Ion Yields in UV-MALDI Mass Spectrometry As a Function of Excitation

Article pubs.acs.org/ac

Ion Yields in UV-MALDI Mass Spectrometry As a Function of Excitation Laser Wavelength and Optical and Physico-Chemical Properties of Classical and Halogen-Substituted MALDI Matrixes Jens Soltwisch,†,§ Thorsten W. Jaskolla,†,‡,§ Franz Hillenkamp,† Michael Karas,‡ and Klaus Dreisewerd*,† †

Institute of Medical Physics and Biophysics, University of Münster, Robert-Koch-Strasse 31, 48149 Münster, Germany Cluster of Excellence Macromolecular Complexes, Institute of Pharmaceutical Chemistry, Goethe-University Frankfurt, Max-von-Laue-Strasse 9, 60438 Frankfurt, Germany

‡

S Supporting Information *

ABSTRACT: The laser wavelength constitutes a key parameter in ultraviolet-matrix-assisted laser desorption ionization-mass spectrometry (UV-MALDI-MS). Optimal analytical results are only achieved at laser wavelengths that correspond to a high optical absorption of the matrix. In the presented work, the wavelength dependence and the contribution of matrix proton affinity to the MALDI process were investigated. A tunable dye laser was used to examine the wavelength range between 280 and 355 nm. The peptide and matrix ion signals recorded as a function of these irradiation parameters are displayed in the form of heat maps, a data representation that furnishes multidimensional data interpretation. Matrixes with a range of proton affinities from 809 to 866 kJ/mol were investigated. Among those selected are the standard matrixes 2,5-dihydroxybenzoic acid (DHB) and α-cyano-4-hydroxycinnamic acid (HCCA) as well as five halogen-substituted cinnamic acid derivatives, including the recently introduced 4-chloro-αcyanocinnamic acid (ClCCA) and α-cyano-2,4-difluorocinnamic acid (DiFCCA) matrixes. With the exception of DHB, the highest analyte ion signals were obtained toward the red side of the peak optical absorption in the solid state. A stronger decline of the molecular analyte ion signals generated from the matrixes was consistently observed at the low wavelength side of the peak absorption. This effect is mainly the result of increased fragmentation of both analyte and matrix ions. Optimal use of multiply halogenated matrixes requires adjustment of the excitation wavelength to values below that of the standard MALDI lasers emitting at 355 (Nd:YAG) or 337 nm (N2 laser). The combined data provide new insights into the UV-MALDI desorption/ ionization processes and indicate ways to improve the analytical sensitivity.

A

In this study, we employed a wavelength-tunable dye laser to investigate the wavelength and laser power dependence of the MALDI process for the analysis of peptides between 280−355 nm. Particular care was taken to control all relevant irradiation parameters including the laser spot size and the applied pulse energies. We compared DHB, HCCA, and a set of five halogensubstituted cinnamic acid (CCA) derivatives as matrixes, exhibiting proton affinity differences of almost 50 kJ/mol and including the recently introduced 4-chloro-α-cyanocinnamic acid (ClCCA) and α-cyano-2,4-difluorocinnamic acid (DiFCCA).6,7 These matrixes provided an improvement over standard HCCA in the limit of detection (LOD).8 Moreover, in the positive ion mode, various analyte compounds could only be detected with the new matrixes.9,10 These findings can be attributed to the reduced proton affinities (PA) exhibited by the halogen-substituted derivatives. However, the introduction of electron-withdrawing halogens into the molecular system

sensitive analysis of nonvolatile biomolecules by ultraviolet matrix-assisted laser desorption ionization mass spectrometry (UV-MALDI-MS) requires that the matrix exhibits a high optical absorption at the excitation laser wavelength.1−3 The wavelengths of the two standard MALDI lasers of 337 nm (N2 laser) and 355 nm (frequency-tripled Nd:YAG laser) match well the absorption bands of popular matrixes like 2,5-dihydroxybenzoic acid (DHB) and α-cyano-4hydroxycinnamic acid (HCCA) in the solid state.4,5 Studies in which further discrete excitation wavelengths were tested, e.g., that of the frequency-quadrupled Nd:YAG laser (266 nm) and that of the XeCl excimer laser (308 nm), have demonstrated that the MS performance generally deteriorates when the excitation wavelength corresponds to a lower absorption of the matrix compound.4 As a consequence, if the standard MALDI wavelengths of 337/355 nm are employed many compounds with potentially advantageous matrix properties cannot be utilized under optimal conditions. Despite the significance of this irradiation parameter, the influence of the laser wavelength on MALDI analyte and matrix ion signal intensities has not been comprehensively examined. © 2012 American Chemical Society

Received: March 27, 2012 Accepted: July 5, 2012 Published: July 5, 2012 6567

dx.doi.org/10.1021/ac3008434 | Anal. Chem. 2012, 84, 6567−6576

Analytical Chemistry

Article

the investigated matrix compounds will be presented elsewhere (manuscript in preparation). Preparation of Matrix and Analyte Solutions. CCA derivatives were dissolved to 20 mM solutions in 70% ACN/ 0.1% trifluoroacetic acid (TFA). DHB was prepared as a 130 mM aqueous solution. The peptides angiotensin I (MW 1295.677 Da) or angiotensin II (MW 1045.534 Da), substance P (MW 1346.728 Da), fibrinopeptide A (MW 1535.685 Da), neurotensin (MW 1671.910 Da), and renin substrate (MW 1757.925 Da) were premixed to concentrations of 10 μM/ peptide in 30% ACN/0.1% TFA (all MW values refer to the neutral monoisotopic species). To determine the limit of detection (LOD), an initial peptide solution containing angiotensin II, substance P, neurotensin, and renin substrate at 200 nM/peptide in 30% ACN/0.1% TFA was diluted four times by factors of 10, each, with 30% ACN/0.1% TFA (v/v) and directly spotted after dilution. Solvent composition and matrix concentrations were varied for some of the matrix compounds (see the Results and Discussion). Sample Preparation for MALDI-MS. All samples were prepared in at least quadruplicates using the dried droplet method. Neat (matrix-only) preparations were prepared by spotting 0.5 μL of matrix solution on a stainless steel sample plate, followed by air-drying. Analyte-matrix samples were prepared by mixing 0.5 μL each of analyte and matrix solution on the stainless steel sample plate followed by air-drying. Mass Spectrometry. The employed orthogonal time-offlight (o-TOF) mass spectrometer has been described in detail previously.13 The instrument is equipped with a modified oMALDI 2 ion source (AB Sciex, Concord, Canada). A wavelength-tunable dye laser, which was adjusted from 280 to 355 nm in increments of 5 nm serves as the laser light source. The emission wavelengths were determined with an accuracy of ≤0.2 nm using a fiber optic spectrometer (UV-50 green-wave, StellarNet, Tampa, FL). The laser beam was coupled into a 10 m long fused silica fiber (SFM200/220Y, Fiberguide, Stirling, NJ; core/clad diameter, 200/220 μm; numerical aperture, 0.12). The end surface of the optical fiber was imaged onto the sample plate at an angle of 60° relative to the surface normal using a custom-built 1:1 telescope. In this way, an approximately flat-top beam intensity profile with a spot size of about 200 μm × 400 μm (±20 μm) was produced, corresponding to an area A of ∼6.3 × 10−8 m2. Focal spot sizes were determined by means of the scanning knife-edge method using a custom-made sample plate.14 Care was taken to adjust the telescope for compensation of the wavelength dispersion such that equal laser spot sizes were produced at all wavelengths. Because of an increasing absorption of the core material in the low-UV range, the UV-light transmitting fiber was useful only for wavelengths ≥280 nm. This allowed the coverage of a wide fluence range from ion detection thresholds up to about 10 times higher fluences. Laser pulse energies were adjusted using a variable dielectric filter for 532 nm positioned in the main amplifier pump beam of the dye laser. Output pulse energies were monitored online during the experiments using a beam splitter and a pyroelectric detector built in-house. The pyroelectric detector was calibrated against a commercial energy meter (PEM100, LaserTechnik Berlin, Germany) by measuring the pulse energies exiting the telescope optics before and after each single wavelength setting. Laser pulse energies EP reported in this paper correspond to values delivered to the sample. Fluences F can be calculated by dividing the pulse energy at the sample by the irradiated area A.

results in a hypsochromic (blue) shift of the optical absorption. As a consequence, a suboptimal MALDI-MS performance is achieved if a 355 nm-Nd:YAG laser is employed.6,7 Higher halogenated derivatives that provide even lower PA values and could potentially offer further improvements (e.g., in the analysis of analytes with a particular low PA) generally exhibit an even stronger hypsochromic shift in their absorption profiles. Therefore, even a 337 nm-N2 laser, as used in most of the previous works with the ClCCA and DiFCCA compounds, might not be suitable anymore. Here, we carried out initial tests with three such highly reactive CCA derivatives, α-cyano-2,4,6-trifluorocinnamic acid (TriFCCA), α-cyano-4trifluoromethylcinnamic acid (F 3 CCCA), and α-cyano2,3,4,5,6-pentafluorocinnamic acid (PentaFCCA), all of which exhibit significantly blue-shifted absorption profiles compared to HCCA.

■

EXPERIMENTAL SECTION Materials. All chemicals and solvents were from SigmaAldrich (Schnellendorf, Germany). Water was purified by a Synergy Ultrapure Water System (Millipore, Schwalbach, Germany). Synthesis of α-Cyanocinnamic Acid Derivatives. 4Chloro-α-cyanocinnamic acid (ClCCA, MW = 207.01 g/mol), α-cyano-2,4-difluorocinnamic acid (DiFCCA, MW = 209.03 g/ mol), α-cyano-2,4,6-trifluorocinnamic acid (TriFCCA, MW = 227.03 g/mol), α-cyano-4-trifluoromethylcinnamic acid (F3CCCA, MW = 241.03 g/mol), and α-cyano-2,3,4,5,6pentafluorocinnamic acid (PentaFCCA, MW = 263.00 g/ mol) were synthesized according to Jaskolla et al.9 All compounds were recrystallized four times from methanol/ water (1:1, v/v), except for PentaFCCA that was prepared as a yellow oil due to weak intermolecular interactions of the perfluorinated phenyl groups. After concentration in vacuo and subsequent dissolving in hot methanol PentaFCCA was precipitated as white crystals upon addition of cold HPLCgrade water. This recrystallization step was repeated four times. Overall yields were between 60−85% of theory. For structures of the employed CCA-derivatives, see Figure S-1 in the Supporting Information. Spectrophotometry. Solution absorption spectra were recorded using a dual beam spectrophotometer (UV-2102PC, Shimadzu, Duisburg, Germany). Solid phase absorption spectra were measured in diffusive reflection mode using an integrating sphere (ISR-260, Shimadzu). In total, 50 mg of matrix powder was mixed with 500 mg of BaSO4, finely ground in a ball mill (Perkin-Elmer, Waltham, MA), and pressed to a pellet. A neat BaSO4 pellet served as the reference. Calculation of Proton Affinities. Visualization and initial building of the compounds were performed using ChemDraw Pro 10.0 and Chem3D Ultra 10.0 (both from PerkinElmer, Waltham, MA). For calculation of proton affinities of the investigated CCA derivatives the corresponding self-consistent field energies, zero-point energies (ZPE), and vibrational energies of the geometrically optimized neutral and protonated species were calculated by means of density functional theory calculations with the hybrid B3LYP method and the strongly extended triple-ζ 6-311++G(3df,3pd) basis set. The vibrational energies and ZPEs were corrected by a factor of 0.9877.11 Gaussian 09W and GaussView 5.0.8 (Gaussian, Inc., Wallingford, CT) were used for all computational calculations.12 A detailed description of the calculation of optimized geometries, PAs, sodium cation affinities, and ionization energies (IE) of 6568

dx.doi.org/10.1021/ac3008434 | Anal. Chem. 2012, 84, 6567−6576

Analytical Chemistry

Article

Typical MALDI-MS laser fluences are about several ten to a few hundred J m−2 for irradiation areas in the 100 μm laser spot diameter range.15 In the present work, a value of 150 J m−2 corresponds to a laser pulse energy EP of about 10 μJ. With regard to the broad laser wavelength range investigated, going along with strongly differing matrix absorptivities, mass spectra were recorded at laser pulse energies of 1.5, 3, 6, 12, 17, and 23 μJ, except for the two shortest utilized wavelengths of 280 and 285 nm, for which the reduced transmission of the optical fiber did not allow taking data points for 23 μJ (285 nm) or 17 and 23 μJ (280 nm). Taking into account the residual uncertainty in adjustment of the spot size and fluctuations in laser pulse energy, the actual fluence values on the sample have an uncertainty of ±15%. Mass spectra were acquired in the positive ion mode by irradiating the sample preparations with 300 laser pulses. During the data acquisition, the sample plate was moved continuously such that different random sample positions were irradiated. In the case of DHB, for which a less homogeneous sample spot was produced, large crystal needles which typically produce the most intense peptide ion signals were irradiated. The N2 buffer gas pressure in the oMALDI 2 ion source was set to 0.7−0.8 mbar. The lower cutoff of the transfer quadrupole was set to m/z = 130 except for determining the LOD, where a cutoff of m/z = 500 was applied to increase the ion transmission for the analyte species. Ions were detected with a four anode microchannel plate detector using a time-to-digital (TDC) conversion. Control experiments with both higher and less concentrated analyte-matrix samples resulted in accordingly higher and lower peptide ion signals over the full experimental wavelength-fluence range. Therefore, a saturation of the TDC detection system can be excluded. Mass spectra were recorded using tof ma or tof multi software (both courtesy of Vic Spicer and Werner Ens, University of Manitoba, Canada); the tof multi program was further modified in-house to allow for an automated determination of signal intensities (see below). Additional measurements were carried out with a Reflex III axial-TOF instrument in reflector mode (Bruker Daltonik, Bremen, Germany). Wavelength-Tunable Dye Laser. A frequency-doubled Nd:YAG laser (Surelite II; Continuum, Santa Clara, CA; λ = 532 nm; laser pulse duration, 7 ns; maximum pulse energy, ∼200 mJ; pulse repetition frequency, 10 Hz) was used for pumping a wavelength-tunable dye laser system (FL2001, Lambda Physik, Göttingen, Germany) exhibiting a line width of ≤2 nm; the pulse duration of the output pulse is ∼6 ns (fwhm). The laser dyes used (all from Radiant Dyes, Wermelskirchen, Germany) were fluorescein 27 (emission wavelength range, 560−580 nm); rhodamine 6G (580−590 nm); 10:1 rhodamine B/rhodamine 100 mix (590−610 nm); DCM (610−650 nm); 4:1 DCM/pyridine I mix (650−675 nm); pyridine I (675−720 nm); styryl 8 (720−760 nm). All dyes were dissolved in ethanol except for the DCM/pyridine I mix which was dissolved in isopropanol and fluorescein 27 (basic methanol). BBO1 (560−615 nm) and BBO2 (615−760 nm) crystals were used for second harmonic generation. The nonconverted fraction of the laser light was removed from the output beam using cutoff filters (Schott, Mainz, Germany). Data Analysis. A self-written routine was used for deriving the intensities of signals of interest. The signal intensity of a peak was determined as the integrated peak area (full width at quarter maximum definition). Three modes of data processing were used: (i) determination of the intensity of selected matrix

and analyte ion signals (monoisotopic peaks were evaluated as provided in a reference list) with a minimum signal-to-noise ratio (S/N) for peak picking of 3 and a maximum error tolerance for specific monoisotopic matrix and analyte peaks of 150 ppm, (ii) determination of the intensity of all discernible analyte or matrix ion peaks with S/N ≥ 3 with m/z ranging from 100−2100, and (iii) determination of the total ion count (TIC) in the m/z range of 100−2800 (thus extending somewhat above the mass of the largest test peptide, renin substrate). The TIC represents the intensity sum of all signals generated by the detector during the time-of-flight measurement, including chemical noise; this is assumed to be proportional to the sum of all gaseous ions generated in the MALDI process. Residual ion counts were negligible above m/z 2800 for the chosen test system. Ion signal intensities of peptide fragments were determined using mode (i). Only a-, b-, and y-type fragment ions (nomenclature according to ref 16) were evaluated; all other fragment classes yield significantly lower signal intensities. The fragments were identified on the basis of their m/z value, calculated using Protein Prospector (version 5.9.4, University of California, CA) and by comparison with mass spectra recorded from neat matrix samples. Relative yields of fragmentation were calculated by dividing the combined fragment ion intensities by the sum of the signal intensities of the intact precursor ions [A + H]+ plus the respective a-, b-, and y-fragment sum intensities. 3D surface plots (“heat maps”) were generated using OriginPro 8.1 (OriginLab, Northampton, MA). All 3D plots are based on 93 experimental signal intensity values, representing a data matrix of 16 wavelengths × 6 laser pulse energies (as described above, only 5 EP values could be recorded at 285 nm and 4 at 280 nm). Safety Hazard Note. The employed dye laser is of laser safety class 4. Safety precautions have to be taken when working with free beams of such lasers by wearing protective goggles.

■

RESULTS AND DISCUSSION Absorption Profiles. Solution and solid phase absorption spectra of the investigated matrixes are displayed in Figure 1. Peak absorption values are summarized in Table 1. The profiles of all compounds are broadened in the solid state and peak values wavelength-shifted for some of the substances because of intermolecular forces. Because of solvent-dependent effects in the case of the solution-absorpton spectra we will strictly use the solid-state absorption curves to evaluate the wavelengthdependence of the MALDI ion signals generated from the seven matrixes. In addition, an enhanced degree of scattering in the solid state toward longer wavelengths could add to the peak broadening.5 Mass Spectra. Tuning the laser to wavelengths of high optical absorption, high-quality mass spectra of the peptide standards mixture were generated with all seven matrixes. Signals of the singly protonated molecular peptide species [A + H]+ generally formed the base peaks in the mass spectra (see Figure S-2 in the Supporting Information for a set of typical mass spectra and text below for characterization of optimal irradiation conditions). Variations in the relative signal strengths among the individual peptides observed in the mass spectra acquired with the individual matrixes could be related to the different PA values exhibited by the compounds and/or a differing cocrystallization behavior. Moreover, the extent of 6569

dx.doi.org/10.1021/ac3008434 | Anal. Chem. 2012, 84, 6567−6576

Analytical Chemistry

Article

signal intensities at concomitantly lowest threshold fluences are obtained between ∼340 and 355 nm, in line with its use at the two standard MALDI wavelengths of 337/355 nm. On the basis of its broad absorption band, presumably HCCA could even be used for some 20 nm beyond the experimental wavelength limit of this study (355 nm). Such a finding was reported by Chen et al.17 in a work that probed the wavelength range between 370 and 450 nm. For DHB, as the second investigated standard MALDI matrix, the highest [A + H]+ ion signals were obtained between 320 and 335 nm (bottom graph in Figure 2a), therefore, somewhat below the peak absorption of 348 nm. However, the more limited pulse energy range that could be probed for this matrix with respect to the ion detection threshold may not be sufficient to allow revealing fine features in the heat maps, in particular toward the upper wavelength limit. For the recently introduced ClCCA and DiFCCA, optimal MALDI performances (i.e., lowest threshold fluences) are found at 310−325 and 300−320 nm, respectively and, therefore, somewhat below the N2 laser wavelength of 337 nm. All other halogenated CCA-derivatives cannot be employed at the standard MALDI wavelengths as a consequence of their blue-shifted absorption profiles. Compared to the optical absorption profiles in the solid state (Figure 1b), an asymmetry in the [A + H]+ contour plots is notable for all CCA-derived matrixes (Figure 2a). Despite similar absorption coefficients found on the rising and falling flanks of the absorption band and similar applied laser pulse energies, considerably lower abundances of molecular peptide ions are generated at the “blue” side. This effect starts at wavelengths only slightly below the optical solid state maxima. For the most part, this asymmetry can be attributed to an increased fragmentation of the analyte ions at the short wavelength side. The yield of a characteristic subgroup of identified peptide fragments, consisting of a-, b-, and y-type ions, is plotted in Figure 2b relative to the total intensity sum of all analyte-related ion signals (molecular ions plus fragments). Complementary to the molecular peptide ion signals, significantly higher fragment ion yields are obtained toward shorter wavelengths as well as toward higher laser energies; a-, b-, and y-type fragment species are typically also observed in high abundance in the MALDI “postsource decay” (PSD) analysis of peptides.18 Therefore, these fragment ions likely indicate a thermally driven fragmentation process occurring at higher energy conditions. For DHB, such a correlation is less

Figure 1. Absorption profiles of the investigated α-cyanocinnamic acid derivatives and of DHB in (a) solution and (b) in the solid state, cf. Table 1 for extinction coefficients and abbreviations. Absolute absorption coefficients cannot be derived in the diffusive reflection geometry used to obtain the solid state absorption profiles; therefore, these data are presented as relative values normalized to the peak absorption.

matrix cluster and peptide-matrix adduct formation varies with the matrix. Ion Signal Intensities As a Function of Laser Wavelength and Pulse Energy. The combined UV-MALDI-oTOF-MS signal intensities [A + H]+ of the five protonated peptide molecules generated from the seven investigated matrixes are plotted in Figure 2a as a function of laser wavelength and laser pulse energy. For all matrixes, the heat maps reveal distinct wavelength−pulse energy combinations at which highest molecular analyte signal intensities are obtained. For HCCA, probably the most widely employed MALDI matrix for peptide mass fingerprinting, comparable analyte

Table 1. Decadic Molar Extinction Coefficients ε of the Investigated CCA-Derivatives and of DHB at the Standard MALDI Laser Wavelengths of 337 and 355 nm and at the Wavelengths of the Strongest Solution Phase Absorption εmax (Data Derived from Figure 1a), Wavelengths of Peak Absorption in the Solid State λmax, solid (Derived from Figure 1b), Wavelength Regions of Maximum Molecular Peptide Ion Generation λmax, ion (Derived from Figure 2a), and Calculated Proton Affinities (PA; cf. Experimental Section for Computational Method)a matrix

HCCA

ClCCA

DiFCCA

TriFCCA

F3CCCA

PentaFCCA

DHB

ε355 [L·mol−1·cm−1] ε337 [L·mol−1·cm−1] εmax [L·mol−1·cm−1] λ [nm] λmax, solid [nm] λmax, ion [nm]b PA [kJ/mol]

11 200 25 700 29 000 326 346 340−355 866

290 2900 27450 301 307 310−325 842

60 1500 17790 292 292 300−320 837