General Contribution of Nonspecific Interactions to Fluorescence

Anal. Chem. 2006, 78, 3699-3705

General Contribution of Nonspecific Interactions to Fluorescence Intensity Eva M. Ga´lvez,† Muriel Matt,*,‡ Vicente L. Cebolla,*,† Francisco Fernandes,‡ Luis Membrado,† Fernando P. Cossı´o,§ Rosa Garriga,| Jesu´s Vela,| and M. Hassan Guermouche⊥

Instituto de Carboquı´mica, CSIC, P.O. Box 549, 50080 Zaragoza, Spain, Laboratoire de Chimie Applique´ e, Faculte´ des Sciences, Universite´ de Metz, 1 Boulevard Arago, Metz Technopoˆ le, 57078 Metz, France, Kimika Fakultatea, Euskal Herriko Unibertsitatea, P. K. 1072, 20080 San Sebastia´ n-Donostia, Spain, Departamentos de Quı´mica Orga´ nica y Quı´mica Fı´sica, y Quı´mica Analı´tica, Universidad de Zaragoza, 50009 Zaragoza, Spain, and Institut de Chimie, USTHB, BP32, Bab Ez zouar, Alger, Algerie

Many chemical compounds, including nonfluorescent ones, induce changes in the fluorescence spectra of certain probes, such as berberine cation and Reichardt’s betaine, both in the absence and the presence of solvent, that affect almost exclusively emission intensity. In this work, the application of fluorescence detection by intensity changes (FDIC) to HPLC and TLC chromatographic systems with fluorescence detectors has been studied. FDIC detection is of special interest in detecting nonfluorescent analytes, either in HPLC or in TLC mode. It does not involve covalent interactions, and the dielectric permittivity (E) of the medium plays an important role. The balance between nonspecific and specific interactions produces either an increase or a decrease in fluorescence intensity. Therefore, the influence of chromatographic conditions and chemical structure of analytes on the sign and magnitude of fluorescence peaks for sample detection in HPLC and TLC systems has been discussed. In general, probe nature and concentration determine response and detection sensitivity for a given sample in TLC and HPLC. As solubility and fluorescence properties in solvents determine the operating conditions for a FDIC probe in HPLC mode, nature and flows of mobile phase and solvent are important for chromatographic response and detection sensitivity. Many chemical compounds do not have the structural and electronic requirements to be detected by fluorescence.1 Derivatization is the most common way to make them detectable, and it usually implies changes in the molecular structure of analytes and involves covalent bond formation.2 * To whom correspondence should be addressed. E-mail: vcebolla@ carbon.icb.csic.es; [email protected]. † Instituto de Carboquı´mica. ‡ Universite´ de Metz. § Euskal Herriko Unibertsitatea. | Universidad de Zaragoza. ⊥ Institut de Chimie. (1) Schulman, S. G.; Di, Q. Q.; Juchum, J. Organic Chemistry Applications of Fluorescence Spectroscopy. In Encyclopedia of Spectroscopy and Spectrometry; Lindon, J. C., Tranter, G. E., Holmes, J. L., Eds.; Academic Press: London, 2000; Vol. III, pp 1718-1725. 10.1021/ac058045b CCC: $33.50 Published on Web 04/28/2006

© 2006 American Chemical Society

When HPLC-based chromatographic separations are involved, a reaction with a derivatizing agent, either nonfluorescent or a fluorophore, yields a chemical entity that shows fluorescent properties. In the case of thin-layer chromatography (TLC), these compounds are usually termed revealing agents. Although a high number of these agents has been described, they are usually efficient only for a particular class of analytes. Therefore, the development of procedures for fluorescence detection with a wide range of application in HPLC, TLC, or both systems is of particular interest. In previous papers,3,4 we reported that berberine cation, an aromatic heterocyclic alkaloid, behaves as a fluorescent molecular sensor,5 which allows saturated hydrocarbons and other nonfluorescent, low-polarity molecules to be detected with high sensitivity. Fluorescence changes involving saturated hydrocarbons was an unexpected and interesting phenomenon. We have now found that this phenomenon is not restricted only to berberine cation and saturated hydrocarbons. Significant, consistent, and structure-related changes in fluorescence intensity have been found for many kinds of compounds on berberine and other probes. All these phenomena can be observed both in solution and in the absence of solvent. Therefore, many chemical compounds induce an enhancement or quenching in the intensity of the fluorescence spectrum of berberine cation and that of other fluorophores. This seems a quite general phenomenon.6-12 (2) Toyo’oka, T. Modern Derivatization Methods for Separation Sciences; Wiley: Chichester, UK, 1999. (3) Cossio, F. P.; Arrieta, A.; Cebolla, V. L.; Membrado, L.; Vela, J.; Garriga, R.; Domingo, M. P. Org. Lett. 2000, 2, 2311-2313. (4) Cossio, F. P.; Arrieta, A.; Cebolla, V. L.; Membrado, L.; Domingo, M. P.; Henrion, P.; J. Vela. Anal. Chem. 2000, 72, 1759-1766. (5) Definition of fluorescent molecular sensor according to B. Valeur: Molecular Fluorescence: Principles and Applications; Wiley-VCH: Weinheim, 2002; p 274. (6) Pang, X.; Letourneau, R.; Rozniecki, J. J.; Wang, L.; Theoharides, T. CC. Neuroscience 1996, 73, 889-902. (7) Mikes, V.; Dadak, V. Biochim. Biophys. Acta 1983, 723, 231-239. (8) Kovar, J.; Skursky, L. Eur. J. Biochem. 1973, 40, 233-244. (9) Tan, Y.; Xie, J. Zhongyao Zazhi 1996, 21, 175-177. (10) Ulrichova, J.; Kovar, J.; Simanek, V. Collect. Czech. Chem. Commun. 1985, 50, 978-983. (11) Mikes, V.; Kovar, J. Biochim. Biophys. Acta 1981, 640, 341-351.

Analytical Chemistry, Vol. 78, No. 11, June 1, 2006 3699

The general character of these changes in fluorescent emission is of analytical interest. Many compounds can be quantitatively detected by the variation they produce on the emission intensity of the probe. Moreover, fluorescence detection by intensity changes (FDIC) may be adapted to high-performance liquid chromatography (HPLC) and planar chromatographic systems (TLC/HPTLC) given that this phenomenon occurs both in the absence and the presence of solvent. The described procedures are based on noncovalent interactions. It should be recalled that this phenomenon was first identified in the case of saturated hydrocarbons when it was demonstrated that they produced an enhancement in berberine emission through a weak electrostatic interaction.3,4 In this work, we study the viability of using this phenomenon in conventional column HPLC and planar TLC systems for detection of analytes by fluorescence. Incorporation of probes, as well as the variables involved in detection, has been studied. Examples of detection of nonfluorescent analytes are reported in both normal-phase and reversed-phase conditions. Procedures have been tested using pure standards, standard mixtures, and real samples. EXPERIMENTAL SECTION Probes and Samples. FDIC Probes. Berberine sulfate (95+%, from Acros Chimica, Geel, Belgium) and Reichardt’s betaine (RD) (C41H29NO2‚2H2O, 90+%, from Sigma, Madrid, Spain) were used. Chemical structures as well as information of standards used as pure analytes were reported elsewhere.4,13 Lipid Mixtures. Two mixtures with compounds from SigmaAldrich (Madrid, Spain) were used. The first one consisted of the following: cholesterol, cholesteryl oleate (C18:1, cis-9), oleic acid (C18:1, cis-9), oleic acid methyl ester (C18:1, cis-9), and triolein (C18:1, cis-9). The second one consisted of L-R-phosphatidylcholine, L-R-phosphatidylethanolamine, ammonium salt of L-Rphosphatidylinositol, and L-R-lisophosphatidylcholine. Gas Oil Sample. A dye-free, olefin-free, straight-run gas oil whose properties were reported elsewhere14 was used. Planar Chromatography Experiments. Either conventional silica gel TLC plates (aluminum sheets, 20 × 20 cm; 5-25-µm particle size; 60-Å pore size; 0.2-mm-thick layer) or highperformance silica gel TLC plates (HPTLC plates, on glass, 10 × 10 cm; 3-10-µm particle size; 60-Å pore size; 0.2-mm-thick layer) from Merck (Darmstadt, Germany), and Macherey-Nagel (Du¨ren, Germany) were used. Plates subjected to chromatographic development were preor postimpregnated with the corresponding FDIC probe. In the case of berberine, a solution of berberine sulfate in methanol (MeOH) (standard conditions: 2-6 mg‚100 mL-1) was used for impregnation during 20 s. In the case of RD, the concentration in MeOH ranged from 12 to 50 mg‚100 mL-1. Plates were subsequently dried overnight at 40 °C. (12) Lee, S. H. Chemical compositions, their use as cytochemical agents and methods for the detection of steroid hormone receptors in human tissue. Eur. Pat. Appl., 1979. (13) Reichardt, C. Chem. Rev. 1994, 94, 2319-2358. (14) Cebolla, V. L.; Membrado, L.; Matt, M.; Ga´lvez, E. M.; Domingo, M. P. Thin-Layer Chromatography for Hydrocarbon Characterization in Petroleum Middle Distillates. In Analytical Advances in Hydrocarbon Research; Hsu, C. S., Ed.; Kluwer Academic/Plenum Publishers: New York, 2002; Chapter 5, pp 95-112.

3700

Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

Samples were dissolved in an appropriate solvent and applied, at in least triplicate, onto the impregnated TLC plates using a bandsprayer Linomat IV sample applicator (from Camag, Muttenz, Switzerland), as 2-mm bands. Sample application point is at 85 mm in HPTLC plates and 185 mm in TLC plates. Samples were developed using either a horizontal development chamber (Camag) or a conventional, vertical, standard TLC tank (22 cm × 25 cm). A CS9301 TLC scanner (Shimadzu) was used in the fluorescence mode. In berberine experiments, λexc ) 365 nm and detection in the >450-nm zone were used, while in RD experiments conditions changed to λexc ) 300 nm and detection in the zone of >400 nm. Linear scanning was used in both cases, with an 1.0 × 1.0 mm beam size. Peak area data were collected, displayed, and stored using Shimadzu CS9310 PC software. Further details on equipment used were reported elsewhere.14 The mixture of neutral lipids (2.8 µg of each) or, alternatively, pure cholesterol (0.02-to 0.7 µg) was applied on a conventional silica gel TLC plate, which was previously impregnated with a solution of berberine sulfate in MeOH (1 mg‚100 mL-1). Development was carried out in a conventional vertical tank using petroleum ether (bp ) 50-70 °C)-diethyl ether-acetic acid (80: 20:1) as eluant (20 min), as reported elsewhere.15,16 The mixture of phospholipids (0.02-5 µg of effective mass, in dichloromethane) was applied on a silica gel TLC plate. Sample was developed in a conventional vertical tank using chloroformMeOH-water (65:25:4) as eluant (15 min), as reported elsewhere.16,17 Postimpregnation with a solution of berberine sulfate in MeOH (1 mg‚100 mL-1) before scanning was carried out. Gas oil sample (5-17.5 µg) was applied on a HPTLC silica gel plate that was previously impregnated with a solution of RD in MeOH (40 mg‚100 mL-1). Development was carried out using n-hexane (5 min). High-Performance Liquid Chromatography Experiments. Runs were performed using two HPLC systems, each one consisting of a pump for mobile-phase elution, a second pump for pre- or postcolumn introduction of a solution of the corresponding probe, a cell for mixing mobile phase and probe solution, and a fluorescence detector. One of the HPLC systems was an integrated SFM 25 model from Kontron Instruments (Buckinghamshire, UK). The other system was a Shimadzu RF 530 (Croissy-Beaubourg, France), equipped with a Shimadzu LC 6A pump (Croissy-Beaubourg, France) for mobile-phase elution and a Milton Roy CP 3000 pump (Interchim, Montluc¸ on, France) for introduction of the probe solution. Acquisition and data treatment was performed in this case using Borwin software (Varian, Les Ulis, France). In general, samples were injected using 5- or 20-µL loops, either pure or dissolved in the mobile-phase solvent. Every sample was injected in triplicate. Elution was isocratic, and mobile phase flows usually ranged from 0.1 to 0.9 mL‚min-1. A silica gel HPLC column (Hypersil silica, 5 µm, 250 × 4.6 mm) and an octadecyl-silica HPLC column (ODS, 5 µm, 250 × (15) H. K. Mangold. Aliphatic lipids. In Thin-Layer Chromatography, 2nd ed.; Stahl, E., Ed.; Springer-Verlag: New York, 1969; pp 363-421. (16) B. Fried, J. Sherma. Lipids. In Thin-Layer Chromatography, 4th ed. Revised and Expanded; Chromatographic Science Series 81; Marcel Dekker: New York, 1999; Chapter 16. (17) Wagner, H.; Horhammer, L.; Wolff, P. Biochem. Z. 1961, 334, 175-184.

4.6 mm), both from Supelco, Inc. (Bellefonte, PA), were used depending on the type of analyte. GPC colums (two PLGel 50 columns, 5 µm, 50 Å, 7.5 mm × 30 cm, Touzart et Matignon, France; plus one Microstyragel 500, 7.5 mm × 30 cm, Waters) connected in series were also used. Probe concentration was usually between 25 and 200 µg‚mL-1 with flows typically in the range from 0.1 to 1 mL‚min-1. Unless otherwise stated, the solvent or mixture of solvents used for FDIC solution was the same as that used as mobile phase. However, a different solvent can be used for dissolving the probe. Since berberine cation interacts strongly with silica gel, only postcolumn introduction was used with this FDIC probe. The mobile phase and probe solution mixture will be referred to as “detection eluant” throughout this work. MeOH (1 vol %) was added to dichloromethane (DCM) and trichlorotrifluoroethane (TCTFE) to improve berberine cation solubility in these solvents. These mixtures will be referred to as DCM* and TCTFE*. When using these modified solvents, DCM and TCTFE were used respectively as mobile phases. Excitation wavelengths used were 350 nm in the case of berberine cation and 260 and 300 nm for RD. Emission wavelengths were 520 and 375 nm for berberine and RD, respectively. RESULTS AND DISCUSSION Changes in Emission Intensity in the Absence of Solvent. It was reported that, when a silica gel plate is impregnated with a solution of berberine cation, the application onto the layer of a paraffinic compound produces an increase in the fluorescence signal (between 450 and 550 nm) when the system is irradiated with UV light at 365 nm.3,4 This fluorescence was exclusively due to the sample in the presence of berberine, and its intensity was in proportion to sample load and alkylic chain length. This phenomenon has been used to develop analytical procedures for detecting saturated hydrocarbons in diesel fuels and other fossil fuel-derived products.14,18-21 Many other molecules induce a change in berberine fluorescence baseline. Lipids also provide an enhanced emission signal,4 whereas molecules with higher polarity induce decreases of fluorescence intensity, as shown in Figure 1 in the case of the studied amino acids, proteins, and rifampicine, an antibiotic. Certain hydrophilic proteins (bovine serum albumin and thrombin) give negative peaks. Hydrophobic amino acids such as tyrosine, proline, and phenylalanine induce a fluorescence enhancement, while hydrophilic ones, like histidine, induce a fluorescence quenching. In the case of alkanes, the enhancement of intensity was explained taking into account that nonspecific, electrostatic, alkane-probe interactions can create a microenvironment that isolates the fluorescent probe from other nonfluorescent decay mechanisms.3 The Einstein coefficient of spontaneous emission, which depends inversely on , the dielectric constant of the (18) Cebolla, V. L.; Membrado, L.; Domingo, M. P.; Henrion, P.; Garriga, R.; Cossio, F. P.; Arrieta, A.; Vela, J. J. Chromatogr. Sci. 1999, 37, 219-226. (19) Cebolla, V. L.; Matt, M.; Ga´lvez, E. M.; Membrado, L.; Domingo, M. P.; Vela, J.; Beregovtsova, N.; Sharypov, V.; Kuznetsov, B. N.; Marin, N.; Weber, J. V. Chromatographia 2002, 55, 87-93. (20) Bacaud, R.; Cebolla, V. L.; Membrado, L.; Matt, M.; Pessayre, S.; Ga´lvez, E. M. Ind. Eng. Chem. Res. 2002, 41, 6005-6014. (21) Matt, M.; Ga´lvez, E. M.; Cebolla, V. L.; Membrado, L.; Bacaud, R.; Pessayre, S. J. Sep. Sci. 2003, 26, 1665-1674.

Figure 1. TLC chromatograms, on berberine-impregnated (6 mg‚ 100 mL-1 methanol) HPTLC silica gel plates: (A) Rifampicine (1 µg); (a) fluorescence at λexc ) 365 nm; (b) UV at 254 nm. (B) (a) Bovine serum albumin; (b) thrombin, both by fluorescence at λexc ) 365 nm. (C) Histidine (2 µg); (a) fluorescence at λexc ) 365 nm; (b) UV at 254 nm. Samples were applied and detected in the application point. (Response in arbitrary units.)

medium, will be increased. Also, the quantum yield will depend on the interaction energy between the fluorescent probe and the Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

3701

Figure 3. HPTLC-scanning fluorescence chromatogram of a gas oil (80 µg). Development conditions: n-hexane, 5 min. Preimpregnation of plate: solution of Reichardt’s dye in MeOH (40 mg‚100 mL-1). Elution order: alkanes (62 mm); naphthenes (69 mm); negative peak corresponds to aromatics; positive peak at the sample application point (85 mm) corresponds to heavy condensed hydrocarbons. Detection conditions: λexc ) 300 nm; λem ) 400-450 nm. (Fluorescence response in arbitrary units.)

Figure 2. (A) TLC-scanning fluorescence chromatogram of a mixture of neutral lipids: (a) cholesterol, (b) cholesteryl oleate (C18: 1, cis-9), (c) oleic acid (C18:1, cis-9), (d) oleic acid methyl ester (C18: 1, cis-9), and (e) triolein (C18:1, cis-9). Development conditions: petroleum ether (bp ) 50-70 °C)-diethyl ether-acetic acid (80:20: 1), 20 min. Preimpregnation of plate: solution of berberine sulfate in MeOH (1 mg‚100 mL-1). (B) HPLC-scanning fluorescence chromatogram of a mixture of phospholipids: (a) L-R-phosphatidylcholine, (b) L-R-phosphatidylethanolamine, (c) ammonium salt of L-R-phosphatidylinositol, and (d) L-R-lisophosphatidylcholine. Development conditions: chloroform-MeOH-water (65:25:4), 15 min. Postimpregnation of plate: solution of berberine sulfate in MeOH (1 mg‚100 mL-1). Detection conditions: λexc ) 365 nm; λem ) 450-550 nm. (Fluorescence response in arbitrary units.)

alkane. For an ion-induced dipole interaction, as is the case between berberine and alkanes, that directly depends on the polarizability of the alkane, a linear correlation between alkane size and fluorescence increase can be expected. This was in fact experimentally verified.3 Data corresponding to other molecules different from alkanes in the absence of solvent seem to fit reasonably well in this theoretical framework (see Appendix). The resulting intensity can be explained as a balance between radiative and nonradiative processes.3,4 FDIC Detection for Planar Chromatography. Figure 2 shows TLC and HPTLC chromatograms corresponding to separa3702 Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

tion of mixtures of neutral lipids and phospholipids using classical conditions of lipid development15-17 and FDIC detection. Preimpregnation of berberine cation was used in the case of neutral lipids, while postimpregnation was required for phospholipids, as some components of their elution mixtures (aqueous solvents, acetone, MeOH) also dissolve berberine. In both cases, and with the other variables being the same (sample load, impregnation, and TLC conditions), fluorescent intensities for neutral lipids were higher than those for phospholipids. In the case of cholesterol determination, the use of a solution containing 1 mg of berberine‚100 mL-1 of MeOH allows 20 ng of cholesterol to be quantified. Sensitivity seems adequate and can be controlled further through modification of impregnating conditions, as was demonstrated in the case of TLC-fluorescence scanning of alkanes.18 Use of RD as a probe is illustrated in Figure 3, which shows the HPTLC fluorescence scanning (λexc ) 300 nm) chromatogram of a gas oil that has been separated into its hydrocarbon classes on a RD-impregnated silica gel plate (40 mg‚100 mL-1 of MeOH) and detected by fluorescence scanning densitometry. A similar gas oil was separated and detected in a similar manner using berberine as a probe (6 mg‚100 mL-1 of MeOH), as reported elsewhere.21 Separated alkanes and naphthenes show two positive fluorescent peaks. After these, a broad and negative peak corresponding to aromatics appears. Detection of this peak by UV at 254 nm confirms its aromatic nature. Its sign can be explained as a result of an inner filter effect because of their absorbing activity in the UV-visible. FDIC Detection for HPLC: Response Can Be Tailored in Function of Polarity Environment. Experiments performed showed that the previously reported phenomenon of probeinduced fluorescence also occurs in solution. Its use for HPLC detection was studied. HPLC baseline is determined by the nature and flows of the solvents used (mobile phase and probe solvent)

Figure 5. Representation of HPLC-FDIC area peak vs mass of n-heptane (n-C7). Eluant: Tetrahydrofuran as mobile phase (1 mL‚min-1) and berberine (50 µg‚mL-1) in DCM* (1 mL‚min-1). GPC colums (set of three connected in series; two PLGel 50 and one Microstyragel 500). Detection conditions: λexc ) 355 nm; λem ) 520 nm. (Area in arbitrary units.)

Figure 4. Evolution of berberine fluorescence in binary solvent mixtures as a function of DCM molar fraction (χDCM). (A) Intensity of berberine fluorescence (4 µg‚mL-1) in DCM/MeOH mixtures (b) or DCM/n-decane mixtures (∆). (B) HPLC-FDIC detection: (a) MeOH (4 µg); (b) n-hexane (13 µg). Mobile phase: DCM (1 mL‚min-1). Berberine postcolumn: 200 µg‚mL-1 in DCM* (0.3 mL‚min-1). The arrows show injection time of solutes. (Fluorescence response and intensity in arbitrary units.)

and by the nature and concentration of the probe. Solvents influence the berberine fluorescence spectrum: the lower the  of solvent, the higher the intensity level of the baseline. Therefore, the higher the water content of the solvent, the lower the intensity level of baseline. A further injection of an analyte gives rise to a positive or negative peak. Magnitude and sign of FDIC response seem to depend on the following variables through a series of different effects: mobile-phase nature and flow; probe nature and concentration; probe solvent; probe solution flow; fluorescence detection (λexc, λem, etc.) parameters; sample load; and analyte chemical structure. Variations of the peak area and even sign for a given analyte can be obtained through controlled modification of the above-mentioned variables. Conversely, a detailed explanation for a particular series of data can be rather difficult to justify as there are too many variables and possible opposite effects to consider. Figure 4A illustrates this point in the case of berberine in binary solvent mixtures. Fluorescence intensity has been measured in solutions of DCM with protic (MeOH) or nonprotic (ndecane) cosolvent. In MeOH/DCM mixtures, it grows to a maximum before decreasing at molar fractions of DCM close to 1. In n-decane/DCM mixtures, there is a direct relationship between fluorescence intensity and molar fraction of DCM in the composition range allowed by the low solubility of berberine in the alkane. A tentative explanation of these results might take

into account the high solubility of berberine in MeOH and a preferential solvation22 effect that produces a microenvironment of a polar solute with more molecules of one solvent than the other, which could easily lead to nonlinear effects regarding MeOH concentration. MeOH gives then a positive peak while a negative one is obtained for n-hexane when the molar fraction of DCM is close to 1 (Figure 4B) because of the increase or decrease in berberine solubility influenced by preferential solvation. Particular electronic effects in the interaction between berberine and DCM could also be involved. Other experiments have shown that flow changes of berberine solution, using both TCTFE* and ACN as detection eluants, have a great influence on the area and even sign of peaks. For instance, the n-heptane peak sign can be inverted varying exclusively the flow of berberine solution, using TCTFE* as detection eluant, as is the case for MeOH using ACN. When using berberine, FDIC detection works with adequate sensitivity when low -solvents are used (Figure 5). DCM/DCM* as detection eluant seems the best compromise between detection sensitivity and berberine-solvent compatibility. Berberine (200 µg‚mL-1) also allows detection of nonfluorescent poly(methyl methacrylate) (Mw ) 4700) with this combination of solvents. Intensity of berberine fluorescence is low in aqueous solvents, and further addition of an analyte produces modest increases in fluorescence intensity. For any analyte, areas decrease with  and water content in the solvent system. (22) Do, M. A.; Silva, R.; Da Silva, D. C.; Machado, V. G.; Longhinotti, E.; Frescura, V. L. A. J. Phys. Chem. A 2002, 106, 8820-8826. (23) Valeur, B. Molecular Fluorescence: Principles and Applications; Wiley: Weinheim, 2002; pp 72-124, and references therein. (24) Lakowicz, J. R. Principles of fluorescence spectroscopy, 2nd ed.; Kluwer Academic/Plenum Publishers: New York, 1999; pp 237-239, and references therein. (25) Reichardt, C. Solvents and Solvent Effects in Organic Chemistry, 3rd ed., Updated and Enlarged; Wiley-VCH: Weinheim, 2003; p 13. (26) Debye, P. Phys. Z. 1920, 21, 178; 1921, 22, 302.

Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

3703

FDIC can also be used to quantify molecules by HPLC. Figure 5 shows the increase of area peak versus mass of n-heptane (nC7), a saturated hydrocarbon, in DCM/DCM* eluant under the chromatographic conditions specified therein, after detector signal inversion without modifying any chromatographic variable. Linearity of n-C7 response allows quantitative determination to be carried out. Molecular size also influences response.3,4 For a given class of compounds, response increases with molecular weight. Figure 6A shows the increase in fluorescent response factor (expressed in area‚mol-1) versus n-alkane chain length, from n-pentane (nC5) to n-triacontane (n-C30). Polystyrenes show also a similar pattern (Figure 6B). As FDIC does not involve covalent interactions, but weak interactions between analyte, probe, and solvent, it should not be considered as a derivatization procedure. It is a nondestructive method from the point of view of the analyte. Therefore, recovery of solvent and probe may be also possible. Choosing a Probe for HPLC-FDIC. Probes other than berberine that fulfill several requirements may be used. The choice of an FDIC probe is determined by its solubility properties and its native fluorescence intensity in solvents that determine response and sensitivity for a given analyte. As an example, RD (25 µg‚mL-1) has been used for detecting amounts of a polystyrene (MW) 3 220 000) in HPLC using a silica gel column. Poly(methyl methacrylates), n-alcohols, and n-alkanes have also been detected with RD either in flow conditions or in HPLC mode. Organic solvents have only been used because of their low solubility in water. Under the same conditions and concentration, RD gives lower sensitivity than berberine cation.

Figure 6. Influence of molecular size on HPLC-FDIC response factor (area‚mol-1) of (A) n-alkanes (n-pentane, n-C5; n-heptane, n-C7; n-nonane, n-C9; n-decane, n-C10; n-undecane, n-C11; n-tridecane, n-C13; n-hexadecane, n-C16; n-triacontane, n-C30) vs n-alkane chain length. Eluant: DCM as mobile phase (0.3 mL‚min-1) and berberine (200 µg‚mL-1) in DCM* (0.8 mL‚min-1). Silica gel column. Detection conditions: λexc ) 350 nm; λem ) 520 nm. (B) different polystyrenes (Mw: 2700, 4136, 5050, 7000, 11 300, 28 500) vs molecular weight.

Although the use of aqueous eluants or buffered solutions is not the best alternative when using berberine as a probe, fluorescence signal can still be adequate sometimes. So, fructose can be detected using ACN/H2O (70/30, v/v) and berberine (200 µg‚mL-1) under HPLC conditions. Moreover, other molecules such as salts, carboxylic acids, ascorbic acid, and cysteine have been detected in aqueous or buffered solvents in experiments under flow conditions. HPLC-FDIC: A Technique for Determining Nonfluorescent Compounds. FDIC use in HPLC is suitable for a wide variety of compounds, but seems especially interesting for the detection of nonfluorescent analytes. Compounds detected with adequate sensitivity include nonfluorescent polymers, saturated hydrocarbons, sugars, salts, and biomolecules. This may be of interest in the fields of petrochemistry and polymer characterization, among others. 3704

Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

CONCLUSIONS The application of FDIC detection, incorporating probes to commercially available chromatographic separation systems, has been studied. HPLC and planar chromatography have been used as FDIC takes place both in the presence and the absence of solvent. FDIC, either in HPLC or TLC, is operationally simple and a useful analytical tool for detecting a wide variety of analytes, especially the nonfluorescent ones. Saturated hydrocarbons and nonfluorescent polymers in HPLC and neutral lipids, phospholipids, or fossil fuel samples in TLC have been detected. The fact that weak, noncovalent interactions are involved in this kind of fluorescence is an additional advantage with regard to derivatizing reactions. Solvent selection is very important in HPLC mode. Probe solubility and fluorescence properties in a particular solvent determine its suitability for a particular analysis. Aqueous and organic solvents can be used as mobile-phase or detection solvent. Although only berberine and RD have been used in this work, other molecules may also be used as probes. In TLC mode, FDIC detection does not involve solvents. Peak response can be tailored through controlled modification of probe nature and concentration and of sample load. ACKNOWLEDGMENT Authors thank the Spanish Ministerio de Ciencia y Tecnologı´a (MCyT) and Ministerio de Educacio´n y Ciencia for financial support (projects PPQ2001-2388 and CTQ2005-00227/PPQ). E.M.G.

acknowledges MCyT for a grant. The authors also thank Dr. Julia´n Pardo and Profs. Marisa Peleato, Javier Galba´n, and Susana de Marcos for technical support.

be the sum of the RD base fluorescence and that of the RD fraction combined to an analyte, the variation in fluorescent intensity ∆IRD,i of RD induced by the analyte i should follow a general linear relationship that can be described by the following expression:

APPENDIX Nonspecific analyte-probe interactions can create a microenvironment that isolates the fluorescent probe from other nonfluorescent decay mechanisms. In contrast, specific donoracceptor Coulombic or exchange interactions result in a quenching of the fluorescence intensity.23,24 When there is a strong or weak coupling between the interacting molecules, it is assumed that the rate associated with these de-excitation processes is proportional to the interaction energy.23,24 Similarly, in the case of weak, nonspecific probe-analyte interactions contributing to the efficiency of the fluorescence emission, we can assume that the intensity enhancement will also be proportional to the interaction energy. As we reported in our previous paper, when the fluorescent probe is ionic in nature (berberine), the most important probe-analyte interaction is of the type ion-induced dipole. For a ionic probe P and analyte i,25

∆IRD,i ) ARD + BRDCRD,i

2

UP,i ) -

where CRD,i is a combined parameter between RD and i, defined as

CRD,i ) RRDµi2 + RiµRD2

(5)

We have measured the ∆IRD,i values for different alcohols (using 8 µg of alcohol and an RD concentration of 400 mg‚L-1) and we have found that the experimental data are in agreement with eqs 4 and 5, with r ) 0.9941, ARD ) -616, and BRD ) 0.24. In the case of polarizable analytes with no permanent dipole moment, CRD,i is given by

CRD,i ) RiµRD2

(6)

2

Z pe Ri 1 2 (4π0) 2r4

(2)

and therefore eq 4 transforms into

∆IRD,i ) A′RD + B′RDR where r is the average probe-analyte intermolecular distance, ZP is the charge of the fluorescent probe, 0 is the dielectric permittivity of the medium, and Ri is the polarizability of i. In the case of dipolar probes (Reichardt’s dye), we should use the Debye equation,25,26 and the equation for the interaction energy will become

URD,i ) -

(4)

RiµRD2 + RRDµi2 1 (4π0)2 r6

(3)

where symbols have the same meaning as in eq 2, RRD and Ri are the polarizabilities of RD and i, respectively, and µRD and µi are the corresponding permanent dipole moments. Since at a given temperature the average values of r must be similar for structurally related analytes, and the total intensity will

(7)

which is formally equivalent to our previously reported expression for changes in fluorescent intensity governed by ion-induced dipole interactions in the case of berberine and alkanes.3,4 We have also measured the ∆IRD,i values for different n-alkanes (using 5 µg of alkane and an RD concentration of 400 mg‚L-1) and we have found that they follow eq 7 with r ) 0.996, A′RD ) 571, and BRD′ ) 45. Therefore, our model to describe the changes in fluorescence intensity for polar probes such as RD may be extended to different types of analytes, including the nonpolar ones, as in the case of berberine cation. Received for review November 13, 2005. Accepted March 26, 2006. AC058045B

Analytical Chemistry, Vol. 78, No. 11, June 1, 2006

3705