Liver-Derived S100A6 Propels β-Cell Dysfunction in NAFLD Free

Lipotoxicity due to an imbalance of de novo lipogenesis (DNL) and lipid disposal coupled to systemic insulin resistance are crucial pathological events in nonalcoholic fatty liver disease (NAFLD) (1,2). Type 2 diabetes mellitus (T2DM) and NAFLD are often closely associated with driving synergistic adverse outcomes (3–5), and a two- to fourfold increase in the risk of incidence of T2DM in the NAFLD population has been documented (6). Thus, within a multiorgan framework, hepato-pancreatic crosstalk in meditating systemic glucose homeostasis and insulin sensitivity in metabolic diseases is being explored (7–10). Recent studies have underscored the importance of secreted bioactive peptides/proteins or hepatokines, including fetuin-A, FGF-21, and kisspeptin, in regulating pancreatic β-cell growth and function (11–14). However, much remains unknown about the contribution of lipid stress-induced secretion of dysregulated hepatokines in the alteration of hepato-pancreatic communication and resulting in glucose intolerance and diabetes.

The Ca²⁺-binding S100 protein family of damage-associated molecular patterns mediates cellular response in multiple diseases, including cancer, NAFLD, hepatic fibrosis, coronary artery disease, and inflammatory and rheumatic diseases (15). By binding to a wide range of membrane receptors, proteins in this family modulate signal transduction, Ca²⁺ sensing, and extracellular factors. Notably, S100A8/A9 acts as a TLR4 ligand that triggers a signaling cascade modulating inflammation, cell proliferation, cellular metabolism, differentiation, and tumor development via nuclear factor-κB activation (16). S100A4, S100A8/9, S100A11, and S100A12 are upregulated in the synovial tissue, synovial fluid, or serum of patients with rheumatoid arthritis (17,18). In the liver, S100A11 regulates lipid metabolism through the receptor for the advanced glycation end products (RAGE)–mediated AKT-mTOR signaling pathway, and S100A14 may be a diagnostic marker for hepatocellular carcinoma (HCC) and its progression (16,19).

Our aim here was to explore the pathophysiological relevance of S100A6 protein in fatty liver–associated metabolic stress. In this study, we depleted hepatic S100A6 to explore the function of S100A6 in hepato-islet communication in dietary and genetic mice models of NAFLD, and we also surveyed S100A6 levels in the serum of human patients with NAFLD. We show that S100A6 expression is increased in serum of patients with NAFLD, and ChREBP regulates hepatic S100A6 expression, and suggest S100A6 antagonism improves glucose-stimulated insulin secretion (GSIS) and glycemia in NAFLD. Thus, we provide evidence that hepatic S100A6 is a messenger protein mediating, to our knowledge, a heretofore unknown interorgan communication between liver and pancreas upon intrahepatic lipid stress.

For this study, conducted in the Department of Hepatology, Institute of Postgraduate Medical Education and Research (IPGME&R), Kolkata, India, 28 patients with biopsy-proven NAFLD and 20 individuals without NAFLD (healthy control participants) were recruited. NAFLD was diagnosed according to histopathological parameters, including steatosis, hepatocellular ballooning, lobular inflammation, and stage of fibrosis (20), and the NAFLD activity score was calculated by one liver pathologist. Fasting blood samples were collected from both healthy control participants and patients with NAFLD. The bioclinical parameters included levels of plasma glucose, insulin, total cholesterol and triglycerides, serum aspartate aminotransferase, alanine aminotransferase, which were measured according to the manufacturer’s protocol. HOMA was performed to calculate HOMA for insulin resistance (HOMA-IR) and HOMA-β for β-cell function. HOMA-IR was calculated as (fasting insulin × fasting glucose)/22.5; and HOMA-β was calculated as (20 × fasting insulin)/(fasting glucose − 3.5). Plasma levels of S100A6 were measured using a commercially available ELISA kit according to the manufacturer’s protocol (ABclonal). The study was approved by the human ethics committee of IPGME&R hospital, and all the volunteers gave written informed consent.

All experiments were approved by the Institutional Animal Ethics Committee at the Council of Scientific and Industrial Research–Indian Institute of Chemical Biology, approved by the Committee for Control and Supervision of Experiments on Animals, the Ministry of Environment and Forest, and the Government of India. C57BL/6 male mice, 6–8 weeks old, were housed in a 12 h light/dark cycle at 22 ± 1°C with free access to food and water. Mice were fed a 60% high-fat diet (20% protein, 60% fat, 20% carbohydrate, and with energy density of 5.21% Kcal/g; D12492; Research Diets) for 8 weeks. For each in vivo experiment, mouse cohorts were randomized, then followed by the respective treatment of adenoviruses for both ChREBP and shS100A6 via intravenous injections.

The recombinant adenovirus containing the shRNA construct for S100A6 was generated by BLOCK-iT Adenoviral Expression System (Invitrogen) using the sequence GGACCGTAACAAGGATCAGGA, following the manufacturer’s instructions. Recombinant adenoviruses were purified using PureVirus Adenovirus Purification Kit (Cell Biolabs Inc.) and administered through tail-vein injection at a titer of 2.5 × 10¹² pfu/mouse.

Oral glucose tolerance tests (OGTTs) and intraperitoneal insulin tolerance tests (ITTs) were performed per the previously described protocol by Mouse Metabolic Phenotyping Centers (23). For the OGTT, mice were fasted for 12 h and the fed glucose (2 g/kg). For ITTs, insulin (0.5 IU/kg body weight; Actrapid, Novo Nordisk) was intraperitoneally injected after 6 h fasting; mice had free access to water. Fasting blood glucose levels were detected at 30, 60, 90, and 120 min after injection from the tail-tip cut with a glucometer (Accu-Check Aviva).

The pancreas was inflated with 3 mL of collagenase P solution (Roche; 1 mg/mL in Hanks’ balanced salt solution) and excised and digested in a water bath at 37°C for 30 min. After digestion, islets were washed three times with cold Hanks’ balanced salt solution and cultured overnight in RPMI 1640 medium supplemented with 10% FBS, glucose (5.5 mmol/L), penicillin (100 U/mL), and streptomycin (100 U/mL) (HiMedia). On the following day, after dithizone staining, isolated islets (20/well) were handpicked in a 24 well plate under a stereoscopic microscope with the aid of a pipette and transferred to RPMI 1640 medium for further biochemical analysis.

Islets were incubated in 500 µL of Krebs buffer (118 mmol/L NaCl, 4.7 mmol/L KCl, 1.2 mmol/L KH2PO4, 1.2 mmol/L MgSO₄, 4.2 mmol/L NaHCO₃, 2 mmol/L CaCl₂, and 10 mmol/L HEPES, 0.1% (w/v) BSA, pH 7.4) containing 2.8 mmol/L glucose for 1 h at 37°C. Islets were then further incubated for 1 h in 200 µL Krebs buffer containing 2.8 mmol/L or 16.7 mmol/L glucose. After incubation, insulin concentrations in the medium were measured by ELISA (RayBioMouse Insulin Elisa Kit) and were normalized by total protein content.

For in vivo GSIS, mice were weighed at the start of the experiment, blood was collected from the tail vein, and 2 mg/kg body weight glucose was injected intraperitoneally. Blood was collected from the tail vein at 10, 20, 30, and 60 min. ELISA for the measurement of plasma insulin levels was performed using RayBioMouse Insulin Elisa Kit (EML-Insulin).

At least three Bouin’s solution–fixed and paraffin-embedded sections per mouse, 100 μm apart, were immunostained for insulin and analyzed for islet/β-cell mass. β-Cell mass for each pancreas was measured as the total β-cell area over the total pancreas area and the pancreas weight before fixation. The β-cell size was calculated by dividing the islet size by the corresponding number of β cells. Imaging was performed on a fluorescence microscope (Zeiss), and images were analyzed with Zeiss imaging software.

Primary hepatocyte isolation was performed with C57BL/6 male mice (6–8 weeks old) according to the protocol previously described (21) and maintained in Hepatozyme (Gibco Life Technologies) supplemented with l-glutamine and 1% penicillin–streptomycin. Hepatocytes were then treated for 48 or 72 h with adenoviruses at a multiplicity of infection of 40. Human HCC cell-line HepG2 and rat insulinoma INS1 832/13 cells were cultured in, respectively, DMEM–high glucose supplemented with 10% FBS and 1% penicillin–streptomycin and in RPMI medium supplemented with 11 mM glucose, 2 mM l-glutamine, 1 mM sodium pyruvate, 10 mM HEPES, 50 μM β-mercaptoethanol, 10% FBS, and 1% penicillin–streptomycin. Plasmids and the siRNA were transiently transfected using Lipofectamine 3000 and Lipofectamine RNAiMAX reagents, respectively, following the manufacturer’s protocol.

Mouse primary hepatocytes were transfected with Ad-EGFP and Ad-ChREBP. After 48 h, cell supernatant was collected and centrifuged to remove the dead cells and cellular debris. Primary islets then were treated with the conditioned medium thus obtained for GSIS assay.

HepG2 was transfected with plasmid-cloning DNA (pcDNA)–enhanced green fluorescent protein (eGFP) and pcDNA–ChREBP. After 48 h, cell supernatant was collected and centrifuged to remove the dead cells and cellular debris. The conditioned media thus obtained was then treated onto INS1 for GSIS assay and proteomic analysis.

As mentioned, the conditioned medium collected was subjected to peptide and protein analysis. The cell supernatant collected (after overexpression) was treated with 6 M Gn-HCl supplemented with protease inhibitors and then sonicated to break nucleic acids. The supernatant was thus carefully collected after centrifugation and protein concentration was quantified using the bicinchoninic acid assay method. Samples were first reduced with 5 mmol/L tris (2-carboxyethyl) phosphine, further alkylated with 50 mmol/L iodoacetamide, and digested for 16 h at 37°C. Digests were then cleaned using a C18 silica cartridge and dried using a speed vacuum. The dried pellet was resuspended in Buffer A (5% acetonitrile, 0.1% formic acid). The peptide mixtures were analyzed via mass spectrometry using EASY-nLC 1000 system (Thermo Fisher Scientific) coupled to a QExactive mass spectrometer (Thermo Fisher Scientific) equipped with a nanoelectrospray ion source. The peptide mixture was resolved using a 15 cm PicoFrit column (360 µm outer diameter, 75 μm inner diameter, 10 μm tip) filled with 1.8 µm of C18-resin (Dr. Maeisch). The peptides were loaded with Buffer A and eluted with a 0–40% gradient of Buffer B (95% acetonitrile, 0.1% formic acid) at a flow rate of 300 mL/min for 100 min, and then finally equilibrated with Buffer A. Mass spectrometry data were acquired using a data-dependent, top-10 method that dynamically chooses the most abundant precursor ions from the survey scan (300–1650 Th) for higher-energy collisional dissociation fragmentation with dynamic exclusion duration of 60 s. The raw files were processed using Proteome Discoverer (version 2.2) against the Uniprot human database.

INS1 832/13 cells were seeded in a 24 well plate and treated for 14 h with 1:5 diluted conditioned medium (as described previously in RESEARCH DESIGN AND METHODS) in RPMI medium containing 5.5 mM glucose. The cells were then washed with secretion assay buffer (114 mM NaCl, 4.7 mM KCl, 1.2 mM KH₂PO₄, 1.16 mM MgSO₄, 20 mM HEPES, 2.5 mM CaCl₂, 0.2% FFA–BSA, pH 7.4) and stimulated with 2.5 mM glucose for 1.5 h and then in 16.7 mM glucose for 1 h. The cell supernatant was then collected so insulin levels could be measured, and cells were harvested for protein extraction. Insulin levels were measured using Rat Insulin ELISA Kit (Thermo) and normalized using total cellular protein.

After 48 h of overexpression, the total cellular RNA was extracted using RNA-Xpress Reagent (Himedia) from cell homogenates. For cDNA preparation using iScript cDNA Synthesis kit (Bio-Rad), 1 μg of RNA was used. The gene expression was measured using SYBR Green (Bio-Rad) in a real-time PCR thermocycler (Bio-Rad) and normalized by b-actin/18S as an internal control, using the comparative cycle threshold method. The primer sequences are listed in the Supplementary Information.

Total protein from the cells and tissue was isolated using radioimmunoprecipitation assay buffer supplemented with a 1× protease inhibitor cocktail (Sigma). The protein concentration was quantified using the bicinchoninic acid assay method. The samples were resolved on 10% SDS-PAGE and transferred onto a 0.2-µm polyvinylidene difluoride membrane (Bio-Rad). For immunoblotting, the membrane was blocked with 5% nonfat dry milk in Tris-buffered saline with 0.5% Tween 20 for 1 h, followed by incubation with specific primary antibodies overnight at 4°C and then with respective secondary antibodies for 1 h at room temperature. The signals were detected by chemiluminescence (Clarity ECL reagent, Bio-Rad). Antibodies used in the study are enlisted in the Supplementary Information.

Primary hepatocytes or HepG2 cells were cultured in Hepatozyme (Gibco Life Technologies) or DMEM medium. The assay was performed using an Active Motif kit, strictly following the manufacturer’s protocol. Primer sequences are shared in the Supplementary Information.

The Seahorse XFp Extracellular Flux Analyzer was used to assess various metabolic parameters of the cellular oxygen consumption rate (OCR) by real-time monitoring. INS-1 832/13 cells were cultured at a density of 15,000 cells/well in RPMI medium containing 11 mM glucose in six wells of Agilent Seahorse XF Cell Culture Microplate. The following day, cells were treated with recombinant S100A6 protein in 5.5 mM glucose-containing RPMI medium for 14 h. Cells were washed with prewarmed Seahorse XF DMEM medium supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose (pH 7.4) and incubated in the same medium for 1 h in a non-CO₂ incubator. Three baseline measurements were recorded at the start of the assay before the addition of any of the inhibitors. The mitochondrial efficiency of the cells in control and treatment conditions was assessed after injection of specific compounds or inhibitors for ATP-linked respiration (1 uM oligomycin), maximal respiration (1 uM carbonyl cyanide p-trifluoromethoxyphenylhydrazone), and proton leak (0.5 µM antimycin A + 0.5 µM rotenone). The OCR data were plotted as absolute values.

The intracellular cAMP levels were measured in islets and in INS-1 cells using a luminescence-based assay from Promega (catalog V1501). The cAMP levels were then measured following the manufacturer’s protocol. The absolute values were plotted as relative luminescence units in GraphPad Prism 8.

ATP content was determined by using the ATP Determination Kit (Thermo Fisher Scientific), and ATP content was determined in terms of luminescence using Synergy ¹H Hybrid reader (BioTek). ATP content was normalized to total protein content.

INS1 cells were seeded on a coverslip and allowed to adhere overnight. The cells were then treated with the conditioned media from eGFP and ChREBP-overexpressed HepG2 cells for 14 h and then incubated with the JC-1 dye (5 µM) for 2 h. The cells were then washed with PBS and mounted using DAPI mounting media. The imaging was done using a Zeiss fluorescence microscope.

For the reanalysis of the human hepatic transcriptome, the fastq samples were mapped to the Homo sapiens GRCm38 genome using STAR (version 2.27.2b) to create bam files and further processed using the Rsamtools, Rsubread, and Genomic Alignments R packages to create the abundance table. The count table was then normalized and differentially abundant genes were identified using the DESeq2 R package. Genes with a greater than twofold change with P < 0.05 and adjusted P < 0.05 were considered significant.

The integrated response of glucose-stimulated calcium dynamics in INS1 cells was measured using fluorescent dye Fura-2 AM. Readings (excitation, 340 and 380 nm; emission, 508 nm) were taken at 5 min intervals from the bottom plane for 30 min. For fura 2/iCa2–specific signal, the 340 to 380-nm ratio was calculated.

Data are presented as mean ± SEM. Graphpad Prism 8 software was used for statistical analysis. We used t tests and ANOVA as appropriate, followed by Bonferroni post hoc test. P < 0.05 was considered statistically significant. Two-tailed Fisher exact test was performed for contingency table analysis to compute the exact P value.

All study data are included in the manuscript and/or supporting information.

Hepatokines regulate an array of systemic metabolic functions, including the metabolism of fatty acids, carbohydrates, amino acids, along with the synthesis and secretion of several lipoproteins and hormones (14,22,23). A subgroup of hepatokines, the Ca²⁺-binding S100A family proteins, performs numerous cellular and molecular functions by regulating calcium balance, hepatocellular apoptosis, energy metabolism, and inflammation. To determine the relevance of S100A proteins in NAFLD, we first examined the expressions of S100A family genes using publicly available hepatic transcriptome data for both healthy people and patients with nonalcoholic steatohepatitis (NASH) (24,25). We found that expressions of three S100A proteins—S100A4, S100A6, and S100A14—were significantly upregulated in the livers of patients with NASH (Fig. 1A). Although S100A4 and S100A14 have already been reported to be associated with hepatic pathologies such as fibrosis and HCC (19,26,27), the role of S100A6 (log2 fold change, 1.2768; adjusted P = 1.16 × 10⁻⁸) in NASH and associated metabolic diseases is not known (Fig. 1B). A distinct gain of correlations of S100A6 expression with insulin pathway, fatty acid metabolism, and lipogenesis genes (r < −0.5, r > 0.5) further underscores the importance of S100A6 in NASH (Supplementary Fig. 1A).

Next, we collected blood samples from a cohort patients with biopsy-proven NASH (n = 28) and age- and sex-matched control participants (Table 1). Consistent with the transcriptomic data, we found that circulatory S100A6 levels were considerably increased in patients with NASH (P < 0.001) (Fig. 1C). Interestingly, plasma S100A6 levels were negatively correlated with HOMA-β, representing pancreatic β-cell function, but not with HOMA-IR, representing insulin resistance (Fig. 1D), whereas such correlations were absent in the control group. Collectively, these results indicate that enhanced S100A6 levels in NASH are associated with increased lipogenesis and might adversely affect β-cell function.

Toward establishing the causal links between aberrant lipid metabolism and lipogenesis with S100A6 expression, we generated an HFD-induced preclinical model of NAFLD. As shown in Fig. 1E and F, expression of both the S100A6 gene and proteins were upregulated in the liver with a corresponding increase in the serum. Consistently, the expression of glucose-sensing transcription factor and the major driver of hepatic DNL, ChREBP, was also enhanced (Fig. 1E and F). ChREBP often contributes to whole-body energy balance by regulating hepatic expression and secretion of hepatokines (26). In silico prediction of the carbohydrate response element (ChORE) site within the S100A6 promoter indicated ChREBP can potentially bind at 281 bp upstream of the transcription start site (Fig. 1G, top panel).

We next examined whether ChREBP can directly induce hepatic S100A6 expression. Chromatin immunoprecipitation followed by the quantitative PCR analysis of ChoRE revealed a fivefold-enriched glucose-dependent occupancy of ChREBP on the S100A6 promoter in primary hepatocytes and HepG2 cells (Fig. 1G bottom panel, and Supplementary Fig. 1E). In agreement with the chromatin immunoprecipitation results, overexpression of ChREBP, as well as high-glucose treatment, led to robust induction of S100A6 expression in HepG2 cells (Supplementary Fig. 1B). Conversely, siRNA-mediated knockdown of ChREBPα in HepG2 cells significantly suppressed high glucose–dependent, elevated S100A expression (Supplementary Fig. 1C). Knockdown was specific, as reflected on RT-PCR by a significant reduction of ChREBPα expression but not of ChREBPβ expression.

To validate the cell-autonomous effect of ChREBP-induced S100A6 expression, ChREBP was ectopically expressed in primary murine hepatocytes by adenovirus (Ad-ChREBP). As expected, overexpression of ChREBP led to enhanced lipid-droplet accumulation (Supplementary Fig. 1D) with elevated S100A6 protein expression (Fig. 1H). Finally, adenoviral-mediated ChREBP overexpression in the murine liver also significantly induced S100A6 expression (Fig. 1I). Taken together, the data showed serum S100A6 level is increased in patients or experimental animals with NAFLD, and ChREBP regulates hepatic S100A6 expression.

Taking cues from human data on the negative association of S100A6 and HOMA-β, we next examined the contribution of S100A6 in regulating β-cell function in the context of hepatic ChREBP overexpression. To execute this, S100A6 was depleted in the mouse liver by adenovirus (Ad-shS100A6), followed by overexpression of ChREBP (Fig. 2A). As expected, enhanced ChREBP expression led to induction of S100A6, an effect substantially blunted upon S100A6 knockdown (Fig. 2B). S100A6 knockdown also significantly reduced serum S100A6 levels (Fig. 2B). The livers from the Ad-ChREBP group were paler than those of the control (Ad-EGFP) group, depicting significant lipid accumulation, whereas lipid accumulation was modestly recovered in the sh-S100A6 cohort (Fig. 2C, upper panel). Consistently, the liver histology also showed enhanced lipid accumulation in Ad-ChREBP mice, which was somewhat restored upon S100A6 depletion (Fig. 2C, lower panel).

Consistently, ChREBP overexpression led to an increase in liver triglyceride levels; however, these remained unaltered by S100A6 knockdown (Fig. 2D). A similar phenomenon was previously observed by Benhamed et al., whereby mice overexpressing ChREBP developed greater hepatic steatosis than did control mice but did not develop insulin resistance (27). Furthermore, ChREBP overexpression led to increased liver weight and ratio of liver to body weight, which was reduced after S100A6 knockdown, suggesting its role in hepatic lipid accumulation (Fig. 2E). However, we did not find any significant difference either in pancreatic weight, histological features, β-cell size, or β-cell mass (Supplementary Fig. 2A–D). Consistently, the OGTT results revealed no change in glucose excursion among three experimental mice groups (Supplementary Fig. 2E). In contrast, ChREBP overexpression caused impaired insulin secretion, whereas depletion of S100A6 substantially enhanced glucose-induced insulin secretion (Fig. 2F). These data support the hypothesis that ChREBP-driven upregulation of hepatic and circulatory S100A6 impairs insulin secretion from the pancreatic β cells, albeit to an extent that did not cause glucose intolerance in the short term.

On the basis of our results, we surmise that hepatic ChREBP elicits two critical stimuli: enhanced DNL-dependent intrahepatic lipid accumulation and induction of hepatic S100A6 production and secretion. To determine the direct impact of ChREBP-driven hepatic S100A6 on pancreatic β-cell dysfunction, we generated an in vitro model of HepG2 cells expressing ChREBP (Fig. 3A). Proteomic analysis of the culture supernatant revealed a substantial enrichment number of proteins including S100A6 (Fig. 3B and Supplementary Table 1). We next incubated INS1 cells with the conditioned medium of ChREBP-overexpressed HepG2 cells and assayed for GSIS. As shown in Fig. 3C, GSIS was substantially reduced when cells were treated with ChREBP-overexpressed conditioned medium. Furthermore, conditioned-medium treatment from S100A6 knockdown in ChREBP-overexpressed HepG2 cells rescued them from impaired GSIS, in contrast to ChREBP-overexpressed conditioned medium–treated INS1 cells (Fig. 3D). Similarly, treatment of ChREBP-overexpressed conditioned medium from primary hepatocytes on islets also led to a significant reduction in GSIS (Fig. 3E2). As observed in INS1 cells, the reduction in GSIS in islets was abolished when S100A6 was concomitantly knocked down in primary hepatocytes (Fig. 3E2).

For validating the adverse impact of S100A6 on GSIS, treatment of recombinant S100A6 protein (rS100A6) in nanomolar concentrations also caused a decrement in insulin secretion in INS1 cells (Fig. 3F). Consequently, we sought to determine the possible mechanism for S100A6-mediated impairment of GSIS. S100 proteins can interact with RAGE (28), which are abundantly expressed in pancreatic β cells and are involved in regulating the apoptotic and inflammatory pathways (29,30). To specifically investigate the functional role of RAGE on β cells in mediating S100A6 action on GSIS, we pretreated INS1 cells and islets with a RAGE antagonist, FPS-ZM1 (31), followed by rS100A6 treatment for 14 h. FPS-ZM1 significantly abrogated rS100A6-mediated abatement of GSIS (Fig. 3G and H), suggesting S100A6 suppresses insulin secretion by engaging with RAGE.

Insulin secretion from the β cells involves a sequence of intracellular events in response to various exogenous mediators (32,33), including cytosolic Ca²⁺ levels, cAMP signaling, and mitochondrial respiration (34–36). To elucidate the S100A6-mediated pathways influencing insulin secretion, we performed a series of experiments. Mouse islets were used to measure intracellular cAMP levels in response to S100A6 treatment. Furthermore, islets were cotreated with the GLP1R agonist exendin-4 (E4) and the phosphodiesterase inhibitor IBMX in the presence and absence of recombinant S100A6 protein and RAGE antagonist FPS-ZM1. S100A6 impaired islet cAMP production in response to the long-acting GLP1 analog E4, a widely used antidiabetic agent that binds and activates GLP1R on β cells and potentiates GSIS by stimulating β-cell cAMP synthesis. Although S100A6 suppressed intracellular cAMP production in response to E4 stimulation, antagonizing RAGE significantly abrogated the reduction in cAMP levels (Fig. 4A), suggesting the potential involvement of S100A6 in either inhibiting GLP1R or stimulating the activity of phosphodiesterase. Consistently, S100A6 also led to a decrement in E4-induced protein kinase A (PKA) activity in primary islets, a readout of the cAMP-signaling axis, which was also abrogated upon RAGE antagonism (Fig. 4B). Consistently, recombinant S100A6 protein treatment led to a significant reduction in IBMX and forskolin-induced intracellular cAMP and PKA levels in INS1 cells in a RAGE-sensitive manner (Supplementary Fig. 3A and B). Moreover, S100A6 was unable to diminish cAMP levels in INS1 cells treated with 8-bromo cAMP, a phosphodiesterase-resistant, cell-permeable cAMP analog (Supplementary Fig. 3C).

We next investigated the effect of S100A6 protein in regulating intracellular calcium dynamics in pancreatic β cells. The recombinant S100A6 protein was incubated with INS1 cells and intracellular calcium was measured using Fura-2 AM. Interestingly, the glucose-stimulated intracellular calcium response was unaffected upon S100A6 recombinant protein treatment in high-glucose concentrations (Supplementary Fig. 3D).

We further examined the glucose-dependent mitochondrial metabolism on insulin secretion (37–39). The recombinant S100A6 protein treatment on primary islets led to a significant reduction in intracellular ATP levels as observed via decreased luminescence (Fig. 4C). Expectedly, high-glucose treatment led to the hyperpolarization of the mitochondrial membrane potential as depicted by the red fluorescence of the JC-1 J dimers, whereas ChREBP-overexpressed conditioned medium from HepG2 cells led to attenuation in the red to green fluorescence-intensity ratio depicting mitochondrial membrane depolarization (Fig. 4D and E). Both basal and maximal OCRs, a readout of mitochondrial respiration, were also reduced by rS100A6 protein (Fig. 4F). The ATP synthase activity was also attenuated, as depicted by the diminished respiratory response upon oligomycin addition. Rotenone and antimycin treatment, inhibiting complexes 1 and 3 of the electron transport chain, respectively, showed a similar pattern of suppression in control and protein-treated INS-1 cells, revealing no mitochondrial leak (Fig. 4G). Collectively, our results indicate that S100A6 reduces β-cell cAMP production, impairs mitochondrial metabolism, causes a reduction in intracellular ATP levels required for insulin secretion, and may impair incretin action on the cAMP synthesis and GSIS potentiation from β cells.

We next examined the role of hepatic S100A6 on the insulin secretory capacity in a preclinical model of NAFLD. To this end, mice were fed a HFD for 8 weeks and the S100A6 gene was knocked down in the liver via adenovirus expressing S100A6-specific shRNA (Ad-shS100A6) (Fig. 5A). HFD-fed mice gained substantial body weight at a faster rate than did the chow diet-fed group, and the depletion of S100A6 did not cause any difference in body weight (Fig. 5B). Knockdown of the S100A6 expectedly led to decreased hepatic and serum levels of S100A6 (Fig. 5C), indicating that the liver is the predominant source of circulating S100A6. Importantly, Ad-shS100A6 treatment did not affect S100A6 expression in adipose tissue (Supplementary Fig. 4A). Moreover, knockdown of S100A6 did not alter liver weight, liver to body weight ratio, hepatic fat accumulation, epididymal fat mass, pancreatic weight (Fig. 5D and E), and morphology of the pancreas, as evident by the hematoxylin and eosin (H-E) staining (Supplementary Fig. 4B).

Mice receiving an HFD for 8 weeks predictively developed glucose intolerance and impaired GSIS, as compared with chow diet-fed control mice (Fig. 5F and G). Interestingly, hepatic S100A6 depletion in HFD-fed mice resulted in significantly lower fasting glucose levels as well as improved glucose tolerance, and also resulted in restored GSIS in the HFD-fed mice (Fig. 5G). Conversely, after knockdown of S100A6, a similar trend in insulin-mediated glucose-lowering capacity excludes the possibility of differential insulin sensitivity or a compensatory increase in insulin secretion in the face of altered insulin resistance as the underlying mechanism for improved glucose tolerance (Supplementary Fig. 4C). Moreover, pancreas morphometric parameters were also unaltered after 8 weeks of HFD feeding, thereby excluding differences in β-cell mass and size to account for the differences in glucose homeostasis (Supplementary Fig. 4D and E).

Taken together, our data strongly suggest that HFD-induced fatty liver impairs insulin secretion via S100A6. Diminution of circulatory S100A6 would improve metabolic outcomes by enhancing insulin secretion and glucose tolerance in HFD-induced NAFLD.

Using complementary human patient data and in vitro, ex vivo, and in vivo experiments, we identified hepatic S100A6 as a messenger mediating β-cell dysfunction in NAFLD. We show that intracellular lipid stress induced either by treatment with high glucose in vitro, by feeding an HFD to mice, or overexpression of ChREBP led to upregulation of hepatic as well systemic S100A6 levels. Augmented systemic S100A6, in turn, perturbed intracellular cAMP synthesis and inhibited mitochondrial ATP production in β cells, leading to inhibition of GSIS. Knockdown of S100A6 in both HFD-fed and in Adv-ChREBP mice resulted in improved GSIS in vivo, uncovering the liver as a site of regulated S100A6 production and release, and providing the first evidence, to our knowledge, that secreted S100A6 is a critical contributor behind NAFLD-driven T2DM. Thus, in NAFLD, the liver is exposed to two critical stimuli elicited by ChREBP action: intrahepatic lipid accumulation by enhanced DNL and S100A6 production that impairs GSIS in pancreatic β cells. These observations propose a direct sequential role of enhanced DNL-induced NAFLD progression in driving the development of T2DM. Furthermore, the results assign what we believe is a heretofore unknown role of S100A6 in liver-to-islet endocrine signaling (Fig. 6).

Endocrine communication between the liver and other distal organs plays a critical role in maintaining whole-body energy homeostasis. For example, enhanced hepatic extracellular signal-regulated kinase activity in obese mice is associated with pancreatic β-cell proliferation (40). Authors of another study deciphered the role of hepatocyte-derived factor(s) in promoting β-cell proliferation using in vivo parabiosis (41). Hepatocyte-derived kisspeptin1 was reported to reciprocally regulate pancreatic β-cell function (13). Studies have also demonstrated the role of HFD-induced, hepatocyte-derived extracellular vesicles in promoting the proliferation of β-cells, but these vesicles have no significant effect on insulin secretion (42). However, there is still no evidence, to our knowledge, for whether hepato-pancreatic crosstalk has any role in the accelerated occurrence of glucose intolerance and T2DM in patients with NAFLD.

In physiology, insulin secretion is augmented by glucose levels in a biphasic manner, with an early first-phase burst of insulin, followed by a second, more sustained release of insulin (43). Patients with T2DM and those at risk for developing the disease present with diminished first-phase insulin secretion (44). Therefore, targeting β-cells to improve their function and survival is of utmost importance for improving diabetes management by developing newer therapies for the disease. Experiments with RAGE antagonists suggest that RAGE signaling transduces extracellular S100A6’s effects in inhibiting GSIS in β cells. Apart from the epidemic of ectopic hepatic fat accumulation, an increase in the prevalence of obesity has also led to a concomitant increase in the prevalence of T2DM. Results from a series of studies conducted in patient samples and obese mice models have hinted about secretory factors (from nutrient-induced metabolically stressed tissues) that link obesity with insulin resistance and β-cell dysfunction. At this juncture, we do not know where S100A6 is also involved in this signaling process. Our data from the HFD-mice model hinted at a pivotal role for S100A6 in restoring β-cell function and achieving diabetes remission in the obese mice model, However, the exact participation of S100A6 in obesity-induced β-cell dysfunction remains to be determined.

In addition to an exploration of molecular and cellular events in DNL-driven S100A6 expression and its adverse impact on β-cell insulin secretory function, this study could also be relevant for translational medicine. First, elevated levels of S100A6 expression are associated with several diseases, such as cancer, Alzheimer’s disease, and inflammatory and autoimmune diseases (45–47), and now we have unraveled its pathogenetic role in NAFLD. Although the exact mechanism of how S100A6 is secreted from the fat-laden hepatocytes is not well understood, elevated serum levels of S100A6 may serve as a biomarker in identifying patients with a heightened risk of developing β-cell dysfunction or T2DM. Second, our study also supports that neutralizing circulating S100A6 and/or antagonizing its receptor, RAGE, could be therapeutic in restoring β-cell function in NAFLD.

Acknowledgments. We sincerely thank the BioX Center of Indian Institute of Technology Mandi for the facilities. S.D. thanks the DST, India for her research fellowship.

Funding. This research was supported by the Department of Biotechnology, Government of India (grant BT/PR27786/MED/30/1980/2018 to P.M.), the Science and Engineering Research Board (grant CRG/2019/004006 to P.M.), and by the Council of Scientific and Industrial Research, India (grant MLP138 to P.C.). S.D. was supported by a research fellowship from DST, India.

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

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author Contributions. S.D., D.D., A.P., and S.K.M. contributed to the study methodology, validation, data curation and formal analysis, and investigation. S.D. and D.D. contributed to data visualization. S.D. wrote the original draft of the manuscript; D.D. reviewed and edited the manuscript. P.R. and P.V.D. contributed to the investigation. K.D. and S.M. contributed to the study methodology and investigation. P.C. and P.M. conceived of and designed the study, researched data and methodology, reviewed and edited the manuscript, and contributed to project administration and funding acquisition. P.C. and P.M. are the guarantors of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.