Molecular Aspects of Fructose Metabolism and Metabolic Disease

Abstract

Excessive sugar consumption is increasingly considered a contributor to the emerging epidemics of obesity and associated cardiometabolic disease. Sugar is added to the diet in the form of sucrose or high-fructose corn syrup, both of which are comprised of nearly equal amounts of glucose and fructose. Unique aspects of fructose metabolism and properties of fructose-derived metabolites allow for fructose to serve as a physiological signal of normal dietary sugar consumption. However, when fructose is consumed in excess, these unique properties may contribute to the pathogenesis of cardiometabolic disease. Here, we review the biochemistry, genetics and physiology of fructose metabolism and consider mechanisms by which excessive fructose consumption may contribute to metabolic disease. Lastly, we consider new therapeutic options for the treatment of metabolic disease based upon this knowledge.

Introduction:

The fixation of carbon dioxide into sugars via photosynthesis forms the basis of Earth’s food web. Heterotrophs cannot synthesize their own sugars and therefore evolved to use plant-derived sugars for energetic and synthetic purposes. Sucrose, the predominant circulating sugar in plants, is composed of equal parts glucose and fructose. Glucose is the major circulating sugar in animals, and the amount of fructose is negligible by comparison. Glucose is a primary energetic and synthetic fuel for most tissues and cell types in the body. In contrast, the fructose component of ingested sucrose is rapidly cleared by the intestines and liver and catabolized for energetic purposes, converted to glucose and its polymeric storage form – glycogen, or converted to fatty acids and stored as triglycerides. Despite its low levels in the circulation, fructose serves as a signal of ingested dietary sugar. The differences by which glucose and fructose are sensed and metabolized are of fundamental importance to the responses to normal physiological sugar consumption, as well as pathophysiological consequences of excessive ingestion.

We are amid a major emergence of obesity, diabetes and associated cardiometabolic diseases, including non-alcoholic fatty liver disease (NAFLD). The etiology of these epidemics is multifactorial and depends on interactions between genetics and environmental factors including diet and physical activity. Concerns that excessive consumption of sugary foods and beverages might contribute to the development of obesity and diabetes date as far back as the first millennium BCE and remerge periodically throughout modern history (Emerson and Larimore, 1924; Johnson et al., 2017). In the last half-century, enhanced industrial processes and corn subsidies increased the availability of sugar, particularly in the form of sweetened beverages. This has correlated with an explosion in the prevalence of cardiometabolic diseases refocusing public health attention on this issue (Bray et al., 2004; Malik et al., 2010).

Sugar is added to beverages and food products in the form sucrose or the industrial product high-fructose corn syrup (HFCS), which is manufactured by digestion by microbial enzymes of corn starch to its glucose monosaccharides. Microbial xylose isomerase then catalyzes the conversion of some portion of the glucose monosaccharides to fructose. “HFCS 55”, the most common formulation added to beverages, consists of 55% fructose and 45% glucose. Sucrose is ingested as a disaccharide and requires cleavage to its monosaccharide constituents, glucose and fructose, prior to absorption. This is catalyzed by the enzyme sucrase-isomaltase which is localized to the intestinal lumen’s brush border membrane (Hauri et al., 1979). In contrast, the sugars in HFCS are ingested as monosaccharides and require no further processing prior to absorption. The differences in mono- vs disaccharide content of HFCS versus sucrose could differentially affect the rate of absorption of these sugars, which might result in distinct biological effects (Le et al., 2012). However, compelling data that sucrose and HFCS impart distinct metabolic effects in humans is lacking (Stanhope et al., 2008).

The contribution of natural dietary sugar and added sugars to the epidemics of cardiometabolic disease remains controversial (Khan and Sievenpiper, 2016; Stanhope, 2016; Ter Horst and Serlie, 2017). The epidemiological correlations between different sugar exposures and outcomes including energy intake, weight gain, type 2 diabetes (T2D), dyslipidemia, insulin resistance, hypertension, hyperuricemia and dental caries tend to be less consistent for measurements of total sugars and more consistent for measures of free and added sugars (Dhingra et al., 2007; Green et al., 2014; Haslam et al., 2020; Imamura et al., 2015; Jayalath et al., 2015; Khan and Sievenpiper, 2016; de Koning et al., 2012; Mela and Woolner, 2018). Consumption of fruit, the major natural source of dietary sugar, associates with healthful outcomes although fruit consumption often associates with other healthy lifestyle choices (Khan and Sievenpiper, 2016). Sugar-sweetened beverages (SSBs) are the largest contributor to added sugar and fructose intake in children and adults (U.S. Department of Agriculture and U.S. Department of and Health and Human Services; Virani et al., 2020), and their consumption consistently associates with increased cardiometabolic risk (Dhingra et al., 2007; Green et al., 2014; Haslam et al., 2020; de Koning et al., 2012; Ma et al., 2015, 2016; Malik et al., 2013; McKeown et al., 2018; Xi et al., 2015; Yoshida et al., 2007; Zhang et al., 2020), including the excess cardiometabolic deaths associated with poor diet (Coffee and Caffeine Genetics Consortium et al., 2015).

Recommendations from public health organizations and policy makers to reduce sugar consumption have met with modest success over the last decade. Almost 50% of adults and 100% of youths report consuming at least one SSB per day, a level of sugar consumption that associates with adverse cardiometabolic risk factors (Bailey et al., 2018; Rosinger et al., 2017; Virani et al., 2020). Moreover, consumption of added sugars even at the upper limits of recent dietary recommendations may adversely impact health (Stanhope et al., 2011). Populations with greater food insecurity and certain ethnic or racial subgroups report even higher intakes of SSB (Virani et al., 2020; Zagorsky and Smith, 2020). SSB consumption has increased during the COVID-19 pandemic (Flanagan et al., 2021). While public health efforts to reduce sugar consumption are warranted, improved understanding of the mechanisms by which SSBs and its constituent sugars contribute to disease are paving the way for new therapeutic strategies. These mechanisms and strategies will be the focus of this review.

Sweet taste perception

Animals have evolved complex mechanisms to sense sugar and motivate its consumption. Monosaccharides including glucose and fructose as well as disaccharides such as sucrose potently activate the G-protein coupled receptors, Tas1R2 and Tas1R3, which are located on epithelial cells of the tongue and palate (Nelson et al., 2001; Zhao et al., 2003). These receptors stimulate neuronal circuitry projecting to regions of the amygdala involved in processing hedonic and aversive stimuli (Wang et al., 2018). Some investigators have proposed that sugar has addictive properties by modulating mesolimbic dopaminergic reward circuitry (Avena et al., 2008; Leigh and Morris, 2018; Lustig, 2010). Indeed, the sweetness of both caloric and non-caloric sweeteners can induce or increase consumption of neutral or aversive substances, such as alcohol in strains of mice that otherwise avoid it (Yoneyama et al., 2008). Nevertheless, compelling data for the ‘addictive’ potential of sugar in humans is lacking (Leigh and Morris, 2018; Markus et al., 2017).

Mechanisms in addition to sweet taste contribute to the motivation for sugary food consumption. Mice with deficiency of both Tas1R2 and Tas1R3 cannot taste sugar or other sweeteners yet learn to prefer foods that are high in glucose content, but not fructose nor other non-caloric sweeteners (de Araujo et al., 2008; Sclafani et al., 2014). This form of sugar preference requires the intestinal sodium-glucose cotransporter (SLC5A1, also known SGLT1), the principal transporter expressed by both enterocytes and enteroendocrine cells that is responsible for uptake of glucose and galactose across the luminal intestinal membrane (Dyer et al., 2005; Gorboulev et al., 2012; Tan et al., 2020). The SGLT1-dependent effects on sugar consumption are mediated by signals carried by the vagus nerve to activate neurons in the caudal nucleus of the solitary tact, a region previously implicated in transducing gut sensations to the brain (Han et al., 2018; Kaelberer et al., 2018; Tan et al., 2020). However, recent work in an alternative “tasteless” genetic mouse model missing the P2X2 and P2X3 purinergic receptors that are required for taste cell signaling to gustatory nerves suggest alternative forms of post-ingestive sugar sensing (Andres-Hernando et al., 2020a; Sclafani and Ackroff, 2021). These studies show that fructose can enhance sugar consumption independently of taste. The combination of sweet taste sensation and post-ingestive sugar sensing likely interact to robustly reinforce preferences for sugar rich foods.

Excessive fructose consumption can cause metabolic disease

In the 1960’s, observations that diets high in fructose can induce hypertriglyceridemia in animals and humans in a matter of days much more robustly than diets containing comparable amounts of starch or glucose led to interest in fructose as a contributor to metabolic disease (Macdonald, 1966; Nikkilä and Ojala, 1965). Carbohydrate- and sugar-induced hypertriglyceridemia was also strongly associated with hyperinsulinemia in humans and in animal models of obesity and overnutrition (Eaton and Kipnis, 1969; Farquhar et al., 1966; Reaven et al., 1979; Reiser and Hallfrisch, 1977). Subsequent studies showed that very high-fructose diets could induce numerous dysmetabolic features including hypertension that were typical of the “metabolic syndrome”, more recently known as a syndrome of cardiometabolic disease (Hwang et al., 1987; Reaven, 1988; Sleder et al., 1980; Tobey et al., 1981). While these studies were instrumental in demonstrating the potential for fructose as a cause of metabolic disease, they often used fructose exposures that exceeded 50% of total energy, far exceeding typical consumption by humans. Fructose accounts for ~ 9% of energy intake in the US and even consumers in the 95^th percentile consume ~ 15% of their energy as fructose (Marriott et al., 2009). Nevertheless, diets unphysiologically high in fructose may be useful for identifying molecular mechanisms relevant for common forms of diet-induced metabolic disease.

In recent decades, interventional studies spanning weeks to months have shown that overfeeding moderate to high doses of sugars containing fructose can adversely impact metabolic outcomes in humans. In the longest such interventional study, Maersk and colleagues found that adding one liter of SSB daily for 6 months increased visceral, liver and ectopic fat (Maersk et al., 2012). These increases were not observed in those consuming isocaloric skim milk, non-caloric diet soda or water (Maersk et al., 2012). Adding SSBs ranging from 10% to 25% of energy requirements dose-dependently increased cardiovascular risk factors including lipids and uric acid over a two-week period (Stanhope et al., 2015).

With respect to potential adverse effects of fructose, Taskinen et al. reported that a moderate increase in fructose consumption (75 g or 300 kcal per day) in men with obesity for 12 weeks increased body weight, liver fat, hepatic de novo lipogenesis (DNL) and other cardiovascular risk factors (Taskinen et al., 2017). A pivotal study by Stanhope et al. compared the effects of adding isocaloric glucose- and fructose-sweetened beverages constituting 25% of basal energy requirement over 10 weeks in adults who were overweight or obese (Stanhope et al., 2009). Both interventions increased body weight to a similar degree. However, fructose, but not glucose, increased visceral adiposity, DNL, an atherogenic dyslipidemia and indices of insulin resistance (Stanhope et al., 2009). Whereas this study provides evidence that added fructose is more deleterious than added glucose, a recent report from the same group revealed that fructose and glucose in combination may be more harmful than fructose alone (Hieronimus et al., 2020). Fructose supplementation to 25% of energy requirement produced the strongest effect to increase circulating triglycerides, but the combination of fructose and glucose in HFCS elicited a greater increase in LDL-cholesterol and circulating apolipoprotein B levels. This result demonstrates the potential for complex interactions between distinct sugars. Similar interactions also likely occur between sugars and other dietary macronutrients, including fat and protein to impact metabolic health (Softic et al., 2018).

The most compelling clinical evidence that dietary sugar, when consumed in amounts typical of Western diets, contributes to adverse metabolic health comes from restriction studies, particularly those performed in children and adolescents. Some, but not all studies show that interventions aimed at reducing SSB consumption reduce weight gain, adiposity, liver fat and indices of insulin resistance, depending on the study population, the degree of sugar restriction and the duration of the study (Ebbeling et al., 2012; de Ruyter et al., 2012; Schwarz et al., 2017; Schwimmer et al., 2019).

While the evidence that added sugars containing fructose can negatively impact health indices is compelling, whether these adverse effects are mediated exclusively through the effects of increased energy intake to enhance weight gain or also through mechanisms independent of adiposity is less clear. Complexity, cost and ethical issues limit large-scale, long-term dietary interventions that are needed to confirm effects on metabolic and cardiovascular outcomes and to fully disentangle the effects of added sugar on weight gain and adiposity from other deleterious effects. Because of the difficulty in conducting long-term dietary studies in ‘free range’ humans, studies in animal models remain valuable. Non-human primates with physiologies that closely approximate our own may be an important alternative model. Indeed, supplementation of diets of rhesus macaques with 300 kcal per day of fructose-sweetened beverages, similar to that of the Taskinen et al. study, for up to one year produced many features of the metabolic syndrome, including increased body weight, fat mass, insulin resistance, dyslipidemia with hypertriglyceridemia and decreased HDL cholesterol, and in some cases overt diabetes (Bremer et al., 2011). Fructose supplementation in non-human primates can induce dyslipidemia, steatosis and the progression of NAFLD (Butler et al., 2019, 2020; Cydylo et al., 2017; Kavanagh et al., 2013) and may provide a more faithful model of the pathogenesis of human cardiometabolic disease.

While the literature largely focuses on the health risks of increasing amounts of sugar, the mode of sugar exposure may also affect its biological impact. In animal studies, sugar provided in liquid form may be more deleterious than when incorporated into solid food (Jang et al., 2020; Togo et al., 2019). Additionally, sugar ingested as a single large daily bolus may be more detrimental than frequent ingestion of smaller amounts of fructose (Jang et al., 2020). These variables are not typically assessed in current epidemiological studies and may contribute to the variable associations of dietary sugar exposures with cardiometabolic health.

Reduced physical activity is another environmental change often invoked to explain our current obesity and diabetes epidemics. Exercise acutely increases energy requirements and mobilizes metabolic substrates to support muscle contraction. Based on evidence that fructose provides some energetic advantage when supplied during exercise, Tappy and Rosset speculated that the adverse effects of fructose are magnified in sedentary people (Tappy, 2018; Tappy and Rosset, 2017). Spontaneous running attenuates sucrose-induced hypertriglyceridemia in rats (Zavaroni et al., 1981). Modern hunter-gatherers of the Hadza tribe consume from 8% to 16% of their energy intake as honey, which is largely fructose and glucose (Pontzer et al., 2012). This is similar to the 50^th and 95^th percentile of sugar consumers, respectively, in Western societies. Yet Hadza foragers have low prevalence of cardiovascular risk factors (Pontzer et al., 2012; Raichlen et al., 2017). Their high levels of physical activity may mitigate the adverse effects of sugar consumption.

Because of the growing concerns regarding the role of sugar consumption in cardiometabolic disease, public health organizations have invested in measures aimed at reducing sugar consumption, including dietary recommendations concerning ‘safe’ limits for sugar consumption (Johnson et al., 2009; U.S. Department of Agriculture and U.S. Department of and Health and Human Services; World Health Organization, 2015). Given the lack of consensus regarding health risks and safe thresholds, recommendations regarding sugar consumption vary substantially. Aside from dietary recommendations, public health studies have largely focused on behavioral interventions to reduce SSB and sugar consumption in at-risk school age children with modest benefits (Abdel Rahman et al., 2018; Vercammen et al., 2018). Interventions aimed at altering the physical or social environment to reduce selection of SSBs, such as warning labels, in-store promotions of healthier beverages, price increases on SSBs, government food benefits programs to disincentivize SSB purchases, increasing availability of low-calorie beverages in the home and community campaigns targeting SSBs, have all shown some success (von Philipsborn et al., 2020). Population-scale efforts to reduce SSB consumption through SSB taxes appear effective (Fernandez and Raine, 2019; Redondo et al., 2018). However, whether any of these efforts enhance health outcomes remains uncertain (Pfinder et al., 2020).

Intestinal fructose absorption

Following ingestion, fructose travels the gastrointestinal tract to the small intestine where it is absorbed via transporters expressed on the brush border of intestinal epithelial cells. GLUT5 (also known as Slc2a5) has marked specificity for fructose over glucose, although a wide range of Km for fructose (6 mM to 15 mM) has been reported depending on the assay system (Burant et al., 1992; Douard and Ferraris, 2008). It is expressed at high levels on the luminal membrane of the small intestine with lesser expression on the basolateral membrane (Burant et al., 1992; Davidson et al., 1992; Ferraris et al., 2018; Rand et al., 1993). Its expression and activity are highest in the proximal duodenum and decline along the length of the small intestine (Rand et al., 1993). GLUT5 is essential for intestinal absorption of dietary fructose. GLUT5 knockout mice are healthy and fertile on standard rodent chow diets, which contain little or no fructose (Barone et al., 2009). However, when exposed to diets containing fructose, these mice develop a severe, morbid malabsorption syndrome characterized by diarrhea and intestinal distension (Barone et al., 2009) (Table 1).

Table 1. Metabolic effects resulting from genetic manipulation of fructolytic enzymes and related factors in animals challenged with high dietary fructose.

KO = knockout; KD = knockdown; OX = overexpression. All genetic interventions were performed in mice unless otherwise specified.

Gene Target	Intervention	Phenotype	References
GLUT5 (SLC2A5)	Global KO	Malabsorption syndrome associated with intestinal distension, diarrhea, and marked weight loss within one week	(Barone et al., 2009)
KHKa/c	Global KO	Reduced weight gain and adiposity; Reduced glycemia and insulin; Reduced hepatic steatosis; Mildly reduced ad libitum fructose consumption; Increased urinary fructose excretion.	(Diggle et al., 2009; Ishimoto et al., 2012)
	Liver KD – ASO	Reduced weight gain; Reduced hepatic steatosis; Improved glucose tolerance.	(Softic et al., 2018)
	Liver KO	Reduced weight gain; Reduced hepatic steatosis; Reduced insulin; Increased urinary fructose excretion.	(Andres-Hernando et al., 2020b)
	Intestine KO	Reduced ad libitum fructose consumption; Despite reduced fructose consumption, persistent increases in body weight and hepatic steatosis.	(Andres-Hernando et al., 2020b)
KHKa	Global KO	Increased weight gain and adiposity; Increased insulin and hepatic steatosis.	(Diggle et al., 2010; Ishimoto et al., 2012)
KHKc	Intestine KO	Increased portal fructose levels; Increased fructose delivery to microbiota; Increased hepatic DNL; Increased circulating triglycerides and hepatic steatosis.	(Jang et al., 2020)
	Intestine OX	Reduced portal fructose levels; Reduced DNL, Reduced ad libitum fructose consumption.	(Jang et al., 2020)
ALDOB	Global KO	Steatosis on fructose free diets and rapid development of severe steatosis and inflammation with fructose exposure; Morbidity and mortality within one week of fructose exposure. Rescue with pharmacological inhibition of KHK.	(Lanaspa et al., 2018a; Oppelt et al., 2015)
TKFC	Global KO		(Liu et al., 2020)
ChREBP (MLXIPL)	Global KO	Weight loss, hypothermia, and severe morbidity and mortality within one week of fructose challenge. Increased hepatic glycogen and hexose-phosphates.	(Iizuka et al., 2004)
	Global KO	Prevents fructose-induced expression of the hepatic fructolytic program. Reduced endogenous glucose production. Impaired glycogenolysis. Increased hepatic hexose-phosphates.	(Kim et al., 2016)
	Global KO	Fructose and sucrose malabsorption associated with diarrhea, intestinal distension, and alterations in microbiome.	(Kato et al., 2018; Oh et al., 2018)
	Global KO	Weight loss; Liver inflammation.	(Zhang et al., 2017)
	Liver KD – ASO (rat)	Reduced circulating triglycerides; Enhanced peripheral insulin sensitivity; Reduced DNL; Increased hepatic glycogen. Mild transaminitis without liver inflammation	(Erion et al., 2013)
	Liver KO	Reduced weight gain and adiposity; Reduced circulating insulin; Reduced DNL; Increased hepatic glycogen. Preserved glycerol tolerance. Mild transaminitis without liver inflammation	(Kim et al., 2017)
	Liver KO	Marked hepatic glycogen accumulation associated with transaminitis without liver inflammation	(Shi et al., 2020)
	Intestine KO	Weight loss with fructose malabsorption and intestinal distension associated with severe morbidity and mortality within one week of fructose challenge.	(Kim et al., 2017)

GLUT5 is a facilitative transporter and fructose transport into the enterocyte is therefore proportional to the concentration gradient across the enterocyte luminal membrane. This contrasts with intestinal glucose absorption which is mediated by the sodium-linked co-transporter, SGLT1, which can efficiently pump glucose into the enterocyte against a concentration gradient. This explains in part why intestinal glucose absorption is faster and more complete than that of fructose (Cori, 1925). Robust fructose phosphorylation within the enterocyte may be important to maintain a steep luminal-enterocyte fructose gradient to facilitate more complete fructose absorption (Patel et al., 2015a). In contrast with glucose, which is nearly fully absorbed within the small intestine, the intestinal capacity for fructose transport can be saturated and some proportion of an ingested fructose load may escape the small intestine and reach the large intestine where it is readily catabolized by gut microbiota (Zhao et al., 2020).

While GLUT5 expression is responsible for intestinal fructose absorption, it is also expressed in other tissues and cell types, including skeletal muscle, pre-adipocytes, the prostate gland, spermatozoa and erythrocytes (Burant et al., 1992; Concha et al., 1997; Kayano et al., 1990; Reinicke et al., 2012). However, the importance of fructose transport into these tissues remains unclear.

Mechanisms mediating efflux of both glucose and fructose across the enterocyte basolateral membrane are less certain. Both glucose and fructose efflux may be facilitated by GLUT2 (also known as Slc2a2) which is expressed at high levels on basolateral membranes in enterocytes of the proximal small intestine (Davidson et al., 1992; Ferraris et al., 2018). GLUT2 is a low-affinity, high-capacity facilitative transporter for both glucose and fructose with similar affinities for both sugars (Km ~ 11 mM) (Manolescu et al., 2007). However, glucose efflux does not appear to be markedly impaired in either GLUT2-deficient humans or mice (Santer et al., 2003; Stümpel et al., 2001). As an alternative, exocytosis via a microsomal membrane-based transport pathway has been proposed to play a major role in transcellular monosaccharide efflux (Stümpel et al., 2001).

At birth, expression of intestinal GLUT5 is negligible and increases at weaning (Douard and Ferraris, 2008; Ferraris, 2001). This induction is enhanced when such diets include fructose (Ferraris, 2001). Fructose-mediated induction of intestinal GLUT5 expression is dependent on intestinal fructose metabolism (Patel et al., 2015b). Induction of intestinal GLUT5 expression in response to dietary fructose is mediated by Carbohydrate Responsive-Element Binding Protein (ChREBP, also known as Mlxipl), a carbohydrate sensing transcription factor expressed at high levels in the intestinal epithelia and other key metabolic tissues (Kato et al., 2018; Kim et al., 2017; Oh et al., 2018). A critical role for ChREBP in organismal fructose metabolism was demonstrated by observations that in mice with global or intestine-specific ChREBP knockout, exposure to high-fructose diets failed to induce intestinal GLUT5 expression, resulting in a malabsorption syndrome and morbidity within one week (Iizuka et al., 2004; Kato et al., 2018; Kim et al., 2017; Oh et al., 2018). This morbidity is similar to the malabsorption phenotype observed in GLUT5 KO mice (Barone et al., 2009).

Fructose absorption is highly variable in human children and adults, depending on age, dietary sugar exposure and likely other unidentified genetic, dietary and environmental factors (reviewed in (Ferraris et al., 2018)). Fruit juices, such as apple juice, which contain large amounts of fructose, are commonly consumed in large quantities by young children and may cause nonspecific gastrointestinal symptoms that resolve with reductions in juice consumption (Ushijima et al., 1995). Fructose malabsorption can be measured via detection of colonic flora-derived gasses, such as hydrogen or methane, after ingestion of an oral fructose load (Rao et al., 2007). Some, but not all, studies using such methods find associations between fructose malabsorption and functional gastrointestinal symptoms (Fernández-Bañares et al., 1993; Melchior et al., 2014). Reducing consumption of dietary fructose, in addition to other highly fermentable but poorly absorbed carbohydrates and polyols, may be useful to treat a range of gastrointestinal symptoms and conditions (Gibson and Shepherd, 2005).

Although genetic ablation of GLUT5 causes fructose malabsorption in mice, fructose malabsorption in humans is not necessarily associated with reduced intestinal GLUT5 expression (Wasserman et al., 1996). Thioredoxin-interacting protein (TXNIP) is another ChREBP-regulated, fructose-induced protein expressed in key metabolic tissues, including the liver and intestine (Dotimas et al., 2016; Iizuka et al., 2004). TXNIP facilitates the localization of facilitative hexose transporters to plasma membranes, including localization of GLUT5 to enterocyte luminal membranes (Shah et al., 2020). Thus, factors regulating proper GLUT5 localization and function, in addition to its expression, may also be important determinants of dietary fructose absorption.

It is not known whether natural variation in intestinal fructose absorption or hepatic fructose metabolism contributes to fructose-induced cardiometabolic disease. Walker and colleagues noted that African-Americans had higher rates of fructose malabsorption associated with lower liver fat compared to Hispanic controls and that within African-Americans, fructose malabsorption correlated inversely with liver fat (Walker et al., 2012). Another small study indicated that children with NAFLD may have enhanced fructose absorption compared to lean controls and that children with either NAFLD or obesity may have enhanced fructose metabolism compared to their lean counterparts (Sullivan et al., 2015). Intriguingly, recent work has shown that GLUT5 expression and glucose production are increased from intestinal stem cell-derived enteroids generated from intestinal biopsies obtained from patients who had obesity compared to lean controls (Hasan et al., 2021). While these studies suggest the possibility that innate differences in intestinal function might contribute to fructose-related metabolic phenotypes, it remains unclear whether these observed differences might be causes or consequences of obesity and NAFLD.

Organismal fructose metabolism and the importance of the intestine and liver

The innate differences in mammalian glucose versus fructose metabolism are reflected in the marked differences in their circulating blood levels. Whereas normal fasting glucose concentrations in peripheral blood are ~5 mM, fructose circulates at less than ~0.02 mM in fasted conditions (Chen et al., 2020; Francey et al., 2019). Ingestion of a large, isolated glucose load as performed during a glucose tolerance test produces peripheral glycemia of ~7.5 mM in healthy individuals. In contrast, an equivalently large oral fructose load transiently increases peripheral circulating fructose levels with peaks rarely exceeding ~1 mM. This is followed by a rapid return to low micromolar levels within hours. The absence of a comparatively large increase in peripheral circulating fructose following ingestion of a large oral fructose load illustrates the gut’s ability to metabolize most of the orally-ingested fructose. Relatively little escapes to the peripheral circulation (Figure 1). Using a sophisticated dual-tracer approach in humans, Fancey et al. recently demonstrated that after ingesting a drink containing 30 g fructose and 30 g glucose, only 15% of the fructose escaped first-pass gut metabolism (Francey et al., 2019).

Figure 1. Organismal fructose metabolism and first pass extraction in the gut.

Following ingestion of a large oral fructose load, fructose is absorbed into intestinal epithelial enterocytes. A portion of this fructose is phosphorylated by KHK within the enterocyte and is converted to glucose, lactate, glycerate and other organic acids which travel via to the portal vein to the liver. Portal fructose concentrations can transiently reach concentrations as high as ~1 mM. Fructose reaching the liver is efficiently extracted by hepatocytes and phosphorylated by KHK where it can be used for glucose production, lipogenesis, glycogen synthesis and energetic purposes. Peripheral blood fructose concentrations transiently peak at levels ~10-fold lower than peak portal levels.

For the last half century, conventional wisdom suggested that the liver accounted for most of the fructose clearance by the gut. This perspective originated from classical studies by Denker et al., Holdsworth et al. and others that measured ~1 to 2 mM fructose in the portal vein of human study participants following oral administration of fructose (Cook, 1969; Dencker et al., 1972; Holdsworth and Dawson, 1965). Cotemporaneous work in baboons demonstrated that gastric installation of 2 g/kg body mass sucrose achieved fructose concentrations of 2 mM in the portal vein while simultaneous measurements in femoral arterial blood transiently peaked at 0.7 mM (Crossley and Macdonald, 1970). Similar results were obtained in other animals (Topping and Mayes, 1971). The large decrement in fructose concentrations from portal blood to peripheral arterial blood speaks to the efficiency of hepatic fructose extraction confirmed in liver perfusion experiments (Mayes, 1993).

Recent work in mice from Jang et al. refocused attention on the relative importance of liver versus intestinal fructose metabolism (Jang et al., 2018). Using a stable isotope approach, Jang and colleagues reported that in mice gavaged with fructose at a dose of 0.5 g/kg body mass, ~90% of the absorbed fructose was metabolized within the small intestine and relatively little reached the liver. Most of the fructose metabolized within the intestine appeared in the portal vein as glucose or other fructose-derived metabolic products including lactate, alanine, glycerate and other organic acids. The small intestine expresses the full complement of gluconeogenic enzymes and is capable of robust conversion of ingested fructose into circulating glucose (Bismut et al., 1993; Ginsburg and Hers, 1960; Jang et al., 2018). While studies show that robust conversion of fructose to glucose is readily detectable in intestines from many species, earlier studies suggested that this did not occur in humans (Cook, 1969; Holdsworth and Dawson, 1965). However, reexamination of isotopic tracer data in studies performed in human children is consistent with extensive conversion of ingested fructose to glucose in human intestines as well (Bismut et al., 1993; Gopher et al., 1990).

Whether intestinal fructose metabolism is dominant over liver fructose metabolism in animals other than mice is less clear. The relative importance of intestinal fructose metabolism may by different among species (Ginsburg and Hers, 1960). Intestinal GLUT5 expression is markedly higher in rats and humans compared to mice (Kim et al., 2007). These differences may contribute to observations that portal fructose levels peaked at ~0.3 mM in mice after fructose gavage, whereas peak portal fructose concentrations are 3- to 5-fold higher in other animals including humans (Crossley and Macdonald, 1970; Holdsworth and Dawson, 1965; Topping and Mayes, 1971). Additional detailed studies are required to clarify the importance of intestinal versus liver fructose metabolism in larger animals including humans.

Nevertheless, these studies led to a novel and important concept - that intestinal fructose metabolism might shield the liver from excessive exposure to dietary fructose where its metabolism may be particularly deleterious (Jang et al., 2018). This issue will be explored in more detail later in this review.

Cellular and intermediary fructose metabolism:

Fructose transport across cell membranes relies on facilitative hexose carriers, such as GLUT5, in enterocytes as there is no known active transporter. In the liver, the other major site of fructose metabolism, GLUT5 is not expressed at high levels, and GLUT2 is likely the major membrane transporter for both glucose and fructose. GLUT8 (also known as Slc2a8), which also has affinity for fructose, has been reported to contribute to fructose transport in both enterocytes and hepatocytes (DeBosch et al., 2012, 2014; Schmidt et al., 2009). However, it predominately localizes to endosomal and lysosomal compartments rather than the cell surface (Alexander et al., 2020; Schmidt et al., 2009).

Upon transport into the cell fructose is metabolized by a cascade of three ‘fructolytic’ enzymes originally delineated by Henri Hers during his graduate studies (Hers, 1957) (Figure 2). These fructolytic enzymes efficiently convert the hexose fructose into triose-phosphates, which are incorporated into the cellular pools of triose-phosphates generated via glycolysis and gluconeogenesis. The first step in fructolysis is the rapid and irreversible phosphorylation of fructose to fructose-1-phosphate (F1P) catalyzed by ketohexokinase (KHK, also known as fructokinase). This metabolite is specific to the fructolytic pathway and is not shared with glycolysis or gluconeogenesis. Compared to glucose, fructose is a poor substrate for hexokinases I, II, and III (Km glucose < 0.1 mM; Km fructose ~1 – 10 mM) (Diggle et al., 2009; Grossbard and Schimke, 1966; Middleton, 1990). At typical circulating concentrations, glucose potently inhibits fructose metabolism by these hexokinases (Diggle et al., 2009; Froesch and Ginsberg, 1962). This is particularly true for glucokinase (GCK), the predominate hexokinase in hepatocytes and pancreatic beta-cells (Km for glucose ~5 mM, Km fructose > 100 mM) (Diggle et al., 2009; Froesch and Ginsberg, 1962; Middleton, 1990; Pollard-Knight and Cornish-Bowden, 1982).

Figure 2. Fructolysis and associated biochemistry.

Fructose is transported into enterocytes and hepatocytes via GLUT5 and GLUT2, respectively. Upon entering the cells, fructose is phosphorylated by KHK to F1P. Energy depletion resulting from robust fructose phosphorylation leads to activation of AMPD2 and uric acid production. F1P is cleaved by ALDOB to DHAP and GA. GA is phosphorylated by triose-kinase (TKFC) to GA3P. Both DHAP and GA3P mix with triose-phosphates common to the glycolytic and gluconeogenic carbon pools. In hepatocytes, F1P allosterically inhibits PGYL to enhance glycogen synthesis and disrupts the interaction between GCK and GCKR allowing GCK to translocate from the nucleus to the cytoplasm and catalyze phosphorylation of glucose further increasing the hexose- and triose-phosphate carbon pools. Fructose-derived substrate has numerous fates, including use in de novo lipogenesis (DNL) both through direct and indirect pathways via microbiome derived acetate. KHK, ketohexokinase; AMPD3, adenosine deaminase; IMP, inosine monophosphate; ALDOB, aldolase B; TKFC, triokinase and FMN cyclase; GA, glyceraldehyde; DHAP, dihydroxyacetone phosphate; GA3P, glyceraldehyde 3-phosphate; PYGL, glycogen phosphorylase L; GYS2, glycogen synthase 2; PKLR, pyruvate kinase, liver and red blood cell; PEP, phosphoenolpyruvate.

Unlike many other hexokinases, KHK is neither inhibited allosterically by ATP, other signals of cellular energy sufficiency, nor by its immediate product. The capacity of KHK to phosphorylate fructose is comparable to the capacity of GCK to phosphorylate glucose in rodent livers and 10-fold higher in human livers (Heinz et al., 1968). Moreover, whereas KHK is constitutively active, hepatic GCK is sequestered in the nucleus in an inhibited state by glucokinase regulatory protein (GCKR) (Agius, 2008; Niculescu et al., 1997). The sequestration and inhibition of GCK by GCKR and the stabilization of this inhibited state by the glycolytic intermediate fructose-6-phosphate (F6P) limits net hepatic glucose uptake and hepatic glucose clearance after oral ingestion of an isolated glucose load (Agius, 2008; Pollard-Knight and Cornish-Bowden, 1982; Van Schaftingen, 1994). In contrast, because of the low Km and high activity of KHK for fructose and the lack of regulation of its activity, the majority of fructose that arrives at the liver via the portal circulation is readily extracted and little reaches the systemic circulation.

The consumption of ATP by KHK-mediated fructose phosphorylation can be so robust that intravenous infusions of large fructose loads cause an acute decline in the hepatocellular ATP-to-AMP ratio and a marked decline in free phosphate as it is sequestered in F1P (Bode et al., 1973; Mayes, 1993). F1P can rapidly accumulate to millimolar concentrations in hepatocytes (Woods et al., 1970). These changes in F1P levels and cellular energy status are detectable by non-invasive ³¹P magnetic resonance spectroscopy in human livers (Oberhaensli et al., 1986). Following fructose infusion, hexose-phosphate levels increased 7-fold within minutes and steadily declined to basal levels after ~ 20 minutes. Hepatic ATP levels declined by more than 80% within minutes and recovered to only ~ 50% of basal levels after one hour (Oberhaensli et al., 1986). Whereas intravenous infusions of fructose can cause acute and profound changes in hepatic metabolite levels and energy status, oral fructose administration results in more modest changes, likely due to the substantial metabolism of fructose within the intestine (Niewoehner et al., 1984). Nevertheless, a recent NMR study indicates that ATP depletion can be detected in human liver following an oral 75 g fructose challenge indicating that hepatic fructose metabolism in humans is substantial (Bawden et al., 2016).

Phosphate potently inhibits the enzyme AMP Deaminase (Ampd2), which catalyzes the rate-limiting step in purine degradation and uric acid production (van den Berghe et al., 1977). Thus, the decline in free phosphate resulting from fructose phosphorylation potently activates Ampd and uric acid production. Fructose may also stimulate purine synthesis (Raivio et al., 1975). Crystallization of uric acid in joints causes the painful rheumatological condition gout. Thus, fructose-induced uric acid production may contribute to the associations between SSB consumption and a risk for gout, but also likely underlie the associations between hyperuricemia and cardiometabolic disease (Jamnik et al., 2016; Yoo et al., 2005).

Although the committed phosphorylation steps in glucose and fructose metabolism are catalyzed by distinct enzymes, the hepatic metabolism of glucose and fructose are not entirely independent of each other. F1P, the product of the KHK-mediated fructose phosphorylation, potently relieves the inhibitory effect of GCKR on GK permitting its translocation to the cytosol and enhancing hepatic glucose uptake and metabolism (Niculescu et al., 1997). Thus, ‘catalytic’ amounts of fructose can markedly enhance hepatic glucose (Petersen et al., 2001; Van Schaftingen, 1994). F1P may also activate pyruvate kinase, the terminal step in glycolysis (Eggleston and Woods, 1970), and can inhibit glycogen phosphorylase (Kaufmann and Froesch, 1973; Thurston et al., 1974; Van Den Berghe et al., 1973). Through these F1P-mediated activities, the fructose contained in dietary sugars as part of a meal may act as a signal to modulate hepatic fuel metabolism.

Recent work has firmly established additional roles for F1P as a key signaling molecule. Taylor et al. recently demonstrated that F1P can bind the M2 isoform of pyruvate kinase (PKM2) and limit formation of highly active PKM2 tetramers (Taylor et al., 2021). PKM2 monomers and dimers are capable of binding and transactivating HIF-1α, a transcription factor essential for cellular adaptation to hypoxia (Luo et al., 2011). This protection against hypoxia increases enterocyte cell survival, leads to elongation of the intestinal villus and increases absorptive capacity. Thus, fructose feeding, by increasing the intestinal absorptive surface may enhance energy harvest and potentially contribute to obesity in calorie-rich environments.

Following fructose phosphorylation, F1P is cleaved by Aldolase B (ALDOB) to dihydroxyacetone phosphate (DHAP) and glyceraldehyde. Glyceraldehyde is then phosphorylated by a triokinase (TKFC, Triokinase and FMN Cyclase) to glyceraldehyde 3-phosphate (G3P). DHAP and G3P then enter the glycolytic/gluconeogenic carbon pools.

Although cellular and organismal fuel status does not regulate fructolytic flux, fuel status does impact the fate of fructose-derived triose-phosphates that enter the central carbon pools (Mayes, 1993). For instance, in fasted or starved animals, when PFK activity is limited by decreased fructose 2,6-bisphosphate levels, fructose-derived triose-phosphates will be routed towards glucose production (Exton and Park, 1967; Exton et al., 1966). In contrast, in fed animals, fructose-derived triose-phosphates may be preferentially metabolized to pyruvate and released as lactate or used as substrate for lipogenesis (Topping and Mayes, 1976). Nutrients co-ingested with fructose may also affect its metabolic fate. For instance, insulin that is secreted when glucose and fructose are ingested together may be important for the full effects of fructose to stimulate glycogen synthesis (Topping and Mayes, 1976). Thus, the fate and effects of ingested fructose on systemic fuel homeostasis depend on systemic fuel status.

Ketohexokinase: genetic lessons in humans and mice and therapeutic implications

KHK catalyzes the committed step in fructose metabolism phosphorylating fructose to F1P. Loss-of-function KHK mutations in humans produces the condition Benign Essential Fructosuria (Asipu et al., 2003; Steinmann et al.). A large and persistent increase in circulating fructose is produced when people with this condition consume foods containing fructose (Laron, 1961; Steinmann et al., 1975). As the name of the condition suggests, in people with this condition ~20% of the fructose is excreted in the urine following an oral or intravenous fructose load (Laron, 1961; Steinmann et al., 1975). The remainder is presumably phosphorylated and cleared by canonical hexokinases, possibly in adipose tissue and skeletal muscle (Froesch and Ginsberg, 1962). Prior to widespread use of glucose oxidase in assays to measure glucose levels in blood and urine samples, cases of Benign Essential Fructosuria came to clinical attention when use of Benedict’s reagent incidentally detected high levels of reducing sugars in the urine, and glucose was subsequently excluded with more specific testing. Because Benedict’s reagent is no longer in regular use, and because there are no known or reported adverse clinical consequences associated with loss of KHK activity, individuals with this condition likely go undetected. While this condition appears to be uncommon, review of data from exome sequencing projects indicates that there is no strong selection against missense or loss of function variants in the KHK gene in human populations (Johnston et al., 2021).

Despite increased circulating fructose in people with Benign Essential Fructosuria, there is no increase in blood insulin and there are no reports of diabetic complications nor increases in %HbA1c (Petersen et al., 1992; Steinmann et al., 1975). The absence of pathology indicates that fructose metabolism via KHK is essential for fructose-induced metabolic disease. This hypothesis is further supported by a mouse with global inactivation of KHK (Ishimoto et al., 2012). Although KHK knockout mice showed a decreased preference for fructose-supplemented drinking water compared to controls, when matched for total energy and fructose intake, KHK KO mice were fully protected from fructose-induced increases in body weight, fat mass, serum insulin and hepatic steatosis (Ishimoto et al., 2012). Thus, fructose metabolism via KHK is essential for fructose-induced metabolic disease.

KHK is expressed as two distinct isoforms from a single gene (Hayward and Bonthron, 1998). Mutually exclusive splicing of exons 3a and 3c in KHK pre-mRNA produce distinct mRNAs that encode KHKa and KHKc protein isoforms, respectively. The Km of KHKc for fructose is ~0.5 mM, whereas the Km of KHKa for fructose is more than an order of magnitude higher (Diggle et al., 2009). KHKa is expressed at low levels ubiquitously. In contrast, KHKc is expressed at high levels selectively in key metabolic tissues, including the liver, small intestine and kidney (Diggle et al., 2009). In addition to a knockout mouse where both KHK isoforms were globally ablated, Ishimoto and colleagues generated a mouse in which the KHKa isoform was selectively deleted. Selective ablation of KHKa exacerbated fructose-induced disease, indicating that in spite of its low level KHKa likely does contribute to systemic fructose metabolism (Ishimoto et al., 2012). In contrast, selective deletion of KHKc revealed that it is primarily responsible for fructose-induced metabolic disease. This conclusion is further supported by recent work from Nikolaou et al. identifying antagonistic actions of RNA-binding proteins A1CF and hnRNPH1/2 that regulate KHKa/c splicing (Nikolaou et al., 2019). A1CF knockout reduces expression of the KHKc isoform and protects mice from fructose-induced metabolic disease (Nikolaou et al., 2019). Regulated splicing of KHK and altered fructose metabolism has also been implicated in other diseases, including hepatocellular carcinoma and heart failure (Li et al., 2016; Mirtschink et al., 2015).

Together, these studies indicate that fructose metabolism in a tissue that expresses the KHKc isoform is critical for fructose-induced metabolic disease. As previously noted, KHKc is expressed in both the intestine and the liver and the intestine is capable of substantial fructose metabolism. Knockout of intestinal KHK exacerbates fructose-induced metabolic disease, whereas liver-specific knockout or knockdown of KHK protects against fructose-induced metabolic disease (Andres-Hernando et al., 2020b; Jang et al., 2020; Softic et al., 2018). These results support the hypothesis that intestinal fructose metabolism may shield the liver from fructose exposure where its metabolism may be particularly harmful. Mechanisms by which hepatic metabolism may contribute to metabolic disease will be addressed in subsequent sections of this review.

Given the evidence that fructose metabolism via KHK is required for fructose-induced metabolic disease and that inactivating mutations in KHK are not associated with adverse clinical effects in humans, targeting inhibition of KHK as a therapeutic strategy seems to hold great promise. Recently, pharmaceutical companies have initiated programs aimed at developing novel potent and specific KHK inhibitors, which are beginning to provide data in pre-clinical and early phase clinical trials (Futatsugi et al., 2020; Gutierrez et al., 2021; Kazierad et al., 2021). Pre-clinical studies in rats demonstrated that the KHK inhibitor PF-06835919 dose-dependently attenuated hyperinsulinemia, hypertriglyceridemia and steatosis provoked by a high-fructose diet. These protective effects were associated with marked increases in circulating fructose and fructosuria and were accompanied by attenuated induction of hepatic ChREBP activity and DNL (Gutierrez et al., 2021). Importantly, PF-06835919 also enhanced metabolic indices in rats on an “American diet”, indicating that KHK inhibition may be healthful outside of extreme fructose exposures. To this end, a randomized, double-blind, placebo-controlled Phase2a study (NCT03256526) in adults with baseline NAFLD, PF-06835919 produced an 18.7% reduction in liver fat after 6 weeks treatment with a 300 mg, but not a 75 mg, daily dose (Kazierad et al., 2021). Treatment with the KHK inhibitor resulted in increased fructosuria consistent with robust KHK inhibition in vivo. This was also associated with improvements in inflammatory markers and increases in circulating adiponectin levels. The drug was well-tolerated. This short-term study provides optimism that inhibiting fructose metabolism may provide a new therapeutic strategy for treating cardiometabolic disease soon.

Additional fructolytic enzymes: genetic lessons in humans and mice

Whereas global KHK loss of function mutations are benign, loss of function mutations in Aldob, the second step in fructolysis, causes Hereditary Fructose Intolerance (HFI), an autosomal recessive Mendelian disorder with significant morbidity (Ali et al., 1998; Hers and Joassin, 1961; Perheentupa et al., 1972). Individuals with this condition tolerate carbohydrates containing glucose or galactose, whereas ingestion of sucrose or fructose triggers abdominal pain, nausea and vomiting accompanied by acute hypoglycemia. This condition often presents in infants as failure to thrive accompanied by liver toxicity after weaning from breast milk or introduction of fructose-containing formula. Aversion to sweet tasting foods is also common (Chambers and Pratt, 1956). Persistent exposure to fructose ultimately leads to liver and kidney failure and ultimately death (Baerlocher et al., 1978). Similar morbidity and mortality can also be provoked by administration of sorbitol, a metabolite that can be used to produce fructose endogenously via the polyol pathway (Ali et al., 1998). People with this condition often learn to avoid all dietary sources of fructose, and are often devoid of dental caries supporting evidence that dietary sucrose is a major contributor to poor dental health (Ali et al., 1998; Marthaler and Froesch, 1967; Touger-Decker and van Loveren, 2003). Elimination of all dietary sucrose, fructose and sorbitol is the mainstay of treatment and improved health can be observed within days of fructose restriction (Mock et al., 1983; Steinmann et al.). Properly treated, HFI patients may experience normal health, growth and lifespan.

Lean and healthy adults with HFI on low-fructose diets have increased intrahepatic triglyceride and glucose intolerance (Simons et al., 2019). This speaks to the importance of F1P as a metabolic signal and indicates that increased fructolytic flux per se is not necessary for the pathogenesis of these features. Additional support for the importance of F1P as a signaling molecule comes from recent studies in adults heterozygous for HFI (hHFI) and controls (Debray et al., 2021). Whereas a high-fructose diet increased fasting uric acid and insulin resistance similarly in both hHFI patients and controls, in the post-prandial condition, plasma uric acid and liver insulin resistance increased only in the hHFI patients (Debray et al., 2021).

Fructose toxicity in HFI is attributed in part to the deleterious effect of the marked and prolonged accumulation of F1P in hepatocytes, enterocytes, and epithelial cells of the proximal tubule (Oberhaensli et al., 1987; Steinmann et al.; Van Den Berghe et al., 1973). ATP depletion and uric acid production are also posited to contribute to toxicity in affected cell types (Steinmann et al.; Van Den Berghe et al., 1973). The hypoglycemia that occurs acutely after fructose consumption in HFI may be due to the effects of F1P to activate glucokinase and also the effects of very high levels of F1P to inhibit glycogenolysis and gluconeogenesis (Rambaud et al., 1973; Steinmann et al.; Van Den Berghe et al., 1973). This may be mediated by an effect of F1P to inhibit glycogen phosphorylase (Van Den Berghe et al., 1973).

Although fructose metabolism in the liver, intestine and kidney is greatly impaired in HFI, less than 20% of a fructose load is recovered in the urine in these patients indicating that the majority of the fructose is metabolized in other tissues (Landau et al., 1971). Adipose tissue metabolism of fructose to fructose-6-phosphate by hexokinase II may be the major mode of fructose metabolism in this condition as may also be the case in Benign Essential Fructosuria (Froesch and Ginsberg, 1962).

The molecular physiology of HFI has been confirmed in a mouse model with global knockout of Aldob (Oppelt et al., 2015). This model recapitulates features observed in human patients including morbidity and mortality when exposed to high-fructose diets. On a fructose-free diet, these mice develop steatosis as is the case for humans with HFI. This may be in part due to signaling effects of increased F1P generated from endogenously produced fructose (Oppelt et al., 2015). The increase in steatosis in Aldob knockout mice can be reversed by inhibiting KHK which further supports the hypothesis that F1P is a critical signaling molecule in the pathogenesis of this condition (Lanaspa et al., 2018a).

As the adverse effects of HFI depend on fructolytic flux, inhibition of KHK could potentially be therapeutic for HFI as it is for other forms of fructose-induced disease. In support of this strategy, Lanaspa and colleagues noted that Osthole, a plant derived coumarinic derivative with pleiotropic biological effects including KHK inhibition, is able to prevent adverse metabolic symptoms in Aldob knockout mice (Lanaspa et al., 2018a; Le et al., 2016). These results provide hope that KHK inhibitors may be useful to markedly expand food options in people with this condition and warrant trials with newly developed potent and specific KHK inhibitors.

The Aldob cleavage reaction generates two metabolites from F1P: DHAP and glyceraldehyde. DHAP requires no further metabolism to join the glycolytic triose-phosphate pool. However, glyceraldehyde, a potentially reactive and toxic metabolite, is phosphorylated by Tkfc to the glycolytic intermediate glyceraldehyde-3-phosphate, completing the final step in fructolysis. Tkfc has been studied less intensively than the preceding two steps, in part, because there are no known clinical conditions associated with mutations in this enzyme. Mice with liver-specific or global Tkfc knockdown demonstrate liver inflammation and evidence of fructose malabsorption when challenged with high-fructose diets, potentially related to toxicity of fructose-derived glyceraldehyde (Liu et al., 2020; Sillero et al., 1969). These mice demonstrated increased glycerate and serine synthesis suggesting metabolism of glyceraldehyde via aldehyde dehydrogenase (Sillero et al., 1969). The recent observation in mice that a substantial fraction of ingested fructose is exported from enterocytes into portal blood as glycerate suggests Tkfc activity may be limiting for fructolysis in some conditions (Jang et al., 2018).

Molecular mediators of fructose on lipid homeostasis

As previously noted, the potential deleterious effects of fructose on metabolism were initially recognized through its adverse effects on lipid homeostasis. This included the ability of extremely high fructose exposures far exceeding that of typical Western diets to provoke fasting hypertriglyceridemia and steatosis and to enhance prandial lipemia within a matter of days. These changes in liver and circulating triglyceride levels consistently associate with increased hepatic DNL – the biochemical process by which new fatty acids are synthesized from precursor molecules (Geidl-Flueck et al., 2021; Lê et al., 2009; Wakil et al., 1983). Indeed, fructose ingestion can increase the amount of newly synthesized fatty acids found within circulating triglyceride in hours (Hudgins et al., 2011). Moreover, newly synthesized fatty acids are increasingly recognized as major contributors to hepatic fat accumulation associated with obesity and NAFLD (Donnelly et al., 2005; Lambert et al., 2014; Smith et al., 2020).

Fructose may facilitate DNL through multiple mechanisms (Figure 3). Fructose-derived metabolites, such as F1P, act as signaling molecules to promote DNL (Kim et al., 2016). As noted above, this is supported by evidence in patients and animal models of HFI where increased F1P but decreased fructolytic flux are associated with increased steatosis. Nevertheless, increased fructolytic flux may also provide substrate and reducing equivalents to support fatty acid synthesis (Wakil et al., 1983). Additionally, uric acid generated as a result of rapid fructose catabolism has been suggested to promote DNL (Lanaspa et al., 2012).

Figure 3. Fructose activates metabolic gene expression.

Fructose derived metabolites act as signaling molecules to activate metabolic transcriptional programs. F1P derived from fructose metabolism activates GCK and the combination of fructose- and glucose-derived hexose phosphates activate ChREBP, which coordinately regulates enzymes involved in fructolysis, glycolysis, glucose production, lipogenesis and VLDL packaging and export. Fructose-derived metabolites may also activate SREBP1c. Both ChREBP and SREBP1c are coactivated by PGC1β to further coordinate and enhance transcription of metabolic programs. ACLY, ATP citrate lyase; ACACA, acetyl-CoA carboxylase α; FASN, fatty acid synthase; GPAT, glycerol-3-phosphate acyltransferases; AGPAT, acylglycerol-3-phosphate acyltransferase; DGAT, diacylglycerol acyltransferase; MTTP, microsomal triglyceride transfer protein; TM6sf2, Transmembrane 6 Superfamily, Member 2; TAG, triacylglycerol; VLDL, very-low density lipoprotein.

The predominant mechanism by which fructose induces DNL in the liver is through its effects to increase hepatocellular F1P, which causes dissociation of GCK from its inhibitory binding partner, GCKR (Agius, 2008; Van Schaftingen, 1994). This serves to increase glycolytic flux and increase the concentration of hexose- and triose-phosphates within the hepatocyte (Kim et al., 2016). One or more of these glucose metabolites activates ChREBP, leading to coordinate upregulation of the full complement of enzymes required for DNL (Dentin et al., 2004; Iizuka et al., 2004; Katz et al., 2021; Kim et al., 2016). Increased glycolytic flux via activation of GCK also provides additional substrate and reducing equivalents supporting both DNL and glycogen synthesis.

ChREBP is an evolutionarily conserved, carbohydrate-sensing transcription factor that is highly expressed in key metabolic cell types, including enterocytes, hepatocytes, adipocytes, pancreatic beta-cells and proximal tubule cells of the kidney (Iizuka et al., 2004; Katz et al., 2021; Uyeda and Repa, 2006). ChREBP is required for carbohydrate-mediated induction of glycolytic and lipogenic enzymes in the liver (Iizuka et al., 2004). It also transactivates expression of the complement of fructolytic enzymes (Iizuka et al., 2004; Ma et al., 2006). It was discovered based upon its ability to mediate upregulation of the glycolytic enzyme liver pyruvate kinase (Pklr) independently of insulin signaling in primary hepatocytes exposed to high glucose culture media (Yamashita et al., 2001). However, in vivo, fructose and not glucose gavage acutely and robustly upregulates expression of hepatic ChREBP-beta, a potent ChREBP isoform, along with its fructolytic, glycolytic and lipogenic targets (Kim et al., 2016). Liver-specific ChREBP knockdown or knockout prevents sucrose- and fructose-mediated induction of DNL enzymes and DNL activity (Erion et al., 2013; Kim et al., 2017; Linden et al., 2018).

In humans, genetic variants in the ChREBP locus strongly associate with increased circulating triglyceride levels and reduced HDL-cholesterol levels, the dyslipidemia characteristically associated with the metabolic syndrome (Kooner et al., 2008). Fructose consumption acutely increases ChREBP activity, which has been observed in livers of humans with obesity, diabetes and fatty liver disease (Eissing et al., 2013; Kursawe et al., 2013). This suggests that ChREBP links hepatocellular hexose availability to both DNL as well as VLDL packaging and export and may mediate the DNL and hypertriglyceridemia after fructose feeding in individuals with obesity and insulin resistance. Recent evidence demonstrates that genetic variants in the ChREBP locus interact with sugar consumption to impact circulating lipids in human populations further supporting a role for ChREBP as a central mediator of sugar’s effects on cardiometabolic lipid risk factors in people (Haslam et al., 2021). ChREBP may serve a similar function to promote DNL and chylomicron packaging in enterocytes after fructose feeding (Haidari et al., 2002).

Liver ChREBP knockdown in fructose-fed rats reduced circulating triglyceride levels confirming a role for ChREBP in fructose mediated dyslipidemia (Erion et al., 2013). The effects of ChREBP on circulating triglyceride are likely mediated through its regulation of key enzymes involved in VLDL packaging and export including Microsomal Triglyceride Transfer Protein (Mttp) and Transmembrane 6 Superfamily, Member 2 (Tm6sf2) (Lei et al., 2020; Niwa et al., 2018). Hepatic ChREBP may also limit circulating triglyceride clearance through its ability to stimulate production of proteins, such as Apolipoprotein C3 (Apoc3) and angiopoietin-like 8 (Angptl8), which inhibit Lipoprotein Lipase (Lpl) activity and clearance of circulating triglycerides (Caron et al., 2011; Fu et al., 2014). While liver-specific ChREBP knockout or knockdown fully abrogates DNL, under some conditions this has little or no effect on steatosis (Erion et al., 2013; Kim et al., 2017; Linden et al., 2018). Linden et al. suggested that the reduction in DNL in ChREBP KO mice is balanced by a reduction in VLDL packaging export leading to little net change in liver fat content (Linden et al., 2018). Additionally, though DNL is a significant contributor to steatosis, particularly in patients with obesity and NAFLD, the majority of lipid in steatotic livers originates from circulating free fatty acids rather than DNL (Donnelly et al., 2005; Smith et al., 2020). These results highlight that DNL and steatosis are regulated by distinct processes and can be dissociated from each other. The importance of DNL to both NAFLD and the dyslipidemia of metabolic syndrome requires further investigation.

ChREBP does not act alone in regulating the expression of DNL enzymes and other mediators of lipid homeostasis. Sterol regulatory element-binding protein 1c (SREBP1c) is another master regulator of lipogenic enzymes that responds to both nutrients and hormones (Shimano and Sato, 2017). The effects of ChREBP and SREBP1c are synergistic on lipogenic targets (Linden et al., 2018). Insulin strongly promotes SREBP1c activity via a pathway that is dependent on AKT and mechanistic target of rapamycin complex 1 (mTORC1) (Shimano and Sato, 2017). While fructose does not directly stimulate insulin secretion, increased adiposity and hyperinsulinemia associated with chronic consumption of fructose might enhance SREBP1c activity.

Along with activation of ChREBP, fructose feeding can activate SREBP1c independently of hepatic insulin signaling in liver-specific insulin receptor-knockout mice (Haas et al., 2012). Recent work indicates that DHAP, a product of both glycolysis and fructolysis, provides a signal to activate mTORC1 which is important in activating SREBP1c (Orozco et al., 2020). However, fructose feeding appears to inhibit, rather than activate, mTOR (Hu et al., 2018). The mechanisms mediating fructose-induced activation of SREBP1c remain uncertain. Additional recent evidence in humans supports the hypothesis that increased lipogenic enzyme expression and lipogenesis in NAFLD patients is dependent on substrate-derived signals, rather than systemic hormonal cues, activating both ChREBP and SREPB1c, (Ter Horst et al., 2020).

Some, but not all studies suggest that fructose consumption promotes ER stress which may also induce proteolytic cleavage and activation of SREPB1c (Flamment et al., 2012; Kammoun et al., 2009; Kim et al., 2017; Marek et al., 2015). ER stress can also activate the transcription factor XBP1 which may enhance expression of lipogenic enzymes independently of other lipogenic transcription factors (Lee et al., 2008; Marek et al., 2015).

PPAR-γ coactivator 1β (PGC-1β) is a transcriptional co-activator that can enhance transcriptional activity of both ChREBP and SPREBP1c, as well as other important metabolic transcription factors including PPAR-γ, PPAR-α, estrogen-related receptors (ERRs) and liver X receptor (LXR) (Chambers et al., 2013; Lin et al., 2002, 2005a, 2005b). In fructose-fed rats, knockdown of hepatic PGC-1beta prevented increases in adiposity, lipogenesis, steatosis, hyperinsulinemia and hypertriglyceridemia (Nagai et al., 2009). As a result, PGC-1β appears to promote multiple adverse sugar-mediated phenotypes and poses interesting potential as a therapeutic target.

Recent work has recast the role of fructose as a lipogenic substrate within hepatocytes. The path by which fructose-derived intermediates can be used as substrates for fatty-acid synthesis within hepatocytes requires synthesis of citrate in the mitochondria, which is then exported from the mitochondria and cleaved by ATP Citrate Lyase (Acly) to generate cytosolic acetyl-CoA, the primary precursor for new fat synthesis (Sul et al., 1984). Contrary to conventional wisdom, Zhao and colleagues showed that fructose-derived carbons could be incorporated into newly synthesized fatty acids in the liver in Acly knockout mice, demonstrating that fructose-derived citrate is not required for its use in DNL (Zhao et al., 2020). In this study, fructose carbons were used as lipogenic substrate via an indirect path. Fructose catabolized by microbiota in the distal gut to acetate was absorbed, transported via the portal vein, taken up by hepatocytes, and converted to acetyl-CoA by an acetyl-CoA synthetase (Acss2). When fructose was consumed more gradually, both the conventional, direct citrate pathway and the microbiota-acetate pathway appeared to contribute fructose carbons to newly synthesized fatty acids (Zhao et al., 2020).

Fructose-derived metabolites may provide signals that regulate additional aspects of intermediary metabolism. For instance, the second step in DNL is mediated by Acetyl-CoA Carboxylase (ACACA) which catalyzes the carboxylation of acetyl-CoA, the rate-limiting step in DNL, to generate malonyl-CoA (López-Casillas et al., 1988). Malonyl-CoA serves as an important signaling molecule as it potently inhibits carnitine palmitoyltransferase 1A (Cpt1a), an enzyme required for translocation of fatty acids into the mitochondria (McGarry, 2002). Thus, fructose-derived signals also serve to inhibit fatty acid oxidation (Topping and Mayes, 1972). Recent evidence suggests that fructose-derived metabolites might also be used to post-translationally modify mitochondrial proteins, including Cpt1a, which may also contribute to inhibition of fatty acids (Softic et al., 2019). Inhibiting fatty acid oxidation is another mechanism by which fructose-derived signals may contribute to the development of steatosis.

Fructose and NAFLD

Hepatic steatosis constitutes the earliest stage of NAFLD and is diagnosed when more than 5% of liver cells contain lipid droplets on biopsy or, more commonly, by specific estimates of liver fat content based on imaging. In 10–15% of individuals, steatosis progresses to non-alcoholic steatophepatitis (NASH), which is characterized by inflammation and fibrosis, and in turn may evolve further to cirrhosis and increased risk for hepatocellular carcinoma. In individuals with NAFLD, steatosis strongly associates with increased DNL and variably associates with changes in VLDL secretion and fat oxidation (reviewed in (Ter Horst and Serlie, 2017)). Because of fructose’s ability to promote DNL and inhibit fat oxidation, investigators have posited that fructose consumption is a major driver of NAFLD (Jensen et al., 2018; Lim et al., 2010). While studies administering very high doses of fructose support the possibility that fructose promotes steatosis in humans, so far, evidence that fructose when consumed at levels typical of Western diets associates with development of NAFLD is mixed. This pertains to both prospective epidemiological cohorts evaluating sugar and fructose consumption, as well as short-term intervention studies comparing addition of isocaloric fructose versus glucose (Agebratt et al., 2016; Chiu et al., 2014; Chung et al., 2014; Kanerva et al., 2014; Ma et al., 2015; Ouyang et al., 2008; Schwarz et al., 2015). Two fructose restriction studies support the possibility that reducing fructose consumption, even isocalorically, reduces DNL and liver fat in humans, but these studies lacked control arms which limit definitive conclusions (Schwarz et al., 2017; Volynets et al., 2013). Recently, a small well-controlled sugar restriction study was performed in adolescent boys with histological evidence of NAFLD (Schwimmer et al., 2019). Free sugars were restricted to less than 3% of daily calories for 8 weeks and this was compared to ‘regular’ diet controls. Sugar restriction associated with greater reduction in liver fat as well as greater reduction in alanine aminotransferase, a circulating marker of liver injury. In the last year, another small, well-controlled restriction study was completed in overweight adults (Simons et al., 2021). Dietary fructose was restricted to less than 10 g per day and individuals were randomly assigned to supplementation with fructose to achieve fructose intake similar to baseline (control group) or isocaloric glucose over 6 weeks (restriction group). The fructose-restricted group achieved small, but statistically significant, reductions in liver fat compared to controls (Simons et al., 2021). Perhaps the most compelling evidence that fructose is a common contributor to NAFLD derives from a clinical study in which participants with steatosis were administrated a KHK inhibitor, thus blocking fructose metabolism in liver, intestines and kidney (Kazierad et al., 2021). Participants with a mean baseline liver fat of 15% were asked to maintain their normal diets without fructose supplementation for six weeks during which time they were treated with either of two doses of KHK inhibitor or placebo. At the end of the study, study participants who took the higher dose of KHK inhibitor demonstrated a greater than 18% reduction in liver fat compared to placebo. These data strongly support the proposal that quantities of fructose typical of a modern Western diet contribute significantly to the increasing prevalence of NAFLD.

A minority of patients with NAFLD progress to more advanced forms of liver injury involving inflammation and fibrosis. Some evidence suggests that increased fructose consumption may promote progression to more advanced forms of NAFLD in children, adolescents and adults (Abdelmalek et al., 2010; Mosca et al., 2017). The ability of fructose to promote progression could be due to its effects to enhance lipogenesis and steatosis, but other mechanisms related to unique aspects of its metabolism could also contribute. Emerging data suggests that hepatocellular energy homeostasis is abnormal in NAFLD, and these energetic derangements may be exacerbated by fructose. As previously noted, the rapid increase in hexose-phosphates and depletion of ATP that can result from fructose phosphorylation within hepatocytes can be detected non-invasively by ³¹P magnetic resonance spectroscopy (MRS). Using this and related techniques, the majority of MRS studies detect alterations in energy status in livers of patients with NAFLD (Abrigo et al., 2014; Bawden et al., 2016; Cortez-Pinto et al., 1999; Karczmar et al., 1989; Nair et al., 2003; Traussnigg et al., 2017). A pilot study suggested that after administration of fructose, recovery of energy status is delayed in livers of patients with NASH (Cortez-Pinto et al., 1999). However, follow-up studies in participants with obesity showed mixed results (Bawden et al., 2016; Nair et al., 2003). These human studies are complemented by studies performed in fructose-fed mice (Lanaspa et al., 2012; Softic et al., 2019). Fructose feeding produced evidence of altered mitochondrial morphology, an altered mitochondrial proteome and increased reactive oxygen species that could potentially adversely impact mitochondrial function and contribute to changes in energy status and liver disease (Lanaspa et al., 2012; Softic et al., 2019).

Recent investigations have focused not only on the direct effects of fructose metabolism within the liver, but indirect effects potentially mediated by changes in the microbiome and gut barrier function. Complex associations between changes in the gut microbiome and NAFLD are of increasing interest (see (Sharpton et al., 2021)). Unabsorbed fructose serves as fuel for intestinal microbiota and may contribute to a wide range of changes in intestinal microbial communities (reviewed in (Lambertz et al., 2017)). Fructose-dependent alterations in intestinal microbiota or direct effects on the epithelial cells may disrupt barrier function leading to progression of NAFLD due to increased intestinal permeability (Bergheim et al., 2008). This may increase TNF-alpha and endotoxin in the portal blood, effects that can be attenuated with antibiotics (Bergheim et al., 2008; Todoric et al., 2020). Recent works suggests that prolonged exposure to 30% fructose in drinking water causes ER stress within colonic enterocytes leading to barrier breakdown and endotoxemia in mice (Todoric et al., 2020). This endotoxemia induces TNF-alpha in liver myeloid cells which can induce ER stress in hepatocytes leading to cleavage and activation of SREBP1 (Kim et al., 2018; Todoric et al., 2020). However, in normal physiological circumstances, little ingested fructose reaches the colon. Indeed, two weeks of fructose supplementation in a small preliminary prospective cohort of adults with obesity did not show any changes in microbiome, fecal metabolites or indices of gut permeability, inflammation or endotoxemia (Aleman et al., 2020).

Fructose effects on glucose homeostasis

Along with changes in liver fat and circulating triglycerides, short-term hypercaloric sucrose or fructose feeding can rapidly induce fasting hyperinsulinemia in humans and animal models (Lê et al., 2009; Reaven, 1988; Ter Horst et al., 2016). This occurs despite the fact that fructose in isolation does not stimulate insulin secretion from pancreatic beta cells (Curry, 1989). Presumably, this hyperinsulinemia is an indicator of liver and/or peripheral insulin resistance. In rats, the dose and duration of sugar exposure has significant effects on which tissues are affected. Compared to glucose, very high doses of fructose and sucrose (> 60% of total energy) can induce both hepatic and peripheral insulin resistance within 2 months. However, more moderate doses (~18% of total energy) for twice as long elicited isolated hepatic insulin resistance (Pagliassotti and Prach, 1995; Pagliassotti et al., 1994; Thresher et al., 2000). As is the case in rodents, isocaloric and moderate hypercaloric fructose feeding in humans has preferential effects to induce hepatic over peripheral insulin resistance (Aeberli et al., 2013; Schwarz et al., 2015; Ter Horst et al., 2016).

Pagliassotti and colleagues noted that gluconeogenic capacity was increased in primary hepatocytes isolated from high sucrose and fructose feeding, possibly due to a fructose-dependent increase of glucose-6-phosphatase (G6pc), the final enzymatic step of glucose production (Bizeau et al., 2001; Wei et al., 2004). ChREBP is essential for fructose-mediated upregulation of both intestinal and hepatic G6pc (Kim et al., 2016; Pedersen et al., 2007). The ChREBP-mediated increase of G6pc is dominant over the antagonistic action of insulin to suppress it (Kim et al., 2016). This ChREBP-G6pc signaling axis is conserved in humans and likely contributes to hepatic insulin resistance in humans in the setting of high sugar diets (Kim et al., 2016). Hepatic ChREBP knockdown in rats also attenuated peripheral insulin resistance by an unknown mechanism (Erion et al., 2013). Recent work demonstrated that knocking out KHK in mice prevented fructose-induced changes in adipose inflammation, insulin resistance and a decline in circulating adiponectin (Marek et al., 2015). Mechanisms linking fructose metabolism to changes in adipose function are uncertain. Fructose may also cause hepatic insulin resistance by impairing hepatic insulin signaling (Softic et al., 2020). Fructose induced ER-stress and associated JNK kinase activation and signaling has been invoked as a cause of hepatic insulin resistance (Sun et al., 2015; Wei and Pagliassotti, 2004; Wei et al., 2005). As fructose feeding may cause steatosis and steatosis is commonly associated with hepatic insulin resistance, one or more of the many putative lipid mediators implicated in lipid-mediated insulin resistance may be involved (Petersen and Shulman, 2018). For instance, fructose feeding has been associated with diacylglycerol accumulation, PKC activation, and impaired insulin-mediated AKT2 activation (Jurczak et al., 2012; Nagai et al., 2009). One other distinct mode of fructose-mediated hepatic insulin resistance has been described. Coate and colleagues noted that hepatic glucose uptake was impaired in both high-fat and high-fructose fed dogs (Coate et al., 2010, 2014). Whereas fat impaired hepatic insulin action and AKT phosphorylation, fructose did not. Rather, fructose impaired hepatic Gck protein expression, Gck activity, and hepatic glycogen synthesis. The mechanism mediating fructose’s negative effects on hepatic Gck are unknown.

Recent work indicates that fructose metabolism in the liver may also regulate systemic amino acid metabolism. In particular, fructose-mediated activation of ChREBP upregulates and downregulates BCKDK and PPM1K, respectively (White et al., 2018). BCKDK is a kinase that phosphorylates and inhibits BCKDH, the rate-limiting step in branched-chain amino acid (BCAA) oxidation (White and Newgard, 2019). This phosphorylation and inhibition are reversed by the phosphatase PPM1K. As liver is a major site of BCAA oxidation, this mechanism may contribute to the long-standing observation that increased circulating branched-chain amino acids associate with insulin resistance and impaired glucose homeostasis in humans (Newgard et al., 2009; White and Newgard, 2019).

Effects of fructose on appetite, macronutrient preference, and obesity

The hedonic reward derived from consuming sugars containing fructose contributes to its overconsumption, leading to excessive energy intake, overweight and obesity. Additional mechanisms independent of its hedonic value have also been invoked to explain why fructose might be particularly obesogenic. For instance, fructose ingestion may have distinct effects on anorexigenic and orexigenic hormones, such as leptin and ghrelin, that impact feeding behaviors. Meal-associated increases in leptin are diminished by fructose compared to glucose and fructose may induce leptin resistance (Chotiwat et al., 2007; Shapiro et al., 2008; Teff et al., 2004). Fructose ingestion less potently suppresses ghrelin secretion compared to glucose ingestion (Teff et al., 2004). All these effects could promote excessive food intake and weight gain.

Recent evidence derived from human genetics and non-clinical model organisms has suggested a negative feedback loop whereby consumption of fructose suppresses further sugar consumption. FGF21 is a liver-derived hormone that regulates systemic fuel homeostasis (Flippo and Potthoff, 2021). In humans, non-human primates and rodents, fructose consumption acutely and robustly increases hepatic production of FGF21 in a ChREBP-dependent manner (Dushay et al., 2015; Iizuka et al., 2009; Kim et al., 2017; Talukdar et al., 2016). FGF21 signals via neural circuitry, including through glutamatergic neurons in the ventromedial hypothalamus, to suppress further carbohydrate intake (von Holstein-Rathlou et al., 2016; Jensen-Cody et al., 2020; Talukdar et al., 2016). Moreover, common genetic variants in the FGF21 locus associate with sugar versus fat preferences in human populations (Chu et al., 2013; Tanaka et al., 2013).

The effect of fructose to suppress further carbohydrate consumption may confound interventional trials that aim to study the metabolic effects of added sugars. Indeed, multiple rigorous studies have found that supplemental sugar markedly reduced spontaneous consumption of other carbohydrates, limiting differences in total sugar consumption between experimental and control paricipants (Ebbeling et al., 2020; Maersk et al., 2012). This strong negative feedback loop must be considered when designing and interpreting dietary intervention trials.

Fructose contributions to dysregulated blood pressure

Whereas extreme fructose-enriched diets can induce hypertension in rodents, associations between SSB consumption and hypertension in humans is weaker, particularly when compared to other cardiometabolic risk factors (Hwang et al., 1987; Khan and Sievenpiper, 2016). Potential mechanisms by which fructose may affect blood pressure include effects on intestinal salt absorption, renal salt absorption and function, or endothelial function (reviewed in (Klein and Kiat, 2015)). Further, based on epidemiological associations and interventional studies in animal models, Johnson and colleagues have proposed that fructose-induced hyperuricemia is a major contributor to the deleterious metabolic effects of fructose, including its adverse effects on kidney function, blood pressure and cardiovascular risk (Ahsan Ejaz et al., 2020). However, most, but not all Mendelian randomization studies indicate that genetic variants that causally effect uric acid levels and risk of gout, do not associate with other adverse cardiometabolic traits (Ge et al., 2020; Jordan et al., 2019; Tin et al., 2019; Yang et al., 2010). Moreover, recent randomized, prospective interventional trials in humans aimed at reducing uric acid production did not show benefits in patients with type 1 diabetes or chronic kidney disease (Badve et al., 2020; Doria et al., 2020). Thus, while uric acid production remains an excellent marker of fructose metabolism, its causal role in fructose-associated pathologies remains controversial.

Fructose feeding may promote hypertension by enhancing intestinal salt absorption (Barone et al., 2009; Singh et al., 2008). Fructose induced the intestinal anion exchanger Slc26a6 in a GLUT5-dependent manner, and Slc26a6 knockout mice are resistant to fructose-induced hypertension but not other adverse metabolic effects of fructose feeding (Barone et al., 2009; Singh et al., 2008). Fructose consumption suppressed renin expression consistent with salt overload, and reduced renal salt excretion, which may further contribute to fructose-induced hypertension. The effects of fructose on increased blood pressure may be synergistic with high-salt diets (Cabral et al., 2014). Recent work suggests that fructose may also sensitize the proximal tubule to angiotensin II and promote renal salt reabsorption (Gonzalez-Vicente et al., 2018; Yang et al., 2020). Additionally, KHK-mediated fructose metabolism within the proximal tubule may enhance salt reabsorption by activating the sodium hydrogen exchanger (Slc9a3) (Hayasaki et al., 2019; Queiroz-Leite et al., 2012). As fructolytic enzymes are expressed at high levels in the renal proximal tubule, the role of renal fructose metabolism as it relates to renal salt handling will likely be of growing interest.

Endogenous production as a modifiable exposure?

While nearly all public health attention and most scientific research has focused on metabolic effects of excessive dietary fructose, fructose is also synthesized endogenously from glucose through the sorbitol (polyol) pathway. In this pathway glucose is reduced by aldose reductase (AR, also known as Akr1b1 in humans and Akr1b3 in mice) to sorbitol, which is then oxidized to fructose by sorbitol dehydrogenase (SORD). Endogenously produced fructose appears to play a role in fertility as it is a major energy source for sperm (Frenette et al., 2006; Jayaraman et al., 2014; Martini et al., 2010). Sorbitol, the immediate precursor for fructose production, is also produced by the placenta (Hwang et al., 2015). In fasted adult humans, circulating plasma fructose is detectable in the low micromolar range, consistent with some degree of endogenous production, and it tends to be higher in men than women possibly related to production in the testes (Chen et al., 2020).

Recently, Francey et al. measured endogenous fructose production rates of ~5.6 −16.7 g/day in vivo in healthy human participants (Francey et al., 2019). Fructose production acutely increased with sugar consumption (Francey et al., 2019). An endogenous fructose production rate of 17 g/day is approximately one-third of the average daily consumption of fructose in the US population and thus a substantial fraction of total fructose exposure in many people (Marriott et al., 2009). Indeed, this is roughly equivalent to the fructose in a can of regular soda – an amount that when consumed daily associates with increased cardiometabolic risk (Green et al., 2014).

Increased fasting fructosemia associates with higher fasting glucose and risk of incident T2D (Chen et al., 2020). Because fructose is produced endogenously from glucose, this may suggest that increased endogenous fructose production is a marker of worsening glycemia. However, even after adjustment for baseline glucose levels and SSB consumption, fasting blood fructose levels associate with increased risk of incident T2D in a dose-dependent manner, possibly due to a pathogenic “feed-forward” loop whereby increased blood glucose stimulates fructose production, which in turn exacerbates T2D and cardiometabolic risk (Chen et al., 2020).

The increased serum, urine and tissue sorbitol concentrations present in individuals with diabetes have been implicated in the development of diabetic microvascular complications (Brownlee, 2005; Lorenzi, 2007; Oates, 2002; Preston and Calle, 2010). Activity of polyol pathway may vary significantly within the population (Yoshii et al., 2001), raising the possibility that people with increased sorbitol and fructose production may be at higher risk for the development of T2D and its complications. While sorbitol levels decrease in response to improved glycemic control in T2D, some studies suggest that elevated sorbitol and fructose levels are more refractory to normalization of glycemia in some groups of patients again indicating that increased endogenous fructose production may contribute to the pathogenesis of T2D (Yoshii et al., 2001). Further studies are needed to better understand the causal relationship between circulating glucose concentrations and endogenous fructose production. However, current knowledge is compatible with a model whereby increased flux through the polyol pathway, leading to increased endogenous fructose production, may present a significant, unappreciated risk factor for development of T2D and its complications.

Studies in animal models also provide evidence of substantial endogenous fructose production. Indeed, micromolar circulating fructose can be measured in GLUT5 knockout mice on fructose-free diets (Patel et al., 2015c). Moreover, inhibition of fructose metabolism by global deletion of KHK in mice leads to a 30-fold increase in circulating fructose on fructose-free diets (Patel et al., 2015c). Findings in animal models also support a pathologic role for endogenous fructose production in metabolic disease (Lanaspa et al., 2013, 2014; Oppelt et al., 2015). Mice exposed to 10% glucose in drinking water develop features of metabolic syndrome including increased visceral adiposity, hyperinsulinemia and fatty liver (Lanaspa et al., 2013). This effect is blunted in both AR-deficient and KHK-deficient mice (Lanaspa et al., 2013). As previously noted, mice with mutations in Aldob, the enzyme required for catabolism of F1P, develop significant steatosis even on diets free of fructose indicating that endogenously produced fructose is sufficient to contribute to metabolic dysfunction in this model (Oppelt et al., 2015).

Sorbitol is a major intracellular osmolyte in the renal medulla, where its production can be increased by hypertonicity through induction of AR expression (Bagnasco et al., 1987). Theoretically, high-salt diets might enhance endogenous fructose production within the kidney through this mechanism. Unexpectedly, Lanaspa and colleagues have observed that gavaging mice with hypertonic fluids induced AR expression in the liver and hypothalamus and have proposed that endogenous fructose production is a conserved cellular program in response to hypertonicity in multiple tissues and that high-salt diets may contribute to metabolic disease through this mechanism (Lanaspa et al., 2018b). Additional potential links between fructose metabolism, the regulation of volume and hydration status, and metabolic disease include the observation that fructose feeding can induce vasopressin secretion and that vasopressin activity through the vasopressin V1B receptor (Avpr1b) may contribute the adverse metabolic effects of fructose (Andres-Hernando et al., 2021).

Conclusions

The last five years have witnessed major advancements in our understanding of the molecular physiology of fructose metabolism and its potential contribution to metabolic disease. One conclusion, firmly established in recent years is that excessive fructose metabolism within the liver is a major source of fructose-associated morbidity. Over the next five years, we anticipate further progress in defining key molecular mechanisms by which hepatic fructose metabolism participates in disease pathogenesis. We are hopeful that mechanisms defined through these studies will be useful not only to understand fructose-associated disease but may generalize to other forms of metabolic disease associated with obesity and overnutrition. Recent work has also established that intestinal fructose metabolism may be protective. Defining mechanisms and pathways to enhance intestinal fructose metabolism seems a promising alternative or complementary therapeutic approach. Whereas fructose metabolism within the intestine may be protective, the role of other fructose metabolism in other tissues, such as the kidney, remains to be firmly established. With the availability of conditional genetic mouse models, we expect that this and related questions will be addressed in short order.

Advances in targeting fructose metabolism pharmacologically for therapeutic purposes have paralleled the advancements in our understanding of its molecular physiology. Initial results in early-stage clinical trials appear promising. Larger and longer-term studies will be required to determine efficacy and safety in diverse clinical settings with potential applications in populations with NAFLD and diabetes.

The pharmacological and physiological data continues to support recommendations to limit sugar consumption. Along with the advancements in molecular physiology and pharmacology, establishing the efficacy of public health efforts to reduce sugar consumption will remain an important goal for the foreseeable future.