Automated serum peptide profiling using novel magnetic C18 beads off-line coupled to MALDI-TOF-MS

Transcription

1 598 DOI /prca Proteomics Clin. Appl. 2007, 1, TECHNICAL BRIEF Automated serum peptide profiling using novel magnetic C18 beads off-line coupled to MALDI-TOF-MS Connie R. Jimenez 1, 2, Zineb El Filali 1, Jaco C. Knol 1, Klaas Hoekman 1, Frank A. E. Kruyt 1, Giuseppe Giaccone 1, August B. Smit 2 and Ka Wan Li 2 1 OncoProteomics Laboratory, Department of Medical Oncology, VU Medical Center, Amsterdam, The Netherlands 2 Department of Molecular and Cellular Neurobiology, Center for Medical Systems Biology/Neurogenomics and Cognitive Research, Vrije Universteit, Amsterdam, The Netherlands Serum peptide profiling by MS is an emerging approach for disease diagnosis and biomarker discovery. A magnetic bead-based method for off-line serum peptide capture coupled to MALDI- TOF-MS has been recently introduced. However, the reagents are not available to the general scientific community. Here, we developed a protocol for serum peptide capture using novel magnetic C18 beads, and automated the procedure on a high-throughput magnetic particle processor. We investigated bead equilibration, peptide binding and peptide elution conditions. The method is evaluated in terms of peaks counts and reproducibility of ion intensities in control serum. Overall, the DynaBead-RPC18-based serum sample processing protocol reported here is reproducible, robust and allows for the detection of,200 peptides at m/z of serum that was allowed to clot for 1 h. The average intra-experiment %CV of normalized ion intensities for crude serum and 0.5% TFA/0.15% n-octyl glucoside-treated serum, respectively, were 12% (range 2 38%) and 10% (3 21%) and the inter-experiment %CVs were 24% (10 53%) and 31% (10 59%). Importantly, this method can be used for serum peptide profiling by anyone in possession of a MALDI-TOF instrument. In conjunction with the KingFisher 96, the whole serum peptide capture procedure is high-throughput (,20 min per isolation of 96 samples in parallel), thereby facilitating large-scale disease profiling studies. Received: September 22, 2006 Revised: January 12, 2007 Accepted: March 12, 2007 Keywords: Automation / MALDI-TOF-MS / Magnetic beads / Reproducibility / Serum In the past few years, MS is increasingly used to profile the low molecular weight proteome of serum in a variety of diseases [1, 2]. The general aim is to identify biomarker patterns that can be used for diagnosis, prognosis or monitoring of disease. Current efforts in working towards this goal are focused on the development of standard procedures for preanalytical sample handling and reliable procedures for serum peptide sample processing and profiling. Ultimately, Correspondence: Dr. Connie R. Jimenez, OncoProteomics Laboratory, Department of Medical Oncology, VU Medical Center, De Boelelaan 1117, 1081HV Amsterdam, The Netherlands c.jimenez@vumc.nl Fax: the impact in the field of oncology may be significant, where early diagnosis and therapy-response monitoring can make a significant difference in outcomes. So far, most studies employed TOF mass spectrometers with SELDI, where peptide binding is performed on chromatographic surfaces [2, 3]. The disadvantages of this approach are the low resolution of the SELDI-TOF instruments and the limited binding capacity of a surface. More recently, a few studies have used high-resolution instruments with MALDI coupled to a TOF ion detector. These MALDI-TOF-MS studies mainly made use of a magnetic bead-based method for off-line serum peptide capture [4 7]. Unfortunately, these magnetic beads are not generally available [6] or only to those with a mass spectrometer from a

2 Proteomics Clin. Appl. 2007, 1, particular manufacturer [4, 5, 7]. Recently, Dynal Bead-Based Separations (Invitrogen) released novel monodisperse superparamagnetic beads that have a surface modified with C18 alkyl groups, optimized for the adsorption of peptides (Dynabeads RPC 18) and that are available to any user. As a first step toward the establishment of multi-center studies of serum protein profiling, we developed a protocol for serum peptide capture using these magnetic C18 beads and implemented the serum sample processing protocol on a highthroughput magnetic particle processor, the KingFisher 96 (Thermo Fisher Scientific). For Dynal bead-based serum processing, we first manually optimized the serum protocol for Dynabeads RPC 18, which was based in part on a previous report [6]. In short, the manual sample processing protocol is as follows: serum and optional binding buffer were mixed in 500-ml Eppendorf tubes and premixed, and the equilibrated bead slurry was added, mixed and left for 2 min. Subsequently, the tubes were placed in the manual magnetic particle concentrator (Dynal MPC, Dynal Bead-Based Separations, Invitrogen) and the supernatant aspirated, the tubes removed and 200 ml wash solution added. Tubes, with wash solution, were placed back into the separator and each tube was rotated in the holder 20 times, followed by aspiration of the supernatant. This washing procedure was repeated three times. After the final washing step, bound peptides and proteins were eluted by incubation with 2 volumes 50% ACN for 2 min before collecting the elution solution using the separator. For MALDI-TOF-MS, 1 volume of sample was manually mixed with 2 volumes of matrix (2 mg CHCA in 1 ml 80% ACN/10 mm diammonium citrate) in a tube from which 0.7 ml was deposited on a stainless steel MALDI plate. MALDI-TOF-MS was performed in reflectron positive mode on the 4800 MALDI-TOF/TOF mass spectrometer (Applied Biosystems). As previous studies showed that most ions occur in the low mass region, the analyzed mass range in most experiments was Da. Linear mode spectra were acquired from m/z from samples spotted in the same matrix in 50% ACN/10 mm diammonium citrate. The acquisition laser power was set just above the threshold. Data acquisition was automated through the software, and programmed to accumulate 2500 shots per spectrum (reflectron) and 5000 shots for linear mode (100 laser shots per sub-spectrum). The instrument was externally calibrated using a calibrant peptide mixture. The samples analyzed were human serum from Sigma, and control serum that was in-house collected according to a standard protocol (1-h clotting at room temperature, prior to centrifugation at room temperature and freezing of the aliquoted supernatant at 807C). For each tested condition, a serum sample was processed in three to five independent isolations. The following conditions were tested: (i) beadprotein ratio; (ii) bead equilibration prior to sample incubation; (iii) serum peptide binding conditions; and (iv) peptide elution conditions. For the two best conditions, four (crude serum) and three (0.5% TFA/0.15% n-octyl glucoside-treated serum) independent experiments were performed using the KingFisher 96, a device for automated magnetic particle processing. The total number of peaks detected was used as the criterion to select a particular condition. To determine peak counts, molecular ions with a S/N ratio.10 were exported from Data Explorer software (Applied Biosystems) to a spreadsheet. Average peak counts and SDs were normalized to the condition with the greatest peak number. Further data handling was performed in Markerview (Applied Biosystems) that allows for import of the spectra in batch mode (200 ppm bin size) and aligns the mass spectra along the m/z axis, thereby facilitating subsequent export of the data as a matrix of m/z values and relative intensities. For assessment of reproducibility of relative ion intensities and to allow for calculation of the %CV of selected peptides, each absolute intensity was normalized against the total intensity in the spectrum and multiplied by a scaling factor of We performed a series of experiments to optimize the conditions for serum sample processing: (i) Bead-protein ratio: Different amounts of Dynabeads RPC 18 were tested against a fixed amount of protein. In the manual protocol, we first tested 0.1, 0.2, 0.5, and 1 mg beads versus 6 ml Sigma serum (,240 mg protein). Because the minimum volume for KingFisher handling is 20 ml, we also tested 0.5, 1, 2.5, and 5 mg C18 beads versus 20 ml Sigma serum (,800 mg protein). Figure 1A reveals that mg is sufficient for serum peptide capture from 6 ml serum and 1 mg beads for 20 ml serum. The latter sample-to-bead ratio was used in subsequent experiments. (ii) Bead equilibration prior to sample incubation: The following buffer compositions were tested: 200 mm NaCl/0.1% TFA and 0.15% n-octyl glucoside/0.1% TFA. Figure 1B shows that the equilibration buffer containing 200 mm NaCl/0.1% TFA gave better performance in terms of peak counts. (iii) Serum peptide binding conditions: Some controversy exists in the literature with respect to the best serum peptide binding condition for RP magnetic beads. Villanueva et al. [6] advocate the use of crude, untreated serum, whereas in the ClinProt protocol an acidic proprietary buffer is used in a sample-to-buffer ratio of 1:2 [4, 5, 7]. Here we tested crude serum as well as serum diluted in various buffers in three different sample-to-buffer ratios (1:2, 1:4 or 1:9). The following binding buffers were tested: 0.1% TFA (1:9), 0.1% TFA/0.15% n-octyl glucoside (1:9), 0.5% TFA (1:2), 0.5% TFA/0.15% n-octyl glucoside (1:2), 0.5% TFA/0.15% n-octyl glucoside/100 mm NaCl (1:4), 0.1% TFA/ 0.15% n-octyl glucoside/100 mm NaCl (1:4 and 1:9) and compared to undiluted serum. The serum in all conditions was incubated for 5 min at room temperature, with mild vortexing, prior to loading to the beads. When performing a total peak count over the whole assessed mass range ( Da), in this experiment a slightly higher peak count was obtained for serum diluted in 0.5% TFA/0.15% n-octyl glucoside ( ions) than for crude serum ( ions) (Fig. 1C), although the average total peak count of m/z of three different experiments was similar (crude serum: ions and 0.5% TFA/0.15% n-octyl glucoside

3 600 C. R. Jimenez et al. Proteomics Clin. Appl. 2007, 1, Figure 1. Optimization of serum sample processing using Dynabeads RPC 18. (A) An investigation of effect of sample-to-magnetic bead ratios on serum peptide processing and display. Equal volumes (6 and 20 ml) of sigma serum (equivalent to,240 and 800 mg protein) were incubated with different weight amounts of Dynabeads RPC 18. (B) Effect of bead equilibration conditions. Beads were equilibrated in 200 mm NaCl/0.1% TFA or 0.1% TFA/0.15% n-octyl glucoside prior to serum sample incubation. (C) Effect of peptide binding conditions on serum peptide profiling using Dynabeads RPC 18 and MALDI-TOF-MS. The following binding conditions (5 min, room temperature, with mild vortexing) were tested in triplicate using 20 ml serum as starting material: 0.1% TFA (1:9), 0.1% TFA/0.15% n-octyl glucoside (1:9), 0.5% TFA (1:2), 0.5% TFA/0.15% n-octyl glucoside (1:2), 0.5% TFA/0.15% n-octyl glucoside/100 mm NaCl (1:4), 0.1% TFA/0.15% n-octyl glucoside/ 100 mm NaCl (1:4 and 1:9) and compared to undiluted serum. Equal volumes of treated serum as well as crude serum were incubated with fixed-weight amounts of Dynabeads RPC 18 (1 mg). (D, E) Peptide profiles are shown for three selected binding conditions [crude serum and serum diluted in 0.5% TFA (1:2) and 0.5% TFA/0.15% n-octyl glucoside (1:2)]. (E) Zoom view of the mass range of Da (indicated by a line in A) is shown. Arrowheads indicate some differential ions. (F) Effect of elution conditions. Elution with 100% ACN was compared to elution in 50% ACN. After each treatment, beads were washed and eluted with 50% ACN, and the eluates analyzed by MALDI- TOF-MS. Relative peak counts (PC), with respect to the largest number are plotted. The absolute peak count at 100% was with S/N.10 and depends on the sample source and duration of clotting time at room temperature (data not shown). The data are the mean of three independent measurements, and SDs are shown. serum ions) (Fig. 2C). When analyzing sub mass ranges ( Da and Da), it is apparent that incubation of serum in 0.5% TFA/0.15% n-octyl glucoside yields enhanced detection of peptides over m/z,2500 (crude serum: ions and 0.5% TFA/0.15% n-octyl glucoside serum ions) (Figs. 1D and 2). This effect is also apparent in linear mode (Figs. 3A and B) that yielded a higher peak count in the m/z range for serum

4 Proteomics Clin. Appl. 2007, 1, Figure 2. Comparison of crude and 0.5% TFA/0.15% n-octyl glucoside-treated serum peptide profiles obtained by Dynabeads RPC 18- mediated sample processing and MALDI-TOF-MS. (A, B) Examples of the serum peptide profiles of in-house collected serum obtained using untreated serum (A) and serum in a 1:2 dilution in 0.5% TFA/0.15% n-octyl glucoside (OG) (B). Overlapping but differential profiles are obtained with the highest peak complexity in the mass range when serum is diluted in 0.5% TFA/0.15% n-octyl glucoside buffer (compare zoom views of A and B). (C) Bar graphs displaying the absolute inter-experiment peak counts for the entire reflectron-mode mass range ( ) and two subdivisions of the mass range ( and Da). The average absolute values with SDs from four (crude serum) and three (TFA/0.15% n-octyl glucoside-treated serum) experiments performed over the course of 10 weeks are shown. samples incubated in 0.5% TFA/0.15% n-octyl glucoside ( ions) than for crude serum ( ions) (Fig. 3C). Interestingly, the MALDI mass profiles of crude and 0.5% TFA/0.15% n-octyl glucoside diluted serum contained overlapping but different sets of peptides (Figs. 1B, 2 and 3), implying that a combination of both binding conditions would yield the highest number of peptides. Note that, to avoid serum protein precipitation, it is important to lower the ph in the sample to,ph 2. We noticed this is only achieved with a 0.5% acid concentration in the 1:2 and 1:4 dilutions. In the condition in which we used 0.1% acid (0.1% TFA/0.15% n-octyl glucoside/100 mm NaCl, ratio 1:4), the ph was only lowered to ph 5 and some protein precipitation was observed. This explains the relative low peak counts obtained in this binding condition (Fig. 1C). (iv) Peptide elution: We compared elution in 50% ACN with elution in 0.1% TFA/50% ACN. In line with Villanueva et al. [6], we observed (slightly) higher peak counts when eluting in 50% ACN with a higher reproducibility of peak detection (Fig. 1E). Automated sample handling is essential for large scale cancer diagnosis and therapy-response studies. We tested the usefulness of the newly launched apparatus, the King- Fisher 96 (Thermo), a magnetic particle processor that makes use of 96 magnetic rods to move particles through the processing steps. Only samples and reagents need to be loaded onto the 96-well plates. We implemented the optimized manual protocol and tested 20-, 40- and 100-ml elution volumes. In addition, we optimized robot parameters with respect to binding time and speed (data not shown). Table 1 shows details on optimized conditions and robot parameters. Figure 4A shows that the serum peptide profiles of the manual and automated procedure are comparable. In addition, similar peak counts were obtained in the manual and KingFisher 96-processed samples (Fig. 4B). Moreover, using only 20 ml serum samples, reproducible elution can be obtained in 20 ml 50% ACN and at better sensitivity, which results in higher relative peak counts than elution in 40 and 100 ml (data not shown). In a last series of experiments, the reproducibility of KingFisher 96-assisted serum sample processing of different aliquots of the same serum sample on the same day and over the course of 8 weeks was determined (four experiments for crude serum and three different experiments for 0.5% TFA/0.15% n-octyl glucoside-treated serum). Figure 3A shows an example of the serum peptide profiles obtained in

5 602 C. R. Jimenez et al. Proteomics Clin. Appl. 2007, 1, Figure 3. Reproducibility of serum peptide profiles in linear mode. In-house collected serum was used for testing (four independent samples were processed as described and spotted in CHCA matrix in 50% ACN; matrix-analyte ratio of 1:1). (A) Serum peptide profiles of crude (untreated) serum. (B) Profiles obtained after 5-min incubation in 0.5% TFA/0.15% n-octyl glucoside buffer. The inserts show a magnification of the protein clusters at,m/z 9287 (the mass range is indicated by the line underneath the m/z axis). (C) Bar graph displaying the average peak count in the linear mode serum spectra (n = 4, S/N 10). Table 1. Summary of KingFisher 96 parameters and optimized conditions KingFisher plate Plate content Time/speed Plate tip plate: Tip comb & no reagent Plate 1: Dynabeads Plate 20 ml RPS-PRO beads 30 s, medium 180 ml 200 mm NaCl/ 0.1% TFA Plate 2: Dynabeads equilibration 200 ml 200 mm NaCl/ 0.1% TFA 60 s, medium Plate 3: Sample plate 20 ml serum 2 min, bottom slow 40 ml 0.15% n-octyl glucoside/0.5% TFA Plate 4 6: 36washing 200 ml 0.1% TFA 60 s, bottom slow Plate 7: Elution ml 50% ACN 2 min, bottom slow three manual independent isolations on the same day and three parallel isolations on the KingFisher. The profiles are very similar. Figure 2C contains the inter-experiment peak count data (already described above) and Table 2 contains the average molecular weight data with SDs as well as the average normalized ion intensities with intra- and interexperiment %CV values of selected peptides across the mass range of Da, and Table 3 across the mass range of Da. These peptides were chosen as a representative set that covered different ion intensities from high to low, and m/z ratios from across the entire acquired mass range. In reflectron mode, the intra-experiment %CV (n = 5) of the different peptides in crude serum ranged from 4% to 38% (average 12%) and in 0.5% TFA/0.15% n-octyl glucoside-treated serum from 2% to 21% (average 10%) (Table 2). The inter-experiment %CV (n = 5; n = 3; n = 3; n = 4) in crude serum ranges from 10% to 53% (average 24%) and in 0.5% TFA/0.15% n-octyl glucoside serum from 10% to 47% (average 31%) (Table 2). In linear mode, the intra-experiment variation on the normalized ion intensities was slightly higher than in relfectron mode (Table 3) (crude serum: range 11 22%, average CV 15%; 0.5% TFA/0.15% n- octyl glucoside-treated serum: range 9 31%, average CV 19%). These values are in the range of values published in

6 Proteomics Clin. Appl. 2007, 1, Figure 4. Automation by King- Fisher 96 and reproducibility of sample processing for serum peptide analysis by MALDI-TOF- MS. A. Automation by King- Fisher 96 yields comparable serum peptide profiles as manual processing. Using the optimized analytical protocol, different aliquots of the same crude serum sample were independently processed either manually or using Dynabeads RPC 18 in conjunction with the King- Fisher 96. Segments of the MALDI-TOF mass spectra corresponding to peptides in the kDa mass range are shown. (B) Intra-experiment reproducibility of absolute peak counts. King- Fisher-assisted processing is similar in performance as compared to manual processing. Table 2. Reproducibility of selected peptide ions in crude and TFA/octyl glucoside-treated serum a) Intra-exp (n = 5) Crude serum OG/TFA Inter-exp (n = 3) Crude serum OG/TFA m/z SD m/z Norm. %CV Norm. %CV m/z SD m/z Norm. %CV Norm. %CV Average a) Norm., normalized ion intensity; OG, n-octyl glucoside. previous studies using the ClinProt kit [4, 5] and are acceptable for semi-quantitative profiling studies. Together, these data show that serum sample processing using the KingFisher 96 is at least as reproducible as manual processing with the added advantage of speed, parallel processing and a performance that is user independent (as opposed to manual processing). These features are especially important for processing of large series of crude serum that may contain on-going proteolytical activity. It is possible that perfectly controlled laboratory temperature and matrix crystallization conditions may further improve reproducibility. Table 3. Reproducibility of selected peptide ions in linear mode (n =4) m/z SD Crude %CV OG/TFA %CV Average 15 19

7 604 C. R. Jimenez et al. Proteomics Clin. Appl. 2007, 1, In summary, the C18-DynaBead-based serum sample processing protocol reported here is reproducible and robust and can be used for serum peptide profiling by anyone in possession of a MALDI-TOF instrument. In conjunction with the KingFisher 96, the whole serum peptide capture procedure is high-throughput and takes only,20 min for isolation of 96 samples in parallel, thereby facilitating largescale disease profiling studies. The authors wish to thank Dr. D. Gillooly (Dynal Bead Based Separations, Invitrogen) and Merja Mehto (Thermo Electron) for useful discussions and providing materials for testing. Dr. Matthias Glueckmann (Applied Biosystems) is acknowledged for assistance with data pre-processing and Dr. Ron Bonner (MDS Sciex) for early access to MarkerView. The Center for Medical Systems Biology is acknowledged for providing funding. References [1] Liotta, L. A., Petricoin, E. F., J. Clin. Invest. 2006, 116, [2] Petricoin, E. F., Liotta, L. A., Curr. Opin. Biotechnol. 2004, 15, [3] Issaq, H. J., Conrads, T. P., Prieto, D. A., Tirumalai, R., Veenstra, T. D., Anal. Chem. 2003, 75, 148A 155A. [4] Baumann, S., Ceglarek, U., Fiedler, G. M., Lembcke, J. et al., Clin. Chem. 2005, 51, [5] de Noo, M. E., Tollenaar, R. A., Ozalp, A., Kuppen, P. J. et al., Anal. Chem. 2005, 77, [6] Villanueva, J., Philip, J., Entenberg, D., Chaparro, C. A. et al., Anal. Chem. 2004, 76, [7] West-Nielsen, M., Hogdall, E. V., Marchiori, E., Hogdall, C. K. et al., Anal. Chem. 2005, 77,