miR-33 deletion in hepatocytes attenuates NAFLD-NASH-HCC progression

The complexity of the mechanisms underlying non-alcoholic fatty liver disease (NAFLD) progression remains a significant challenge for the development of effective therapeutics. miRNAs have shown great promise as regulators of biological processes and as therapeutic targets for complex diseases. Here, we study the role of hepatic miR-33, an important regulator of lipid metabolism, during the progression of NAFLD and the development of hepatocellular carcinoma (HCC). We report that miR-33 is elevated in the livers of humans and mice with NAFLD and that its deletion in hepatocytes (miR-33 HKO ) improves multiple aspects of the disease, including steatosis and inflammation, limiting the progression to non-alcoholic steatohepatitis (NASH), fibrosis and HCC. Mechanistically, hepatic miR-33 deletion reduces lipid synthesis and promotes mitochondrial fatty acid oxidation, reducing lipid burden. Additionally, absence of miR-33 alters the expression of several known miR-33 target genes involved in metabolism and results in improved mitochondrial function and reduced oxidative stress. The reduction in lipid accumulation and liver injury resulted in decreased YAP/TAZ pathway activation, which may be involved in the reduced HCC progression in HKO livers. Together, these results suggest suppressing hepatic miR-33 may be an effective therapeutic approach to temper the development of NAFLD, NASH, and HCC in

Despite the global health and economic burden associated with NAFLD/NASH, approved therapies are still lacking.Therefore, there is a pressing need to identify potential therapeutic options to halt the progression of the disease and its rapid growth (15,16).Several studies have associated NAFLD with multiple metabolic alterations (9,(17)(18)(19) and impaired mitochondrial function is one of the most prominent observed in NAFLD.Mitochondria are integral organelles for oxidative energy production, a process that encompasses numerous pathways including fatty acid ß-oxidation (FAO), tricarboxylic acid (TCA) cycle, electron transport chain (ETC) and adenosine triphosphate (ATP) generation.Mitochondrial dysfunction can vary depending on the stage of NAFLD, but in advanced disease it frequently includes alterations in mitochondrial number, mtDNA, mitochondrial biogenesis, mitochondrial dynamics, and mitochondrial recycling (18,(20)(21)(22)(23)(24)(25)(26).Coordinated regulation of these processes is essential to enhance mitochondrial activity without producing detrimental effects associated with excess mitochondrial-derived reactive oxygen species (ROS) formation and redox imbalance.Conversely, targeting de novo lipogenesis (DNL) has also arisen as a therapeutic option to mitigate NAFLD pathogenesis (27)(28)(29)(30).
MicroRNAs (miRNAs) have shown great promise as potential therapeutic targets for the treatment of metabolic disease, due to their ability to target numerous mRNAs and pathways simultaneously (31,32).Previous research from our group and others identified miR-33 as an intronic miRNA hosted within the sterol regulatory element-binding protein 2 (Srebf2) gene (33)(34)(35).miR-33 plays a pivotal role in metabolism through the regulation of mRNA transcripts involved in a wide variety of metabolic processes, including lipid and glucose metabolism (33)(34)(35)(36)(37)(38)(39)(40).Notably, miR-33 coordinates the expression of genes associated with mitochondrial function and homeostasis (37,41) and increased miR-33 levels in the liver (42) and serum (43) have been associated with NAFLD in humans.
Here, we elucidate a role of hepatocyte miR-33 in regulating obesity-driven NAFL-NASH-HCC progression.Genetic ablation of miR-33 in hepatocytes (HKO) improves metabolic function in the liver, enhancing glucose tolerance and insulin sensitivity and attenuating dyslipidemia, steatosis, and NASH.
These improvements contribute to long term reductions in liver injury and the development HCC.
Mechanistically, we found that hepatocyte deletion of miR-33 increases mitochondrial fatty acid oxidation (FAO) and alters mitochondrial dynamics, correlating with increased expression of miR-33 target genes, such as CPT1a, PGC1a and AMPKa.miR-33 regulation of AMPKa contributes to the regulation of a subset of downstream targets and pathways which have been recently implicated in NASH progression (44)(45)(46)(47)(48)(49).Additionally, attenuation of lipid and cholesterol accumulation in the liver reduces hepatic injury, protecting from NAFLD induced HCC development in the long term.Overall, this work indicates that the deletion of miR-33 in hepatocytes is sufficient to regulate several pathways altered throughout the development of NAFL/NASH/HCC, impeding the progression of the disease.

Loss of hepatic miR-33 improves glucose tolerance, insulin sensitivity and dyslipidemia during obesity-driven NAFLD.
To study the specific role of hepatic miR-33 in NAFL and its progression to NASH and HCC, we employed the conditional miR-33 knock-out murine model (miR-33 loxP/loxP ) bred with an Albumin-Cre to induce the deletion of miR-33 in hepatocytes (HKO) (50).WT and HKO littermates were then fed a choline-deficient, high-fat diet (CD-HFD) for 3, 6 and 15 months to induce simple steatosis/NAFL, steatohepatitis/NASH and HCC, respectively (Supplemental Figure 1A), as previously described (51).
To investigate the role of hepatic miR-33 in systemic metabolism and liver function during NAFLD progression, we analyzed miR-33 levels after 3 and 6 months on a CD-HFD.qPCR analysis of freshly isolated hepatocytes confirmed miR-33 deletion in HKO mice, while revealing increased miR-33 levels in diet-induced NAFL and NASH in control mice (Fig. 1A).These findings align with recent studies showing enhanced SREBP2 (the host gene of miR-33a) transcriptional activation in humans and other mouse models of NAFL (52).We further confirmed this observation by measuring SREBP2 and SREBP1 and miR-33a/b levels in core liver biopsies from obese non-steatotic (BMI; 36-61, NAS = 0), obese steatotic (BMI; 36-61, NAS = 1-2) and obese NASH (BMI; 36-61; NAS>5, fibrosis score = 1-2) patients.The results shown that both SREBP1 and SREBP2, as well as their intronic miRNAs (miR-33b and miR-33a, respectively), were elevated in obese steatosis and obese NASH subjects compared to obese healthy individuals (Supplemental Figure 1 and Fig. 1B).In agreement with the coordinated expression of SREBPs and miR-33a/b isoforms, a positive correlation was observed in their expression levels in the liver of those patients (Supplemental Figure 1C).As fatty liver and CD-HFD-induced NAFLD models have been associated with metabolic dysfunctions such as obesity, dyslipidemia and insulin resistance, we first sought to determine whether miR-33 deficiency in hepatocytes influenced obesity-driven NAFLD progression (51).To this end we analyzed the development of obesity, dyslipidemia and insulin resistance in WT and HKO mice after 3 and 6 months on a CD-HFD.While no changes in body weight were observed (Fig. 1C), the slight decreased in body fat accumulation in HKO mice (Fig. 1D) was gradually attenuated over the time, indicating no relevant changes in body weight or fat accumulation in our model.Circulating lipids, including total cholesterol and HDL-cholesterol were also reduced in HKO mice in NAFL and NASH while no changes were observed in circulating triglycerides (TAGs) (Fig. 1E-H).Finally, we assessed the regulation of glucose homeostasis and insulin sensitivity in WT and HKO mice by glucose and insulin tolerance tests (GTT and ITT).We found that HKO mice showed improved glucose metabolism after both 3 and 6 months on a CD-HFD (Fig. 1I-L).These results are consistent with our previous study showing improved systemic metabolism in HKO mice and reinforces the metabolic benefit of depleting miR-33 in hepatocytes, independent of the underlying dietary factors driving fatty liver progression (50).

Genetic ablation of miR-33 in hepatocytes reduces liver steatosis by enhancing FAO and decreasing fatty acid synthesis
Excess hepatic lipid accumulation results from the dysregulation of one or more pathways leading to an imbalance between lipid uptake, synthesis and oxidation (9).Our results showed a reduction in steatosis after feeding mice a CD-HFD for 3 and 6 months, which was confirmed by histological analysis, including H&E and Oil Red O staining (Fig. 2A and B).Additionally, liver/body weight ratio and TAG content in the livers were also reduced in HKO mice (Fig. 2 C and D).miR-33 is an important posttranscriptional regulator of numerous genes that participate in FAO (36,53); thus, we first sought to determine if the regulation of FAO was occurring in our model of NAFLD.Ex vivo analysis of the rate of [ 14 C]-palmitate oxidation showed increased liver FAO in HKO mice (Fig. 3A).We further characterized the contribution of miR-33 to mitochondrial metabolism by measuring the respiratory capacity of freshly isolated hepatocytes from CD-HFD fed WT and HKO mice, confirming the increase in mitochondrial respiration in hepatocytes lacking miR-33 (Fig. 3B and C).Mechanistically, we observed that carnitine O-octanoyltransferase (CROT) and the mitochondrial fatty acid (FA) transporter, CPT1a, both bona fide targets of miR-33 and key molecules that participate in FAO, were significantly upregulated in HKO livers (Fig. 3D and Supplemental Figure 2D).
Next, we aimed to determine whether hepatocyte miR-33 deficiency influenced DNL during NAFLD progression.To this end, we assessed the activities of fatty acid synthase (FASN) (the enzyme involved in the synthesis of FAs from acetyl-CoA and malonyl-CoA) and HMG-CoA reductase (HMGCR) (the rate-limiting enzyme for cholesterol synthesis) in freshly isolated liver homogenates from WT and HKO mice.The results showed decreased activity of both enzymes in HKO livers (Supplemental Figure 2A and B).Additionally, ex vivo measurement of DNL was assessed in WT and HKO livers, confirming a decreased rate of acetate incorporation into lipids in HKO livers after CD-HFD feeding (Supplemental Figure 2C).Consistently, we observed that HKO livers had increased Ser79 phosphorylation of Acetyl-CoA carboxylase (ACC) (Fig. 3D and Supplemental Figure 2D).The increased hepatic FAO and suppression of DNL observed in HKO mice correlated with a significant increase in AMPKa activation (as assessed by phosphorylation) (Fig. 3D and Supplemental Figure 2D).In contrast, the expression and phosphorylation of AMPKb was not altered (Figure 3D and L and Supplemental Figure 2D).Thus, the attenuation of NAFLD progression mediated by miR-33 deletion in hepatocytes is through the regulation of multiple metabolic pathways.
Given the profound metabolic alterations observed in miR-33 HKO livers, we next assessed global transcriptional changes by RNA-seq analysis in the livers of WT and HKO NAFL mice aiming to identify specific genes or upstream regulators involved in these functions.We found 1082 differentially expressed genes (DEGs) (421 up-regulated and 661 down-regulated in HKO, Padj.<0.05), indicating the broad effect that miR-33 deficiency has in the liver during steatosis initiation.Interestingly, genes involved in metabolic functions and pathways altered in obesity-driven NAFLD, revealed that gene signatures associated with FA uptake, FA synthesis and cholesterol homeostasis were altered in HKO livers (Fig. 4A and B).Of interest was the upregulation of Abca1 and Cyp7a1 observed in HKO livers, given their participation in cholesterol and bile acid (BA) metabolism and their direct regulation by miR-33 (33,54).Thus, we sought to determine whether hepatocyte deletion of miR-33 was playing a role in cholesterol and BA metabolism in the liver.ABCA1 and CYP7A1 upregulation in NAFLD HKO livers was confirmed by western blot analysis (Supplemental Fig. 4A).Moreover, RNA-seq analysis also revealed upregulation of other genes related to BA metabolism, including Abcb11, Cyp27a1, Abcg8, Abcg5, Atp11c and Atp8b1 (Fig. 4B).Given the observed regulation of BA metabolic genes in our model, we measured bile acid profiles in livers from WT and HKO mice fed a chow or a CD-HFD.In mice fed a chow diet, differences were only found in the levels of cholic acid (CA), with no other differences observed for other bile acid species or total bile acid content (Supplemental Figure 4B).By contrast, when mice were fed the CD-HFD, a more profound dysregulation of the hepatic bile acid profile was observed.Total BAs were increased in WT mice but maintained at standard basal levels in HKO livers (Supplemental Figure 3B).Additionally, deoxycholic acid (DCA), one of the most toxic BAs, were reduced in HKO livers compared to WT livers fed the CD-HFD (Supplemental Figure 3B).Analysis of different species related with liver pathologies, including 12aOH/NON-12aOH, unconjugated/conjugated and secondary/primary ratios, revealed different trends in the relative amounts of these BAs in WT livers under CD-HFD feeding but not in HKO livers (Fig. 4C), correlating with insulin resistance and liver injury (55)(56)(57)(58)(59)(60).Finally, we further interrogated our RNA-seq data for changes in well-known specific processes associated with NAFLD progression, including inflammatory, profibrogenic and CYP450 associated functions (61,62).
We observed downregulation of genes associated with inflammation and fibrogenesis in HKO livers, while repression of CYP expression was prevented (Fig. 4D).To gain better insight into the direct impact of miR-33 on liver gene expression signature, we sought to identify potential genes directly targeted by miR-33 in the liver in healthy and NAFLD conditions.To do this we assessed the overlap between genes upregulated in HKO livers from chow (50) and CD-HFD mice and the top 2000 genes predicted to be miR-33 targets by TargetScan7.2.While only 6 genes including Abca1, Ski and Atp11c were found in the intersection of all 3 conditions, more overlapping of genes (133) were found in the intersection between miR-33 targets and those upregulated in HKO under NAFLD than in chow diet fed mice (Fig. 4E and Supplemental Table 1).These findings highlight the potential direct role of several miR-33 targets, including genes known to regulate metabolism and liver injury, such as Abca1, Ski, Atp11c, Slc25a51, Atp8b1, Atp11a, Slc30a1, Irs2, Acadsb, Serpind1, Klf15 and Edem1, in the progression of the disease.
Moreover, these results suggest that most of the observed changes in the HKO mice are not a consequence of basal gene regulation mediated by miR-33 deletion, but to changes that occur during the progression of the disease.Overall, our analysis suggests miR-33 HKO mice are protected from NAFLD progression through the regulation of metabolic function at multiple levels, resulting in increased FAO and mitochondrial function and decreased DNL, cholesterol metabolism.

miR-33 HKO mice are protected from diet-induced NASH and fibrosis
The adverse outcomes associated with NASH and the subsequent fibrosis encompass the progression to cirrhosis and end-stage liver disease or HCC (9).Based on the improved metabolic function and the beneficial changes in the gene expression profile in HKO livers, we examined how miR-33 deletion specifically affects the development of liver fibrosis during NASH.Sirius Red staining of liver sections revealed a strong decrease in collagen content in HKO livers (Fig. 5A).Moreover, histopathological analysis of H&E-stained liver sections from Figure 2A revealed decreased macrovesicular fat content and hepatocyte ballooning in HKO livers (Fig. 5B).Consistently, we observed a significant reduction in liver fibrosis markers including Fibronectin (FN1) and Collagen type a1 (COL1a1), as well as total hydroxyproline content in HKO mice (Fig. 5C and D).Attenuation of liver fibrosis in the absence of hepatic miR-33 was not accompanied by significant reduction in liver inflammation (Supplemental Figure 4A-E).Reduction in liver injury in mice lacking miR-33 in hepatocytes was also confirmed by reduced serum levels of alanine aminotransferase (ALT) (Fig. 5E).
Together, our findings suggest that miR-33 deficiency in hepatocytes protects CD-HFD-fed mice from diet-induced liver injury and progression to fibrosis.

Loss of miR-33 regulates miR-33 target genes exclusively in hepatocytes and triggers metabolic changes and cellular cross communication in the liver.
In order to understand the potential regulatory effects of miR-33 regulation in the different cell populations in the liver during the advanced stages of the disease, we performed single cell RNAsequencing (scRNA-seq) in WT and HKO livers with NASH (Fig. 6A-C).UMAP analysis of liver cells showed that most evident changes between WT and HKO cells were found in the population identified as hepatocytes (Fig. 6C).Unbiased analysis of changes within hepatocytes identified downregulation of pathways related to obesity, protein translation, and unfolded protein response (UPR) in HKO hepatocytes, while pathways related to EIF2 signaling, inhibition of activin and activation of RXR were upregulated in these cells (Fig. 6D and E).Additionally, analysis of hepatocytes confirmed that alterations in metabolic and fibrogenic genes, as well as miR-33 bona fide target genes (Fig. 6F-I).Moreover, we analyzed different cell populations and interactions involved in liver fibrosis, aiming to understand how miR-33 deletion in hepatocytes could affect other cells.To monitor for hepatic stellate cell (HSC) activation, the genetic profile of HSCs was analyzed based on different characteristics known to contribute to their activation and liver fibrosis (63).HSCs were classified and named according to their activated phenotype and pro-or anti-fibrogenic function as follows: activated myofibrolast HSC (Myo) (pro-fibrogenic), cytokine-producing HSC (Cy&Gr) (anti-fibrogenic), classically activated (Activ) (profibrogenic), quiescent (Quies) (anti-fibrogenic), and apoptotic HSC (Death) (anti-fibrogenic) (63).Myo and Activ markers were increased in WT HSCs, whereas markers of Cy&Gr, Quies, and Death of HSC were higher in HSCs derived from HKO livers, indicating reduced activation of HSCs in this group (Fig. 7A).On the other hand, although anti-inflammatory markers of macrophages showed a trend towards to increased activation in HKO livers, analysis of inflammatory markers aligned with our previous data showing no clear regulation in HKO livers (Fig. 7B).Finally, we analyzed different cell types from our scRNA-seq looking for potential changes in miR-33 target genes that could suggest an existing communication between different cell types and an indirect gene regulation from hepatocyte-derived miR-33.With that purpose, bona fide targets of miR-33 (Abca1, Crot and Cpt1a) were analyzed in different cell types, including HSC, endostellate cells, macrophages and cholangiocytes.While downregulation of some of those genes were observed in some of the groups, the clear lack of upregulation of these bona fide miR-33 target genes in the HKO group, suggests no direct effect of miR-33 deletion in these cells when miR-33 was removed from the hepatocytes (Fig. 7C-F).We also performed cell-cell communication analysis to get an overall picture of the different interactions occurring in WT and HKO livers.Cell-cell communication analysis showed increased total number and strength of cell-cell communication in WT livers compared to HKO livers (Supplemental Figure 5 A-C), with hepatocytes being the main cell type contributing to cell communication in the liver (Supplemental Figure 5C).Cell-cell communication analysis also revealed that the most relevant differences in WT vs. HKO livers were found in pathways related to leukocyte activation and markers of liver fibrosis mostly derived from the hepatocytes (Supplemental Figure 5D and E (Supplemental 5D and E).Finally, we attempted to analyze specific ligand-receptor differences in WT vs. HKO cell-cell communication from hepatocytes to other liver cells, however no differences in terms of signaling were found.Taken together, our results indicate that the main effect produced by miR-33 deletion in hepatocytes is specific to these cells and the effect observed in recruitment and activation of inflammatory and hepatic stellate cells are a consequence of changes produced initially in hepatocytes.

miR-33 deficiency in hepatocytes prevents mitochondrial dysfunction associated with NAFLD/NASH progression.
Mitochondrial dysfunction is a common feature underlying NAFLD-NASH progression (18,(20)(21)(22)(23)(24)(25)(26).Previous work from our group and others identified miR-33 as a critical regulator of mitochondrial function through the targeting of multiple factors, including PGC1a and AMPKa (40,41,71).Thus, we next aimed to further characterize the molecular mechanisms that mediate the improved mitochondrial function observed in miR-33 deficient hepatocytes.We found increased mitochondrial content in hepatocytes from HKO NASH livers, as assessed by the protein levels of different complexes of the electron transport chain (ETC) and by the mitochondrial to nuclear DNA ratio (mtDNA/nDNA) (Fig. 8A    and B).These findings were further supported by electron microscopy analysis of hepatocytes from NASH mice, which revealed an increase in the coverage and density of mitochondria, as well as mitochondrial elongation in HKO mice (Fig. 8D-H).We also observed enhanced mitochondrial ETC activity of Complex I and Complex II (Fig. 8C).The increase in mitochondrial mass found in hepatocytes from HKO mice was correlated with elevated levels of PGC1α, which is a direct target of miR-33 (72,73).Moreover, its downstream target, TFAM, was also upregulated in livers from miR-33-deficient hepatocytes (Fig. 8I).Together, these results demonstrate that absence of miR-33 in hepatocytes improves mitochondrial function by increasing mitochondrial mass and ETC activity.
Mitochondrial homeostasis is critical for the control of mitochondrial health and metabolism (18,24,74).Mitochondrial quality control mechanisms include mitochondrial biogenesis and dynamics, a process that involves fusion and fission of mitochondrial membranes and is dysregulated in NAFLD.
Mitochondrial number and size are also controlled through the balance of mitochondrial dynamics (22,24,74,75).Thus, we sought to characterize mitochondrial dynamics in our NASH model.We found an increase in fusion related proteins MFN2 and OPA1, but no relevant changes in fission proteins (Fig. 8I and Supplemental Figure 6A).Importantly, the increased MFN2 levels is consistent with the changes in mitochondrial shape observed by EM and correlates with the increased respiratory capacity of these mice.
Lipid overload and excessive mitochondrial activity have been linked with mitochondrial dysfunction in NAFLD.Besides the inability to sustain metabolic needs, mitochondrial dysfunction is responsible for the production of large amounts of reactive oxygen species (ROS), which increases mitochondrial damage and can eventually lead to cell death (76).Although the increased mitochondrial number and activity in HKO mice could lead to higher ROS production and damage, changes in mitochondrial dynamics can also play a role in ROS regulation, membrane potential and other downstream processes related to mitochondrial stress (24,74).To determine whether miR-33 levels in hepatocytes influence ROS production in obesity-driven NAFLD/NASH, we monitored ROS accumulation in liver sections by dihydroethidium (DHE) and observed a decrease in HKO mice (Fig. 9A).Liver lipid peroxidation measured by assessing malondialdehyde (MDA) as a readout of ROS damage also showed a similar decrease in livers from HKO mice (Fig. 9B).Although no major changes were found in the oxidized or reduced forms of glutathione, or their ratio, (Supplemental Figure 6B), an increase in glutathione-reductase activity was found in HKO livers, suggesting that changes in the recycling rather than the synthesis of glutathione may contribute to reduced oxidative stress in these livers (Fig. 9C).
Finally, oxidative stress markers including 4-Hydroxynonenal (4-HNE), and OxyBlot, confirmed increased levels of oxidative stress in WT livers during NASH (Fig. 9D and E).Considering the close link between mitochondrial dynamics, lipid overload and ER stress, we interrogated HKO NASH livers for changes in ER stress response.However, only partial regulation of ATF4 and PERK was found in those livers, indicating a limited involvement of miR-33 in the regulation of ER-stress (Supplemental Figure 6C).
Lastly, the ultimate cellular consequence of mitochondrial dysfunction and oxidative stress, the induction of cell death, was also attenuated in the HKO mice under NASH conditions, as seen by caspase3/6 activity and TUNEL staining (Supplemental Figure 6D-G).Notably, analysis of gene expression changes related to decreased hepatocyte cell death and increased mitochondrial function were also observed in HKO hepatocytes identified from scRNA-sequencing (Supplemental Figure 6H).Together, these findings indicate that miR-33 deficiency in hepatocytes improves mitochondrial quality control by enhancing mitochondrial biogenesis and mitochondrial dynamics to sustain high rates of oxidative metabolism without increasing mitochondrial injury and oxidative stress during lipid overload, thereby protecting against hepatocyte cell death.The activation of mitochondrial biogenesis and dynamics in miR-33 deficient hepatocytes through upregulation of PGC1a and AMPKa is in line with previous studies identifying these two genes as direct targets of miR-33.
AMPK signaling pathway is increased in miR-33 HKO livers.
AMPK is a master regulator of metabolism and mitochondrial homeostasis (77).Our previous results showed that AMPK activation is increased in HKO mice compared to WT mice in both NAFL and NASH stages, counteracting the progressive decrease otherwise evident in NAFLD (44).These results prompted us to characterize additional posttranscriptional mechanisms regulating AMPK in our model.Notably, we found that the activation of liver kinase B1 (LKB1), a kinase that controls AMPK activity, was enhanced as shown by the increased phosphorylation of LKB1 at serine 428 in HKO livers (Fig. 10A).
LKB1 activation is regulated by its subcellular compartmentalization through deacetylation and phosphorylation (78)(79)(80), correlating with increased levels of sirtuins, a family of histone and protein deacetylases.In accordance, we found increased levels of SIRT1, SIRT2, SIRT3, SIRT7 and a trend toward upregulation of SIRT6 in HKO livers (Fig. 10B).Sirtuin activity is dependent not only on expression levels but also on the availability of NAD + .We found that total NAD and NAD + as well as the NAD + /NADH ratio were increased in HKO livers (Fig. 10C and Supplemental Figure 7A).These results point towards the increased activation of upstream regulators of AMPKa.As previously shown in Figure 2, we found increased FAO and decreased FAs, with increased AMPKa/ACC phosphorylation indicating a broad rewiring of metabolism mediated by AMPK signaling in HKO livers.Consistent with our observations on the effects of AMPK, we found that phosphorylation of ULK1, a downstream target of AMPKa, was also increased in HKO livers, along with the increased levels of LC3bII and ATG5, suggesting a role of AMPKa in the promotion of increased autophagy in HKO livers (Fig. 10D).Likewise, HKO livers displayed increased phosphorylation of Caspase 6, which has been described to be regulated by AMPK, consistent with reduced hepatocyte cell death in NASH (44) (Supplemental Figure 6F).

miR-33 deficiency in hepatocytes reduces NAFLD progression to HCC.
To analyze whether the improved metabolic function and protection against NAFL-NASH progression attenuates the development of HCC in HKO mice, we fed WT and HKO mice a CD-HFD for 15 months.While 75% of WT mice were found to develop tumors, only 40% of the HKO mice did (Fig. 11A).Tumor quantification revealed a decrease in the average tumor number per mouse in HKO mice (Fig. 11B), particularly pronounced for large tumors (volume >20mm 3 ) (Fig. 11C).In agreement with reduced tumor incidence, serum levels of a-fetoprotein (AFP) were significantly reduced in HKO mice compared to WT mice (Fig. 11D).Histological analysis of WT and HKO tumors also revealed a decrease in proliferative Ki67 positive cells in tumors from HKO mice compared to WT mice (Fig. 11E).Analysis of miR-33 expression by qPCR in livers fed the CD-HFD compared to a standard chow diet for 15 months showed a modest increase in miR-33 expression in NASH, which was not statistically significant (Supplemental Figure 8A).Additionally, increased expression of miR-33 was detected in tumor samples (T) when directly compared with adjacent healthy liver tissue (NT) from the same mice in WT samples (Supplemental Figure 8B), suggesting a potential role of miR-33 directly within the tumor microenvironment.Moreover, miR-33 levels correlated with increased levels of Ki67, a proliferation marker (Supplemental Figure 8B-E).
Recent studies have highlighted the role of cholesterol in the activation of the gene regulator TAZ and the importance of this for the severity of NASH and progression to HCC (45,46,49,81,82).YAP/TAZ are transcriptional coactivators of the Hippo pathway that participate in the initiation and progression of different cancers (82)(83)(84).Specifically, TAZ levels in HCC have been associated with its initiation and prognosis (49,81,85).TAZ upregulation in NASH and HCC has been associated with both increased cholesterol levels and decreased AMPKa activity and it has been described to participate in the transcriptional regulation of several genes involved in fibrosis, proliferation, superoxide formation and regulation of metabolism.Furthermore, its upregulation has been described in the pre-tumor NASH stage (49,81,85).Given the role of mIR-33 as a key regulatory molecule that controls cholesterol efflux through the targeting of ABCA1 and ABCG1, we hypothesized that the absence of miR-33 in hepatocytes would decrease cholesterol content, potentially attenuating YAP/TAZ activation.Upregulation of TAZ levels in NASH livers that was partially abrogated in mice lacking miR-33 in hepatocytes (Fig. 12A and Supplemental Figure 8F).We further confirmed the activation of TAZ in NASH livers by cellular fractionation and immunoblotting for its nuclear localization (Fig. 12B).In line with this, expression of downstream TAZ target genes was decreased in HKO livers (Fig. 12C).In line with the role of miR-33 in the regulation of cholesterol homeostasis, a decrease in total and free cholesterol content was found the livers from HKO mice compared to WT mice fed with a CD-HFD (Fig. 12D).To strengthen the potential link between liver cholesterol and miR-33 levels with YAP/TAZ activation we cultured AML12 mouse hepatocytes in the absence or presence of LDL (120µg/ml) and analyzed changes in YAP/TAZ signaling.
Analysis of the regulation of YAP/TAZ protein levels confirmed that cholesterol loading of hepatocytes is sufficient to induce YAP/TAZ stability (Supplemental Figure 8H-G).Thus, we aimed to determine whether the regulation of miR-33 in hepatocytes modulates YAP/TAZ activation in response to cholesterol-LDL.Overexpression of miR-33 in AML12 cells resulted in increased cholesterol content when cells were loaded with LDL (120µg/ml) (Supplemental Figure 8I and J).However, we did not observe differences in cholesterol sensing genes (Supplemental Figure 8K).In line with the increased cellular cholesterol content, YAP/TAZ activation was increased in response to miR-33 overexpression, in accordance with the increase cellular content of cholesterol (Supplemental Figure 8L).Overall, these results suggest that miR-33 regulates YAP/TAZ activation in hepatocytes through the regulation of cholesterol content, linking cholesterol accumulation within the cells with activation of proliferative pathways associated with hepatocellular carcinoma.
Taken together, our results suggest miR-33 deletion in hepatocytes improves mitochondrial metabolic function, restraining NAFLD/NASH progression and in the long-term, preventing the development of HCC (Fig. 13).

DISSCUSION
The rise in NAFLD associated with overnutrition has reached epidemic levels, but its complexity has hindered the development of effective treatments.This study shows that miR-33 is increased in hepatocytes at different stages of NAFLD in human patients and mouse models.Deletion of miR-33 in hepatocytes improves liver function, reducing lipid accumulation progression of the disease.Numerous studies have demonstrated roles of miR-33 in metabolism, including cholesterol homeostasis, triglyceride metabolism, autophagy, glucose metabolism and mitochondrial function (33)(34)(35)(36)(37)(38)(39)(40)(41), showing that inhibiting miR-33 reduces atherosclerosis in mice and non-human primates (37,38,(86)(87)(88)(89)(90)(91).However, due to the promiscuous nature of miRNAs, the whole-body deficiency of miR-33 has been associated with obesity, dyslipidemia and insulin resistance (92,93).These detrimental effects prompted us to study the role of miR-33 in other metabolic diseases, such as NAFLD.Recent strategies have emerged to overcome potential undesired effects of miRNA therapies, shedding light on the cell-specific functions of miR-33 and its therapeutic potential.Some of these studies, have demonstrated the efficiency of delivering miR-33 inhibitors inside pH low insertion peptides (pHLIP) to the kidney and atherosclerotic lesions (94,95).
Additionally, in a recent study, using a different strategy, our group also demonstrated the safety and efficiency of specifically removing miR-33 from hepatocytes to improve cholesterol and FA metabolism, underscoring the role of hepatic miR-33 in liver metabolism and fibrosis (50).This previous study not only showed that miR-33 suppression in hepatocytes was not responsible for the adverse metabolic effects observed in whole-body deficient mice, but that liver specific loss of miR-33 improved whole body metabolism under hyperlipidemic conditions (50).Building upon this model, our current work has focused into the metabolic advantages associated with miR-33 deficiency in hepatocytes, during NAFLD/NASH/HCC development to investigate the long-term alterations in liver function under this chronic inflammatory disease.
The initial characterization of miR-33 HKO mice in this CD-HFD model showed a clear impact on the regulation of cholesterol and glucose metabolism, with minimal effects on body weight, consistent with previous results found for HKO mice on other diets (50).The reduction in steatosis was associated with the modulation of several pathways involved in liver lipid accumulation, including FA uptake, DNL and FAO.Interestingly, metabolic benefits observed in HKO mice were already observed after feeding the mice for 3 months with the CD-HFD and mostly sustained over the time, suggesting that early protection can prevent from the progression of the disease.Taken together, our data suggest that the benefits resulting from miR-33 deletion in hepatocytes are apparent in the context of diet induced obesity and NAFLD.Minimal changes were observed in mice fed a standard chow diet, further suggesting that the improvements in liver metabolism that alleviate lipid buildup are primarily responsible for ameliorating liver injury and progression of the disease.Indeed, most predicted miR-33 targets upregulated in HKO livers were associated with known metabolic genes and functions.Additionally, our results suggest that miR-33 deletion in hepatocytes could affect BA metabolism and homeostasis.Although the lack of major changes in BA content in chow fed WT and HKO livers was surprising given the role of miR-33 in regulating CYP7A1 and other bile acid transporters, the changes observed in NASH livers were in line with miR-33 playing a prominent role in the regulation of BA and cholesterol metabolism (54,60).
One of the most common features in NAFLD is the inability to sustain mitochondrial adaptation to nutrient status (18,(20)(21)(22)(23)(24)(25)(26)75).Mitochondria are highly dynamic organelles with the ability to undergo functional and structural changes in response to environment and energy requirements (74).However, in NAFLD, as with other metabolic diseases, increased FAO rates to counteract lipid accumulation are thought to result in increased oxidative stress and ER stress, resulting in mitochondrial injury, and impaired oxidative phosphorylation (18,26,75,96).The analysis of hepatic mitochondria from WT and HKO mice fed a CD-HFD for 6 months showed that miR-33 deficiency was associated with significant changes in mitochondrial quantity and morphology, suggesting a broader role of miR-33 in metabolism beyond regulation of cholesterol and FAO.The observed changes in mitochondria suggest miR-33 HKO mice have increased mitochondrial biogenesis and mitochondrial dynamics, mechanisms directed by PGC1a, AMPKa and MFN2 among other markers, a phenotype associated with enhanced oxidative capacity coupled with reduced ROS production, resulting in protection from liver injury (24)(25)(26)74).The approaches used here to study mitochondrial turnover and dynamics suggest both a positive regulation of mitochondrial biogenesis and mitochondrial fusion; however, as these processes are related, we cannot exclude the two events being interdependent.Several studies have described the direct role of miR-33 in repressing mitochondrial biogenesis and FAO through the targeting of PGC1a and AMPK, which are known to participate both in mitochondrial biogenesis and regulation of fusion/fission processes, suggesting that miR-33 could directly regulate these functions in our model through the downregulation of these two targets.Despite employing different approaches to elucidate potential miR-pathways, no direct evidence of miR-33 action was found, suggesting that oxidative stress reduction is a consequence of improved mitochondrial biogenesis and dynamics.
The deficiency of miR-33 in hepatocytes was sufficient to protect from HCC carcinoma development in our diet model.Previous studies aiming to determine the potential oncogenic role of miR-33 in hepatocellular carcinoma and cell proliferation have reported conflicting results, ranging from a role in inhibiting cell proliferation to anti-oncogenic roles of miR-33 in HCC cell lines.Similarly inconsistent results were reported in murine models and human samples of HCC, muddling the conclusions about the role of miR-33 in HCC (97-100).While the higher levels of miR-33 detected in tumor samples compared to adjacent healthy tissue and their correlation with Ki67 suggest that miR-33 could be directly involved in promoting tumor progression, the direct role of miR-33 within these tumors should be the subject of further research.Finally, our results align with previous studies showing that YAP/TAZ activation in hepatocytes plays a central role in liver fibrosis and the transition to hepatocellular carcinoma in response to increased cellular cholesterol levels (46).Given the role of miR-33 in cholesterol regulation, we hypothesized that miR-33 could also be contributing to attenuation of HCC development through this mechanism.Cell culture experiments in AML12 hepatocytes indicate the involvement of miR-33 and cholesterol accumulation in activation of YAP/TAZ pathway in hepatocytes.However, given the multiple adaptations regulated in HKO mice, the decrease observed in YAP/TAZ activation could also be a consequence of AMPKa phosphorylation.
This study contributes to a deeper understanding of the mechanisms involved in NAFLD/NASH progression and the potential applicability of miR-33 therapeutic approaches for its treatment.The beneficial or detrimental of enhanced FAO in NAFLD/NASH has been extensively discussed in the past, and our present results bring insight into the beneficial role of FAO in the disease (15,25,101,102).Our findings indicate that the regulation of hepatocyte metabolism by miR-33 is involved in the progression of NAFLD/NASH, as well as NAFLD/NASH-derived HCC.The direct deletion of miR-33 in hepatocytes protects from the progression of this disease.This may be particularly relevant for the use of approaches such as N-acetylgalactosamine-conjugated antisense oligonucleotides, which have been demonstrated to effectively promote targeted delivery of inhibitors to the liver (103).In the context of human pathology, it is important to note, that while mice possess only the miR-33a isoform of miR-33, humans express both miR-33a and miR-33b isoforms, encoded within the SREBF2 and SREBF1 genes, respectively, which are regulated by different mechanisms (104,105).Recent observations of the transcriptional activity of both SREBF1 and SREBF2 in both murine and human NAFLD (52) suggest that in humans, miR-33b may also contribute to the development of the disease, supporting the therapeutic potential of targeting hepatic miR-33 in human pathology.

STUDY LIMITATIONS
This work demonstrates that the depletion of miR-33 in the hepatocytes is sufficient to increase hepatic metabolic activity and reduce lipid accumulation in the liver protecting from NAFLD and NASH.However, in this study we did not determine whether a pharmacological approach to revert the progression of the disease would have a similar impact.Future studies should investigate whether pharmacological approaches can replicate the metabolic benefits observed in this study by selectively inhibiting miR-33 in hepatocytes.Although scRNA-seq data did not reveal major alterations in miR-33 target genes in cell types other than hepatocytes, it will be important to confirm if the Albumin Cre used in this study might affect miR-33 expression in liver cholangiocytes.While the study focused on changes in described and predicted miR-33 target genes, future investigations directly should assess the interaction of miR-33 with mRNA in hepatocytes under NAFLD conditions to better understand the precise mechanisms involved.
Finally, this study suggests that the protection from HCC development is mainly mediated through the steady long-term effect of miR-33 deletion in hepatocytes.However, based on our data, we cannot discard a potential intrinsic role of miR-33 within the tumor microenvironment, promoting tumor growth and proliferation.Further exploration is needed to elucidate the exact role of miR-33 within the tumor in the context of obesity-induced HCC.By addressing these avenues, future research can deepen our understanding of the intricate roles of miR-33 in liver metabolism, and potential therapeutic interventions for NAFLD, NASH, and HCC.These investigations will not only advance our knowledge of molecular mechanisms but also pave the way for targeted, precision medicine approaches in the clinical management of these complex liver diseases.

Figure 4 .
Figure 4. RNA-seq in NAFL livers reveals global changes in gene expression regulated by miR-33.(A, B and D) Heatmaps of pathways relevant to NAFLD progression in livers from WT and HKO mice.Cutoff values were settled as Fold change > Log21.5 and Padj < 0.05.(n=4) (C) Comparison of liver bile acids and bile acid distribution of the 12α-hydroxy/non-12α-hydroxy ratio in WT and HKO livers.(n=4-5).(E) Venn diagram depicting the overlap between miR-33 predicted targets and genes upregulated in HKO livers vs. WT livers in mice fed a chow and CD-HFD.Data represent the mean ± SEM (*P≤0.05,**P≤0.01,***P≤0.001 compared with WT animals, unpaired Student's t test for 2 group comparisons and 2-way ANOVA followed by multiple comparison.

Figure 6 .
Figure 6.Hepatic loss of miR-33 attenuates liver injury through the crosstalk of different cell types.UMAP (A) and heatmap (B) representation of cell clusters identified from sc-RNA-seq analysis.(C) UMAP projection of single cell profiles from WT (gray) and HKO (red) mice hepatocytes identified from sc-RNA-seq analysis.(D) Canonical pathways represented by Z score among differentially expressed genes in scRNA-seq analysis of hepatocytes from WT and HKO mice.Red bars indicate pathways in which genes are upregulated in HKO, and gray indicates downregulated pathways on the predicted Z score.All represented pathways were significantly changed with a -log P value > 1.5.(E-I) Violin plots representing the top upregulated and downregulated genes significantly altered in the indicated pathways in hepatocytes.

Figure 7 .
Figure 7. miR-33 deficiency attenuates liver fibrosis without regulating miR-33 target genes in non-hepatocytes liver cells.(A, B) Heatmap showing hepatic stellate cell activation (A) and macrophage inflammatory/non-inflammatory markers (B) identified from sc-RNA-seq from WT and HKO mice livers.Color codes referred to Z score.(C-F) Violin plot showing expression changes of miR-33 target genes in nonhepatocyte cells, including HSC (C), endostellate cells (D), macrophages (E) and cholangiocytes (F).

Figure 8 .
Figure 8. Hepatic miR-33 deficiency improves mitochondrial function and homeostasis.(A) Western blot and densitometric analysis of different mitochondrial subunits blotted with the Total OXPHOS Rodent WB Antibody Cocktail (ab110413) and housekeeping standard VINCULIN in WT and hepatocyte specific miR-33 knockout (HKO) livers from mice fed with CD-HFD for 6 months (n=6).(B) qPCR analysis of mitochondrial DNA and nuclear DNA in WT and HKO livers.Data represented as mtDNA / nDNA (n=6).(C) Activity of the ETC Complex I and Complex II in NASH livers.Enzyme activities are expressed as change in absorbance / minute / µg protein / CS activity (n=4-6).(D) Representative electron micrographs of mitochondria profiles in WT and HKO hepatocytes from NASH livers.(E-H) Mitochondrial coverage (E), mitochondrial density (F) and cumulative distribution and mean of mitochondrial area (G) and mitochondria aspect ratio (H) from WT and HKO hepatocytes (n=3-4).(I) Western blot of PGC1a and TFAM; MFN2, MFN1 and housekeeping standard VINCULIN or GAPDH in WT and HKO livers.Data represent the mean ± SEM [*P≤0.05,**P≤0.01,***P≤0.001 compared with WT animals, unpaired Student's t test for 2 group comparisons and 2-way ANOVA followed by multiple comparison (B, C, I)].

Figure 11 .
Figure 11.Hepatic miR-33 deficiency reduces diet-induced tumor incidence.(A) Representative images of WT and hepatocyte specific miR-33 knockout (HKO) livers after 15 months of CD-HFD, dashed line used to outline tumors and relative number of mice with and without tumor.(B-D) Graphical representation of (B) total number of tumors/mouse and (C) number of tumors larger >20mm 3 /mouse.(D) Circulating AFP levels (left panel) (n=13-12).(E) Representative images of Ki67 staining in liver tumors from WT and HKO mice after 15months of CD-HFD.Indicated quantification on the right (n=3).Data represent the mean ± SEM. *P≤0.05,**P≤0.01,***P≤0.001 compared with WT animals, unpaired Student's t test for 2 group comparisons and 2-way ANOVA followed by multiple comparison.