High-performance liquid chromatography with diamond ATR FTIR detection for the determination of carbohydrates, alcohols and organic acids in red wine

Size: px

Start display at page:

Download "High-performance liquid chromatography with diamond ATR FTIR detection for the determination of carbohydrates, alcohols and organic acids in red wine"

Collin York
5 years ago
Views:

1 Anal Bioanal Chem (2003) 376 : DOI /s ORIGINAL PAPER Andrea Edelmann Josef Diewok Josefa Rodriguez Baena Bernhard Lendl High-performance liquid chromatography with diamond ATR FTIR detection for the determination of carbohydrates, alcohols and organic acids in red wine Received: 30 November 2002 / Revised: 6 February 2003 / Accepted: 7 February 2003 / Published online: 29 March 2003 Springer-Verlag 2003 Abstract A horizontal diamond attenuated total reflection (ATR) element has been incorporated in a flowthrough cell with low dead volume and used for on-line mid-ir detection in high-performance liquid chromatography. The chemical inertness of the ATR element permitted the use of a strongly acidic mobile phase in the isocratic separation. The hyphenation was used for the analysis of organic acids, sugars and alcohols in red wine. In the case of co-eluting analytes multivariate curve resolution alternating least squares (MCR ALS) was successfully employed for quantitative analysis. Keywords HPLC FTIR Diamond ATR Wine analysis MCR ALS Introduction A. Edelmann J. Diewok J. R. Baena B. Lendl ( ) Institute of Chemical Technologies and Analytics, Vienna University of Technology, Getreidemarkt 9/164-AC, 1060 Vienna, Austria blendl@mail.zserv.tuwien.ac.at The coupling of high-performance liquid chromatography (HPLC) and Fourier transform infrared spectroscopy (FTIR) is very powerful analytical tool. Since most compounds absorb in the mid-infrared region, FTIR can be regarded a general detector for liquid chromatography and has been recently reviewed [1]. This is of special interest in the case of poorly UV absorbing substances like carbohydrates and alcohols. Furthermore, FTIR spectroscopy provides detailed qualitative information on the separated analytes along with a unique fingerprint that often enables confirmation of the identity of the separated analytes. This has been shown on the example of soft drink analysis where an ion-exchange column (Ca 2+ counter ion) and distilled water as mobile phase have been used under isocratic conditions. Under these conditions glucose, fructose and sucrose could be quantified as well as the presence of taurin and ethanol in the sample confirmed by comparison of the spectra of the unknown with reference spectra [2]. For the simultaneous separation of carbohydrates, alcohols and organic acids in wine, however, an acidic mobile phase (0.005 M H 2 SO 4 ) had to be used along with an ion-exchange column (H + counter ion). To cope with the ph of the mobile phase the IR-transparent CaF 2 windows of the flow cell had to be coated with a thin polymer film to protect them from acidic degradation [3]. Despite the fact that this concept proved to be successful, the necessity for frequent replacement of the protective coating hindered routine application of this technology. However, this problem can be overcome when more robust and non-corroding window materials can be applied. In this context diamond, which has recently become a popular material for attenuated total reflection (ATR) spectroscopy [4], is particularly promising due to its chemical inertness. Whereas the use of attenuated total reflection cells mostly based on cylindrical ZnSe rods (Circle cell) has been already reported in combination with HPLC separations [5, 6, 7] the use of a diamond-based ATR cell has not been reported yet. In this work we have integrated a ninereflection horizontal diamond ATR cell in a flow cell and attached it to an isocratic HPLC system using M H 2 SO 4 as mobile phase. For on-line FTIR detection, liquid-phase isocratic separation conditions must be used because gradients would result in a significant change in background absorption that would be extremely difficult to compensate for. This experimental limitation makes full separation of the analytes difficult because of the reduced freedom in setting optimum physicochemical separation conditions. In the case of full-spectrum recording, post-run chemometric treatment of chromatographic data sets presents a powerful solution for mathematically resolving overlapping or co-eluting peaks. The techniques used for this purpose are commonly known as curve resolution (CR) approaches [8]. Various CR methods have been proposed that differ in their principles and requirements to the data structure. Multivariate curve resolution alternating least squares (MCR ALS) [9] is a very

2 flexible method that does not demand identical peak shapes in different runs and allows the simultaneous analysis of several (standard/calibration and real) samples. In this work MCR ALS and the high information content of the mid-ir spectra of the sugar and acid analytes have been used to separate the individual contributions of several overlapping or co-eluting analytes and to predict their concentrations in real wine samples. Experimental Reagents All standard substances were of analytical grade (>99%) and purchased from standard suppliers. Standards were prepared by dissolving the analyte in distilled water and stabilized with some mg L 1 of NaN 3. The mobile phase consisted of M H 2 SO 4 thereby providing all organic acids undissociated. The wine samples (all red wines from the 2000 harvest) were provided from collaborating wineries located in Carnuntum, Austria. All wine samples were filtered with disposable syringe filters (45-µm pore size, polypropylene, Cameo, MSI, Inc., Westborough, USA) prior to analysis. Apparatus The HPLC system consisted of a Merck/Hitachi L71000 isocratic pump (flow rate 0.5 ml min 1 ) and a Rheodyne 7725 injection valve (injection volume 50 µl). The analytical column (VA 300/7.8 NUCLEOGEL SUGAR H, mm) was obtained from Macherey Nagel, Düren, Germany. The stationary phase was a polystyrene divinylbenzene anionic-exchange resin with H + counter ion (particle size 8 µm, crosslinkage 8%). An in-line filter and a mm guard column with the same stationary phase was used in order to protect the analytical column. The column was used at room temperature. A horizontal diamond ATR cell (Dura SamplIR, SensIR Technologies, Danbury, USA) with a circular surface of 4.3-mm diameter and nine internal reflections was employed with an in-housebuilt flow-through accessory (see Fig. 1). This enabled the pumping through of solutions over the ATR diamond surface with minimal dead volume (ca. 3 µl). A Bruker IFS-55 spectrometer equipped with liquid nitrogencooled mercury cadmium telluride (MCT) detector was used for spectra acquisition. For the control of the spectrometer the software package OPUS 3.0/IR (Bruker) was used. Spectra were collected in the range 1, cm 1 with 8-cm 1 resolution and 50 coadded scans each (apodisation function: Blackman-Harris-3-term). The scanner velocity was 100 khz (HeNe frequency). The total acquisition time including Fourier transformation and further data processing was 4.9 s, providing a time resolution of 12 spectra per minute. Data processing was carried out using OPUS 3.0/IR. Multivariate curve resolution alternating least squares (MCR ALS) For resolution and quantitation of heavily overlapping chromatographic peaks multivariate curve resolution alternating least squares (MCR ALS) was applied [9]. The general goal of curve resolution [8] applied to co-elution regions of HPLC data with full-spectrum recording is to decompose the mixed spectral matrix D (elution time wavenumber) into the pure contributions of the co-eluting compounds, according to: D = CS T + E (1) where matrix C contains columns of elution profiles for all compounds and S T contains the corresponding compound spectra in rows. MCR ALS solves Eq. (1) in an iterative, alternating leastsquares manner by minimizing the error matrix E. Theory and application of curve resolution, and MCR ALS in special cases, have been described in detail elsewhere [8, 9, 10]. For optimal resolution, HPLC runs of mixture samples (calibration or real wine samples) were analysed together with HPLC runs of pure standard samples. From the integrated areas of MCR-resolved elution profiles and the known concentrations of calibration samples linear calibration for each analyte could be established. Pretreatment of HPLC FTIR data prior to MCR ALS consisted of subtracting the average baseline spectrum recorded directly before co-elution of the analytes and calculating first-derivative spectra in order to remove baseline drifts. Data treatment was performed in Matlab 5.3 (The Math Works Inc., Natick, MA, 1999). 93 Fig. 1 Schematic view of the applied flow cell attached to the ATR device Fig. 2 Spectra of standards containing 10 g L 1 of each compound recorded with the ATR cell after elution from the column: 1 citric acid, 2 glucose, 3 tartaric acid, 4 fructose, 5 malic acid, 6 lactic acid, 7 glycerol, 8 acetic acid, 9 ethanol. For better visualization appropriate offsets were added

3 94 Fig. 3a, b Chromatogram of a standard sample with analyte concentration 10 g L 1. a Selective trace for the carbohydrates and alcohols (integration limits 1,080 1,020 cm 1 ), b selective trace for the organic acids (integration limits 1,290 1,230 cm 1 ). 1 citric acid, 2 glucose, 3 tartaric acid, 4 fructose, 5 malic acid, 6 lactic acid, 7 glycerol, 8 acetic acid, 9 ethanol. The negative peak at 4.8 min displays the decrease in sulfate absorption (maximum at 1,105 cm 1 ) due to sample injection (compared to the M H 2 SO 4 of the eluent) For MCR ALS analysis, the freely available program (Matlab code) by A. de Juan and R. Tauler was used [11]. Results and discussion FTIR spectra of standard substances and recorded chromatograms The spectra of all investigated pure compounds are depicted in Fig. 2. The single pure substances were injected (injection volume 50 µl) and a chromatographic run was performed. The spectra were then extracted from the 3D data set of the chromatogram at the peak maximum and were used as reference spectra for the identification of the substances. It is important to note that the bands of IR spectra, in particular those of organic acids, depend on the degree of dissociation and are therefore particularly sensitive to ph shifts. Thus, in the recording of reference spectra, the same conditions as in all other experiments were used. As illustrated in Fig. 3, not all of the analytes could be separated under the chromatographic conditions used in this work. The analytes fructose and malic acid were sufficiently separated, as were acetic acid and ethanol. For these analytes the calibration and data evaluation was performed directly on the recorded spectra using appropriate integration boundaries. However, citric acid, glucose and tartaric acid were insufficiently separated for direct data treatment; lactic acid and glycerol completely co-eluted in one peak. Therefore these five analytes were subjected to MCR ALS for resolution of their peaks and quantitative analysis. Quantitative analysis The concentration range of the synthetic multi-analyte calibration samples was 0.5, 1, 3, 5, 8 and 10 g L 1 for all analytes except ethanol which was additionally calibrated at 80, 90, 100 and 110 g L 1 in order to meet the concentrations found in wines. All samples were injected in triplicate. In the case of the sufficiently separated compounds (ethanol, fructose, acetic acid and malic acid) the areas obtained by integrating the respective peak of the elution profiles were subjected to linear regression. The overlapping analyte peaks (citric acid/glucose/tartaric acid and lactic acid/glycerol) were resolved by MCR ALS. Then the MCR-resolved elution profiles from the calibration samples were integrated and the areas also linearly regressed on the known concentrations. The precision of the method was evaluated by the standard deviation of the Table 1 Parameters of calibration Analyte Equation of linear regression R 2 Integration limits Standard Limit for trace (cm 1 ) deviation a s xo of detection b (g L 1 ) (gl 1 ) Acetic acid y=0.7153x c ,310 1, Fructose y=0.0259x c ,096 1, Malic acid y=0.5631x c ,300 1, Ethanol y=0.4121x c ,070 1, Glycerol y= x d ,400/MCR Glucose y= x d ,400/MCR Citric acid y= x d ,400/MCR Lactic acid y= x d ,400/MCR Tartaric acid y= x d ,400/MCR Concentration range: 0.5, 1, 3, 5, 8 and 10 g L 1 for all analytes except ethanol, which was calibrated additionally at 80, 90, 100 and 110 g L 1 a According to ISO (1990) b According to DIN (Equation 15, 1994) [12] c x concentration (g L 1 ), y interval integrated absorbances (arbitrary units) d x concentration (g L 1 ), y integrated MCR elution profiles (arbitrary units)]

4 95 Fig. 4 3D plot of co-eluting lactic acid and glycerol from HPLC FTIR run of wine sample S5 Fig. 5a, b MCR-ALS results for lactic acid glycerol co-elution peak in wine sample S5. Glycerol (dash-dot), lactic acid (solid). a Resolved elution profiles for HPLC FTIR run of glycerol standard, lactic acid standard and wine sample. b First-derivative spectra for glycerol and lactic acid obtained from MCR ALS method s xo, calculated as the ratio of the residual standard deviation, s y, and the slope (according to ISO ). The limit of detection was estimated according to DIN (Equation 15, 1994) [12]. The calibration functions, standard deviations and detection limits are listed in Table 1. The real wine samples were subjected to MCR analysis simultaneously with pure standard samples for optimal resolution. Then, the resolved areas were used for analyte prediction via the previously established calibration functions. In Fig. 4, a 3D plot of the co-eluting glycerol and

5 96 Table 2 Contents of carbohydrates and organic acids in red wines S1 S7 determined by the newly developed method Standard deviation is given in brackets Values of citric acid and malic acid below limit of detection a Calibration based on peak areas of integration traces b Calibration based on MCR results Analyte S1 (g L 1 ) S2 (g L 1 ) S3 (g L 1 ) S4 (g L 1 ) S5 (g L 1 ) S6 (g L 1 ) S7 (g L 1 ) Glucose b * 0.55 (0.01) (0.21) (0.19) (0.02) (0.14) (0.04) (0.10) Tartaric acid b (0.11) (0.17) (0.16) (0.17) (0.16) (0.31) (0.15) Fructose a (0.08) (0.17) (0.06) (0.10) (0.11) (0.20) (0.13) Lactic acid b (0.05) (0.05) (0.11) (0.16) (0.12) (0.14) (0.08) Glycerol b (0.09) (0.02) (0.01) (0.01) (0.18) (0.12) (0.22) Acetic acid a (0.05) (0.14) (0.18) (0.03) (0.17) (0.16) (0.19) Ethanol a (0.17) (1.05) (0.19) (0.51) (0.04) (1.63) (0.08) lactic acid in wine sample S5 is given as an example of overlapping chromatographic peaks. It is obvious that the two analytes cannot be quantified by conventional approaches as they overlap heavily in the elution time and spectral direction. When the wine sample is subjected to MCR ALS simultaneously with pure standard samples the individual contributions of the analytes can be resolved (Fig. 5). From the obtained elution profiles (Fig. 5a) the concentration of both glycerol and the completely embedded lactic acid can be predicted accurately. Results of the analysis of red wine samples The results of the analysis of red wines together with the corresponding standard deviations are shown in Table 2. As can be seen the concentrations of malic and citric acid are below the limit of detection of the method. Very low concentrations of these analytes have been expected, since all red wines of this study have undergone complete malolactic fermentation. In this step, malic acid is converted to lactic acid by lactic acid bacteria such as Oenococcus oeni [13]. Also, citric acid, which is naturally already present only in very small amounts, is further degraded during this process. For this reason its concentration in the analysed wines is very low. The concentration of the remaining organic acids tartaric, acetic and lactic acid were in the normal range [14], as were glycerol and ethanol concentrations. The concentrations found for glucose and fructose are generally low, which is expected for dry red table wines. In addition, it is known that during malolactic fermentation residual sugar can be metabolised by lactic acid bacteria, hence producing off-flavours [15]. Therefore, wine producers generally aim to complete the alcoholic fermentation and to obtain low remaining sugar concentrations prior to malolactic fermentation. This fact further explains the found concentration levels for glucose and fructose. Conclusions A dedicated flow cell comprising a diamond ATR cell was used for on-line FTIR detection in the analysis of organic acids, sugars and alcohols in red wines that have been separated by HPLC. The chemical inertness of the employed diamond enabled the use of a strongly acidic mobile phase. Despite the fact that baseline separation was not achieved for all analytes quantitative analysis of all poorly separated target analytes could be carried out successfully using multivariate curve resolution alternating least squares (MCR ALS) in post-run data analysis. The chemical inertness of the developed flow cell presents an important improvement in the robustness of HPLC FTIR. Therefore, we consider this hyphenation applicable also for routine analysis. Acknowledgement Financial support from the Austrian Science Foundation within the project is acknowledged. References 1. Somsen GW, Visser T (2000) Liquid chromatography/infrared spectroscopy. In: Meyers RA (ed) Encyclopedia of analytical chemistry. John Wiley, Chichester 2. Vonach R, Lendl B, Kellner R (1997) Anal Chem 69: Vonach R, Lendl B, Kellner R (1998) J Chrom A 824: Fitzpatrick J, Reffner JA (2002) Macro and micro internal reflection accessories. In: Chalmers JM, Griffiths PR (eds) Handbook of vibrational spectroscopy, Vol 2. Wiley, Chichester 5. Sabo M, Gross J, Wang J, Rosenberg IE (1985) Anal Chem 57: Rein A, Wilks P Jr (1982) Am Lab 14:152; McKittrick PT, Danielson ND, Katon JE (1991) J Liq Chromatogr 14: Massart DL, Vandeginste BGM, Buydens LMC, de Jong S, Lewi PJ, Smeyers-Verbeke J (1998) Handbook of chemometrics and qualimetrics, Part B. Elsevier, Amsterdam, Chap Tauler R (1995) Chemom Intell Lab Syst 30:

6 De Juan A, Casassas E, Tauler R (2000) Soft-modeling of analytical data. In: Meyers RA (ed) Encyclopedia of analytical chemistry, Vol 11. Wiley, New York 11. Tauler R, de Juan A (1999) Multivariate curve resolution alternating least squares (MCR ALS), MATLAB code, University of Barcelona, Barcelona, Spain. mcr/mcr.htm. Cited 29 Nov Deutsches Institut für Normung (1994) DIN , Nachweis, Erfassungs- undbestimmungsgrenze 13. Ribéreau-Gayon P, Dubourdieu D, Donéche B, Lonvaud A (2000) Handbook of enology, Vol 1. Wiley, Chichester, Chap Cabanis JC, Cabanis, MT (1998) In: Flanzy C (ed) Oenologie. Lavoisier Tech & Doc, Paris, Chap Ribéreau-Gayon P, Dubourdieu D, Donéche B, Lonvaud A (2000) Handbook of enology, Vol 1. Wiley, Chichester, Chap 3.8.2

RESOLUTION OENO 33/2004 DETERMINATION OF SHIKIMIC ACID IN WINE BY HPLC AND UV-DETECTION

RESOLUTION OENO 33/2004 DETERMINATION OF SHIKIMIC ACID IN WINE BY HPLC AND UV-DETECTION DETERMINATION OF SHIKIMIC ACID IN WINE BY HPLC AND UV-DETECTION The GENERAL ASSEMBLY, Considering Article paragraph iv of the agreement establishing the International organisation of vine and wine Upon