Anal. Chem. 2004, 76, 2869-2877
Derivatization for LC-Electrospray Ionization-MS: A Tool for Improving Reversed-Phase Separation and ESI Responses of Bases, Ribosides, and Intact Nucleotides Anders Nordstro 1 m,† Petr Tarkowski,†,‡ Danuse Tarkowska,†,‡ Karel Dolezal,‡ Crister A ° stot,§ Go 1 ran Sandberg,† and Thomas Moritz*,†
Umeå Plant Science Center, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, SE-901 87 Umeå, Sweden, Laboratory of Growth Regulators, Palacky University and Institute of Experimental Botany ASCR, Slechtitelu 11, 783 71, Olomouc, Czech Republic, and The Swedish Defence Research Agency (FOI), NBC Defence, 901 82 Umeå, Sweden
We have developed a method for analyzing polar compounds by reversed-phase LC-ESI-MS following esterification of the analytes’ free hydroxyl groups with propionyl or benzoyl acid anhydride. The method was applied to members of the plant hormone group cytokinins, which includes adenine bases, ribosides/glycosides, and nucleotides substituted at N-6 with an isoprenoid side chain, spanning a wide range of polarity. It was also used to analyze other compounds of biological importance, e.g., the nucleotides AMP, ADP, and ATP. The formation of more hydrophobic derivatives had a significant impact on two aspects of the analysis. The retention on a reversedphase material was greatly increased without the use of any acetate/formate buffer or ion pairing reagent, and the ESI response was enhanced, due to the higher surface activities of the derivatives. Detection limits of propionylated cytokinins were in the high-attomole to low-femtomole range, an improvement by factors of 10-100 compared to previously reported figures. Using an automated SPE-based purification method, 12 endogenous cytokinins were quantified in extracts from 20- to 100mg samples of leaves (from the plant Arabidopsis thaliana) with high accuracy and precision. Furthermore, the chromatographic properties of the benzoylated AMP, ADP, and ATP in the reversed-phase LC-MS system were much better in terms of retention, separation, and sensitivity than those of their underivatized counterparts, even without the use of any ion pairing reagent. Our data show that derivatization followed by LC-ESI-MS is an effective strategy for analyzing low molecular weight compounds, enabling compounds with a wide range of polarity to be determined in a single-injection LC-MS analysis. The development of sensitive analytical methods for determining trace levels of different compounds has become essential for * Corresponding author. E-mail:
[email protected]. † Swedish University of Agricultural Sciences. ‡ Palacky University and Institute of Experimental Botany ASCR. § The Swedish Defence Research Agency (FOI). 10.1021/ac0499017 CCC: $27.50 Published on Web 04/20/2004
© 2004 American Chemical Society
elucidating the role and function of specific metabolites in diverse biological systems. However, many biochemical events are not determined merely by the concentration of a single compound but also by rates of biosynthesis and catabolism. Therefore, it is important to be able to analyze and quantify entire metabolic pathways in biological samples rather than just single compounds. However, analysis of these pathways often demands a method that can handle compounds ranging from hydrophobic to polar or ionic. To resolve fundamental biological questions, the need to analyze individual tissues, cells, and even different cell compartments has also become apparent. Furthermore, the analytical methodology must be not only sensitive but also rapid since sample throughput demands are increasing. Combined liquid chromatography-electrospray ionization-mass spectrometry (LC-ESI-MS) has become the analytical tool of choice for identification and quantification of low molecular weight compounds. LC-MS analysis combines the separation capacities of the LC system with the sensitivity and (especially with MS/ MS) the specificity of detection provided by MS systems. However, two major problems are associated with LC-MS analysis. First, no universal LC column packing material can be used for all possible kinds of analytes. Second, no eluent system is compatible with both all possible analytes and ESI. A compromise has to be made at some level regarding the packing material, eluent system, or analyte response. Positive ESI is an excellent interface for reversed-phase chromatography. The solvents used, acidic acetonitrile/MeOH-H2O mixtures, are suitable for the desolvation process. However, many biologically important compounds do not separate readily on reversed-phase packing material due to their high polarity. The need for acetate or formate buffers or ion pairing reagents often hampers the ESI response.1 Derivatization chemistry has an important history in separation science, and with that knowledge in mind, an alternative would be to make the analytes more hydrophobic, thus improving both the reversedphase HPLC separation and the ESI process. Nonpolar analytes can separate well with higher proportions of organic modifiers in the eluent, and for positive ESI, both a stable spray and LC effluents with suitable evaporative properties can be obtained with (1) Temesi, D.; Law, B. LC GC North Am. 1999, 17, 626-632.
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004 2869
50% organic modifier.2 Furthermore, it has been shown that it is possible to predict ESI responses from chromatographic retention times and that increasing the hydrophobicity increases the ESI response.3 Increasing the retention time also avoids potential problems associated with suppression effects due to salts and interfering compounds eluting early in the chromatographic analysis. The analytes’ affinity for the droplet surface (surface activity) strongly affects the ESI response,4,5 and several studies have shown that derivatization can increase this response. For example, Okamoto et al.6 increased the response for maltopentaose 5000-fold by reacting it with trimethyl (p-aminophenyl)ammonium, while Bleicher and Bayer7 showed that the retention of an oligonucleotide could be greatly enhanced by adding a dimethoxytrityl group at the 5′ end, thus boosting the ESI response. LC-ESI-MS analyses of nucleotides usually involve reversedphase separation, because of the problems that the nonvolatile buffers used in anion-exchange separations would cause.8 Although ion pair separation of nucleotides9-11 and oligonucleotides12 on various types of reversed-phase materials has been used in combination with ESI-MS, ion pair agents can suppress the ESI signal. Different strategies to circumvent this problem have been tested,13,14 but it is still important to be aware that improvements in the separation and signal intensity for one analyte might have opposite effects for another analyte. Therefore, single-run LCMS analysis of such important types of compounds as adenine bases, nucleosides, and nucleotides presents serious difficulties. Cytokinins are an important class of growth-promoting substances, controlling a wide range of growth and development processes in plants.15 The most intensively studied types of cytokinins are adenine bases substituted at N-6 with an isoprenoid side chain.16 Cytokinins occur as free bases, ribosides (nucleosides), nucleotides, and glucosides, usually at very low concentrations (0.1-100 pmol g-1 of fresh weight). Their wide range of physicochemical properties combined with their low abundance makes analysis of cytokinins a difficult task. Several different LCMS methods for analysis of cytokinins have been described, including LC-ESI-MS/MS,17,18 LC-frit-FAB-MS/MS,19 and LC-ESI(2) Cech, N. B.; Enke, C. G. Mass Spectrom. Rev. 2001, 20, 362-387. (3) Cech, N. B.; Krone, J. R.; Enke, C. G. Anal. Chem. 2001, 73, 208-213. (4) Tang, L.; Kebarle, P. Anal. Chem. 1993, 65, 3654-3668. (5) Cech, N. B.; Enke, C. G. Anal. Chem. 2001, 73, 4632-4639. (6) Okamoto, M.; Takahashi, K. I.; Doi, T. Rapid Commun. Mass Spectrom. 1995, 9, 641-643. (7) Bleicher, K.; Bayer, E. Chromatographia 1994, 39, 405-408. (8) Huber, C. G.; Oberacher, H. Mass Spectrom. Rev. 2001, 20, 310-343. (9) Witters, E.; VanDongen, W.; Esmans, E. L.; VanOnckelen, H. A. J. Chromatogr., B 1997, 694, 55-63. (10) Aussenac, J.; Chassagne, D.; Claparols, C.; Charpentier, M.; Duteurtre, B.; Feuillat, M.; Charpentier, C. J. Chromatogr., A 2001, 907, 155-164. (11) Tuytten, R.; Lemiere, F.; Van Dongen, W.; Esmans, E. L.; Slegers, H. Rapid Commun. Mass Spectrom. 2002, 16, 1205-1215. (12) Huber, C. G.; Krajete, A. Anal. Chem. 1999, 71, 3730-3739. (13) Apffel, A.; Fischer, S.; Goldberg, G.; Goodley, P. C.; Kuhlmann, F. E. J. Chromatogr., A 1995, 712, 177-190. (14) Kuhlmann, F. E.; Apffel, A.; Fischer, S. M.; Goldberg, G.; Goodley, P. C. J. Am. Soc. Mass Spectrom. 1995, 6, 1221-1225. (15) Mok, M. C. In Cytokinins Chemistry, Activity and Function; Mok, D. W. S., Mok, M. C. Eds.; CRC: Boca Raton, FL, 1994; Chapter 12. (16) Straw, G. I, In Cytokinins Chemistry, Activity and Function; Mok, D. W. S., Mok, M. C. Eds.; CRC: Boca Raton, FL, 1994; Chapter 2. (17) Prinsen, E.; Redig, P.; Vandongen, W.; Esmans, E. L.; Vanonckelen, H. A. Rapid Commun. Mass Spectrom. 1995, 9, 948-953.
2870
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
MS,20 but none of these methods has the capacity to analyze all types of cytokinins in a single chromatographic run. As plant extracts are very complex, the sample preparation prior to the final MS analysis is also extremely important. Immunoaffinity or solid-phase extraction (SPE) purification methods are often used, usually separating nucleotides from ribosides and free bases. Sample purification involving only C18- and mixed-mode SPE cartridges have been reported,21 and one advantage of this type of purification is the possibility it offers to automate and thereby increase the sample throughput. In this report, we show that cytokinins can be analyzed from minute amounts of plant material, using an automated SPE purification system, combined with LCMS/MS analysis. By using a derivatization procedure the ESI signal and chromatographic behavior of the cytokinins were dramatically improved. Similar improvements were found for a number of other biochemically important compounds spanning a range of polarity, e.g., AMP, ADP, and ATP, showing that the principle used for cytokinin analyses can also be applied to other types of compounds. EXPERIMENTAL SECTION Chemicals. Stable isotope-labeled cytokinins were purchased from Apex Organics (Devon, U.K.). Stock solutions (500 ng µL-1) were prepared using an analytical balance (Mettler AT 20, resolution 2 µg) by dissolving the crystallized cytokinins in dimethylformamide. Ado, AMP, ADP, ATP, FMN, and UDPG were all obtained from Sigma (St. Louis, MO). Stock solutions of AMP, ADP, and ATP (500 ng µL-1) were prepared in water. Ado, FMN, and UDPG stock solutions (500 ng µL-1) were prepared in methanol.water (1:1 v/v). For derivatization, the following reagents were used: 99% propionic anhydride (Aldrich, Milwaukee, WI), 95% benzoic anhydride (Sigma), and 98-99% N-methylimidazole (Sigma-Aldrich, Steinheim, Germany). In addition, the following solvents, acids, and bases were used: HPLC grade acetonitrile (Fisher), HPLC grade methanol (J.T. Baker, Denventer, Holland), Milli-Q Plus water (Millipore, Bedford, MA), 98% formic acid (J.T. Baker), and 25% ammonium hydroxide (J.T. Baker. Sample Purification and Derivatization. Arabidopsis thaliana wild-type variant Colombia plants were grown under short-day conditions (9-h photoperiod/day) with fluorescent light. After three months, they were harvested and ground in liquid N2 using a mortar and pestle and then stored in a -80 °C freezer until analysis. Samples of 20-100 mg were individually placed, with 1 mL of extraction medium (MeOH/H2O/HCOOH, 750:200:50) including stable isotope internal standards, into 1.5-mL Eppendorf tubes. They were then extracted using an MM 301 vibration mill (Retsch GmbH & Co. KG, Haan, Germany) at a frequency of 30 Hz for 2 min after adding 3-mm tungsten carbide beads (also from Retsch) to each tube to increase the extraction efficiency. The tubes were then placed in a -20 °C freezer for 2 h. After centrifugation in an Eppendorf centrifuge (model 5417C) for 10 min at 14 000 rpm, the supernatants were transferred to glass (18) Witters, E.; Vanhoutte, K.; Dewitte, W.; Machackova, I.; Benkova, E.; Van Dongen, W.; Esmans, E. L.; Van Onckelen, H. A. Phytochem. Anal. 1999, 10, 143-151. (19) Astot, C.; Dolezal, K.; Moritz, T.; Sandberg, G. J. Mass Spectrom. 1998, 33, 892-902. (20) Novak, O.; Tarkowski, P.; Tarkowska, D.; Dolezal, K.; Lenobel, R.; Strnad, M. Anal. Chim. Acta 2003, 480, 207-218. (21) Dobrev, P. I.; Kaminek, M. J. Chromatogr., A 2002, 950, 21-29.
Figure 1. Schematic diagram of the LC setup. W, waste; C, pump (AcN 1% HCOOH); A/B, pumps A and pump B used for gradient elution (A, H2O/3% HCOOH; B, AcN/3% HCOOH). Dashed squares represent the column (10 × 1 mm i.d. Betamax Neutral from Thermosil Keystone, Bellafonte, PA).
tubes (10 × 75 mm) and placed in an Aspec XL4 robot (Gilson S.A.S). The robot was configured with 48-position vial plates. The first purification step involved passage through a 1-mL C18 100-mg Bond Elute cartridge (Varian, Palo Alto, CA), conditioned with 1 mL of MeOH at a flow rate of 2 mL/min (this flow rate was used for all steps) followed by 1 mL of extraction buffer. The samples, 1 mL, were then withdrawn and loaded onto the C18 cartridge. The flow-through was collected and transferred to a Speed-Vac concentrator (Savant Instrument, Framingdale, NY) for evaporation to water phase. After this, 3 mL of 1 M formic acid solution was added to each of the sample vials, which were transferred back to the SPE robot. The second purification step involved use of 1 mL/30 mg Oasis MCX columns (Waters, Milford, MA) conditioned with 1 mL of MeOH, followed by 1 mL of 1 M formic acid. Three milliliters of each sample was loaded onto the SPE column, which was then washed with 1 mL of 1 M formic acid and 1 mL of MeOH. The sample was eluted with 1 mL of 0.35 M NH4OH in MeOH (4:6 v/v) into 1-mL total recovery LC autosampler vials (Waters). After evaporation in the SpeedVac concentrator, the samples were derivatized. For propionylation, samples were dissolved in 10 µL of acetonitrile, 6 µL of N-methylimidazole, and 3 µL of propionic anhydride and then heated at 37 °C in an oven for 30 min before being evaporated to dryness in the Speed-Vac concentrator. Samples for benzoylation were dissolved in 10 µL of 0.66 M benzoic anhydride in acetonitrile, with 6 µL of N-methylimidazole, and heated at 37 °C for 30 min. The reaction mixtures were then evaporated to dryness. Prior to LC-MS analysis, propionylated samples were dissolved in 1.3 µL of acetonitrile containing 3% formic acid followed by 11.7 µL of aqueous 3% formic acid. Benzoylated samples were first dissolved in 2 µL of acetonitrile and then 9 µL of H2O was added. The excess benzoic anhydride precipitated as white flakes, and the addition of 2 µL of acetonitrile generated a clear solution. In this way, benzoylated samples were dissolved in 30% acetonitrile. A 10-µL sample was injected on the column, representing ∼77% of the sample volume. Liquid Chromatography-Mass Spectrometry. A schematic diagram of the LC setup is shown in Figure 1. The autosampler was a Spark model 920 (Spark, AJ Emmens, Holland), with a 25µL syringe, 2.4-µL needle, and 10-µL injection loop fitted in the autosampler. The LC-MS system was a Micromass CAP-LC (Waters, Manchester, U.K.), coupled to a Quattro Ultima triple-
Table 1. Elution Gradient Used and Positions of the External Valvea time (min)
B (A/B) (%)
flow rate (µL/min)
C (flow rate) (µL/min)
external valve position
0 1.5 2 3.5 3.6 10 12 12.1 13 13.1 17 17.1 17.5 19 20
5 5 5 b b 55 55 80 80 100 100 5 5 5 5
10 10 10 10 10 10 10 10 10 40 40 20 20 20 10
0 0 0 0 40 40 40 40 40 0 0 0 0 0 0
1 2 2 2 2 2 2 2 2 2 2 2 1 1 1
a B (% of solvent B), C (flow rate, auxiliary pump C), for external valve position; see Figure 1. b Linear gradient between 2 and 10 min.
stage quadrupole mass spectrometer (Waters). A 10 × 1 mm BetaMax Neutral drop-in guard cartridge with 5-µm particles (Thermo Hypersil-Keystone) was used as an analytical column. This was placed in a 10-mm drop-in cartridge holder (also from Thermo Hypersil-Keystone). The needle wash solvent was 15% acetonitrile with 1% formic acid. The gradient elution buffers were (A) H2O and (B) acetonitrile, both containing 3% formic acid, and the flow rate during elution was 10 µL min-1. To wash the injection loop during the gradient elution, 1% formic acid in acetonitrile was pumped through it, between 3.6 and 13 min at a flow rate of 40 µL min-1, by auxiliary pump C. The elution gradient is shown in Table 1. To wash the column, tubing, and connections following elution of the analytes, the flow rate was increased to 40 µL min-1 and the solvent composition to 100% B at 13.1 min. The data shown in Figures 4-6 were gathered with a longer isocratic period at 80% B (13-17 min), because of the later elution of the more hydrophobic benzonylated derivatives. For the direct flow injection experiments, the mobile phase consisted of 50% B at a flow rate of 40 µL min-1. The injections in these cases were performed with a 5-µL sample loop, introduced into the built-in injection valve of the mass spectrometer. The capillary voltage, cone, and desolvation gas flows and temperatures, resolutions, and ion energies were optimized with Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2871
200 pg µL-1 PRO-ZR delivered at a flow of 10 µL min-1 using a syringe pump. Cone voltage and collision cell energy were optimized for each individual compound using the same setup. The mass spectrometer settings were as follows: capillary voltage 3 kV, source temperature 90 °C, desolvation gas temperature 250 °C; cone gas flow 12 L h-1; desolvation gas flow 800 L h-1; LM 1+2 resolution 13; HM 1+2 resolution 12; ion energy 1, 2 V, ion energy 2, 2 V; entrance -5 V; exit 1 V; multiplier 650 eV. Collision cell pressure was kept between 2 and 2.5 mbar with argon gas. The cone voltage and collision cell energies were as follows, (cone voltage (V))/collision cell energy (eV)): Z (60/14), ZR (60/19), ZMP (80/20), FMN (50/20), UDPG (50/20), Ado (50/20), AMP (50/20), ADP (50/20), iP (60/15), PRO-Z (60/14), PRO-DHZ (60/ 17), PRO-ZR (60/19), PRO-iPA (80/17), PRO-ZOG (80/25), PROiPMP (80/18), PRO-ZMP (80/20), PRO-DHZR (60/19), PROGZ7G,Z9G (100/22), PRO-ZROG (85/27), PRO-Ado ((60/15), PROAMP (70/17), PRO-ADP (75/20), PRO-FMN (100/25), PROUDPG (50/25), benzoyl-Z (50/18), benzoyl-iPMP (80/21), benzoylZMP (80/25), benzoyl-iPA (80/22), benzoyl-ZR (80/26), benzoylZOG (80/30), benzoyl-Z7G,Z9G (80/30), benzoyl-AMP (90/25), benzoyl-ADP (90/25) and benzoyl-ATP (100/25). Method Validation. To validate the method for analyzing cytokinins from Arabidopsis using propionylation as a derivatization step prior to LC-MS, the following quality parameters were estimated: limit of detection (LOD), recovery (during purification), relative standard deviation (RSD) for the retention time, RSD for the concentration determination, and analytical accuracy at different spiking concentrations. LOD was determined by injecting varying amounts of a standard mixture and determining the concentration at which the signal-to-noise ratio for the respective cytokinins was 3. For the other parameters 7 g of the plant material was ground in liquid nitrogen with a mortar and pestle, and then three sets of experiments were performed (designated A, B, and C). For each of these three sets, the extraction buffer contained an internal standard (IS) mixture of the cytokinins with 500 pg mL-1 of each individual IS, all samples were subject to centrifugation at 25400g in a Beckman J2-MC centrifuge (Fullerton, CA), and the extraction time was 2 h. For the A set of analyses, 1 g of the plant material was extracted in 10 mL of extraction buffer. After centrifugation, the supernatant was divided into 10 aliquots, each corresponding to 100 mg of plant tissue extract in 1 mL of extraction buffer. These aliquots were analyzed with no further additions or modifications. For the B set of analyses, five 1-g portions of the plant material were extracted separately in 10 mL of extraction buffer, after spiking them with cytokinins in a series of concentrations (100, 250, 500, 750, and 1000 pg mL-1 extraction buffer). After extraction and subsequent centrifugation, these 5 preparations were each split into 10 aliquots and analyzed as described above. For the C set of experiments, a 1-g portion of the plant material was extracted in 10 mL of extraction buffer. After extraction, the sample was split into 10 aliquots and purified as described. Before derivatization of the 10 samples, 1000 pg of cytokinins mL-1 of extraction buffer was added. In all of these cases, each aliquot was 1 mL in volume and contained extracts from 100 mg of plant tissue. By dividing cytokinin peak areas obtained from the B samples spiked at the 2872
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
1000-pg level with corresponding peak areas from the C samples, the recovery value was obtained for each cytokinin. The content and RSD for the retention time of each cytokinin were determined from the A samples. The accuracy at each spiking level was determined from the B set of samples, by dividing the determined concentration by the concentration derived from the A set of samples plus the spiked amount. Concentrations of the different cytokinins were determined from the area ratios between endogenous cytokinins and their respective stable isotope-labeled counterparts. The response ratios were entered as Y values in the calibration curves, established in the form of Y ) kX + m. The calibration curves were obtained from five separate injections of mixtures with the following endogenous/IS ratios (values in ng): 0.02/0.5, 0.15/0.5, 0.25/0.5, 0.5/0.5, 1.25/0.5, 2/0.5, 2.5/0.5, and 3/0.5. The calibration curves, standard deviations of k and m, and the corresponding confidence limits were determined according to standard procedures.22 RESULTS AND DISCUSSION The analysis of nucleotides and other structurally related compounds that occur in biological matrixes is usually very difficult due to the incompatibility of good LC separation conditions and ESI. In the investigation presented here, the aim was to develop a strategy allowing nucleotides and structurally related compounds to be analyzed rapidly, with high sensitivity, by reversed-phase LC-ESI-MS. We chose to use the plant hormone group cytokinins as model compounds, since they represent a wide spectrum of analytes, ranging from polar nucleotides to nonpolar adenine bases. A further purpose was to show the usefulness of the developed method for other biologically interesting compounds that are normally difficult to analyze by LC-MS under optimal MS conditions, e.g., ATP and AMP. Sample Purification and the Chromatographic System. Immunoaffinity purification has been commonly used in many of the previously described purification methods for cytokinins, and the fact that nucleotides have been analyzed after dephosphorylation has created bottlenecks in the development of highthroughput analysis techniques for this group of plant components. Here, using a modified version of the SPE-based sample preparation method described by Dobrev and Kaminek,21 the entire purification protocol was automated with a SPE robot, and the cytokinin bases, ribosides, and nucleotides were collected in a single fraction. Theoretically, 96 samples could be purified daily. However, ∼70 samples/day is the practical number of samples processed. To avoid reducing the throughput at the final stage of the analysis, a cycle time of ∼20 min is required for the LC-MS analysis. A schematic view of the LC system is shown in Figure 1. During gradient elution, the C-pump washes the sample loop to ensure that sample carryover is very low (almost undetectable). We utilized a short (10 × 1 mm i.d.) column packed with a very hydrophobic silica-based material. Although the column we used is actually manufactured as a guard column (and thus is relatively low cost), it was possible to separate all 12 cytokinins within 13 min using it (data not shown). As no precolumn/valve switching is needed, less time is needed for the analysis. Furthermore, as washing and equilibration can be performed at higher flow rates, (22) Miller, J. C.; Miller, J. N. Statistics for Analylical Chemistry, 3rd ed; Ellis Horwood Limited: Bodmin, Great Britain, 1993; pp 109-112.
Figure 2. (A) Structure of the compounds used in this study. A notable feature is that only internal standards of ZMP and iPMP are labeled with 15N on the purine ring. (B) Conditions used for the derivatization through esterification of hydroxyl groups via reaction with anhydride and basic catalyst: 10:6:3 and 10:6 represents the volumetric ratios of the derivatization agents.
the cycle time was also decreased by using the 1-mm-i.d. column. High formic acid concentrations were not needed for the chromatographic separation, although a small reduction in peak tailing for the nucleotides was observed when using 3% compared with 0.7%. The chromatographic setup can be easily adapted for longer and smaller i.d. columns if lower flow rates or better separations are needed. Derivatization Improves Both the ESI Response and the Chromatography. We chose to use propionyl and benzoyl ester formation as the mode of derivatization. Propionyl esters have previously been shown easily formed at a high yield,19 and benzoylation was chosen to investigate how different sizes of derivatives can affect ESI response. The physicochemical diversity of the compounds studied is illustrated in Figure 2A. The analytical
difficulties caused by their wide range of polarity were overcome by esterification of the hydroxyl groups (Figure 2B). This improved the separation and ESI response because it both reduced the polarity range of the compounds and made them more hydrophobic. The daughter ion spectra from four of the compounds analyzed are shown in Figure 3A-D. Propionylated zeatin loses its propionyl ester side chain through a neutral loss of its corresponding acid, resulting in the base peak fragment at m/z 202 (Figure 3A). The propionylated zeatin riboside loses its ribose moiety, resulting in the base peak fragment at m/z 276 (Figure 3B). The fragmentation mechanism for the investigated glucosides is similar to that described by A° stot et al.,19 (data not shown). Propionylated ZMP (Figure 3C) and propionylated ADP (Figure 3D) are fragmented by the same mechanism. The phospho group Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2873
Figure 3. (A-D) Daughter ion spectra of selected propionylated cytokinins. Fragment ions with percentage of base peak area in parentheses. (A) Propionylated zeatin (M + H)+ m/z 276; 220 (1), 202 (100), 185 (34), 141 (10), 136 (14). (B) Propionylated ZR (M + H)+ m/z 576; 576 (57), 301(9), 276 (100), 202 (28), 153 (21), 141 (5). (C) Propionylated ZMP (M + H)+ m/z 600; 600 (12), 276 (100), 202 (14). (D) Propionylated ADP (M + H)+ m/z 540; 540 (1), 460 (4), 153 (24), 136 (100). (E-L) Effect of propionylation on ESI response. Five nanograms of each compound was analyzed by direct flow injection. The TIC values of the daughter ion spectra are presented. (E) Z, (F) propionylated Z, (G) ZR, (H) propionylated ZR, (I) ZMP, (J) propionylated ZMP, (K) ADP, and (L) propionylated ADP. The intensity scale is the same for each underivatized/derivatized pair of compounds.
Figure 4. Multiple reaction monitoring (MRM) traces of (A) propionylated Z (m/z 276-202), propionylated ZR (m/z 576-276), propionylated ZMP (m/z 600-276). (B) Benzoylated Z, (m/z 324-202), benzoylated ZR (m/z 768-324), benzoylated ZMP (m/z 744-324). (C) Underivatized Z, (m/z 220-202), ZR (m/z 352-220), ZMP (m/z 416-220). The intensity scale in (A)-(C) is the same.
is lost through cleavage of the ester bond at the 5′ position on the riboside. The resulting base peak fragment consists of the purine ring plus the attached side chain. The fragmentation mechanism is the same for mono-, di-, and triphosphates (data not shown). The effect of derivatization on the ESI response was also investigated by direct flow injections of propionylated and nonpropionylated compounds (Figure 3E-L). The results show that the improvement in the ESI response is correlated with the increase in hydrophobicity associated with the derivatization. Propionylation of Z resulted in only a modest increase in response (Figure 3E,F). However, for ZR (Figure 3G,H) and ZMP (Figure 3I,J), propionylation increased the response 5-6-fold. The nucleotide ADP also showed a high increase in response after
propionylation (Figure 3K,L). Cech and Enke5 stated that analysis with ESI is likely to be most successful for analytes with the highest affinity for the droplet surface. However, many compounds present in biological organisms are not hydrophobic and therefore not particularly surface active. The results shown in Figure 3E-L suggest that higher hydrophobicity is achieved by propionylation of the hydroxyl groups, resulting in higher surface activity and thus better ionization and ESI responses. It should also be mentioned that the derivatization also results in a considerable increase in molecular weight; e.g., the propionylated ZR has 60% higher mass than nonderivatized ZR. Thus, since the chemical background is higher in the lower m/z region, the signal-to-noise ratio may also be improved by the reduction in background noise. The drawback with derivatization is the increase in total analysis time. However, one way to reduce the total derivatization
2874 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
Figure 6. Selected cytokinin MRM traces from a 20-mg extract of A. thaliana (A) propionylated and (B) benzoylated. The intensity scale in (A) and (B) is the same.
Figure 5. Effect of derivatization on chromatography and ESI response for different compounds of biological interest. The amount of analyte injected onto the column is shown in parentheses. (A) Propionylated UDPG (550 fmol), propionylated ADP (460 fmol), propionylated FMN (80 fmol), propionylated AMP (5 fmol), and propionylated Ado (10 fmol) (B) UDPG (880 fmol), FMN (1 pmol), Ado (1.8 pmol), ADP (1.2 pmol), and AMP (1.4 pmol). (C) Benzoylated ATP (14 pmol), benzoylated ADP (300 fmol), and benzoylated AMP (70 fmol). MRM transition stages according to Table 4.
time is to avoid the evaporation step after derivatization by direct dilution of the samples with injection buffer. In this method, large volumes (100 µL) can easily be injected, since we use a 1-mm-i.d. column and can load the sample at high flow rates, thus avoiding the need to increase the analysis time and diffusion problems. Figure 4 shows a comparison between derivatized (Figure 4AB) and nonderivatized (Figure 4C) Z, ZR, and ZMP, with the chromatographic setup used in this study. Even with the short column, excellent separation is achieved for derivatized cytokinins, with shorter retention times for the propionylated (Figure 4A) than the benzoylated compounds (Figure 4B). With the short column and no acetate or formate buffer added to the mobile phase, the nonderivatized cytokinin standards could not be separated (Figure 4C).
To further illustrate the wide range of compounds that can be simultaneously determined using this kind of derivatization strategy, we also analyzed other compounds of biological significance (Figure 5). In Figure 5A, MRM traces of propionylated UDPG, FMN, Ado, AMP, and ADP are shown. Separation of AMP and ADP normally requires the addition of ion pair reagents,11 but after propionylation, no ion pair reagent is needed. For comparison, we injected the same set of compounds (1 pmol of each) without prior propionylation (Figure 5B). The difference in the results was striking: for UDPG, FMN, and ADP, essentially no signals were obtained, although the injected amounts were 2-10 times higher than the amounts of their propionylated counterparts (Figure 5A). For AMP and Ado, signals were observed, but there was no retention of the compounds. The detection limits derived for PRO-AMP (5 fmol) and PRO-ADP (460 fmol) are better than the previously reported 1 pmol using LCMS/MS with an ion pair reagent and MRM detection.11 However, the retention for PRO-ADP was relatively poor, and for PRO-ATP, no retention at all was observed. Therefore, we also investigated the possibility of benzoylating AMP, ADP, and ATP (Figure 5C). The standards were dissolved, injected in 30% acetonitrile, and eluted at ∼40% acetonitrile in the mobile-phase gradient. This was unexpectedly high, considering that no ion pair reagents were used. The benzoyl groups probably shield the ionic phosphate group, so that hydrophobic interactions between mobile and stationary phase can occur and the compounds are retained. No separation was obtained, using this setup, between benzoyl-ADP and benzoyl-ATP. Analysis of Cytokinins in Small Amounts of Plant Material. The purification and derivatization protocol developed here was tested for LC-MS/MS analysis of cytokinins from the model plant A. thaliana. To demonstrate the strength of derivatization as an analytical approach for detecting polar compounds, we Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2875
Table 2. Parameters Derived in the Method Validation for the LC-MS/MS Analysis of Propionylated Cytokininsa
CK
diagnostic transiton (m/z)
LOD (fmol)
recovery (%)
RT RSD (%)
iP Z DHZ ZR iPA ZMP iPMP ZOG Z7G Z9G ZROG DHZR
204-136 276-202 278-204 576-276 504-204 600-276 528-204 606-202 662-276 662-276 906-606 578-278
25 18 18 0.17 0.2 0.8 0.9 0.8 0.8 1.5 110 0.17
67 63 70 71 62 43 65 65 114 70 45 59
7.2 5.3 6.7 0.4 0.4 0.6 1.4 1.4 0.4 0.4 0.2 1.2
content fmol/100 mg of FW
content RSD (%)
n
50 202
21 23
7 7
133 59 1888 1030 1232 2618 823 253
15 10 7 5 5 16 7 19
9 10 10 10 10 10 10 10
100 pg
analytical accuracy (%) 250 500 750 pg pg pg
1000 pg
56 109 156 113 104 105 102 115 123 101 92 76
90 113 114 105 104 94 91 116 130 105 100 171
69 95 143 108 98 100 96 102 112 105 92
67 110 114 104 102 89 90 107 119 99 93 147
83 123 108 112 106 91 85 105 131 103 129 178
a Diagnostic transition, MRM transition monitored; LOD, limit of detection; RT RSD, retention time RSD); content, determined concentration in a 100-mg aliquot of a pooled extract; n, number of aliquots analyzed; and analytical accuracy at the different spiking concentrations. See Experimental Section Method Validation, B set of samples, for more details.
performed the analysis on small samples of 20 mg of fresh weight. Twelve cytokinin bases, ribosides, and nucleotides were simultaneously separated and detected within a total run time of 20 min (data not shown). The results for a limited number of cytokinin metabolites are shown in Figure 6. The response for benzoylated cytokinins is generally weaker than for the corresponding propionylated cytokinins. The difference is pronounced for ZMP, the response for the propionylated derivative being ∼6 times higher than for the benzoylated derivative. However, the most striking difference appears for Z. In the analysis of 20 mg of plant tissue after propionylation, Z appears as an undefined peak in the area of the chromatogram with high background noise (Figure 6A). Benzoylation of samples results in a better-defined peak for Z (Figure 6B) and thus improves the detection limits. Benzoyl derivatives of ZMP and iPMP separate better than the corresponding propionylated derivatives, probably due to the hydrophobic benzoyl derivatives having stronger reversed-phase interactions than the propionyl derivatives. The weaker responses for most of the benzoylated cytokinin derivatives might be due to differences in stability and reaction yields or to their incomplete desolvation in the injection solvent. Although benzoylation is the best choice as derivatization mode for analysis of Z, propionylation was chosen as the derivatization method for general cytokinin analysis, partly because the other propionylated cytokinins showed higher responses and partly because the larger glycoside/riboside conjugates were difficult to dissolve in 30% acetonitrile. However, when 100-mg samples were analyzed, Z could be quantified after propionlyation, so propionylation is a good compromise for general cytokinin analysis. If necessary, an alternative would be to split the samples during purification in the SPE robot, and then benzoylate and analyze Z separately. Method Validation. The method for analyzing cytokinins from plant tissues was validated (Table 2), and the recovery values were found to be slightly lower than those obtained using the similar purification method described by Dobrev and Kaminek.21 This is probably due to the automation of the method. The derivatization yield was not determined separately but was included in the recovery measurements. It has previously been found to be >90%.19 The stability of the LC system was determined by calculating the precision of the retention times. The results show 2876 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
Table 3. Calibration Curves for Propionylated Cytokininsa compd
calibration equation (confidence limits)
R2
dynamic range (fmol)
iP Z ZR DHZ DHZR iPA ZMP iPMP ZOG Z7G Z9G ZROG
Y ) 1.063 ( 0.023x + 0.0019 ( 0.038 Y ) 0.721 ( 0.015x - 0.0029 ( 0.025 Y ) 1.045 ( 0.059x + 0.014 ( 0.036 Y ) 0.671 ( 0.017x - 0.0051 ( 0.027 Y ) 0.743 ( 0.189x + 0.235 ( 0.307 Y ) 0.885 ( 0.042x - 0.003 ( 0.012 Y ) 1.287 ( 0.031x - 0.068 ( 0.051 Y ) 1.611 ( 0.127x - 0.053 ( 0.207 Y ) 1.081 ( 0.029x + 0.017 ( 0.048 Y ) 0.854 ( 0.012x + 0.002 ( 0.019 Y ) 0.747 (0.022x + 0.005 ( 0.036 Y ) 0.906 ( 0.218x + 0.032 ( 0.357
0.9995 0.9996 0.9991 0.9994 0.9392 0.9998 0.9994 0.9938 0.9993 0.9998 0.9991 0.9446
100-14000 70-10000 30-2000 70-10000 35-5000 40-1000 30-5000 40-5700 30-4900 30-4500 30-4500 20-3300
a Confidence limits, correlation coefficients, and dynamic range for propionylated cytokinins.
low precision for the early-eluting cytokinin bases. However, as deuterated internal standards are used for the quantification, the observed difference in retention times is considered to be a minor problem. Calibration curves were established for all of the measured cytokinins (Table 3), and a high degree of linearity over a wide dynamic range was found for all of them. The detection limits are all in the high-attomole to low-femtomole range except for PRO-ZROG (Table 2). For many of the cytokinins, this is an improvement of the detection limit by a factor of 10-100 compared to other published methods.18,20 The analytical precision was also determined to be between 5 and 15%, except for the bases, which are present at very low levels, resulting in lower precision. The accuracy of the analytical procedure was determined by spiking aliquots of plant extract with varied amounts of the analytes. The results show an accuracy of 80-120% in the range investigated, which is satisfactory for this kind of analysis of trace components in a complex matrix.23 A limited validation of the LC-MS method for the noncytokinin compounds was also performed (Table 4). The detection (23) Van Rhijn, J. A.; Heskamp, H. H.; Davelaar, B.; Jordi, W.; Leloux, M. S.; Brinkman, U. A. Th. J. Chromatogr., A 2001, 929, 31-42.
Table 4. Validation Parameters for the Non-Cytokinin Compounds Studieda compound
diagnostic transition (m/z)
LOD (fmol)
RT (min)
RT RSD (%)
n
PRO-AMP PRO-ADP PRO-Ado PRO-FMN PRO-UDPG benzoyl-AMP benzoyl-ADP benzoyl-ATP
460-136 540-136 436-136 625-527 903-183 556-136 636-136 716-136
5 460 1 40 140 35 250 14000
4.3 2.4 7.2 7.7 8.8 7.2 5.8 6
7.6 2 1.5 0.5 0.7 0.7 1.2 1.2
5 5 4 4 6 5 5 5
a Diagnostic transitions, LOD, retention time, and relative standard deviation for the retention time (RSD) n (number of injections).
limits varied from low-femtomole to low-picomole concentrations. The highest detection limits were found for ADP and ATP. In the analysis of benzoylated ATP, signals in the ADP and AMP channels were also detected. It is well known that ATP and ADP are labile and very frequently partially or fully dephosphorylated. Therefore, for the quantification of AMP, ADP, and ATP, differentially labeled internal standards should be used. However, in the present investigation, the ambition was only to demonstrate that ionic compounds that are not normally associated with this type of reversed-phase chromatography can be analyzed using our derivatization approach. CONCLUSIONS The need for rapid and sensitive analytical methods in experimental biology to quantify compounds spanning large ranges of hydrophobicity calls for new analytical strategies. LCMS is generally recognized for its greater ability (compared to
GC/MS) to separate and detect polar compounds without the need for any derivatization. However, it is well known that the sensitivity of LC combined with fluorescence or UV detection can often be increased by derivatization. Nevertheless, the derivatization of analytes for LC-ESI-MS detection is not widely practiced. As shown in the present study, both the chromatography and the response of a wide group of compounds can be improved with derivatization. Using precolumn derivatization, we have demonstrated that it is possible to simultaneously analyze compounds with a wide polarity range, many of which have previously been impossible to analyze simultaneously in a reversed-phase system without ion pairing reagents. The formation of propionyl or benzoyl esters can be performed under mild conditions and requires no further sample cleanup. The resulting derivatives show less diversity in polarity and are more hydrophobic. This improves the separation of the analytes on a reversed-phase column, and significantly enhances ionization during the ESI process. The utility of the precolumn derivatization strategy was demonstrated by the analysis of cytokinins in 20-100-mg samples of Arabidopsis, showing that it can be used to determine trace components, with a wide range of polarity, in complex biological matrixes. The method represents an improvement from previous methods, enabling detailed studies of cytokinins in specific plant tissues, which was not previously possible. ACKNOWLEDGMENT We thank the Swedish Research Council and the Strategic Foundation for Research for financial support. Received for review January 15, 2004. Accepted March 12, 2004. AC0499017
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2877
