The role of excess fructose intake in the pathogenesis of non-alcoholic fatty liver disease (NAFLD) has recently received increasing attention, but the pathophysiology of this relationship has been only partly elucidated.
The role of excess fructose intake in the pathogenesis of non-alcoholic fatty liver disease (NAFLD) has recently received increasing attention, but the pathophysiology of this relationship has been only partly elucidated.
To provide an overview of the potential role played by fructose in the pathogenesis of NAFLD by focusing on both indirect and direct harmful effects.
Experimental and clinical studies which investigated the relation of fructose with NAFLD are reviewed.
Several factors may potentially contribute to fructose-induced NAFLD, including the induction of the metabolic syndrome, copper deficiency, bacterial translocation from the gut to the liver, the formation of advanced glycation endproducts and a direct dysmetabolic effect on liver enzymes.
Experimentally-increased fructose intake recapitulates many of the pathophysiological characteristics of the metabolic syndrome in humans, which may in turn lead to NAFLD. However, the majority of experimental studies tend to involve feeding excessively high levels of fructose (60–70% of total energy intake) which is not reflective of average human intake. Hopefully, the combination of in vivo, in vitro and genetic research will provide substantial mechanistic evidence into the role of fructose in NAFLD development and its complications.
Non-alcoholic fatty liver disease (NAFLD) is a common but often silent chronic liver disease characterised by the accumulation of triglycerides in hepatocytes occurring in people who consume little or no alcohol.[1-3] This condition comprises a wide spectrum of histological lesions ranging from simple steatosis to non-alcoholic steatohepatitis (NASH), a parenchymal liver inflammation which can develop further to fibrosis, cirrhosis and hepatocellular carcinoma.[4, 5] NAFLD is strongly associated with obesity and insulin resistance and is currently conceptualised as the hepatic manifestation of metabolic syndrome (MS).[6-8] Growing evidence suggests that the fast-growing and alarming epidemic of NAFLD is closely intertwined with the Westernisation of dietary patterns with an increasing intake of simple sugars, especially fructose.[9-14]
Fructose is an isomer of glucose with a hydroxyl group on carbon-4 reversed in position. Recent decades have witnessed an enormous rise in fructose consumption. Studies on ancestral diets have shown that 8000–10 000 years ago the average intake of fructose per capita was around 2 kg per year. Significant increases observed after the abolition of sugar tax in the 1870's saw fructose consumption increase to 45 kg by 1950. Notably, by 1997 consumption had risen consistently to 69 kg per annum.[13, 16] Currently, fructose is massively used in the food industry due to its sweeter taste and its lack of inhibition of satiety compared with other sugars. From a metabolic standpoint, fructose is absorbed by the small intestine and is transported across the epithelial barrier into cells and the bloodstream by the fructose-specific GLUT5 transporter. Entry of fructose into cells is not insulin-dependent and does not promote insulin secretion unlike glucose. Absorbed fructose is transported in plasma via the hepatic portal vein to the liver where fructose is predominantly metabolised via its phosphorylation; only a small amount of fructose is metabolised by hexokinase in muscle and adipose tissue.[13, 16] Further metabolism of fructose produces 3-carbon intermediates in the glycolytic pathway that are identical to those derived from glucose.
Starting from the 1960s, a number of animal and human studies have reported associations with excessive fructose consumption and adverse metabolic effects, which may have important hepatic consequences (including the development of NAFLD). The potential role of fructose in the pathogenesis of NAFLD has thus recently gained mounting attention and has been previously reviewed in detail.[9-17] However, the extent to which fructose consumption is playing a role in liver damage and failure has not yet been investigated fully. Theoretically, fructose consumption can be linked to NAFLD via two main pathophysiological mechanisms: the first can be defined as ‘indirect’ (i.e. fructose can lead to prominent metabolic adverse effects which can in turn increase the risk of developing NAFLD), whereas the second one is more ‘direct’ and could involve a direct hepatotoxic damage (Figure 1).
This article provides an overview of the potential role played by fructose in the pathogenesis of NAFLD by focusing on both indirect and direct harmful effects. The list of pathophysiological mechanisms provided in this review is not intended to be exhaustive; rather, a brief summary of some key links between fructose and NAFLD is provided.
The MS is a cluster of well-defined metabolic disorders (obesity, hypertension, hyperglycaemia and dyslipidaemia) associated with the development of cardiovascular diseases.[18, 19] The MS and associated insulin resistance – which are universally considered the main risk factors for NAFLD – are massive health problems in Westernised countries.[20, 21] Several studies have shown that fructose may predispose to fatty liver infiltration by creating an adverse metabolic milieu. From an indirect standpoint, fructose may predispose to NAFLD by promoting an increase in fasting and postprandial triglycerides, which can in turn result in liver steatosis.[22-24] In addition, hypertriglyceridaemia caused by excessive fructose intake has been shown to be a precursor of insulin resistance in experimental studies.[25, 26] Although fructose feeding in the short term does not significantly increase insulin secretion, long-term administration can cause hyperinsulinaemia. Ackerman et al. have shown that rats given fructose-enriched diet had an increase in hepatic triglycerides and hepatic cholesterol, hypertriglyceridaemia and hypertension, ultimately leading to the development of macrovesicular and microvesicular fat deposits. Tetri et al. demonstrated that male C57BL/6 mice fed relevant amounts of a high-fructose corn syrup equivalent specifically developed impaired insulin sensitivity and severe hepatic steatosis associated necroinflammatory changes. The authors reasoned that feeding rats with high fructose doses can induce histological features of NASH in the context of a metabolic profile similar to that observed in the MS. Sánchez-Lozada et al. reported that fructose-fed rats develop early features of the MS when compared with control animals. Such changes in turn resulted in overt hepatic alterations, including increased hepatic triglyceride accumulation and fatty liver, an increase in uric acid content in the liver, as well as an increase in hepatic levels of proinflammatory markers such as monocyte chemoattractant protein-1 and tumour necrosis factor-alpha measured in liver homogenates. Interestingly, Botezelli et al. have shown that exercise may counteract the onset of insulin resistance in fructose-fed rats, a finding similar to that observed in MS patients. In a clinical study, Abdelmalek et al. analysed the association of fructose intake with metabolic and histological features of fatty liver in a large cohort of 427 adults with NAFLD. In keeping with the results obtained in animal studies,[27, 28] the authors found that increased fructose consumption in humans was univariately associated with several components of the MS, including hypertriglyceridaemia, low high-density lipoprotein cholesterol and hyperuricemia. In recent years, studies have also elucidated the molecular mechanisms underlying the onset of fructose-induced metabolic alterations. An important study by Nagai et al. demonstrated that the transcriptional factor peroxisome proliferator-activated receptor gamma coactivator-1 beta (PGC-1 β) plays a crucial role in the pathogenesis of fructose-induced insulin resistance in Sprague–Dawley rats. Based on their findings, the authors concluded that the pharmacological inhibition of PGC-1 β could play an important role for preventing fructose-induced hypertriglyceridaemia and insulin resistance. Compelling evidence from both experimental and animal studies was therefore found that high fructose consumption can lead to insulin resistance and associated metabolic disorders (especially hypertension, hyperuricaemia and dyslipidaemia), all of which appear to be underlying mechanisms involved in NAFLD.
Continuous fructose ingestion may impose a metabolic burden on the liver through the induction of fructokinase and fatty acid synthase. In the liver, fructose is metabolised to fructose-1-phosphate by fructokinase, which consumes ATP.[13, 16] As a consequence, a massive incorporation of fructose into liver metabolism can lead to high levels of metabolic stress via ATP depletion. In an experimental study in the rat, Vilà et al. have shown that fructose-induced fructokinase hyperexpression in the liver can be reduced (by 0.6-fold) by the hydroxymethyl-glutaryl-CoA reductase inhibitor atorvastatin. Of note, clinical studies have shown that atorvastatin can improve liver injury in NAFLD patients with hyperlipidaemia.[34, 35] Fatty acid synthase catalyses the last step in the fatty acid biosynthetic pathway and is a key determinant of the maximal capacity of the liver to synthesise fatty acids by de novo lipogenesis. In a clinical study, Ouyang et al. reported that increased fructose consumption in patients with NAFLD is associated with hyperexpression of hepatic mRNA for fatty acid synthase, suggesting that this molecular derangement could play a crucial role in fructose-induced fatty liver infiltration.
Copper is a redox-active transition metal which has been implicated in fatty liver infiltration through mechanisms involving redox-sensitive signalling pathways. NAFLD patients have been shown to have lower hepatic copper concentrations than control subjects and should be therefore considered copper deficient. In addition, patients with NASH have lower hepatic copper concentrations than those with simple steatosis. In experimental animals, restriction of dietary copper induces hepatic steatosis and insulin resistance. Recently, evidence has suggested that a high fructose intake may be one of the potential mechanisms of copper deficiency in NAFLD. In this regard, Song et al. showed that fructose feeding impaired copper status in Sprague–Dawley rats. Of note, liver injury and fat infiltration were significantly induced in marginal copper deficient rats exposed to fructose as evidenced by robustly increased plasma aspartate aminotransferase and hepatic fatty infiltration. The hepatic antioxidant defence system was suppressed and lipid peroxidation was increased by marginal copper deficiency and fructose feeding. In addition, the authors found that fructose feeding led to iron overload, a common finding in the setting of NAFLD. Taken together, the results indicated that high fructose-induced NAFLD may be due, at least in part, to decreased copper absorption, and subsequent copper deficiency. As an essential trace element, copper is required as a co-factor for several enzymes, e.g. cytochrome c oxidase, copper-zinc superoxide dismutase, ceruloplasmin and tyrosinase. However, copper seems to be implicated in fructose-induced NAFLD mainly via its redox-activity.[38-40] Besides affecting copper metabolism, fructose has been shown increase iron retention and absorption in experimental studies. Such subtle alterations in iron metabolism may have as well important pathophysiological effects in NAFLD as iron is one of the putative elements that interacts with oxygen radicals in inducing hepatoxicity and insulin resistance.
Endotoxin, also referred to as lypopolysaccharide, represents the outer cell wall membrane of Gram-negative bacteria. Increased endotoxin levels have been previously observed in conditions which disrupt intestinal barrier function and have been associated with subtle chronic inflammation and metabolic abnormalities. Evidence suggests that fructose ingestion is associated with intestinal bacterial dysbiosis and increased gut permeability; the resulting low-grade endotoxaemia may in turn ignite chronic inflammation and immune dysregulation in the liver. In an experimental study, Bergheim et al. reported that endotoxin levels in portal blood are significantly higher in fructose-fed mice. Notably, concomitant treatment of fructose-fed mice with antibiotics (e.g. polymyxin B and neomycin) markedly reduced hepatic lipid accumulation in fructose-fed animals by lowering endotoxin levels. The authors concluded that high fructose consumption may not only lead to liver damage through overfeeding but also may exert a proinflammatory action by increasing the intestinal translocation of endotoxin. In a pathophysiological study, Spruss et al. subsequently confirmed the hypothesis that the onset of fructose-induced NAFLD in mice is associated with intestinal bacterial overgrowth and increased intestinal permeability, ultimately leading to an endotoxin-dependent activation of hepatic Kupffer cells. Haub et al. subsequently reported that a loss of intestinal serotonin reuptake transporter is a critical factor in fructose-induced impairment of intestinal barrier function and subsequent endotoxin-induced experimental steatosis. Interestingly, Volynets et al. also reported that bile acids prevent fructose-induced hepatic steatosis in mice through mechanisms involving protection against the fructose-induced translocation of intestinal bacterial endotoxin. From a molecular standpoint, Wagnerberger et al. reported that several toll-like receptors (TLR) – a series of genetically conserved and stable cell-surface receptors that have emerged as key players in the initiation of cellular innate immune responses in response to endotoxin – may be involved in the onset of fructose-induced NAFLD. TLR activation by endotoxin leads to the release of various host defence mediators from hepatic immune cells which may influence lipid metabolism, insulin signalling and cell survival, ultimately modulating the inflammatory response in the hepatic parenchyma. The potential role of endotoxaemia in the pathogenesis of human NAFLD has been elucidated in the clinical setting by Thuy et al., who showed that endotoxin plasma concentrations are significantly higher in NAFLD patients than in controls and positively associated with dietary intakes of carbohydrates, including fructose. The authors speculated that low-grade endotoxaemia in human NAFLD could possibly be the result of fructose-induced increased gut permeability. Importantly, a high burden of fructose-induced bacterial translocation may activate a proinflammatory secretory phenotype in visceral adipose tissue depots via innate immune pathways.
Adiposity has been traditionally divided into two main compartments comprising subcutaneous (SAT) and visceral/intraabdominal (VAT) fat with the former located under the skin and the latter surrounding the internal organs of the abdominal viscera. VAT is generally considered more metabolically adverse than SAT, and its presence has been repeatedly associated with diabetes, inflammation and cardiovascular disease. In addition, waist circumference, a surrogate anthropometric measure for VAT, is one of the diagnostic criteria for the MS. Recently, several studies have attempted to elucidate the relative importance of visceral fat type and subcutaneous fat types as a risk factor for NAFLD. In this regard, Sogabe and co-workers have shown that men with visceral fat type MS are more likely to have dyslipidaemia, fatty liver, and liver dysfunction than those with subcutaneous fat type MS. In a study of subjects who underwent a health check-up and measurement of visceral fat accumulation, Eguchi et al. demonstrated that visceral fat accumulation in patients with advanced NASH was greater than that in patients with early NASH. However, an inherent caveat of the investigation was the lack of assessment of SAT. Jun et al. investigated whether or not visceral, subcutaneous abdominal, intramuscular and subcutaneous thigh adipose tissue could be associated with NAFLD. The authors found that a larger subcutaneous fat area was negatively associated with NAFLD after adjustment for visceral fat and abdominal subcutaneous fat areas in women, but not in men. This result suggested that low femoral subcutaneous fat amounts were an independent correlate of fatty liver disease in women. In a small clinical study of 35 obese patients undergoing bariatric surgery, O'Connell et al. reported that both omental and subcutaneous adipocyte size measured in biopsy sample were significantly associated with the degree of liver steatosis. However, only omental adipocyte size was an independent predictor of the presence or absence of fibrosis. Taken together, the results of the available studies suggest that both VAT, SAT and other fat compartments of interest should be included to further investigate the association between specific fat depots and liver fat content. The evidence that the metabolic perturbations resulting from fructose consumption may result in an altered fat partitioning in visceral and subcutaneous compartments remains controversial. Dietary fructose may cause an increase in visceral adiposity in overweight/obese adults. However, the addition of 150 g of fructose or glucose per day for 4 weeks to a balanced weight-maintaining diet has no effect on visceral fat, subcutaneous abdominal fat and intramyocellular lipids of the tibialis anterior muscle in 20 healthy subjects. In contrast with these findings, Pollock et al. recently reported that the association of higher fructose consumption with multiple markers of cardiometabolic risk in adolescents is mediated by visceral obesity. In a cross-sectional analysis of 791 non-Hispanic white men and women, Odegaard et al. examined how fructose-sweetened beverage consumption habits could influence overall and abdominal adiposity measures. The authors observed significant increases in waist circumference and the proportion of visceral to subcutaneous abdominal adipose tissue, with no change in total body fat or body mass index with increasing frequency of fructose-sweetened beverages. The authors concluded that fructose-sweetened beverage consumption habits are associated with a more adverse abdominal adipose tissue deposition pattern. Based on these findings, there is a clear lack of definite evidence about the complex interplay between fructose consumption, alterations in SAT and VAT, and the risk of NAFLD. Additional prospective studies may be helpful to shed more light on this intricate issue.
Recent years have witnessed an enhanced interest in ectopic fat as one of missing link between the MS and the development of cardiovascular disease. Epicardial adipose tissue (EAT) is a type of visceral fat adjacent to the heart in the thoracic cavity, located between the surface of myocardium and the epicardium. Interestingly, the coronary arteries are embedded in EAT for the majority of their course. Coronary arteries adjacent to EAT appear to be more susceptible to develop atherosclerotic plaques and it is noteworthy that intramyocardial coronary artery segments do not generally develop atherosclerosis. Most importantly, EAT functions as an active secretory organ that releases hormones, inflammatory cytokines and chemokines in the bloodstream; therefore, it is currently conceptualised as a type of secretory visceral fat that may have a systemic metabolic impact. We have recently shown that EAT thickness is significantly higher in patients with NAFLD than in controls, a finding mirrored by an increase in EAT-derived adipokines vaspin and chemerin. In addition, we found that EAT thickness, serum vaspin and the severity of liver fibrosis were independent predictors of coronary flow reserve – a surrogate marker of coronary endothelial function – in NAFLD patients. In turn, Axelsen et al. have recently shown that Sprague–Dawley rats fed a 60 kcal/100 kcal fat diet with 10% fructose in the drinking water show significant intramyocardial lipid accumulation, ultimately resulting in cardiac hypertrophy. These preliminary data should prompt more research into the link between fructose intake, EAT thickness and increased fat accumulation in the liver.
The role played by active brown adipose tissue (BAT) – a key thermogenic tissue – in adult humans has recently become a hot topic in human metabolic disorders.[71, 72] Although the major depot of BAT in newborn humans (between the shoulder blades) regresses rapidly after birth, the presence of other deposits of functional BAT in adult humans has been widely confirmed. Evidence suggests an inverse correlation of BAT with both body mass index and body fat percentage.[74, 75] In general, BAT abundance is significantly lower in obese subjects compared with that in lean individuals. In addition, BAT activity is negatively correlated with both glucose and age, and women display higher activity than men. Recently, we have shown that the absence of BAT in adult life is associated with an increased risk of fatty liver as defined by computed tomography (CT) results. In addition, BAT activity was found to be inversely associated CT measures of hepatic fat content. It was concluded that presence of BAT in adulthood is independently associated with a lower likelihood of CT-proven NAFLD. A classical animal study by Boll et al. demonstrated that the activities of lipogenic enzymes in rat BAT can be influenced by high-carbohydrate diets, including high fructose feeding. In addition, Young et al. have shown that glucose and fructose, but not galactose, can activate the sympathetic nervous system in rat interscapular BAT. The authors concluded that oral intake of fructose can activate sympathetic mechanisms in specific adipose tissue depots which may contribute to insulin resistance under some circumstances. Studies on the association between fructose intake and the presence of BAT in adult life are eagerly awaited.
Direct evidence of fructose involvement in hepatic damage comes from individuals with hereditary fructose intolerance, an inborn error of fructose metabolism caused by a deficiency of the glycolytic enzyme, aldolase B, in the liver and kidney. Central to the pathogenesis of the disease is the depletion of the intracellular phosphate pool and the accumulation of fructose-1-phosphate in the hepatocytes; the latter phenomenon leads to the formation of toxic intracellular precipitates which ultimately result in hepatocyte cell death and liver failure.Although hereditary fructose intolerance is rare from an epidemiological standpoint, it has been hypothesised that functional common polymorphisms of the aldolase B gene may reduce the activity of the enzyme to varying degrees in the general population. Although the carriage of these polymorphisms is unlikely to cause major health problems, they may potentially be one of the multiple genetic factors that influence NAFLD risk, especially in people consuming large amounts of fructose.
Accumulation of tissue advanced glycation endproducts (AGEs) in long-lived tissues is an important marker of cumulative oxidative stress. Fructose is a potent reducing sugar and has the capacity to glycate proteins at a rate 17 times faster than glucose. AGEs are protein adducts formed when fructose is non-enzymically linked to proteins via the Maillard reaction. Dietary AGEs caused by fructose can interact with tissues and serum proteins and interfere with cellular mechanisms. Recently, elevated levels of serum AGEs have been reported in patients with NASH; in addition, AGEs seem to be involved in the generation of reactive oxygen species followed by the proliferation and activation of hepatic stellate cells, a major contributor to liver fibrosis. Importantly, atorvastatin has been shown to decrease the serum levels of AGEs in NASH patients with dyslipidaemia, suggesting that a reduction in AGEs can serve as a biomarker for the attenuation of NASH. Recently, Gaens et al. have reported that the accumulation of the AGE N(ε)-(carboxymethyl)lysine in the liver is significantly associated with the grade of hepatic steatosis, the grade of hepatic inflammation and gene expression levels of inflammatory markers in liver biopsies of obese individuals. However, the degree of fructose intake was not directly measured. It is feasible to speculate that rate of both AGEs formation and intracellular accumulation induced by fructose may contribute to liver tissue pathology in NAFLD. Of interest, AGE-mediated damage may be neutralised by soluble RAGE (sRAGE), a circulating truncated variant of the RAGE isoform that may compete with cell surface RAGE for ligand binding; notably, sRAGE levels have been shown to be reduced in NAFLD. Importantly, Takeuchi et al. have recently developed an assay for immunological detection of fructose-induced AGEs. Immunohistochemical studies of fructose-induced AGEs in liver biopsy are needed to shed more light on this issue.
On the basis of the data reviewed it can be acknowledged that experimentally increased fructose intake recapitulates many of the pathophysiological characteristics of the MS in humans. A state of hypertriglyceridaemia in particular seems to be one of the most consistently replicated fructose-induced metabolic alteration. This is believed to be one main reason for considering fructose consumption as a risk factor for the MS and NAFLD. It should be noted, however, that the majority of experimental studies tend to involve feeding excessively high levels of fructose (60–70% of total energy intake) which is not reflective of average human intake in Western diets. Recent years have witnessed a great alarmism about the metabolic risks of the introduction and mass commercial production of high fructose corn syrups (HCFS), which is a mixture of fructose and glucose in varying amounts produced by enzymatic isomerisation of dextrose to fructose.[89, 90] Three main varieties of HFCS (HFCS-45, HFCS-55 and HFCS-95 – the numerical value indicating the percentage of fructose relative to glucose within the HFCS) are currently available. The use of HFCS has increased by 1000% between 1970 and 1990, subsequently raising the consumption of added or extrinsic sugars. This increase has been linked to the continuing obesity epidemic. Although HFCS have received most of the attention as a ‘sweet hepatic toxin’, high fruit juice intake has been also linked to childhood obesity.Although the American Heart Association has called for a reduction in added sugars intake to reduce the burden of obesity and type 2 diabetes (which may in turn reduce the onset on NALFD), no clinical human study has directly compared the hepatotoxic effects of HFCS with those of fruit-, honey-, or vegetable- derived fructose. This issue is not trivial for designing future dietary guidelines to help quell the incidence of NAFLD. It is also noteworthy that the majority of modern fruits and vegetables have been bred to have much higher fructose content than the wild plants they are descended from. However, it can be hypothesised that the potentially hepatotoxic effects of natural fructose from fruits and honey may be naturally counterbalanced as these foods also have important antioxidant and cytoprotective properties due to their high content in simple phenolic compounds. Although fructose generates a markedly higher amount of reactive oxygen species than glucose, this phenomenon may be partially attenuated by other antioxidants present in fruits, vegetables and honey.
Another point that merits consideration is that the mechanistic aspects of gender-specific metabolic effects of fructose should be analysed in detail. In this regard, it should be noted that an association between fructose consumption and an increased risk of type 2 diabetes mellitus has been reported in women but not in men. This effect may be due to lower leptin concentrations and higher ghrelin levels elicited by fructose in women, ultimately resulting in a reduced satiety feeling. A more complete understanding of how a high fructose diet affects mechanisms that cause hepatic insult and NAFLD may aid in the prevention of steatohepatitis, fibrosis and severe progressive liver damage. However, NAFLD is clearly multifactorial in nature and to pinpoint one mechanism or pathway linking fructose, the MS and later NAFLD development is not possible with many mechanisms explored to date still needing additional investigation. The present review has started to elucidate possible indirect and direct pathophysiological routes by which the development of NAFLD may be linked to high fructose intake, but leaves more questions unresolved than answered. Experimental studies have shown that species-related variations for metabolic effects of dietary fructose exist. Similarly, epidemiological and clinical studies have not produced definite data regarding the hepatic correlates of high fructose consumption in humans. Several factors may potentially contribute to fructose-induce hepatotoxic injury, including copper deficiency, the formation of AGEs, a direct dysmetabolic effect on liver enzymes and bacterial translocation from the gut to the liver. Microbial translocation and endotoxaemia can also increase and trigger metabolic disorders that eventually enhance ectopic fat deposition.
In conclusion, there is sufficient evidence from both clinical and human studies that fructose may be one of the key players in the pathogenesis and development of NAFLD, potentially acting as a silent ‘sweet killer’. Future studies should attempt to identify the cut-off level of fructose consumption where concrete risks of developing metabolic complications exist. This is a crucial point around which any future research agenda on fructose and NAFLD might be designed. Indeed, it seems still difficult to translate the evidence derived from experimental and clinical studies to everyday dietary recommendations. In addition, the combination of in vivo, in vitro and genetic research will hopefully provide substantial mechanistic evidence to identify causal regulatory pathways in the pathogenesis and role of fructose in NAFLD development and its complications.
Declaration of personal and funding interests: None.