Structural Origins of the Insulin-mimetic Activity of Bis(acetylacetonato)oxovanadium(IV)*

We have investigated the interaction of bis(acetylacetonato)oxovanadium(IV) (VO(acac)2) with bovine serum albumin (BSA) by EPR and angle-selected electron nuclear double resonance, correlating results with assays of glucose uptake by 3T3-L1 adipocytes. EPR spectra of VO(acac)2 showed no broadening in the presence of BSA; however, electron nuclear double resonance titrations of VO(acac)2 in the presence of BSA were indicative of adduct formation of VO(acac)2 with albumin of 1:1 stoichiometry. The influence of VO(acac)2 on uptake of 2-deoxy-d-[1-14C]glucose by serum-starved 3T3-L1 adipocytes was measured in the presence and absence of BSA. Glucose uptake was stimulated 9-fold in the presence of 0.5 mm VO(acac)2, 17-fold in the presence of 0.5 mm VO(acac)2 plus 1 mm BSA, and 22-fold in the presence of 100 nm insulin. BSA had no influence on glucose uptake, on the action of insulin, or on glucose uptake in the presence of VOSO4. The maximum insulin-mimetic effect of VO(acac)2 was observed at VO(acac)2:BSA ratios less than or equal to 1.0. Similar results were obtained also with bis(maltolato)oxovanadium(IV). These results suggest that the enhanced insulin-mimetic action of organic chelates of VO2+ may be dependent on adduct formation with BSA and possibly other serum transport proteins.

In the past several years the clinical potential of vanadium compounds in the treatment of type II diabetes has changed from low to high because of the introduction of an organic chelate of oxovanadium(IV) known as KP-102 into phase I trials (1). Studies in both laboratory animals and in humans have now convincingly demonstrated the lowering effects of vanadium compounds on blood glucose levels (2)(3)(4)(5). It has also been shown that the organic chelated vanadyl (VO 2ϩ ) compounds illustrated in Fig. 1 exhibit significantly enhanced insulin-mimetic activity in diabetic laboratory animals compared with that of inorganic VO 2ϩ introduced as VOSO 4 (6 -8). Because the capacity of organic chelates of VO 2ϩ to lower blood glucose is equivalent whether administered gastrointestinally or intraperitoneally (6), the enhanced insulin-mimetic action compared with that of VOSO 4 cannot be due only to increased lipophilicity, facilitating transport across the intestinal wall. The insulin-mimetic action of these organic chelates must be the result of how the organic moiety modulates the intrinsic chemical properties of the VO 2ϩ ion. While pH-dependent speciation of organic chelates observed on the basis of EPR spectra has been ascribed to rearrangements of the organic ligand moieties and displacement by solvent molecules (9 -11), it is not known whether these equilibria influence insulin-mimetic action. Furthermore, the physiologically active form of chelated VO 2ϩ in the blood stream is not established.
An important observation made by Chasteen and co-workers (12) shows that VO 2ϩ , when given as VOSO 4 by gastric intubation to laboratory rats, distributes itself in circulating plasma between the two major isoforms of the serum transferrins in proportion to the amount administered. Although serum albumin and transferrin bind VO 2ϩ tightly in the micromolar range (13)(14)(15)(16), it is difficult to ascribe the enhanced glucose-lowering capacity of organic VO 2ϩ complexes simply to the "stripping" out of the VO 2ϩ ion from its chelate ligand environment to form protein-bound VO 2ϩ in the blood stream. Such action would be likely to render organic VO 2ϩ complexes no more potent in glucose-lowering capacity than VOSO 4 itself. Since the serum transport proteins albumin, transferrin, and transthyretin bind a variety of organic ligands, e.g. fatty acids, steroids, and thyroxine hormone as carrier molecules in circulating blood (17)(18)(19)(20), we would argue that the organic moiety of VO 2ϩ chelates likely also facilitates binding to serum transport proteins. For this reason in these initial studies, we have investigated the potential of organic chelates of VO 2ϩ , namely VO(acac) 2 , 1 compound b in Fig. 1, to form adducts with serum albumin of defined stoichiometry. We have also analyzed the spectroscopic properties of VO(acac) 2 by EPR and ENDOR spectroscopy to determine the stoichiometry of the organic ligand bound to VO 2ϩ as a function of pH. To investigate whether serum proteins have an influence on the insulin-mimetic action of VO 2ϩ chelates, we have compared glucose transport, measured as the uptake of 2-deoxy-D-[1-14 C]glucose by 3T3-L1 adipocytes stimulated by VO(acac) 2 , in the absence and presence of albumin. The results disprove previous interpretations of the pH-dependent speciation of these organic chelates (10,11) and demonstrate that there is no change in the stoichiometry or coordination geometry of the bound organic ligand at low pH. Not only does VO(acac) 2 form a tightly bound adduct with albumin of 1:1 stoichiometry, but also a ratio of VO 2ϩ chelate to protein of Յ1.0 elicits maximal enhancement of insulin-mimetic activity in cultured 3T3-L1 adipocytes. * This work was supported by National Institutes of Health Grant DK20959. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Cell Culture and Experimental Treatment-A protocol was developed to avoid the presence of serum or albumin while measuring the influence of insulin and insulin-mimetic compounds on the uptake of 2-deoxy-[1-14 C]glucose by 3T3-L1 adipocytes. 3T3-L1 fibroblasts were maintained and differentiated into adipocytes as reported previously (21). Cells were used 4 -9 days after completion of the differentiation protocol when Ͼ95% of the cells contained lipid droplets. Prior to insulin stimulation or treatment with VO 2ϩ compounds, cells were washed twice with phosphate-buffered isotonic saline at 37°C and serum-starved for 2.5 h at 37°C in 1 ml/well KRBH with 5 mM glucose and 25 mM HEPES, pH 7.4. The basal and insulin-stimulated rates of glucose transport in adipocytes treated in this manner were identical to cells serum-starved in KRBH/glucose plus 0.5% BSA and were similar to previous results when Dulbecco's modified Eagle's medium plus 0.5% fetal bovine serum was used (22). The serum starvation medium was then removed, the cells were washed two times with phosphate-buffered isotonic saline (37°C), and the adipocytes were placed in 0.5 ml/well KRBH containing insulin, the VO 2ϩ compound desired (accordingly in the absence or presence of BSA), or no insulin-mimetic compound. After 30 min at 37°C, 20 M 2-deoxy-D-[1-14 C]glucose (ϳ20 cpm/pmol) was added to all wells. After 5 min at room temperature, the assay was terminated by addition of 50 l of 200 mM 2-deoxy-D-glucose and washing the cells three times with phosphate-buffered isotonic saline on ice. Adipocytes were collected in 0.5 ml of distilled water, and 2-deoxyglucose uptake was determined by liquid scintillation counting.
Metabolic Assays-Concentrated stock solutions of VO(acac) 2 or VO-(malto) 2 for metabolic studies were made by dissolving the crystalline compound in a small volume of absolute ethanol followed by dilution so that the final solution contained 150 mM sodium chloride buffered to pH 7.4 with 10 mM HEPES. The buffered saline had been previously purged with nitrogen gas. Aliquots of the buffered saline suspension of the organic chelate were then added to the wells in the presence or absence of BSA. (The ethanol content of incubation mixtures was no greater than 1%.) Before use, BSA had been exhaustively dialyzed against HEPES-buffered saline at pH 7.4 containing 10 mM EDTA followed by dialysis against HEPES-buffered saline without EDTA to remove possible contaminant vanadium (23). No vanadyl species could be detected in the BSA solution by EPR. Concentrated stock solutions of VOSO 4 were prepared by dissolving vanadyl sulfate hydrate in a small volume of H 2 O under a nitrogen atmosphere, and aliquots were added to a solution of BSA in buffered isotonic saline to result in 1 mM final concentration of each. This solution was then used for suspension of adipocytes for glucose uptake measurements.
EPR and ENDOR Studies-Stock solutions of VO(acac) 2 for acidbase titrations monitored by EPR were made by dissolving the crystalline compound in methanol or in nitrogen-purged water. EPR spectra of aqueous or methanol solutions of VO(acac) 2 at ambient room temperature were collected with the sample in quartz capillaries. For titrations of VO(acac) 2 in aqueous solution, small aliquots of concentrated HCl or NaOH were added. Before and after spectral recording, the pH was measured with a Radiometer PHM82 standard pH meter equipped with a glass electrode to ensure that the pH of the solution had not changed. The pH meter was generally calibrated with two standard pH solutions bracketing the pH of the test solution. The volume change after addition of concentrated acid or base was less than 1%.
Acid-base titrations of VO(acac) 2 in methanol were carried out by two different methods. Equivalent results were obtained with both methods. Either small aliquots of concentrated aqueous HCl or NaOH were added to the methanolic solution, the volume of the aqueous component remaining Ͻ1% of the total volume. To completely avoid possible effects of the aqueous component, aliquots of methanol saturated with dry, gaseous HCl or NH 3 were added to the methanolic solution of VO(acac) 2 to alter [H ϩ ]. The pH* was measured with a glass electrode prior to and after spectral recording according to Bates (24), allowing adequate time for equilibration of the electrode. The nominal pH reading was converted to pH* according to the relationship pH* ϵ pa H * ϭ pH Ϫ ␦, where pH is the nominal reading of the electrode and ␦ has the value of Ϫ2.34 for 100% methanol as solvent (24).
EPR and ENDOR spectra were recorded with an X-band Bruker ESP 300E spectrometer equipped with a cylindrical TE 011 ENDOR cavity, an Oxford Instruments ESR910 liquid helium cryostat, and Bruker ENDOR digital accessory as described previously (25,26). The ESP 300E spectrometer was equipped with a complete computer interface (Bruker ESP3220 data system) for spectrometer control and data acquisition and processing. Typical experimental conditions for ENDOR measurements were: temperature, 20 K; microwave frequency, 9.45 GHz; incident microwave power, 64 microwatts (full power, 640 milliwatts at 0 dB); rf power, 50 -70 watts; rf modulation frequency, 12.5 kHz; and rf modulation depth, 10 -20 kHz. The static laboratory magnetic field was not modulated for ENDOR. EPR spectra were simulated with use of the program WINEPR2.11 (Bruker Instruments, Inc., Bellerica, MA) as described previously (27).

RESULTS AND DISCUSSION
The pH Dependence of the EPR Spectrum of VO(acac) 2 - Fig.  2 compares representative EPR spectra of VO(acac) 2 at ambient room temperature in aqueous solution with those observed in methanol. The absorption intensity of the EPR spectrum of the S ϭ 1/2 oxovanadium(IV) ion in solution at ambient temperatures is distributed over eight components due to the hyperfine coupling of the unpaired electron with the (I ϭ 7/2) 51 V nucleus. While the centrally located components for the different spectral species overlap heavily with each other over the titratable pH range, the low field and high field components are separated from each other at extremes of protonic activity. In aqueous solutions we have observed the reversible formation of four spectrally distinct species, labeled A-D, while in methanol only three were observed. Vertical lines have been drawn, therefore, in Fig. 2 at low field and high field positions identifying each species. For both solvent systems, it is seen that the vertical lines for species B and BЈ in aqueous and methanol systems, respectively, identify a weak shoulder that is not part of the spectrum for species A or AЈ. It is seen in Table I that species A-D and AЈ-CЈ in aqueous and methanol solutions, respectively, differ from each other primarily on the basis of A o values. Comparable observations have been made earlier by others for VO(acac) 2 (11) and VO(malto) 2 (9). The values of the spectroscopic parameters g ʈ , g Ќ and A ʈ , A Ќ for these species obtained from frozen solution spectra, summarized in Table I, are nearly identical, indicating very similar structural environments of the oxovanadium(IV) ion. Corresponding values for VO(acac) 2 have not been reported from frozen solution spectra by others. Our results for VO(acac) 2 from frozen solution spectra are comparable to those reported for VO(malto) 2 (9).
In Table I it is seen that the g-and A-values for species A, B, and C in aqueous solution are within experimental measurement identical to the g-and A-values of species AЈ, BЈ, and CЈ in methanol solution, respectively. Only in aqueous solution is the fourth species D observed. Its spectrum yields g-and A-values identical to those of the [VO(H 2 O) 5 ] 2ϩ ion (9,26,28). However, no previous study has shown the equivalence of species A, B, and C in aqueous solution to AЈ, BЈ, and CЈ, respectively, in methanol. Crans and co-workers (10, 11) have stated that for VO(acac) 2 freshly prepared in aqueous solutions the conversion of species A to species C is time-dependent, requiring up to 11 days at ambient temperature. The spectral changes have, therefore, been interpreted to reflect kinetically sluggish, time-dependent alterations in coordination geometry and displacement of an equatorial organic ligand by water. With freshly prepared, unbuffered solutions of species A, we have similarly observed this phenomenon after 11 days but find that the change in spectra was accompanied by a corresponding decrease in pH. On the other hand, direct adjustment of the pH of a solution of species A elicits the conversion of species A to C immediately within the mixing time. Since the g-and A-values provide a signature of elemental composition of the four equatorial donor-ligand atoms of the complex (29), we conclude that species A, B, and C in aqueous solutions are identical to their AЈ, BЈ, and CЈ counterparts in methanol, respectively. Fig. 3 illustrates titration curves constructed on the basis of the EPR spectra of VO(acac) 2 in aqueous solution, demonstrating the reproducibility of the data and complete reversibility of ionizations. Data of equivalent precision were collected for titrations in methanol. The titration curves could be constructed only through careful measurement of the peak-to-peak amplitudes of the hyperfine components for each spectral species. Exquisite care was taken to avoid introduction of air into the solution to prevent oxidation of the oxovanadium(IV) ion. Species A, viz. AЈ, is readily defined under conditions of low acidity; however, as [H ϩ ] increases, species B, viz. BЈ, appears with closely overlapping spectral features before species C, viz. CЈ, is detectable. The spectral features of species A and B remained closely overlapping, and titration data could be analyzed only by considering the summed total of species A(ϩ B) distinct from species C. The titration curves, therefore, were calculated as a function of the composition of A(ϩ B), C, and D in aqueous solution and, correspondingly, AЈ(ϩ BЈ) and CЈ in methanol. In aqueous solution, as illustrated in Fig. 3, one ionization was observed with a pK a value of ϳ3.1, governing the reversible interconversion between species A(ϩ B) and C, while the second ionization with a pK a value of ϳ1.7 governs the equilibrium between species C and D. In methanol only one ionization was observed as a function of protonic activity with a pK a * value of ϳ5.1, governing the interconversion between species AЈ(ϩ BЈ) and CЈ.
The spectral speciation of VO(acac) 2 (10,11) and of VO(malto) 2 (9) has been previously noted by others on the basis of EPR spectra; however, the values of ionization constants governing their interconversion have not been estimated, and the spectroscopic equivalence of species A, B, and C for VO(acac) 2 in The symbols indicate the following directions of the acid-base titrations: ‫,ء‬ from low pH to high pH; ‚ and E, from high pH to low pH. The species identified in Fig. 2 are indicated by color: A(ϩ B), black; C, red; and D, green. Since the EPR spectrum of a solution of VO(acac) 2 at any given pH results from the summed contributions of the individual species present, the relative contributions of each identified species to the composite spectrum were evaluated by spectral fitting using spectra of the "pure" species observed at the extremes of pH according to the relationship A ϫ f A (H) ϩ (1 Ϫ A ) ϫ f C (H) ϭ f obs (H), where A represents the fraction of species A(ϩ B), and f A (H) represents the spectrum of pure A(ϩ B) as a function of magnetic field position H, and f obs (H) represents the experimentally observed spectrum at a defined pH, the other quantities having analogous definitions. This approach was extended to include the spectral contributions of species D as the penta aquo VO 2ϩ cation at low pH in aqueous solutions. The solid lines represent the distribution of each species as a function of pH calculated for single ionizations. It is seen that there is a small but systematic deviation of the data from the calculated titration curves in the pH range 2-5. We attribute this small deviation to the fact that we were unable to separate the spectral features of species B from those of species A. The ordinate axis represents the fraction (f) of each species present according to the ionization equilibria. a pH* in methanol solutions was measured as described by Bates (24). Values of g are estimated to within Ϯ0.001; values of A are estimated to within Ϯ0.5 G. The microwave frequency was 9.432 GHz. Anisotropic spectroscopic parameters (g ʈ , g -, A ʈ , and A -) were determined by spectral fitting of frozen solution spectra (cf. Ref. 27), and isotropic parameters (g o and A o ) were determined from fluid solution spectra.
b Overlapping resonance components of species A, AЈ allows only an approximate estimate.
aqueous solutions and of AЈ, BЈ, and CЈ in methanol solutions has not been demonstrated hitherto. While the molecular origins of the two ionizations in aqueous solution are under further investigation in this laboratory through use of perdeuterated VO(acac) 2 , the results do demonstrate that the metalloorganic chelate VO(acac) 2 is stable in aqueous solutions of pH ϳ2. This observation, together with its expected greater lipophilicity, undoubtedly underlies in part its enhanced insulinmimetic activity when introduced orally to laboratory animals compared with that of VOSO 4 (6). ENDOR Characterization of VO(acac) 2 -Isomerizations of the organic ligand, replacement of carbonyl oxygen atoms by solvent molecules, and other similar structural changes have been attributed to underlie the pH-dependent spectral speciation of VO(acac) 2 (10,11). However, the values of the spectroscopic parameters in Table I extracted from spectra can be related only to the average elemental composition of the equatorial donor-ligand atoms with appropriate changes in vanadium-oxygen covalency (29). The spectroscopic effect of a solvent oxygen atom is essentially indistinguishable from that of a carbonyl or hydroxyl oxygen. Angle-selected ENDOR, therefore, becomes the method of choice to assign the coordination geometry of the acetylacetonate ligand.
The underlying principles of angle-selected ENDOR of VO 2ϩ have been described in earlier studies from this laboratory (25, 30 -33) and are only briefly summarized here. When H 0 is set to the Ϫ7/2 ʈ feature of the EPR absorption spectrum and is, therefore, parallel to the symmetry axis or the VϭO bond of the complex, a proton located along the symmetry axis gives rise to a parallel hyperfine (hf) resonance coupling (A ʈ ), while A Ќ or the perpendicular hf coupling is observed for a proton in the molecular x, y plane. On the other hand, when the field is set to the Ϫ3/2 Ќ absorption feature, an axial proton gives rise to a perpendicular hf coupling A Ќ , and the equatorial proton gives rise to a combination of parallel and perpendicular hf couplings. The combination of parallel and perpendicular hf couplings is observed only for a proton in the equatorial plane for the Ϫ3/2 Ќ setting of H 0 . On the other hand, only an axial proton gives rise to a single pair of ENDOR features when H 0 is set to the Ϫ3/2 Ќ absorption feature. This variation in the resonance pattern, dependent on the orientation of the magnetic field H 0 with respect to magnetic axes in the molecule, is the essence of angle-selected ENDOR as first observed by Rist and Hyde (34).
The detection of ENDOR features of the VO 2ϩ ion is heavily dependent on the nearby solvent environment and the quality of glass formation (25, 30 -33). For this reason, because of the near identity of the g-and A-values of species A, B, and C to those of AЈ, BЈ, and CЈ, respectively, we have collected ENDOR spectra of species AЈ and CЈ only in methanol because of its glass-forming properties. Water does not form a glass upon freezing, and therefore, broadening of ENDOR lines occurs because of variations in the crystalline field, particularly for small molecule VO 2ϩ complexes, preventing detection of ENDOR absorptions. Fig. 4 illustrates the proton ENDOR spectra of VO(acac) 2 in perdeuterated methanol. Only nonexchangeable, covalently attached hydrogens in the acetylacetonate ligand are detected by ENDOR under these conditions. With H 0 set to the Ϫ3/2 Ќ component of the EPR spectrum, both A ʈ and the A Ќ hf couplings are observed for species AЈ formed at high pH* and for species CЈ formed at low pH* as defined by the spectra in Fig. 2. On the other hand, with H 0 set to the Ϫ7/2 ʈ component of the EPR spectrum, only the A Ќ hf couplings were detected. This pattern of resonances is seen only when the ENDOR-detected hydrogens are in the equatorial plane. This result, therefore, indicates that there is no change in geometry of the bound acetylacetonate ligand with change in [H ϩ ]. Since the peak-to-peak amplitudes and line widths of the resonance features for both AЈ and CЈ species are identical within experimental measurement, the ENDOR spectra also indicate that there is no change in stoichiometry of bound acetylacetonate ligand with change in pH*. These observations were confirmed by collecting ENDOR spectra of VO(acac) 2 in [ 2 H 3 ]methanol (data not shown). No resonance features characteristic of equatorial OH groups (30) could be observed for species AЈ or CЈ that would have resulted from displacement of an acetylacetonate oxygen by a methanolic hydroxyl group according to the equilibria proposed (10,11). Also since the ENDOR spectra for both species AЈ and CЈ have their origin only in equatorial hydrogens covalently attached to the organic ligand, proposed isomerizations of the ligand into axial positions (10,11) are excluded. The molecular origins of the ionizations described in Figs. 2 and 3, therefore, must derive from changes in the covalency of vanadium-oxygen interactions induced through protonation-dependent changes of outer sphere solvent molecules hydrogen bonded to the equatorial carbonyl oxygens or to the vanadyl oxygen. Fig. 5 compares proton ENDOR spectra of VO(acac) 2 in deuterated aqueous buffer in the presence of BSA. The spectra were collected with H 0 set to both the Ϫ3/2 Ќ and Ϫ7/2 ʈ components of the EPR spectrum of VO(acac) 2 . While the underlying ENDOR features of the acetylacetonate ligand remained unchanged indicating that the organic chelating ligand was not displaced, there are additional features near the Larmor frequency that have their origin in protein residues. Since exchangeable protons on the serum albumin molecule will have been substituted by deuterons from the solvent, the new resonance features can be ascribed only to covalent hydrogens of nearby amino acid residues. The ENDOR splittings of these FIG. 4. ENDOR spectra of VO(acac) 2 of species A (solid line spectrum) and C (dotted line spectrum) in methanol. The prominent resonance features in both the Ϫ7/2 parallel and Ϫ3/2 perpendicular spectra are due to methyl groups, while the weak features in the Ϫ3/2 perpendicular spectrum arise from the bridge proton. It is seen that the pair of features observed in the Ϫ7/2 parallel spectrum also appear in the Ϫ3/2 perpendicular spectrum with identical ENDOR splitting. On the other hand, in the Ϫ3/2 perpendicular spectrum the added pair of prominent ENDOR lines exhibit the shape expected for parallel hf interactions (29 -32) and are associated with an ENDOR splitting twice that of the inner pair. These observations, therefore, assign the prominent ENDOR features to A ʈ and A Ќ hf coupling components of the methyl groups. The sample volume, concentration, and geometry of the sample in the ENDOR cavity were kept constant for collection of ENDOR spectra of species AЈ and CЈ. Therefore, the relative amplitudes of the resonance features and line widths can be considered directly proportionate to the number of covalent hydrogens of the organic ligand contributing to the spectra for each species. While the unchanging peak-to-peak amplitudes and line widths of the resonance features demonstrate that there is no change in the stoichiometry of the bound organic ligand, the observation of only two pairs of resonance features for the methyl groups requires the complex to have 2-fold symmetry. The abscissa indicates the ENDOR shift where 0 MHz represents the Larmor frequency. T ϭ 10 K. resonance features arise from amino acid hydrogens over a 5-10-Å distance from the vanadium nucleus (25, 30 -33); they, therefore, clearly indicate binding of the organic chelate to the protein molecule. The plot in Fig. 6 shows that binding occurs to form an adduct of 1:1 stoichiometry. Because the underlying features of the acetylacetonate ligand are observed in the ENDOR spectra in Fig. 6, the spectra indicate that the VO(acac) 2 complex remains intact upon binding to the protein.
The vanadium hf components in the EPR spectra of VO(acac) 2 showed no broadening in the presence of BSA. Comparable observations have been reported by others (35). Since unresolved ligand hf broadening measurably adds to the EPR line width only for hydrogens covalently attached to equatorial donor-ligand atoms (36), we conclude that the oxovanadium(IV) moiety is bound axially to the protein either through a residue hydrogen bonded to the axial water molecule or through a residue directly coordinated to the vanadium ion, having displaced the axial water. At present we cannot distinguish between these two possibilities.
Measurement of the Insulin-mimetic Action of VO(acac) 2 -Albumin is often added to cell culture medium as a neutral, protectant macromolecule (cf. Ref. 7); therefore, we have used serum-starved, differentiated 3T3-L1 adipocytes for metabolic assays to avoid albumin as a complicating factor in the assay medium as described under "Experimental Procedures." Cells were washed twice with phosphate-buffered isotonic saline (37°C) and serum-starved for 2.5 h in KRBH lacking BSA or serum. The medium was removed, and cells were incubated for 30 min in KRBH with the indicated additions prior to measurement of glucose transport rates. On this basis, we were able to separate the intrinsic influence of a VO 2ϩ compound alone on the uptake of 2-deoxy-D-[1- 14 C]glucose from that in the presence of added BSA. Comparative effects are summarized in Table II for four VO 2ϩ -containing systems.
As shown in Table II, added BSA had no influence on the action of insulin or on the basal rate of glucose uptake. However, not only did the addition of BSA have a pronounced effect on the insulin-mimetic activity of VO(acac) 2 , but also this varied according to the molar ratio of added BSA to VO 2ϩ chelate. Fig. 7 compares in histogram form the uptake of 2-deoxy-D-[1-14 C]glucose as a function of the VO(acac) 2 :BSA molar ratio. It is seen that the near maximal influence of BSA on the insulinmimetic effect of VO(acac) 2 occurs at a VO(acac) 2 :BSA ratio of ϳ1.0. Addition of VO(acac) 2 in excess of this ratio resulted in a decrease in the enhancement of glucose uptake. Since the results of ENDOR titrations in Fig. 5 indicate that the VO 2ϩ ion is not removed from its organic chelate environment, we believe that the lower activity observed at VO(acac) 2 :BSA ratios Ն1.0 is due to the intrinsically lower activity of the free, unbound portion of   VO(acac) 2 (cf. Table II). Although comparable effects were observed also for VO(malto) 2 , the influence of added BSA was less pronounced. In contrast to the enhancement of the insulinmimetic action of VO(acac) 2 or VO(malto) 2 by added BSA, no influence for [N-(2-hydroxyethyl)-iminodiacetato]oxovanadium(IV) was observed above the basal level for the free VO 2ϩ ion added as VOSO 4 (cf . Table II).
Conclusions-The molecular mechanisms by which VO 2ϩ and its chelates exert their insulin-mimetic effects in vivo and in vitro have not been fully clarified. It has been demonstrated that both inorganic salts and organic chelates of VO 2ϩ exert their insulin-mimetic effect, at least in part, by stimulating lipogenesis in adipocytes via membrane-associated phosphotyrosine phosphatases and a cytosolic, non-insulin receptor protein tyrosine kinase (7). Other enzyme systems may also be involved. Since enhancement of lipogenesis by VO(acac) 2 has been shown to be far greater than that by VOSO 4 in the studies by Shechter and co-workers (7), the structure and affinity of the organic ligand for VO 2ϩ are likely to be critical in the expression of insulin-mimetic activity.
It has been previously postulated on the basis of EPR spectra of VO(acac) 2 that isomerizations and displacement of equatorial oxygen-donor atoms by solvent underlie their pH-dependent spectral speciation (10,11). Our ENDOR results demonstrate in contrast that there is no change in coordination geometry or stoichiometry of the bound acetylacetonate ligand and that the ionizations can be ascribed only to outer sphere solvent molecules hydrogen bonded to equatorial or axial oxygens of the VO 2ϩ chelate. Of particular interest, therefore, is the observation through Figs. 5 and 6 that VO(acac) 2 formed a specific adduct with BSA of 1:1 stoichiometry. Protein-chelate adduct formation was correlated with enhanced insulin-mimetic activity of VO(acac) 2 over that with other organic chelates of VO 2ϩ and over that of VO 2ϩ bound to BSA. These results suggest that formation of protein-chelate complexes in the bloodstream may be an important step for the insulinmimetic action of these compounds in vivo.
Insulin regulates blood glucose levels through the stimulation of glucose uptake and storage by peripheral tissues and suppression of hepatic glucose output. In type II diabetes, these tissues become resistant to the action of insulin, resulting in chronically elevated plasma glucose levels and numerous secondary complications. The insulin-mimetic effect of chelated VO 2ϩ observed in streptozocin-diabetic laboratory animals is attributed to a reduction in the level of insulin needed to maintain euglycemia (6,37). To this end, we have examined the influence of VO(acac) 2 on glucose uptake by 3T3-L1 adipocytes stimulated by insulin in the presence of BSA. The results showed that the effect of insulin is enhanced in the presence of low levels of VO(acac) 2 by an amount that is greater than that due to VO(acac) 2 alone in the presence of BSA. 2 These observations thus suggest that the two agents influence the same pathway and act synergistically. Furthermore, there was a saturating effect on the synergy that appeared to be dependent on both concentration of VO(acac) 2 and the concentration of added insulin. These conditions are being explored further to identify the enzyme pathways involved.
Our observations that complex formation of serum albumin with VO(acac) 2 further stimulates insulin-mimetic activity over that of the organic chelate alone, thus identify a factor that may be important in controlling drug potency and distribution of organic chelates of VO 2ϩ to insulin-sensitive tissues, namely binding of the organic chelate to serum transport proteins. This factor has not been considered heretofore. Differential enhancement of the insulin-mimetic action of VO 2ϩ chelates by BSA reported in Table II may be due to variation in the specific activity of the protein-chelate adduct, a different binding affinity of the chelates for BSA, or the stripping out of the VO 2ϩ ion from its chelate environment. It is also possible that other serum transport proteins such as transferrin and transthyretin may form specific adducts with organic chelates of VO 2ϩ with differential enhancement of their insulin-mimetic action. Our results thus point to a variety of protein-chelate interactions that may be associated with enhancement of their insulin-mimetic action. Of particular interest is that they may lead to multiple routes for development of pharmacologic agents for treatment of type II diabetes.