Thiago NogueiraClaudimir Lucio do Lago

Departamento de QuímicaFundamental,Instituto de Química,Universidade de São Paulo,São Paulo – SP, Brazil

Received January 23, 2007Revised March 15, 2007Accepted March 19, 2007

Research Article

Determination of caffeine in coffeeproducts by dynamic complexation with3,4-dimethoxycinnamate and separationby CZE

A method based on the formation of p-complexes with chlorogenate-like species was pro-posed for the determination of caffeine in regular (nondecaffeinated) and decaffeinatedcoffee. Both caffeate and 3,4-dimethoxycinnamate were able to transform caffeine – a neu-tral species in aqueous solutions – into an anionic species. The usage of 3,4-dimethoxy-cinnamate in the running electrolyte is advantageous, because of its greater chemical sta-bility and the improved resolution of the peaks of caffeine, theobromine, and theophylline.Negative peaks were registered with a capacitively coupled contactless conductivity detectorwhen solutions of these alkylxanthines were analyzed with a BGE composed of 20 mmol/L3,4-dimethoxycinnamic acid and pH adjusted to 8.5 with Tris. This behavior was expected,because the complex is larger and thus should move slower than the free anion. Caffeinewas determined in ground and instant coffee with precision and accuracy that meet Bra-zilian norms about such products. The LOD was estimated as 33 mg/L, which correspondsto 0.8 and 0.3 mg of caffeine per gram of dry instant coffee and ground coffee, respectively.For the case of decaffeinated coffee, ten times preconcentration with dichloromethane wascarried out to allow the quantitation of caffeine, which should not exceed the concentrationof 1 mg/g in dry matter.

Keywords:

Alkylxanthines / Caffeine / Chlorogenic acid / Coffee / Contactless conductivitydetection DOI 10.1002/elps.200700039

3570 Electrophoresis 2007, 28, 3570–3574

1 Introduction

Coffee is the most popular drink around the world, and theproduction of the coffee beans is in the base of the economyof several tropical countries. This popularity is due in part tocaffeine, a well-known stimulant drug. Thus, the qualitycontrol of this commodity is important from both commer-cial and public health standpoints.

Several analytical methods have been proposed, butHPLC became the method for routine analysis of caffeine incoffee products as well as in other drinks and pharmaceuticalformulations.

In the last few decades, CE was also proposed as a tech-nique for the determination of caffeine and related com-

pounds in several matrices. Although soluble in water, caf-feine is a neutral species in a broad pH range. Thus, MEKChas been proposed as the most appropriated approach [1–8].

Hyotylainen and co-workers [2] could not separate caf-feine from other neutral drugs in several kinds of BGE withpH from 3.0 to 11. In that study, not only pH was changed,but the chemical composition (glycine, acetate, Tricine,CAPS, and borate) with no effect on the caffeine migration.However, good resolution was obtained by adding SDS to themedium. Similar result was obtained by Blanco and Valverdewith SDS and bile salts. Alternatively, the authors also triedto use carboxymethyl-b-CD (CM-b-CD) to accomplishseparation. Although caffeine could be mobilized due to itsinclusion in the anionic CM-b-CD, poor resolution betweenneutral xanthines discourages such an approach. On theother hand, Wang and co-workers [9] had success using astructure other than micelles. They proposed the use of car-boxylic single-walled carbon nanotubes to resolve caffeineand theobromine peaks. Huang and co-workers [8] demon-strated that heptane–SDS microemulsion might also be usedfor the determination of caffeine.

Correspondence: Professor Claudimir Lucio do Lago, Departa-mento de Química Fundamental, Instituto de Química, Universi-dade de São Paulo, Av. Prof. Lineu Prestes 748, CEP 05508-000,São Paulo – SP, BrazilE-mail: claudemi@iq.usp.brFax: 155-11-3091-3781

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Electrophoresis 2007, 28, 3570–3574 CE and CEC 3571

In spite of the fact that caffeine is a neutral species inaqueous medium, some authors have used CZE without asuitable modifier to allow its electrophoretic migration [10–15]. In these cases, however, caffeine is carried by the EOFand, thus, is not separated from other species. Instead ofthat, other ionic species are separated from caffeine andother neutral species. If the sample matrix is simple enoughand well controlled, like a pharmaceutical formulation, thisapproach may be acceptable, but it should not be genericallyrecommended.

Although we were able to show that capacitively coupledcontactless conductivity detection (C4D) may be used inMEKC [16], the sensitivity is low in some cases, and, thus, aCZE counterpart should be used [17]. That was also the casefor caffeine. Our second attempt to impose charge to caffeinewas trying to complex it with metal ions. Although suchcomplexes are described in the literature [18], the addition ofZn21 and Ca21 to the running buffer was not effective.However, another kind of complex was investigated: a p-complex with an ionic species.

Caffeine is found in coffee seeds associated to chloroge-nic acids, which are esters of trans-cinnamic acids and quinicacid. The most common and commercially available one is5-O-caffeoyl-quinic acid, which is conventionally denomi-nated chlorogenic acid [19].

Basically, caffeine can freely permeate through mem-branes and it is accumulated inside the cells by formingcomplex with one of the chlorogenic acids [20]. These naturalcomplexes inspired this work, which proposes the use ofchlorogenate analogs to convert caffeine into an anionicspecies to be separated by CE.

2 Materials and methods

All reagents and solvents were of analytical grade. Caffeine(1,3,7-trimethylxanthine), theobromine (3,7-dimethyl-xanthine), theophylline (1,3-dimethylxanthine), caffeic acid,and 3,4-dimethoxycinnamic acid were obtained from Sigma(St. Louis, USA). Tris, sodium hydroxide, ethyl acetate,methanol, and dichloromethane were obtained from Merck(Darmstadt, Germany). All the BGE and standard solutionswere prepared by dissolving the reagents or diluting stocksolutions in deionized water (Barnstead, Dubuque, IA,USA).

All electropherograms were performed using homemadeCE-C4D equipment [21, 22]. The detector was operated at600 kHz and 2 Vpp. Although the resulting signal is propor-tional to the conductance of the solution, it is a commonpractice to give the resulting voltage without conversion toconductance or conductivity scales. However, to preventmisunderstandings, an arbitrary unit scale (based on themillivolt scale) is displayed in the figures.

Sixty-centimeter long (effective length of 50 cm) and75 mm id6375 mm od fused-silica capillaries (J&W Scientif-ic, Folsom, CA, USA) were used. The samples were hydro-

dynamically introduced by elevation of the anodic reservoirat 10 cm for 30 s. A constant voltage of 25 kV was used for allexperiments.

The BGE was prepared daily by dissolving given amountsof caffeic acid or 3,4-dimethoxycinnamic acid in deionizedwater and adjusting the pH with Tris solution. After that, thefinal volume was completed with deionized water. The capil-lary column was preconditioned before use by washing with0.1 mol/L NaOH solution (10 min), followed by deionizedwater for 10 min and BGE (20 min). Before the first injec-tion, high voltage was applied for approximately 20 min.

The method described by Trugo and co-workers [23] wasemployed for HPLC analysis. Comparative measurementswere performed in a Shimadzu (Model LC-10A) liquid chro-matograph equipped with a piston pump and an UV–Visabsorbance detector. A Shim-pack G-ODS (4) RP18(1064 mm id, 5 mm) precolumn and a Shim-pack G-ODS(4) RP18 (25064.6 mm id, 5 mm) analytical column wereemployed. Water/methanol 60:40 was used as mobile phaseat a flow rate of 0.9 mL/min and UV absorption was mon-itored at 272 nm.

Coffee samples were purchased at the local market. Ali-quots of 1.000 g of ground coffee and 0.400 g of regular ordecaffeinated instant coffee were used. Infusions were pre-pared adding 10 mL of deionized water to the dried materialand heating under stirring at 907C in a water-bath for10 min. For the decaffeinated instant coffee samples, apreconcentration step, based on the liquid–liquid extractionof caffeine using dichloromethane as described by Ashoorand co-workers [24], was used. The pH of the decaffeinatedcoffee solution was adjusted to 10 with 1 mol/L NaOH solu-tion and transferred to separation funnel and extracted threetimes with 20 mL of solvent. Then, the organic phase wascompletely evaporated in a glass beaker and then resus-pended in 1 mL of deionized water. All samples were filteredthrough a 0.22 mm pore membrane filter.

3 Results and discussion

Figure 1 shows the structures of the chemical species ofinterest in the following discussion.

The caffeine–chlorogenate complex was described byGorter [25] one century ago. Since there is no central metalelement in the complex, this kind of complex is usuallyformed by hydrogen or hydrophobic bonds. In 1972, Hor-man and Viani [26], based on NMR studies, proposed thatcaffeine–chlorogenate complex might be described as a 1:1hydrophobically bound p-molecular complex. In this samestudy, the authors determined the association constant ofother analogous species with caffeine to try identifying theportion of the chlorogenate ion that is responsible for thecomplex formation.

Although chlorogenate could be used to compose a run-ning buffer that would impose a negative charge to caffeine,chlorogenic acid is too expensive to be used in routine anal-

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

3572 T. Nogueira and C. L. do Lago Electrophoresis 2007, 28, 3570–3574

Figure 1. Structures of caffeine (a), theobromine (b), theophyl-line (c), 5-O-caffeoyl-quinic acid or chlorogenic acid (d), caffeicacid (e), 3,4-dimethoxycinnamic acid (f).

ysis. Thus, caffeic acid was initially chosen, because the qui-nate moiety, which differentiates caffeic and chlorogenicacids, has a little effect on the complexing ability: the associ-ation constants are 16.9 and 12.2 kg/mol for chlorogenateand caffeate, respectively [26].

Figure 2 shows the electropherogram of a standard solu-tion of caffeine (500 mg/L) with a BGE composed of30 mmol/L caffeic acid, whose pH was adjusted to 8.0 withTris. The pKa values of the three dissociable groups of caffeicacid are 4.44, 7.6, and 11.85 at 207C [27]. Thus, at pH 8, caffe-ate has the carboxylate group totally dissociated, while onephenol group is partially dissociated. This anionic speciescomplexes caffeine, which migrates away from the sampleplug (EOFpeak) and allows it to be detected as a negative peak.This negative peak registered by a conductivity detector isexpected, because the complex is larger than caffeate aloneand, thus, its mobility should be smaller [28, 29].

As shown, caffeic acid could be successfully used in CEanalysis of caffeine. However, as a catechol analog, caffeicacid is prone to oxidation [27]. Thus, caffeate was substitutedby 3,4-dimethoxycinnamate that is less prone to oxidationand has an even higher association constant with caffeine(15.7 kg/mol) [26].

Figure 3 shows the electropherogram of a standard solu-tion of caffeine (200 mg/L), theobromine (200 mg/L), andtheophylline (100 mg/L) with a BGE composed of 20 mmol/L3,4-dimethoxycinnamic acid, whose pH was adjusted to 8.5with Tris. This condition allows baseline separation of thethree methylxanthines and becomes a promising approachfor the determination of such species in other matrices. Inthe present paper, however, theobromine and theophyllineare not systematically studied, because they are not presentin an appreciable extent in coffee products. In the following

Figure 2. Electropherogram of a standard solution of caffeine(500 mg/L) with a BGE composed of 30 mmol/L caffeic acid/Tris(pH 8.0).

Figure 3. Electropherogram of a standard solution of caffeine(200 mg/L), theobromine (200 mg/L), and theophylline (100 mg/L)with a BGE composed of 20 mmol/L 3,4-dimethoxycinnamic acid/Tris (pH 8.5).

sections, the BGE based on 3,4-dimethoxycinnamate wasused for the determination of caffeine in ground (non-instant), instant, and decaffeinated instant coffee.

3.1 Method validation

Precision was evaluated by five consecutive runs of standardsolution of caffeine and five consecutive runs of a real sam-ple of coffee. Figure 4 shows the superposed electro-pherograms. The RSD for the migration time and peak areawere not greater than 0.18 and 1.1%, respectively.

Intermediate precision was evaluated by determining theconcentration of caffeine in a real sample in three differentdays. The results for each day were 31.2, 32.6, and 30.3 mg/gwith RSD for triplicates of 1.2, 0.8, and 0.5%, respectively.The global RSD was 3.7%.

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Electrophoresis 2007, 28, 3570–3574 CE and CEC 3573

Figure 4. Electropherograms of a standard solution of caffeine(a) and a real sample of coffee (b) prepared as described in thetext. Five electropherograms of each solution were superposed.

The accuracy was evaluated by calculating the percentage ofrecovery and by comparison with HPLC. Samples of fourinstant coffee and two of ground coffee of different originswere spiked with caffeine to render two different levels ofconcentration (50 and 100 mg/L) in the solution to be injec-ted. The results, summarized in Table 1, show that it is notpossible to detect loss of material in the procedure.

Table 2 shows the results of a comparative study betweenthe proposed method and the one based on HPLC for sixdifferent kinds of coffee. The results show a good agreementbetween both methods.

Table 1. Recovery test of caffeine

Coffee samples Added concentration(mg/L)

Recovery (%)

Ground 50 97–105100 92–107

Instant 50 92–104–98–99100 109–97–93–103

Table 2. Concentration of caffeine obtained by CE and HPLC

Coffee samples Concentrationby CEa)

(mg/g coffee)

Concentrationby HPLCa)

(mg/g coffee)

Instant 1 31.5 (60.4) 31.91 (60.09)Instant 2 33.1 (60.2) 34.06 (60.16)Instant 3 33.0 (60.6) 33.84 (60.10)Instant 4 29.5 (60.3) 30.36 (60.12)Ground 1 12.7 (60.3) 12.66 (60.04)Ground 2 14.2 (60.2) 14.68 (60.04)

a) SD expressed in parenthesis (n = 3).

The analytical curve in the range from 100 to 500 mg/Lshows good linearity (r = 0.9998, n = 5) and the LOD(S/N.3) was estimated as 33 mg/L. Taking into account thedilution along the procedure, the LOD for caffeine in drymatter is 0.8 and 0.3 mg/g of dry instant coffee and groundcoffee, respectively.

3.2 Decaffeinated coffee

The LOD obtained for the proposed method is enough to attendBrazilian norm about decaffeinated coffee (Anvisa ResolutionRDC number 277 from September 22, 2005), which limits in 1and 3 mg of caffeine per gram of dry matter for ground andinstant coffee, respectively. However, a preconcentration stepwas included in the cases of decaffeinated products to allow notonly the detection, but also quantitation of caffeine.

In general, the industrial decaffeination is carried outwith the coffee beans, before the roasting process, by using asolvent [30]. The extraction can be done in two ways. The firstone is the direct extraction, in which the solvent is put indirect contact with the beans. In the second way, water isused as an extracting phase, which is then put in contact withan organic solvent to remove the caffeine. In this case, waterhelps caffeine to be released from the chlorogenate complexand to permeate through the cell membranes [31].

These two approaches were used in the caffeine pre-concentration. However, the direct extractions with ethyl ac-etate and dichloromethane gave poor recovery of caffeine.The extraction with water followed by dichloromethane(Table 3) was better. Although caffeine is partially recovered(�89%), it is possible to identify a product that did not con-form to the norm. The aqueous phase was kept at pH 10 toprevent organic acids [32] from being also preconcentratedby dichloromethane. Figure 5 shows that the preconcentra-tion did not lead to the rising of other peaks than the caffeineone in that region, which could be deleterious to the method.

Table 3. Determination of caffeine and recovery test in dec-affeinated samples

Decaffeinatedcoffee

Concentrationa)

(mg/g of coffee)Recoverya)

(%)

Sample A 0.71 (60.01) 94 (62)0.72 (60.02) 97 (61)0.76 (60.02) 89 (62)

Sample B 0.35 (60.01) 97 (62)

a) SD expressed in parenthesis (n = 3).

4 Concluding remarks

The use of caffeate or 3,4-dimethoxycinnamate in the run-ning buffer allows the effective electrophoretic migration ofcaffeine by forming anionic complexes. To the best of our

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

3574 T. Nogueira and C. L. do Lago Electrophoresis 2007, 28, 3570–3574

Figure 5. Electropherograms of samples of regular instant coffee(a) and decaffeinated instant coffee before (b) and after (c) tentimes preconcentration.

knowledge, this is the first time that caffeine is mobilized inCE without the use of micelles or other macromolecularstructure.

The buffer based on 3,4-dimethoxycinnamate not onlyallows the quantitation of caffeine in coffee products, but isstable and cost-effective. Moreover, caffeate and 3,4-dime-thoxycinnamate are promising candidates for CE methodsfor other alkylxanthines that could complex with the planarp-system of such anions.

This work was supported by Fundacão de Amparo à Pesquisado Estado de São Paulo (FAPESP) and Conselho Nacional deDesenvolvimento Científico e Tecnológico (CNPq). The authorsalso thank CNPq for the fellowships and Dr. Z. G. Richter for theEnglish revision.

5 References

[1] Burton, D. E., Powell, L. L., Xi, X. B., J. Microcol. Sep. 1994, 6,5–10.

[2] Hyotylainen, T., Siren, H., Riekkola, M. L., J. Chromatogr. A1996, 735, 439–447.

[3] Zhao, Y. P., Lunte, C. E., J. Chromatogr. B 1997, 688, 265–274.

[4] Watanabe, T., Nishiyama, R., Yamamoto, A., Nagai, S., Ter-abe, S., Anal. Sci. 1998, 14, 435–438.

[5] Blanco, M., Valverde, I., J. Chromatogr. A 2002, 950, 293–299.

[6] Huang, W. S., Lin, S. J., Wu, H. L., Chen, S. H., J. Chromatogr.B 2003, 795, 329–335.

[7] Pomilio, A. B., Trajtemberg, S. P., Vitale, A. A., J. Sci. FoodAgric. 2005, 85, 622–628.

[8] Huang, H. Y., Lien, W. C., Chiu, C. W., J. Sep. Sci. 2005, 28,973–981.

[9] Wang, Z. H., Luo, G. A., Chen, J. F., Xiao, S. F., Wang, Y. M.,Electrophoresis 2003, 24, 4181–4188.

[10] Hurst, W. J., Martin, R. A., Analysis 1993, 21, 389–391.

[11] Horie, H., Mukai, T., Kohata, K., J. Chromatogr. A 1997, 758,332–335.

[12] Peri-Okonny, U. L., Wang, S. X., Stubbs, R. J., Guzman, N. A.,Electrophoresis 2005, 26, 2652–2663.

[13] Sombra, L. L., Gomez, M. R., Olsina, R., Martinez, L. D., Silva,M. F., J. Pharm. Biomed. Anal. 2005, 36, 989–994.

[14] Regan, F., Shakalisava, Y., Anal. Chim. Acta 2005, 540, 103–110.

[15] Amini, A., Barclay, V., Rundlof, T., Jonsson, S. et al., Chro-matographia 2006, 63, 143–148.

[16] da Silva, J. A. F., do Lago, C. L., Electrophoresis. 2000, 21,1405–1408.

[17] Felix, F. S., Quintino, M. S. M., Carvalho, A. Z., Coelho, L. H.G. et al., J. Pharm. Biomed. Anal. 2006, 40, 1288–1292.

[18] Kolayli, S., Ocak, M., Kucuk, M., Abbasoglu, R., Food Chem.2004, 84, 383–388.

[19] Clifford, M. N., J. Sci. Food Agric. 1999, 79, 362–372.

[20] Waldhauser, S. S. M., Baumann, T. W., Phytochemistry 1996,42, 985–996.

[21] da Silva, J. A. F., do Lago, C. L., Anal. Chem. 1998, 70, 4339–4343.

[22] da Silva, J. A. F., Guzman, N., do Lago, C. L., J. Chromatogr.A 2002, 942, 249–258.

[23] Trugo, L. C., Macrae, R., Dick, J., J. Sci. Food Agric. 1983, 34,300–306.

[24] Ashoor, S. H., Seperich, G. J., Monte, W. C., Welty, J., J.Assoc. Off. Anal. Chem. 1983, 66, 606–609.

[25] Gorter, K., Ann. Chem. 1907, 358, 327.

[26] Horman, I., Viani, R., J. Food Sci. 1972, 37, 925.

[27] Jovanovic, S. V., Steenken, S., Tosic, M., Marjanovic, B.,Simic, M. G., J. Am. Chem. Soc. 1994, 116, 4846–4851.

[28] Brito-Neto, J. G. A., da Silva, J. A. F., Blanes, L., do Lago, C.L., Electroanalysis 2005, 17, 1198–1206.

[29] Brito-Neto, J. G. A., da Silva, J. A. F., Blanes, L., do Lago, C.L., Electroanalysis 2005, 17, 1207–1214.

[30] Ramalakshmi, K., Raghavan, B., Crit. Rev. Food Sci. Nutr.1999, 39, 441–456.

[31] Toci, A., Farah, A., Trugo, L. C., Quim. Nova 2006, 29, 965–971.

[32] Galli, V., Barbas, C., J. Chromatogr. A 2004, 1032, 299–304.

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com