Toxin Screening in Phytoplankton: Detection and Quantitation Using

Quantitative Analysis of Antiretroviral Drugs in Lysates of Peripheral Blood Mononuclear Cells Using MALDI-Triple Quadrupole Mass Spectrometry. Jeroen...
2 downloads 0 Views 219KB Size
Download PDF
Anal. Chem. 2005, 77, 1509-1517

Toxin Screening in Phytoplankton: Detection and Quantitation Using MALDI Triple Quadrupole Mass Spectrometry Lekha Sleno and Dietrich A. Volmer*

Institute for Marine Biosciences, National Research Council, 1411 Oxford Street, Halifax, NS, Canada B3H 3Z1

The investigation of a MALDI triple quadrupole instrument for the analysis of spirolide toxins in phytoplankton samples is described in this study. A high-frequency (kHz) laser was employed for MALDI, generating a semicontinuous ion beam, thus taking advantage of the high duty cycle obtained in sensitive triple quadrupole MRM experiments. Initially, several experimental parameters such as type of organic matrix and concentration, solvent composition, and matrix-to-analyte ratio were optimized, and their impact on sensitivity and precision of the obtained ion currents for a reference spirolide, 13-desmethyl-C, was studied. In all quantitative experiments, excellent linearities in the concentration range between 0.01 and 1.75 µg/mL were obtained, with R2 values of 0.99 or higher. The average precision of the quantitative MALDI measurements was 7.4 ( 2.4% RSD. No systematic errors were apparent with this method as shown by a direct comparison to an electrospray LC/MS/MS method. Most importantly, the MALDI technique was very fast; each sample spot was analyzed in less than 5 s as compared to several minutes with the electrospray assay. To demonstrate the potential of the MALDI triple quadrupole method, its application to quantitative analysis in several different phytoplankton samples was investigated, including crude extracts and samples from mass-triggered fractionation experiments. 13-Desmethyl spirolide C was successfully quantified in these complex samples at concentration levels from 0.05 to 90.4 µg/mL (prior to dilution to have samples fall within the dynamic range of the method) without extensive sample preparation steps. The versatility of the MALDI triple quadrupole method was also exhibited for the identification of unknown spirolide analogues. Through the use of dedicated linked scan functions such as precursor ion and neutral loss scans, several spirolide compounds were tentatively identified directly from the crude extract, without the usual time-consuming chromatographic preseparation steps. Moreover, high-quality CID spectra were obtained for lowabundant spirolides present in the phytoplankton samples. Toxic components in food products destined for human consumption are usually either food-borne pathogenic microor* To whom correspondence should be addressed. Tel. (902) 426-4356. Fax (902) 426-9259. E-mail: [email protected]. 10.1021/ac0486600 CCC: $30.25 Published 2005 Am. Chem. Soc. Published on Web 01/28/2005

ganisms (bacteria, viruses, and parasites) or natural toxins.1 Natural toxin accumulation in fish originates mostly after feeding on toxic phytoplankton blooms.2 There are also a few naturally occurring toxins in certain fish species such as puffer fish.3 The U.S. Food and Drug Administration has established specific tolerance levels for toxins in fish tissue; for example, for paralytic shellfish poisons and neurotoxic shellfish poisons, a value of 0.8 ppm is set, and for diarrhetic shellfish poisons, an allowed concentration of 0.2 ppm is indicated.4 Unfortunately, validated and rapid analytical methods that allow commercial testing of every fish are not available.4 To guarantee consumer safety and protect the aquaculture industry from financial damages, however, sensitive and effective monitoring tools for contaminated seafood are required. Analytical protocols for the monitoring of toxic components in biological samples are usually composed of several steps. After extensive sample extraction and cleanup, the extracts are then subjected to a chromatographic separation. If the aim of the work is discovery or structural elucidation of novel or unknown toxins, a bioassay-guided or mass-triggered (semi-) preparative fractionation step is often additionally performed prior to chromatography.5 Liquid chromatography (LC) is most commonly employed for separating polar toxins.6 Ideally, LC is combined with specific mass spectrometry detection (LC/MS). While soft ionization techniques such as atmospheric-pressure chemical ionization,7 sonicspray ionization,8 or fast-atom-bombardment9 have been successfully implemented in toxin assays, most researchers presently use electrospray ionization (ESI) as a means of ionizing polar and thermally labile toxin molecules.10 Of course, subsequent tandem mass spectrometry (MS/MS) analysis is then needed to (1) Hallegraeff, G. M. Phycologia 1993, 32, 79-99. (2) Food and Drug Administration. Bad Bug Book: Foodborne Pathogenic Microorganisms and Natural Toxins; FDA, Center for Food Safety and Applied Nutrition, Washington, regularly updated at http://vm.cfsan.fda.gov/ ∼mow/intro.html. (3) Field, J. J. Accident Emergency Med. 1998, 15, 334-336. (4) Food and Drug Administration. Fish and Fishery Products Hazards and Controls Guidance, 3rd ed.; FDA, Center for Food Safety and Applied Nutrition: Washington, DC, 2001; p 73-82. (5) Sleno L.; Chalmers, M.; Volmer, D. A. Anal. Bioanal. Chem. 2004, 378, 977-986. (6) Quilliam, M. A. J. Chromatogr., A 2003, 1000, 527-548. (7) Dickey, R.; Jester, E.; Granade, R.; Mowdy, D.; Moncreiff, C.; Rebarchik, D.; Robl, M.; Musser, S.; Poli, M. Nat. Toxins 1999, 7, 157-65. (8) Hashimoto, T.; Nishio, S.; Nishibori, N.; Yoshioka, S.; Noguchi, T. J. Food Hyg. Soc. Jpn. 2002, 43, 144-147. (9) Ofuji, K.; Satake, M.; McMahon, T.; Silke, J.; James, KJ.; Naoki, H.; Oshima, Y.; Yasumoto, T. Nat. Toxins 1999, 7, 99-102.

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005 1509

produce fragment ions for reliable confirmation of a particular toxin’s presence or for supporting the elucidation of an unknown structure.11 An ionization technique that has seen virtually no application in the toxin field as of yet is matrix-assisted laser desorption ionization (MALDI). Except for some studies in which MALDI showed the presence of protonated molecules of toxins in sample extracts and LC fractions,12-15 and a few reports on MALDI MS/MS for obtaining sequence information on biologically active peptide toxins,16,17 there appears to be no detailed study on the application of MALDI for identification, quantitation, and structural characterization of non-peptide toxins. The reasons for the limited use of MALDI for toxins are, of course, based on its limitations: many non-peptide toxins have molecular masses well below 1000 Da; i.e., there is a significant potential for interferences with fluctuating ion signals from the organic matrix used for desorption and ionization in MALDI.18 Also, signal intensities are notoriously irreproducible18 due to nonhomogeneous cocrystallization of the analyte with the matrix. Moreover, hyphenated LCMALDI is still predominantly conducted with off-line spotting devices,19 as no commercial on-line interfaces are available. Recently, however, new instrumental developments, novel matrices, and specialized sample preparation procedures have extended the application range of MALDI. It has been demonstrated18 by several researchers that MALDI can be used for extremely rapid and quantitative analysis of low molecular weight pharmaceutical drugs. In particular, the combination of MALDI with a triple quadrupole (QqQ) MS has shown great promise in drug analyses that are currently dominated by ESI methods.20 For the present work, there are two specific advantages of MALDIQqQ: first, the availability of dedicated QqQ scan modes (precursor ion and neutral loss scans), for selective class-specific screening of toxin analogues; second, the considerably higher duty cycle of a QqQ instrument in the multiple reaction monitoring (MRM) mode in comparison to, for example, a quadrupole timeof-flight analyzer, for a significant sensitivity increase in quantitative measurements. MRM also provides a convenient way of circumventing interfering MALDI background signals in the lowmass range. These advantages are further enhanced by using a high-frequency (kHz) laser20,21 for ion generation instead of the more common N2 lasers with typical repetition rates of 10 Hz. A kHz laser generates a semicontinuous ion beam, which is ideal (10) Quilliam, M. A. In Applications of LC-MS in Environmental Chemistry; Barcelo, D., Ed.; Elsevier: Amsterdam, 1996; p 415. (11) Sleno, L.; Volmer, D. A.; Kovacevic, B.; Maksic, Z. B. J. Am. Soc. Mass Spectrom. 2004, 15, 462-477. (12) Ranasinghe, C.; Akhurst, R. J. J. Invertebr. Pathol. 2002, 79, 51-58. (13) Welker, M.; Fastner, J.; Erherad, M.; von Dohren, H. Environ. Toxicol. 2002, 17, 367-374. (14) Nasri, A. B.; Bouaicha, N.; Fastner, J. Arch. Environ. Contam. Toxicol., 2004, 46, 197-202. (15) Oliveira, J. S.; Pires, O. R.; Morales, R. A. V.; Bloch, C.; Schwartz, C. A.; Freitas, J. C. J. Venomous Anim. Toxins Trop. Dis. 2003, 9, 76-88. (16) Kalume, D. E.; Stenflo, J.; Czerwiec, E.; Hambe, B.; Furie, B. C.; Furie, B.; Roepsdorff, P. J. Mass Spectrom. 2000, 35, 145-156. (17) Rigby, A. C.; Lucas-Meunier, E.; Kalume, D. E.; Czerwiec, E.; Hambe, B.; Dahlqvist, I.; Fossier, P.; Baux, G.; Roepstorff, P.; Baleja, J. D.; Furie, B. C.; Furie, B.; Stenflo, J. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 5758-5763. (18) Cohen, L. H.; Gusev, A. I. Anal. Bioanal. Chem. 2002, 373, 571-586. (19) Gusev, A. I. Fresenius J. Anal. Chem. 2000, 366, 691-700. (20) Hatsis, P.; Brombacher, S.; Corr, J.; Kovarik, P.; Volmer, D. A. Rapid Commun. Mass Spectrom. 2003, 17, 2303-2309. (21) McLean, J. A.; Russell, W. K.; Russell, D. H. Anal. Chem. 2003, 75, 648654.

1510

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

Figure 1. Structures of 13-desmethyl spirolide C (Des-C) and gymnodimine (GYM). Other known spirolides have similar structures with changes in R1, R2 (H or CH3) with or without double bond at position 2, 3.24

for mass analysis in a beam-type instrument such as the triple quadrupole MS. Statistically meaningful results with several thousand laser shots per sample can be obtained in only a few seconds, the averaging of which dramatically improves the overall precision.20 The aim of the present work was to exploit the specific advantages of MALDI and QqQ mass spectrometry and carry out a comprehensive analytical study of a representative class of marine toxins in phytoplankton samples, including screening for known and unknown analogues, quantitation of toxin concentrations, and preliminary structural characterization, based on lowenergy collision-induced dissociation (CID) spectra. We are aware that a MALDI assay would immediately be scrutinized for its ability to compete with the established ESI methods for toxins. Therefore, we compared and cross-correlated the results with data from ESI LC/MS/MS analyses. The selected compound class for this study is a group of spirolide toxins. The spirolides, a novel class of bioactive macrocyclic imines, were initially discovered in shellfish samples from Nova Scotia.22 These compounds were found to be toxic in the mouse bioassay and originally derive from planktonic sources of Alexandrium ostenfeldii. Their mode of action is seen by rapid death following neurological symptoms, including convulsions. Human toxicity is still unknown; however, vague symptoms have been documented, including gastric distress and tachycardia, following shellfish consumption from Nova Scotian sites.23 The spirolides have been identified as polyketides with cyclic imine and spiro-linked tricyclic ether moieties (Figure 1). Their mass spectra, obtained under electrospray CID conditions in triple quadrupole and ion trap instruments as well as infrared multiphoton photodissociation in a Fourier transform ion cyclotron resonance MS, have been recently described,24 exhibiting several common product ions and neutral losses between known and (22) Hu, T.; Curtis, J. M.; Oshima, Y.; Quilliam, M. A.; Walter, J. A.; WatsonWright, W. M.; Wright, J. L. C. J. Chem. Soc., Chem. Commun. 1995, 21592161 (23) Richard, D.; Arsenault, E.; Cembella, A.; Quilliam, M. A. Harmful Algal Blooms 2000, 383-386. (24) Sleno, L.; Windust, A.; Volmer, D. A. Anal. Bioanal. Chem. 2004, 378, 969976.

Figure 2. Schematic representation of MALDI QqQ analysis. Sampling was done with slow rastering down each column of the sample plate, which translates into intensity traces for each MRM transition (Des-C and GYM) with subsequent integration for quantitative results. Total analysis time for the above nine sample spots was less than 1.5 min.

unknown spirolide structures. This consistency in the CID spectra provided an opportunity for class-specific screening experiments of biogenetically linked spirolide species, which was particularly important in the present study. EXPERIMENTAL SECTION Chemicals. Acetonitrile, methanol (Caledon, Georgetown, ON, Canada) and Milli-Q organic-free water (Millipore, Bedford, MA) were used as solvents. Trifluoroacetic acid (TFA), ammonium formate, formic acid, sinapinic acid (SA), 2,5-dihydroxybenzoic acid (DHB), and R-cyano-4-hydroxycinnamic acid (CHCA) were purchased from Sigma-Aldrich (Mississauga, ON, Canada). Standard solutions of 13-desmethyl spirolide C (CRM-SPX1, 10.2 µM) and gymnodimine (CRM-GYM, 10.0 µM) in methanol with 0.05% TFA were obtained through IMB’s Certified Reference Materials Program (Halifax, NS, Canada). Crude phytoplankton extract and fractions used as unknown samples in these experiments were prepared from A. ostenfeldii culture as previously described.5,24 Sample Preparation. For the optimization of the MALDI analyses, 13-desmethyl spirolide C (Des-C) samples were prepared in different solvent compositions (% methanol/acetonitrile). The calibration curves for Des-C were determined in the concentration range between 0.05 and 2.5 µg/mL, with a constant gymnodimine concentration of 0.625 µg/mL. Several calibration curves were analyzed at different matrix concentrations as well as for constant matrix-to-analyte ratios. For quantitative experiments in phytoplankton samples, a calibration curve for CDes-C ) 0.01-1.75 µg/mL (CGYM ) 0.625 µg/mL) was used, with CCHCA ) 2.5 mg/ mL and 75% methanol (with 0.1% TFA). The phytoplankton samples for quantitation of Des-C were prepared in the same manner with appropriate dilution factors, after semipreparative mass-triggered LC fractionation of a crude methanolic extract prepared from a culture of A. ostenfeldii.5 Five samples (samples 8, 11-14) were prepared by solid-phase extraction by loading 500 µL of diluted sample onto preconditioned cartridges (Oasis HLB 1 cm3, Waters, Bedford, MA), followed by a washing step (1.5 mL 5% methanol) and elution with 1.5 mL of either 70 (sample 8) or 80% (samples 11-14) methanol. Samples 8 and 14 both consisted of crude extract samples with slightly different cleanup procedures and recoveries. Eluted samples were dried in a SpeedVac (Thermo-Savant, Holbrook, NY) and reconstituted in 50% methanol. Each calibration standard and sample (1 µL) was spotted three times onto the MALDI plate and allowed to air-dry

prior to analysis. The calibration curve used for the electrospray quantitation included standards from 0.0005 to 0.15 µg/mL Des-C with 0.0625 µg/mL GYM with 10% organic content in final samples. Unknown samples were prepared with an additional 1/10 dilution factor over those used for the MALDI experiments. MALDI Mass Spectrometry. MALDI experiments were performed on an MDS Sciex (Concord, ON, Canada) API 3000 triple quadrupole mass spectrometer equipped with a prototype orthogonal MALDI source with a high-repetition rate (1000 Hz) frequency-tripled (355 nm) Nd:YAG laser and a laser power of 4.1 (40%) or 5.1 µJ (50%) per pulse, depending on the laser’s condition. The sample plate (10 × 10 spots, stainless steel, Perseptive Biosystems, Framingham, MA) was positioned in the q0 region at a pressure of ∼8 mTorr. All spectra were acquired in positive ion mode. MRM transitions for 13-desmethyl spirolide C and gymnodimine used were m/z 692.9 f 164.3 and 508.3 f 174.2, respectively, with a dwell time of 250 ms each. These experiments used a collision energy of 60 eV, a CAD (N2) gas pressure of 12 (arbitrary units), and precursor ion selection in Q1 at 0.6 Da resolution. Quantitative results were obtained by slow vertical raster sampling and, subsequently, plotting peak area ratios of Des-C/GYM versus concentration of Des-C. A schematic representation of the MALDI QqQ analysis is shown in Figure 2. Liquid Chromatography. An Agilent 1100 (Palo Alto, CA) binary pump was employed for the LC separation of 13-desmethyl spirolide C and gymnodimine during the electrospray analyses. Gradient elution was achieved using the following buffer system:25 (A) water and (B) 95% acetonitrile, each with 2 mM ammonium formate and 50 mM formic acid. A Phenomenex (Torrance, CA) Luna C18 50 × 2.0 mm column with 3-µm particles separated these two compounds in a 5-min run (initial 0.5 min held at 30% B, followed by a linear increase to 70% B in 4.5 min), with a 5-min reequilibration step, at a flow rate of 0.25 mL/min. Electrospray Mass Spectrometry. Electrospray data were acquired using an MDS Sciex API 4000 triple quadrupole mass spectrometer in positive ion mode. The following operating source parameters were used: spray voltage 5.0 kV, declustering potential 50 V, and source temperature 450 °C. MRM transitions monitored were m/z 692.5 f 164.2 for Des-C and m/z 508.6 f 174.3 for (25) Cembella, A. D.; Lewis, N. I.; Quilliam, M. A. Nat. Toxins 1999, 7, 197206.

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

1511

GYM, each with a dwell time of 500 ms and unit resolution, set for both Q1 and Q3. Collision energy was maintained at 60 eV, and ultrapure nitrogen was used as a collision gas at a pressure setting of 12 (arbitrary units). RESULTS AND DISCUSSION The investigation of a MALDI triple quadrupole instrument for the analysis of spirolide toxins in complex samples involved several steps. Initially, the MALDI sample preparation and ionization parameters were addressed to ensure the appropriate conditions were used for optimal results. Once the optimization procedure was complete, a full quantitative study was performed for the analyte Des-C, the main toxin component present in the crude extract and the only spirolide compound for which a pure, reference standard is obtainable. To validate the quantitative results for MALDI, the same samples were also subjected to LC/ MS analysis with electrospray ionization. A comparison of these two methods provided good evidence for the possibility of obtaining accurate quantitative data from MALDI experiments from complex samples. Finally, the unique features possible from the combination of MALDI with a triple quadrupole instrument are illustrated with results from specialized scan functions, such as precursor ion and neutral loss experiments. Species from the same compound class can be selected through the use of these specific scans. Also, it is possible to obtain CID spectra from lowabundant species in complex samples with good signal-to-noise ratios. These results are summarized in three main sections below. Optimization of MALDI analysis. The internal standard chosen for the quantitative experiments was GYM (Figure 1), a marine toxin with important structural similarities to the spirolides. Gymnodimine was thought to act similarly to Des-C in the crystallization process and subsequent MALDI analysis. It has also been shown that Des-C and GYM exhibit similar ionization efficiencies under ESI conditions.5 Since spirolide toxins had never before been analyzed by MALDI, several factors in both sample preparation and instrumentation were initially optimized. For sample preparation, three different organic matrixes were tested, namely, CHCA, DHB, and SA, with different organic solvents as well as percent organic content. After selecting the best matrix and solvent composition from these experiments, different matrix concentrations and matrix-to-analyte ratios were compared. In the matrix optimization experiments, CHCA gave the highest signal intensities for Des-C and GYM. It also exhibited the least amount of possible interference peaks with the analyte and internal standard. Both methanol and acetonitrile were investigated as organic solvent for preparing the samples, and methanol samples exhibited an increase in intensity for Des-C and GYM. Better sensitivity was achieved for the 75% methanol samples, when compared to 25 and 50% content. For further optimization of the sample preparation procedure, Des-C and GYM MH+ peak intensities were monitored using matrix-to-analyte ratios of 10, 100, 1000, and 10 000. The best analyte sensitivity versus matrix interference was achieved using a ratio of 1000. Also, constant matrix concentrations of 2.5 and 5 mg/mL were employed for a three-point analysis of Des-C samples at 1, 2, and 5 µg/mL prepared with a final composition of 75% methanol (+ 0.1% TFA). Ultimately, 2.5 mg/mL CHCA gave more intense signals for both analyte and internal standard. When these results were compared, it was noticed that the GYM signal was 1512

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

suppressed as the concentration of Des-C was increased above the concentration of the internal standard. This phenomenon of IS signal suppression at high analyte concentration was not important unless the concentration of Des-C was in great excess (∼10×) compared to gymnodimine. Therefore, this trend was carefully considered when concentrations were chosen for the quantitative analysis experiments. The highest ratio of Des-C/ GYM in this study was 2.8, at which the ion suppression effect was not apparent. Additionally, certain instrumental parameters were important for the MALDI analysis, namely, laser pulse frequency and energy. The laser was run at 1000 Hz for the quantitative experiments in order to ensure maximal ion currents. A laser energy of 5.1 µJ (50%) was chosen as a compromise to achieve good ionization efficiency with minimal fragmentation. At lower laser power settings, poor intensities were observed; however, for values higher than 50%, significant in-source fragmentation reduced signal intensities for the protonated molecule. Quantitative experiments were performed with a constant matrix concentration of 2.5 mg/mL for different analyte concentrations and a constant matrix-to-analyte ratio of 1000, to compare the potential importance of maintaining a constant ratio throughout the analyte concentration range on quantitative results. It has been suggested previously that this does in fact play an important role in quantitative data for MALDI.26 Figure 3 illustrates a summary of the results from the quantitation of Des-C with and without the use of the internal standard. The SIM calibration curves are shown using a constant matrix concentration in Figure 3a, with an inset exhibiting the curves using absolute signal intensities for Des-C. The same is shown in Figure 3b, but a constant matrixto-analyte ratio was maintained throughout the entire concentration range. It is clearly seen that linearities were improved when the ratios of Des-C/GYM peak areas were considered. Also, interday variations are decreased when an internal standard is employed in the analysis. It is therefore imperative that the internal standard be used, especially in the case of MALDI analysis, which is continuously plagued with signal irreproducibilities due to inhomogeneous sample spots.27,28 The calibration curves using the constant matrix ratio gave slightly better R2 values (average values of 0.998 vs 0.993); however, the constant matrix concentration curves still yielded excellent results. It was therefore decided to go forward with the constant matrix concentration, as otherwise quantitation of unknown samples would be much more difficult. Ensuring a constant matrix-to-analyte ratio would require an additional analysis step to estimate the concentration in unknown samples. Subsequent quantitative experiments were performed using MRM conditions. The CID spectra of the protonated molecule of Des-C (m/z 692.5) exhibited very similar fragment ions from both MALDI and ESI (Figure 4a and b). These spectra were acquired with a collision energy of 45 eV, because these conditions yielded good structural information. Two structurally significant ions were identified at m/z 164.3 and 444.6. For the MRM experiments, the collision energy was increased to 60 eV for higher sensitivity of the transition from m/z 692 to 164. Above this collision energy, (26) Kang, M.-J.; Tholey, A.; Heinzle, E. Rapid Commun. Mass Spectrom. 2000, 14, 1972-1978. (27) Cheng, Y.; Hercules, D. M. Microchem. J. 2002, 72, 255-267. (28) Mims, D.; Hercules, D. Anal. Bioanal. Chem. 2003, 375, 609-616.

Figure 4. Product ion spectra following CID in QqQ instrument with MALDI (a) and electrospray ionization (b). Spectra shown are at collision energy of 45 eV. CID conditions were later optimized for MRM quantitative analysis (formation of m/z 164).

Figure 3. Optimization of MALDI quantitation. Each experiment was performed 3-fold (interday variation: 2 day 1; [ day 2; 9 day 3). (a) SIM detection with constant matrix concentration (inset without IS); (b) SIM detection with constant analyte/matrix ratio (inset without IS); and (c) MRM detection with constant matrix concentration.

scattering losses become predominant and the signal for this transition was consequently reduced. Under the same CID conditions, the most predominant fragment ion in the product ion spectra of GYM (m/z 508) was seen at m/z 174; therefore, the transition from m/z 508 to 174 was monitored as the internal standard signal for quantitative results. The standards were prepared with a constant matrix concentration of 2.5 mg/mL and the same analyte and internal standard concentrations as previously used for the SIM experiments. The MRM method was obviously much more selective for quantitative experiments;

therefore, the analysis of unknown samples employed the MRM method. Figure 3c shows three interday analyses of Des-C standard curves with GYM as the internal standard and a constant matrix concentration. Linearities were excellent and very consistent with an averaged R2 value of 0.995. The slopes of the three linear regression curves deviate slightly due to the fluctuation of the CAD gas pressure in the collision cell of the instrument. In general, MRM provides a convenient way of avoiding interferences from MALDI background signals in the low-mass range. Another important advantage of the triple quadrupole instrument is its ability to obtain MRM data with a higher duty cycle in comparison to other tandem instruments, such as a quadrupole time-of-flight MS.20 MALDI Quantitation of Spirolide Toxins. Once the optimization procedure for the MALDI analysis of these toxins was complete, a full quantitative study was performed for several unknown samples derived from a crude extract. The samples employed in this study were prepared with semipreparative masstriggered LC fractionation of a crude methanolic extract from a culture of A. ostenfeldii.5 Several fractions resulted with different concentrations and compositions of spirolide toxins. All fractions (samples 1-7) with substantial Des-C concentrations were chosen as unknown samples for obtaining quantitative results. The concentration of Des-C was also determined in the crude extract with two separate dilution factors (samples 9 and 10). All samples contained other spirolide analogues; however, these did not interfere with the Des-C measurements. Also, four samples (samples 8, and 11-14), including the crude extract, were chosen and cleaned-up with an additional solid-phase extraction procedure (see Experimental Section). These SPE prepared samples were added in order to determine whether additional sample cleanup would improve the quantitation results from the MALDI analysis. A standard curve of Des-C covering the concentration range of 0.01-1.75 µg/mL as well as appropriately diluted unknown samples were prepared with a constant matrix concentration of 2.5 mg/mL, gymnodimine concentration of 0.625 µg/mL, and final solvent composition of 75% methanol with 0.1% TFA (vide supra). Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

1513

Table 1. Summary of Quantitative Results for Unknown Samples from MALDI and ESI Experiments MALDI quantitation

a

ESI quantitation

sample

concn of run sample (µg/mL)

concn of original samplea (µg/mL)

concn of run sample (µg/mL)

concn of original samplea (µg/mL)

1 2 3 4 5 6 7 8 9 10 11 12 13 14

0.01 0.45 0.22 0.21 0.28 0.25 0.13 0.34 0.07 0.15 0.58 0.39 1.04 0.12

0.05 90.4 44.2 20.6 27.6 1.00 2.63 3.41 7.21 7.41 23.2 15.6 41.7 4.70

0.0028 0.039 0.019 0.018 0.022 0.023 0.012 0.027 0.0079 0.014 0.063 0.054 0.135 0.015

0.07 83.5 38.2 18.3 22.4 0.94 2.4 2.8 7.3 6.8 27.2 22.9 58.6 5.81

Different dilution factors were applied in order to have the samples fall within the dynamic range of each method.

Each standard and sample was spotted 3-fold on the MALDI plate with each spot being sampled twice. The resulting calibration curve exhibited excellent linearity (R2 ) 0.999) throughout the concentration range. Average values for the ratio of Des-C/GYM peak areas from the six repetitive measurements gave very good relative standard deviations for each of the unknown samples, all but one of which were under 10% (Table 1). Importantly, the RSD values did not improve after the SPE step was added to the sample preparation. Therefore, the MALDI method yielded excellent quantitative results directly from the semipurified fractions and the crude extract. The corresponding concentrations for the undiluted samples were calculated and are represented in Table 1. To further validate the results from this method, the same samples were also quantified using an LC/MS method with electrospray ionization. LC/ESI-MS in the MRM mode represents the most common method for the quantitation of small polar molecules, such as pharmaceutical drugs. The same MRM transitions were monitored as for the MALDI analysis while also maintaining identical CID conditions. The ESI standard curve extended from 0.0005 to 0.15 µg/mL, and all samples were diluted an extra 10-fold compared to the MALDI samples. Des-C and GYM were well separated with better than baseline resolution at elution times of 1.5 and 0.9 min, respectively. This is important as it has been suggested previously that coelution of analyte and internal standard may cause ion suppression in ESI methods.29 The ESI calibration curve also displayed excellent linearity (R2 ) 0.998). A significant correlation between these two methods would allow the conclusion that accurate data could be achieved with MALDI even with complex samples. The calculated concentrations obtained from the ESI data are summarized alongside those from the MALDI results in Table 1. The average RSD for the MALDI analysis was 7.4 ( 2.4%. The cross-correlation plot in Figure 5 illustrates good agreement between the two methods. Ideally, the comparison would result in a linear plot with a slope of 1. The corresponding curve deviates only slightly from the ideal situation, thus providing excellent confidence in the accuracy of the MALDI method. The correlation between the two methods did not improve when the SPE samples were considered alone. It was therefore (29) Sojo, L. E.; Lum, G.; Chee, P. Analyst 2003, 128, 51-54.

1514

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

Figure 5. Cross-correlation diagram comparing MALDI and ESI quantitation results and accuracy of methods. (Note that the line in the diagram represents the ideal behavior; i.e., slope, 1.)

concluded that the SPE step was not required for obtaining quantitative MALDI data. Additionally, this method benefits from several inherent advantages of MALDI over electrospray. It has been reported that this ionization method is much more tolerant of buffers and salts than ESI and, therefore, is potentially more amenable to the direct analysis of biological samples.30,31 Also, repetitive sampling is a clear advantage for MALDI. For example, one could readily run a data-dependent experiment, starting with an initial full scan experiment, followed by targeted product ion scan analyses on precursor ions of interest, as well as precursor ion and neutral loss scans (vide infra), all on the same spot. We could therefore easily obtain much more information from one MALDI spot than is possible for a single sample injection in ESI LC/MS, since after the elution of the chromatographic peak, the analyte is no longer present and a further injection is required. When sample amounts are limited, MALDI has a clear advantage. Both MALDI and ESI (30) Kast, J.; Parker, C. E.; van der Drift, K.; Dial, J. M.; Milgram, S. L.; Wilm, M.; Howell, M.; Borchers, C. H. Rapid Commun. Mass Spectrom. 2003, 17, 1825-1834. (31) Jonsson, A. P. Cell. Mol. Life Sci. 2001, 58, 868-884.

Figure 6. MALDI experiments on crude phytoplankton extract. (a) full scan experiment (i) m/z 150-900 and (ii) 500-750; (b) precursor ion scans (PIS) for fragment ions at (i) m/z 164 (CE 60 eV) and (ii) 444 (CE 45 eV); (c) neutral loss scans (NLS) of (i) 528 and (ii) 248 Da. The boldface m/z values represent protonated spirolide molecules, whereas the others are derived from water losses occurring in the MALDI source.

calibration curves yield excellent linearities; however, the ESI curve extends to concentrations of an order of magnitude lower. Therefore, the ESI method is perceived as having lower detection limits. Yet, the absolute amount sampled is much less for the MALDI analysis. Although there are similar amounts applied in both methods (injected vs spotted), the electrospray method consumes 100% of the sample with each analysis whereas the MALDI method samples a very minimal area from the spot. Previous analysis of spirolide toxins in A. ostenfeldii cultures have estimated 13-desmethyl spirolide C to have a concentration of 54 pg‚cell-1.32 Therefore, we suggest that single-cell analysis is feasible with the present method, since 10 pg on target was easily detected and quantified. Single-cell analysis is of great interest, as it would allow the differential toxin profiles and spirolide content to be studied for different cells in the same culture. Moreover, the limit of detection for the MALDI method could be improved substantially with a longer sampling time on each spot; this would, however, decrease the throughput of the (32) Cembella, A. D.; Lewis, N. I.; Quilliam, M. A. Phycologia 2000, 39, 67-74.

method. Ultimately, the high-throughput nature of MALDI is the most important advantage over electrospray. The present method sampled each spot within 3-5 s with the entire sample set being analyzed in under 1 h. Conversely, the ESI method needed over 10 min for each sample due to chromatographic separation, column equilibration, and other overhead steps. As a result, analyzing the entire set of samples using the ESI method required ∼12 h. Of course, slight improvements on the throughput of the ESI method could be achieved; however, it would still need a much longer analysis time than the MALDI experiment. Similar methods for the quantitation of small molecules in complex biological matrixes would be extremely attractive for the pharmaceutical industry, where a great need for increased productivity and efficiency exists. Unique Advantages of MALDI Triple Quadrupole MS. Another aspect of this study represents the use of certain dedicated scan functions of a triple quadrupole not previously employed in MALDI analysis. Precursor ion and neutral loss scans are particularly useful for the identification of specific Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

1515

Figure 7. (a) Full scan experiment (m/z 500-750) from purified fraction of phytoplankton extract. (b) Product ion spectra from individual spirolide components at m/z 692, 694, 706, and 708 with collision energy of 45 eV (* denotes the precursor ion for CID).

compound classes. Several common fragment ions and neutral losses are observed within the spirolide toxin family. Their characteristic fragmentation behavior can aid in their detection as well as elucidating their structure. The most informative ions in the CID spectra of Des-C are seen at m/z 164 and 444. We can easily identify spirolide-type compounds in a crude extract using these ions for specific linked scans. A full scan spectrum, shown in Figure 6a, shows a wide range of ions, including those related to the organic matrix as well as other interferences present in the extract. Precursor ion scans were performed for both m/z 164 and 444 ions (Figure 6b). For the precursor ion spectrum of m/z 164, we can identify peaks at m/z values of 692, 694, 706. and 708, which represent previously identified spirolide compounds (seen in boldface type in the figure),22,33 as well as some other peaks as a result of in-source fragmentation of either m/z 692 or 706. A precursor ion at m/z 598 was also identified in this experiment, which represents an unknown compound thought to be a spirolide analogue present in the phytoplankton extract.5 A similar experiment for identifying precursors of the fragment ion at m/z 444 gave signals for the protonated molecule of Des-C (m/z 692), a fragment ion corresponding to a single water loss (m/z 674), and a very small signal for m/z 694. The latter corresponds to that of 13-desmethyl spirolide D, a minor component present in the extract. None of the other spirolide species present in the crude extract yield a (33) Hu, T.; Burton, I. W.; Cembella, A. D.; Curtis, J. M.; Quilliam, M. A.; Walter, J. A.; Wright, J. L. C. J. Nat. Prod. 2001, 64, 308-312.

1516 Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

product ion at m/z 444. Instead, spirolide C and D give analogous peaks at m/z values of 458, for their protonated molecules at m/z 706 and 708, respectively. Two other minor spirolide toxins, with m/z values of 694 and 708, have previously been detected in the extract and exhibit fragment ions at m/z 460.5 Corresponding neutral loss scans were also monitored for the dissociation pathways forming the two main structurally informative ions of Des-C. A neutral loss scan of 528 Da gave a sole signal at m/z 692; however, different spirolides were detected through a neutral loss scan of 248 Da. The experiment for detecting precursors exhibiting neutral losses for 528 Da would allow the identification of spirolide analogues with structural changes on the cyclic imine portion of the molecule.24 The latter experiment, for the neutral loss of 248 Da, revealed precursor ions at m/z values of 692, 706, and 604, corresponding to the protonated molecules of 13desmethyl spirolide C, spirolide C, and an unknown spirolide analogue, respectively.5 In the next step, we looked at the CID spectra of the tentatively identified spirolides from the precursor ion and neutral loss scans. A full scan spectrum (m/z 500-750 region) for one of the fractions containing several of the toxins is shown in Figure 7a. Protonated spirolide compounds at very different concentration levels were detected at m/z 692, 694, 706, and 708. Each component was selected to undergo CID (Figure 7b). Even the minor components resulted in structurally informative CID spectra with excellent signal-to-noise ratios. The spectra consequently classified the different spirolide species as 13-desmethyl spirolide C (m/z 692),

13-desmethyl spirolide D (m/z 694), spirolide C (m/z 706), and spirolide D (m/z 708). CONCLUSION The performance of a MALDI triple quadrupole mass spectrometer for rapid identification and quantitation of spirolide toxins in complex phytoplankton extracts has been investigated. To our knowledge, this study represents the first extensive report on MALDI analysis of nonpeptidic toxins. The employed highfrequency (kHz) laser generated a semicontinuous ion beam from the MALDI sample plates, thus taking advantage of the high duty cycle obtained in sensitive triple quadrupole MRM experiments. Following extensive quantitative analyses and comparison with an electrospray method, the present MALDI method has been established as a very accurate and precise approach. 13-Desmethyl spirolide C was successfully quantified in a range of complex samples without extensive sample preparation steps. In all experiments, excellent linearities were obtained over at least 2 orders of magnitude and a precision of 7.4 ( 2.4% RSD. No systematic errors were apparent in the MALDI method as shown by a direct comparison to an electrospray LC/MS/MS method. Most importantly, the MALDI technique was very fast; each sample spot was analyzed in less than 5 s as compared to several minutes with the electrospray assay. In addition to determining concentrations, the combined MALDI triple quadrupole method was also very useful for identifying unknown spirolide analogues. We were able to demonstrate that, through the use of dedicated linked scan functions such as precursor ion and neutral loss scans, several spirolide compounds could be tentatively identified directly in

complex samples, without the usual time-consuming chromatographic preseparation steps, as conducted, for example, in most electrospray LC/MS assays of toxins. Moreover, high-quality CID spectra were obtained for low-abundant spirolides present in the crude extract. We believe that the results shown in this report are of general interest to researchers in many fields of analytical mass spectrometry, in particular to laboratories in the drug industry, where increased speed and efficiency of analyses are very often directly profitable. Thus, the combination of MALDI and a triple quadrupole analyzer would be particularly advantageous in the drug discovery process. A commercially available instrument of this type could find use in many other fields as well, where fast and quantitative monitoring of biological or environmental samples is required. ACKNOWLEDGMENT The authors thank Suzanne Bos (NRC, Halifax) for assistance with the experiments during the optimization of the MALDI method. We also thank MDS Sciex for donation of the prototype MALDI triple quadrupole instrument. L.S. acknowledges financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC) and the National Research Council’s Graduate Student Scholarship Supplement Program (GSSSP).

Received for review December 13, 2004.

September

9,

2004.

Accepted

AC0486600

Analytical Chemistry, Vol. 77, No. 5, March 1, 2005

1517