Anal. Chem. 2001, 73, 453-457
An Inductively Coupled Plasma Carbon Emission Detector for Aqueous Carbohydrate Separations by Liquid Chromatography Heather L. Peters, Keith E. Levine,† and Bradley T. Jones*
Department of Chemistry, Wake Forest University, Winston-Salem, North Carolina 27109
An inductively coupled plasma atomic emission spectrometer is used to detect carbon-containing compounds following separation by high-performance liquid chromatography. A calcium form ligand exchange column with distilled and deionized water as the mobile phase is used to separate carbohydrates. The eluting species are detected by monitoring the carbon atomic emission line at 193.09 nm. The mass detection limits using a photomultiplier tube for sucrose and glucose are 50 ng, while that for fructose is 60 ng. The carbon emission detector should provide the same detection limit for any compound with a similar mass percent of carbon, whether or not the compound exhibits appreciable absorption characteristics. While the carbon emission detector will universally detect any organic compound, it will discriminate against species with high molar absorptivity that may be present at low concentration. Such species may act as interferences in chromatograms generated with conventional UVvisible absorption detectors. To demonstrate the utility of the carbon emission detector, three sugars (glucose, fructose, sucrose) are determined in apple, crangrape, and orange juice. Carbohydrates are essential components of all biological and botanical organisms. The metabolic breakdown of food sugars is perhaps the most important role played by these compounds because sugars provide the energy used to power biological processes. As a consequence, the determination of carbohydrates is of considerable importance to the food and beverage industry. In addition to providing the nutritional data offered by regulatory agencies, the quantitation of sugars also yields information related to the geographical origin and maturity of foodstuffs. The need for carbohydrate determinations in food and beverages has fueled the search for rapid and accurate techniques of analysis. Methods involving high-performance liquid chromatography (HPLC) are heavily employed in the literature because they allow for the rapid characterization of many nutritionally significant sugar species. Extensive research into various HPLC modes employed for carbohydrate separations has prompted the publication of recent review articles. Progress in application of high* Corresponding author: (e-mail)
[email protected]; (fax) (336) 758-5889. † Present address: Research Triangle Institute, P.O. Box 12194, Research Triangle Park, NC 27709. 10.1021/ac000902i CCC: $20.00 Published on Web 12/22/2000
© 2001 American Chemical Society
performance anion exchange chromatography (HPAEC),1 reversedphase high-performance liquid chromatography (RP-HPLC),2 hydrophilic interaction chromatography (HILIC),3 and ligand exchange chromatography (LEC)4 for the separation of carbohydrate species has been summarized. Although vast differences exist between the modes of HPLC, problems associated with the detection of carbohydrates are common to all. The popular UV detector is generally not employed because sugar species lack chromophores that result in absorption above 200 nm. Carbohydrates also do not exhibit native fluorescence, which precludes the use of highly sensitive and selective fluorescence detection systems. One approach often used to overcome these difficulties is to make carbohydrates more amenable to detection through a chemical derivatization procedure. Chen and Novotny5 reported the analytical properties of a new fluorescence-tagging reagent (PDFAc) for the enhanced detection of aminated carbohydrates, while derivatization procedures employed by Strydom (PMP)6 and Alpert et al.7 (2aminopyridine) resulted in improved UV detection. The utility of these reactions has prompted reviews of carbohydrate derivatization procedures used to both enhance separation and improve detection (precolumn)8 or solely to facilitate detection (postcolumn).9 Although frequently employed, derivatization reactions are often time-consuming and can add complexity to a sugar determination procedure. As a result, “universal” detection systems are often used to detect underivatized carbohydrate species. The universal refractive index detector has been used extensively for the determination of major sugars present in food and beverage samples. Olive plants, strawberries, apple juices, wines, legume seeds, and citrus juices are a few of the substances recently assayed for carbohydrates using this mode of detection.10-15 The (1) Lee, Y. C. J. Chromatogr. 1996, 720, 137-149. (2) Rassi, Z. E. J. Chromatogr. 1996, 720, 93-118. (3) Churms, S. C. J. Chromatogr. 1996, 720, 75-91. (4) Stefansson, M.; Westerlund, D. J. Chromatogr. 1996, 720, 127-136. (5) Chen, P.; Novotny, M. V. Anal. Chem. 1997, 69, 2806-2811. (6) Strydom, D. J. J. Chromatogr. 1994, 678, 17-23. (7) Alpert, A. J.; Shukla, M.; Shukla, A. K.; Zieske, L. R.; Yuen, S. W.; Ferguson, M. A. J.; Mehlert, A.; Pauly, M.; Orlando, R. J. Chromatogr. 1994, 676, 191-202. (8) Hase, S. J. Chromatogr. 1996, 720, 173-182. (9) Honda. S. J. Chromatogr. 1996, 720, 183-199. (10) Romani, A.; Baldi, A.; Tattini, M.; Vincieri, F. F. Chromatographia 1994, 39, 35-39. (11) Mangas, J. J.; Moreno, J.; Sua´rez, B.; Picinelli, A.; Blanco, D. Chromatographia 1998, 47, 197-202.
Analytical Chemistry, Vol. 73, No. 3, February 1, 2001 453
practical utility of refractive index detection, however, is generally limited to major sugar determinations because of poor sensitivity. When the vulnerability to minor mobile-phase and environmental changes is coupled to the poor sensitivity of refractive index systems, this mode of detection is often “the choice of last resort”.16 Another frequently employed HPLC mode of detection used for the determination of carbohydrate species is pulsed amperometry. Detection limits for this sensitive electrochemical technique are generally 2-3 orders of magnitude lower than those obtained using refractive index detection. One limitation of pulsed amperometric detection is that it requires electroactive analyte species. Since carbohydrates are weakly acidic, they are best detected by this method at an elevated pH. A mode of HPLC well suited for an alkaline mobile-phase requirement is HPAEC. As a consequence, pulsed amperometric detection coupled with HPAEC is one of the most frequently employed techniques used to determine carbohydrates in the literature. Groups have employed this method to determine sugars in legume seeds, citrus juices, instant coffees, streamwater, Maillard reaction products, and marine water.14,15,17-20 The increased sensitivity and practical utility of pulsed amperometric detection is offset by a few drawbacks. Perhaps the most serious disadvantage of this detection method is that electrochemical reaction products can deposit on the working electrode surface. Frequent electrode calibration is required as well as labor-intensive cleaning procedures.21,22 In general, there exists a need for a simple, sensitive detector for underivatized carbohydrate species. This need has driven the search for alternative modes of detection. Vonach et al.23 recently used Fourier transform infrared spectroscopy (FT-IR) as a molecular-specific detection system for carbohydrates following separation by HPLC. Herring and Piepmeier24 used an atmospheric pressure argon glow discharge to oxidize eluting carbohydrate species prior to their conductivity detection. Early work by Yoshida et al.25 demonstrated the potential utility of inductively coupled plasma atomic emission spectroscopy (ICP-AES) as a mode of universal detection for selected elements. Amino acids were detected after HPLC separation with a cation exchange column by simultaneously monitoring carbon and sulfur emission lines. Carbon emission detection limits of 30-50 µg/mL were reported for amino acids. This work was extended by Jinno et (12) Gomis, D. B.; Alvarez, M. D. G.; Alonso, J. J. M.; Vallina, A. N. Chromatographia 1988, 25, 701-706. (13) Calull, M.; Marce´, R. M.; Borrull, F. J. Chromatogr. 1992, 590, 215-222. (14) Frias, J.; Hedley, C. L.; Price, K. R.; Fenwick, G. R.; Vidal-Valverde, C. J. Liq. Chromatogr. 1994, 17 (11), 2469-2483. (15) White, D. R., Jr.; Widmer, W. W. J. Agric. Food Chem. 1990, 38, 19181921. (16) Ewing, G. W., Ed. Analytical Instrumentation Handbook, 2nd ed.; Marcel Dekker: New York, 1997; Chapter 22. (17) Bernal, J. L.; Del Nozal, M. J.; Toribio, L.; Del Alamo, M. J. Agric. Food Chem. 1996, 44, 507-511. (18) Gremm, T. J. Limnol. Oceanogr. 1997, 42, 385-393. (19) Ge, S. J.; Lee, T. C. J. Agric. Food Chem. 1996, 44, 1053-1057. (20) Borch, H. H.; Kirchmann, D. L. Marine Chem. 1997, 57, 85-95. (21) Strobel, H. A.; Heineman, W. R. Chemical Instrumentation: A Systematic Approach, 3rd ed.; John Wiley and Sons: New York, 1989; Chapter 26. (22) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development, 2nd ed.; John Wiley and Sons: New York, 1997; Chapter 3. (23) Vonach, R.; Lendl, B.; Kellner, R. Anal. Chem. 1997, 69, 4286-4290. (24) Herring, C. J.; Piepmeier, E. H. Anal. Chem. 1997, 69, 1738-1745. (25) Yoshida, K.; Hasegawa, T.; Haraguchi, H. Anal. Chem. 1983, 55, 21062108.
454
Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
al.26 to carbohydrate detection after microcolumn gel permeation chromatography. Although chromatograms were obtained, sugar detection limits (3σ) were still in the microgram range. The focus of the present study was to use a commercially available ICP-AES instrument as a sensitive, universal detector for carbon in carbohydrates following an HPLC separation procedure. The utility of the system was demonstrated through the determination of sugars in a variety of juice samples. LEC was selected as the mode of HPLC because it lacked many of the potential difficulties encountered by interfacing the other modes to the ICP. Both HILIC and HPAEC were not employed because of the frequent use of carbon-containing compounds during separation procedures. Organic mobile phases are required for a HILIC method while acetate “pushing agents” are typically used to elute analyte species from the column. RP-HPLC was not selected because of the difficulties encountered in separating structurally similar monosaccharides with just water as the mobile phase. EXPERIMENTAL SECTION Apparatus. A Hewlett-Packard series 1100 HPLC (Agilent Technologies, Palo Alto, CA) with vacuum degasser, quaternary pump, autosampler, thermostated column compartment, and photodiode array (PDA) detector separated the sugars. The flow rate was isocratic at 0.6 mL/min. Sugar samples (20 µL) were injected by the autosampler and separated by a 300-mm × 7.8mm Supelcogel calcium carbohydrate column (Supelco, Bellefonte, PA). A 50-mm × 4.6-mm Supelcogel calcium and C611 carbohydrate guard column preceded the separatory column and was connected by a 25-cm length of 0.178-mm-i.d. × 1.59-mmo.d. polyetheretherketone (PEEK) tubing. The column compartment was maintained at 80 °C. For comparison purposes, chromatograms were collected from the PDA at 193 and 254 nm. A 1-m segment of previously described PEEK tubing interfaced the chromatographic components of this system to the atomic emission detector. Using a shorter length would be advantageous to reduce band broadening; however, the proximity of the two instruments to each other determined the minimum length. Also, the inner diameter of the PEEK tubing should be minimized to prevent band broadening. A Leeman Labs Direct Reading Echelle ICP-AES generated the plasma emission (Hudson, NH). Column eluent was introduced through a Hildebrand grid nebulizer with argon at 45 psi. The spray chamber eliminated large particles from the resulting aerosol, which were evacuated by a peristaltic pump operating at 0.6 mL/min. Small, reproducible aerosol particles that successfully navigated the spray chamber flowed into the horizontally positioned plasma generated by 1.2 kW. Two types of detectors were evaluated. The first was a 0.35-m scanning monochromator (McPherson model 270, Acton, MA) with a photomultiplier tube (PMT). A fused-silica lens (25-mm diameter, 300-mm focal length) was placed 600 mm from the plasma so that it produced a 1:1 image of the plasma on the monochromator slit. The monochromator employed a 1200 grooves/mm grating, providing a reciprocal linear dispersion of 2 nm/mm. The entrance and exit slits were set at a width of 30 µm, providing a spectral band-pass of 0.06 nm at the carbon (26) Jinno, K.; Nakanishi, S.; Nagoshi, T. Anal. Chem. 1984, 56, 1977-1979.
Figure 2. Calibration curves for (0) sucrose, (2) glucose, and (O) fructose.
Figure 1. (a) UV-visible PDA. (b) ICP-AES PMT. (c) ICP-AES CCD. Chromatograms a and b are the same injection. The solution contained 4.8-6.2 mg/mL of each sugar (2.5 mg/mL carbon from each sugar). The solution for chromatogram c contained 7.2-9.0 mg/ mL each sugar (3.6 mg/mL carbon from each sugar).
emission wavelength (193.09 nm). A personal computer equipped with an A/D converter collected and stored the chromatogram. A data point was collected every second, using a 1-s time constant, and five successive points were averaged for each stored data point. This collection rate of one point every 5 s, is potentially too slow for many modern, high-efficiency applications, but it provides the best signal-to-noise ratio for the current system without significantly degrading the resolution. The second detector evaluated was a miniature CCD spectrometer (Ocean Optics, Inc., S2000, Dunedin, FL). The spectrometer was equipped with an 1800 grooves/mm grating, providing a reciprocal linear dispersion of 8.8 nm/mm, and a spectral band-pass of approximately 0.6 nm. A
fused-silica lens (25-mm diameter, 75-mm focal length) was placed 150 mm from the plasma producing a 1:1 image on the entrance slit of the spectrometer. The spectrometer was powered and controlled by a computer, and a complete spectrum was collected every 333 ms. Nine consecutive spectra were averaged and the result stored, providing a data point every 3 s in the chromatogram. The isocratic mobile phase used throughout this investigation was distilled and deionized water from a Millipore MILLI-Q system (Bedford, MA) with a resistivity of 18 MΩ/cm. The mobile phase was filtered (0.45 µm) before pouring into the HPLC solvent dispensing bottles. Water was placed in all four channels to reduce carbon contamination in the vacuum degasser. Reagents. A carbohydrate stock solution containing fructose, glucose, ribitol, melezitose, (Sigma, St. Louis, MO), and sucrose (Fisher, Fair Lawn, NJ) was prepared by dissolving these analytical grade reagents in distilled, deionized water. Calibration standards for these sugars were prepared from serial dilutions of the stock solution. Three different fruit juices were procured for this experiment: Tropicana Pure Premium Grovestand orange juice (lot 48EW0403) in a 64-oz paper carton, Oceanspray Cran-Grape Juice (lot 091200CT2210910CE) in a 64-oz plastic jug, and White House Apple Juice (lot 9250112030) in a 7-oz glass bottle. Samples were prepared by diluting 5 mL to 50 mL with distilled and deionized water, and the resultant mixture was then filtered through a 0.45µm syringe filter into a 1.5-mL crimp-top autosampler vial. Procedure. The column compartment was heated to 80 °C and the mobile phase was pumped through the column at 0.6 mL/ min for ∼2 h until a flat baseline was achieved. Equilibration occurred much more rapidly if the pump was continuously run at a reduced flow rate (0.1 mL/min) when the system was not in use. The PEEK tubing from the PDA flow cell was connected to the ICP nebulizer just prior to ignition of the plasma. The method was run with parameters as defined in the apparatus section with a 2-min postrun time to allow the column to equilibrate between injections. RESULTS AND DISCUSSION A mixture of test compounds was analyzed using HPLC-ICPAES. Three modes of detection were compared: UV-visible at 193 nm and ICP carbon emission at 193.09 nm using a PMT and Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
455
Figure 3. Three-dimensional UV-visible chromatogram of melezitose, sucrose, glucose, fructose, and ribitol (right to left). Above 210 nm, only fructose is minimally detected. Table 1. Detection Limits (µg)
melezitose sucrose glucose fructose ribitol
RI15
RI27
RI14
UV-visible PDA
ICP-AES PMT
ICP-AES CCD
5 5 5
60 60 30
45 130
0.008 0.007 0.007 0.002 0.003
0.05 0.05 0.05 0.06 0.05
0.2 0.2 0.2 0.2 0.3
a CCD. Resulting chromatograms are in Figure 1. Standards were made to contain equal concentrations of carbon for each sugar. Hence the peaks for ICP-AES are relatively the same height versus various heights for the UV-visible detector, which depends on the molar absorptivity of the sugar. The observed elution order is as expected since separations by the employed analytical column are based on both size exclusion and ion exchange mechanisms. The pores of the column exclude the larger saccharides more than the smaller ones. Melezitose is an oligosaccharide containing three saccharides, sucrose is a disaccharide, glucose and fructose are monosaccharides, and ribitol is a sugar alcohol. The separation of monosaccharides occurs due to different amounts of interaction with calcium ions inside the resin beads. The efficiency of this separation can be expressed by the number of theoretical plates based on the peak resulting from ribitol (N ) 11 700). This value is in agreement with that reported by the column manufacturer (N > 7000). A calibration curve spanning the concentration range 1-15 mg/mL was prepared for each sugar. These curves are shown in Figure 2. Note that the sensitivity (slope of the calibration curve) is nearly identical for each sugar, since they have very similar weight percents of carbon. In this case, a single calibration curve would suffice for all three sugars. Furthermore, if a calibration curve is prepared as carbon concentration versus emission signal, then that curve should work for all compounds, so long as they had similar nebulization efficiencies in the ICP. Then, if the identity of the analyte was known, one could calculate its concentration based upon its weight percent of carbon. Such a technique would 456 Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
be an efficient time-saving device that reduces preparation and instrument run time for calibration standards to one universal set. The limits of detection (LODs, 3σ) measured for the carbon emission detector and the UV-visible detector are compared to those reported in the literature for the refractive index detector (Table 1). The LOD observed for the carbon emission detector is nearly identical for each species, varying only slightly where the blank noise (σ) varied slightly. This amount (0.05 µg) should be the LOD for any compound containing ∼40% carbon. For comparison, Jinno et al.26 in 1984 reported a 2.8-µg LOD for raffinose using their carbon emission detector. Raffinose has the same molecular formula as melezitose. The carbon emission detector, with both the PMT and CCD, had significantly lower detection limits for sucrose, glucose, and fructose than those reported for refractive index detection. Even at the best refractive index detection limits reported (5 µg), the PMT was 100 times better, while the low-cost CCD was 25 times better. Surprisingly, the UV-visible PDA had nearly 1 order of magnitude lower detection limit than the carbon emission detector. The UV-visible detector, however, required operation well below 200 nm. Detection at higher wavelengths, such as the commonly used 254-nm line, would fail to detect any of the sugars. Figure 3 demonstrates the wavelength dependence of the UV-visible detector for sugars. Wavelengths below 190 nm, if accessible, should provide even better UV-visible sensitivity. The carbon emission detector, on the other hand, should provide 0.05-µg LODs for compounds with even smaller molar absorptivities than the sugars used in this example. Another advantage of the carbon emission detector over UVvisible is a cleaner sample chromatogram (Figure 4). This of course arises not from detector selectivity but from changes in relative sensitivity. A compound’s carbon emission signal is dependent directly upon its carbon content and not upon the strength of any chromophore. As a result, small but very highly absorbing molecules are effectively “selected against” with the (27) Shaw, P. E.; Wilson, C. W. J. Sci. Food Agric. 1983, 34, 109-113.
Table 2. Sugars in Juices (mg/mL/% RSD)a. juice apple crangrape orange a
detector
sucrose
glucose
fructose
UV-visible ICP-AES UV-visible ICP-AES UV-visible ICP-AES
1.4/5.5 2.1/3.2
1.6/5.4 3.6/0.7 8.5/0.3 7.7/0.6 1.9/4.8 2.4/2.5
6.6/0.2 5.1/1.9 8.0/0.4 6.6/3.7 1.8/2.0 2.2/4.2
5.2/2.2 6.4/3.8
All samples were diluted to 10%.
by carbon emission detection. The two-point baseline correction method was employed, and then this background was subtracted before the peaks were integrated in a spreadsheet program. These data were obtained without an extraction procedure to clean the samples and yet precision, expressed as percent relative standard deviation, was less than 6% in all cases. The chromatograms were based on three consecutive replicate measures of each juice sample. All juices were diluted 10:1 with distilled and deionized water.
Figure 4. Three consecutive replicates of orange juice: (a) UVvisible (193 nm); (b) ICP-AES (193.09 nm).
ICP system when compared to the conventional detection scheme. Real samples, such as fruit juice, may contain many species that may be present at very low concentration but that possess very high molar absorptivities. These species can pose significant matrix interference or peak crowding in the resulting chromatogram (Figure 4a). The carbon emission detector is relatively insensitive to such species and hence provides a simpler chromatogram (Figure 4b). The average sugar concentrations for three fruit juice samples were determined using the described HPLC-ICP-AES system. Results are presented in Table 2. Since baseline resolution was not achieved with the UV-visible detector, a line was manually drawn at the base of each peak prior to integration by the HP Chem Station software. Baseline resolution was nearly achieved
CONCLUSIONS A carbon emission detector may be the best option for analytes having low molar absorptivity in the UV-visible region, especially if these species are present in complex samples containing many species of high molar absorptivity. The analytes must be soluble in water since the mobile phase is restricted to 100% aqueous solution. Organic solvents, if employed, would cause a significant increase in the background carbon signal. The carbon emission detector will provide similar detection limits for compounds having similar mass percent carbon. This could lead to rapid and simple calibration using a single calibration curve for multiple analytes. The carbon emission detector also eliminates the requirement for pre- or postcolumn derivatization of those analytes that are difficult to detect by conventional means. ACKNOWLEDGMENT This work was funded by Leeman Labs, Inc., Hudson, NH 03051, and by grants from the National Science Foundation (CHE9710218) and the National Institutes of Health (1R41RR13245-01). Received for review August 3, 2000. Accepted November 15, 2000. AC000902I
Analytical Chemistry, Vol. 73, No. 3, February 1, 2001
457
