Due to the important role of glucose as an energy source, glucose homeostasis is of central importance for the function of an organism. Before we can understand the regulation of glucose homeostasis in whole organisms, it is essential to characterize the processes that control the various components at the cellular level. Three principal components determine glucose homeostasis within a cell: (i) transport across the plasma membrane, (ii) biosynthesis and metabolism, and (iii) compartmentation. It is conceivable that all three processes are regulated and that both extracellular and intracellular glucose levels as well as downstream metabolic intermediates can affect the individual metabolic and transport reactions. To model glucose metabolism, it would be useful to know all of the relevant parameters, such as concentration and flux. However, it is difficult to measure the key metabolites directly. While tracer studies with nonmetabolizable analogs provide indications for affinity and velocity of the uptake process, these parameters may not represent the true values due to differences in recognition of substrate and analog and due to slower metabolism of the tracer. The efflux parameters are even more difficult to determine. One way to analyze the parameters would be to preload the cells with substrate and to determine the release, but this approach is complicated by the difficulty of manipulating intracellular analyte levels. Compartmentation is also difficult to measure. The isolation of organelles may affect the transport and metabolic properties of the organelle, or purification of the compartment may be partial, especially in the case of the ER, where, typically, microsomal fractions are used as proxies. Moreover, regulatory metabolites may be lost during purification. Therefore, the regulation of the three components, despite their relevance for disease, is still poorly understood.
Noninvasive methods that dynamically report glucose levels in the cytosol as well as within cell compartments are thus required. FRET-based sensors provide the opportunity to measure all parameters simultaneously in different compartments, since these genetically encoded sensors can be targeted to organelles and report glucose levels with high temporal and spatial resolution (21
). In previous studies, FRET sensors for glucose were developed and used to measure glucose homeostasis in COS-7 cells (14
) The sensor was targeted to both the cytosol and the nucleus, showing that steady-state glucose levels were identical in the two compartments and suggesting that nuclei can be used as proxies for the levels in the cytosol (11
). In this study, the FLIPglu-600μ sensor was expressed in the cytosol or targeted to the ER lumen to address the question of glucose flux across the ER membrane in vivo.
As outlined above, hepatocytes function as transient glucose buffering systems. When serum levels drop, the liver compensates by releasing glucose. For reasons not fully understood, the ER plays a crucial role during gluconeogenesis and thus for the release of glucose into the bloodstream from these cells. According to “Arion's model,” the ER membrane G6P translocator and the ER lumen G6P phosphatase are essential for glucose production from glycogen (2
). According to this model, which has been sustained by measurements of glucose transport across microsomal vesicles, the ER membrane also contains a glucose transport system for efflux of glucose produced in the ER lumen to the cytosol. This predicted ER transport system for glucose may thus represent a potential target for drug development aiming at control of blood glucose levels.
As a first step towards a better understanding of glucose homeostasis in liver, HepG2 cells were used as a model. HepG2 cells are liver derived but, as with many immortalized cell cultures, display properties of tumor cells, especially the upregulation of GLUT1 gene expression (15
). The advantage for imaging studies is that HepG2 cells are easy to culture and transform, adhere to coverslip surfaces, and thus can be perfused with various glucose and inhibitor concentrations. Such an analysis may be used to unravel the principal components and regulatory networks involved in glucose homeostasis, to identify novel genes involved in transport, metabolism, and regulation, and to develop drug targets, but the results may not be directly applicable to normal hepatocytes in the context of the liver. Insights into altered glucose homeostasis in immortalized cells may also be highly relevant for the development of treatments for cancer. Cancer cells typically show elevated uptake and metabolism of glucose (15
). Many studies of cancer tissue support the hypothesis that the control over glycolytic flux resides primarily at the glucose uptake and phosphorylation steps.
Glucose export from the ER.
Mammalian cells are unique in that gluconeogenesis involves import of G6P into the ER, where it is dephosphorylated, and an unknown function, named the T3 component according to Arion's model, that exports glucose from the ER back into the cytosol (reviewed in references 3
). However, glucose transport across the ER membrane has been a matter of debate. Some studies suggested that the ER (as derived from microsome tracer studies) was not permeable to glucose, while others found evidence for glucose transport with properties differing from those of GLUTs (4
). Moreover, while glycogen storage disease cases can be related to mutations in either the G6P translocator or the G6P phosphatase, no mutations affecting T3 have been described. An alternative model was suggested by Guillam et al., based on findings with GLUT2 knockout mice (17
). While facilitated diffusion of 3-O
-methylglucose in glut2
hepatocytes was reduced by 95%, efflux of d
-glucose was indistinguishable from that of the wild type (19
). This finding led to the suggestion of a vesicular pathway from ER effluxing glucose from the cells (33
). This pathway appeared different from the typical Golgi vesicle trafficking since it was Brefeldin A independent and nocodazole sensitive. Nocodazole was, however, shown to affect GLUT function directly; thus, the existence of this ER-based vesicular pathway is still controversial (26
Vesicular secretion pathways typically make use of active transporters for loading the vesicles. One would thus expect a concentrative step of the ER for efficient export of glucose from the cell to the serum. Concentration of glucose inside the ER lumen can be achieved by the combined actions of the G6P translocator and the G6P phosphatase. Therefore, the presence of a facilitative glucose transport system like that posited in Arion's model would represent a leak, leading to reduction of the glucose gradient and the production of ER-derived vesicles with glucose concentrations similar to that in the cytosol. One of the aims of this study was therefore to determine whether such a leak can be detected by directly comparing cytosolic and luminal ER glucose levels.
Targeting of a glucose nanosensor to the ER lumen.
To measure glucose transport across the ER membrane the FRET glucose sensor FLIPglu-600μ was targeted to the ER lumen of HepG2 cells. ER signal sequences have been extensively used to target heterologous proteins, including GFP variants, to the ER lumen and to retain them in the ER by using the KDEL retention signal (22
). Confocal images support colocalization of FLIPglu-600μER
with an ER marker, while no colocalization was observed with a Golgi marker. Permeabilization of the cells with digitonin leads to rapid loss (<2 min) of fluorescence in cells expressing the sensor in the cytosol, while cells expressing the senor in the ER retain fluorescence in vesicle-like structures that form after addition of digitonin for longer periods (>10 min). Moreover, the difference in galactose accumulation between ER and cytosol supports a differential localization. The data do not exclude the possibility that cotranslational import of the sensor is blocked after interaction with the recognition particle, leading to partial import and thus exposure of sensor domains to the cytosol. However, since GFP and other proteins are efficiently imported into the ER, one would expect that at least the N-terminally fused cyan fluorescent protein should be able to enter the ER lumen. Subsequently, one may envisage that sequences in the periplasmic glucose binding protein lead to a block of transfer. However, in such a scenario, the sensor would not be folded correctly, thus being unable to undergo the conformational change that leads to a FRET change, and would be nonfunctional. The data presented here regarding both localization and functionality indicate that the sensor FLIPglu-600μER
is efficiently imported and retained in a fully functional conformation in the ER lumen.
Properties of the glucose transport system detected at the ER membrane.
The observed changes in steady-state glucose levels in response to perfusion of the cells with glucose shown here permit the following conclusions: (i) the ER of HepG2 cells contains a function for mediating uptake of glucose; (ii) glucose uptake across the ER membrane is faster than transport across the plasma membrane, since no difference is observed between accumulation rates of glucose in the cytosol and the ER; and (iii) the predominant mechanism for efflux of glucose from the ER is facilitated diffusion of glucose.
The hypothesis that ER efflux is mediated by facilitated diffusion across the ER membrane rather than vesicular efflux is based on a number of observations. Given that the glucose concentrations in the ER and cytosol as derived from fractional saturation of the sensor in vivo are similar, two possible scenarios can be envisaged. If efflux through an independent vesicular efflux pathway as suggested by Hosokawa and Thorens exists (19
), then this efflux must compensate for concentrative uptake into the ER; otherwise, we would observe higher or lower glucose levels inside the ER than in the cytosol. Alternatively, rapid equilibration occurs by a facilitator, as suggested by Banhegyi and Mandl (3
). While the measured concentrations and kinetics do not exclude a vesicular route, the similarity in uptake rates at the different external levels does not seem highly probable. More importantly, also the rates of decline of glucose levels after removal of extracellular glucose are highly similar for ER and cytosol. The decline of glucose levels after withdrawal of glucose from the medium in the cytosol is driven by cellular efflux and metabolism via hexokinase I. In contrast, the ER does not contain known enzymes for glucose metabolism; thus, the decline in glucose levels during withdrawal is assumed to be due to efflux through the vesicular or ER transport systems. Again, here it does not seem probable that the vesicular efflux component shows rates similar to the combined efflux from the cytosol and cytosolic metabolism. If the facilitator and vesicular paths coexist, as suggested by Hosokawa and Thorens (19
), fluxes out of the ER should be additive; thus, the decline in ER levels is expected to be faster than decline in the cytosol. The observed relative glucose concentrations and kinetics can however be simply explained by the presence of a high-capacity bidirectional glucose facilitator.
Evidence for glucose transport across the ER membrane was derived from biochemical studies using microsomal membrane fractions from homogenized hepatocytes. The data were interpreted as evidence for pore-like glucose transporter selecting solutes on the basis of size (25
). Another group defined at least two kinetic components for glucose transport across the ER membrane (4
): the slower component had affinities for influx and efflux of glucose in the range of 100 mM, was insensitive to CytB, and was more selective in discriminating between glucose and galactose. Moreover, both data sets concluded that rat liver microsomal vesicles are heterogeneous concerning their glucose transport properties.
The data presented here suggest that the ER glucose facilitator is insensitive to CytB but may also be permeable to galactose. Galactose uptake and efflux kinetics of the ER membrane were significantly slower than the transport across the plasma membrane, suggesting that affinity and Vmax of the facilitator differ for glucose and galactose. Alternatively, it is possible that the ER contains separate transporters for glucose and galactose. Taken together, the differences in the kinetics between glucose and galactose transport would argue against the presence of a nonselective pore.
Numerical model for glucose homeostasis.
The ratio changes and steady-state ratios measured by the nanosensors reflect all components acting on metabolite levels, offering an opportunity to come to a better understanding of homeostasis by building and testing models. The perfusion curves with external glucose could be simulated on the basis of a very simple model comprising reversible transport of glucose across the plasma and ER membranes together with phosphorylation inside the cytosol. Transport across the ER membrane was faster than transport across the plasma membrane, which exceeded phosphorylation of glucose. In this model, the ER constitutes a buffer compartment expanding the accessible space for glucose inside the cell. The decrease in glucose levels inside the cytosol and ER under conditions where transport across the plasma membrane is blocked could be explained simply by phosphorylation of glucose in the cytosol. Internal glucose levels in HepG2 cells were ~75 to 80% of external supply and thus higher than in COS-7 cells, where cytosolic levels were ~50% of external glucose (14
). This is in agreement with the rates of increase in ratio measured for both cell lines after loading with glucose and blocking export across the plasma membrane. The model can be further improved by determining the actual subcellular volumes in HepG2 cells rather than estimating these values by comparison to measurements made for other cell types.
Molecular nature of the predicted glucose facilitator in the ER membrane.
Glucose uniporters have been identified for Saccharomyces cerevisiae
and mammals as members of the MFS (m
). Many of these yeast Hxt and mammalian GLUT transporters are located on the plasma membrane. Plasma membrane targeting of transporters occurs typically via the ER-Golgi pathway; thus, the transporters transiently pass through the ER. If these transport proteins are active in the ER, they might well be responsible for the observed glucose uniport across the ER membrane. Moreover, some members of the GLUT family were found in the ER (33
The redundancy of GLUTs could also explain why mutants in the T3 component of “Arion's cycle” have not been found. GLUTs were found to be differentially sensitive to CytB and to differ in selectivity. However, the CytB insensitivity of ER glucose transport and the differences between plasma membrane and ER transport between glucose and galactose suggest that either a different set of proteins is responsible for ER glucose transport or, at least, the sets of GLUTs contributing to plasma membrane and ER transport are different (Fig. ). Direct analysis of cells lacking GLUT transporters may be a way to characterize the potential role of different GLUT members in this function.
FIG. 7. Glucose (glc) transport across the ER membrane is catalyzed by (A) a different subset of GLUT transporters or (B) a transport protein different from that for transport across the plasma membrane. (C) Import into and export from the ER are catalyzed by (more ...) Potential role of Arion's cycle.
The rapid equilibration of glucose between the cytosol and the ER abolishes differences in glucose levels between the compartments, which result from the different subcellular localizations of glucose-6-phosphatase and hexokinases. Thus, one may speculate that the distinct localizations of G6P forming and hydrolyzing activities are instead involved in compartmentalization of G6P. G6P is not only an intermediate of glucose utilization and production but also a regulator of gene expression and enzyme activity (1
). This hypothesis would be consistent with studies that provided indirect evidence for multiple G6P pools (8
Taken together, the use of genetically encoded nanosensors for glucose targeted to the cytosol and ER provides a unique tool permitting measurements of glucose homeostasis in both compartments and analysis of the exchange of glucose between the compartments. The analysis provides direct evidence for the presence of high-capacity glucose transport systems for import into and export from the ER. The identification of a bidirectional high-capacity glucose transport system suggests that glucose produced in the ER during gluconeogenesis is exported into the cytosol before it is released into circulation. The characterization of ER glucose transport provides a basis for identifying the responsible proteins, which may represent important drug targets for diabetes treatment. One may hypothesize that the ER leak detected here with HepG2 cells is suppressed in the liver and that a vesicular glucose efflux pathway is activated in liver cells. The next step will thus be the analysis of primary hepatocytes or even liver tissue from mice transformed with the sensor. Further optimization of the dynamic range of the glucose sensors has been achieved by semirational design, providing the opportunity to detect more subtle changes (9
). Moreover, novel sensors for metabolites downstream of glucose, i.e., G6P and pyruvate, may be useful to obtain deeper insights into homeostasis. These novel sensors may be engineered on the basis of existing nanosensor scaffolds, using rational redesign of the binding pocket to change the substrate specificity (24
). Such a set of FRET sensors will undoubtedly help to unravel the open questions regarding sugar homeostasis and its regulation and provide a new means for the development of drugs using high-content screens.