Multivariant optimization, validation, and application of capillary electrophoresis for simultaneous determination of polyphenols and phenolic acids in Brazilian wines



A method for the simultaneous determination of the stilbene resveratrol, four phenolic acids (syringic, coumaric, caffeic, and gallic acids), and five flavonoids (catechin, rutin, kaempferol, myricetin, and quercetin) in wine by CE was developed and validated. The CE electrolyte composition and instrumental conditions were optimized using 27-3 factorial design and response surface analysis, showing sodium tetraborate, MeOH, and their interaction as the most influential variables. The optimal electrophoretic conditions, minimizing the chromatographic resolution statistic values, consisted of 17 mmol/L sodium tetraborate with 20% methanol as electrolyte, constant voltage of 25 kV, hydrodynamic injection at 50 mbar for 3 s, and temperature of 25°C. The R2 values for linearity varied from 0.994 to 0.999; LOD and LOQ were 0.1 to 0.3 mg/L and 0.4 to 0.8 mg/L, respectively. The RSDs for migration time and peak area obtained from ten consecutive injections were less than 2% and recoveries varied from 97 to 102%. The method was applied to 23 samples of inexpensive Brazilian wines, showing wide compositional variation.

1 Introduction

Moderate consumption of wine has been associated with reduced risk of cardiovascular diseases 1–4 and cancer 5, as well as beneficial effects on the immune system 6 and cognitive functions 7, 8. These health-promoting properties have been attributed to the wine's phenolic compounds, particularly the flavonoids and the stilbene resveratrol. Phenolic compounds, such as phenolic acids, catechins, and other flavonoids, also have an important role in wine quality, contributing to flavor and color 9.

Determination of the quantitative composition and investigation of the factors affecting the composition of bioactive substances, using reliable methods, is considered a priority. Some inconsistencies in the findings of epidemiological studies can be due, at least in part, to analytical and natural variability of the data on the levels of these compounds. In their review article, German and Walzem 10 acknowledged that compositional factors complicated the experimental conclusions on the health effects of wine. There is also increasing recognition that the levels of bioactive substances in the diet can be maximized through agriculture, food technology, and nutrition. For such a strategy, the compositional variation in the food chain has to be known. Reviewing the occurrence and analysis of phenolics in foods, Nacsk and Shahidi 11 concluded that there is a need to develop more robust HPLC and capillary electrophoretic methodologies for the simultaneous determination of important classes of these compounds.

HPLC has been the method of choice for the analysis of phenolic compounds in wine, but CE has been increasingly used for this purpose. The application of CE in the analysis of beverages 12 and foods 13–15, including wine, and specifically for resveratrol in wine 16 has been reviewed. CZE, usually with phosphate or borate-based electrolytes, has been used for the quantitative analysis of phenolic acids 17–21, resveratrol 19, 22, flavonols 18, 19, catechins 18, 20–22, and different flavonoids 23. CZE was also applied in combination with ITP for the separation and quantification of 14 flavonoids and phenolic acids 24. Preferred for complex matrices and large numbers of analytes, MEKC with SDS was used for the determination of phenolic acids 25, 26 and flavonoids 26, 27. Some CE studies in wine have focused solely on the determination of the phytoalexin resveratrol 28–33. The diode array detector (DAD) has been generally utilized, but electrochemical detection 22, 33 has also been employed.

Comparison of the quantitative data obtained for phenolic compounds in wine by HPLC and CE methods has been carried out. While some small differences could be seen in the results obtained for some phenolics by Garcia-Viguera and Bridle 18 and by Wang and Huang 23, Andrade et al. 21 found no significant qualitative and quantitative differences in the results obtained by the two techniques. CE has the advantages of high speed, high resolution, low operational cost, low consumption of chemicals, and convenience.

In Brazil, the production of wine is concentrated in the southern region (state of Rio Grande do Sul). Data about phenolic compounds in Brazilian wine is still very limited. Souto et al. 34 quantified resveratrol in 36 samples of red wines by HPLC and Minussi et al. 20 determined phenolic acids and catechins in seven samples of white and red wines by CZE.

The objective of this study was to develop and evaluate a CE method for the simultaneous determination of important phenolic compounds in wine and apply it to various Brazilian wines.

2 Materials and methods

2.1 Instrumentation

Method development and evaluation, as well as sample analyses, were conducted in a capillary electrophoresis system model HP3DCE (Agilent Technologies, Palo Alto, CA, USA), equipped with a DAD set at 280 nm for quantification. The temperature was controlled and stabilized at 25°C. For data acquisition, a Chemstation Software (Rev A.06.01) (Agilent Technologies) installed in a personal computer supplied by the manufacturer was used. Samples and standard solutions were injected hydrodynamically (50 mbar for 3 s) and constant voltage of 25 kV was employed. The column was a fused-silica capillary (Polymicro Technologies, Phoenix, AZ, USA) with dimensions of 48.5 cm total length, 40.0 cm effective length, 75 μm id, and 375 μm od.

The software used for the factorial design and response surface analysis was Minitab 14 (Minitab, State College, PA, USA).

2.2 Reagents and solvents

All reagents were of analytical grade and solvents of chromatographic grade, used without previous purification. Methanol was purchased from Tedia Company (Fairfield, OH, USA), HCl from Merck KGaA (Darmstad, Germany), sodium tetraborate (STB), and SDS from Riedel-de Haen (Seelze, Germany) and Brij35 from Aldrich Chemical Company (Milwaukee, WI, USA). Water was purified by deionization (18 mΩ) (Milli-Q System, Millipore, Bedford, MA, USA). The standards resveratrol, gallic acid, p-coumaric acid, quercetin, rutin, myricetin, caffeic acid and catechin were purchased from Aldrich (St. Louis, MO, USA); syringic acid and kaempferol were from Fluka (St. Louis, MO, USA). Stock standard solutions of each of the ten phenolic compounds were prepared at 1000 mg/L in 60:40 v/v water/ethanol. Working solutions were prepared by mixing appropriate volumes of the stock solutions with 60:40 v/v water/ethanol. The standard solutions were stable for three months.

2.3 Sample preparation

Twenty-three samples of different types (red, white, and blended) and brands of wine were purchased from a supermarket in São Paulo, Brazil. Year of production and types of wine are listed in Table 5. These wines were chosen to get a good representation of wines produced in the region of Serra Gaucha, which are sold at low (<US$1) prices, thus reaching a greater part of the population.

All wines were stored in the dark at 4°C until analysis. The contents of two freshly opened bottles were mixed and 1 mL was extracted with 1.6 mL of ethyl ether, followed by acidification with 100 μL HCl for 15 min with magnetic stirring. The organic phase was separated from the aqueous phase, dried under nitrogen, and dissolved with 2.5 mL of ethanol/water 60:40 v/v. The samples were filtered through a 0.45 μm membrane (Millex LCR PTFE) (Millipore, Sao Paulo, Brazil).

The analytes were identified by comparison of the migration times with those of standards and with wine spiked with standards under identical conditions, along with the spectra of the migrated solutes obtained with the DAD.

2.4 Capillary electrophoresis procedure

New capillaries were preconditioned by 20 psi flushes with 1 mol/L NaOH (30 min) followed by deionized water (30 min). On each day of analysis, the capillary was conditioned by flushing with 1 mol/L NaOH for 5 min, followed by purified water (MilliQ System, Millipore) for 5 min, and the electrolyte solution for 30 min. In between runs, the capillary was flushed with the electrolyte solution (1 min). The electrolyte solution was prepared daily and filtered through a 0.45 μm membrane. All standards and samples were injected in triplicate.

3 Results and discussion

3.1 Optimum electrolyte composition and CE conditions

During method development and optimization, the chromatographic resolution statistic equation (CRS) was used. This is a mathematical function developed originally for chromatographic separations 35; the CRS tends to a minimum value for chromatograms having well resolved and uniformly spaced peaks. It has been used in the CE area for optimizing the separation of fatty acids 36 and food dyes 37.

STB-based buffers, MeOH as organic modifier, SDS or Brij35 as surfactants, time/pressure of injection, voltage, and temperature were investigated. So that the effects of various factors could be simultaneously investigated, including interaction of factors, a factorial design was employed rather than the one-factor-at-a-time approach.

STB as the electrolyte was assessed from 10 to 30 mmol/L. The lower level was established considering the need to maintain the pH in all the analyses. Very low levels of STB do not provide good buffering capacity. The upper level was limited by the electrical current. High amounts of STB combined with high levels of SDS may result in a high current, and temperature gradients can occur in the capillary due to Joule heating.

MeOH as organic modifier was evaluated from 0 to 10%. One of the first effects is a decrease in the electrical current, reducing Joule heating and peak band broadening. The second effect is increasing the migration time and the resolution of the peaks. Modification of the pKa of the analytes can occur with MeOH, thereby increasing solubilization.

The third variable was SDS, evaluated from 0 to 10 mmol/L. The formation of micelles was the main objective in selecting this variable because each compound can partition to the micelle core, thereby increasing selectivity. Brij35 from 0 to 5 mmol/L was also included because the presence of both SDS and Brij in the electrolyte results in a mixed micelle system constituted by anionic monomers (SDS) and neutral monomers (Brij35). The mixed micelle might provide better selectivity.

Time of injection from 1 to 3 s at 10 mbar, operation voltage from 10 to 25 kV, and temperature from 25 to 35°C were also evaluated.

Optimization of the CE method was initially carried out, using a factorial design, for the selection of the factors with direct influence on the separation. After selection of these factors, they were optimized by a response surface methodology. Table 1 shows the studied variables for the factorial design as well as the low (−1) and high (+1) levels.

Table 1. 27-3 Factorial design variable levels
Coded levelsSTB (mmol/L)MeOH (% v/v)SDS (mmol/L)Brij 35 (mmol/L)Injection time (s/mbar)Voltage (kV)Temp (°C)
  1. a

    STB, sodium tetraborate.


The results of the 27-3 fractionary factorial design are shown in Table 2, revealing that the effects of the Brij35 and SDS levels were not significant. Likewise, the time of injection did not have any significant effect during the studied interval and the high level was chosen in order to obtain better sensitivity. The temperature was kept at a low level to avoid bubbles that could be formed by the evaporation of the organic solvent. The most significant effects came from the variables STB, MeOH, and their interaction. These variables were therefore studied further through the response surface methodology, keeping the voltage at 25 kV, injection for 3 s at 10 mbar, and the temperature at 25°C, without SDS and Brij35.

Table 2. Principal effects obtained from the factorial design 27-3
Variable% Contribution
Injection time8.3
STB:injection time9.6

According to Table 3, the best condition (lowest CRS value) was achieved with 20% MeOH and 10 mmol/L STB. However, these conditions did not give good separation of the phenolic compounds of wine. Thus, response surface analysis, using the reciprocal of the CRS value, was used to obtain the optimum separation condition: 20% MeOH, 17 mmol/L STB, 25 kV, 3 s inj. 10 mbar at 25°C (Fig. 1).

Figure 1.

Response surface diagram for electrolyte composition and CE conditions. Optimum conditions: 20% MeOH, 17 mmol/L STB, 25 kV, 3 s injection 10 mbar, 25°C. STB, sodium tetraborate; CRS, chromatographic resolution statistic.

Table 3. Response function results of the response surface analysis
Experiment% MeOH v/vSTB (mmol/L)CRS
  1. a

    STB, sodium tetraborate; CRS, chromatographic resolution statistic.


3.2 Method performance

Linearity was evaluated by the injection of five standard solutions of each compound at concentrations varying from 2.5 to 220 mg/L. LOD and LOQ were calculated as the solution concentrations that gave signal-to-noise ratios of 3 and 10, respectively. Peak and migration time precision was evaluated by ten consecutive injections of the standard solutions. Recovery was studied at three levels, each level injected in triplicate.

Table 4 shows the performance characteristics of the proposed method. At the concentration range tested (2.5–220 mg/L), the method presented good R2 values (from 0.994 to 0.999). The LOQ varied from 0.4 to 0.8 mg/L and the LOD from 0.1 to 0.3 mg/L. Instrumental precision was shown by RSDs of ten consecutive injections of standard solutions being lower than 2% for both peak area and migration time. Recoveries varied from 97% (myricetin and quercetin) to 102% (rutin). The method is therefore considered suitable for the analysis of phenolic compounds in wine.

Table 4. Performance characteristics of the CE method developed
CompoundConcentration range (mg/L)a)InterceptSlopeR2a)LOQa) (mg/L)LODa) (mg/L)Recovery (%)b)Repeatability (%RSD)c)
        Peak areaMigration time
  • a)

    a) Seven concentrations, each injected in triplicate.

  • b)

    b) At three levels, each level in triplicate.

  • c)

    c) Ten consecutive injections.

Syringic acid5.0–601.5411.7920.9980.70.399±11.71.3
Coumaric acid18–2202.3757.3210.9960.50.298±10.91.5
Caffeic acid4.0–601.0914.1310.9940.70.2101±11.31.6
Gallic acid4.0–602.2345.4970.9940.50.199±11.71.7

3.3 Polyphenols of Brazilian wines

Obtained under the optimum conditions established, Fig. 2 shows the separation of the ten phenolic compounds used as standards. Figures 3 and 4 present the typical electropherograms of white and red wines, demonstrating the now well-known much greater phenolic concentrations in red wine. Table 5 shows the concentrations of the ten phenolic compounds determined in 23 inexpensive Brazilian wines. The wide variations in the phenolic levels indicate that better quality control for these wines is warranted.

Figure 2.

Electropherogram of a mixture of phenolic compound standards. Peak identification: 1, resveratrol; 2, catechin; 3, rutin; 4, syringic acid; 5, kaempferol; 6, p-coumaric acid; 7, myricetin; 8, quercetin; 9, caffeic acid; 10, gallic acid. CE conditions are described in the text.

Figure 3.

Electropherogram of the phenolic compounds in red wine. Peak identification: 1, resveratrol; 2, catechin; 3, rutin; 4, syringic acid; 5, kaempferol; 6, p-coumaric acid; 7, myricetin; 8, quercetin; 9, caffeic acid; 10, gallic acid. CE conditions are described in the text.

Figure 4.

Electropherogram of the phenolic compounds in white wine. Peak identification: 1, resveratrol; 2, catechin; 3, rutin; 4, syringic acid; 6, p-coumaric acid; 8, quercetin; 9, caffeic acid; 10, gallic acid. CE conditions are described in the text.

Table 5. Phenolic compound concentrations (mg/L)a) in Brazilian wines obtained by the CE method developed
SampleResveratrolCatechinRutinSyringic acidKaempferolCoumaric acidMyricetinQuercetinCaffeic acidGallic acid
  • a)

    a) Mean and standard deviations of triplicate analyses.

  • b)

    b) Estimated by extrapolation of the calibration curve; nd, not detected (<LOD).

Cabernet Sauvignon red 20003.1±0.1122.6±0.112.5±0.24.9±0.114.7±0.8nd12.3±0.111.7±0.111.7±0.129.1±0.1
Cabernet Sauvignon red 20013.5±0.344.7±0.123.6±0.44.4±0.117.4±0.6nd12.8±0.11.0b)17.1±2.031.0±3.0
Cabernet Sauvignon red1.4±0.2111.2±0.123.5±0.54.5±0.316.1±0.512.7±0.112.3±0.114.2±0.115.9±0.315.0±1.0
Cabernet Sauvignon red3.1±0.182.3±0.225.5±0.24.9±0.118.3±0.4nd12.3±0.111.7±0.111.7±0.129.1±0.1
Cabernet Franc red 20002.3±0.144.3±0.232.4±0.44.5±0.119.2±0.513.6±0.112.1±0.215.6±0.417.9±0.330.8±1.0
Cabernet Franc red 20013.4±0.1132.1±0.124.5±0.54.4±0.112.1±0.512.9±0.111.9±0.114.8±0.316.4±0.737.6±0.1
Gammay red 20023.7±0.181.5±0.123.8±0.7nd12.9±0.4nd11.1±0.210.6±0.3b)14.0±0.445.0±0.4
Gammay red 20031.0±0.1b)117.8±0.124.8±0.54.5±0.315.4±0.712.5±0.211.7±0.315.1±0.515.2±1.656.4±2.1
Merlot red 20012.0±0.172.1±0.121.4±0.6ndndnd11.2±0.311.5±0.217.7±0.168.7±0.2
Merlot red1.5±0.2183.2±0.118.9±0.8ndndnd11.7±0.213.5±0.117.7±0.169.9±0.1
Shiraz red 20021.4±0.272.9±0.221.2±0.6nd12.5±0.610.0±1.012.0±0.115.9±0.217.0±0.157.7±0.1
Blended red0.01b)nd12.4±0.6nd8.9±0.4nd14.6±0.1nd14.8±0.253.1±0.1
Chardonnay white 2002nd12.4±0.48.9±0.2ndndnd1.6±0.11.7±0.4b)1.8±0.31.6±0.4
Riesling white 2000nd13.6±0.4ndndndndndnd2.2±0.12.2±0.2
Riesling white 20010.9±0.2b)15.7±0.33.2±0.11.3±0.1nd2.1±0.3ndnd4.4±1.52.5±1.2
Riesling white 2002nd18.9±0.4ndndndndndnd3.4±0.32.5±0.5
Riesling whitend15.5±0.3ndndndndndnd3.3±0.32.4±0.2
Sauvignon Blanc white 2002nd23.4±0.4ndndndndndnd2.9±0.22.9±0.2
Ugni Blanc white 2002nd21.3±0.2ndndndndndnd2.2±0.42.8±0.2

In the comparison that follows, the blended red wine is not included. Gallic (15.0–69.9 mg/L for red wine; 1.6–2.9 mg/L for white wine) and caffeic (11.7–17.9 mg/mL for red wine; 1.8–4.4 mg/L for white wine) acids were the only phenolic compounds found in all the wines. German and Walzem 10 put together published data on phenolic compounds of wine and the ranges were 26–320 mg/L gallic acid and 3–18 mg/L caffeic acid for red wines. The Brazilian wines, therefore, are in the lower end for gallic acid but with caffeic acid in the upper end of these ranges.

Myricetin was found in all red wines (11.1–12.8 mg/L) and one white wine (1.6 mg/L). Quercetin was also encountered in all red wines (1.0–15.9 mg/L) and in one white wine (1.7 mg/L). According to German and Walzem, myricetin (2–45 mg/L) and quercetin (5–53) were found only in red wines, the Brazilian values falling within these ranges.

Syringic acid was detected in 7 (4.4–4.9 mg/L) out of 11 red wines and in 1 white wine (1.3 mg/L). The range reported by German and Walzem was 4.2–5.9 mg/L for red wine, in agreement with the data of the present work.

Coumaric acid was greater in blended wines (14.6–19.9); the range in five red wines was 10.0–13.6 mg/L. These ranges agree with that of German and Walzem (7.5–22 mg/L for red wine), but this acid was not detected in six red wines.

The concentration of one of the most important phenolic compounds in wine, resveratrol, ranged from 1.0 to 3.7 mg/L in red wine. The range in blended wines was 0.4–1.4 mg/L; in white wine, only one had resveratrol at a low concentration (0.9 mg/L). Souto et al. 28 found 0.82 to 5.75 mg/L in eight types of Brazilian red wines, in agreement with the values obtained in the present work, slightly higher than the range of German and Walzem (0.1–2.3 mg/L for red wine).

For catechin, the red wines had a range of 44.3–183.2 mg/L, while the white wines had 13.6–23.4 mg/L. The ranges in German and Walsem were 27–191 mg/L for the former and 3–35 mg/L for the latter. Rutin had a range of 12.5–32.4 mg/L in Brazilian red wines, higher than the range of German and Walzem of 0.5–10.8 mg/L. This compound was not detected in white wine by German and Walzem; it was found in two Brazilian white wines at 8.9 and 3.2 mg/L. Kaempferol had only one value in German and Walzen's table for red wine: 18 mg/L, which is within the range encountered in the present study of 12.1–19.2 mg/L for nine red wines, not being found in two red wines. In both work, this compound was not detected in white wine.

4 Concluding remarks

A simple, versatile, and low-cost CE method, which utilizes methanol, STB, and silica capillaries, was optimized by factorial design and response surface methodologies. Optimization of the electrolyte composition and CE conditions resulted in excellent selectivity. The method provided good LODs and LOQs, as well as linearity, peak area, and migration time repeatability and good recovery in the concentration levels studied. Applied to Brazilian wines, wide variation in the composition of phenolic compounds was observed, indicating the need for better quality control in the production of these wines. The method herein developed can serve as a valuable tool for such a task.


The authors have declared no conflict of interest.