Patulin Alters Insulin Signaling and Metabolic Flexibility in HepG2 and HEK293 Cells

Non-communicable diseases (NCDs) have risen rapidly worldwide, sparking interest in causative agents and pathways. Patulin (PAT), a xenobiotic found in fruit products contaminated by molds, is postulated to be diabetogenic in animals, but little is known about these effects in humans. This study examined the effects of PAT on the insulin signaling pathway and the pyruvate dehydrogenase complex (PDH). HEK293 and HepG2 cells were exposed to normal (5 mM) or high (25 mM) glucose levels, insulin (1.7 nM) and PAT (0.2 μM; 2.0 μM) for 24 h. The qPCR determined gene expression of key enzymes involved in carbohydrate metabolism while Western blotting assessed the effects of PAT on the insulin signaling pathway and Pyruvate Dehydrogenase (PDH) axis. Under hyperglycemic conditions, PAT stimulated glucose production pathways, caused defects in the insulin signaling pathway and impaired PDH activity. These trends under hyperglycemic conditions remained consistent in the presence of insulin. These findings are of importance, given that PAT is ingested with fruit and fruit products. Results suggest PAT exposure may be an initiating event in insulin resistance, alluding to an etiological role in the pathogenesis of type 2 diabetes and disorders of metabolism. This highlights the importance of both diet and food quality in addressing the causes of NCDs.


Introduction
Non-communicable diseases (NCDs), including diabetes and cancer, have quickly become the leading cause of mortality worldwide, with disproportionately higher impacts on public health in developing countries [1]. This has raised interest in causative agents, diet and nutritional guidelines. Mycotoxins are secondary metabolites produced by fungi. These potentially damaging xenobiotics are found in food, with a higher incidence in developing countries [2,3].
Patulin (PAT) is a mycotoxin produced by Penicillium, Bissochlamys and Aspergillus sp. [4]. These molds are found in overripe or rotting fruit and apple products. A safety level of 50 µg/L PAT in consumables was established following evidence of adverse effects in exposed humans and animals [3,5]. Despite this regulation and the assertion that PAT levels are a measure of poor product quality, there are vast variations in PAT concentrations in apple products worldwide [5]. PAT induces oxidative stress, compromises mitochondrial function and causes cell death via depletion of antioxidant glutathione (GSH) [6][7][8][9][10][11]. PAT is also associated with changes in lipid metabolism and inflammation [2,12]. These deleterious

PAT Alters Expression of Enzymes Involved in Glucose Generation
Previous studies have alluded to PAT as a diabetogenic agent [15]. In a preliminary screening, HEK293 and HepG2 cells were exposed to a high glucose (25 mM) medium and PAT for 24 h. qPCR was then used to measure gene expression of enzymes vital to glucose homeostasis.
PFK-2, a key enzyme in glycolysis, was not altered by PAT in HEK293 cells ( Figure 1A; p = 0.7270) but decreased significantly in HepG2 cells ( Figure 1B; p = 0.0231; 10-fold). Although overall statistical significance was found for glycogenolysis initiator, PYGL, in both cell lines (HEK293 ( Figure 1A; p = 0.0137)), post-hoc tests found that only the distinct two-fold increase in HepG2 cells was significant when compared to the control ( Figure 1B; p = 0.0218)]. PCK-1, a rate limiting enzyme in gluconeogenesis, was significantly altered by PAT at 2 µM in both cell lines, though trends differed between them (HEK293 cells ( Figure 1A; p = 0.0112) and HepG2 cells ( Figure 1B; p = 0.0214)). This outcome provides a preliminary indication that PAT caused a shift from glycolysis to gluconeogenesis and glycogenolysis in HepG2 cells with HG media. Most significant changes were observed at 2 µM PAT-above the established safety concentration for PAT in consumables.

PAT Alters Expression of Enzymes Involved in Glucose Generation
Previous studies have alluded to PAT as a diabetogenic agent [15]. In a preliminary screening, HEK293 and HepG2 cells were exposed to a high glucose (25 mM) medium and PAT for 24 h. qPCR was then used to measure gene expression of enzymes vital to glucose homeostasis.
PFK-2, a key enzyme in glycolysis, was not altered by PAT in HEK293 cells ( Figure  1A; p = 0.7270) but decreased significantly in HepG2 cells ( Figure 1B; p = 0.0231; 10-fold). Although overall statistical significance was found for glycogenolysis initiator, PYGL, in both cell lines (HEK293 ( Figure 1A; p = 0.0137)), post-hoc tests found that only the distinct two-fold increase in HepG2 cells was significant when compared to the control ( Figure  1B; p = 0.0218)]. PCK-1, a rate limiting enzyme in gluconeogenesis, was significantly altered by PAT at 2 µM in both cell lines, though trends differed between them (HEK293 cells ( Figure 1A; p = 0.0112) and HepG2 cells ( Figure 1B; p = 0.0214)). This outcome provides a preliminary indication that PAT caused a shift from glycolysis to gluconeogenesis and glycogenolysis in HepG2 cells with HG media. Most significant changes were observed at 2 µM PAT-above the established safety concentration for PAT in consumables. PAT alters expression of enzymes involved in glucose production in hyperglycemic conditions. qPCR was used to assess expression of glucose homeostasis enzymes PYGL, PFK-1 and PCK-1 in (A) HEK239 and (B) HepG2 cells exposed to 25 mM glucose and PAT for 24 h (* p < 0.05 relative to control).

Glucose Availability Determines PAT-Associated Insulin Signaling Defects in HepG2 Cells
IR activation triggers phosphorylation of the receptor and IRS-1. Under NG conditions, PAT stimulated IRS-1 phosphorylation (pIRS-1) (Figure 2A; p = 0.0388), though pIR showed no statistical change (Figure 2A; p = 0.0775). This trend was potentiated in the presence of insulin shown by a significant two-fold increase in pIR ( Figure 2B; p = 0.0180) and pIRS-1 ( Figure 2B; p = 0.0145) in PAT-exposed NG i+ HepG2 cells, exhibiting a similar effect to the positive control, metformin.
This observation was dramatically reversed under hyperglycemic conditions. PAT caused significant two-fold decreases in both pIR ( Figure 2C; p = 0.0178) and pIRS-1 (Figure 2C; p = 0.0213) in HG HepG2 cells. This result was consistent following insulin pathway stimulation (HG i+ HepG2) where both pIR ( Figure 2D; p = 0.0182) and pIRS-1 ( Figure  2D; p = 0.0188) were decreased in PAT treatments. These findings are suggestive of defects in insulin signaling under HG conditions following PAT exposure. PAT alters expression of enzymes involved in glucose production in hyperglycemic conditions. qPCR was used to assess expression of glucose homeostasis enzymes PYGL, PFK-1 and PCK-1 in (A) HEK239 and (B) HepG2 cells exposed to 25 mM glucose and PAT for 24 h (* p < 0.05 relative to control).

Glucose Availability Determines PAT-Associated Insulin Signaling Defects in HepG2 Cells
IR activation triggers phosphorylation of the receptor and IRS-1. Under NG conditions, PAT stimulated IRS-1 phosphorylation (pIRS-1) (Figure 2A; p = 0.0388), though pIR showed no statistical change (Figure 2A; p = 0.0775). This trend was potentiated in the presence of insulin shown by a significant two-fold increase in pIR ( Figure 2B; p = 0.0180) and pIRS-1 ( Figure 2B; p = 0.0145) in PAT-exposed NG i+ HepG2 cells, exhibiting a similar effect to the positive control, metformin.
This observation was dramatically reversed under hyperglycemic conditions. PAT caused significant two-fold decreases in both pIR ( Figure 2C; p = 0.0178) and pIRS-1 ( Figure 2C; p = 0.0213) in HG HepG2 cells. This result was consistent following insulin pathway stimulation (HG i+ HepG2) where both pIR ( Figure 2D; p = 0.0182) and pIRS-1 ( Figure 2D; p = 0.0188) were decreased in PAT treatments. These findings are suggestive of defects in insulin signaling under HG conditions following PAT exposure.

PAT-Altered ERK/MAPK Signaling Is Influenced by Glucose and Insulin Availability
Insulin-activated ERK phosphorylation is connected to the downstream regulation of transcription factors involved in metabolic homeostasis. Under NG conditions, PAT increased ERK activation in both HEK293 and HepG2 cells evidenced by increased pERK relative to total ERK (NG HEK293 ( Figure 3A; pERK: p = 0.00610; ERK; p = 0.0032); NG HepG2 ( Figure 3C; pERK: p = 0.0072; ERK: p = 0.0788)).
PAT-altered ERK signaling trends were dissimilar to positive control metformin results, indicating a potential alternate mechanism of action. These data suggest that PAT stimulates ERK/MAPK signaling under NG conditions and inhibits signaling under HG conditions. These effects can, however, be counteracted by stimulating the insulin signaling pathway ( Figure 3).

PAT-Altered ERK/MAPK Signaling Is Influenced by Glucose and Insulin Availability
Insulin-activated ERK phosphorylation is connected to the downstream regulation of transcription factors involved in metabolic homeostasis. Under NG conditions, PAT increased ERK activation in both HEK293 and HepG2 cells evidenced by increased pERK relative to total ERK (NG HEK293 ( Figure 3A; pERK: p = 0.00610; ERK; p = 0.0032); NG HepG2 ( Figure 3C; pERK: p = 0.0072; ERK: p = 0.0788)).
PAT-altered ERK signaling trends were dissimilar to positive control metformin results, indicating a potential alternate mechanism of action. These data suggest that PAT stimulates ERK/MAPK signaling under NG conditions and inhibits signaling under HG conditions. These effects can, however, be counteracted by stimulating the insulin signaling pathway ( Figure 3).

HepG2 (
These results show PAT stimulated PI3K/Akt signaling under NG conditions and hibited PI3K/Akt activation under HG conditions despite insulin action. This indicat possible PAT-induced defect in response to elevated glucose levels and insulin-stimul signaling.  PAT-modified GSK-3 activation corresponded with PI3K/Akt signaling trends while GLUT2 expression was widely compromised by PAT. Western blotting established PAT significantly increased GSK-3 inhibition by phosphorylation relative to total GSK-3 expression with no change GLUT2 in NG HEK293 (A) and NG HepG2 (C) cells. This was reversed in HG HEK293 (B) and HG HepG2 (D) cells with decreases in GLUT2, pGSK-3, and increased GSK-3 following 24 h PAT exposure. NG i+ cells (E) exposed to PAT showed no significant changes while HG trends were maintained in the presence of insulin (F) (* p < 0.05 relative to control).

Figure 5.
PAT-modified GSK-3 activation corresponded with PI3K/Akt signaling trends while GLUT2 expression was widely compromised by PAT. Western blotting established PAT significantly increased GSK-3 inhibition by phosphorylation relative to total GSK-3 expression with no change GLUT2 in NG HEK293 (A) and NG HepG2 (C) cells. This was reversed in HG HEK293 (B) and HG HepG2 (D) cells with decreases in GLUT2, pGSK-3, and increased GSK-3 following 24 h PAT exposure. NG i+ cells (E) exposed to PAT showed no significant changes while HG trends were maintained in the presence of insulin (F) (* p < 0.05 relative to control).

PAT Contributes to Metabolic Inflexibility by PDK-1 Elevation and PDH Inhibition under NG and HG Conditions
Active PDH, stimulated by insulin, catalyzes the conversion of pyruvate to acetyl-CoA. Inhibition of PDH is catalyzed by PDK-mediated phosphorylation on the E1α subunit.
Elevated pPDH E1α suppresses aerobic pyruvate oxidation and can aggravate the Warburg effect. This pathway, characterized predominantly in cancerous cells, favors the anaerobic conversion of pyruvate to lactate. This may explain the distinct differences in PDH E1α results between HepG2 (cancerous) and HEK293 (non-cancerous) cells. Interestingly, lactate levels increased significantly in controls under all HG conditions relative to NG (Table 1; (HEK293; p = 0.0006) (HepG2; p = 0.0013) (HepG2 i+; p = 0.0190)). The highly significant increase observed in HEK293 cells relative to HepG2 cells may be related to metabolic differences in cancerous and non-cancerous cells as well as the pH sensitivity of HEK293 cells. Interestingly, a significant decrease in lactate levels was observed in PAT treatments relative to the controls in NG treatments and HEK293 cells (Table 1; NG HEK293 (p = 0.0222); HG HEK293 (p = 0.0069); NG HepG2 (p = 0.0142); NG i+ (p = 0.0069)). HG HepG2 and HG i+ HepG2 treatments however, (Table 1; NG HepG2 (p = 0.4336); NG i+ (p = 0.7427)) were not statistically changed. This lack of lactate elevation despite changes to the PDH axis in PAT treatments suggests pyruvate was possibly shunted back towards gluconeogenesis or to an alternative fate instead of reduction to lactate by LDH.

Discussion
The key finding in this study is that exposure to PAT impairs insulin signaling under HG conditions and alters metabolic flexibility after 24 h exposure in an in vitro model. This finding was consistent following insulin stimulation, suggestive of an insulinresistant mechanism.
A recent study showed PAT was cytotoxic to rat pancreatic β cells but did not affect insulin production [24,25] prompting examination of the insulin signaling cascade in this study. The IR, a tyrosine kinase, undergoes autophosphorylation catalyzing the phosphorylation and activation of downstream cellular proteins including IRS-1, the PI3K/Akt pathway and ERK/MAPK pathway. These pathways act in concert to positively regulate glycolysis, glucose storage as glycogen and lipids and repress glucose synthesis and release by inhibiting glycogenolysis and gluconeogenesis. Insulin action is attenuated by dephosphorylation of the receptor and its substrates. Phospho-tyrosine residues indicated PAT activated the insulin signaling cascade under NG conditions but suppressed activation under HG conditions in HepG2 cells (Figure 2). Defects in this pathway have been linked to impaired glucose tolerance and the pathogenesis of type 2 diabetes [22,28].
The ERK/MAPK pathway is activated in a stepwise sequence within the insulin signaling cascade. pERK catalyzes the activation of transcription factors to initiate cell proliferation. Inhibition of this pathway observed in PAT exposed HG HepG2 and HEK293 cells ( Figure 3) prevents insulin-stimulated cell growth. PAT-induced alterations in ERK signaling and proliferation have been reported in the literature, but metabolic consequences have been neglected [4,23]. The findings in this study imply PAT-mediated ERK changes are dependent on insulin and glucose availability (Figure 3). This may be of consequence to glucose metabolism under conditions of PAT-inflicted β-cell death [24].
Proximal insulin signaling activates the PI3K/Akt pathway in parallel to the ERK/MAPK pathway. pIRS-1 recruits PI3K to activate Akt by phosphorylation. pAkt-mediated phosphorylation of GSK-3 (ser 9) relieves GS inhibition and promotes glycogen synthesis. Under NG conditions, PAT-activated pIRS-1 (Figure 2), PI3K/Akt activation ( Figure 4) and increased pGSK-3 ( Figure 5) in HepG2 cells positively regulating glycogen synthesis. Under HG conditions, however, PAT diminished pIRS-1 (Figure 2) in HepG2 cells and PI3K/Akt (Figure 4) mediated GSK-3 ( Figure 5) phosphorylation in both HepG2 and HEK293 cells. These findings reveal distinct differences in PAT-mediated signaling depending on glucose availability. PAT-altered PI3K/Akt signaling has been reported previously in other cell lines [23]; the findings in this study, however, reveal novel metabolic implications. These trends identified in HepG2 cells were observed consistently in both the presence and absence of insulin, implicating a non-insulin dependent mechanism. Elevated total GSK-3 is associated with impaired glucose disposal and glycogen synthesis and type 2 diabetes [20]. Similar dysfunctions in reduced pIRS-1/PI3K/Akt signaling as observed in this study, have been associated with insulin resistance and impaired glucose tolerance [21,29].
Previous studies describe PAT-inflicted glycogen loss and serum glucose elevation in animals, postulating that PAT increases gluconeogenesis and glycogenolysis [15,30]. In this study, active GSK-3, a prominent feature in glycogenolysis, is elevated by PAT under HG conditions in both HEK293 and HepG2 cell lines ( Figure 5). PAT also increased the gene expression of glucose-producing enzymes PYGL and PCK-1 and inhibited glycolytic PFK-1 (Figure 1) under the same conditions in HepG2s. These classic indicators of glycogenolysis and gluconeogenesis, usually inhibited by insulin and HG conditions [28], were increased by PAT under the same conditions. Glucose homeostasis relies on the detection of varying glucose availability. GLUT2, expressed in the kidney and liver, plays a crucial role in glucose sensing and uptake. GLUT2 suppression, induced by PAT in both HEK293 and HepG2 cells ( Figure 5), is associated with cellular glucose efflux, impaired activity of glucose sensitive genes, reduced glucose uptake and promoting type 2 diabetes pathogenesis and organ damage [18,31]. While organ damage has been described in previous findings on PAT, this novel finding offers a possible explanation for the increased gluconeogenic markers, impaired glycogen synthesis and glycolysis in PAT treatments in this study.
These findings indicate that PAT compromised cellular metabolic processes in adaption to glucose availability. PDH, the principal enzyme controlling nutrient adaption and metabolic flexibility, is activated by insulin and glucose via dephosphorylation on the E1α subunit. PDH metabolizes pyruvate at the junction between glucose, fatty acid metabolism and the TCA cycle [17]. PDK-1 phosphorylates and inhibits PDH E1α under hypoxic conditions, shifting metabolism to glycolytic processes [32]. PAT significantly increased PDK-1 and pPDH E1α expression in HepG2 cells ( Figure 6). In cancerous cells such as HepG2s, this is associated with aggravation of the Warburg effect. The Warburg effect posits cancerous cells inhibit mitochondrial ATP production and increase rates of glycolysis and cytoplasmic conversion of pyruvate to lactate [33,34]. This pathway is linked to glucose scarcity, synthesis and inhibition of pyruvate conversion to acetyl-CoA. Mitochondrial inhibition in cancerous cells is achieved by PDK and pPDH elevation. These actions block excessive mitochondrial ROS production, linked extensively to PAT in previous studies [14]. Thus, PAT-induced pPDH and PDK-1 elevation may be an energy-and mitochondrial conservation strategy [4,10,11]. In addition, lactate levels were decreased in PAT treatments (Table 1), despite PDK-1 and pPDH E1α (Figure 6) elevation. While the lower levels of lactate in i+ treatments suggest insulin may have stimulated mitochondrial function, the high levels in the HEK293 control cells may explain the potent toxicity of PAT in the cell line as PAT loses biological activity in alkaline media [35,36]. Taken together, the results imply pyruvate did not undergo lactic acid fermentation as suggested in the Warburg effect theory. Alternate fates of pyruvate under PDH inhibition include the TCA cycle, fatty acid synthesis and gluconeogenesis [17]. While there is currently insufficient evidence in this study and the literature to conclusively rule out the TCA cycle or fatty acid synthesis these findings, together with gluconeogenic enzyme mRNA levels (Figure 1), GLUT2 and GSK-3 expression (Figure 5), infer pyruvate was shunted back toward gluconeogenesis despite the availability of glucose and insulin. Decreased PDH and metabolic inflexibility are initiating events in impaired glucose oxidation, associated with insulin resistance and type 2 diabetes [14,16,[37][38][39].
Type 2 diabetes is characterized by elevated glucose levels, impaired glucose tolerance, insulin resistance, defects in the insulin signaling pathway, oxidative stress and mitochondrial dysfunction [22]. The literature shows PAT causes oxidative stress and mitochondrial dysfunction which is associated with both causes and consequences of insulin resistance [10,11,26,40]. This study presents novel findings indicating PAT opposes insulin action under HG conditions, suppresses components of the insulin signaling pathway and prevents metabolic flexibility in HepG2 and HEK293 cells. Given that PAT targets the kidney and liver, and the role of these organs in systemic glucose and energy homeostasis, this has strong implications for PAT-induced toxicology and disorders of metabolism.

Conclusions
Our findings show PAT caused defects in the insulin signaling pathway under HG conditions, stimulated glucose production pathways and inhibited PDH activity, contributing to metabolic inflexibility. Collectively, these results suggest PAT exposure may be an initiating event in insulin resistance with an etiological role in pathogenesis of type 2 diabetes and other disorders of metabolism. Significant results were observed above and below the safety level of PAT.

Materials and Methods
HEK293 and HepG2 cells were obtained from the American Type Culture Collection (Johannesburg, South Africa). Culture reagents were purchased from Lonza BioWhittaker (Basel, Switzerland). Western Blotting and qPCR reagents and consumables were purchased from Bio-Rad (Hercules, CA, USA). Patulin (P1639) was purchased from Sigma-Aldrich (St. Louis, MO, USA). All other reagents were obtained from Merck (Darmstadt, Germany) unless otherwise stated.

Cell Culture
HepG2 cells were cultured in complete culture medium (CCM) containing Eagle's minimum essential medium (EMEM) supplemented with 2 mM l-glutamine, 1% pen-strepfungizone (500 units potassium penicillin, 500 µg streptomycin, 1.25 µg amphotericin B/5 mL flask) and 10% fetal bovine serum in 25 cm 3 flasks at 37 • C. HepG2 cells were derived from hepatocellular carcinoma. Despite the limitations of being a cancerous line (relative to primary human hepatocytes), HepG2 cells were used as they exhibit phenotypic stability, genotypic features of normal liver cells and exhibit high glucose consumption and active energy metabolism [41,42]. HEK293 cells were cultured in CCM in comprising Dulbecco's modified essential medium (DMEM) supplemented with 2 mM l-glutamine, 1% pen-strep-fungizone (500 units potassium penicillin, 500 µg streptomycin, 1.25 µg amphotericin B/5 mL flask), 10 mM HEPES and 10% fetal bovine serum in 25 cm 3 flasks at 37 • C; 5% CO 2 . HEK293 cells are established from primary human embryonic kidney and have the capacity to switch metabolic processes from glucose consumption and lactate production to glucose-lactate co-consumption [43]. This characteristic makes HEK293 cells an ideal model to study potential changes to glucose metabolism as a result of PAT exposure.

Dosage Information
PAT (5 mg) was dissolved in 1ml 100% dimethyl sulfoxide (DMSO) (32 mM PAT). Smaller PAT stock solutions were prepared in 0.1 M PBS to a final concentration of 1 mM. When 90% confluent, cells were serum starved for 16 h (h) and then treated with PAT (0.2 µM; 2.0 µM) in 5 mL CCM for 24 h. These PAT concentrations are below and above the safety level of PAT in consumables, respectively, and were determined from incidence and monitor studies [3,5]. These studies indicate that despite a safety level of 50 µg/L (0.3 µM), PAT exposure and levels in consumables can vary from 0.1 µM to 4.0 µM. Cells grown and treated under these conditions (normoglycemic: 5 mM glucose) are referred to as NG in the results.
To determine the effect of PAT on glucose tolerance, once 90% confluent HepG2 and HEK293 cells were serum starved for 16 h and exposed to high glucose CCM (HG) (25 mM) and PAT (0.2 µM; 2.0 µM) for 24 h.
Finally, to determine PAT effect on insulin signaling, 90% confluent HepG2 cells were serum starved for 16 h, exposed to NG (5 mM glucose) or HG (25 mM glucose) media and PAT (0.2 µM; 2.0 µM) for 24 h. Cells were then stimulated with 10 ng/mL (1.7 nM) insulin for 5 min before the media was removed. These are represented by NG i+ and HG i+ in the results. HEK293 cells show a limited insulin signaling profile and were excluded from this parameter in the study.
Metformin (5 mM) was used as a positive control in all protein expression experiments; this concentration was selected from the literature [44].
Results are representative of three independent experiments completed in triplicate.

Statistical Analysis
Results are represented as mean fold change ± standard deviation (SD) relative to normalized control. Statistical significance was assessed using one-way ANOVA with appropriate post hoc comparisons on GraphPad Version 5.0 Software. p values less than 0.05 were considered significant.
Author Contributions: Conceptualization, A.C. and Y.P.; methodology, Y.P. and S.N.; lab investigation and formal analysis, Y.P.; funding and resources, A.C. and Y.P.; writing-original draft preparation, Y.P.; writing-review and editing, A.C. and S.N.; supervision, A.C. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: Ethical review was waived as this study did not involve human or animal subjects.