Mass Spectrometry Analysis

27 May 2005 - The caffeine content of three diet soft drinks (Diet Coke, Diet Cherry Coke, ... Topographical Imaging and Nanometer-Scale Resolved Prox...
0 downloads 0 Views 110KB Size
Anal. Chem. 2005, 77, 4385-4389

Quantitative Thin-Layer Chromatography/Mass Spectrometry Analysis of Caffeine Using a Surface Sampling Probe Electrospray Ionization Tandem Mass Spectrometry System Michael J. Ford,† Michael A. Deibel,‡ Bruce A. Tomkins, and Gary J. Van Berkel*

Organic and Biological Mass Spectrometry Group, Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6131

Quantitative determination of caffeine on reversed-phase C8 thin-layer chromatography plates using a surface sampling electrospray ionization system with tandem mass spectrometry detection is reported. The thin-layer chromatography/electrospray tandem mass spectrometry method employed a deuterium-labeled caffeine internal standard and selected reaction monitoring detection. Up to nine parallel caffeine bands on a single plate were sampled in a single surface scanning experiment requiring 35 min at a surface scan rate of 44 µm/s. A reversedphase HPLC/UV caffeine assay was developed in parallel to assess the mass spectrometry method performance. Limits of detection for the HPLC/UV and thin-layer chromatography/electrospray tandem mass spectrometry methods determined from the calibration curve statistics were 0.20 ng injected (0.50 µL) and 1.0 ng spotted on the plate, respectively. Spike recoveries with standards and real samples ranged between 97 and 106% for both methods. The caffeine content of three diet soft drinks (Diet Coke, Diet Cherry Coke, Diet Pepsi) and three diet sport drinks (Diet Turbo Tea, Speed Stack Grape, Speed Stack Fruit Punch) was measured. The HPLC/UV and mass spectrometry determinations were in general agreement, and these values were consistent with the quoted values for two of the three diet colas. In the case of Diet Cherry Coke and the diet sports drinks, the determined caffeine amounts using both methods were consistently higher (by ∼8% or more) than the literature values. Thin-layer chromatography (TLC) can be an adaptable and inexpensive approach to analytical or preparative separation challenges.1 As a result of its versatility and affordability, TLC is used routinely in the chemical and life sciences.1 Since its inception, the technology underlying TLC has changed significantly with considerable attention given to detection techniques * Corresponding author. Phone: 865-574-1922. Fax: 865-576-8559. E-mail: [email protected]. † Present address: Division of Systems Toxicology, National Center for Toxicological Research, Jefferson, AR 72079. ‡ Permanent address: Department of Chemistry, Earlham College, Richmond, IN 47374-4095. (1) Fried, B.; Sherma, J. Thin-Layer Chromatography, 4th ed.; Marcel Dekker: New York; 1999. 10.1021/ac050488s CCC: $30.25 Published on Web 05/27/2005

© 2005 American Chemical Society

capable of identifying the components present in separated bands.2,3 Coupling TLC with mass spectrometry (MS)3-5 pairs a very simple and robust separation method with a detector that exhibits the ability to detect selectively a very wide variety of analytes at trace to ultratrace levels. Highly specific quantitative TLC is one additional advantage made possible with the coupling of TLC to MS detection. There have been several quantitative TLC/MS methods reported that have used different methods for producing gas-phase analyte ions directly from the surface of a TLC plate.6-10 Many of these literature reports are lacking the analysis of real samples or lacking a comparison of the resulting quantitative values compared with those generated using other analytical techniques. Hercules et al.6 demonstrated low-level quantitation of cocaine on normal and reversed-phase TLC plates using matrix-assisted laser desorption/ ionization (MALDI)-MS and cocaine-d3 as an internal standard. The method was linear from 1 to 16 ng spotted with an estimated detection limit of 60 pg. However, they reported only the quantitation of analyte standards. Brown and Busch8 coupled TLC with fast atom bombardment-tandem mass spectrometry (MS/ MS) to quantify the diuretic amiloride hydrochloride at clinical concentrations spiked in urine. The method was linear over 2 orders of magnitude, and the estimated detection limits for data obtained using selected reaction monitoring (SRM) were comparable to existing literature HPLC/UV values (1-10 ng). Banno et al.9 reported a method for quantifying nicergoline, a drug used in the treatment of senile dementia, in which TLC was coupled with secondary ion mass spectrometry (SIMS). This TLC/SIMS method featured a chlorinated nicergoline internal standard and demonstrated a linear response from 50 to 5000 ng. Only standards were analyzed. Clench et al.10 have reported a quantitative TLC/ MALDI-MS method for the analysis of piroxicam, a nonsteroidal (2) Poole, C. J. Chromatogr., A 1999, 856, 399-427. (3) Poole, C. J. Chromatogr., A 2003, 1000, 963-984. (4) Busch, K. L. J. Chromatogr., A 1995, 692, 275-290. (5) Wilson I. D. J. Chromatogr., A 1999, 856, 429-442. (6) Nicola, A. J.; Gusev, A. I.; Hercules D. M. Appl. Spectrosc. 1996, 50, 14791482. (7) Li, L.; Lubman, D. M. Anal. Chem. 1989, 61, 1911-1915. (8) Brown, S. M.; Busch K. L. J. Chromatogr. 1991, 4, 189-193. (9) Banno, K.; Matsuoka, M.; Takahashi, R. Chromatographia 1991, 32, 79181. (10) Crecelius, A.; Clench, M. R.; Richards, D. S.; Parr, V. J. Pharm. Biomed. Anal. 2004, 35, 31-39.

Analytical Chemistry, Vol. 77, No. 14, July 15, 2005 4385

antiinflammatory, making use of a structural analogue internal standard predeveloped over the plate. The calibration curve with best precision was linear from 400 to 800 ng. We report here a quantitative method for the determination of caffeine using a surface sampling probe/electrospray (ES) ionization system to couple TLC with tandem mass spectrometric detection. We have previously reported on this TLC/ES-MS coupling for the qualitative detection and characterization of a variety of different analytes separated on hydrophobic reversedphase C8 and C18 plates.11-13 Caffeine was chosen for this quantitation demonstration for three reasons. First, it is a small polar molecule that is particularly suited both to positive ion electrospray ionization and to separation on reversed-phase TLC. Second, it is an ingredient in a wide variety of commercial products (e.g., beverages). Finally, a low-cost stable isotope internal standard is available. An HPLC method with UV detection was developed in parallel as a check on the TLC/ES-MS/MS results. The method was applied to a small selection of diet colas and diet sport drinks. The quantitative results, limits of detection, dynamic ranges, and other figures of merit are calculated, compared, and contrasted between the two quantitative methods. EXPERIMENTAL SECTION Chemicals. HPLC grade methanol and water were purchased from Burdick & Jackson (Muskegon, MI). Acetic acid (doubly distilled, PPB/PTFE grade) was from Sigma Aldrich (Milwaukee, WI). Caffeine, CAS registry no. [58-08-02], 99% purity was obtained from Sigma Aldrich. Caffeine-d3 (1-methyl-d3), CAS registry no. [2635-03-11], 99.8% atom D was purchased from C/D/N Isotopes, Inc. (Pointe Claire, PQ, Canada). Standards. Standards for the HPLC determinations containing 1.0, 2.5, 5.0, 10, 25, and 50 ng of caffeine/µL were prepared by diluting a 5000 ng of caffeine/µL caffeine stock solution with 50/ 50 (v/v) methanol/water. The standards prepared for the TLC/ ES-MS/MS determinations contained the same concentrations as those for HPLC analyses, with an additional 50 ng/µL caffeine-d3 as the internal standard. Samples. Six beverages were analyzed for their caffeine content. One single-serving can (355 mL) each of Diet Coke, Diet Cherry Coke (Coca-Cola Co., Atlanta, GA), and Diet Pepsi (PepsiCo, Inc., Purchase, NY) and one single-serving bottle (532 mL) each of Diet Turbo Tea, Speed Stack Grape, and Speed Stack Fruit Punch (American Body Building Products, Walterboro, SC) were purchased over the counter locally. Prior to use, the carbonated cola samples were degassed for ∼15 min in an ultrasonic bath. Portions of the three colas and Diet Turbo Tea were diluted 1/10 and the two Speed Stack drinks were diluted 1/25, respectively, with 50/50 (v/v) methanol/water. Aliquots of these dilutions were filtered through a 0.2-µm Teflon syringe filter and stored at -20 °C prior to analysis. The samples prepared for the TLC/ES-MS/MS determinations were prepared using the same dilutions as those for HPLC analyses, with an additional 50 ng/µL caffeine-d3 as the internal standard. Thin-Layer Chromatography. TLC was performed using 10 × 10 cm glass-backed RP C8 TLC plates (P/N 13725/5, EM (11) Van Berkel, G. J.; Sanchez, A. D.; Quirke, J. M. E. Anal. Chem. 2002, 74, 6216-6223. (12) Ford, M. J.; Van Berkel, G. J. Rapid Commun. Mass Spectrom. 2004, 18, 1303-1309. (13) Ford, M. J.; Van Berkel, G. J.; Kertesz V. J. Mass Spectrom. In press.

4386

Analytical Chemistry, Vol. 77, No. 14, July 15, 2005

Science, Gibbstown, NJ.). TLC plates were developed using 70/ 30 (v/v) methanol/water. Samples and standards were spotted manually with a micropipet (1.0 µL), horizontally spaced (1-cm separation) in parallel development lanes. The plates were developed until the solvent front was ∼4 cm from the origin (caffeine Rf ≈ 0.45). Developed plates were dried in an oven at 110 °C for 15 min just prior to analysis. Surface Sampling System. The surface sampling principle and experimental setup have been described in detail elsewhere.11-13 Briefly, a surface sampling electrospray probe was mounted on a 4000 QTrap hybrid triple quadrupole linear ion trap mass spectrometer (MDS Sciex, Concord, ON, Canada). A robotic x,y,z sample handling platform (Applied Scientific Instrumentation Inc., Eugene, OR), modified to accommodate a 10 × 10 cm TLC plate, held and positioned the plate in the x,y plane perpendicular to the z plane of the sampling probe. A syringe pump (Harvard Apparatus, Holliston, MA) delivered the 60/40 (v/v) methanol/ water (0.1% acetic acid) eluting solvent to the surface sampling probe from a 1.0-mL glass syringe at 15 µL/min. A short-wavelength UV lamp (λ ) 214 nm) was used to locate the caffeine bands on the TLC plate, and a marker pen was used to draw a black line through the bands on the back of the glassbacked TLC plate. Supplemental illumination used to visualize the sampling probe liquid microjunction with the plate produced a silhouette of the sampling probe tip that could be observed through the glass-backed TLC plate. The silhouette of the sampling probe was aligned with the black line marked through the caffeine bands and a liquid junction formed at a position near the plate edge at the Rf value where the caffeine bands appeared. The surface was scanned under computer control horizontally relative to the stationary sampling probe across the caffeine bands in parallel development lanes at a constant scan rate of 44 µm/s. The z position of the surface, which controlled the liquid microjunction thickness, was operated manually to maintain a 20-50µm thickness for optimum surface sampling. In this manner, as many as nine 1-cm equidistant parallel caffeine bands across a 10 × 10 cm TLC plate could be analyzed in 35 min. Mass Spectrometric Analysis. The mass spectrometer was operated in SRM mode using Analyst Software version 1.4. The detection of caffeine and caffeine-d3 was optimized by aspirating through the surface sampling probe an equimolar mixture of caffeine and caffeine-d3 (0.1 µM each). The optimized instrument parameters were as follows: ion spray voltage, 4.5 kV; declustering potential, 85 V; exit potential, 5 V; collision energy, 30 eV; collision exit potential, 10 V; collision activation gas density, medium; nebulizing gas, 21 (arbitrary units); countercurrent gas, 15 (arbitrary units). The SRM transitions monitored were m/z 195 f 138 and 195 f 110 for caffeine and m/z 198 f 138 and 198 f 110 for caffeine-d3. The dwell time for each transition was 250 ms. Following completion of the surface scan, the areas under these peaks were integrated using the quantitation mode of the Analyst Software package. HPLC/UV. All HPLC determinations of caffeine were performed using an Agilent 1100 capillary LC system equipped with a continuous degasser, capillary LC pump, autosampler, column heater, and diode array detector. System operation and data processing were carried out using ChemStation Revision A.10.01 software. Prior to use, the HPLC eluent (50/50/0.5 (v/v/v)

Table 1. Caffeine Determined in Six Beverages Using HPLC/UV and TLC/ES-MS/MS caffeine per container (mg) HPLC/UV

TLC/MS

sample

lit.a

mean

mean vs lit. (%)

std dev

RSD (%)

replicates

mean

mean vs lit. (%)

std dev

RSD (%)

replicates

Diet Coke Diet Pepsi Diet Cherry Coke Diet Turbo Tea Speed Stack Grape Speed Stack Fruit Punch

45 36 34 90 250 250

46.0 35.0 36.9 120.8 276.2 284.4

+2.2 -2.8 +8.5 +34 +10 +14

0.3 0.2 0.3 0.7 0.7 0.7

0.7 0.7 0.9 0.6 0.2 0.25

4 4 4 4 4 4

43.2 34.7 38.0 119.8 270.0 278.0

-4.0 -3.6 +12 +33 +8.0 +11

0.7 0.8 0.8 0.8 11 4.3

1.6 2.3 2.1 0.7 1.5 4.4

3 3 3 3 3 3

a Literature values for Diet Coke, Diet Pepsi, and Diet Cherry Coke taken from ref 15. Quoted values for other beverages taken from the manufacturer’s label.

methanol/water/acetic acid) was filtered through a 0.45-µmporosity nylon filter. Caffeine (0.5-µL injection) was eluted at a flow rate of 25 µL/min from a 250 × 1 mm column packed with Partisil ODS-3 (C18, 5-µm particle diameter, Thermo Electron Corp., Bellefonte, PA). Under these conditions, the solvent front was detected at ∼7 min and caffeine eluted at ∼9.3 min. UV spectra were collected between 190 and 400 nm for all peaks exhibiting a minimum absorbance of 0.250 mAU. Caffeine quantitation was based on the absorbance measured at 275 nm using a 360-nm absorbance reference. Four aliquots of each diluted and filtered standard or sample were analyzed. The peak area for each replicate trial was determined using the default integration parameters of the instrument, and the average of all four trials was used for quantitation. RESULTS AND DISCUSSION Caffeine Determined by HPLC. A method featuring HPLC separation with UV detection was employed to establish “reference” values for the amount of caffeine in the six beverages evaluated for comparison with both label or literature values and the TLC/ES-MS/MS determinations. HPLC methods for quantitation of caffeine other than that used here are described in the literature.14 Caffeine content values for the cola beverages listed in Table 1 were obtained from the web site of the American Beverage Association,15 while those for the “sport drinks” were taken from the manufacturer’s label. We do not know the error associated with the label and literature values for the caffeine content, nor the details of the methods used to determine those values. A typical chromatogram from our HPLC method for a 0.5-µL injection of a 50 ng/µL caffeine standard is shown in Figure 1. The HPLC calibration data from a series of standards injected (1.0-50 ng/µL) were evaluated using a least-squares regression and fit the model y ) (111.31 ( 0.18)x + (-5.2 ( 2.1), where y is the integrated peak area and x is the amount (ng) of caffeine injected (r2 > 0.9999). From this calibration curve, the detection limit was estimated (3sx/y/slope, where sx/y, the standard error of the y value estimates is assumed to approximate the standard (14) Kusch, P.; Knupp, G. Chem. Educ. 2003, 8, 201-205. (15) American Beverage Association, Nutrition & Health, Ingredients, Caffeine. http://www.nsda.org/health/caffeinecontent.asp. Last accessed January 13, 2005.

Figure 1. HPLC/UV chromatogram (275 nm) from a 2.5-ng injection of a caffeine standard. Conditions described in the text. The UV spectrum of caffeine (inset) was taken from the chromatographic peak at 9.3 min.

deviation of the blank, sB) to be 0.20 ng injected.16 Table 1 shows the caffeine values for the six representative caffeine-containing beverages calculated from four replicate determinations. The standard deviations on the calculated values reflect the error on the replicate analyses and were all less than 1% RSD. The agreement between the expected and HPLC/UV measured caffeine amounts for Diet Coke and Diet Pepsi were within 3% of the expected values. In the case of Diet Cherry Coke, the HPLC/UV method was higher (+8.5%) than the expected value. In the case of the diet sports drinks, the determined caffeine amounts were considerably higher (by 10% or more) than the label values. Diet Turbo Tea was the extreme case in which the HPLC/ UV value was 34% higher than the label value. These deviations from the expected values demanded further investigation. A matrix affect assessment was made with a series of spike recovery experiments. For each sample type, a spike recovery was performed by adding a caffeine spike (final concentration 10 ng/ µL) to an aliquot of the diluted sample prior to filtration. Spike recoveries of caffeine ranged between 97 and 105% and were consistent with the simple sample preparation performed. These results confirm the absence of any contribution from the sample matrix or the analytical procedure itself to the quantitative determination of caffeine. The UV spectrum taken from the peak at 9.3 min in the Diet Turbo Tea chromatogram matched the spectrum of the caffeine in the standards, which is shown as an inset in Figure 1. An instrumental evaluation of “peak purity” (16) Miller, J. C.; Miller, J. N. Statistics for Analytical Chemistry, 2nd ed.; Ellis Horwood Ltd: Chichester; 1988; pp 110-115.

Analytical Chemistry, Vol. 77, No. 14, July 15, 2005

4387

Figure 2. Structure and mass-to-charge ratio for the protonated molecules of caffeine (m/z 195) and caffeine-d3 (m/z 198) as well as proposed fragmentation giving rise to major product ion (m/z 138) used in the quantitative SRM transition for both compounds.

clearly demonstrated that there were no additional components that had coeluted with caffeine that would contribute a bias to the results in Table 1. As the TLC/ES-MS/MS analyses gave similar results (see below), we suspect the amount of caffeine in these diet sports drinks, which contain caffeine from plant extracts, is greater than the amount indicated on the manufacturer’s label. Caffeine Determined by TLC/ES-MS/MS. SRM with an isotopically labeled internal standard is the accepted practice for accurate and precise quantification of target compounds by MS/ MS.17 Addition of an internal standard provides a means to compensate for minor variations in sample preparation and instrument parameters, thereby improving the accuracy of the quantitation. In the present work, the internal standard also allowed for reliable quantitation when the analyte bands on the TLC plate were not sampled from within the same regions of the bands among the bands in the separate development lanes (i.e., all bands not sampled “dead center”). The TLC/ES-MS/MS method uses SRM and caffeine-d3 as an internal standard for quantitative analysis of caffeine. Caffeine and caffeine-d3 both are observed as protonated molecules, m/z 195 and 198, respectively, under positive ion ES-MS conditions. The base peak in the respective product ion spectra of the protonated molecules was observed at m/z 138 (collision energy, 30 eV). The second most abundant product ion was m/z 110 for both caffeine and caffeine-d3. The fragmentation mechanisms leading to these product ions have been examined by others.18 Thus, the SRM transitions m/z 195 f 138 (caffeine) and 198 f 138 (caffeine-d3) were used for quantitation (Figure 2), while the abundance (∼20% of the former transitions) for the transitions m/z 195 f 110 (caffeine) and 198 f 110 (caffeine-d3) were used to further confirm the detection of caffeine. Panels a and b in Figure 3 show examples of SRM chromatograms used for quantitation (m/z 195 f 138 (caffeine, blue line) and 198 f 138 (caffeine-d3, red line)) that were obtained from the analysis of nine and eight caffeine bands, respectively, in parallel development lanes. The data set in Figure 3a shows the results from the analysis of six caffeine standard bands (1.0, 2.5, (17) Willoughby, R. C.; Sheehan, E. W.; Mitrovich, S. M. A Global View of LC/ MS; Global View Publishing: Pittsburgh, PA, 2002. (18) Thevis, M.; Opfermann, G.; Krug, O.; Scha¨nzer, W. Rapid Commun. Mass Spectrom. 2004, 18, 1553-1560.

4388 Analytical Chemistry, Vol. 77, No. 14, July 15, 2005

Figure 3. (a) TLC/ES-MS/MS SRM ion current chromatograms (caffeine m/z 195 f 138, blue trace; caffeine-d3 m/z 198 f 138, red trace) from the analysis of (from left to right) 1.0-, 2.5-, 5.0-, 10-, 25-, and 50-ng caffeine standard bands and Diet Pepsi (DP), Diet Coke (DC), and Diet Cherry Coke (DCC) sample bands. (b) TLC/ES-MS/ MS SRM ion current chromatograms (caffeine m/z 195 f 138, blue trace; caffeine-d3 m/z 198 f 138, red trace) from the analysis of (from left to right) a 50-ng caffeine standard, three replicate DCC sample bands, a 50-ng caffeine standard, and three replicate DC sample bands. All sample and standard solutions spotted contained 50 ng of caffeine-d3 as an internal standard.

5.0, 10, 25, and 50 ng spotted) and one spot each of three beverage sample bands (Diet Coke, Diet Pepsi, Diet Cherry Coke) all of which were spiked with 50 ng of caffeine-d3. The data collection for these nine bands was completed in 35 min. Figure 3b shows the results from the analysis of three bands each from Diet Cherry Coke and Diet Coke samples and two 50-ng caffeine standard bands, all of which were spiked with 50 ng of caffeine-d3. Equivalent data to that in Figure 3b were collected for all six beverage samples. The caffeine standard data in Figure 3a were also collected in triplicate for construction of a calibration curve. The TLC/ES-MS/MS calibration data were evaluated using a least-squares regression and fit the model y ) (1.094 ( 0.005)x + (-0.0024 ( 0.0023), where y is the ratio of the integrated analyte peak area to the integrated internal standard peak area and x is the caffeine standard amount to internal standard amount ratio (r2 > 0.9998). From this calibration curve, the detection limit was estimated (3sx/y/slope) to be 1.0 ng spotted on the plate.16 This detection level is a factor of ∼5 poorer than that for our HPLC/ UV method. However, one must consider that the 0.635-mm-wide sampling probe actually samples material from only ∼16% or less of the area of the near-circular bands (∼5-mm diameter) on the plate.11 In any case, this detection level posed no problem in the determination of the caffeine content of the beverages under study,

which, when diluted by over 1 order of magnitude for analysis, still contained 10 ng/µL or more caffeine. The data in Table 1 for the six beverages were calculated with the regression equation above. As with the HPLC/UV data, the standard deviations listed were calculated from the replicate analyses. Compared to the HPLC data, the precision of the TLC/ ES-MS/MS determinations (reflected by the RSD) was poorer, but still less than 2.5% RSD, except for the case of Speed Stack Fruit Punch (4.4% RSD). This poorer RSD may relate to manual spotting of samples and to the fact that replicate standards for the calibration curve and replicate samples were run on different plates. Consistent with the HPLC/UV data, the mean caffeine value determined for Diet Coke and Diet Pepsi were very close to the expected values (within 4%). Also consistent with the HPLC/ UV data, the values determined for Diet Cherry Coke and each of the sports drinks were significantly higher than the expected values. For these samples, the HPLC/UV data and TLC/ES-MS/ MS values were more comparable to one another than to the expected values. This close agreement between the caffeine content of these beverages using two completely independent analytical procedures strongly suggests that they may be closer to the true values. The actual error in the literature determination is unknown, so a more critical discussion of the accuracy of our determination is not possible. Furthermore, it is not possible to evaluate the reliability of the label values for the caffeine content of the three sport drinks without a more detailed understanding of the analytical method(s) used to generate those values, as well as their uncertainties. We strongly suspect, but cannot prove, that the quality control and analytical procedures employed for quantifying caffeine in the colas were substantially more sophisticated and tightly controlled than that for the sport drinks, which contained caffeine supplied from one or more plant extracts. The reason for the discrepancy with the literature value for Diet Cherry Coke was not clear, but was consistent with the value determined by HPLC/UV. A matrix affect assessment for the TLC/ES-MS/MS was made with a spike recovery experiment. For the Diet Cherry Coke sample, a spike recovery was performed by adding a caffeine spike (final concentration, 10 ng/µL) to an aliquot of the diluted sample prior to filtration. The spike recovery of caffeine was 106%, confirming the absence of any significant contribution from the sample matrix or the analytical procedure itself to the quantitative determination of caffeine. CONCLUSIONS We demonstrated here that quantification of caffeine at the low-nanogram level can be performed directly from a surface of

a TLC plate using a surface sampling probe, ES mass spectrometry system employing SRM detection, and an isotopically labeled internal standard spotted with the samples. The resulting calculated analyte concentrations exhibited accuracy comparable to the HPLC/UV method that was developed in parallel. The 50-fold linear range of the TLC/ES-MS/MS procedure, ranging between 1.0 and 50 ng of caffeine spotted on a TLC plate, allowed this analyte to be quantified successfully in six commercially available beverages using only minimal sample preparation. The limit of detection (1.0 ng spotted) and 50-fold dynamic range compare very well with the best quantitative TLC/MS methods published to date using other sampling/ionization methods. Few of those prior methods actually analyzed real samples and all required relatively involved postseparation processing of the plates, which is avoided in the present method. Certainly the quantification of other analytes with suitable isotopically labeled internal standards would be possible with this surface sampling system. Future improvements to the precision and accuracy of the measurements might be obtained though the analysis of all standards and samples on a single TLC plate and the use of an automated sample spotting system. ACKNOWLEDGMENT M.J.F. acknowledges an ORNL appointment through the ORNL Postdoctoral Research Associates Program. M.A.D. acknowledges an appointment to the U.S. Department of Energy (DOE) Higher Education Research Experiences (HERE) Program for Faculty at the Oak Ridge National Laboratory (ORNL), administered by the Oak Ridge Institute for Science and Education with support from the Oak Ridge Science Semester Program administered by Denison University, Granville, OH. The micro ion spray head, associated positioner, and ion source mount were supplied by MDS Sciex through a Cooperative Research and Development Agreement (CRADA ORNL02-0662). Initial development of the TLC/ES-MS readout system was sponsored by the Laboratory Directed Research and Development Program of ORNL. ORNL Technology Transfer and Economic Development (TTED) Royalty Funds supported the current project. ORNL is managed by UTBattelle, LLC for the U.S. Department of Energy under Contract DE-AC05-00OR22725.

Received for review March 23, 2005. Accepted April 28, 2005. AC050488S

Analytical Chemistry, Vol. 77, No. 14, July 15, 2005

4389