Selective Enhancement of Carbohydrate Ion Abundances by Diamond

15 Mar 2013 - Diamond Nanoparticles for Mass Spectrometric Analysis. Chieh-Lin Wu, ... The best enhancement effect was achieved when matrix, DNP, and...
0 downloads 0 Views 409KB Size
Technical Note pubs.acs.org/ac

Selective Enhancement of Carbohydrate Ion Abundances by Diamond Nanoparticles for Mass Spectrometric Analysis Chieh-Lin Wu,† Chia-Chen Wang,‡,† Yin-Hung Lai,† Hsun Lee,† Jia-Der Lin,† Yuan Tseh Lee,† and Yi-Sheng Wang*,†,‡ †

Genomics Research Center, Academia Sinica, Taipei 115, Taiwan, ROC Institute of Biochemistry and Molecular Biology, National Yang-Ming University, Taipei 112, Taiwan, ROC



S Supporting Information *

ABSTRACT: Diamond nanoparticles (DNPs) were incorporated into matrix-assisted laser desorption/ionization (MALDI) samples to enhance the sensitivity of the mass spectrometer to carbohydrates. The DNPs optimize the MALDI sample morphology and thermalize the samples for thermally labile compounds because they have a high thermal conductivity, a low extinction coefficient in UV−vis spectral range, and stable chemical properties. The best enhancement effect was achieved when matrix, DNP, and carbohydrate solutions were deposited and vacuum-dried consecutively to form a trilayer sample morphology. It allows the direct identification of underivatized carbohydrates mixed with equal amount of proteins because no increase in the ion abundance of proteins was achieved. For dextran with an average molecular weight of 1500, the trilayer method typically improves the sensitivity by 79- and 7-fold in comparison to the conventional dried-droplet and thin-layer methods, respectively.

E

unsuitable for complicated biological mixtures and in situ analyses. Mixing an analyte/matrix with frozen solutions, on the other hand, can reduce phase transition temperature in MALDI and improve sensitivity.17 Such experiments require a modified instrument that is unavailable commercially. In addition, the change in sensitivity using this method should be a general result, so the diagnosis of carbohydrates in complex mixtures is still difficult. To facilitate efficient and convenient MS examinations of native carbohydrates, selective enhancement of their ionization efficiency is urgently needed. We report here a convenient method to selectively enhance the ion abundance of carbohydrates with MALDI. The samples are prepared by incorporating diamond nanoparticles (DNPs) to change sample morphology and facilitate fast thermalization. The most important function of DNPs was to reduce direct heat exchange between analytes and matrixes. The best signal improvement effect occurred when DNPs were sandwiched between matrixes and carbohydrates, which we defined herein as the trilayer sample. Diamond is selected because it has the highest thermal conductivity (κ) among all natural materials to thermalize its surroundings. The κ of type I diamonds is 8.95 W/cm K at room temperature,18 which is 3 orders of magnitude higher than that of matrixes (typically 0.002 W/ cm K).19 Additionally, diamonds are highly transparent in the near-UV region that is typically used for MALDI, with

xaminations of carbohydrates are important to the study of the construction of cellular structures, the direction of immune processes, and even signal transduction in biological systems.1−3 Traditional chemical and enzymatic methods are inefficient because of the heterogeneous structures and indistinguishable chemical properties of carbohydrates.4 Chromatographic analysis of such molecules is also difficult due to their low extinction coefficient in the UV−vis spectral range. Although chemical derivatizations can change the hydrophobicity or optical properties of carbohydrates to improve the efficiency of analysis,5−9 they spend time and increase sample consumption. Among other analytical methods, mass spectrometry (MS) has been the most important technique for carbohydrate analysis after the invention of electrospray ionization (ESI) and matrix-assisted laser desorption/ionization (MALDI) methods.7,10−12 However, ESI and MALDI are less effective for carbohydrates than for proteins because the proton affinity of carbohydrates is low.7,12 Another complication is that the nonvolatile and thermally labile properties of carbohydrates result in a low ion abundance under high-energy or hightemperature ionization conditions; the peak temperature in a typical MALDI reaction is roughly 1000 K.13−16 Thus, identifying carbohydrates in complex samples by MS is still highly challenging. To increase the ion abundance of carbohydrates, permethylation and peracetylation of carbohydrates are typically conducted.6−9 Although such chemical derivatizations can improve the sensitivity by roughly 10- to 100-fold, they are © 2013 American Chemical Society

Received: December 16, 2012 Accepted: March 15, 2013 Published: March 15, 2013 3836

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841

Analytical Chemistry

Technical Note

extinction coefficients on the order of 10−7.18 Therefore, embedded DNPs in matrix crystals do not absorb the laser energy for matrix ionization. Finally, diamonds are chemically inert dielectric materials that neither react nor compete for charges with analytes. The strategy is generally applicable because it optimizes ionization condition thermodynamically. The preparation method was discussed and compared with the conventional dried-droplet (DD) and thin-layer (TL) preparation methods.20,21

Fluorescence images were recorded using a Leica TCS SP8 confocal microscope (Mannheim, Germany). DNPs and DHB were simultaneously excited by using 635 and 405 nm laser photons, respectively, and the fluorescence signal within 640− 676 nm for DNPs and 412−473 nm for DHB was collected. The vertical scanning of the fluorescence images was performed every 0.95 μm, with a detection thickness of roughly 1.05 μm. The infrared emission from samples was detected by an InGaAs PIN photodiode to estimate the temperature, as described elsewhere.16,22 All mass spectra were recorded using an Applied Biosystems Voyager-DE MALDI-time-of-flight-mass spectrometer (TOFMS) (Foster City, CA). It was operated in the linear mode, with an acceleration voltage of +20 kV and typical laser energies for routine analysis ranging from 6.5 to 10 μJ. The mass spectra were acquired using 200 laser shots. Imaging mass spectrometric analyses were conducted using a Bruker Daltonics Ultraflex TOF/TOF mass spectrometer (Bremen, Germany). Every image was obtained from 1300 sampling spots, and every spot was subjected to 100 laser shots.



EXPERIMENTAL SECTION 2,5-Dihydroxybenzoic acid (DHB), 2,4,6-trihydroxyacetophenone (THAP), butylamine, dextran [H(C6H10O5)nOH, with an average molecular weight of 1500 Da], β-cyclodextrin (β-CD), and ACTH fragment 18-39 (ACTH) were purchased from Sigma-Aldrich (St. Louis, MO). Type Ib synthetic DNPs were purchased from Element Six (Luxembourg). In this work, 500 nm DNPs were typically used unless specified otherwise. In addition, 30−50 nm titanium (Ti) and 30 nm silver (Ag) nanoparticles were purchased from UniRegion Bio-Tech (Taiwan). Acetone and acetonitrile were purchased from Merck (Whitehouse Station, NJ). Distilled deionized water (DDW) was used throughout the experiment. To remove impurities and nondiamond carbon from the DNPs, the DNPs were dispersed evenly on a Petri dish and heated in an oven at 400 °C for 2 h. After returning to room temperature, the pretreated DNPs were collected and cleaned with acetone by using a vortex, and the supernatant was removed after centrifugation. The cleaning procedures were repeated 5 to 10 times until the color of the solution was whitish. After being dried in a fume hood, the DNPs were resuspended in DDW to required concentrations (typically 0.6−10 mg/mL) and dispersed thoroughly by the vortex before use. Because DNP surfaces contain both hydrophilic and hydrophobic sites, they did not precipitate during sample preparation. Metallic nanoparticles were dispersed in DDW by using the vortex and deposited rapidly on the sample surface to minimize the precipitation. The trilayer samples contained the matrix, DNP, and analyte layers. Every layer was deposited and immediately vacuumdried for 5−20 min. The first and the second layer was prepared by using 0.5 μL of matrix solution (0.1 M in 50% acetonitrile aqueous solution) and 1 μL of the DNP solution, respectively. An ionic liquid matrix layer used in this work was 2,5-dihydroxybenzoic acid butylamine (DHBB) prepared by dissolving butylamine into a DHB solution with the identical concentration to DHB. The third layer contained 0.5 μL of analyte solution, which was 1 μM aqueous dextran in most experiments. The analyte solution for evaluating the performance of various nanoparticles or the sensitivity with various sample composition was 5 μM aqueous dextran or 2 μM β-CD, respectively. In the experiment of carbohydrate/protein mixture, the analyte solution consisted of 0.25 μL each of 20 μM aqueous dextran and 20 μM aqueous ACTH. The aqueous ACTH also contained 0.1% formic acid to improve the solubility of ACTH. Care was always taken to prevent the crystals from changing their structures by pipet tips during preparation. The conventional DD and TL samples were prepared without using DNPs. In the DD method, 0.5 μL each of analyte and matrix solutions were premixed before deposition and vacuum-drying, whereas in the TL method 0.5 μL each of matrix and analyte solutions were successively deposited and vacuum-dried.



RESULTS AND DISCUSSION Sample Morphology. The trilayer samples consist of a matrix as the bottom, DNPs as the middle, and analytes as the top layers, as shown in Figure 1a. The DHB layer typically

Figure 1. Morphology of trilayer and other samples. (a) Schematic of the configuration of matrix, DNP, and analyte in trilayer samples, (b) image of a dried-droplet sample, (c) image of a thin-layer sample, and (d) image of a trilayer sample containing 3 μg DNPs. The scale bars represent 1 mm.

formed a circular spot containing apparent crystals at the rim and a microcrystalline film covering the center, similar to that of DD (Figure 1b) and TL samples (Figure 1c). Such morphology became the substrate of the subsequent layers, and the sample containing DNPs was whitish (Figure 1d). The addition of dextran did not result in further color change. As can be seen from the pictures, the thickness of the trilayer sample was thicker than the DD and the TL samples. In trilayer samples, the population of DNPs near the crystal surface was more than that in the bottom. Because DDW was used to suspend DNPs and dissolve analytes, the redissolvation of matrix crystals (less soluble in water) was minimized. It was also found that the best morphology for signal enhancement was obtained when the deposited solutions were immediately 3837

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841

Analytical Chemistry

Technical Note

Figure 2. Fluorescence images (242 × 242 μm) of DNPs (green) and DHB (red) in various layers of a trilayer sample. The numbers indicate the sequences of the layers below the crystal surface, and every layer was separated from adjacent layers by 0.95 μm.

vacuum-dried. Because DNPs and DHB fluoresce at different spectral ranges, their spatial distribution in every layer of the sample can be unambiguously analyzed. Figure 2 shows the morphology in the outer half of the microcrystalline region that typically produces a high carbohydrate ion abundance. The sample contained 3 μg of DNPs, and the thickness of this area was roughly 14 μm. In accordance with the result, DNPs mainly populated from 4 to 7 μm below the crystal surface, and no apparent DNPs were found outside 3 to 8 μm. Nonfluorescent dextran was presumed to exist mainly above the DNPs, since it was the last layer deposited. Such a thin DNP layer near the surface of the crystal was important and could not be obtained by premixing the DNPs into the analyte/matrix solution. In a premixed sample of 14 μm thick, the fluorescence images indicated that DNPs populated from 4 to 13 μm below the crystal surface and have the highest abundance between 5 to 10 μm (Figure S1 of the Supporting Information). The wide DNP population was found to considerably suppress the ion abundance (see discussion later). Sample Temperature. The trilayer samples resulted in lower peak temperatures than DD and TL samples during laser excitation. Direct evidence of this was the reduction in infrared emissions from the surface when increasing the amount of DNPs (Figure S2 of the Supporting Information). When fixing the laser fluence at 229 J/m2, the infrared emissions were 11.2, 7.8, 3.5, 2.8, and 2.4 mV as the amount of DNPs increased from 0.0, 0.3, 0.9, 3.0 to 6.0 μg, respectively. Presuming blackbody radiation, those IR emissions correspond to roughly 771, 753, 714, 705, and 698 K, respectively. Such a cooling effect was mainly due to a decrease in the effective optical absorbance, which subsequently reduced deposited energy density, and the efficient thermalization by DNPs. Because analytes are partially separated from matrixes, the exact temperature of analytes is

expected to be lower than the value derived from the IR emissions. Analysis of Carbohydrate. With trilayer samples, the signal-to-noise (S/N) ratio of dextran ions in mass spectra was considerably higher than that with DD and TL samples. In the spectrum of the DD sample (Figure 3a), the dextran signals merged with background and chemical noises. An average of 11-fold improvement was achieved with TL samples for the representative sodiated dextran at m/z 851, 1013, and 1499 (Figure 3b), but the features were still too low to be distinguished easily from noises. With the trilayer sample, on the other hand, these signals increased with increasing amounts of DNPs (i.e., from 0.6 to 3.0 μg) and leveled at about 79- and 7-fold improvement over DD and TL samples, respectively, when 3.0 to 6.0 μg DNPs was used (Figure 3, panels c−f). The dextran signals finally declined with 10.0 μg of DNPs, likely due to obstruction of matrix ions by the thick DNP layer and insufficient temperature for ion desorption. These results corresponded to detection limits of roughly 0.78, 0.10, and 0.01 pmol for DD, TL, and trilayer methods, respectively. The mass resolving power obtained from the trilayer samples (about 1000 to 1300 in Figure 3, panels c−f) was also increased by roughly 2.5-fold in comparison to the TL sample (about 450 in Figure 3b). Increasing DNP in the trilayer samples also reduced the background and chemical noises. Interestingly, no signal can be detected by premixing 3.0 μg of DNPs with dextran and matrix solutions before deposition or when no matrix layer was prepared. Table 1 summarizes the normalized S/N ratios of those representative features and noise with various sample preparation methods. The enhancement of the carbohydrate signal with trilayer samples is generally effective. We used THAP and DHBB, a neutral matrix for acidic carbohydrates and a known ionic liquid matrix providing superior sensitivity, respectively, as matrixes to ionize 1 pmol of β-CD. In the DD samples prepared with 3838

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841

Analytical Chemistry

Technical Note

Figure 4. Comparison of signal intensity of β-CD with four sample conditions. (a) Dried-droplet sample with THAP, (b) dried-droplet sample with DHBB, (c) trilayer sample with THAP, and (d) trilayer sample with DHBB.

Analysis of Carbohydrate/Protein Mixture. A unique feature of this method is the selective improvement of ion abundance of thermally labile molecules. By contrast, for analytes resistant to high temperatures, a decrease in desorption efficiency due to thermalization is dominant.13,14,23 For example, mixing 5 pmol each of dextran and ACTH in DHB strongly favors the production of protonated ACTH at m/z 2467. Figure 5a shows the result from such a sample, in which dextran signals at the low m/z region were barely detectable (S/N ratios less than 9). In the TL sample, the S/N ratio of dextran ions and ACTH all increased by 20−50% (Figure 5b), but the identification of dextran was still difficult. A significant

Figure 3. Mass spectra of 0.5 pmol dextran obtained from various sample preparation methods, including (a) dried-droplet method, (b) thin-layer method, and the trilayer method with (c) 0.6, (d) 1.5, (e) 3.0, and (f) 6.0 μg DNPs. The numbers indicate the degrees of polymerization (n) of dextran with sodium or potassium attached.

Table 1. Average S/N Ratios of Three Representative Sodiated Dextran Features and Relative Noise with Various Sample Preparation Methods average S/N ratio preparation method

m/z 851

m/z 1013

m/z 1499

relative noise

dried-droplet thin-layer trilayer (μg DNPs) 0.6 0.9 1.5 3.0 6.0 10.0

4 22

1 23

3 15

0.94 0.85

70 108 106 164 180 86

69 131 91 159 161 87

38 53 53 90 116 77

0.97 1.00 0.94 0.82 0.78 0.58

THAP (Figure 4a) and DHBB (Figure 4b), the typical S/N ratio of sodiated β-CD was approximately 75 and 600, respectively. When 3.0 μg of DNPs was used with THAP in the trilayer sample (Figure 4c), the S/N ratio increased to approximately 730. This result indicates that such a method can roughly improve the sensitivity by 1 order of magnitude, regardless of the property of the matrix and the carbohydrate. Furthermore, using 3.0 μg of DNPs and DHBB in the trilayer sample increased the S/N ratio to approximately 2400 (Figure 4d). Although the improvement was only 4-fold in comparison with the DD method, the result implies that thermalization plays an essential role in the ionization process of carbohydrates.

Figure 5. Mass spectra of 5 pmol each of dextran and ACTH fragment 18−39 obtained from various samples, including (a) dried-droplet, (b) thin-layer, and (c) trilayer samples. The numbers indicate the degrees of polymerization (n) of sodiated dextran. 3839

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841

Analytical Chemistry

Technical Note

be functionalized with carbohydrate receptors, it has the potential to become a tool to solve other glycochemistry problems, such as to enrich glycoproteins or induce immune reactions prior to mass analysis.

change in spectral patterns was obtained using the trilayer method with 3.0 μg of DNPs, as shown in Figure 5c. Although the S/N ratio of ACTH roughly increased by 40% due to a reduction of noise, the S/N ratio of dextran increased by 10- to 12-fold to dominate the spectrum. Thus, this sample preparation method is ideal for diagnosing carbohydrates in protein mixtures. Compensating the sensitivity difference between carbohydrates and proteins also provides more accurate quantity information of complex samples than conventional methods. Notably, the region producing a carbohydrate signal is the largest in trilayer samples in comparison to others. The imaging MS showed that the ACTH ion (m/z 2467) has the highest abundance on the rim in all sample morphologies (Figure S3 of the Supporting Information). The dextran ion (m/z 851), however, populated differently in the three morphologies: in the DD and the TL samples, it populated in the outer half of the microcrystalline region and on the rim, respectively; in the trilayer sample, it populated not only in both regions but also with the larger area, suggesting that this method reduces the sample inhomogeneity problem of MALDI.24,25 Particle Size and Material. The size of DNPs is not an essential factor in the performance of the trilayer samples. We have tested DNPs ranging from 50 to 500 nm and found that the signal intensity only slightly increased as the particle size increased (Figure S4 of the Supporting Information), but the changes were within the experimental error of MALDI. Although many metals have good κ, they cannot result in the same improvement. We have compared the mass spectra of 2.5 pmol dextran obtained with various sample preparation methods (Figure S5 of the Supporting Information), including trilayer samples with DNP replaced by Ag (4.29 W/cm K) or Ti (0.22 W/cm K) nanoparticles. As a reference, the spectrum from DD (Figure S5a of the Supporting Information) and TL (Figure S5b of the Supporting Information) samples showed dextran features but with significant background and chemical noises. Apparent improvement was achieved with 0.9 and 3.0 μg of DNPs (Figure S5, panels c and d of the Supporting Information). In contrast, except slight improvement achieved with 0.9 μg of Ag (Figure S5g of the Supporting Information), the dextran ion abundances always reduced dramatically when Ag (Figure S5h of the Supporting Information) and Ti (Figure S5, panels e and f of the Supporting Information) were used. The minor improvement with low amount of Ag is probably due to the reduction of surface temperature, whereas the side reactions from the metal surface induced by laser may be responsible for the signal reduction in most cases.



ASSOCIATED CONTENT



AUTHOR INFORMATION

S Supporting Information *

Additional material includes fluorescence images, IR emissions, mass spectrometry images, and mass spectra with various sizes of diamond and metallic nanoparticles. This material is available free of charge via the Internet at http://pubs.acs.org.

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the Genomics Research Center, Academia Sinica, and the National Science Council of Taiwan (Contract 101-2113-M-001-003-MY3). We thank the excellent technical assistance on confocal microscopy analysis provided by Li-Wen Lo.



REFERENCES

(1) Lehninger, A. L.; Nelson, D. L.; Cox, M. M. Lehninger Principles of Biochemistry, 4th ed.; W.H. Freeman: New York, 2005. (2) Costello, C. E. Biophys. Chem. 1997, 68, 173−188. (3) Caroff, M.; Karibian, D. Carbohydr. Res. 2003, 338, 2431−2447. (4) Holme, D. J.; Peck, H. Analytical Biochemistry, 3rd ed.; Addison Wesley Longman Limited: Harlow, Essex, United Kingdom, 1998. (5) Waterval, J. C. M.; Lingeman, H.; Bult, A.; Underberg, W. J. M. Electrophoresis 2000, 21, 4029−4045. (6) Lamari, F. N.; Kuhn, R.; Karamanos, N. K. J. Chromatogr., B 2003, 793, 15−36. (7) Harvey, D. Mass Spectrom. Rev. 1999, 18, 349−450. (8) Ciucanu, I.; Kerek, F. Carbohydr. Res. 1984, 131, 209−217. (9) Hakomori, S. I. J. Biochem. (Tokyo, Jpn) 1964, 55, 205−208. (10) Reinhold, V. N.; Reinhold, B. B.; Costello, C. E. Anal. Chem. 1995, 67, 1772−1784. (11) Harvey, D. J. J. Am. Soc. Mass Spectrom. 2000, 11, 900−915. (12) Zaia, J. Mass Spectrom. Rev. 2004, 23, 161−227. (13) Dreisewerd, K. Chem. Rev. 2003, 103, 395−425. (14) Lai, Y.-H.; Wang, C.-C.; Lin, S.-H.; Lee, Y. T.; Wang, Y.-S. J. Phys. Chem. B 2010, 114, 13847−13852. (15) Allwood, D. A.; Dyer, P. E.; Dreyfus, R. W. Rapid Commun. Mass Spectrom. 1997, 11, 499−503. (16) Koubenakis, A.; Frankevich, V.; Zhang, J.; Zenobi, R. J. Phys. Chem. A 2004, 108, 2405−2410. (17) Liang, C.-W.; Chang, P.-J.; Lin, Y.-J.; Lee, Y.-T.; Ni, C.-K. Anal. Chem. 2012, 84, 3493−3499. (18) Lide, D. R. CRC Handbook of Chemistry and Physics: A ReadyReference Book of Chemical and Physical Data, 87th ed.; CRC Press: Boca Raton, FL, 2006. (19) Handschuh, M.; Nettesheim, S.; Zenobi, R. Appl. Surf. Sci. 1999, 137, 125−135. (20) Hillenkamp, F.; Peter-Katalinic, J. MALDI MS: a Practical Guide to Instrumentation, Methods and Applications; Wiley-VCH: Weinheim, Germany, 2007. (21) Vorm, O.; Roepstorff, P.; Mann, M. Anal. Chem. 1994, 66, 3281−3287. (22) Lai, Y.-H.; Wang, C.-C.; Chen, C. W.; Liu, B.-H.; Lin, S.-H.; Lee, Y. T.; Wang, Y.-S. J. Phys. Chem. B 2012, 116, 9635−9643. (23) Dreisewerd, K.; Schurenberg, M.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Processes 1995, 141, 127−148.



CONCLUSIONS Diamond nanoparticles are used to adjust MALDI sample morphology and facilitate thermalization for selective enhancement of carbohydrate ion abundances. The samples are prepared using matrix, diamond nanoparticles, and analyte solutions to form three partially separated layers. These trilayer samples compensate for the difference in ionization efficiency between carbohydrates and proteins, allowing for the direct diagnosis of underivatized carbohydrates from protein mixtures. It is convenient and can be applied generally with most matrixes and standard instruments without empirical trial. Further study is necessary to evaluate different types of diamond and sample morphology. Applying this method for the analysis of carbohydrate derivatives is also important and is currently under investigation. Because the diamond surface can 3840

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841

Analytical Chemistry

Technical Note

(24) Stahl, B.; Thurl, S.; Zeng, J. R.; Karas, M.; Hillenkamp, F.; Steup, M.; Sawatzki, G. Anal. Biochem. 1994, 223, 218−226. (25) Garden, R. W.; Sweedler, J. V. Anal. Chem. 2000, 72, 30−36.

3841

dx.doi.org/10.1021/ac3036469 | Anal. Chem. 2013, 85, 3836−3841