International Journal of Mass Spectrometry 226 (2003) 117–131
Protein incorporation into MALDI-matrix crystals investigated by high resolution field emission scanning electron microscopy夽 Verena Horneffer a , Rudolf Reichelt b , Kerstin Strupat a,∗ a
Institute for Medical Physics and Biophysics, Laser Mass Spectrometry Group, University of Münster, Robert-Koch-Str. 31, D-48149 Münster, Germany b Institute for Medical Physics and Biophysics, Electron Microscopy and Analysis Group, University of Münster, Hüfferstr. 68, D-48149 Münster, Germany Received 19 September 2001; accepted 15 July 2002
Abstract In this study, high resolution field emission scanning electron microscopy (FE-SEM) was tested for its feasibility to investigate analyte incorporation into and analyte distribution in slowly grown crystals of 2,5-dihydroxybenzoic acid (2,5-DHB) and 2,6-dihydroxybenzoic acid (2,6-DHB); both compounds function as a matrix for matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS). To investigate matrix–analyte interaction in 2,5-DHB crystals two Au-labels (20 nm colloidal gold and Nanogold) were examined as potential protein markers to visualize proteins in matrix solids by SEM. For this purpose, analyte-doped 2,5-DHB crystals were mechanically cleaved and SEM-micrographs of the opened inner faces were recorded. During the course of the Au-label study, crystal defects became apparent for crystals grown from analyte-free and analyte-doped matrix solutions; these defects are interpreted as ‘fluid’ or ‘liquid inclusions.’ The size and amount of liquid inclusions depend on the individual cooling rate during crystal growth and on the size of analyte molecules added to the matrix solution prior to crystal growth. Considering the results obtained for both matrix compounds, we assume that analyte incorporation occurs via a ‘solid solution’ in the case of 2,5-DHB as earlier proposed and not via phase defects such as liquid inclusions. Whether liquid inclusions—presumably filled with saturated mother solution—help to facilitate the desorption/ionization event remains open. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Matrix–analyte interaction; MALDI-mass spectrometry; Scanning electron microscopy; Gold label
1. Introduction Since the first MALDI results in the late 1980s [1], it is an open question how matrix and analyte molecules interact with each other to enable the des夽 This work is in partial fulfillment of the Ph.D. thesis V.H. in Experimental Physics at the University of Münster. ∗ Corresponding author. E-mail address:
[email protected] (K. Strupat).
orption/ionization of intact analyte molecules from the condensed matrix material. Having in mind that even today only a few empirically found organic compounds work as a matrix for MALDI, the question of overriding importance is ‘What goes to make up a MALDI-matrix?’ Knowledge about the matrix– analyte interaction on a molecular level is, therefore, understood to be the key for better and more appropriate matrices.
1387-3806/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII S 1 3 8 7 - 3 8 0 6 ( 0 2 ) 0 0 9 7 9 - X
118
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
Earlier investigations addressed the questions of how matrix and analyte molecules interact with each other prior to desorption/ionization, and if certain features of an organic compound qualify it as a functional matrix. In the past, various kinds of MALDI preparations were analyzed by different techniques such as scanning electron microscopy (SEM) [2–6], Raman spectroscopy [2], X-ray photoelectron spectroscopy [7], polarization microscopy [8], phase contrast microscopy [8], fluorescence microscopy [6,9], confocal laser scanning microscopy [10,11], and mass spectrometric imaging [12,13] with the aim of understanding the matrix–analyte interaction especially the mechanisms and functions of analyte incorporation into matrix solids. Mainly, different sample morphologies induced by the presence or absence of analyte molecules and by different sample preparation techniques are described in those papers. The question of matrix–analyte interaction in matrix crystals on a molecular basis could, however, not be answered in a satisfying way. To investigate the matrix–analyte interaction in more detail, studies were carried out with the aim to qualify and quantify incorporation of analyte molecules into matrix crystals. We and others have studied the incorporation of proteins into well-working matrices [14,15]. These investigations demonstrated that protein molecules are embedded (incorporated) into slowly grown single crystals of MALDI matrices such as 2,5-dihydroxybenzoic acid (2,5-DHB) or 3,5-dimethoxy-4-hydroxy-cinnamic acid (sinapic acid). The incorporation of protein was proven by MALDI-mass analysis of the crystals in depth and at cleaved surfaces as well as by re-dissolving crystals for a concentration determination of the protein by UV-absorption spectrophotometry. X-ray structure analysis of protein-doped crystals revealed the undisturbed crystal structure of the neat matrix [16,17]. It was concluded from these previous results that incorporation of analyte molecules into matrix solids upon matrix crystallization is a necessary prerequisite for a successful MALDI-mass analysis [14]. The image of a ‘solid solution’ [18], i.e., analyte molecules are solvated by matrix molecules in the solid and matrix
crystallization takes place completely uninfluenced by analyte molecules (because MALDI-relevant molar analyte-to-matrix ratios are too small to change the long-range order of the crystal structure of the matrix compound)—fitted best into the early experimental findings. Later investigations showed both incorporation of analyte molecules in the case of the IR-matrix succinic acid [19,20] and exclusion of analyte molecules from the crystal lattice in the case of 2,6-dihydroxybenzoic acid (2,6-DHB) [21]; both compounds work as a MALDI-matrix. These results and a recently published ionization model [22] suggest a differentiation between analyte attachment to and analyte incorporation into matrix solids [23]. The former, so-called ‘matrix support effect’ [23], seems to be limited in mass (<30 ku) and enables for a desorption/ionization of intact analyte molecules at a sufficiently large surface-to-volume ratio [24,25]. The physico-chemical properties of the analyte attachment is studied in a joined project by the Münster and the Frankfurt group at present and will be published elsewhere [26]. The latter, so-called ‘true MALDI,’ is supposed to require analyte molecule incorporation [23] and is functional for a larger mass range. Residual solvent content in matrix crystals is discussed to have an influence on this process [27,28]. Taking previous SEM investigations of MALDIpreparations into account [3,19], SEM is considered to be a suitable tool for investigation of slowly grown analyte-doped matrix crystals. One aim of the present study is to image individual analyte molecules by SEM using suitable labels such as Au-labels in order to learn about the localization and distribution of analyte molecules in matrix crystals. Two gold labels, colloidal gold and Nanogold, are used for these studies: colloidal gold is the standard gold label in biological SEM. Nanogold is a relative new alternative to colloidal gold and has the advantage of a covalent binding to the analyte molecules. In the course of this study, defects—trapping of mother solution leading to the formation of ‘fluid’ or ‘liquid inclusions’—in matrix crystals become visible which were paid no or low attention to the field of MALDI before [29].
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
Although such defects are well-known in crystal growth literature [30–32], a greater emphasis is placed on the defects here with respect to MALDI-samples (large crystals), dependencies on crystal growth rate, and addition of impurities (analyte molecules). In particular, the question where analyte molecules are localized—in the liquid inclusions or in the solid matrix—is followed by our study and discussed taking recent results of crystal growth into account [33–36]. 2. Experimental 2.1. Materials 2,5-DHB was purchased from Fluka, Buchs, Switzerland, and 2,6-DHB was obtained from Janssen Chimica, Geel, Belgium. Both matrices were purified by activated carbon and recrystallized from an aqueous solution, because this is the standard procedure in the lab prior to all MALDI-preparations for all matrices. Aliquots of horse heart cytochrome c, CC (Fluka, Buchs, Switzerland, 12.6 ku), Texas-Red® -labeled avidin (Molecular Probes, Leiden, The Netherlands), bovine serum albumin (BSA) (Sigma, St. Louis, MO, USA, 66 ku), and ␥-globulin, GG (Serva, Heidelberg, Germany, 150 ku) were used as test substances. BSA labeled with 20 ± 0.5 nm colloidal gold (BSA-Au) was purchased from British BioCell International, Cardiff, Great Britain. Suitable aliquots of BSA were labeled in house with sulfo-succinimido Nanogold® , Nano-Au (Nanoprobes, New York, USA, 1.4 nm in size). Label procedure was similar to the protocol given by the company. To achieve a suitable labeling, a Nano-Au-to-protein ratio of 5:1 was chosen. Column gel chromatography with Sephadex G-50 material (Amersham Pharmacia, Freiburg, Germany) as stationary phase and bi-distilled water as mobile phase was used to separate unbound Nano-Au (waste) from unlabeled and labeled BSA proteins (sample investigated); the separation was performed according to the protocol of the manufacturer. Separation of labeled and unlabeled protein was not performed. Nano-Au-labeled BSA was analyzed by MALDI-MS and SDS–PAGE.
119
2.2. Crystal growth Crystal growth was carried out similar to the method described in a previous report [19]. 2,5-DHB was solved in bi-distilled water in a concentration of 60 g/L, 2,6-DHB was dissolved to saturation concentration of 10 g/L in bi-distilled water. Both matrix solutions were placed in an ultrasonic bath for 15 min for faster dissolution, and for release of solvated gas molecules from the solutions. Afterwards both solutions were placed in a water bath at 36 ◦ C for 1 h to insure a complete dissolution. Analyte molecules were dissolved in bi-distilled water separately and added to the matrix solution to give a molar analyte-to-matrix ratio of 10−4 for all protein samples. These and pure matrix solutions were continuously cooled from 36 ◦ C down to 4 ◦ C without stirring in 1, 24, or 72 h (cooling rates of 32, 1.3, and 0.44 K/h, respectively). Crystal harvesting was performed by filtration. The crystals were washed with bi-distilled water at 4 ◦ C and dried in an oven at a temperature of 30 ◦ C for 24 h. 2.3. Preparation For SEM analysis, single crystals of sufficient size (≥1 mm in each dimension) were chosen under the optical polarization microscope (Olympus SHZ, Olympus, Hamburg, Germany). Prior to investigations by SEM, mass spectrometry of crystals was carried out to verify the incorporation of analyte molecules. Crystals of 2,5-DHB were cleaved mechanically with a scalpel parallel to their largest surface; further details are given in the result part (Section 3.1). 2,6-DHB crystals were cleaved parallel to the smallest crystal surface (Section 3.4). Subsequently, the freshly opened inner faces were mounted on aluminum specimen stubs with a small amount of electrically conductive carbon glue (PLANO, Wetzlar, Germany). Each crystal was carefully adjusted in the viscous glue with cleaved surface upwards and easily accessible for scanning electron microscopic observations. After 2 h air-drying of the glue the samples were rotary shadowed at room temperature with about 2 or 5 nm Pt/C at an elevation angle of 65◦ to
120
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
provide a sufficient electrical conductivity at their surface. The vapor deposition was performed in the high vacuum chamber of a freeze-etch device (BAF 300 with turbo molecular pump, Balzers/Liechtenstein). The thickness of the Pt/C film was measured with a quartz crystal film thickness monitor (Balzers QSG 201D, Balzers/Liechtenstein). For some crystals, the other half was analyzed by X-ray diffraction to investigate the quality of single crystal patterns. The sharpness of the diffraction pattern shows the averaged extent of long-range order for the whole analyzed crystal. 2.4. Instrumentation MALDI-mass analysis was performed on a modified VISION 2000 (Finnigan MAT, Bremen, Germany) equipped with a nitrogen laser (LSI Inc., Franklin, USA) emitting an laser pulse with pulse duration of 3 ns at a wavelength of 337 nm. Linear mode with delayed extraction as well as reflectron mode were used with an ion acceleration of 25 and 10 kV, respectively. Ions were post-accelerated to a total ion kinetic energy of 30 keV onto a conversion dynode in front of a secondary electron multiplier. The detector signal was pre-amplified and acquired on a transient recorder (LeCroy 9545, Chestnut, NY, USA). The coated crystals were examined in a high resolution field emission scanning electron microscopy (FE-SEM; ‘in-lens’ type, model S-5000 [37], Hitachi Ltd., Tokyo, Japan) in high vacuum (P = 4×10−5 Pa) at room temperature. Micrographs were recorded at acceleration voltages from 3 up to 8 kV using secondary electron (SE) and back scattered electron (BSE) mode, respectively. SE imaging and electron energies = 5 keV preferentially were used to monitor the topography of the surface taking advantage of the so-called ‘mass thickness contrast’ [38]. The information depth obtained in the SE mode is in the order of 5–10 nm for our studies. For the unambiguous detection of colloidal gold particles at the surface and within a thin surface layer, respectively, BSE was used because of a strong ‘Z-contrast’ (for further methodical details of the FE-SEM and detection of colloidal
gold see [39–41]). The thickness of the thin surface layer which contribute to the BSE signal is roughly given by 1/3 of the electron range R in the specimen which amounts to approximately R = 1.5 m for 10 keV electrons in carbonaceous matter [42]. For the results shown below the information depth obtained in the BSE mode is in the range of several 100 nm. All micrographs were recorded using Agfapan APX100 film. X-ray diffraction patterns were investigated by a Siemens P3 diffractrometer (λ = 0.71073 Å, Mo K␣). The amount of gold in the colloidal gold sample was quantitatively determined by inductively coupled plasma mass spectrometry (ICP-MS), ELAN 500 (PE Sciex). Quantitative determination of the BSA concentration in the colloidal gold sample was performed by UV-absorption spectrophotometry (UV-2101 PC, Shimadzu) at 290 nm wavelength. Spectral absorption of gold was taken from the literature [43] while the BSA-absorption was determined using standard solutions of neat BSA.
3. Results 3.1. 2,5-DHB crystals 2,5-DHB forms crystals of orthorhombic habit with a size from several tens of microns up to 1 mm. Mass spectrometry of analyte-doped single crystals of 2,5-DHB of the same lot confirmed the incorporation of analyte molecules into 2,5-DHB. This is in good agreement with former studies [11,14,20,21]. Under air, all crystals cleaved easily through the middle using a scalpel (at the crystal edge). 2,5-DHB crystals could be cleaved along all surfaces. In the work described here the largest surface was chosen which is the (1 0 0) face of the crystal [16]. 3.2. Au-labeled proteins Without further treatment analyte and matrix molecules cannot be distinguished in SEM due to their compositional similarity. Suitably labeled analyte molecules are therefore needed. Gold enables an
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
121
Fig. 1. Micrographs of opened inner faces of a 2,5-DHB crystal doped with 20 nm colloidal gold labeled bovine serum albumin (BSA-Au) in back scattered electron mode (A), and secondary electron mode (B), respectively. Scale bar in the images corresponds to 1 m. The crystal was grown with a cooling rate of 0.44 K/h. Both micrographs were recorded at an electron acceleration voltage of 8 kV.
unambiguous detection especially in BSE mode due to the strong Z-contrast. To localize analyte molecules in the opened inner face of a cleaved 2,5-DHB crystals, BSA was labeled with two different gold particles (a) conventional colloidal gold with an averaged diameter of 20 nm and (b) Nano-Au, 1.4 nm in size. Fig. 1 compares micrographs of BSE mode (Fig. 1A) and SE mode (Fig. 1B) of an opened inner face of a 2,5-DHB crystal doped with BSA-Au. An electron acceleration voltage of 8 kV was used for SEM. A molar analyte-to-matrix ratio of 10−4 was adjusted in the mother solution, a cooling rate of 0.44 K/h was applied in crystal growth. Both images show a planar structure with defects of different sizes and rectangular shape. The defects appear in dark contrast, meaning less or no release of back scattered or SEs, respectively, which is equivalent to a lack of material; in SE mode edges appear brighter than planar structures due to higher SE yield [38]. Therefore, these defects are interpreted as voids or cavities. A more detailed description can be found below (Section 3.2). In Fig. 1A, the colloidal gold particles are easily detectable as bright spots with nearly circular shape.
A more or less homogeneous distribution of gold particles is observed over the inspected surface, but accumulations can also be found (data not shown). Most gold particles are incorporated in an edge of the voids or cavities. Exceptions are gold particles 6, 8, and 10 (see Fig. 1A), which are located between the voids. Nearly all particles can be observed well in the SE mode, too, even though their location at edges complicates their identifications (Fig. 1B). For a better visualization, insets of the gold particles are shown in Fig. 1B enlargements and some Au particles are carefully circled by a dotted line in the insets. The fact that the colloidal gold particles can be detected in both imaging modes, especially in SE mode (information depth approximately 5–10 nm), leads to the conclusion that Au particles are incorporated in the upper layer of the opened inner face with a thickness equal to the information depth. The BSA-Au sample was investigated by MALDIMS using a standard preparation of DHBs-matrix. MALDI-mass spectra of this sample show signals of the unlabeled BSA only. There are potentially several reasons for the failure in BSA-Au detection: either
122
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
an unfavorable labeling rate (BSA-Au:BSA 1), a loss of the label during mass analysis, or—most probably—a too high mass of the colloidal Au particles to be detected by MALDI-MS. The latter argument is certainly quite reasonable, because the molecular mass of a colloidal gold particle 20 nm in diameter calculates to around 50 × 106 u. In order to learn about the gold particle-BSA labeling rate, the sample was investigated by ICP-MS for Au content and by UV-absorption spectrophotometry for protein content. These quantitative determinations revealed that the ratio of gold particle to BSA is around 3.7 × 10−6 only. Therefore, a gold particle density of 2.3 m−3 can be expected assuming a homogeneous distribution of BSA-Au in the crystal; this value is about three times lower than the observed gold particle density of around 6 m−3 (see Fig. 1). The low labeling rate of 3.7 × 10−6 , unfortunately, suggests that analyte incorporation in MALDI matrices cannot be investigated reasonably with such samples. Nano-Au was tested as label (1.4 nm in size, approximately 15 ku in weight) and covalently bound to BSA in house. Control of labeling rate was performed by MALDI-MS and SDS–PAGE. While the gold label is not stable under MALDI-conditions (labeled peak as well as neat BSA and Nano-Au fragment appear in the mass spectrum), SDS–PAGE reveals that Nano-Au
is covalently bound to the protein. A quantification of the labeling rate was, however, not performed. SEM of 2,5-DHB crystals doped with Nano-Aulabeled BSA reveals voids similar to those found in the colloidal Au-case (data not shown). Detection of the much smaller Nano-Au particles with SEM, however, requires a higher magnification. Unfortunately, the significantly higher electron doses destroys 2,5-DHB crystals. Therefore, no micrographs of Nano-Au-labeled BSA incorporated into crystals of 2,5-DHB could be obtained with sufficient magnification to detect Nano-Au. This is astonishing because Nano-Au is used in biological SEM, too, without destruction of the sample; it is supposed that the organic matrix is much more labile than typical biological tissue and therefore does not survive the electron radiation under the required conditions. 3.3. Crystal defects 3.3.1. Influence of the rate of crystal growth Fig. 2 shows scanning electron micrographs of opened inner faces of cleaved CC-doped 2,5-DHB crystals of different cooling rates used for crystal growth. SE mode and an electron acceleration voltage of 5 kV were used. Aqueous mother solutions contained a molar analyte-to-matrix ratio of 10−4 , and a cooling rate of 1.3 K/h was applied to the crystal
Fig. 2. Secondary electron micrographs of opened inner faces of CC-doped 2,5-DHB crystals. Crystals were grown with a cooling rate of 1.3 K/h (A) and 32 K/h (B), respectively. Bar in images corresponds to 30 m. Micrographs were recorded at an electron acceleration voltage of 5 kV.
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
shown in Fig. 2A while Fig. 2B shows a crystal grown with a cooling rate of 32 K/h. The bars in the images correspond to lengths given in figures. Both surfaces show a planar structure with edges and voids similar to Fig. 1B. In Fig. 2A, large edges cross the image in the middle from right to left. Along these edges rectangular voids can be found in sizes of ca. 1 m in width and 3 m in length. All voids lie next to each other and have, therefore, a parallel orientation. Some separate larger voids can be seen in the plane, starting from one cleavage edge and going up the micrograph. These voids have sizes of 3 m × 6 m. They have an parallel orientation to those voids at the edges. Applying a higher cooling rate (Fig. 2B) results in many more edges all over the image. Voids of similar size, but in larger number can be found everywhere in the plane. The shape of the voids is not as uniform as observed for voids in Fig. 2A. Indeed, there is less
123
orientation. The optical extinction of both crystals between crossed polarizers of suitable orientation (optical microscopy) indicates the crystals as single crystals of sufficient quality. No X-ray diffraction pattern analysis of these crystals was performed. In order to verify that the defects can indeed be interpreted as voids, the spatial structures of defects of BSA-doped crystals were inspected by stereopair SE micrographs. Images were recorded at medium and high magnification with tilt angles of ±5◦ (data not shown). Using a special home-built stereo-viewer the spatial shape of voids and cavities, respectively, became visually accessible; a determination of the depths of these defects was not performed. Micrographs of opened inner faces of 2,5-DHB crystals doped with GG (molar analyte-to-matrix ratio 10−4 ) are shown in Fig. 3. An electron acceleration voltage of 3 kV and SE mode were used. For
Fig. 3. Secondary electron micrographs of opened inner faces of GG-doped 2,5-DHB crystals at different magnifications (bars in images correspond to given lengths). Crystals were grown with a cooling rate of 1.3 K/h (A and C) or of 32 K/h (B and D). Micrographs were recorded with an electron acceleration voltage of 3 kV.
124
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
micrographs shown on the left hand side (Fig. 3A and C, magnification of (A)) the solution was cooled applying a cooling rate at 1.3 K/h; right hand side (Fig. 3B and D, magnification of (B)) the cooling rate was 32 K/h. Planar structures can be found for the slower grown crystal in Fig. 3A with voids of 1 m × 3 m in size. Compared to the opened inner faces of CC-doped crystals in Fig. 2, these voids are less symmetric in shape, smaller in size, and larger in number. However, they still have an orientation to each other even though it is less pronounced. A higher magnification of the plane between the voids (Fig. 3C), shows a porous structure of material, cavities, with dimensions of 880 down to 20 nm. No planar structure with voids can be seen for the crystal cooled down with a larger cooling rate of 32 K/h at low magnification (Fig. 3B). The edges show no preferred orientation. At higher magnification a structure similar to a ‘Swiss cheese’ appears (Fig. 3D). The surface is pervaded with porous cavities, from 100 down to 10 nm in size, no planar structure exits in between. It is worth mentioning that both crystals shown in Fig. 3 were inspected by optical polarization microscopy and by X-ray diffraction; none of the crystals
could be classified as single crystals by these techniques. The above described observations can be summarized: an increase of the cooling rate (faster crystal growth) results an increase of the number of voids, in a decrease of the size of the voids and in a decrease of crystal orientation. 3.3.2. Influence of the size of analyte molecules Fig. 4 compares structures of opened inner faces of 2,5-DHB crystals doped with analyte molecules of different size. An electron acceleration voltage of 3 kV and SE mode were used. All crystals were grown in the same cooling run with a cooling rate of 0.44 K/h. Fig. 4A and D show the opened inner face of a neat 2,5-DHB crystal, Fig. 4B and E reflect the face of a 2,5-DHB crystal doped with CC (12,360 u), and Fig. 4C and F of a BSA-doped (66,430 u) 2,5-DHB crystal, respectively. In the top row (low magnification) an overview of the opened inner face of the different crystals is given. Areas marked by the square are magnified and shown in the bottom row of Fig. 4. Voids can be found in each crystal. The size of the voids decreases from neat matrix crystals (largest
Fig. 4. Secondary electron micrographs of opened inner faces of crystals of neat 2,5-DHB (A and D), doped with CC (B and E), and doped with BSA (C and F), respectively. Different magnifications are shown (bars in images correspond to given lengths). All crystals were grown with a cooling rate of 0.44 K/h. The micrographs were recorded at an electron acceleration voltage of 3 kV.
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
void in Fig. 4A: 14 m × 18 m) to BSA-doped crystals (largest void in Fig. 4C: 2.1 m × 0.5 m). Their number meanwhile increases: for neat 2,5-DHB the occurrence of voids is approximately 0.04 per 1 m2 , for the CC-doped crystal it is increased to 0.4 per 1 m2 , and up to 10 per 1 m2 for BSA as analyte molecule. For a neat 2,5-DHB crystal no further voids appear in the image with higher magnification (Fig. 4D). The voids are rectangular and oriented to each other. It should be mentioned that the structures inside the upper two voids (Fig. 4D) are damaged due to the electron beam irradiation during scanning. In Fig. 4E, smaller voids with sizes of 500 nm × 500 nm down to 200 nm × 200 nm become apparent for the CC-doped 2,5-DHB crystal inspected at higher magnification. The shape of the voids is still rectangular and edges are of the same orientation. The BSA-doped 2,5-DHB crystal in Fig. 4F shows lots of voids with sizes down to some 50 nm in each dimension. The characteristics of the voids are similar to those of the 2,5-DHB crystal doped with BSA-Au as seen in Fig. 1B. The larger voids with rectangular shapes are oriented parallel to each other. For the smaller voids an asymmetrical shape is found. Within a certain lot (same, identical mother solution, analyte content, and cooling run) voids were reproducible in size and number for different crystals. Keeping all parameters the same but performing the crystals growth at different days, voids vary in size, shape and number, but by far not as dramatically as for changes made by the application of different cooling rates or the variation of the size of analyte molecules. Despite the large number of voids, this crystal (Fig. 4C and F) is characterized as a single crystal by X-ray diffraction. To all of our knowledge it is not important for a successful MALDI-mass analysis, whether the analysis is performed from a single crystal (100 m in size) or from a polycrystal from the same size. However, other analytical results obtained by, e.g., CLSM are most likely quite difficult to interpret due to voids or other crystal defects, because they introduce complicated optical characteristics in the crystals [44].
125
The above described observations can be summarized: an increase of the analyte size results in an increase of the number of the voids, in a decrease of the sizes of the voids, and in less oriented crystal growth. 3.4. 2,6-DHB crystals As described previously [11,21], 2,6-DHB matrix forms long thin needles and excludes protein molecules added to the mother solution from its crystal lattice. Although it was tedious and not highly reproducible to obtain large enough crystals of 2,6-DHB, a crystal of 2,6-DHB of sufficient size (≥1 mm in each dimension) was obtained from a mother solution containing Texas-Red® -labeled avidin in a molar analyte-to-matrix ratio of 10−4 . Optical polarization microscopy indicates the crystal as a single crystal. The 2,6-DHB crystal could be cleaved easily only along the smallest crystal surface into two parts. For the other two orientations the crystal fractured into several small pieces; therefore, only the first was taken into account for further considerations. Strong electrical charging occurred with a Pt/C film of 2 nm while investigating the opened inner face with SEM even at low acceleration voltages and low magnification (low doses). After a further coating with 3 nm Pt/C, micrographs as shown in Fig. 5 could be obtained using SE mode and an electron acceleration voltage of 3 kV. Fig. 5 shows two micrographs of this opened inner face of 2,6-DHB single crystal at different magnifications. Instead of an almost smooth plane as for 2,5-DHB thin needles stick out of the plane and a polycrystalline morphology appears between the needles with two different structures (Fig. 5A). The needles have a parallel epedical shape and different sizes. They are oriented randomly over the plane. A higher magnification of the area between the needles shows both a more flat and crystalline structure as well as a rough and grainy structure as seen in Fig. 5B. The more flat structure shows edges and voids of nearly symmetrical shape. These structures are comparable to those found for 2,5-DHB. The grainy structure has no symmetry or orientation. None of the structures is
126
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
Fig. 5. Secondary electron micrograph of an opened inner face of a 2,6-DHB crystal (A) and a magnification of the area marked by the square (B). Mother solution contained Texas-Red® -labeled avidin as analyte molecule. The crystal was grown with a cooling rate of 0.44 K/h. Scale bars in images correspond to given lengths. The micrographs were recorded at an electron acceleration voltage of 3 kV.
compact; there are voids for the flat structure and gaps for the grainy structure.
4. Discussion 4.1. Au-labeled proteins Large crystals of 2,5-DHB were doped with BSA which had been labeled with two different gold labels, colloidal gold and Nano-Au. Conventional colloidal gold of 20 nm in size can be detected by high resolution field emission SEM without destruction of the organic substrate (matrix crystal). The gold particles are clearly seen as light dots in the electron micrographs in BSE mode (see Fig. 1A). Nevertheless, the high amount of unlabeled protein (label rate 3.7 × 10−6 ), the unfavorable ratio of gold label size to analyte size (5:1, diameter wise; 125:1 volume wise), and the more or less unknown mechanism of protein binding under the used conditions of crystal growth, e.g., the quite acidic pH value of the mother solution (matrix) compared to the typical conditions used for Au-labeling such as physiological pH, disables colloidal gold labeling for our intention to study matrix–analyte interaction. The ostensible observation that Au-labeled proteins reside preferentially in the voids and are incorporated in the cleavage plane cannot be generalized with respect to analyte distribution in the matrix
crystal, because most of the protein molecules in the matrix crystal remained unlabeled. On the other side, taking the below discussed Nano-Au results into account, the colloidal gold results show the principle feasibility to detect gold particles of sufficient size in a sensible sample, such as a MALDI-matrix crystal. A more promising candidate seemed to be Nano-Au which had been covalently bound to the protein. The covalent binding and the smaller size of Nano-Au promises that it is a much more faithful, reliable possibility for our goal. Unfortunately, for an unambiguous detection of these small Au particles higher magnifications and consequently higher electron doses had to be applied in SEM so that the matrix crystal (organic and non-conductive material), was immediately destroyed during irradiation. Under non-destructive conditions (at small magnification) it was not possible to detect Nano-Au particles. The results obtained and discussed show that it is a challenging job to optimize the experimental conditions for high resolution field emission SEM and/or to develop more appropriate labeling methods for SEM to investigate matrix–analyte interaction in MALDI-samples. 4.2. Crystal growth The observation of crystal defects (in our case voids and cavities) in opened inner faces of cleaved 2,5-DHB
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
crystals was surprising. A closer inspection of images of previous reports reveals that crystal defects occur in slowly grown single crystals of MALDI matrices [3,19,45]. SEM-investigations of crystal surfaces show cavities [3], using optical microscopy to inspect single crystals of succinic acid inner defects can be visualized in the crystals [19,45]. Crystal defects have been known for decades and are well described in the literature of crystallography and crystal growth [30,33]. Defects can be of various types, e.g., defects are induced by the incorporation of guest particles in crystal structures on the atomic scale or by layer displacements of macroscopic size. Also defects can occur due to different mechanisms, depending on the growth technique. In the following we exclusively discuss formation of so-called ‘fluid’ or ‘liquid inclusions’ meaning the trapping of supersaturated mother solution into the crystal lattice, a very well-known macroscopic phase defect phenomenon for crystals grown from solution [30–35]. A liquid inclusion means a volume inside the crystal lattice which is not crystallized to solid state, but contains mother solution. Because crystals are grown from solution in our case the formation of liquid inclusions seems to be the most obvious interpretation for all observed defects (voids and cavities). Inclusions of gas bubbles can (most likely) be excluded due to the chosen preparation procedure (Section 2.2). 4.2.1. Influence of the rate of crystals growth In this context some general comments on crystal growth are mentioned first: crystals have many faces which have individual growth rates depending on the conditions of the surrounding medium. The growth rate is responsible for the morphology of a crystal and it is obvious that the face with the fastest growth rate tends to be the smallest surface, while the face with the slowest growth rate tends to be the largest surface. It is well known that growing faces are not flat but have steps of different heights, so-called macrosteps. Based on these macrosteps Chernov developed a model to explain the formation of liquid inclusions parallel to a slowly growing face. This model is briefly presented here in [31]. During crystal growth, macrosteps of a
127
slowly growing face grow towards each other. Due to a disturbance an area of a fast growing face between two macrosteps is slowed down or stopped while the rest remains undisturbed. A more slowly growing face can be fast enough to grow over the disturbance resulting in an overhanging surface layer. An inclusion (hole) is built, if the layer reaches another macrostep to cocrystallize. The inclusion is filled with the surrounding media, i.e., mother liquid; so-called ‘liquid’ or ‘fluid inclusions’ are built. The mechanism strongly depends on the hydrodynamic effects in the solution such as locally different concentrations, temperatures, or pH values, and can occur for both, guest molecule-free systems and guest molecule containing mother solutions. It might be deduced from the Chernov model that the faster the crystal grows the more likely is the formation of such liquid inclusions. Applying these considerations to our case of slowly grown (analyte-free or analyte-doped) matrix crystals, our results are in reasonable agreement with the Chernov model (see Figs. 2 and 3): the larger the cooling rate the greater the supersaturation of the solution, the faster the crystal growth, and the larger the amount and the smaller the size of liquid inclusions (Section 3.3.1). With this respect, the observation of voids in slowly grown matrix crystals is not surprising. For a typical MALDI-sample preparation, such as dried droplet or thin layer preparation, the growth conditions are even rougher. Lots of liquid inclusions containing mother solution (solved matrix and analyte molecules) are expected in crystals of standard preparation as mentioned by Figueroa et al. [29]. The interpretation of the Chernov model that voids are filled with matrix–analyte solution is also in reasonable agreement with recent 1 H NMR results of Krüger et al. who could show that free organic solvent (acetonitrile) is indeed found in crystals of 2,5-DHB [27]. Whether there is an influence of the liquid inclusions on the MALDI process and whether liquid inclusions are essential for the MALDI process is yet another question. 4.2.2. Influence of the size of analyte molecules The influence of the size of the analyte molecules on the crystal growth is another aspect to be discussed.
128
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
Whether analyte molecules are considered as an impurity (or guest molecules) to host matrix crystals or whether matrix and analyte molecules build a new crystal structure, depends, in first approximation, on the analyte-to-matrix ratio. For the molar analyte-to-matrix ratio used in this study (10−4 ) analyte molecules can be considered as impurity/guest molecules; this is substantiated by previous results which showed that the crystal structure detected by X-ray diffraction patterns is not disturbed by analyte molecules incorporated using the typical, large molar matrix excess [14,19]. It should be mentioned in this context, however, that the original matrix crystal structure can be replaced by a new crystal structure of both matrix and analyte if analyte molecules cannot be regarded longer as an impurity due to equal molar amounts. This was described recently by Mele and Malpezzi [46] (ratio: 1:5) and Kinsel et al. [47] (ratio: 1:1) who found that the original crystal structure of the matrix compound (2,5-DHB) is substituted by a new crystal lattice containing matrix and analyte molecules, cyclodextrin and prolin, respectively. Taking a high matrix excess into account as given in our studies analyte molecules act as an impurity; the effects of guest molecules on crystal growth can be of very different types and have been recently discussed in the crystal growth literature [34,36]. It is speculated that analyte molecules can act in two opposing ways (a) as growth promoter or (b) as growth inhibitor, and it assumed that both alternatives can be responsible for the formation of liquid inclusions. The crystal growth rate is increased by analyte molecules if analytes act as growth promoters. Analyte molecules acting as growth promoters are supposed to typically lead to an increased nucleation and less oriented, dendritic growth. On the other hand, analyte molecules act as growth inhibitors if crystal growth rate is decreased. Analyte molecules acting as growth inhibitors are expected to typically lead to a delay of nucleation, to a higher supersaturation of the mother solution, and to a different mechanism resulting in a different and less oriented crystal growth [31,34,36]. Our results are in good agreement with these considerations: the larger the size of the analyte molecules
the less oriented the crystals grow, the larger the voids become in number and smaller in size (Section 3.3.2). The non-oriented, dendritic growth is clearly observable for the fast grown GG-doped 2,5-DHB crystal (Fig. 3D) while more oriented voids are found for CC-doped 2,5-DHB crystal applying the same cooling rate (Fig. 2B). This analyte size dependency is obviously seen in Fig. 4, too. Few large symmetric voids are observed for analyte-free 2,5-DHB crystals (Fig. 4A and D) whereas voids become less symmetric, large in number, and smaller in size for BSA-doped 2,5-DHB crystals (Fig. 4C and E). Our studies are insufficient so far to differentiate between the influence of analyte molecules as growth promoters or inhibitors. Further investigations— comparison of nucleation time of pure and analytedoped matrix solution—need to be undertaken to elucidate the actual mechanism in the case of MALDI-samples in general and of protein-doped 2,5-DHB-matrix crystals in special. 4.3. Cleavage of 2,6-DHB crystals As discussed above, the formation of liquid inclusions is one possible mechanism for guest molecule incorporation into host crystals. A second discussed mechanism is the formation of a so-called ‘solid solution’ [34]. Such a solid solution describes the incorporation and interaction of guest and host molecules on a molecular scale (e.g., zero-dimensional or point defects) as opposed to the above discussed more macroscopic scale of liquid inclusions (three-dimensional or phase defects). Several considerations have been undertaken to understand the mechanism of a solid solution formation [36,48]. The term ‘solid solution’ had been proposed in the MALDI community [18] without the context of the theoretical background of crystal growth. The experimentally observed phenomenon of analyte incorporation into matrix crystals could be described best by the model of a solid solution meaning the separation of guest molecules from each other and their solvation by surrounding matrix molecules in a solid state [14,15,20,49,50].
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
To examine whether liquid inclusions are of importance for (quantitative) analyte incorporation, 2,6-DHB was investigated for liquid inclusions by SEM. This matrix is known to behave quite differently to 2,5-DHB matrix: in contrast to 2,5-DHB which incorporates analyte molecules into the crystal quantitatively (meaning the molar analyte-to-matrix ratio prepared in the mother solution was found in the grown crystals) [14,20], 2,6-DHB excludes analyte molecules from incorporation [11,21]. Micrographs reveal that liquid inclusions are also found in opened inner faces of a 2,6-DHB crystal grown from a solution containing analyte molecules. This result does not support the idea that liquid inclusions are sufficient or even required for (quantitative) analyte incorporation into MALDI matrix crystals: 2,6-DHB shows liquid inclusions but excludes analyte molecules from its interior up to the detection limits of the used methods [11,21]. For 2,5-DHB a quantitative analyte incorporation was found applying different growth rates, 0.19 up to 1.3 K/h [14,20,11]. It is reasonable to assume— according to the experimental data in the present study—that the number of liquid inclusions varies for the different crystal growth rates while the incorporation rate remains the same. In other words, the number and size of liquid inclusions depend on the cooling rate while the incorporation rate stays constant. Therefore, it is concluded that the majority of analyte molecules is incorporated on the basis of a solid solution into 2,5-DHB crystals.
5. Conclusions The feasibility of high resolution field emission SEM to investigate the matrix–analyte interaction in MALDI samples has been examined by the present study. Large crystals of 2,5-DHB doped with analyte molecules different in label and size, respectively, were mechanically cleaved and SEM-micrographs of the opened inner face were recorded with different detection modes. Our results show that the—at first glance—quite straight forward and promising looking
129
Au-labeling of analyte molecules and its detection by SEM is obviously not the method of choice to inspect matrix–analyte interaction in MALDI-samples. Meanwhile the appearance of voids and cavities in matrix crystals was observed and examined in dependence of the rate of crystal growth and analyte size. It is concluded that these voids and cavities contain mother solution prior to cleavage and are known as ‘liquid’ or ‘fluid inclusions’ in crystal growth literature. As liquid inclusions do not seem to be the main mechanism for (quantitative) incorporation, there is reason to conclude that analyte molecules are incorporated as a solid solution formation into 2,5-DHB matrix crystals. The mechanism of molecular interaction during incorporation and the location of analyte molecules in the matrix crystal lattice, respectively, remain open. These subjects are of great interest for a further understanding of the ‘true MALDI’ process. The results and interpretation might be taken as motivation for more sophisticated investigations of the molecular contact between matrix and analyte molecules of both, matrices that incorporate and matrices that exclude analyte molecules from the crystal lattice. Further examination of the influence of analyte molecules on nucleation and on crystal growth rate as well as observation of crystal faces by atomic force microscopy (AFM) may be of help to increase our understanding of incorporation mechanism and matrix–analyte interaction. Whether the residual solvent content trapped in liquid inclusions may have some influence on the plume dynamics and chemistry of both, ‘matrix support effect’ and the ‘true MALDI’ process, is yet an open question.
Acknowledgements We gratefully acknowledge Professor Dr. Franz Hillenkamp for the opportunity to perform the presented experiments in his laboratories, for the many discussions on the topic matrix–analyte interaction, and giving many good hints in this field. Thanks to Professor Dr. Andrew Putnis, Institute for Mineralogy,
130
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131
University of Münster, for his discussions and literature on crystals growth mechanism and crystal defects. Professor Dr. B. Krebs and Ms. D. Wolf, Inorganic Institute, University of Münster, are thanked for X-ray diffraction measurements and help with the interpretation. We thank Ms. U. Keller and Ms. U. Malkus for the specimen preparation and the scanning electron microscopic investigations and Ms. G. Kiefermann (IMPB) for photographic work. Thanks to Ms. K. Zerf (IMPB) for the help with the SDS–PAGE of the Nano-Au-labeled BSA sample.
References [1] M. Karas, F. Hillenkamp, Anal. Chem. 60 (1988) 2299. [2] D.T.W. Chan, A.W. Colburn, P.J. Derrick, Organic Mass Spectrom. 27 (1992) 188. [3] A. Westman, T. Huth-Fehre, P. Demirev, B.U.R. Sundqvist, J. Mass Spectrom. 30 (1995) 206. [4] A.I. Gusev, W.R. Wilkinson, A. Proctor, D.M. Hercules, Anal. Chem. 67 (1995) 1034. [5] P. Onnerfjord, S. Ekstrom, J. Bergquist, J. Nilsson, T. Laurell, G. Marko-Varga, Rapid Commun. Mass Spectrom. 13 (1999) 315. [6] R.R. Hensel, R.G. King, K.G. Owens, Rapid Commun. Mass Spectrom. 11 (1997) 1785. [7] D.S. Jones, K.S. Pobinsons, S.P. Thompson, P. Humphrey, in: Proceedings of the 43rd ASMS Conference, 1995, p. 692. [8] S.J. Doktycz, P.J. Savickas, D.A. Krueger, Rapid Commun. Mass Spectrom. 5 (1991) 145. [9] R.C. King, K.G. Owens, in: Proceedings of the 42nd ASMS Conference, 1994, p. 977. [10] Y. Dai, R.M. Whittal, L. Li, Anal. Chem. 68 (1996) 2494. [11] V. Horneffer, A. Forsmann, K. Strupat, F. Hillenkamp, U. Kubitscheck, Anal. Chem. 73 (2001) 1016. [12] B. Spengler, M. Hubert, R. Kaufmann, in: Proceedings of the 42nd ASMS Conference, 1994, p. 1041. [13] R.W. Garden, J.V. Sweedler, Anal. Chem. 72 (2000) 30. [14] K. Strupat, M. Karas, F. Hillenkamp, Int. J. Mass Spectrom. Ion Process. 111 (1991) 89. [15] R.C. Beavis, J.N. Bridson, J. Phys. D: Appl. Phys. 26 (1993) 442. [16] M. Haisa, S. Kashino, S.I. Hanada, K. Tanaka, S. Okazaki, M. Shibagaki, Acta Cryst. B38 (1982) 1480; M. Haisa, S. Kashino, S.I. Hanada, K. Tanaka, S. Okazaki, M. Shibagaki, Acta Cryst. B38 (1982) 2984. [17] K. Strupat, M. Karas, F. Hillenkamp, in: Proceedings of the 12th International Mass Spectrometry Conference (IMSC), 26–30 August, Amsterdam, The Netherlands, 1991. [18] F. Hillenkamp, M. Karas, R.C. Beavis, B.T. Chait, Anal. Chem. 63 (1991) 1193A.
[19] J. Kampmeier, K. Dreisewerd, M. Schürenberg, K. Strupat, Int. J. Mass Spectrom. Ion Process. 169/170 (1997) 31. [20] K. Strupat, J. Kampmeier, V. Horneffer, Int. J. Mass Spectrom. Ion Process. 169/170 (1997) 43. [21] V. Horneffer, K. Dreisewerd, H.-C. Lüdemann, F. Hillenkamp, M. Läge, K. Strupat, Int. J. Mass Spectrom. 185–187 (1999) 859. [22] M. Karas, M. Glückmann, S. Schäfer, J. Mass Spectrom. 35 (1) (2000) 1. [23] M. Glückmann, A. Pfenninger, R. Krüger, M. Thierolf, M. Karas, V. Horneffer, F. Hillenkamp, K. Strupat, Int. J. Mass Spectrom. 210/211 (2001) 121. [24] M. Glückmann, A. Pfenninger, M. Karas, V. Horneffer, F. Hillenkamp, K. Strupat, in: Proceedings of the 34th Discussion Meeting of the German Society for Mass Spectrometry (DGMS), 4–7 March, München, Germany, 2001. [25] V. Horneffer, F. Hillenkamp, K. Strupat, M. Glückmann, A. Pfenninger, M. Karas, in: Proceedings of the 34th Discussion Meeting of the German Society for Mass Spectrometry (DGMS), 4–7 March, München, Germany, 2001. [26] Manuscript in preparation. [27] R. Krüger, A. Pfenninger, I. Fournier, M. Glückmann, M. Karas, in: Proceedings of the 49th ASMS Conference, 27–31 May, Chicago, IL, USA, 2001. [28] M. Karas, M. Glückmann, R. Krüger, A. Pfenninger, I. Fournier, in: Proceedings of the 49th ASMS Conference, 27–31 May, Chicago, IL, USA, 2001. [29] I.D. Figueroa, O. Torres, D.H. Russell, Anal. Chem. 70 (1998) 4527. [30] K.T. Wilke, Kristallzüchtung, Harri Deutsch, Thun, 1988. [31] A.A. Chernov, Contemp. Phys. 30 (4) (1989) 251. [32] A. McEwan, R.I. Ristic, B.Y. Shekunov, J.N. Sherwood, J. Crystal Growth 167 (1996) 701. [33] A.A. Chernov, Acta Cryst. A54 (1998) 859. [34] G.G.Z. Zhang, D.J.W. Grant, Intern. J. Pharma. 181 (1999) 61. [35] N. Saito, M. Yokota, T. Fujiwara, N. Kubota, Chem. Eng. J. 79 (2000) 53. [36] K. Sangwal, Prog. Crystal Growth Charact. 32 (1996) 3. [37] M. Yamada, M. Nakagawa, M. Satoh, Y. Nakaizumi, Hitachi Instrument News, 27th ed., 1995, p. 26. [38] D.C. Joy, J. Microsc. 36 (2) (1984) 241. [39] M. Müller, P. Walther, R. Hermann, P. Schwarb, in: A.J. Verkleij, J.L.M. Leunissen (Eds.), Immunogold Labeling in Cell Biology, CRC Press, Boca Raton, FL, 1989, p. 199. [40] D.C. Joy, J.B. Pawley, Ultramicroscopy 47 (1992) 80. [41] L. Reimer, Image formation in low-voltage scanning electron microscopy, in: D.C.O Shea (Ed.), SPIE Optical Engineering Press, Bellingham, Washington, vol. TT12, 1993. [42] L. Reimer, in: P.W. Hawkes (Ed.), Springer Series in Optical Sciences, 2nd ed., 1998, Springer, Berlin. [43] P. Baudhuin, P. Van der Smissen, S. Beauvois, P.J. Courtoy, Colloidal Gold, Academic Press, 1991 (Chapter 1). [44] V. Horneffer, Ph.D. Thesis in Experimental Physics, University of Münster, Germany, 2002.
V. Horneffer et al. / International Journal of Mass Spectrometry 226 (2003) 117–131 [45] J. Kampmeier, Diploma Thesis in Experimental Physics, University of Münster, Germany, 1993. [46] A. Mele, L. Malpezzi, J. Am. Soc. Mass Spectrom. 11 (2000) 228. [47] G.R. Kinsel, Q. Zhao, D.S. Marynick, J. Hardesty, in: Proceedings of the 49th ASMS Conference, 27–31 May, Chicago, IL, USA, 2001.
131
[48] B. Kahr, R.W. Gurney, Chem. Rev. 101 (4) (2001) 893. [49] I. Fournier, R.C. Beavis, J.C. Blais, J.C. Tabet, G. Bolbach, Int. J. Mass Spectrom. Ion Process. 169/170 (1997) 19. [50] I. Fournier, M. Glückmann, M. Karas, in: Proceedings of the 49th ASMS Conference, 27–31 May, Chicago, IL, USA, 2001.