Targeting senescent hepatocytes for treatment of metabolic dysfunction-associated steatotic liver disease and multi-organ dysfunction

Du, Kuo; Umbaugh, David S.; Wang, Liuyang; Jun, Ji Hye; Dutta, Rajesh K.; Oh, Seh Hoon; Ren, Niansheng; Zhang, Qiaojuan; Ko, Dennis C.; Ferreira, Ana; Hill, Jon; Gao, Guannan; Pullen, Steven S.; Jain, Vaibhav; Gregory, Simon; Abdelmalek, Manal F.; Diehl, Anna Mae

doi:10.1038/s41467-025-57616-w

Download PDF

Article
Open access
Published: 28 March 2025

Targeting senescent hepatocytes for treatment of metabolic dysfunction-associated steatotic liver disease and multi-organ dysfunction

Nature Communications volume 16, Article number: 3038 (2025) Cite this article

14k Accesses
11 Citations
12 Altmetric
Metrics details

Subjects

An Author Correction to this article was published on 06 May 2025

This article has been updated

Abstract

Senescent hepatocytes accumulate in metabolic dysfunction-associated steatotic liver disease (MASLD) and are linked to worse clinical outcomes. However, their heterogeneity and lack of specific markers have made them difficult to target therapeutically. Here, we define a senescent hepatocyte gene signature (SHGS) using in vitro and in vivo models and show that it tracks with MASLD progression/regression across mouse models and large human cohorts. Single-nucleus RNA-sequencing and functional studies reveal that SHGS+ hepatocytes originate from p21+ cells, lose key liver functions and release factors that drive disease progression. One such factor, GDF15, increases in circulation alongside SHGS+ burden and disease progression. Through chemical screening, we identify senolytics that selectively eliminate SHGS+ hepatocytes and improve MASLD in male mice. Notably, SHGS enrichment also correlates with dysfunction in other organs. These findings establish SHGS+ hepatocytes as key drivers of MASLD and highlight a potential therapeutic strategy for targeting senescent cells in liver disease and beyond.

Bioinformatic analysis of molecular expression patterns during the development and progression of metabolic dysfunction-associated steatotic liver disease (MASLD)

Article Open access 01 March 2025

A male mouse model for metabolic dysfunction-associated steatotic liver disease and hepatocellular carcinoma

Article Open access 02 August 2024

Spatial genomics: mapping human steatotic liver disease

Article 23 April 2024

Introduction

Although metabolic dysfunction-associated steatotic liver disease (MASLD) has become the leading cause of chronic liver disease worldwide, its pathogenesis remains unclear. This knowledge gap has limited development of effective pharmacotherapies and hence, MASLD is driving the health and economic burden of liver disease^1,2. The problem is projected to worsen because the world’s older population is growing, and aging exacerbates MASLD^3,4. A key hallmark of aging is cellular senescence; organs progressively lose functional mass and become fibrotic as these non-regenerative cells accumulate. While this helps to explain the adverse effects of aging on MASLD^5,6, it has not yet translated into more effective therapies despite the availability of various senolytic agents.

Cellular senescence is a permanent state of cell cycle arrest induced by various stressors, including oxidative stress, DNA damage, and metabolic disturbances^7,8,9. However, it is difficult to use this definition to identify truly senescent cells because progression through the cell cycle can also be halted transiently to permit cellular repair and/or reprogramming^10,11 and determining the stability of cell cycle arrest is challenging. Further, the factors mediating cell cycle arrest vary depending on where in the cell cycle growth is arrested and why¹¹. Hence, relying on a single parameter (e.g., cell cycle arrest) to define senescent cells obscures important phenotypic differences in cell state. In turn, this limits insight about how cells that are in different growth-arrested states influence recovery of health in individual organs, as well as the net impact of senescence in one organ on systemic homeostasis. These points merit emphasis because senescent cells exhibit robust secretory phenotypes (known as senescence-associated secretory phenotypes - SASPs) that are diverse, although broadly characterized by the secretion of pro-inflammatory cytokines, chemokines, growth factors, and proteases^7,8,9. In healthy organs, senescence is a homeostatic autocrine tumor-suppressive mechanism that triggers compensatory proliferation of non-transformed cells that retain regenerative capacity. However, in chronically diseased organs, the SASPs of senescent cells generate signals that drive tissue dysfunction and, thereby, contribute both to disease progression and carcinogenesis^7,8,9. Understanding how cellular senescence impacts disease pathogenesis is further confounded by the fact that senescence can occur in multiple cell types depending on the injury stimulus and stage of liver disease^12,13,14.

Senolytics emerged as potential therapies for age-related chronic diseases in 2015¹⁵ and have demonstrated beneficial effects in certain liver diseases. However, their efficacy in targeting senescent hepatocytes has been inconsistent. For instance, senolytic CAR-T cells reversed MASLD and alleviated biliary fibrosis in mice by selectively ablating senescent cholangiocytes¹⁶. On the other hand, widely used senolytics such as dasatinib and quercetin (D + Q) failed to reduce hepatocyte senescence in models of acute liver injury, including acetaminophen hepatotoxicity¹⁷. Further, they exacerbated liver disease in other contexts, such as in the DEN/HFD mouse model¹⁸. Similarly, Bcl-xl inhibitors A-1331852 and ABT263 were ineffective in reducing hepatocyte senescence or fibrosis in CCl₄-induced liver injury¹⁹. These seemingly contradictory findings are consistent with recent reviews which emphasize that the phenotypes of senescent cells are dynamic and vary with cell type and senescence-inducing stressor^20,21. This heterogeneity has resulted in a paucity of data regarding the pathobiological relevance of different senescent states in diverse cell types^22,23,24,25, and highlights the need for novel senolytics specifically designed to target senescent hepatocytes in MASLD.

Dysregulated hepatocyte senescence is thought to be a key pathogenic mechanism in MASLD, as hepatocytes constitute the majority of hepatic cells and are crucial for whole-body energy metabolism and homeostasis^26,27. Notably, markers of senescence in hepatocytes correlate with diabetes prevalence, fibrosis severity, and adverse outcomes in MASLD patients^6,28,29. These observations justify efforts to identify and characterize senescent hepatocytes, especially at the single-cell level in diseased livers. However, challenges associated with the lack of specific markers that reliably differentiate among hepatocytes that are in different states of permanent growth arrest, or that can distinguish permanently senescent hepatocytes from hepatocytes that are being held in non-replicative states temporarily^20,30, have hindered the field. Current methods for identifying senescent cells in diseased livers rely mostly on histological or immunostaining analyses of proteins that inhibit cell cycle progression (e.g., p16 and p21) without characterizing these cells comprehensively. Hence, the molecular mechanisms driving and maintaining hepatocyte senescence remain unclear, and whether targeting specific subpopulations of senescent hepatocytes is a promising and safe approach for treating MASLD is not known.

Here we address these challenges by using in vitro and in vivo models of hepatocyte senescence to construct a senescent hepatocyte gene signature (SHGS) that tracks with MASLD progression and treatment outcomes. We demonstrate how SHGS+ hepatocytes promote MASLD pathogenesis, reveal a serum biomarker that correlates with the liver burden of these cells, and identify novel senolytic agents that improve MASLD by eliminating SHGS+ hepatocytes.

Results

Accumulation of senescent cells is a conserved mechanism during MASLD pathogenesis

Our recent RNA-seq analysis of 368 liver biopsies in the Duke MASLD cohort revealed that genes involved in cellular senescence are enriched in MASLD patients compared to individuals without MASLD, and demonstrated that this enrichment is more pronounced in MASLD patients with advanced fibrosis than in those with mild fibrosis²⁵. Importantly, immunohistochemistry showed that the standard senescence markers, P21 and P16, are upregulated at the protein level in hepatocytes and positively correlate with MASLD severity, suggesting that hepatocyte senescence has a role in MASLD pathogenesis²⁵. To assess the generalizability of our findings, here we perform Gene Set Enrichment Analysis (GSEA) in diverse MASLD patient cohorts with different demographics. Genes involved in various senescence-associated pathways (Supplementary Fig. 1A–D), including CDKN1A (P21) and CDKN2A (P16) (Supplementary Fig. 1E–H), are enriched in the liver transcriptomes of MASLD patients versus healthy controls, MASH versus MASLD patients, and MASLD patients with advanced fibrosis (F3F4) versus mild fibrosis (F0F1).

Development of Senescence Hepatocyte Gene Signature (SHGS) to identify and characterize senescent hepatocytes during MASLD pathogenesis

Identifying and characterizing senescent hepatocytes in injured liver, especially at the single-cell level, has been challenging due to their heterogeneous nature and lack of canonical markers^20,30. To develop a robust gene panel associated with hepatocyte senescence, hepatocyte senescence was induced in vitro and in vivo to mimic chemotherapy-induced senescence and oncogene-induced senescence, respectively. In the in vitro model, Huh7 cells were treated with the CDK4/6 inhibitor palbociclib for 8 days (Fig. 1A). Palbociclib-treated Huh7 cells exhibited a robust senescent phenotype, as indicated by strong SA-β-Gal staining, cell cycle arrest, and increased expression of SASP factors and DNA damage markers such as γH2AX (Supplementary Fig. 2A–E). In the in vivo model, 12-week-old mice were injected with NRASG12V-IRES-GFP plasmid and CMV-SB13 through hydrodynamic tail vein injection. Six days later, GFP-positive (NRASG12V overexpressing hepatocytes) and GFP-negative (non-transduced hepatocytes) were isolated via low-pressure fluorescence-activated cell sorting following liver perfusion and subjected to RNA-seq analysis (GSE145642) (Fig. 1B). GSEA confirmed that the transcriptomes of both models are depleted with genes involved in liver-specific functions but enriched with genes associated with cellular senescence, SASP, oxidative stress, and DNA damage-induced senescence (Fig. 1C, D; Supplementary Fig. 2F–H). The paradoxical upregulation of MYC and E2F targets in senescent hepatocytes likely reflects an early stress response that supports metabolic adaptation and SASP activation despite stable cell cycle arrest^31,32. In addition, DEG profiles of both models significantly overlap with those from MASLD patients, and distinguish MASLD patients from healthy controls (GSE33814) and advanced fibrosis from mild fibrosis (GSE49541) by unsupervised clustering (Supplementary Fig. 3). By overlapping the upregulated differentially expressed genes (DEGs) from both datasets, we found that DEGs in the in vitro model with adjusted P < 0.01 and log2 FC > 1 and in the in vivo model with adjusted P < 0.05 and log2 FC > 0 had the best balance between gene set size and high Jaccard similarity index. The intersection resulted in 100 DEGs, which we defined as the Senescent Hepatocyte Gene Signature (SHGS) (Fig. 1E; Suppl. Table 1). Annotation of the SHGS genes for subcellular location revealed that the majority encode factors that are located in the plasma membrane, cytosol, nucleoplasm, or vesicles (Fig. 1F). The top 10 KEGG pathways enriched in SHGS included cellular senescence, cytokine-cytokine receptor interaction, extracellular matrix-receptor interactions, and TGFβ and PI3K-AKT signaling (Fig. 1F). Thus, SHGS captures the detrimental effects of cellular senescence that promote tissue fibrosis and inflammation, and signaling pathways that promote MASLD pathogenesis. Deconvolution analyses using SHGS in single-nucleus (sn) RNA sequencing data from 57 human samples across different disease stages further revealed that hepatocytes bearing SHGS increase in human livers during MASLD progression, with 8.1% of hepatocytes in livers with advanced fibrosis being SHGS+, compared with only 0.13% in healthy livers (Fig. 1G; Supplementary Fig. 4A, B). Hepatocyte enrichment with the SHGS is also seen in a mouse model of advanced MASH, where mice fed with a choline-deficient L-amino acid-defined high-fat diet (CDA-HFD) or control diet for 22 weeks showed 8.6% versus 0.7% SHGS+ hepatocytes, respectively (Fig. 1H; Supplementary Fig. 4C). The levels of senescent hepatocyte enrichment in human and mouse livers with MASLD fibrosis aligns with previous studies, which reported ~9% senescent epithelial cells in bleomycin-treated lungs and 3–6% senescent cells across tissues in aged mice³³. As expected, the transcriptomes of the SHGS+ hepatocytes are enriched with genes involved in cellular senescence and SASP (Fig. 1I), and GSEA further revealed that the transcriptomes of SHGS+ hepatocytes are relatively depleted of genes and pathways associated with hepatocyte-specific metabolic functions (e.g., bile acid metabolism, xenobiotic metabolism, fatty acid metabolism), but display enrichment with genes and pathways related to inflammation, extracellular matrix formation, cell reprogramming and remodeling (e.g., epithelial-to-mesenchymal transition (EMT), axon guidance), cell death and stress pathways (e.g., hypoxia, apoptosis, and MAPK signaling), and liver-related and systemic organ dysfunction (e.g., cardiomyopathy and cancer) (Fig. 1J, K; Supplementary Fig. 4D, E). To further evaluate the physiological relevance of the SHGS with respect to that of the widely used senescence marker P21, we classified hepatocytes into three categories (P21-, P21+, and SHGS+) and compared relative enrichment for senescence-defining factors as specified in recent guidelines³⁴. The SHGS+ hepatocyte population exhibits greater overall enrichment for hallmarks of senescence than the P21+ hepatocyte population (i.e., lower levels of cell proliferation and metabolic function and higher levels of genes involved in cell cycle arrest, SASP, oxidative stress, lysosomal alterations, DNA damage), indicating that the SHGS+ hepatocyte population is generally more senescent than the P21+ population (Supplementary Fig. 5). These results complement the above analysis in bulk RNA-seq datasets and together, strongly suggest that senescent hepatocytes contribute to metabolic dysfunction, injury, and maladaptive repair in livers with MASLD. It is important to note that neither CDKN1A nor CDKN2A is included in the SHGS gene list. However, both genes are highly significant (p < 0.001, with log2FC values of 2.33 and 2.7, respectively) in SHGS+ versus SHGS- hepatocytes. In addition, we did not use our recently developed AHGS (Aging Hepatocyte Gene Signature)³⁵ to construct the SHGS, as the characteristics of senescent cells can vary depending on the stimulus that induces senescence, and the AHGS was derived from hepatocytes in the context of healthy aging. Furthermore, while aging is associated with an increase in senescent cells, it is a continuous process that begins at birth and evolves over time. Thus, we reasoned that the AHGS captures a mixed population of hepatocytes, including those that are not senescent and may never become senescent, as well as senescent hepatocytes. Consistent with this concept, compared to the AHGS+ hepatocytes³⁵, we found that SHGS+ hepatocytes localize in the same clusters in MASLD livers, but only a small subset of AHGS+ hepatocytes are also SHGS+ (Fig. 1G, H). Thus, the aggregate data support the hypothesis that the SHGS specifically captures terminally senescent hepatocytes, whereas the AHGS reflects a broader transcriptomic profile encompassing a spectrum of aging-related changes in hepatocytes.

**Fig. 1: Development of Senescent Hepatocyte Gene Signature (SHGS) to identify and characterize senescent hepatocytes during MASLD pathogenesis.**

SHGS tracks with MASLD progression, regression, and clinical outcomes

Recent studies have demonstrated that the development of disease mechanism-driven gene signatures serves as a valuable tool for disease diagnosis, prognosis, monitoring, and risk assessment^{36,37,38,39,40}. Since our findings in both bulk and snRNA-seq analyses demonstrate that the SHGS captures pathogenic mechanisms of MASLD (e.g., inflammation, extracellular matrix formation, cell death), we hypothesized that SHGS enrichment might increase with disease progression and decrease with treatment response. To test this, we first performed deconvolution analysis in the bulk liver RNA-seq data from the Duke MASLD cohort and found that SHGS is significantly enriched in MASLD patients compared to healthy controls (Fig. 2A). Interestingly, we observed a positive correlation between BMI and liver enrichment with SHGS in both controls and patients with MASLD (Supplementary Fig. 6A). This was anticipated because obesity is known to promote hepatic metabolic stress that can induce senescence^41,42. Unexpectedly, while cellular senescence is a hallmark of aging, liver enrichment with SHGS did not increase with chronological age in controls but did in MASLD patients (Supplementary Fig. 6B), suggesting that SHGS captures MASLD-related pathogenic responses that are sensitive to the aging process. In support of this, SHGS enrichment scores correlated negatively with albumin levels and positively with serum AST and Fib4 scores (Fig. 2B). The scores also increased progressively with steatosis, hepatocyte ballooning, portal inflammation, and fibrosis severity, demonstrating that SHGS tracks closely with MASLD progression (Fig. 2C).

To determine the generalizability of SHGS, we used it to deconvolute diverse independent MASLD transcriptomic datasets from Germany, Japan, the USA, and Europe. In all these cross-sectional datasets, SHGS is more enriched in the transcriptomes of MASH vs MASLD patients, and this enrichment is more pronounced in MASLD patients with advanced fibrosis compared to those with mild fibrosis (Fig. 2D). Additional analyses revealed that SHGS has prognostic relevance as it was more enriched in MASLD patients with a higher risk for either primary or recurrent HCC (Fig. 2E). To assess if SHGS tracks with outcomes of MASLD therapeutic interventions, we applied it to a longitudinal dataset acquired from MASLD patients who underwent bariatric surgery⁴³. Improvement in MASLD after surgery was paralleled by reduction of SHGS enrichment scores in the liver transcriptomes (Fig. 2F). SHGS enrichment scores also were somewhat reduced in a different cohort of patients undergoing statin treatment, which has demonstrated beneficial effects on MASLD histology⁴⁴ (Fig. 2G). We next assessed whether interventions that change MASLD severity influence the SHGS in preclinical models. Aged mice display exacerbated MASH compared to young mice in a CDA-HFD-induced MASH model, and this can be reversed by the ferroptosis inhibitor Ferrostatin-1 (Fer1)³⁵. Re-analysis of that dataset shows that SHGS enrichment scores are also dramatically higher in old mice with CDA-HFD diet-induced MASH than young mice fed the same diets and indicates that SHGS enrichment is reversed by Fer1 treatment (Supplementary Fig. 6C), accurately reflecting Fer1-induced changes in liver phenotype. Importantly, we find that SHGS scores are also elevated in the GAN-MASH model⁴⁵. Additional analysis revealed that the SHGS is reduced by dietary intervention and GLP1 agonist semaglutide, treatments that improve MASH in people⁴⁶. The PPARα/γ agonist lanifibranor also showed a trend toward reducing SHGS in the GAN-MASH model, although this change did not reach statistical significance (Supplementary Fig. 6D).

Because 13 previously-reported senescence- and aging-related gene signatures are available to track cell senescence, it was important to compare the gene composition and performance of SHGS to that of the other signatures. SHGS has relatively small gene set size and exhibits minimal overlap with these other signatures (Fig. 3A, B; Supplementary Fig. 7A). Given that SenMayo is one of the most widely recognized senescence gene signatures⁴⁷, we performed an in-depth comparison between SHGS and SenMayo. The two signatures shared only 7 genes, highlighting their distinct composition. However, pathway enrichment analyses revealed strong overlaps in biological processes (Supplementary Figs. 7B, C), indicating that SHGS captures broader senescence mechanisms while retaining liver-specific relevance. Importantly, SHGS consistently outperformed SenMayo, SenNet, and other state-of-the-art senescence signatures in predicting MASLD occurrence and progression (Fig. 3C–J). Across eight comparisons spanning four independent MASLD cohorts, SHGS achieved the highest AUC values in four comparisons, ranked second in two, and placed fourth and fifth in the remaining two, demonstrating both its robustness and superior predictive performance (Fig. 3K). This benchmarking analysis is the first comprehensive evaluation of senescence signatures in MASLD cohorts, further solidifying SHGS as a robust and clinically relevant marker of hepatocyte senescence. Although this does not resolve whether accumulation of SHGS+ hepatocytes is a cause or a consequence of MASLD, the strength of the association suggests that secreted factors from SHGS+ hepatocytes might provide noninvasive biomarkers to track histological responses to clinical interventions for MASLD. The aggregate findings also justify further research to determine if (and why) the benefits of these diverse interventions derive from their ability to reduce the liver burden of SHGS+ hepatocytes.

**Fig. 3: Comparative evaluation of SHGS against 13 senescence and aging-related gene signatures across MASLD cohorts.**

SHGS+ hepatocytes are major source of GDF15, a serum protein that reflects MASLD severity

Cellular senescence is an adaptive stress response that is ultimately implemented by autocrine factors. In turn, cells with SASPs generate arrays of juxtacrine, paracrine, and endocrine signals that reconfigure the phenotypes other cells to orchestrate integrated stress responses within and among tissues^7,8,9. We used cell-to-cell interaction analysis⁴⁸ to characterize the SASP of SHGS+ hepatocytes and identify its interactome (Supplementary Fig. 8A). Interestingly, we found that SHGS+ hepatocytes generate a signaling factor, growth differentiation factor 15 (GDF15), that is able to engage all liver cell types by interacting with TGFBR2 on the signal-receiving cells (Supplementary Fig. 8A). Notedly, SHGS+ hepatocytes are the primary source of GDF15 signaling (Supplementary Fig. 8B). These cells utilize GDF15 for both autocrine and paracrine communication, with particularly strong paracrine interactions observed between SHGS+ hepatocytes and endothelial cells, as well as among SHGS+ hepatocytes themselves (Supplementary Fig. 8B). This cellular cross-talk may have pathobiological relevance because GDF15 is a mitokine that is induced during integrated mitochondrial stress responses in MASH⁴⁹, as well as a conserved component of SASPs^50,51,52. Thus, by releasing GDF15, SHGS+ hepatocytes convey hepatocyte distress to other cell types. Further, GDF15 is a generally accepted biomarker of biological aging because its serum levels increase with chronological age and aging-exacerbated diseases^41,53. Consistent with these concepts, snRNA-seq analysis demonstrates that GDF15+ hepatocytes accumulate in MASLD livers with advanced fibrosis (Supplementary Fig. 8C), and GSEA demonstrates that the transcriptomes of GDF15+ hepatocytes are significantly enriched with pathways involved in cellular senescence and the SASP (Supplementary Fig. 8D). Furthermore, GDF15 enrichment mainly occurs in the subpopulation of hepatocytes that are SHGS+ in MASLD (Supplementary Fig. 8E) and expression of GDF15 mRNA is significantly increased by SHGS+ induced hepatocytes in vitro (Supplementary Fig. 9A). Importantly, analysis of human bulk RNA seq datasets indicates that liver enrichment with GDF15+ hepatocytes tracks with accumulation of SHGS+ hepatocytes as both the liver burden of SHGS+ hepatocytes and GDF15 mRNA levels increase with MASLD severity, decrease with improvement of MASLD after bariatric surgery and increase with HCC risk (Supplementary Figs. 9B–D). Further, serum levels of GDF15 protein reflect changes in liver GDF15 mRNA levels, as they are higher in patients with MASH than in patients with MASL and highest in those with F3-4 fibrosis (Supplementary Fig. 9E). Notably, serum GDF15 levels differentiate MASLD patients from healthy controls (AUC: 0.79), and those with advanced fibrosis versus control (AUC: 0.85) or at-risk MASLD F2 liver (AUC: 0.78) (Supplementary Fig. 9F). Together, these new data suggest that changes in serum GDF15 levels parallel senescent hepatocyte accumulation and thus, might be a predictive non-invasive biomarker for senescence-related pathology, including progressive liver fibrosis. Interestingly, GDF15 was recently identified as a predictor of age-related mortality^52,53,54,55. However, GDF15 has been reported to exert both beneficial^56,57,58 and detrimental effects^59,60,61 on the liver, underscoring the need for further investigation. Given that the secretome of SHGS+ hepatocytes is complex, additional research is essential to elucidate how GDF15 interacts with other factors in these secretomes to modulate metabolic stress responses across diverse liver cell populations.

Senescent hepatocytes induce pathogenic reprogramming of neighboring cells

To examine the paracrine effects of SGHS+ hepatocytes, we subjected Huh7 cells to the CDK4/6 inhibitor (palbociclib) protocol that we used to develop the SHGS signature (Fig. 1A). Results were compared to recipient cell cultures treated with conditioned medium (CM) from vehicle-treated Huh7 cells. The recipient cells included non-senescent hepatocytes (Huh7), hepatic stellate cells (LX2), macrophages (Raw264.7), and primary human liver sinusoidal endothelial cells (LSECs) (Fig. 4A). While CM from vehicle-treated cells had no effect on the recipient cell cultures, we found that CM from SHGS+ hepatocytes induced secondary senescence in recipient Huh7 cultures, as evidenced by altered expression of cell cycle/senescence markers, strong SA-β-Gal staining, and increased expression of SASP factors (Fig. 4B). These results indicate that hepatocyte senescence can self-reinforce and be transmitted to neighboring non-senescent hepatocytes. Additionally, CM from SHGS+ cells induced fibrogenesis in hepatic stellate cells (HSCs), as indicated by increased mRNA and protein expression of proliferative and profibrogenic markers (Fig. 4C, D). It also promoted macrophage activation, evidenced by increased expression of proliferative and proinflammatory markers (Fig. 4E, F), and caused capillarization in liver sinusoidal endothelial cells (LSECs), as indicated by altered expression of LSEC functional markers and impaired tube formation (Fig. 4G, H). Together, these findings demonstrate pathogenic paracrine effects of SHGS+ hepatocytes that are predicted to drive MASLD progression. Because MASLD is one component of a systemic metabolic disease that broadly disrupts cellular resiliency, it is important to assess whether modulating the burden of SHGS+ hepatocytes influences disease progression in intact organisms.

**Fig. 4: Senescent hepatocytes induce pathogenic reprogramming of neighboring hepatic cells through paracrine mechanisms.**

SHGS+ hepatocytes emerge from p21+ hepatocytes and represent an advanced stage of cellular senescence

To address this question, we began by exploring the relationship between P21+ hepatocytes and SHGS+ hepatocytes. Previous investigations have described P21 as an early event to halt cell cycle progression, particularly in the liver^17,62. Given that we used aggressive models of cell senescence to derive the SHGS signature, we reasoned that a hepatocyte trajectory may exist along this P21+ hepatocyte (e.g., early senescence) to SHGS+ hepatocyte (e.g., late senescence). To validate this, we evaluated the percentage of SHGS+ and SHGS- hepatocytes across fibrosis stage finding a robust increase in F4 human livers (Fig. 1C, 0.134% in fibrosis stage 0 and 8.11% in fibrosis stage 4). As such, we extracted the senescent hepatocytes (e.g., SHGS+ and/or p21+ hepatocytes), reprocessed them, and then took the UMAP embeddings as input for trajectory analysis with monocle3⁶³. Using this biologically informed approach, we mapped a trajectory finding that p21+/SHGS- hepatocytes give rise to two subpopulations of SHGS+ hepatocytes: an SHGS+ population that loses P21 expression and a SHGS+ population that remains p21+ (Supplementary Figs. 10A–C). We found that the SHGS+ populations expanded across fibrosis stages with the emergence of the SHGS/P21+ hepatocytes almost exclusively in patients with F4 fibrosis (Supplementary Figs. 10A–C). To evaluate the idea that these SHGS/P21+ hepatocytes represent an advanced stage of cellular senescence, we examined the ten hallmarks of senescence³⁴, finding a stronger induction of cell cycle inhibition, erosion of the nuclear envelope, activation of alarmins, DNA damage, a SASP, and oxidative damage, and lysosomal alterations compared to either P21+ or SHGS+ populations (Supplementary Fig. 10D). Collectively, this data indicates that there is heterogeneous population of senescent hepatocytes that progress along a senescence trajectory which can be mapped across human fibrosis stage.

P21 deficiency protects against MASLD

Consistent with bulk RNA seq evidence that P21+ hepatocytes accumulate with MASLD progression in people (Supplementary Fig. 1), snRNA-seq analyses of human cohorts and CDA-HFD fed mice identified some p21+ hepatocytes in control livers and demonstrated robust enrichment with p21+ hepatocytes in both human and mouse livers with MASLD fibrosis. In contrast, enrichment with p16 expressing hepatocytes was much less pronounced (Fig. 5A, B; Supplementary Fig. 11A–D). To further clarify the relationship between P21 and senescence, we carried out additional analysis of the snRNA-seq human datasets. The results show that SHGS+ hepatocytes comprise a greater percentage of P21+ rather than P21- cells. Conversely, the P21+ hepatocyte population is more enriched with SHGS+ hepatocytes than the SHGS- population (Supplementary Fig. 11E). GSEA indicates that similar to SHGS+ hepatocytes, p21+ hepatocyte populations are relatively enriched for transcripts linked to senescence and inflammation, and depleted of transcripts related to liver-specific functions in both human and mouse livers (Fig. 5C; Supplementary Figs. 12A–D). Together, these data suggest that P21+ hepatocytes contribute to metabolic dysfunction and impaired liver repair in MASLD by enabling accumulation of senescent hepatocytes. Supporting this concept, most cells positive for P21 and/or SHGS are found in MASLD livers with advanced fibrosis (Figs. 1G, H; 5A, B; Supplementary Fig. 11A–D).

**Fig. 5: Deficiency of cellular senescent program protects against MASLD.**

To examine the role of P21 in MASLD pathogenesis more directly, we fed mice with global genetic deficiency of p21 (p21 knockout mice) and littermate control WT mice with either a chow diet or CDA-HFD for 3 months (Fig. 5D). CDA-HFD feeding induced p21 expression, hepatic senescence (indicated by β-gal staining and DNA damage marker γ-H2AX), steatosis (indicated by H&E and Oil-Red-O staining), liver fibrosis (indicated by Sirius red and αSMA staining), and inflammation (shown by F4/80 staining) in WT mice (Fig. 5E–H). Although p21 knockout did not affect liver fat accumulation, it decreased markers related to hepatic senescence, fibrosis, and inflammation (Fig. 5E–H). The expression of pro-inflammatory and pro-fibrogenic SASP factors including TGF-β, IL-1β, and TNF-α were also decreased (Fig. 5H). Together, the human and mouse data demonstrate that during metabolic stress, up-regulation of p21 enables hepatic accumulation of SHGS+ cells with a SASP that promotes maladaptive repair responses that drive MASLD progression.

One mechanism whereby p21 induction might expand SHGS+ populations in livers with MASLD is by enhancing the survival of SHGS+ cells. This concept is supported by evidence that p21 induction enhances survival of hepatocytes in healthy livers during metabolic stress⁶⁴. Because we recently reported that senescent hepatocytes in MASLD livers upregulate p21 and exhibit ferroptotic stress and removing these cells improved MASLD³⁵, we examined whether p21 status affects accumulation of hepatocytes with ferroptotic stress. We found that in mice with diet-induced MASLD, p21 knockout dramatically decreased the liver burden of hepatocytes expressing ferroptotic stress markers such as MDA and proteins involved in iron import and metabolism (Supplementary Fig. 12E). p21 knockout also decreased markers of apoptosis in mice with MASLD (Supplementary Fig. 12E). Together, these findings are consistent with reports that p21 can function as a survival factor for metabolically-stressed hepatocytes and suggest that p21 inhibitors may function as senolytics in this context. More precise targeting of senescent hepatocytes might be advantageous, however, as our data indicate that MASLD progression tracks with the burden of SHGS+ hepatocytes and these cells comprise a relatively small proportion of p21+ hepatocytes. Targeting SHGS+ hepatocytes per se would also inhibit accumulation of p21-, senescent hepatocytes.

Chemical library screening identified novel senolytics for senescent hepatocytes

Thus far, few senolytic interventions have improved liver disease outcomes by targeting hepatocytes^{17,18,22,24,65}. These failures reflect evidence that the efficacy of senolytics varies in different senescent cell populations and disease contexts²⁰, and highlight the need for novel senolytics designed to target senescent hepatocytes. Therefore, we used hepatocytes to screen chemical libraries to identify agents that kill senescent cells without inducing cytotoxicity in non-senescent cells. A prior publication that had compared senescence characteristics of 16 different hepatoma cell lines demonstrated that Huh7 cells treated with the CDK4/6 inhibitor palbociclib displayed the most robust and irreversible senescent phenotype⁶⁶. As noted earlier, we used this strategy to develop the SHGS because it induces complete cell cycle arrest, strong SASP expression, and near 100% β-Gal positivity (Supplementary Fig. 2). Further, we confirmed the model’s pathobiological relevance by showing that conditioned medium (CM) from these senescent hepatocytes induces pathogenic reprogramming of multiple hepatic cell types (Fig. 4) and the DEG profiles of these senescent Huh7 cells highly overlap with those from MASLD liver (Supplementary Fig. 3). Therefore, while the Huh7 model does not fully replicate the heterogeneity and complexity of in vivo senescent hepatocytes, it provides a robust platform for studying key aspects of senescence in MASLD. The aggregate findings support its use in screening chemical libraries to identify novel senolytics for MASLD treatment.

To perform chemical screenings, Huh7 cells were treated with either vehicle control or palbociclib for 8 days, washed and seeded into drug-stamped 384-well plates (Selleck Bioactive Compound library, 2100 chemicals) for 72 h, and cell viability was assessed (Fig. 6A). Triplicate treatment wells were averaged and normalized to vehicle control wells from the same plate position to account for differences in growth rates. Examining the direct target categories of the top 50 chemical hits that are most cytotoxic towards senescent hepatocytes revealed the top 3 categories target proteasome, HDAC, and PI3K/mTOR/autophagy pathways, all known to be critical for senescent cell survival^67,68,69,70 (Fig. 6B). We considered the most efficacious senolytics to be those hits with high selectivity towards senescent hepatocytes (S) and minimal cytotoxicity towards proliferating hepatocytes (P). A higher P/S ratio in cell viability indicates a relatively higher selectivity to senescent cells, while a higher difference in ΔP-S indicates a relatively lower cytotoxicity to proliferating cells. After plotting them on the X-Y axis, we identified Dp44mT, an iron and copper chelator thiosemicarbazone compound with antitumor activity^71,72,73, as the most selective and cytotoxic hit to senescent Huh7 cells (Fig. 6C). Widely used senolytics like ABT263, ABT737, or Dasatinib showed no selectivity towards senescent hepatocytes (Fig. 6C), consistent with recent studies that they failed to eliminate senescence in either acute liver injury, liver fibrosis or xenograft HCC^18,24,65. DpC, a Dp44mT analog and second-generation thiosemicarbazone with higher oral potency and favorable pharmacokinetics^74,75, was also highly cytotoxic and selective towards senescent Huh7 cells (Fig. 6D, E). Prior studies reported that copper accumulates in senescent cells^76,77. Consistent with this, we observed higher copper levels in senescent Huh7 cells (Fig. 6F, G). Interestingly, the cytotoxicity was largely rescued by the copper chelator Tetrathiomolybdate (TM) but not by the iron chelator Desferoxamine (DFO) (Fig. 6H). Moreover, cytotoxicity was exacerbated by exogenous copper but not by supplementing iron in the culture medium (Fig. 6H), indicating that both Dp44mT and DpC display their senolytic effect by binding to copper, but not iron. RNA-seq analyses further revealed that Dp44mT increased the expression of genes linked to cellular response to copper ion and apoptosis, suggesting that DpC may bind to copper forming redox-active complexes that selectively induce apoptosis in senescent cells (Fig. 6I). This mechanistic insight highlights DpC’s specificity and therapeutic promise for senescence-driven liver diseases.

**Fig. 6: Chemical library screening identified novel senolytics for senescent hepatocytes.**

Novel senolytic DpC improves MASLD

To evaluate the therapeutic potential of our newly identified senolytic, DpC, for MASLD, mice were fed a CDA-HFD for 10 weeks to induce MASH, and DpC or vehicle was given by gavage two times per week for the last 4 weeks (Fig. 7A). DpC significantly reduced the senescent burden in MASH livers, as indicated by markedly decreased β-Gal and p21 staining (Fig. 7B, C). While DpC treatment did not affect the body weight of the mice, it did reduce the liver/body weight ratio (Fig. 7D; Supplementary Fig. 13A). ALT and AST, serum markers of liver injury, were also reduced, although the reduction in AST did not reach statistical significance (p = 0.056) (Fig. 7D). Although serum triglyceride levels remained unchanged, serum cholesterol levels trended to be lower (p = 0.052) (Supplementary Fig. 13B), suggesting a beneficial effect of DpC on both MASH and its associated cardiovascular complications. Consistent with the changes in liver/body weight ratio, H&E staining also indicated reduced steatosis in DpC-treated mice, and this is further corroborated by Oil Red O staining (Fig. 7B, E). Importantly, liver fibrosis and inflammation, as indicated by Sirius Red and F4/80 staining, respectively, were also significantly reduced (Fig. 7B, E). Western blot analysis of liver tissues supported these histological findings, showing decreased levels of fibrosis markers such as collagen 1 (Col1), TGF-β, and MMP2 (Fig. 7F). Senescence markers and SASP factors, including p21, IL-1β and HMGB1 were also reduced (Fig. 7F). Additionally, as noted in p21 knockout mice fed with CDA-HFD, there was an increase in ferroptosis inhibitor GPX4, a dramatic decrease in ferroptotic stress markers MDA and ACSL4, as well as the apoptosis marker cleaved caspase 3 (Supplementary Fig. 13C). GSEA also revealed that DpC decreased the expression of genes and pathways involved in extracellular matrix formation, inflammation, senescence and apoptosis, but enriched the genes and pathways associated with hepatocyte-specific metabolic functions (e.g., bile acid metabolism, xenobiotic metabolism, fatty acid metabolism) (Fig. 7G; Supplementary Fig. 13D, E). Notably, DpC also lowered the enrichment score of SHGS, indicating a reduction in the accumulation of SHGS+ senescent hepatocytes (Fig. 7H). Together, these data indicate that DpC, a novel senolytic for senescent hepatocytes, significantly reduces markers of hepatic senescence, fibrosis, inflammation, cell stress, and death in a mouse model of MASLD. These findings suggest that DpC holds promise as a therapeutic agent for treating MASLD, providing a basis for future investigation into its clinical potential.

SHGS correlates with multi-organ dysfunction

MASLD is increasingly recognized as a systemic metabolic disorder, with failure of other organs such as the pancreas, heart, adipose tissue, and kidneys being major contributors to patient mortality^78,79. While SHGS was developed to identify senescent hepatocytes, we hypothesized that the pathways it captures - particularly those linked to SASPs, fibrosis, inflammation, and integrated stress responses—might also reflect senescence-driven dysfunction in these other organs commonly affected by MASLD. This hypothesis provides an opportunity to investigate whether senescence-associated mechanisms identified by SHGS are conserved across different tissues under metabolic stress. The SHGS captures SASPs that cause maladaptive repair of metabolically-stressed livers, including signaling that orchestrates integrated stress responses and that promote tissue fibrosis and inflammation (Fig. 1; Supplementary Fig. 4). Hence, we hypothesized that SHGS might also be enriched in the transcriptomes of other organs that become progressively dysfunctional in this context. Evidence that liver enrichment with SHGS increases with BMI (Supplementary Fig. 6A) suggests that obesity-related metabolic stress accelerates cellular senescence in the liver. Given that obesity is also a significant risk factor for hyperinsulinemia and type 2 diabetes⁸⁰, we aimed to explore the connection between hepatocyte senescence and dysfunction in other organs more directly. To do so, we applied SHGS to deconvolute RNA datasets from four additional organs commonly affected in MASLD patients: adipose, pancreas, kidney, and heart. In visceral adipose tissue (VAT), reanalysis of bulk RNA-seq data from lean individuals and individuals with obesity revealed significant SHGS enrichment in individuals with obesity compared to their lean counterparts (Fig. 8A). For subcutaneous abdominal adipose tissue, we reanalyzed two bulk RNA-seq datasets: GSE244118, which included 53 adults categorized as metabolically healthy individuals with obesity (MHO), metabolically unhealthy individuals with obesity (MUO, defined by prediabetes and hepatic steatosis), and metabolically healthy lean (MHL); and GSE156906, which consisted of 66 individuals similarly categorized as MHL, MHO, and MUO. In both datasets, we found that SHGS was more enriched in MHO individuals compared to MHL, with the enrichment score even more pronounced in patients with MUO (Fig. 8B, C). For the pancreas, we reanalyzed bulk RNA-seq data from human pancreatic islets of 89 donors with varying levels of insulin sensitivity and hyperglycemia⁸¹ (Fig. 8D). The SHGS score strongly correlated with glycated hemoglobin (HbA1c) and was significantly enriched in donors with T2D and impaired glucose tolerance (IGT) (Fig. 8D). For the heart analysis, we examined publicly available RNA-seq data from 45 healthy and 100 failing hearts⁸². Both heart failure with preserved ejection fraction (HFpEF) and reduced ejection fraction (HFrEF), conditions strongly associated with obesity-related systemic metabolic dysfunction and MASLD, were enriched with SHGS (Fig. 8E). In the kidney analysis, we used microarray-based transcriptome expression data from 79 normal and fibrotic human kidneys, segregated into four groups by Kang and colleagues⁸³. Compared with healthy controls, the three disease groups—diabetic chronic kidney disease (D-CKD), diabetes, and hypertension—each showed significant enrichment with SHGS (Fig. 8F). Merging data from the diabetes and hypertension groups with healthy controls still showed significant SHGS enrichment in the D-CKD group relative to this combined control group (Fig. 8F). While these studies do not clarify whether hepatocyte senescence induces metabolic stress that triggers dysfunction in other organs, or vice versa, the analyses across all four organs demonstrate that the SHGS generally differentiates diseased from healthy organs and indicates that cellular senescence is a conserved mechanism contributing to MASLD and its associated multi-organ dysfunction. Importantly, DpC treatment reduced the expression of genes and pathways linked to cardiomyopathy, diabetes, and cancer in the mouse model of MASLD (Fig. 8G), suggesting that this newly discovered senolytic may not only improve MASLD but also mitigate associated dysfunction in other organs.

**Fig. 8: Hepatocyte senescence correlate with multi-organ dysfunction.**

Discussion

Our study establishes the significant role of cellular senescence, especially hepatocyte senescence, in MASLD pathogenesis, as well as its broader systemic impact. Through RNA-seq analysis of liver biopsies, we identified a general enrichment of senescence-associated genes in MASLD patients and demonstrated that senescent cell accumulation consistently correlates with MASLD severity. Defining the roles of senescent cells in MASLD pathogenesis requires more precise phenotyping, however, because senescence is one eventual outcome of diverse adaptive processes that are initiated by various cellular stressors and thus, senescing cell populations are inherently heterogeneous. By inducing senescence in hepatocytes in vitro and in vivo, we developed the SHGS, a robust gene panel that captures conserved features of cells in late stages of senescence. We leveraged single cell analytical approaches to determine the origins of SHGS+ hepatocytes and found that most derive from hepatocytes that express p21, a cyclin dependent kinase inhibitor that blocks transition from G1 into S phase of the cell cycle. However, p21 has multiple functions and we noted that a small percentage of hepatocytes also express p21 in healthy liver. Further, the vast majority of p21+ hepatocytes in diseased livers do not score as positive for the SHGS. ln addition we discovered that SHGS+ hepatocytes do not always express p21. Therefore, we used multiple complementary strategies to determine whether the SHGS identifies senescent cells that drive key processes in MASLD pathogenesis. We found that liver enrichment with the SHGS tracks with severity of liver damage and fibrosis in several cross-sectional human studies, and improves with liver histology after various therapeutic interventions. Parallel analysis of data sets from different mouse models of MASLD confirmed that the SHGS increases with MASLD progression and decreases with interventions that lead to MASLD regression. Further, we showed that changes in SHGS enrichment map to hepatocyte populations in both human and mouse single nuclei RNA seq data sets. Together, these findings demonstrate that the SHGS not only identifies senescent cells but also provides a valuable tool for studying MASLD pathogenesis. To assess the functional implications of the robust correlations between the SHGS and maladaptive liver repair, we treated cell cultures with conditioned medium from SHGS+ cells and demonstrated that SHGS+ hepatocytes release factors that induce secondary senescence in hepatocytes, promote fibrogenesis in HSCs, activate macrophages, and cause capillarization in liver sinusoidal endothelial cells. These results illustrate the broad impact of hepatocyte SASPs on the liver microenvironment and MASLD progression. To pinpoint specific paracrine interactions for future development of diagnostic biomarkers and therapeutic interventions, we performed further analysis of our single nuclei RNA seq data sets using programs that map cell-cell interactions. This revealed an array of signaling factors produced by SGHS+ hepatocytes and identified various types of signal-receiving cells. Interestingly, we learned that SHGS+ hepatocytes regulate themselves and interact with all of the other liver cell types examined via GDF15-TGFBR2 signaling. Given that GDF15 is key component of the integrated response to mitochondrial stress, we propose that this cross-talk permits senescent hepatocytes to telegraph their distress to other cells. Although GDF15 can have both context-dependent detrimental and protective functions^52,84, it is an established biomarker of biological aging as its serum levels consistently increase in aging-related diseases and predict mortality^52,85. Our new evidence that serum levels of GDF15 increase and decrease in parallel with changes in hepatic SHGS enrichment supports the importance of SHGS+ hepatocytes in determining MASLD outcomes and suggests a role for unresolved mitochondrial stress in the process.

Delineating the roles of senescent hepatocytes in MASLD-related pathology requires tools to manipulate these cells directly. Through chemical screening of 2,100 bioactive compounds, we identified DpC as a novel senolytic that selectively targets SHGS+ senescent hepatocytes. DpC significantly reduced markers of senescence, fibrosis, inflammation, and cell death in a mouse model of MASLD. Mice with global deletion of p21 exhibited similar protection from MASLD. Given evidence that the relatively small subpopulation of SHGS+ hepatocytes in MASLD livers derives from a larger p21+/SHGS- hepatocyte population, these findings indicate that MASLD pathogenesis is driven by senescent hepatocytes that are SHGS+ and identify new insights into mechanisms that control hepatic accumulation of these cells. Our studies show that DpC’s senolytic actions are copper-dependent but target hepatocytes that are evading ferroptosis, a type of regulated cell death that results from iron-mediated peroxidation of membrane lipids³⁵. Human genetic evidence supports the importance of both iron and copper in MASLD pathogenesis/progression: certain polymorphisms of the iron homeostatic HFE gene increase the risk for cirrhosis and liver cancer in MASLD patients, while polymorphisms in genes that regulate copper transport can cause steatohepatitis^86,87. However, to the best of our knowledge, the possibility that these two metals mediate MASLD progression by regulating conserved senescence mechanisms has not been explored. Hence, our discovery provides a fresh perspective on the role of divalent metal cations in hepatocyte senescence and MASLD pathogenesis, opening opportunities for future research. Such studies are likely to be fruitful because although copper uniquely controls cuproptosis, both metals have been shown to regulate various other cell death programs, including apoptosis, autophagy, pyroptosis, necroptosis, and ferroptosis⁸⁸. Further, there appears to be cross-talk between copper-sensitive and iron-sensitive homeostatic mechanisms. For example, copper regulates the expression of proteins that control iron homeostasis and iron is a critical cofactor for enzymes that repair copper-mediated DNA damage⁸⁸. Importantly, both copper and iron regulate biology that is dysregulated in MASLD, including redox reactions, electron transport, oxygen bioavailability, energy metabolism, and protein function. Levels of hepatocyte-generated proteins that regulate iron and copper bioavailability locally and systemically (e.g., hepcidin, ferroportin, ferritin, and ceruloplasmin) also correlate with expression of GDF15⁸⁹.

As noted earlier, we found that serum GDF15 levels reflect the liver burden of SHGS+ hepatocytes in MASLD and others have reported that high serum levels of GDF15 predict mortality from various aging-related diseases. We discovered that the SHGS broadly captures this deleterious senescent state by applying the SHGS to transcriptomic datasets from other organs that commonly become dysfunctional and damaged in MASLD patients. Diseased adipose tissue, pancreatic islets, kidneys, and hearts demonstrate SHGS enrichment and DpC treatment decreased genes/pathways linked to cardiomyopathy, diabetes, and cancer, suggesting that conserved mechanisms drive chronic organ deterioration. Hence, phenotyping SHGS+ cells may reveal processes that control the evolution of multi-organ dysfunction, and targeting senescence could have therapeutic implications beyond the liver. Our data show that DpC consistently suppressed genes and pathways linked to inflammation, fibrosis, and multiple-organ dysfunction, supporting the concept that senolytics like DpC may be particularly promising as MASLD treatments since they benefit not only the liver, but also other organs that become dysfunctional in these patients. This approach is particularly appealing given that senolytics can be administered intermittently in a ‘hit-and-run’ manner^8,90, thereby minimizing potential toxic effects resulting from continuous dosing and increasing adherence among MASLD patients, who are often afflicted with multiple morbidities and thus require multiple medications.

Our findings suggest that SHGS, initially developed to identify senescent hepatocytes, may provide broader insights into systemic metabolic stress and its effects on multiple organs. MASLD is increasingly recognized as part of a systemic metabolic disorder contributing to the gradual degeneration of organs, including the pancreas, heart, adipose tissue, and kidneys^78,79. Building on prior studies showing aging-sensitive mechanisms, such as those captured by AHGS³⁵, are enriched in failing metabolic syndrome target tissues, we leveraged SHGS to identify correlations between the burden of SHGS+ cells and dysfunction in non-hepatic tissues. Our analysis revealed significant positive correlations between SHGS activity and damage in metabolic syndrome target tissues, including adipose depots. Additionally, our novel senolytic agents not only reduced SHGS+ hepatocytes and improved MASLD but also suppressed the expression of genes implicated in extra-hepatic dysfunction, such as cardiomyopathy and diabetes. These findings suggest that senescence-associated pathways captured by SHGS may represent conserved mechanisms driving tissue damage during metabolic stress, offering a unified therapeutic target for MASLD and its comorbidities. While it remains unclear whether SHGS+ hepatocytes directly induce senescence in other tissues, recent studies have proposed a role for hepatocyte-derived signals in propagating systemic dysfunction^91,92. Our findings provide a foundation for future research to elucidate the mechanisms by which SHGS+ hepatocytes contribute to multi-organ senescence and dysfunction.

While our study primarily focuses on hepatocytes, we acknowledge that senescence is not limited to these cells in the liver. Obesity-related senescence has been observed in HSCs, cholangiocytes, LSECs, and macrophages, each of which has distinct roles in liver pathophysiology⁹³. For example, senescent HSCs may limit fibrosis^94,95, while senescence in LSECs and macrophages modulates the liver microenvironment and inflammation⁹⁶. These observations underscore the complexity of hepatic senescence and the need for cell-type-specific tools. Our findings suggest that SHGS successfully distinguishes senescent hepatocytes and supports the hypothesis that these cells are key drivers of maladaptive repair in MASLD livers. Experiments using senescent hepatocyte-conditioned medium revealed that these cells orchestrate maladaptive repair processes, including HSC activation, macrophage recruitment, and LSEC capillarization. This highlights their central role in MASLD pathogenesis, although further research is needed to clarify the interplay between senescence in different hepatic cell types and their collective contribution to liver dysfunction.

In conclusion, our study underscores the pivotal role of hepatocyte senescence in MASLD and its complications. The development of SHGS provides a valuable tool for identifying senescent cells and tracking disease progression. The discovery of DpC as a selective senolytic offers a promising therapeutic approach for MASLD and its associated multi-organ dysfunction. The potential of intermittent senolytic administration to minimize toxic effects and enhance patient adherence further highlights the clinical relevance of our findings. These results pave the way for future research and clinical applications, emphasizing the importance of targeting cellular senescence to improve outcomes in MASLD and related metabolic disorders.

Methods

Animal studies

Mice

Eight-week-old C57BL/6J male mice were obtained from Jackson Laboratories (Bar Harbor, ME). Heterozygous p21 (p21+/-) mice were generously provided by Dr. David Kirsch (Department of Radiation & Oncology, Duke University). These p21+/- mice were used for breeding to generate p21 knockout (p21-/-) and littermate control wild-type (WT, p21+/+) mice. All mice were housed in a barrier facility with a 12-h light/dark cycle, controlled temperature (20–25 °C), humidity (30–70%), and free access to water and standard rodent diets. All experimental protocols were approved by the Duke University Institutional Animal Care and Use Committee (IACUC) and adhered to the guidelines of the National Institutes of Health for humane animal care.

CDA-HFD studies

The choline-deficient, L-amino acid-defined, high-fat diet (CDA-HFD) induces robust histologic and biochemical indicators of fibrosing MASH in a relatively short timeframe and mimics the effects of dietary choline deficiency known to promote MASLD in humans. To investigate the impact of cellular senescence on MASLD, 8–12 weeks-old p21 knockout (p21-/-) and littermate control WT (p21+/+) male mice were fed either a chow diet or CDA-HFD (Research Diets, Inc, New Brunswick, NJ) for 3 months (chow-WT, n = 3; chow-p21KO, n = 3; CDAHFD-WT, n = 11; CDAHFD-p21KO, n = 8). To test the effects of DpC on MASLD, 10-week-old C57BL/6J male mice were fed CDA-HFD for 10 weeks, and DpC (5 mg/kg) or vehicle (10% cavitron™ w7 hp7 pharma cyclodextrin) was given by gavage twice a week for the last 4 weeks (CDAHFD-Veh: n = 7; CDAHFD-DpC: n = 6). Mice were not fasted before sacrifice, which was performed between 1 and 4 pm. At the end of the diet administration, mice were euthanized using isoflurane inhalation in a closed chamber until deep anesthesia was achieved, as indicated by the loss of righting reflex and absence of response to toe pinch. Blood was collected, and liver tissues were either fixed in optimal cutting temperature (OCT) compound or phosphate-buffered formalin for histological analysis or flash-frozen in liquid nitrogen and stored at −80 °C.

Cell culture

Huh7 cells (a gift from Dr. Charles M. Rice, Rockefeller University) were cultured and maintained in DMEM medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). To induce senescence, the cells were treated with 1 µM palbociclib for 8 days. Afterward, the cells were treated with various concentrations (1 nM to 1 µM) of Dp44mT, DpC, or vehicle (0.1% DMSO) for up to 72 h, and cell viability was assessed using CCK8 assays. To confirm the senolytic effects of DpC or Dp44mT, both vehicle and palbociclib-treated Huh7 cells were harvested, stained with FITC Annexin V Apoptosis Detection Kit I (BD Pharmingen™ #556547), and analyzed by flow cytometry according to the manufacturer’s instructions. Briefly, cells were washed twice with cold PBS, resuspended in 1X Binding Buffer at 1 × 10⁶ cells/mL, and 100 µL of the suspension (1 × 10⁵ cells) was stained with 5 µL of FITC Annexin V and 5 µL of propidium iodide (PI). After gentle vortexing, the samples were incubated in the dark at room temperature for 15 min, followed by the addition of 400 µL of 1X Binding Buffer. Flow cytometry was performed within 1 h, with fluorescence emission data collected for FITC Annexin V (FITC channel) and PI (PE-Cy5 channel). Compensation settings were optimized using single-stain controls to differentiate FITC and PI signals accurately. Viable cells (FITC Annexin V-negative, PI-negative), early apoptotic cells (FITC Annexin V-positive, PI-negative), and late apoptotic or necrotic cells (FITC Annexin V-positive, PI-positive) were quantified to assess the selective cytotoxicity of DpC and Dp44mT in senescent versus non-senescent hepatocytes. To determine the role of copper and iron in DpC or Dp44mT-induced cytotoxicity, cells were treated with the iron chelator deferoxamine (DFO, 5 µM), the copper chelator tetrathiomolybdate (TM, 1 µM), or supplemented with exogenous FeSO4 or CuSO4 (0.01 µM). Cell viability was then determined using CCK8 assays after the treatments. Intracellular copper levels were measured using Coppersensor 1 staining (MedchemExpress #HY-141511) and Copper Microplate Assay Kit (MyBioSource #MBS8292800) following the manufacturer’s instructions. To examine the paracrine effects of senescent Huh7 cells, the cells were treated with 1 µM palbociclib or its vehicle for 8 days. After thorough washing with PBS, the cells were incubated with fresh medium (without palbociclib) for 24 h. Conditioned medium (CM) was then harvested and used to culture recipient cells, including non-senescent hepatocytes (Huh7 cell line), hepatic stellate cells (LX2 cell line), macrophages (Raw264.7 cell line), and primary human liver sinusoidal endothelial cells (LSECs), for 72 h.

Chemical library screening

To identify novel senolytics, we performed a high-throughput chemical screening using Selleck Bioactive Compound library (2100 compounds, 1 µM). Huh7 cells were first cultured and induced into a senescent state by treating with 1 µM palbociclib for 8 days. Non-senescent (treated with vehicle 0.1% DMSO) and senescent (palbociclib-treated) Huh7 cells were then seeded into 384-well plates pre-stamped with the chemical library. The cells were incubated with the compounds for 72 h, and cell viability was assessed using the Cell Titer Glo assay (Promega, Madison, WI) according to the manufacturer’s instructions. This assay measures ATP levels as an indicator of metabolically active cells. Triplicate treatment wells were averaged and normalized to vehicle control wells from the same plate position to account for differences in growth rates. The primary screening identified compounds that preferentially reduced the viability of senescent Huh7 cells compared to proliferating cells. Hits from the primary screen were further validated by dose-response studies and additional assays to confirm their senolytic activity.

RNA-seq library construction and analyses

Total mRNA was extracted using the Trizol reagent, converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, Waltham, MA), and sequenced on an Illumina NovaSeq sequencer with 150 bp paired-end reads by Novogene. Raw reads were processed by trimming adapters, and low-quality reads, aligning cleaned reads to the reference genome using STAR v2.7.5, and quantifying gene counts using featureCounts. Differential expression genes (DEGs) were identified using DESeq2, and pathway gene enrichment was analyzed using Gene Set Enrichment Analysis (GSEA) software (version 4.2.2). RNA-seq raw files and gene counts have been uploaded to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database.

Development and application of Senescence Hepatocyte Gene Signature (SHGS)

SHGS development

To develop a robust gene panel associated with hepatocyte senescence, hepatocyte senescence was induced in vitro and in vivo. In the in vitro model, Huh7 cells were treated with the CDK4/6 inhibitor palbociclib (1uM) for 8 days, followed by RNA isolation and RNA-seq analysis. In the in vivo model, 12-week-old mice were injected with NRASG12V-IRES-GFP plasmid and CMV-SB13 via hydrodynamic tail vein injection. Six days later, GFP-positive (NRASG12V overexpressing hepatocytes) and GFP-negative (non-transduced hepatocytes) cells were isolated via fluorescence-activated cell sorting and subjected to RNA-seq analysis (GSE145642). Differentially expressed genes (DEGs) were quantified using DESeq2⁹⁷ and those that were up-regulated in both models with different cutoffs of adjusted p-values (0.05 or 0.01) and log₂FoldChange (from 0 to 5) were identified. By overlapping the upregulated differentially expressed genes (DEGs) from both datasets (Supplementary Table 1), we found that DEGs in the in vitro model with adjusted P < 0.01 and log₂ FC > 1 and in the in vivo model with adjusted P < 0.05 and log2 FC > 0 had the best balance between gene set size and high Jaccard similarity index (Fig. 2E). This overlapping resulted in 100 DEGs, which we defined as the Senescent Hepatocyte Gene Signature (SHGS) (Supplementary Table 2).

The Jaccard index was calculated using following equation:

$$J=\frac{{n}_{o}}{{n}_{1}+{n}_{2}-{n}_{o}}$$

Where ${n}_{o}$ represents number of overlapping DEGs, ${n}_{1}$ represents number of DEGs in dataset 1, and ${n}_{2}$ represents number of DEGs in dataset 2. The $J$ of 1 suggests the two datasets completely overlap, and $J$ of 0 suggests no overlapping.

Deconvolution of independent bulk RNA-seq data sets using SHGS

Enrichment scores of AHGS were calculated using the Gene Set Variation Analysis (GSVA)⁹⁸ based on normalized gene expressions per cohort. The enrichment score of this gene signature was firstly evaluated in MASLD cohorts below:

1.
Duke MASLD cohort⁹⁹ (GSE213623; n = 368). This cohort included individuals with obesity but no MASLD (n = 69) and those with obesity and biopsy-proven MASLD (n = 299). Liver, blood, and clinical data were collected as part of the Duke University Health System (DUHS) MASLD Clinical Database and Biorepository. SHGS enrichment was calculated from bulk RNA-seq data and correlated with clinical markers such as serum albumin, AST, and Fib4 score.
2.
German MASLD cohort¹⁰⁰ (GSE33814; n = 44). Transcriptomic analysis was performed on liver tissue samples (normal, n = 13; simple steatosis, n = 18; steatohepatitis, n = 12) using Affymetrix HG-U133 Plus 2.0 array. Raw “.CEL” files were downloaded using the GEOquery package in R and normalized using the gcrma method for SHGS enrichment analysis..
3.
Japanese MASLD cohort¹⁰¹ (GSE167523; n = 98). This retrospective cohort comprised 51 patients with simple steatosis and 47 with non-alcoholic steatohepatitis (NASH). The gene count matrix from bulk RNA-seq was downloaded from NCBI GEO and normalized using DESeq2 before SHGS enrichment analysis.
4.
European MASLD cohort³⁶ (GSE135251; n = 216). Bulk RNA-seq data were generated from liver biopsies in 206 MASLD patients and 10 healthy controls from France, Germany, Italy, and the UK. Fibrosis severity in MASLD patients ranged from F0 to F4. DESeq2 was used to normalize the raw counts for SHGS deconvolution analysis.
5.
MASLD-HCC primary risk cohort (GSE193066; n = 106)³⁷. This dataset included 106 noncirrhotic, HCC-naïve NAFLD patients who underwent diagnostic liver biopsy at Hiroshima University between May 2003 and February 2015. Patients were followed regularly using ultrasound every 6 months for a median of 8.9 years (IQR, 5.1 to 11.9), during which six patients developed HCC. Time to HCC development was defined as the interval between biopsy and HCC diagnosis or last follow-up. SHGS enrichment was correlated with HCC risk, which was determined using an etiology-agnostic prognostic liver signature (PLS) developed by Dr. Yujin Hoshida’s group.
6.
MASLD-HCC recurrence risk cohort (GSE193080; n = 59)³⁷. This dataset included 59 NAFLD patients who had previously undergone curative surgical resection for early-stage HCC at Toranomon Hospital or Kumamoto University between January 2003 and November 2011. These patients were followed postoperatively for a median of 1.8 years (IQR, 0.6 to 3.5), during which 32 patients experienced HCC recurrence. Time to HCC recurrence was defined similarly to the primary risk cohort. SHGS enrichment was evaluated for its correlation with HCC recurrence risk.
7.
Bariatric surgery follow-up cohort (GSE83452; n = 25)⁴³. Microarray data were obtained for 25 patients at baseline and at a 1-year follow-up after bariatric surgery. Transcriptomic analysis was performed on liver tissue samples using the Affymetrix HG-U133 Plus 2.0 array. Raw “CEL” files were downloaded using the GEOquery package in R and normalized using the gcrma method for SHGS enrichment analysis.
8.
Statin treatment cohort (GSE130991; n = 157)¹⁰². Microarray data were obtained from liver tissue samples of 157 statin-treated and 157 non-statin-treated patients of the matched cohort. Data analysis was conducted using ArrayStudio software (version 10.0.1.75, Omicsoft). Raw data from Affymetrix microarrays were normalized using the robust multi-array average (RMA) method and log2-transformed.

We also tested the SHGS in 4 other dysfunctional human organ systems:

1.
Visceral adipose tissue cohort (GSE235696; n = 8). SHGS enrichment was calculated using bulk RNA-seq data from visceral adipose tissue (VAT) of lean individuals (n = 4) and individuals with obesity (n = 4). DESeq2 was used to normalize the data, and GSVA was applied for SHGS deconvolution analysis.
2.
Subcutaneous adipose tissue cohort 1 (GSE244118; n = 53). Bulk RNA-seq was performed on subcutaneous abdominal adipose tissue from individuals categorized as metabolically healthy lean (MHL; n = 15), metabolically healthy individuals with obesity (MHO, defined as normal fasting glucose, glucose tolerance, HbA1c, intrahepatic triglyceride, serum triglycerides, and whole-body insulin sensitivity; n = 19), and metabolically unhealthy individuals with obesity (MUO, defined as prediabetes and hepatic steatosis; n = 19). DESeq2 was used for data normalization, and GSVA was applied for SHGS enrichment analysis.
3.
Subcutaneous adipose tissue cohort 2 (GSE156906; n = 56). SHGS enrichment was analyzed in subcutaneous abdominal adipose tissue from individuals in three categories: MHL (n = 14), MHO (n = 25), and MUO (n = 17). DESeq2 was used for normalization, and GSVA was applied for SHGS deconvolution analysis.
4.
Heart failure cohort⁸² (Zenodo; n = 145). Gene count data were obtained from the Zenodo database https://zenodo.org/record/4287217. Bulk RNA-seq was performed on left or right ventricular septal endomyocardial biopsies from patients with heart failure with preserved ejection fraction (HFpEF; n = 41), heart failure with reduced ejection fraction (HFrEF; n = 59), and healthy controls (Normal; n = 45). Raw counts were normalized using DESeq2, and SHGS enrichment was calculated using GSVA for deconvolution analysis.
5.
Kidney fibrosis cohort⁸³ (E-MTAB-2502; n = 79). Microarray data were obtained from the EMBL-EBI website https://www.ebi.ac.uk/biostudies/arrayexpress/studies/E-MTAB-2502. The dataset includes gene expression data from human kidney tubule samples classified into four groups: Control (n = 20), diabetes-chronic kidney disease (D-CKD; n = 19), diabetes (n = 21), and hypertension (n = 19). Gene probe annotation was based on the Affymetrix HG-U133A_2 array. For genes with multiple probes, the mean expression of all probes was used for deconvolution analysis. SHGS enrichment was performed using GSVA.
6.
Pancreatic Islets cohort⁸¹ (GSE50244; n = 89). Bulk RNA-seq was conducted on human pancreatic islets from 89 donors. Following Fadista et al., the cohort was divided into three groups based on HbA1c levels: Normal (HbA1c < 6%; n = 51), impaired glucose tolerance (IGT; 6% ≤HbA1c < 6.5%; n = 15), and type 2 diabetes (T2D; HbA1c ≥ 6.5%; n = 11). DESeq2 was used to normalize the raw counts, and GSVA was applied for SHGS deconvolution analysis.

Benchmarking of senescence gene signatures

To assess the predictive power of SHGS in MASLD cohorts, we compiled a total of 14 senescence or aging signatures, including published gene sets, curated gene panels from the literature, and senescence/aging pathways from established databases. The evaluated senescence signatures included: AHGS³⁵, SenMayo⁴⁷, CellAge¹⁰³, GeneAge3¹⁰³, SenNet¹⁰⁴, SeneQuest_HsLiver¹⁰⁵, SASP¹⁰⁶, Reactome_SASP (MsigDB), Freund2010¹⁰⁷, CSGene_Zhao¹⁰⁸, Casella2019¹⁰⁹, Reactome_CellSenescence (MsigDB), and GO_CellSenescence (MsigDB). These signatures ranged in size from 41 to 866 genes. We observed varied overlap among the genes in these signatures, with shared gene counts ranging from 0 to 248 (Supplementary Fig. 7A). To quantify the similarity between gene sets, we calculated the Jaccard similarity index. Most signature pairs exhibited a low similarity score (<0.1), while only four signature pairs showed intermediate similarity scores (0.26–0.52) (Fig. 3B). Given that these signatures were derived from different studies, tissues, and experimental contexts, evaluating their performance in MASLD cohorts was essential. We next conducted receiver operating characteristic (ROC) analyses to benchmark the performance of these 14 senescence signatures in four MASLD cohorts representing diverse ancestries. Area under the curve (AUC) values were calculated for comparisons between MASLD occurrence and healthy controls, as well as MASH versus MASL, and advanced fibrosis stages (F3/F4) versus early fibrosis stages (F0/F1). To determine the best-performing senescence signature, we ranked the 14 signatures based on their AUC values for each comparison. The signature with the highest average rank across all benchmarking tests was deemed the most robust and relevant for predicting MASLD-related outcomes.

Human and mouse single nuclei RNA sequencing datasets

Human single nuclei data were obtained from GSE202379 and our in-house Duke Human Dataset. Mouse snRNA-seq was obtained from GSE262939 and processed as described in our previous publication³⁵. UMAP representations of the datasets are shown in Supplementary Fig. 3: GSE202379 is presented in panel A, the Human Dataset in panel B, and GSE262939 in panel C. Basic cell annotation labels were added from the original publication for GSE202379¹¹⁰ or using manual cell marker annotation for our Duke Dataset. GSE202379 consists of 47 liver biopsies from patients across the spectrum of metabolic dysfunction associated steatotic liver disease. After standard QC processing as previously described^110,111, this dataset included 99,809 cells consisted of B-cells, cholangiocytes, hepatocytes, endothelial cells, lymphocytes, macrophages, neutrophils, and stellate cells (additional information in Figure Legends). Our Duke Human snRNA-seq data consisted of 10 liver biopsies from lean and healthy controls, as well as across fibrosis staging. Written informed consent was obtained from all participants and this study was IRB approved under protocol no. Pro00005368. After standard QC processing, this dataset included 68,615 cells consisting of B-cells, cholangiocytes, endothelial cells, hepatocytes, HSCs, macrophages, and T-cells. We then harmonized the cell labels and combined the two datasets to yield a total set of 168,424 consisting of 57 samples. The initial processing and annotation were carried out in R (v 4.4.0) using Seurat (v 5.1.0)¹¹².

Integration of hepatocyte snRNA-seq and trajectory analysis

As we were interested in understanding the senescence trajectory in hepatocytes, from the combined dataset we extracted the hepatocyte population which consisted of 117,485 hepatocytes (n = 69426 from GSE202379 and n = 48059 from our Duke Dataset). We then re-processed the data following standard Seurat processing¹¹² with integration using Harmony¹¹³. Our constructed SHGS gene signature was applied separately to the human and mouse hepatocytes using AUCell¹¹⁴ which defined SHGS+ hepatocytes based on AUC scores across expression ranking of SHGS genes per cell with an adjusted cutoff.

Gene Set Enrichment on human and mouse snRNAseq data

Seurat’s FindMarkers function was used to compare SHGS+ vs SHGS- and P21+ vs P21- hepatocytes in both the human and mouse snRNAseq datasets. The gene lists were sorted in descending order based on the average log₂ fold change. The Reactome, Hallmark, and KEGG pathways were acquired from the Molecular Signatures Database (R package msigdbr). ClusterProfiler⁶³ was used to perform gene set enrichment analysis (GSEA) on specific pathways including senescence and SASP pathways.

Analysis of serum proteins

The O-link target panels, cardiometabolic, inflammation, neurology, and oncology, were used to assess serum proteins from 80 patients (n = 40 fibrosis stage 0/2 and n = 40 fibrosis stage 3/4). Normalized protein expression (NPX) was used to determine protein abundance (a dataset published before at https://www.jci.org/articles/view/180310#sd). OlinkAnalyze was used to determine differentially expressed proteins between the two groups with a Benjamini–Hochberg adjusted p-value cutoff of 0.05. We used ggplot2 and customized OlinkAnalyze functions to visualize the proteomics data with a volcano plot representation. Written informed consent was obtained from all participants and this study was IRB approved under protocol no. Pro00005368. Additionally, a comparison of serum proteins between MASH (n = 37) and MASLD (n = 20) patients was performed using data derived from GSE251855. In this dataset, protein levels were measured with the SOMAscan Assay Kit for human serum 1.3 k, which quantifies the expression of 1,305 proteins in accordance with the manufacturer’s standard protocol (SomaLogic; Boulder, CO). Data quality control, normalization, and calibration were carried out as per the manufacturer’s instructions.

Quantitative real time PCR (qRT-PCR)

Total RNA was isolated from whole liver tissue or cultured cells using Trizol reagent (ThermoFisher Scientific). After determining RNA concentration and purity, complementary DNA (cDNA) was synthesized using Superscript II Reverse Transcriptase (Life Technologies, Carlsbad, CA) according to the manufacturer’s instructions. mRNA was quantified by qRT-PCR using SYBR Green Super-mix (Life Technologies) and primers listed in Supplementary Table 3 on a QuantStudio™ 6 Real-Time PCR System (Thermo Fisher). Results were normalized to the housekeeping gene S9 based on the threshold cycle (Ct), and relative fold change was determined using using the 2-ΔΔCt method.

Immunoblot

Proteins were extracted from isolated cell pellets or whole liver tissue using RIPA buffer with protease inhibitors (Sigma-Aldrich). Equal quantities of protein were loaded onto SDS-PAGE gels (4%-20% Criterion gels; BioRad, Hercules, CA) and subjected to electrophoresis. Subsequently, proteins were transferred onto PVDF membranes and probed with primary antibodies listed in Supplementary Table 4. All primary antibodies used were validated for compatibility with mouse and human species and intended applications either by the manufacturers or through previous publications, as documented in Citeab (https://www.citeab.com). Additionally, all antibodies were individually tested and titrated on relevant positive or negative biological controls. Blots were incubated with HRP-conjugated secondary antibodies and visualized using Image Studio™ Lite Ver 5.2 (LI-COR Biosciences). The uncropped and unprocessed scans were provided as Source Data file in the Supplementary Information.

Histopathological analysis, immunohistochemistry (IHC) and immunocytochemistry (ICC)

Liver tissue samples were fixed in formalin, embedded in paraffin, and sectioned. Sections were stained using various techniques for histopathologic evaluation as detailed in previous publications³⁵. Hematoxylin and eosin (H&E) staining was performed to assess overall liver histopathology. Picrosirius Red staining (Sigma-Aldrich, 365548) was used to evaluate fibrosis intensity.

For IHC, sections were dewaxed, hydrated, and treated with 3% hydrogen peroxide for 10 min to block endogenous peroxidase activity. Antigen retrieval was performed by heating sections in 10 mmol/L sodium citrate buffer (pH 6.0) for 10 min. Sections were then blocked with Dako protein block solution (Agilent Technologies, Santa Clara, CA) for 1 h and incubated overnight at 4 °C with specific primary antibodies listed in Supplementary Table 4. Polymer-horseradish peroxidase secondary antibodies were applied for 1 h at room temperature, followed by detection using the Dako 3,3’-Diaminobenzidine Substrate Chromogen System. To detect liver DNA damage, sections were incubated with anti-gamma H2A.X (Abcam, ab11174) and anti-8-Hydroxy-2’-deoxyguanosine (Abcam, ab48508) primary antibodies, followed by Alexa Fluor™ 488 and Alexa Fluor™ 594 secondary antibodies (ThermoFisher Scientific) for fluorescence detection. Images were acquired and processed using Leica Microsystems. Positively stained areas were quantified using ImageJ.

Frozen liver tissue samples were cut at a thickness of 20 μm, fixed with 10% formalin, and stained with Oil Red O (Sigma-Aldrich, O0625) for 10 min to visualize lipid accumulation. Cellular senescence in the liver was assessed by SA-β-gal staining using a commercially available kit (Cell Signaling, 9860), following the manufacturer’s instructions. Results were examined using light microscopy, and positively stained areas were quantified using ImageJ.

For ICC, cells were washed, fixed in 4% paraformaldehyde, permeabilized, blocked with normal goat serum, and incubated overnight with anti-gamma H2A.X (Abcam, ab11174) primary antibodies. After washing with PBS, cells were incubated with Alexa Fluor™ 488 and Alexa Fluor™ 594 secondary antibodies (ThermoFisher Scientific) for 1 h at room temperature. Nuclei were visualized using 4’,6-diamidino-2-phenylindole (DAPI). Images were acquired and processed using Leica Microsystems, and positively stained areas were quantified using ImageJ.

Statistics and reproducibility

All experimental data were statistically analyzed using GraphPad Prism (version 10.2.1) and Microsoft Office Excel (Microsoft Office 2019). No statistical methods were used to predetermine sample sizes; however, our sample sizes are similar to those reported in previous publications. Mice were randomized and allocated to the experiments, but data collection and analysis were not randomized. Investigators were not blinded to the experimental conditions. No data points were excluded from our analysis, except for statistical outliers identified through Grubbs’ test (α = 0.05). These excluded data points are clearly specified and described in the corresponding figure legends. Data distribution was assumed to be normal but was not formally tested. Statistical significance between two groups was evaluated using the two-tailed student’s t test, while comparisons of multiple groups were assessed by one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparisons. A p ≤ 0.05 was considered statistically significant. RNA-seq dataset plots were generated using R ggplot2. The Wilcox non-parametric test or t-test was used to compare enrichment scores among groups, as implemented in R ggpubr software.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data generated in this study are included in this article and its supplementary materials. Source data are also provided with this article. The single-nucleus RNA-seq (snRNA-seq) data from the 22-week chow vs. CDA-HFD-fed mice have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE262939. Bulk RNA-seq data from senescent vs. non-senescent Huh7 cells (n = 3 samples/group), Dp44mT- vs vehicle- treated senescent Huh7 cells (n = 4 samples/group), and DpC-treated vs. vehicle-treated mice fed with CDA-HFD (n = 5 mice/group), have been also uploaded to GEO with accession code GSE288534 Other publicly available transcriptomic datasets used in this study can be accessed through the GEO database with the following accession numbers: GSE213623 (Duke MASLD cohort), GSE33814 (German MASLD cohort), GSE167523 (Japanese MASLD cohort), GSE135251 (European MASLD cohort), GSE193066 (primary MASLD-HCC) and GSE193080 (recurrent MASLD-HCC), GSE50244 (pancreatic islet cohort), and GSE202379 (human snRNA-seq cohort). The RNA-seq dataset for heart failure patients was obtained from Zenodo at https://zenodo.org/record/4287217. The RNA-seq dataset for kidney fibrosis patients was sourced from E-MTAB-2502 on the EMBL-EBI website at. Additionally, microarray data for the bariatric surgery cohort are available under GSE83452, and data for the statin treatment cohort can be found under GSE130991. The processed, integrated snRNAseq hepatocyte data consisting of 120,000 cells from 57 human samples with different disease stages is available at https://doi.org/10.5281/zenodo.14804772. Source data are provided with this paper.

Code availability

The algorithms used for the analysis in this study are all publicly available. All code is available at https://github.com/dumbaugh/MASH_SenescentHepatocytes.

Change history

06 May 2025
A Correction to this paper has been published: https://doi.org/10.1038/s41467-025-59084-8

References

Sanyal, A. J. et al. Prospective study of outcomes in adults with nonalcoholic fatty liver disease. N. Engl. J. Med. 385, 1559–1569 (2021).
Article CAS PubMed PubMed Central Google Scholar
Paik, J. M. et al. Global burden of NAFLD and chronic liver disease among adolescents and young adults. Hepatology 75, 1204–1217 (2022).
Article CAS PubMed Google Scholar
Estes, C. et al. Modeling NAFLD disease burden in China, France, Germany, Italy, Japan, Spain, United Kingdom, and United States for the period 2016-2030. J. Hepatol. 69, 896–904 (2018).
Article PubMed Google Scholar
Golabi, P. et al. Prevalence and long-term outcomes of non-alcoholic fatty liver disease among elderly individuals from the United States. BMC Gastroenterol. 19, 56 (2019).
Article PubMed PubMed Central Google Scholar
Lopez-Otin, C., Blasco, M. A., Partridge, L., Serrano, M. & Kroemer, G. Hallmarks of aging: an expanding universe. Cell 186, 243–278 (2023).
Article CAS PubMed Google Scholar
Papatheodoridi, A. M., Chrysavgis, L., Koutsilieris, M. & Chatzigeorgiou, A. The role of senescence in the development of nonalcoholic fatty liver disease and progression to nonalcoholic steatohepatitis. Hepatology 71, 363–374 (2020).
Article CAS PubMed Google Scholar
Paez-Ribes, M., Gonzalez-Gualda, E., Doherty, G. J. & Munoz-Espin, D. Targeting senescent cells in translational medicine. EMBO Mol. Med. 11, e10234 (2019).
Article CAS PubMed PubMed Central Google Scholar
Kirkland, J. L. & Tchkonia, T. Senolytic drugs: from discovery to translation. J. Intern. Med. 288, 518–536 (2020).
Article CAS PubMed PubMed Central Google Scholar
van Deursen, J. M. The role of senescent cells in ageing. Nature 509, 439–446 (2014).
Article ADS PubMed PubMed Central Google Scholar
Branzei, D. & Foiani, M. Regulation of DNA repair throughout the cell cycle. Nat. Rev. Mol. Cell Biol. 9, 297–308 (2008).
Article CAS PubMed Google Scholar
Stallaert, W. et al. The molecular architecture of cell cycle arrest. Mol. Syst. Biol. 18, e11087 (2022).
Article CAS PubMed PubMed Central Google Scholar
Aravinthan, A. D. & Alexander, G. J. M. Senescence in chronic liver disease: is the future in aging? J. Hepatol. 65, 825–834 (2016).
Article CAS PubMed Google Scholar
Huda, N. et al. Hepatic senescence, the good and the bad. World J. Gastroenterol. 25, 5069–5081 (2019).
Article CAS PubMed PubMed Central Google Scholar
Ferreira-Gonzalez, S., Rodrigo-Torres, D., Gadd, V. L. & Forbes, S. J. Cellular senescence in liver disease and regeneration. Semin. Liver Dis. 41, 50–66 (2021).
Article CAS PubMed Google Scholar
Zhu, Y. et al. The Achilles’ heel of senescent cells: from transcriptome to senolytic drugs. Aging Cell 14, 644–658 (2015).
Article CAS PubMed PubMed Central Google Scholar
Amor, C. et al. Senolytic CAR T cells reverse senescence-associated pathologies. Nature 583, 127–132 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Umbaugh, D. S., Nguyen, N. T., Smith, S. H., Ramachandran, A. & Jaeschke, H. The p21(+) perinecrotic hepatocytes produce the chemokine CXCL14 after a severe acetaminophen overdose promoting hepatocyte injury and delaying regeneration. Toxicology 504, 153804 (2024).
Article CAS PubMed Google Scholar
Raffaele, M. et al. Mild exacerbation of obesity- and age-dependent liver disease progression by senolytic cocktail dasatinib + quercetin. Cell Commun. Signal 19, 44 (2021).
Article CAS PubMed PubMed Central Google Scholar
Du, K. et al. Targeting YAP-mediated HSC death susceptibility and senescence for treatment of liver fibrosis. Hepatology 77, 1998–2015 (2023).
Article PubMed Google Scholar
Cohn, R. L., Gasek, N. S., Kuchel, G. A. & Xu, M. The heterogeneity of cellular senescence: insights at the single-cell level. Trends Cell Biol. 33, 9–17 (2023).
Article CAS PubMed Google Scholar
Hernandez-Segura, A. et al. Unmasking transcriptional heterogeneity in senescent cells. Curr. Biol. 27, 2652–2660 e2654 (2017).
Article CAS PubMed PubMed Central Google Scholar
Hense, J. D. et al. MASLD does not affect fertility and senolytics fail to prevent MASLD progression in male mice. Sci. Rep. 14, 17332 (2024).
Article CAS PubMed PubMed Central Google Scholar
Ogrodnik, M. et al. Cellular senescence drives age-dependent hepatic steatosis. Nat. Commun. 8, 15691 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Kovacovicova, K. et al. Senolytic cocktail dasatinib+quercetin (D+Q) does not enhance the efficacy of senescence-inducing chemotherapy in liver cancer. Front Oncol. 8, 459 (2018).
Article PubMed PubMed Central Google Scholar
Maeso-Diaz, R. et al. Targeting senescent hepatocytes using the thrombomodulin-PAR1 inhibitor vorapaxar ameliorates NAFLD progression. Hepatology 78, 1209–1222 (2023).
Article PubMed Google Scholar
Jensen-Cody, S. O. & Potthoff, M. J. Hepatokines and metabolism: deciphering communication from the liver. Mol. Metab. 44, 101138 (2021).
Article CAS PubMed Google Scholar
Rui, L. Energy metabolism in the liver. Compr. Physiol. 4, 177–197 (2014).
Article PubMed PubMed Central Google Scholar
Aravinthan, A. et al. Hepatocyte senescence predicts progression in non-alcohol-related fatty liver disease. J. Hepatol. 58, 549–556 (2013).
Article CAS PubMed Google Scholar
Engelmann, C. & Tacke, F. The potential role of cellular senescence in non-alcoholic fatty liver disease. Int. J. Mol. Sci. 23, https://doi.org/10.3390/ijms23020652 (2022).
Hernandez-Segura, A., Nehme, J. & Demaria, M. Hallmarks of cellular senescence. Trends Cell Biol. 28, 436–453 (2018).
Article CAS PubMed Google Scholar
Aksoy, O. et al. The atypical E2F family member E2F7 couples the p53 and RB pathways during cellular senescence. Genes Dev. 26, 1546–1557 (2012).
Article CAS PubMed PubMed Central Google Scholar
Campaner, S. et al. Myc, Cdk2 and cellular senescence: old players, new game. Cell Cycle 9, 3655–3661 (2010).
Article CAS PubMed Google Scholar
Biran, A. et al. Quantitative identification of senescent cells in aging and disease. Aging Cell 16, 661–671 (2017).
Article CAS PubMed PubMed Central Google Scholar
Ogrodnik, M. et al. Guidelines for minimal information on cellular senescence experimentation in vivo. Cell 187, 4150–4175 (2024).
Article CAS PubMed Google Scholar
Du, K. et al. Aging promotes metabolic dysfunction-associated steatotic liver disease by inducing ferroptotic stress. Nat. Aging 4, 949–968 (2024).
Article CAS PubMed Google Scholar
Govaere, O. et al. Transcriptomic profiling across the nonalcoholic fatty liver disease spectrum reveals gene signatures for steatohepatitis and fibrosis. Sci. Transl. Med. 12, https://doi.org/10.1126/scitranslmed.aba4448 (2020).
Fujiwara, N. et al. Molecular signatures of long-term hepatocellular carcinoma risk in nonalcoholic fatty liver disease. Sci. Transl. Med. 14, eabo4474 (2022).
Article CAS PubMed PubMed Central Google Scholar
Qian, T. et al. Molecular signature predictive of long-term liver fibrosis progression to inform antifibrotic drug development. Gastroenterology 162, 1210–1225 (2022).
Article CAS PubMed Google Scholar
Verschuren, L. et al. Development of a novel non-invasive biomarker panel for hepatic fibrosis in MASLD. Nat. Commun. 15, 4564 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Chakraborty, S., Anand, S. & Bhandari, R. K. Sex-specific expression of the human NAFLD-NASH transcriptional signatures in the liver of medaka with a history of ancestral bisphenol A exposure. bioRxiv, https://doi.org/10.1101/2024.05.19.594843 (2024).
Narasimhan, A. et al. Cellular senescence in obesity and associated complications: a new therapeutic target. Curr. Diab Rep. 22, 537–548 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ogrodnik, M. et al. Obesity-induced cellular senescence drives anxiety and impairs neurogenesis. Cell Metab. 29, 1233 (2019).
Article CAS PubMed PubMed Central Google Scholar
Lefebvre, P. et al. Interspecies NASH disease activity whole-genome profiling identifies a fibrogenic role of PPARalpha-regulated dermatopontin. JCI Insight 2, https://doi.org/10.1172/jci.insight.92264 (2017).
Ayada, I. et al. Dissecting the multifaceted impact of statin use on fatty liver disease: a multidimensional study. EBioMedicine 87, 104392 (2023).
Article CAS PubMed Google Scholar
Mollerhoj, M. B. et al. Hepatoprotective effects of semaglutide, lanifibranor and dietary intervention in the GAN diet-induced obese and biopsy-confirmed mouse model of NASH. Clin. Transl. Sci. 15, 1167–1186 (2022).
Article PubMed PubMed Central Google Scholar
Newsome, P. N. et al. A placebo-controlled trial of subcutaneous semaglutide in nonalcoholic steatohepatitis. N. Engl. J. Med 384, 1113–1124 (2021).
Article CAS PubMed Google Scholar
Saul, D. et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun. 13, 4827 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12, 1088 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Kim, K. H. & Lee, M. S. GDF15 as a central mediator for integrated stress response and a promising therapeutic molecule for metabolic disorders and NASH. Biochim. Biophys. Acta Gen. Subj. 1865, 129834 (2021).
Article CAS PubMed Google Scholar
Ha, G. et al. GDF15 secreted by senescent endothelial cells improves vascular progenitor cell functions. PLoS ONE 14, e0216602 (2019).
Article PubMed PubMed Central Google Scholar
Guo, Y. et al. Senescence-associated tissue microenvironment promotes colon cancer formation through the secretory factor GDF15. Aging Cell 18, e13013 (2019).
Article CAS PubMed PubMed Central Google Scholar
Conte, M. et al. GDF15, an emerging key player in human aging. Ageing Res. Rev. 75, 101569 (2022).
Article ADS CAS PubMed Google Scholar
Evans, D. S. et al. Proteomic analysis of the senescence-associated secretory phenotype: GDF-15, IGFBP-2, and cystatin-C are associated with multiple aging traits. J. Gerontol. A Biol. Sci. Med. Sci. 79, https://doi.org/10.1093/gerona/glad265 (2024).
Wang, D. et al. GDF15: emerging biology and therapeutic applications for obesity and cardiometabolic disease. Nat. Rev. Endocrinol. 17, 592–607 (2021).
Article CAS PubMed Google Scholar
Galuppo, B. et al. Growth differentiation factor 15 (GDF15) is associated with non-alcoholic fatty liver disease (NAFLD) in youth with overweight or obesity. Nutr. Diab. 12, 9 (2022).
Article CAS Google Scholar
Li, D., Zhang, H. & Zhong, Y. Hepatic GDF15 is regulated by CHOP of the unfolded protein response and alleviates NAFLD progression in obese mice. Biochem. Biophys. Res. Commun. 498, 388–394 (2018).
Article CAS PubMed Google Scholar
Kim, K. H. et al. Growth differentiation factor 15 ameliorates nonalcoholic steatohepatitis and related metabolic disorders in mice. Sci. Rep. 8, 6789 (2018).
Article ADS PubMed PubMed Central Google Scholar
Zhang, Z. et al. ARRB1 inhibits non-alcoholic steatohepatitis progression by promoting GDF15 maturation. J. Hepatol. 72, 976–989 (2020).
Article CAS PubMed Google Scholar
Qi, P., Ma, M. Z. & Kuai, J. H. Identification of growth differentiation factor 15 as a pro-fibrotic factor in mouse liver fibrosis progression. Int. J. Exp. Pathol. 102, 148–156 (2021).
Article CAS PubMed PubMed Central Google Scholar
Myojin, Y. et al. Hepatic stellate cells in hepatocellular carcinoma promote tumor growth via growth differentiation factor 15 production. Gastroenterology 160, 1741–1754 e1716 (2021).
Article CAS PubMed Google Scholar
Dong, G. et al. Treatment-damaged hepatocellular carcinoma promotes activities of hepatic stellate cells and fibrosis through GDF15. Exp. Cell Res. 370, 468–477 (2018).
Article CAS PubMed Google Scholar
Sturmlechner, I. et al. p21 produces a bioactive secretome that places stressed cells under immunosurveillance. Science 374, eabb3420 (2021).
Article CAS PubMed PubMed Central Google Scholar
Wu, T. et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation 2, 100141 (2021).
CAS PubMed PubMed Central Google Scholar
Tinkum, K. L. et al. Forkhead box O1 (FOXO1) protein, but not p53, contributes to robust induction of p21 expression in fasted mice. J. Biol. Chem. 288, 27999–28008 (2013).
Article CAS PubMed PubMed Central Google Scholar
Du, K. et al. Targeting YAP-mediated hepatic stellate cell death susceptibility and senescence for treatment of liver fibrosis. Hepatology 77, 1998–2015 (2023).
Bollard, J. et al. Palbociclib (PD-0332991), a selective CDK4/6 inhibitor, restricts tumour growth in preclinical models of hepatocellular carcinoma. Gut 66, 1286–1296 (2017).
Article CAS PubMed Google Scholar
He, Y. et al. Using proteolysis-targeting chimera technology to reduce navitoclax platelet toxicity and improve its senolytic activity. Nat. Commun. 11, 1996 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Al-Mansour, F. et al. Characterization of the HDAC/PI3K inhibitor CUDC-907 as a novel senolytic. Aging 15, 2373–2394 (2023).
Article CAS PubMed PubMed Central Google Scholar
Ribierre, T. et al. Targeting pathological cells with senolytic drugs reduces seizures in neurodevelopmental mTOR-related epilepsy. Nat. Neurosci. 27, 1125–1136 (2024).
Article CAS PubMed PubMed Central Google Scholar
Wakita, M. et al. A BET family protein degrader provokes senolysis by targeting NHEJ and autophagy in senescent cells. Nat. Commun. 11, 1935 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Nam, S. Y., Han, N. R., Yoon, K. W., Kim, H. M. & Jeong, H. J. Di-2-pyridylketone 4,4-dimethyl-3-thiosemicarbazone (Dp44mT), an anticancer agent, exerts an anti-inflammatory effect in activated human mast cells. Inflamm. Res. 66, 871–879 (2017).
Article CAS PubMed Google Scholar
Krishan, S., Richardson, D. R. & Sahni, S. The anticancer agent, di-2-pyridylketone 4,4-dimethyl-3-thiosemicarbazone (Dp44mt), up-regulates the AMPK-dependent energy homeostasis pathway in cancer cells. Biochim. Biophys. Acta 1863, 2916–2933 (2016).
Article CAS PubMed Google Scholar
Lovejoy, D. B. et al. Antitumor activity of metal-chelating compound Dp44mT is mediated by formation of a redox-active copper complex that accumulates in lysosomes. Cancer Res. 71, 5871–5880 (2011).
Article CAS PubMed Google Scholar
Sestak, V. et al. Novel and potent anti-tumor and anti-metastatic di-2-pyridylketone thiosemicarbazones demonstrate marked differences in pharmacology between the first and second generation lead agents. Oncotarget 6, 42411–42428 (2015).
Article PubMed PubMed Central Google Scholar
Guo, Z. L. et al. The novel thiosemicarbazone, di-2-pyridylketone 4-cyclohexyl-4-methyl-3-thiosemicarbazone (DpC), inhibits neuroblastoma growth in vitro and in vivo via multiple mechanisms. J. Hematol. Oncol. 9, 98 (2016).
Article PubMed PubMed Central Google Scholar
Masaldan, S. et al. Copper accumulation in senescent cells: interplay between copper transporters and impaired autophagy. Redox Biol. 16, 322–331 (2018).
Article CAS PubMed PubMed Central Google Scholar
Wiley, C. D. & Campisi, J. The metabolic roots of senescence: mechanisms and opportunities for intervention. Nat. Metab. 3, 1290–1301 (2021).
Article CAS PubMed PubMed Central Google Scholar
Lim, S., Kim, J. W. & Targher, G. Links between metabolic syndrome and metabolic dysfunction-associated fatty liver disease. Trends Endocrinol. Metab. 32, 500–514 (2021).
Article CAS PubMed Google Scholar
Byrne, C. D. & Targher, G. NAFLD: a multisystem disease. J. Hepatol. 62, S47–S64 (2015).
Article PubMed Google Scholar
Klein, S., Gastaldelli, A., Yki-Jarvinen, H. & Scherer, P. E. Why does obesity cause diabetes? Cell Metab. 34, 11–20 (2022).
Article CAS PubMed PubMed Central Google Scholar
Fadista, J. et al. Global genomic and transcriptomic analysis of human pancreatic islets reveals novel genes influencing glucose metabolism. Proc. Natl. Acad. Sci. USA 111, 13924–13929 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Hahn, V. S. et al. Myocardial gene expression signatures in human heart failure with preserved ejection fraction. Circulation 143, 120–134 (2021).
Article CAS PubMed Google Scholar
Kang, H. M. et al. Defective fatty acid oxidation in renal tubular epithelial cells has a key role in kidney fibrosis development. Nat. Med. 21, 37–46 (2015).
Article CAS PubMed Google Scholar
Li, Y. et al. Growth differentiation factor 15: emerging role in liver diseases. Cytokine 182, 156727 (2024).
Article CAS PubMed Google Scholar
Buendgens, L. et al. Growth differentiation factor-15 is a predictor of mortality in critically Ill patients with sepsis. Dis. Markers 2017, 5271203 (2017).
Article PubMed PubMed Central Google Scholar
Corradini, E. et al. Ceruloplasmin gene variants are associated with hyperferritinemia and increased liver iron in patients with NAFLD. J. Hepatol. 75, 506–513 (2021).
Article CAS PubMed Google Scholar
Stattermayer, A. F. et al. Hepatic steatosis in Wilson disease—Role of copper and PNPLA3 mutations. J. Hepatol. 63, 156–163 (2015).
Article PubMed Google Scholar
Li, Y. et al. Iron and copper: critical executioners of ferroptosis, cuproptosis and other forms of cell death. Cell Commun. Signal. 21, 327 (2023).
Article CAS PubMed PubMed Central Google Scholar
Wu, J. R. et al. GDF15-mediated upregulation of ferroportin plays a key role in the development of hyperferritinemia in children with hemophagocytic lymphohistiocytosis. Pediatr. Blood Cancer 60, 940–945 (2013).
Article CAS PubMed Google Scholar
Dolgin, E. Send in the senolytics. Nat. Biotechnol. 38, 1371–1377 (2020).
Article CAS PubMed Google Scholar
Kiourtis, C. et al. Hepatocellular senescence induces multi-organ senescence and dysfunction via TGFbeta. Nat. Cell Biol. https://doi.org/10.1038/s41556-024-01543-3 (2024).
Du, K., Umbaugh, D. S., Dutta, R. K. & Diehl, A. M. A systemic effect for liver senescence. Nat. Cell Biol. https://doi.org/10.1038/s41556-024-01520-w (2024).
Sanfeliu-Redondo, D., Gibert-Ramos, A. & Gracia-Sancho, J. Cell senescence in liver diseases: pathological mechanism and theranostic opportunity. Nat. Rev. Gastroenterol. Hepatol. 21, 477–492 (2024).
Article PubMed Google Scholar
Krizhanovsky, V. et al. Senescence of activated stellate cells limits liver fibrosis. Cell 134, 657–667 (2008).
Article CAS PubMed PubMed Central Google Scholar
Du, K., Jun, J. H., Dutta, R. K. & Diehl, A. M. Plasticity, heterogeneity, and multifunctionality of hepatic stellate cells in liver pathophysiology. Hepatol. Commun. 8, https://doi.org/10.1097/HC9.0000000000000411 (2024).
Omori, S. et al. Generation of a p16 reporter mouse and its use to characterize and target p16(high) cells in vivo. Cell Metab. 32, 814–828 e816 (2020).
Article CAS PubMed Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Hanzelmann, S., Castelo, R. & Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinform. 14, 7 (2013).
Article Google Scholar
Chen, T. et al. Hepatocyte smoothened activity controls susceptibility to insulin resistance and nonalcoholic fatty liver disease. Cell Mol. Gastroenterol. Hepatol. 15, 949–970 (2023).
Article CAS PubMed Google Scholar
Starmann, J. et al. Gene expression profiling unravels cancer-related hepatic molecular signatures in steatohepatitis but not in steatosis. PLoS One 7, e46584 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Kozumi, K. et al. Transcriptomics identify thrombospondin-2 as a biomarker for NASH and advanced liver fibrosis. Hepatology 74, 2452–2466 (2021).
Article CAS PubMed Google Scholar
Margerie, D. et al. Hepatic transcriptomic signatures of statin treatment are associated with impaired glucose homeostasis in severely obese patients. BMC Med. Genom. 12, 80 (2019).
Article Google Scholar
de Magalhaes, J. P. et al. Human Ageing Genomic Resources: updates on key databases in ageing research. Nucleic Acids Res. 52, D900–D908 (2024).
Article PubMed Google Scholar
Suryadevara, V. et al. SenNet recommendations for detecting senescent cells in different tissues. Nat. Rev. Mol. Cell Biol. 25, 1001–1023 (2024).
Article CAS PubMed Google Scholar
Gorgoulis, V. et al. Cellular senescence: defining a path forward. Cell 179, 813–827 (2019).
Article CAS PubMed Google Scholar
De Cecco, M. et al. L1 drives IFN in senescent cells and promotes age-associated inflammation. Nature 566, 73–78 (2019).
Article ADS PubMed PubMed Central Google Scholar
Freund, A., Orjalo, A. V., Desprez, P. Y. & Campisi, J. Inflammatory networks during cellular senescence: causes and consequences. Trends Mol. Med. 16, 238–246 (2010).
Article CAS PubMed PubMed Central Google Scholar
Zhao, M., Chen, L. & Qu, H. CSGene: a literature-based database for cell senescence genes and its application to identify critical cell aging pathways and associated diseases. Cell Death Dis. 7, e2053 (2016).
Article CAS PubMed PubMed Central Google Scholar
Casella, G. et al. Transcriptome signature of cellular senescence. Nucleic Acids Res. 47, 7294–7305 (2019).
Article CAS PubMed PubMed Central Google Scholar
Gribben, C. et al. Acquisition of epithelial plasticity in human chronic liver disease. Nature 630, 166–173 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Robarts, D. R. et al. The essential role of O-GlcNAcylation in hepatic differentiation. Hepatol. Commun. 7, https://doi.org/10.1097/HC9.0000000000000283 (2023).
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 e3529 (2021).
Article CAS PubMed PubMed Central Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
Article CAS PubMed PubMed Central Google Scholar
Aibar, S. et al. SCENIC: single-cell regulatory network inference and clustering. Nat. Methods 14, 1083–1086 (2017).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

This work was supported by the NIDDK K01DK135793-01A1 and 2021 AASLD Pinnacle Award to K.D., and National Institutes of Health grants R01 AA010154, R01 DK077794, R56 DK134334 and Sponsored Research Study Agreement 337521 supported by Boehringer Ingelheim Pharmaceuticals, Inc, awarded to A.M.D. The authors are grateful to the patients who donated liver tissue for analysis, to Dr. Cynthia Guy and the clinical staff and coordinators who created and maintain the Duke NAFLD Biorepository, Emily Hocke, Stephanie Arvai, and Karen Abramson from the Duke Molecular Genetics Core for single nuclei RMA sequencing of control and MASLD liver tissues, Dr. Steven Pullen’s team at Boehringer Ingelheim Pharmaceuticals, Inc. for their assistance in the bioinformatics analysis, Dr. Christopher C. Pan from Duke Department of Molecular Genetics and Microbiology (MGM) for assistance in preparing RNA-seq samples of in vitro model of hepatocyte senescence, and Dr. Shengli Song from Duke Department of Surgery for technical assistance in flow cytometry.

Author information

Authors and Affiliations

Department of Medicine, Duke University, Durham, NC, USA
Kuo Du, David S. Umbaugh, Ji Hye Jun, Rajesh K. Dutta, Seh Hoon Oh, Niansheng Ren & Anna Mae Diehl
Department of Molecular Genetics and Microbiology, Duke University, Durham, NC, USA
Liuyang Wang & Dennis C. Ko
Department of Neurology, Duke University, Durham, NC, USA
Qiaojuan Zhang
Boehringer Ingelheim Pharmaceuticals Inc, Ridgefield, CT, USA
Ana Ferreira, Jon Hill, Guannan Gao & Steven S. Pullen
Duke Molecular Physiology Institute, Duke University, Durham, NC, USA
Vaibhav Jain & Simon Gregory
Department of Medicine, Mayo Clinic, Rochester, MN, USA
Manal F. Abdelmalek

Authors

Kuo Du
View author publications
Search author on:PubMed Google Scholar
David S. Umbaugh
View author publications
Search author on:PubMed Google Scholar
Liuyang Wang
View author publications
Search author on:PubMed Google Scholar
Ji Hye Jun
View author publications
Search author on:PubMed Google Scholar
Rajesh K. Dutta
View author publications
Search author on:PubMed Google Scholar
Seh Hoon Oh
View author publications
Search author on:PubMed Google Scholar
Niansheng Ren
View author publications
Search author on:PubMed Google Scholar
Qiaojuan Zhang
View author publications
Search author on:PubMed Google Scholar
Dennis C. Ko
View author publications
Search author on:PubMed Google Scholar
Ana Ferreira
View author publications
Search author on:PubMed Google Scholar
Jon Hill
View author publications
Search author on:PubMed Google Scholar
Guannan Gao
View author publications
Search author on:PubMed Google Scholar
Steven S. Pullen
View author publications
Search author on:PubMed Google Scholar
Vaibhav Jain
View author publications
Search author on:PubMed Google Scholar
Simon Gregory
View author publications
Search author on:PubMed Google Scholar
Manal F. Abdelmalek
View author publications
Search author on:PubMed Google Scholar
Anna Mae Diehl
View author publications
Search author on:PubMed Google Scholar

Contributions

K.D. and A.M.D. conceived of the experiments. K.D. and J.H.J. performed experiments. K.D., D.S.U, L.W., J.H.J., R.K.D., S.H.O., N.R., Q.Z, D.C.K., A.F., J.H., G.G., S.S.P., V.J., S.G., M.F.A. and A.M.D. analyzed data. K.D., D.S.U., and A.M.D. wrote the manuscript. K.D. and A.M.D. secured funding for the study. Everyone reviewed and approved the manuscript. Correspondence to A.M.D. or K.D.

Corresponding authors

Correspondence to Kuo Du or Anna Mae Diehl.

Ethics declarations

Competing interests

A.M.D., S.G., and M.F.A. received grant funding from Boehringer Ingelheim paid to Duke University to support human transcriptomics analysis. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Data processing file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Du, K., Umbaugh, D.S., Wang, L. et al. Targeting senescent hepatocytes for treatment of metabolic dysfunction-associated steatotic liver disease and multi-organ dysfunction. Nat Commun 16, 3038 (2025). https://doi.org/10.1038/s41467-025-57616-w

Download citation

Received: 19 December 2024
Accepted: 23 February 2025
Published: 28 March 2025
DOI: https://doi.org/10.1038/s41467-025-57616-w