The senescence-associated secretome of hedgehog-deficient hepatocytes drives MASLD progression

The burden of senescent hepatocytes correlates with the severity of Metabolic dysfunction associated steatotic liver disease (MASLD) but mechanisms driving senescence, and how it exacerbates MASLD are poorly understood. Hepatocytes experience lipotoxicity and become senescent when Smoothened (Smo) is deleted to disrupt Hedgehog signaling. We aimed to determine if the secretomes of Smo-deficient hepatocytes perpetuate senescence to drive MASLD progression. RNA seq analysis of liver samples from human and murine cohorts with MASLD confirmed that hepatocyte populations of MASLD livers are depleted of Smo(+) cells and enriched with senescent cells. When fed choline deficient amino acid restricted high fat diet (CDA-HFD) to induce MASLD, Smo(-) mice had lower antioxidant markers and developed worse DNA damage, senescence, steatohepatitis and fibrosis than Smo(+) mice. Sera and hepatocyte-conditioned medium from Smo(-) mice were depleted of thymidine phosphorylase (TP), a protein that maintains mitochondrial fitness. Treating Smo(-) hepatocytes with TP reduced senescence and lipotoxicity; inhibiting TP in Smo(+) hepatocytes had the opposite effects and exacerbated hepatocyte senescence, steatohepatitis, and fibrosis in CDA-HFD-fed mice. We conclude that inhibiting Hedgehog signaling in hepatocytes promotes MASLD by suppressing hepatocyte production of proteins that prevent lipotoxicity and senescence.


Introduction
Over the last fifty years, much of the world's population has transitioned into a state of relative energy surplus due to improvements in food availability and declining demands for physical activity.This shift has imposed challenges on homeostatic mechanisms that maintain tissue energy balance, spawning a global increase in diseases caused by metabolic dysfunction (1).
Metabolic dysfunction-associated steatotic liver disease (MASLD) is now a leading cause of chronic liver disease in most countries (2).Liver-related mortality in MASLD is greatest in people with MASLD cirrhosis and associates with increased mortality from other causes (3).
MASLD cirrhosis is the end result of a chronic degenerative process that develops when constitutive regenerative mechanisms to replace dead hepatocytes are overwhelmed, leading to progressive replacement of functional parenchyma with fibrous scar.Thus, development of interventions to prevent, arrest and/or reverse MASLD progression requires deeper understanding of the inherent mechanisms that normally maintain liver homeostasis.
The risk for MASLD cirrhosis and other metabolic dysfunction-related degenerative diseases varies among individuals but generally increases with age (4).Older age increases the incidence and prevalence of degenerative diseases because epigenetic modifications that promote cellular metabolic dysfunction accumulate with aging.This amplifies the tissue burden of dysfunctional cells that are unable to regenerate, but which secrete factors that perpetuate maladaptive repair.Thus, cellular senescence and related senescence associated secretory phenotypes (SASPs) are acknowledged hallmarks of aging tissues (5).MASLD and other chronic liver diseases induce hepatocyte DNA damage and senescence, and MASLD severity correlates with the burden of senescent hepatocytes (6).However, the mechanisms driving hepatocyte senescence, and how this exacerbates MASLD, are poorly understood.
Hedgehog, a lipid-regulated morphogenic signaling pathway, is known to inhibit senescence and promote longevity (7).Hedgehog signaling is tightly coupled to primary cilia and thus, nutrient sensing (8).Hedgehog maintains organelle quality control mechanisms and optimizes energy retrieval while repressing expression of cell cycle inhibitors (9).Hence, growing and regenerating tissues, as well as many cancers, exhibit high Hedgehog pathway activity (10).
Previous studies demonstrated increased Hedgehog signaling in chronically injured livers and liver cancers, but barely detectable pathway activity in healthy livers, consistent with evidence that hepatocytes replicate very infrequently in adulthood (11).Also, as mentioned earlier, Hedgehog signaling is tightly coupled to cilia, and cilia are demonstrated in fewer than 10% of adult hepatocytes at a given time (12).Together, these findings led to the assumption that the Hedgehog pathway is dormant in the vast majority of healthy adult hepatocytes and thus, plays a negligible role in regulating hepatocyte function in uninjured livers.However, work by us and others challenge this concept.As summarized in two recent reviews (13,14), profound dysregulation of fatty acid, cholesterol, bile acid and glucose metabolism ensue when the Hedgehog pathway is disrupted in hepatocytes.Further, our recent analysis of hepatocyte RNA seq data sets from healthy young and old mice revealed substantial age-related differences in hepatocyte Hedgehog pathway activity.Basal Hedgehog signaling was repressed in hepatocytes from old mice.Conversely, hepatocytes from young mice had higher pathway activity and quickly became senescent when Smoothened (Smo, an obligatory Hedgehog pathway component) was disrupted experimentally (15).Although the mechanisms involved have not yet been defined, we found that disrupting Hedgehog signaling in hepatocytes evokes multiple defects that are known to trigger cell senescence, including impaired growth factor signaling, mitochondrial dysfunction, autophagy inhibition, increased oxidative stress, DNA damage and telomere attrition.Remarkably, this was sufficient to induce MASLD in healthy young chow-fed mice.Concomitant analysis of archived human liver samples showed that the burden of Smo-depleted hepatocytes is markedly greater in patients with MASLD than age-matched controls and parallels senescent hepatocyte accumulation and progression to cirrhosis in people with MASLD (16).Together, these findings indicate that inhibiting Hedgehog pathway activity in hepatocytes accelerates liver aging.This suggests that Hedgehog-deficient hepatocytes may be fundamentally responsible for the maladaptive regenerative responses that drive progressive hepatic dysfunction and degeneration in MASLD.
Similar to other senescent cell types, Smo-depleted hepatocytes remain viable but their state change induces SASPs that robustly effect their secretomes.SASPs critically shape repair responses and cumulative aging-related changes in SASPs can progressively corrupt regenerative efforts, resulting in tissue degeneration (17).Hence, in the current study, we evaluated the hypothesis that the secretome of Hedgehog-deficient hepatocytes perpetuates senescence to drive MASLD progression.Unexpectedly, our initial efforts to profile the secreted proteome of Smo-depleted hepatocytes demonstrated striking depletion of thymidine phosphorylase (TP), a protein that is normally induced by antioxidant mechanisms and functions to inhibit senescence by preserving mitochondrial fitness and optimizing repair of damaged DNA (18).Loss of function mutations of Tymp, the human gene that encodes TP, cause a rare mitochondriopathy that is typically fatal by third-fourth decade of life (19).Our findings identify a role for Smo in the regulation of TP and thus, suggest a mechanism whereby Hedgehog signaling enables hepatic metabolic resiliency.Importantly, the results demonstrate that perturbations of these basal defenses accelerate liver aging and promote pathogenesis/progression of MASLD.

Hepatocyte-specific deletion of Smo exacerbates MASLD in mice
Smo is an obligatory component of the Hedgehog pathway.To determine if (and how) changes in hepatocyte Hedgehog activity influence MASLD progression, we fed Smo flox/flox mice a MASH inducing diet for 6 weeks and used viral vectors to delete Smo selectively in hepatocytes during the final week.Results in mice with hepatocyte-specific deletion of Smo (Smo KO) were compared to Smo flox/flox mice that were treated with control vectors (Fig. 1A).
Our previous work in chow-fed Smo flox/flox mice demonstrated that this gene targeting approach selectively depletes Smo in hepatocytes by ~90% within 48-72 hours (20).Principle component analysis of bulk RNA-seq data sets of livers from four CDA-HFD-fed control mice and four CDA-HFD-fed mice with Smo-depleted hepatocytes (Smo KO) mice confirms clustering of transcriptomes within each experimental group and reveals substantial differences in the transcriptomes of the control versus Smo KO livers (Supple Figs.1A and   B).Further Gene Set Enrichment Analysis (GSEA) revealed that genes related to adipogenesis and fatty acid metabolism are up-regulated, while those involved in cholesterol homeostasis are down-regulated in Smo KO livers (Supple Fig. 1C).Also, although serum glucose levels are similar in CDA-HFD fed control mice and mice with Smo-depleted hepatocytes, serum insulin levels and HOMA-IR are substantially higher in the latter group, demonstrating that deleting Smo in hepatocytes promotes insulin resistance (Supple Fig. 1D).
Activating Smo is known to stabilize Gli2 protein and enable nuclear accumulation of this transcription factor (21).Consistent with this, we found that deleting Smo in hepatocytes decreased nuclear accumulation of Gli2 in CDA-HFD-fed mice (Figs.1B, C).Mice with decreased hepatocyte Hedgehog signaling also exhibited increased hepatic inflammation (e.g., F480) and fibrosis (e.g., ASMA, Col1, Vimentin, and Desmin) (Figs. 1B-D).We reported previously that the severity of fibrosing steatohepatitis parallels hepatic expression of senescence markers, p16, p21, and B galactosidase activity in people with MASLD (15,16), and all of these markers were increased by deleting Smo in mouse hepatocytes during CDA-HFD exposure (Figs.1D-F).The increased senescent cell burden likely reflects exacerbated DNA damage in Smo-depleted hepatocytes because accumulation of rH2AX and 8-hydroxy-2'-deoxyguanosine (8OHDG) also increased in hepatocyte nuclei.In addition, disrupting hepatocyte Hedgehog signaling increased other markers of hepatic lipotoxicity, as evidenced by quantitative histochemistry for TUNEL and steatosis (Oil Red O), as well as increased lipid peroxidation markers (malondialdehyde, MDA; 4-hydroxynonenal, 4HNE) in liver and serum (Figs.1E-G).Together, these results indicate that disrupting hepatocyte Hedgehog signaling lowers the threshold for lipotoxicity, accelerates senescence and exacerbates fibroinflammatory responses in liver cells that together, result in maladaptive repair and promote MASLD progression.

Hepatocyte Smo-KO secretome is depleted of factors that promote antioxidant defense
During the senescence process, cells acquire various senescence associated secretory phenotypes (SASPs) that shape regenerative responses (22).Our data indicate that deleting Smo in hepatocytes corrupts regeneration in injured livers.To identify differentially-induced or repressed proteins that might mediate these maladaptive repair responses we used a commercially available platform (Proteome Profiler Mouse Cytokine Array) to compare the secretomes in sera and hepatocyte-conditioned medium from chow-fed Smo-KO and control mice.Various cytokines (e.g., TNFA, IL1A, IL1B, IL6, IL15, IL33, GMCSF, MCSF, and IFNr) were substantially higher in sera and/or conditioned medium from the Smo KO groups compared to the control groups.Interestingly, one protein, TP, was maximally down-regulated in both the sera and conditioned medium from the Smo-KO group relative to the control groups (Figs.2A, B).TP is a pro-angiogenic factor that promotes mitochondrial fitness, reduces oxidant stress and inhibits senescence (23).Loss of function polymorphisms in Tymp (the gene that encodes TP) cause mitochondrial neurogastrointestinal encephalomyopathy (MNGIE) syndrome, a rare human mitochondriopathy that is typically fatal by mid-adulthood (24).Because our current work demonstrate that Smo depletion causes mitochondrial dysfunction, oxidative stress, DNA damage and senescence, we elected to determine whether TP depletion had a role in the phenotype of mice with Smo-depleted hepatocytes.
Immunoblot analysis demonstrated that the protein expression of TP is highest in chow-fed control livers, declines in livers of CDA-HFD fed mice with MASLD, and is lower in Smo KO livers than the respective controls both before and during CDA-HFD exposure (Fig. 2C).
Immunohistochemistry demonstrated that TP localizes in both hepatocyte cytoplasm and nuclei in control livers and is markedly depleted in both compartments in Smo KO livers (Figs.2D, E).Serum levels of TP are also lower in Smo KO mice than controls (Fig. 2F).
To assess MASLD-related changes in human hepatocyte expression of SMO and TYMP, we re-analyzed publicly available single nucleus (sn) RNA seq data sets from 2 control human livers and 2 human livers with advanced MASLD (GSE174748).Our previous deconvolution of bulk liver RNA seq data from a large cohort of MASLD patients (GSE213623) indicated that hepatocyte Smo expression inversely correlates with MASLD severity in people (16).The present snRNA seq analysis confirms that Smo(+) cells are lower in the hepatocyte population of livers with MASLD than in control livers (Supple Fig. 2A).TYMP-expressing hepatocytes are also less abundant in the MASLD livers than control livers in this snRNA seq data set (Supple Fig. 2A).To localize expression of TP protein, we performed immunostaining in a representative subset of the human livers that had been processed to generate our previously published bulk liver RNA seq data (16) (GSE213623).As observed in the mouse livers (Figs.2C-F), nuclei and cytosol of many hepatocytes stained strongly for TP in control human livers and TP protein decreased in hepatocytes as the severity of MASLD fibrosis increased (Figs. 2G, H).Together, these findings support the concept that livers with worse MASLD have fewer TP-expressing hepatocytes.However, serum levels of TP protein are not decreased and whole liver expression of TYMP mRNAs is upregulated in MASLD patients relative to controls (Supple Fig. 2B, D).These discrepancies may reflect cell type-specific disease-related differences in TYMP expression because liver macrophage populations express TYMP and the relative abundance of these cells increases in MASLD (data not shown).
TP inhibits oxidant stress and its expression is up-regulated when antioxidant defense mechanisms are induced by nuclear factor erythroid 2-related factor 2 (Nrf2), a master transcriptional regulator of antioxidant defense responses (25).The responsible mechanisms remain unclear, however, as the Tymp promoter lacks obvious Nrf2 binding sites.In a previously reported study of lung cancer cells (26), levels of TP protein were increased by over-expressing HO1, the protein encoded by Hmox1, a direct transcriptional target of Nrf2.
However, in those studies, deleting HO1 only partially diminished Nrf2's positive effects on TP expression, while dose-related increases in TP expression resulted when the cells were treated with N-acetyl cysteine.The researchers concluded that yet-to-be-determined Nrf2sensitive factors antioxidant factors (including, but not limited to, HO1) promote TP accumulation.To determine if/how the Nrf2/HO1/TP axis is impacted in MASLD, we leveraged both the Duke MASLD patient bulk RNA-seq dataset (GSE213623) and the publicly available single nucleus RNA seq data set (GSE174748) to compare expression of NFE2I2 (the gene that encodes Nrf2) and HMOX1 in whole liver tissue and hepatocyte populations of human livers with and without MASLD, respectively.Expression of N FE2I2-and HMOX1 were decreased in both liver tissues and hepatocytes in MASLD livers (Supple Figs.2A, D).We also used quantitative immunohistochemistry and western blotting to determine how changes in TP protein relate to changes in protein levels of Nrf2 and HO1 and found that changes in hepatocyte accumulation of Nrf2 and HO1 parallel changes in TP, i.e., levels of all three proteins decrease with increasing severity of MASLD in humans (Figs.2G, H) and mice (Figs.

2C-F, 3A-B).
Impaired mitochondrial fitness in Smo-KO hepatocytes with decreased TP, Nrf2, and HO1 Coordinated expression of TP, Nrf2 and HO1 proteins is not a unique attribute of hepatocytes as this was demonstrated previously in cancer cell lines (26).In many cancers, TP is overexpressed and loss of TP promotes mitochondrial dysfunction and oxidant stress (27).
Therefore, we asked if TP, Nrf2 and HO1 might be involved in the mitochondrial integrated stress response (ISRmt) in hepatocytes.The ISRmt optimizes mitochondrial fitness by inducing adaptive mechanisms to mitigate stress at the cellular level.During the ISRmt, damaged mitochondria release factors that trigger retrograde signaling to the nucleus to induce transcription of nuclear genes that encode mitokines, such as growth/differentiation factor 15 (GDF15).The resultant increases in serum levels of GDF protein are used clinically as predictive biomarkers of morbidity and mortality in mitochondrial diseases, including those that induce mitochondrial DNA damage by disrupting the thymidine salvage pathway (28).
Because GDF15 mRNA and protein levels are tightly coupled to the intensity of the ISR, we predicted that decreased accumulation of proteins that limit oxidative stress-related damage to mitochondria would amplify up-regulation of GDF15.Analysis of the snRNA seq data set showed that MASLD livers tended to accumulate more GDF15(+) hepatocytes than healthy livers (Suppl Fig. 2A).Further, the bulk liver RNA seq data from the large MASLD cohort, as well as the O-link proteomic analysis of sera from that cohort, indicate that levels of GDF15 mRNA and protein increase with the severity of MASLD fibrosis in humans (Supple Figs.2C,   D).Importantly, levels of GDF15 protein were also increased in conditioned medium from Smo KO hepatocytes and sera of Smo KO mice (Figs.2A, B), suggesting that Smo-depleted hepatocytes with the ISRmt are an important source of GDF15 in MASLD.
Although the molecular mechanisms whereby Smo might regulate the ISRmt remain to be determined, interactions between the Hedgehog pathway and Nrf2 are well-documented.
For example, Nrf2 has been reported to activate transcription of the gene that encodes Sonic hedgehog ligand in liver cancer stem cells and replenishing Shh restores Hedgehog signaling and re-establishes the stem cell phenotype in Nrf2-depleted cells (29).Interestingly, Marin-Hurtado and colleagues demonstrated that deletion of the Nfe2l2/Nrf2 gene in fibroblasts compromises their antioxidant defenses and suppresses expression of multiple genes that encode cilia-associated factors, including components of the Hedgehog pathway, leading both to impaired ciliognesis and decreased Hedgehog signaling (30).Recent studies in astrocytes also demonstrate that cilia homeostasis is disrupted by mitochondrial dysfunction and resultant oxidant stress, leading the authors to propose that ciliary signaling is part of the ISRmt in those cells (31).The aggregate findings support other evidence that Hedgehog pathway activity is mito-protective (32), and suggest that Nrf2-senstive factors may be involved.Indeed, Hedgehog signaling is activated when HO1 is over-expressed in cancer cells (33) and as noted above, enforcing HO1 expression promotes rapid and progressive accumulation of TP protein (26).Our data show that deleting Smo both disrupts Hedgehog signaling and reduces accumulation of TP, HO1 and Nrf2.Therefore, to identify mechanisms whereby Smo might enhance accumulation of these proteins, we performed immunoprecipitation experiments in extracts from primary hepatocytes.Remarkably, we found that Smo, TP, Nrf2, and HO1 physically interact (Fig. 3C).This finding suggests that Smo might be involved in post-transcriptional mechanisms that enable cellular accumulation of factors that critically mediate antioxidant defense and conversely, that factors that promote antioxidant defense help to maintain Hedgehog pathway integrity and thus, ciliary homeostasis.TP maintains nucleoside pools for mitochondrial DNA synthesis and thus, depleting TP compromises repair of mitochondrial DNA, and leads to mitochondrial dysfunction and increased mitochondrial ROS production (34).To determine whether Smo-depleted hepatocytes that are unable to form complexes of TP/Nrf2/HO1 exhibit impaired mitochondrial fitness, we isolated mitochondria from livers of CDA-HFD-fed Smo-KO and control mice and performed immunoblot analysis for TP, Smo, Nrf2, HO1, OXPHOS components, and other mitochondrial markers.Deleting Smo decreased mitochondrial accumulation of TP and HO1, and markedly down regulated OXPHOS complexes 2, 3, and 5, as well as mitochondrial markers (e.g., succinate dehydrogenase complex flavoprotein subunit A, SDHA; pyruvate dehydrogenase complex, PDH; prohibitin 1, PHB1; heat shock protein 60, HSP60; voltage-dependent anion channel, VDAC; and superoxide dismutase, SOD) (Fig. 3D).Smo and Nrf2 proteins were not detected in mitochondria of Smo KO or control mice.However, expression of the nuclear DNA-encoded mitochondrial biogenesis marker and Nrf2-target gene, peroxisome proliferator-activated receptor gamma coactivator 1 a (PGC1a), was substantially decreased in the CDA-HFD Smo KO livers (Fig. 3E), and nuclear accumulation of Nrf2 is markedly decreased in hepatocytes of these mice (Fig. 3A).In addition, the Smo-KO group had lower mitochondrial levels of apoptosis-inducing factor (AIF) and second mitochondriaderived activator of caspase (Smac) but higher cytosolic levels of these proteins (Fig. 3F), consistent with increased mitochondrial membrane permeability and activation of mechanisms that promote apoptosis.Hence, the collective data suggest a conceptual model for maintenance of hepatocyte resiliency whereby Smo, an obligatory signaling component of the Hedgehog pathway, regulates the bioavailability of ISRmt-related antioxidant proteins to maintain mitochondrial fitness and assure hepatocyte metabolic flexibility and viability (Fig 3G).Therefore, depleting Smo is predicted to enhance hepatocyte susceptibility to lipotoxicity but limit hepatocyte regenerative capacity and thus, exacerbate MASLD.

TP protects hepatocyte cultures from lipotoxicity and senescence
Lipotoxicity promotes senescence and accumulation of senescent hepatocytes parallels progression of liver damage in MASLD.To determine whether TP directly influences hepatocyte susceptibility to senescence, we treated AML-12 cells (a mouse hepatocyte cell line) with palbociclib (a selective inhibitor of cyclin-dependent kinases CDK4 and CDK6) for five days to induce senescence and then supplemented the culture medium with either recombinant TP protein or vehicle for an additional 48 hours (Supple Fig. 3A).Results in AML-12 cells that had been cultured in palbociclib-supplemented medium for 7 days were compared to cultures of AML-12 cells that had been maintained in normal growth-promoting medium for the same duration.As expected, cultures treated with palbociclib and vehicle had substantially fewer hepatocytes than control cultures.Remarkably, treating palbociclib cultures with TP restored cell numbers to levels of control cultures that had not been exposed to palbociclib (Supple Fig. 3B).Also, compared to control cultures, palbociclib cultures that were treated with vehicle exhibited increased steatosis (Oil Red O), lipotoxicity (4HNE) and senescence (B galactosidase).Treating palbociclib cultures with TP substantially reduced Oil Red O staining, 4HNE and B galactosidase activity (Supple Figs.3B, C).Together, these data show that TP acts directly on hepatocytes to inhibit lipotoxicity and senescence.

Manipulating TP in cultured hepatocytes modulates antioxidants and susceptibility to lipotoxicity and senescence
Next, we tested the effects of TP in an in vitro model of hepatocyte lipotoxicity in which Huh7 cells were cultured with oleate and palmitic acid (OPA) for a total of 4 days.During the last two days of culture under these lipotoxic conditions, half the cultures were treated with recombinant TP protein and half were treated with a pharmacologic TP inhibitor (TPI).Results in the OPA (lipotoxic) cultures were compared to Huh7 cells that had been cultured for 4 days in normal growth medium (Supple Fig. 4A).As expected, OPA cultures exhibited substantially more steatosis (Oil Red O) and senescence (B galactosidase activity) than control cultures (Supple Fig. 4B).Relative to control cultures, OPA cultures also had decreased expression of Nrf2 and HO1 and increased expression of p21 (Supple Fig. 4C), supporting the concept that lipotoxic cultures had decreased antioxidant defense and increased cell cycle arrest.
Remarkably, treating lipotoxic cultures with recombinant TP protein reduced Oil Red O staining and -galactosidase activity, increased Nrf2 and HO1, and decreased p21, while inhibiting TP had the opposite effect on each of these parameters (Supple Figs.4B, C).Hence, the collective results of these studies in Huh7 cells complement and extend the studies in AML-12 cells and together, provide strong evidence that TP can act directly on hepatocytes to maintain their resiliency during lipotoxic stress by boosting antioxidant defenses that inhibit lipotoxicity and senescence.

Replenishing TP restores mitochondrial fitness and rescues Smo-depleted hepatocytes from lipotoxicity and senescence
TP production is reduced in Smo KO hepatocytes (Figs.2B, C) and mice with Smo-depleted hepatocytes develop worse MASH when challenged with diets that promote hepatic lipotoxicity (Fig. 1).To determine if treating Smo KO hepatocytes with TP can protect them from lipotoxicity, we isolated primary hepatocytes from control and Smo KO mice and cultured the cells in OPA-enriched, serum-depleted medium that was supplemented with vehicle or recombinant TP protein for 48 hours.Results in the OPA cultures were compared to cells cultured in serum-depleted medium without OPA (Fig. 4A).As expected, culture in OPAenriched medium increased steatosis (Oil Red O) in both control and Smo KO hepatocytes and Smo KO cells were more steatotic than control cells under both culture conditions (Figs. 4B, C).Although numbers of hepatocytes were similar in Smo KO-and control-OPA cultures (Fig. 4D), only Smo KO cultures demonstrated enhanced accumulation of oxidized lipids (MDA) when exposed to OPA (Fig. 4E).Treating Smo KO cultures with TP completely protected them from the steatosis-inducing effects of OPA (Figs. 4B, C), prevented them from OPA-induced accumulation of MDA (Fig. 4E) and increased cell numbers by about 10% despite the serum-depleted culture conditions (Fig. 4D).Evidence that replenishing TP can rescue Smo-depleted hepatocytes from lipotoxicity indicates that TP operates downstream of Smo to protect hepatocytes from lipid-related stress and supports the concept that Smo mainly enhances TP stability, although more research is needed to validate this hypothesis.To determine how these Smo/TP-sensitive differences in lipotoxicity relate to differences in mitochondrial fitness, experiments were repeated and Seahorse analysis was done to compare mitochondrial oxygen consumption rates (OCR) (Fig. 4F).The OCR in Smo KO hepatocytes was lower than in controls basally.Although OPA suppressed these functions in both control and Smo KO hepatocytes, OPA-treated Smo KO hepatocytes also demonstrated the lowest OCRs when cultured in lipotoxic conditions.Importantly, replenishing TP partially protected Smo KO hepatocytes from OPA-induced mitochondrial dysfunction and substantially improved the OCR in Smo KO hepatocytes during exposure to lipotoxic stress (Fig. 4F).To verify and further clarify the effects of TP during Smo gene disruption, we transfected Huh7 cells with siRNA-Smo and cultured these cells in OPA-enriched medium for 4 days, adding either vehicle or recombinant TP protein for the final 2 days of culture (Fig. 4G).As noted in our earlier studies of Smo KO mice (Fig. 1) (16), disrupting the Smo gene in hepatocytes is sufficient to induce senescence basally and dramatically exacerbates senescence (B galactosidase activity) that is provoked by lipotoxic stress (induced here by culture in OPAenriched medium) (Fig. 4I).Remarkably, TP treatment completely protected siRNA-Smotreated Huh7 cells from OPA-induced senescence, as evidenced by reduced B galactosidase staining (Fig. 4I) and decreased expression of the cell cycle inhibitor, p21, on immunoblots (Fig. 4H).These changes in hepatocyte senescence were reciprocally related to changes in accumulation of HO1 and mitochondrial OXPHOS complexes (Fig. 4H), suggesting that TP treatment reversed depletion of the Smo-regulated factors that maintain mitochondrial fitness in metabolically stressed hepatocytes.Consistent with this concept, Seahorse analysis showed that Smo siRNA decreased the OCR in Huh7 cells and demonstrated that this parameter was more sensitive to OPA-suppression in Smo-depleted cultures than in Huh7 controls.Importantly, similar to Smo-KO primary hepatocytes (Fig. 4F), Smo-depleted Huh7 cells were rescued from the mito-inhibitory effects of OPA by TP treatment (Fig. 4J).

Inhibiting TP exacerbates MASH and reduces antioxidant defense and mitochondrial fitness in livers of CDA-HFD mice
We used adenoviral vectors to disrupt Smo in hepatocytes of Smo flox/flox mice and both viral vectors and gene manipulation can have off-target effects.Therefore, to assure that the responses we observed in Smo KO mice result from decreased abundance of TP, we fed wild type mice CDA-HFD for 6 weeks to induce MASH and treated them with vehicle or TPI during the final week of diet exposure (Fig. 5A).TPI had no effect on body weight but markedly increased liver weight and thus, liver/body weight ratio (Supple Figs.5A, B and Fig. 5B).
WT mice treated with TPI during the final week of CDA-HFD administration also demonstrated hyperglycemia, hyperinsulinemia and increased HOMA-IR at sacrifice (Supple Fig 5C and Fig. 5E).Thus, similar to Smo KO mice, TPI-treated WT mice developed systemic insulin resistance when fed CDA-HFD.Insulin resistance is thought to play an important role in MASLD pathogenesis.Consistent with this, hepatic steatosis was increased in TPI-treated mice as visualized on both H&E and Oil Red O stained liver sections (Fig. 5C), confirming other evidence which showed that inhibiting TP promotes steatosis in cultured hepatocytes (Supple Figs. 3, 4).TPI-treated mice also had higher serum levels of AST and ALT (Figs. 5D), consistent with results of hepatocyte culture studies which showed that TP directly regulates hepatocyte susceptibility to lipotoxicity (Supple Fig. 3).Lipotoxicity induces senescence in cultured hepatocytes and TP regulates this process (Supple Fig. 4).Further, senescent hepatocytes incite liver inflammation and fibrosis (35).Consistent with these facts, livers of TPI-treated mice accumulated more senescent cells (evidenced by greater B galactosidase activity and increased expression of p16 and p21) (Figs.5C, F).They also exhibited greater inflammation (F480 staining, Fig. 5C) and worse fibrosis, as evidenced by increased expression of fibrosis markers (Vimentin and Desmin) on immunoblots (Supple Fig. 5D) and more collagen fibrils on Sirius red-stained liver sections (Fig. 5C).Together, these results demonstrate that inhibiting TP exacerbates diet-induced MASH in mice by promoting liver lipotoxicity and senescence.
Senescence can be triggered by chronic mitochondrial damage and resultant oxidant stress.
TP is important for maintaining mitochondrial fitness and forms a complex with Nrf2 and HO1 (Fig. 2C), factors that have critical roles in antioxidant defense.Most importantly, our data reveal that Smo can physically interact with TP and its binding partners (Fig. 3C).Therefore, we analyzed liver samples from vehicle and TPI-treated CDA-HFD-fed mice to determine if/how TPI treatment impacts these relationships.We found that TPI treatment reduced both serum and liver levels of TP (Figs. 5G, H).Liver levels of Nrf2, HO1 and Smo proteins were also substantially reduced in the TPI-treated group (Fig. 5H), demonstrating that inhibiting TP recapitulates the liver phenotype of mice with Smo-depleted hepatocytes and supporting the concept that TP is needed to stabilize these proteins and maintain proper antioxidant defense.Indeed, TPI treatment depleted components of OXPHOS complexes (NDUFB8 and ATP5A) in mice fed MASH-inducing diets (Fig. 5I, Supple Figs. 5E, F).Importantly, levels of GDF15 protein were also increased in hepatocytes of TPI-treated mice (Supple Fig. 5G), suggesting that inhibiting TP induces ISRmt in hepatocytes leading to the exacerbation of MASLD.

Discussion
We recently discovered that adult livers rely upon the Hedgehog pathway to shield hepatocytes from lipotoxicity and suggested that this slows their biological aging by enabling metabolic flexibility that is necessary to maintain liver homeostasis and prevent MASLD (15,16,20).These concepts have not been widely accepted despite growing evidence that hepatocyte Hedgehog signaling broadly regulates metabolism and regenerative capacity in these cells (13,14).Thus, the current studies are important because they provide additional proof that the Hedgehog pathway critically controls the liver aging process and hence, susceptibility to MASLD.Further, we have now identified a Smo-dependent factor, TP, that has a major role in preventing hepatocyte aging and showed that this benefit accrues because TP promotes mitochondrial fitness.Although more research is needed to delineate the mechanisms involved, TP is known to facilitate repair of mitochondrial DNA and this reduces the risk for mitochondrial dysfunction and oxidative stress (24).We have shown that Smo also promotes co-incident accumulation of Nrf2 and HO1, two other proteins involved in antioxidant defense (30), and demonstrated that all four proteins directly interact.More research is needed to determine if (and how) these complexes control protein stability and/or localization but our data show that deleting Smo blocks nuclear localization of Nrf2 and reduces mitochondrial accumulation of TP and HO1, while inhibiting TP decreases Smo accumulation and recapitulates all the negative consequences of hepatocyte Smo depletion, including MASLD and insulin resistance.
The findings suggest a self-reenforcing mechanism whereby Hedgehog signaling promotes accumulation of TP and in turn, TP promotes accumulation of Smo to sustain Hedgehog signaling.Hedgehog signaling is coupled to primary cilia, a key nutrient sensing organelle, by bi-directional signaling: Hedgehog pathway components traffic on and off primary cilia to control Hedgehog signaling; conversely, ciliogenesis itself is regulated by Hedgehog pathway activity (36).As mentioned earlier, the significance of cilia (and by extension, Hedgehog signaling) in adult hepatocytes has been debated because cilia have been demonstrated in fewer than 10% of hepatocytes at a given point in time (12,37).However, more research is needed to clarify if/how cilia might mediate the striking phenotype that results when Smo is depleted in these cells.The issue is quite complicated for three main reasons.First, cilia are dynamic structures and they are much smaller in hepatocytes than cholangiocytes (12).Therefore, it may simply be easier to identify cholangiocyte cilia than hepatocyte cilia, leading to underestimation of ciliary abundance in hepatocytes.Second, hepatocytes and cholangiocytes themselves exhibit significant plasticity -each cell type can transition to become the other cell type and then revert back to its original phenotype (38).This confounds efforts to determine whether the seemingly small subpopulation of hepatocytes that express primary cilia at any moment in time are in the process of transitioning into cholangiocytes (or derive from cholangiocytes that have nearly become hepatocytes).Third, in either situation, cilia-linked Smo activity may persist (or emerge) even after (or before) the state transition is obvious.In any case, our data complement and extend growing evidence that the Hedgehog pathway is coupled to mitochondria, the main energy-producing organelle in cells (16).Our findings suggest that the Hedgehog pathway (i.e., Smo) links cilia to mitochondria and thus, couples nutrient sensing to energy production.This is a critical insight because defective nutrient sensing is a key driver of tissue degeneration related to aging and metabolic dysfunction (39), the major risk factors for MASLD and other tissue damage associated with obesity and type 2 diabetes (40).The model also helps to explain the pathobiology of inherited ciliopathies and mitochondrial diseases, both of which are characterized by dysregulated morphogenesis, progressive metabolic dysfunction, premature tissue degeneration and decreased longevity.
The Hedgehog pathway has long been known to promote mitochondrial fitness and increase longevity in model organisms (7).Our data show that TP is readily detected in the livers of healthy people and mice.Conversely, inherited deficiency of TP causes systemic mitochondrial dysfunction, i.e., MNGIE syndrome, and substantially shortens the life span of afflicted individuals (24).TP is a key enzyme in the thymidine salvage pathway and thus, TP deficiency compromises repair of mitochondrial DNA deletions that occur regularly in healthy cells as a consequence of routine metabolic activity.This leads to mitochondrial dysfunction that triggers the ISRmt.The ISRmt is a dynamic, and potentially progressive, process that aims to "right-size" the mitochondrial network either to meet the energy demands required to optimize cell viability, or assure elimination of terminally wounded cells to 'make room' for healthier replacements so that tissue integrity is restored and organismal life span is preserved (41).The intensity of the ISRmt is reflected by serum levels of mitokines and other factors released from cells with dysfunctional mitochondria -rising mitokine levels are indicative of an unsuccessful ISRmt and thus, persistent mitochondrial dysfunction.Hence, serum levels of the mitokine, GDF15, are used clinically as a predictor of morbidity and mortality in people with mitochondrial diseases (42).Early work showed that steatotic hepatocytes in MASLD livers have mitochondrial abnormalities (43).We discovered that GDF15 production is increased in Smo-deficient hepatocytes and up-regulated in livers of people with MASLD.
These results indicate that loss of Smo, and resultant decreases in TP, promote the ISRmt.Thus, the findings complement and extend earlier reports showing that both Smo-depleted senescing hepatocytes and GDF15 levels increase with age in people, and are particularly high in patients with advanced fibrosis related to MASLD (44).
Our comparison of conditioned medium from control and Smo-depleted hepatocytes demonstrate differences in GDF15, TP and multiple other proteins.The relative abundance of many, but not all, of these factors differed similarly in sera collected from the respective groups of mice, suggesting that hepatocytes were major sources of most of the proteins.However, some differentially abundant proteins in sera were not identified by the hepatocyte-conditioned medium analyses.While technical issues may have contributed to these discrepancies, it is also conceivable that the differences reflect changes in the secretomes of other types of cells in liver and extra-hepatic tissues that were caused by deleting Smo in hepatocytes.Defining mechanisms that orchestrate inter-cellular and inter-organ crosstalk to enable systemic adaptations to metabolic stress has become a focus of research that aims to identify diagnostic, prognostic and therapeutic targets for MASLD and other metabolic dysfunction associated diseases.More research is needed to unravel how Smo-dependent changes in hepatocyte mitochondrial stress fits into this complex pathobiology.Single cell and multi-omics analytical approaches will likely be needed to map and integrate time-sensitive changes in multiple targets because studies of GDF15 and TP exemplify the fact that stress-related factors have pleiotropic actions that are context-dependent (45).
Case series of patients with MNGIE syndrome provide the most direct evidence that loss of per se promotes progressive steatotic liver disease.A recent consensus conference of MNGIE experts reported 'hepatopathy' (i.e., 'liver steatosis evolving to cirrhosis') in 22% of MNGIE patients based on their meta-analysis of published literature, but this figure may underestimate the prevalence of liver dysfunction as the authors also noted that 'death is mainly due to GI and liver complications', including 'liver failure' (24).Further, because the liver is a major source of circulating TP, liver transplantation has been done to treat MNGIE patients, and these experts recommended liver replacement as a preferred treatment in selected MNGIE patients.Insulin resistance occurs in MNGIE syndrome and disrupting Tymp blocks insulin signaling in cultured adipocytes (46), suggesting that adipocyte-related systemic metabolic dysfunction contributes to MNGI-related hepatic steatosis.However, given that gastrointestinal dysmotility (a dominant clinical feature of MNGIE syndrome) promotes intestinal dysbiosis, gut liver axis pathobiology may also have a role in liver damage induced by TP deficiency.Interestingly, we have reported that deleting Smo in hepatocytes dysregulates bile acids and rapidly provokes intestinal dysbiosis in mice (20).Our current studies show that Smo promotes accumulation of TP and vice versa, raising the possibility that primary loss of either protein may trigger hepatocyte mitochondrial dysfunction which, in turn, dysregulates bile acid homeostasis and the gut-liver axis to drive progression of steatotic liver disease.Therefore, although we found no reports linking MNGIE syndrome with Hedgehog pathway inhibition, our data clearly demonstrate that deleting Smo in hepatocytes reduces hepatic production of TP and show that loss of TP activity recapitulates the negative effects of Smo deletion on hepatocyte mitochondria, antioxidant defense, oxidative damage, lipotoxicity and senescence.Since all these defects are reported to occur in other cell types when TP is inhibited (18) and Smo-deficient hepatocytes accumulate as MASLD progresses (16), the available data suggest that MASLD may be one facet of a MNGIE-like syndrome.This concept is supported by a recent study that used radio-labeled TP and positron emission tomography (PET) to localize TP in control mice and mice with diet-induced MASLD.TP mainly accumulated in the livers of control mice and hepatic TP content was substantially reduced in mice with MASLD (47).Thus, while seemingly heretical, the possibility that loss of Tymp/TP function promotes MASLD pathogenesis is a worthy topic for future research given acknowledged associations of MASLD with hepatic mitochondrial dysfunction, disordered gastrointestinal motility, sarcopenia, and neurodegenerative disorders (48).
Effect of TP on MASLD.To study the role of TP in MASLD progression, C57BL/6J male mice (The Jackson Laboratory) were fed CDA-HFD diet for 6 weeks and intraperitoneally injected with 1.7 mg/kg of tipiracil-hydrochloride, a TP inhibitor (TPI-HCl; n = 10) (MedChemExpress, Mounmouth Junction, NJ) or its vehicle (PBS; n = 10) three times per week during the final week before sacrifice to obtain blood and liver.
Assay details are provided in Supplemental Methods.

Cell culture studies
Smo and TP were manipulated in primary mouse hepatocytes and cell lines (AML-12, Huh7) to determine effects on senescence and lipotoxicity (see Supplemental Methods).

Human studies
Details provided in Supplemental Methods.

Statistics
Data were expressed as mean ± SEM.Statistical significance between two groups was evaluated using the student's t test, while comparisons of multiple groups were assessed by one-way analysis of variance (ANOVA), followed by Student-Newman-Keul's test.p ≤ 0.05 was statistically significant.

Study approval
We affirm that our research with human samples was conducted in accordance with the Declarations of Helsinki and Istanbul, and approved by Duke University Health System (DUHS) Institutional Review Board (Pro00005368), with written consent given from all subjects.Deidentified liver sections from patients with different stages of liver fibrosis were analyzed.Also, all animal studies were approved by the Duke University Institutional Animal Care (A200-21-09) and fulfilled National Institutes for Health and Duke University IACUC requirements for humane animal care.

Figure 2 .
Figure 2. Hepatocyte Smo-KO secretome is depleted of factors that promote antioxidant defense.Heat-map through proteome profiler analysis in (A) serum of Smo KO mice and (B) conditioned medium of oleic and palmitic acid treated Smo KO primary hepatocytes.n=3 per group.Thymidine phosphorylase (TP) protein expression by (C) immnunoblot (n=3 per group), (D and E) immunohistochemistry in mice total liver (n=5 controls, n=4 Smo KO), and (F) ELISA using mice serum (n=9 control mice; n=10 Smo KO mice).(G) Representative staining for TP, Nrf2, and HO1 and (H) the quantification of the positively stained area in MASLD patients versus healthy controls (n=3 individuals per group).P values were calculated using one-way ANOVA.Data are graphed as mean ± SEM. * p < 0.05.

Figure 3 .
Figure 3. Hepatocyte Smo-KO induces mitochondrial dysfunction.(A) Representative staining for Nrf2 and the quantification of the positively stained area in Smo KO compared to the control mice liver tissues (n=5 mice per group).(B) HO1 protein expressions in Smo KO versus to the control mice total liver by immunoblot (n=7 mice per group).(C) Interaction between TP, Smo, Nrf2, and HO1 in mouse primary hepatocytes by immunoprecipitation (IP).(D) Protein expressions of TP, HO1, OXPHOS complexes, succinate dehydrogenase complex flavoprotein subunit A (SDHA), pyruvate dehydrogenase (PDH), prohibitin 1 (PHB 1), heat shock protein 60 (HSP 60), voltage-dependent anion channel (VDAC), superoxide dismutase (SOD), and cytochrome c (Cytc) by immunoblot in isolated mitochondria from Smo KO mice total liver (n=7 mice per group).(E) Representative staining for PGC1a and the quantification of the positively stained area in Smo KO compared to the control mice liver tissues (n=9 control mice; n=10 Smo KO mice).(F) Protein expressions of apoptosis-inducing factor (AIF) and second mitochondria-derived activator of caspase (Smac) in mitochondria or cytoplasm from Smo KO and control mice total liver by immunoblot (n=7 mice per group).P values were calculated using one-way ANOVA.Data are graphed as mean ± SEM. * p < 0.05.(G) Hypothetical design.

Figure 4 .
Figure 4. Replenishing TP restores mitochondrial fitness and rescues Smo-depleted hepatocytes from lipotoxicity and senescence.(A) Experimental scheme.(B, C) Oil Red O staining and (D) Cell number of Smo KO primary hepatocytes by CCK-8 assay.(E) MDA concentration in conditioned media of Smo KO primary hepatocytes.(F) Oxygen consumption rate (OCR) using a Seahorse extracellular flux analyzer in Smo KO primary hepatocytes.(G) Experimental scheme.(H) Protein expression by immunoblot and corresponding morphometric quantification in siRNA-Smo transfected Huh7 cells.(I) B galactosidase staining and corresponding morphometric quantification in siRNA-Smo transfected Huh7 cells.(J) OCR of siRNA-Smo transfected Huh7 cells using a Seahorse extracellular flux analyzer.Data from triplicate experiments are graphed.P values were calculated using one-way ANOVA.Data are graphed as mean ± SEM. * p < 0.05.

Figure 5 .
Figure 5. Inhibiting TP exacerbates diet-induced MASH and liver fibrosis in wild type mice.(A) Wild type mice were fed with a CDA-HFD diet for 6 weeks.Mice were intraperitoneally injected with tipiracil-HCl (TPI) or its vehicle three times.(B) Liver and body weight ratio in TPI treated versus vehicle treated mice (n=10 mice per group).(C) Representative stainings for H&E, Oil Red O, Sirius Red, F480, B gal and corresponding densitometric analysis of positively stained areas.Serological results of hepatic function markers, (D) ALT, AST, and (E) HOMA-IR (n=10 mice per group).(F) Expression of senescence markers, p16 and p21, detected in TPI treated mice total liver by immunoblot (n=7 mice per group).(G) Concentration of TP in serum of vehicle or TPI treated mice (n=10 mice per group).(H) Protein expressions of TP, Smo, Nrf2, and HO1 detected in TPI treated mice total liver by immunoblot (n=7 mice per groups).(I) Expression of mitochondrial OXPHOS complex markers, NDUFB8 and ATP5A, detected in mitochondria from vehicle or TPI treated mice total liver by immunoblot (n=7 mice per group).P values were calculated using one-way ANOVA.Data are graphed as mean ± SEM. * p < 0.05.