Figure S6. Differential pathway gene set enrichment analysis among DKOAA vs DKO in all cell types, visualized as networks via aPEAR.

Figure S20. DE across OS models, stratified by pathologic subtype for which samples were available (including Osteo and Fibro-like, but not chondro-like).

Table S9. Compositional analysis (differential abundance analysis) comparing proportions of cell subtypes from each genotype using the Propeller test from the Speckle package.

Table S8. Marker genes for all celltype subclusters.

Figure S1. Clustering and markers of transgenic OS tumors.

Figure S4. Leading edge genes from GSEA showing downregulation of invasive phenotypes in TKO and DKOAA relative to DKO.

Upregulated Myc target activity in TKO and DKOAA malignant cells. A, SCENIC transcription factor activity scores of E2F family transcription factors. B, Expression of apoptosis-related genes derived from the Reactome apoptosis gene set. C and D, GSEA plot of one of the “Hallmark Myc targets V1” gene set in TKO and DKOAA relative to DKO, respectively. E, Heatmap of GSEA leading-edge genes in the “Hallmark Myc targets V1” gene set. The intersect of leading-edge genes from the TKO and DKOAA enrichments is shown. F, Compositional analysis of Myc-high responder cells. G, Violin plot for malignant cells only of signature scores calculated from genes in the “Hallmark Myc targets V1” gene set. H, Violin plot for malignant cells of Ccna2, a Myc target gene. I, Violin plot of the Myc signature scores in all cells.

Renal cell carcinoma (RCC), especially clear cell RCC (ccRCC), is characterized by metabolic reprogramming, notably disordered lipid metabolism and prominent intracellular lipid droplet accumulation. In addition to abnormal triglyceride and cholesterol ester storage, lipid droplets promote tumor proliferation, survival, and drug resistance by supplying energy, membrane components, and signaling platforms, thereby representing potential therapeutic targets. We integrated TMU-RNAseq and TCGA-KIRC datasets and performed proximity-labeling mass spectrometry targeting PLIN2/3 to identify perilipin-related genes (PRGs). Based on TCGA-KIRC, we established a lipid droplet-related model (LDM), and the LDM score (LDMS) effectively predicted ccRCC prognosis. Validation in TCGA-KIRC and three external cohorts (GSE22541, E-MTAB-1980, and E-MTAB-3267) showed superior prognostic performance over conventional clinical parameters. High LDMS was associated with increased mutation frequency and an immunosuppressive microenvironment. Drug sensitivity analysis suggested differential responses to Sorafenib, Cediranib, and Saracatinib. Functional assays demonstrated that GNA14 inhibits ccRCC progression and enhances lipid accumulation via PLIN2 upregulation. GNA14 overexpression combined with Lovastatin further strengthened antitumor effects. Overall, LDM is a robust prognostic tool, and GNA14 plus Lovastatin may offer a potential therapeutic strategy for ccRCC.On one hand, at a general level, overexpression of GNA14 in RCC significantly inhibits the proliferation, migration, and invasion of renal cancer cells, thereby exerting a suppressive effect on RCC. On the other hand, GNA14 interacts with PLIN2, leading to an increase in PLIN2 expression, which subsequently results in an enlargement of lipid droplet quantity and volume. This causes a notable rise in cholesterol and triglyceride levels within RCC cells, further facilitating RCC progression. Excitingly, lovastatin can partially block the detrimental pathway by which GNA14 promotes lipid accumulation. Therefore, combining lovastatin with GNA14 overexpression yields a more effective suppression of RCC.

Abstract Osteosarcoma is the most common primary pediatric bone malignancy. One promising new target is SKP2, encoding a substrate recognition factor of the SCF E3 ubiquitin ligase that targets p27 for proteasomal degradation, driving cellular proliferation. Knockout (KO) of Skp2 in an immunocompetent transgenic mouse model of osteosarcoma improved survival, drove apoptosis, and induced antitumor immunity. In this study, we applied single-cell RNA-sequencing (scRNA-seq) to primary osteosarcoma tumors from Osx-Cre conditional Rb1/Trp53 KO mice. We further compared with models of Skp2 disruption: Skp2 KO or disruption of the Skp2–p27 interaction (resulting in p27 overexpression). We report that murine osteosarcoma models recapitulate the tumor heterogeneity and microenvironment complexity observed in patient tumors. Skp2 disruption led to reduction of T-cell exhaustion and upregulation of interferon (IFN) signaling, as well as induction of cell type–specific replicative and endoplasmic reticulum stress, which we validated with proteomics analysis. Furthermore, we showed that IFN induction was correlated with improved survival in patients with osteosarcoma. Additionally, our scRNA-seq analysis uncovered decreased expression of metastasis-related gene signatures in Skp2-disrupted osteosarcoma, which we validated by a strong reduction in lung metastasis in the Skp2 KO mice. Finally, we report several mechanisms potentially used by osteosarcoma to escape from Skp2 targeting, including upregulation of Myc targets, induction of genomic instability, overexpression of alternative E3 ligases, and lineage plasticity. These mechanistic insights into osteosarcoma tumor biology and Skp2 function suggest novel targets for new, synergistic therapies, whereas the data and our comprehensive analysis may serve as a public resource for further big data–driven osteosarcoma research. Significance: Our single-cell study of murine osteosarcoma models uncovers Skp2 function in metastasis, genomic instability, and immune activation and reveals additional target pathways to overcome resistance to Skp2 disruptions.

Table S3. Markers of celltypes after annotation. Wilcox test from Seurat (v5) FindAllMarkers function was used.

Induced immune activation in the form of IFN pathway activity and reduction of T-cell exhaustion. A, Dotplot of Hallmarks gene sets significantly upregulated in TKO relative to DKO across cell types. B, GSEA plots showing enrichment of IFN response pathways in genes differentially expressed in TKO relative to DKO among immune cells. C, UMAP showing subclustering of T cells. D, Canonical and data-driven markers of T-cell subclusters. E, Signature gene scores of T-cell states derived from marker genes in a published meta-analysis of tumor-infiltrating T cells (47). F, Compositional analysis of T-cell subclusters across 3 osteosarcoma tumors. UMAP, Uniform Manifold Approximation and Projection.

Figure S15. E2f and Myc upregulation in TKO and DKOAA malignant cells.

Background Proton pump inhibitors (PPIs) have been widely used for over 35 years. However, in recent decades, numerous adverse drug reactions (ADRs) have been reported, with evidence often inconsistent and heterogeneous across studies. Objective To conduct an overview of systematic reviews/meta-analyses (SR/MAs) to provide a contemporary review of the evidence for the safety of PPIs in patients using them for treating or prophylaxis, to summarise the outcome data, assess the methodological quality and rate the certainty of the evidence and to provide references for clinical decision-making and the subsequent formulation of evidence. Methods We conducted an overview of reviews following the Preferred Reporting Items for Overviews of Reviews guidelines. We identified SR/MAs regarding PPI safety through a search of multiple databases, including the China National Knowledge Infrastructure, Chinese Science and Technology Journal Database and Wanfang Database (WanFang) on 6 December 2025 as well as Embase, PubMed and the Cochrane Library database on 7 December 2025. The participants were human populations who had been administered PPIs; the intervention groups used PPI-based regimens and the control group received different PPI regimens, no-PPIs, H2-receptor antagonists (H2RAs), potassium-competitive acid blockers (P-CAB) or placebo; the outcomes mainly included the specific ADRs associated with PPI use. We used the A Measurement Tool to Assess Systematic Reviews-2 (AMSTAR-2) tool to assess the methodological quality of the included SR/MAs and employed the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework to evaluate the quality of evidence for the reported outcomes. The focus of the data presentation was descriptive, featuring detailed tabular presentations of characteristics and results at both the review level and the primary studies level. Main results A total of 940 studies were retrieved from various databases according to the search strategies. After eliminating duplicates and the screening process, 36 SR/MAs were finally included. Overall, there was a slight overlap (corrected covered area, CCA of 0.91%) in 483 primary studies included in the 36 SR/MAs. AMSTAR-2 evaluation results showed that 34 SR/MAs were of low quality, and the other two were of critically low quality. The GRADE evaluation results indicated that the certainty of evidence for all outcomes was low or very low . The use of PPIs might be associated with an increased risk of acute kidney injury (AKI) (relative risk (RR) 1.75, 95% CI 1.40 to 2.19), chronic kidney disease (CKD) (RR 1.35, 95% CI 1.15 to 1.56), gastric cancers (GC) (RR 1.67, 95% CI 1.39 to 2.00) and community-acquired pneumonia (CAP) (OR 1.37, 95% CI 1.22 to 1.53). PPI therapy was associated with a higher recurrence rate of Clostridioides difficile infection (CDI) (24% vs 18%) and might increase CDI risk in renal transplant recipients (RR 2.33, 95% CI 1.07 to 5.07). The use of PPIs might heighten the risk of fractures in young patients (aged <29 years) (RR 1.20, 95% CI 1.12 to 1.29). PPI therapy was also linked to hypomagnesaemia in haemodialysis patients (1/36, 2.78%) and renal transplant recipients (1/36, 2.78%). Conclusions This overview evaluates the safety profile of PPIs, noting associated risks including renal impairment, GC, fractures and infections. Most evidence comes from observational studies, resulting in low or very low certainty evidence that limits definitive causal conclusions. Clinicians and pharmacists may consider greater vigilance with PPI use, and these ADRs may not require de-escalation but de-implementation when inadequate (eg, long-term use or special group use). Future research should focus on high-quality prospective studies and investigate underlying mechanisms to better establish causality and quantify risks. PROSPERO registration number CRD420251059575.

Genome instability in osteosarcoma malignant cells. A, Heatmap showing HMM results of CNVs as determined by InferCNV. The top heatmap shows CNV states in macrophages (reference nonmutated cells), whereas the bottom heatmap shows malignant cells from each sample (rows). The columns represent chromosomes. B, Violin plots showing the number of genes affected by “extreme CNVs” (2× deletions or 2×> amplification). The left shows all extreme CNVs, the top-right shows extreme amplifications, and the bottom-right shows extreme deletions. C, PCA plot of malignant cells using inferred CNVs. D, Same PCA split but by OS models and colored by samples. E, Venn Diagram showing comparison of two differential CNV identification methods with differential expressed genes of TKO vs. DKO. F, Venn diagram showing comparison of two differential CNV identification methods with differential expressed genes of DKOAA vs. DKO. G, Heatmap showing expression patterns of significantly DNA copy number–amplified genes that were also significantly overexpressed in TKO and DKOAA.

Pathologic subtype and lineage infidelity in mouse osteosarcoma tumor. A and B, UMAP showing malignant cells from OS tumors, colored by sample (A) and pathologic subtype (B) as inferred via label transfer from a murine nonmalignant bone atlas dataset. C, The same UMAP, colored by label transfer scores. D, Violin plots of signature scores computed for markers of various bone cells. E, Dot plot of markers of malignant subtype classifications. F, Bar plot showing proportions of malignant cells annotated to each pathologic subtype in each sample. The subtype with the highest proportion is considered the inferred pathologic classification of the indicated tumor. G, UMAP of all OS tumor cells similar to Fig. 1B but split according to sample-wise inferred pathologic classification. H, Proportion of cell types in each pathologic subtype. I, Expression patterns of genes related to macrophage and osteoclast differentiation among malignant cells from each subtype. UMAP, Uniform Manifold Approximation and Projection.

Table S4. Cell metadata and celltype annotations.

Relating mouse osteosarcoma models to patients. A and B, UMAPs of human OS patient scRNA-seq data integrated from two publications, colored by data sources (A) and cell types (B). C, Violin plot showing the number of genes affected by extreme CNVs (≥2× deletion or amplification) across cell types. D, Canonical and data-driven markers of cell types in integrated human osteosarcoma. E, Heatmap showing expression scores of murine OS cell type markers in patient tumor cells. Rows correspond to murine marker genes, and columns refer to human OS cell types. F, Forest plot illustrating survival analysis of murine cell type signatures in the NCI Target OS cohort. G, KM plot showing survival association of an expression score calculated from genes in the Hallmark IFNα and IFNγ response gene sets that were also upregulated in TKO immune cells relative to DKO. H, KM plot showing survival association of an expression score calculated from genes in the Hallmark IFNα and IFNγ response gene sets that were also upregulated in DKOAA immune cells relative to DKO. UMAP, Uniform Manifold Approximation and Projection.

Table S5. Differential Expression results for all genes, celltypes, and comparisons

Complex microenvironment in transgenic murine osteosarcoma tumors. A, Mouse models for scRNA-seq analysis. Numbers indicate the mice included, with one tumor isolated from each mouse for scRNA-seq. B, UMAP of integrated cells colored by cell types. C, Canonical and data-driven markers of cell types shown as a bubble plot. D, UMAP colored by genotype. E, Violin plot of number of genes affected by “extreme CNVs” (deep 2× deletions, or 2×> amplification) in individual cell types. F, PCA plot of all cells using their CNV profiles inferred from scRNA-seq data. UMAP, Uniform Manifold Approximation and Projection.

Table S7. Mass spectrometry differentially expressed protein results.