Determination of Major Carboxylic Acids in Wine by an Optimized HPLC Method with Linear Gradient Elution

Transcription

1 Determination of Major Carboxylic Acids in Wine by an Optimized HPLC Method with Linear Gradient Elution R. M. MARCE ~, M. CALULL 2, F. BORRULL 3. and F. X. RIUS 4 An optimized method to determine carboxylic acids in wine based on pre-column derivatization with phenacyl bromide is reported. The determination of the different acids has been carried out using a linear gradient elution involving water and methanol. The method has been optimized by the Simplex method using a modification of CRF (chromatographic response factor) as the criterion for selectivity optimization. After validation, the method was applied to determine carboxylic acids in red and white wines and grape must of the Tarragona region. KEYWORDS: carboxylic acids, wine, optimization, simplex method, linear gradient Carboxylic acids have an important influence on the organoleptic properties of wines, and they act as substrates or products in various enzymatic transformations. Therefore, the nature and concentration of these compounds in wines are of interest in severa] aspects of wine chemistry (19). Enzymatic methods are currently used to determine carboxylic acids in wine, but their applications are time-consuming and tedious. To overcome these drawbacks, the analysis of organic acids in wines by HPLC using ion exchange (5,18) or ion exclusion (6) were developed recently. However, most chromatographic systems employed in the direct determination of organic acids in wine do not solve the problem of coelution of neutral compounds along with the analyses of interest. Several alternative clean-up procedures have been developed to improve the accuracy but at the expense of speed and simplicity (9,15,23). To solve this problem, several authors have used a derivatization method to determine carboxylic acids in wine. Some of them (4,10) take advantage of their conversion into phenacyl esters, obtaining good resolution by applying reversed-phase liquid chromatography. In a previous paper (7), the different parameters which influence the derivatization reaction were optimized. In the present paper, the chromatographic separation and quantification of the different major carboxylic acids present in wine have been optimized using a modification of the original Simplex method. Starting composition of the linear elution program and rate of mobile phase change were tested to maximize the response, in this case, a modification of the CRF criterion for selectivity optimization. After validation, the method was applied to deter- 1,2,3.4Dep. Quimica, Universitat de Barcelona, Imperial Tarraco 1,43005 Tarragona (Spain). *Author to whom correspondence should be addressed. Acknowledgements: The authors wish to thank the CICYT (project PA ) for financial support, and R. M. Marc~ gratefully acknowledges a research grant from CIRIT of Catalonia. Manuscript submitted for publication 13 November Copyright 1990 by the Amreican Society for Enology and Viticulture. All rights reserved. 289 mine carboxylic acids in wines and grape musts. Materials and Methods Equipment: A Hewlett-Packard 1050 modular chromatograph with a Hewlett-Packard series 1050 variable wavelength ultraviolet detector and a HP 3396A integrator was used. The chromatographic separations were carried out with a Spherisorb ODS2 column (250 X 4.6 mm i.d.), 5-pm particle size with a precolumn (30 X 3.9 mm) filled with Bondapack C18/ Corasil (37-50 pm particle size) (Waters). Reagents and standards: The standard solutions of the acids were prepared by dissolving the major carboxylic acids frequently found in wine (tartaric, malic, lactic, acetic, succinic, and citric acids) (Aldrich Chemicals) in water to provide a concentration of 0.5 g/ L. The methylmalonic acid was used as internal standard because of its adequate retention time and its near absence in wine samples (4). Phenacyl bromide (Fluka, 77450) (reagent), 18- crown-6 (Fluka, 28125) (catalyst), phosphate buffer solution (ph 6.8), and acetone (solvent) were used in the derivatization reaction. As a mobile phase for the chromatographic separation, methanol and water were used. All solvents in the derivatization process and in the chromatographic separation were Merck HPLC quality, and water was MilliQ quality. Chromatographic conditions: After derivatization of the carboxylic acids, the separation was carried out by a linear solvent gradient. The chromatographic conditions adopted were as follows: flow-rate, I ml/min; detection, uv absorption, 254 nm; volume injected, 5 pl, temperature constant at 30 C; and mobile phase, water as solvent A and methanol as solvent B and a linear gradient, optimized in the present work, from 32% of B to 92% of B in 20 minutes. Derivatization procedure: The following derivatization process was adopted. To a 2-mL Pyrex screwcap test tube, 200 L of standard solution or wine sample previously neutralized with KOH, 200 L of the phosphate buffer solution ofph 6.8,400 pl of 0.2 M phenacyl bromide solution in acetone, 400 pl of 0.02 M 18-crown-

2 290 m MAJOR CARBOXYLIC ACIDS 6 solution in acetone, and 400 pl of acetone were added. The test tube was heated at 90 C for 90 minutes. The solution was cooled before chromatographic analysis. All carboxylic acids present in wine were sterified, although due to their low concentration level and the sensitivity of the present procedure they would not interfere in the determination of the major carboxylic acids studied. Results and Discussion An elution program with a polarity gradient is necessary to obtain the best separation and quantification of the different carboxylic acids present in wine. Among the many different gradients to be applied, the easiest is the linear gradient. To develop a linear gradient method, the composition of the mobile phase at initial time and the increment of the solvent B (organic solvent) during the chromatographic separation must be fixed prior to the chromatographic elution. The different values given for these two parameters by different authors (4,10) prompted us to optimize these two variables by a mathematical method. In this paper, the sequential Simplex method was used to optimize simultaneously the starting mobile phase composition and the solvent strength change with a reduced number of experiments. A modification of the original Nelder and Mead algorithm (11), which significantly increases the method efficiency, was used. In the present work, the selection of the initial conditions of the simplex method was based on the criterion given by Yarbro and Deming (22). A quality criterion must be defined in order to evaluate the quality of the results in the optimization process. In this case, the selection of this criterion was based on the indications given by P. Schoenmakers (16) which are.5 compiled by A. Peeters et al. (13) in their expert system for selectivity optimization. In general, chroma- 3 tographic responses to be optimized using sequential experi- 2.5 mental designs (e.g., the Simplex method) are based upon sums of terms, usually taking into account 2 resolution and analysis time. The CRF (chromatographic response 1.5 function) was selected as response criterion to optimize the selectiv- 1 ity by the Simplex method. The CRF term has different expressions depending on the variables involved: gradient parameters and flow-rate (12); concentration 0 of modifier, buffer, and ph (21), concentration of organic modifier and ph (12); composition of ter- nary mobile phase, temperature, flow-rate and ph (1); or composi- 0.5 tion of ternary phase (2). In this work, a modification of the criteria employed by Smilde et al. (17) was used as optimized criterion: CRF = a ~ (R,j) + b (t m - t,) where a and b are weighting factors, R is the resolution 1J. between the different peaks, t m the maximum acceptable retention time of the last peak and t~ the retention time of the last peak. The parameter range was selected on the basis of several bibliographical data (9,15), and it was defined as: starting composition from 30% to 45% of methanol and solvent strength change from 0.5% to 3% methanol/ minute. The optimization of the separation of the carboxylic acid was carried out using an actual red wine sample. Due to the appearance of several system peaks, solvent and reagent peaks and several minor organic acids in real derivatized wine samples, these peaks must be taken into account in the optimization process. For the calculation of the CRF term, values of the coefficients a = b - 1 were chosen, since the same importance has been allotted to the two factors, sum of resolutions and difference of time between the last eluted peak and the maximum acceptable analysis time; the latter was established at 20 minutes in the present work. The resolution between the different peaks was calculated using the well-known expression R~ 2 = 2 (t 2 - t~)/(w 2 + W~). The term E R. was obtained considering the main different peaks of t~he chromatogram (system Sequential Simplex Solvent strength change (% methanol/min) _ 1 7/8 I I 1,,, I Starting solvent composition (% methanol) Fig. 1. Graphical representation of the distinct experimental conditions assayed during the sequential optimization process using the Simplex method.

3 MAJOR CARBOXYLIC ACIDS m 291 peak and derivatized carboxylic acids peaks). In this study, there were appreciable differences among the resolution values obtained for some pairs of peaks. Some of them were higher than 6 while other values are less than 1.5. This difference of the R values introduces a distortion effect on the CRF va~ue, since the same increment of the resolution has much less influence onto two well-separated peaks than onto two near peaks. Therefore, a modification in the CRF expression was introduced; only the peaks with resolution less than 4 were considered in the new E R.. term. 1j After Simplex optimization, the following obtained experimental conditions give a maximum CRF value: starting composition 32% of methanol and mobile phase strength change 3% of methanol/minute. A graphic representation of the different experiments undertaken to develop the best response by the Simplex method can be seen in Figure 1. After the first four experiments, all subsequent experimental points are situated at high solvent strength change values (near the maximum) and at starting composition between 30% and 35% methanol. In this study, it is apparent that the change of percent methanol per minute in the mobile phase was a stronger influence than the starting composition. The next six experiments considered (not labeled in the figure), all of them at 3% ,,,6 Ill '1" O~ ry~ f~ T 0,1 IYJ L1 I li A P) ~f' C~ ~0 B 45 9 CO e~j Oh 8~ O0 J o r f%, o 10 O) M) '11 1 I,'3 ~1", i~.~ ilia,.7~ u) Fig. 2. HPLC separation of standard carboxylic acids and a red wine according to the described procedure. A, standard sample B, red wine. 1, solvent peak; 2, reagent peak; 3, system peak; 4, lactic acid; 5, acetic acid; 6, tartaric acid; 7, malic acid 8, succinic acid 9, methyl malonic acid (internal standard) 10, citric acid.

4 292- MAJOR CARBOXYLIC ACIDS b-~ b') f') ~r -'r,~0 n ii A B! -i t~"o i,.~ i{ lo r/l! IA3 2, j ~JIi ~ I _. _. _. _ Fig. 3. Typical chromatogram of a white wine and a grape must. A, white wine B, grape must. 1, solvent peak; 2, reagent peak; 3, system peak; 4, lactic acid 5, acetic acid 6, tartaric acid 7, malic acid 8, succinic acid; 9, methyl malonic acid (internal standard) 10, citric acid. gradient increase and different values of starting composition, were carried out to reach the best value at 32% of solvent starting composition. The chromatograms obtained with the optimum conditions found corresponding to a standard mixture of carboxylic acids and to a typical red wine sample can be seen in Figure 2. The larger peaks correspond to the solvent peak (acetone) (peak 1), the phenacyl bromide peak (reagent) (peak 2), and a system peak (product of the derivatization process) (peak 3). The resolution between peaks 1 and 3 was not considered in the optimization process, and only the resolution values composed for the interest peaks and the adjacent ones were considered. The rest of the peaks correspond to the major acids present in wine: lactic (peak 4), acetic (peak 5), tartaric (peak 6), malic (peak 7), succinic (peak 8), and citric (peak 10). Method of validation: To validate the method, several studies were carried out. First, to determine the linearity of the response, the presence of possible influential points in the calibration lines drawn for each derivative carboxylic acid studied was checked by means of Cook's test (14). The ranges studied for tartaric, lactic, and malic acids are between 2 and 0.05 g/l; for succinic and acetic acids, between I and 0.04 g/l; and for citric acid, between 0.3 and 0.02 g/l. At higher concentrations of these compounds, deviations from linearity were observed. In some cases, the concentration of the carboxylic acids studied in the wine samples are higher than this value, and a dilution of the wine sample should be done before the derivatization process. In the studied range, a good linearity for all the

5 MAJOR CARBOXYLIC ACIDS m 293 carboxylic acids studied was obtained, with a regression coefficient higher than and without the presence of outliers in all cases. On the other hand, the effect of the matrix interferences was tested by comparing the slope of the aqueous calibration line to the slope of the calibration line obtained by using the standard addition technique. The Table 1. Accuracy of the optimized method. Carboxylic Present method Enzymatic method acid n = 6 n = 6 tcal mean (g/l) RSD (%) Mean (g/l) RSD (%) Lactic Malic Succinic Two-sided tta b ((X, = 0.001) = 4.52 tta b (a = 0.01 ) = 3.12 tta b ((z = 0.05) = 2.22 Table 2. Study of repeatability and reproducability between days of the optimized method. Carbo xyl ic Repeatab i I ity Reprod u ci b il ity acid n = 10 n = 10 Mean (g/l) RSD (%) Mean (g/l) RSD (%) Lactic Acetic Tartaric Malic Succinic Citric Slopmod program (8)was used, and the results obtained were compared with the tabulated t-test. When no significant difference between variances of the two regression lines existed, the usual t-test was used, and when a significant difference was observed, the Cochran approach was applied (8). No interferences were observed for tartaric and lactic acids when (z = 0.05, malic acid when a = 0.01, and acetic, succinic, and citric acids when a = The mean values obtained with the developed method were compared to the results obtained with the standard enzymatic method (3). Six replicate determinations were carried out for malic, lactic, and succinic acids which were analyzed in a unique white wine sample. The results are summarized in Table 1. A t-test was carried out between the results obtained for each carboxylic acid by both methods, and no significant differences were observed when a A stability study of the derivative compounds was carried out over ten days. No significant variation of the response was observed during this period, and the results were similar to those obtained in the repeatability studies (between 5% and 10% of the RSD). The repeatability (n - 10) and the reproducibility between days (during 10 days, the same red wine sample) of the method were also studied when applied to wine. The results can be seen in Table 2. Recovery studies were performed using a red wine. Each carboxylic acid was spiked at three different concentrations, and the results can be seen in Table 3. All the results obtained are higher than 90%, except for the citric acid that is nearer to 80%. The optimized method was applied to determine carboxylic acids in samples of white and red wines and must with similar results. A typical chromatogram for Table 3. Recovery study of organic acids. Acid Amount in wine (g/l) Amount Cal. Found added (g/l) (g/l) (g/l) % Recovery Lactic Lactic Tartaric Malic S u cci nic Citric

6 294- MAJOR CARBOXYLIC ACIDS Red wine White wine Grape must Table 4. Quantitative analysis of the carboxylic acids (g/l). nd= not detected. Tartaric Malic Lactic Acetic Citric Succinic acid acid acid acid acid acid nd nd 0.17 nd each application can be seen in Figure 3, and the results are summarized in Table 4. Conclusions The presented method can be considered an alternative to other HPLC methods based either on ion exchange or reversed-phase chromatography (with or without derivatization). The optimization of the chromatographic conditions to determine carboxylic acids in wine, carried out by using the simplex method, allows fast and reliable identification and quantification of the derivitized compounds in wine and grape must. The optimized method does not require any complicated sample preparation, so the method may be used routinely. The precision obtained is slightly lower than that obtained using enzymatic methods, although the presented method offers the advantage of reduced cost if the equipment is available and a shorter time of analysis. Literature Cited 1. Berridge, J.C. Unattended optimization of reverse-phase high performance liquid chromatographic separation using modified simplex algorithm, J. Chromatogr. 244:1-14 (1982). 2. Berridge, J.C., A. F. Fell, and A. W. Wright. Sequential simplex optimization and multichannel detection in HPLC: application to method development. Chromatographia 24: (1987). 3. Boehringer Mannheim GmbH, West Germany. Methods of Enzymatic Food Analysis 82/83. Gebr. Parcus KG. Munich (1982). 4. Caccamo, F., G. Carfagnini, A. Di Corcia, and R. Samperi. Improved high-performance liquid chromatographic assay for determining organic acids in wines. J. Chromatogr. 362:47-53 (1986). 5. Frayne, R. F. Direct analysis of the major organic components in grape must and wine using high performance liquid chromatography. Am. J. Enol. Vitic. 37:281-7 (1986). 6. Haginaka, J., J. Wakai, H. Yasuda, and T. Nomura. Ion-exclusion chromatography of carboxylic acids with conductivity detection. J. Chromatogr. 447: (1988). 7. Marc6, R. M., M. Calull, F. Borrull, and F. X. Rius. Optimization of the derivatization method for the liquid-chromatographic determination of carboxylic acids in wines. Anal. Chim. Acta. (In press, 1990). 8. Massart, D. L., J. Smeyers-Verbeke, and F. X. Rius. Method validation: software to compare the slopes of two calibration lines with different residual variance. Trends Anal.Chem. 8(2):49-51 (1989). 9. McCord, J. D., E. Trousdale, and D. D. Y. Ryu. An improved sample preparation procedure for the analysis of major organic components in grape must and wine by high performance liquid chromatography. Am. J. Enol. Vitic. 35:28-9 (1984). 10. Mentasti, E., M. C. Gennaro, C. Sarzanini, C. Baiocchi, and M. Savigliano. Derivatization, identification and separation of carboxylic acids in wine and beverages by high-performance liquid chromatography. J. Chromatogr. 322: (1985). 11. Nelder, J. A., and R. Mead. A simplex method for function minimisation. Computer J. 7: (1965). 12. Nickel, J.H., and S. N. Deming. Use of the sequential simplex algorithm for improved separations in automated liquid chromatographic methods development. Liq. Chromatagr. Mag., 1: (1983). 13. Peeters, A., L. Buydens, D. Massart, and P. J. Schoenmakers. An expert system for the selection of criteria for selectivity optimization in highpressure liquid chromatography. Chromatographia, 73:101-9 (1988). 14. Rius, F. X., J. Smeyers-Verbeke, and D. L. Massart. Method validation: software to plot calibration lines and their response residuals, and to detect outliers according to Cook's distance. Trends Anal. Chem. 8(1 ):8-11 (1989). 15. Schneyder, J., and W. Flak. Determination quantitative des principaux composants acides du vin par chromatographie liquide haute pression. F.V. OIV 749:1-5 (1982). 16. Schoenmakers, P. J. (Ed.). Optimization of chromatographic selectivity. A guide to method development. In:Journal of Chromatography Library (Vol. 35). Elsevier, Amsterdam (1986). 17. Smilde, A.K., A. Knevelman, and J. Coenegracht. Introduction of multi-criteria decision making in optimization procedures for high-performance liquid chromatographic separations. J. Chromatogr. 389:1-10 (1986). 18. Sneider, A., V. Gerbi, and M. Redoglia. A rapid HPLC method for separation and determination of major organic acids in grape must and wines. Am. J. Enol. Vitic., 38:151-5 (1987). 19. Tomasset, L. U. Chimica Enologica, AEB (2nd ed.). Brescia (1985). 20. Watson, M.W., and P. W. Carr. Simplex algorithm for the optimization of gradient elution high-performance liquid chromatography, Anal. Chem., 51: (1979). 21. Wegscheider, W., E. P. Lankmayr, and K. W. Budna. A chromatographic response function for automated optimization of separations. Chromatographia 15: (1982). 22. Yarbro, L. A., and S. N. Deming. Selection and preprocessing of factors for simplex optimization. Anal. Chim. Acta. 73:391-8 (1974). 23. Yokosuka, K., T. Matsudo, and T. Kushida. Pretreatment of wines with charcoal for determination of organic acids by high performance liquid chromatography with ultraviolet detection. J. Inst. Enol. Vitic. 18:7-13 (1983).