cases from GSE39582, 177 cases from GSE17536, and 452 cases from TCGA-COAD) were selected to test differentially expressed genes screened from our clinical samples (GSE113296). Based on the optimal cutoff value calculated using the R package maxstat (maximally selected rank statistics with several P-value approximations), each gene in the cases was classified as high or low expression, and the difference in overall survival (OS) between the two groups was evaluated using R package survival. Genes with a Hazard Ratio (HR) of 95 % CI excluding 1 and log-rank test P-value <0.05 were considered statistically significant. The function of candidate genes was predicted using GSEA software (v3) combined with the subset c5.go.bp.v7.4. symbols.gmt downloaded from the Molecular Signatures Database. With five minimum gene sets, 5000 maximum genes, and 1000 resamplings, those satisfying P-value <0.05 and FDR <0.3 were considered to have statistically significant differences. Candidate functions with normalized P-value <0.05 and |NES| >1 were selected for further validation.
Variables in the statistical graphs are shown as mean ± SD. The significance of differences in qPCR, flow cytometry, Laurdan GP value, Thalf (FRAP), cell count (Transwell), enrichment fold (RIP), ACC activity, and xenografts-related measurements were assessed by two-tailed Student’s t-test. Paired t-test was used to analyze the relative protein levels (western blot). The significance of differences in curves (FRAP) was assessed using two-way ANOVA. *, ** and *** represented P-value less than 0.05, 0.01, and 0.001, respectively.

Ethical approval
All experiments involving patients were approved by the Ethics Committee of Shunde Hospital, Southern Medical University (No. 20210813) and complied with the Declaration of Helsinki. Informed consent was not required because the data were analyzed anonymously. All animal experiments were performed according to the ethical guidelines approved by Southern Medical University Animal Care and Use Committee (No. SMUL202310033).

Credit author statement
LZ designed the study and prepared the manuscript. M-HJ, L-JX and W-DL performed the main experiments and analyzed the data. W-WL performed paraffin sections and H&E staining. W-DL and HW assisted with tissue sample collection. Y-JZ performed the statistical analyses. All authors reviewed the manuscript

Funding
National Natural Science Foundation of China (82173172 and 81902946), Guangdong Basic and Applied Basic Research Foundation (2021B1515120001 and 2020A1515011389) and Beijing Xisike Clinical Oncology Research Foundation (Y-Roche2019/2-0025).

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.