Hepatocellular senescence induces multi-organ senescence and dysfunction via TGFβ

Abstract

Cellular senescence is not only associated with ageing but also impacts physiological and pathological processes, such as embryonic development and wound healing. Factors secreted by senescent cells affect their microenvironment and can induce spreading of senescence locally. Acute severe liver disease is associated with hepatocyte senescence and frequently progresses to multi-organ failure. Why the latter occurs is poorly understood. Here we demonstrate senescence development in extrahepatic organs and associated organ dysfunction in response to liver senescence using liver injury models and genetic models of hepatocyte-specific senescence. In patients with severe acute liver failure, we show that the extent of hepatocellular senescence predicts disease outcome, the need for liver transplantation and the occurrence of extrahepatic organ failure. We identify the TGFβ pathway as a critical mediator of systemic spread of senescence and demonstrate that TGFβ inhibition in vivo blocks senescence transmission to other organs, preventing liver senescence induced renal dysfunction. Our results highlight the systemic consequences of organ-specific senescence, which, independent of ageing, contributes to multi-organ dysfunction.

Fig. 1. Hepatocellular senescence results in senescence and dysfunction in other organs.

a, A schematic of the experimental approach; 8–12-week-old Mdm2^E5/E6fl; R26^{LSL-tdTomato/LSL-tdTomato} mice were intravenously injected with 2 × 10¹¹ GC of AAV-Cre or AAV-Null and culled 4 days later. The downstream targets of MDM2 are highlighted. b, Representative images of p21 IHC in liver cells of ΔMdm2^Hep and control mice; n = 16/19 control/ΔMdm2^Hep mice, respectively. c, Automated quantification of p21⁺ liver cells; n = 4/5 control/ΔMdm2^Hep mice, respectively; unpaired two-tailed t-test. d, Representative images of p21 IHC on mouse kidney sections. e, Manual quantification of p21⁺ renal tubular cells; the data are presented per field of view (FOV), n = 16/19 control/ΔMdm2^Hep mice (additional controls shown in Extended Data Fig. 1g); Welch’s two-tailed t-test. f, Representative images of p21 IHC on brain sections. g, Manual quantification of p21⁺ brain cells; n = 4/6 control/ΔMdm2^Hep mice, respectively; two-tailed Welch’s t-test. h, Representative images of p21 IHC in lung sections. i, Automated quantification of p21⁺ lung cells; n = 6/9 control/ΔMdm2^Hep mice, respectively; unpaired two-tailed t-test. j, Representative images of p21 IHC in liver sections of Kras^G12D/Kras^WT mice. k, Automated quantification of p21⁺ liver cells; n = 8/11 Kras^WT/Kras^G12D mice, respectively; two-tailed Mann–Whitney test. l, Representative images of p21 IHC on kidney sections of Kras^G12D/Kras^WT mice. m, Manual quantification of p21⁺ renal tubular cells; n = 14/16 Kras^WT and Kras^G12D mice, respectively; two-tailed Mann–Whitney test. n, The plasma levels of cystatin C in ΔMdm2^Hep or control mice 4 days post AAV-Cre or AAV-Null injection; n = 9/10 control/ΔMdm2^Hep mice, respectively; two-tailed Welch’s t-test. o, The urine levels of alanine and threonine in ΔMdm2^Hep mice pre and post AAV-Cre injection; the dots represent the average peak area of n = 3/4/5 mice at days −2 and 0, 4 and −1 and 3, respectively; two-way ANOVA comparing each timepoint to induction day (day 0). p, The proportion of time spent by the ΔMdm2^Hep/control mice in the new arm of the Y maze at day 4; n = 7/9 control/ΔMdm2^Hep mice, respectively; unpaired two-tailed t-test. q, The share of stable versus unstable oscillations of hippocampal brain slices; n = 13/14 brain slices from four control/ΔMdm2^Hep mice each, respectively. All the bars are mean ± s.e.m., each dot represents one sample per mouse and the numbers are P values. The scale bars are 50 μm and 5 μm in the inset magnifications. The source numerical data are available in Source data. Source data

Fig. 2. Hepatocellular senescence in human severe acute indeterminate hepatitis correlates with development of subsequent renal dysfunction.

a, A schematic of patient stratification and outcomes. b, Multiplex immunofluorescence staining for hepatocytes (HNF4α, red), senescent cells (p21, green), DNA damage (γH2AX, yellow) and nuclei (DAPI, grey) in human patient liver tissue. c, Quantification of p21⁺ and γH2AX⁺ hepatocytes in survivors versus non-survivors. n = 17 patients in both groups; two-tailed Welch’s t-test. d, The serum creatinine levels in survivors versus non-survivors over the course of 28 days post hospital admission. n = 16/14, 13/13, 13/14, 12/12, 10/6 and 9/5 patients on days 0, 2, 3, 7, 14 and 28 for the survivors/non-survivors groups, respectively; two-tailed paired t-test. The bars are the mean ± s.e.m. e, Change in serum creatinine from first (D0) to last recorded (DLast) day of admission; two-tailed paired t-test. n = 16 in both groups. f, Quantification of p21⁺ hepatocytes in patients who developed hepatic encephalopathy (HE) versus those ones who did not. N = 10 and n = 24 in the ‘HE’ and ‘no HE’ groups, respectively; two-tailed Mann–Whitney test. On all violin graphs, each dot represents one biological sample from individual patients, and the numbers are P values. Scale bars, 50 μm. The source numerical data are available in Source data.

Fig. 3. scRNA-seq reveals transcriptional changes, including amino acid transporter expression within the proximal tubular compartment in response to liver senescence.

a, A schematic of the scRNA-seq experiment. Three kidneys from control mice and three kidneys from ΔMdm2^Hep mice were dissociated, and droplet-based scRNA-seq was performed on them on a 10x Chromium chip. b, UMAP plots of all 24,215 cells (control and ΔMdm2^Hep). The cells are coloured on the basis of the broad clusters (left) or experimental group (right) (green, control; blue, ΔMdm2^Hep). c, UMAP plots showing the distribution of the cells that have a positive score for the p21 gene signature in the control (top) and the ΔMdm2^Hep (bottom) cells. The inset image is a representative kidney section from a ΔMdm2^Hep mouse stained for p21 by IHC with highlighted tubules (dashed lines). d, Pie charts showing the share of p21^high, TGFβ^high PTCs and JAK–STAT^high PTCs and mesenchymal cells in the control and ΔMdm2^Hep samples. Contingency was tested with the chi-square test (one-tailed). e, A bar chart showing the total number of cells with a positive score for the senescence transcriptional signature in the control and ΔMdm2^Hep samples (chi-square test (one-tailed)). f, A heat map showing the significantly differentially expressed Slc transporter genes in the PTC, DTC and LOH compartments. The gene IDs in the red frames are genes encoding transporters associated with amino acid transport. g, A schematic of the p21^KO experiment. Six Mdm2^E5/E6fl; p21^WT and six Mdm2^E5/E6fl; p21^KO mice were injected with AAV-Cre and culled 4 days later. h, The plasma levels of cystatin C in Mdm2^E5/E6fl; p21^WT and Mdm2^E5/E6fl; p21^KO mice 4 days post AAV injection. n = 6 for both groups; unpaired two-tailed t-test. The bars are the mean ± s.e.m., and the numbers on the graphs are P values. The source numerical data are available in Source data.

Fig. 4. Hepatic SASP in plasma induces senescence.

a, A schematic of the in vitro plasma treatment. WT MEFs and neuronal cells were treated with plasma from control/ΔMdm2^Hep mice and stained for SA β-Gal. b, Manual quantification of SA β-Gal⁺ WT MEFs. Each dot represents one technical replicate. Each of the two biological replicates (one mouse’s plasma sample, with 1/3 technical replicates shown as red/black points, respectively) from both control/ΔMdm2^Hep; n = 3 technical replicates for no plasma control. One-way ANOVA compares all the technical replicates. c, Manual quantification of SA β-Gal⁺ NS cell-derived neuronal cells. Each dot represents the average of the technical triplicates of one biological replicate for each group (n = 6 biological replicates). For the ‘no plasma’ groups, each bar is the mean of four technical replicates (n = 1 biological replicate), one-way ANOVA. d, The plasma levels of TGFβ1, TGFβ2, TGFβ3, CCL2 and LIF. For the TGFβ ligands: n = 9/7, CCL2: n = 9/10, LIF: n = 7/10 control/ΔMdm2^Hep plasma samples, respectively; unpaired two-tailed t-test, Welch’s t-test and Mann–Whitney U test for TGFβ1, TGFβc3 and TGFβ2 and CCL2 and LIF, respectively. Each dot represents one biological replicate (one mouse’s plasma sample). e, Quantification of western blot band signal intensity (Extended Data Fig. 8c) normalized to β-actin; n = 5 each, two-tailed Welch’s t-test/Mann–Whitney for pSMAD2/SMAD2/pSMAD3/SMAD3. f, A schematic of TGFβ inhibitor treatment. g, Plasma-treated neuronal cells were treated with either TGFβR1i (AZ12601011, AstraZeneca) or vehicle (DMSO), stained for SA β-Gal and manually quantified. Each dot represents the average of a technical triplicate from one biological replicate (one mouse’s plasma sample); n = 5 each, one-way ANOVA. h, A schematic of the transfer of conditioned media (CM) in murine slice culture experiments. The livers of Mdm2^E5/E6fl/R26^{LSL-tdTomato/LSL-tdTomato} mice were collected 2 days after AAV-Null/AAV-Cre (n = 6 each). CM was collected 48 h later and added, with TGFβR1i or vehicle, to ex vivo-cultured WT murine kidney slices. i, Manual quantification of kidney slice p21⁺ cells, n = 6 each; unpaired two-tailed t-tests. j, A schematic of a human kidney slice experiment. k, Human kidney slices were treated with TGFβ1 with or without TGFβR1i ex vivo, and p21⁺ renal cells were manually quantified; n = 6 each group, one-way ANOVA. All the bars are the mean ± s.e.m., and the numbers on the graphs are the P values. The source numerical data are available in Source data.

Fig. 5. Inhibition of TGFβ signalling prevents induction of senescence in the extrahepatic organs.

a, A schematic of the in vivo treatment with TGFβ receptor inhibitor (TGFβR1i). ΔMdm2^Hep mice were treated with TGFβR1i or vehicle by oral gavage twice daily, starting 24 h post AAV-Cre. b, A quantification of western blot band signal intensity (Extended Data Fig. 8a); n = 5 each group, unpaired two-tailed t-tests. c, Representative images of p21 IHC in kidney, brain and lung sections of ΔMdm2^Hep mice treated with vehicle or TGFβR1i (n = 3–13 mice for each group as stated below). d, A manual quantification of p21⁺ renal cortical cells; the data are presented as mean p21⁺ cells per field of view (FOV) for each mouse; n = 13 (biological replicates) mice for each group, two-tailed Mann–Whitney test. e, a Manual quantification of p21⁺ brain cells; the data are presented as p21⁺ cells per square millimetre, n = 3/5 vehicle/TGFβR1i treated, respectively; two-tailed Welch’s t-test. f, Automated quantification of p21⁺ lung cells; n = 7/8 vehicle/TGFβR1i treated, respectively; two-tailed Mann–Whitney test. g, Plasma cystatin C of vehicle or TGFβR1i treated at cull. Each dot represents the data from one mouse; n = 7 mice for each group, unpaired two-tailed t-test. h,i, The levels (peak area ratio to creatinine) of alanine (h) and threonine (i) in the urine of ΔMdm2^Hep mice before and after AAV-Cre injection; n = 4/5 vehicle/TGFβR1i treated, respectively, per timepoint, two-tailed Welch’s t-test comparing vehicle- or TGFβR1i-treated mice at day 4. j, A schematic of the genetic ablation of TGFβ signalling in the kidney. TGFBR1(Alk5)^fl/fl ± AhCre^+/− mice were administered either 2 × 10¹¹ GC ml⁻¹AAV-Cre (ΔAlk5^Liver) or 3 x 80 mg kg⁻¹β-NF (ΔAlk5^Kidney), respectively, or control induction (2 × 10¹¹ GC ml⁻¹AAV-TBG-Null/vehicle, respectively) and were collected 4 days later. k,l, Automated quantification of p21⁺ liver cells (k) and manual quantification of p21⁺ renal cortical cells (l) in ΔAlk5^Liver and ΔAlk5^Kidney mice; the data are presented as percentage of total liver cells or as p21⁺ cells per FOV. n = 6/8 for ΔAlk5^Liver/ΔAlk5^Kidney mice, respectively; Brown–Forsythe and Welch ANOVA test (liver) or one-way ANOVA (kidney). The bars are the mean ± s.e.m., and the numbers on the graphs are P values. The scale bars are 50 μm and 5 μm in the inset magnifications. The source numerical data are available in Source data.

Extended Data Fig. 1. The AAV8-TBG-Cre vector is hepatocyte-specific and does not result in notable genetic recombination of extra-hepatic tissues.

(a) Representative IHC images of liver (left), kidney (middle) and brain (right) sections of control and ΔMdm2^Hep mice stained for RFP and p53; mice were culled 4 days post AAV injection. n=4 and 5 control and ΔMdm2^Hep mice, respectively. (b), (c), (d) Automated quantification of p53⁺ cells and RFP⁺ area on liver, renal cortex and brain tissue respectively, n=4 and 5 control and ΔMdm2^Hep mice, respectively, in all graphs. For panel (b), unpaired two-tailed t-test (RFP) and Mann-Whitney test (p53) were used. For panels (c) and (d), two-tailed Welch’s t-test (RFP) and two-tailed Mann-Whitney test (p53) were used. (e) Representative images from ISH-stained kidney sections with a probe that specifically detects the junction of murine Mdm2 exons 5 and 6. (f) Normalised RFP transcript counts (FPKMs) from bulk liver and bulk kidney RNA sequencing data. n=4 except n=3 in liver control; unpaired two-tailed t-test. (g) Manual quantification of P21⁺ renal tubular cells from either 8–12 weeks old male WT or male Mdm2^E5/E6fl; R26^{LSL-tdTomato/LSL-tdTomato} (Mdm2^flox) mice, untreated (-) or intravenously injected with either 2x10¹¹ GC/mouse AAV8-TBG-Cre (Cre) or with AAV8-TBG-Null (Null) and culled 4 days later: additional control data from Fig. 1e are presented as p21⁺ tubular cells per field of view (FOV), n=4 and 8 for untreated vs Cre in WT mice and n=16 and 19 Null and Cre treated ΔMdm2^Hep mice; one way Anova, adjusted p value denotes Mdm2 Cre vs Null; all other comparisons p>0.99. Throughout all bars are mean ± S.E.M and the numbers on the graphs are p values. Scale bars are 50μm. Source numerical data are available in source data.

Extended Data Fig. 2. Liver senescence is associated with renal senescence.

(a) Representative images of control and ΔMdm2^Hep liver sections stained for SA β-Gal; n=8 mice per group. (b) Targeted Gene Set Enrichment Analysis (GSEA) for a senescence signature on the whole liver RNA-seq dataset performed on whole liver RNA isolated from day 4 kidneys in the ΔMdm2^Hep model and its control. (c) RT-qPCR analysis of Cdkn1a, Cdkn2a, Cdkn2b, Bcl2, Tgfb1, Tgfb2, Tgfb3 and Lif expression in whole liver lysates in the ΔMdm2^Hep model. n=4 mice for Tgfb1, Tgfb2 and Tgfb3 and n=5 in all others; n=4 in Tgfbs due to misloading of one sample. Unpaired two-tailed t test for Cdkn1a, Cdkn2a, Tgfb1, Tgfb3 and Bcl2; two-tailed Welch’s t test for Tgfb2, Lif and Cdkn2b (unselected samples from larger cohorts; see Fig. 1 for details). (d) Representative images of control and ΔMdm2^Hep model kidney sections stained for SA β-Gal. n=8 mice per group (unselected samples from larger cohorts). (e) Targeted GSEA for a senescence signature on the whole kidney RNA-seq dataset performed on whole liver RNA isolated from day 4 kidneys in the ΔMdm2^Hep model and its control. (f) Cytokine arrays on whole kidney lysates for a range of SASP factors in the ΔMdm2^Hep model. n=4 control mice and n=5 ΔMdm2^Hep mice (unselected samples from larger cohorts). Unpaired two-tailed t-test (IL-17, LIF, GM-CSF, CXCL-1 and CCL3) or two-tailed Mann-Whitney test (CCL-2, CXCL-5). (g) Alanine aminotransferase (ALT) and alkaline phosphatase (ALP) in the plasma of ΔMdm2^Hep and control mice, n=7 and 14 control and ΔMdm2^Hep mice (unselected samples from larger cohorts; samples failing quality control were excluded), respectively, for both graphs; two-tailed Mann-Whitney test (ALT) or two-tailed Welch’s t-test (ALP). (h) Plasma bilirubin in ΔMdm2^Hep and control mice, n=9 and 11 control and ΔMdm2^Hep mice respectively (unselected samples from larger cohorts, samples failing quality control were excluded); two-tailed Mann-Whitney test. (i) Representative images of cleaved caspase 3 (CC3) IHC on liver sections of ΔMdm2^Hep and control mice. Arrowheads highlight CC3⁺ cells. (j) Automated quantification of CC3 IHC on liver sections: data are presented as CC3⁺ area as a percentage of total liver area, n=6 and 7 control and ΔMdm2^Hep mice respectively (unselected samples from larger cohorts); two-tailed Welch’s t-test. (k) Schematic of the CCL₄ treatment of WT mice. C57Bl/6 mice were injected with CCL₄ and were sacrificed 2, 3, 7 or 14 days post injection. (l) Representative images of p21-stained liver and kidney sections of CCL₄-injected mice. (m) Automated quantification of p21⁺ liver cells on liver sections or (n) manual quantification of p21⁺ renal tubular cells on kidney sections of the CCL₄-injected mice. n=5 mice for each timepoint except day 0 where n=4; One-way ANOVA. (o) Schematic of AAV-mediated induction of the Kras^G12D mice. 8–12 weeks old mice were injected with either AAV-Null or AAV-Cre and were sacrificed 7 days post induction. (p) ALT and ALP levels in the plasma of Kras^WT and Kras^G12D mice. n=7 Kras^WT and n=6 Kras^G12D mice for both ALT and ALP. Two-tailed Mann-Whitney test (ALT) or unpaired two-tailed t-test (ALP). (q) Plasma bilirubin in Kras^WT and Kras^G12D mice, n=7 Kras^WT and n=6 Kras^G12D mice; two-tailed Mann-Whitney test. Normalised liver (r) and kidney (s) Cdkn1a reads in Kras^WT versus Kras^G12D mice from bulk RNAseq; n=5 mice in both groups (unselected samples from the larger cohort); unpaired two-tailed t-tests. (t) Targeted GSEA for a senescence signature on the whole RNA-seq dataset of Kras^WT versus Kras^G12D kidneys. (u) Top-16 upregulated pathways in an unsupervised GSEA performed on Kras^WT versus Kras^G12D kidneys. For all graphs, mice were culled 4 days post AAV injection. Throughout all bars are mean ± S.E.M. and the numbers on the graphs are p values. Scale bars are 50μm. Source numerical data are available in source data.

Extended Data Fig. 3. The intensity of the renal senescence-like phenotype in ΔMdm2^Hep mice is proportionate to the level of liver senescence.

(a) Schematic of AAV dose titration in ΔMdm2^Hep mice in the recovery-Mdm2^Hep. 8–12 weeks old male Mdm2^E5/E6fl; R26^{LSL-tdTomato/LSL-tdTomato} mice were either injected with 5x10¹⁰ GC/mouse of AAV-Cre and then harvested 4, 7, 30 or 62 days post injection, or they were injected with 5x10⁹ GC of AAV-Cre and harvested 7 days post injection. (b) Representative images of p21-stained liver and kidney sections from the AAV dose titration experiment. (c) Automated quantification of p21⁺ liver cells on liver sections from the AAV dose titration experiment; data are presented as percentage of total liver cells. n=4,5,4,7,8,4 and 9 from left to right;; subgroup members of the 2x10¹¹ AAV-Cre and controls were analysed, and all member of the recovery model are included. (d) Manual quantification of p21⁺ renal tubular cells: data are presented as mean p21⁺ tubular cells per 20 fields of view (FOV); n=16,19, 4,4,3,3 and 9 from left to right; whole cohorts of 2x10¹¹ AAV-Cre and controls are included (data for day 4 is shown also in Fig. 1e), the recovery model was performed in smaller cohorts with n= 3-9 with all biological replicates assessed included in analyses. (e) Serum levels of ALT, ALP and Bilirubin in the mice of the AAV dose titration experiment. Biological replicates (separate animals) n=7, 14, 3, 4, 3 and 4 for ALT and ALP and 9, 11, 3, 4, 3 and 4 for Bilirubin from conditions from left to right (f) Urine levels of Alanine, Phenylalanine and Tryptophan identified by liquid chromatography-mass spectrometry (LC-MS) in ΔMdm2^Hep mice pre- and post- AAV-Cre injection: n=4 mice. Throughout all bars are mean ± S.E.M. Scale bars are 50μm. Source numerical data are available in source data.

Extended Data Fig. 4. Renal and brain dysfunction in response to liver senescence.

(a) Urine levels of Glutamine, Serine and Valine identified by liquid chromatography-mass spectrometry (LC-MS) in ΔMdm2^Hep mice pre- and post- AAV-Cre injection: dots represent the average peak area of n=3 mice at days -2 and 0, n=4 mice at day 4 and n=5 mice at days -1 and 3; 2-way ANOVA comparing each time point to induction day (day 0). (b) Proportion of time spent by control and ΔMdm2^Hep mice in the novel arm of the Y-maze 4 days before and 4 days after AAV injection: dots represent the average percentage of time spent in the novel arm for n=7 and 9 for control and ΔMdm2^Hep mice respectively; 2-way ANOVA. Area under the curve (area power) (c) and frequency (Hz) (d) of the brain slice oscillations after stimulation with carbachol. n=14 brain slices (except n=12 at times 0,15,60 and 105) from 4 control mice and n=14 brain slices from 4 ΔMdm2^Hep mice for every time point. Throughout all bars are mean ± S.E.M. and the numbers on the graphs are p values. Source numerical data are available in source data.

Extended Data Fig. 5. Sub-clustering and cell type assignment in the scRNA-seq data.

(a) scRNA-seq analysis of 3 kidneys from control mice and 3 kidneys from ΔMdm2^Hep mice taken 4 days after induction. After single cell sequencing cells were filtered for quality based upon the number of genes expressed and percentage of mitochondrial gene reads in each event. Bar charts are shown, by each sequencing batch, total number of reads per cells (left), with cells filtered out shown in grey and those included in analyses shown in purple based upon number of reads (centre; <75 or >3000) and % mitochondrial genes/total (right; >50%). (b) UMAP plot of 24,215 single-cell transcriptomes from 6 mouse kidneys. (c) Table with violin plots showing the expression levels of marker genes across the 8 clusters. (d) UMAP plot of the cells following sub-clustering performed upon the mixed cluster shown in Fig. 3b. (e) Table with violin plots showing the expression levels of marker genes across the 6 subclusters within the mixed (#3) cluster. (f) Bar plot showing the number of RFP UMIs per cell in control and ΔMdm2^Hep cells. Source numerical data are available in source data.

Extended Data Fig. 6. Liver senescence results in transcriptional changes in the renal proximal tubular cell compartment.

(a) Dot plot of pathways resulting from unsupervised GSEA of differentially expressed genes between ΔMdm2^Hep and control PTCs. Differential expression analysis was performed between ΔMdm2^Hep and control PTCs and the ranked gene set was used to perform unsupervised GSEA against Gene Ontology (GO) and KEGG pathways. The size of the dots represents the number of downregulated (suppressed, left) or upregulated (activated, right) genes in the ΔMdm2^Hep PTCs for each pathway (count) with colour representing statistical significance; Permutation test. (b) Representative images from a single batch of dual immunofluorescence staining for p21 and renal tubular epithelial markers (LRP2 or CALB1) on ΔMdm2^Hep kidneys at day 4; LRP2 is expressed on the apical brush boarder of the PTCs; n-5/6 for control/ ΔMdm2^Hep respectively. Scale bars are 50μm and 5μm in inset magnification. (c) UMAP plot showing the clusters of the cells (red versus grey) of the scRNA-seq data from analysis of 3 kidneys from control mice and 3 kidneys from ΔMdm2^Hep mice taken 4 days after induction highlighting cells with a positive score for the senescence signature (sen) from O’Sullivan et al.. (d) UMAP plot of the same cells shown in panel c highlighting co-occurrence (yellow) of the p21 signatures (Fig. 3c) and the renal tubular senescence signature from O’Sullivan et al.. (e) Heatmap of relative expression of Slc transporter genes in the PTC compartment grouped by p21 signature score in each experimental condition; PTCs without the p21 signature in both conditions are merged as the left-hand column. Source numerical data are available in source data.

Extended Data Fig. 7

Liver-derived SASP factors affect the TGFΒ and LIF/JAK-STAT signalling pathways in the kidney. (a) GSEA plot for a SASP gene set on the significant differentially expressed genes from the bulk liver RNA-seq data in the ΔMdm2^Hep model versus control at day 4. (b) UMAP plots showing the distribution of the cells that have a positive score for the TGFβ gene signature in the control (left) and the ΔMdm2^Hep model (right) renal cells. Inset images are representative kidney sections stained for Smad7 by ISH (RNAscope) n=3/3 for control/ ΔMdm2^Hep respectively, dashed lines highlight renal tubules. (c) Representative images of dual p21 IHC/Smad7 ISH in the kidney; pink arrows highlight examples of Smad7 detection in red, whilst p21 is detected by brown 3,3′-Diaminobenzidine (DAB) staining; n=3/3 for control/ ΔMdm2^Hep respectively. (d) UMAP plots showing the distribution of the cells that have a positive score for the JAK-STAT gene signature in the control (left) and the ΔMdm2^Hep model (right) renal cells. Inset images are representative kidney sections stained for pSTAT3 by IHC; n=3/3 for control/ ΔMdm2^Hep respectively. Scale bars are 50μm and 5μm in inset magnification. (e) Automated quantification of pSTAT3⁺ in renal cortical cells; data are presented as percentage of total cortical cells, n=6 and 7 in control and ΔMdm2^Hep mice respectively; two-tailed Mann-Whitney test. Bars are mean ± S.E.M. and the numbers on the graphs are p values. (f) GSEA plot for a TGFβ signalling pathway gene set on the significant differentially expressed genes from the bulk liver RNA-seq in the ΔMdm2^Hep model versus control at day 4. (g) Western blots for pSMAD2 and pSMAD3 and their respective total protein on whole kidney lysates from ΔMdm2^Hep model and control mice at day 4. n=5 mice in each group; see experimental schematic in Fig. 1a. 1 gel was run for pSMAD2/SMAD2 and another one for pSMAD3/SMAD3. β-actin was used as a loading control on both gels. (h) Representative images of ISH (RNAScope) for Tgfb1 and Tgfb2 on liver sections of ΔMdm2^Hep and control mice detected by DAB (brown); n=3/3 for control/ ΔMdm2^Hep respectively for each ISH. (i) Representative images of dual p21 IHC/Tgfb1 ISH (RNAScope) in the liver; Tgfb1 ligand detection in red, whilst p21 is detected by brown DAB staining; n=3/3 for control/ ΔMdm2^Hep respectively. Normalised Tgfbr1, Tgfbr2 and Tgfbr3 transcript counts (FPKMs) from bulk liver (j) and kidney (k) at day 4 in the ΔMdm2^Hep model versus control, n=4 except n=3 in liver control; both unpaired two-tailed t-test. All bars are mean ± S.E.M and the numbers on the graphs are p values. (l) Representative images of ISH (RNAScope) for Tgfbr1 on kidney sections of ΔMdm2^Hep model and control mice at day 4 detected by brown DAB staining; n=3/3 for control/ ΔMdm2^Hep respectively. Dashed lines highlight renal tubules. Scale bars are 50μm. Source numerical data and unprocessed blots are available in source data.

Extended Data Fig. 8. Systemic inhibition of the TGFβ signalling pathway does not affect liver senescence; see experimental schematic in Fig. 5a.

(a) Western blots for pSMAD2, pSMAD3 and their non-phosphorylated forms on whole kidney lysates from vehicle- and TGFβR1i-treated ΔMdm2^Hep mice. n=5 mice in each group. 1 gel was run for pSMAD2/SMAD2 and another one for pSMAD3/SMAD3. β-actin was used as a loading control on both gels. (b) Representative images of p21 IHC on liver sections of vehicle- and TGFβR1i-treated ΔMdm2^Hep mice. (c) Automated quantification of p21⁺ liver cells on liver sections of vehicle- and TGFβR1i-treated ΔMdm2^Hep mice: data are presented as percentage of total liver cells, n=6 mice for each group; unpaired two-tailed t-test. (d) Representative images of p21 IHC on kidney sections of vehicle- and TGFβR1i-treated Kras^G12D mice. (e) Manual quantification of p21⁺ renal tubular cells in vehicle- and TGFβR1i-treated Kras^G12D mice: data are presented as mean p21⁺ tubular cells per 20 FOVs, n=6 and 7 vehicle- and TGFβR1i-treated Kras^G12D mice respectively; two-tailed Welch’s t-test. (f) Urine levels of Serine in ΔMdm2^Hep model over time, before and after AAV-Cre injection (Day 0). n=4 and 5 vehicle- and TGFβR1i-treated mice, respectively, per time point; two-tailed Welch’s t-test comparing vehicle to TGFβR1i-treated mice at day 4. (g) Automated quantification of p21⁺ liver cells on liver sections of vehicle and rapamycin-treated ΔMdm2^Hep model mice 4 days after AAV induction (2x10¹¹ GC/mouse); data are presented as percentage of total liver cells, n=5 and n=6 vehicle- and rapamycin-treated mice respectively; two-tailed Mann-Whitney test. (h) Manual quantification of p21⁺ renal cortical cells on kidney sections of vehicle- and rapamycin-treated ΔMdm2^Hep mice: n=5 and n=6 vehicle- and rapamycin-treated mice respectively; two-tailed Welch’s t-test. (i) Automated quantification of p21⁺ lung cells on lung sections of vehicle- and rapamycin-treated ΔMdm2^Hep mice: data are presented as percentage of total lung cells, n=5 and n=6 vehicle- and rapamycin-treated mice respectively; two-tailed Welch’s t-test. Throughout all bars are mean ± S.E.M. and the numbers on the graphs are p values. Scale bars are 50μm. Source numerical data and unprocessed blots are available in source data.