Protein Phosphatase 1 Regulatory Subunit 3C integrates cholesterol metabolism and isocitrate dehydrogenase in chondrocytes and neoplasia

Significance

Enchondromas, common cartilage tumors driven by IDH1/2 mutations, can progress to chondrosarcoma. We identified PPP1R3C as a key regulator linking glycogen and cholesterol metabolism in IDH-mutant chondrocytes. PPP1R3C was shown to modulate the neoplastic phenotype in vitro and in vivo. The level of cholesterol and the activation of intracellular cholesterol synthesis regulate PPP1R3C differently, suggesting potential synergistic and antagonistic combination therapies in IDH-mutant tumors. We also report on genetically modified mice that can be used to study neoplastic processes caused by Idh mutations.

Abstract

Enchondromas are common bone tumors composed of chondrocytes originating from growth plate cells which can progress to malignant chondrosarcoma. Mutations in the genes encoding isocitrate dehydrogenase (IDH1 and IDH2) are identified in a large proportion of these tumors. IDH enzymes convert isocitrate to alpha-ketoglutarate (α-KG), an essential component of the citric acid cycle. While mutant IDH enzymes produce 2-hydroxyglutarate, which has epigenetic effects important in tumor initiation, cell maintenance and growth rely on additional factors. Prior work shows that intracellular cholesterol and glycogen are upregulated in mutant IDH chondrocytes. Here, we show that Protein Phosphatase 1 Regulatory Subunit 3C (PPP1R3C, previously termed Protein Targeting to Glycogen or PTG) is highly expressed in chondrocytes harboring a mutant IDH. Furthermore, Sterol Regulatory Element-Binding Proteins (SREBPs), transcriptional regulators of sterol biosynthesis, regulate PPP1R3C expression. We found that PPP1R3C regulates glycolysis and glycolytic capacity in chondrocytes. Depletion of PPP1R3C in mouse chondrocytes in vivo suppresses the neoplastic phenotype. The growth plate phenotype associated with the genetic inhibition of cholesterol biosynthesis is partially rescued by PPP1R3C overexpression. Taken together, our data show that PPP1R3C integrates cholesterol metabolism and isocitrate dehydrogenase in growth plate and neoplastic chondrocyte metabolism by regulating intracellular glycogen levels.

Growth plate chondrocytes undergo a tightly regulated differentiation program. Dysregulation of this differentiation process by somatic mutations in chondrocytes can result in enchondromas which are cartilaginous tumors at the ends of long bones. These tumors can cause deformity, pain, and progress to malignant chondrosarcomas. Chondrocytes in the growth plate, enchondromas, and chondrosarcomas utilize glycogen, promoting tumor cell growth (1). Cholesterol is also known to play an important role in these cell types, regulating longitudinal bone growth (2), enchondroma formation, and sarcoma cell viability (3).

Studies in the liver show that cholesterol biosynthesis can stimulate glycogen synthesis (4), but the mechanism connecting intracellular cholesterol biosynthesis and glycogen is not known. Protein Phosphatase 1 Regulatory Subunit 3C (PPP1R3C), previously termed Protein Targeting to Glycogen, or PTG, could mediate the accumulation of glycogen in this context. This gene encodes a carbohydrate-binding protein that is a subunit of the protein phosphatase 1 (PP1) complex. PPP1R3C functions primarily to regulate glycogen biosynthesis by activating glycogen synthase and reducing glycogen phosphorylase activity limiting glycogen breakdown (5–10). Overexpression of PPP1R3C increases basal and glycogen synthesis in cell lines, human muscle cell cultures, and in mice (6, 9–11). Depleting Ppp1r3c in mice rescues the phenotype of the glycogen-storage disorder, Lafora disease (12, 13). PPP1R3C might play a role in neoplasia, as it is differentially expressed in some cancer cell lines (14–17).

Somatic mutations in IDH1 and IDH2, the genes encoding isocitrate dehydrogenase proteins, are present in the majority of enchondromas and in about half of chondrosarcomas (18–20). The IDH genes encode for enzymes involved in the Krebs cycle, and mutant IDH enzymes produce a novel product, D-2-hydroxyglutarate (D2-HG) at the expense of alpha-ketoglutarate (α-KG) (21). IDH1 and 2 reside in the cytoplasm and mitochondria respectively. Mutations in these genes were initially identified in gliomas and leukemia (22, 23). IDH1 and IDH2 mutations in tumors are heterozygous, because their wild-type activities are essential for cellular respiration and metabolic function (18, 19, 22, 23) and IDH1 Arg132, IDH2 Arg172, and Arg140 account for the majority of alterations. D2-HG is sometimes called an “oncometabolite” (21), and has epigenetic effects related to histone and DNA hypermethylation; stabilization of Hif-1α protein; impairing cellular differentiation; increasing cell proliferation; and increasing the expression of stem cell markers (24–27). In chondrocytes, expression of a mutant IDH also increases intracellular glycogen, and inhibiting glycogen synthesis suppresses enchondroma formation in mice (1).

Here, we investigated the role of PPP1R3C in glycogen accumulation in chondrocytes and its regulation by genes controlling intracellular cholesterol biosynthesis.

Results

IDH Mutations Increase PPP1R3C Expression in Growth Plate Chondrocytes and Chondrosarcoma.

To investigate how mutant IDH affects the regulation of glycogen in chondrocyte development, we generated conditional mutant IDH1 (IDH1R132C) and IDH2 (IDH2R172S) transgenic mice in which the mutant allele was inserted in the Rosa26 locus. These are frequent mutations observed in chondrosarcomas (18, 28). We crossed these mice with Col2a1Cre transgenic mice to activate IDH1R132C (R26^IDH1R132C) or IDH2R172S (R26^IDH2R172S) expression in chondrocytes. In these animals, the expression of the mutant IDH protein (IDH1R132C or IDH2R172S) is activated by Cre-mediated excision of LoxP-STOP-LoxP (LSL) sequence (Fig. 1A). We found that D2-HG levels were significantly elevated in the IDH mutant growth plate chondrocytes isolated from embryonic day 18.5 (E18.5) animals (Fig. 1B). These mice developed an identical fetal phenotype with that found in mice expressing the mutant Idh1R132Q allele with an expansion of the growth plate cartilage and type X collagen (Col X)-producing chondrocytes (Fig. 1C) (1, 20, 29). Similar to animals expressing the mutant Idh1R132Q allele, they exhibited perinatal lethality (2/9 mice in IDH1R132C; 22.2%, 4/7 mice in IDH2R172S: 57.1%) and dwarfism with a short tail and dysplasia of the respiratory system at 4 wk (SI Appendix, Fig. S1). Enchondroma-like lesions were observed in these animals at 3 mo, similar to data in which the Idh1R132Q allele was driven by Col2a1Cre/ERT2 (Fig. 1D)(20).

Fig. 1.

PPP1R3C is upregulated in IDH mutant chondrocytes and deleting PPP1R3C partially rescues the aberrant differentiation of chondrocytes caused by mutant IDH. (A) A schematic representation of the targeting vector containing the LoxP-flanked STOP (LSL) cassette, the human IDH mutation (IDH1R132C or IDH2R172S), and IRES-eGFP is shown. The LSL cassette is composed of the neomycin resistance gene, a splicing acceptor sequence that stops transcription at the insertion point. Although the presence of LSL inhibits the expression of the mutant IDH (IDH1R132C or IDH2R172S) protein, Cre-mediated excision of LSL allows the expression of the protein. (B) D2-HG levels measured by LC–MS–MS in growth plate cartilage harvested from Cre (−), Col2Cre; R26^IDH1R132C, and Col2Cre; R26^IDH2R172S genotypes at E18.5. Each data point represents an individual animal; one-way ANOVA. (C) Representative images of Col X staining in femurs at E18.5 from the three genotypes. (D) Representative Safranin O staining images of distal femurs (Top) and proximal tibias (Middle) from the three genotypes at 3 mo. The Bottom figures show proximal tibia growth plates at a higher magnification (×10). (E) (Top) Representative in situ hybridization images with the Ppp1r3c probe in proximal tibias at E17.5. (Bottom) Higher magnification (×120) of the resting chondrocytes (RC), proliferating chondrocytes (PC), prehypertrophic chondrocytes (Pre-HC), and hypertrophic chondrocytes (HC) for each genotype. (F) Western blot analysis of PPP1R3C in deidentified patient-derived primary cell cultures. A total of six different cell cultures were used, including 711 (IDH1R132C), 713 (IDH1R132C), 725 (IDH1R132H), 731 (IDH2R172S), 740 (IDH2R172G), and 743 (IDH Wt). (G) D2-HG levels of 743 parental cells (IDH Wt), Dox inducible 743−Flag−IDH1R132C, and 743−IDH2R172M−Flag cells. LC–MS–MS measured the levels after Dox 1 µg/mL treatment for 72 h; one-way ANOVA. (H) Relative expression of the PPP1R3C gene in Dox inducible 743−Flag−IDH1R132C, and 743−IDH2R172M−Flag cells treated with Dox 1 µg/mL for 72 h; Student’s t test. (I) Western blot analysis of PPP1R3C and Flag (mutant IDH1/2) in Dox inducible 743−Flag−IDH1R132C and 743−IDH2R172M−Flag cells treated with Dox 1 µg/mL for 72 h. Vinculin was used as a loading control. The abbreviation is as follows: mIDH, mutant IDH. (J) Intracellular glycogen levels of Dox inducible 743−Flag−IDH1R132C and 743−IDH2R172M−Flag cells treated with Dox 1 µg/mL for 72 h; Student’s t test. (K) (Top) Representative images of PAS staining in proximal tibias at E18.5 from six genotypes. (Bottom) Higher magnification (×120) of the RC (Upper Left), PC (Upper Right), Pre-HC (Lower Left), and HC (Lower Right) for each genotype. (L) Quantification of PAS-stained area normalized to a total number of cells in growth plate chondrocytes of proximal tibia; one-way ANOVA. (M) Representative images of Col X staining in proximal tibias at E18.5 from six genotypes. (N) Quantification of the length of Col X-positive area; one-way ANOVA. (O) Representative images of Alcian blue staining in tibias at E18.5 from six genotypes. (P) Quantification of the length of the tibia (Top), the percentage of proximal growth plate relative to the total tibia (Lower Left), and the percentage of ossified bone relative to the total tibia (Lower Right); one-way ANOVA. (Q) Representative skeletal prep images of the sternal manubrium of the mice at 4 wk from six genotypes. (R) Quantification of the Alcian blue positive area in sternal manubrium; one-way ANOVA. (S) Representative images of Safranin O/fast green staining in proximal tibial epiphysis at 4 wk from six genotypes. Yellow arrows show a representative enchondroma lesion. (T) Quantification of the percentage of secondary ossification center (SOC) in proximal tibial epiphysis; one-way ANOVA. The abbreviation is as follows: SOC, secondary ossification center. (U) Representative images of Safranin O/fast green staining in the distal femoral (Top) and proximal tibial (Bottom) growth plate at 6 mo from four genotypes. Large enchondroma-like lesions are observed in Col2a1Cre/ERT2; R26^IDH1R132C and Col2a1Cre/ERT2; R26^IDH2R172S animals. (V) Quantification of the area (Left) and number (Right) of enchondroma-like lesions in the distal femoral and proximal tibial growth plate at 6 mo from six genotypes; one-way ANOVA. Each data point represents an individual animal. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

We then investigated the expression of Ppp1r3c in the growth plate chondrocytes. Through in situ hybridization of proximal tibias, we found that Ppp1r3c was mainly expressed in the round cell/RC, PC, and Pre-HC of control animals. Mutant IDH increased the expression, especially in the PC and Pre-HC (Fig. 1E). Ppp1r3c was also highly expressed in Col2Cre; Idh1R132Q knock-in embryos (SI Appendix, Fig. S2).

Previous studies in chondrosarcoma found that glycogen accumulation and glycolysis were increased in IDH mutant chondrosarcomas (1, 30). Published microarray data from a chondrosarcoma cohort showed that PPP1R3C expression was 40% higher in IDH1 mutant chondrosarcomas (IDH1 mutant vs. IDH wild-type, P = 0.019) (31). Western blot analysis in patient-derived primary chondrosarcoma cells revealed that PPP1R3C expression was high in IDH mutant cells (Fig. 1F). To further validate the increase of PPP1R3C expression with mutant IDH, we generated Doxycycline (Dox)-inducible IDH1 (IDH1R132C) or IDH2 (IDH2R172M) mutant overexpressing cells using a primary patient-derived wild-type IDH chondrosarcoma cell. After 72 h of Dox treatment, the levels of D2-HG increased significantly along with the expressions of Flag-tagged mutant IDH1 or IDH2 (Fig. 1 G and H). Both mutant IDH1 and mutant IDH2 increased the levels of PPP1R3C and glycogen (Fig. 1 H–J). In contrast, the mutant IDH1 inhibitor, AG-120, significantly reduced the expression of PPP1R3C (SI Appendix, Fig. S3).

Deletion of Ppp1r3c Partially Rescues the Growth Plate Phenotype Caused by a Mutant IDH1 or IDH2.

We then investigated the growth plate phenotype in the IDH1 and IDH2 mutant fetal mice lacking Ppp1r3c. PAS staining was positive in the proximal tibia growth plate of E18.5 control mice (Fig. 1K). In IDH1 and IDH2 mutant mice, staining was more prominent in the PC and Pre-HC regions, indicating increased glycogen accumulation (Fig. 1 K and 1 M). We then focused on the function of Ppp1r3c by crossing Ppp1r3c^fl/fl mice (32) with Col2a1Cre, Col2a1Cre; R26^IDH1R132C and Col2a1Cre; R26^IDH2R172S animals, conditionally deleting Ppp1r3c in chondrocytes. Col2a1Cre; Ppp1r3c^fl/fl chondrocytes showed decreased expression of Ppp1r3c in growth plate chondrocytes and reduced glycogen accumulation, confirming efficient deletion of PPP1R3C in these mice (Fig. 1 E, K, and M). Ppp1r3c deletion decreased the PAS staining intensity in the PC and Pre-HC of the IDH mutant animals. (Fig. 1 K and M).

Examination of fetal mice expressing mutant IDH1 or IDH2 (Col2a1Cre; R26^IDH1R132C and Col2a1Cre; R26^IDH2R172S) exhibited an increase in the area of Col X staining and growth plate cartilage (Fig. 1 C, L, and N). Deleting the Ppp1r3c gene reduced the extended Col X staining area in IDH mutant mice but had no effect on the phenotype in wild-type IDH mice. (Fig. 1 L and N). The Col2a1Cre; R26^IDH1R132C and Col2a1Cre; R26^IDH2R172S fetal limbs were shorter than controls, with a higher proportion of cartilage compared to bone (Fig. 1 O and P). In contrast, Col2a1Cre; R26^IDH1R132C; Ppp1r3c^fl/fl and Col2a1Cre; R26^IDH2R172S; Ppp1r3c^fl/fl embryos showed almost the same limb length and proportion of cartilage as control animals not expressing a mutant IDH (Fig. 1 O and P). Animals with IDH mutation have a disordered columnar structure of chondrocytes similar to hypoxic chondrocytes, which was not rescued by Ppp1r3c deletion (SI Appendix, Fig. S4).

At 4 wk, the dwarfism caused by the mutant IDH1/2 was partially rescued by deleting Ppp1r3c (SI Appendix, Fig. S5 A–D). Additionally, in IDH mutant mice, there was a delay in the secondary ossification of the sternal manubrium and ramal epiphysis at 4 wk, but deleting Ppp1r3c partially improved this (Fig. 1 Q and R and SI Appendix, Fig. S5 E and F). Safranin O/ Fast green staining of the normal proximal tibia revealed that the epiphyseal region was mostly occupied by trabecular bone, and the formation of the SOC was almost completed (Fig. 1S). However, delayed development of SOC was observed in IDH mutant mice, which was partially improved by the deletion of Ppp1r3c (Fig. 1 S and T). Although IDH mutant mice had a disordered columnar structure of the proximal tibial growth plate, this was partially restored by deleting Ppp1r3c (SI Appendix, Fig. S5G). Moreover, to examine the impact of deleting Ppp1r3c in postnatal mutant IDH enchondroma-like lesion formation, we generated mice with induced mutant IDH1/2 and Ppp1r3c deletions simultaneously through tamoxifen administration in Col2a1 expressing cells. Enchondroma-like lesion formed in 6 mo in IDH mutant mice (Fig. 1U). However, the absence of Ppp1r3c resulted in a reduction in the size and number of the lesions (Fig. 1 U and V). Overall, these results indicate that PPP1R3C regulates glycogen synthesis in growth plate chondrocytes, and that this process is partially responsible for the IDH mutant enchondroma and growth plate phenotype.

PPP1R3C Mediates IDH Mutant Chondrosarcoma Tumor Growth.

To investigate the effect of PPP1R3C in chondrosarcoma, an IDH mutant chondrosarcoma cell line, JJ012 (IDH1R132G) (33), was investigated. Dox-inducible PPP1R3C overexpressing (JJ012−HA−PPP1R3C) cells and PPP1R3C deletion cells were generated (Fig. 2 A and F). Intracellular glycogen levels and cell viability correlated with the level of PPP1R3C expression (Fig. 2 B, C, G, and H and SI Appendix, Fig. S6 C and D). Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured to examine the metabolic profile in various chondrosarcoma cell lines and primary patient cells (SI Appendix, Fig. S6 A and B). The results demonstrated a high ECAR in IDH1 mutant cells, consistent with data that showed that IDH mutant chondrosarcoma uses glycogen and glycolysis as an energy source (1, 30). PPP1R3C overexpression increased glycolysis and glycolytic capacity (Fig. 3 D and E), whereas PPP1R3C gene depletion significantly impaired them, which is similar to the effect of a pharmacological blockade of glycogen utilization with CP-91149 in a prior study (1) (Fig. 2 I and J and SI Appendix, Fig. S6 E and F). We then transfected a Dox-inducible shRNA plasmid (nontargeting control or PPP1R3C) into a primary chondrosarcoma cell culture harboring the IDH1 mutation (IDH1R132H). The knockdown efficacy of the three shRNAs with Dox at a concentration of 1 µg/mL for 72 h was nearly identical (SI Appendix, Fig. S6G). As a result, these cells were mixed in equal proportions, totaling 1.0 × 10⁶ cells, and then injected subcutaneously into NOD scid gamma null mice (Fig. 2K). Continuous Dox diet administration significantly impaired tumor progression and reduced tumor weight with PPP1R3C knockdown, without any obvious side effects in the mice such as body weight loss (Fig. 2 L–N and SI Appendix, Fig. S6H). Lower PPP1R3C expression and glycogen levels were confirmed in the harvested tumors (Fig. 2 O and P). Additionally, histological analysis showed a decrease in Ki67 positive cells with PPP1R3C knockdown, suggesting a reduction in proliferation (Fig. 2Q). These findings indicate that IDH mutant chondrosarcoma relies on glycogen as a source of energy for tumor growth. Moreover, our results suggest that PPP1R3C could be an effective therapeutic target for treating this tumor type. In support of this concept, we reanalyzed published data from a cohort of human chondrosarcomas (31), finding a 5-y survival advantage of 18% for patients with low PPP1R3C expression.

Fig. 2.

PPP1R3C regulates glycogen synthesis and utilization in IDH mutant chondrosarcoma, playing a crucial role in tumor progression. (A) Western blot analysis of HA (PPP1R3C) in Dox inducible JJ012−HA−PPP1R3C cells treated with Dox 0.01, 0.1, and 1 µg/mL for 24 h. Vinculin was used as a loading control. (B) Intracellular glycogen levels of Dox inducible JJ012−HA−PPP1R3C cells treated with Dox 1 µg/mL for 24 h; Student’s t test. (C) The relative cell viability of Dox inducible JJ012−HA−PPP1R3C measured by CellTiter-Glo 2.0 3 and 6 d after Dox 1 µg/mL treatment; two-way ANOVA. (D) ECAR measured in Dox inducible JJ012−HA−PPP1R3C cells when challenged by 1) glucose (fuel for glycolysis), 2) oligomycin (an ATP synthase blocker), and 3) 2-DG (an inhibitor of glycolysis). Dox 1 µg/mL treatment was performed for 24 h before the measurement (n = 8 per group). (E) Quantification of glycolysis (Left) and glycolytic capacity (Right) in Dox inducible JJ012−HA−PPP1R3C cells. Glycolysis represents the resting state and glycolytic capacity defines the total ability of the cell to perform glycolysis when mitochondria is compromised by oligomycin; Student’s t test. (F) Western blot analysis of PPP1R3C protein extracted from knockout JJ012 cells with CRISPR/Cas9. Vinculin was used as a loading control. The abbreviation is as follows: ko, knockout. (G) Intracellular glycogen levels of PPP1R3C-knockout JJ012 cells. The original LentiCRISPRv2 vector was transfected into JJ012 cells as a control, without sgRNA targeting PPP1R3C; Student’s t test. (H) The relative cell viability of PPP1R3C-knockout JJ012 cells by CellTiter-Glo 2.0; two-way ANOVA. (I) ECAR measured in PPP1R3C-knockout JJ012 cells (n = 4 per group). (J) Quantification of glycolysis (Left) and glycolytic capacity (Right) in PPP1R3C-knockout JJ012 cells; Student’s t test. (K) Experimental schematic of the in vivo study. A total of 1.0 × 10⁶ cells (725: IDH1R132H) transduced with Doxycycline-inducible shRNA (nontargeting control or PPP1R3C) were suspended in 100 μL of 50% Matrigel prepared in PBS and subcutaneously inoculated into the left flank of 6-wk-old female mice. The 625 mg/kg Doxycycline diet was fed continuously starting at 3 d after transplantation. (L) Representative images of the control tumors and the tumors in which PPP1R3C was knocked down. (M) Tumor volume in mice subcutaneously transplanted with 725 cells transduced with Doxycycline-inducible shRNA (nontargeting control or PPP1R3C) (n = 8 per group); two-way ANOVA. (N) Tumor weight in mice subcutaneously transplanted with 725 cells transduced with Doxycycline-inducible shRNA (nontargeting control or PPP1R3C). Each data point represents an individual animal; Student’s t test. (O) Western blot analysis of PPP1R3C protein extracted from the harvested tumor cells. (P) Intracellular glycogen levels of the harvested tumors. Each data point represents an individual animal; Student’s t test. (Q) Representative images of the immunohistochemistry of Ki67 (Left) and quantification of the percentage of Ki67-positive cells (Right). Each data point represents an individual animal; Student’s t test. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

Fig. 3.

ChIP-seq data reveal that SREBF2 directly regulates PPP1R3C. (A) ChIP-qPCR of SREBF2 at the locus of the PPP1R3C gene in JJ012 cells. Background signals in each region were evaluated by ChIP using rabbit IgG (green); Student’s t test. Three primers were designed around the SREBF2 binding site located near TSS (#1), −1 kb from TSS (#2), and −6 kb from TSS (#3), respectively. (B) Western blot analysis of HA (active form of SREBF2) in Dox inducible HT1080−HA−SREBF2 cells treated with Dox 0.01, 0.1, and 1 µg/mL for 24 h. For parental HT1080 cells, only the pLVX-EF1a-Tet3G vector was transduced. Vinculin was used as a loading control. (C) Total cholesterol levels in HT1080 parental control cells and Dox inducible HT1080−HA−SREBF2 cells treated with Dox 1 µg/mL for 72 h; Student’s t test. The abbreviation is as follows: OE, overexpression. (D) Relative expression of cholesterol pathway target genes and PPP1R3C from HT1080 parental control cells and Dox inducible HT1080−HA−SREBF2 cells treated with Dox 1 µg/mL for 72 h; Student’s t test. The abbreviation is as follows: OE, overexpression. (E) ChIP-seq analysis of HA (exogenous SREBF2) in Dox inducible HT1080−HA−SREBF2 cells and parental control cells treated with Dox 1 µg/mL for 72 h and immunoprecipitated with HA antibodies. Data are presented as normalized tag counts of HA peaks; Student’s t test. (F) Tornado plots displaying HA peaks within ± 2 kb of the transcription start site (TSS). (G) Representative Integrated Genomics Viewer (IGV) diagrams showing the distribution of ChIP-seq read densities for HA peaks at the locus of PPP1R3C promoter in Dox inducible HT1080−HA−SREBF2 cells (red) and parental control cells (black) treated with Dox 1 µg/mL for 72 h. (H) (Top) A schematic representation of the primer pairs and the SREBF2 binding sites at the locus of the PPP1R3C promoter region. Three primers were designed around the SREBF2 binding site located near TSS (#1), −1 kb from TSS (#2), and −6 kb from TSS (#3), respectively. (Bottom) ChIP-qPCR of HA at the locus of the PPP1R3C gene in Dox inducible HT1080−HA−SREBF2 cells (red) and parental control cells (black) treated with Dox 1 µg/mL for 72 h. Background signals in each region were evaluated by ChIP using rabbit IgG (green); one-way ANOVA. (I) Relative cell viability of JJ012 cells treated with different concentrations of Lovastatin for 72 h. The Y-axis shows the luminescence measured by CellTiter-Glo 2.0 Cell Viability Assay as a percentage relative to DMSO-treated cells. Data are expressed as the mean of triplicates (±SD). (J) Relative expression of SREBF2, SCAP, and PPP1R3C genes from JJ012 cells treated with Lovastatin 1 µM for 48 h; one-way ANOVA. (K) ChIP-qPCR of SREBF2 at the locus of the PPP1R3C gene in JJ012 cells treated with Lovastatin 1 µM for 48 h (blue) or DMSO (black). Background signals in each region were evaluated by ChIP using rabbit IgG (green); one-way ANOVA. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

To determine the consequences of PPP1R3C in chondrosarcoma in an unbiased manner, RNA sequencing was performed in chondrosarcoma cell lines expressing a doxycycline-inducible PPP1R3C overexpression construct, or a cell line infected with a lenti-CRISPR single-guide RNA (sgRNA) targeting PPP1R3C. Overexpression increased the expression of genes implicated in metabolic pathways, while depletion regulated genes implicated in cell adhesion and proliferation. (SI Appendix, Fig. S7) Expression of genes implicated in metabolic pathways is consistent with the function of PPP1R3C and changes in cell adhesion and proliferation are likely responsible for how PPP1R3C regulates tumor cell behavior.

SREBF2 Regulates PPP1R3C in Chondrocytes and Chondrosarcoma.

To determine whether intracellular cholesterol biosynthesis regulates PPP1R3C expression, we analyzed the Ppp1r3c mouse promoter sequence (34). Using the GPMiner tool (35), we identified several putative sterol regulatory element-binding factor (SREBF) binding sites. SREBF binding sites around PPP1R3C promoter region were also found in the public ChIP-seq database of SREBF2 in other human cell types (36, 37). SREBF transcription activates cholesterol biosynthesis, and the major SREBF in chondrocytes is SREBF2 (2). Our previous study shows that cholesterol biosynthesis in mesenchymal cells regulates long bone growth and chondrocyte homeostasis (2). Furthermore, mutant IDH1 promotes cholesterol biosynthesis in growth plate chondrocytes and chondrosarcoma (3). ChIP-qPCR revealed that SREBF2 was recruited around the PPP1R3C promoter region in chondrosarcoma cell lines, JJ012 (IDH1R132G) and SW1353 (IDH2R172S) (Fig. 3A and SI Appendix, Fig. S8A). To further study the transcriptional regulation of PPP1R3C, we generated Dox-inducible HA-SREBF2 overexpressing chondrosarcoma cells. We first confirmed the exogenous band of the active form of SREBF2 in the Western blot after 24 h of Dox treatment (Fig. 3B and SI Appendix, Fig. S8B). Overexpression of SREBF2 led to increased cholesterol levels (Fig. 3C and SI Appendix, Fig. S8C) and enhanced the expression of target genes in the cholesterol pathway, such as LDLR and HMGCR, while also promoting the expression of PPP1R3C (Fig. 3D and SI Appendix, Fig. S8D). To investigate whether this is a direct effect, we undertook a ChIP-seq analysis using a Human influenza hemagglutinin (HA)—tagged SREBF2 antibody, enabling us to utilize HA antibodies' specificity and affinity for exogenous SREBF2. We observed an increase in normalized tag counts of HA peaks after Dox treatment (Fig. 3E), and the peaks increased significantly within ±2 kb around the TSS (Fig. 3F). IGV showed that HA-SREBF2 binding was highly enriched at PPP1R3C promoters and several target genes, including HMGCR, LDLR, mevalonate kinase (MVK), and fatty acid synthase (FASN) (Fig. 3G and SI Appendix, Fig. S8E). These findings are consistent with previous ChIP-seq analysis of SREBF2 occupancy in other cell types (36, 37) (SI Appendix, Fig. S8F). ChIP-qPCR confirmed that HA-SREBF2 was recruited around the PPP1R3C promoter region after Dox treatment (Fig. 3H and SI Appendix, Fig. S8G).

To further validate the interaction between SREBF2 and PPP1R3C, we used a pharmacological approach with Lovastatin, a 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitor. Lovastatin treatment suppressed cell viability with low concentrations, similar to our previous study (Fig. 3I) (3). In addition, two primary chondrosarcoma cell cultures harboring the IDH1 mutation showed greater sensitivity to lower concentrations of Lovastatin compared to IDH WT primary cells (SI Appendix, Fig. S8A). SREBF2 expression was elevated 48 h after 1 µM of Lovastatin treatment due to feedback from the depletion of sterols (Fig. 3J and SI Appendix, Fig. S9 B and C) (38). In this situation, the PPP1R3C expression also significantly increased, and ChIP-qPCR revealed that SREBF2 was recruited more around the PPP1R3C promoter region with lovastatin treatment (Fig. 3 J and K and SI Appendix, Fig. S9 B and C). These results support the notion that SREBF2 directly regulates PPP1R3C and that there is a strong connection between cholesterol biosynthesis and glycogenesis.

PPP1R3C Partially Rescues Aberrant Chondrogenic Differentiation Caused by a Lack of Intracellular Cholesterol Biosynthesis.

To determine whether the regulation of PPP1R3C by intracellular cholesterol biosynthesis would alter glycogen or the growth plate phenotype associated with disordered intracellular cholesterol biosynthesis, we next examined mice we previously generated in which intracellular cholesterol biosynthesis was inhibited (2). SREBF cleavage-activating protein (SCAP) regulates SREBF activity in response to low intracellular cholesterol (38). Col2a1Cre; Scap^fl/fl embryos had severe dwarfism with ectopic hypertrophic cells and exhibited perinatal lethality (2). These animals exhibited a disrupted columnar structure in the proliferating zone (PZ) and a limited number of HC at E17.5, while Col2a1Cre; Ppp1r3c^OE; Scap^fl/fl embryos partially restored the columnar structure and increased the number of HC (Fig. 4 A and B). This resulted in a higher proportion of the bone compared to chondrocytes in the tibia despite no change in tibia length (Fig. 4 C and D). PAS staining revealed that Ppp1r3c overexpression increased glycogen accumulation and overcame the deficient levels caused by Scap deletion (Fig. 4E).

Fig. 4.

Overexpression of PPP1R3C partially ameliorates the aberrant chondrogenic differentiation caused by cholesterol biosynthesis inhibition. (A) Representative images of Alcian blue and Col X staining in proximal tibias at E17.5 from four genotypes. (B) Quantification of the length of the Col X-positive zone relative to the area of the growth plate. Each data point represents an individual animal; one-way ANOVA. (C) Representative images of Alcian staining in tibias at E17.5 from four genotypes. (D) Quantification of the length of the tibia (Left), the percentage of proximal growth plate relative to the total tibia (Middle), and the percentage of ossified bone relative to the total tibia (Right). Each data point represents an individual animal; one-way ANOVA. (E) (Top) Representative images of PAS staining in proximal tibias at E17.5 from four genotypes. (Middle) Higher magnification (×120) of the prehypertrophic chondrocytes for each genotype. (Bottom) Quantification of PAS-stained area normalized to a total number of cells. Each data point represents an individual animal; one-way ANOVA. (F) Relative expression of Scap gene from ATDC5 cells and Dox-inducible ATDC5−HA−PPP1R3C cells treated with Dox 1 µg/mL for 72 h. Both cells were transduced with Dox-inducible shRNA (nontargeting control or Scap); one-way ANOVA. The abbreviation is as follows: OE, overexpression. (G) (Left) Western blot analysis of HA (PPP1R3C) in Dox-inducible ATDC5−HA−PPP1R3C cells treated with Dox 0.01, 0.1, and 1 µg/mL for 24 h. Vinculin was used as a loading control. (Right) Intracellular glycogen levels of Dox-inducible ATDC5−HA−PPP1R3C cells treated with Dox 1 µg/mL for 72 h; Student’s t test. (H) Representative images of Alcian blue staining in ATDC5 cells (sh control, sh Scap, sh Scap & PPP1R3C^OE) taken at 1, 2, 3, and 4 wk after chondrocyte differentiation with 1% ITS. Chondrocyte differentiation started 3 d after the cells were seeded. Scap knockdown and Ppp1r3c overexpression were induced simultaneously with Dox 1 µg/mL treatment. (I) Relative expression of chondrogenic-related genes from ATDC5 cells (sh control, sh Scap, PPP1R3C^OE, sh Scap & PPP1R3C^OE) at each time point. For each gene, the baseline expression level of the sh control (ITS−) after 1 wk of chondrocyte differentiation was set to 1; two-way ANOVA. (J) Relative expression of SCAP and PPP1R3C genes from JJ012 cells transduced with shRNA (nontargeting control or SCAP); one-way ANOVA. The abbreviation is as follows: ctrl, control. (K) The relative cell viability of JJ012 cells transduced with shRNA (nontargeting control or SCAP) and Dox-inducible HA−PPP1R3C measured by CellTiter-Glo 2.0. The cell viability of the sh control group after 3 d of Dox treatment was set to 1; two-way ANOVA. (L) Relative expression of SREBF1, SREBF2, LDLR, and PPP1R3C genes from JJ012 cells treated with different concentrations of Fatostatin for 48 h; one-way ANOVA. (M) Relative cell viability of JJ012 cells treated with different concentrations of Fatostatin for 48 h. The values are represented as percent of viable cells where the vehicle-treated cells were regarded as 100%; two-way ANOVA (Šidák). (N) A schematic diagram illustrating the overall pathway linking cholesterol biosynthesis and glycogen, mediated by PPP1R3C, in mutant IDH growth plate chondrocytes and mutant IDH chondrosarcoma. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.0001.

To validate these findings in another system, ATDC5 cells were utilized. The cells were modified with Dox-inducible shRNA constructs, which targeted either nontargeting control or Scap, followed by Dox-inducible HA-PPP1R3C. Subsequently, chondrogenic differentiation was initiated, and Dox treatment was commenced simultaneously. The chondrogenic differentiation of the cells was then observed for 4 wk. Initially, we confirmed reduced expression of Scap and overexpression of PPP1R3C with Dox treatment (Fig. 4 F and G). Glycogen levels were also found to be elevated with PPP1R3C overexpression (Fig. 4G). Control cells steadily increased proteoglycan production detected by Alcian blue staining, while Scap knockdown considerably reduced the positive area and intensity (Fig. 4H and SI Appendix, Fig. S10B). Conversely, PPP1R3C overexpression partially restored the Alcian blue staining, although not as strong as the control (Fig. 4H and SI Appendix, Fig. S10 A and B). The expression of chondrogenic-related genes, including Sox9, Col2a1, Aggrecan, and Col10a1, was also decreased with Scap knockdown, but increased with PPP1R3C overexpression (Fig. 4I). Together, these results demonstrate that cholesterol biosynthesis regulates PPP1R3C, and its overexpression partially restores the aberrant chondrocyte differentiation caused by Scap deletion.

We then investigated whether overexpressing PPP1R3C can restore cell viability suppressed by inhibition of cholesterol biosynthesis in chondrosarcoma cells. Dox-inducible PPP1R3C overexpressing cells (JJ012−HA−PPP1R3C) were modified with shRNA constructs, targeting nontargeting control or SCAP. We confirmed overexpression of PPP1R3C with Dox treatment and reduced expression of SCAP (Figs. 2A and 4J). Knockdown of SCAP resulted in the decreased PPP1R3C expression (Fig. 4J). SCAP knockdown significantly reduced the cell growth, and PPP1R3C overexpression overcame the inhibitory effect of SCAP deletion (Fig. 4K).

To determine how IDH mutations and SREBF activity interact in chondrosarcoma, we next used Fatostatin, an inhibitor of SCAP and thus SREBF transcriptional activity (39, 40) as well as Lovastatin, which inhibits HMG-CoA reductase enzyme, lowers cholesterol by inhibiting Mevalonate production, but increasing SREBF transcriptional activity as part of a feedback loop. Treatment with Fatostatin reduced PPP1R3C expressions and decreased cell viability in a dose-dependent manner (Fig. 4 L and M). However, when PPP1R3C was overexpressed, we saw a partial reversal of the effect of Fatostatin (Fig. 4M). We obtained similar results with another cell line, HT1080 (SI Appendix, Fig. S11). This is in contrast to treatment with Lovastatin, which blocks the production of Mevalonate, an early step in cholesterol production upstream of the effect of SREBF. As a feedback loop, this upregulates SREBF transcriptional activity (3, 41) and PPP1R3C expression as well (Fig. 3).

We next determined how cholesterol metabolism modulating drugs and a drug targeting the effect of a mutant IDH might interact in chondrosarcoma cell viability. Studies of combination therapy, could shed light into the relative contributions of cholesterol, glycogen, and 2-HG on cell viability. We used AG-120, a mutant IDH targeting drug in use in patients. When combining Fatostatin and AG-120, We observed a modest synergistic effect in targeting cell viability in two cell lines compared to using Fatostatin alone. However, Lovastatin, which increases PPP1R3C expression, had an opposing effect to AG-120 alone (SI Appendix, Fig. S12). This confirms our overall findings, and suggests that combining Lovastatin with a 2-HG inhibitor will not have synergistic effects in clinical use.

Taken together, these data show a connection between cholesterol biosynthesis and glycogen mediated by PPP1R3C. In the context of the growth plate chondrocytes, part of the phenotypic effect of cholesterol dysregulation can be rescued by PPP1R3C, and in the context of IDH mutant chondrosarcomas, PPP1R3C can integrate the effects of cholesterol and glycogen in the regulation of tumor cell growth (Fig. 4N).

Discussion

Here, we show that PPP1R3C mediates glycogen accumulation in growth plate chondrocytes. Its expression is elevated in chondrocytes harboring a mutant IDH and SREBF regulates its expression. Since SREBF regulates PPP1R3C, this mechanism acts as a link between intracellular cholesterol biosynthesis and glycogen accumulation. Such a mechanism likely plays a more generalized role in a variety of situations in which SREBF is transcriptionally active.

We generated two new Idh mutant mice. Both developed a similar phenotype, showing that mutations in either Idh1 or Idh2 can cause enchondromas. They also have a similar developmental phenotype and generate similar levels of 2-HG, Ppp1r3c, and glycogen, showing that mutant isocitrate dehydrogenase enzymes in the cytoplasm and mitochondria can have a similar function in metabolic regulation.

It has been known for decades that glycogen is present in growth plate chondrocytes (42), but its function in this context has only recently been studied. Inhibition of glycogen synthesis by genetic deletion of Gys1 in growth plate chondrocytes is not reported to cause an embryonic phenotype, but there is a phenotype when Gys1 is depleted in growth plates also harboring a mutant IDH (1). This suggests that under physiologic conditions there can be compensation from other energy sources to allow normal growth without glycogen, and our data from Ppp1r3c deletion are consistent with this notion.

The regulation of Ppp1r3c by genes implicated in cholesterol biosynthesis may explain why supplementing chondrocytes lacking intracellular cholesterol biosynthesis with exogenous cholesterol does not rescue their phenotype (2). In this instance, cholesterol supplementation would not alter SREBF activity, and thus PPP1R3C expression and glycogen levels would not be altered. Such a mechanism may play a role in other contexts as well. For instance, a correlation between cholesterol biosynthesis and glycogen level was found in the liver (4, 43).

Our data raise a potential limitation in the use of Lovastatin in the treatment of chondrosarcoma. This drug inhibits HMG-CoA reductase enzyme activity, which blocks the production of Mevalonate, an early step in cholesterol production. Because cholesterol levels fall with Lovastatin treatment, SREBF2 expression increases as part of a feedback loop. Thus, Lovastatin treatment increases expression of SREBF2 target genes including PPP1R3C (44). This suggests that Lovastatin therapy is limited in its ability to treat chondrosarcoma as it will not target intracellular glycogen, which plays a role in chondrosarcoma cell energetics. This concept is supported by our data showing that adding a statin with a 2-HG inhibitor will increase chondrosarcoma cell viability over the use of a 2-HG inhibitor alone. This is in contrast to Fatostatin treatment, where PPP1R3C expression is directly inhibited.

PPP1R3C is differentially expressed in some tumor types (14–17). Glycogen is also implicated as an energy source in some cancers (45). As such, targeting glycogen might be developed as a therapeutic approach. Since there are multiple mechanisms that might regulate glycogen levels, a central regulatory mechanism would be an enticing therapeutic target. Thus, an approach to downregulate PPP1R3C, perhaps by modulating SREBF activity which can be targeted by several drugs already in use in patients, is a potential approach in select tumor types.

Materials and Methods

Animals.

All animals were used according to the approved protocol by the Institutional Animal Care and Use Committee at Duke University. Mice used in this study include Ppp1r3c^fl/fl mice (32), overexpressing Ppp1r3c (Ppp1r3c^OE) mice (46), Scap^fl/fl mice (JAX:004162) (2, 47), Idh1R132Q-KI (Idh1^LSL/+) mice (20), R26^IDH1R132C mice, R26^IDH2R172S mice, Col2a1Cre mice (JAX:003554) (48), Col2a1Cre/ERT2 mice (29, 49), and interleukin-2 receptor gamma chain (gamma)-null NOD/SCID (NSG) mice. Idh1^LSL/+ mice bear a conditional knock-in of the point mutation Idh1R132Q, as previously clarified (20). Regarding R26^IDH1R132C and R26^IDH2R172S mice, the targeting vector containing the LoxP-flanked STOP (LSL) cassette, the IDH mutation (IDH1R132C or IDH2R172S), and IRES-eGFP is shown in Fig. 1A. Cre-mediated excision of LSL allows the expression of the protein. Since the Ppp1r3c expression cassette was introduced into the Hprt locus in the X chromosome, and to avoid variability due to female X chromosome inactivation (46, 50), the studies related to this mouse were conducted in male animals. All the mice studied were littermates.

Cell Lines and Culture.

The following chondrosarcoma cell lines were used: HT1080 (IDH1R132C, RRID: CVCL_0317), JJ012 (IDH1R132G, RRID: CVCL_D605), SW1353 (IDH2R172S, RRID: CVCL_0543), and OUMS27 (IDH Wt, RRID: CVCL_3090). HT1080 and SW1353 cells were obtained from the American Type Culture Collection (ATCC; Manassas, VA), and OUMS27 was obtained from the Japan Health Science Research Resources Bank (Osaka, Japan). The cells were cultured in DMEM supplemented with 10% FBS. The ATDC5 cell line (RRID: CVCL_3894, MilloporeSigma, Burlington, MA, US), a mouse chondrogenic cell line, was incubated in DMEM/F-12 medium (Gibco, Waltham, MA) supplemented with 5% FBS. To induce chondrogenic differentiation, subconfluent cultures were incubated in a medium containing 1% Insulin-Transferrin-Selenium (ITS) (Gibco). Patient-derived chondrosarcoma primary cell cultures were derived from histologically confirmed intermediate and high-grade chondrosarcoma. Cell origin was authenticated and mycoplasma-free cells were used in all experiments (Mycostrip, Invivogen, San Diego, CA).

D2-HG Measurement and Glycogen Quantification.

D2-HG was analyzed by liquid chromatography with tandem mass spectrometry (LC–MS–MS) at Duke PK/PD Core Laboratory. Growth plate cartilages and chondrosarcoma cells were lysed in RIPA Lysis and Extraction Buffer (Thermo Scientific) with Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Scientific). After adding 2-HG-²H₄ (internal standard), the sample was dried under nitrogen and derivatized by (+)-O-O′-diacetyl-L-tartaric anhydride (DATAN) for measurement. Glycogen levels were measured using Glycogen Assay Kit II (Abcam, Cambridge, MA) following the manufacturer’s protocol. Glycogen was also detected in the sections of growth plates using a PAS Kit (#395B, Sigma-Aldrich, St. Louis, MO) according to the manufacturer’s instructions (Standard procedure).

Histological Analysis.

Bone histomorphometry was performed on hindlimbs fixed in 10% neutral-buffered formalin (NBF), followed by decalcification with 14% EDTA (Sigma). Skeletons were embedded in paraffin and sectioned. Safranin O/Fast green and Alcian blue staining were performed. Quantification was done using the image processing software QuPath and ImageJ. To determine the size of the SOC, the area containing bone or marrow but no cartilage was measured.

In Situ Hybridization and Immunohistochemistry.

The paraffin sections were deparaffinized and rehydrated, and then fixated with 4% PFA. The sections were incubated with hybridization buffer with a Digoxigenin-labeled RNA probe (Ppp1r3c) at 58 °C overnight. Sections were developed with BM Purple at room temperature until color developed. Immunohistochemistry was performed on 5 μm paraffin-sectioned limbs. For type X collagen, antigen retrieval was performed by 1 mM EDTA incubation at 85 °C for 15 min and hyaluronidase digestion at 10 mg/mL (Sigma) at 37 °C for 30 min. For Ki67 staining, antigen retrieval was performed by citrate-based antigen unmasking solution for 20 min. The specimens were blocked with 2% horse serum at room temperature for 30 min, followed by incubation with antibodies for Col X (1:500, Thermo Fisher Scientific, 14-9771-82) or Ki67 (1:500, Abcam, ab15580) overnight at 4 °C.

Alizarin Red and Alcian Blue Whole-Mount Skeletal Staining.

Whole-mount skeletal staining was performed according to the previous report (51). Sample images were captured by Zeiss Lumar V12 stereomicroscope, and quantification was performed with ImageJ.

Quantification of Enchondroma-Like Lesions.

Tamoxifen was administered daily for 10 d at 100 mg/kg body weight/day via intraperitoneal injection starting at 4 wk of age, and mice were killed at 6 mo of age. After decalcification, skeletons were embedded in paraffin and sectioned at 5 µm thickness. Enchondroma-like lesions were first identified by Safranin O staining, which was performed on one slide (2 sections, 5 µm/section) in every 10 consecutive slides (10 µm). We then selected one slide for each genotype to be evaluated, which is close to the PCL attachment of the femur. The area and number of lesions were quantified using the image processing software Fiji ImageJ (29). Animals of both sexes were used for analysis. All analyses were undertaken by an observed blinding to the genotype.

RNA Isolation and qRT-PCR.

RNA was isolated using the Direct-zol RNA Miniprep Plus Kit (Zymo Research, Irvine, CA). Total RNA was reverse-transcribed to synthesize cDNA using the iScript Reverse Transcription (RT) Supermix (Biorad). qRT-PCR was performed using QuantStudio 6 Flex cycler (Applied Biosystems) with TaqMan Fast Advanced Master Mix (Applied Biosystems) or SsoAdvanced Universal SYBR Green Supermix (Biorad). The following probes/primers were used: ACTB (Hs03023943), PPP1R3C (Hs01921501), SREBF1 (Hs01088680), SREBF2 (Hs01081784), SCAP (Hs00378725), LDLR (Hs01092524), Actb (Mm01205647), Scap (Mm01250176), Sox9 (Mm00448840), Col2a1 (Mm01309565), Acan (Mm00545794), Col10a1 (Mm00487041), and HMGCR Fwd, CCCCTCTCCAGGTGTTCACA, and Rev, AATTGAGGTAGGTTTCATAGAGATGCT. All genes were assayed in triplicate.

RNA Sequencing and Expression Profile analysis.

RNA sequencing was undertaken in triplicate. Messenger RNA was purified from total RNA using poly-T oligo-attached magnetic beads. After fragmentation, the first strand cDNA was synthesized, followed by the second strand cDNA synthesis using dTTP. A cDNA library was prepared after end repair, A-tailing, adapter ligation, size selection, amplification, and purification. The libraries were sequenced (150 bp paired-end reads) using an Illumina NovaSeq 6000 sequencer at Novogene (Beijing, China). Read mapping was conducted with HISAT2 (v2.0.5) using the Homo sapiens (human) genome assembly GRCh38 (hg38) as the reference. The number of reads mapped to each gene was quantified using featureCounts (v1.5.0-p3). TMM normalization and differential gene expression analysis were performed using the R packages edgeR (v4.2.2) and limma (v3.60.2), respectively. Gene set enrichment analysis was performed with the R package fgsea (v1.30.0) and the gene sets were imported with msigdbr (v7.5.1). Genes were ranked based on values obtained by multiplying the negative log of the P-value from the differential gene expression analysis by the sign of the log fold change. Enrichment scores for Hallmark, KEGG, and Gene Ontology gene sets were then calculated. RNA sequencing data have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession GSE286965. For analysis of published microarray information, the rdata normalized using the RMA algorithm with the “oligo” package (52). For survival analysis, samples were grouped into high/low expression based on median gene expression using the “survminer” package.

Western Blotting.

Western analysis was undertaken using anti-Vinculin (MAB3574; Millipore); anti-PPP1R3C (PA5-112237; Invitrogen); anti-DYKDDDDK Tag (2368; Cell Signaling); or anti-HA High Affinity (11867423001; Roche) antibodies.

Cell Viability Assay.

Cell viability was evaluated using CellTiter-Glo 2.0 Cell Viability Assay (Promega). Luminescence was then measured using Spark (TECAN, Männedorf, Switzerland).

Bioenergetic Measurements.

XF Glycolysis Stress Kit (Agilent, Santa Clara, CA) and XF Mito Stress Kit (Agilent) were used, respectively. Metabolic flux analysis was performed using the Seahorse XFe96 Analyzer (Agilent), and the outputs were recorded as ECAR and OCR. Basal ECAR (glyco ATP) was calculated as ECAR_Glucose − ECAR_2DG when glucose was given as fuel. Basal OCR (mito ATP) was taken as the average of the first three measurements subtracted by the OCR measurements after rotenone and antimycin D injections.

Knockout with CRISPR/Cas9.

To deplete the PPP1R3C gene, we used a Lentinus to infect three different single-guide RNA (sgRNA) targeting PPP1R3C (exon 2: CTGTCTTAGGAGACGTCTGG) generated by the Duke Core Facility. The LentiCRISPRv2 (Addgene plasmid # 52961; https://n2t.net/addgene:52961; RRID: Addgene_52961) (53) vector expressing sgRNA was generated from supernatants after transfection in Lenti-X 293 T cells (Takara). Chondrosarcoma tumor cells were then infected with lentiviruses for 24 h. Deletion was validated by western blot analysis against PPP1R3C.

Doxycycline-Inducible Overexpression.

Lenti-X Tet-On 3G Inducible Expression System with EF1a regulator vectors was obtained from Takara. pDONR223_IDH1_p.R132C and pDONR223_IDH1_p.R172M were gifts from Jesse Boehm & William Hahn & David Root (Addgene plasmid # 81726; https://n2t.net/addgene:81726; RRID:Addgene_81726) (54). The full-length SREBF2 and PPP1R3C cDNAs were purchased from GenScript (Piscataway, NJ) and Origene (Rockville, MD), respectively. Flag-IDH1R132C, IDH2R172M-Flag, HA-SREBF2, and HA-PPP1R3C constructs were synthesized and inserted into lentiviral vectors downstream of the TRE3G doxycycline-inducible promoter. The pLVX-EF1a-Tet3G vector was used to express the Tet-On 3G transactivator protein from the human EF1 alpha promotor. Cells were sequentially transduced with lentivirus and selected with G418 (pLVX-EF1a-Tet3G) and puromycin (Flag-IDH1R132C, IDH2R172M-Flag, HA-SREBF2, PPP1R3C-HA).

ShRNA.

shERWOOD UltramiR Lentiviral shRNA plasmids (pZIP-TRE3G-ZsGreen-Blast) targeting SCAP (TLHSU1419) and shERWOOD UltramiR Lentiviral Inducible shRNA plasmids (pZIP-TRE3G-ZsGreen-Blast) targeting Scap (TLMSU2331) were obtained from Transomic Technologies (Huntsville, AL). Nontargeting plasmids (TLNSU4419 and TLNSU4501) were used for negative controls. SMARTvector Inducible Lentiviral shRNA plasmids (hEF1a-TurboGFP-Puro) targeting PPP1R3C (V3SH11252) were purchased from Horizon Discovery (Cambridge, UK). A nontargeting plasmid (SVC17010402) was used for negative control. Lentivirus was produced by transfecting Lenti-X 293 T cells with the shRNA plasmids and packaging plasmids (psPAX2 and pMD2.G). Cells were then infected with lentiviruses for 24 h using Polybrene (final concentration of 10 µg/mL) in high-glucose DMEM supplemented with 10% FBS and 1% P/S. Infected cells were selected with blasticidin or puromycin 48 h after transduction.

Alcian Blue Staining for ATDC5 Cells.

ATDC5 cells were seeded in 12-well plates at a density of 5 × 10⁴ cells/well. After 3 d, the cells were differentiated in a medium containing 1% ITS for 7, 14, 21, and 28 d. Dox 1 µg/mL treatment was also started simultaneously on day 3. The cells were stained with 0.3% Alcian blue 8GX (Sigma-Aldrich) in 0.1 N HCl. For quantitative analyses, the amount of extracted dye was measured after extraction with 200 µl of 6 M guanidine HCl (#SRE0066, Sigma-Aldrich) for 6 h at room temperature. Optical densities were measured at 620 nm using a microplate reader.

Measurement of Cholesterol Levels.

Total cholesterol levels were measured in chondrosarcoma cells using the Cholesterol/Cholesteryl Ester Quantitation Assay Kit (ab65359, Abcam) following the manufacturer’s protocol.

Chromatin Immunoprecipitation (ChIP) Assay.

ChIP assay was performed using the SimpleChIP Plus Sonication Chromatin IP Kit (Cell Signaling Technology) according to the manufacturer’s instructions with modification. The following antibodies were used for immunoprecipitation: anti-HA (ab9110; Abcam), anti-SREBF2 (10007663; Cayman), and rabbit IgG (2729S; Cell Signaling Technology). The following primers used for RT-PCR: PPP1R3C#1 (TSS) Fwd, GTCGCTGGGAGAGACTGAG, and Rev, ATTGGTCCCAGGGATCGG; PPP1R3C#2 (TSS−1 kb) Fwd, GGGCTAGTTCCCAAGTTTCTAC, and Rev, GGCACAGAGTAAAGGCTCAA; PPP1R3C#3 (TSS−6 kb) Fwd, GTGCAGTGACGCGATCT, and Rev, ACCTGTAGTCCCAGCTACTC; HMGCR Fwd, CTTATTGGTCGAAGGCTCGT, and Rev, CTCACTAGAGGCCACCGAAC; Nontarget Fwd, TGCCCAGCCTCAGTTTCTTA, and Rev, GCAACCAAACCATGAGCTGA. For ChIP-seq, duplicate samples were analyzed in each condition (control and HA-SREBF2). The qualified libraries were sequenced (150 bp paired-end reads) using an Illumina Novaseq 6000 sequencer at Novogene (Beijing, China). ChIP-seq data have been deposited in the NCBI GEO under accession GSE270026.

Patient-Derived Xenograft (PDX).

The study was performed with the approval of the institutional review board at Duke University. Informed consent was obtained from all participants. Chondrosarcoma cells were obtained from each patient’s tumor and maintained subcutaneously in vivo in NSG mice. A total of 1.0 × 10⁶ cells (725: IDH1R132H) transduced with Doxycycline-inducible shRNA (control or PPP1R3C) were suspended in 100 μL of 50% Matrigel prepared in PBS and subcutaneously inoculated into the left flank of 6-wk-old female mice. The administration of 625 mg/kg of Doxycycline diet (Inotive) was started 3 d after injection. The tumor volume was measured weekly with calipers and calculated as 1/2 × (tumor length) × (tumor width)². Tumors were harvested, weighted, and processed for histological analysis.

Quantification and Statistical Analysis.

Statistical analyses were performed using GraphPad Prism 10 software (La Jolla, CA). Data were presented as mean values with error bars representing SD. Two-tailed Student’s t test and one-way or two-way ANOVA with post hoc comparisons using Tukey’s honest significant difference (HSD) test for multiple comparisons were used for statistical analysis and determination of statistical significance.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNA-seq data and ChIP-seq data have been deposited in the NCBI GEO under accession GSE286965 (55) and GSE270026 (56), respectively. All other data are included in the article and/or SI Appendix.