Glucose Activates Mitogen-activated Protein Kinase (Extracellular Signal-regulated Kinase) through Proline-rich Tyrosine Kinase-2 and the Glut1 Glucose Transporter*

Glucose serves as both a nutrient and regulator of physiological and pathological processes. Presently, we found that glucose and certain sugars rapidly activated extracellular signal-regulated kinase (ERK) by a mechanism that was: (a) independent of glucose uptake/metabolism and protein kinase C but nevertheless cytochalasin B-inhibitable; (b) dependent upon proline-rich tyrosine kinase-2 (PYK2), GRB2, SOS, RAS, RAF, and MEK1; and (c) amplified by overexpression of the Glut1, but not Glut2, Glut3, or Glut4, glucose transporter. This amplifying effect was independent of glucose uptake but dependent on residues 463–468, IASGFR, in the Glut1 C terminus. Accordingly, glucose effects on ERK were amplified by expression of Glut4/Glut1 or Glut2/Glut1 chimeras containing IASGFR but not by Glut1/Glut4 or Glut1/Glut2 chimeras lacking these residues. Also, deletion of Glut1 residues 469–492 was without effect, but mutations involving serine 465 or arginine 468 yielded dominant-negative forms that inhibited glucose-dependent ERK activation. Glucose stimulated the phosphorylation of tyrosine residues 402 and 881 in PYK2 and binding of PYK2 to Myc-Glut1. Our findings suggest that: (a) glucose activates the GRB2/SOS/RAS/RAF/MEK1/ERK pathway by a mechanism that requires PYK2 and residues 463–468, IASGFR, in the Glut1 C terminus and (b) Glut1 serves as a sensor, transducer, and amplifier for glucose signaling to PYK2 and ERK.

Glucose, the ubiquitous carbohydrate nutrient, activates intracellular signaling systems and thereby alters physiological and pathological processes. Effects of glucose on signaling are generally thought to be effected through the uptake and subsequent intracellular metabolism of glucose. For example, in rat adipocytes, glucose is rapidly converted, via glycolysis and generation of glycerol-3-PO 4, to phosphatidic acid and diacylglycerol (DAG) 1 (1)(2)(3), which, in turn, activates DAG-sensitive protein kinase C (PKC) isoforms (4,5). Also, in pancreatic islet ␤-cells, glucose, via glycolysis and ATP generation, closes K ϩ channels, causing plasma membrane depolarization, Ca 2ϩ -gating via L-type channels (6 -8), hydrolysis of phosphatidylinositol (PI) 4,5-(P0 4 ) 2 (9), mobilization of Ca 2ϩ by inositol (PO 4 ) 3 , generation of DAG, and activation of DAG-sensitive PKCs. Increases in PKC activity by either mechanism can activate RAS or RAF and subsequent signaling to MEK1 and extracellular signal-regulated kinases, ERK1 and 2 (10,11). Activation of either PKC or ERK by glucose can alter physiological processes, such as gene expression and cellular proliferation, differentiation, and survival. Moreover, in diabetes mellitus, inordinate increases in glucose-dependent PKC and ERK signaling play important roles in the pathogenesis of both diabetic complications (12)(13)(14) and acquired insulin resistance (15)(16)(17)(18). Such aberrant, diabetes-related, glucose-stimulated signaling through PKC to ERK has generally been attributed to increased de novo synthesis of phosphatidic acid and DAG. However, the latter hypothesis, although plausible, is neither proven nor exclusive, and other signaling mechanisms may also be operative. Presently, we found, in rat adipocytes, 3T3/L1 fibroblasts, L6 myotubes, and several other cell types, that glucose activates ERK independently of both glucose metabolism and PKC activation but dependent upon the activation of nonreceptor tyrosine kinase, proline-rich tyrosine kinase (PYK2). We also found that the Glut1 glucose transporter, apparently through specific amino acid residues in its C terminus, binds PYK2 and amplifies ERK activation by glucose.
Cell Culture Preparations and Incubation Conditions-3T3/L1 fibroblasts, 3T3/L1 adipocytes, and L6 myotubes were cultured as described (23,24). Rat A-10 vascular smooth muscle cells and mouse kidney mesangial cells were obtained from American Type Culture Collection. After growth and differentiation as needed, all cells were incubated in KRP medium as in the rat adipocyte experiments described above.
Assays of Immunoprecipitable ERK-As described (20,25), after incubation, adipocytes or other cells were sonicated in buffer containing 40 mM ␤-glycerophosphate (pH 7.3), 0.5 mM dithiothreitol, 0.75 mM EGTA, 0.15 mM Na 3 VO 4 , 5 g/ml leupeptin, 5 g/ml aprotinin, 0.1 mM phenylmethylsulfonyl fluoride, and 5 g/ml trypsin inhibitor. After and then treated for 10 min without or with agonists, 20 mM glucose, 1 M TPA, or 10 nM insulin. Note that although not shown, these inhibitors did not alter basal ERK activity, and glucose effects on ERK were virtually identical in cells preincubated for either 15 or 180 min. In B, adipocytes were incubated overnight in DMEM containing 1 M TPA to deplete DAG-sensitive PKCs (see Refs. 20 and 21), following which, the cells were washed and incubated in glucose-free KRP medium for 10 min without (CON) or with 10 nM insulin (INS), 1 M TPA, or 20 mM glucose (GLU). After incubation, immunoprecipitable ERK activity was measured. Values in A and B are the means Ϯ S.E. of 4 -16 determinations. In C and D, adipocytes were incubated first for 15 min in glucose-free KRP medium without or with indicated concentrations of cytochalasin B. In C, the cells were then treated for 10 min without (CONTROL) or with 20 mM glucose or 10 nM insulin before measuring immunoprecipitable ERK activity. In D, 20 mM D-glucose and 10 Ci of [6-3 H]D-glucose were added, and uptake of label over 30 min was measured. Values in C and D are the means Ϯ S.E. of four determinations. centrifugation at 700 ϫ g for 10 min, the fat cake, cell debris, and nuclei were removed. Post-nuclear supernatants were supplemented with 0.154 M NaCl, 1% Triton X-100, and 0.5% Nonidet, and equal amounts of lysate protein (200 -500 g) were subjected to overnight immunoprecipitation at 4°C with mouse monoclonal anti-ERK2 antibodies (Santa Cruz Biotechnologies) or with anti-epitope (HA, Myc) antibodies in co-transfection studies. Precipitates were collected on protein AG-agarose beads, washed, and incubated for 10 min at 30°C in 50 l of buffer containing 25 mM ␤-glycerophosphate (pH 7.3), 0.5 mM dithiothreitol, 1.25 mM EGTA, 0.5 mM Na 3 VO 4 , 10 mM MgCl 2 , 1 mg/ml bovine serum albumin, 1 M okadaic acid, 0.1 mM [␥-32 P]ATP (PerkinElmer Life Sciences; 1,500,000 dpm/nmol), and 50 g of myelin basic protein (Sigma). After incubation, aliquots were spotted on p81 filter paper, washed, and counted for 32 P radioactivity, as described (20,25). Blank values were obtained by substituting a nonimmune antibody preparation for anti-ERK2 antibodies or by omitting myelin basic protein sub-strate (results were similar). As noted previously (25), the Santa Cruz anti-ERK2 mouse monoclonal antibodies, although recognizing only ERK2 in Western analyses, precipitated ERK1, as well as ERK2, and the ratio of ERK2 to ERK1 in immunoprecipitates was approximately 1.4 -1.5/1, as measured by Western analysis, using a Santa Cruz rabbit polyclonal antiserum that recognizes both ERK1 and ERK2. Treatment with glucose and other carbohydrates did not alter the levels of either ERK2 or ERK1, or the ratio of ERK2 to ERK1, in ERK2, immunoprecipitates (also see Ref. 25).
Transfection Studies-Rat adipocytes and 3T3/L1 fibroblasts were transiently co-transfected as described (21,23,25,26). In brief, for transient co-transfection experiments with rat adipocytes, 0.4 ml of cells were suspended and electroporated in an equal volume of Dulbecco's modified Eagle's medium (DMEM) containing 5% bovine serum albumin and 3.3 g of pCEP4 encoding HA-tagged ERK2 or 3.3 g of pCMV5 encoding Myc-tagged ERK2 (both kindly supplied by (j) pCMV5 encoding either HA-tagged Glut1/Glut4 chimera containing a human Glut1 N terminus (equivalent to residues 1-462 in mouse Glut1) and the 29 C-terminal residues of rat Glut4 (residues 481-509, QISATFRRTPSLLEQEVKP-STELEYLGPD), or HA-tagged Glut4/Glut1 chimera containing residues 1-478 of the rat Glut4 N terminus and the 30 C-terminal residues of human Glut1 (identical to mouse residues 463-492, IASGFRQG-GASQSDKTPEELFHPLGADSQV) (both HA-tagged chimeras were kindly supplied by Dr. Michael Czech; see Ref. 29) (note that these HA tags are in exofacial loops and, when present in the plasma membrane, are accessible with extracellular antibodies); (k) pCDNA3 encoding human Glut1; (m) pCB6 encoding mouse Myc-Glut1; (l) pCB6 encoding mouse Myc-⌬489 -492-Glut1 (missing the last C-terminal 4 amino acids); (m) pCB6 encoding mouse Myc-⌬469 -492-Glut1 (missing the last C-terminal 24 amino acids); and (n) pCB6 encoding mouse Myc-Glut1 mutants, Myc-S465A-Glut1, Myc-A464S/S465A/G466A-Glut1 (note that Glut1 and Glut4 contain ASG and SAA, respectively, in these corresponding positions), or Myc-R468G-Glut1. 3T3/L1 fibroblasts were transfected with pCB6 encoding mouse Myc-tagged Glut1 using Lipo-fectAMINE (Life Technologies, Inc.) as described (23). After overnight (rat adipocytes) or 48 h (3T3/L1 cells) of incubation to allow time for expression, cells were washed and incubated, in most cases for 10 min, in glucose-free KRP medium, without or with addition of glucose, insulin or TPA, as described in the text. After incubation, HA-ERK2 or Myc-ERK2 was immunoprecipitated with anti-HA mouse monoclonal antibody (Covance) or rabbit polyclonal anti-Myc antiserum or mouse monoclonal anti-Myc antibodies (Upstate Biotechnologies Inc.) and assayed for myelin basic protein phosphorylation, as described above. In some experiments, plasma membrane levels of HA-tagged or Myctagged glucose transporters were quantified by measuring the cell surface content of HA or Myc epitopes, using anti-HA or anti-Myc mouse monoclonal primary antibodies (Covance or Upstate Biotechnologies Inc.) and 125 I-labeled rabbit anti-mouse IGG second antibody, as described (21,26). In some experiments, Myc-Glut1 was immunoprecipitated and examined for associated immunoreactive PYK2.
Adenoviral Gene Transfer Studies-In initial studies, we found that rat adipocytes were damaged (i.e. leaky with respect to glucose transport) by adenoviral infection and were considered to be unsuitable for use in the present study. We therefore used 3T3/L1 fibroblasts, because they have only Glut1 glucose transporters, are not damaged significantly by adenoviral infection, and are highly responsive to glucose and other saccharides in terms of increasing ERK activity independently of PKC and subsequent metabolism (see below). Fully confluent 3T3/L1 fibroblasts were infected with adenovirus alone or adenovirus encoding Glut1, Glut2, or chimeras in which their C termini were transposed, viz. N(1-457)Glut1/C(490 -525)Glut2 and N(1-489)Glut2/C(458 -492)Glut1; as described previously (30). Each of these viral constructs encode glucose transporters that are functional with respect to hexose transport (see Ref. 30; also documented presently, see below). After 48 h to allow time for expression, cells were placed into serum-free DMEM for 3 h, then equilibrated in glucose-free KRP medium for 20 min, and subsequently treated for10 min with or without 20 mM glucose. After incubation, cell lysates were examined for total immunoprecipitable ERK activity. Cellular uptake of 3 H-labeled D-glucose over 10 min was also monitored to compare the effects of expressed glucose transporters on cytochalasin B-inhibitable glucose uptake to effects on ERK.

FIG. 5. Effects of wild-type and chimeric forms of Glut1 and
Glut4 glucose transporters on activation of epitope-tagged ERK2 in rat adipocytes. Adipocytes (0.8 ml of 50% adipocyte suspension) were co-transfected with: (a) 6.6 g of pCDNA3 encoding Glut1, or pCIS2 encoding wild-type HA-Glut4 (N-Glut4/C-Glut4), pCMV5 encoding HA-Glut1/Glut4 chimera containing a Glut4 C terminus (N-Glut1/ C-Glut4), or HA-Glut4/Glut1 chimera containing a Glut1 C terminus (N-Glut4/C-Glut4) and (b) 3.3 g of pCMV5 encoding Myc-ERK2. Total DNA was kept constant by varying the amount of vector (VEC). After overnight incubation in DMEM to allow time for expression, cells were washed and equilibrated for 20 min in glucose-free KRP medium and then treated for 10 min with or without 20 mM glucose or 10 nM insulin in A and C or for 30 min with or without 10 nM insulin in B, as indicated. After incubation, in A and C, Myc-ERK2 was immunoprecipitated and assayed. In B, cell surface levels of HA-Gluts were measured as described under "Experimental Procedures." B shows the comparison of plasma membrane levels of wild-type N-Glut4/C-Glut4, chimeric N-Glut4/C-Glut1, and chimeric N-Glut1/C-Glut4 glucose transporters in transiently transfected rat adipocytes treated with or without insulin. C shows the effects of overexpression of the Glut4 glucose transporter and subsequent insulin (INS) treatment on basal and glucose (GLU)-stimulated Myc-ERK2 activation. All values are the means Ϯ S.E. of four determinations.

Studies in Rat Adipocytes
Effects of Glucose and Other Monosaccharides on ERK-Glucose, over a range of 2.5-30 mM, provoked rapid progressive increases in immunoprecipitable ERK activity (Fig. 1, A and B). Effects of 10 and 20 mM glucose on ERK activity were comparable and nonadditive to those of submaximal and maximally effective insulin (Fig. 1, A and B). The latter was surprising, because glucose uptake in rat adipocytes, which increases steadily over the concentration range of 2.5-20 mM in the absence of insulin (19), is increased further by insulin uptake throughout this glucose concentration range (19).
The failure of insulin to augment glucose effects suggested that the uptake and intracellular metabolism of glucose may not be the major determinant for ERK activation. Interestingly, transported, but nonmetabolizable sugars, 3-O-methylglucose (3OMG), which is not phosphorylated, and 2-DOG, which is converted only to glucose-6-PO 4 , were as effective as glucose in activating ERK (Fig. 1D). Also, both mannose, which is transported and metabolized like glucose, and glucosamine, which is transported similarly to but metabolized differently from glucose (i.e. is primarily used to synthesize glycolated products (31)), activated immunoprecipitable ERK as potently as glucose (Fig. 1D). In contrast, L-glucose and mannitol had little or no effect on ERK activity (Fig. 1, B and D). Thus, glucose-stimulated ERK activation was not due to changes in medium osmolality and required specific structural features. The ability of each of these carbohydrates to activate ERK was paralleled by their ability to inhibit 2-DOG uptake (not shown), through isotopic dilution or competitive interaction at the level of the glucose transporter (see below).
Requirements for Downstream Signaling Factors in Glucoseinduced Activation of ERK-Glucose effects on ERK, unlike those of phorbol esters, were not influenced by PKC inhibitors, GF109203X and GO6976 ( Fig. 2A), or by down-regulation of DAG-sensitive PKCs by prolonged phorbol ester treatment (Fig. 2B). Also, glucose effects on ERK, unlike those of insulin, were not inhibited by the PI 3-kinase inhibitor, wortmannin ( Fig. 2A), or expression of kinase-inactive PKC- (Fig. 3D), both of which inhibit insulin-induced activation of ERK2 in rat adipocytes (Figs. 2 and 3 and Ref. 25). On the other hand, glucose effects on ERK, like those of insulin (25), were inhibited by: (a) the tyrosine kinase inhibitor, genistein ( Fig. 2A); (b) a GRB2-SH2 domain inhibitor ( Fig. 2A); (c) dominant-negative forms of GRB2, SOS, RAS and c-RAF-1 (Fig. 3, A-C); and (d) the MEK1 inhibitor, PD98059 ( Fig. 2A). As with glucose, effects of 3OMG and 2DOG on ERK were independent of PKC in rat adipocytes (not shown).
Requirement for the Glut1 Glucose Transporter in Glucoseinduced Activation of ERK-The activation of ERK by transported sugars, glucose, mannose, glucosamine, 3OMG, and 2-DOG, suggested that one or more glucose transporters is required for ERK activation. Accordingly, cytochalasin B, which binds to Glut1 and Glut4 glucose transporters and noncompetitively blocks the binding and subsequent uptake of glucose and other saccharides (32)(33)(34), inhibited both ERK activation by glucose (Fig. 2C) and other sugars (see below), and cellular uptake of 3 H-labeled D-glucose (Fig. 2D). Halfmaximal and maximal inhibitory effects of cytochalasin B on both processes were observed at approximately 1 and 30 -100 M, respectively. Note that (a) 20 mM glucose was used in experiments shown in Fig. 2 (C and D) and (b) the same concentrations of cytochalasin B were required to inhibit the uptake of 20 mM glucose (Fig. 2D) and 50 M 3 H-labeled 2-DOG (not shown). Thus, assuming that glucose uptake is dependent on binding to glucose transporters, the K d values for binding of cytochalasin B and glucose to glucose transporters (see below and Fig. 7) are markedly different. Also note that cytochalasin B did not inhibit insulin-induced ERK activation (Fig. 2C).
Overexpression of increasing amounts of the Glut1 glucose transporter (Fig. 4D) led to progressive increases in (a) plasma membrane levels of such expressed (epitope-tagged) Glut1 (Fig.  4C), and (b) glucose-induced (Fig. 4A), but not phorbol esterinduced (not shown), activation of Myc-ERK2. In contrast, overexpression of the Glut4 glucose transporter inhibited glucose-dependent Myc-ERK2 activation (Figs. 4B, 5C, and 6A), even when insulin was added to increase plasma membrane Glut4 content and subsequent glucose transport (Fig. 5C). Thus, because Glut4 expression and insulin treatment did not amplify glucose effects on ERK and because plasma membrane levels of epitope-tagged Glut1 and Glut4 were similar (not shown), simple changes in glucose uptake did not account for observed effects of glucose transporters on ERK. (Note that Glut4 translocation in transfected cells is increased only 2-3fold by insulin, as compared with 5-10-fold increases in 2-DOG uptake observed in freshly incubated rat adipocytes. As discussed (21,26), this reflects an artifactual increase in basal translocation/transport caused by electroporation and overnight culture of rat adipocytes.) Molecular Requirements for Amplifying Effects of Glut1 on Glucose-dependent ERK Activation-Expression of a Glut1/ Glut4 glucose transporter chimera that contained a Glut4 C terminus, like wild-type Glut4, inhibited glucose effects on Myc-ERK2 (Fig. 5A). In contrast, a Glut4/Glut1 glucose transporter chimera that contained a Glut1 C terminus, like wildtype Glut1, amplified glucose effects on Myc-ERK2 (Fig. 5A). This difference in ERK amplification was apparent, despite the fact that plasma membrane levels of these chimeras were only slightly different, and both chimeras, like Glut4, were present in agonist-responsive pools (Fig. 5B). Of further note, expression of Glut3, unlike Glut1, but like Glut4, did not amplify glucose effects on ERK (Fig. 6A).

FIG. 8. Effects of glucose, other carbohydrates, and insulin either on ERK activation in 3T3/L1 fibroblasts (A, B, and E), 3T3/L1 adipocytes (B), L6 myotubes (C), and rat A-10 vascular smooth muscle cells (D) or on [ 3 H]2-DOG uptake in 3T3/L1 fibroblasts (F). In
Findings with chimeras suggested that the last 30 C-terminal amino acids of Glut1 (residues 463-492) were required for amplifying glucose effects on ERK. However, deletion of the last 4 or 24 amino acids of the Glut1 C terminus did not compromise the amplifying effects of Glut1 on glucose-dependent ERK (Fig. 6B). Thus, residues 463-468, viz. IASGFR, were required for amplifying effects of Glut1 on glucose-dependent ERK activation (note that this sequence is identical in human, mouse, and rat Glut1 glucose transporters). Because the corresponding human, mouse, and rat Glut4 sequences are ISAAFH, ISAAFR, and ISATFR, respectively, we speculated that AS may be required for amplifying effects of Glut1, and, contrariwise, SAA may be required for inhibitory effects of GLUT4, on glucose-dependent ERK activation. We also questioned whether the basic arginine 468 residue was required for amplifying effects of Glut1 on glucose-dependent ERK. Indeed, mutation of Glut1 residues 464 -468, IASGFR, to produce IAAGFR, ISAAFR, and IASGFG sequences not only led to a loss of amplifying effects of Glut1 but yielded dominant-negative forms of Glut1 that, like Glut4, inhibited glucose effects on ERK (Fig. 6C). Because the dominant-negative effects of these Glut1 mutants suggested that they inhibited endogenous Glut1 that resided in the plasma membrane, it was of interest to find that plasma membrane levels (as per exofacial Myc epitopes) of Myc-S465A-Glut1, Myc-A464S/S465A/G466A-Glut1, and Myc-R468G-Glut1 mutants were sizeable as comparable to that of Myc-WT-Glut1 (Fig. 6D), which potentiated glucose effects on HA-ERK (Fig. 6C). We also documented that the wild-type form and each of these mutant forms of Glut1 were expressed to about the same extent in transfected adipocytes, as measured by blotting for Myc immunoreactivity migrating at the level of the Glut1 glucose transporter on SDS-PAGE (not shown). In view of the plasma membrane levels of these GLUT1 mutants and because substitution of corresponding Glut4 residues in the Glut1 IASGFR sequence would not be expected to impair transport activities and, moreover, because arginine-468 is not required for transport activity of Glut1 (35), it seems clear that effects of the presently used Glut1 mutants on glucose-dependent ERK activation cannot be attributed to alterations in glucose uptake.
Effects of Disaccharides on ERK Activity and 2-DOG Uptake-The above findings were compatible with two possibilities, viz. saccharides that activate ERK must either enter the cell or interact with a cell surface protein, such as Glut1. To evaluate these possibilities, we used maltose and lactose, which contain galactose or glucose connected to glucose through a 1:4 glycosidic linkage and which, like bis-mannose, should interact with Glut1 or other proteins but not be transported. Indeed, both disaccharides, like the more potent D-glucose, activated ERK by a cytochalasin B-dependent mechanism (Fig. 7, B and C) and concomitantly inhibited 2-DOG uptake (Fig. 7, A and D). Moreover, the K i for inhibition of 2-DOG uptake and the K m for activation of ERK by both nontransportable disaccharides were only slightly different, viz. 20 mM versus 26 mM, respectively (Fig. 7, C and D). This small difference in K i and K m values was similar to that observed with glucose, viz. 4 and 10 mM (Fig. 7,  C and D), and may reflect the fact that 2-DOG uptake and ERK activation are decidedly different processes, and ERK activation may require a higher level of ligand binding than glucose transport through Glut1.

Activation of ERK by Glucose and Other Sugars in 3T3/ L1 Fibroblasts, 3T3/L1 Adipocytes, L6 Myotubes, and Rat A-10 Vascular Smooth Muscle Cells and Other Cell Types
In addition to rat adipocytes, ERK was activated by glucose, 3OMG, and 2-DOG in 3T3/L1 fibroblasts and rat A-10 vascular smooth muscle cells, which contain only Glut1 glucose transporters, and 3T3/L1 adipocytes and L6 myotubes, which contain both Glut1 and Glut4 glucose transporters (Fig. 8, A-D). In 3T3/L1 fibroblasts and L6 myotubes, glucose effects on ERK were largely independent of PKC, as evidenced by the failure of GF109203X to inhibit these effects. In rat A-10 vascular smooth muscle cells, glucose effects on ERK were partly independent and partly dependent on PKC. In mouse kidney mesangial cells (which contain Glut1 but not Glut4 glucose transporters), glucose effects on ERK were largely PKC-dependent, and 3OMG had no effect on ERK (data not shown). Thus, there is considerable hetereogeneity of mechanisms used by glucose to activate ERK in various cell types, and the simple presence FIG. 9. Effects of adenovirally transferred Glut1 and Glut2 on glucose-dependent activation of ERK and glucose uptake in L6 myotubes. Confluent, fully differentiated/fused myotubes in 24-well and 100-mm plates were use for studies of glucose uptake and ERK activation, respectively. The cells were infected with 4 multiplicity of infection adenovirus alone or adenovirus encoding Glut1 or Glut2 as indicated. After 48 h of incubation to allow time for expression, cells were equilibrated and incubated for 10 min in KRP medium containing indicated concentrations of D-glucose, along with, in B, 2 Ci of 3 Hlabeled D-glucose for determination of glucose uptake. After incubation, cell lysates were examined for immunoprecipitable ERK activity and cytochalasin B-inhibitable glucose uptake (calculated by dividing cpm in cells by the specific activity of medium glucose). Bars and brackets reflect the means Ϯ S.E. of four determinations. Shown below are immunoblots reflecting changes in contents of immunoreactive Glut1 and Glut2 (both blotted with rabbit polyclonal antisera) in virus-infected cells (note-Glut1 levels were not reproducibly altered by expression of Glut2). Although not shown, levels of ERK 1 and 2 were not altered by expression of viral constructs. of Glut1 glucose transporters does not necessarily imply the presence of metabolism/PKC-independent effects of glucose on ERK.
Interestingly, 3T3/L1 fibroblasts appeared to be more responsive to glucose but less responsive to insulin for ERK activation, as compared with 3T3/L1 adipocytes (Fig. 8B). Also, 3OMG, 2-DOG, and maltose, as well as glucose, activated ERK by a cytochalasin B-sensitive mechanism and concomitantly inhibited 2-DOG uptake in 3T3/L1 fibroblasts (Fig. 8, E and F). Thus, the Glut1 glucose transporter, operating in the absence of the Glut4 glucose transporter, as in 3T3/L1 fibroblasts, is sufficient to mediate the effects of metabolizable and nonmetabolizable and transported and nontransported sugars on both hexose uptake and ERK activation.

Effects of Adenoviral Gene Transfer of Glut1 and Glut2 and Their Chimeras on Glucose-dependent Activation of ERK and Glucose Uptake in L6 Myotubes and 3T3/L1 Fibroblasts
To more definitively evaluate the question of whether or not the amplifying effects of the Glut1 glucose transporter on glucosedependent ERK activation are caused by glucose uptake, we compared the effects of expression of wild-type forms of Glut1 and Glut2, and chimeric forms in which the C termini of Glut1 and Glut2 had been transposed, on both glucose-stimulated ERK activation and glucose uptake, using adenoviral gene transfer methods in L6 myotubes and 3T3/L1 fibroblasts. Whereas viral-mediated expression of Glut1 in L6 myotubes amplified glucose-dependent increases in ERK activity, expression of Glut2 had little or no effect on glucose-dependent ERK activation, despite the fact that glucose uptake was increased in cells expressing Glut2 to an extent slightly greater than that observed in cells expressing Glut1 (Fig. 9). Similarly, in 3T3/L1 fibroblasts, viral-mediated expression of either wild-type Glut1 or, particularly at the higher viral multiplicity of infection, the chimera containing the Glut1 C terminus led to an amplification of glucose-dependent effects on ERK (Fig. 10). In contrast, wild-type Glut2 and the chimera containing the Glut2 C terminus had little or no effect on glucose-dependent ERK activation in 3T3/L1 fibroblasts, despite the fact that glucose uptake was increased more effectively in cells expressing Glut2 or the chimera containing the Glut2 C terminus (Fig. 10). These findings provided further evidence that (a) glucose uptake is not, of itself, sufficient to increase ERK activity in either L6 myotubes or 3T3/L1 fibroblasts, (b) amplifying effects of Glut1 and chimeras containing the Glut1 C terminus on glucose-stimulated ERK are not explicable on the basis of alterations in glucose uptake, and (c) the Glut1 C terminus is required for amplifying effects on glucose-dependent ERK activation.

Linkage of Glut1 to the ERK Pathway via
Nonreceptor Tyrosine Kinase, PYK2 The above findings suggested that a tyrosine kinase was required for glucose-induced activation of ERK. Initial examination of cell lysates for phosphotyrosine (pY)-containing proteins revealed increases in the pY content of proteins migrating at approximately 125 kDa following acute glucose treatment of rat adipocytes and 3T3/L1 fibroblasts (not shown). Because the nonreceptor tyrosine kinase, PYK2, migrates at 125 kDa and because PYK2 is activated by pY-dependent mechanisms, autophosphorylates, and signals to GRB2 (28), it was of interest to find that, in both rat adipocytes and 3T3/L1 fibroblasts, glucose provoked acute increases in the phosphorylation of specific pY residues in PYK2, viz. Tyr 402 , the autophosphorylation site, and Tyr 881 , the site of interaction with the SH2 domain of GRB2 (Fig. 11).
Of further interest, PYK2 co-immunoprecipitated with Myctagged Glut1 obtained from extracts of 3T3/L1 fibroblasts transiently co-transfected with plasmids encoding wild-type Myctagged Glut1 and wild-type PYK2 (the latter used to augment PYK2 availability for co-immunoprecipitation, and 3T3/L1 cells used because of relatively high transient transfection rates). FIG. 10. Effects of adenovirally transferred Glut1, Glut2, and chimeric forms of Glut1 and Glut2 on glucose-dependent activation of ERK and glucose uptake in 3T3/L1 fibroblasts. As in Fig. 8, fully confluent cells were infected with 3 (left panels) or 12 (right panels) multiplicity of infection (MOI) adenovirus alone or adenovirus encoding wild-type Glut1, wild-type Glut2, chimeric N-Glut1/C-Glut2, or chimeric N-Glut2/C-Glut1. After 48 h to allow time for expression, cells were washed, equilibrated, and incubated for 10 min in glucose-free KRP medium without or with 20 mM D-glucose and, in B and D, 2 Ci of 3 H-labeled D-glucose to measure glucose uptake. After incubation, cell lysates were examined for glucose-dependent ERK activity and cytochalasin B-inhibitable glucose uptake (see Fig. 8 for other details). Values are the means Ϯ S.E. of four determinations. As shown by the representative immunoblots, expression of ERK 1 and 2 was not altered by infections with adenoviral constructs.
Most interesting, glucose acutely increased the interaction between Myc-Glut1 and PYK2 (Fig. 11). No such co-immunoprecipitation was seen in cells in which S465A or R468G mutant forms of Myc-tagged Glut1 were used instead of wild-type Myctagged Glut1. DISCUSSION It was surprising to find that, like glucose, nonmetabolizable saccharides, 3OMG, 2-DOG, lactose, and maltose activated ERK in several cell types. This suggested that certain nonmetabolizable saccharides can, presumably like D-glucose, act as ligands for a receptor that activates the ERK pathway. Moreover, because these nonmetabolizable saccharides activated ERK by a cytochalasin B-sensitive mechanism and concomitantly inhibited 2-DOG uptake through Glut1, the Glut1 glucose transporter itself may serve as a receptor for these and other saccharides. However, further studies are needed to verify that: (a) nonmetabolizable and nontransported saccharides activate ERK by the same mechanism used by glucose and (b) inhibitory effects of cytochalasin B on ERK activation are due to inhibition of binding of these saccharides to the Glut1 glucose transporter rather than to another protein.
In keeping with the suggestion that the cellular uptake of glucose was not required for glucose-dependent activation of ERK in the indicated cell types, the overexpression of apparently functional Glut4, even in the presence of insulin, failed to augment glucose effects on epitope-tagged ERK. In fact, Glut4 overexpression inhibited glucose-dependent ERK activation. The reason for this inhibitory effect of Glut4 is uncertain, but findings with dominant-negative Glut1 mutants suggested that specific Glut4 C-terminal sequences, perhaps via competition with endogenous wild-type Glut1 C-terminal sequences, can impair the activation of upstream components of the ERK signaling pathway.
More convincing support for the notion that glucose uptake could not account for amplifying effects of overexpression of Glut1 on glucose-stimulated ERK activation was derived from adenoviral gene transfer studies in L6 myotubes and 3T3/L1 fibroblasts. In these studies, the expression of Glut2 was, if anything, more effective than Glut1 overexpression for promoting glucose uptake, but, unlike Glut1 overexpression, Glut2 expression did not enhance glucose-dependent ERK activation.
Taken together, findings with truncated forms and chimeric forms of Glut1 suggested that C-terminal residues 463-468, viz. IASGFR, were responsible for amplifying effects of Glut1 on glucose-dependent ERK activation. Of these amino acids, the presence and position of serine 465 proved to be important, because simple replacement with alanine or swapping the positions of alanine 464 and serine 465 in the ISAAFR mutant (i.e. assuming that the concomitant replacement of glycine 466 with alanine was without effect) caused not only a loss of amplifying effects of Glut1 but generated dominant-negative forms of GLUT1 that inhibited glucose effects on ERK. In addition to serine 465, arginine 468 was important, because replacement with glycine also yielded a dominant-negative form of Glut1. This finding with the R468G mutant is of particular interest, because mutation of arginine 468 to leucine does not alter glucose transporting properties of Glut1 (35). Of further note, the ISAAFR sequence in the A464S/A465S/ G466A-Glut1 triple mutant is identical to that found in mouse Glut4, and there is no reason to think that this mutant would be deficient in transport. In this regard, as discussed above, just as alterations in transport did not appear to account for the stimulatory effects of Glut1 on ERK, it is unlikely that alterations in transport could account for inhibitory effects of the mutant forms of Glut1 on glucose-dependent ERK activation.
In any event, the dominant-negative inhibitory effects of S465A, A464S/S465A/G466A and R468G Glut1 mutants suggested that these mutants inhibited the ability of endogenous wild-type Glut1 to either transmit and/or amplify signals from glucose to the ERK pathway.
Our findings suggested that PYK2 functioned downstream of glucose and in conjunction with Glut1 during ERK activation. In A, cells were incubated in glucose-free KRP medium for indicated times with or without 20 mM glucose, and lysates were resolved by SDS-PAGE and blotted for immunoreactivity with phosphospecific anti-pY-PYK2 antibodies (using rabbit polyclonal antisera obtained from BIOSOURCE) as indicated. Representative blots are shown here. Similar results were observed in at least four experiments. In B, adipocytes were transiently co-transfected with plasmids encoding HA-ERK2 and WT or DN forms of PYK2 and subsequently incubated for 10 min in KRP medium with or without 20 mM glucose, following which, HA-ERK2 was immunoprecipitated and assayed. KD, kinase-defective; PRNK, PYK2-related noncatalytic kinase, i.e. C-terminal domain that lacks ability to phosphorylate substrate and therefore serves as effective dominant-negative. Insets show expression-related increases in contents of immunoreactive Wild-type and kinase-defective forms of PYK2 in total cellular lysates (note that approximately 5-10% of adipocytes are successfully transfected and relative increases in these transfected cell are greater than those seen in total cell extracts). Values are the means Ϯ S.E. of (n) determinations (where n is the number in parentheses). In C, 3T3/L1 fibroblasts were transiently cotransfected with plasmids encoding Myc-tagged Glut1 and PYK2 and subsequently incubated in KRP medium for 10 min with or without 20 mM glucose as indicated. After incubation, lysates were immunoprecipitated with anti-Myc mouse monoclonal antibodies (from Upstate Biotechnologies Inc.) and, after resolution by SDS-PAGE, blotted for immunoreactivity with anti-PYK2 antibodies (28). Shown here are representative blots from two experiments. Similar results were obtained in four experiments.
PYK2 was tyrosine-phosphorylated during acute glucose treatment and, moreover, co-immunoprecipitated with Myc-labeled Glut1 in a glucose-dependent manner. In addition, overexpression of wild-type PYK2 potentiated and dominant-negative forms of PYK2 inhibited glucose effects on ERK. The inhibitory effects of dantrolene also suggested that PYK2 was required for glucose effects on ERK (see Ref. 28).
The mechanism whereby glucose provoked increases in the pY content of PYK2 is uncertain. Phosphorylation of the autophosphorylation site in PYK2, Tyr 402 , implied that PYK2 was itself activated. However, it is uncertain whether PYK2 was activated as a result of binding to Glut1 or whether it was activated by other factors, including SRC family tyrosine kinases, e.g. FYN and YES. In any event, pY residues in PYK2 are capable of interacting with and activating upstream components of the ERK pathway, e.g. GRB2 via pY881-PYK2 and/or SHC via pY402-PYK2 (28).
With respect to dantrolene sensitivity, we have previously reported that glucose does not activate phospholipase C-dependent hydrolysis of PI 4,5-(PO 4 ) 2 in rat adipocytes (4), and we presently ruled out the involvement not only of DAG-dependent and atypical PKCs but also of Ca 2ϩ influx via L-type channels, because 10 M nifedipine had no effect on glucose stimulation of ERK (data not shown). In addition, glucose does not increase cytosolic Ca 2ϩ in rat adipocytes (36), and the Ca 2ϩ ionophore A32187 was without effect on ERK (not shown). Accordingly, although simple increases in cytosolic Ca 2ϩ cannot explain glucose effects on ERK, the inhibitory effects of dantrolene suggest a requirement for Ca 2ϩ during PYK2 activation.
The presently observed effects of glucose on ERK in the rat adipocyte are different from those seen in pancreatic ␤-cells, where glucose effects are dependent on both glycolysis and Ca 2ϩ influx via nifedipine-sensitive L-type channels (37)(38)(39), and those seen in kidney mesangial cells, where glucose effects are largely PKC-dependent (Ref. 40 and present findings). We are presently examining various cell types to see whether glucose activates ERK by a mechanism that is independent of metabolism and PKC. To date, as reported here, such activation has been seen in rat adipocytes, 3T3/L1 fibroblasts, 3T3/L1 adipocytes, L6 myotubes, and rat A-10 vascular smooth muscle cells, and preliminary studies suggest that glucose activates ERK independently of metabolism and PKC in rat soleus muscles.
From our findings, we speculate that glucose interacts with the Glut1 glucose transporter and/or other cell surface proteins that recruit/phosphorylate/activate PYK2 at the inner surface of the plasma membrane. PYK2 in turn activates GRB2 and/or SHC and other components of the ERK pathway. The recruitment of PYK2 and the amplifying effect of Glut1 on ERK activation apparently require and may be mediated through residues 463-468, IASGFR, in the Glut1 C terminus. Accordingly, Glut1 may act as a sensor, transducer, and amplifier during glucose signaling to PYK2 and the ERK pathway. Although further studies are needed to test this hypothesis, our findings are likely to be important for understanding the physiological and pathological actions of glucose.