Abstract
Three ω-18F-fluoro-n-alkyl-β-d-glucosides (alkyl = ethyl (5a)), n-butyl (5b), and n-octyl (5c)) were synthesized and evaluated as potential substrates for the sodium/d-glucose cotransporter SGLT1. Methods: The ligands were prepared by 18F-fluoride displacement of the corresponding tetraacetyl-protected tosylate alkylglucoside precursors in CH3CN, followed by hydrolysis of the protective acetate esters with NaOMe/MeOH. Transport of the nonradioactive analogs 5a, 5b, and 5c by the human sodium-d-glucose cotransporter hSGLT1 was characterized in vitro in oocytes of Xenopus laevis that expressed hSGLT1. The biodistribution of the tracers was determined in mice and the presence of metabolites in the blood was investigated. Compound 5a was also evaluated in mice pretreated with phlorizin. The intrarenal distribution of 5a in mice kidney was visualized using autoradiography. Results: The radiochemical yield of 5a, 5b, and 5c was in the range of 8%–15% (end of bombardment) with a total synthesis time of 90 min. The in vitro evaluation after expression of the hSGLT1 showed that 2′-fluoroethyl-β-d-glucoside (5a) was transported with similar Michaelis-Menten Km and Vmax (maximum velocity) values as compared with methyl-α-d-glucopyranoside (αMDG). The more lipophilic compounds 5b and 5c were not transported but inhibited transport of αMDG. In vivo tissue distribution in mice revealed that 5a, 5b, and 5c were cleared mainly by the renal pathway and that 5a showed a significantly higher accumulation in the kidneys and a slower renal excretion as compared with 5b and 5c. Compound 5a was retained mainly in the renal outer medulla containing S3 segments of proximal tubules and the accumulation could be blocked by phlorizin pretreatment. Compound 5c passed the blood-brain barrier to some extent. Conclusion: The data indicate that 2′-18F-fluoroethyl-β-d-glucoside (5a) is a specific tracer of Na+-dependent glucose transport that may be used to visualize this transport activity in the S3 segments of renal proximal tubules.
Secondary active d-glucose transporters (SGLTs) play a pivotal role in the absorption of d-glucose in small intestine and in the reabsorption of d-glucose in renal proximal tubules (1,2). Up to 6 isoforms of SGLT (SGLT1–SGLT6) have been identified that differ in tissue distribution and functional properties and exhibit some species differences (1–3). SGLT1 is expressed in the luminal membrane of small intestinal enterocytes, where it is responsible for the absorption of d-glucose, and in the S3 segment of renal proximal tubules, where it mediates high-affinity reabsorption of d-glucose that has not been reabsorbed earlier. SGLT2 is a low-affinity sodium-d-glucose cotransporter that is expressed in the S1 and S2 segments of renal proximal tubules and reabsorbs the bulk of ultrafiltered d-glucose (2). SGLT3 from pig has been identified as a sodium-d-glucose cotransporter that is expressed in small intestine but not in kidney (3,4); however, no transport function could be detected for human SGLT3 (5). Transcription of SGLT1 has also been observed in several tumor cell lines and various human carcinomas (6,7).
The SGLT transporters exhibit a variety of important physiologic functions and are of biomedical importance. Genetic defects of the SGLT1 and SGLT2 result in glucose-galactose malabsorption (8) or familial renal glycosuria (9), respectively. In a rat model of short bowel syndrome, uptake of galactose could be enhanced by transfection of SGLT-1 into colon mucosa (10). Compounds that inhibit SGLTs are currently explored for the treatment of noninsulin-dependent diabetes mellitus with the aim of increasing the urinary excretion of excess plasma glucose (11). The decrease of methyl-α-d-glucopyranoside (αMDG) uptake in isolated proximal tubules correlates with nephrotoxicity of xenobiotics (12).
Recently it has been shown that SGLT1 is also expressed in neurons of the central nervous system (CNS) and in endothelial cells of brain arteries and capillaries (13,14; H. Koepsell, unpublished data, 2003). This was a surprising observation since it is generally believed that transport of d-glucose in the CNS is only mediated by facilitated diffusion through GLUT1 and GLUT3 transporters. It has been established that GLUT1 facilitates transport of glucose over the blood-brain barrier (BBB) and that GLUT3 mediates transport of glucose into neurons (15,16). The functional role of SGLT1 in brain is not well understood. Poppe et al. (13) raised the hypothesis that SGLT1 may account for the increased availability of glucose during metabolic stress in the brain. They found a local accumulation of 14C-αMDG after inducing a seizure in rat brain. They argued that the increased uptake of αMDG may reflect the induction of SGLT1 by the increased demand for d-glucose, since αMDG is a substrate of SGLT1 but not of GLUT1 and GLUT3 (17).
Given the fact that 2-deoxy-d-glucose is a poor substrate for SGLT1 (18), PET studies with 18F-FDG do not allow evaluation of the function of SGLT1 in brain or other tissues. The aim of this study was to prepare an 18F-labeled substrate with high affinity for SGLT1 that would allow studying the function of the SGLT1 transporter with PET in vivo. Glucosides with an alkyl moiety in the β-position appeared to be appropriate starting compounds, as it has been reported that they have a higher affinity for SGLT1 compared with d-glucose (19). Some of these compounds were found to inhibit uptake of d-glucose but are not transported themselves. This may be due to the longer alkyl side chain that increases the affinity for SGLT1 but prevents the translocation step (19).
In this study, we describe the synthesis and evaluation of 2′-18F-fluoroethyl-β-d-glucoside (5a), ω-18F-fluoro-n-butyl-β-d-glucoside (5b), and ω-18F-fluoro-n-octyl-β-d-glucoside (5c). The corresponding nonradioactive compounds 5a, 5b, and 5c were also prepared. In vitro, 5a, 5b, and 5c were evaluated for their ability to be transported by human SGLT1 (hSGLT1) and for their inhibition of transport of αMDG by hSGLT1. The biologic behavior of 5a, 5b, and 5c in vivo was determined in mice.
MATERIALS AND METHODS
Acetobromoglucose (1) was obtained from Sigma. β-Bromoethyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside was prepared as described by Dahmén et al. (20). 1H and 13C NMR spectra were recorded on a Gemini 200-MHz spectrometer (Varian). Chemical shifts are reported in parts per million relative to tetramethylsilane (δ = 0). Exact mass measurements were performed on a time-of-flight spectrometer (LCT; Micromass) equipped with a standard electrospray ionization (ESI) interface. Samples were infused in 2-propanol/water (1:1 v/v) with a Harvard 22 syringe pump (Harvard Instruments). For accurate mass calculations, a lock mass solution of 0.1 mg/mL solution of Kryptofix 2.2.2. (Merck) at a flow rate of 1 μL/min was coinfused. Preparative column chromatography was performed on silica gel 60 (Merck). Batches of 18F-fluoride were prepared by an 18O(p,n)18F reaction by irradiation of 2 mL 18O-enriched water (Cortec) in a titanium target with 10-MeV protons using a Cyclone 10/5 cyclotron (IBA). Quality control of labeled 3a, 3b, and 3c was performed on a Carbohydrate Amino column (10 μm, 4.1 × 300 mm; Alltech), eluted with a linear gradient of H2O, CH3CN, and 0.1 mol/L NaF. The gradient was composed as follows (v/v/v): 0 min, 10:90:0; 6 min, 0:50:50; 7.5 min, 0:0:100. The column effluent was monitored with a refractive index detector, an ultraviolet detector set at 254 nm, and a 2-in. (5.1 cm) NaI(Tl) scintillation detector coupled to a single-channel analyzer. Output of the detectors was analyzed by a RaChel analysis program (Lablogic). Radiochemical yields were corrected for decay. The identity of the 18F-labeled compounds was confirmed on high-performance liquid chromatography (HPLC) by coelution of authentic material.
Synthesis
2′-Tosyloxyethyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (3a).
A solution of 0.49 g (1.1 mmol) 2′-bromoethyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (20) and 0.60 g (2.1 mmol) silver tosylate in 25 mL CH3CN was refluxed for 5 d shielded from light. The solvent was evaporated and the residue was applied on a silica column eluted with a gradient of ether/hexane (1:1 v/v increased to 2:1 v/v). Yield was 0.37 g (0.7 mmol, 63%) of a white solid. 1H NMR (CDCl3) δ 2.00, 2.02, 2.05, 2.08 (4 × s, 12H), 2.46 (s, 3H), 3.68 (m, 1H), 3.90 (dm, 2H), 4.10 (dd, 1H), 4.15 (t, 2H), 4.24 (dd, 1H), 4.51 (d, 1H, 3J = 7.8 Hz), 4.95 (dd, 1H), 5.04 (t, 1H), 5.19 (t, 1H), 7.37 (d, 2H), 7.78 (d, 2H).
2′-Fluoroethyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (4a).
To a solution of 3.9 g (10 mmol) 1,2,3,4,6-penta-O-acetyl-β-d-glucose and 0.8 mL (13.6 mmol) 2-fluoroethanol in 20 mL CH2Cl2 at 0°C was added 6.2 mL (50 mmol) boron trifluoride etherate. The mixture was stirred overnight at room temperature. The reaction mixture was poured onto ice and neutralized with saturated NaHCO3 solution. The water layer was extracted with CH2Cl2 and the organic layer was dried over MgSO4. Purification was performed on a silica column eluted with ether/hexane (1:1 v/v). The desired product was allowed to crystallize from the eluent, yielding 1.3 g (3.3 mmol, 33%) of a white crystalline product. 1H NMR (CDCl3) δ 2.01, 2.03, 2.05, 2.09 (4 × s, 12H), 3.72 (m, 1H), 3.87 (dm, 2H), 4.15 (dd, 1H), 4.28 (dd, 1H), 4.54 (dt, 2H, 2JHF = 47.6 Hz), 4.60 (d, 1H, 3J = 8.0 Hz), 5.04 (dd, 1H), 5.14 (t, 1H), 5.23 (t, 1H). 13C NMR (CDCl3) δ 20.46, 61.79, 68.28, 68.74 (d, 2JCF = 21.4 Hz), 71.08, 71.87, 72.63, 82.45 (d, 1JCF = 169.4 Hz), 100.98, 169.47.
2′-Fluoroethyl-β-d-Glucoside (5a).
To a solution of 1.3 g (3.3 mmol) 2′-fluoroethyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (4a) in 25 mL methanol was added 5 mL of a solution of 100 mg sodium in 20 mL methanol and stirring was continued for 30 min. An amount of 2 g silica was added and the solvent was evaporated. The silica impregnated with the crude product was applied on a silica column that was eluted with acetonitrile, yielding 0.6 g (2.6 mmol, 80%) of a colorless oil. 1H NMR (CD3OD) δ 3.1–3.4 (m, 4H), 3.6–4.2 (m, 4H), 4.32 (d, 1H, 3J = 7.4 Hz), 4.59 (dt, 2H, 2JHF = 47.6 Hz). 13C NMR (CD3OD) δ 62.60, 68.13 (d, 2JCF = 18.3 Hz), 70.34, 73.65, 76.99, 77.17, 83.40 (d, 1JCF = 164.8 Hz), 103.21. Exact mass (ESI-MS) for C8H15FO6 [M+Na]+: found 249.0754, calculated 249.0750.
ω-Hydroxy-n-Butyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (2b).
An amount of 5 mL (56 mmol) 1,4-butanediol was added to 200 mL CH2Cl2, followed by a sufficient amount (10 mL) of dioxane to result in a homogeneous solution. Next, 1.8 g (8.3 mmol) yellow HgO, 0.14 g (0.39 mmol) HgBr2, and 6 g MgSO4 were added and, after stirring for 15 min, 4.2 g (10.2 mmol) 2,3,4,6-tetra-O-acetyl-α-d-glucopyranosyl bromide (1) was added in one portion. The flask was shielded from light and stirring was continued overnight. The suspension was filtered over celite, and the filtrate was concentrated in vacuo and applied on a silica column eluted with a gradient of ethyl acetate/hexane (1:2 v/v increased to 1:1 v/v). Yield was 3.7 g (8.7 mmol, 85%) of a colorless oil. 1H NMR (CDCl3) δ 1.65 (m, 4H), 2.00, 2.03, 2.05, 2.09 (4 × s, 12H), 3.55 (dt, 1H), 3.63 (t, 2H), 3.72 (m, 1H), 3.93 (dt, 1H), 4.14 (dd, 1H), 4.27 (dd, 1H), 4.53 (d, 1H, 3J = 7.8 Hz), 4.98 (dd, 1H), 5.08 (t, 1H), 5.22 (t, 1H). 13C NMR (CDCl3) δ 20.36, 25.65, 29.02, 61.76, 62.01, 68.29, 69.81, 71.17, 71.60, 72.66, 100.64, 169.44, 169.45, 170.32, 170.75.
ω-Tosyloxy-n-Butyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (3b).
To a solution of 3.2 g (8.1 mmol) ω-hydroxy-n-butyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (2b) and 5 mL (35 mmol) triethylamine in 100 mL CH2Cl2 at 0°C was added 3.1 g (16.3 mmol) p-toluenesulfonyl chloride. After stirring the solution overnight at room temperature, the solution was extracted twice with water. The solution was dried over MgSO4 and concentrated in vacuo, and the residual oil was applied on a silica column eluted with hexane/ethyl acetate (2:1 v/v). Yield was 1.6 g (2.9 mmol, 36%) of an oil that solidified on standing. 1H NMR (CDCl3) δ 1.65 (m, 4H), 2.00, 2.02, 2.05, 2.08 (4 × s, 12H), 2.46 (s, 3H), 3.48 (dt, 1H), 3.68 (m, 1H), 3.85 (dt, 1H), 4.03 (t, 2H), 4.12 (dd, 1H), 4.26 (dd, 1H), 4.46 (d, 1H, 3J = 7.6 Hz), 4.96 (dd, 1H), 5.02 (t, 1H), 5.19 (t, 1H), 7.36 (d, 2H), 7.79 (d, 2H). 13C NMR (CDCl3) δ 20.48, 21.52, 25.34, 61.86, 68.35, 68.86, 70.05, 71.23, 71.49, 72.78, 100.70, 127.89, 129.90, 132.70, 144.65, 169.53, 170.38
ω-Fluoro-n-Butyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (4b).
An amount of 6.7 g (16.4 mmol) ω-hydroxy-n-butyl 2,3,4,6-tetra-O-acetyl-β-d-glucoside (2b) was dissolved in 320 mL CH2Cl2 containing 2.6 mL (32 mmol) pyridine and cooled to 0°C. To this solution was added 2.3 mL (18.2 mmol) diethylaminosulfur trifluoride (DAST) dissolved in 20 mL CH2Cl2 and stirring was continued at the same temperature for 20 min. The solution was poured into 400 mL saturated NaHCO3 solution at 0°C, and the organic layer was separated and washed twice with 100 mL water. After drying over MgSO4, the solution was evaporated and the residue was applied on a silica column eluted with hexane/ethyl acetate (3/1 v/v), yielding 1.1 g (2.6 mmol, 16%) of a white solid. 1H NMR (CDCl3) δ 1.70 (m, 4H), 2.01, 2.02, 2.04, 2.09 (4 × s, 12H), 3.54 (dt, 1H), 3.70 (m, 1H), 3.92 (dt, 1H), 4.14 (dd, 1H), 4.25 (dd, 1H), 4.45 (dt, 2H, 2JHF = 48.0 Hz), 4.51 (d, 1H, 3J = 7.8 Hz), 4.99 (dd, 1H), 5.08 (t, 1H), 5.21 (t, 1H).
ω-Fluoro-n-Butyl-β-d-Glucoside (5b).
The same procedure for the hydrolysis was used as described for 2′-fluoroethyl-β-d-glucoside (5a). Yield was 0.5 g (1.4 mmol, 53%) of a colorless oil. 1H NMR (CD3OD) δ 1.6–1.9 (m, 4H), 3.1–3.4 (m, 4H), 3.5–3.7 (m, 2H), 3.8–4.0 (m, 2H), 4.27 (d, 1H, 3J = 7.6 Hz), 4.45 (dt, 2H, 2JHF = 48.0 Hz). 13C NMR (CD3OD) δ 26.57, 28.16, 62.72, 70.18, 71.58, 75.04, 77.83, 78.0, 84.78 (1JCF = 164.8 Hz), 104.30. Exact mass (ESI-MS) for C10H19FO6 [M+Na]+: found 277.1068, calculated 277.1063.
ω-Hydroxy-n-Octyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (2c).
The title compound was prepared from 1,8-octanediol as described for ω-hydroxy-n-butyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (2b) (addition of dioxane was not required). Yield was 3.9 g (8.6 mmol, 84%) of a crystallized oil. 1H NMR (CDCl3) δ 1.30 (m, 8H), 1.56 (m, 4H), 2.01, 2.03, 2.04, 2.09 (4 × s, 12H), 3.48 (dt, 1H), 3.60 (t, 2H), 3.70 (m, 1H), 3.87 (dt, 1H), 4.13 (dd, 1H), 4.27 (dd, 1H), 4.50 (d, 1H, 3J = 7.6 Hz), 4.97 (dd, 1H), 5.07 (t, 1H), 5.20 (t, 1H). 13C NMR (CDCl3) δ 20.27, 20.40, 25.40, 28.92, 29.05, 32.38, 61.76, 62.43, 68.26, 69.90, 71.14, 71.45, 72.66, 100.58, 169.23, 169.36, 170.20, 170.63.
ω-Tosyloxy-n-Octyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (3c).
The same procedure was used as described for ω-tosyloxy-n-butyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (3b). Yield was 0.93 g (1.5 mmol, 36%) of a colorless oil that solidified on standing. 1H NMR (CDCl3) δ 1.3 (m, 8H), 1.5 (m, 4H), 2.01, 2.03, 2.04, 2.09 (4 × s, 12H), 2.46 (s, 3H), 3.47 (dt, 1H), 3.70 (m, 1H), 3.86 (dt, 1H), 4.02 (t, 2H), 4.14 (dd, 1H), 4.28 (dd, 1H), 4.50 (d, 1H, 3J = 7.6 Hz), 4.98 (dd, 1H), 5.08 (t, 1H), 5.21 (t, 1H), 7.36 (d, 2H), 7.79 (d, 2H). 13C NMR (CDCl3) δ 20.42, 21.42–29.14, 61.85, 68.38, 70.02, 70.50, 71.23, 71.60, 72.75, 100.73, 127.80, 129.78, 132.70, 144.65, 169.30, 169.45, 170.30, 170.69.
ω-Fluoro-n-Octyl-2,3,4,6-Tetra-O-Acetyl-β-d-Glucoside (4c).
The title compound was prepared as described for ω-fluoro-n-butyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (4b). Yield was 1.4 g (3.1 mmol, 30%) of a white solid. 1H NMR (CDCl3) δ 1.3–1.8 (m, 12H), 2.01, 2.03, 2.04, 2.09 (4 × s, 12H), 3.49 (dt, 1H), 3.69 (m, 1H), 3.87 (dt, 1H), 4.13 (dd, 1H), 4.29 (dd, 1H), 4.44 (dt, 2H, 2JHF = 47.4 Hz), 4.50 (d, 1H, 3J = 7.6 Hz), 4.99 (dd, 1H), 5.08 (t, 1H), 5.21 (t, 1H). 13C NMR (CDCl3) δ 20.27, 24.80–32.32, 61.79, 68.29, 69.93, 71.17, 71.53, 72.66, 83.92 (1JCF = 164.8 Hz), 100.64, 169.17, 169.36, 170.20, 170.60.
ω-Fluoro-n-Octyl-β-d-Glucoside (5c).
The title compound was prepared as described for ω-fluoro-n-butyl-β-d-glucoside (5b). Yield was 0.5 g (1.6 mmol, 52%) of a white solid. 1H NMR (CD3OD) δ 1.3–1.8 (m, 12H), 3.1–3.4 (m, 4H), 3.53 (dt, 1H), 3.66 (dd, 1H), 3.86 (dd, 1H), 3.93 (t, 1H), 4.24 (d, 1H, 3J = 7.8 Hz), 4.40 (dt, 2H, 2JHF = 47.6 Hz). 13C NMR (CD3OD) δ 26.93, 30.27–31.70, 62.75, 70.85, 71.64, 75.10, 77.89, 78.13, 84.89 (1JCF = 163.2 Hz), 104.39. Exact mass (ESI-MS) for C14H27FO6 [M+H]+: found 311.1882, calculated 311.1869.
2′-18F-Fluoroethyl-β-d-Glucoside (5a).
After cyclotron irradiation, the content of the 18F-fluoride-target was passed over a disposable anion-exchange column (Accell plus QMA; Waters) to trap 18F-fluoride. The column was eluted with a solution of 18 mg Kryptofix 2.2.2. and 1.6 mg K2CO3 in 0.75 mL CH3CN/H2O (95:5 v/v) and the eluate was dried by 2 consecutive evaporations with 0.3 mL CH3CN. A solution of 5 mg 2′-tosyloxyethyl-2,3,4,6-tetra-O-acetyl-β-d-glucoside (3a) in 0.5 mL CH3CN was added to 5,000 MBq Kryptofix/K2CO3/18F-fluoride and the mixture was heated in a closed vial at 60°C during 20 min. The mixture was cooled, eluted over a silica Sep-Pak column (Waters) with 5 mL ether, and concentrated, yielding 630 MBq (17%, 40 min) of the crude tetraacetate. Preparative HPLC was performed on a μBondapak column (7.8 × 300 mm; Waters), eluted with CH3CN/H2O 45:55 (v/v) at 4 mL/min (retention time [tR] = 7 min 50 s). After dilution of the effluent to 10% CH3CN in H2O, the solution was applied on a C18 light Sep-Pak column (Waters). The tetraacetate was eluted from the Sep-Pak with 2 mL methanol and, after addition of 0.2 mL of a 0.22 mol/L solution of NaOCH3/CH3OH, the mixture was stirred at room temperature for 30 min. The solution was applied on a silica Sep-Pak followed by 3 mL CH3CN and the solvent was evaporated. The residue was dissolved in 0.9% NaCl solution, yielding 220 MBq (87%, 50 min) 2′-18F-fluoroethyl-β-d-glucoside (5a). On average, the total synthesis time was 90 min and the overall decay-corrected yield was 11% ± 3% (n = 4). The radiochemical purity of 5a was assessed by HPLC on a Carbohydrate Amino column and was consistently found to be >95% (tR = 6 min 15 s).
ω-18F-Fluoro-n-Butyl-β-d-Glucoside (5b).
The title compound was prepared and analyzed as described for 5a, except for the purification of the final product that was performed on a C18 column (Hypersil BDS C18 = 5 μm; Alltech) eluted with a mixture of CH3CN/H2O (5:95 v/v, 1 mL/min). Compound 5b eluted after 5 min 30 s; the radiochemical yield was 10% ± 5% (n = 5). The radiochemical purity of 5b was determined on a Carbohydrate Amino column and was consistently found to be >95% (tR = 4 min 20 s).
ω-18F-Fluoro-n-Octyl-β-d-Glucoside (5c).
The same procedure was applied as described for 5b. The final product was purified on a C18 column eluted with the following gradient (CH3CN/H2O v/v, 1 mL/min): 0–7 min 25:75; 10 min 80:20; 30 min 90:10. Compound 5c eluted after 12 min. The radiochemical yield was 9% ± 5% (n = 5). The radiochemical purity of 5c was determined on a Carbohydrate Amino column and was consistently found to be >95% (tR = 2 min 5 s).
Evaluation for Interaction with hSGLT1
Sense complementary RNA (cRNA) of hSGLT1 was synthesized as described previously (6). For expression, defolliculated Xenopus laevis oocytes were injected with 50 nL water containing 10 ng of hSGLT1 cRNA per oocyte. The oocytes were incubated for 3 d at 19°C in Ori buffer (5 mmol/L Hepes-Tris, pH 7.4, 100 mmol/L NaCl, 3 mmol/L KCl, 2 mmol/L CaCl2, and 1 mmol/L MgCl2) containing 50 mg/L gentamycin.
For tracer flux measurements, Xenopus oocytes were incubated for 1 h at room temperature with Ori buffer containing 10 μmol/L 14C-αMDG (11.25 GBq/mmol; Amersham Pharmacia Biotech) plus different concentrations of either 2′-fluoroethyl-β-d-glucoside (5a), ω-fluoro-n-butyl-β-d-glucoside (5b), or ω-fluoro-n-octyl-β-d-glucoside (5c). The incubation was performed in the absence or presence of 100 μmol/L phlorizin. Then the oocytes were washed and analyzed for radioactivity (6). The uptake of 14C-αMDG was calculated as the difference of its uptake in 7–10 oocytes in the absence and presence of phlorizin.
Electrical measurements of glycoside transport by hSGLT1 were performed using the 2-microelectrode voltage-clamp technique as described previously (6). Oocytes clamped at −50 mV were continuously superfused at room temperature (∼3 mL/min) with Ori buffer, and inward currents that were induced by addition of different concentrations of glycosides were measured. The measurements were performed with at least 4 different oocytes from 2 different batches of oocytes.
The maximal currents that are induced by αMDG and 5a were determined in 5 oocytes expressing hSGLT1 that were clamped at −50 mV. These oocytes were superfused respectively with Ori buffer, with Ori buffer containing 5 mmol/L αMDG, again with Ori buffer, and finally with Ori buffer containing 5 mmol/L 5a.
Biodistribution in Mice
Animal studies were performed according to the Belgian code of practice for the care and use of animals. The solution of the 18F-labeled sugars were diluted with normal saline to a concentration of 4 MBq/mL. Male NMRI mice (body mass, 25–30 g) were sedated by intramuscular injection of 0.05 mL of a 1:4 diluted solution of Hypnorm (Duphar). Then, 0.1 mL of the tracer solution was injected in each of the mice (n = 5 at each time point) via a tail vein. The mice were killed by decapitation at 10 or 30 min after injection. Five mice were pretreated with an intraperitoneal dose of 800 mg/kg body mass phlorizin (200 mg/mL solution in 1,3-propanediol/water 60:40 v/v) 2 h before injection with 5a and killed after 30 min. Blood was collected in tared tubes and weighed. All organs and other body parts were dissected and their 18F radioactivity was determined in a Wizard γ-counter (Wallac). Corrections were made for background radiation and physical decay during counting. For calculation of total blood activity, blood mass was assumed to be 7% of body weight.
In Vivo Metabolism
An amount of 2 MBq 5a, 5b, or 5c was injected in each of 3 mice via a tail vein. The animals were decapitated after 30 min and blood and urine were collected. The hematocrit value was determined in duplicate. The blood samples were weighed and their radioactivity was counted in a γ-sample changer. After centrifugation, plasma was collected, weighed, and diluted with 3 parts of acetonitrile. After filtration through a 0.22-μm filter, the sample was applied on HPLC and 1.5-mL fractions were collected and their radioactivity was counted in a γ-sample changer. The residual radioactivity in the plasma pellet and the filter was also determined. The urine samples were treated and analyzed on HPLC as described for the plasma samples. To compare the uptake of 5a, 5b, and 5c in red blood cells, an ANOVA with a Tukey honest significant difference post hoc test was used.
Autoradiography
To determine the in vivo intrarenal distribution, 0.1 mL of a saline solution containing 6 MBq 2′-18F-fluoroethyl-β-d-glucoside (5a) was injected intravenously in a sedated NMRI mouse. The animal was killed by decapitation at 30 min after injection and a kidney was rapidly removed and immediately frozen in 2-methylbutane (cooled to −25°C with liquid nitrogen). Serial frozen sections of 25 μm were cut and mounted on microscope slides. The sections were immediately air dried at 50°C and exposed overnight to a high-performance storage phosphor screen (Packard) that was scanned in a Phosphor Imager Scanner (Packard Cyclone). Later, the sections were fixed by 20 min incubation with 4% (w/v) paraformaldehyde dissolved in phosphate-buffered saline, stained with hematoxylin, and examined by light microscopy.
RESULTS
Chemistry
2-Bromoethylglucoside was prepared as described by Dahmén et al. (20) and converted to the corresponding tosylate 3a with silver tosylate in CH3CN. The precursors for the synthesis of 18F-fluorobutyl- (5b) and 18F-fluorooctylglucoside (5c) were prepared using Koenigs-Knorr reaction conditions followed by tosylation of the ω-alcohol function (overall yield, 30%; Fig. 1). The Koenigs-Knorr synthesis is applied widely for the synthesis of alkyl glycosides using silver or mercury salts as catalyst (21). We found the best results with mercury(II)bromide as catalyst (22). The synthesis of the butyl derivative (2b) required the addition of a small amount of 1,4-dioxane to the reaction mixture to prevent separation of the 2 phases and inactivation of the mercury catalyst. Under these conditions, the β-isomers were isolated in good yields. The fluoro-derivatives 5b and 5c were prepared by reaction of DAST on the corresponding alcohols. To prevent hydrolysis of the 1-acetal, 2 equivalents of pyridine were added to the reaction mixture. The labeling reactions with 18F were performed using standard conditions with Kryptofix/K2CO3/18F-fluoride in CH3CN at 60°C (Fig. 2). Hydrolysis of the protective acetate groups was performed with NaOMe/MeOH at 20°C. Hydrolysis of the 1-acetal was not found using the basic conditions, although it could be hydrolyzed under acidic conditions (1N HCl, 130°C, 10 min). The purity and identity of the final products were assessed on HPLC before injection in mice. In each case, the radiochemical purity was >95%. The specific activity at the end of synthesis was estimated to be >15 GBq/μmol. Nonradioactive impurities as determined with mass spectrometry were virtually absent (<10 nmol).
Uptake Measurements with hSGLT1 Expressed in Xenopus Oocytes
To characterize the interactions of 2′-fluoroethyl-β-d-glucoside (5a), ω-fluoro-n-butyl-β-d-glucoside (5b), and ω-fluoro-n-octyl-β-d-glucoside (5c) with SGLT1, the human transporter hSGLT1 was expressed in Xenopus oocytes. The inhibition of expressed 14C-αMDG uptake by these compounds and the currents that were induced by 5a, 5b, or 5c were determined using the 2-electrode voltage-clamp technique. The measurements were performed with hSGLT1 since species differences between SGLT transporters have been described and we wanted to make sure that the obtained data are relevant for the human. In Figure 3, we tested whether the phlorizin-inhibitable uptake of 14C-αMDG in hSGLT1-expressing oocytes was inhibited by different concentrations of 5a, 5b, or 5c. All 3 compounds were able to inhibit αMDG uptake by hSGLT1. Fitting the Michaelis-Menten equation to the data, Ki values of 1.51 ± 0.23 mmol/L (5a), 1.60 ± 0.20 mmol/L (5b), and 66 ± 2.9 μmol/L (5c) were found. In Figure 4, we determined whether 5a, 5b, and 5c were transported by hSGLT1. When Xenopus oocytes expressing hSGLT1 clamped at −50 mV were superfused with 5 mmol/L of αMDG, 5a, 5b, or 5c, inward currents were induced by αMDG (23) and 5a but not by 5b and 5c. This indicates that hSGLT1 mediates cotransport of Na+ and 5a (as has been described for d-glucose as well as αMDG), whereas 5b and 5c are nontransported inhibitors. To determine the maximal currents that are induced by αMDG and 5a, 5 oocytes expressing hSGLT1 that were clamped at −50 mV were examined. The inward currents induced by αMDG and 5a were 52.5 ± 3.0 nA and 45 ± 4.0 nA (P < 0.05 for difference), respectively. This indicates that the maximal transport rate of hSGLT1 for 5a is slightly lower than that of αMDG. In Figure 5, we measured the substrate dependence of inward currents in hSGLT1-expressing Xenopus oocytes clamped at −50 mV that could be induced by 5a. Fitting the Michaelis-Menten equation to the data, a Km value of 2.6 ± 0.2 mmol/L was obtained. This value is in the same range as the apparent Ki value determined for the inhibition of αMDG uptake by 5a.
In Vivo Evaluation
The results of the tissue distribution study in mice are shown in Table 1. All 3 18F-fluoroalkyl glucosides were cleared mainly by the kidneys. Interestingly, the excretion of 5b and 5c into the urine was more rapid compared with the shorter-chain 2′-18F-fluoroethyl-β-d-glucoside 5a.
To study the intrarenal distribution of 5a in more detail, slices of a kidney obtained at 30 min after injection of 5a were examined with autoradiography and light microscopy after fixation and staining. The resulting autoradiographic images (Fig. 6) revealed that 5a was mainly concentrated in the outer stripe of the outer medulla. This kidney region contains the S3 segments of the proximal tubules where SGLT1 is expressed (2).
Pretreatment of the mice with phlorizin 2 h before injection of 5a showed a marked increase of urinary excretion. Phlorizin has a >100-fold higher affinity compared with d-glucose (24) and will partially block SGLT1 transport resulting in glucosuria. The observed increased urinary excretion of 5a and the decreased retention in kidney (Table 1) indicate that tubular reabsorption of 5a is indeed achieved by SGLT1.
HPLC analysis of plasma and urine samples collected at 30 min after injection of 5a or 5b revealed that >90% of the radioactivity in plasma and urine was in the form of intact tracer (Table 2). After injection of ω-18F-fluoro-n-octyl-β-d-glucoside (5c), only 38% was recovered in plasma as unchanged tracer, and in urine 72% of the radioactivity was present as unchanged tracer. The remainder of the radioactivity was found to be in the form of a more polar metabolite.
At 30 min after injection, only 19% ± 4% (n = 4) of 5a in the blood was detected in red blood cells. This was significantly lower than the fractions of 5b (38% ± 6%, n = 2, P < 0.02) and 5c (47% ± 3%, n = 2, P < 0.002) after this time period. This may be due to a higher passive diffusion of 5b and 5c compared with 5a and suggests that 5a is not transported by GLUT1 that is expressed in erythrocytes (1).
DISCUSSION
The in vitro experiments in Xenopus oocytes showed that 5b and 5c were not transported by hSGLT1 but inhibited transport of αMDG. In contrast, 5a was transported almost as efficiently as αMDG by hSGLT1. The in vivo experiments in mice, however, showed that an appreciable fraction of the injected dose of 5a was excreted with the urine, whereas 11C-αMDG revealed only a negligible urinary excretion (23). These apparent discordant results between in vitro and in vivo results may be caused by a different transport rate of 5a by hSGLT1 versus mouse SGLT1 or by a different transport rate of 5a and 11C-αMDG by SGLT2. SGLT2 is a low-affinity, high-capacity transporter located in the S1 and S2 segments of the proximal tubules, whereas SGLT1 is a high-affinity, low-capacity transporter located in the S3 segment of the proximal tubules (2). The maximal transport rates of hSGLT1 for d-glucose, αMDG, and 5a are similar (see above) (23). In human and mouse, αMDG maybe a good substrate for both SGLT1 and SGLT2 and it may be removed almost completely from the glomerular filtrate during its passage through the S1 and S2 segments. If 5a has a lower affinity for SGLT2, it will pass through the S1 and S2 segments and then will be reabsorbed by SGLT1 in the S3 segment. The low capacity of the SGLT1 transporter may not be sufficient to clear 5a completely from the glomerular filtrate, resulting in a partial urinary excretion. Glucoside 5a may leave the proximal tubular cells only slowly, as it is probably a poor substrate for the GLUT2 transporter in the basolateral membrane. This would explain the transiently higher concentration in kidney at 30 min of 5a compared with 5b and 5c and the slower decrease of concentration of 5a in the blood.
The autoradiographic images of the intrarenal distribution (Figs. 6A and 6C) are consistent with these hypotheses. Glucoside 5a was accumulated mainly in the outer part of the outer medulla. This kidney region, termed the “outer stripe,” contains the S3 segments of renal proximal tubules where SGLT1 is expressed (2). The “inner stripe” contains no proximal tubules but pars rectae of distal tubules and vascular bundles with descending and ascending vasae rectae. At variance with the distribution of 5a, 11C-αMDG was mainly accumulated in the renal cortex that contains S1 and S2 segments of renal proximal tubules (23). No accumulation of 11C-αMDG was observed in the outer stripe of the outer medulla because most 11C-αMDG was cleared from the glomerular filtrate in the S1 and S2 segments (Fig. 6D).
The uptake of 5a in kidney was also evaluated after pretreatment with phlorizin. Phlorizin is an inhibitor of SGLT1 and SGLT2 and its application leads to glucosuria. In our experiment, the pretreatment resulted in a reduced retention of 5a in the kidneys and an increased excretion into the urine. The blood level was also markedly reduced. This is proof of the involvement of SGLT1 or SGLT2 in the accumulation of 5a in the kidneys. Taken together, the data suggest that the activity of SGLT1 in humans can be visualized by 5a, whereas the activity of both SGLT1 and SGLT2 can be visualized by 11C-methyl-α-d-glucoside (23).
During acute renal failure, the S3 segments of proximal tubules and the thick ascending limbs of Henle are primarily and most strongly damaged (25). Since compound 5a appears to be a selective marker for functional activity of SGLT1 and SGLT1 is expressed in the S3 segment, PET with compound 5a may be used to detect early renal defects during acute renal failure.
In brain, the concentrations of 5a and 5b were very low and only a small fraction of 5c was detected. The latter may be due to passive diffusion of the lipophilic compound through the plasma membranes of the BBB. The low uptake in brain suggests that 5a, 5b, and 5c are not transported by GLUT1 that is expressed in the BBB (15,16). Recently we were able to immunolocalize SGLT1 in luminal membranes of rat brain capillaries, suggesting that SGLT1 plays a role in the migration of d-glucose across the BBB (H. Koepsell, unpublished data, 2003). However, the concentration of SGLT1 in the brain is lower than that in the kidney and, until now, it is not clear whether SGLT1 expressed in brain is functional in normal physiologic conditions. This may explain the observed absence of accumulation of 5a in brain.
CONCLUSION
Three 18F-labeled ligands—that is, 2′-18F-fluoroethyl-β-d-glucoside (5a), ω-18F-fluoro-n-butyl-β-d-glucoside (5b), and ω-18F-fluoro-n-octyl-β-d-glucoside (5c) were prepared and evaluated in an in vitro and in vivo model.
In vitro, 5a was transported almost as efficiently as αMDG by SGLT1, whereas 5b and 5c were not transported but inhibited transport of αMDG. In vivo, 5a, 5b, and 5c were cleared mainly by the kidneys, and 5a was retained in the outer medulla of the kidney. The accumulation of 5a could be blocked by pretreatment with phlorizin. Brain uptake of 5a and 5b was low, whereas 5c showed limited brain uptake. The data indicate that 5a appears to be a valid tracer for the specific in vivo visualization of SGLT1 activity. The accumulation of 5a in the kidney may be used to visualize the outer stripe of the inner medulla with intact proximal tubules containing functional sodium/d-glucose cotransporter 1.
Acknowledgments
This research was supported by the National Fund for Scientific Research-Flanders, Belgium, and by the Deutsche Forschungsgemeinschaft (SFB 487/C1).
Footnotes
Received Apr. 4, 2003; revision accepted Aug. 28, 2003.
For correspondence or reprints contact: Tjibbe de Groot, PhD, Laboratory of Radiopharmaceutical Chemistry and Radiopharmacy, Universitaire Ziekenhuizen Gasthuisberg, Herestraat 49, B-3000 Leuven, Belgium.
E-mail: tjibbe.degroot{at}uz.kuleuven.ac.be