PMVK drives hepatocarcinogenesis via SP1 phosphorylation–mediated lipid metabolic reprogramming Authors: Ruiyang Liu, Caini Huang, Zhijie Xu, Fuyuan Xu, Xinyi Wen, Zhiju Zhao, Fei Xiao Published: 2026 ABSTRACT Metabolic reprogramming is a hallmark of cancer, yet the critical drivers and mechanisms in hepatocellular carcinoma (HCC) remain incompletely understood. Through a metabolism-focused CRISPR screen, we identified phosphomevalonate kinase (PMVK), a mevalonate pathway enzyme, as a key regulator of HCC stemness and progression. PMVK directly phosphorylates the transcription factor SP1 at Thr355, which enhances SP1’s DNA binding affinity and promotes its interaction with the master lipid regulator SREBP1/2, driving a transcriptional program essential for de novo cholesterol and fatty acid synthesis. Clinically, PMVK levels are positively correlated with phospho-SP1 (Thr355) levels and predicate poor prognosis in patients with HCC. Functionally, we demonstrate that genetic or pharmacologic inhibition of the PMVK-SP1 axis, including using the SP1 inhibitor mithramycin A (MTA), significantly suppresses HCC tumorigenesis. In conclusion, our findings uncover the PMVK-SP1 axis as a central driver of lipid metabolic reprogramming and hepatocarcinogenesis, highlighting it as a compelling therapeutic target in HCC. INTRODUCTION Metabolic reprogramming is an established hallmark of cancer. However, the precise molecular mechanisms that drive this reprogramming, particularly during tumor initiation and progression, remain incompletely understood (1). A pathway central to this metabolic shift is the mevalonate (MVA) pathway, which consumes acetyl–coenzyme A, reduced form of nicotinamide adenine dinucleotide phosphate, and adenosine 5′-triphosphate to produce sterols and isoprenoids that are critical for tumor growth and survival. Consequently, the MVA pathway is frequently up-regulated across diverse cancer types, where its activity is strongly linked to accelerated tumor tumorigenesis (2, 3). Cholesterol, a major product of the MVA pathway, supports cancer cell proliferation by fueling membrane biogenesis and acting as a precursor for essential signaling molecules. Enhanced cholesterol biosynthesis is a prominent feature of metabolic reprogramming in multiple tumors, including hepatocellular carcinoma (HCC) (4, 5). In HCC, it promotes tumor stemness and progression by activating key oncogenic pathways such as Hedgehog and mTOR pathways (6–8), while also inhibiting autophagic degradation of receptor tyrosine kinase in a DOLM1-dependent manner (9). In parallel, dysregulated fatty acid (FA) biosynthesis represents another major metabolic alteration driving HCC initiation and progression (10). Elevated expression of FA biosynthetic enzymes, including ATP citrate lyase (ACLY) Acetyl-CoA carboxylase (ACC), Fatty Acid Synthase (FASN), and Stearoyl-CoA Desaturase (SCD), is associated with poor patient outcomes, and inhibition of these enzymes has been shown to reduce hepatocarcinogenesis and enhance therapeutic responses (11–13). Despite these insights, how cholesterol and FA biosynthesis are coordinately regulated to support HCC stemness and progression remains a central unanswered question. Sterol regulatory element-binding proteins (SREBPs), particularly SREBP1 and SREBP2, are master transcription factors of FA and cholesterol biosynthesis, respectively (14). Their activity is regulated at multiple levels and in different subcellular compartments (14). The transcription factor SP1, a known oncogene in HCC, has been reported to interact with SREBP1/2 and coregulate a subset of their target genes (15, 16). SP1 contributes to tumor proliferation, migration, invasion, and metastasis by regulating both protein-coding and noncoding RNA (17, 18). Furthermore, its transcriptional activity is known to be modulated by post-translational modifications, including phosphorylation and acetylation (19, 20). However, the role of these modifications in regulating lipid metabolic programs, particularly within liver cancer stem cells (LCSCs), remains entirely unexplored. Here, through a metabolic CRISPR screen, we demonstrate that phosphomevalonate kinase (PMVK), a kinase within the MVA pathway (2), as a key driver of HCC stemness and progression. We demonstrate that PMVK directly phosphorylates SP1 at Thr355, thereby enhancing SP1’s DNA binding affinity and its physical interaction with nuclear SREBP1/2. This PMVK-SP1-SREBP1/2 axis drives the expression of a coordinated lipid biosynthetic program. Critically, inhibition of SP1 abolishes the tumor-promoting effects of PMVK, establishing this axis as a fundamental mechanism of lipid metabolic reprogramming in hepatocarcinogenesis. RESULTS PMVK is essential for HCC stemness and malignant phenotypes To systematically identify metabolic regulators of HCC stemness and progression, we transduced PLC/PRF/5 (PLC) cells with a lentiviral, metabolism-focused CRISPR-Cas9 gRNA library. We then performed two parallel screens: one for stemness and one for proliferation. For the stemness screen, we leveraged the strong tumor-initiating capacity of CD24+CD133+ LCSCs (21). After puromycin selection, we sorted CD24+CD133+ and CD24−CD133− populations to identify genes regulating the CSC state. For the proliferation screen, we cultured the knockout pool for 14 days and collected the surviving cells. We then used next-generation sequencing to quantify gRNA abundance in each population (Fig. 1A). PMVK is essential for HCC stemness and malignant progression. (A) Schematic of the metabolism-focused CRISPR-Cas9 screen to identify regulators of HCC stemness and proliferation. (B) Gene ontology (GO) analysis of screen hits, showing significant enrichment in lipid metabolic processes. (C) Overlap of candidate genes required for both HCC stemness and proliferation, as determined by robust rank aggregation (RRA) analysis of the dual CRISPR screen. PMVK and other selected known regulators are highlighted. (D) Flow cytometry analysis of CD24+CD133+ HCC stem cells in HCC cells transduced with scrambled control shRNA (shScramble) or shRNAs against PMVK, TPl1, and SEPSH1. (E) IncuCyte proliferation assay showing the effect of PMVK, TPl1, or SEPSH1 knockdown on HCC cell proliferation. (F) Limiting dilution transplantation assay of control or PMVK-knockdown PLC cells in NOD/SCID mice (n = 6 mice per group). (G) Viability of patient-derived HCC organoids after PMVK knockdown. Data are from three biological replicates. Scale bars, 50 μm. (H) Representative images (left) and quantification (right) of lung metastatic nodules in mice injected with control or PMVK-knockdown HCC cells via the tail vein. Scale bars, 50 μm. (I) Hydrodynamic transfection of Myc/sgp53/Pmvk plasmids to construct a Pmvk-overexpressing (PmvkOE) mouse HCC model. (J) Representation image of mouse livers following mediated Myc overexpressing and p53 knockout with or without constitutive Pmvk overexpression (n = 6). (K) Livers weights and rations of liver to body weight in control and PmvkOE mice. (L) Representative IHC staining for PMVK, Ki-67, and AFP in liver tissues from control and PmvkOE mice. Scale bars, 50 μm. Data represents at least three biological replicates [(D), (E), (G), (K), and (L)]. In all statistical plots, data are shown as mean ± SEM. P values were determined by two-tailed unpaired Student’s t test [(D), (E), (G), (K), and (L)]. Model-based analysis of genome-wide CRISPR-Cas9 knockout (MAGeCK) analysis identified several known regulators of cancer stemness and proliferation, including INPP5D, GRIK2, PRPS1, and PKM (fig. S1A) (22). Gene ontology (GO) analysis revealed that the screen hits were significantly enriched for genes involved in lipid metabolism (Fig. 1B). By overlapping the hits from both screens, we identified a core set of genes essential for both HCC stemness and proliferation, including TPI1, PYROXD1, AGK, SEPHS1, and PMVK (Fig. 1C). Functional screening by fluorescence-activated cell sorting (FACS) and cell proliferation assays revealed that knockdown of PMVK, TPI1, or SEPHS1 significantly suppressed HCC stemness and proliferation. Among these, PMVK knockdown exhibited the most pronounced inhibitory effects, leading us to focus on PMVK for subsequent studies (Fig. 1, D and E). PMVK, an enzyme in the MVA pathway, is essential for cholesterol synthesis. Analysis of The Cancer Genome Atlas (TCGA) data confirmed that PMVK expression is significantly up-regulated in HCC tissues compared to normal liver and that high PMVK expression correlates with poor patient survival (fig. S1, B and C). On the basis of these findings, we selected PMVK for further investigation. To validate PMVK’s functional role, we knocked it down in HCC cells (fig. S1D). PMVK knockdown significantly reduced the CD24+CD133+ CSC population and impaired spheroid formation (fig. S1, E and F). In vivo, PMVK knockdown markedly reduced tumor-initiating capacity and CSC frequency, as assessed by limiting dilution and serial transplantation assays in NSG mice (Fig. 1F). Furthermore, PMVK knockdown suppressed malignant phenotypes in vitro, including proliferation, migration, and invasion (fig. S1, G to I). This tumor-suppressive effect was confirmed in primary liver cancer organoids, where PMVK knockdown impaired tumoroid formation and viability (Fig. 1G). In a lung metastasis model, PMVK knockdown drastically reduced metastatic incidence (Fig. 1H). We next evaluated the oncogenic potential of Pmvk overexpression in vivo using a hydrodynamic transfection model. We codelivered a p53 sgRNA and c-Myc overexpression plasmids with either a Pmvk overexpression or control plasmid (Fig. 1I). Compared to controls, PmvkOE mice developed significantly larger liver sizes, with higher liver–to–body weight ratios (Fig. 1, J and K). Immunohistochemistry (IHC) staining revealed that tumors from PmvkOE mice had higher percentage of cells positive for the HCC marker AFP and proliferation marker Ki-67 (Fig. 1L). These results are consistent with a previous report that liver-specific PMVK deletion inhibits HCC growth (23). In summary, our genetic screens and functional validation demonstrate that PMVK is a pivotal driver of HCC stemness, tumor growth, and metastasis. PMVK drives HCC stemness and malignant phenotypes by enhancing cholesterol and FA biosynthesis independent of MVA-5PP To elucidate the mechanisms by which PMVK promotes HCC stemness and malignant phenotypes, we performed RNA sequencing (RNA-seq) analysis following PMVK knockdown. This revealed that PMVK silencing down-regulated key oncogenes (e.g., FOS, PIK3CA, and HIF1A) and up-regulated tumor suppressors (e.g., BAX, RRAP1, and HNF4A) (Fig. 2A). GO analysis further indicated that PMVK knockdown suppressed genes involved in lipid metabolism, including FA metabolism, steroid biosynthesis, and phosphatidylinositol signaling (Fig. 2B). The strongly suggested that lipid metabolic reprogramming is central to PMVK’s oncogenic function. PMVK drives cholesterol and FAs biosynthesis. (A) Volcano plot of differentially expressed genes in control versus PMVK-knockdown HCC cells. Selected oncogenes (red) and tumor suppressors (blue) are highlighted. (B) KEGG pathway enrichment analysis of genes down-regulated upon PMVK knockdown. (C) Heatmap showing expression of cholesterol and FA biosynthesis genes in control and PMVK-knockdown HCC cells. (D) Western blot analysis of key lipid metabolism regulators in control and PMVK-knockdown cells. (E) Luciferase reporter assays measuring transcriptional activity of HMGCR, SQLE, FASN, and SCD promoters in control and PMVK-knockdown cells. (F) Intracellular total cholesterol (T-CHO) and triglyceride (TG) levels in control and PMVK-knockdown HCC cells. (G) Representative IHC staining of HMGCR, SQLE, FASN, and SCD1 in livers from control and PmvkOE mice. Scale bars, 50 μm. (H) Bioluminescence imaging of orthotopic tumors in mice injected with PMVK-overexpressing HCC cells with or without HMGCR/SCD knockdown. (I) Total cholesterol (T-CHO) and triglyceride (TG) levels in the orthotopic tumors from (H). Data represent at least three biological replicates [(A) to (I)]. In all statistical plots, data are shown as mean ± SEM. P values were determined by two-tailed unpaired Student’s t test [(E) to (I)]. Although PMVK is canonically known to catalyze the production of 5-diphosphomevalonate (MVA-5PP) in the MVA pathway (24), our GO analysis suggested a broader role in regulating FA synthesis. Consistent with this, PMVK knockdown significantly reduced the expression of critical genes in both the MVA pathway (HMGCR and SQLE) and FA synthesis (SCD and FASN) (Fig. 2C and fig. S2C). We confirmed these reductions at the protein level by Western blot and at the transcriptional level using luciferase reporter assays (Fig. 2, D and E). Functionally, PMVK knockdown led to a substantial decrease in intracellular cholesterol, FAs, lipid droplet content, and overall lipid species, as shown by enzymatic assays and high-throughput targeted lipidomics (Fig. 2F and fig. S2, B to F). Conversely, tumors from PmvkOE mice exhibited increased expression of FASN, SCD1, 3-Hydroxy-3-Methylglutaryl-CoA Reductase (HMGCR), and Squalene Epoxidase (SQLE) (Fig. 2G). These complementary loss- and gain-of-function data establish PMVK as a key driver of lipid biosynthesis in HCC. To test whether lipid biosynthesis is functionally required for PMVK’s effects, we performed rescue experiments by knocking down HMGCR or SCD in PMVK-overexpressing cells. Co-knockdown of HMGCR or SCD significantly reversed PMVK-induced enhancement of HCC stemness, spheroid formation, proliferation, migration, and invasion (fig. S3, A to E). In an orthotopic model, PMVK overexpression accelerated tumor growth, and this effect was markedly attenuated by knockdown of HMGCR, SCD, or both (Fig. 2H). Correspondingly, the high intracellular cholesterol and FA levels in PMVK-overexpressing tumors were also suppressed by HMGCR or SCD knockdown (Fig. 2I). Last, we examined whether PMVK exerts its effects through its catalytic product, MVA-5PP (fig. S4A). Notably, supplementation with MVA-5PP only marginally restored stemness and proliferation in PMVK-knockdown cells, whereas cholesterol effectively reversed these defects (fig. S4, A to C). In summary, PMVK promotes HCC stemness and tumorigenesis primarily by orchestrating a transcriptional program that drives cholesterol and FA biosynthesis, through a mechanism largely independent of its catalytic product MVA-5PP. PMVK drives HCC stemness and malignant phenotypes in a kinase activity-dependent manner We next investigated whether PMVK’s oncogenic functions depend on its catalytic activity. We reconstituted PMVK-knockdown HCC cells with either wild-type (WT) PMVK or a kinase-dead mutant (K22M). Western blotting confirmed comparable overexpression of both constructs (fig. S5A). Re-expression of WT PMVK largely rescued the CD24+CD133+ CSC population and spheroid formation capacity impaired by PMVK knockdown, whereas the K22M mutant produced a markedly weaker rescue (Fig. 3, A and B), indicating that PMVK’s kinase activity is crucial for maintaining HCC stemness. Furthermore, in a limiting dilution assay, WT PMVK significantly restored the tumor-initiating frequency, while the K22M mutant failed to do so (Fig. 3C). The functional rescue by WT PMVK, but not the K22M mutant, extended to the restoration of proliferation, growth, migration, and invasion (Fig. 3D and fig. S5, B to D). Similarly, in patient-derived organoids, WT PMVK effectively restored viability, an effect that was significantly blunted by the K22M mutation (Fig. 3E). PMVK drives HCC stemness and progression in a kinase-dependent manner. (A) Flow cytometry analysis of CD24+CD133+ HCC CSC population in PMVK-knockdown HCC cells reconstituted with wild-type (WT) PMVK or the kinase-dead (K22M) mutant. (B) Representative images (left) and quantification (right) of spheroid formation in the indicated cell groups. Scale bars, 200 μm. (C) Limiting dilution assay of PMVK-knockdown PLC cells reconstituted with WT PMVK or PMVK K22M in NOD/SCID mice (n = 6 mice per group). (D) Bioluminescence imaging of orthotopic tumors in mice injected with PMVK knockdown PLC cells reconstituted with WT PMVK or PMVK K22M. (E) Viability of HCC tumoroids with PMVK knockdown, reconstituted with WT PMVK or PMVK K22M. Data represent three biological replicates [(A), (B), (D), and (E)]. In all statistical plots, data are shown as mean ± SEM. P values were determined by two-tailed unpaired Student’s t test [(A), (B), (D), and (E)]. We then assessed whether PMVK’s kinase activity is required for lipid biosynthesis. WT PMVK re-expression restored intracellular lipid droplet content, total cholesterol, triglyceride, HMGCR, SQLE, SCD, and FASN levels in knockdown cells, whereas the K22M mutant was ineffective at this rescue (fig. S5, E to H). Collectively, these data demonstrate that PMVK promotes HCC stemness, tumorigenicity, and lipid biosynthesis in a manner strictly dependent on its kinase activity. PMVK directly interacts with phosphorylate SP1 at Thr355 To identify the direct substrates through which PMVK regulates HCC stemness and progression, we performed mass spectrometry on PMVK immunoprecipitates. The analysis revealed a specific interaction between PMVK and the transcription factor SP1 (Fig. 4A). Immunofluorescence confirmed that PMVK localizes to both the cytoplasm and the nucleus, with clear colocalization with SP1 in the nucleus of HCC cells (fig. S6A). Proximity ligation assay revealed significantly stronger interaction signals in HCC tissues than in normal liver tissues (fig. S6B). We further confirmed direct physical interaction using coimmunoprecipitation (co-IP), surface plasmon resonance (SPR), and glutathione S-transferase (GST) pulldown assays (Fig. 4, B to D, and fig. S6C). Notably, PMVK knockdown reduced SP1 phosphorylation without altering its protein levels (Fig. 4E). This phosphorylation was restored by re-expressing WT PMVK but not the kinase-dead (K22M) mutant (Fig. 4F). Most definitively, an in vitro kinase assay demonstrated that purified WT PMVK, but not the K22M mutant, directly phosphorylates SP1 (Fig. 4G). PMVK directly interacts and phosphorylates SP1 at T355. (A) Tandem mass spectrometry identification of SP1 phosphorylation sites from PMVK immunoprecipitates. (B) Endogenous coimmunoprecipitation (co-IP) demonstrating the interaction between PMVK and SP1 in HCC cells. (C) Surface plasmon resonance (SPR) sensorgrams showing the direct binding affinity between purified PMVK and SP1 proteins. (D) GST pulldown assay confirming the direct PMVK-SP1 interaction. (E) Endogenous co-IP and Western blot (WB) showing that PMVK knockdown reduces SP1 phosphorylation without affecting total SP1 levels. (F) Co-IP and WB analysis of PMVK-knockdown cells reconstituted with wild-type (WT) PMVK or the kinase-dead (K22M) mutant, demonstrating that SP1 phosphorylation requires PMVK’s catalytic activity. (G) In vitro kinase assay showing direct phosphorylation of SP1 by WT PMVK, but not the PMVK K22M mutant. (H) Co-IP mapping the PMVK interaction domain on SP1 using a series of FLAG-tagged SP1 truncations. (I) Co-IP and WB identifying the critical phosphorylation site by analyzing SP1 phospho-levels in cells expressing WT SP1 or the indicated threonine-to-alanine mutants. (J) Co-IP and confirming T355 as the primary phosphorylation site, showing that PMVK fails to phosphorylate the SP1 T355A mutant. Next, we explored the specific phospho-acceptor site on SP1. By testing a series of SP1 truncations, we mapped the PMVK-interaction domain to the D2 fragment (amino acids 251 to 459) (Fig. 4H and fig. S6D). Within this region, we systematically mutated three candidate threonine residues and identified Thr355 as the critical site, as its mutation to alanine (T355A) abolished PMVK-induced SP1 phosphorylation (Fig. 4I and fig. S6E). We validated this finding in HCC cells, where coexpression of PMVK with WT SP1, but not the SP1 T355A mutant, robustly increased SP1 phosphorylation (Fig. 4J). Then, we found that PMVK knockdown inhibited the phosphorylation at the Thr355 site, as detected by a custom pSP1(T355) antibody (fig. S6F). In summary, we demonstrate that PMVK directly binds to and phosphorylates the transcription factor SP1 at Thr355, a specific modification essential for PMVK’s oncogenic functions. SP1 drives HCC stemness and progression by reprogramming lipid metabolism We next sought to define the functional role of SP1 in HCC stemness and lipid metabolism. Consistent with a pivotal oncogenic role, SP1 knockdown significantly impaired multiple facets of tumor aggressiveness. It reduced the CD24+CD133+ CSC population, spheroid formation capacity, tumor-initiating frequency, and CSC frequency in vivo (fig. S7, A to C). Furthermore, this SP1 knockdown significantly reduced the viability of patient-derived HCC organoids (fig. S7D). Meanwhile, SP1 knockdown suppressed core malignant phenotypes, including proliferation, migration, and invasion (fig. S7, E to G). Critically, we found that SP1 is a central regulator of lipid metabolic reprogramming in HCC. SP1 knockdown led to a significant reduction in intracellular cholesterol, FAs, lipid droplet content, and the expression of key biosynthetic enzymes (fig. S7, H to J). Notably, SP1 knockdown did not affect PMVK expression, positioning SP1 downstream of PMVK in the oncogenic signaling cascade. In summary, these results establish SP1 as a crucial downstream effector of PMVK, driving hepatocarcinogenesis by orchestrating a lipid metabolic program essential for maintaining HCC stemness and promoting malignant progression. PMVK drives HCC stemness and progression through phosphorylation of SP1 at Thr355 To define the functional significance of SP1 Thr355 phosphorylation, we performed rescue experiments in PMVK-knockdown HCC cells by reconstituting with WT SP1 or a phosphomimetic mutant (SP1 T355D). Notably, the SP1 T355D mutant significantly restored the CD24+CD133+ CSC population and spheroid formation capacity, whereas WT SP1 had only a minimal effect (fig. S8, A and B). This functional rescue by SP1 T355D extended to the restoration of tumor-initiating capacity and CSC frequency in vivo, as measured by limiting dilution assays (Fig. 5A), and to the viability of patient-derived HCC organoids (Fig. 5B). Furthermore, the phosphomimetic mutant, but not WT SP1, markedly rescued the impairments in proliferation, colony formation, migration, and invasion caused by PMVK knockdown (fig. S8, C to E). These data establish that phosphorylation of SP1 at Thr355 is the critical mechanism by which PMVK drives HCC stemness and progression. We further examined the potential of PMVK, WT SP1, and the SP1 mutants (T355D and T355A) to drive CSC formation when overexpressed in CD24−CD133− non-CSCs. Overexpression of PMVK, WT SP1, or SP1 T355D mutant successfully induced CSC formation, with PMVK and SP1 T355D exhibiting the strongest effects. In contrast, overexpression of the SP1 T355A mutant failed to induce CSC formation (fig. S8, F to H). These results indicate that SP1 phosphorylation at Thr355 is critical for PMVK-mediated CSC induction. PMVK drives HCC stemness and progression via phosphorylating SP1 at T355. (A) Limiting dilution assay of PMVK-knockdown PLC cells reconstituted with WT SP1 or the phosphomimetic SP1 T355D mutant in NOD/SCID mice (n = 6 mice per group). (B) Viability of PMVK-knockdown HCC tumoroids reconstituted with WT SP1 or SP1 T355D. Scale bars, 50 μm. (C) Intracellular total cholesterol (T-CHO) and triglyceride (TG) levels in PMVK-knockdown HCC cells reconstituted with WT SP1 or SP1 T355D. (D) Western blot analysis of key lipid metabolism regulators in PMVK-knockdown HCC cells reconstituted with vector, WT SP1, or SP1 T355D. (E) Average CUT&Tag signal profiles and binding heatmaps WT SP1, SP1 T355A, and SP1 T355D at transcription start sites (TSS ±3 kb) in HCC stem cells. (F) Representative CUT&Tag tracks showing the occupancy of WT SP1, SP1 T355A, and SP1 T355D at the genomic loci of key lipid biosynthetic genes. (G) KEGG pathway enrichment analysis of genes whose promoter binding by SP1 was lost in the T355A mutant, as identified by CUT&Tag in HCC stem cells. (H) Luciferase reporter assays measuring the transcriptional activity of FASN, SCD, HMGCR, and SQLE promoters in cells expressing WT SP1, SP1 T355A, or SP1 T355D. (I) Bioluminescent imaging of orthotopic tumors from mice injected with PMVK-knockdown HCC cells reconstituted with WT SP1, SP1 T355A, or SP1 T355D. (J) Representative IHC staining images of PMVK, pSP1 (T355), and lipid metabolism enzymes (HMGCR, SQLE, FASN, and SCD1) in orthotopic tumor tissues from the indicated groups. Scale bars, 50 μm. Data represent three biological replicates [(A), (B), (C), and (H)]. In all statistical plots, data are presented as mean ± SEM. P values were determined using a two-tailed unpaired Student’s t test [(A), (B), (C), and (H)]. Given PMVK’s role in lipid metabolism (as shown in Fig. 2), we asked whether SP1 T355D could rescue the lipid-depleted phenotype. SP1 T355D overexpression, but not WT SP1, significantly restored intracellular cholesterol, triglyceride, and lipid droplets levels in PMVK-knockdown cells (Fig. 5C and fig. S8, I and J). Western blot analysis confirmed that SP1 functions downstream of PMVK and that the T355D mutant restores the expression of key cholesterol and FA biosynthetic enzymes (Fig. 5D). To elucidate the underlying transcriptional mechanism, we performed CUT&Tag in HCC stem cells expressing WT SP1, SP1 T355D, or a phosphorylation-deficient mutant (SP1 T355A). To elucidate the transcriptional mechanism, we performed CUT&Tag in HCC stem cells expressing WT SP1, a phosphomimetic mutant (SP1 T355D), or a phosphorylation-deficient mutant (SP1 T355A). We first found that SP1 T355D significantly enhanced SP1’s genomic DNA binding affinity, while the T355A mutant severely impaired it (Fig. 5E). Focusing on lipid metabolism, we demonstrated that SP1 T355D, but not T355A, specifically enriched at the promoters of key biosynthetic genes, including HMGCR, SQLE, FASN, and SCD (Fig. 5F). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the shared targets confirmed that this phosphorylation-dependent binding program is strongly associated with FA and steroid biosynthesis (Fig. 5G). We next confirmed the functional consequences of this enhanced binding. Electrophoretic mobility shift assays verified that SP1 T355D markedly increased protein-DNA complex formation with the promoters of HMGCR, SQLE, SCD, FASN, and SREBF1/2, while T355A disrupted it (fig. S9A). Consequently, SP1 T355D overexpression significantly boosted the transcriptional activity of these genes, an effect that was minimal with WT SP1 or the T355A mutant, as shown by luciferase reporter assays and quantitative polymerase chain reaction (fig. S9, B and C). The critical role of this PMVK-SP1(T355) axis was validated in vivo. In an orthotopic model, SP1 T355D, but not WT or T355A, rescued the tumor growth suppressed by PMVK knockdown (Fig. 5I and fig. S9D). IHC analysis of these tumors revealed a coordinated reduction in pSP1(T355), FASN, SCD, HMGCR, and SQLE upon PMVK knockdown, which was specifically restored by SP1 T355D (Fig. 5J and fig. S9E). The WNT pathway is critical for HCC stemness and progression. Previous studies have also shown that PMVK promotes WNT pathway activation (23). Knockdown of PMVK inhibits WNT pathway activity, as evidenced by reduced expression of β-catenin, c-Jun, and c-Myc. Notably, overexpression of the phosphomimetic SP1 T355D mutant partially restored the expression of these WNT pathway components in PMVK-knockdown HCC cells. These findings suggest that p-SP1 (Thr355) levels are partially associated with WNT signaling activation, indicating a potential cross-talk between the PMVK-SP1 axis and WNT signaling in HCC (fig. S9F). To explore the potential role of the PMVK-SP1 axis in other malignancies, we extended our analysis to prostate cancer. Analysis of TCGA data showed that PMVK expression was elevated in prostate cancer tissues compared to adjacent normal tissues (fig. S9G). In prostate cancer cells, PMVK knockdown specifically reduced phosphorylation of SP1 at Thr355 without affecting total SP1 protein levels (fig. S9H). Furthermore, PMVK knockdown significantly impaired sphere formation capacity, and this impairment was rescued by overexpression of the SP1 T355D mutant (fig. S9, I and J). Collectively, these data suggest that the PMVK-SP1 regulatory axis may be conserved across multiple cancer types in promoting stem-like properties. Last, to establish that lipid biosynthesis is the essential downstream effector, we knocked down HMGCR or SCD in SP1 T355D-overexpressing cells. This ablation reversed the protumorigenic effects of SP1 T355D, confirming that lipid synthesis is the primary functional output (fig. S10, A to F). In conclusion, these data delineate a coherent pathway in which PMVK-mediated phosphorylation of SP1 at Thr355 enhances its DNA binding affinity and transcriptional activity, driving a lipid biosynthetic program that is essential for HCC stemness and tumorigenesis. Phosphorylation of SP1 at Thr355 enhances nuclear SREBPs recruitment, transcriptionally activating key enzymes in cholesterol and FA biosynthesis SREBP1/2 are master transcription factors that regulate lipid biosynthesis genes expression and serve as critical hubs that integrate and branch out within comprehensive biological signaling networks, orchestrating a wide range of physiological and pathophysiological processes (15). SP1 has been reported as a cofactor for SREBP1/2 and recruits nuclear SREBP1/2 to SREs sequences, leading the transcription of a subset of SREBP1/2 targets (15, 25, 26). We investigated whether SP1 phosphorylation affects the affinity of SP1 and SREBP1/2 and the targets expression of SREBP1/2. Co-IP assays confirmed that WT SP1 interacts with nuclear SREBP1/2 (Fig. 6A), and this direct interaction was further confirmed using GST pulldown and SPR binding assays (Fig. 6, B and C). Then, we compared SP1 variants. Co-IP results showed that SP1 T355D significantly enhanced its interaction with nuclear SREBP1/2, while SP1 T355A showed weaker interaction compared to WT SP1 (Fig. 6D). GST pulldown and SPR binding assays further demonstrated that SP1 T355D exhibited higher affinity for nuclear SREBP1/2, while SP1 T355A almost did not bind with nuclear SREBP1/2 compared to WT SP1 (Fig. 6, E and F). Furthermore, CUT&Tag assays indicated that SREBPs’ genomic binding ability was disrupted in SP1 T355A–overexpressing LCSCs but enhanced in SP1 T355D–overexpressing cells (Fig. 6G). This increased binding was particularly evident in the promoter regions of key lipid biosynthesis genes, including HMGCR, SQLE, FASN, and SCD (Fig. 6H). GO analysis of cotargets regulated by SP1 T355D and nuclear SREBP1/2 showed significant involvement in lipid metabolism and HCC progression (Fig. 6I). In conclusion, phosphorylation of SP1 at Thr355 acts as a cofactor of nuclear SREBP1/2, facilitating their recruitment, leading to the transcriptional activation of genes regulating cholesterol and FA biosynthesis, thus promoting HCC stemness and progression. Phospho-SP1 (T355) promotes lipid biosynthesis by enhancing interaction with nuclear SREBP1/2. (A) Endogenous co-IP demonstrating the interaction between SP1 and nuclear SREBP1/2 (nSREBP1/2) in HCC cells. (B) GST pulldown assay confirming the direct interaction between purified SP1 and nSREBP1/2 proteins. (C) Surface plasmon resonance (SPR) analysis quantifying the binding affinity between SP1 and nSREBP1/2. (D) Co-IP assay comparing the interaction of nSREBP1/2 with wild-type (WT) SP1, phosphodeficient (T355A), or phosphomimetic (T355D) SP1 mutants. (E) GST pulldown assay assessing the direct binding strength between nSREBP1/2 and the indicated SP1 mutants. (F) SPR assay quantifying the binding affinity of nSREBP1/2 for WT SP1, SP1 T355A, and SP1 T355D. (G) Average nSREBP1/2 CUT&Tag signals profiles (left) and binding heatmaps (right) for nSREBP1/2 at transcription start sites (TSS ±3 kb) in HCC stem cells expressing WT SP1, SP1 T355A, or SP1 T355D. (H) Representative CUT&Tag tracks showing nSREBP1/2 occupancy at the genomic loci of key lipid biosynthetic genes in HCC stem cells expressing the indicated SP1 variants. (I) GO analysis of genes cobound by the SP1 T355D mutant and nSREBP1/2, showing significant enrichment for lipid metabolic processes. Elevated PMVK and SP1 T355 phosphorylation predict poor outcome in patients with HCC To evaluate the clinical relevance of our findings, we analyzed the expression of PMVK and phospho-SP1 (T355) in a cohort of 112 HCC patients and 37 paired adjacent normal tissues by IHC (Fig. 7A). We found that both PMVK and phospho-SP1 (T355) were significantly up-regulated in HCC tissues compared to normal counterparts (Fig. 7B). Critically, high levels of either PMVK or phospho-SP1 (T355) were strongly associated with shorter overall survival in patients with HCC (Fig. 7C). Furthermore, elevated expression of both proteins correlated with aggressive clinicopathological features, including advanced tumor stage, higher TNM classification, and increased microvascular invasion (tables S1 and S2). Both univariate and multivariate Cox regression analyses identified PMVK and pSP1(T355) as independent prognostic factors for poor survival (tables S3 and S4). Elevated PMVK and SP1 T355 phosphorylation are clinically associated with poor prognosis in HCC. (A) Representative immunohistochemistry (IHC) images of PMVK, phospho-SP1 (T355), and key lipid enzymes (HMGCR, SQLE, FASN, and SCD) in human HCC and paired adjacent normal tissue (ANT). Scale bars, 50 μm. (B) Quantification of IHC scores for PMVK and pSP1(T355) in paired adjacent normal (n = 37) and HCC (n = 112) tissues. (C) Kaplan-Meier analysis of overall survival in the HCC cohort stratified by high or low expression of PMVK or pSP1(T355). (D) Correlation analysis (Pearson’s) demonstrating positive associations between PMVK expression levels and the levels of SP1, pSP1(T355), and its downstream lipid metabolic targets in human HCC specimens. P values were determined using a two-tailed unpaired Student’s t test (B) and log-rank test (C). To confirm the activity of the PMVK-SP1 axis in human tumors, we performed correlation analysis, which revealed a significant positive association between the levels of PMVK, phospho-SP1 (T355), and their key downstream targets (HMGCR, SQLE, FASN, and SCD) in patient specimens (Fig. 7D). In summary, these clinical data firmly establish the PMVK-pSP1(T355) axis as a powerful driver of HCC aggressiveness and a promising independent prognostic biomarker, underscoring its translational relevance. The SP1 inhibitor mithramycin A suppresses HCC stemness, progression, and PMVK-driven lipid biosynthesis Our work identifies the PMVK-SP1 axis as a critical driver of HCC, suggesting that its components are promising therapeutic targets. Although a specific PMVK inhibitor (PMVKi5) has been reported (23), targeting the transcription factor SP1 offers broader potential, as it integrates oncogenic signals from both PMVK and other metabolic processes (17, 20, 27, 28). Therefore, we evaluated the therapeutic efficacy of mithramycin A (MTA), a well-characterized SP1 inhibitor, in HCC models. MTA treatment potently inhibited HCC malignancy across multiple functional assays. In vitro, MTA dose-dependently reduced the CD24+CD133+ CSC population and impaired spheroid formation (Fig. 8, A and B, and fig. S11, A and B). It also suppressed proliferation, migration, and invasion (fig. S11, C to E). This efficacy extended to in vivo and ex vivo models, where MTA significantly suppressed tumor growth in mouse xenografts and reduced the viability of patient-derived tumoroids (Fig. 8, C to E). The SP1 inhibitor mithramycin A suppresses HCC tumorigenesis and PMVK-driven lipid biosynthesis. (A) Flow cytometry analysis of the CD24+CD133+ CSC population in control or PMVK-overexpressing HCC cells treated with mithramycin A (MTA). (B) Representative images and quantification of spheroid formation in control or PMVK-overexpressing HCC cells treated with MTA. Scale bars, 200 μm. (C) Representative images of subcutaneous tumor xenografts from nude mice injected with PLC cells (control or PMVK-overexpressing) and treated with vehicle or MTA. (D) Quantification of tumor volume (left) and tumor weight (right) from the xenograft experiment in (C). (E) Viability of patient-derived HCC tumoroids treated with MTA. Scale bars, 50 μm. (F) Western blot analysis of key lipid biosynthetic enzymes in control or PMVK-overexpressing HCC cells treated with MTA. (G) Intracellular total cholesterol (T-CHO) and triglyceride (TG) levels in control or PMVK-overexpressing HCC cells treated with MTA. Data represent three biological replicates [(A), (B), and (E) to (G)]. In all statistical plots, data are presented as mean ± SEM. P values were determined using a two-tailed unpaired Student’s t test [(A), (B), (D), and (G)]. We next determined whether MTA’s action specifically counteracts the PMVK-SP1 axis. MTA treatment effectively reversed the enhanced stemness, proliferation, and invasion conferred by PMVK overexpression (Fig. 8, A to E, and fig. S11, A to E), confirming that SP1 is the essential downstream mediator of PMVK’s oncogenic effects. Last, we established that disruption of lipid metabolism is a key mechanism underlying MTA’s efficacy. MTA treatment significantly reduced intracellular cholesterol and FA levels (Fig. 8F). Critically, it also abolished the prolipogenic effect of PMVK overexpression, concurrently suppressing the expression of key lipid biosynthetic enzymes and reducing lipid droplet accumulation (Fig. 8, F and G, and fig. S11F). We also determined the effect of MTA on PMVK expression and found that MTA did not affect PMVK expression, suggesting that SP1 inhibition preferentially targets tumors with elevated SP1 expression (fig. S12A). To assess the safety profile of MTA, we examined its potential off-target effects on major organs. Histological analysis revealed no significant abnormalities in liver, heart, lung, kidney, or spleen following MTA treatment, indicating minimal off-target toxicity (fig. S12B). Furthermore, we investigated the regulatory relationship between PMVK and SREBP1/2. CUT&Tag revealed that SREBP1/2, but not SP1, directly bound to the PMVK promoter (fig. S12C). Consistently, overexpression of nuclear SREBP1/2 (nSREBP1/2) up-regulated PMVK expression, whereas SP1 overexpression had no effect (fig. S12C). These findings suggest the existence of a feed-forward regulatory circuit in HCC CSCs: PMVK-p-SP1-SREBP activation-PMVK up-regulation. On the basis of this model, we hypothesized that dual targeting of PMVK and SP1 could overcome potential compensatory up-regulation of SREBP activity. Combined treatment with PMVK shRNA and the SP1 inhibitor MTA exhibited the most potent inhibition of HCC stemness (fig. S12, D to G). In conclusion, targeting the PMVK-SP1-lipid metabolism axis effectively suppresses HCC tumorigenesis by highlighting a promising translational strategy for a therapeutically challenging cancer. To further validate that SP1 inhibition suppresses HCC stemness and progression by targeting PMVK-driven lipid synthesis, we performed rescue experiments by supplementing MTA-treated HCC cells with cholesterol or MVAPP. Cholesterol supplementation significantly reversed the inhibitory effects of MTA on HCC stemness and progression, whereas MVAPP supplementation did not (fig. S13). These findings confirm that PMVK drives HCC stemness and progression in an MVAPP-independent manner. DISCUSSION Metabolic reprogramming is a well-established hallmark of cancer, yet the molecular drivers that link specific metabolic pathways to HCC stemness and progression remain incompletely defined (29, 30). Through a metabolism-focused CRISPR screen, we identified PMVK as a critical regulator of HCC. While PMVK is canonically known to catalyze the production of MVA-5PP in the MVA pathway, our work reveals a noncanonical, kinase-dependent mechanism that is largely independent of this metabolic product. We found that MVA-5PP supplementation only minimally rescued the tumor-suppressive effects of PMVK knockdown, whereas cholesterol fully restored HCC stemness and progression. This positions PMVK as a driver of hepatocarcinogenesis through a mechanism that extends beyond its traditional role in the MVA pathway. Many metabolic enzymes, including PKM2, PGK1, HK, PCK1, and TK1, have been reported to function as protein kinases, phosphorylating substrates involved in gene expression, metabolic reprogramming, tumor progression, and immune evasion (31). However, this moonlighting activity remains controversial, primarily regarding experimental rigor and physiological relevance. In terms of experimental rigor, observed kinase activity may stem from contamination with authentic kinases; cocrystallization is required to confirm substrate binding, yet some studies have failed to reproduce PKM2’s kinase activity (32, 33). Regarding physiological relevance, the substrate selection mechanisms by which metabolic enzymes function as protein kinases remain incompletely understood. The regulatory networks governing these activities are complex and require more systematic investigation, and the functions may vary across cell types and tissues. Moreover, the protein kinase activities observed in vitro may lack significance or biological relevance under physiological conditions in vivo. For example, PKM2 is dispensable for tumor cell proliferation, its inactive state marks proliferating tumor cells, whereas nonproliferating cells require active pyruvate kinase (34). These findings highlight the need to consider the functional changes of metabolic enzymes in the in vivo context, and to carefully evaluate the safety of targeting them therapeutically. Our in vitro kinase assays were performed using purified recombinant PMVK and SP1 proteins, which minimizes the likelihood of contaminating kinases. The PMVK kinase-dead mutant abolished SP1 phosphorylation, strongly suggesting that PMVK’s intrinsic kinase activity is required. The positive correlation between PMVK expression and p-SP1(T355) levels in HCC tissues, without changes in total SP1, is consistent with PMVK acting as a direct upstream kinase. A cocrystal structure of PMVK bound to SP1 should be further determined to define how a target amino acid in SP1 is presented or accommodated in the active site for phosphate transfer. A substantial challenge in targeting lipid metabolism for cancer therapy is the compensatory cross-talk between FA and cholesterol biosynthesis pathways; inhibiting one often up-regulates the other, limiting efficacy and potentially driving resistance (13, 35). Our discovery that PMVK coordinately regulates both cholesterol and FA synthesis is therefore of considerable therapeutic interest. By acting as a central node for dual lipid control, PMVK inhibition presents a strategy to simultaneously suppress both arms of lipid biosynthesis, potentially overcoming the limitations of targeting rate-limiting enzymes like HMGCR alone. This positions PMVK as a compelling target not only for HCC but potentially for other lipid-associated pathologies, including metabolic dysfunction–associated liver disease. Mechanistically, we unraveled how PMVK achieves this coordinated regulation. We identified the transcription factor SP1 as a direct phosphorylation target of PMVK at Thr355. While SP1 is known to influence lipid genes and can interact with SREBPs (26, 36–40), its role in the lipid reprogramming of LCSCs was unknown. Our work demonstrates that PMVK-mediated phosphorylation of SP1 at Thr355 acts as a molecular switch, markedly enhancing SP1’s DNA binding affinity and its recruitment of nuclear SREBP1/2 to the promoters of key lipogenic genes. This PMVK-SP1-SREBP axis drives a transcriptional program essential for de novo lipogenesis and HCC tumorigenesis. This finding not only elucidates a key mechanism of metabolic rewiring in LCSCs but also substantially advances our understanding of how specific phosphorylation events can dictate the oncogenic function of SP1. Therapeutically, these findings highlight the promise of disrupting the PMVK-SP1 axis. We demonstrate that the SP1 inhibitor MTA effectively suppresses HCC stemness, tumor growth, and the protumoral lipid metabolism driven by PMVK. This confirms the functional importance of this axis and suggests its potential as a therapeutic vulnerability. In summary, our study delineates a signaling pathway in which PMVK, through the direct phosphorylation of SP1 at Thr355, co-opts the SREBP transcriptional machinery to drive a lipid biosynthetic program that fuels HCC stemness and progression. Targeting the PMVK/phospho-SP1 (T355) axis represents a promising therapeutic strategy to disrupt the metabolic underpinnings of HCC. Limitations of the study While our findings demonstrate that SP1 T355 phosphorylation enhances its genomic DNA binding activity and affinity for nuclear SREBP1/2, the precise structural mechanism by which this single modification elicits these effects remains to be determined. Future structural biology studies, such as crystallography or cryo–electron microscopy of the protein complexes, will be essential to visualize how Thr355 phosphorylation alters SP1’s conformation and interaction interfaces. MATERIALS AND METHODS Human tissue specimens A total of 112 human HCC tissues and 37 adjacent normal liver tissues were collected from the Fifth Affiliated Hospital of Sun Yat-sen University for IHC analysis. Informed consent was obtained from all participants, and ethical approval was granted by the Research Committee of The Fifth Affiliated Hospital, Sun Yat-sen University (approval no. K126-1). Human CRISPR metabolic gene knockout library screen and data analysis The human CRISPR metabolic gene knockout library (#110066) was obtained from Addgene. Amplification of the libraries was performed using Endura cells (catalog no. 60242, Lucigen) and the Gemini X2 system (BTX), following the manufacturer’s protocol. The workflow of the CRISPR gRNA pooled screen is depicted in Fig. 1A. The library plasmids were packaged into lentivirus and used to infect PLC cells at a multiplicity of infection of 0.2, ensuring effective barcoding of individual cells. To preserve the representation of sgRNAs during the screen, the number of cells was maintained at least 1000 times the number of sgRNAs in the library. After 48 hours of transduction, the cells were selected using puromycin (2.5 μg/ml) for 7 days to establish a mutant cell pool. CD24+CD133+ LCSCs and CD24−CD133− non-CSCs populations were then sorted using FACS. Genomic DNA from the sorted cells was extracted using the Quick-DNA Microprep Plus Kit (catalog no. D4074, Zymo Research) according to the manufacturer’s protocol. Sequencing libraries were prepared and sequenced by Genewiz. The sequencing data were processed and analyzed using MAGeCKFlute software (41). Mice model Four- to 6-week-old male NOD/SCID mice and BALB/c Nude mice were obtained from the Shanghai Model Organisms Center Inc. The mice were housed in an Individually Ventilated Caging system in a specific pathogen–free (SPF) grade animal room. They were fed a standard laboratory diet with access to food and water ad libitum. All animal experiments were approved by the Animal Ethics Committee of the Fifth Affiliated Hospital of Sun Yat-Sen University (approval no. 2022052501) and conducted in accordance with NIH Guidelines. To assess tumorigenicity, 1 × 104 or 5 × 104 PLC cells mixed with Matrigel (catalog no. 356231, Corning) were injected subcutaneously into 5- to 6-week-old male NOD/SCID mice (n = 6 per group). At least eight mice were included in each experimental cohort. Mice were euthanized when tumor size reached 1.0 to 1.5 cm in diameter. The tumor-initiating frequency was determined using the Extreme Limiting Dilution Analysis software available at http://bioinf.wehi.edu.au/software/elda/. For the xenograft assay, 5 × 106 cancer cells were suspended in Matrigel and injected subcutaneously into 5- to 6-week-old BALB/c male nude mice (n = 6 per group). Once the tumor size reached 0.5 cm, the mice were randomly divided into three treatment groups and fed with dimethyl sulfoxide and mithramycin (0.2 mg/kg per day; catalog no. T5530, TargetMol). Tumor growth was monitored by measuring tumor length (L) and width (W) every 2 days, and the tumor volume (V) was calculated using the formula: V = L × W2 × 0.5. For the orthotopic transplantation model, HCC cells were transplanted into the livers of 5- to 6-week-old male BALB/c-nude mice (n = 6). Subsequently, 28 days after transplantation, the tumor growth capacity in the liver was observed using bioluminescence, and the mice were then euthanized. For the hydrodynamic transfection model, 4- to 5-week-old male C57BL/6J mice (n = 6 per group) were used. sgp53 and Myc overexpression plasmids along with SB transposase were diluted in sterile Ringer’s solution with a volume equal to 10% of body weight. The mixture was injected via the tail vein within 5 to 7 s. sgp53 and Myc overexpression and Pmvk overexpression plasmids along with SB transposase were injected using the same protocol. Mouse livers were collected approximately 4 to 5 weeks after injection to evaluate HCC development. Statistical analysis and reproducibility Data analysis was conducted using GraphPad Prism 8 and SPSS version 19.0 (IBM, USA). Kaplan-Meier survival curves were plotted to evaluate overall survival and disease-free survival, and statistical differences were assessed with the log-rank test. Differences between groups were assessed using Student’s t test for two-group comparisons and analysis of variance (ANOVA) for multiple groups. The relationship between PMVK levels and clinical characteristics of patients was evaluated using the Pearson’s χ2 test. Univariable and multivariable regression analyses were performed to determine the prognostic factors for survival outcomes. All experiments were performed in triplicate, and data were reported as mean ± SD. Statistical significance was determined at P < 0.05.