Direct Profiling and Imaging of Epicuticular Waxes on - American

Mar 16, 2009 - sepals were physically dissected from the flowers and individually attached to the sample plate for metabolite profiling. Three replica...
1 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2009, 81, 2991–3000

Direct Profiling and Imaging of Epicuticular Waxes on Arabidopsis thaliana by Laser Desorption/ Ionization Mass Spectrometry Using Silver Colloid as a Matrix Sangwon Cha,†,‡ Zhihong Song,§ Basil J. Nikolau,§ and Edward S. Yeung*,†,‡ Ames Laboratory, U.S. Department of Energy, Ames, Iowa, and Department of Chemistry, Department of Biochemistry, Biophysics and Molecular Biology, Iowa State University, Ames, Iowa 50011 Colloidal silver laser desorption/ionization (LDI) mass spectrometry (MS) was employed to directly profile and image epicuticular wax metabolites on a variety of different surfaces of Arabidopsis thaliana leaves and flowers. Major cuticular wax compounds, such as very long-chain fatty acids, alcohols, alkanes, and ketones, were successfully detected as silver adduct ions. The surface metabolites of different flower organs (carpels, petals, and sepals) were profiled for the first time at a spatial resolution of ∼100 µm. In addition, mass spectral profiles and images were collected from wild type and a mutant strain, which carried alleles that affect the surface constituents of this organism. One of these mutant alleles (cer2-2) is in a gene whose biochemical functionality is still unclear, although its effect on normal epicuticular wax deposition was the characteristic that led to its original identification. Variations of wax products between different spatial locations for wild type and for a mutant strain were investigated by normalizing the ion intensities to a reference peak ([107Ag + 109Ag]+). The spatially resolved surface metabolite profiling data of this mutant has provided new insights into the complexity of epicuticular wax deposition at the cellular-resolution scale. This MS-based metabolite imaging technology has the potential to provide valuable data for dissecting metabolism in multicellular organism at the level of single cells. As a barrier to the abiotic environment, plants produce the cuticle, which protects them from external stresses, such as water loss, insect herbivores, and fungal pathogens. The cuticle is composed of two hydrophobic components, cutin and cuticular waxes.1,2 A fatty acid-based polyester, cutin serves as the structural backbone of the cuticle, and it is covered and blended with aliphatic long-chain fatty acids and their derivatives, cuticular waxes. Molecular genetic and metabolite profiling studies have * To whom correspondence should be addressed. E-mail: yeung@ ameslab.gov. † Ames Laboratory, U.S. Department of Energy. ‡ Department of Chemistry, Iowa State University. § Department of Biochemistry, Biophysics and Molecular Biology, Iowa State University. (1) Heredia, A. Biochim. Biophys. Acta 2003, 1620, 1–7. (2) Nawrath, C. Curr. Opin. Plant Biol. 2006, 9, 281–287. 10.1021/ac802615r CCC: $40.75  2009 American Chemical Society Published on Web 03/16/2009

been carried out on a number of plant species, primarily maize and Arabidopsis thaliana, and achievements in this field have been reviewed recently.3 Wax-deficient mutants have been identified primarily by examining the visual appearance of plant surfaces, and many mutant loci have been identified in Arabidopsis, called eceriferum (cer),3,4 and maize, called glossy.5 In addition to visual screening, condensed GC methods have been employed to analyze these mutant lines.6 Twenty-four cer loci have been identified to date. The wax profiles of cer mutants have been analyzed by gas chromatography mass spectrometry (GC/MS) and, on the basis of differences in the carbon chain lengths and amounts of wax constituents, the function of CER genes and their products have been hypothesized.6-15 In a few instances, molecular cloning of the cer locus has provided further insights into the biochemical function of each CER protein; however, this has not always been possible; thus, the specific biochemical functions of most CER genes are still unknown.3 GC-based methods that are used to analyze cuticular wax components are always concomitant with chemical extraction of the wax compounds with organic solvents, and derivatization, to protect functional groups and enhance volatility of the compounds. Therefore, these profiling procedures are labor-intensive. More(3) Samuels, L.; Kunst, L.; Jetter, R. Annu. Rev. Plant Biol. 2008, 59, 683– 707. (4) Koornneef, M.; Hanhart, C. J.; Thiel, F. J. Hered. 1989, 80, 118–122. (5) Schnable, P. S.; Stinard, P. S.; Wen, T. J.; Heinen, S.; Weber, D.; Schneerman, M.; Zhang, L.; Hansen, J. D.; Nikolau, B. J. Maydica 1994, 39, 279–287. (6) Rashotte, A. M.; Jenks, M. A.; Ross, A. S.; Feldmann, K. A. Planta 2004, 219, 5–13. (7) Goodwin, S. M.; Rashotte, A. M.; Rahman, M.; Feldmann, K. A.; Jenks, M. A. Phytochemistry 2005, 66, 771–780. (8) Hannoufa, A.; McNevin, J.; Lemieux, B. Phytochemistry 1993, 33, 851– 855. (9) Jenks, M. A.; Rashotte, A. M.; Tuttle, H. A.; Feldmann, K. A. Plant Physiol. 1996, 110, 377–385. (10) Jenks, M. A.; Tuttle, H. A.; Eigenbrode, S. D.; Feldmann, K. A. Plant Physiol. 1995, 108, 369–377. (11) Jenks, M. A.; Tuttle, H. A.; Feldmann, K. A. Phytochemistry 1996, 42, 29– 34. (12) Lai, C.; Kunst, L.; Jetter, R. Plant J. 2007, 50, 189–196. (13) McNevin, J. P.; Woodward, W.; Hannoufa, A.; Feldmann, K. A.; Lemieux, B. Genome 1993, 36, 610–618. (14) Rashotte, A. M.; Jenks, M. A.; Feldmann, K. A. Phytochemistry 2001, 57, 115–123. (15) Rashotte, A. M.; Jenks, M. A.; Nguyen, T. D.; Feldmann, K. A. Phytochemistry 1997, 45, 251–255.

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

2991

over, spatial information concerning the distribution of cuticular wax component at the tissue and cellular level is lost due to sample sizes, which is determined primarily by the sensitivity of the GC methods used in these analyses. In addition, chemical extraction protocols cannot be exclusively restricted to epicuticular waxes, and the degree of extraction is also ambiguous. Recently, laser desorption/ionization mass spectrometry (LDI MS) has been employed to analyze various wax compounds and alkane oligomers.16-19 In these studies, silver-based matrixes yielded adduct ions of alkanes and their derivatives. The advantage of silver-based matrixes over other transition metal-based matrixes is that the reactivity of silver to alkanes is lower than that of any other transition metal. Therefore, silver mostly generates intact silver adduct ions without fragmentation.20 Direct LDI MS analysis of wax compounds on Arabidopsis stems and leaves with silver colloid as a matrix was successfully performed in our laboratory.21 Recently, silver colloids were also utilized as a matrix for the LDI MS analysis of various olefin-containing compounds.22 The advantage of colloidal silver over other silver matrixes, such as ultrafine silver powder, silver nitrate, and AgTFA, is that colloidal silver solution can be applied more easily and more homogeneously onto hydrophobic surfaces than other silver matrixes. Because colloidal silver does not form crystals when it dries, good reproducibility of the mass spectral signals is expected, making it possible to examine localization of the compounds of interest on the plant surface. Here we present the improved colloidal silver-LDI MS methodology for the direct profiling and imaging of epicuticular wax metabolites. By optimizing the colloidal silver spraying, and by using the LDI linear ion trap mass spectrometer with a programmable rastering capability, spatially resolved mass spectral profiles and chemically selective images were successfully generated for Arabidopsis leaves and flowers. After normalization of the mass spectra with respect to a reference peak, relative abundance information was obtained at a high spatial resolution not possible by traditional GC/MS analysis. In this study the cuticular wax compositional variations between a wild type and a mutant Arabidopsis genetic stock that carried a mutation in a cuticular wax biosynthetic gene (cer2) were compared for the first time by silver-LDI MS, and these results were compared to GC/MS analysis. EXPERIMENTAL SECTION Chemicals. Standards such as n-alkanes, alcohols, esters, ketones, and fatty acids, consisting of 13-32 carbon atoms, N,Obis(trimethylsilyl)trifluoroacetamide with trimethylchlorosilane (BSTFA/TMCS), and Murashige and Skoog basal salt mixture were obtained from Sigma-Aldrich (St. Louis, MO). A mixture of n-alkanes (C12-C60) was purchased from Supelco (Bellefonte, (16) Dutta, T. K.; Harayama, S. Anal. Chem. 2001, 73, 864–869. (17) Kuhn, G.; Weidner, S.; Just, U.; Hohner, G. J. Chromatogr., A 1996, 732, 111–117. (18) Pruns, J. K.; Vietzke, J. P.; Strassner, M.; Rapp, C.; Hintze, U.; Konig, W. A. Rapid Commun. Mass Spectrom. 2002, 16, 208–211. (19) Yalcin, T.; Wallace, W. E.; Guttman, C. M.; Li, L. Anal. Chem. 2002, 74, 4750–4756. (20) Chen, R.; Li, L. J. Am. Soc. Mass Spectrom. 2001, 12, 367–375. (21) Sluszny, C.; Yeung, E. S.; Nikolau, B. J. J. Am. Soc. Mass Spectrom. 2005, 16, 107–115. (22) Sherrod, S. D.; Diaz, A. J.; Russell, W. K.; Cremer, P. S.; Russell, D. H. Anal. Chem. 2008, 80, 6796–6799.

2992

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

Figure 1. Arabidopsis flower organ samples prepared on an LDI target plate for silver-LDI mass spectral profiling and imaging.

PA). Water-based colloidal silver (20 ppm) was purchased from Purest Colloids, Inc. (Westhampton, NJ). Double-deionized water was used from a MilliQ water purification system (Framington, MA). All other chemicals were purchased from Fisher Scientific (Fairlawn, NJ). Plant Growth Conditions. A. thaliana ecotype Landsberg erecta (Ler-0) and the genetic stock CS8, which carried the mutant alleles cer2-2, ap2-1, and bp1, were obtained from the Arabidopsis Biological Resource Center (ABRC) (Columbus, OH). Seeds were sterilized and sown on Murashige and Skoog basal salt mixture media in 10 cm × 10 cm square Petri dishes. The dishes were first placed at 4 °C for 4 days to break seed dormancy and then in a growth room for 10 days. On the 15th day the seedlings were transferred to soil in 5 cm × 5 cm pots and placed in a growth room. Leaves were collected on the 26th day, and flowers were collected on the 34th day. The growth room was kept at 24 °C with continuous illumination at 85 µE m-2 s-1 and ambient relative humidity. Plant Sample Preparation for LDI Mass Spectral Profiling and Imaging. Collected flower and leaf samples from plants were stored in the ice before attaching them to sample plates. They were attached onto a stainless steel target plate of similar dimensions as a microscope glass slide using a conductive doublesided tape (3M, St. Paul, MN). To avoid damage to the plant surface, contact with forceps was minimized and air pressure from a nitrogen gas cylinder was used for attaching samples firmly onto the sample plate. All samples were dried in moderate vacuum (∼50 Torr) for 30-60 min. As shown in Figure 1, carpels, petals, and sepals were physically dissected from the flowers and individually attached to the sample plate for metabolite profiling. Three replicates of leaves and whole flowers and five replicates of flower parts for each genotype were prepared. The spraying device for applying colloidal silver solution was composed of a computer-controlled syringe pump with a 500 µL syringe (Kloehm Ltd., Las Vegas, NV), a MicroFlow PFA-ST nebulizer (Elemental Scientific Inc., Omaha, NE) with a 0.25 mm i.d. sample uptake capillary, and a helium gas cylinder. Spraying time and speed was programmed by “WinPump” the software, which was provided by Kloehm. To obtain a homogeneous coverage of silver particles, several spraying intervals with a small volume of colloidal silver solution were applied per application. This was more desirable than applying the colloidal silver solution in a one-time spraying with a large volume. The reason is that

spraying large amounts caused big droplets, which led to silver particles aggregating when they were dried. Therefore, the parameters for spraying were optimized as 100 µL/min flow rate with 12.5 µL of spraying volume and eight spraying periods; the sample was air-dried after each application. The distance between the nebulizer tip and the sample was 12.4 cm, and the area covered by colloidal silver on the sample plate was about 4.5 cm2. Spraying onto the leaf or the flower was performed without moving the sample plate because sample sizes were much smaller than the spray area. Plant Sample Preparation for Gas Chromatography/Mass Spectrometry. Flowers or leaves were each collected from plants and collectively weighed. An aliquot of the internal standard (hexadecane) was added to the surface of the plant material. The plant material was completely immersed in chloroform for 60 s. The cuticular wax extract was dried under a stream of nitrogen gas. Samples were derivatized using BSTFA/TCMS (65 °C, 30 min) for GC/MS analysis. Three replicates of leaves and flowers were prepared. Mass Spectrometry. For LDI mass spectral profiling and imaging, a Thermo Finnigan LTQ linear ion trap mass spectrometer equipped with a vMALDI source (Mountain View, CA) was used. A fiber-optic guided nitrogen laser (337 nm, maximum energy of 280 µJ/pulse, and maximum frequency of 20 Hz) was used as a laser source. The laser spot size was 100 µm at the target plate surface. The intermediate-pressure (0.17 Torr) sampling environment was kept by nitrogen gas flow control, and this allows softer ionization compared to high-vacuum environments (∼10-6 Torr). All mass spectra were collected in the positiveion mode, and the scanning range was set to m/z 200 to m/z 1000. For unknown peaks acquired directly from plant samples, putative peak identification was carried out by matching masses with those detected from GC/MS experiment. For overlapped peaks due to two stable silver isotopes and similar monoisotopic masses, the major compounds detected were determined by comparing the first-generation product ion spectra from unknown peaks with those from standards or from literature data. GC/MS analysis was performed with an Agilent 6890 GC interfaced to a 5973 mass spectrometer (Agilent Technologies). An HP-5 ms column (30 m × 0.25 mm i.d. coated with a 0.25 µm film, Agilent Technologies) was used, and temperature gradient was programmed from 80 to 320 °C at 5 °C/min with He flow rate at 2.0 mL/min. Operating parameters were set to 70 eV (electron ionization) for ionization voltage and 280 °C for interface temperature. The GC/MS data files were deconvoluted by NIST AMDIS software and searched against an in-house mass spectral library and the NIST mass spectral library. Gas chromatograms of cuticular wax extracts and mass spectra for representative wax metabolites are shown in Supporting Information Figures S3 and S4, respectively. LDI Mass Spectra Collection for Profiling and Imaging. For leaves and flower organs, the number of laser shots and the laser intensity scale were examined by collecting mass spectra in a small area (usually 5-10 rastering points) by turning on the automatic gain control feature (AGC, which keeps the ion amounts in the trap at a similar range by varying the number of laser shots) of the mass spectrometer. After optimizing the laser parameters,

a fixed number of laser shots without AGC was used for collecting mass spectra over the whole sample. For whole flower imaging, however, varying numbers of laser shots that were controlled by AGC were applied because each flower part had different surface characteristics and wax loads which could affect ion yields. The sample was scanned with a step size which ranged from 50 to 100 µm. Mass spectral images for whole flowers were processed by using the custom software from Thermo (ImageQuest 1.0). The normalized intensity value which is defined as the fractional peak intensity compared to the total ion current (TIC) of each mass spectrum was used for presentation of chemical abundance information. The mass window was set to 0.8-1.0 Da. For comparing relative abundances among samples, intensities relative to the peak intensity at m/z 215.8 ([107Ag + 109Ag]+) were used. RESULTS AND DISCUSSION Silver-LDI MS Profiling of Metabolites from Arabidopsis Leaf Surfaces. Supporting Information Figure S1 shows the averaged silver-LDI mass spectrum taken from a wild-type Arabidopsis leaf surface. As shown in the inset of Supporting Information Figure S1, the molecular ion at m/z 215.8 corresponding to [107Ag + 109Ag]+ had the highest intensity, and this was consistent through all mass spectra collected under our experimental conditions. Therefore, for quantitative purposes, the relative intensities of all peaks were scaled with respect to the peak intensity at m/z 215.8, and these normalized data were used in interpreting the mass spectra of cuticular wax metabolites. Because the two stable isotopes of silver are present with similar abundances (107Ag, 51.839%, and 109Ag, 48.161%), each metabolite produces a group of silver adduct ion peaks including two major ions [monoisotopic mass of the metabolite + 107Ag or 109Ag]+. Although this group of peaks aids to distinguish sample peaks from noise, the multiple metabolite peaks introduced spectral complexity due to peak overlap of compounds that differ in molecular weight by 2 Da. Despite the complexity introduced by these peak overlaps, many of the wax compounds can be identified by focusing on one of the two silver isotope adduct ion peaks as shown in Table 1. In the m/z range of 571-575, for example, C33 alkane and C32 alcohol ion species overlap at m/z 573 as [C33 alkane + 109 Ag]+ and [C32 alcohol + 107Ag]+, but each can be uniquely detected at m/z 571 as [C33 alkane + 107Ag]+ and at m/z 575 as [C32 alcohol + 109Ag]+, respectively. For isobaric ions associated with a single m/z value, the major compounds can be putatively identified by examining fragmentation patterns in the first-generation product ion mass spectra. For example, major ion species at m/z 503, 559, and 587 were identified as long-chain fatty acids because loss of water (18 Da) and loss of formic acid (46 Da) in their product ion spectra was always predominantly observed with very similar ratios to those observed from fatty acid standards. However, isobaric ions such as C29 alkane and C28 aldehyde could not be distinguished by fragmentation patterns with the current experimental conditions, and this is the main limitation of the current silver-LDI MS method. Also, as indicated in Table 1, in the range of m/z 529-535, two silver adduct ions of C28 fatty acid (at m/z 531 and 533) overlap with [C29 ketone + 109Ag]+ (at m/z 531) and Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

2993

Table 1. Possible Cuticular Wax Metabolites on Arabidopsis (Ler) Leaves Catalogued by to Mass-to-Charge Ratio as Detected by Silver-LDI MS m/z range

possible ion speciesa

475-477 C24 fatty acid (475) 487-491 C27 alkane (487) C26 aldehyde (487) C26 alcohol (489) 503-505 C26 fatty acid (503) C27 secondary alcohol (503) 515-519 C29 alkane (515) C28 aldehyde (515) C28 alcohol (517) 529-535 C29 ketone (529) C28 fatty Acid (531) C29 secondary alcohol (531) unknown contaminant (533) 543-547 C31 alkane (543) C30 aldehyde (543) C30 alcohol (545) 559-561 C30 fatty acid (559) C31 secondary alcohol (559) 571-575 C33 alkane (571) C32 alcohol (573) 585-589 Unk-A1(585)f C32 fatty acid (587) 597-603 Unk-B1(597) C35 alkane (599) C34 alcohol (601) 613-617 Unk-A2 (613) C34 fatty acid (615) 625-629 Unk-B2 (625) 641-645 Unk-A3 (641) 653-657 Unk-B3 (653) 669-673 Unk-A4 (669)

wax load LDI relative (nmol/g fresh weight)b intensity (%)c 0.6 ± 0.1 24.9 ± 1.0 0.7 ± 0.1 4.3 ± 1.2 2.0 ± 0.4 0.4 ± 0.1 180.0 ± 3.2 8.8 ± 0.3 9.0 ± 1.0 0.4 ± 0.04 5.7 ± 0.4 0.3 ± 0.01 n.a.e 196.8 ± 8.3 6.8 ± 0.5 2.6 ± 0.6 0.3 ± 0.1 0.3 ± 0.01 74.7 ± 3.7 1.3 ± 0.2 n.a. 0.1 ± 0.02 n.a. 2.7 ± 0.4 0.08 ± 0.01 n.a. 0.1 ± 0.01 n.a. n.a. n.a. n.a.

10.0 ± 1.09 15.6 ± 1.41 20.3 ± 2.19 36.0 ± 1.37 8.4 ± 0.18d

32.8 ± 1.56 6.1 ± 0.29 20.5 ± 1.33 6.3 ± 0.23 14.1 ± 0.77 4.6 ± 0.29 12.0 ± 1.14 5.7 ± 0.01 6.1 ± 0.47 4.7 ± 0.87

a Cuticular wax compounds (nominal m/z values of 107Ag adduct ions) listed are from identified compounds in parallel GC/MS analysis. b Wax load values are from GC/MS for three replicates with ±standard error. c Relative intensity (%) is defined as the fractional intensity of the intensity at m/z 215.8 which corresponds to [107Ag + 109Ag]+, and the values for the corresponding m/z ranges are listed with ±standard error. d Due to the overlap with unknown contaminants, only relative intensities at m/z 529 and 531 were included for m/z 529-535. e n.a.: not applicable. f Unk: Unknown ion species. Unknown species were grouped by fragmentation patterns in their first-generation product ion spectra.

[unidentified contaminant + 107Ag]+ (at m/z 533), respectively. Therefore, these overlaps prohibit unambiguously extracting abundance information for this fatty acid. Accurate mass measurement with high mass resolution could be the solution for these problems. Finally, there were several unknown compounds detected in m/z 585-673 range (annotated as “Unk-A” and “Unk-B” in Table 1). Unk-A was detected at m/z 585, 613, 641, and 669, and Unk-B was detected at m/z 597, 625, and 653, as 107Ag adduct ions. Ions in each group are 28 Da apart from each other, and this difference may correspond to a -C2H4- group. For Unk-B, losses of multiples of 42 Da, single loss to quadruple losses were observed, which could correspond to propene. For UnkA, the product ion spectra were very complicated due to overlap with minor unknown ions. Comparison of Metabolite Profiles Obtained by Silver-LDI MS and GC/MS. Table 1 compares the abundance of cuticular wax metabolites as determined by GC/MS analysis and the sum of the relative intensities from silver-LDI mass spectra in each m/z range. These comparisons indicate that the two methods generate data that have a similar trend but are not in total agreement. There are several reasons for this phenomenon. One 2994

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

of the main reasons maybe that the sampling depth by the LDI method is shallower than that of the chemical extraction protocol used for GC/MS analysis. Therefore, the silver-LDI mass spectral profiles obtained directly from cuticular wax surfaces do not represent the total amount of each cuticular wax compound on the surface. The shallower sampling depth by silver-LDI may provide the means of serially profiling cuticular wax compounds relative to the depth of the cuticle layer. For example, plant sterols and triterpenoids were extracted and found to constitute about 6% (∼1.0 µmol/g per fresh weight) of the cuticular wax load as determined by GC/MS method. In contrast, although colloidal silver-LDI could detect plant sterol standards with a very low detection limit, ∼200 amol/10 000 µm2 (sampling area of the laser), these metabolites were not detectable in the LDI mass profiles obtained directly from leaf surfaces. This may be due to the fact that plant sterols are not at the outermost layer of the epicuticular wax surface, which is sampled by silver-LDI method. In addition, as previously reported from our laboratory,21 there are differences in the ionization efficiencies of different cuticular wax compounds depending on their chemical functionality. Therefore, the relationship between the observed signal intensity to the actual amount of each cuticular wax metabolite may differ for different metabolite classes and/or carbon chain length of each metabolite. For example, in silver-LDI MS, metabolite standards of 18-32 carbons showed good peak reproducibility in terms of intensities, but compounds that have fewer carbon atoms, such as C16 fatty acid, showed variable relative intensity depending on laser intensities and concentrations. Similar to the previous study,21 alkanes showed the lowest ionization yields and fatty acids showed the highest ionization yields for the same molar amount. These trends appear to be consistent with the observations of plant surfaces by silver-LDI. For example, the relative amounts of metabolites in the m/z range of 475-477, 503-505, and 559-561 (which contain fatty acids) were higher in silverLDI than those analyzed by GC/MS, whereas in the m/z regions 515-519 and 543-547 (which contain alkanes) the relative abundances were much lower by silver-LDI than by GC/MS analysis. Comparison Cuticular Waxes of Wild-Type and Mutant Leaf Surfaces. On the basis of parallel analyses of leaf cuticular waxes extracted and analyzed by GC/MS, the most abundant metabolites within each overlapping isobaric species were identified (Supporting Information Table S1). To provide a basis for comparing the data generated by the two methods, these analyses were conducted for leaves from both a wild-type Arabidopsis plant (ecotype Ler-0) and an isogenic mutant stock (CS8) that carries three mutant alleles, one of which (cer2-2) specifically affects the normal deposition of cuticular waxes.4 The relative intensities of each metabolite as determined by silver-LDI MS and GC/MS were compared. In order to validly compare these two data sets, metabolite abundance data were all converted to log-ratios (i.e., for each metabolite, Mx, we calculated the value of log2 ([Mx]CS8/ [Mx]Ler), where [Mx]CS8 is the abundance of that metabolite in the CS8 mutant and [Mx]Ler is the abundance of that metabolite in the wild-type Ler control (Figure 2 and Supporting Information Table S1).

Figure 2. Ratios of targeted cuticular wax metabolites in Arabidopsis wild-type (Ler) and CS8 mutant leaves. Ratios from silver-LDI MS (left) and GC/MS (right) analyses were scaled as log2 values. Data points in each colored box correspond to the compound class indicated in the box. The number of carbon atoms in each metabolite is indicated for each data point. Detailed ion species assignments and their relative intensities from silver-LDI MS and the amounts from GC/MS are listed in Supporting Information Table S1. Each silver-LDI MS and GC/MS analysis has three replicates, and error bars corresponding to standard errors are shown.

Upon the basis of these log-ratios (Figure 2), several observations can be summarized. First, with both methods, all alkanes and the C29 ketone were detected at significantly lower intensities or lower amounts on the mutant leaves than wild-type leaves, although the decreases were more pronounced by the GC/MS method than the silver-LDI method. Second, both methods indicate that differences in the amounts of primary alcohols between wild type and the mutant were less significant than the other classes of compounds. Third, the two methods showed the most incongruent data for the fatty acids, with C30 and C32 fatty acids being more abundant on the mutant leaves by both methods, but the relative abundances for C24 and C26 fatty acids being opposite as determined by the two methods. However, it should be noted that fatty acids are minor constituents of the cuticular waxes and their relative abundances are likely to be most susceptible to variation in extraction efficiencies. Silver-LDI MS Imaging of Arabidopsis Flower Surfaces. To demonstrate the utility of LDI MS as a rapid-imaging metabolite profiling tool, we imaged cuticular wax metabolites on Arabidopsis flowers. Flowers were chosen for these analyses because they are minute and are composed of several different organs that have even smaller dimensions (e.g., carpels, petals, sepals, and stamens are each less than 0.3 mm in size). Mass spectral images of wildtype (Ler) flowers are shown in Figure 3. Because different flower organs have different surface characteristics and cuticular waxes, the TIC, which is directly affected by ion yield, varies among the different flower organs. Therefore, all chemically selective images of flowers were processed as normalized intensity, defined as the peak intensity for each metabolite as compared to the TIC of each mass spectrum. In Figure 3, the ion species at m/z 529, which mainly corresponds to [C29 ketone + 107Ag]+, was highly localized to the flower stem and the carpel. Major alkane species, [C29 alkane + 107 Ag]+ and [C31 alkane + 107Ag]+ at m/z 515 and 543,

respectively, were the major species on carpels, sepals, and petals, but their abundance was relatively lower in the tips of petals. Two major cuticular wax metabolites, C31 alkane and C30 alcohol, contribute to mass signals in the m/z range of 543-547. Both of these metabolites contributed to the intensity at m/z 545 as [C31 alkane + 109Ag]+ and [C30 alcohol + 107Ag]+. As shown in Figure 3, this overlapping image was clearly separated into two images by choosing different silver isotope ions, i.e., [C31 alkane + 107Ag]+ at m/z 543 and [C30 alcohol + 109Ag]+ at m/z 547. By comparing the images of the m/z 543 and 547 ions, it is clear that the intensity on the petals and at the central part of the sepals for the m/z 545 image is primarily due to the distribution of the C31 alkane, and not the C30 alcohol. At the tips of petals, fatty acid [C26 fatty acid + 107Ag]+ at m/z 503 was found at higher amounts than other wax compounds. We further assessed the utility of LDI MS in generating chemically selective images of metabolite distributions by comparing the cuticular wax compositions of flowers between wild-type and the mutant stock. Serial mass spectral images were generated with a 0.8 Da mass window scanned at 2.0 Da step sizes through the m/z ranges where cuticular wax compounds are detected. Figure 4A shows chemically selective images for four m/z values that showed distinct differences between the wild-type and mutant flowers. The images of the C29 alkane ([C29 alkane + 107Ag]+ at m/z 515) and C29 ketone ([C29 ketone + 107Ag]+ at m/z 529) show that these metabolites occur at much lower abundance in the mutant flower than in the wild type. In contrast, C30 ([C30 fatty acid + 107Ag]+ at m/z 559) and C32 ([C32 fatty acid + 109 Ag]+ at m/z 589) fatty acids occur at higher abundance levels in the mutant than in the wild-type flowers. Moreover, both fatty acids are asymmetrically distributed on the petal, occurring at higher abundance at the tip of the petal. On the carpel, however, only the C30 fatty acid was found in high abundance (Figure 4B). Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

2995

Figure 3. Chemically selective images of Arabidopsis wild-type (Ler) whole flowers. The step size for data collection was set to 50 µm for both x and y directions. Dimension of the image is 5550 µm high × 4960 µm wide. The value in each image corresponds to the nominal m/z value of the corresponding ions. All images were processed as normalized intensities and shown as a percentage of the total ion current. All images were smoothed linearly. Major ions detected at m/z 529 correspond to [C29 ketone + 107Ag]+. Ions detected at m/z 515 and 543 are mainly silver adduct ions of C29 and C31 alkanes. Ions at m/z 547 are from C30 alcohol, and the image for the peak at m/z 545 corresponds to the overlapped image of C31 alkane and C30 alcohol as [C31 alkane + 109Ag]+ and [C30 alcohol + 107Ag]+. Images for silver adduct ions of C26 fatty acids (at m/z 503) and an unknown compound (at m/z 625) are also shown.

Imaging intact flowers provides spatial distribution data concerning cuticular wax metabolites among and on different organs of the flower, without the need to conduct laborious, manual separation of these organs. Although these data are generated in a short time period, there are limitations to the interpretation of these data sets. One of these issues is the need to normalize data acquisition. For example, the data presented in Figures 3 and 4 each metabolite ion intensity was normalized relative to TIC; however, there are several potential complications with such normalization. First, TIC may be affected by both surface characteristics and other metabolite signals in a mass profile. Second, in our experiments the physical proximity between adjoining organ structures may have disturbed the ability to obtain clear mass spectral profiles of individual organs. Therefore, we profiled cuticular wax compounds using silverLDI MS on individually separated flower organs (Figure 5), and generated quantitative information of metabolites based on relative intensities referenced to the peak at m/z 215.8 ([107Ag + 109Ag]+) (Supporting Information Table S2). Three Arabidopsis flower organs, carpels, petals, and sepals, were separately dissected, and mass spectra were serially collected from these separated organs with a 50 µm step movement. In contrast to leaf cuticular wax profiles, identification of metabolites on separated flower organs was more straightforward because fewer silver adduct ion peaks were observed per flower organ. Comparison of these observations with those made by imaging whole flower surfaces showed a consistency in the metabolites that were detectable by the two approaches, such as the decrease in C29 ketone on the carpels of the mutant (Figure 5A), decrease in C29 alkane on all organs of this mutant, and increased accumulation of C30 and C32 fatty acids on the mutant petals (Figure 5B). 2996

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

Because it is difficult to validate the findings of the silver-LDI MS method via comparison to prior data of such high spatial resolution or by an independent method, we also generated cuticular wax composition data for flowers, using the traditional “extraction” and GC/MS analytical method. These data are shown in Supporting Information Figure S2. It should be noted that at least 75 whole flowers (25 flowers per sample, three replicates of both genotypes) were needed to obtain reproducible cuticular wax metabolite profiles via GC/MS, whereas, in silver-LDI MS profiling, only five flowers were used for collecting the wax metabolite profiles of each kind of flower organ. In this study, we compared the alterations in abundance of each cuticular wax metabolite between the wild-type and the mutant flowers as determined by each method (silver-LDI MS and GC/MS) (Figure 6). It should be emphasized that even though the GC/MS results are from intact flowers and not from the individual flower organs (as provided by the LDI MS data), and thus cannot be directly compared, several features could be summarized from silver-LDI MS results and contrasted with the GC/MS results. First, both methods indicate that the CS8 mutations affect the accumulation of alkanes (C27, C29, and C31), which are the most abundant cuticular wax components on the wild-type flower organs and are most significantly affected by the mutations. However, the higher spatial resolution afforded by LDI MS indicates organ-specific differences in the accumulation of some of these metabolites; for example, the C31 alkane showed the smallest decrease on mutant sepals and even increased on the carpels and petals of this mutant. By GC/MS analysis, the C31 alkane showed the smallest change in abundance among the alkane species. Second, in both the silver-LDI MS and GC/MS analyses the C29 ketone was

Figure 4. (A) Chemically selective images of wild-type (Ler) and mutant (CS8) Arabidopsis flowers. The step size for data collection was set to 50 µm for both x and y directions. Dimensions of the images are 5550 µm high × 4960 µm wide (Ler) and 4100 µm high × 4770 µm wide (CS8). The C29 alkane (at m/z 515 as [C29 alkane + 107Ag]+) showed higher abundance on sepals and carpels of the wild type than on the mutant. The localization of C29 ketone (at m/z 529 as [C29 ketone + 107Ag]+) was easily detectable on the carpels of the wild-type flower but not on the mutant. Higher abundances of the C30 and C32 fatty acids at m/z 559 as [C30 fatty acid + 107Ag]+ and at m/z 589 as [C32 fatty acid + 109 Ag]+) were found in the mutant. (B) 3D presentations of the normalized intensities of chemically selective images of C30 and C32 fatty acids (at m/z 559 and at m/z 589) on the mutant flower. The yellow arrows in panels A and B are for aligning the viewing angle. Both fatty acids showed high abundances in the tips of petals. C30 fatty acid showed high abundance in the carpel, but C32 fatty acid did not.

obviously lower in all mutant organs than the corresponding wild-type organs. Third, via both silver-LDI MS and GC/MS analyses, there are relatively small differences in abundance of the primary alcohols between the mutant and wild-type organs. Forth, of the free fatty acid components, the most pronounced difference between the mutant and wild-type flower organs is in the accumulation of C30 and C32 fatty acids, which accumulate to higher abundances on the mutant flowers than on the wild type for all flower organs. Silver-LDI MS analysis indicates that the increase in the C30 fatty acid was most pronounced in the carpel and petals rather than the sepal. Perspectives on CER2 Gene Function in Cuticular Wax Deposition. The spatially resolved metabolite profiling data presented herein was conducted on two Arabidopsis lines that differed from each other by a number of genetic lesions, one of which was the nature of the allele carried at the cer2 locus; the mutant line used in these studies carried the cer2-2, ap2-1, and bp1 mutant alleles. Although the triple mutant genotype may confound the interpretation of these data sets, because only the cer2-2 allele affects cuticular wax deposition, the metabolic differences we have imaged between these two lines likely provide insights into the function of the CER2 gene in cuticular wax deposition. The cer2 gene was initially identified as a mutation that affects the normal deposition of cuticular waxes on stems

and siliques of Arabidopsis.4 Initial chemical analysis of the cuticular waxes of this mutant by traditional GC/MS analysis has indicated that this gene may function in the elongation of fatty acids from C28 to C30.8 The subsequent molecular cloning of this gene23,24 resulted in the identification of the gene product as a nuclear localized protein that is primarily expressed in the outer layer of cells, the epidermis of the aerial organs of Arabidopsis, leading to the hypothesis that CER2 protein may have a regulatory role in cuticular wax biosynthesis.25 Subsequent independent studies of the biosynthesis of fragrant volatile esters uncovered sequence homology between acetyl-transferases and CER2,26 which may suggest that the latter is involved in an acetylation reaction that regulates cuticular wax deposition. Although the data generated herein cannot specifically address the mechanistic function of the CER2 protein per se, the higher spatial resolution data that we present relative to the metabolites whose abundance is altered by the absence of the CER2 function indicates a complexity to the function of this gene that has not (23) Negruk, V.; Yang, P.; Subramanian, M.; McNevin, J. P.; Lemieux, B. Plant J. 1996, 9, 137–145. (24) Xia, Y. J.; Nikolau, B. J.; Schnable, P. S. Plant Cell 1996, 8, 1291–1304. (25) Xia, Y. J.; Nikolau, B. J.; Schnable, P. S. Plant Physiol. 1997, 115, 925– 937. (26) Dudareva, N.; D’Auria, J. C.; Nam, K. H.; Raguso, R. A.; Pichersky, E. Plant J. 1998, 14, 297–304.

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

2997

Figure 5. Positive-ion mode silver-LDI mass spectra taken from wild-type (Ler) and CS8 mutant Arabidopsis flower organs: (A) carpel, (B) petal, and (C) sepal. Each spectrum is the average of all scans obtained from each flower organ. The peak intensities were normalized to the reference peak (m/z 215.8, which corresponds to the ion [107Ag + 109Ag]+).

been possible to realize with prior analytical technologies. This pattern of altered metabolite profiles can be compared with the detailed expression pattern of this gene previously dissected by staining plants carrying the CER2::GUS reporter transgene.24,25 For example, consistent with the finding that CER2 expression is more prevalent in flowers than leaves, we have found that changes in the abundance of cuticular wax metabolites is most strongly affected on the surface of the former organ than in the latter, as was also indicated by the initial comparison of these metabolites in the cer2-1 mutant.9 Moreover, the data generated with the silver-LDI method provides metabolite distribution data with a spatial resolution 2998

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

comparable to the resolution obtained with reporter-gene constructs (i.e., CER2::GUS transgene).25 Therefore, comparison of these two data sets should provide the means for better evaluating the function of the CER2 gene in cuticular wax deposition. This would be particularly revealing when a gene’s expression pattern is highly differential at high spatial resolution, as is the case for CER2. Specifically, the CER2 gene is expressed in the three floral organs, sepals, carpels and petals,25 and indeed silver-LDI imaging demonstrate that in the absence of this gene function the abundance of the cuticular wax metabolites are reduced on each these organs; this insight was revealed by imaging the intact flower and further confirmed by profiling metabolites from the

Figure 6. Relative abundance of targeted cuticular wax metabolites on flower organs of Arabidopsis as determined by silver-LDI MS and GC/MS. For each flower organ imaged by silver-LDI MS (carpels, petals, and sepals), and the GC/MS data obtained from whole flowers, the relative abundance of each targeted metabolite is expressed as a log2-ratio of the wild-type (Ler)/CS8 mutant. Data points in each colored box correspond to the compound class indicated in the box. The number of carbon atoms in each metabolite is indicated for each data point. Detailed ion species assignments and their relative intensities for the three flower organs are listed in Supporting Information Table S2. Data collected from five replicates of each flower organ are averaged, and error bars correspond to standard errors. Whole flower analysis by GC/MS has three replicates, and the error bars correspond to standard errors.

physically separated organs. However, the effect of the mutations on the cuticular waxes deposited on each of these organs is distinct among the different organs, which would suggest that the functionality of the CER2 protein is sensitive to the organ context in which it is normally expressed. This would be consistent with the regulatory function of the CER2 gene, rather than a metabolic function; in the latter functional model the effect of the cer2-2 mutation may be expected to be similar among the different organs. We therefore argue that these high-resolution metabolite imaging data of cuticular wax metabolites on the flower organs of the cer2-2 mutant further supports the regulatory model for CER2 function, but these data also indicate that this functionality

is more complex than was contemplated from lower spatial resolution metabolite profiling data. This finding therefore argues for the need to develop technologies that can image metabolite distribution at a high spatial resolution, similar to that illustrated by the silver-LDI method we have described herein. CONCLUSIONS The high spatial resolution imaging of plant surface metabolites (cuticular waxes) was made possible by the reproducible spraying of colloidal silver solution onto the plant’s surface using a controllable spraying device, which generated homogeneous and constant surface coverage of colloidal silver particles. This made Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

2999

it possible to use the intensity of silver dimer ions as a reference to normalize all mass spectra. Cuticular wax metabolite profiling by LDI MS with colloidal silver as a matrix has several advantages over GC/MS approaches. First, sample preparation is much simpler because it does not require any extraction and derivatization steps. Second, the amounts of sample required are about 20-25 times smaller than that required by GC/MS. Third, localization information of metabolites can be preserved because of the direct sampling capability. We demonstrated the utility of silver-LDI MS by successfully imaging for the first time the distribution of cuticular wax metabolites on each flower organ. However, there are still challenging issues associated with the use of this method. These include unresolved metabolite peaks due to the overlap between isobaric ions, which was specifically illustrated in our case study with alkanes and aldehydes. This limitation may be resolved by employing exact mass measurements with high mass resolution. Another limitation is associated with the different ionization efficiencies for different classes of compounds, which makes it difficult to compare actual abundances between two different classes of metabolites directly from the mass spectra. This may be resolved by systematic studies on the response factors as a function of metabolite classes. Finally, because a mass spectral database of metabolites for silver-LDI

3000

Analytical Chemistry, Vol. 81, No. 8, April 15, 2009

MS does not currently exist, identification of metabolites is less confident than for GC/MS. Despite these limitations, by comparing the relative abundances of metabolites between wild-type and mutant plants we have illustrated that the high spatial resolution metabolite profiling data provided by silver-LDI MS provides considerable more power in interpreting the biological functionality of a gene product. ACKNOWLEDGMENT E.S.Y. thanks the Robert Allen Wright Endowment for Excellence for support. The Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under contract no. DE-AC02-07CH11358. This work was supported by the Director of Science, Office of Basic Energy Sciences, Division of Chemical Sciences. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review December 11, 2008. Accepted February 19, 2009. AC802615R