Hepatic Notch Signaling Correlates With Insulin Resistance and Nonalcoholic Fatty Liver Disease

The study conforms to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the institutional review board and ethical committee of the Fondazione Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) Ca’ Granda. Each subject gave written informed consent. Demographic and anthropometric features, arterial blood pressure, medical history, and medications were recorded for all patients and are summarized in Table 1. A needle liver biopsy was performed in all patients. The specimen was formalin-preserved for histological and immunofluorescence analysis. Part of the sample was included in RNAlater (Ambion, Carlsbad, CA), immediately frozen in liquid nitrogen, and stored at −80°C for RNA analysis.

We recruited 44 of 48 consecutive patients who underwent bariatric surgery at the IRCCS Ca’ Granda Ospedale Policlinico di Milano, between 2006 and 2008. Indications for bariatric surgery included BMI >40 kg/m² or BMI >35 kg/m² in the presence of metabolic complications (T2D, uncontrolled hypertension, severe dyslipidemia, obstructive sleep apnea). We excluded subjects with alcohol consumption >30 g/day for men and 20 g/day for women (n = 1) and chronic viral hepatitis (n = 3). Fasting glucose, HDL and total cholesterol, triglycerides, alanine aminotransferase (ALT), and aspartate aminotransferase (AST) were assessed the day of surgery, and needle liver biopsy (16-gauge) was performed during the bariatric surgery. Insulin-resistance status was classified according to fasting glucose levels and oral glucose tolerance test results. These patients are part of a previously reported cohort in which we characterized insulin-dependent signaling and the regulation of lipid metabolism according to liver histology (27).

We recruited 38 unrelated patients followed up at the Metabolic Liver Diseases outpatient service, Fondazione IRCCS Ca’ Granda, who underwent percutaneous liver biopsy because of suspected NAFLD due to persistently abnormal liver enzymes/serum ferritin test results or a history of steatosis associated with severe metabolic abnormalities, between January 2011 and January 2012. Other causes of liver disease were excluded, including increased alcohol intake (>30 g/day for men or 20 g/day for women), viral and autoimmune hepatitis, hereditary hemochromatosis, and α1-antitrypsin deficiency. Fasting glucose and insulin levels, HDL and total cholesterol, triglycerides, and ALT and AST levels were assessed the day of the biopsy. Patients were classified as insulin-sensitive based on homeostasis model assessment–insulin resistance (HOMA-IR) <2.5 ([fasting insulin (μU/mL) × fasting glucose (mmol/L)]/22.5) (28) without a history of impaired glucose tolerance or impaired fasting glucose, insulin resistance (HOMA-IR >2.5, or impaired fasting glucose or impaired glucose tolerance, but no diagnosis of diabetes) or T2D.

A single expert pathologist, unaware of gene expression and immunofluorescence data, evaluated all biopsy specimens according to Kleiner et al. (29), based on the determination of the NAFLD activity score (NAS) as the result of the sum of steatosis severity (0–3), intralobular necroinflammation (0–3), and hepatocellular ballooning (0–2). Steatosis percentage was determined in at least 10 hepatic lobules per patient. Subjects were classified in three groups according to liver histology: histologically normal liver, SS, and NASH.

Paraffinized sections were deparaffinized, rehydrated, and stained as previously described (30). To determine Notch signal activation, tissues were stained with Hey1 (#5714, 1:100 dilution) or HeyL (#10094, 1:150 dilution) antibodies from Millipore and detected with donkey anti-rabbit Alexa Fluor 488 (1:1,000 dilution) or donkey anti-mouse Alexa Fluor 594 (1:1,000) from Invitrogen. Slides were mounted with Vectashield with DAPI (Vector Laboratories). Images were captured with a Nikon ECLIPSE E800 microscope and a Nikon DXM 1200 digital camera and processed with Image ProPlus software. Staining was quantitated in a blinded fashion from 1 (low expression) to 4 (high expression), independently by three investigators, and scores were averaged.

We isolated RNA with Trizol (Invitrogen), synthesized cDNA with Superscript III RT (Invitrogen), and performed quantitative PCR with a DNA Engine Opticon 2 System (Bio-Rad) and GoTaq SYBR Green (Promega). Absolute mRNA levels were determine for each gene using species- and primer-specific standard curves, then normalized to 18S, and are presented as relative transcript levels (fg/ng 18S). Primer sequences are available upon request.

Hepa1c1c7 cells were transfected with the Rbp-Jk reporter luciferase construct, as previously described (8), and then incubated in serum-free medium with variable glucose content in the presence or absence of 10 nmol/L insulin or 10 nmol/L glucagon.

We isolated and cultured primary mouse hepatocytes as described (8) and obtained human primary hepatocytes from Invitrogen. For gene and protein expression studies, we treated hepatocytes with variable concentrations of glucose, and/or 10 nmol/L insulin (Sigma-Aldrich) for 3 h, with all experiments completed by 24 h after isolation.

Results are shown as mean ± SEM. ANOVA was used for comparison of means among groups. Gene expression levels were correlated by the Pearson correlation. The association between HES1 expression (above the median) and insulin resistance of T2D was evaluated by multivariate logistic regression analysis adjusted for age, steatosis, and BMI. Independent predictors of HES1 expression were evaluated at multivariate regression analysis (generalized linear model), including NAS and insulin levels, the variables most significantly associated at univariate analysis. Differences were considered significant at P < 0.05 (two-tailed).

To clarify the role of Notch signaling in the postdevelopment human liver, we characterized liver Notch and ligand expression in insulin-sensitive subjects with a histologically normal liver. We found that in these subjects, all four Notch receptors were expressed, with relative higher levels of NOTCH1 and NOTCH2 (Fig. 1A, left panel). In addition, all five Notch ligands were detectable by quantitative PCR, with JAG1 and JAG2 showing greater expression than Delta-like ligands. Of note, this expression pattern is broadly similar to the distribution observed in murine liver (Fig. 1A, right panel) as well as in mouse hepatocytes (8). Generally, expression of NOTCH1 and other Notch genes covaried (Fig. 1B) as well as with HES/HEY family target genes (Fig. 1C), but showed no correlation with housekeeping genes (ACTB, GAPDH) and a nonsignificant trend toward a negative correlation with Notch ligand expression (data not shown). These data show that Notch signaling components are present in mature liver and that relative expression is evolutionarily conserved.

Because Notch proteins and ligands are expressed in human liver, we hypothesized that Notch signaling may be increased in obesity and insulin resistance as in mouse models. We performed a pilot study in 42 morbidly obese patients undergoing liver biopsy at time of gastric banding. Full demographic and clinical information is summarized in Table 1, but most of the subjects were female (76%), with SS or NASH (88%), and nondiabetic (79%). Expression of canonical Notch target genes of the HES/HEY family positively correlated with hepatic expression of G6PC and PCK1 (Table 2 and Fig. 2A). Of note, we saw no correlation in severely obese patients between Notch signaling and age, sex, or BMI (data not shown), which suggests that the correlations between HES/HEY and G6PC/PCK1 were independent of overall adiposity, but correlations persisted or even strengthened when diabetic patients were excluded from the analysis (Table 2).

Liver consists primarily of hepatocytes, but also nonhepatocyte residents, including endothelial, phagocytic Kupffer, and stellate cells (31). The relative contribution of hepatocytes to overall hepatic Notch signaling was unclear from gene expression studies from the whole liver, so we performed immunofluorescence for Notch targets on a representative subset of patients across a range of glycemic control. We observed predominantly hepatocyte staining of Notch targets HEY1 and HEYL (Fig. 2B). HEY1 hepatocyte staining and HEY1 gene expression was very well correlated (Fig. 2C), as was HEYL staining and HEYL expression (not shown), suggesting hepatic gene expression is a good surrogate for hepatocyte protein levels. Further, HEY1 gene expression strongly correlated with HES1, HEY1, and other Notch targets (Fig. 2D and data not shown), allowing HEY1 staining as a surrogate for global hepatocyte Notch activation. Having validated the technique, we next examined sections from age- and BMI-matched normoglycemic and diabetic patients and noted a marked increase in HEY1 and HEYL staining in hyperglycemic patients (Fig. 2E and F and data not shown). These data establish that Notch signaling is present in adult hepatocytes and increased in T2D.

The positive correlation of HES/HEY family genes with G6PC/PCK1 in obese patients (Table 2), coupled with increased HEY1 staining in hyperglycemia, suggested that Notch signaling is increased in insulin-resistant liver. To determine whether hepatic Notch signaling was similarly induced in leaner patients and might precede development of frank diabetes in insulin resistance, we analyzed specimens from 38 consecutive outpatients undergoing percutaneous liver biopsy. Full demographics can be found in Table 1, but this group was 97% male, with an average BMI in the overweight range (25–30 kg/m²). In these patients, we again noted that HES/HEY family genes correlated with G6PC and PCK1, as well as with each other (data not shown), suggesting that Notch and gluconeogenic gene expression coregulation is not specific to obese patients. As in the bariatric surgery cohort, Notch signaling did not vary by age, BMI, or abdominal circumference, but as predicted, we found a significant positive correlation with Notch targets (HES1) and plasma insulin levels (not shown) and HOMA-IR, independent of confounding factors (Table 3), which was not mitigated by exclusion of diabetic patients (data not shown). Multivariate logistic regression analysis showed increased HES1 expression was positively associated with the presence of insulin resistance of diabetes (odds ratio [OR] 2.11 [95% CI 1.5–38]), together with age (OR 1.17 [95% CI 1.05–1.40]) and steatosis (OR 5.5 [95% CI 1.3–53]), independently of BMI. These data establish that Notch activation is found in the insulin-resistant liver, preceding frank hyperglycemia and the development of diabetes.

To test the hypothesis that the apparent regulation of hepatic Notch signaling by insulin resistance and hyperglycemia is direct and cell-autonomous, we transfected hepatoma cells with a Notch-reporter luciferase construct and found increased Notch activation in hyperglycemic as opposed to basal conditions (Fig. 3A). Similarly, endogenous Notch target expression was higher in primary mouse or human hepatocytes transiently exposed to hyperglycemia (Fig. 3B and C). Insulin treatment had the opposite effect on Notch signaling, with insulin-treated cells showing decreased reporter activation (Fig. 3D). Interestingly, the inhibitory effect of insulin, but not a synergistic effect of glucagon, was lost in cells chronically cultured in hyperglycemic conditions (Fig. 3E), Further, insulin was no longer able to repress Hes1 expression in primary hepatocytes derived from mice lacking Rbp-Jk (8), the common transcriptional effector of Notch1–4 signaling (Fig. 3F). Similarly, pharmacologic application of a novel Notch antagonist, an ectodomain “decoy” that quenches ligand-dependent Notch signaling (24), abrogated hyperglycemia-induced Notch-reporter activity (Fig. 3G). In sum, these data suggest a cell-autonomous, dynamic regulation of hepatic Notch signaling by metabolic stimuli.

Beyond the positive correlation between HES/HEY gene expression and measures of insulin resistance and hyperglycemia, we predicted that Notch signaling would correlate with hepatic lipid content and markers of necroinflammation. Indeed, when patients are subdivided by liver histology, we find that HES1 and other Notch target genes are strongly upregulated across the spectrum from normal liver to SS to NASH, and further still in insulin resistance (Fig. 4A) or in patients with T2D (Fig. 4B). Notch target staining was similarly increased in hepatocytes of NASH patients compared with SS patients (Fig. 4C).

In fact, HES1 expression more closely correlated to measures of hepatic inflammation than to steatosis, with stronger coefficients of correlation with NAS and ALT levels than with the percentage of steatosis (Fig. 5A–C and Table 3). As such, patients with a NAS score of 0–2, correlating with a low risk of NASH (29,32), had lower Notch, ligand, and HES/HEY family gene expression than those with scores of 3 or higher (Fig. 5D–F). Furthermore, the NAS score was associated with HES1 expression independently of insulin resistance (Table 3). In sum, these data suggest that hepatic Notch signaling is elevated in insulin resistance as well as in NASH.

Notch signaling has been extensively studied in the context of differentiation or cancer (10), but its metabolic functions are novel. Rodent studies have demonstrated a postdevelopment role for Notch in the regulation of obesity-induced insulin resistance/diabetes (8,26) and in liver fat accumulation (9), but whether this would translate to human disease was unclear. This work answers two important questions: hepatic Notch signaling is 1) present in human liver/hepatocytes, and 2) its activation is tied to the metabolic state of the organism, with an independent positive correlation with measures of insulin resistance and hepatic steatosis/inflammation. Notch signaling appears to be inappropriately reactivated in the insulin-resistant liver, reprising its developmental role and reassociating with its molecular partners from differentiation that drives the metabolic effects of Notch signaling. For instance, the Notch transcriptional effector Rbp-Jk binds to and activates FoxO1 (33), a key transcriptional activator of hormone-stimulated hepatic glucose production (2,34), increasing functional hepatic insulin resistance (8). Similarly, Notch signaling activates the nutrient-sensitive mTorc1 pathway in the liver, increasing de novo lipogenesis and hepatic triglyceride (9). Mouse models with reduced hepatic glucose production often have compensatory fatty liver due to reoriented carbon flux; for instance, FoxO-knockout mice (35,36) or mice treated with glucokinase activators (37). Reduced Notch action allows the dissociation of insulin-signaling pathways in the liver, redressing insulin resistance without causing undue nutrient sensitivity, allowing for this rare dual therapeutic benefit independent of effects on body weight or adiposity.

It is intriguing that Notch activation more strongly correlates with markers of steatohepatitis, including the NAS and ALT levels, than with steatosis itself. With better imaging techniques (38), clinicians can increasingly diagnose excess hepatic fat (5). NAFLD is extraordinarily common, with a prevalence approaching 30% in many populations (39), leaving a therapeutic dilemma because only a subset of patients progress to NASH (40). As such, increased Notch signaling may represent a biomarker or even a causative factor for progression from SS to NASH. It is noteworthy that Notch pathway activation has similarly been found in human hepatocellular carcinoma (41) and causes hepatic fibrosis and tumor formation in mice (41–43), which potentially explains the predisposition of patients with NASH to develop cirrhosis and hepatocellular carcinoma (44,45).

One of the major limitations of this type of observational study is that we are able to detect Notch signaling, as well as markers of insulin sensitivity and hepatic fat, at one moment in time. Longitudinal studies (46) are required to determine whether altered Notch signaling predates or predicts the development of worsening steatosis or progression to NASH and hepatocellular carcinoma. Similarly, whether increased Notch signaling precedes development of insulin resistance or heralds the transition to frank diabetes is unknown. In addition, it would be of interest to know whether interventions to reduce insulin resistance, or steatosis and associated inflammation (47,48), also reduce hepatic Notch signaling. Finally, whether higher Notch signaling seen in NASH patients reflects both hepatocyte and nonhepatocyte contribution requires clarification, because this may shed light on how Notch signals are transduced. These future studies will inform the question whether Notch may be a therapeutic target for insulin-resistance/diabetes or NASH.

It is premature, given these lingering questions and potential safety concerns, to propose a clinical trial with the use of Notch inhibitors, which are already in advanced clinical development for cancer (49,50), for treatment of T2D or NASH. This temptation exists, however, in the current era of pandemic obesity. Our data suggest the possibility of alternative uses for these existing therapeutics to combat the various faces of obesity-related metabolic disease in the 21st century.

The work was partly funded by National Institutes of Health grants DK-093604 (U.B.P.) and 5R01-CA136673 (C.J.S.), a Columbia Diabetes and Endocrinology Research Center Pilot and Feasibility Grant (U.B.P.), the Edward Mallinckrodt, Jr., Foundation Grant (U.B.P.), and Ricerca Corrente Fondazione IRCCS Ca’ Granda, Milano (L.V.).

No potential conflicts of interest relevant to this article were reported.

L.V. and C.J.S. analyzed data and wrote the manuscript. R.M.M. and C.K. designed and performed experiments and analyzed data. R.R. collected and managed biological samples for RNA analysis. M.M. selected and collected histological samples and reviewed histological samples. U.B.P. designed and performed experiments, analyzed data, and wrote the manuscript. U.B.P. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Domenico Accili and members of the Accili Laboratory at Columbia University, as well as members of the Metabolic Liver Diseases Laboratory at the University of Milan, for insightful discussion of the data, and Robin Goland at Columbia University for assistance in study design and implementation. The authors acknowledge excellent technical support from Jessie Lee and Taylor Lu at Columbia University, and Rosa Lombardi, Anna Ludovica Fracanzani, Enrico Mozzi, and Silvia Fargion at the University of Milan for collection of clinical data and biological samples.