Cell cycle duration determines oncogenic transformation capacity

Abstract

Oncogenic mutations are widespread in normal human tissues¹. Similarly, in murine chimeras, cells carrying an oncogenic lesion contribute normal cells to adult tissues without causing cancer^2-4. How lineages that escape cancer via normal development differ from the minority that succumb is unclear. Tumours exhibit characteristic cancer hallmarks; we therefore searched for hallmarks that differentiate cancer-prone lineages from resistant lineages. Here we show that total cell cycle duration (T_c) predicts transformation susceptibility across multiple tumour types. Cancer-prone Rb- and p107-deficient retina (Rb is also known as Rb1 and p107 is also known as Rbl1) exhibited defects in apoptosis, senescence, immune surveillance, angiogenesis, DNA repair, polarity and proliferation. Perturbing the SKP2-p27-CDK2/CDK1 axis could block cancer without affecting these hallmarks. Thus, cancer requires more than the presence of its hallmarks. Notably, every tumour-suppressive mutation that we tested increased T_c, and the T_c of the cell of origin of retinoblastoma cells was half that of resistant lineages. T_c also differentiated the cell of origin in Rb^-/- pituitary cancer. In lung, loss of Rb and p53 (also known as Trp53) transforms neuroendocrine cells, whereas Kras^G12D or Braf^V600E mutations transform alveolar type 2 cells^5-7. The shortest T_c consistently identified the cell of origin, regardless of mutation timing. Thus, relative T_c is a hallmark of initiation that distinguishes cancer-prone from cancer-resistant lineages in several settings, explaining how mutated cells escape transformation without inducing apoptosis, senescence or immune surveillance.

Fig. 1. Tumour suppression without increasing cell death, senescence or immune infiltration.

a–c, Kaplan–Meier curves showing how Skp2^+/− and Skp2^−/− (a), p27^KI/+, p27^KI/KI and Skp2^+/−;p27^CK−/CK− (b) and Cdk2^−/− and Cdk1^+/−;Cdk2^−/− (c) genotypes affect tumorigenesis in the DKO mouse retina. P values by log-rank (Mantel–Cox) test comparing indicated genotypes with DKO. d, Quantification of AP2A⁺ (amacrine), ONECUT2-expressing (OC2⁺) (horizontal) and SOX9⁺ (Müller) cells in P10 retina of indicated mouse genotypes as a percentage of all retinal cells. e, Active caspase-3⁺ cells in retinas of indicated ages and genotypes. f, Percentage of AP2A⁺ active caspase-3⁺ cells among all AP2A⁺ amacrine cells. g, Counts of senescence-associated β-galactosidase-expressing (SA-β-gal⁺) cells per retinal section. h, Representative flow cytometry histograms showing the proportion of B220⁺ B cells, CD3⁺ T cells, NKp46⁺ natural killer (NK) cells, CD11b⁺ macrophage/microglial cells and CD45⁺ leukocytes in P8 retina and spleen. i, Quantification of GAL3⁺ microglia of P8 retina as a percentage of all retinal cells. Data in d–g,i are mean ± s.d. (n = 6 mice per cohort), and asterisks indicate a significant difference between indicated genotypes and DKO; one-way analysis of variance (ANOVA) with Bonferroni correction. *P < 0.05; **P < 0.01. Source data

Fig. 2. Tumour suppression without altering DNA damage, aberrant angiogenesis or apical polarity defects.

a, Horizontal retinal sections of retina from P8 mice of indicated genotypes were stained for nuclei (DAPI, blue) and DNA damage (γH2AX, green). GCL, ganglion cell layer; INL, inner nuclear layer; ONL, outer nuclear layer. Scale bars, 50 µm. b, Quantification of γH2AX⁺ cell density in retina from a. c–e, Quantification of mean vessel length (c), vessel coverage area (d) and lacunarity (e) of isolectin B4 (IB4)-stained whole-mount P8 retinas of indicated genotypes. SVP, superficial vascular plexus. f, Horizontal retinal sections from mice of the indicated ages and genotypes were stained for DAPI (blue) and markers that detect F-actin at adherens junctions (green), amacrine cells (calretinin, green) and photoreceptor precursors (CRX, green). Arrows indicate breaks in the outer limiting membrane at E14 or E17-P0, or rosettes at P8. NBL, neuroblast layer. Scale bars, 50 µm. g, Whole-mount retinas of P0 mice of the indicated genotypes were stained for F-actin (green). Scale bars, 200 µm. h, Area of disrupted F-actin staining in g as a percentage of the whole retina. Experiments were repeated independently with similar results at least three times in a,f and two times in g for each mouse. Data are mean ± s.d. (n = 6 mice (b–e) and 9–29 mice (h) per group as indicated). Asterisks indicate significant difference compared with DKO by one-way ANOVA with Bonferroni correction. Source data

Fig. 3. Tumour suppression without altering cell lineage.

a, UMAP analysis of scRNA-seq data from P8 DKO and DKO-Skp2^+/− retinas. b, Heat map of top five DEGs. Dendrogram shows the relatedness of the seven clusters. log₂FC, log₂ fold change. c, Proportion of cells in each cluster. d, Seurat analysis of the proportions of cells in each phase of the cell cycle. The red outline highlights the AmP cluster. e, Enriched gene lists in DKO versus DKO-Skp2^+/− AmP, Mu and Neu clusters. BioPlanet available at https://tripod.nih.gov/bioplanet/. f, Enrichment of cell cycle gene signatures among genes that are up-regulated in DKO retina. g, P0 Rb^f/f;p107^−/−;Skp2^+/− pups received subretinal injection of low-titre Cre–GFP retrovirus to knockout Rb, and cell types in GFP⁺ clones were scored at P21. Data are mean ± s.d. (n = 3 mice per group). Source data

Fig. 4. Tumour suppression extends T_c, and the cell of origin has the shortest T_c.

a, Proliferation index in retina from mice of the indicated genotypes and ages. b, Percentage of BrdU⁺ cells in P8 retinas of the indicated genotypes. c, Total cell cycle (T_c) and S phase (T_s) duration of all dividing (Ki67⁺) cells in retina of the indicated genotypes at P8. d, T_c and T_s in amacrine (Ama), Müller (Mu) or horizontal (Ho) cells in retina of the indicated genotypes at P8. Data are mean ± s.d. (n = 6 mice per cohort as indicated by bars). Asterisks indicate significant difference compared with DKO by one-way ANOVA with Bonferroni correction. Exact P values are indicated for comparisons with amacrine (AP2A⁺) and Müller (SOX9⁺) or horizontal (ONECUT2⁺) cells in DKO retinas by two-tailed unpaired t-test. Source data

Fig. 5. Shortest T_c as a general hallmark of cancer susceptibility.

a, T_c and T_s of MSHα⁺ cells in the intermediate (IL) and anterior (AL) pituitary lobes of Rb-null (creERT2;LSL-tdTomato;Rb^f/f) mice injected with tamoxifen at 4 or 8 weeks of age and assessed at 6 or 10 weeks of age, respectively. b, T_c and T_s of lung CGRP⁺ neuroendocrine (NE), SPC⁺ AT2 and CCSP⁺ club cells from mice that are deficient in Rb and p53 (hSPC-rtTA;tet-cre;Rb^f/f;p53^f/f) that were injected with doxycycline (Dox) at plug formation and assessed at E18, and from Rb- and p53-deficient (creERT2;LSL-tdTomato;Rb^f/f;p53^f/f) mice injected with tamoxifen at 4 weeks of age and assessed at 8 weeks of age. c, T_c and T_s of lung NE, AT2 and club cells of Kras^G12D (creERT2;Kras^G12D;LSL-tdTomato mice) mice injected with tamoxifen at 4 weeks, 4 weeks, 5 weeks or 10 weeks of age and assessed at 6 weeks, 8 weeks, 7 weeks or 13 weeks of age, respectively. d, T_c and T_s of lung NE, AT2 and club cells of Braf^CA (creERT2;LSL-tdTomato;Braf^CA) mice that were injected with tamoxifen at 4 weeks of age and assessed at 6 weeks of age. Data are mean ± s.d. (n = 6 mice (a, b (right),c,d); n = 4 mice (b (left)) per group as indicated). Exact P values for the comparison between MSHα⁺ intermediate lobe versus anterior lobe cells (a), NE versus AT2 or club cells (b) or AT2 versus NE or club cells (c,d) by one-way ANOVA with Bonferroni correction (b, c (first three graphs),d) or two-tailed unpaired t-test (a,c (right)) are shown. Source data

Extended Data Fig. 1. Disrupting the Skp2-p27-Cdk axis can suppress retinoblastoma without increasing cell death, senescence, or immune infiltration.

a) Representative Skp2 Western blot from retinas of the indicated genotypes (upper panel) and quantification (lower panel). b) Representative p27 Western blot from P8 retinas of the indicated genotypes (upper panel) and quantification (lower panel). c) Horizontal retinal sections from P10 retinas of the indicated genotypes were stained for nuclei (DAPI, blue), p27 (green) and amacrine cells (Ap2a, red). d) The percentage of p27⁺ amacrine (Ap2a⁺) cells from (c). e) Thickness of the retinas of indicated ages and genotypes. f) Quantification of Brn3⁺, Pkca⁺, Rhodopsin (Rho)⁺ and Cone arrestin (Arr3) ⁺ cells per mm² of P21 retina, normalized to p107 KO retinas. g) Representative active caspase 3 Western blot from P8 retinas of the indicated genotypes (left panel) and quantification (right panel). h) Horizontal retinal sections of the indicated ages and genotypes were stained for SA-β-gal. White squares were blown up to show positive SA-β-gal staining (Arrows). i) Horizontal P8 retinal sections of the indicated genotypes were stained for nuclei (DAPI, blue) and Galectin-3 (Gal3) to label microglia cells (green). Experiments were repeated independently with similar results at least three times in c, h and i for each animal. Data are presented as mean ± s.d. in a, b, d, e-g from n = 6 mice per cohort as indicated by bars, and asterisks indicate a significant difference between retinas of DKO and the indicated genotypes (*, p < 0.05; **; p < 0.01, one-way ANOVA, Bonferroni correction). Scale bar in c, h, i: 50 µm. ONL: outer nuclear layer. INL: inner nuclear layer. GCL: ganglion cell layer. Source data

Extended Data Fig. 2. An in vivo strategy to identify a cancer cell-of-origin hallmark.

The top panel summarizes the four steps. Step 1 introduces mutations to suppress cancer (here, retinoblastoma). Step 2 assesses multiple (here, nine) cancer hallmarks to determine whether any are always altered upon tumor suppression; only cell cycle length (Tc) fulfilled this criterion. Step 3 asks whether Tc is shorter in the cancer-prone (here, amacrine cell) vs. cancer-resistant lineages (Müller glia, horizontal cells). Tc in cancer-prone or resistant lineages is indicated on the left, and the effect of multiple tumor-suppressing mutations on Tc is shown on the right. Step 4 compares Tc of cancer-prone vs. resistant lineages in multiple other contexts. Tc was always shortest in the cancer-prone lineage independent of oncogenic lesion or tissue (indicated by green check marks).

Extended Data Fig. 3. Tumor suppression without rescuing aberrant angiogenesis and apical polarity.

a) IB4 staining of P8 whole-mount retinas of indicated genotypes. Selected areas are blown up to show the superficial vascular plexus (SVP) and deep vascular plexus (DVP). b) IB4 (green) and DAPI (blue) staining of P8 retinal sections of indicated genotypes. c) E16 whole-mount retina of αCre;Rb^f/f;p107^−/−;Z/red mouse stained for F-actin (green). Red fluorescence indicates Cre-activity. Arrows indicate OLM breaks in Cre⁺ region labelled by F-actin. d) Horizontal retinal sections from P0 mice of the indicated genotypes were stained for nuclei (DAPI, blue) and amacrine cells (Ap2a, green), or apical polarity complex components (aPkcι, red; phospho-aPkcι/λ: green), or adherens junctions (N-cadherin, green). Arrows as in (c). Dotted lines in a and c indicate the boundary between no α-Cre expression (central retina) and α-Cre expression areas (peripheral retina). Experiments were repeated independently with similar results at least two times in a, b and d for each animal. Scale bars: 50 µm in b, d; 200 µm in a, c. ONL: outer nuclear layer. INL: inner nuclear layer. GCL: ganglion cell layer. NBL: Neuroblast layer. ON: optic nerve head.

Extended Data Fig. 4. Tumor suppression independent of cell lineage changes.

a) UMAP plot (left panel) showing astrocytes (green) and immune cells (blue) based on scRNAseq of P8 DKO and DKO-Skp2^+/− retinas, and quantification (right panel). b) The six steps to analyze the scRNAseq data. c) The initial clustering UMAP of the scRNAseq data of P8 DKO (orange) and DKO-Skp2^+/− (blue) retinas. d) Cells annotated with three cell cycle phases, including G1 (blue), G2/M (orange) and S (green). Left panel: distribution before regressing out cell cycle genes. Right panel: distribution after regressing out cell cycle genes to focus on lineage. e) Clustering UMAP of the scRNAseq data of P8 DKO (orange) and DKO-Skp2^+/− (blue) retinas, after regressing out cell cycle genes. f) The 12 Leiden clusters in (e). g) Dot plot of the top 5 marker genes in each Leiden cluster. Based on this analysis, the 12 Leiden clusters (numbers on the right) were consolidated into 7 clusters (names on the left). h) Known marker genes of 11 mouse retinal cell populations (left panel) and dot plot of the scores of each cell population in the scRNAseq data of DKO and DKO-Skp2^+/− retinas (right panel). i) Cluster UMAP showing the 11 cell populations during normal murine retinal development (left panel) from Clark et al., and the allocation of these cell types in the UMAP (middle panel) in (f). The graph (right panel) shows the proportion of each cell population in the 12-Leiden clusters.

Extended data Fig. 5. Tracking proliferation index and cell cycle length in multiple clusters.

a) Proliferation index (%Ki67⁺ cells) in the indicated cell types expressing the indicated markers in P8 retinas of indicated genotypes. b) Dot plot indicating expression of the markers used in (a) in the seven annotated clusters. c) Examples of feature maps for some markers used in (a). The expression of the markers in DKO or DKO-Skp2 +/− retina is shown on the two left panels, while expression in normal development is shown on the right panel. d) Tc and Ts of cells expressing the indicated marker proteins. Data are presented as mean ± s.d. in a and d (n = 6 mice per cohort as indicated by bars). Asterisks: significant difference between DKO and the indicated genotypes (*, p < 0.05; **; p < 0.01, one-way ANOVA, Bonferroni correction). The exact p-values between amacrine precursors (Ptf1a⁺) and mitotic neural progenitors (p21-Cdkn1a⁺) or mitotic neural progenitor and amacrine precursors (Mycn⁺) in DKO retinas by two-tailed unpaired t-test are indicated in the figures. Source data

Extended data Fig. 6. Proliferation (Ki67) index of Rb KO pituitary and Rb/p53 DKO E18 lung.

a) Proliferation index in the intermediate (IL) and anterior (AL) pituitary lobes of WT (LSL-tdTomato;Rb^f/f) or Rb null (CreERT2;LSL-tdTomato;Rb^f/f) mice injected with tamoxifen at 4 or 8 wk and assessed at 6 or 10 wk of age, respectively. b) Proliferation index in WT or Rb/p53 DKO E18 lung Cgrp⁺ neuroendocrine (NE), Spc⁺ Alveolar type 2 (AT2) and Ccsp⁺ club cells. Data are presented as mean ± s.d. (n = 6 mice for a, n = 4 mice for b per cohort as indicated by bars); The exact p-values between WT/Control vs. mutants by two-tailed unpaired t-test are indicated in the figures. Source data

Extended Data Fig. 7. tdTomato (tdTmt) index and proliferation (Ki67) index of tamoxifen-induced Rb/p53 DKO, Kras^G12D and BrafCA mouse lung cells.

a) Control (CreERT2;LSL-tdTomato) and Rb/p53 DKO animals (CreERT2;LSL-tdTomato;Rb^f/f;p53^f/f) injected with tamoxifen at 4 wk and assessed at 8 wk of age. b-e) Control (CreERT2;LSL-tdTomato) and Kras (CreERT2;LSL-tdTomato;Kras^G12D) mice injected with tamoxifen at: 4 wk and assessed at 6 wk of age (b); 4 wk and assessed at 8 wk of age (c); 5 wk and assessed at 7 wk of age (d); 10 wk and assessed at 13 wk of age (e). f) Control (CreERT2;LSL-tdTomato) and BrafCA animals (CreERT2;BrafCA;LSL-tdTomato) injected with tamoxifen at 4 wk and assessed at 6 wk of age. Schematics on the left summarize the experiments. Graphs in the center show tdTmt index of the indicated cell types (grey control, blue oncogenic lesion induced). Graphs on the right show Ki67 index of the indicated cell types for control tdTmt⁺ or tdTmt⁻ cells in CreERT2;LSL-tdTomato mice or control unrecombined tdTmt⁻ cells in CreERT2;LSL-tdTomato; Rb/p53, Kras^G12D, or Braf^CA mice (blue bars), or induced tdTmt⁺ cells in the latter (red bars). Data are presented as mean ± s.d. (n = 6 mice per cohort as indicated by bars); asterisks: significant difference between control vs. mutated mice (* p < 0.05, ** p < 0.01. two-tailed unpaired student’s t-test). Source data