Anal. Chem. 1999, 71, 4753-4757
Solid-Phase Extraction/MALDI-MS: Extended Ion-Pairing Surfaces for the On-Target Cleanup of Protein Samples Li Zhang and Ron Orlando*,†
Complex Carbohydrate Research Center, Department of Biochemistry & Molecular Biology, and Department of Chemistry, University of Georgia, 220 Riverbend Road, Athens, Georgia 30602-4712
The surface of a matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) target can be covalently modified so that it behaves like a medium for solid-phase extraction (SPE). These modified targets are capable of binding molecules of interest, but not contaminants, from sample solutions placed on them. This allows the analyte to be cleaned up on the probe surface by simply washing the target to remove the contaminants prior to MALDI-MS analysis. A limitation of the current SPE/MALDI-MS targets is that they have a fairly low binding capacity, since the coating on these targets is based upon self-assembled monolayers (SAMs). To overcome this limitation, we have investigated new surface coatings for SPE/MALDI-MS that will have a higher binding capacity than targets modified with SAMs. Here, we describe the development of new SPE/MALDI-MS surfaces that have very high molecular weight (>300 000) polylysine chains attached to them. Targets modified in this manner are capable of binding peptides/proteins by ion-pairing interactions and have approximately 100 times the binding capacity of the SAM-based targets. Furthermore, these polylysine targets can capture over 60% of a protein from a highly contaminated solution. Consequently, polylysine SPE/MALDI-MS targets offer a practical solution for analyzing very small volumes (300 kDa) polylysine to the MALDI probe. We have demonstrated that these surfaces possess approximately 100 times more analyte binding sites than SAM based SPE/MALDI targets for proteins and can extract >60% of the protein from a highly contaminated solution. Consequently, polylysine SPE/MALDI probes appear to be a practical solution for analyzing very small volumes (300 kDa) at pH 8.0 in 50 mM triethylamine (TEA) with 0.5 M NaCl and allowed to react for 1 h at room temperature. After that, the targets were washed with water and dried. SAM modified targets were made using previously published procedures.9 Basically, the sputtered gold targets were incubated for 30 min in an anhydrous ethanol solution saturated with 2-aminoethanethiol hydrochloride. After incubation, the targets were first washed in ethanol then in water to remove unbound thiolates. Sample Preparation. Stock solutions containing 20 pmol/ µL of bovine serum albumin (BSA) or carbonic anhydrase were 4754
Analytical Chemistry, Vol. 71, No. 20, October 15, 1999
prepared in both distilled water and a solution saturated in sodium acetate that also contained 8 M urea and 20% Tween solution. The yeast protein VTi1p expressed from Escherichia coli of 1.9 pmol/ µL was present in a solution containing 25 mM HEPES, 200 mM NaCl, 15% glycerol, and 10% NP-40 (a nonionic surfactant). Solutions of angiotensin-III at a concentration of 1 pmol/µL were also prepared in both distilled water and a solution saturated with sodium acetate and urea that also contained 20% Tween. On-target SPE/MALDI-MS Procedures. The basic steps of an SPE/MALDI-MS experiment are to deposit a 0.3-0.5 µL aliquot of the analyte solution on the modified probe and let it stand for a few minutes. Next, it is washed with deionized water to remove any contaminants. Finally, a solution containing the matrix is deposited on the sample regions, the matrix is allowed to dry, and the target is inserted into the mass spectrometer. This entire process adds only a few minutes to sample preparation yet allows successful MALDI-MS analysis of samples whose concentrations of salts, chaotropic agents, and/or detergents/surfactants would have otherwise required extensive sample purification prior to MALDI/MS analysis. MALDI-TOF Measurements. Mass spectra were obtained on both a Hewlett-Packard G2030A MALDI-TOF mass spectrometer and a Kratos SEQ MALDI-TOF mass spectrometer. The HP instrument was operated at an accelerating voltage of 28 kV, an extractor voltage of 7 kV, and a pressure of ∼7 × 10-7 Torr. The Kratos SEQ was operated in linear mode using pulsed-ion extraction, with an accelerating voltage of 20 kV and a pressure less than 1 × 10-6 Torr. In both instruments, the samples were desorbed/ionized from the probe tip using a nitrogen laser source with an output wavelength of 337 nm. Calibration was performed using mixtures of peptides/proteins with known molecular masses. We used 3,5-dimethoxy-4-hydroxy-cinnamic acid (sinapanic acid) dissolved in 70% aqueous acetonitrile with 0.01% trifluoroacetic acid as the MALDI matrix for experiments performed on the HP instrument, while R-cyano-4-hydroxycinnamic acid dissolved in 70% aqueous acetonitrile with 0.01% trifluoroacetic acid was used as the MALDI matrix for experiments performed on the Kratos instrument. “Conventional” MALDI refers to a typical MALDI preparation in which the analyst mixes a volume of matrix and a volume of sample and then dries the resulting mixture onto the probe surface. This method is the technique with which we wish to compare our method using SPE/MALDI-MS technologies. This procedure is referred to throughout the text as conventional MALDI. Materials. The poly-L-lysine samples (with molecular weight ranges of 4-15, 30-70, 150-300, and >300 kDa), hydrochloric acid, carbonic anhydrase, BSA, angiotensin-III, and Tween were obtained from Sigma (St. Louis, MO). NaCl, TEA, trifluoroacetic acid, sinapinic acid, R-cyano-4-hydroxycinnamic acid, 2-aminoethanethiol hydrochloride, and nitric acid were obtained from Aldrich (Milwaukee, WI). THF (Aldrich) was dried by treating with molecular sieves (Aldrich; baked at 400 °C for 4 h and stored in a dry container) and redistilled before use. DSP was obtained from Fluka (Buchs, Switzerland). Sodium acetate, acetonitrile, and methanol were obtained from T. J. Baker (Phillipsburg, NJ). Ethanol and urea were purchased from Fisher (Fair Lawn, NJ). Unless previously described, all of these were used as provided.
Figure 1. Mechanistic illustration of analyte binding to an SPE/ MALDI target depicting (A) a small analyte on a SAM surface, (B) a large analyte on a SAM surface, and (C) a large analyte on an extended surface.
RESULTS AND DISCUSSION The binding capacity of SAM based SPE/MALDI targets is fairly small and is a major limitation of these targets. For example, in the case of hydrophobic SPE/MALDI targets, the capacity is approximately 3 pmol for a 1.7 kDa peptide per sample region regardless of the amount of peptide applied.10 Our current theory is that this behavior results from the SAM coating, which presumably will only bind a single monolayer of analyte (Figure 1A). This theory also explains why SAM based SPE/MALDI targets are less successful in the analysis of proteins than they are with that of peptides, since increasing the analyte’s size decreases the number of molecules in the monolayer (Figure 1B). On the basis of this theory, we predicted that surfaces with extended coatings would allow a larger number of analyte molecules to be captured by the probe (Figure 1C). An easy way to manufacture this type of surface is to attach high-molecularweight polylysine to the MALDI target using existing chemistry.11 Using this methodology, we envisioned that one or more of the amino groups on the polylysine would be attached to the MALDI target, while the remainder would be available to bind analyte molecules by ion-pairing interactions. Thus, we expected that targets modified with polylysine would have a larger number of analyte binding sites than SAM based targets. A series of experiments was performed to evaluate if polylysine SPE/MALDI targets do indeed have a greater binding capacity for proteins than SAM modified targets and to determine the effect of the polylysine chain length on binding capacity. If the binding is limited by surface area, one would expect that increasing the polylysine chain length would increase the binding capacity of the target. In these experiments, 0.5 µL of a carbonic anhydrase solution (20 pmol/µL, dissolved in a solution saturated with sodium acetate, with 8 M urea and 20% Tween) was placed on the SPE/MALDI target. This was allowed to stand for 10 min following which the target was washed with deionized water. Finally, 1 µL of a matrix solution containing 1 pmol/µL of myoglobin (the internal standard used to normalize these spectra) was placed on the sample region. The variable in this experiment was the length of the polylysine chain. As can be clearly seen from these spectra (Figure 2), the amine-terminated SAM modified SPE/MALDI target has a very low binding capacity for this proteinsthe very problem we are trying to overcome with the (11) Brockman, A. H.; Orlando, R. Anal. Chem. 1995, 67, 4581.
Figure 2. SPE/MALDI-MS analysis of carbonic anhydrase (20 pmol/ µL, dissolved in a solution saturated with sodium acetate, with 8 M urea and 20% Tween) on targets modified with (A) an amineterminated SAM, (B) 30-70 kDa polylysine, (C) 150-300 kDa polylysine, and (D) >300 kDa polylysine. In these spectra, 1 pmol of myoglobin was added to each sample region in the matrix solution. The myoglobin served as an internal standard, and the spectra were normalized to its intensity.
development of the polylysine targets. This data also demonstrates that all of the polylysine SPE/MALDI targets have a greater binding capacity than the SAM based target and that the binding capacity increases with the length of the polylysine chain. Similar experiments were performed using various other proteins as the analyte and provided similar results (data not shown for brevity). This data strongly supports our theory that surface area limits the binding capacity of SPE/MALDI targets and that by using a chemically rough surface the binding capacity of the probe can be increased by at least 2 orders of magnitude. The extraction efficiency of the polylysine (>300 kDa) was evaluated in a second set of experiments. Here, 0.5 µL of a carbonic anhydrase solution (20 pmol/µL, dissolved in a solution saturated with sodium acetate, with 8 M urea and 20% Tween) was analyzed by SPE/MALDI on a polylysine (>300 kDa) target (Figure 3A), while a sample of carbonic anhydrase (20 pmol/µL, dissolved in deionized water) was analyzed by conventional MALDI (Figure 3B). In both of these experiments, the matrix solution contained 1 pmol/µL of myoglobin so that we could compare the abundance of the carbonic anhydrase peaks. These spectra indicate that over 60% of the carbonic anhydrase from the highly contaminated solution was captured by the SPE/MALDI target, since the intensity of the carbonic anhydrase signal observed in this experiment is just over 60% of that observed by conventional MALDI. Results from a similar experiment using bovine serum albumin (BSA) also demonstrated a capture efficiency of approximately 60% from a highly contaminated sample (data not shown). One concern we had was that the polylysine coatings on the SPE/MALDI targets would degrade the resolution and mass accuracy of the MALDI. To evaluate this potential pitfall, a peptide was analyzed by both SPE/MALDI and conventional MALDI. A peptide was selected for this study since decreases in mass resolution and/or accuracy would be more easily seen with a lower mass sample. In the SPE experiments, angiotensin-III, dissolved to a concentration of 1 pmol/µL in a solution saturated with sodium acetate and urea that also contained 20% Tween, was Analytical Chemistry, Vol. 71, No. 20, October 15, 1999
4755
Table 1. Successful Analysis of Peptides/Proteins in the Presence of Buffers, Salts, Detergents, Stabilizing Agents, and Chaotorpesa by SPE/MALDI
Figure 3. Analysis of carbonic anhydrase solutions (20 pmol/µL) by (A) SPE/MALDI-MS using a >300 kDa polylysine-modified target and (B) conventional MALDI. In the SPE/MALDI-MS experiment, the protein solution was saturated with sodium acetate and contained 8 M urea and 20% Tween, while in the conventional MALDI-MS experiment the protein was dissolved in deionized water. In these experiments, 1 pmol of myoglobin was added to each sample region in the matrix solution. The myoglobin served as an internal standard, and the spectra were normalized to its intensity.
buffers (to saturated solutions) potassium phosphate sodium phosphate potassium acetate sodium acetate Tris Tris*EDTA TRICINE BICINE PBS HEPES salts (to saturated solutions) sodium chloride potassium chloride detergents (to 20% v/v) Triton X-100 Tween 20 Np-40 stabilizing agents (to 15% v/v) glycerol chaotorpes (to 8 M) urea guanidinium HCl a
And mixtures of the above contaminants.
Figure 4. Analysis of angiotensin-III solutions (1 pmol/µL) by (A) SPE/MALDI-MS using a >300 kDa polylysine-modified target and (B) conventional MALDI. In the SPE/MALDI-MS experiment, the peptide solution was saturated with sodium acetate and urea and contained 20% Tween, while in the conventional MALDI-MS experiment the peptide was dissolved in deionized water.
Figure 5. Analysis of a “real world” sample (VTi1p) by (A) conventional MALDI-MS and (B) SPE/MALDI-MS using a >300 kDa polylysine-modified target. This protein solution contained 25 mM HEPES, 200 mM sodium chloride, 15% glycerol, and 10% NP-40.
analyzed on a polylysine (>300 kDa) target (Figure 4A), while a sample of angiotensin-III (1 pmol/µL, dissolved in deionized water) was analyzed by conventional MALDI (Figure 4B). As can be clearly seen by comparing these two spectra, the mass resolution and accuracy are unaffected by the polylysine coating. We would also like to point out that in these experiments the instrument was calibrated externally by conventional MALDI-MS. Consequently, changing from an unmodified target to an SPE/MALDI target has no effect on the mass assignments. Experiments were also performed to evaluate the range of contaminants that could be removed by polylysine SPE targets (Table 1). The only class of common protein additives that we have found to interfere with this type of target is ionic surfactants (such as sodium dodecyl sulfate, SDS). This presumably results from the ability of SDS to ion-pair with the polylysine surface. Because these surfactants are much more abundant than the analyte (8 M vs 10 µM), the ionic surfactants will be the
predominate species captured by the positively charged amine groups on the target’s surface. Various controls were used to ensure that the polylysine coatings on the MALDI target were in fact the cause of our successful isolation of the proteins in the above solutions. We were unsuccessful in obtaining spectra from protein solutions containing a high level of any of the contaminants listed in Table 1 when an underivatized probe was substituted for a polylysine probe. These results show that an underivatized surface does not share the binding characteristics of ion-pairing surfaces. Also, we were unable to acquire a spectrum from these protein solutions by conventional MALDI, that is, by mixing a 1-µL aliquot of sample with 1 µL of matrix, depositing this mixture on the probe, and directly analyzing. The absence of signal in these experiments confirmed that conventional preparatory methods are not suitable for this level of contamination.
4756 Analytical Chemistry, Vol. 71, No. 20, October 15, 1999
The success of polylysine (>300 kDa) SPE/MALDI targets on known samples led us to analyze a series of “real world” samples using this approach. For brevity, we will only discuss one of these samples: VTi1p, a yeast protein that has been overexpressed in E. coli. When submitted to us for MS analysis, the protein was dissolved in 25 mM HEPES, 200 mM sodium chloride, 15% glycerol, and 10% NP-40 solution. Hence, this sample provided an excellent test for the ability of our polylysine targets. Conventional MALDI-MS of this sample (Figure 5A) did not detect any signals from this protein because of the high degree of contamination. Alternatively, the SPE/MALDI spectrum of this sample (Figure 5B) shows an intense signal from the protein. This experiment clearly demonstrates the power of polylysine-modified targets for the on-target cleanup of highly contaminated real world protein samples. In summary, a new class of SPE/MALDI-MS targets has been developed that provide a significantly higher binding capacity for proteins than SPE targets based on SAMs. These new targets
enable protein samples with high concentrations of salts, nonionic detergents, denaturing agents, and low volatility solvents to be quickly analyzed by SPE/MALDI-MS. The ion-pairing nature of these surfaces may be useful for a broad range of applications such as the analysis of oligonucleotides and oligosaccharides. These surfaces also offer the potential for on-probe separations. ACKNOWLEDGMENT This work was supported by grants from the National Science Foundation (NSF Grant no. 9626835) and the National Institutes of Health (NIH Grant no. 2-P41-RR05351) which made this project possible. We would like to thank Professor Leigh Ann Lipscomb (Department of Biochemistry and Molecular Biology, University of Georgia) for the sample of recombinant VTi1p. Received for review March 29, 1999. Accepted August 11, 1999. AC990328E
Analytical Chemistry, Vol. 71, No. 20, October 15, 1999
4757
