In vivo CRISPR activation screen identifies acyl-CoA binding protein as a driver of bone metastasis

Abstract

One of the most common sites of cancer metastasis is to the bone. Bone metastasis is associated with substantial morbidity and mortality, and current therapeutic interventions remain largely palliative. Metastasizing tumor cells need to reprogram their metabolic states to adapt to the nutrient environment of distant organs; however, the role and translational relevance of lipid metabolism in bone metastasis remain unclear. Here we used an in vivo CRISPR activation screening system coupled with positive selection to identify acyl-CoA binding protein (ACBP) as a bone metastasis driver. In non-metastatic and weakly metastatic cancer cells, overexpression of wild-type ACBP, but not the acyl-CoA binding deficient mutant, stimulated fatty acid oxidation (FAO) and bone metastasis. Conversely, knockout of ACBP in highly bone-metastatic cancer cells abrogated metastatic bone colonization. Mechanistically, ACBP-mediated FAO increased ATP and NADPH production, reduced reactive oxygen species, and inhibited lipid peroxidation and ferroptosis. We found that ACBP expression correlated with metabolic signaling, bone-metastatic ability, and poor clinical outcomes. In mouse models, pharmacological blockade of FAO or treatment with a ferroptosis inducer inhibited bone metastasis. Altogether, our findings reveal the role of lipid metabolism in tumor cell adaptation and thriving in the bone and identify ACBP as a key regulator of this process. Agents that target FAO or induce ferroptosis represent a promising therapeutic approach for treating bone metastases.

One Sentence Summary:

ACBP and lipid metabolism invigorate tumor cells to adapt and thrive in the bone, which can be targeted by an FAO blocker or a ferroptosis inducer.

INTRODUCTION

Distant metastases are the major cause of mortality associated with solid cancers (1, 2). The metastatic cascade involves multiple steps, including local invasion, intravasation, survival in circulation, extravasation, and colonization (3). Metastasis is regulated by genetic and epigenetic alterations of cancer cells as well as microenvironmental cues (4). To survive and thrive, metastasizing tumor cells need to acquire nutrients and energy, reduce reactive oxygen species (ROS), and alleviate cellular stress resulting from detachment from the extracellular matrix (5). Moreover, these cells need to adapt their metabolic state to the nutrient environment of the circulatory system and distant organs (6).

A growing body of evidence has linked lipid metabolism to metastasis (7, 8). Whereas elevated uptake of long-chain fatty acids through the fatty acid transporter CD36 promotes lymph node metastasis of oral squamous cell carcinoma (9), de novo fatty acid synthesis through fatty acid synthase enables breast cancer brain metastasis, reflecting an adaptation to reduced lipid availability in the brain relative to other organs (10). Moreover, increased fatty acid oxidation (FAO) has been shown to promote omental metastasis in ovarian cancer (11) and lung metastasis in several cancers (12–14). However, the role of lipid metabolism in bone metastasis remains poorly understood.

Over the past two decades, technological advancements have led to the development of new methods for discovering metastasis genes. Forward genetics screens have the advantage of identifying functional mediators of a phenotype in an unbiased manner. Two types of forward genetics screens have been used in metastasis research, including loss-of-function screens, such as RNAi screens and CRISPR knockout screens (15–18), and gain-of-function screens, such as microRNA library screens and cDNA library screens (19–21). These screens have identified metastasis suppressor and metastasis promoter genes. To date, the library used in reported cDNA library screens for metastasis genes is the cDNA isolated from a highly metastatic cancer cell line, such as the 4T1 mouse mammary tumor cell line (19, 20), and thus is limited to genes that are abundantly expressed in a selected cell line. Another limitation is that previous forward genetics screens have identified lung or liver metastasis genes (17–21), but not bone metastasis genes.

In this study, we employed a CRISPR activation (CRISPRa) system that activates endogenous gene expression (22), coupled with positive selection, to identify drivers of metastasis. Our in vivo screen using a CRISPRa guide RNA (gRNA) sub-library identified DBI (encoding the extracellular protein diazepam binding inhibitor and the intracellular protein acyl-CoA binding protein, ACBP, which have the identical amino acid sequence) as a previously undescribed bone metastasis promoter gene. Our studies revealed that ACBP promotes FAO and metastatic bone colonization through its acyl-CoA-binding ability, and that in the presence of long-chain fatty acids, ACBP-mediated FAO supplies ATP and NADPH, which in turn protects against lipid peroxidation and ferroptosis. We found that FAO blockade or treatment with a ferroptosis inducer inhibited bone metastasis in mouse models, suggesting translational potential.

RESULTS

An in vivo CRISPRa screen identifies ACBP as a driver of bone metastasis

Highly multiplexed pooled library screening platforms have emerged as powerful tools for forward genetics screens (23). The fusion of nuclease-dead Cas9 (dCas9) to domains that activate transcription (CRISPRa) or to domains that inhibit transcription (CRISPRi) has enabled large-scale screens (22, 23). To generate a cell line suitable for positive selection and in vivo screening, we transduced the non-metastatic human non-small cell lung cancer cell line H1299 with lentiviral FU-luciferase-CRW/RFP and dCas9-VPR (22) (fig. S1, A to C). The resulting H1299-Luc-RFP-dCas9-VPR cell line was tested with lentivirus expressing CRISPRa single guide RNA (sgRNAs) for solute carrier family 7 member 11 (SLC7A11, encoding a cystine transporter), which validated the CRISPRa system (fig. S1, D and E). We then performed intracardiac injections with 1 × 10⁶ cells per mouse and confirmed that H1299-Luc-RFP-dCas9-VPR cells formed no detectable metastases after extended periods of time (270 days; fig. S1F).

Next, we amplified a pooled CRISPRa gRNA sub-library of kinases, phosphatases, and drug targets (genes targeted: 2,320; targeting gRNAs: 12,780; control gRNAs: 250) (22). After a 1:20,000 dilution, approximately 1,500 colonies were obtained. Thus, the library coverage = 1500 × 20000 ÷ 13030 (total number of gRNAs) = 2302. We then transduced H1299-Luc-RFP-dCas9-VPR cells with the sub-library at a multiplicity of infection of approximately 0.3 (Fig. 1A and fig. S1G) and selected the cells with puromycin for 7 days to ensure that >90% of surviving cells incorporated single lentivirus particles. After expansion and intracardiac injection of these cells into nude mice (1 × 10⁶ cells per mouse, 15 mice per group), the control group (15 of 15 mice) remained metastasis-free (fig. S1H), whereas four of 15 mice in the sub-library group showed metastases on day 77 (Fig. 1B).

Figure 1. An in vivo CRISPRa screen identifies ACBP as a driver of bone metastasis.

(A) Workflow of in vivo CRISPRa-based screening for metastasis-promoting genes. (B) Bioluminescence imaging of 5-week-old male nude mice after intracardiac injection with 1 × 10⁶ H1299-Luc-RFP-dCas9-VPR cells transduced with a pooled gRNA library of kinases, phosphatases, and drug targets. Day 0 imaging was 1 hour after injection. Color scale, radiance. Hits identified as the encoded proteins. n = 15 mice. (C) Representative immunoblotting of ACBP, β-actin, and GAPDH in H1299-Luc-RFP-dCas9-VPR cells transduced with control sgRNA (ac-sgControl) or 3 independent sgRNAs targeting the expression of ACBP by CRISPRa (ac-sgACBP). (D) Bioluminescence imaging on day 93 after intracardiac injection with 1 × 10⁶ H1299-Luc-RFP-dCas9-VPR cells transduced with ac-sgControl or 3 independent ac-sgACBP (upper panel). Number of mice with metastasis per number of animals in each group and metastasis sites (lower panel). n = 7, 8, 6, and 8 mice. (E) Representative H&E staining and ACBP immunohistochemical staining of the bones from the mice described in (D). Scale bars, 4 mm (low magnification) and 200 μm (high magnification). Data represent n = 4.

Because most cancer cells did not survive after intracardiac injection (Fig. 1B and fig. S1H, day 0 vs. day 14), we reasoned that the metastases observed in the sub-library group were clonal and that we could use Sanger sequencing to identify the metastasis-inducing sgRNAs. At the endpoint, we isolated genomic DNA from individual visible metastases (all were bone metastases). The genomic DNA fragments encompassing the sgRNA target sites were amplified by PCR, and the PCR products were cloned into the TOPO vector. For each visible metastasis, we analyzed 12 clones by Sanger sequencing. From three of the four visible bone metastases, we identified two different sgRNAs targeting the gene DBI, which encodes ACBP. From the other bone metastasis, we identified two hits, natriuretic peptide receptor C (NPR3) and integrin alpha 2B (ITGA2B, encoding CD41), with one sgRNA targeting each gene. All clones from mouse #2 and mouse #8 (Fig. 1B) had the same sgRNA sequence 5′-GGGGAAAGGTGGCACCCTTC-3′ (table S1), which targets DBI; all clones from mouse #12 (Fig. 1B) had the same sgRNA sequence 5′-GGGGACGCCGGGAACAGGTG-3′ (table S1), which also targets DBI; the clones from mouse #13 (Fig. 1B) had one of the two sgRNA sequences, 5′-GGGCGCGATGGGTAGTGAAG-3′ (table S1), which targets NPR3, and 5′-GGCTAGAATTGCCAGGAAGT-3′ (table S1), which targets ITGA2B. Compared with H1299-Luc-RFP-dCas9-VPR cells transduced with control sgRNA, cells derived from the bone metastases from this screen showed higher expression of ACBP protein (fig. S1I).

The top hit from our screen, DBI, was initially identified as a small, secreted polypeptide in the brain, which can displace diazepam from the benzodiazepine binding site on the GABA_A receptor (24). Subsequently, an identical polypeptide was identified as a 10 kDa cytosolic protein capable of binding long-chain fatty acyl-CoA esters (LCACoAs) with high affinity (K_D: 1–10 nM), and was thus named ACBP (25). As an extracellular protein, ACBP/DBI regulates neurogenesis, neuronal survival, cognition, and behavior (26). As an intracellular protein, ACBP/DBI can shield LCACoAs from hydrolysis, extract LCACoAs from membranes, and deliver LCACoAs to glycerolipid, glycerophospholipid and ceramide synthesis, to fatty acid elongation, and to mitochondrial β-oxidation (26). The role of ACBP in cancer metastasis has not been reported. To confirm our screening result, we constructed additional ACBP-targeting CRISPRa cell lines by transducing H1299-Luc-RFP-dCas9-VPR cells with three independent sgRNAs (Fig. 1C). Consistently, all three ACBP sgRNAs induced bone metastasis; in addition, these sgRNAs exhibited varying abilities to promote metastatic outgrowth at other sites, including the brain, eye, pancreas, and adrenal gland (Fig. 1, D and E, and fig. S2, A and B).

ACBP promotes bone metastasis through its acyl-CoA-binding ability

To determine whether ACBP promotes metastatic bone colonization through its fatty acyl-CoA-binding ability, we generated a Y32A mutant of human ACBP, which retains the 3D structure of the protein but exhibits diminished binding affinity to LCACoA (27–29). In H1299 cells, overexpression of wild-type ACBP, but not of the Y32A mutant, induced bone colonization (Fig. 2, A to D, and fig. S3, A and B), suggesting that the bone metastasis-promoting function of ACBP is dependent on the ACBP-LCACoA interaction. Aside from the analysis of macroscopic metastases, we performed additional intracardiac injection with 1 × 10⁶ control or ACBP-overexpressing H1299 cells and euthanized the mice at 14 days (when the animals showed the lowest bioluminescence signals). Immunohistochemical staining of human-specific CA19–9 demonstrated that mice injected with wild-type ACBP-overexpressing H1299 cells had more single-cell and microscopic metastases in the bone, compared with mice injected with control or Y32A-mutant ACBP-overexpressing H1299 cells (fig. S3, C and D), suggesting that ACBP may promote extravasation or early colonization of the bone, or both. In contrast, ACBP overexpression did not affect the in vitro proliferation, wound-healing ability, migration across Transwell membranes, sphere-forming ability, or 3D growth of H1299 cells (fig. S4, A to E). Moreover, when H1299 cells were injected subcutaneously into nude mice, the growth of primary tumors was not affected by ectopic expression of either wild-type ACBP or the Y32A mutant (fig. S4, F to H).

Figure 2. ACBP promotes bone metastasis through its acyl-CoA-binding ability.

(A) Representative immunoblotting of ACBP, HSP90, and GAPDH in H1299-Luc-RFP cells transduced with the control vector, wild-type (WT) human ACBP, or the Y32A mutant. (B) Bioluminescence imaging of male nude mice on day 64 after intracardiac injection with 1 × 10⁶ H1299-Luc-RFP cells transduced with the control vector, wild-type ACBP, or the Y32A mutant (left panel). Summary of metastasis incidences and sites (right panel). n = 7, 8, and 7 mice. (C) Quantification of photon flux of the mice described in (B). Data are presented as mean ± SEM. Analysis by two-way ANOVA followed by Tukey’s post hoc test. **P < 0.01. n = 7, 8, and 7 mice. (D) H&E staining and ACBP immunohistochemical staining of the bones from the mice described in (B). Scale bars, 4 mm (low magnification) and 200 μm (high magnification). (E) Immunoblotting of ACBP and GAPDH in MDA-MB-231, LM2, and BoM-1833 cell lines. (F) Immunoblotting of Cas9 and GAPDH in BoM-1833 cells with or without Cas9 expression. (G) Immunoblotting of ACBP, HSP90, and GAPDH in 1833-Cas9 cells transduced with ACBP sgRNAs. (H) Immunoblotting of ACBP, HSP90, and GAPDH in control and ACBP-knockout BoM-1833 cells (generated by CRISPR-Cas9) with or without ectopic expression of ACBP (wild-type or the Y32A mutant). (I and J) Bioluminescence imaging (I) and quantification of photon flux (J) of female nude mice with intracardiac injection of 2 × 10⁵ control or ACBP-knockout BoM-1833 cells with or without ectopic expression of ACBP (wild-type or the Y32A mutant). Data are presented as mean ± SEM. Analysis by two-way ANOVA followed by Dunnett’s post hoc test. *P < 0.05. n = 6 mice. (K) H&E staining, immunohistochemical staining of ACBP and human-specific vimentin, and micro-CT imaging of the bones. Red stars indicate osteolytic lesions. Scale bars, 4 mm (first row) and 200 μm (second to fourth rows). n = 6 mice. Data in (D) to (H) represent n = 3.

Bone is one of the most frequent sites for metastasis in both lung cancer and breast cancer. To determine the effect of ACBP on breast cancer cells, we expressed wild-type ACBP and the Y32A mutant in the weakly metastatic breast cancer cell line MDA-MB-231 (fig. S5A). After injecting these cells into the mammary fat pad of female nude mice, we observed no difference in tumor growth (fig. S5, B to D). In contrast, after intracardiac injection, we found that overexpression of wild-type ACBP, but not the Y32A mutant, markedly promoted bone colonization (fig. S5, E to G). Next, we investigated whether ACBP is required for bone colonization by highly metastatic cancer cells. Compared with the parental MDA-MB-231 cell line, ACBP was found to be upregulated in the highly bone-metastatic subline BoM-1833 (30), but not in the lung-metastatic subline LM2 (31) (Fig. 2E). To determine the loss-of-function effect of ACBP, we knocked out ACBP in BoM-1833 cells using the CRISPR-Cas9 method (Fig. 2, F to H), which abolished bone metastasis formation, as gauged by bioluminescence imaging, micro-CT imaging, hematoxylinn & eosin (H&E) staining, and immunohistochemical staining of ACBP and human-specific vimentin (Fig. 2, I to K, and fig. S6). Re-expression of wild-type ACBP, but not the acyl-CoA-binding deficient mutant Y32A, restored the bone-metastatic ability of ACBP-knockout BoM-1833 cells (Fig. 2, H to K, and fig. S6).

ACBP is associated with metabolic signaling, bone metastatic ability, and poor clinical outcomes

Neither overexpression of ACBP in MDA-MB-231 cells (fig. S7A) nor knockout of ACBP in BoM-1833 cells (fig. S7B) affected the proliferation of these cells cultured in standard growth medium. In the presence of oleic acid (an 18-carbon monounsaturated fatty acid), overexpression of ACBP in MDA-MB-231 cells increased, and knockout of ACBP in BoM-1833 cells decreased cell proliferation (fig. S7, C and D). To gain insight into ACBP-regulated genes and pathways in the presence of long-chain fatty acid, we performed high-throughput RNA sequencing (RNA-seq) of control and ACBP-knockout BoM-1833 cells cultured in the presence of oleic acid. We analyzed the differentially expressed genes (DEGs; fold-change (FC) > 1.5; false discovery rate (FDR) < 0.05) and identified 341 downregulated genes and 389 upregulated genes upon ACBP depletion (data file S1). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of DEGs showed that genes downregulated in ACBP-knockout cells were enriched for molecular concepts related to metabolic signaling, electron transport chain, and extracellular matrix organization (Fig. 3A). Moreover, gene set enrichment analysis (GSEA) revealed that fatty acid metabolism signaling activity was enriched in control BoM-1833 cells, whereas oxidative stress-induced senescence signaling activity was enriched in ACBP-knockout BoM-1833 cells (Fig. 3B). qPCR analysis confirmed that depletion of ACBP upregulated the expression of oxidative stress genes, such as JUN, E2F2, and RBBP4, and downregulated the expression of the lipid metabolism gene CYP2J2 (Fig. 3C), validating our RNA-seq analysis. Also, compared with their isogenic cell lines, the bone-metastatic sublines BoM-1833 and PC-3M (a bone-metastatic subline of the PC-3 human prostate cancer cell line (32)) showed lower expression of JUN, E2F2, and RBBP4 as well as higher expression of CYP2J2 (fig. S8, A and B), which was consistent with the RNA-seq results. Furthermore, we collected bone metastasis samples from mice injected with control or ACBP-expressing MDA-MB-231 cells (fig. S5, E to G), finding that ACBP-overexpressing bone metastases had lower expression of JUN, E2F2, and RBBP4 as well as higher expression of CYP2J2 (fig. S8C). In addition, based on our RNA-seq analysis, IL1B (encoding interleukin 1β, IL-1β) and MMP9 were downregulated in ACBP-knockout BoM-1833 cells, which we confirmed by qPCR (fig. S8D). Whereas IL-1β has been shown to facilitate breast cancer cell extravasation and homing to the bone marrow (33), MMP9 can cleave type I, IV, and V collagen and plays an important role in the degradation of bone matrix (34).

Figure 3. ACBP is associated with metabolic signaling, bone metastatic ability, and poor clinical outcomes.

(A) Gene ontology (GO; red) and Kyoto Encyclopedia of Genes and Genomes (KEGG; blue) analyses of downregulated genes in ACBP-knockout BoM-1833 cells. Top enriched molecular concepts are shown on the y-axis, and the x-axis indicates enrichment significance. n = 3 biological replicates. (B) Gene set enrichment analysis (GSEA) for fatty acid metabolism signaling activity (left panel) and oxidative stress-induced senescence signaling activity (right panel) in control versus ACBP-knockout BoM-1833 cells. NES, normalized enrichment score; FDR, false discovery rate. n = 3 biological replicates. (C) qPCR of JUN, E2F2, RBBP4, and CYP2J2 in control and ACBP-knockout BoM-1833 cells. n = 3 biological replicates. (D and E) qPCR of Dbi (D) and immunoblotting of ACBP and GAPDH (E) in mouse mammary tumor cell lines. n = 3 biological replicates in (D). (F and G) qPCR of DBI (F) and immunoblotting of ACBP and GAPDH (G) in human prostate cancer cell lines. n = 3 biological replicates in (F). (H) TNMplot analysis of DBI mRNA expression (encoding ACBP) in normal (n = 242), tumor (n = 7,569), and metastatic (n = 82) tissues from breast cancer patients. Data are presented as violin plots. Analysis by the Kruskal-Wallis test followed by Dunn’s test. (I) Kaplan-Meier curves of relapse-free survival of breast cancer patients stratified by DBI mRNA expression (encoding ACBP), based on KM plotter analysis. n = 3,258 patients, DBI-low; n = 1,671 patients, DBI-high. HR, hazard ratio. (J) Kaplan-Meier curves of overall survival of lung cancer patients stratified by DBI mRNA expression, based on KM plotter analysis. n = 543 patients, DBI-low; n = 1,623 patients, DBI-high. (K) Kaplan-Meier curves of overall survival of breast cancer patients stratified by ACBP protein expression, based on KM plotter analysis. n = 39 patients, ACBP-low; n = 26 patients, ACBP-high. (L) Immunohistochemical staining of ACBP in primary tumors and bone metastases from breast and lung cancer patients. Scale bars, 200 μm (low magnification) and 50 μm (high magnification). (M) H scores of ACBP in primary tumors and bone metastases from breast cancer (n = 26 primary tumors and 40 bone metastases) and lung cancer (n = 35 primary tumors and 60 bone metastases) patients. Data are presented as mean ± SEM, dots represent individual replicates. Analysis by two-tailed unpaired Student’s t-test in (C) and (M). Analysis by one-way ANOVA followed by Tukey’s post hoc test in (D) or Holm-Sidak post-hoc test in (F). Analysis by the log-rank test in (I) to (K). **P < 0.01, ****P < 0.0001. Data in (E) and G) represent n = 3.

To expand the correlation analysis of ACBP with bone metastatic ability, we examined ACBP protein expression in additional cell lines. In 4T1.2 cells, a bone-metastatic subline of the 4T1 mouse mammary tumor cell line (35, 36), we observed higher expression of ACBP compared with the parental 4T1 cell line and its isogenic cell lines that cannot form macrometastases (37), including 67NR, 168FARN, and 4T07 (Fig. 3, D and E). In addition, ACBP expression was much higher in PC-3M cells, compared with the parental PC-3 cell line and two weakly metastatic human prostate cancer cell lines, LNCaP and DU145 (Fig. 3, F and G).

Next, we analyzed publicly available gene expression data from normal, tumor, and metastatic tissues from patients with breast cancer with the TNMplot tool (38). DBI mRNA was upregulated in tumor tissues compared with normal tissues and was further increased in metastatic tissues, albeit not specific to bone metastasis (Fig. 3H). To analyze the correlation between gene expression and clinical outcomes, we used the KM plotter tool to evaluate published data (39, 40), finding that high DBI mRNA expression correlated with poor relapse-free survival in breast cancer (Fig. 3I) and poor overall survival in lung cancer (Fig. 3J). Moreover, higher expression of ACBP protein was associated with shorter overall survival in patients with breast cancer (Fig. 3K). To determine the relevance of ACBP in bone metastasis, we performed immunohistochemical staining to examine ACBP expression in primary tumors and bone metastases from patients with lung or breast adenocarcinoma. In both cancer types, bone metastases showed higher immunostaining intensity of ACBP than primary tumors (Fig. 3, L and M). Thus, ACBP expression is associated with bone metastasis in human cancers.

Similar to metastasis to other organs, metastasis to the bone is highly inefficient. Upon arrival in the bone marrow, tumor cells interact with various bone-resident cells during colonization. Among these bone-resident cells, crosstalk between cancer cells and osteoclasts or osteoblasts has been extensively studied (41–43). Bone marrow adipocytes account for 50%–70% of the total volume of bone marrow in adults (44–47) and these adipocytes release free fatty acids, which can regulate osteoblasts, osteoclasts, and immune cells in the bone niche (48–51). Oleic acid is the topmost monounsaturated fatty acid in the human diet and the second most abundant in human tissues (52). Previous work has shown that oleic acid may protect metastasizing cells by reducing oxidative stress (53). By performing gas chromatography-mass spectrometry (GC-MS) analysis (54), we found that a panel of medium- and long-chain fatty acids (table S2), including oleic acid, were more abundant in bone tissues than in other common sites of metastasis, including the liver, brain, lung, and kidney (fig. S8, E and F), which may contribute to the preferential outgrowth of ACBP-overexpressing tumor cells in the bone.

The regulation of ACBP expression is poorly understood. To find the potential regulators of ACBP, we performed a computational analysis of the ChIP-seq datasets curated in the Cistrome Data Browser, which identified the transcriptional regulators that exhibited at least one ChIP-seq peak within ±1 kb of the transcriptional start site of the human DBI gene (table S3). Among the 69 identified candidates, androgen receptor (AR), RELA subunit of NF-κB, and the E1A-associated cellular p300 transcriptional co-activator protein (EP300) have been directly implicated in metastasis (55–57) and exhibited upregulation in BoM-1833 cells compared with MDA-MB-231 cells (fig. S9A). Knockdown of each of these three pro-metastatic transcription factors in BoM-1833 cells led to a significant decrease in ACBP expression (P < 0.05) (fig. S9, B and C), suggesting that AR, RELA, and EP300 may contribute to the upregulation of ACBP in bone-metastatic cells.

ACBP promotes long-chain fatty acid oxidation through its acyl-CoA-binding ability

To survive and thrive, metastasizing tumor cells need to acquire nutrients and energy, reduce reactive oxygen species (ROS), and alleviate cellular stress resulting from detachment from the extracellular matrix (5). Some cancer cells overcome this challenge by elevating FAO-associated production of energy and antioxidants. The first rate-limiting step of FAO is the transport of LCACoA esters into mitochondria through carnitine palmitoyltransferase 1 (CPT1), and ACBP can bind LCACoA to increase its affinity to CPT1 (26, 58–60) (Fig. 4A). After entering the mitochondria, LCACoA undergoes multiple rounds of β-oxidation to generate NADH, FADH₂, and acetyl-CoA. NADH and FADH₂ then enter the electron transport chain to generate ATP (8, 61) (Fig. 4A). Although FAO does not directly produce NADPH, it can lead to an increase in cytoplasmic NADPH. This occurs as acetyl-CoA enters the tricarboxylic acid (TCA) cycle to generate isocitrate, which is exported to the cytoplasm and participates in NADPH-producing reactions through the isocitrate–α-ketoglutarate shuttle (62) (Fig. 4A).

Figure 4. ACBP promotes long-chain fatty acid oxidation through its acyl-CoA-binding ability.

(A) Schematic representation of the FAO pathway. (B) Mitochondrial bioenergetics of control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells cultured with or without BSA-conjugated palmitate (PA). Oxygen consumption rates (OCR) were measured by the Seahorse metabolic analyzer. n = 6 wells. FCCP, carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone. (C and D) Maximal respiration rates (C) and ATP production rates (D) of the samples described in (B). n = 6 wells. (E) Immunoblotting of ACBP, HSP90, and GAPDH in control and ACBP CRISPR knockout 4T1.2 cells with or without ectopic expression of ACBP (wild-type or the Y32A mutant). (F) Mitochondrial bioenergetics of control, ACBP-knockout, and ACBP-restored 4T1.2 cells cultured with or without BSA-conjugated palmitate (PA). OCR were measured by the Seahorse metabolic analyzer. n = 6 wells. (G and H) Maximal respiration rates (G) and ATP production rates (H) of the samples described in (F). n = 6 wells. (I to K) Fluorescence images (I), flow cytometry plots (J), and data quantification (K; by MFI, mean fluorescence intensity) from control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells that were cultured in the presence of 33.3 μM oleic acid (OA)-albumin for 24 hours before staining with BODIPY 493/503 for 15 minutes. Scale bars, 100 μm. n = 3 biological replicates. (L to N) Fluorescence images (L), flow cytometry plots (M), and data quantification (N) from control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells that were cultured in the presence of 33.3 μM palmitic acid (PA)-albumin for 24 hours before staining with BODIPY 493/503 for 15 minutes. Scale bars, 100 μm. n = 3 biological replicates. Data are presented as mean ± SEM, dots represent individual replicates. Analysis by one-way ANOVA followed by Tukey’s post hoc test in (C), (D), (G), (H), (K), and (N). *P < 0.05, **P < 0.01, ****P < 0.0001. Data in (E), (I), and (L) represent n = 3.

We reasoned that overexpression of ACBP in cancer cells could increase the rate of mitochondrial FAO, which would be beneficial for metastasizing cells in the presence of fatty acids (Fig. 4A). To test this, we used the Seahorse XF96 Analyzer to gauge mitochondrial bioenergetics of control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells cultured with or without a fatty acid substrate, BSA-conjugated palmitate. Compared with the BSA control, the addition of palmitic acid (a 16-carbon saturated fatty acid) significantly increased the respiration rate and mitochondrial ATP production rate in H1299 cells overexpressing wild-type ACBP (P < 0.02), but not in the control H1299 cells or H1299 cells overexpressing the acyl-CoA-binding-deficient mutant (Fig. 4, B to D), and these changes were abolished by treatment with the FAO blocker etomoxir (fig. S10, A to C), which is an inhibitor of CPT1 (63). Conversely, in the bone-metastatic mammary tumor cell line 4T1.2, CRISPR-Cas9-mediated depletion of ACBP abrogated palmitic acid-induced respiration rate and mitochondrial ATP production rate, which was rescued by ectopic expression of wild-type ACBP but not the Y32A mutant (Fig. 4, E to H).

Next, we incubated the cells with 1 μM BODIPY-conjugated palmitate (BODIPY FL C16) (64, 65) for 5 minutes, finding that overexpression of ACBP in H1299 cells did not affect the uptake rate of long-chain fatty acid, as gauged by immunofluorescent staining and flow cytometry (fig. S10, D to F). We then used BODIPY 493/503 staining to monitor fatty acid consumption, as described previously (66), the cells were cultured with oleic acid, palmitic acid, or butyrate (a short-chain fatty acid) for 24 hours before staining with BODIPY 493/503. We found that in H1299 cells, overexpression of wild-type ACBP, but not the Y32A mutant, significantly reduced cellular accumulation of long-chain fatty acids such as oleic acid (P < 0.01) (Fig. 4, I to K) and palmitic acid (P < 0.01) (Fig. 4, L to N), but not accumulation of short-chain fatty acids such as butyrate (P > 0.05) (fig. S10, G to I).

To further substantiate these results, we performed ACBP gain-of-function experiments in two mammary tumor cell lines, MDA-MB-231 and Hs578T, which have low expression of endogenous ACBP (fig. S11, A and B), as well as loss-of-function and rescue experiments in three bone-metastatic cancer cell lines, BoM-1833, 4T1.2, and PC-3M (Figs. 2H and 4E, and fig. S11C). In these cell lines, ACBP overexpression or knockout did not alter long-chain fatty acid uptake (fig. S11, D to H). In contrast, overexpression of ACBP (but not the Y32A mutant) in MDA-MB-231 and Hs578T cells increased long-chain fatty acid consumption (fig. S12, A to F), and knockout of ACBP in BoM-1833, 4T1.2, and PC-3M cells decreased cellular consumption of long-chain fatty acids, which was reversed by re-expression of wild-type ACBP but not the Y32A mutant (fig. S13, A to F, and fig. S14, A to C). Collectively, our data demonstrate that ACBP promotes long-chain fatty acid oxidation/consumption through its LCACoA-binding ability.

ACBP upregulates NADPH, downregulates ROS, and inhibits lipid peroxidation and ferroptosis

Only a small fraction of disseminated tumor cells survive in circulation and extravasate into distant organs to form secondary tumors. Cell death can be triggered by detachment from the extracellular matrix, which can lead to catastrophic metabolic alterations, including defective glucose uptake, diminished pentose phosphate pathway (PPP) flux, reduced cellular ATP, and increased ROS (5, 6). Considering that ACBP promotes metastatic bone colonization and FAO through its acyl-CoA-binding ability, we reasoned that ACBP-mediated activation of FAO leads to an increase in cytoplasmic NADPH to counteract ROS. Indeed, in the presence of oleic acid, overexpression of ACBP in H1299, MDA-MB-231, and Hs578T cells reduced the NADP+/NADPH ratio (Fig. 5A and fig. S15, A and B), suggesting that ACBP increases NADPH under long-chain fatty acid-rich conditions. Conversely, knockout of ACBP in BoM-1833, 4T1.2, and PC-3M cells cultured with oleic acid increased the NADP+/NADPH ratio, which was reversed by re-expression of ACBP (Fig. 5B and fig. S15, C and D).

Figure 5. ACBP upregulates NADPH, downregulates ROS, and inhibits lipid peroxidation and ferroptosis.

(A) NADP⁺/NADPH ratios of control and ACBP-overexpressing H1299 cells cultured in the serum-starved medium (1% FBS) for 12 hours with or without 10% FBS and 33.3 μM oleic acid (OA)-albumin for 4 additional hours. (B) NADP⁺/NADPH ratios of control, ACBP-knockout, and ACBP-restored BoM-1833 cells cultured in the serum-starved medium for 8 hours with or without 10% FBS and 33.3 μM oleic acid (OA)-albumin for 3 additional hours. (C) Relative GSH content in control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells cultured with 10% FBS or 1% FBS with or without and 33.3 μM oleic acid (OA)-albumin for 9 hours. GSH content (light unit) was normalized to the cell number. (D) Relative GSH content in control, ACBP-knockout, and ACBP-restored BoM-1833 cells cultured with 10% FBS or 1% FBS with or without and 33.3 μM oleic acid (OA)-albumin for 9 hours. GSH content (light unit) was normalized to the cell number. (E) Quantification of flow cytometry results from control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells that were cultured with 200 μM H₂O₂ and 33.3 μM oleic acid (OA)-albumin, alone or in combination, before staining with the ROS-reactive fluorescent probe 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) for 30 minutes. (F to H) Control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells cultured with 5% charcoal-stripped FBS (F), 10% FBS (G), or 5% charcoal-stripped FBS plus 33.3 μM oleic acid (OA)-albumin (H) were treated with different doses of RSL3 for 12 hours. Cell viability was determined by a CCK8 assay. (I) Control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells cultured with 10% FBS were treated with 0.3 μM RSL3 for 12 hours, with or without co-treatment with 1 μM ferrostatin-1 (Fer-1) or 50 μM DFO. Cell viability was determined by a CCK8 assay. (J) Flow cytometry plots (left panel) and data quantification (right panel; by MFI, mean fluorescence intensity) from control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells that were cultured with 33.3 μM oleic acid (OA)-albumin and 0.3 μM RSL3 for 6 hours, before staining with the lipid peroxide indicator BODIPY C11 for 30 minutes. (K and L) The percentage of annexin V and 7-AAD double-negative population in control, wild-type ACBP-, or Y32A mutant ACBP-overexpressing H1299 cells that were cultured with 33.3 μM oleic acid (OA)-albumin and 0.3 μM RSL3 (K) or 8 μM erastin (L) for 6 hours, with or without co-treatment with 1 μM ferrostatin-1 (Fer-1) or 50 μM DFO. (M) Flow cytometry plots (left panel) and data quantification (right panel) from control, ACBP-knockout, and ACBP-restored BoM-1833 cells that were cultured with 33.3 μM oleic acid (OA)-albumin and 0.4 μM RSL3 for 8 hours, before staining with the lipid peroxide indicator BODIPY C11 for 30 minutes. (N) Control, ACBP-knockout, and ACBP-restored BoM-1833 cells cultured with 5% charcoal-stripped FBS plus 33.3 μM oleic acid (OA)-albumin were treated with different doses of RSL3 for 12 hours. Cell viability was determined by a CCK8 assay. Analysis by two-way (A to D) or one-way (E to N) ANOVA followed by Sidak post-hoc test (A and E), Dunnett post hoc test (B), or Tukey’s post hoc test (C, D, and F to N). Data are presented as mean ± SEM, dots represent individual replicates. *P < 0.05, **P < 0.01, ****P < 0.0001. n = 3 biological replicates in all panels except for (E) (n = 4 biological replicates).

NADPH-dependent conversion of oxidized glutathione (GSSG) to glutathione (GSH) by glutathione reductase is critical for counteracting oxidative stress (8) (Fig. 4A). Consistent with increased NADPH, H1299 cells overexpressing wild-type ACBP, but not the Y32A mutant, showed upregulation of GSH when cultured with oleic acid (Fig. 5C). Conversely, depletion of ACBP in BoM-1833 cells cultured with oleic acid reduced GSH, which was rescued by wild-type ACBP but not the Y32A mutant (Fig. 5D). Moreover, in H1299 cells cultured in the presence of long-chain fatty acid (oleic acid) and H₂O₂, overexpression of wild-type ACBP, but not of the Y32A mutant, reduced ROS (Fig. 5E).

A recent study (53) showed that oleic acid in the lymphatic fluid protects disseminated melanoma cells from ferroptosis, a non-apoptotic form of cell death characterized by lipid peroxidation (67), which in turn facilitates metastasis. The essence of ferroptosis is the reaction between oxidative free radicals and membrane lipid polyunsaturated fatty acids (PUFAs) – but not monounsaturated fatty acids (MUFAs) – which generates excessive lipid peroxides leading to membrane damage and cell death (68–71). In H1299 cells cultured in the presence or absence of oleic acid, overexpression of ACBP (wild-type or Y32A) did not affect expression of SLC7A11 and CPT1 proteins (fig. S15E). When cultured with full serum or oleic acid, but not charcoal-stripped serum, H1299 cells expressing wild-type ACBP, but not the Y32A mutant, showed reduced sensitivity to the classic ferroptosis inducer RSL3 (Fig. 5, F to H); the cell death induced by RSL3 could be rescued by co-treatment with the ferroptosis inhibitor ferrostatin-1 (67) or the iron chelator deferoxamine (DFO) (72) (Fig. 5I). We also observed the same effects with another ferroptosis inducer, erastin (fig. S15, F and G). The hallmark of ferroptosis, lipid peroxidation, can be gauged by BODIPY C11 staining (67). Overexpression of wild-type ACBP, but not of the Y32A mutant, decreased lipid peroxidation (Fig. 5J and fig. S15H) and increased the surviving fraction (as gauged by flow cytometry analysis of the 7-AAD/annexin V double-negative population; Fig. 5, K and L, and fig. S15, I and J) in RSL3- or erastin-treated H1299 cells cultured with oleic acid. Similarly, overexpression of ACBP in MDA-MB-231 (fig. S16, A to D) and DU145 (fig. S16, E to H) cells reduced lipid peroxidation and decreased sensitivity to ferroptosis inducers. Conversely, knockout of ACBP in BoM-1833 cells increased lipid peroxidation (Fig. 5M) and enhanced sensitivity to RSL3 (Fig. 5N), which was reversed by wild-type ACBP but not the Y32A mutant (Fig. 5, M and N). Taken together, these data suggest that in the presence of oleic acid, ACBP can protect cancer cells from lipid peroxidation and reduce their sensitivity to ferroptosis.

FAO blockade or ferroptosis inducer treatment inhibits bone metastasis

CPT1, which localizes to the outer membrane of mitochondria, facilitates the transport of fatty acyl-CoA into mitochondria by converting acyl-CoAs into acylcarnitines, an essential intermediate of FAO (73). To investigate the role of CPT1 in ACBP-induced bone metastasis, we used two independent shRNAs to deplete CPT1 in ACBP-overexpressing H1299 cells (Fig. 6, A and B). This resulted in the reversal of ACBP-induced bone metastasis formation after these cells were implanted into mice through intracardiac injection (Fig. 6, C to E, and fig. S17, A and B), suggesting that ACBP drives metastatic bone colonization through CPT1. This finding aligns with our data indicating the dependence of ACBP’s bone metastasis-promoting function on its acyl-CoA-binding ability.

Figure 6. FAO blockade or ferroptosis inducer treatment inhibits bone metastasis.

(A) Immunoblotting of CPT1 and β-actin in H1299 cells transduced with control shRNA or three shRNAs targeting CPT1. (B) Immunoblotting of CPT1, ACBP, and β-actin in control and ACBP-overexpressing H1299-Luc-RFP cells with or without knockdown of CPT1. (C) Summary of metastasis incidences and sites in male nude mice on day 110 after intracardiac injection with 1 × 10⁶ control or ACBP-overexpressing H1299-Luc-RFP cells with or without knockdown of CPT1. (D) Quantification of photon flux of the mice described in (C) at the indicated times. n = 10 mice. (E) H&E staining and ACBP immunohistochemical staining of the bones from the mice described in (C). Scale bars, 4 mm (low magnification) and 200 μm (high magnification). (F) Workflow of the treatment of the BoM-1833 bone metastasis model with vehicle, etomoxir, or IKE. (G) Quantification of photon flux of the mice described in (F) at the indicated times. n = 5 mice. (H) H&E staining and ACBP immunohistochemical staining of the bones from the mice described in (F). Scale bars, 2 mm. (I) Workflow of the treatment of the ACBP-overexpressing H1299 bone metastasis model with vehicle, etomoxir, or IKE. (J and K) Bioluminescence imaging (J) and summary of metastasis incidences and sites (K) in the mice described in (I), on day 75 after intracardiac injection with 1 × 10⁶ ACBP-overexpressing H1299 cells. (L) Quantification of photon flux of the mice described in (I) at the indicated times. n = 7 mice. (M) H&E staining and ACBP immunohistochemical staining of the bones from the mice described in (I). Scale bars, 2 mm. Analysis by two-way ANOVA followed by Tukey’s post hoc test in (D), (G), and (L). Data are presented as mean ± SEM. Dots in (G) and (L) represent individual animals. **P < 0.01, ****P < 0.0001. Data in (A), (B), (E), (H), and (M) represent n = 3.

In addition to CPT1 knockdown, we evaluated the effect of etomoxir, a pharmacological inhibitor of CPT1, on bone colonization by BoM-1833 breast cancer cells, which are high ACBP-expressing cells. Concurrently, a separate group of BoM-1833 tumor cell-bearing mice were treated with imidazole ketone erastin (IKE), a well-characterized in vivo ferroptosis inducer that inhibits the activity of the cystine transporter SLC7A11 (74–78). The tumor cells were implanted into nude mice through intratibial injection. From day 8 post-tumor cell implantation, the mice received one week of daily etomoxir (30 mg/kg, i.p.) or daily IKE (50 mg/kg, i.p.) (Fig. 6F). Bioluminescence imaging revealed that compared with the vehicle group, mice treated with etomoxir had significantly lower signals in the tibiae (P < 0.0001; Fig. 6G and fig. S17, C and D). IKE treatment exhibited inhibitory effects on bone metastasis comparable to those of etomoxir treatment (Fig. 6G and fig. S17, C and D). Moreover, H&E staining of bone sections and immunohistochemical staining of ACBP demonstrated markedly fewer bone metastases in etomoxir- or IKE-treated mice (Fig. 6H).

Compared with the vehicle group, mice treated with etomoxir or IKE had increased staining intensity of 4-hydroxy-2-noneal (4-HNE), a widely used marker of lipid peroxidation, in bone metastases (fig. S17E), suggesting that ferroptosis may be relevant to the observed drug effect. On the other hand, neither drug reduced the expression of SLC7A11 (fig. S17E). In parallel, a toxicity assessment during a one-week daily treatment with etomoxir (10–30 mg/kg) or IKE (10–50 mg/kg) showed no significant changes in body weight or tissue histology of livers and kidneys (P > 0.05) (fig. S18, A to D). In addition, we examined multiple ACBP-overexpressing cell lines (H1299, MDA-MB-231, and DU145) treated with etomoxir or IKE, finding no substantial effects of these two drugs on ACBP protein expression (fig. S18E).

We further tested these two drugs in mice that received intracardiac injection of ACBP-overexpressing H1299 cells (fig. S19). From day 15 post-tumor cell implantation, the mice received daily intraperitoneal injections of either 30 mg/kg etomoxir or 50 mg/kg IKE, and the treatment lasted one week (Fig. 6I). All mice treated with vehicle (7/7) developed evident metastases, with the bone being the predominant metastasis site (Fig. 6, J to M). In contrast, no mice treated with etomoxir or IKE (0/7 for either drug) showed detectable metastases, as gauged by bioluminescence imaging of live animals, H&E staining of bone sections, and immunohistochemical staining of ACBP (Fig. 6, J to M). Collectively, FAO blockade or ferroptosis inducer treatment inhibits bone metastasis in preclinical models of breast cancer and lung cancer.

DISCUSSION

The premise of forward genetics screening is to identify genes responsible for a phenotype in an unbiased manner. Here, leveraging an in vivo positive selection system based on CRISPR activation, we conducted a forward genetics screen that led to the discovery of a bone metastasis driver, ACBP, whose role in cancer metastasis has not been reported previously. We subsequently validated the metastasis-promoting role of ACBP in lung and breast cancer cells by independent CRISPRa gRNAs, cDNA overexpression, CRISPR knockout, and rescue experiments, thereby supporting the reliability of our screening approach. The role of lipid metabolism in bone metastasis remains elusive. In this study, we found that ACBP regulates lipid metabolism, boosting FAO while protecting against lipid peroxidation and ferroptosis, in lung and breast cancer cells, which is linked to their bone-metastatic ability. Taken together, our findings demonstrate the power of coupling positive selection with next-generation forward genetics screening in vivo, and uncover the pivotal role of lipid metabolism in enabling tumor cells to adapt and thrive in the bone microenvironment.

The distribution of distant metastases to specific organs is a non-random process known as metastatic organotropism, which is regulated by multiple factors including cancer subtypes, molecular features of tumor cells, the microenvironment, and crosstalk and interactions of cancer cells with niche cells (79). The metabolic microenvironment (such as lipid content and adipocytes) is one of the multiple aspects of the metastatic microenvironment. Our experiments performed with the acyl-CoA binding deficient mutant of ACBP (Y32A) in both overexpression and rescue settings demonstrate that ACBP promotes bone metastasis in a fatty acyl CoA-dependent manner; however, FAO (and the lipid content at the metastatic site) is probably not the only factor contributing to ACBP-induced metastasis. For example, regulation of IL-1β and MMP9 expression may provide an additional explanation aside from lipid metabolism as to why ACBP promotes bone metastasis. Therefore, future work is needed to elucidate other mechanisms of action of ACBP in regulating malignant progression.

We observed a significant correlation of ACBP expression with bone metastases (P < 0.0001) and poor clinical outcomes (P < 0.05) in human breast and lung cancers. Lung cancer is the second most common cancer in both genders, after prostate cancer in men and breast cancer in women (80). Current therapeutic interventions remain largely palliative in patients with bone metastases. For instance, treatment with denosumab, a receptor activator of NF-κB ligand (RANKL)-neutralizing antibody, can inhibit bone resorption, reduce the risk of skeletal-related events, and alleviate bone pain; however, its clinical use for bone metastasis is insufficient (81). Therefore, understanding and targeting the molecular underpinnings of bone metastasis represent one of the most pressing challenges in cancer therapy. Based on our preclinical data, therapeutic agents that target FAO or induce ferroptosis have the potential as drugs for treating bone metastasis. Although ACBP rendered cancer cells less sensitive to ferroptosis inducers in vitro, at higher doses, ACBP-overexpressing cells still underwent cell death. Consistent with this, we observed upregulation of the lipid peroxidation marker 4-HNE in bone metastases from mice treated with etomoxir or IKE compared with mice treated with vehicle.

This study has several limitations. Although our screening strategy is a straightforward way of discovering strong metastasis drivers, it is a low-yield method due to its high stringency. Because we used a cancer cell line with no detectable metastatic ability after intracardiac injection into nude mice, and because only a very small population of circulating tumor cells in the sub-library group can colonize the bone, the theoretical redundancy of the screening for metastatic bone colonization is not sufficient to guarantee a fully unbiased screen. Second, through functional experiments, mechanistic studies, and clinical validation, we characterized the role of ACBP in breast and lung cancers; although we also observed upregulation of ACBP expression in highly bone-metastatic prostate cancer cells, whether ACBP governs lipid metabolism and bone metastasis in prostate cancer and other cancer types warrants further investigation. Third, ACBP has cytoplasmic and secreted forms. In our study, we compared wild-type ACBP and the acyl-CoA binding deficient mutant (Y32A), and their distinct phenotypes indicate the functional dependence on the binding of ACBP to fatty acyl-CoA, which occurs in the cytoplasm. However, we cannot exclude the possibility that ACBP secreted by ACBP-high tumor cells can be taken up by ACBP-low tumor cells, thus propagating FAO activation and promoting metastasis. Alternatively, extracellular ACBP may bind to surface proteins such as cell membrane receptors on tumor cells or niche cells to modulate cell signaling. Finally, the FAO blocker etomoxir and the ferroptosis inducer IKE have been tested in animal studies; however, their efficacy and safety in cancer patients are unknown. Clinical translation may require the development of safer and more robust drugs.

MATERIALS AND METHODS

Study design

This study aimed to discover valid targets for bone metastasis treatment by harnessing the power of next-generation forward genetics screening. To this end, we conducted an in vivo metastasis screen by using H1299-Luc-RFP-dCas9-VPR lung cancer cells transduced with a pooled CRISPRa gRNA sub-library of 2,320 genes (Addgene, 83980), which identified a previously undescribed bone metastasis driver, ACBP. We then substantiated the metastasis-promoting role of ACBP in lung cancer and breast cancer cells by independent CRISPRa gRNAs, cDNA overexpression, CRISPR knockout, and rescue experiments (rescuing its loss with wild-type ACBP or its fatty acyl-CoA-binding deficient mutant), validating our screening. Bone metastases in mice were evaluated by bioluminescence imaging, autopsy, H&E staining, and immunohistochemical staining. Osteolytic lesions were assessed by micro-CT imaging. Fatty acid oxidation and consumption rates were gauged by a Seahorse metabolic analyzer and BODIPY 493/503 staining. Lipid peroxidation was measured by BODIPY C11 staining. Cell death induced by ferroptosis inducers, including RSL, erastin, and ML210 (with or without co-treatment with the ferroptosis inhibitor ferrostatin-1 or the iron chelator deferoxamine), was evaluated by a CCK8 assay and 7-AAD/annexin V double staining. High-throughput RNA-sequencing, Gene Ontology analysis, and KEGG pathway analysis were performed with ACBP-knockout BoM-1833 cells cultured in the presence of oleic acid. To determine the relevance of our findings in human cancer, we analyzed samples from patients with lung or breast cancer and performed correlation analyses by using publicly available datasets. To evaluate the therapeutic potential in bone metastasis, we treated bone metastasis-bearing mice with an FAO blocker or a ferroptosis inducer. The sample size was chosen based on the results of previous studies. Mice were randomly assigned to different treatment groups. Investigators were not blinded. No data points were excluded. Sample sizes are indicated in the figure legends.

Mice

All animal studies were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee of MD Anderson Cancer Center (protocol number: 00001012-RN03). Animals were housed at 21.1 °−23.3 °C (set point: 22.2 °C) with 40%–55% humidity (set point: 45%). The light cycle of animal rooms is 12 hours of light and 12 hours of dark. Nude mice were purchased from the Jackson Laboratory (stock number: 002019). Male mice were used as recipients of H1299 cells and female mice were used as recipients of MDA-MB-231 and BoM-1833 cells. Mice were randomly assigned to different groups and received intracardiac or intratibial injections of tumor cells at 5 weeks of age. Animal procedures were performed in a specific pathogen-free facility. Sample sizes were determined based on our preliminary experiments. Blinding was not performed.

Cell lines

All cell lines were cultured in a 37 °C incubator in a humidified, 5% CO2 atmosphere. The HEK293T, H1299, Hs578T, and MDA-MB-231 cell lines were from the American Type Culture Collection (ATCC). The luciferase-expressing LM2 and BoM-1833 cell lines were from X. Zhang (Baylor College of Medicine). The LNCaP, DU145, PC-3, and PC-3M cell lines were from D.Z.’s lab stock. The 67NR, 168FARN, 4T07, 4T1, and 4T1.2 cell lines were from L.M.’s lab stock. The H1299 and LNCaP cell lines were cultured in RPMI-1640 supplemented with 10% fetal bovine serum (FBS) and 10,000 U/mL penicillin-streptomycin. The MDA-MB-231, LM2, BoM-1833, 67NR, 168FARN, 4T07, 4T1, 4T1.2, Hs578T, PC-3M, and HEK293T cell lines were cultured in Dulbecco modified Eagle medium (DMEM) supplemented with 10% FBS and 10,000 U/mL penicillin-streptomycin. DU145 cells were cultured in Eagle’s Minimum Essential Medium (EMEM) supplemented with 10% FBS and 10,000 U/mL penicillin-streptomycin. The PC-3 cell line was cultured in Kaighn’s Modification of Ham’s F-12 medium supplemented with 10% FBS and 10,000 U/mL penicillin-streptomycin. Short tandem repeat profiling (for cell line authentication) and mycoplasma tests were done by ATCC and MD Anderson’s Cytogenetics and Cell Authentication Core.

Lentiviral transduction

Lentiviruses were produced in HEK293T cells by co-transfection with the viral vector and packaging plasmids (pMD2.G: Addgene, 12259; psPAX2: Addgene, 12260). Two days after transfection, the viral supernatant was harvested, filtered through a 0.45 μm filter, and added to the target cells in the presence of polybrene reagent (Sigma-Aldrich, R-1003-G) at 4 μg/mL. The infected cells were selected with puromycin, hygromycin B, or blasticidin.

Sub-library amplification

We amplified a CRISPRa gRNA sub-library of kinases, phosphatases, and drug targets (Addgene, 83980; genes targeted: 2,320; targeting gRNAs: 12,780; control gRNAs: 250) (22) following Jonathan Weissman’s lab protocol with minor modifications. The gRNA sequences and gene list of the sub-library can be downloaded from Addgene (related to Addgene, 83980) (22). Briefly, 1 μL of the pooled CRISPRa gRNA sub-library was electroporated into Endura Electrocompetent Cells (Lucigen, 60242–2) with a transformation efficiency of >1 × 10¹⁰ cfu/μg DNA. Electroporation was performed according to the manufacturer’s instructions (1.8 kV with 10-μF constant capacitance and 600-Ohm resistance, in a 0.1-cm cuvette), and 1 mL SOC medium was added to the transformed bacteria immediately (within 10 s), followed by shaking at 37 °C for 1.5 hour. Then, 5 μL was used for serial dilutions and plating on LB agar plates. After a 1:20,000 dilution, approximately 1,500 colonies were obtained. Thus, the library coverage = 1500 × 20000 ÷ 13030 (total number of gRNAs) = 2302. The remaining recovered bacteria were added to 500 mL LB with carbenicillin, followed by shaking at 37 °C for 16 hours. Extraction of the sub-library plasmids was performed by using the Qiagen Megaprep Kit (Qiagen, 12981).

In vivo CRISPRa screening

H1299 cells were transduced with lentiviral FU-luciferase-CRW/RFP and then subjected to fluorescence-activated cell sorting (FACS). The top 25% RFP-positive cells (H1299-Luc-RFP) were transduced with dCas9-VPR (Addgene, 96917) and selected with 15 μg/mL blasticidin (Sigma-Aldrich, 15205–100MG). Subsequently, 2 × 10⁸ H1299-Luc-RFP-dCas9-VPR cells were transduced with the CRISPRa gRNA sub-library (on a pCRISPRia-v2 vector expressing blue fluorescent protein) (22) at a multiplicity of infection of approximately 0.3 and selected with 2 μg/mL puromycin (Life Technologies, A1113803) for 7 days. Next, 5 × 10⁷ cells were passaged every 48 hours at a density of 4 × 10⁶ cells per 15-cm dish in RPMI 1640. After 14 days, cells were introduced through intracardiac injections into 5-week-old male nude mice at a dose of 1 × 10⁶ cells per mouse, 15 mice per group. Thus, the theoretical redundancy of the screening = 1.5 × 10⁷ ÷ 13030 = 1151. The mice were monitored by bioluminescence imaging for metastases. At the endpoint, genomic DNA was isolated from individual visible metastases (all were bone metastases) by using the QIAamp Blood & Cell Culture DNA Mini Kit (Qiagen, 13343). Then, the genomic DNA fragments encompassing the sgRNA target sites were amplified by PCR, and the PCR products were cloned into the TOPO vector by using the TOPO TA Cloning Kit (ThermoFisher Scientific, 450030). The forward primer used for PCR was 5′-GCACAAAAGGAAACTCACCCT-3′, and the reverse primer used for PCR was 5′-TCGACTCGGTGCCACTTTTTC-3′. For each visible metastasis, we analyzed 12 clones by Sanger sequencing. The sequencing primer was 5′-GTAAAACGACGGCCAGTG-3′. The sequencing results are presented in table S1.

Metastasis assays, bioluminescence imaging, and drug treatment

For ACBP gain- or loss-of-function experiments, 1 × 10⁶ ACBP-overexpressing H1299 cells, 2 × 10⁵ ACBP-overexpressing MDA-MB-231 cells, or 2 × 10⁵ ACBP-knockout BoM-1833 cells in 100 μL PBS were injected into the left cardiac ventricle of anesthetized athymic nude mice by using a 26G syringe. One hour after injection, the mice were injected intraperitoneally with the luciferase substrate D-luciferin (PerkinElmer, 122799–5) and placed in the IVIS 200 Imaging System for whole-body imaging. Successful intracardiac injection was indicated by bioluminescence signals distributed throughout the body. Only mice with satisfactory injections were used for the following procedures. Weekly (for BoM-1833 and MDA-MB-231 cells) or biweekly (for H1299 cells) bioluminescence imaging was performed to assess metastasis till the indicated endpoints. For drug treatment experiments, two models were used: (1) 2 × 10⁵ BoM-1833 cells in 50 μL PBS were injected into the tibia of anesthetized athymic female nude mice, and we started the drug treatment one week after tumor cell implantation; (2) 1 × 10⁶ ACBP-overexpressing H1299 cells in 100 μL PBS were injected into the left cardiac ventricle of anesthetized athymic male nude mice, and we started the drug treatment two weeks after tumor cell implantation. In both models, the mice received daily intraperitoneal injections of either 50 mg/kg imidazole ketone erastin (IKE, dissolved in 5% DMSO, 40% PEG300, 5% Tween 80, and 50% ddH2O) or 30 mg/kg etomoxir (dissolved in 10% DMSO, 40% PEG300, 5% Tween 80, and 45% ddH2O). The treatment lasted one week. Bioluminescence imaging was performed to assess metastasis in live animals.

Generation of CRISPR knockout or CRISPR activation cell lines

For CRISPR knockout, we cloned sgRNAs targeting the gene of interest into lentiGuide-Puro and used a two-vector knockout system (lentiGuide-Puro, Addgene, 52963, and lentiCas9-Blast, Addgene, 52962) for knockout cell line construction. For CRISPR activation, we cloned sgRNAs targeting the gene of interest into pCRISPRia-v2 and used a two-vector activation system (pCRISPRia-v2 and dCas9-VPR) for CRISPR activation cell line construction. The gRNA sequences are listed in table S4.

Overexpression and knockdown

Human ACBP cDNA was amplified by PCR from a commercial vector (DNASU, clone ID: HsCD00353026) and cloned into the pDONR201 vector through the BP reaction. The Y32A mutant of ACBP in the pDONR201 vector was generated by using the QuikChange Kit (Agilent Technologies) and validated by sequencing. Full-length wild-type ACBP and the Y32A mutant, as well as their knockout-resistant forms, were subcloned into pLenti-CMV-puromycin and pLenti-CMV-hygromycin destination vectors through the LR reaction. CPT1 knockdown was performed by using predesigned shRNAs from Sigma-Aldrich and the Clone IDs are TRCN0000036281, TRCN0000036282, and TRCN0000036283. For siRNA-mediated knockdown, one negative control and two predesigned gene-specific siRNAs were purchased from Sigma-Aldrich. siRNAs were transfected into cells by using the Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher Scientific, 13778150). Total RNA was extracted 48 hours after transfection for RT-qPCR analysis of knockdown efficiency. siRNA sequences are listed in table S4.

Human samples and immunostaining

The study was approved by the institutional review board of Shanghai General Hospital (protocol number: 2018KY081). Written informed consent for the collection of tissue samples and clinical data was obtained from the patients or their legal guardians. Paraffin-embedded tissue samples were obtained from Shanghai General Hospital. The lung cancer cohort included 35 primary lung tumor samples and 60 bone metastasis samples. The breast cancer cohort included 26 primary breast tumor samples and 40 bone metastasis samples. Immunohistochemical staining of tissue sections was performed by using a kit from Dako (Copenhagen). A primary antibody targeting ACBP (1:300, Sigma-Aldrich, HPA051428, RRID: AB_2681482) was used for immunostaining. Hematoxylin (Vector Laboratories) was used for nuclear counterstaining. The immunostaining intensity of each slide was evaluated by two pathologists using the histochemistry score (H-score) method. The H-score was calculated as Σpi(i + 1), where ‘I’ denotes the intensity score and ‘pi’ represents the percentage of cells exhibiting that particular intensity.

Micro-computed tomography (micro-CT)-based bone scanning

Mice were scanned on a Bruker micro-CT SkyScan 1276 system (Bruker) with a source voltage of 55 kV, a source current of 200 μA, a filter setting of Al 0.2 mm, and a pixel size of 13 μm at 2016 × 1344. We used 435 ms exposure time and the step and shoot mode with a rotation step of 0.400 degrees. The reconstruction of scanned images was performed by using Insta-Recon software (Bruker microCT). The parameters for reconstruction were windowing 0–0.08 intensity, ring artifact reduction of 5, beam hardening of 23%, and automatic post-alignment correction. The scanned region was analyzed by using CTAn software (version 1.18 8.0+, Bruker microCT), and the 3D model was created by using CTVox software (version 3.3.0, Bruker microCT) to visualize osteolytic areas.

RNA isolation and qPCR

Total RNA was extracted by using TRIzol reagent (Invitrogen, 15596026) or the PureLink RNA Mini Kit (Invitrogen, 12183018A). cDNA was synthesized from 1 μg of total RNA by using the iScript cDNA Synthesis Kit (Bio-Rad, 1708891). Real-time PCR and data collection were performed by using SYBR Green Supermix (Bio-Rad, 1725124) on a CFX96 instrument (Bio-Rad). Data were normalized to ACTB or GAPDH. Primer sequences are listed in table S5.

RNA-seq and data analysis

Total RNA was isolated by using the RNeasy Mini Kit (Qiagen, 74104) and was treated with DNase I (Qiagen, 79254). RNA-seq libraries were prepared from 3 μg of total RNA by using the TruSeq Stranded mRNA Library Prep Kit (Illumina, 20020594), according to the manufacturer’s instructions. The libraries were sequenced on an Illumina NextSeq 500 instrument at MD Anderson’s Advanced Technology Genomics Core. The RNA-seq reads were first trimmed by using Trim Galore (v0.6.5) (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), a wrapper around two tools: cutadapt v2.8 (https://github.com/marcelm/cutadapt/) and FastQC v0.11.5 (https://github.com/chgibb/FastQC0.11.5/; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and were then mapped to the human genome (GRCh38) by using hisat2 (v2.1.0) (82). The gene-level raw read counts were calculated by using featureCounts from the Subread package (83), based on the aligned and sorted bam files. Normalization of read counts and differentially expressed gene analysis were performed by using DESeq2 (1.22.2) (84). The threshold values of base mean > 1, log₂|(Fold-Change)| ≥ log₂(1.5), and FDR < 0.05 were used to define differentially expressed genes for downstream analysis. Gene Ontology analysis and KEGG pathway analysis were performed by using the R package clusterProfiler (85).

TNMplot and KM plotter analyses

To compare DBI mRNA expression in normal, tumor, and metastatic tissues from patients with breast cancer, we used the TNMplot tool (38) to analyze gene expression data from GEO, TCGA, TARGET, and GTEx repositories. Statistical significance was determined by the Kruskal-Wallis test followed by Dunn’s test. To evaluate the correlation between DBI mRNA or ACBP protein expression and clinical outcomes, we used the KM plotter tool to analyze published lung cancer and breast cancer data (39, 40). Statistical significance was determined by the log-rank test.

Seahorse FAO assay

H1299 or 4T1.2 cells were cultured in RPMI 1640 or DMEM with charcoal-stripped FBS. Before measurement, the medium was replaced with substrate-limited medium: DMEM with no glucose, glutamine, or phenol red (Gibco, A14430–01), which was supplemented with 0.5 mM D-(+)-Glucose (Sigma-Aldrich, G8769–100ML),1 mM GlutaMAX (Gibco, 35050061), 1% FBS, and 0.5 mM L-carnitine (Sigma-Aldrich, C0158–25G). H1299 or 4T1.2 cells were cultured for 12 or 8 hours and were then dissociated. 3,000 or 8,000 cells/well were seeded in the Seahorse XF96 cell culture plate coated with poly-L-lysine (EMD Millipore, A-005-C) to improve cell attachment, and cells were centrifuged at 100 × g at 37 °C for 15 minutes. Then, 45 minutes before measurement, the cell culture medium was replaced with FAO assay medium: XF RPMI medium (pH 7.4; Agilent Technologies, 103576–100) or XF DMEM medium (pH 7.4; Agilent Technologies, 103575–100), which was supplemented with 2.5 mM glucose and 0.5 mM L-carnitine. Immediately before the assay, 30 μL of BSA control or palmitate-BSA (Agilent Technologies, 102720–100) was added to each well. The Agilent Seahorse XF96 Analyzer was used to obtain measurements of oxygen consumption rates according to the manufacturer’s instructions, and drugs (etomoxir, Sigma-Aldrich, 236020, and components from the Seahorse XF Cell Mito Stress Test Kit, Agilent Technologies, 103015–100) were added during the experiment as indicated: etomoxir, 40 μΜ; oligomycin, 2 μΜ; FCCP, 3 μΜ (H1299 cells) or 2 μM (4T1.2 cells); and rotenone-antimycin, 1 μΜ.

BODIPY 493/503 staining

For measuring fatty acid consumption with BODIPY493/503 (ThermoFisher Scientific, D3922), H1299 (2 × 10⁵), BoM-1833 (4 × 10⁵), PC-3M (4 × 10⁵), MDA-MB-231 (4 × 10⁵), 4T1.2 (4 × 10⁵), or Hs578T (4 × 10⁵) cells were seeded in each well of 6-well plates and cultured for 24 hours in the presence of BSA control, 33.3 μM oleic acid-albumin (Sigma-Aldrich, O3008), 33.3 μM palmitic acid-albumin (Cayman Chemical, 29558), or 500 μM sodium butyrate (Sigma-Aldrich, TR-1008). Then, cells were washed with PBS and incubated with 2 μM BODIPY 493/503 staining solution (prepared in PBS) in the dark at 37 °C for 15 minutes, followed by flow cytometry analysis.

BODIPY C16 staining

For measuring fatty acid uptake with BODIPY-conjugated palmitate (BODIPY FL C16, ThermoFisher Scientific, D3821), H1299 (2 × 10⁵), BoM-1833 (4 × 10⁵), PC-3M (4 × 10⁵), MDA-MB-231 (4 × 10⁵), 4T1.2 (4 × 10⁵), or Hs578T (4 × 10⁵) cells were seeded in each well of 6-well plates, incubated for 12 hours, and serum-starved for 12 hours. Then, cells were washed and incubated with 1 μM BODIPY FL C16 staining solution (diluted in sterile RPMI-1640 or DMEM) for 5 minutes, followed by microscopy or flow cytometry analysis.

Lipid peroxidation measurement

Lipid peroxidation was measured by using BODIPY 581/591 C11 dye (Invitrogen, D3861) according to the manufacturer’s instructions. Briefly, H1299 (2 × 10⁵) or BoM-1833 (4 × 10⁵) cells were seeded in each well of 6-well plates, and treated with 0.3 μM RSL3 or 8 μM erastin in the presence or absence of 33.3 μM oleic acid-albumin. Staining was performed by incubating cells with 5 μM BODIPY 581/591 C11 dye (Invitrogen, D3861). After incubation at 37 °C for 30 minutes, cells were washed twice with PBS and trypsinized, followed by flow cytometry analysis.

GSH measurement

Relative GSH content was measured by using the GSH-Glo Glutathione Assay Kit (Promega, V6911) according to the manufacturer’s instructions. Briefly, H1299 (1 × 10³) or BoM-1833 (2 × 10³) cells were seeded in each well of 96-well plates one day before analysis. The following day, the medium was replaced with fresh basic medium supplemented with 1% FBS, 10% FBS, or 1% FBS + 33.3 μM oleic acid-albumin. After 9 hours, the culture medium was removed from the wells, and 100 μL of 1 × GSH-Glo Reagent was added to each well, followed by incubation at room temperature for 30 minutes. Next, 100 μL of the reconstituted luciferin detection reagent was added to each well and mixed briefly on a plate shaker. After 20 minutes, luminescence was measured by using a Gen5 Microplate Reader (Biotek).

NADP+ and NADPH measurement

The intracellular NADPH and total NADP (NADPH and NADP+) were measured as previously described (86, 87) with modifications. Briefly, cells were cultured in 6-well plates overnight. On the next day, the cells were lysed in 300 μL of extraction buffer (20 mM nicotinamide, 20 mM NaHCO₃, and 100 mM Na₂CO₃). After centrifugation, the supernatant was split into two 150 μL aliquots. For total NADP measurement, 20 μL of cell supernatant from one aliquot was added to a 96-well plate and mixed with 80 μL of NADP-cycling buffer (100 mM Tris-HCl, pH 8.0, 0.5 mM thiazolyl blue, 2 mM phenazine ethosulfate, and 5 mM EDTA) containing 0.75 U of G6PD enzyme (Sigma, G4134). After incubation in the dark at 30 °C for 1 minute, 20 μL of 10 mM glucose-6-phosphate was added to the mixture, and the change in absorbance at 570 nm was measured every minute at 30 °C for 6 minutes on a microplate reader. For NADPH measurement, the other 150 μL of supernatant was incubated at 60 °C for 30 minutes (to destroy NADP+ without affecting NADPH), followed by the same procedures as for total NADP measurement. The concentration of NADP⁺ was calculated by subtracting [NADPH] from [total NADP].

ROS measurement

One day before measurement, 2 × 10⁵ H1299 cells were seeded in one well of a 6-well plate. After 12 hours, the medium was replaced with fresh medium with or without 33.3 μM oleic acid-albumin. Cells were treated with 200 μM H₂O₂ in culture medium for 3 hours and were then incubated with 4 μM CM-H2DCFDA (ThermoFisher Scientific, C6827). After staining at 37 °C for 30 minutes, cells were washed twice with PBS and trypsinized, followed by flow cytometry analysis.

Statistical analysis

Statistical tests were performed with GraphPad Prism software (version 9.0.0), and biological replicates were used. Except for the animal and patient studies (one time) and RNA-seq analysis (one time), each experiment was repeated three times with similar results. Sample sizes are reported in the figure legends for each study. Statistical analysis of each plot is described in figure legends. Unless otherwise noted, data are presented as mean ± SEM. Normal distribution of the data was confirmed by the Shapiro-Wilk test. When comparing two groups of samples, we used Student’s t-test (two-tailed, unpaired). When comparing three or more groups, we used one-way or two-way ANOVA, followed by a multiple comparisons test recommended by GraphPad Prism software for each case (Tukey’s, Holm-Sidak’s, Sidak’s, or Dunnett’s post hoc test). No data points were excluded from the analysis. P < 0.05 was considered statistically significant. All individual-level data is available in data file S2. Uncropped blots are presented in data file S3.

Supplementary Material

List of Supplementary Materials

Materials and Methods

Figures S1 to S19

Tables S1 to S5

MDAR reproducibility checklist

Data files S1 to S3

Data and materials availability:

All data associated with this study are presented in the paper or the Supplementary Materials. The RNA-sequencing data have been deposited in NCBI’s Gene Expression Omnibus (GEO) with the accession number GSE250240.