Abstract

   Solute carrier family 27 member 5, a key enzyme in fatty acid transport
   and bile acid metabolism in the liver, is frequently expressed in low
   quantities in patients with hepatocellular carcinoma, resulting in poor
   prognosis. However, it is unclear whether SLC27A5 plays non‐canonical
   functions and regulates HCC progression. Here, an unexpected
   non‐canonical role of SLC27A5 is reported: regulating the alternative
   splicing of mRNA to inhibit the metastasis of HCC independently of its
   metabolic enzyme activity. Mechanistically, SLC27A5 interacts with
   IGF2BP3 to prevent its translocation into the nucleus, thereby
   inhibiting its binding to target mRNA and modulating PIP4K2A pre‐mRNA
   splicing. Loss of SLC27A5 results in elevated levels of the PIP4K2A‐S
   isoform, thus positively regulating phosphoinositide 3‐kinase signaling
   via enhanced p85 stability in HCC. SLC27A5 restoration by AAV‐Slc27a5
   or IGF2BP3 RNA decoy oligonucleotides exerts an inhibitory effect on
   HCC metastasis with reduced expression of the PIP4K2A‐S isoform.
   Therefore, PIP4K2A‐S may be a novel target for treating HCC with
   SLC27A5 deficiency.

   Keywords: alternative splicing; HCC; nonmetabolic functions; PIP4K2A‐S,
   SLC27A5
     __________________________________________________________________

   A mechanism is proposed that SLC27A5 plays a noncanonical role in
   inhibiting HCC progression. SLC27A5 inhibits the conversion from
   PIP4K2A‐L to PIP4K2A‐S via interaction with IGF2BP3, and suppresses the
   PI3K/AKT pathway. These findings highlight PIP4K2A‐S is a potential
   marker of diagnosis and prognosis for HCC. IGF2BP3 RNA decoy is a
   promising therapy for HCC patients with SLC27A5 deficiency.

   graphic file with name ADVS-11-2305374-g003.jpg

1. Introduction

   Metabolism can be intricately linked to multiple malignant biological
   processes through aberrantly regulated metabolic enzymes and their
   metabolic products during disease progression. Thus, metabolic
   reprogramming plays a key role in cancer metastasis.^[ [46]^1 ^] Recent
   studies indicate that some metabolic enzymes in tumor cells acquire
   non‐canonical functions and directly regulate gene expression via
   protein‐protein interactions, protein modifications, or protein kinase
   activity to promote cancer progression.^[ [47]^2 , [48]^3 , [49]^4 ,
   [50]^5 ^] For instance, UDP‐glucose 6‐dehydrogenase interacts with Hu
   antigen R and increases Snail mRNA stabilization to promote lung cancer
   metastasis.^[ [51]^6 ^] In addition, NAD(P)H quinone oxidoreductase‐1
   binds to SREBP1 and promotes the progression and metastasis of
   hepatocellular carcinoma (HCC).^[ [52]^7 ^] Therefore, understanding
   these noncanonical roles of metabolic enzymes in cancer can provide
   novel strategies for treating neoplasms.

   Solute carrier family 27 member 5 (SLC27A5, also known as FATP5), which
   is exclusively expressed in the liver, is involved in fatty acid
   transport and bile acid metabolism.^[ [53]^8 ^] SLC27A5 has been
   demonstrated to function as a novel tumor suppressor in HCC and is
   positively correlated with HCC prognosis.^[ [54]^9 , [55]^10 ^]
   Further, SLC27A5 deficiency can activate the NRF2/TXNRD1 pathway by
   increasing lipid peroxidation in HCC.^[ [56]^11 ^] Nevertheless,
   whether SLC27A5 plays non‐metabolic‐specific functions and regulates
   HCC progression remains unclear.

   Alternative splicing (AS) is a pivotal biological process for
   regulating gene expression and generating proteomic diversity.^[
   [57]^12 ^] Aberrant AS events provide potential molecular hallmarks for
   malignancies.^[ [58]^13 , [59]^14 ^] Accumulating evidence indicates
   that some RNA binding proteins (RBPs), which contain abnormal splicing
   factors (SFs), promote cancer progression via AS with the generation of
   tumor‐dependent variants.^[ [60]^15 , [61]^16 , [62]^17 , [63]^18 ,
   [64]^19 ^] Dysregulation of splicing variants in many types of cancer
   is strongly associated with tumor invasion, metastasis, and poor
   prognosis.^[ [65]^16 , [66]^20 ^] Therefore, clarifying the functional
   impact of AS variants, their regulators, and the associated signaling
   pathways is critical for evaluating the effects of aberrant splicing
   isoform expression involved in hepatocarcinogenesis and/or
   metastasis.^[ [67]^21 , [68]^22 ^]

   As a significant RBP, insulin like growth factor 2 mRNA binding protein
   3 (IGF2BP3) exhibits infrequent nuclear translocation and is involved
   in mRNA processing.^[ [69]^23 , [70]^24 , [71]^25 ^] Additionally, it
   contributes to cell proliferation and metastasis in various cancers.^[
   [72]^26 , [73]^27 , [74]^28 ^] Our study identified IGF2BP3 as a new
   potential SLC27A5‐interacting protein in hepatoma cells. We performed
   high‐throughput RNA sequencing combined with RIP sequencing and
   verified a SLC27A5/IGF2BP3‐modulated AS event of PIP4K2A pre‐mRNA
   involving in HCC metastasis. This variant of PIP4K2A‐S (lack of exon 5)
   may serve as a potential prognostic marker; further, targeting mRNA
   alternative splicing might be a promising strategy for HCC therapy.

2. Results

2.1. SLC27A5 Directly Binds to IGF2BP3 Proteins

   To detect potential SLC27A5‐interacting proteins in HCC, we performed
   immunoprecipitation mass spectrometry (IP‐MS) (Figure [75] 1A). The
   results indicated that 294 proteins, including 36 RNA‐binding proteins
   (RBPs) were likely to interact with SLC27A5. ClueGO‐enrichment analysis
   revealed that these potential interacting proteins primarily played a
   role in regulating RNA‐related biological processes (Figure [76]1B,C).
   Additionally, our findings indicated that SLC27A5 exerted inhibitory
   effects on HCC cell metastasis in vitro (Figure [77]S1A–E, Supporting
   Information). To identify metastasis‐associated RBPs that may interact
   with SLC27A5, we performed a combined analysis of the metastatic
   datasets of liver cancer ([78]GSE36376, [79]GSE112790), the RBP
   database (available at [80]http://attract.cnic.es) and
   SLC27A5‐interacting proteins. Venn diagram analysis indicated that
   SLC27A5 may play an important role in promoting HCC metastasis through
   interaction with IGF2BP2 and/or IGF2BP3 (Figure [81]1D). Subsequently,
   co‐immunoprecipitation (Co‐IP) confirmed the interaction between
   SLC27A5 and IGF2BP3 (Figure [82]1E,F), whereas the interaction between
   SLC27A5 and IGF2BP2 was relatively weak (data not shown). Meanwhile,
   SLC27A5 and IGF2BP3 interactions were confirmed in vitro using the
   purified proteins (Figure [83]1G). Confocal analysis also indicated
   that SLC27A5 and IGF2BP3 were colocalized in the cytoplasm of PLC/PRF/5
   and HepG2 cells (Figure [84]1H; Figure [85]S2A, Supporting
   Information).

Figure 1.

   Figure 1
   [86]Open in a new tab

   SLC27A5 binds directly to IGF2BP3 in hepatocellular carcinoma. A)
   Schematic of IP‐MS analysis for the SLC27A5‐interaction proteins. B)
   Based on ClueGO‐enrichment analysis with a two‐sided hypergeometric
   test, the potential SLC27A5‐interaction proteins were significantly
   enriched in RNA splicing and mRNA processing. The p value was adjusted
   using the Bonferroni method and p <0.05 was considered to indicate
   significant enrichment. C) Potential SLC27A5‐ interaction proteins
   obtained from IP‐MS; the most relevant proteins fall in the first
   quadrant, proteins with coverage (%) ≥20 and peptide count ≥10 were
   marked. D) SLC27A5‐interaction proteins discovered by IP‐MS,
   RNA‐binding proteins (RBPs) obtained from the ATtRACT database, and
   HCC‐associated metastasis genes acquired from the GSE database
   ([87]GSE36376, [88]GSE112790) are illustrated by Venn diagram. E) IP
   and WB assays of SLC27A5 with IGF2BP3 in HepG2 cells. F) Co‐IP assay
   was performed with anti‐Myc (left) or anti‐Flag (right) in PLC/PRF/5
   cells transfected with Flag‐SLC27A5 and Myc‐IGF2BP3. Immunoblotting
   with anti‐Flag or anti‐Myc was performed to analyze the
   immunoprecipitates. lgG served as a control. G) GST pull‐down assays
   showing interaction between the GST‐tagged SLC27A5 and 6xHis‐IGF2BP3.
   GST was used as a negative control. The left panel is the western blot
   showing the levels of GST‐tagged SLC27A5 and 6xHis‐IGF2BP3. The right
   panel shows Coomassie blue staining of GST protein and GST‐tagged
   SLC27A5 used for the experiment. H) Subcellular colocalization of
   SLC27A5 and IGF2BP3 in PLC/PRF/5 cells was determined by
   immunofluorescence staining. (Scale bar: 20 µm). I,J) Images of the
   full‐length forms and mutant variants of IGF2BP3 and SLC27A5. Truncated
   mutants of IGF2BP3, comprising amino acids (aa) 1–194, 195–344, or
   345–579, were designated as RRM1‐2, KH1‐2, and KH3‐4, respectively (I).
   Truncated mutants of SLC27A5, comprising amino acids (aa) 1–77 or
   78–690, were designated as ΔC and ΔN, respectively(J). Co‐IP assays of
   Interactions between SLC27A5 and the full‐length IGF2BP3, RRM1‐2,
   KH1‐2, or KH3‐4 in PLC/PRF/5 cells (I). Co‐IP assays of interactions
   between IGF2BP3 and the full‐length SLC27A5, ΔN, or ΔC in PLC/PRF/5
   cells (J).

   To verify whether the interaction between SLC27A5 and IGF2BP3 depends
   on its enzyme activity, we constructed an inactivated mutant of SLC27A5
   (S296A),^[ [89]^29 , [90]^30 ^] and IP analysis confirmed that their
   interaction was not affected by enzyme activity (Figure [91]S2B,
   Supporting Information). To further map the interaction sites of
   SLC27A5 with IGF2BP3, we generated three truncated forms of IGF2BP3
   (Figure [92]1I) and found that the C‐terminal region of IGF2BP3
   (195–579aa), particularly in the KH3‐4 domain (345‐579aa), showed a
   strong interaction with SLC27A5. Moreover, the IGF2BP3‐interacting
   domain was retained in △N of SLC27A5 (78‐ 690aa) (Figure [93]1J).

   To further investigate the impact of the SLC27A5 enzyme inactivating
   mutant (S296A) on its biological function, we evaluated the
   proliferation and metastasis ability of HCC cells expressing the
   wild‐type (WT) or S296A mutant SLC27A5. The results showed that,
   compared to the WT SLC7A5, the S296A mutant exhibited weaker ability to
   suppress cell proliferation and migration in HCC cells
   (Figures [94]S3A–E, Supporting Information). Taken together, these
   results indicate that SLC27A5 can suppress liver cancer progression by
   interacting with IGF2BP3, independent of its enzyme activity.

2.2. SLC27A5 Interacts with IGF2BP3 to Regulate PIP4K2A Pre‐mRNA Splicing

   As a prominent RBP, upon translocated to the nucleus and occupying the
   relevant binding sites, IGF2BP3 plays a crucial role in regulating the
   alternative splicing, and stability of target mRNA.^[ [95]^24 , [96]^25
   , [97]^27 , [98]^31 ^] We identified that SLC27A5 interacted with
   IGF2BP3 in cytoplasm, whereas depletion of SLC27A5 promoted the nuclear
   translocation of IGF2BP3 (Figure [99]S4A,B, Supporting Information). To
   further explore the role of SLC27A5 deletion in liver cancer
   metastasis, we performed RNA sequencing (RNA‐seq) on invasive liver
   cancer tissues in WT mice and Slc27a5^−/‐ mice. The results revealed
   that there were only 65 differentially expressed genes but 722
   differential alternative splicing (AS) events occurring in Slc27a5^−/‐
   mice (Figure [100] 2A,B ; Figure [101]S5A–C, Supporting Information).
   Therefore, we postulated that SLC27A5 may modulate mRNA splicing events
   by interacting with IGF2BP3 during HCC progression.

Figure 2.

   Figure 2
   [102]Open in a new tab

   SLC27A5 interacts with IGF2BP3 to regulate PIP4K2A splicing. A)
   Workflow of RNA‐seq and subsequent mRNA splicing analysis for
   WT/Slc27a5^−/− mouse liver tumors. B) Pre‐mRNA alternative splicing
   events regulated by SLC27A5 were analyzed by RNA splicing analysis.
   A3SS, alternative 3′ splicing site; SE, skipping exon; RI, retention
   intron; MXE, mutually exclusive exon; A5SS, alternative 5′ splicing
   site. C) Schematic showing that the IGF2BP3‐recognized mRNA discovered
   by RIP‐seq was analyzed by mRNA splicing analysis in SLC27A5‐knockout
   (SLC27A5‐KO) cells. D) All pre‐mRNA splicing events identified in RNA
   splicing analysis. E) IGF2BP3‐recognized mRNA splicing skipping events
   identified by RIP‐seq in SLC27A5‐KO cells. F) Splicing variants of
   PIP4K2A differentially display exon skipping events in the parental
   (Blue) and SLC27A5‐KO (Red) cells. Sashimi plots of the exon 5 junction
   from the aligned RNA‐seq data; the supported junction reads were marked
   between exons. G) IF and RNA FISH assays of IGF2BP3 and PIP4K2A mRNA in
   parental or SLC27A5‐KO SNU449 cells. IF/FISH revealed co‐localization
   of PIP4K2A mRNA (red), IGF2BP3 protein (green), and DAPI (blue)‐stained
   nuclei (Scale bars: 20 µm). H) RNA pull‐down assays using
   biotin‐labeled PIP4K2A pre‐mRNA. RNAs with random sequences were
   included as the negative control. Precipitated IGF2BP3 proteins were
   examined by immunoblotting. I) RIP assay of PIP4K2A mRNA recognized by
   IGF2BP3 in parental or SLC27A5‐KO SNU449 cells. J) PSI of PIP4K2A were
   analyzed using agarose gel electrophoresis of PCR products in
   SLC27A5‐KO HepG2 cells and SNU449 cells transfected with PIP4K2A‐S
   shRNA or negative control shRNA. Gel densitometry was analyzed using
   Image J to calculate the percentage‐splice‐in (PSI =
   splice_in/(splice_in+splice_out)) (n = 3, ^* p <0.05; ^** p <0.01). (K)
   PIP4K2A variants in ten pairs of tissues from patients with HCC. Gel
   densitometry was analyzed using Image J to calculate the PSI; the
   relative PSI of the variants in each sample pair was quantified (n =
   10, two‐tailed, paired t‐test used for analysis) and the average PSI in
   ten pairs is shown (n = 10, blue column indicates matched normal
   tissues [N]) and red column indicates tumor [T].

   Next, we performed RNA immunoprecipitation followed by sequencing
   (RIP‐seq) and confirmed that SLC27A5 regulates AS via IGF2BP3
   (Figure [103]2C,D). Among the key AS genes, we noted that the skipping
   exon was the most common event, accounting for 71% (Figure [104]2D).
   Therefore, we combined the non‐differentially expressed genes,
   differential transcripts, and AS events of skipping exons from RIP‐seq
   data for analysis. We found that 15 mRNAs may serve as splicing targets
   for the SLC27A5/IGF2BP3 complex (Figure [105]2E). Additionally, KEGG
   analysis revealed that the binding mRNAs of the SLC27A5/IGF2BP3 complex
   were mainly enriched in the phosphatidylinositol signaling pathway
   (Figure [106]S5D, Supporting Information). This pathway is involved in
   almost all aspects of cell death and fate by regulating the activity of
   phosphoinositide kinases, including phosphoinositide 3‐kinases (PI3Ks)
   and phosphatidylinositol phosphate kinases (PIPKs).^[ [107]^32 ,
   [108]^33 , [109]^34 ^] Among these 15 mRNAs, only PIP4K2A was involved
   in regulating the phosphatidylinositol signaling pathway. Therefore, we
   focused on PIP4K2A, a potential target for treating cancer and
   inflammation, as it catalyzes the production of PI(4.5)P2 and activates
   the PI3K signaling pathway.^[ [110]^35 , [111]^36 ^] Furthermore,
   RNA‐seq splicing analysis of SLC27A5‐KO HepG2 cells was performed, and
   the Sashimi plot indicated that the exon skipping occurred at the exon
   5 junctions of PIP4K2A. Meanwhile, SLC27A5‐KO obviously decreased the
   exon 5 reads but elevated the reads across exons 4 and 6
   (Figure [112]2F), indicating that SLC27A5 deletion was necessary for
   exon 5 skipping of PIP4K2A.

   To determine whether SLC27A5 regulates PIP4K2A splicing via interaction
   with IGF2BP3, RNA‐fluorescence in situ hybridization (RNA FISH) along
   with a protein immunofluorescence (IF) assay were performed. We found
   that PIP4K2A mRNA was strongly co‐localized with IGF2BP3 protein in the
   nucleus of SLC27A5‐KO cells (Figure [113]2G; Figure [114]S5E,
   Supporting Information). Conversely, binding between PIP4K2A mRNA and
   IGF2BP3 was significantly weakened in parental cells (Figure [115]2G;
   Figure [116]S5E, Supporting Information). In addition, the interaction
   between PIP4K2A mRNA and IGF2BP3 was also validated by RNA pulldown
   assay (Figure [117]2H). Moreover, RIP‐qPCR confirmed that SLC27A5
   deficiency enhanced the binding of IGF2BP3 to PIP4K2A mRNA
   (Figure [118]2I). The results suggested that SLC27A5 deletion may
   promote IGF2BP3 translocation into the nucleus and its binding with
   PIP4K2A mRNA.

   To investigate the impact of the SLC27A5/IG2BP3 complex on PIP4K2A
   pre‐mRNA alternative splicing, we utilized RT‐qPCR and western blot
   assays to assess the expression levels of both PIP4K2A‐L (including
   exon 5) and PIP4K2A‐S (lacking exon 5). PIP4K2A‐L was upregulated in
   MHCC‐97H and PLC/PRF/5 cells overexpressing SLC27A5 (Figure [119]S6A,
   Supporting Information), whereas PIP4K2A‐S was the predominant isoform
   in SLC27A5‐KO HepG2 and SNU449 cells and in the liver tumor tissues of
   Slc27a5^−/− mice (Figure [120]2J; Figure [121]S6B–D, Supporting
   Information). Furthermore, knockdown of IGF2BP3 partially decreased
   PIP4K2A‐S expression in SLC27A5‐KO cells (Figure [122]2J;
   Figure [123]S6B,E, Supporting Information). Therefore, these data
   demonstrated that SLC27A5 regulated the splicing of PIP4K2A pre‐mRNA
   via interaction with IGF2BP3.

   Finally, to investigate whether regulation of PIP4K2A splicing by
   SLC27A5 was dependent on its enzymatic activity, we transfected the
   S296A mutant plasmid of SLC27A5 into MHCC‐97H or PLC/PRF/5 cells and
   examined the splicing of PIP4K2A pre‐mRNA. The results showed that
   S296A mutant exhibited the ability of regulating PIP4K2A splicing
   similar to the wild type of SLC27A5 (Figure [124]S6F,G, Supporting
   Information). In conclusion, regulation of SLC27A5 on PIP4K2A pre‐mRNA
   alternative splicing through IGF2BP3‐interaction was independent of its
   enzymatic activity.

2.3. PIP4K2A‐S Promotes HCC Progression In Vitro and In Vivo

   To investigate the biological functions of PIP4K2A‐L and PIP4K2A‐S,
   PIP4K2A mRNA splicing was validated in ten pairs of adjacent and HCC
   tissues (Figure [125]2K; Figure [126]S6H, Supporting Information). Our
   data indicated that PIP4K2A‐S was enriched in HCC tissues, suggesting
   that PIP4K2A‐S expression may functionally contribute to HCC
   progression. We then examined the expression levels of the two PIP4K2A
   isoforms (PIP4K2A‐L and PIP4K2A‐S) in HCC cell lines (Figure [127]S7A,
   Supporting Information). The results indicated that PIP4K2A‐S was
   highly expressed in SNU449 and HepG2 cells. Since the endogenous
   PIP4K2A‐L and PIP4K2A‐S expression levels were low in PLC/PRF/5 and
   MHCC‐97H cells, we overexpressed PIP4K2A‐L and PIP4K2A‐S in these cells
   (Figure [128]S7B, Supporting Information). The results showed that
   PIP4K2A‐S, but not PIP4K2A‐L, promoted cell proliferation, colony
   formation, migration, and invasion (Figure [129] 3A–D;
   Figure [130]S7C–F, Supporting Information). Importantly, PIP4K2A‐S also
   significantly enhanced hepatocarcinogenesis (Figure [131]3E;
   Figure [132]S7G, Supporting Information) and lung metastasis
   (Figure [133]3F; Figure [134]S7H, Supporting Information) in the
   orthotropic HCC mouse model. Notably, knockdown of PIP4K2A‐S markedly
   restored cell growth and cell metastasis in SLC27A5‐KO HepG2 and SNU449
   cells (Figure [135]3G–J; Figure [136]S8A–H, Supporting Information).
   Moreover, PIP4K2A‐S knockdown significantly inhibited
   hepatocarcinogenesis (Figure [137]3K; Figure [138]S8I, Supporting
   Information) and lung metastasis (Figure [139]3L; Figure [140]S8J,
   Supporting Information) in nude mice bearing SLC27A5‐KO tumor
   xenografts. Overall, the aforementioned results suggest that PIP4K2A‐S
   may serve as an oncogenic driver in HCC progression.

Figure 3.

   Figure 3
   [141]Open in a new tab

   PIP4K2A‐S significantly promotes the proliferation and migration of
   hepatoma cells. A–D) Cell Counting Kit‐8 (CCK8) (A), colony formation
   (B), transwell migration (C), and invasion assays (D) for PLC/PRF/5 and
   MHCC‐97H cells infected with AdPIP4K2A‐L, AdPIP4K2A‐S, or AdGFP. Data
   are expressed as the mean ± SD for three independent experiments. ^*
   p <0.05; ^** p <0.01.E,F) To establish a lung metastasis model of nude
   mice, MHCC‐97H cells with PIP4K2A‐L or PIP4K2A‐S overexpression were
   orthotopically injected into their left liver lobe. After 8 weeks
   post‐injection, the livers and lungs were collected. Graphs (mean ± SD)
   showing tumor size, tumor numbers, liver weight (E), and metastatic
   nodules numbers (F). n = 6 (^* p <0.05; ^** p <0.01; ^*** p <0.001).
   G–J), CCK8 (G), colony formation (H), migration (I), and invasion
   assays (J) for SLC27A5‐KO HepG2 cells and SNU449 cells transfected with
   PIP4K2A‐S shRNA or negative control shRNA. Data are expressed as the
   mean ± SD for three independent experiments. ^* p <0.05; ^** p <0.01.
   K,L) SLC27A5‐KO or parental SNU449 cells with concomitant knockdown of
   PIP4K2A‐S were injected into nude mice via the tail vein. Graphs (mean
   ± SD) showing tumor number, liver weight (K). and number of metastatic
   nodules (L). n = 6 (^** p <0.01; ^*** p <0.001).

2.4. PIP4K2A‐S Stabilizes p85 Expression and Activates PI3K‐AKT Signaling

   A previous study has shown that PIP4K2A binds to P85 and promotes its
   ubiquitination in glioblastoma cells to inhibit the PI3K/AKT signaling
   pathway.^[ [142]^36 ^] In our results, we also discovered that the
   SLC27A5/IGF2BP3 complex was involved in regulating the PI3K/AKT
   signaling pathway (Figure [143] 4A). To verify the functional role of
   the SLC27A5/IGF2BP3 complex in regulating the PI3K/AKT signaling
   pathway through modulation of PIP4K2A pre‐mRNA alternative splicing, we
   first examined the expression of PI3K target genes in SLC27A5‐KO cells.
   The results showed that p85, p110α, p110β, and phosphorylated AKT
   (p‐AKT) were significantly unregulated when SLC27A5 was depleted
   (Figure [144]4B; Figure [145]S9A, Supporting Information). Moreover,
   SLC27A5‐KO resutled in the conversion of PIP4K2A‐L to PIP4K2A‐S
   (Figure [146]4B). In contrast, SLC27A5 increased the expression of
   PIP4K2A‐L and downregulated p85, p110α, p110β, and phosphorylated AKT
   (p‐AKT) (Figure [147]4C). Interestingly, in MHCC‐97H and PLC/PRF/5
   cells overexpressing PIP4K2A‐L, p85 and p‐AKT were downregulated,
   whereas overexpression of PIP4K2A‐S showed the opposite effects
   (Figure [148]4D,E). Moreover, we found that PIP4K2A‐L, but not
   PIP4K2A‐S, promoted ubiquitination‐mediated proteasomal degradation of
   p85 (Figure [149]4F; Figure [150]S9B, Supporting Information). Notably,
   PIP4K2A‐L specifically bound to p85, whereas PIP4K2A‐S lacking exon 5
   could not bind to p85 (Figure [151]4G; Figure [152]S9C, Supporting
   Information). Further, colocalization of PIP4K2A‐L and p85 in the
   cytoplasm was evidenced by IF and western blot, whereas PIP4K2A‐S
   showed predominant expression in the nucleus (Figure [153]4H;
   Figure [154]S9D, Supporting Information). These results indicated that
   the binding between p85 and PIP4K2A depended on exon 5. Taken together,
   SLC27A5 interacted with IGF2BP3 and blocked its binding to PIP4K2A
   pre‐mRNA, thereby inhibiting the alternative splicing of exon 5 in
   PIP4K2A and increasing PIP4K2A‐L expression. The up‐regulated PIP4K2A‐L
   bound to p85 and promote its ubiquitination, thus inhibiting activation
   of the PI3K/AKT signaling pathway (Figure [155]4I).

Figure 4.

   Figure 4
   [156]Open in a new tab

   PIP4K2A‐S positively regulates PI3K signaling by stabilizing p85. A)
   KEGG pathway enrichment analysis of IGF2BP3‐binding genes in SLC27A5‐KO
   cells. B,C) Western blot analysis of the PI3K complex (p85, p110α, and
   p110β), phosphorylated‐AKT (p‐AKT), AKT, and PIP4K2A in SLC27A5‐KO (B)
   and SLC27A5‐OE (C) cells. β‐actin was used as a loading control. D)
   Western blot analysis of p85, p‐AKT, AKT, and PIP4K2A in hepatoma cells
   infected with AdGFP, AdPIP4K2A‐L, or AdPIP4K2A‐S. β‐actin was used as a
   loading control. E) Immunofluorescence staining of p‐AKT in hepatoma
   cells transduced with a control vector or PIP4K2A‐L‐ or
   PIP4K2A‐S‐expressing vector (scale bar: 20 µm). F) Effects of PIP4K2A‐L
   or PIP4K2A‐S on the ubiquitination of p85. PLC/PRF/5 cells were
   transduced with the indicated plasmids (5 µg each). Cells were lysed
   and the lysates were immunoprecipitated with anti‐HA. Co‐IP and
   immunoblot analysis were performed with the indicated antibodies. G)
   Co‐IP assay was performed with anti‐HA (left) or anti‐Flag (right) in
   PLC/PRF/5 cells transfected with Flag‐PIP4K2A‐L, Flag‐PIP4K2A‐S, and
   HA‐P85. lgG served as a control. H) Immunoblot analysis of nuclear and
   cytoplasmic fractions in PLC/PRF/5 cells transfected with
   PIP4K2A‐L‐Flag, PIP4K2A‐S‐Flag showing the levels of PIP4K2A‐L/S.
   LaminB1 and β‐tubulin served as the nuclear and cytoplasmic markers,
   respectively. I) Model of degradation of the p85 regulatory subunit in
   PI3K by PIP4K2A‐L, but not by PIP4K2A‐S.

2.5. AAV‐Mediated Restoration of SLC27A5 Alleviated HCC Metastasis in
Slc27a5‐KO Mice by Regulating PIP4K2A Splicing

   In order to observe the role of SLC27A5 in liver cancer metastasis, we
   restored the expression of SLC27A5 in invasive liver cancer via tail
   vein injection of AAV8‐TBG control and AAV8‐TBG‐Slc27a5 (n = 6) in WT
   or Slc27a5^−/‐ mice (Figure [157] 5A). AAV‐mediated restoration of
   hepatic SLC27A5 decreased the liver weight and tumor numbers in
   Slc27a5^−/‐ mice (Figure [158]5B). Moreover, the number of lung
   metastasis in SLC27A5‐treated mice was significantly lower than that in
   Slc27a5^−/‐ mice (Figure [159]5C). Consistent with our in vitro data,
   depletion of SLC27A5 induced partial IGF2BP3 translocation to the
   nucleus and caused an isoform switch of PIP4K2A‐L to PIP4K2A‐S in the
   liver tumors, whereas restoration of SLC27A5 in Slc27a5^−/‐ mice
   increased the PIP4K2A‐L expression but repressed PIP4K2A‐S expression
   (Figure [160]5D–G). Importantly, activation of PI3K/AKT signaling and
   enhancement of HCC migration were observed in Slc27a5 ^−/− mice, which
   were reversed by restoring SLC27A5 (Figure [161]5F,G; Figure [162]S9E,
   Supporting Information). These results indicated that SLC27A5
   alleviated HCC metastasis in Slc27a5‐KO mice by regulating PIP4K2A
   splicing.

Figure 5.

   Figure 5
   [163]Open in a new tab

   Restoration of SLC27A5 in Slc27a5^−/− mice blunts HCC metastasis. A)
   Schematic diagram of the experimental procedures in WT and Slc27a5^−/−
   mice. Mice were injected intraperitoneally with 75 mg kg^−1 DEN and
   received phenobarbital through diet at a concentration of 0.05% for 20
   weeks, followed by injection with AAV‐control or AAV‐Slc27a5. The liver
   and lung tissues were collected at 40 weeks post‐DEN/PB treatment. B,C)
   Gross appearances and H&E staining (scale bar: 20 µm) of samples from
   the livers (B) and lungs (C) with tumors (tumor nodules, n = 6/group).
   Representative images of tumor‐bearing livers are provided, and the
   tumor number and liver weight were calculated. Data indicate the mean ±
   SD. ^** p <0.01, one‐way analysis of variance followed by Tukey's test.
   D,E) qPCR assays of the indicated PIP4K2A‐L or PIP4K2A‐S mRNA
   expression in liver tumors. Gel densitometry analysis using Image J to
   calculate PSI (D). (n = 3, data are expressed as the mean ± SD,
   two‐tailed unpaired t test). Data are presented as the mean ± SD (E).
   (n = 3) ^* p <0.05. F,G) The indicated proteins in liver tumors were
   assessed by IHC labeling (scale bar: 20 µm) (F) and western blot (G).

2.6. SLC27A5 Deficiency with PIP4K2A‐S Upregulation Promotes HCC Metastasis

   Next, we determined the expression of SLC27A5, PIP4K2A‐L, PIP4K2A‐S,
   and p‐AKT in 40 pairs of HCC and adjacent tissues (Table [164]S1,
   Supporting Information). Compared to adjacent tissues, both SLC27A5
   expression and the PIP4K2A‐L/S ratio were downregulated in HCC tissues,
   whereas p‐AKT expression was significantly upregulated
   (Figure [165]S10A,B, Supporting Information). Notably, a positive
   correlation was observed between SLC27A5 expression and the ratio of
   PIP4K2A‐L/S (Figure [166] 6A), whereas SLC27A5 expression
   (Figure [167]6B) and the ratio of PIP4K2A‐L/S (Figure [168]6C) was
   negatively correlated with p‐AKT expression. Furthermore, patients with
   high PIP4K2A‐S expression levels displayed poorer overall survival
   based on data from TCGA Splice‐Seq dataset (Figure [169]6D). Univariate
   and multivariate Cox regression analysis revealed that SLC27A5
   deficiency and high PIP4K2A‐S expression are independent risk factors
   for death (Figure [170]6E).

Figure 6.

   Figure 6
   [171]Open in a new tab

   The variant PIP4K2A‐S is enriched in HCC and associated with poor
   prognosis. A‐C) Correlation analyses of the PIP4K2A‐L/S ratio and
   SLC27A5 mRNA expression (A), SLC27A5 and p‐AKT expression (B),
   PIP4K2A‐L/S ratio and p‐AKT expression (C) in tumor tissues from 40
   patients with HCC. Data are presented as the mean ± SEM. D)
   Kaplan–Meier survival analysis of overall survival of PIP4K2A‐L and
   PIP4K2A‐S in HCC patients from TCGA Splice Seq – liver hepatocellular
   carcinoma dataset (PIP4K2A‐L n = 324, PIP4K2A‐S, n = 43). Overall
   survival of PIP4K2A‐L and PIP4K2A‐S were calculated by Kaplan‐Meier
   method, and statistical significance was measured by log‐rank test. E)
   The univariate and multivariate Cox regression model of the clinical
   characteristics and SLC27A5‐mRNA expression levels. Univariate and
   multivariate Cox proportional hazards regression models were used to
   estimate the independent factors for overall survival. F)
   Representative IHC staining of SLC27A5 and PIP4K2A‐S in an HCC tissue
   microarray. Scores (0–3) were calculated according to the percentage of
   stained cells and intensity. (Scale bar: 20 µm). G,H) Staining scores
   of SLC27A5 and PIP4K2A‐S in tumor tissues from patients with HCC, with
   (n = 41) or without metastatic recurrence (n = 26). The percentage of
   non‐metastatic or metastatic recurrence of HCC (n = 67) was stratified
   by SLC27A5 and PIP4K2A‐S expression using the “survminer” R package
   (version 3.6.3). I) Kaplan–Meier analysis of overall survival of
   PIP4K2A‐S in the tissue microarray cohort (n = 67), HCC patients were
   grouped according to the PIP4K2A‐S expression level in tumor. J)
   Kaplan–Meier analysis of overall survival rate of SLC27A5 and PIP4K2A‐S
   in an HCC tissue microarray cohort (n = 67), stratified by SLC27A5 and
   PIP4K2A‐S expression.

   Moreover, we surveyed the expression levels of SLC27A5 and PIP4K2A‐S
   using a cohort of samples from patients with HCC (Table [172]S2,
   Supporting Information) by performing immunohistochemistry (IHC)
   assays. As shown in Figure [173]6F, the patients were stratified by
   high versus low SLC27A5 and/or PIP4K2A‐S levels. SLC27A5 expression was
   negatively correlated with PIP4K2A‐S (Figure [174]S10C, Supporting
   Information). Notably, low SLC27A5, high PIP4K2A‐S, and low SLC27A5
   levels coupled with high PIP4K2A‐S levels were strongly associated with
   HCC metastasis (Figure [175]6G,H) and negatively correlated with
   survival duration (Figure [176]6I; Figure [177]S10D,E, Supporting
   Information). Furthermore, patients with low SLC27A5 and high PIP4K2A‐S
   expression levels had the poorest survival (Figure [178]6J). Therefore,
   PIP4K2A‐S may represent an independent factor for poor prognosis and
   serve as a promising target for treating HCC.

2.7. Decoy Oligonucleotides Specifically Binding to IGF2BP3 Inhibit HCC
Progression In Vivo

   Previous studies defined the RNA binding consensus of IGF2BP3 as
   GGC/CA‐ through in vitro SELEX and in vivo iCLIP‐seq approaches.^[
   [179]^37 ^] Based on these studies, we designed RNA decoy
   oligonucleotides,^[ [180]^38 ^] containing the GGC/CA‐ sequence, to
   block the RNA binding activity of IGF2BP3 by interacting with the RRM
   domains (Figure [181] 7A). Transfection of the IGF2BP3 decoy
   (IGF2BP3‐D) oligonucleotides into HCC cell lines partially blocked cell
   proliferation and migration (Figure [182]S11A,B, Supporting
   Information). Moreover, PIP4K2A splicing was inhibited by transfection
   of the IGF2BP3‐D oligonucleotide (Figure [183]7B,C; Figures [184]S11C
   and [185]S12A, Supporting Information). Notably, splicing of PIP4K2A
   lasted for at least 96 h, indicating the enduring biological effect of
   the decoy oligonucleotides (Figure [186]S11C, Supporting Information).
   Furthermore, in SLC27A5‐KO cells, treatment with IGF2BP3‐D
   oligonucleotides induced an isoform switch from PIP4K2A‐S to PIP4K2A‐L,
   suppressed the expression of p85 and p‐AKT, and inhibited cell
   migration and wound healing compared to that in the scrambled group
   (Figure [187]7B–J; Figure [188]S12A–E, Supporting Information).
   Importantly, IGF2BP3‐D oligonucleotides reduced lung metastasis and
   inhibited the liver tumor growth induced by SLC27A5 depletion
   (Figure [189]7K; Figure [190]S12F, Supporting Information). Moreover,
   IHC analysis indicated that SLC27A5‐KO enhanced the expression of p85
   and p‐AKT but repressed PIP4K2A‐L, whereas the IGF2BP3‐D
   oligonucleotide significantly increased PIP4K2A‐L expression in vivo
   (Figure [191]7L). Taken together, these data indicate that RNA decoy
   oligonucleotides, specifically bound to IGF2BP3, reduce the
   cancer‐related splicing events of PIP4K2A and inhibit tumor growth and
   metastasis (Figure [192]7J).

Figure 7.

   Figure 7
   [193]Open in a new tab

   IGF2BP3 decoy oligonucleotides block SLC27A5‐induced PIP4K2A splicing.
   A) Schematic representing the inhibition of RNA binding proteins by
   decoy oligonucleotides. B) The PSI of PIP4K2A was analyzed using
   agarose gel electrophoresis of PCR products in SLC27A5‐KO SNU449 cells
   transfected with scrambled or IGF2BP3 decoy oligonucleotides. PSI was
   calculated by gel densitometry using Image J (^* p <0.05; ^** p <0.01).
   C – E) SLC27A5‐KO cells were treated as described in (B). Cells were
   subjected to western blot analysis (C) or IF staining (D and E) with
   the indicated proteins (scale bar: 20 µm). F – I) Effects of the
   scrambled or IGF2BP3 decoy oligonucleotides on cell migration and wound
   healing in SLC27A5‐KO cells. All data were expressed as mean ± standard
   deviation (x ± s), and analyzed using one‐way ANOVA for comparing
   between groups (n = 3). ^* p <0.05; ^** p <0.01. J) Schematic showing
   the IGF2BP3 RNA decoy oligonucleotides inhibiting tumor growth and
   metastasis. K) SLC27A5‐KO SNU449 cells transfected with scrambled or
   IGF2BP3 decoy oligonucleotides (scrambled, n = 6; IGF2BP3‐D, n = 6)
   were orthotopically injected into the left liver lobes of nude mice.
   After 8 weeks post‐injection, the liver and lung tissues were obtained
   for H&E staining. The metastatic nodules were counted (^* p <0.05; ^**
   p <0.01). L) The indicated proteins in liver tumors were evaluated by
   IHC labeling (scale bar: 20 µm).

3. Discussion

   SLC27A5 deficiency plays a critical role in increasing polyunsaturated
   lipids and promoting cancer progression.^[ [194]^9 , [195]^11 ,
   [196]^39 ^] Although a few previous studies indicated that the
   potential role of SLC27A5 in the transcriptional regulation of gene
   expression does not depend on its enzyme activity,^[ [197]^10 ^] most
   studies on SLC27A5‐mediated suppression of HCC progression have focused
   on its enzymatic role in lipid or bile acid metabolism.^[ [198]^11 ,
   [199]^39 , [200]^40 ^] In this study, we found that SLC27A5 directly
   interacted with IGF2BP3 to inhibit the interaction of PIP4K2A mRNA with
   IGF2BP3, leading to decreased expression of the PIP4K2A‐S variant to
   inhibit HCC progression (Figure [201] 8 ).

Figure 8.

   Figure 8
   [202]Open in a new tab

   Mechanistic model: SLC27A5 plays a noncanonical role in HCC. Normally,
   SLC27A5 interacted with IGF2BP3. In HCC, SLC27A5 depletion induced
   IGF2BP3 translocation into nuclear and generated PIP4K2A‐S isoform with
   exon 5 skipping, which promoted cell metastasis via enhancing p85
   protein stability to activate PI3K/AKT signaling pathway. PIP4K2A‐S may
   represent an independent factor for poor prognosis and serve as a
   promising target for the treatment of HCC.

   IGF2BPs are primarily cytoplasmic proteins; however, upon translocating
   to the nucleus and occupying relevant binding sites, they play a
   crucial role in the post‐transcriptional regulation of target mRNA.^[
   [203]^41 , [204]^42 ^] Alternative splicing is a crucial process in the
   post‐transcriptional processing of pre‐mRNA. Previous studies show that
   IGF2BP3 may also play a role in regulating pre‐mRNA alternative
   splicing.^[ [205]^25 ^] The eCLIP analysis revealed that IGF2BP3 not
   only interacted with the 3′UTR of target transcripts but also bound to
   the intronic regions and splice sites of target transcripts. In
   leukemia, IGF2BP3 directly binds to target mRNA and is involved in
   regulating alternative splicing events.^[ [206]^25 ^] Additionally,
   IGF2BP3 directly binds to PKM mRNA and regulates its splicing in lung
   cancer.^[ [207]^24 ^] Consistent with previous research,^[ [208]^23 ^]
   a portion of IGF2BP3 translocates to the nucleus in the absence of
   SLC27A5, regulates the alternative splicing of PIP4K2A mRNA by directly
   binding to it, and is involved in the progression of liver cancer. In
   future studies, it is necessary to characterize the mechanism governing
   IGF2BP3 nuclear localization and the splicing function of IGF2BP3

   A recent study showed that PIP4K2A physically interacted with p85 and
   induced proteasome‐mediated degradation, thereby regulating PI3K.^[
   [209]^36 ^] Similarly, we revealed that PIP4K2A‐L bound to and induced
   p85 degradation. In contrast, PIP4K2A‐S was incapable of binding to p85
   owing to the loss of exon 5, which resulted in absence of
   PIP4K2A‐mediated p85 ubiquitination. These results suggest that exon 5
   of PIP4K2A was the binding domain for p85. Further, in other lipid
   kinases, the activation loop in the kinase domain determines substrate
   specificity and is also involved in membrane recruitment.^[ [210]^43 ,
   [211]^44 ^] Exon 5 of PIP4K2A contains 51 amino acids, located in the
   ATP‐binding domain of PIP4K2A.^[ [212]^32 ^] Deletion of exon 5 may
   have altered the conformation of the activation loop and modified its
   subcellular localization.^[ [213]^32 , [214]^45 ^] Interestingly, we
   found that the subcellular distribution of PIP4K2A‐S differed from that
   of PIP4K2A‐L; PIP4K2A‐S was mainly distributed in the nucleus whereas
   PIP4K2A‐L was localized in the cytoplasm. Future studies on the
   mechanism underlying the nuclear accumulation of PIP4K2A‐S are required
   for deeper insights into the roles of PIP4K2A variants.

   The results of clinical studies targeting PIP4Ks are promising, but
   this strategy may not be appropriate for all cancers.^[ [215]^36 ^] In
   a recent study, the selective and potent PIP4K2A inhibitors (BAY‐091
   and BAY‐297) developed by Bayer did not induce an antiproliferative
   effect on p53 null tumor cells, suggesting that PIP4K inhibitors must
   potently inhibit both PIP4K2A and other PIP4K isoforms to exert an
   antitumor effect.^[ [216]^46 ^] Thus, attenuation of the antitumor
   effect of PIP4K2A inhibitors may be related to the splicing variants
   produced by alternative splicing. Here, we identified a new splicing
   variant, PIP4K2A‐S, which provides a promising alternative for
   additional PIP4K targets and may be a potential therapeutic target for
   HCC.

   Recent studies indicate that targeting alternative mRNA splicing is a
   promising strategy for cancer therapy.^[ [217]^13 ^] Thus, various
   approaches have targeted AS events to inhibit the initiation of poor AS
   variants. One strategy is the application of decoy oligonucleotides to
   compete for the natural binding targets of splicing factors for
   reducing their activity.^[ [218]^47 ^] In addition, engineered U7
   snRNAs and CRISPR‐based approaches for gene editing have the ability to
   target specific alternative splicing events.^[ [219]^48 , [220]^49 ^]
   RNA decoy oligonucleotides offer the following two advantages: 1) they
   have fewer side effects than those of complete knockdown^[ [221]^47 ^]
   and 2) they are single stranded RNA molecules, which can be delivered
   to the organs more efficiently than double stranded siRNA molecules.^[
   [222]^50 , [223]^51 ^] In our study, we designed RNA decoy
   oligonucleotides targeting IGF2BP3 and used them to block splicing
   factor activity by interacting with the RRM domains. Treatment with
   IGF2BP3 RNA decoy oligonucleotides caused isoform switching of
   PIP4K2A‐S to PIP4K2A‐L and suppressed the expression of p85 and p‐AKT,
   which inhibited tumor growth and metastasis both in vitro and in vivo.
   These findings indicate that RNA decoy oligonucleotides targeting
   tumor‐specific splicing‐related proteins may be a more promising
   clinical strategy for treating HCC.

   In summary, our findings indicate that SLC27A5 interacts with IGF2BP3,
   partially blocking the nuclear translocation of IGF2BP3 and its binding
   to PIP4K2A mRNA, thereby inhibiting conversion from PIP4K2A‐L to
   PIP4K2A‐S and suppressing activation of the PI3K/AKT pathway. This is
   an unrecognized noncanonical mechanism for SLC27A5 to inhibit tumor
   progression independently of its enzymatic activity. These findings
   highlight a potential prognostic marker for HCC and suggest that
   targeting IGF2BP3 activity is a promising strategy for treating
   patients with HCC as well as SLC27A5 deficiency.

4. Experimental Section

Cell Culture and Cell Transfection

   MHCC‐97H, PLC/PRF/5, HEK293, and HepG2 cells were cultured in DMEM
   (HyClone, Logan, UT, USA) mixed with 1% penicillin/streptomycin
   (HyClone) and 10% fetal bovine serum (FBS, Gibco, NY, USA) at 37 °C
   with 5% CO[2]. SNU449 cells were cultured in Roswell Park Memorial
   Institute (RPMI) 1640 medium (Gibco, Grand Island, NY, USA) mixed with
   1% penicillin/streptomycin (MedChemExpress, Monmouth, NJ, USA) and 10%
   FBS at 37 °C with 5% CO[2].

Animal Studies

   Slc27a5 ^−/− C57 BL/6J mice were generated from crosses between Slc27a5
   ^+/‐ mice and Slc27a5^+/− mice using the CRISPR/Cas9 system (European
   Mouse Mutant Archive, KOCMP‐04385‐Slc27a5), and WT mice were used as
   controls (n = 5 per group). After treatment with DEN (75 mg kg^−1) and
   PB (at a concentration of 0.05%), HCC was induced in these mice as
   described previously.^[ [224]^52 , [225]^53 ^] Tumor volume (V) was
   calculated as: V [cm^3]  =  (length [cm])  ×  (width [cm]  ×  (width
   [cm])/2. Metastasized tumor nodules and foci in the lungs were counted
   through microscopic observation.^[ [226]^16 ^]

   For SLC27A5 overexpression, AAV8‐TBG‐control or AAV8‐TBG‐Slc27a5 (Gene
   ID: 26459, NM[‐]009512.2, OBiO Technology, Corp., Ltd. Shanghai, China)
   was injected via tail vein into WT and Slc27a5 ^−/‐ mice after 20‐weeks
   of DEN/PB treatment (2 × 10^11 genome copies per mouse). After 40
   weeks, the mice were euthanized and their liver/lung tissues with
   tumors were examined.

   BALB/c nude mice were randomly grouped (n = 6/group) for the orthotopic
   implantation model. MHCC97H cells (1 × 10^5, AdPIP4K2A‐L‐,
   AdPIP4K2A‐S‐, AdGFP‐infected) were suspended in a 50 µL Matrigel
   (356234, BD Biosciences, CA, USA)/phosphate‐buffered saline (PBS)
   mixture (1:1 ratio, volol) and then implanted into the left liver lobe
   of each nude mouse. All mice were euthanized after 8 weeks of
   implantation. All animal experiments were compliant with the Medical
   Ethics Committee of Chongqing Medical University (reference number:
   2022053).

Clinical Specimens

   Human HCC tissues and corresponding paired nontumorous liver tissues
   were collected from 40 patients at the Second Affiliated Hospital of
   Chongqing Medical University, approved by the IRB of Chongqing Medical
   University (reference number: 2022053).

CRISPR/Cas9‐Mediated Gene Editing

   Single guide RNA sequences were designed to target the SLC27A5 and GSR
   genes (refer to Table [227]S3, Supporting Information). The sequences
   were designed using the E‐CRISP tool
   ([228]http://www.e‐crisp.org/E‐CRISP/designcrispr.html) and
   subsequently cloned into the CRISPRv2 lentiviral vector expressing
   Cas9, which was kindly provided by Prof. Ding Xue from Tsinghua
   University, Beijing, China. Lentiviruses were then packaged, and stable
   knockout cell lines were selected following previously established
   protocols.^[ [229]^11 ^]

Plasmid Construction

   The cloned human vector cDNAs for SLC27A5 were ligated into the
   TO4‐3Flag expression vector to produce Flag‐tagged proteins. The
   mutants of SLC27A5 (S296A) were constructed by site‐directed
   mutagenesis using wild‐type (WT) SLC27A5‐3xFlag plasmid as the
   template.

Cell Transfection and Reagent Treatment

   SNU449 cells (parental or SLC27A5‐KO) were transfected with 200 nm RNA
   decoys using Lipofectamine RNAiMAX (Thermo Fisher Scientific). SNU449
   and HepG2 cells were transfected with 50 nm shRNAs using Lipofectamine
   3000 (Thermo Fisher Scientific). PLC/PRF/5 and 293 cells were treated
   with MG132 (HY13259, MedChemExpress) for 8 h.

IP‐MS Assay

   PLC/PRF/5 cells, which were pre‐transfected with SLC27A5‐TO4‐3Flag or
   TO4‐3Flag plasmid, were incubated overnight at 4°C with ANTI‐FLAG M2
   affinity gel (A2220, Sigma). Coomassie blue was used to stain the
   protein complexes. The excised protein gel bands were submitted to
   Shanghai Applied Protein Technology Co., Ltd. for identification of
   proteins that bind to SLC27A5. After undergoing reduction reaction and
   alkylation treatment, the samples were incubated with trypsin (mass
   ratio, 1:50). The digested samples were then dried and analyzed by mass
   spectrometry after injecting 10 or 20 µL of each sample. Mass
   spectrometry (MS) analysis was performed on a Lumos Tribrid Orbitrap
   Mass Spectrometer (Thermo Scientific) equipped with Ultimate 3000
   (Thermo Scientific) nano‐high‐performance liquid chromatograph. The raw
   MS data were analyzed using Mascot 2.2 software to identify the
   immunoprecipitating proteins. Further, Gene Ontology enrichment
   analysis was performed using clusterProfiler package in R, with
   Fisher's Exact Test, and the significant p value cutoff was set at
   0.05.

RNA‐Seq Analysis for Identifying Metastasis‐Associated Splicing Events

   Five cases of primary HCC tissue samples were collected from
   Slc27a5^−/− mice and five HCC tissue samples from WT mice for RNA‐seq.
   rMATS were used for splicing analysis in the data set. The expression
   levels of alternative splicing events (ASE) in primary HCC tissues with
   Slc27a5^−/− mice and WT mice were quantified, and the differentially
   expressed isoforms or exons between two groups were analyzed using
   MISO. The ASEs with |Δdifferent| > 0.2 and bayes_factor > 2 were
   considered significantly differentially spliced.

RIP Assay

   The EZ‐Magna RIP kit (17‐701, Millipore) was used to perform RIP. In
   short, ≈2 × 10^7 SLC27A5‐KO HepG2 cells were lysed with 200 µL RIP
   Lysis Buffer at low temperature for 5 min and centrifuged. The
   supernatants were incubated with magnetic beads (Thermo Fisher
   Scientific) and the appropriate antibodies for 8 h at 4 °C with gentle
   rotation. The beads were washed four times with RIP lysis buffer
   containing RNase inhibitor (Millipore, Massachusetts, USA). The RNA was
   extracted by chloroform and detected by RT‐PCR. Total RNA (input
   controls) and negative IgG controls were detected simultaneously.
   Table [230]S4 (Supporting Information) indicates the specific primers
   used for the RIP assay. Table [231]S5 (Supporting Information) lists
   the antibodies used for RIP assay.

Immunofluorescence Staining

   After plated on coverslips in a 12‐well dish, HepG2 cells were fixed
   with paraformaldehyde, and then blocked using bovine serum albumin. The
   coverslips were then incubated with primary antibody (SLC27A5, Novus
   Biologicals, 1:200; IGF2BP3, Proteintech, 1:400) at 4 °C overnight. The
   cells were then incubated with goat anti‐rabbit IgG/TRITC or goat
   anti‐mouse IgG/FITC for 2 h at room temperature. The cells were viewed
   and photographed under a fluorescence microscope (Leica TCS SP8, Solms,
   Germany).

RNA Fluorescence In Situ Hybridization Combined with Immunofluorescence

   The cells were seeded on the coverslips pre‐coated with 5 µg mL^−1
   poly‐L‐ornithine (P4957, Sigma–Aldrich) for 30 min, and grown for 24 h
   followed by fixation using 4% paraformaldehyde. The cells were then
   washed thrice in PBS and permeated using 0.5% Triton X‐100
   (Sigma–Aldrich). The cell were then blocked with 4% IgG‐free Bovine
   Serum Albumin and incubated with the primary antibody (see
   Table [232]S5, Supporting Information) for 4–16 h. After washing thrice
   with PBS, the cells were incubated with the secondary antibody for 1 h
   and then fixed using 4% paraformaldehyde. After two further PBS washes,
   the cells were examined using a fluorescence in situ hybridization
   (FISH) kit (GenePharma, [233]F16501/50); the RNA FISH probes are listed
   in Table [234]S4 (Supporting Information). The cells were then washed
   once with water, sealed with nail polish, and photographed under a
   fluorescence microscope (Leica TCS SP8, Solms, Germany).

Glutathione S‐Transferase (GST) Pull‐Down Assay

   GST‐SLC27A5 was purified by glutathione agarose beads (GE Healthcare,
   Piscataway, NJ, USA). His‐IGF2BP3 was purified by HisPur
   Ni‐NTAmagneticbeads (Thermo Fisher, MA, USA). The purified proteins
   were incubated with glutathione agarose beads for 3 h. The beads were
   washed with wash buffer (137 mm NaCl, 2.7 mm KCl, 10 mm Na[2]HPO[4],
   2 mm KH[2]PO[4] and 0.5% Triton X‐100), mixed with 2× SDS loading
   buffer and boiled for 8 min. The inputs/elutions were further analyzed
   by Coomassie staining and/or immunoblot analysis.

RNA Pull‐Down Assays

   RNA Pull‐down assays were performed by incubating 0.1 mm biotin‐PIP4K2A
   with purified His‐IGF2BP3 (5 µg) for 0.5 h at 25 °C. RNAs with random
   sequences served as the negative control. Complexes were pulled down
   using Pierce streptavidin agarose beads and then immunoblotted with the
   indicated antibodies.

PIP4K2A‐L and PIP4K2A‐S Antibody Production

   Anti‐PIP4K2A‐L or Anti‐PIP4K2A‐S antibodies generated using the
   NSLTRSAPLPNDSQAR and RFGIDDQDFQYIVEC antigens were purchased from
   ABclonal.

Co‐Immunoprecipitation and Western Blot Analysis

   Cells were resuspended in RIPA lysis buffer mixed with an EDTA‐free
   protease inhibitor cocktail (#C0001, TargetMol, MA, USA). The
   precleared lysate was incubated with magnetic beads (Thermo Fisher
   Scientific, Carlsbad, CA, USA) and then immunoprecipitated overnight at
   4 °C using the indicated antibodies. The protein complex was then
   subjected to western blotting. Briefly, the samples were separated
   using sodium dodecyl‐sulfate polyacrylamide gel electrophoresis and
   then blotted on PVDF membranes (Millipore, Billerica, MA, USA). The
   membrane was then blocked for 30 min with 5% nonfat milk, and incubated
   overnight with primary antibodies. This was followed by incubation with
   horseradish peroxidase‐conjugated secondary antibodies. The signal was
   then visualized using ECL (WBKLS0500, Millipore). GAPDH was used as the
   loading control. The antibodies in the study were listed in
   Table [235]S5 (Supporting Information).

qPCR

   TRIzol reagent (Invitrogen, Carlsbad, CA, USA) was used to extract
   total RNA from tissues or cells. PrimeScript RT Reagent Kit (TaKaRa,
   Shiga, Japan) was used for cDNA synthesis. The target gene expression
   was then detected using real‐time PCR (SYBR green) with specific
   primers. The PCR conditions were as follows: 95 °C for 3 min, followed
   by 40 cycles of 95 °C for 15 s, 60 °C for 15 s, and 72 °C for 20 s,
   concluding with an extension step of 72 °C for 5 min. The data were
   analyzed using the 2^−∆∆Ct method. The primers used were listed in
   Table [236]S4 (Supporting Information).

Statistics and Reproducibility

   The dead animals and euthanized animals were excluded according to the
   IACUC protocol. The sample size was not predetermined by any
   statistical method. Investigators who were blinded to the experimental
   conditions did not perform the data collection and analyses. Animals
   were randomized and treated as former indication for each in vivo. All
   samples were analyzed equally with no sub‐sampling resulting and there
   was no need for randomization in the in vitro experiments. The
   experiments in Figures [237]2I–K; [238]3A–L; [239]5D,E; and [240]7B,F–I
   and in Figures [241]S1A,C,D,E; [242]S3A–E; [243]S6A,C,F;[244]S7C;
   [245]S8B; [246]S10B; [247]S11A,B (Supporting Information) were
   performed at least thrice and similar results were obtained. The figure
   captions indicate the statistical analyses performed. The means ±
   standard error of the mean (SEM) is indicated by the error bars.
   Statistical calculations were performed using GraphPad Prism software
   (version 7). Clinicopathological characteristics of patients with HCC
   were analyzed using χ^2 analysis. Data are expressed as mean ± standard
   deviation (SD).

Ethics Approval Statement

   This study was approved by the Research Ethics Committee of Chongqing
   Medical University (approval number: 2022054). All animal experiments
   were performed under the guidelines of the institutional Animal Care
   and Use Committee at Chongqing Medical University.

Patient Consent Statement

   Informed consents were obtained from all involved participants.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   D.N., X.T., H.J.D. and X.J.Y. contributed equally to this work. D.N.,
   X.T., H.J.D., X.J.Y., J.J.T., F.L.X., Y.L., and K.W. performed
   experiments. H.J.D. and M.Z.C. provided technical and clinical support.
   X.J.Y. evaluated histopathology. K.W., M.Z.C., A.L. and N.T. designed
   and supervised experiments. N.D., K.W., and N.T. wrote the manuscript.
   All authors read and approved the manuscript.

Supporting information

   Supporting Information
   [248]Click here for additional data file.^ (3.1MB, pdf)

   Supporting Information
   [249]Click here for additional data file.^ (13MB, xlsx)

Acknowledgements