- Email: [email protected]
Improved performance in silicon enzyme microreactors obtained by homogeneous porous silicon carrier matrix
Talanta 56 (2002) 341– 353 www.elsevier.com/locate/talanta
Improved performance in silicon enzyme microreactors obtained by homogeneous porous silicon carrier matrix Martin Bengtsson a, Simon Ekstro¨m a, Gyo¨rgy Marko-Varga b, Thomas Laurell a,* a
Department of Electrical Measurements, Lund Institute of Technology, Uni6ersity of Lund, P.O. Box 118, 22100 Lund, Sweden b Department of Analytical Chemistry, Uni6ersity of Lund, P.O. Box 124, 22100 Lund, Sweden Received 17 April 2001; received in revised form 9 July 2001; accepted 10 July 2001
Abstract The catalytic performance of porous silicon (PS) micro enzyme reactors (mIMER) is strongly dependent on the PS matrix morphology for enzyme immobilisation. PS was achieved in the mIMER by anodisation in a HF– ethanol mixture. PS etching of structured silicon surfaces commonly results in an inhomogeneous pore formation. The deep channel microreactors described herein have previously suffered from these phenomena, yielding non-optimised mIMERs. In order to obtain a homogeneous PS layer on the deep microreactor channel walls, different reactor geometries (channel wall thicknesses of 50 and 75 mm) were anodised at 10 and 50 mA cm − 2 for anodisation times ranging between 0 and 50 min. The mIMERs were evaluated by immobilising two types of enzymes, glucose oxidase (GOx) and trypsin, and the resulting catalytic turnover was monitored by a colorimetric assay. It was found that reactors with a homogeneous PS matrix displayed improved performance. The trypsin mIMERs were used to digest a protein, b-casein, in an on-line format and the digest was analysed by MALDI-TOF MS. The importance of tailoring the mIMER geometry and the PS-matrix is crucial for the protein digestion. Successful protein identification after only 12 s. digestion was demonstrated for the best reactor, 75 mm channel wall, 25 mm channel width, anodised at 50 mA cm − 2 for 10 min. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Porous silicon; IMER; Enzyme; Carrier matrix
1. Introduction In systems for biochemical process monitoring, immobilised enzyme reactors (IMER) are commonly used tools for biospecific detection. In such reactors one of the most crucial factors is the * Corresponding author. E-mail address: [email protected] (T. Laurell).
effective immobilised area compared with the volume of the reactor. To increase the effective area in the reactor, porous material such as packed porous glass beads, and silica gel [1], is commonly used. In the current trend of miniaturising bioanalytical technology enzyme reactors has also been the focus for microsystem integration. The development of chip integrated microIMERs (mIMER) have fol-
0039-9140/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 9 1 4 0 ( 0 1 ) 0 0 6 0 0 - 2
342
M. Bengtsson et al. / Talanta 56 (2002) 341–353
lowed strategies of either loading enzyme activated porous microbeads into chip based flowthrough microstructures [2], or merely utilised enzymes coupled to the silicon dioxide surface available in an anisotropically etched flowchannel [3]. In order to improve these mIMER designs, our group developed a silicon microstructured parallel channel reactor [4]. The central feature was the increased surface area as a result of vertically etched parallel flow-channels, i.e. a high aspect ratio channel structure. Since this IMER was made of a non-porous material, the catalytic effect was directly dependent of the linewidth in the microfabrication process. Narrower channels and thinner walls increase the effective area to volume ratio as well as it shortens the diffusion length for the reacting species to the channel wall. However, to achieve a surface to volume ratio in the same order as a porous material proved impossible with standard lithography methods and conventional anisotropic etching. To further improve the efficiency of the high aspect ratio channel microreactor, porous silicon (PS), was introduced as a strategy to enhance the available surface area in the IMER [5].
PS has thereafter been investigated as a carrier matrix for immobilised enzymes in chemical microsystems, and was shown to increase the catalytic performance up to 350 times when compared with a non-porous surface [6]. Fig. 1 shows a cleaved mIMER with the inlet basin and the parallel channel structure. It has been shown for nonstructured (planar) PS samples that the most favourable porous depth is around 20 mm [7]. A deeper porous matrix does not increase the turnover rate since the deep pores are not fully utilised due to diffusion transport limitations. For high aspect ratio structured silicon however, it is difficult to achieve a homogeneous porous depth with pores as deep as 20 mm, since the porous etch process is governed by the arrival of charge carriers, holes, to the silicon/hydrofluoric acid electrolyte interface. The migration of charge carriers, in the bulk silicon is mainly diffusion driven and in the vicinity of the concave structure of the deep microreactor flow channels the charge carrier transport is controlled by the electrostatic field imposed on the silicon structure during the anodisation process. Thus, the porous layer gets deeper in concave corners of the structure. For the longer
Fig. 1. Overview of a cleaved microreactor with the inlet basin and parallel channels.
M. Bengtsson et al. / Talanta 56 (2002) 341–353
343
layer will get very shallow in these regions, Fig. 2. In order to get a more homogeneous porous depth along the height of the channel walls a more recent study was made where the dependency of the different wall dimensions was investigated [8]. It was shown that an increased wall thickness gave a more homogeneous porous layer over the full wall height. Thus, an increased catalytic performance was demonstrated as compared with reactors with thinner walls, yet having the same initial geometric reactor area prior to anodisation. This paper demonstrates homogeneous pore formation and investigates the optimal porous depth for high aspect ratio channel mIMERs having wall/channel geometries of 75/25 and 50/25 mm, the two best candidates from our previous study [8]. The catalytic performance of these mIMER geometries are also investigated with respect to the pore morphology obtained at different anodisation times and current densities. The reactor performance was evaluated for two enzyme categories using (a) a fast diffusing low molecular metabolite, glucose and immobilised glucose oxidase (GOx) and (b) large slow diffusing protein, b-casein and immobilised trypsin. 2. Materials and methods
2.1. Anisotropic etching of the channel structure
Fig. 2. Cross-section of porous reactor wall. The arrow indicates the cut off region for whole migration to the higher regions of the channel wall. The porous layer at the bottom has almost cut off the wall for the migration of charge carriers.
etch times which are needed to achieve a 20 mm deep porous layer, the standing channel walls of the reactor structure may get cut off from the bulk silicon by the deeper pores formed at the channel base. Therefore, the pore formation in the wall structures will not continue and the porous
All reactors were fabricated in p-type (110)-silicon with a resistivity of 20–70 V cm. To achieve a good electrical contact to the electrolyte in the anodisation process, the wafer backside was boron doped, 1 h boron deposition at 1000 °C followed by 4 h diffusion at 1050 °C. The channel structure was etched in KOH (30 wt.%, 80 °C) to a channel depth of 200 mm with an etchmark of SiO2 patterned with standard lithography methods. The reactor structure comprised 10 mm long channels with an inlet/outlet basin on each side resulting in a total length of 13 mm. All channels were fabricated with an equal width of 25 mm. The walls between the channels were either 75 mm wide, giving 32 channels per reactor, or 50 mm wide, giving 42 channels per reactor, resulting in a total width of the structure of 3.1 mm.
344
M. Bengtsson et al. / Talanta 56 (2002) 341–353
2.2. Anodisation, fabrication of the porous layer The reactors were anodised in an in-house designed cell [9]. In a solution of HF (40%) and ethanol (96%) mixed in a 1:1 ratio, the silicon chip was anodised at a constant current during backside illumination. Two different current densities, 10 and 50 mA cm − 2, were applied for four different anodisation times, 5, 10, 20 and 50 min. After anodisation the reactors were thoroughly rinsed with H2O several times and stored in H2O without drying until immobilisation of the enzyme.
2.3. Enzyme immobilisation The enzyme was immobilised in a three-step process [1]. First all reactors were silanised with 1 ml APTES (3-aminopropyltriethoxysilane) in 9 ml of water; pH was adjusted to 3.5 with 6 M HCl. Silanisation was performed under stirring in a water bath at 75 °C for 2 h, followed by washing in water for 1 h. Secondly the reactors were treated with a 2.5% glutaraldehyde (GA) solution in 0.1 M phosphate buffer saline (PBS). GA activation was performed in room temperature for 2 h, followed by washing with PBS for over 1 h. Finally, the enzymes, GOx for the glucose measurements and trypsin for the protein digestion, were coupled onto the PS matrix. The microreactors were immersed in a solution containing either 10 mg GOx in 2 ml PBS, or 3 mg trypsin per ml PBS. The coupling was performed for 12 h in room temperature. Afterwards the reactors were rinsed and stored in PBS at 8 °C until the measurements. As a reference a non-porous reactor was simultaneously immobilised with each enzyme.
2.4. Enzyme acti6ity monitoring— GOx The colorimetric assay used to measure the catalytic turnover for the GOx reactors have been described elsewhere [10]. Briefly, Trinder reagent, with glucose concentrations varying from 0.5 to 50 mM, was pumped either past the reactor to achieve a baseline or through it. The Trinder reagent yields a colour shift proportional to the H2O2 produced in the enzyme catalysed reaction
and the absorbance shift at 492 nm was measured with an HPLC absorbance detector (Waters 486 Tuneable Absorbance Detector, Millipore Corp., MA, USA). To determine the highest turnover, Ve, for each concentration, So, the flow was gradually increased until no further increase in turnover was detectable (data not shown).
2.5. Enzyme acti6ity monitoring— trypsin Enzyme activity measurements were also made to access which reactor geometry and porosity that provided optimal characteristics when immobilised with trypsin. This was done by injection of a 1 mM Na-p-tosyl-L-arginine methyl ester (TAME) solution in 50 mM Tris, 10 mM CaCl2, pH 8.2. The enzyme activity, i.e. the TAME turnover rate, for the trypsin mIMER’s was then calculated by applying the Lambert–Beer law to the observed absorbance readings at 247 nm. For each reactor design the TAME turnover rates obtained was plotted against flow (data not shown) until a further increase in flow did not provide any increase in turnover.
2.6. Mass spectrometry analysis Previously, it has been shown that these m-reactors immobilised with trypsin can successfully be used to digest proteins [11], generating a peptide map that can be used to identify the protein. Peptide mapping with MALDI-TOF MS is a frequently used technique for identification of proteins and studies of novel post-translational modification [12–16]. Briefly, the protein is digested with a proteolytic enzyme, and the resulting digest mixture is analysed with MALDI-TOF MS, generating a peptide map that is unique for the digested protein. The observed peptide masses are then used to search existing protein and DNA databases. In order to purify and concentrate the samples prior to MALDI-TOF MS a microscale sample purification technique [17–19] was utilised. GELoader (Eppendorf) tips were packed with (0.5–1 ml volume) POROS 20 R2 (Perseptive Biosystems Inc., Framingham, MA, USA). These microcolumns were equilibrated with 0.1% trifl-
M. Bengtsson et al. / Talanta 56 (2002) 341–353
uoroacetic acid (TFA), and the acidified sample was passed through and followed by a washing step using water (Millipore). Finally the bound analytes were eluted with matrix solution directly onto the MALDI-target plate. The matrix solution was prepared by dissolving 12 mg ml − 1 a-cyano-hydroxycinnamic acid (CHCA) in 2:1 acetonitrile (ACN)/0.3% TFA. The matrix-assisted laser desorption/ionisation time of flight mass spectrometry (MALDI-TOF MS) instrument used was a Voyager DE-PRO (Perseptive Biosystems Inc., Framingham, MA, USA). The instrument was equipped with a linear flight tube of 1.1 m, delayed extraction ion source, and used a nitrogen laser (u= 337 nm) with a laser focal spot diameter of approximately 100 mm. For high accuracy peptide mapping reflector mode was used. In order to evaluate the impact of the reactor design upon the protein digestion efficiency, the reactor that provided the highest turnover of TAME during the absorbance measurements (75/ 25 50 mA cm − 2 10 min) and one that exhibited a very low TAME turnover (50/25 10 mA cm − 2 5 min) were used to digest a protein sample in an on-line format. A sample solution containing 200 nM b-casein in 1 M urea, 100 mM NH4HCO3 was passed through the reactors at flows varying from 50 to 1 ml min − 1 and the eluate was collected in micro tubes (Eppendorf AG, Germany) and acidified by addition of TFA. During the protein digestion the IMER was mounted in a holder that allowed it to be heated to 50 °C. The elevated temperature served to unfold the protein, thereby aiding the enzymatic digestion.
3. Chemicals 3-Aminopropyltriethoxysilane, GA (grade II), GOx EC 1.1.3.4, type X-S, Aspergillus niger, trypsin type IX, porcine pancreas, (lot, 25H0359), b-casein (lot, 25H9550), TAME (lot, 68H0898) Trizma® Base (lot, 12H56001), Trizma® HCl (lot, 35H5705), a-CHCA (lot, 51229) were all purchased from Sigma Chemical Co. (St. Louis, MO, USA), NH4HCO3, CaCl2, KH2PO4, Na2HPO4,
345
NaN3, ACN from Merck (Darmstadt, Germany). Urea (lot, 51459) and TFA were obtained from Fluka Chemicals (Buchswald, Switzerland). All buffers used were freshly prepared and pH adjusted.
4. Results
4.1. Porous depth To estimate the porous depth in the reactor one reactor of each was cut in half, sputtered with a 15 nm gold layer and studied in a scanning electron microscope (SEM), Fig. 3. The pore depth was measured in the bottom of the reactor channels, at 50 mm height on the wall and at the top of the wall, Table 1.
4.2. Enzyme acti6ity measurements To determine the maximum turnover for each reactor, Vmax, and the apparent Michaelis constant, Km app, the theoretical equation of a first order enzymatic system, Ve = Vmax/(1+ Km/So) was fitted to the recorded data, Fig. 4. To derive additional estimates of Vmax and Km app, Eadie – Hofstee fits were also made on the acquired data for each reactor type. A more detailed description of the data treatment process has been described elsewhere [6]. A one-way analysis of variance (ANOVA) showed a 98% similarity between the estimated Vmax values from the Eadie–Hofstee fit and the theoretical equation fit. The derived maximum turnover, Vmax for the GOx reactors is plotted versus anodisation time in Fig. 5, and the apparent Michaelis constant, Km app, is listed in Table 2. The maximum turnover for TAME in the trypsin immobilised reactors as a function of anodisation time are plotted in Fig. 6.
4.3. Mass spectrometry analysis On-line microdigestion of b-casein was performed using two different mIMER designs (75/25 mm, 50 mA cm − 2, 10 min, and 50/25 mm, 10 mA cm − 2, 5 min). In order to compare the perfor-
346
M. Bengtsson et al. / Talanta 56 (2002) 341–353
mance of these reactors, 10 ml sample solution from each mIMER that had been digested at a flow of 50 ml min − 1 was subjected to microscale sample purification and eluted on the MALDI target with 2 ml CHCA matrix.
The mass spectra resulting from the MALDITOF MS analysis is presented in Fig. 7. As can be seen it is clear that the mIMER 75/25 mm, 50 mA cm − 2 10 min provided the most effective enzymatic digestion of the protein, which is in accor-
Fig. 3. Cross section of (a) 50/25 reactor anodised at 50 mA cm − 2 for 5 min, (b) 75/25 reactor anodised at 10 mA cm − 2 for 50 min, (c) 75/25 reactor anodised at 50 mA cm − 2 for 10 min. Arrows indicate the measured porous depth at the channel bottom (h), 50 mm up the wall (i) and at the top of the wall (k).
M. Bengtsson et al. / Talanta 56 (2002) 341–353
347
Fig. 3. (Continued)
dance with the reactor performance data in Fig. 6. After database search using PeptiIdent (available on http://www.expasy.ch/tools/peptident.html) ten peaks could be assigned to b-casein, Fig. 7b, whereas the corresponding protein digestion on the 50/25 mm, 10 mA cm − 2, 5 min mIMER provided only two peaks, Fig. 7a, which was insufficient to identify the protein.
5. Discussion The porous layer in the reactors with 50 mm channel wall showed a non-homogeneous profile, with deep pores in the lower region cutting off the charge carriers from migrating to the upper region of the wall, which lead to a very shallow porous layer, less than 5 mm, in these parts. It was also
348
M. Bengtsson et al. / Talanta 56 (2002) 341–353
found that the porous depth at the top of the walls did not increase with etch time. The reactors with thicker walls (75 mm), however, showed a porous profile with equal porous depth over the full height of the wall, even for anodisation times
as long as 50 min and pore depths as deep as half the wall thickness, 35 mm, Table 1. The highest turnover rate, both for GOx and trypsin, is shown for the 75 mm wall reactor anodised at 50 mA cm − 2 for 10 min (Figs. 5 and
Fig. 3. (Continued)
M. Bengtsson et al. / Talanta 56 (2002) 341–353
349
Table 1 Porous depth of the different reactors measured from SEM photographs Wall/channel (mm)
Current density (mA cm−2)
Etchtime (min)
75/25
10
5 10 20 50 5 10 20 5 10 20 50 5 10 20
50
50/25
10
50
Bottom (mm)
Fifty micrometer up (mm)
Top (mm)
13
18
20
57 14 52 62
30 6 35* 35*
29 6 35* 35*
7 19 50 11 38
4 6 8 5 8 **
**
2 4 3 4 3 **
*, Indicates that the wall is completely porous. **, The walls got unstable and fell, due to the deep pores at the base, and no value could be measured.
6). In spite of the differences in enzyme and molecule size of the substrate the combined pore size and porous depth gives an advantageous porous matrix in both cases. For GOx the difference is merely about 10% increase compared with the second best morphology, 50 mA cm − 2, 20 min. For trypsin the increase is more than 200%. Indicating that the trypsin microreactor performance is more sensitive to differences in pore morphology than the glucose measurements.
5.2. 75 /25 vm Reactors, 10 mA cm − 2 The reactors anodised with a current density of 10 mA cm − 2 show a very homogeneous porous layer along the full height of the channel wall with
5.1. 75 /25 vm Reactor, 50 mA cm − 2 For the 75 mm wall reactors it is noticeable that for 50 mA cm − 2 anodisation current, the samples anodised for 10 min and longer are porous throughout the wall thickness. The shorter anodisation time still provides a higher catalytic efficiency for the reactor, indicating that only a certain porous depth contributes to the effective surface enlargement for the immobilisation. The deeper pores are not fully utilised due to diffusion transport limitation. After the most efficient depth is reached, further anodisation depth will leave the catalytic effect constant or even decreased due to loss of surface area at the wall surface as a result of continued etching.
Fig. 4. Measured catalytic turnover for reactors with four different geometry/current density combinations and a nonporous reference and the least square fits to a first order enzyme kinetic system.
M. Bengtsson et al. / Talanta 56 (2002) 341–353
350
Fig. 5. Vmax as a function of anodisation time for the different reactor types and current densities. , Indicates which PS-reactor cross section that are shown in the SEM-photos, Fig. 3. , Indicates which PS-reactor corresponds to the graphs in Fig. 4.
an increased depth for longer anodisation times. However, though the reactors anodised for the longest time have a porous layer almost as deep as the one anodised at 50 mA cm − 2 for 10 min, the catalytic effect is still lower, Figs. 5 and 6, both for GOx and trypsin. This indicates that the pore size obtained at a current density of 10 mA cm − 2 does not provide a pore morphology that gives the same effective surface for enzyme immobilisation and substrate catalysis as the ones obtained at a higher current density. Still, for the 10 mA cm − 2 trypsin immobilised reactors, the turnover increase does not level off at the etch
Fig. 6. Trypsin turnover as a function of anodisation time for the different reactor types and current densities.
times investigated which may imply that the most efficient porous depth is not reached for that pore size and that it may be possible to increase the effect further with even thicker channelwalls.
5.3. 50 /25 vm Reactors Though the 50 mm wall reactors have an initial geometrical area before anodisation that is 30% larger than the 75 mm wall reactors, the catalytic turnover is significant lower than for the reactors with wider walls. Since the porous depth in the thinner walls is non-homogeneous, the topmost part of the channel walls have a catalytic effect much less than the lower part, though the porous depth in the lower part are identical for the
Table 2 Km app from the GOx activity measurements Etch time (min)
0 5 10 20 50
50/25 mm
75/25 mm
10 mA cm−2 (mM)
50 mA cm−2 (mM)
10 mA cm−2 (mM)
50 mA cm−2 (mM)
13.1 22.6 17.7
22.7 11.4 8
11 4.6 5.2 3.8 6.7
2.8 3.4 3.3
M. Bengtsson et al. / Talanta 56 (2002) 341–353
351
Fig. 7. MALDI-TOF MS spectra resulting from digestion of 200 fmol ml − 1 b-casein. (a) On a trypsin mIMER, 50/25, 10 mA, 5 min only two peaks could be observed which was insufficient to correctly identify the protein. (b) On a trypsin mIMER, 75/25, 50 mA, 10 min providing enough peaks to correctly identify the protein as b-casein (Swissprot accession cP05814) k, keratin contamination; *, unidentified peaks.
352
M. Bengtsson et al. / Talanta 56 (2002) 341–353
different reactor geometries. The exception to this is the reactors anodised at 10 mA cm − 2 immobilised with trypsin, Fig. 6. For etch times of 10 and 20 min, the 50 mm wall reactors have a catalytic turnover about 80% higher than the 75 mm wall reactors anodised for the same times. This difference is partly explained by the initial difference in area for the non-porous reactors. For this low current density the shorter etch time gives a shallow porous layer all over the reactor structure, therefore the porous matrix is fully utilised and the inhomogeneity does not have an influence on the catalytic performance. For the longest anodisation times, 50 min, the porous depth in the bottom has grown deeper than the efficient depth and the difference in initial area is compensated by the more homogeneous porous layer in the 75 mm wall reactor.
5.4. Protein digestion 75 /25 vm 10 min; 50 /25 vm 5 min The presented mass spectra (Fig. 7) were obtained at a high flow-rate, 50 ml min − 1 which corresponds to a digestion time of only 12 s. Previously digestion times of 1 min (6 ml min − 1) have been found to provide sufficient digestion to obtain identity of a number of model proteins [11]. The ability to direct the degree of digestion of the protein by varying the flow provides a powerful mean to adapt the IMER to different conditions, since more difficult samples may require longer digestion times. The mass spectra in Fig. 7b also show two phosphopeptides that are characteristic for b-casein. Phosphorylation of peptides can be confirmed either by an enzymatic dephosphorylation step [20,21] or by MS/MS analysis [22]. One potential use for these trypsin mIMERs could be digestion of proteins separated by multidimensional liquid chromatography, 2DHPLC, where the ability to obtain identity of separated proteins is of great interest.
6. Conclusion The importance of a correctly designed mIMER to fully utilise the benefit of integrating PS in
chemical m-systems as a surface enlarging carrier matrix for enzymes has clearly been demonstrated. By tailoring the mIMER channel cross-section geometry, homogeneous PS layers were obtained on high aspect ratio silicon microstructures, e.g. the mIMER, yielding improved biocatalytic performance. In the case of the protein digestion application the outcome of the optimisation work is crucial, enabling high-speed on-line protein digestion. Further work will focus on the microfluidic optimisation of the microreactors and the full applicability of the proteolytic mIMER to the proteomics field.
References [1] H.H. Weetall, in: K. Mosbach (Ed.), Methods in Enzymology, vol. 44, Academic Press, New York, 1976, p. 134. [2] B. Xie, B. Danielsson, P. Norberg, F. Winquist, I. Lundstrom, Sens. Actuat. B-Chem. 6 (1992) 127. [3] M. Suda, T. Sakuhara, Y. Murakami, I. Karube, Appl. Biochem. Biotechnol. 41 (1993) 11. [4] T. Laurell, L. Rosengren, Sens. Actuat. B-Chem. 19 (1994) 614. [5] T. Laurell, J. Drott, L. Rosengren, K. Lindstrom, Sens. Actuat. B-Chem. 31 (1996) 161. [6] J. Drott, L. Rosengren, K. Lindstrom, T. Laurell, Thin Solid Films 330 (1998) 161. [7] J. Drott, L. Rosengren, K. Lindstrom, T. Laurell, Mikrochim. Acta 131 (1999) 115. [8] M. Bengtsson, J. Drott, T. Laurell, Phys. Status Solidi (a) 182 (2000) 533. [9] J. Drott, K. Lindstrom, L. Rosengren, T. Laurell, J. Micromech. Microeng. 7 (1997) 14. [10] T. Laurell, J. Drott, L. Rosengren, Biosens. Bioelectron. 10 (1995) 289. [11] S. Ekstro¨ m, P. O8 nnerfjord, J. Nilsson, M. Bengtsson, T. Laurell, G. Marko-Varga, Anal. Chem. 72 (2000) 286. [12] W.J. Hezel, T.M. Billeci, J.T. Stults, S.C. Wong, C. Grimely, C. Watanabe, Proc. Natl. Acad. Sci. USA 90 (1993) 5011. [13] P. James, M. Quadroni, E. Carafoli, G. Gonnet, Prot. Sci. 3 (1994) 1347. [14] D.J.C. Pappin, P. Hojrup, A.J. Bleasby, Curr. Biol. 3 (1993) 327. [15] M. Mann, P. Hojrup, P. Roepstroff, Biol. Mass Spectrom. 22 (1993) 338. [16] J. Yates, S. Speicher, P.R. Griffin, T. Hunkapiller, Anal. Biochem. 214 (1993) 397. [17] R.S. Annan, D.E. Mculty, S.A. Carr, Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, 1996, p.702.
M. Bengtsson et al. / Talanta 56 (2002) 341–353 [18] J. Gobom, E. Nordhoff, E. Mirgorodsakaya, R. Ekman, P. Roepstorff, J. Mass Spectrom. 34 (1999) 105. [19] M. Kussmann, E. Nordhoff, H. Rahbek-Nielsen, S. Haebel, M. Rossel-Larsen, L. Jakobsen, J. Gobom, E. Mirgorodskaya, A. Kroll-Kristensen, L. Palm, P. Roepstorff, J. Mass Spectrom. 32 (1997) 593.
353
[20] A. Srensballe, S. Andersen, O.N. Jensen, Proteomics 1 (2001) 207. [21] M.R. Larsen, G.L. Sorensen, S.J. Fey, P.M. Larsen, P. Roepstrorff, Proteomics 1 (2001) 223. [22] M.J. Huddleston, R.S. Annan, M.T. Bean, S.A. Carr, J. Am. Soc. Mass Spectrom. 4 (1993) 710.
Improved performance in silicon enzyme microreactors obtained by homogeneous porous silicon carrier matrix Martin Bengtsson a, Simon Ekstro¨m a, Gyo¨rgy Marko-Varga b, Thomas Laurell a,* a
Department of Electrical Measurements, Lund Institute of Technology, Uni6ersity of Lund, P.O. Box 118, 22100 Lund, Sweden b Department of Analytical Chemistry, Uni6ersity of Lund, P.O. Box 124, 22100 Lund, Sweden Received 17 April 2001; received in revised form 9 July 2001; accepted 10 July 2001
Abstract The catalytic performance of porous silicon (PS) micro enzyme reactors (mIMER) is strongly dependent on the PS matrix morphology for enzyme immobilisation. PS was achieved in the mIMER by anodisation in a HF– ethanol mixture. PS etching of structured silicon surfaces commonly results in an inhomogeneous pore formation. The deep channel microreactors described herein have previously suffered from these phenomena, yielding non-optimised mIMERs. In order to obtain a homogeneous PS layer on the deep microreactor channel walls, different reactor geometries (channel wall thicknesses of 50 and 75 mm) were anodised at 10 and 50 mA cm − 2 for anodisation times ranging between 0 and 50 min. The mIMERs were evaluated by immobilising two types of enzymes, glucose oxidase (GOx) and trypsin, and the resulting catalytic turnover was monitored by a colorimetric assay. It was found that reactors with a homogeneous PS matrix displayed improved performance. The trypsin mIMERs were used to digest a protein, b-casein, in an on-line format and the digest was analysed by MALDI-TOF MS. The importance of tailoring the mIMER geometry and the PS-matrix is crucial for the protein digestion. Successful protein identification after only 12 s. digestion was demonstrated for the best reactor, 75 mm channel wall, 25 mm channel width, anodised at 50 mA cm − 2 for 10 min. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Porous silicon; IMER; Enzyme; Carrier matrix
1. Introduction In systems for biochemical process monitoring, immobilised enzyme reactors (IMER) are commonly used tools for biospecific detection. In such reactors one of the most crucial factors is the * Corresponding author. E-mail address: [email protected] (T. Laurell).
effective immobilised area compared with the volume of the reactor. To increase the effective area in the reactor, porous material such as packed porous glass beads, and silica gel [1], is commonly used. In the current trend of miniaturising bioanalytical technology enzyme reactors has also been the focus for microsystem integration. The development of chip integrated microIMERs (mIMER) have fol-
0039-9140/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 9 1 4 0 ( 0 1 ) 0 0 6 0 0 - 2
342
M. Bengtsson et al. / Talanta 56 (2002) 341–353
lowed strategies of either loading enzyme activated porous microbeads into chip based flowthrough microstructures [2], or merely utilised enzymes coupled to the silicon dioxide surface available in an anisotropically etched flowchannel [3]. In order to improve these mIMER designs, our group developed a silicon microstructured parallel channel reactor [4]. The central feature was the increased surface area as a result of vertically etched parallel flow-channels, i.e. a high aspect ratio channel structure. Since this IMER was made of a non-porous material, the catalytic effect was directly dependent of the linewidth in the microfabrication process. Narrower channels and thinner walls increase the effective area to volume ratio as well as it shortens the diffusion length for the reacting species to the channel wall. However, to achieve a surface to volume ratio in the same order as a porous material proved impossible with standard lithography methods and conventional anisotropic etching. To further improve the efficiency of the high aspect ratio channel microreactor, porous silicon (PS), was introduced as a strategy to enhance the available surface area in the IMER [5].
PS has thereafter been investigated as a carrier matrix for immobilised enzymes in chemical microsystems, and was shown to increase the catalytic performance up to 350 times when compared with a non-porous surface [6]. Fig. 1 shows a cleaved mIMER with the inlet basin and the parallel channel structure. It has been shown for nonstructured (planar) PS samples that the most favourable porous depth is around 20 mm [7]. A deeper porous matrix does not increase the turnover rate since the deep pores are not fully utilised due to diffusion transport limitations. For high aspect ratio structured silicon however, it is difficult to achieve a homogeneous porous depth with pores as deep as 20 mm, since the porous etch process is governed by the arrival of charge carriers, holes, to the silicon/hydrofluoric acid electrolyte interface. The migration of charge carriers, in the bulk silicon is mainly diffusion driven and in the vicinity of the concave structure of the deep microreactor flow channels the charge carrier transport is controlled by the electrostatic field imposed on the silicon structure during the anodisation process. Thus, the porous layer gets deeper in concave corners of the structure. For the longer
Fig. 1. Overview of a cleaved microreactor with the inlet basin and parallel channels.
M. Bengtsson et al. / Talanta 56 (2002) 341–353
343
layer will get very shallow in these regions, Fig. 2. In order to get a more homogeneous porous depth along the height of the channel walls a more recent study was made where the dependency of the different wall dimensions was investigated [8]. It was shown that an increased wall thickness gave a more homogeneous porous layer over the full wall height. Thus, an increased catalytic performance was demonstrated as compared with reactors with thinner walls, yet having the same initial geometric reactor area prior to anodisation. This paper demonstrates homogeneous pore formation and investigates the optimal porous depth for high aspect ratio channel mIMERs having wall/channel geometries of 75/25 and 50/25 mm, the two best candidates from our previous study [8]. The catalytic performance of these mIMER geometries are also investigated with respect to the pore morphology obtained at different anodisation times and current densities. The reactor performance was evaluated for two enzyme categories using (a) a fast diffusing low molecular metabolite, glucose and immobilised glucose oxidase (GOx) and (b) large slow diffusing protein, b-casein and immobilised trypsin. 2. Materials and methods
2.1. Anisotropic etching of the channel structure
Fig. 2. Cross-section of porous reactor wall. The arrow indicates the cut off region for whole migration to the higher regions of the channel wall. The porous layer at the bottom has almost cut off the wall for the migration of charge carriers.
etch times which are needed to achieve a 20 mm deep porous layer, the standing channel walls of the reactor structure may get cut off from the bulk silicon by the deeper pores formed at the channel base. Therefore, the pore formation in the wall structures will not continue and the porous
All reactors were fabricated in p-type (110)-silicon with a resistivity of 20–70 V cm. To achieve a good electrical contact to the electrolyte in the anodisation process, the wafer backside was boron doped, 1 h boron deposition at 1000 °C followed by 4 h diffusion at 1050 °C. The channel structure was etched in KOH (30 wt.%, 80 °C) to a channel depth of 200 mm with an etchmark of SiO2 patterned with standard lithography methods. The reactor structure comprised 10 mm long channels with an inlet/outlet basin on each side resulting in a total length of 13 mm. All channels were fabricated with an equal width of 25 mm. The walls between the channels were either 75 mm wide, giving 32 channels per reactor, or 50 mm wide, giving 42 channels per reactor, resulting in a total width of the structure of 3.1 mm.
344
M. Bengtsson et al. / Talanta 56 (2002) 341–353
2.2. Anodisation, fabrication of the porous layer The reactors were anodised in an in-house designed cell [9]. In a solution of HF (40%) and ethanol (96%) mixed in a 1:1 ratio, the silicon chip was anodised at a constant current during backside illumination. Two different current densities, 10 and 50 mA cm − 2, were applied for four different anodisation times, 5, 10, 20 and 50 min. After anodisation the reactors were thoroughly rinsed with H2O several times and stored in H2O without drying until immobilisation of the enzyme.
2.3. Enzyme immobilisation The enzyme was immobilised in a three-step process [1]. First all reactors were silanised with 1 ml APTES (3-aminopropyltriethoxysilane) in 9 ml of water; pH was adjusted to 3.5 with 6 M HCl. Silanisation was performed under stirring in a water bath at 75 °C for 2 h, followed by washing in water for 1 h. Secondly the reactors were treated with a 2.5% glutaraldehyde (GA) solution in 0.1 M phosphate buffer saline (PBS). GA activation was performed in room temperature for 2 h, followed by washing with PBS for over 1 h. Finally, the enzymes, GOx for the glucose measurements and trypsin for the protein digestion, were coupled onto the PS matrix. The microreactors were immersed in a solution containing either 10 mg GOx in 2 ml PBS, or 3 mg trypsin per ml PBS. The coupling was performed for 12 h in room temperature. Afterwards the reactors were rinsed and stored in PBS at 8 °C until the measurements. As a reference a non-porous reactor was simultaneously immobilised with each enzyme.
2.4. Enzyme acti6ity monitoring— GOx The colorimetric assay used to measure the catalytic turnover for the GOx reactors have been described elsewhere [10]. Briefly, Trinder reagent, with glucose concentrations varying from 0.5 to 50 mM, was pumped either past the reactor to achieve a baseline or through it. The Trinder reagent yields a colour shift proportional to the H2O2 produced in the enzyme catalysed reaction
and the absorbance shift at 492 nm was measured with an HPLC absorbance detector (Waters 486 Tuneable Absorbance Detector, Millipore Corp., MA, USA). To determine the highest turnover, Ve, for each concentration, So, the flow was gradually increased until no further increase in turnover was detectable (data not shown).
2.5. Enzyme acti6ity monitoring— trypsin Enzyme activity measurements were also made to access which reactor geometry and porosity that provided optimal characteristics when immobilised with trypsin. This was done by injection of a 1 mM Na-p-tosyl-L-arginine methyl ester (TAME) solution in 50 mM Tris, 10 mM CaCl2, pH 8.2. The enzyme activity, i.e. the TAME turnover rate, for the trypsin mIMER’s was then calculated by applying the Lambert–Beer law to the observed absorbance readings at 247 nm. For each reactor design the TAME turnover rates obtained was plotted against flow (data not shown) until a further increase in flow did not provide any increase in turnover.
2.6. Mass spectrometry analysis Previously, it has been shown that these m-reactors immobilised with trypsin can successfully be used to digest proteins [11], generating a peptide map that can be used to identify the protein. Peptide mapping with MALDI-TOF MS is a frequently used technique for identification of proteins and studies of novel post-translational modification [12–16]. Briefly, the protein is digested with a proteolytic enzyme, and the resulting digest mixture is analysed with MALDI-TOF MS, generating a peptide map that is unique for the digested protein. The observed peptide masses are then used to search existing protein and DNA databases. In order to purify and concentrate the samples prior to MALDI-TOF MS a microscale sample purification technique [17–19] was utilised. GELoader (Eppendorf) tips were packed with (0.5–1 ml volume) POROS 20 R2 (Perseptive Biosystems Inc., Framingham, MA, USA). These microcolumns were equilibrated with 0.1% trifl-
M. Bengtsson et al. / Talanta 56 (2002) 341–353
uoroacetic acid (TFA), and the acidified sample was passed through and followed by a washing step using water (Millipore). Finally the bound analytes were eluted with matrix solution directly onto the MALDI-target plate. The matrix solution was prepared by dissolving 12 mg ml − 1 a-cyano-hydroxycinnamic acid (CHCA) in 2:1 acetonitrile (ACN)/0.3% TFA. The matrix-assisted laser desorption/ionisation time of flight mass spectrometry (MALDI-TOF MS) instrument used was a Voyager DE-PRO (Perseptive Biosystems Inc., Framingham, MA, USA). The instrument was equipped with a linear flight tube of 1.1 m, delayed extraction ion source, and used a nitrogen laser (u= 337 nm) with a laser focal spot diameter of approximately 100 mm. For high accuracy peptide mapping reflector mode was used. In order to evaluate the impact of the reactor design upon the protein digestion efficiency, the reactor that provided the highest turnover of TAME during the absorbance measurements (75/ 25 50 mA cm − 2 10 min) and one that exhibited a very low TAME turnover (50/25 10 mA cm − 2 5 min) were used to digest a protein sample in an on-line format. A sample solution containing 200 nM b-casein in 1 M urea, 100 mM NH4HCO3 was passed through the reactors at flows varying from 50 to 1 ml min − 1 and the eluate was collected in micro tubes (Eppendorf AG, Germany) and acidified by addition of TFA. During the protein digestion the IMER was mounted in a holder that allowed it to be heated to 50 °C. The elevated temperature served to unfold the protein, thereby aiding the enzymatic digestion.
3. Chemicals 3-Aminopropyltriethoxysilane, GA (grade II), GOx EC 1.1.3.4, type X-S, Aspergillus niger, trypsin type IX, porcine pancreas, (lot, 25H0359), b-casein (lot, 25H9550), TAME (lot, 68H0898) Trizma® Base (lot, 12H56001), Trizma® HCl (lot, 35H5705), a-CHCA (lot, 51229) were all purchased from Sigma Chemical Co. (St. Louis, MO, USA), NH4HCO3, CaCl2, KH2PO4, Na2HPO4,
345
NaN3, ACN from Merck (Darmstadt, Germany). Urea (lot, 51459) and TFA were obtained from Fluka Chemicals (Buchswald, Switzerland). All buffers used were freshly prepared and pH adjusted.
4. Results
4.1. Porous depth To estimate the porous depth in the reactor one reactor of each was cut in half, sputtered with a 15 nm gold layer and studied in a scanning electron microscope (SEM), Fig. 3. The pore depth was measured in the bottom of the reactor channels, at 50 mm height on the wall and at the top of the wall, Table 1.
4.2. Enzyme acti6ity measurements To determine the maximum turnover for each reactor, Vmax, and the apparent Michaelis constant, Km app, the theoretical equation of a first order enzymatic system, Ve = Vmax/(1+ Km/So) was fitted to the recorded data, Fig. 4. To derive additional estimates of Vmax and Km app, Eadie – Hofstee fits were also made on the acquired data for each reactor type. A more detailed description of the data treatment process has been described elsewhere [6]. A one-way analysis of variance (ANOVA) showed a 98% similarity between the estimated Vmax values from the Eadie–Hofstee fit and the theoretical equation fit. The derived maximum turnover, Vmax for the GOx reactors is plotted versus anodisation time in Fig. 5, and the apparent Michaelis constant, Km app, is listed in Table 2. The maximum turnover for TAME in the trypsin immobilised reactors as a function of anodisation time are plotted in Fig. 6.
4.3. Mass spectrometry analysis On-line microdigestion of b-casein was performed using two different mIMER designs (75/25 mm, 50 mA cm − 2, 10 min, and 50/25 mm, 10 mA cm − 2, 5 min). In order to compare the perfor-
346
M. Bengtsson et al. / Talanta 56 (2002) 341–353
mance of these reactors, 10 ml sample solution from each mIMER that had been digested at a flow of 50 ml min − 1 was subjected to microscale sample purification and eluted on the MALDI target with 2 ml CHCA matrix.
The mass spectra resulting from the MALDITOF MS analysis is presented in Fig. 7. As can be seen it is clear that the mIMER 75/25 mm, 50 mA cm − 2 10 min provided the most effective enzymatic digestion of the protein, which is in accor-
Fig. 3. Cross section of (a) 50/25 reactor anodised at 50 mA cm − 2 for 5 min, (b) 75/25 reactor anodised at 10 mA cm − 2 for 50 min, (c) 75/25 reactor anodised at 50 mA cm − 2 for 10 min. Arrows indicate the measured porous depth at the channel bottom (h), 50 mm up the wall (i) and at the top of the wall (k).
M. Bengtsson et al. / Talanta 56 (2002) 341–353
347
Fig. 3. (Continued)
dance with the reactor performance data in Fig. 6. After database search using PeptiIdent (available on http://www.expasy.ch/tools/peptident.html) ten peaks could be assigned to b-casein, Fig. 7b, whereas the corresponding protein digestion on the 50/25 mm, 10 mA cm − 2, 5 min mIMER provided only two peaks, Fig. 7a, which was insufficient to identify the protein.
5. Discussion The porous layer in the reactors with 50 mm channel wall showed a non-homogeneous profile, with deep pores in the lower region cutting off the charge carriers from migrating to the upper region of the wall, which lead to a very shallow porous layer, less than 5 mm, in these parts. It was also
348
M. Bengtsson et al. / Talanta 56 (2002) 341–353
found that the porous depth at the top of the walls did not increase with etch time. The reactors with thicker walls (75 mm), however, showed a porous profile with equal porous depth over the full height of the wall, even for anodisation times
as long as 50 min and pore depths as deep as half the wall thickness, 35 mm, Table 1. The highest turnover rate, both for GOx and trypsin, is shown for the 75 mm wall reactor anodised at 50 mA cm − 2 for 10 min (Figs. 5 and
Fig. 3. (Continued)
M. Bengtsson et al. / Talanta 56 (2002) 341–353
349
Table 1 Porous depth of the different reactors measured from SEM photographs Wall/channel (mm)
Current density (mA cm−2)
Etchtime (min)
75/25
10
5 10 20 50 5 10 20 5 10 20 50 5 10 20
50
50/25
10
50
Bottom (mm)
Fifty micrometer up (mm)
Top (mm)
13
18
20
57 14 52 62
30 6 35* 35*
29 6 35* 35*
7 19 50 11 38
4 6 8 5 8 **
**
2 4 3 4 3 **
*, Indicates that the wall is completely porous. **, The walls got unstable and fell, due to the deep pores at the base, and no value could be measured.
6). In spite of the differences in enzyme and molecule size of the substrate the combined pore size and porous depth gives an advantageous porous matrix in both cases. For GOx the difference is merely about 10% increase compared with the second best morphology, 50 mA cm − 2, 20 min. For trypsin the increase is more than 200%. Indicating that the trypsin microreactor performance is more sensitive to differences in pore morphology than the glucose measurements.
5.2. 75 /25 vm Reactors, 10 mA cm − 2 The reactors anodised with a current density of 10 mA cm − 2 show a very homogeneous porous layer along the full height of the channel wall with
5.1. 75 /25 vm Reactor, 50 mA cm − 2 For the 75 mm wall reactors it is noticeable that for 50 mA cm − 2 anodisation current, the samples anodised for 10 min and longer are porous throughout the wall thickness. The shorter anodisation time still provides a higher catalytic efficiency for the reactor, indicating that only a certain porous depth contributes to the effective surface enlargement for the immobilisation. The deeper pores are not fully utilised due to diffusion transport limitation. After the most efficient depth is reached, further anodisation depth will leave the catalytic effect constant or even decreased due to loss of surface area at the wall surface as a result of continued etching.
Fig. 4. Measured catalytic turnover for reactors with four different geometry/current density combinations and a nonporous reference and the least square fits to a first order enzyme kinetic system.
M. Bengtsson et al. / Talanta 56 (2002) 341–353
350
Fig. 5. Vmax as a function of anodisation time for the different reactor types and current densities. , Indicates which PS-reactor cross section that are shown in the SEM-photos, Fig. 3. , Indicates which PS-reactor corresponds to the graphs in Fig. 4.
an increased depth for longer anodisation times. However, though the reactors anodised for the longest time have a porous layer almost as deep as the one anodised at 50 mA cm − 2 for 10 min, the catalytic effect is still lower, Figs. 5 and 6, both for GOx and trypsin. This indicates that the pore size obtained at a current density of 10 mA cm − 2 does not provide a pore morphology that gives the same effective surface for enzyme immobilisation and substrate catalysis as the ones obtained at a higher current density. Still, for the 10 mA cm − 2 trypsin immobilised reactors, the turnover increase does not level off at the etch
Fig. 6. Trypsin turnover as a function of anodisation time for the different reactor types and current densities.
times investigated which may imply that the most efficient porous depth is not reached for that pore size and that it may be possible to increase the effect further with even thicker channelwalls.
5.3. 50 /25 vm Reactors Though the 50 mm wall reactors have an initial geometrical area before anodisation that is 30% larger than the 75 mm wall reactors, the catalytic turnover is significant lower than for the reactors with wider walls. Since the porous depth in the thinner walls is non-homogeneous, the topmost part of the channel walls have a catalytic effect much less than the lower part, though the porous depth in the lower part are identical for the
Table 2 Km app from the GOx activity measurements Etch time (min)
0 5 10 20 50
50/25 mm
75/25 mm
10 mA cm−2 (mM)
50 mA cm−2 (mM)
10 mA cm−2 (mM)
50 mA cm−2 (mM)
13.1 22.6 17.7
22.7 11.4 8
11 4.6 5.2 3.8 6.7
2.8 3.4 3.3
M. Bengtsson et al. / Talanta 56 (2002) 341–353
351
Fig. 7. MALDI-TOF MS spectra resulting from digestion of 200 fmol ml − 1 b-casein. (a) On a trypsin mIMER, 50/25, 10 mA, 5 min only two peaks could be observed which was insufficient to correctly identify the protein. (b) On a trypsin mIMER, 75/25, 50 mA, 10 min providing enough peaks to correctly identify the protein as b-casein (Swissprot accession cP05814) k, keratin contamination; *, unidentified peaks.
352
M. Bengtsson et al. / Talanta 56 (2002) 341–353
different reactor geometries. The exception to this is the reactors anodised at 10 mA cm − 2 immobilised with trypsin, Fig. 6. For etch times of 10 and 20 min, the 50 mm wall reactors have a catalytic turnover about 80% higher than the 75 mm wall reactors anodised for the same times. This difference is partly explained by the initial difference in area for the non-porous reactors. For this low current density the shorter etch time gives a shallow porous layer all over the reactor structure, therefore the porous matrix is fully utilised and the inhomogeneity does not have an influence on the catalytic performance. For the longest anodisation times, 50 min, the porous depth in the bottom has grown deeper than the efficient depth and the difference in initial area is compensated by the more homogeneous porous layer in the 75 mm wall reactor.
5.4. Protein digestion 75 /25 vm 10 min; 50 /25 vm 5 min The presented mass spectra (Fig. 7) were obtained at a high flow-rate, 50 ml min − 1 which corresponds to a digestion time of only 12 s. Previously digestion times of 1 min (6 ml min − 1) have been found to provide sufficient digestion to obtain identity of a number of model proteins [11]. The ability to direct the degree of digestion of the protein by varying the flow provides a powerful mean to adapt the IMER to different conditions, since more difficult samples may require longer digestion times. The mass spectra in Fig. 7b also show two phosphopeptides that are characteristic for b-casein. Phosphorylation of peptides can be confirmed either by an enzymatic dephosphorylation step [20,21] or by MS/MS analysis [22]. One potential use for these trypsin mIMERs could be digestion of proteins separated by multidimensional liquid chromatography, 2DHPLC, where the ability to obtain identity of separated proteins is of great interest.
6. Conclusion The importance of a correctly designed mIMER to fully utilise the benefit of integrating PS in
chemical m-systems as a surface enlarging carrier matrix for enzymes has clearly been demonstrated. By tailoring the mIMER channel cross-section geometry, homogeneous PS layers were obtained on high aspect ratio silicon microstructures, e.g. the mIMER, yielding improved biocatalytic performance. In the case of the protein digestion application the outcome of the optimisation work is crucial, enabling high-speed on-line protein digestion. Further work will focus on the microfluidic optimisation of the microreactors and the full applicability of the proteolytic mIMER to the proteomics field.
References [1] H.H. Weetall, in: K. Mosbach (Ed.), Methods in Enzymology, vol. 44, Academic Press, New York, 1976, p. 134. [2] B. Xie, B. Danielsson, P. Norberg, F. Winquist, I. Lundstrom, Sens. Actuat. B-Chem. 6 (1992) 127. [3] M. Suda, T. Sakuhara, Y. Murakami, I. Karube, Appl. Biochem. Biotechnol. 41 (1993) 11. [4] T. Laurell, L. Rosengren, Sens. Actuat. B-Chem. 19 (1994) 614. [5] T. Laurell, J. Drott, L. Rosengren, K. Lindstrom, Sens. Actuat. B-Chem. 31 (1996) 161. [6] J. Drott, L. Rosengren, K. Lindstrom, T. Laurell, Thin Solid Films 330 (1998) 161. [7] J. Drott, L. Rosengren, K. Lindstrom, T. Laurell, Mikrochim. Acta 131 (1999) 115. [8] M. Bengtsson, J. Drott, T. Laurell, Phys. Status Solidi (a) 182 (2000) 533. [9] J. Drott, K. Lindstrom, L. Rosengren, T. Laurell, J. Micromech. Microeng. 7 (1997) 14. [10] T. Laurell, J. Drott, L. Rosengren, Biosens. Bioelectron. 10 (1995) 289. [11] S. Ekstro¨ m, P. O8 nnerfjord, J. Nilsson, M. Bengtsson, T. Laurell, G. Marko-Varga, Anal. Chem. 72 (2000) 286. [12] W.J. Hezel, T.M. Billeci, J.T. Stults, S.C. Wong, C. Grimely, C. Watanabe, Proc. Natl. Acad. Sci. USA 90 (1993) 5011. [13] P. James, M. Quadroni, E. Carafoli, G. Gonnet, Prot. Sci. 3 (1994) 1347. [14] D.J.C. Pappin, P. Hojrup, A.J. Bleasby, Curr. Biol. 3 (1993) 327. [15] M. Mann, P. Hojrup, P. Roepstroff, Biol. Mass Spectrom. 22 (1993) 338. [16] J. Yates, S. Speicher, P.R. Griffin, T. Hunkapiller, Anal. Biochem. 214 (1993) 397. [17] R.S. Annan, D.E. Mculty, S.A. Carr, Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, 1996, p.702.
M. Bengtsson et al. / Talanta 56 (2002) 341–353 [18] J. Gobom, E. Nordhoff, E. Mirgorodsakaya, R. Ekman, P. Roepstorff, J. Mass Spectrom. 34 (1999) 105. [19] M. Kussmann, E. Nordhoff, H. Rahbek-Nielsen, S. Haebel, M. Rossel-Larsen, L. Jakobsen, J. Gobom, E. Mirgorodskaya, A. Kroll-Kristensen, L. Palm, P. Roepstorff, J. Mass Spectrom. 32 (1997) 593.
353
[20] A. Srensballe, S. Andersen, O.N. Jensen, Proteomics 1 (2001) 207. [21] M.R. Larsen, G.L. Sorensen, S.J. Fey, P.M. Larsen, P. Roepstrorff, Proteomics 1 (2001) 223. [22] M.J. Huddleston, R.S. Annan, M.T. Bean, S.A. Carr, J. Am. Soc. Mass Spectrom. 4 (1993) 710.