Christelle Thibault

Cell-cycle remodeling is fundamental to pluripotency and lineage commitment, yet whether its transcriptional and post-transcriptional architecture is conserved across species and developmental states has remained unresolved. Here we introduce Ciclopes, a biology-informed deep-learning framework that resolves continuous cell-cycle phase and phase-dependent mRNA transcription and degradation directly from single-cell RNA sequencing. Applying Ciclopes across six mouse and human pluripotent stem-cell systems spanning naive and primed states, we uncover striking divergence in transcriptional complexity and oscillatory control: mouse systems sustain elevated baseline expression of core cell-cycle regulators, while human systems trade higher baseline expression for larger oscillatory amplitude. Strikingly, mRNA degradation timing remain far more conserved across systems than transcription timing, exposing post-transcriptional regulation as a stable evolutionary backbone. As human iPSCs differentiate into definitive endoderm, cells progressively exit the cell cycle, cell-cycle-coupled gene networks contract, and surviving regulators oscillate with larger amplitude. Ciclopes establishes a general framework for dissecting how pluripotent cells tune proliferation across evolutionary and developmental transitions.

A total of 3–6% human genome is composed of microsatellite sequences, which are short DNA elements composed of two to six nucleotide motifs repeated in tandem. Expansion of a subset of these microsatellites is the leading cause of >60 diseases. However, most of these mutations are located in sequences annotated as noncoding, which raises questions about their pathogenicity. Here we found that GGC repeat expansions causing oculopharyngodistal myopathy with or without oculopharyngeal myopathy leukoencephalopathy are located within previously unrecognized open reading frames (ORFs), resulting in their translation into new polyglycine-containing proteins. Antibodies developed against these proteins stain the p62-positive inclusions typical of these diseases. Moreover, expression of these polyglycine proteins causes locomotor and skeletal muscle alterations associated with neurodegeneration in cell, fly and mouse models. Finally, we identified a compound, the cationic porphyrin TMPyP4, targeting the expression of these polyglycine proteins, raising hope to develop a therapy for these disorders. Overall, this work highlights the complexity and richness of the human genome and the importance of mutations in yet-unrecognized small ORFs.

Deconvolution of cell types in TCGA-KIRC cohort.

Annotation of supplementary figures

Single-cell RNA sequencing markers for each global cluster and each re-clustering

Deconvolution of cell types in BIONIKK cohort.

General and specific markers for the identification of each sub-cluster in scRNAseq analysis and associated references.

Supplementary Figure 1. Phenotyping of main cell populations. A. Metrics (number of cells, reads, median genes, UMI) after sequencing of each sample and their associated clinical characteristics. B. Expression of common markers for tumour, immune and stromal cells and of epithelial and mesenchymal states represented as percentage of expressing cells after dividing the UMAP into 80 hexagonal bins. Supplementary Figure 2. Diversity in lymphoid, myeloid and endothelial populations. A. UMAP plot of lymphoid sub-clusters (27,223 cells). B. Pseudo-bulk heatmap of identity markers for each lymphoid sub-cluster. Abbreviations: CD4.reg: regulatory CD4-T cells; CD4.eff: effector CD4-T cells; CD4.mem: memory CD4-T cells; CD8.mem: memory CD8-T cells; CD8.eff: effector CD8-T cells; CD8.ex: exhausted CD8-T cells; CD8.stress: stressed CD8-T cells; NKT.inf: inflamed natural killer-like T cells; NKT.IFNG: NKT cells expressing IFNG as top marker; NKT.stress: stressed NKT cells; NK.HBB: natural killer cells expressing HBB as top marker; NK.PTGDS: NK cells expressing PTGDS as top marker; NK.surv: surveillance NK cells; NK.ct: cytotoxic NK cells. C. UMAP plot of endothelial sub-clusters (9,726 cells). Abbreviations: ED.cor: cortical endothelial cells; ED.lymph: lymphatic endothelial cells; ED.RGCC: endothelial cell expressing RGCC as top marker; ED.venous: venous endothelial cells; ED.glom: glomerular endothelial cells; POD: podocytes. D. Pseudo-bulk heatmap of identity markers for each endothelial sub-cluster. E. VISION projection of capillary and arterial signatures coloured by value ranks showing a division of the cortical endothelial sub-cluster. F. UMAP plot of myeloid sub-clusters showing the gradient from M1-like to M2-like macrophages (3,295 cells). Abbreviations: NEUT: neutrophils; MONO.cl: classical monocyte; MONO.at: atypical monocytes; TAM: tumour-associated macrophages; TAM.CD1C: TAM expressing CD1c as top marker. G. Pseudo-bulk heatmap of identity markers of each monocyte and macrophage sub-cluster. H. VISION projection of M1- and M2-like macrophages signatures coloured by value ranks highlighting a gradient in the TAM sub-cluster. Supplementary Figure 3. Heterogeneity in the ccRCC.int populations. A. Pie chart displaying the proportion of each ccRCC subtypes in the total tumour collection. B. UMAP projection of gene expression for TGFBI (red) and EPCAM (green) in ccRCC tumour cells. C. KEGG pathway ontology analysis of specific markers for tumour sub-clusters (number of markers for each cluster indicated in brackets; number of genes found in each pathway noted near the bar). D. VISION projection of the melanoma immune-like signature. E. GSEA analysis of ccRCC.int1 versus ccRCC.int2. F. UMAP plot showing ccRCC.int sub-clustering. Each ccRCC.int is divided into two sub-clusters (ccRCC.int1 in C0 and C2; ccRCC.int2 in C1 and C3). G. Heatmap of the 10-top markers for each sub-cluster. H. DAVID gene ontology analysis of specific markers for the four new formed sub-clusters using BP_DIRECT (number of markers for each cluster indicated in brackets; number of genes found in each pathway noted near the bar). Supplementary Figure 4. Validation of ccRCC tumour populations in dataset PRJNA705464 from Krishna et.al. A. Original labelling superimposed on the UMAP plot obtained after re-analysis of CA9-expressing cells. B. UMAP plot showing identified sub-clusters. C. Bar plot displaying the contribution of 3 tumour samples to each identified cluster. Patient samples comprising less than 20 annotated tumour cells were not included. D. Piechart displaying the global proportion of each identified cluster. E. VISION projections of indicated ccRCC signatures. F. Heatmap representation of the GSVA analysis showing specific hallmarks enriched in each cluster. G. Pseudo-bulk heatmap of 5 specific markers for epithelial, mesenchymal and inflamed-state cells. Supplementary Figure 5. Validation of ccRCC tumour populations in PRJNA671297 from Braun et.al. A. Original labelling superimposed on the UMAP plot obtained after our re-analysis and removal of misannotated cells (Dataset: on dbGaP, accession number phs002252.v1.p1). B. UMAP plot showing identified sub-clusters. C. Bar plot displaying the contribution of 5 tumour samples to each identified cluster. Patient samples comprising less than 20 annotated tumour cells were not included. D. Piechart displaying the global proportion of each identified cluster. E. VISION projections of indicated ccRCC signatures. F. Heatmap representation of the GSVA analysis showing specific hallmarks enriched in each cluster. G. Pseudo-bulk heatmap of 5 specific markers of epithelial, mesenchymal and inflamed-state cells. Supplementary Figure 6. Description of PST clusters. A. Localisation of each identified cluster with respect to nephron structure. B. Heatmap of the 10-top markers for each tubule sub-cluster. C. Differentially expressed genes between PST1 (blue) and PST2 (red). D. Pseudo-bulk heatmap of S1/S2/S3 PST segment markers by Young et.al., and Lake et.al., to assess if different segments have been captured in our dataset. E. VISION projection of gluconeogenesis, glycolysis, oxidative phosphorylation and HIF1a signatures. F. Pseudo-bulk heatmap recapitulating expression of key genes for PST and the indicated ccRCC populations. Supplementary Figure 7. Analysis of cell populations in TCGA-KIRC cohort. A. Heatmap showing deconvolution results of all identified populations inferred by CIBERSORTx and displayed as row-scaled absolute scores on bulk RNA-seq data of 495 tumour samples. B. Reduced heatmap for deconvolution results of indicated populations displayed as column-scaled CIBERSORTx absolute scores. C. Sorted Spearman correlation coefficients between the indicated ccRCC populations and all other identified populations. D. Heatmap of each subpopulation average row-scaled absolute score with samples grouped according to the m1-m4 TCGA-KIRC classification. Supplementary...

The oncogenic mechanisms by which TFE3 fusion proteins drive translocation renal cell carcinoma (tRCC) are poorly characterized. Here, we integrated loss and gain of function experiments with multi-omics analyses in tRCC cell lines and patient tumors. High nuclear accumulation of NONO-TFE3 or PRCC-TFE3 fusion proteins promotes their broad binding across the genome at H3K27ac-marked active chromatin, engaging a core set of M/E-box-containing regulatory elements to activate specific gene expression programs as well as promiscuous binding to active promoters to stimulate mRNA synthesis. Within the core program, TFE3 fusions directly regulate genes involved in ferroptosis resistance and oxidative phosphorylation metabolism (OxPhos). Consequently, human tRCC tumors display high OxPhos scores that persist during their epithelial to mesenchymal transition (EMT). We further show that tRCC tumor aggressiveness is related to their EMT and their associated enrichment in myofibroblast cancer-associated fibroblasts (myCAFs) that are both hallmarks of poor prognostic outcomes. We define tRCC as a novel metabolic subtype of renal cancer and provide unique insights into how broad genomic binding of TFE3 fusion proteins regulates OxPhos and ferroptosis resistance.

Prostate cancer is a heterogeneous disease with a slow progression and a highly variable clinical outcome. The tumor suppressor genes PTEN and TP53 are frequently mutated in prostate cancer and are predictive of early metastatic dissemination and unfavorable patient outcomes. The progression of solid tumors to metastasis is often associated with increased cell plasticity, but the complex events underlying TP53-loss-induced disease aggressiveness remain incompletely understood. Using genetically engineered mice, we show that Trp53 deficiency in Pten-null prostatic epithelial cells (PECs) does not impact early cell proliferation and neoplasia formation, nor growth arrest and senescence entry at a later time. However, Trp53-deficiency enhances invasive adenocarcinoma development and promotes metastatic cell dissemination. Importantly, our single-cell transcriptomic and chromatin accessibility analyses combined with histological examinations uncovered an epithelial cell population characterized by an induction of Jak/Stat3 signaling and displaying mesenchymal features. Moreover, we show that the transcriptomic signature of this cell population is prominent in tumors of patients with high-risk prostate cancer or metastatic disease. In addition, our in vivo and organoid-based experiments provide evidence that PEC plasticity occurs through bi-directional communication with cancer-associated fibroblasts (CAFs). Thus, our study demonstrates that p53 loss induces a protumorigenic crosstalk between PECs and CAFs, and identifies new vulnerabilities that might be targeted to limit cancer progression.