Biosensor-Based Determination of Riboflavin in Milk Samples

Anal. Chem. 2004, 76, 137-143

Biosensor-Based Determination of Riboflavin in Milk Samples Isabelle Caelen,*,† Andras Kalman,† and Lennart Wahlstro 1 m‡

Nestle´ Research Center Lausanne, Nestec Ltd., Vers-Chez-Les-Blanc, 1000 Lausanne 26, Switzerland, and Biacore AB, Rapsgatan 7, 754 50 Uppsala, Sweden

An assay for quantification of riboflavin (Rf) in milk-based products has been developed using the principle of surface plasmon resonance with on-chip measurement. The quantification was done indirectly by measuring excess of Rf binding protein (RBP) that remains free after complexation with Rf molecules originally present in the sample solution. The chip was modified with covalently immobilized Rf in order to bind the RBP in excess. A chemical modification was performed to introduce a reactive ester group at the N-3 position of the natural Rf to bind amino groups present on the chip surface. Calibration solutions were prepared by mixing a range of Rf standard solutions with an optimized concentration of RBP. The Rf content in the milk-based products was then measured by comparison of the response against the calibration. Results obtained were very close to those from an official HPLC-fluorescence procedure. The limit of quantification was determined to 234 µg/L and the limit of detection to 70 µg/L by an injection volume of 160 µL. Riboflavin (Rf) is an essential vitamin in human nutrition occurring in a wide variety of food products. In the body riboflavin is transformed into two active coenzymes, flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD). These two active flavins are directly involved as coenzymes with oxidases and dehydrogenases for the hydrolysis of fatty acids and the degradation of amino acids or pyruvic bases. Second, flavins can transfer electrons or protons from a donor to an acceptor, important in the Krebs cycle and the respiratory chain. Official methods for quantification of Rf in food samples are based usually on HPLC measurement and microbiological assays. Alternative methods involve electrochemical or fluorescent characteristics of flavins for standard solutions,1-3 biological fluids,4,5 and pharmaceutical tablets6 but none for food matrixes. In addition to these approaches, methods based on biological properties of Rf through * To whom correspondence should be addressed. E-mail: isabelle. [email protected]. † Nestec Ltd. ‡ Biacore AB. (1) Nishikimi, M.; Kyogoku, Y J. Biochem. 1973, 73, 1233-42. (2) Sawamoto, H. J. Electroanal. Chem. 1985, 186, 257-65. (3) Wang, J.; Luo, D.-B.; Farias, P. A. M.; Mahmoud, J. S. Anal. Chem. 1985, 57, 158-62. (4) Kozik, A. Analyst 1996, 121, 333-37. (5) Kodentsova, V. M.; Vrzhesinskaya, O. A.; Spirichev, V. B. Ann. Nutr. Metab. 1995, 39, 355-60. (6) Zhang, C.; Qi, H. Anal. Sci. 2002, 18, 819-22. 10.1021/ac034876a CCC: $27.50 Published on Web 11/14/2003

© 2004 American Chemical Society

the binding with Rf binding protein (RBP) from egg white7-9 have been investigated. Interest in developing analytical techniques based on selective biointeractions between biochemical entities increases day after day. Development of a biosensor-based approach, using a specific interaction between Rf and RBP10,11 in molecular ratio 1:1, would appear to be competitive with classical approaches in terms of speed of analysis and simplicity. In particular, the surface plasmon resonance (SPR) technique, which detects changes in mass onto the sensor chip surface provides quantitative, specific, and real-time analysis and allows kinetic and affinity analysis.12 In the case of low molecular weight compounds, such as vitamins, indirect assay is performed.13-17 The chip surface is first modified with Rf derivative covalently immobilized on the surface. Next, RBP is added in Rf containing samples where complexation of the proteins occurs. Finally, the sample is injected onto the surface and the free remaining RBP binds to the Rf derivative molecules on the chip surface. A number of conditions need to be met to obtain a successful SPR assay. First, the chemistry involved in linking ligand molecules to surfaces should fully preserve their biochemical properties. Second, immobilized ligands should be easily accessible to ensure optimum detection and successive quantification of target molecules. Third, the biochemical layer must be stable, providing identical binding properties after many steps of washing the surface and regeneration of the free ligand molecules. In this work, the development of a new Rf assay is described and evaluated for the quantification of Rf in milk-based products, taking into account the three principles mentioned above. This type of product was chosen as a test matrix since more than a quarter of the average Rf intake is from milk and milk products. (7) Becvar, J.; Palmer, G. J. Biol. Chem. 1982, 257, 5607-17. (8) Lotter, S. E.; Miller, M. S.; Bruch, R. C.; White, H. B., III. Anal. Biochem. 1982, 125, 110-7. (9) Ushijima, H.; Okamura, H.; Nishina, Y.; Shiga, K. J. Biochem. 1989, 105, 467-72. (10) Ogasawara, F. K.; Wang, Y.; Bobbitt, D. R. Anal. Chem. 1992, 64, 163742. (11) Yao, T.; Rechnitz, G. A. Anal. Chem. 1987, 59, 2115-8. (12) Lo ¨fas, S.; Johnsson, B. J. Chem. Soc., Chem. Commun. 1990, 1526-8. (13) Caselunghe, M. B.; Lindeberg, J. Food Chem. 2000, 70, 523-32. (14) Grace, T. A.; Stenberg, E. Cereal Foods World 2002, 47, 7. (15) Haines, J.; Aulenta, F.; Patel, P. Leatherhead: LFRA 2001, 1-13. (16) Indyk, H. E.; Evans, E. A.; Caselunghe, M. C. B.; Persson, B. S.; Finglas, P. M.; Woollard, D. C.; Filonzi, E. L. J. AOAC Int. 2000, 83, 1141-8. (17) Indyk, H. E.; Persson, B. S.; Malin, C. B.; Moberg, A.; Filonzi, E. L.; Woollard, D. C. J. AOAC Int. 2001, 85, 73-81.

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004 137

MATERIALS AND METHODS Materials. Riboflavin was purchased from Calbiochem (Darmstadt, Germany); riboflavin binding protein and retinol binding protein (ABP) were from Fluka (Buchs, Switzerland). All the organic solutions were purchased from Merck (Buchs, Switzerland) or Fluka. Methods. Riboflavin Derivative Synthesis. Protection of Hydroxyl Groups. A 5.0-g sample of riboflavin was added to a 400mL beaker containing 200 mL of a 1:1 (v/v) mixture of glacial acetic acid and acetic anhydride. After the dropwise addition of 1 mL of 70% perchloric acid, the mixture was stirred for 30 min at 40 °C.10 The mixture was cooled in an ice bath and diluted with an equal volume of cold water, and the solution was then extracted twice with 400 mL of chloroform. The combined chloroform extracts were washed twice with 200 mL of deionized water and the organic phase dried over anhydrous sodium sulfate.18 After filtration, the solvent was removed under reduced pressure at 50 °C to a yield of 95% with a melting point of 219-220 °C. Derivatization at the N-3 Position. A 1 mol equiv of 2′,3′,4′,5′tetra-O-acetylriboflavin was reacted with bromoacetic acid ethyl ester (5 mol equiv) with stirring for 20 h at room temperature in anhydrous N,N-dimethylformamide containing anhydrous potassium carbonate (5 mol equiv). Solvent was removed by evaporation under reduced pressure with warming, and the residue was redissolved in an equal volume of methylene chloride and washed twice with an equivalent volume of 1 N acetic acid and then twice with water.19 The organic phase was dried over anhydrous sodium sulfate, and the methylene chloride was removed overnight under reduced pressure at 36 °C to a yield of 73%. Deprotection of the Hydroxyl Groups. The residue was suspended in 2 N HCl (equal volume of previous step), refluxed for 2 h and evaporated to dryness.20 Formation of Activated Ester. A 0.50-mmol aliquot of N,N′dicyclohexylcarbidiimide (DCC) was added to a solution of 3-carboxymethylriboflavin (0.55 mmol) and N-hydroxysulfosuccinimide sodium salt (sulfo-NHS; 0.50 mmol) in 15 mL of anhydrous DMF at 0 °C, 48 h in the dark (to prevent photodegradation of the N-10 side chain of Rf) at 4 °C under stirring. Dicyclohexylurea formed during the reaction was removed by filtration. The activated product was isolated by evaporating of DMF at reduced pressure and by washing the resulting residue with ethanol. Any remaining DMF, DCC, or sulfo-NHS was removed by ethanol, a solvent in which the activated ester product is not soluble. The activated ester (powder form) prepared in this manner was stable for at least 6 months when stored below 0 °C.21 NMR Analysis. Each step of organic synthesis was followed by NMR analysis. The samples for NMR spectroscopy were prepared in Wilmad 528-PP 5-mm Pyrex NMR tubes, using heavy water or DMSO as solvent (0.7 mL). The NMR spectra were acquired on a Bruker AM-360 spectrometer, equipped with a quadrinuclear 5-mm probe head, at 360.13 MHz for 1H and at 90.03 MHz for 13C spectra. All shifts are cited in ppm measured relative (18) McCormick, D. B. J. Heterocycl. Chem. 1970, 7, 447-50. (19) Wu, F. Y-H.; MacKenzie, R. E.; McCormick, D. B. Biochemistry 1970, 9, 2219-24. (20) Merril, A. M.; McCormick, D. B. Methods. Enzymol. 1980, 66, 338-44. (21) Cha, G. S.; Meyerhoff, M. E. Anal. Biochem. 1988, 168, 216-27.

138

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

to solvent signal. One-dimensional 1H NMR, 13C NMR, and distortionless enhancement by polarization transfer (DEPT 135), were acquired using standard conditions.22 The spectra were interpreted using the MestRe-C 2.3 software. For the acetylation: 1H NMR (DMSO-d8) δ 1.93-2.42 (6 s, 18H, 4CH3CO, 2CH3), 4.11-4.17 (m, 1H, CH2), 4.27-4.30 (m, 1H, CH2), 4.52-4.55 (m, 1H, CH2), 4.71-4.75 (m, 1H, CH2), 5.13-5.38 (m, 3H, 3 CH), 7.60 (s, 1H, aromatic), 7.74 (s, 1H, aromatic), 11.34 (s, 1H, NH); 13C NMR (DMSO-d8) δ 19.6-21.4 (6CH3, 4CH3CO, 2CH3), 44.5 (NCH2), 62.3 (1CH2, OCH2), 69.6 (CH, OCH), 70.5 (CH, OCH), 73.4 (CH, OCH), 117.1 (CH, aromatic), 131.9 (CH, aromatic), 134.3-137.5 (4Cq, aromatic), 147.2 (1Cq, CdN), 151.3 (1Cq, CdN), 156.2 (1Cq, NCdO), 160.6 (1Cq, NCdO), 170.3171.0 (4 Cq, CdO). For the derivatization on N-3: 1H NMR (DMSO-d8) δ 1.1 (q, 3H, CH3), 1.93-2.46 (6 s, 18 H, 4CH3CO, 2CH3), 3.26 (s, 2H, CH2), 4.03-4.30 (m, 2H, CH2), 4.55-4.75 (m, 2H, CH2), 5.13-5.38 (m, 3H, 3 CH), 7.72 (s, 1H, aromatic), 7.78 (s, 1H, aromatic); 13C NMR (DMSO-d8) δ 14.2 (CH3, 19.5-21.6), (6 CH3, 4CH3CO, 2CH3), 43.2 (NCH2), 44.9 (NCH2), 61.4 (1CH2, OCH2), 63.3 (1CH2, OCH2), 69.5 (CH, OCH), 70.5 (CH, OCH), 73.4 (CH, OCH), 117.1 (CH, aromatic), 131.9 (CH, aromatic), 135.1-137.5 (4Cq, aromatic), 47.2 (1Cq, CdN), 151.3 (1Cq, CdN), 154.8 (1Cq, NCdO), 159.9 (1Cq, NCdO), 168.1 (Cq, Cdo), 170.3-171.0 (4Cq, CdO). For the deprotection: 1H NMR (D2O) δ 2.21 (s, 3H, CH3), 2.39 (s, 3H, CH3), 3.66-3.74 (m, 1H, CH2), 3.82-3.95 (m, 1H, CH2), 4.11-4.37 (m, 4H, 2 CH2), 4.46-4.63 (m, 3H, CH), 7.38 (s, 1H, aromatic), 7.65 (s, 1H, aromatic); 13C NMR (D2O) δ 19.6 (1CH3), 21.4 (1CH3), 43.2 (NCH2), 44.2 (NCH2), 60.1 (1CH2, OCH2), 69.6 (CH, OCH), 72.8 (CH, OCH), 73.7 (CH, OCH), 125.2 (CH, aromatic), 139.7 (CH, aromatic), 130.7-135.0 (4 Cq, aromatic), 148.9 (1Cq, CdN), 151.3 (1Cq, CdN), 156.6 (1Cq, NCdO), 160.8 (1Cq, NCdO), 174.8 (1Cq, CdO). Fluorescence Measurements. A stock solution of 1 M phosphate buffer (pH 7.4) was prepared monthly, and dilutions to 0.1 and 0.01 M were used. A 2.0-mg sample of Rf derivative (13.7 µM) was reacted under agitation for 1 h at room temperature in 230.5 mL of 0.1 M phosphate buffer with 3 mol equiv ethanolamine. This solution of blocked Rf was diluted 4 times (concentration of 3.42 µM) in 0.1 M phosphate buffer. A 95-µL sample of Rf and 5 µL of different concentrations of RBP or ABP were introduced in the wells of a microtiter plate. The analysis was performed with a SpectraFluor Plus (Tecan, Hombrechtikon, Switzerland) for a fluorescent light setup at 360 nm and an emission light read at 535 nm. Preparation of the Chip Surface and SPR Analysis. Surface immobilization and analysis were performed simultaneously using a Biacore Q system (Biacore AB, Uppsala, Sweden) using a carboxymethylated dextran sensor chip (CM5, Biacore AB). HBS-EP (Biacore AB) buffer was used as running buffer for all immobilizations. The measurements are real-time analysis giving curves named “sensorgrams”. These sensorgrams are a plot of response (arbitrary units) as a function of time (seconds). One arbitrary unit (au) increase corresponds to 1 pg/mm2 increase of surface density. (22) Lin, J.; Welti, D. H.; Arce Vera, F.; Fay, L. B.; Blank, I. J. Agric. Food Chem. 1999, 47, 2813-21.

Derivatization of Dextran End Groups. The gold surface of the CM5 chip with carboxyl-end molecules of dextran was washed with deionized water. Solutions of N-(3-dimethylaminopropyl)-N′ethylcarbodiimide/N-hydroxysuccinimide ester (EDC/NHS) (0.4 M/0.1 M in water) from the amine coupling kit (Biacore AB) were injected on the surface (flow rate 10 µL/min, contact time 7 min). Dihydrochloride ethylenediamine solution (0.1 M in 50 mM borate solution) was injected on the surface (flow rate 10 µL/min, contact time 7 min). Remaining ester was deactivated with 1 M ethanolamine in water (pH 8.5) (flow rate 19 µL/min, contact time 3 min). Washing of the surface was done with 1 M NaCl (flow rate 20 µL/min, contact time 4 min) and with water (flow rate 20 µL/min, contact time 4 min). Immobilization of Riboflavin Derivative. Buffer (0.1 M bis-tris propane in 0.01 M phosphate buffer, pH 6.5) was prepared for the immobilization of riboflavin. A 5 mM concentration of Rf derivative in cold immobilization buffer was injected on the surface (flow rate 10 µL/min) for 7 min. The surface was washed with 1 M ethanolamine hydrochloride (pH 8.5) (Biacore AB) at a flow rate of 19 µL/min for 7 min to remove any unbound ester of Rf derivative. RBP Binding. Injection of sample solutions (pure RBP solutions or mixed solutions containing riboflavin and a known amount of RBP) was done for 8 min at a flow rate of 20 µL/min. The solution was mixed automatically in the Biacore Q instrument, using 75% of RBP solution and 25% of sample solutions. Solutions of RBP were prepared daily in 0.1 M phosphate buffer containing 0.6 M NaCl and 0.1% surfactant P20 (Biacore AB). For the surface regeneration, 0.033% SDS23 (pH 5) was used for 50 s at a flow rate of 40 µL/min. Sample Preparation. The sample preparation procedure followed the official French method that describes the determination of vitamins B1 and B2 by reversed-phase high-pressure liquid chromatography (RP-HPLC). This method applies an enzymatic hydrolysis followed by RP-HPLC with fluorescence detection, and it is validated for fortified food products such as infant formulas, infant cereals, milk-based dietetic products, breakfast cereals, and fortified drinks. Based on the French HPLC method, a CEN method is now under approval.24 Our sample preparation procedure was made similarly according to the treatment with Taka-Diastase enzyme (Fluka) at 45-50 °C to break up compounds with high molecular masses. This was followed by acidic hydrolysis for 30 min in boiling water bath with 1 M HCl solution. After cooling the solution to room temperature, it was filtered through a 0.22-µm filter so that the solution was ready for the assay. RESULTS AND DISCUSSION Synthesis of Riboflavin Derivative and Test for Chemical Functionality. Natural riboflavin does not contain chemical groups that are able to bind to the chip surface, and therefore, organic synthesis is performed to introduce reactive ester groups at the N-3 position of the molecule.25 A summary of the four steps synthesis of the Rf derivative is shown in Figure 1. Several groups on the Rf nucleus (compound 1) contribute in a nonselective (23) Murthy, U. S.; Podder, S. K.; Adiga, P. R. Biochim. Biophys. Acta 1976, 434, 69-81. (24) CEN/TC 275 Work programme, pr EN 14 152 (under approval). (25) Merril, A. M.; McCormick, D. B. Anal. Biochem. 1978, 89, 87-102.

manner to the complexation with RBP molecules,7 and hence, they need to be preserved. First, some hydrophobic interactions occur via the two methyl groups at the C-7 and C-8 positions of the Rf molecule. This dimethylbenzenoid portion of the ring is involved in primary interactions of binding and is relatively buried in the interacting RBP, and both methyl groups seem to facilitate a fit with the binding site that maintains the proper position of the flavin ring system. Second, steric limitation is imposed at C-8, N-10, especially C-1′, 2′-hydroxyl group of the ribytil chain, and 4-carbonyl positions.26 Last, the hydroxyl groups of the N-10 side chain contribute in a stereoselective manner by formation of hydrogen bonds. However, substituents at C-2 and N-3 do not compromise binding affinity between the partners, and thus, derivatization can occur at these positions.27 In the case of RBP, a tryptophanyl residue is essential for the binding with Rf, and in addition, a tyrosine group is probably involved in a stack interaction with flavin molecules.28,29 As a consequence, a functional reactive group at the N-3 position was chosen that could bind surface molecules. Briefly, acetylation of hydroxyl groups was performed (compound 2) followed by nucleophilic substitution of the nitrogen atom on a bromoacetic acid ethyl ester (compound 3). The carboxylic acid (compound 4) obtained by hydrolysis of this ester is then converted back into a highly reactive ester (compound 5). Each step of the synthesis was confirmed by NMR analysis. This derivative can subsequently react by forming an amide linkage with free amino groups on the surface. Natural fluorescence of the Rf molecule is quenched by the tryptophanyl residue of RBP during complexation.30 Hence, bioreaction between Rf derivative and RBP can be evaluated by measuring quenching of Rf fluorescence when mixed in solution. Reactive ester groups on the derivative molecules that bind amino groups were inactivated with ethanolamine prior to the addition of RBP to prevent any covalent binding between the two biomolecules. Covalent binding itself could be responsible for a decrease of emitted fluorescence, falsifying the measurement of quenching due to the biointeractions. Fluorescence signals of the native Rf or the Rf derivative were followed by homogeneous assays as shown in Figure 2. Solutions of equal concentrations of Rf and its derivative were introduced separately into the wells of a microtiter plate, with increasing amount of RBP molecules. An increase in the concentration of RBP leads to an increase of fluorescence quenching for both Rf and Rf derivative. Negative tests demonstrate that binding does not occur between the two kinds of Rf molecules and ABP. It was observed that progressively increasing the concentration of ABP in the wells does not lower the level of fluorescence. As a consequence, quenching of fluorescence observed with RBP is certainly due to the biointeraction between RBP and riboflavins. The curves corresponding to Rf and Rf derivative are superposable up to a concentration of RBP of ∼1.5 µM. The fluorescence quenching of Rf and Rf derivative differs at higher concentration of RBP, which could be due to impurities of the solution of Rf derivative. Immobilization Chemistry of Riboflavin Derivative. This step illustrates the covalent immobilization of riboflavin derivative (26) Mifflin, T. E.; Langerman, N. Arch. Biochem. Biophys. 1983, 224, 319-25. (27) Choi, J. D.; McCormick, D. B. Arch. Biochem. Biophys. 1980, 204, 41-51. (28) Monaco, H. L. EMBO J. 1997, 16, 1475-83. (29) Blankenhorn, G. Eur. J. Biochem. 1978, 82, 155-60. (30) Guevara, I.; Zak, Z. J. Protein Chem. 1993, 12, 179-85.

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

139

Figure 1. Synthesis of riboflavin derivative.

Figure 2. Fluorescence quenching of (a) Rf and (b) Rf derivative in homogeneous assay, validating the biological properties of the molecules. Percentage of fluorescence quenching for riboflavin and its derivative is shown as a function of RBP concentration in solution. Molecular ratio between Rf or Rf derivative and RBP is indicated on the top of the curves. The lower curves represent reaction between ABP with (a′) Rf and (b′) Rf derivative.

onto the chip surface on which a dextran polymer bearing carboxyl groups is bound, Figure 3A. Injection of a solution of ethylenediamine (EDA) was performed to transform carboxyl end groups into amino groups via the formation of an amide linkage between the molecules. An excess of EDA prevented the molecules from reacting at both amino ends, leading to an unreactive surface. An increased coupling yield between EDA and carboxyl groups can be obtained by using NHS in conjunction with carbodiimide: O-acylisourea intermediate formed between EDC and carboxyl groups reacts with NHS to produce an active ester with an 140

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

enhanced stability, increasing the time for primary amino groups to react with the ester. In the third step, Rf derivatives were dissolved in cold bis-tris propane buffer to reduce the rate of hydrolysis of the activated ester prior to the reaction with amino groups on the surface, and the solution is injected onto the amine surface. The immobilization procedure was followed after each step by SPR measurement, Figure 3B, showing an increase of the response (value related to the refractive index) from step 1 to step 3. This trend is explained by the variation of surfaceassociated mass, due to successive immobilizations of EDA and Rf derivative molecules: surface density is around 190 pg/mm2 for the amino surface and 350 pg/mm2 for the surface-immobilized Rf. Finally, the surface was intensively washed while keeping the same response, meaning that Rf derivatives were covalently bound on the surface. Detection and Binding of RBP. Injecting different concentrations of the RBP onto the chip surface, Figure 4, was performed to test the ability of the immobilized Rf on the surface to be complexed by RBP. The response is followed in situ by SPR measurement as a function of time after injection of RBP solution, Figure 4A. Surface-associated mass increased with increasing concentration of RBP, for a range of concentrations between 0.1 and 10 µg/mL. Figure 4B shows the final response level, after rinsing of the surface, as a function of RBP concentration. The average values from three different experiments exhibit an exponential relationship when plotted on a logarithmic scale, up to a concentration of RBP of 10 µg/mL. The inset curve presents

Figure 3. Riboflavin derivative molecules covalently bound on the chip surface. (A) Carboxyl-end surface is transformed into amino-end surface after reaction with ethylenediamine. Riboflavin derivative can subsequently react by forming an amide linkage with primary amino groups on the surface. (B) Response associated with the measurement of refractive index at the surface is followed after each step of immobilization, demonstrating an increase of the mass present at the surface from step 1 to step 3.

Figure 5. Response obtained after complexation of RBP (open symbols) is compared to response obtained by direct assay (filled symbols) for two different concentrations of RBP: (A) 500 ng/mL (4, 2) and (B) 10 µg/mL (O, b).

Figure 4. Direct binding assay of RBP to immobilized Rf derivative performed using different concentrations of binding protein. (A) The response is followed in situ after injection of RBP with concentrations of (a) 0.1, (b) 0.2, (c) 0.4, (d) 0.8, (e) 1, (f) 2.5, (g) 5, and (h) 10 µg/mL. (B) Responses obtained after intensive washing of the surface are plotted as a function of RBP concentration. The inset drawing corresponds to linear range of the response up to a concentration of 1000 ng/mL.

linear response as function of RBP concentration up to a concentration level of 1 µg/mL. These results demonstrate, first, that the RBP binds efficiently to the surface-immobilized Rf and, second, that the quantity of Rf derivatives on the surface is adequate to discriminate between different concentrations of RBP molecules from the solution. A regeneration step using acidic

conditions completely releases the RBP molecules from the chip surface, and successive cycles of binding and regeneration may be performed. Assay for Quantification of Riboflavin in Food Samples. The next set of experiments characterizes the complexation reaction of the RBP when mixed with samples containing riboflavin molecules. A comparison was made between injections of RBP molecules directly onto the surface with residual RBP molecules after complexation with Rf molecules, Figure 5. Concentrations of free RBP at 500 ng/mL and 10 µg/mL were chosen, to compare values at low and high response according to the results from Figure 4B. In the case of complexation reaction, an initial concentration of Rf standard solution at 500 ng/mL was mixed in appropriate ratio with a total concentration of the solution of RBP at 20 and 10.5 µg/mL. As a result, the responses obtained for the two experiments are very similar for both low and high free RBP concentrations. The final response value after 8 min for the Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

141

Figure 6. Complexation assay for RBP to quantify Rf in food samples. (A) Calibration curve is obtained for a panel of Rf concentrations obtained by applying standard extraction procedures (HPLC-based) from food matrixes. (B) Standard solutions of Rf were injected at low concentration range from 3 to 400 ng/mL on the surface to determine limits of detection (LOD) and quantification (LOQ).

complexation assay was 10% less than the response obtained with RBP directly injected on the surface. First, this test shows that RBP was efficiently complexed with Rf molecules and, second, that the binding of RBP on the surface is not altered in the presence of these bound molecules in the solution. Figure 6A shows a plot of response as a function of Rf concentration between 200 and 1000 ng/mL, with a total concentration of RBP at 19.3 µg/mL. This range corresponds to free concentrations of RBP from 15.3 to 0 µg/mL. Calibration curves have a sigmoidal shape, and consequently, quantification over a quite large concentration range is possible. The coefficient of variation between two different calibration solutions at the same concentration is below 5%. To minimize the possible matrix effect of the sample on the RBP binding on the chip surface, we chose to dilute three times the sample solution in the RBP solution prior the injection. The experimental responses associated with this range of concentration of Rf are lower than those calculated, based on the results of Figure 4B, except for the value at 1000 ng/mL which should be ∼0. Indeed, binding of RBP on the surface occurs for a Rf concentration of 1000 ng/mL (response ∼100), showing that complexation of RBP is not fully completed. Ideally, the response obtained for the Rf-containing samples should be in the linear region of the curve in order to reach a maximal sensitivity and good separation of values for similar concentrations of Rf. In Figure 6B, limits of detection and quantification were determined. Using the same experimental conditions, solutions with different low concentrations of Rf, from 3 to 400 ng/mL, were injected on the surface with a concentration of RBP of 19.3 ng/mL and an injection volume of 160 µL 1 day after the previous measurement. The values obtained for 200 and 400 ng/mL RBP were higher (∼100 au) than those obtained previously because the initial baseline measured on the chip had an increase of ∼100 au. The matrix on the sensor chip consists of linear dextran, which is swollen in aqueous media, providing an extensively solvated hydrogel. When stocked overnight, the level of hydration of the chip surface changes, and hence the initial baseline can vary, affecting the absolute measurements but not the relative ones. As a consequence, it is advised to run the sample measurements the same day as the calibration curve. Injection of five blank samples was performed, and the corresponding value of standard deviation was 5%. Under those experimental conditions, the limit of detection (LOD) was 70 ng/mL and the limit of quantification (LOQ) was 234 ng/mL for Rf solution. This calibration was applied to the determination of Rf in 142

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

Table 1. Comparison of the Results of SPR Measurement with HPLC Reference Values riboflavin concn in mg/100 g measured by

milk-based product a milk-based product b

SPR

HPLC

1.67 ( 0.09 1.42 ( 0.02

1.69 ( 0.17 1.38 ( 0.14

two milk samples (a, b), and the results were in good agreement with those given by official method as indicated in Table 1. Rf was extracted from the milk-based products according to the official procedure described in Materials and Methods. Sample extracts were diluted to obtain a Rf concentration between 500 and 1000 ng/mL. The samples gave 238 and 365 au in response, respectively (duplicate analysis for each sample), in the region of the calibration curve where it is possible to distinguish close values (see Figure 6A). CONCLUSIONS This work describes the successful development of a SPR assay to quantify riboflavin in milk samples based on complexation of Rf binding protein added to the sample solution. The excess of RBP binds to the ligand molecule, a derivative of Rf that was synthesized and subsequently immobilized on the chip surface. Even though the initial results obtained by SPR are in good agreement with HPLC data, further work was required to optimize the concentration of added RBP in conjunction with a suitable range for Rf calibration in order to decrease the LOD and the LOQ. Another route for the synthesis of Rf derivative was used to increase the purity of the compound and hence the binding capacity of the surface. Under those conditions, the LOD decreased down to 17.1 ng/mL and LOQ down to 42.9 ng/mL. The repeatability of the assay is under 3% and the reproductibility under 4.5%.31 Once the derivative is immobilized on the surface, at least 400 assays can be performed on the same chip, which has a lifetime of 3 months when stored under 4 °C. Conventional assays include microbiological assays (MBA) and HPLC. MBA is limited by poor reproductibility, long sample preparation, and analysis time of 2-3 days. The HPLC method requires lengthly sample preparation and a sample turnover (31) Kalman, A.; O’Kane, A.; Caelen, I.; Trisconi, M.-J.; Wahlstro¨m; L., submitted to New-Food.

limited to 24 samples per 24 h. Biosensor measurements based on SPR principles present a number of advantages: automatic mixing and injection of the solutions make the system friendly to use, and the final concentrations of Rf in the samples are directly calculated by the software, giving fast and reliable measurements. Downscale measurement on chip surfaces economize reagents, shorten analysis time, and offer multidetermination of analytes in a single experiment. Using the optimized procedure described above, 20 samples can be fully analyzed in less than 6 h and up to 88 samples can be measured in a single run. Biosensor-based assays propose an alternative method to traditional analytical techniques.

ACKNOWLEDGMENT We thank Dr. J. Svorc for useful discussions and interest in the work and Dr. O. Heudi for his contribution to the writing of the article. We express our thanks to Mrs. M.-J. Trisconi for her help in the sample preparation. We warmly thank Dr. F. Robert and Mr. W. Matthey-Doret for their precious advise in the organic synthesis. We are grateful to Dr. J.C. Blake for carefully reading the manuscript and for his continuous support. Received for review July 30, 2003. Accepted October 14, 2003. AC034876A

Analytical Chemistry, Vol. 76, No. 1, January 1, 2004

143