Genome-wide CRISPR screen identifies splicing factor SF3B4 in driving hepatocellular carcinoma

Abstract

Although genome sequencings have recognized many cancer-associated genes in hepatocellular carcinoma (HCC), distinguishing their functional effect remains challenging. Leveraging on a genome-wide CRISPR knockout (KO) screening, we uncovered spliceosome factors as major survival essential genes in HCC and up-regulations of ferroptosis suppressors [particularly glutamate-cysteine ligase catalytic subunit (GCLC)] in lenvatinib resistance. Our KO screen in patient-derived HCC organoid showed splicing factor 3b subunit 4 (SF3B4) to be top-ranked, conferring prosurvival signal in HCC organoid and driving tumorigenic potentials in both hepatic progenitor organoids and hydrodynamic tail vein injection HCC murine model. The combined RNA immunoprecipitation sequencing, long-read isoform sequencing, and transcriptome revealed characteristic splicing landscape regulated by SF3B4 and identified T-box transcription factor 3 (TBX3) variant TBX3+2a as a potent downstream effector. Our findings highlighted vital roles of SF3B4 in HCC cell survival and tumor progression, and the phenomenon of ferroptosis resistance in patients unresponsive to first-line agent lenvatinib.

Fig. 1.. Organoid-based CRISPR screen revealed the essential role of spliceosome in HCC growth.

(A) Representative brightfield and histological images of patient-derived HCC tumor organoid H813Torg. Scale bar, 50 μm. (B) Schematic workflow of the genome-wide CRISPR screen using H813Torg. Experiments were performed in four biological replicates. (C) Fold change in sgRNA abundance for predefined high-confidence core essential (n = 625) and nonessential genes (n = 350) highlighted the reliability of the screening results. (D) Rank plot showing that 1470 HCC essential genes were negatively selected from the KO screen (P < 0.05 in negative selection). (E) Over-representation analysis (ORA) identified the top KEGG pathways enriched for HCC essential genes. (F) Volcano plot highlighting distribution of sgRNAs targeting spliceosome genes in the KO screen. Spliceosome genes ranked within the top 30% of essential genes are highlighted in pink. (G) Fold change in expression of spliceosome genes in HCC tumor (T) versus adjacent nontumor (NT) liver tissues across two cohorts of primary tissues. SF3B4 is the most up-regulated spliceosome gene in both The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) and Prince of Wales (PWH)–HCC cohorts. (H) Forest plot showing the association between the expression of the top 6 up-regulated spliceosome genes in HCC tumor and overall survival (n = 364). Patients were stratified by median expression levels of individual genes. (I) Transcriptome sequencing results revealing notable up-regulation of SF3B4 in HCC tumor compared to paired adjacent nontumor liver tissues (n = 27 pairs). (J) qPCR confirmed substantial up-regulation of SF3B4 in 60 pairs of HCC tumor and matched nontumor liver tissues. (K) Immunoblot analysis demonstrated up-regulation of SF3B4 protein levels in HCC tumor tissues (n = 6). Data were plotted as means ± SD. ***P < 0.001 and ****P < 0.0001 by Student’s t test [(C), (I), (J), and (K)]. gDNA, guide DNA.

Fig. 2.. SF3B4 maintained HCC cell survival and promoted liver oncogenicity.

(A) Knockdown of SF3B4 substantially suppressed the growth of two HCC tumor organoids, H813Torg and H688Torg (n = 4). (B) Live/dead staining demonstrating that knockdown of SF3B4 induced spontaneous cell death in HCC tumor organoids. Viable cells were stained with Calcein AM (green). Dead cells were stained with EthD-1 (red) and quantified by MetaXpress software (n = 3). Scale bar, 50 μm. (C) Representative brightfield images showing that overexpression of SF3B4 (SF3B4^OE) conferred growth advantages to non-tumoral liver organoids (Liver. Orgs) compared to the empty vector (EV) control. (D) Representative images of hematoxylin and eosin (H&E) staining revealed that SF3B4 expression induced premalignant morphological changes in liver organoids. (E) Morphological analysis indicated the proportion of single-layered versus tumor-like morphology in liver organoids expressing vector control or SF3B4 (n = 5). (F) Expansion potential of liver organoids showing that vector control organoids ceased propagation within 20 passages, whereas SF3B4-overexpressing organoids continued to grow for over 40 passages. (G) Representative immunohistological (IHC) images demonstrated a marked increase in the staining of proliferation marker Ki67 in SF3B4-overexpressing organoids. (H) Gene set enrichment analysis (GSEA) revealed significant enrichment of a stemness-associated gene set in SF3B4-overexpressing liver organoids. (I) qPCR results showed increased expression of SOX9 and EpCAM in SF3B4-overexpressing liver organoids. (J) Tumor incidence on day 35 revealed that SF3B4 enhanced liver oncogenicity in immunocompetent C57B/L mice. Values above the columns indicate the number of animals. (K) Representative images of HDTVi-induced HCC tumors and quantification result revealed that SF3B4 promoted tumor growth in vivo. Data were plotted as means ± SD. *P < 0.05, ***P < 0.001, and ****P < 0.0001 by two-way analysis of variance (ANOVA) [(A) and (J)], one-way ANOVA (B), or Student’s t test [(E), (I), and (K)]. OE, overexpression. NES, normalized enrichment score.

Fig. 3.. SF3B4 altered mRNA splicing patterns.

(A) Over-representation analysis (ORA) revealed that SF3B4-binding proteins were enriched for RNA splicing and processing. SF3B4-associated binding partners were identified by IP-LC/MS. (B) IP-LC/MS results identified spliceosome components and RNA splicing regulators as the major categories of SF3B4-binding proteins. (C) Density scatter plot showing substantial enrichment of reads aligned to the identified SF3B4-binding regions from RNA RIP-seq compared to the input control. (D) Distribution of SF3B4-bound RNAs across genomic elements, as identified by RIP-seq. (E) Hybrid RNA-seq analysis identified alternative splicing events modulated by SF3B4, with cassette exon events being the most altered category. Other categories include retained intron (RI), alternative 5′ splice site (A5SS), alternative 3′ splice site (A3SS), and mutually exclusive exon (MXE). (F) Scatter plot of transcripts showing fold change in expression and delta percentage spliced in (dPSI) in SF3B4^OE group relative to the vector control. Transcripts with increased dPSI > 0.1 and Log₂ fold change (SF3B4/EV) > 2 are highlighted in red, while transcripts with dPSI < −0.1 and log₂ fold change (SF3B4/EV) < −1 are highlighted in blue. (G) Enrichment analysis revealed that transcripts up-regulated by SF3B4 were enriched for transcription factor binding and DNA binding activities. Gene Ontology (GO) molecular function (GOMF) gene sets were used for the analysis. 5′UTR, 5′ untranslated region. ncRNA, noncoding RNA. hnRNP, heterogeneous nuclear ribonucleoproteins.

Fig. 4.. TBX3+2a is a critical downstream effector of SF3B4.

(A) qPCR analysis showed a significant correlation between TBX3+2a expression and SF3B4 levels in HCC tumors, while TBX3 did not. A cohort of 41 primary HCC specimens was analyzed. (B) RIP-seq identified SF3B4 binding near TBX3 Exon 2a region. (C) RIP-qPCR validation confirmed SF3B4 binding to TBX3 pre-mRNA, with EZH2 antibody used as a negative control for specificity (n = 4). (D) Semi-qPCR showed that SF3B4 preferentially increased TBX3+2a expression in liver organoids (n = 3). (E) qPCR analysis confirmed increased TBX3+2a expression upon SF3B4 overexpression (n ≥ 3). (F) Knockdown of SF3B4 reduced TBX3+2a expression in HCC tumor organoid H688Torg (n = 3). (G) Semi-qPCR showed that ASOs targeting SF3B4-binding regions reduced TBX3+2a production compared to TBX3 (n = 4). (H) Minigene reporter assay confirmed that knockdown of SF3B4 preferably decreased minigene long isoform production (n = 3). L, long isoform representing TBX3+2a; S, short isoform representing TBX3. (I) Cell viability assays revealed that knockdown of TBX3+2a suppressed SF3B4-driven liver oncogenicity (n ≥ 3). (J) Representative image (left) and tumor weight quantification (right) of H813Torg-derived subcutaneous xenografts demonstrated that spliceosome modulator H3B-8800 (H3B) suppressed tumor growth as monotherapy while sensitized HCC xenografts to lenvatinib (n = 6). ODX, organoid-derived xenograft. Data were plotted as means ± SD; ns, not significant (P > 0.05); *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by one-way ANOVA [(C), (F), (G), (H), and (J)], Student’s t test [(D) and (E)], or two-way ANOVA (I). bp, base pair.

Fig. 5.. Organoid-based CRISPR screen identified GCLC as a key mediator of lenvatinib resistance.

(A) Schematic workflow of the CRISPR KO screen used to identify genes associated with lenvatinib resistance. Experiments were performed as four biological replicates. (B) Rank plot showing that GCLC was the most significant gene associated with lenvatinib resistance. Ferroptosis suppressor genes ranked top in the screen and are highlighted in pink. (C) Elevated expression of ferroptosis suppressor genes was associated with progressive disease (PD) compared to complete response (CR) in patients treated with lenvatinib. (D and E) Knockdown of GCLC sensitized HCC organoids to lenvatinib treatment. Median inhibitory concentration (IC₅₀) values were quantified from independent experiments (n = 4). (F) Expression of GCLC conferred lenvatinib resistance to HCC organoid H744Torg. IC₅₀ values were quantified from four independent experiments. (G) Representative images of H813Torg-derived subcutaneous xenografts (left) and quantification of tumor volume (right) showed that knockdown of GCLC-sensitized HCC cells to lenvatinib treatment in vivo (n = 5). (H and I) qPCR analysis showed that knockdown of GCLC increased the expression of ferroptosis markers upon lenvatinib treatment (n = 3). (J) Flow cytometric analysis of BODIPY-C11 staining revealed that knockdown of GCLC exacerbated lenvatinib-induced lipid peroxidation. fluor, fluorescence. (K) qPCR analysis demonstrated that lenvatinib dose dependently increased GCLC expression in H813Torg (n = 3). (L) Representative image (left) and tumor weight quantifications (right) of H813Torg-derived subcutaneous xenografts showed that the l-buthionine-(S,R)-sulfoximine (BSO) or imidazole ketone erastin (IKE) conferred sensitivity to lenvatinb in HCC xenografts (n = 7). ODX, organoid-derived xenograft. Data were plotted as means ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA [(D), (E), (H), (K), and (L)] or Student’s t test [(F), (G), and (I)].