Experimental evidence suggesting a type of glucose uptake regulation prevailing in resting and differentiated cells was surveyed. This type of regulation is characterized by transport-limited glucose metabolism and depends on segregation of glucose transporters on microvilli of differentiated or resting cells. Earlier studies on glucose transport regulation and a recently presented general concept of influx regulation for ions and metabolic substrates via microvillar structures provide the basic framework for this theory. According to this concept, glucose uptake via transporters on microvilli is regulated by changes in the structural organization of the microfilament bundle, which is acting as a diffusion barrier between the microvillar tip compartment and the cytoplasm. Both microvilli formation and the switch of glucose metabolism from “metabolic regulation” to “transport limitation” occur during differentiation. The formation of microvillar cell surfaces creates the essential preconditions to establish the characteristic functions of specialized tissue cells including the coordination between glycolysis and oxidative phosphorylation, regulation of cellular functions by external signals, and Ca2+ signaling. The proposed concept integrates various aspects of glucose uptake regulation into a ubiquitous cellular mechanism involved in regulation of transmembrane ion and substrate fluxes.
- glucose transport
- membrane limitation
- insulin action
a vast quantity of information about transport regulation by insulin, “insulin-like” effectors, and other environmental signals has been accumulated in the past. Yet, the precise biochemical mechanism(s) by which all these effects are mediated is still unclear. The only mechanistic concept proposed by Cushman and Wardzala (36) and Suzuki and Kono (183), the “translocation hypothesis,” explains insulin-induced transport stimulation by exocytic insertion of additional glucose transporters from an intracellular vesicular storage pool into the plasma membrane. Despite a considerable amount of conflicting data and the obvious failure to integrate the great variety of insulin-independent forms of transport regulation into a common mechanistic framework, the translocation hypothesis has remained the sole concept until now. The manner in which various effectors and conditions such as glucose starvation and glucose feeding (glucose-curb), endogenous demand in muscle cells, hydrogen peroxide, SH-reagents, high Ca2+, and physiochemical conditions (high pH, hyperosmolarity, and shear stress) modulate transport rates has remained largely unexplained.
Applying the same experimental techniques used to detect insulin-mediated translocation effects shows that some forms of insulin-independent transport modulation appear to follow the same mechanism of transporter recruitment from a cytoplasmic pool to the plasma membrane. Thus glucose deprivation (73), virus transformation (95), high pH and hyperosmolarity (194), action of tumor promoters (94), and cold treatment (64) apparently cause the same type of exocytic response. However, a number of other effectors and conditions do not induce transporter redistribution, although their action on transport is not additive to the insulin effect. Two examples are stimulation of glucose transport by hyperosmotic exposure (32) and metabolic stress in clone 9 cells (8, 68,166-169). In the latter case, the authors explicitly stated that the predominating mechanism mediating the early response to inhibition of oxidative phosphorylation is activation of GLUT-1 glucose transporters preexisting in the plasma membrane. Dissociation of translocation and transport effects occurs in many experimental systems. Even the total lack of transporter appearance or disappearance on the surface during transport modulation has been reported (49,120, 184). Similarly, Fisher and Frost (49) demonstrated that glucose deprivation increased the rate of glucose transport in 3T3-L1 adipocytes eight- to tenfold without changing the distribution of GLUT-1 and GLUT-4 in the plasma membrane and microsomal fractions. They concluded, “the increase in transport activity associated with glucose deprivation does not result from the translocation of either of the glucose transporters known to exist in 3T3-L1 adipocytes.” Even the numerous techniques for cell surface labeling of glucose transporters did not solve the main problem of the translocation hypothesis, which is that the appearance of new transporters on the cell surface is unable to fully account for the observed activation of glucose transport rate (152, 156,196). This obvious shortcoming is reflected by the use of operative or descriptive terms such as “intrinsic activity” or “masking and demasking effects.”
In a recent study, Zierler (210) analyzed the relation between the insulin-induced translocation and the transport effect reported in five relevant papers. Using a model-free mathematical analysis, he came to the conclusion that the concentration of GLUT-4 in the plasma membrane in response to insulin would have to have been two- to sevenfold greater than that actually observed if translocation alone were to account for the observed increase of glucose uptake. That is, insulin-induced increase in GLUT intrinsic activity accounts for more than half of the observed increase in glucose uptake.
As pointed out by Olefsky (146), tissue-specific expression of the so-called insulin-sensitive glucose transporter GLUT-4 alone is not the only factor conferring insulin-stimulated glucose transport on these tissues, because heterologous expression of GLUT-4 in other tissues does not confer insulin-stimulated GLUT-4 translocation to non-muscle or non-adipose cells. Consequently, additional factors in muscle and adipose cells, related to insulin signaling or vesicle trafficking, must exist. It appears that exocytosis of transporter proteins to the plasma membrane does not completely describe the mechanism by which insulin stimulates glucose transport, because a yet unrecognized mechanistic feature is involved in transport regulation.
More than 10 years ago, a novel mechanism of transport regulation was proposed to be able to explain many of the above-mentioned experimental data (104, 105, 110, 111). This concept was induced by the finding that the GLUT-4 transporters in differentiated 3T3-L1 adipocytes are exclusively located on microvilli. Furthermore, transport of hexoses via these microvillar transporters was found to be low unless shortening of microvilli or disorganization of the microvillar actin filament bundle occurred. Thus the microvillar cytoskeleton was proposed to function as a diffusion barrier between the microvillar tip compartment (entrance compartment) and the cytoplasm. At that time, however, this hypothesis was widely rejected.
In the meantime, compelling evidence in favor of the regulated microvillar pathway has accumulated, which has turned out to be a multifunctional device for modulation of substrate and ion fluxes via transporters and ion channels (see reviews, Refs.101-103). The same mechanistic components are used even by one of the most important cellular signal systems, the Ca2+ signaling pathway (101). In addition, several recent findings complete the general outline of this concept; e.g., the role of phosphatidylinositol (PI) 3-kinase in insulin-stimulated transport, some biochemical aspects of microvilli formation, and, most importantly, the finding that almost all types of glucose transporters are preferentially localized on microvilli of differentiated tissue cells.
The following is a survey on this concept and its ability to integrate a large body of experimental data into the common concept of cellular regulation via microvillar pathways. It should be stated, however, that the proposed mechanistic scheme did not invalidate the translocation concept of insulin action, since it solely concerns the immediate rapid component of transport activation preceding exocytosis. Exocytic recruitment of membrane proteins undoubtedly belongs to the well-established consequences of hormonal and even some nonhormonal forms of cell stimulation and clearly contributes to the overall effects on transport.
REGULATION OF GLUCOSE METABOLISM IN DIFFERENTIATED CELLS
Metabolic Regulation Vs. Membrane Limitation
Depending on the state of cellular differentiation, glucose uptake via the facilitated diffusion system responds to changing cellular energy demands by two different mechanisms (see reviews, Refs.45 and 84). The rate of glucose uptake into undifferentiated, rapidly growing cells such as embryonic and tumor cells or preconfluent cells in culture largely depends on the activity of certain rate-limiting glycolytic enzymes, mainly hexokinase and fructose 6-phosphate 1-kinase or other downstream enzymes, which are regulated by adenine nucleotides and intermediates of the carbohydrate metabolism itself. In such systems, low demand of metabolic energy causes a bank-up of metabolic substrates upstream of the rate-limiting enzyme(s), which finally results in accumulation of intracellular free glucose and decreased net uptake of glucose. This type of glucose uptake regulation, called “metabolic regulation,” is generally characterized by high levels of intracellular free glucose almost reaching the extracellular value, accompanied by considerable lactate production under aerobic conditions.
A completely different way of metabolic regulation is observed in muscle cells, adipocytes, and other types of differentiated tissue cells. Here, metabolic activity is regulated by external and internal signals via modulation of the transport activity across the plasma membrane. The “transport (or membrane)-limited” metabolic type is characterized by very low concentrations of intracellular free glucose. Membrane limitation is an essential prerequisite for regulation of glucose metabolism by exogenous signals.
Although most of the recent work on transport regulation was done using membrane-regulated cell types, especially adipocytes and muscle cells, very little is known about the nature of the membrane-limited state itself and the specific mechanisms that enable cells to switch from metabolic control to membrane regulation. Since the days of Otto Warburg, the problem of why tumor cells, in contrast to their differentiated counterparts, exhibit aerobic glycolysis has not been adequately resolved. One reason for this failure of classical biochemistry is the original suggestive question of defects in tumor cells, which forced research in the wrong direction. In fact, the critical gap in knowledge concerned metabolism of normal cells, not tumor cells, and one should have better asked for the specific organization of energy metabolism in normal tissues, which enable differentiated cells to coordinate mitochondrial oxidative phosphorylation and glycolysis in a way that avoids excessive lactate production. This problem, however, remained largely unnoticed until now.
Membrane-Limited Hexose Utilization
Early studies on endogenous transport regulation have yielded some insight into the nature of the membrane-limited state and led to the assumption that glucose metabolism is regulated on the level of the membrane transport. The starting point of these studies was the observation of a special type of endogenous regulation of hexose uptake occurring under specific experimental conditions in C6 glioma cells, a permanent cell line exhibiting membrane-limited glucose metabolism at confluence (93, 113).
In this cell line, inhibition of the glycolytic ATP production by either pretreatment with iodoacetate or use of 2-deoxyglucose (2-DG) as transport substrate resulted in a conspicuous time-dependent periodic fluctuation of hexose uptake rates (110). Although uptake rates changed up to sixfold, the intracellular concentration of free glucose remained at a very low and completely constant level (1–2 orders of magnitude lower than the extracellular concentration) (113). Constant levels of intracellular free glucose despite rapidly changing uptake rates pointed to an effective coupling between transport and phosphorylation of glucose. This tight coupling between membrane transport and hexokinase reaction strongly pointed to a specific organization of these processes entirely different from that of metabolic regulation.
Further studies revealed that the observed periodic fluctuation of 2-DG uptake is triggered by changes in the ATP content of a special cellular compartment, called the “entrance compartment” for hexose transport. This compartment, the site of hexose entry into the cell, appeared to be localized within microvilli and other lamellar surface protrusions (110). Moreover, experimental evidence based on 2-DG and 3-O-methylglucose (3-OMG) uptake kinetics as well as electron microscopy suggested that this entrance compartment is separated from the cytoplasmic main compartment by a cytoskeletal diffusion barrier formed by the central core of actin filaments of microvilli (110, 111). The effectiveness of this diffusion barrier is regulated via the cytoplasmic ATP/ADP system, allowing substrate uptake in response to the actual energy demand.
The development of membrane-limited hexose metabolism is based on the differentiation-dependent sequestration of glucose transporters (and other integral membrane proteins such ion channels) to special plasma membrane domains localized on the tips of microvilli (105) (Fig. 1).
The entrance compartment is separated from the cell body by a diffusion barrier attenuating the flow of low-molecular-weight solutes into the cytoplasm. This barrier is formed by the microfilament bundle, a dense central structure tightly connected to the covering lipid membrane of the microvillar shaft.
The diffusion resistance between the entrance compartment and the cell body is regulated by structural changes of the cytoskeletal core. Elongation or shortening of the microfilaments is triggered by changes in the ATP and ADP concentrations of both the entrance compartment and the cytoplasm. Exogenous signals such as insulin or other growth factors also induce shape changes of microvilli or even complete integration of the microvillar membrane portion into the plasma membrane via depolymerization and reorganization of the microfilament bundle (104, 105, 111).
The biochemical equipment of the entrance compartment, which allows the autonomous production of glycolytic ATP as well as the assembly and disassembly of microfilament bundles, imparts to this system self-organizing and self-regulating properties responding to environmental hexose supply. High external glucose concentrations cause reversible elongation of the microfilament bundle, thereby enhancing the internal diffusion resistance of the microvillar pathway for glucose and glucose phosphates (111, 163-165).
To simplify the discussion of this system, an idealized cell model is used in which the numerous spaces of the microvillar tips are combined, giving the topologically equivalent form with one large tip compartment (entrance compartment) (Fig.2). However, the unique arrangement of the finely dispersed peripheral tip spaces in the multicompartment system imparts specific properties to the entrance compartment, which must be kept in mind. For instance, the very small internal water spaces of the tip compartments rapidly follow environmental conditions by equilibration via cation/anion channels and transporters located within the tip membranes (see Transport Activation by Hyperosmotic Conditions). Thus the microvillar tip compartment rather reflects the composition of the external medium than that of the cytoplasm. Because of this unique property, the microvillar tip compartment is qualified as a sensor for the rapid detection of relevant environmental changes. Some of the cellular responses on changes of external ionic conditions and glucose supply are discussed in Hyperosmotic treatment coactivates glucose transport and ion fluxes.
Essential Functions of the Microvillar Surface Organization in Differentiated Cells
Microvilli contain a considerable portion of the total surface membrane of the cell. Under hypotonic conditions, this membrane reserve is used for cell enlargement, enabling cells to expand their cytoplasm to several times their original volume without membrane damage. On the other hand, swelling-induced shape changes of microvilli cause activation ion fluxes necessary for regulatory volume changes (103).
Another important differentiation-dependent cytoplasmic parameter is the extremely low cytosolic Ca2+ concentration that is an absolute prerequisite for Ca2+ signaling. Maintenance of the steep Ca2+ gradient between the extracellular and intracellular space is essential for regulation of cellular functions by cytosolic Ca2+. The microvillar diffusion barrier system provides for both the effective blockade of the influx and efflux pathways for cations and the high-affinity Ca2+ storage system to adjust the low basal [Ca2+] levels (102,103, 108) (Fig. 3). As recently proposed, cellular Ca2+ influx via microvillar ion channels is inhibited by the microvillar diffusion barrier (106-108,114, 115). The polyanionic nature of the microfilament bundle, acting as cation exchanger matrix with a high density of fixed charge centers (122, 189; surveyed in Ref. 102), represents a highly effective barrier preventing influx of divalent cations such as Ca2+ and Mg2+. The phospholipase C (PLC)-coupled pathway, known to induce disassembly or reorganization of actin filaments via activation of the phosphatidylinositol 4-phosphate (PIP)/ phosphatidylinositol 4,5-bisphosphate (PIP2)-regulated actin-binding and -severing proteins profilin, gelsolin/villin, and cofilin (59, 60,116), triggers Ca2+ release from microvillar actin filaments (108) and at the same time opens the influx pathway for Ca2+. Because of the biochemical identity of the internal Ca2+ store and the diffusion barrier system, Ca2+ release is functionally coupled with the activation of Ca2+ influx (store-operated Ca2+ pathway). This mechanism is schematically depicted in Fig. 3 (see review, Ref. 101). The morphological aspect of this process has been visualized by scanning electron microscopy (SEM) of vasopressin-stimulated rat hepatocytes (109, 115). Both the mechanism and morphology of receptor-stimulated Ca2+influx via the microvillar influx pathway closely resemble the proposed mechanism of glucose transport regulation.
A further important aspect of cellular differentiation is the specific cellular organization of the glucose metabolism, which is closely related to the expression of microvillar cell surfaces and enables tissue cells to conduct just as much glucose through the glycolytic pathway as necessary for the highly efficient ATP production via the oxidative pathway.
The Membrane-Limited State
Since Warburg's observation that tumor cells differ from their normal counterparts by lactate production under aerobic conditions, the biochemical basis of this and two other metabolic features of malignant cells, inhibition of glycolysis by oxygen (Pasteur effect) and inhibition of respiration by glucose (Crabtree effect), have remained largely unexplained. Pasteur and Crabtree effects are generally explained by competition of the two main metabolic pathways, respiration and glycolysis, for their common phosphorylation substrates, ADP and Pi. However, the question of why normal differentiated cells are lacking this competition has remained unanswered.
With respect to the type of hexose utilization, transition of nondividing differentiated cells to the rapidly proliferating state, regardless of whether induced by oncogenic transformation (74,81) or mitogenic stimulation, can be characterized as change from membrane limitation to metabolic regulation. The membrane-limited state differs from that of metabolic regulation by the special type of glucose uptake regulation at the membrane level, which excludes competition of glycolysis and respiration for adenine nucleotides because the ATP/ADP system acquires a signal status for regulation of glucose entry. Abolition of membrane limitation also can be induced by other conditions including metabolic stress; e.g., by 2,4-dinitrophenol-induced reduction of cellular ATP (110), receptor-mediated stimulation (66, 35, 41), or glucose starvation (73). The reverse process is observed in cell cultures at confluence or under differentiating conditions (111).
Loss of membrane limitation can be detected by three experimental criteria. First, the steady-state content of free intracellular glucose increases up to the extracellular level (93, 113). Second, aerobic lactate production increases (46, 165). Elevated aerobic lactate production, most obvious after oncogenic transformation, also occurs under other experimental conditions such as inhibition of respiration (134) and stimulation by growth factors (35, 40, 41). Third, abolition of membrane limitation is accompanied by shape changes or shortening of microvilli, or even their complete integration into the plasma membrane (46,61, 88, 111, 118, 143, 174).
In membrane-limited systems, the length of microvilli is critically determined by the cytoplasmic ATP/ADP system, which in turn depends on the availability of C3 substrates for mitochondrial ATP production. This way, a complete regulatory circuit precisely adapts glucose uptake to cellular ATP demand (Fig.4). Because of this regulatory principle, pyruvate/lactate, produced via the Embden-Meyerhof pathway, is almost exclusively used for oxidative phosphorylation, yielding 18 times more ATP per mole of glucose than glycolysis alone. Only small amounts of lactate can leave the cell under these conditions.
The morphological aspect of membrane limitation.
During adipose conversion of 3T3-L1 fibroblasts, these cells acquire the membrane-limited state as indicated by the appearance of the endogenous regulation of 2-DG uptake (111). The morphological aspect of this transition is shown in Fig.5. Adipose conversion is accompanied by the outgrowth of numerous microvilli on the cell surface (Fig. 5,B and C) (111). Stimulation of hexose uptake by either endogenous signals such as ATP depletion (Fig.6 A) or insulin treatment (Fig.6 B) is paralleled by conspicuous shape changes of microvilli assuming a voluminous saclike appearance (111). Apart from electron microscopic evidence, enlargement of the internal microvillar compartment also was verified by demonstrating that insulin induces a significant (2-fold) increase of the cellular distribution space for 3-OMG (111).
Furthermore, removal and isolation of microvilli from the cell surface of intact 3T3-L1 adipocytes revealed that microvilli contain nearly 90% of the surface-exposed GLUT-4 transporters comprising the total insulin-sensitive transporter pool of these cells (104). The microvillar localization of tissue-specific glucose transporters has been shown for a number of other cell types.
Localization of glucose transporters on microvilli.
Several different isoforms of transporter proteins are involved in specific tissue functions related to regulation of hexose metabolism (Table 1).
The various isoforms are clearly designed to serve different functions. Those tissues that exclusively depend on glucose as energy-providing substrate, including blood cells, brain, muscle, and fat cells, express the high-affinity transporters GLUT-1, GLUT-3, and GLUT-4. Insulin-regulated tissues such as fat and muscle cells express the GLUT-4 isoform, whereas all these tissue cells and most other cells types during growth (embryonic state) synthesize GLUT-1 alone.
In contrast, tissue cells that are involved in homeostasis of blood glucose express the low-affinity isoform GLUT-2. In these latter cell types, the working point of the transporter is strongly elevated matching the specific tissue function. For instance, pancreatic β-cells carry GLUT-2 on their basal microvilli, which are serving as glucose sensors. Increased glucose uptake in response to hyperglycemia elevates glycolytic ATP within a specific cellular subcompartment and activates insulin secretion via Ca2+influx through voltage-sensitive Ca2+ channels. Similarly, the homeostatic function of liver cells mainly rests on feeding glucose into the blood stream under hypoglycemic conditions. Stimulated glycogenolysis elevates the intracellular glucose concentration to more than 10 mM, sufficient for release of glucose via the low-affinity transporter GLUT-2.
An essential prerequisite of the outlined concept of membrane limitation is the preferred localization of glucose transporters on microvilli. In the past years, almost all transporter isoforms have been demonstrated in microvilli of a number of differentiated cell types using immunoelectron microscopy:
GLUT-1 is localized on microvilli of the rat oviduct (185).
GLUT-1 is localized primarily to basolateral membranes of fully differentiated Caco cells, whereas the differentiation-specific transporter GLUT-3 was found predominantly on apical microvilli (70).
GLUT-1 is localized on the microvillar surfaces of endothelial or epithelial cells of the blood-tissue barriers (186).
GLUT-1 is concentrated at both the microvillous apical and the infolded basal plasma membrane of the syncytiotrophoblast. (91).
GLUT-2 is specifically localized to microvilli of pancreatic β-cells (148).
Reduced GLUT-2 on microvilli of β-cells is correlated with diabetic loss of glucose-stimulated insulin secretion (149).
GLUT-2 is localized in sinusoidal microvilli of hepatocytes (144).
GLUT-5 (fructose transporter) is found on apical microvilli in S3 proximal tubules of the rat kidney (180).
Further evidence has been acquired by studying isolated microvillar membrane fractions with immunoblot analysis and cytochalasin B binding techniques:
GLUT-1 is found in microvilli of rod outer segments of the photoreceptor cells (corresponding to microvillar surfaces) colocalized with hexokinase, aldolase, glyceraldehyde-3-phosphate dehydrogenase, phosphoglycerate kinase, and lactate dehydrogenase (78).
Insulin stimulates hexose uptake in 3T3-L1 cells up to tenfold. Consequently, about 90% of the total surface-localized transporter pool should be localized on microvilli. With the use of the cytochalasin B-binding technique, this type of transporter distribution was found in 3T3-L1 adipocytes (104, 105). The reason for this unexpected clustering of transporter proteins in microvillar membranes turned out to be surprisingly simple and revealed a new, hitherto unknown aspect of cellular differentiation.
Formation of microvilli on differentiated cells.
In rapidly dividing cells, translocation of newly synthesized proteins into the plasma membrane continuously occurs by exocytosis of trans-Golgi vesicles. This process leads to insertion of additional lipids and integral membrane proteins into the cell surface. Subsequently, the newly inserted integral membrane proteins migrate out of their original lipid microenvironment and rapidly distribute over the whole cell surface by lateral diffusion (Fig.7).
This type of surface processing of integral membrane proteins is significantly changed in resting (G1/0 arrested) or differentiated cells. Soon after induction of adipose conversion by 48-h cultivation of 3T3-L1 cells with dexamethasone, isobutylmethylxanthine, and insulin, SEM (Fig. 5, A–C) revealed a large number of small spherical surface protrusions. At later stages of cultivation, these exocytic blebs grow out from the plasma membrane, forming the tips of new microvilli (Fig.5 C).
In nondividing or differentiated cells, exocytosis results in the formation of stable spherical membrane blebs that are not integrated into the plasma membrane. Stabilization of these surface protrusions is due to a proteoglycane coat that inhibits lateral diffusion of the integral membrane proteins located in this membrane domain. After exocytosis, a submembrane actin filament bundle forms the central core structures of microvilli bearing the original exocytic membrane domain at their tips (Fig. 8). Achler et al. (1) described the cell surface events accompanying microvilli formation on intestinal epithelium: “Formation of the basolateral microvilli required polymerization of actin and proceeded at glycocalyx-studded plaques that resembled the dense plaques located at the tips of the apical microvilli.” Because of surface coat stabilization of the microvillar membrane, the original protein composition of exocytic vesicles is preserved in microvillar tip domains, which then contain all those integral membrane proteins that are expressed by the differentiated tissue cell. This type of surface distribution of integral membrane proteins on differentiated cells has been confirmed by biochemical techniques in the following way (105).
Restricted mobility of integral membrane proteins in exocytic membrane domains of differentiated cells.
Some time ago, the existence of a novel type of glucose transporter was discovered in adipocytes and muscle cells (see review, Ref.63). This transporter species, named GLUT-4, is expressed only in insulin-sensitive tissues such as fat and muscle cells. In cultured 3T3-L1 adipocytes, both GLUT-4 and the erythrocyte/brain transporter type GLUT-1 are expressed at the same time. However, in contrast to GLUT-1 that is synthesized in both 3T3-L1 fibroblasts and adipocytes, GLUT-4 does not appear before day 4 after induction of adipose conversion (55, 92). Because at this time the formation of microvilli on the adipocyte surface had started, the distribution of the GLUT-4 protein between plasma membrane and microvilli was used to test the above-postulated inhibition of lateral mobility in the microvilli-forming surface blebs. If it were true that part of the GLUT-1 protein is expressed before adipose conversion, this transporter isoform should occur within both the plasma membrane fraction and microvillar vesicles; in contrast, GLUT-4 should be present in the microvillar fraction only. This type of transporter distribution was indeed observed. Whereas GLUT-1 occurred in equal amounts in both the plasma membrane and the microvilli fraction, GLUT-4 was detected almost exclusively in the microvillar fraction (105).
The result of this study strongly supports the idea of restricted lateral mobility in newly inserted membrane areas of differentiated adipocytes. Moreover, the exclusive localization of transporter proteins on microvilli of differentiated cells is explained in a rather compelling way, since even those integral membrane proteins that have reached the plasma membrane vanish with time because of proteolytic degradation. Restricted lateral mobility of integral membrane proteins depends on the presence of proteoglycans on the cell surface (138, 139). Calvo et al. (22) reported the increased expression of surface coat components in 3T3-L1 fibroblasts during their differentiation to adipocytes. The transmembrane connection between proteoglycans of the surface coat and the microfilaments of microvilli has been demonstrated by Carothers Carraway et al. (26). Thus proteoglycans are intracellularly and extracellularly connected with each other, forming domain-stabilizing cage structures.
Reduced lateral mobility of integral proteins in glycocalyx-covered plasma membrane regions was described more than 20 years ago. Studying the cell surface distribution of insulin receptors on human placental syncytiotrophoblasts, Nelson et al. (138) noted: “Insulin receptors are specifically associated with the glycocalyx region of microvilli on the surface membranes of microvilli. No insulin receptors were detectable in association with the intermicrovillous plasma membrane even though its glycocalyx is in direct continuity with the glycocalyx of microvilli … , which suggests that there is not complete freedom of lateral mobility of the insulin receptors in the surface membrane of this tissue.” The exclusive localization of insulin receptors on microvilli of lymphocytes (30), hepatocytes and hepatoma cells (27), pancreatic β-cells (9), and 3T3-L1 adipocytes (48, 173) was confirmed later. The epidermal growth factor (EGF) receptor is another example of the microvillar localization of cell surface receptors (155, 209).
In proliferating cells, the surface coat is removed by serum proteases such as plasmin and thrombin, which are activated by tissue-specific ectoproteases of the plasmin activator type. These ectoenzymes are tightly bound to the cell surface via proteoglycans. During adipose differentiation in the presence of dexamethasone, the expression of plasmin activator is diminished, and a specific protease inhibitor, plasmin activator inhibitor, is induced. This inhibitor also is bound to the cell surface coat as a complex with its substrate, the plasmin activator (85, 132). The significance of serum proteases for cell differentiation also follows from the fact that cultivation of 3T3-L1 fibroblasts with heparinized medium immediately induces adipose conversion and microvilli formation (17, 208). Similarly, omission of serum is a widely used condition to differentiate cells in culture.
Microvilli formation is generally observed in cultured cells during growth arrest in G0/G1 phases of the cell cycle or in the differentiated state (17, 42, 53, 193,). Under these conditions, newly synthesized functional membrane proteins are located exclusively on microvilli tips and thus are no longer able to support maximal rates of cellular metabolism, because the cytoskeletal diffusion barrier of the microvillar shaft attenuates the flux of hexoses and ions into the cytoplasm. Thus the specific metabolic state of membrane or transport limitation is established, allowing external regulation via receptor-mediated signaling.
Biochemistry of Microvilli Formation
Roles of PI kinases and actin-binding proteins.
The specific organization of the microvillar cell surface essentially depends on the presence of PI 4- and 5-kinases [PI 4(5)-kinases] involved in PIP and PIP2formation. The interaction between PIP/PIP2 and several actin-binding proteins, most importantly those of the gelsolin/cofilin/villin family, cause these capping proteins and associated actin monomers to be translocated to PI 4(5)-kinase-containing membrane domains. These membrane microdomains are thus the destination points for the transfer of actin monomers from the cytoplasm to the plasma membrane (see review, Ref. 179). In the case of gelsolin and villin, the cytoplasmic 1:2 complex consisting of one gelsolin/villin and two actin monomers liberates one actin monomer after binding of the complex to phospholipids. The membrane-bound 1:1 complex acts as a nucleation point for the formation of microfilaments at the cytoplasmic side of the plasma membrane (72).
Actin monomers are thus shuttled from the cytoplasm to the inner face of the plasma membrane and used for the localized assembly of new microfilaments. Gelsolin and villin both have been identified as components of microvilli in epithelial cells (1, 12, 42,72). Expression (53) or microinjection of villin into fibroblasts results in the loss of stress fibers and the formation of large surface microvilli (52). In this way, the occurrence and activation of PI 4(5)-kinases on the inner face of the plasma membrane initiate a rapid and spatially directed reorganization of cytoplasmic actin to the cortical locations of PI 4(5)-kinases (see review, Ref. 179). Most interestingly, PI 4-kinase has been identified as a component of the GLUT-4-containing vesicles from rat adipocytes (38) and of HeLa and Chinese hamster ovary cell microsomes most likely containing microvillus-derived membrane components (75, 204).
Subsequent to the formation of microfilament bundles, the plasma membrane tightly covers the lateral surface of the bundles by formation of specific noncovalent interactions between certain phospholipids and F-actin domains. These interactions have been demonstrated and studied by Gicquaud (58), Grimard et al. (67), and Taylor and Taylor (190). Using electron microscopy, they observed the formation of two-dimensional paracrystalline sheets of parallel fibers in registers or in a netlike organization on lipid layers of phospholipids. The bindings between phospholipids and F-actin depend on the presence of divalent bridging ions such as Mg2+ or Ca2+ or on positively charged lipids. Gicquaud (58) proposed a model in which a limited number of phospholipid molecules interact with specific sites on the actin molecule. In contrast, apical microvilli of the polarized epithelium are stabilized by the presence of membrane-microfilament linker proteins such as villin, fimbrin, and myosin I. Consequently, they exhibit increased mechanical resistance compared with their basal counterparts.
Specific linker proteins are required for stable residence of integral membrane proteins in microvilli.
The preferred or exclusive localization of various types of transporters and ion channels including nonselective cation, Na+, K+, and anion channels on microvilli is documented for a multitude of different cell types. Most likely, stable localization of membrane proteins in microvilli affords direct or indirect binding of their cytoplasmic domains to microvillar actin filaments. Otherwise, unfixed membrane proteins would leave the microvillar membrane domain during transient stimulation.
Recently, several authors, including Yonemura and Tsukita (207), Yonemura et al. (206), Bonilha et al. (18), Matsui et al. (130), Pakkanen and Vaheri (151), Berryman and Bretscher (12), Sato et al. (159), and Oshiro et al. (150), have demonstrated the central role of specific linker proteins such as the ezrin/radixin/moetin (ERM) proteins for the formation and maintenance of microvilli (see review, Ref. 195). According to these studies, the presence of activated (phosphorylated) ERM linker proteins is an essential precondition for the development of microvillar cell surfaces (56). Moreover, Matsui et al. (130) pointed out that activation of moesin requires its phosphorylation and the presence of PIP2. Thus the localized action of both PI 4(5)-kinases and linker proteins is essential for microvilli formation.
An increasing number of integral membrane proteins localized on microvilli, including glucose transporters and ion channels, have been shown to interact with the actin cytoskeleton via specific linker proteins. Presently known examples are the K+ channel (86), the epithelial Na+ channel (171,211), the erythrocyte anion channel (162), the cystic fibrosis transmembrane conductance regulator (CFTR) (170), Na+-K+-ATPase (39), the adhesion receptors CD44, CD43, intercellular adhesion molecule (ICAM) (206), and the Na+/H+ exchanger (44). As it appears, a specific group of functional proteins is selected to reside exclusively within microvillar membranes because of their stable binding to cortical actin filaments, whereas other proteins lacking binding sequences for linker proteins may have a limited residence time in microvilli.
Recently, Kao et al. (90) demonstrated that the insulin-sensitive glucose transporter GLUT-4 (but not GLUT-1) is connected to the cytoskeleton via a specific linker protein identified as the glycolytic enzyme fructose 1,6-bisphosphate (FDP) aldolase. This interaction is modulated by insulin as well as by glycolytic intermediates. The aldolase substrate FDP and, to a much lesser extent, glyceraldehyde 3-phosphate (G3P) inhibit binding of the enzyme to GLUT-4. In intact 3T3-L1 adipocytes, insulin action decreases the interaction between GLUT-4 and aldolase. Moreover, treatment of permeabilized cells with FDP or 2-DG and microinjection of aldolase-specific antibodies inhibits insulin-induced GLUT-4 translocation.
Whereas the interaction of aldolase with actin filaments is textbook knowledge (99, 198), the cytoskeletal association of GLUT-4 via aldolase is a novel aspect with far-reaching consequences. Metabolic regulation of the association between cytoskeleton and aldolase is suggested by the finding that the interaction of F-actin with aldolase is inhibited by FDP, G3P, fructose 1-phosphate, and dihydroxyacetone phosphate (198). Within the range of physiological metabolite concentrations, the binding between aldolase and GLUT-4 appears to be affected predominantly by FDP (90). As it appears, low concentrations of intermediates from the upper part of the glycolytic pathway favor cytoskeletal fixation of glucose transporters within microvillar surface structures. Because low intermediates are characteristic of starvation conditions, it is an intriguing assumption that one mechanistic aspect of deprivation-induced cell differentiation may be the formation of cytoskeleton-transporter interactions to generate or stabilize microvillar structures and to establish or stabilize the differentiated status of membrane limitation.
On the other hand, activation of the microvillar influx pathway via insulin receptor activation usually leads to increased glucose uptake and high metabolite levels. As shown by Kao et al. (90), the cytoskeletal connection to transporters is transiently lost during insulin action. Transient release of the cytoskeletal fixation allows a time-limited downregulation of glucose transport activity via endocytosis of transporter-containing membrane domains. Together, these data suggest that formation of GLUT-4-containing microvilli is determined by several mechanistic aspects:
Directed formation of actin filaments occurs by shuttling of actin monomers by actin-binding and capping proteins to the sites of PIP and PIP2 formation, the location of PI 4-kinase, which is a component of GLUT-4-containing exocytic vesicles (38).
Binding of the transporter protein GLUT-4 to actin filaments via FDP aldolase facilitates the formation of microvilli and allows the restoration of the original microvillar organization after receptor-mediated or starvation-induced activation.
The cytoskeletal association of GLUT-4 is subject to metabolic regulation by the aldolase substrate FDP responding to the state of cellular glucose supply.
INSULIN ACTION ON GLUCOSE UPTAKE
Cell Surface Morphology of Insulin Action
Ten years ago, SEM was used to study changes in the cell surface morphology of 3T3-L1 adipocytes after short (5–10 min) periods of insulin treatment (111). As shown in Fig. 6 B, microvilli lose their original shapes, assuming a saclike appearance. In some cases, the original small upper part of the microvillus is still visible and only its stalk has widened, giving the whole structure an appearance of bagpipes. At later stages of insulin action, microvillar membranes are largely integrated into the adjacent plasma membrane.
Abolition of the microvillar diffusion barrier results in a conspicuous enlargement of the intracellular water space, as shown by 3-OMG equilibration experiments (111). In 3T3-L1 adipocytes, the time course of the insulin-induced increase in the distribution space of 3-OMG follows the same time and dose dependency as that of glucose uptake stimulation by insulin. With the use of the same experimental protocol, even the rapid initial uptake of 3-OMG into the microvillar entrance compartment of unstimulated adipocytes could be demonstrated. Equilibration of the microvillar tip compartment precedes the long-lasting basal influx kinetics. The experiments further showed that the rate of the initial component of basal cells is identical with the uptake rate of insulin-stimulated cells (111). The same biphasic uptake kinetics of 3-OMG have been observed by Whitesell et al. (201) in 3T3-L1 cells, with a rapid initial half-time of equilibration of 1.7 s and a second slower component with a half-time of 23 s. Thus correlated morphological and functional data confirm the diffusion barrier concept for glucose uptake regulation.
GLUT-4-Containing Vesicles Can Be Sheared From the Cell Surface
The mechanism of insulin action on hexose transport was studied further by isolation of a cell surface-derived membrane fraction consisting of microvillus-derived vesicles. This fraction was prepared from intact 3T3-L1 adipocytes by use of a hydrodynamic shearing technique originally developed by Carothers Carraway et al. (25) for isolation of microvilli from tumor cells and later used for microvilli isolation from cultured cells (151). This technique is based on gentle shearing forces exerted on the cell surface by pressing a cell suspension through hypodermic needles. Although devoid of microvilli, most of the cells (80–90%) remained intact after this procedure. The cells were then subjected to homogenization and subcellular fractionation according to the protocol of McKeel and Jaret (131), as modified by Gibbs et al. (57).
As shown by cytochalasin B binding, the microvillar fraction isolated from 3T3-L1 adipocytes contains 90% of the surface-exposed glucose transporters, including the total insulin-sensitive transporter pool of these cells (105, 106). Only 5% of the total cellular transporter protein is found in the plasma membrane, nearly 60% in the microvillar fraction, and 37% in the microsomal fraction (105). Pretreatment of the cells with insulin (10 min) shifts ∼40% of the transporters from the microvillar fraction to the plasma membrane fraction.
Microvillar Vesicles Are Part of Microsomal Fractions
Applying the same sedimentation protocol without the shearing procedure before cell homogenization, the total insulin-sensitive transporter pool was found in the microsomal fraction. Obviously, the cell surface-derived microvillar vesicle fraction, generated by the low-force shearing technique, is also formed during normal homogenization using the Teflon-glass homogenizer. In either case, hydrodynamic shearing forces were used. However, the two methods differ considerably in the strength of the generated shear forces, which depend mainly on the dimension of the aperture through which the cell suspension is pressed. In contrast to the Teflon-glass homogenizer, which usually has a clearance of 90 μm, the needles used for microvilli preparation have an internal diameter of 600 and 500 μm. Consequently, the needle technique exerts much less shear stress on the cells than does the Teflon-glass homogenization. These experiments clearly demonstrate that even shearing forces much smaller than those exerted by the Teflon-glass homogenizer are able to pull off microvilli and related structures from the cell surface and to generate a distinct population of cell surface-derived vesicles exhibiting sedimentation properties identical to those of microsomal fractions.
This finding implies far-reaching consequences. Both fractions contain the same proteins, because the internal membranes are the immediate precursors of the microvillar membranes. As pointed out above, the protein composition of the original exocytic membrane domain is completely preserved in the microvillar tip membrane. Thus the so-called microsomal marker proteins are present in both fractions, and a distinction by “marker enzymes” is impossible, on principle. Consequently, the general agreement about the intracellular location and function of certain important functional proteins, especially about those attributed to the endoplasmic reticulum, has to be questioned seriously.
Until now, two of these important systems with widely accepted intracellular localization, the receptor-regulated Ca2+store/Ca2+ channel (106, 114) and the insulin-sensitive glucose transporter pool (104), have been identified as microvillar components. However, a number of further systems with putative intracellular localization remain to be reevaluated. For instance, one of the most important components of microsomal fractions, the drug-metabolizing cytochrome P-450 system, has been shown to occur and function on microvilli of different cell types (119, 123). A cell surface function of this enzyme system appears to be a convincing alternative to the currently accepted view that detoxication occurs deep within the cell where toxic compounds already have flooded all sensible cellular targets. A possible model for the epithelial handling of xenobiotics, including the peripheral function of the P-450 system, has been proposed (112).
The use of cell fractionation techniques for the isolation of intracellular membranes should have been questioned almost two decades ago when Carothers Carraway et al. (25) for the first time used a hydrodynamic shearing technique for the isolation of microvilli from tumor cells. Later, Pakkanen and Vaheri (151), Lange and Brandt (104, 106), and Lange et al. (114) applied the same technique for microvilli isolation from cultured cells and hepatocytes. Lange and Brandt (105) presented a decisive experiment clearly demonstrating the microvillar origin of a large portion of the microsomal fractions. The use of this gentle variant of normal homogenization by Teflon-glass homogenizers for isolation of an inside-out fraction of microvillus-derived vesicles should have caused all experimenters to seriously doubt the generally accepted, but never soundly proved, assumption that microsomal fractions consist of intracellular membranes only. The unconcerned handling of this problem may be due to the widespread ignorance about the ubiquitous occurrence of microvilli on the surfaces of almost all types of differentiated or resting cells (except neurons).
Differentiation Between Vesicles of Microvillar and Intracellular Origin
Much experimental work has been employed for the purification and characterization of glucose transporter-containing vesicles from the “light microsomal” fraction of adipocytes. Biber and Lienhard (13) introduced an immunoadsorption technique for the isolation of vesicles containing glucose transporters from homogenates of 3T3-L1 adipocytes. These authors used an antiserum raised against a COOH-terminal peptide of the transporter. This part of the transporter molecule is located at the cytoplasmic surface of membranes. If one expects that the transporter-containing vesicles are part of the endoplasmic reticulum or trans-Golgi system, it is right to assume that these vesicles carry the cytoplasmic transporter domains on their external surface. Consequently, vesicle fractions isolated by this technique comprise only vesicles of intracellular origin and are necessarily devoid of microvillus-derived vesicles known to maintain their original outside-out orientation (76). Moreover, microvillar vesicles differ from those isolated by the immunoadsorption technique by their size of 150–200 nm in diameter (104) compared with the microsomal vesicles, which are about 50 nm in diameter (13).
Role of Exocytosis in Insulin Action
The translocation theory is supported by a large body of data obtained by various experimental protocols. Nevertheless, there are disagreements as to the quantitative contribution of recruitment as discussed in a review by Stephens and Pilch (176). In many cases, the rate of insulin-enhanced glucose uptake is much greater than the degree of translocation as assessed by Western blot and cytochalasin B binding.
The assumption of two different transport-enhancing processes proceeding at different time scales offers an explanation for the observed differences between the insulin stimulation of hexose uptake and transporter translocation to the cell surface of 3T3-L1 adipocytes (16, 21, 57, 62) and rat adipocytes (5, 87). A certain part of the observed insulin-induced transport stimulation may indeed be due to translocation of glucose transporters from an intracellular compartment to the cell surface by exocytic events. Intracellular vesicles are inserted into the plasma membrane during insulin-induced exocytosis. However, in contrast to the fast initial effect of insulin on the microvillar pathway occurring with a half-time of about 2 min, exocytic events usually proceed at a much larger time scale with a half-time of 10–20 min. Most likely, the exocytic portion accounts for 35–40% of the total insulin action (210). Transport stimulation within 10 min of insulin action, however, is largely due to activation of the microvillar transport pathway via transporters already present on the cell surface before insulin action.
PI 3-Kinase-Dependent Activation of the Microvillar Influx Pathway
Although morphological studies on 3T3-L1 adipocytes (111) clearly have shown that the structural organization of microvilli is critically altered by the action of insulin, the biochemical mechanism by which insulin receptor activation modulates the cytoskeletal diffusion barrier remains unclear. In contrast to receptor-mediated Ca2+ signaling, which is initiated by F-actin depolymerization via the PLC pathway, insulin signaling critically depends on activation and translocation of cytoplasmic PI 3-kinase to the plasma membrane. Activation of PI 3-kinase is accompanied by extensive membrane dynamics on the cell surface, called membrane ruffling, which is an energy-dependent cytoskeleton-driven process essentially involved in cell motility. The cytoplasmic surface of the ruffling membrane is a site of extensive actin polymerization generating a three-dimensional F-actin network that provides the driving force for the observed membrane ballooning. However, membrane ruffling is only one aspect of the cytoskeletal response to PI 3-kinase activation. Ruffling can be observed readily by light microscopy. The other cytoskeletal aspect of PI 3-kinase activation, which is only perceptible on the electron microscopic level, is the loss of structural organization of microvilli, giving rise to conspicuous shape changes.
With the use of isolated rat hepatocytes, the dependence of the insulin-induced microvillar shape change on PI 3-kinase activation has been demonstrated clearly. Wortmannin completely inhibited insulin-induced shape changes of microvilli on these cells (109). In contrast, the PLC-mediated vasopressin action on microvilli morphology, which closely resembles that of insulin-induced shape changes, is wortmannin insensitive. The involvement of PI 3-kinase in microvillar shape changes is further strengthened by the earlier finding that phenylarsine oxide (PAO) inhibits insulin-stimulated glucose transport in fat and muscle cells. Later, PAO was identified as a potent inhibitor of PI 3-kinase (69).
Recent findings have contributed to a better understanding of the role of PI 3-kinase in cytoskeletal reorganization in differentiated cells with microvillar surfaces.
Complete barbed-end capping of cytoplasmic actin filaments is an essential precondition for the stable existence of microvillar cell surfaces.
Capping proteins play a specific role in the maintenance of the microvillar surface. In the basal (unstimulated) state, microvilli expose only the slowly growing pointed ends of their microfilaments to the cytoplasm. Microfilaments with shielded barbed (high affinity) but exposed pointed (low affinity) ends are stable only at or above the high critical monomer concentration (Cc = 4 μM). At monomer concentrations below this value, monomers dissociate from pointed ends. In contrast, the rapidly growing barbed ends of microfilaments exhibit a 40-fold lower Cc of 0.1 μM. Thus the exposed pointed ends of microvilli are stable only when all cytoplasmic microfilaments are blocked at their barbed ends by capping proteins. Under this condition, the steady-state monomer concentration equals the high pointed-end Cc. Otherwise, the length of microvilli would be reduced by pointed-end dissociation in favor of barbed-end elongation of the cytoplasmic filaments.
The metastable state of the actin cytoskeleton in cells with microvillar surfaces can be used to trigger rapid cytoskeletal changes in response to specific signals acting on the capping state of microfilaments. This system closely resembles a strained spring, able to release large forces via a trigger mechanism operated by a small trigger force. Receptor-mediated decapping of F-actin has been reported for several cellular models includingN-formyl-Met-Leu-Phe-activated neutrophils (3,9, 128), thrombin-stimulated platelets (71), and EGF activation of adenocarcinoma cells (77). Recent findings point to a novel mechanism of PI 3-kinase-dependent decapping of cytoplasmic F-actin, which is able to trigger rapid changes of the microvillar diffusion barrier.
Role of PI 3-kinase in cytoskeletal disorganization of microvilli.
The specific role of actin-binding and -severing proteins such as profilin and gelsolin/cofilin in Ca2+ signaling was briefly described earlier. These proteins are released from the membrane via receptor-mediated activation of PLC to cleave PIP/PIP2 even when these phospholipids are shielded by proteins of the gelsolin/cofilin family (59, 60). Release of actin-binding/capping proteins from their original binding sites at the plasma membrane causes microfilament depolymerization and reorganization as visualized by shortening of microvilli and ballooning of the adjacent plasma membrane. Although insulin signaling is independent of PLC, the morphology of insulin action is quite similar.
A recent study (124) on the downstream targets of PI 3-kinase closed the mechanistic gap between PI 3-kinase activation and insulin-induced reorganization of microvillar F-actin. These authors found that the PI 3-kinase products phosphatidylinositol 3,4-bisphosphate [PI(3,4)P2] and phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3], bind with up to 10-fold [K d = 1.1 μM for PI(3,4)P2] higher affinity to profilin (124) than does PI(4,5)P2. A similar affinity change may apply for gelsolin and other phospholipid-regulated capping proteins. This interpretation is supported by Hartwig et al. (71), who demonstrated that D-3 polyphosphoinositides uncap cytoplasmic F-actin in permeabilized resting platelets, and by Barkalow et al. (7), who showed that thrombin stimulation of platelets results in release of gelsolin and capping protein from barbed ends of cytoplasmic F-actin. Obviously, the high affinity of 3-phosphorylated membrane lipids causes rapid redistribution of PI-regulated actin-binding and capping proteins from cytoplasmic F-actin to the sites of PI 3-kinase action at the inner face of the plasma membrane, most likely at the foot points of microvilli where this enzyme develops its highest activity (seeActivation of PI 3-kinase by high membrane curvature). Thus PI 3-kinase-induced redistribution of capping proteins is the mechanistic basis for both localized membrane ruffling and microvilli shortening.
Decapping of barbed ends on cytoplasmic microfilaments results in extensive monomer addition to these sites, reducing the actin monomer concentration in the cytoplasm to levels far below the Ccof the pointed ends. As shown in Fig. 9, the lowered cytoplasmic monomer concentration causes disassembly of microvillar filaments at their exposed pointed ends accompanied by shortening of microvilli and reduction of the diffusion barrier function.
Decapping of F-actin initiates membrane ruffling.
The second important aspect of insulin-induced cytoskeletal reorganization is translocation of actin to the plasma membrane and induction of membrane ruffling. Because capping proteins carry actin monomers of cytoplasmic origin, formation of D-3 phospholipids also causes submembrane microfilament assembly. In addition, actin monomers liberated from the pointed ends of microvillar F-actin contribute to the formation of the expanding microfilament network below microvilli. Overall, the membrane ruffling response is induced by directed transfer and assembly of actin monomers to the immediate vicinity of activated insulin receptors (see Fig. 13). The physiological relevance of this vectorial activation of cytoskeletal growth may rest in the oriented initiation of cell motility to gradients of activating external signals (chemotaxis).
On the other hand, the formation of a localized submembrane microfilament network gives rise to mechanical membrane stress at the ruffling sites, further contributing to the retraction of microvilli and integration of their membrane portions into the plasma membrane as indicated by SEM of insulin-stimulated 3T3-L1 adipocytes (Fig.6 B) and hepatocytes (see Fig. 12 B). PI 3-kinase-induced membrane ruffling may be interpreted as an auxiliary mechanism necessary for a sufficient reduction of the microvillar diffusion barrier.
Insulin-induced activation of the actin-depolymerizing factor/cofilin system accelerates pointed-end dissociation.
Recent findings point to a further mechanistic component involved in insulin-induced cytoskeletal reorganization downstream of PI 3-kinase. The actin-binding and -depolymerizing factor (ADH)/cofilin belongs to a family of 15- to 20-kDa proteins that are regulated by phosphorylation on serine (145). In the basal state, these proteins were found as inactive phosphoprotein (P-cofilin) in the cytoplasm. After insulin or EGF receptor activation, P-cofilin is dephosphorylated and translocated to the plasma membrane. The mechanism of dephosphorylation is still unclear; however, this reaction appears to depend directly or indirectly on activation of PI 3-kinase, since inhibition of this enzyme by wortmannin completely inhibits dephosphorylation of cofilin (117).
As stated by Carlier et al. (24), the functionally relevant main effect of actin-depolymerization factor/cofilin on F-actin is a 25-fold increase in the rate of monomer dissociation from the pointed ends, whereas dissociation from the barbed ends remains unaffected. Accelerated pointed-end dissociation excellently fits into the picture of cytoskeletal reorganization by insulin-induced activation of PI 3-kinase. Otherwise, unraveling of the microvillar actin filaments from their slowly reacting pointed ends via simple monomer dissociation would be too slow to explain the rapid early effects of insulin on transport and microvillar shapes proceeding with a half-time of about 2 min.
On the other hand, inactivation of cofilin by phosphorylation was shown to depend on Rho-mediated activation of Rho/LIM kinases (129), a pathway common to that of ERM protein activation, which is essential for microvilli formation. ERM phosphorylation also depends on phosphoinositide lipids such as PIP2 and PIP and, thus, on the location of PI 4(5)-kinase activity on lipid membranes (117, 181). Obviously, the Rho-initiated kinase pathway provides for ERM activation and cofilin deactivation, two essential steps involved in microvilli regeneration following receptor-mediated activation.
Activation of PI 3-kinase by high membrane curvature.
A further remarkable property of the PI 3-kinase reaction is the strong dependence of its activity on membrane curvature (80). In 50-nm vesicles, the PI 3-kinase activity is 100 times greater than on 300-nm vesicles. On the surface of differentiated cells, the 50-nm curvature is almost exclusively found on microvilli. A specific type of high membrane curvature occurs within the transition region between the plasma membrane and the microvillar shaft, presumably a region of facilitated transbilayer flipping of membrane lipids during microvilli elongation (112) (Fig.10).
Phospholipids within this region are much more mobile than those located within the highly ordered, even plasma membrane, allowing rapid steric reorientation of these lipids for interaction with the PI 3-kinase. According to this illustration, sites of high membrane curvature should be locations of high PI 3-kinase activity, increased formation of 3-phosphorylated phosphatidylinositols, and, thus, sites of high ruffling activity.
In summary, the action of insulin on glucose transport is the direct consequence of PI 3-kinase activation and translocation to the plasma membrane. Downstream of receptor-mediated PI 3-kinase translocation/activation, four mechanistic components are involved in the activation of the microvillar influx pathway (Fig.11):
PI 3-kinase activity is strongly potentiated by high membrane curvature at the foot points of microvilli.
Capping proteins blocking barbed ends of cytoplasmic F-actin are transferred to D3-phosphorylated PIP and PIP2 moieties in the immediate vicinity of activated insulin receptors/PI 3-kinase. Because of actin monomer addition at the free barbed ends, the cytoplasmic actin monomer concentration is reduced to levels below pointed-end Cc, giving rise to microvilli shortening through monomer dissociation from the pointed ends of microvillar F-actin.
Pointed-end dissociation of actin monomers is specifically accelerated by PI 3-kinase-dependent activation of the ADH/cofilin system.
Both pointed-end dissociation of microvillar F-actin and monomer shuttling to the submembrane binding sites of capping proteins result in extensive formation of a three-dimensional actin network that imparts mechanical strain on the stimulated membrane regions (membrane ruffling) and supports retraction of microvilli.
This interpretation is strongly supported by SEM of insulin-induced changes of the surface morphology of isolated rat hepatocytes (109). Within a few minutes of insulin action, large ballooned plasma membrane areas have formed on which the original microvillar tips are still visible for a short time (Fig. 6 Band 12 B). The surface morphology of insulin action of hepatocytes is almost identical to that of 3T3-L1 adipocytes (Fig. 6 B). The dome- or saclike curvatures of the plasma membrane around the microvilli are due to rapid actin network formation following PI 3-kinase-induced cytoskeletal reorganization (membrane ruffling), assumed to facilitate the shortening process of microvilli by exerting a tangential strain on the plasma membrane. Later, complete integration of the microvillar membrane domains into the ballooning areas occurs.
Insulin receptors are localized on microvilli.
The events following insulin binding to its receptor have been subjected to extensive investigations. In various cell types, unoccupied insulin receptors have been detected on microvilli. Immediately after insulin binding, these receptors appear to leave the microvillar membrane domains, moving to the intervillous part of the cell surface.
Carpentier and McClain (28) summarized the state of knowledge in the following way:
In its unoccupied and unstimulated state, the insulin receptor associates with microvilli on the cell surface.
This preferential association is dependent on the integrity of the cytoplasmic tail of the receptor.
Insulin binding releases the constraint maintaining the receptor on microvilli and does so via receptor kinase activation and autophosphorylation of three tyrosine residues present in the regulatory domain.
The insulin receptor complex, freely mobile on the cell surface, next associates with the internalization gates, the clathrin-coated pits, via signal sequences in the juxtamembrane domain of the cytoplasmic tail of the receptor.
Part of this knowledge comes from studies in which deletion mutants of the insulin receptor expressed in Chinese hamster ovary (CHO) cells (29) were used to study the structural requirements for the ligand-specific redistribution of the receptor. Electron microscopic analysis revealed that there is a preferential initial association of 125J-insulin with microvilli at low temperatures. As a function of the incubation time at 37°C, this association shifted from microvillous to “nonvillous” domains in all those cell lines in which insulin receptors were normally autophosphorylated. In contrast, the receptors of four autophosphorylation-deficient cell lines were unable to leave microvilli after insulin binding, even during incubation at 37°C. The authors concluded that the unstimulated insulin receptor underlies constraints maintaining it on microvilli. After stimulation of receptor kinase and autophosphorylation, the receptor is released from this constraint. The electron micrographs they presented, however, clearly demonstrate that all mutants with intact receptor autophosphorylation, most obviously the wild-type cells, respond to incubation with125J-insulin at 37°C with shape changes of microvilli into large saclike surface protrusions carrying the labeled receptors at their bases (29). This illustration closely corresponds to the insulin-induced shape changes of microvilli on 3T3-L1 adipocytes that we documented using REM techniques (111) and clearly demonstrates that insulin activation of cells with intact receptor autophosphorylation leads to disappearance of microvilli in their original form. Unfortunately, transmission electron microscopy, in contrast with SEM, is an inadequate technique to quantitate the disappearance of microvilli from the cell surface.
The results of these studies are in accord with the presented concept of microvillar transport regulation, confirming some of its most important mechanistic aspects:
Glucose transporters, insulin receptors, and the diffusion barrier are colocalized within microvilli, which may be seen as a functional unit involved in transport regulation.
Insulin receptor activation is a necessary step to initiate microvillar shape changes.
Insulin-induced shape changes of microvilli are due to intracellular processes downstream of receptor tyrosine phosphorylation.
The postulated “constraint maintaining the receptor on microvilli” can be explained by the rigid structural organization of microvilli but not necessarily by direct or indirect linking to the cytoskeleton as proposed by Carpentier et al. (29).
INSULIN-INDEPENDENT FORMS OF GLUCOSE TRANSPORT MODULATION
Transport Activation by Hyperosmotic Conditions
In a recent survey (103), the implications of the diffusion barrier concept on the mechanism of cell volume regulation were discussed. The ability for rapid volume regulation following an anisotonic challenge is a systematic property of cells with microvillar surfaces emerging from the specific distribution of ion and water channels on the surface of differentiated cells. Here, the consequences of hypertonic treatment are shortly described and their relevance for transport modulation is pointed out.
Distribution of ion and water channels on the cell surface.
As described earlier, functional relevant integral membrane proteins of differentiated cells are primarily located on the tips of microvilli, whereas the surface of the cell body is largely depleted from these components. Depending on their specific anchorage to the actin cytoskeleton, various membrane proteins are retained within the microvillar structure, whereas other proteins that are not linked to microfilaments can leave the exocytic membrane domain. Thus a specific set of ion channel proteins, characteristic of the respective tissue type, is located beyond the diffusion barrier on microvillar tips.
According to the present state of knowledge, ion and water channels of epithelial cells are distributed in the following way (Fig.13):
Water channels are localized on the basal plasma membrane (37, 98, 141, 154) but are largely absent from the apical plasma membrane except after vasopressin stimulation (83,142). Basal microvilli are devoid of water channels (37,121, 136, 137).
Epithelial Na+ channel (23, 133, 171, 172,), anion channels with Cl−, ATP, and osmolyte permeability such as MDR (multidrug resistance) P-glycoprotein (43, 82, 182), CFTR (47), MRP/MOAT (multidrug resistance-related protein/multispecific organic anion transporter) (33), MTC (monocarboxylate transporter) (11), and, occasionally, water channels (37,54) are detected on apical microvilli of polarized epithelial cells.
Hyperosmotic treatment coactivates glucose transport and ion fluxes.
Because of the specific arrangement of ion channels on microvilli of differentiated cells, the increase of external osmolarity primarily results in a corresponding increase of the osmotic activity within the entrance compartment via NaCl or osmolyte uptake (Fig.14). Because microvillar membranes are largely devoid of water channels, the hypertonic entrance compartment draws water from the cytoplasm through the diffusion barrier. The finely dispersed surface spaces (see Fig. 2) of the entrance compartment respond much more rapidly to osmotic changes than the cytoplasm does by water exit through the plasma membrane. Consequently, the entrance compartment rapidly swells, accompanied by shortening of the microvillar shafts. The resulting impairment of the diffusion barrier function allows entry of not only Na+ and Cl− but also glucose from the entrance compartment into the cytoplasm. This activation mechanism for microvillar pathways also applies to cells lacking water channels.
Hypertonic enlargement of the microvillar tip compartment is accelerated by microfilament disassembly due to dilution within this compartment, resulting in considerable reduction of the G-actin and ATP concentrations. Dilution of both components critically lowers the stability of actin filaments at their fast-reacting barbed ends exposed to the entrance compartment. Terminal ATP-actin subunits confer stability to the barbed ends of the filament and allow rapid elongation by the addition of further ATP-actin monomers. In contrast, subunits containing ADP instead of ATP (or ADP-Pi) initiate rapid dissociation of monomers, which results in the sudden decomposition of large regions at the barbed ends of microfilaments (dynamic instability). Filament shortening proceeds unless addition of ATP-G-actin again stabilizes the filament ends (see review, Ref.96). Because actin filaments are always oriented with their barbed ends toward the top of microvilli, the filament ends at the tip compartment sensitively respond to swelling-induced changes of ATP and actin monomer concentrations with dynamic instability.
The proposed mechanism of hypertonic volume response is strongly supported by the observations of Winterhager and Stieve (203) and Korten et al. (97), who reported that microvilli do not shrink under hypertonic conditions but, instead, dilate. Hypertonic treatment increases microvillar diameters two- to threefold and decreases the total number of microvilli, most probably due to integration into the plasma membrane.
Moreover, using X-ray microanalysis, Schrärmeyer et al. (161) presented direct evidence that the microvillar compartment reflects the ionic condition of the external medium rather than that of the cytoplasm. Microvilli of photoreceptor cells contain higher Na+ and lower K+ concentrations than the cytosol. The authors concluded that the rhabdomeric plasma membrane is permeable to these ions, allowing rapid equilibration of the microvillar compartment with the extracellular medium.
Similar observations were reported by Lumpkin and Hudspeth (125,126) and by Postma et al. (153). Applying fura 2/confocal laser scanning microscopy, they demonstrated high Ca2+ concentrations in the tip compartments of hair cell microvilli. “Unstimulated hair cells showed a tip blush of enhanced fluorescence at the hair bundle[s] top, which attribute to Ca2+ permeation through transduction channels open at rest. Upon mechanical stimulation, individual stereocilia displayed increased fluorescence that originated near the tips and spread towards their bases” (125).
Further arguments in favor of the proposed mechanism come from studies demonstrating that hyperosmotic stimulation of glucose transport is always accompanied by coactivation of ionic fluxes (14, 31, 34,157, 175, 192, 194). Moreover, hypertonic pretreatment completely prevents any further insulin stimulation of glucose transport (32); similarly, glucose starvation-stimulated transport cannot be further activated by hypertonic treatment (34), indicating at least partial identity of the involved mechanisms.
Hypertonicity as well as insulin receptor activation affects the diffusion barrier function by disorganization of the microvillar cytoskeleton. They do this, however, in different ways. Whereas insulin, via activation of PI 3-kinase, induces shortening of the microvillar actin filaments from their cytoplasmic ends (109), hypertonicity starts disassembly of microfilaments at the tips of microvilli by dilution of the internal water space. An almost identical mechanism underlies starvation-induced activation of glucose transport.
Glucose Starvation-Induced Transport Stimulation
Transport stimulation by glucose starvation induces shortening of the diffusion barrier by disassembly of the microfilament bundles from the top of microvilli. Similarly to hypertonic treatment, glucose starvation critically lowers the ATP concentration within the tip compartment, thereby enhancing the dynamic instability at the barbed filament ends.
Starvation-induced and insulin receptor-mediated transport stimulation can be differentiated by SEM (111). In 3T3-L1 adipocytes, insulin initiates transient structural changes at the basis of microvilli, giving rise to domed membrane areas on which the upper parts of microvilli remained unaffected (Fig. 6 B). This illustration suggests a mechanism of microvilli shortening that starts from the cytoplasmic side. In contrast, glucose starvation ultimately results in the complete loss of the shaft region accompanied by enlargement of the microvillar tips, which then appears to protrude directly from the cell surface (Fig.15). This illustration clearly points to a shortening mechanism that starts from the top of the microvilli as schematically shown in Fig. 16. Because of the loss of cytoskeletal support, the membrane of the upper shaft region is integrated into the tip domain, increasing the microvillar tip diameter from 200 to 400 μm (5) (Fig. 15).
Interestingly, glucose starvation-induced (8- to 10-fold) transport activation is not accompanied by transporter redistribution between plasma membrane and microsomal fractions as assessed by fractionation techniques. Fisher and Frost (49) stated that “the increase in transport activity associated with glucose deprivation does not result from translocation of either of the glucose transporters known to exist in 3T3-L1 adipocytes.”
The changes in the surface morphology of starved 3T3-L1 adipocytes (Fig. 15) yield a compelling explanation for this behavior. In contrast to insulin-treated cells, on which a large portion of microvilli are completely integrated into the ruffling plasma membrane, the surface of glucose-starved cells exhibit intact spherical blebs representing the enlarged tip domains of the original microvilli. Most likely, these surface blebs are not integrated into the plasma membrane because they still contain the stabilizing surface coat and internal cytoskeletal support. The cytoskeletal connection of GLUT-4 is stabilized by low cytoplasmic FDP concentrations as prevailing under starvation conditions. Using Potter-Elvejem techniques of cell homogenization transforms these surface vesicles into a homogeneous vesicle population exhibiting microsomal rather than plasma membrane properties. The same fractionation behavior is observed after hyperosmotic stimulation. The difference in the subcellular fractionation behavior between glucose starvation and insulin action may be due to the removal of surface coat components by insulin receptor-mediated activation of surface proteases, whereas glucose starvation leaves the surface coat intact.
The effect of glucose starvation on the cell surface is closely related to that of hyperosmotic activation. Both conditions cause shortening of the diffusion barrier by disassembly of the microfilament bundles from the top of microvilli. During hyperosmotic challenge, dilation of this compartment is caused by osmotic salt (NaCl) and water entry. Similarly, glucose starvation-induced microvilli shortening leads to cell volume increase (111), most likely due to activation of ionic pathways. Thus glucose starvation as well as hyperosmolarity enhances influx or efflux activities for both glucose and ions via a common mechanism.
Costimulation of transporter and channel activities is a clear consequence of the proposed general role of microvilli in the regulation of transmembrane substrate and ion fluxes. The involvement of microvilli in Ca2+ signaling also explains the observed coactivation of hexose transport and Ca2+ signaling as demonstrated by the action of bombesin on Swiss 3T3 fibroblasts (187, 188) and of insulin on β-cells (205), CHO cells (140), hepatocytes (10), and pancreatic β-cells (4).
On the other hand, the concept of microvillar channel and transporter regulation opens a direct access to the phenomenon of cell swelling coupled to receptor activation as shown for the insulin action. Al-Habori et al. (2) and Baquet et al. (6) demonstrated that insulin-induced stimulation of ionic influx pathways in hepatocytes can be completely mimicked by hyposmotic cell volume increase, which is another way to inactivate the microvillar diffusion barriers and to increase ion and substrate fluxes along this pathway [reviewed by Lange (103)]. Some authors even considered cell volume changes to be a second messenger for transmission of hormonal and environmental signals (100).
Transport Stimulation by Metabolic Depletion
The mechanism of transport modulation by metabolic depletion is another example of transport regulation by modulation of the cytoskeletal diffusion barrier. This type of uptake regulation closely resembles that of hypertonic treatment and glucose starvation. Again, shortening of the cytoskeletal diffusion barrier is caused by ATP depletion; however, because of inhibition of the mitochondrial ATP production, cytoplasmic ATP is involved. Consequently, disassembly of the diffusion barrier is initiated from the cytoplasmic ends of the filament bundle. The morphological consequences of cytoplasmic ATP depletion have been described for various cell types:
Anoxic hepatocytes display globular microvilli and bleb formation (88).
Lowered ATP results in thinning of microvillar actin in renal cultured cells (61).
In hypoxic placental trophoblasts, reduction in number and shortening of microvilli is accompanied by elevated glucose consumption and lactate production (46).
Microvillar actin bundles are disrupted in anoxic proximal tubule cells because of cytoplasmic ATP depletion.
Within individual cells, all microvilli collapse simultaneously and microvillar actin filaments including villin undergo a parallel translocation to a perinuclear location (200).
A large body of experimental data on transport stimulation by metabolic depletion has accumulated for the liver cell line clone 9, which responds to ATP depletion by a significantly (up to 12-fold) increased glucose uptake rate. Again, no change in the transporter distribution between subcellular fractions could be observed (68,166, 168, 169) (Fig. 17).
Together, these data suggest that cytoplasmic ATP depletion results in changes of the microvillar diffusion barrier similar to those caused by hyposmotic treatment or glucose starvation. All three conditions cause disassembly of the microvillar F-actin and, consequently, shortening of microvilli. Because the stabilizing surface coat is not affected, the original microvillar tip domains remain intact, appearing as small bleblike surface protrusions (Fig. 17). After cell homogenization and centrifugation, these surface blebs retain right-side-out orientation and are recovered within the microsomal fraction.
Direct Evidence for the Involvement of the Actin Cytoskeleton in Glucose Transport Modulation
Several recent publications have stressed the role of the actin cytoskeleton in insulin-induced transport stimulation, although the exact function of F-actin in insulin signaling remains unclear. Wang et al. (199) described the inhibitory action of the F-actin-disrupting agents cytochalasin D and latrunculin B on the insulin-induced reorganization of the actin cytoskeleton in 3T3-L1 adipocytes, which is accompanied by a 50% inhibition of insulin-stimulated glucose transport. PI 3-kinase activity is not changed by these agents. Cytochalasin D also decreases by about 50% the insulin-dependent redistribution of GLUT-1 and GLUT-4 transporters to the plasma membrane fraction. Similarly, Omata et al. (147) observed a reduction of insulin-stimulated glucose transport after treatment of isolated rat adipocytes with latrunculin A and the complete inhibition of GLUT-4 translocation to the plasma membrane. The same results were obtained after disorganization of the actin cytoskeleton via Rab or Rho/Rac inactivation (197,160).
Since the work of Sampath and Pollard (158), it has been known that in vitro, cytochalasin D only slows down the rate of actin polymerization but does not completely stop this process. Consequently, there is no reason to assume that application of cytochalasin D to cells may cause microfilament disruption. As these authors pointed out, the main reduction of the polymerization rate occurs at the barbed ends of the filaments, which in the case of microvillar F-actin are completely membrane shielded. Thus an action on the basal transport rate is not to be expected. At the least, the reduction of insulin stimulation of transport by cytochalasin D may indicate that PI 3-kinase-induced F-actin reorganization into submembranous networks (membrane ruffling; see Fig. 11) is a necessary mechanistic component for full activation of glucose transport by insulin.
A novel mechanism of glucose uptake regulation is proposed that emerges as a special case of the general form of influx and efflux regulation for substrates and ions via microvillar pathways. The establishment of the specific cellular surface organization necessary for this type of regulation, the microvillar surface, is characteristic of resting and differentiated cells. The consequences of this type of surface organization meet the requirements of differentiated tissue cells for regulation of essential cellular functions by external signals such as hormones, growth factors, and transmitters and for response to various environmental conditions including high or low levels of glucose supply, metabolic stress, and unusual thermal or osmotic conditions.
One of the most important aspects of this surface reorganization is the differentiation-dependent switch of energy metabolism from metabolic regulation to membrane limitation, which allows the effective integration of oxidative phosphorylation into the ancient framework of glycolytic energy production. Furthermore, the formation of microvillar cell surfaces creates essential preconditions for the function of the Ca2+ signal pathway, the main signaling system of differentiated tissue cells. Actin filaments of microvilli represent two of the main components of this pathway. First, the highly effective diffusion barrier enables the cell to maintain a Ca2+gradient of four orders of magnitude by preventing the entry of divalent cations into the cytoplasm. Second, the ability of F-actin for high-affinity Ca2+ storage allows cytoplasmic Ca2+ to return quickly from high signaling levels to low basal levels. The same cytoskeletal structure that regulates receptor-operated Ca2+ fluxes also is involved in modulation of glucose uptake and in activation of ionic fluxes in response to changes of environmental osmotic conditions (103).
Numerous further adaptations of microvillar functions exist in living systems. The most prominent examples are the various sensory cells mediating sound, light, and taste perception via changes of the membrane potential. Another wide field is electrical signal transduction in the nervous system. Nerve cells do not express microvilli on the cell body but, instead, on dendrites on which the dendritic spines, the postsynaptic receptors of nerve cell, represent the functional analogs of microvilli. Even dendritic spines are able to change their length and shape in response to synaptic activity. A further adaptation of high physiological relevance is the polarized microvillar surface of epithelial cells, which represents a sophisticated defense system against external cytotoxic compounds. It appears that the microvillar structure is an ancient, highly conserved biological device that has been continuously adjusted to the ever-changing demands of life.
The proposed microvillar mechanism of glucose uptake regulation is compatible with most of the established data on this field. It explains by only one common final mechanism such different types of transport modulation as the insulin action, the action of high and low external glucose of hyperosmotic environments, and metabolic depletion. Many other puzzling features such as costimulation of glucose and ion fluxes turn out to be intrinsic properties of the proposed mechanism. The concept completes the translocation mechanism of transport regulation, solves most of its inconsistencies and shortcomings, and may contribute to a better comprehension of diabetes, obesity, and hypertension.
Address for reprint requests and other correspondence: K. Lange, Kladower Damm 25b, D-14089 Berlin, Germany (E-mail:).
- Copyright © 2002 the American Physiological Society