Integrated Sample Preparation and MALDI Mass Spectrometry on a

Anal. Chem. 2004, 76, 345-350

Integrated Sample Preparation and MALDI Mass Spectrometry on a Microfluidic Compact Disk Magnus Gustafsson,† Daniel Hirschberg,‡ Carina Palmberg,‡ Hans Jo 1 rnvall,‡ and Tomas Bergman*,‡

Gyros AB, Uppsala Science Park, SE-751 83 Uppsala, Sweden, and Department of Medical Biochemistry and Biophysics, Karolinska Institutet, SE-171 77 Stockholm, Sweden

High-throughput microfluidic processing of protein digests integrated with matrix-assisted laser desorption/ ionization (MALDI) mass spectrometry on a compact disk (CD) is described. Centrifugal force moves liquid through multiple microstructures, each containing a 10-nL reversed-phase chromatography column. The CD enables parallel preparation of 96 samples with volumes ranging from one to several microliters. The peptides in the digests are concentrated, desalted, and subsequently eluted from the columns directly into MALDI target areas (200 × 400 µm) on the CD using a solvent containing the MALDI matrix. After crystallization, the CD is inserted into the MALDI instrument for peptide mass fingerprinting and database identification at a routine sensitivity down to the 200-amol level. Detection of proteolytic peptides down to the 50-amol level is demonstrated. The success rate of the CD technology in protein identification is about twice that of the C18 ZipTips and standard MALDI steel targets. The CDs are operated using robotics to transfer samples and reagents from microcontainers to the processing inlets on the disposable CD and spinning to control the movement of liquid through the microstructures. Identification of proteins and their posttranslational modifications are important1 to functionally assign the large number of genomic sequences now available. The human genome was previously reported,2,3 and additional genomes including that of the mouse4 are constantly added. Polyacrylamide gel electrophoresis in two dimensions (2-D) combined with mass spectrometry is a major route for correlation of genomic data with those of protein expression5 and for routine investigations of protein levels. Recent developments in narrow pI range gels, protein prefractionation, sensitive mass spectrometers, and careful sample preparation adds the possibility of identifying low-abundance * Corresponding author. Tel: +46-8-5248 7780. Fax: +46-8-337 462. E-mail: [email protected]. † Gyros AB. ‡ Karolinska Institutet. (1) Proteomics in functional genomics; Jolle`s, P., Jo ¨rnvall, H., Eds.; Birkha¨user: Basel, 2000. (2) Lander, E. S.: et al. Nature 2001, 409, 860-921. (3) Venter, J. C.; et al. Science 2001, 291, 1304-1351. (4) Waterston, R. H.; et al. Nature 2002, 420, 520-562. (5) Appella, E.; Arnott, D.; Sakaguchi, K.; Wirth, P. J. Proteomics in functional genomics; Birkha¨user: Basel, 2000; pp 1-27. 10.1021/ac030194b CCC: $27.50 Published on Web 12/09/2003

© 2004 American Chemical Society

proteins, which often are the most interesting targets.6 However, additional technologies to ensure high sensitivity, throughput, and resolution need to be developed for detection of crucial, low-copynumber proteins.7 For characterization of proteins, mass spectrometry (MS) is now routinely performed on tryptic digests extracted from electrophoresis gels. Peptide mass fingerprinting by matrix-assisted laser desorption/ionization (MALDI) mass spectrometry followed by assignments in sequence databases identifies the protein.8 Phosphorylations and further posttranslational modifications can then also be determined with mass spectrometry.9-11 Sample preparation with a minimal loss of material is particularly important when protein amounts are low. Below the picomole level, adsorption of sample molecules to the surfaces of pipets, tubing, walls of gel supports, and chromatographic resins becomes significant. Microfluidic devices are then essential. The key feature is handling of multiple samples and reagents in the nanoliter range.12 Microfluidic chips coupled to electrospray ionization mass spectrometry as a detection method were early demonstrated.13,14 Later, chip-based separation techniques were used with electrophoresis in an open channel containing sample mixed with MALDI matrix, and after crystallization, the chip was inserted into a MALDI instrument for analyte identification.15 However, the combination of electrodriven chips and MALDI-MS is not technically straightforward since crystallization of the sample with matrix before MALDI-MS is not easily achieved in a continuous liquid medium. Consequently, there is a need for a technology that can work at the nanoliter scale and integrate sample concentration, desalting, elution, and crystallization before MALDI analysis. High(6) Herbert, B. R.; Harry, J. L.; Packer, N. H.; Gooley, A. A.; Pedersen, S. K.; Williams, K. L. Trends Biotechnol. 2001, 19, S3-S9. (7) Gygi, S. P.; Corthals, G. L.; Zhang, Y.; Rochon, Y.; Aebersold, R. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 9390-9395. (8) Aebersold, R.; Mann, M. Nature 2003, 422, 198-207. (9) Larsen, M. R.; Roepstorff, P. Fresenius J. Anal. Chem. 2000, 366, 677690. (10) Hirschberg, D.; Rådmark, O.; Jo ¨rnvall, H.; Bergman, T. J. Protein Chem. 2003, 22, 177-181. (11) Knight, Z. A.; Schilling, B.; Row: R. H.; Kenski, D. M.; Gibson, B. W.; Shokat, K. M. Nat. Biotechnol. 2003, 21, 1047-1054. (12) Astorga-Wells, J.; Jo ¨rnvall, H.; Bergman, T. Anal. Chem. 2003, 75, 52135219. (13) Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 1997, 69, 1174-1178. (14) Xue, Q.; Foret, F.; Dunayevskiy, Y. M.; Zavracky, P. M.; McGruer, N. E.; Karger, B. L. Anal. Chem. 1997, 69, 426-430. (15) Liu, J.; Tseng, K.; Garcia, B.; Lebrilla, C. B.; Mukerjee, E.; Collins, S.; Smith, R. Anal. Chem. 2001, 73, 2147-2151.

Analytical Chemistry, Vol. 76, No. 2, January 15, 2004 345

throughput processing at the subfemtomole level with such an integrated system is now reported. EXPERIMENTAL SECTION Materials. Bovine serum albumin (BSA, 67 kDa) from Sigma (98%, product A-3294) was prepared in aliquots for gel application by serial dilution in water from an aqueous stock solution of 9.9 pmol of BSA/µL. The BSA stock concentration was determined by amino acid analysis using a Biochrom 20 Plus instrument (Amersham Pharmacia Biotech) after acid hydrolysis. Water was from a MilliQ purification unit (Millipore). All chemicals were of analytical grade. SDS/Polyacrylamide Gel Electrophoresis. One-dimensional separations employed Bis-Tris-HCl buffered (pH 6.4) precast 10% polyacrylamide gels of 1-mm thickness containing 15 wells (NuPAGE, Novex). Electrophoresis was carried out at 120 V and room temperature for 1-1.5 h until the dye marker (bromophenol blue) had reached the edge of the gel. After electrophoresis, proteins were stained with Coomassie R-250 at 0.1% (w/v) in 40% methanol containing 10% acetic acid for 4 h at room temperature (shaking). Destaining in the methanol/acetic acid solution was carried out until the protein bands were clearly visible against the background, usually overnight at room temperature (shaking). 2-D Gel Electrophoresis. Cell lysates in 8 M urea, detergent, reducing agents, and protease inhibitors were prepared and submitted to isoelectric focusing on 13-cm Immobiline Drystrips, pI gradient 4-7, followed by SDS-PAGE with 10% polyacrylamide gels as described.16 Gel spots were visualized with Coomassie R-250 as above. In-Gel Digestion. After excision with a methanol-cleaned scalpel blade, the gel pieces containing protein bands/spots were digested manually or by the use of a MassPREP robotic protein handling system with a protocol including reduction and alkylation of cysteine residues (Micromass). For manual digestion, gel pieces were placed in Eppendorf tubes (1.5 mL) and cut into small cubes of ∼1-mm side. The gel cubes were washed twice for 45 min (shaking), using 200 µL of 0.2 M ammonium bicarbonate/ acetonitrile (1:1) at 37 °C. They were then shrunk in 100 µL of neat acetonitrile twice (shaking), and dried to crispness in a vacuum centrifuge (10-20 min). Proteins were reduced by addition of 20 mM DTT in 0.2 M ammonium bicarbonate using a volume sufficient to cover the gel pieces (typically 10-20 µL) and incubated at 37 °C for 1 h (shaking), again followed by shrinking and drying as above. The proteins were then alkylated by the addition of 55 mM iodoacetamide in 0.2 M ammonium bicarbonate, again using volumes sufficient to cover the gel pieces (1020 µL), followed by incubation at room temperature for 15 min in the dark. After this treatment, the gel pieces were shrunk again and dried as above. The proteins were then digested with 0.5 µg of trypsin (Promega, modified) added in 1-µL aliquots from a 0.1 µg/µL solution in 0.2 M ammonium bicarbonate to enable slow penetration into the gel pieces. The pieces were covered with 0.2 M ammonium bicarbonate (∼50 µL) followed by incubation overnight at 37 °C (shaking). The digestion was stopped by addition of 0.5-1 µL of neat trifluoroacetic acid (TFA). After recovery of the supernatant into a new Eppendorf tube (0.5 mL), (16) Oppermann, M.; Cols, N.; Nyman, T.; Helin, J.; Saarinen, J.; Byman, I.; Toran, N.; Alaiya, A. A.; Bergman, T.; Kalkkinen, N.; Gonza`lez-Duarte, R.; Jo ¨rnvall, H. Eur. J. Biochem. 2000, 267, 4713-4719.

346 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

Figure 1. Microfluidic structure for on-CD processing of samples of up to 1 µL (96 structures per CD). An individual microstructure element is shown separately for the three consecutive steps of sample application (A), washing/elution (B), and cocrystallization with MALDI matrix (C). The dilute and salt-containing crude sample (blue) is applied at (A) onto the 10-nL reversed-phase column (white). The washing and elution/matrix solutions (yellow) are applied via a common distribution channel at (B), with the liquid volume defined to 200 nL after activation of an overflow channel (not shown). Cocrystallization of the concentrated and desalted sample with MALDI matrix is taking place at (C) in a 200 × 400 µm target area where the crystalline deposit is accessible to the laser beam of the MALDI instrument as shown by a photograph of the crystals in the desorption area at the outlet.

the gel pieces were extracted twice for 45 min at 37 °C (shaking) with 100 µL of 60% acetonitrile containing 0.1% TFA and then extracted two more times for 10 min at 37 °C (shaking), first with 50 µL of 40% acetonitrile/0.1% TFA and then with 50 µL of neat acetonitrile containing 0.1% TFA. All extracts were pooled with the initial supernatant. This combined peptide extract was stored at -20 °C until processed. In-Solution Digestion. The BSA stock solution, or dilutions thereof, was mixed with ammonium bicarbonate, reduced, alkylated, and digested with trypsin (Promega, modified). Typically, 1 nmol of BSA (100 µL of stock solution) was mixed with 80 µL of 50 mM ammonium bicarbonate to which 1 µL of 50 mM DTT was added. After incubation at 50 °C for 15 min, 2 µL of 55 mM iodoacetamide was added and incubation continued for 15 min but at room temperature. To this solution, 1.5 µg (1.5 µL) of trypsin and 15 µL of 0.2 M ammonium bicarbonate was added, and the digestion was carried out overnight at 37 °C. The reaction was quenched by the addition of 2 µL of neat TFA, and the digest was stored frozen at -20 °C until processed. CD Processing of Tryptic Digests. For CD-based sample preparations, a Gyrolab MALDI SP1 CD in a Gyrolab Workstation (Gyros AB, Uppsala, Sweden) was used. The microfluidic steps and components used for concentration, desalting, and crystallization of each peptide sample before MALDI-MS are illustrated in Figure 1. On each CD, 96 microcolumns for parallel concentra-

Figure 2. MALDI mass spectra from analysis of a tryptic peptide extract recovered after in-gel digestion of BSA (maximally 12 fmol/µL based on the amount of protein applied to the gel before electrophoresis, staining, excision, digestion, and extraction). Panel A: 1 µL of the extract processed and analyzed on-CD resulted in 17 identified peptide peaks (asterisks). Panel B: 1 µL of the same extract processed and analyzed in standard manner on a steel target (dried-droplet technique) after mixing with 1 µL of 3 mg/mL CHCA in 70% acetonitrile containing 0.1% TFA, resulted in seven identified peptide peaks (asterisks) using the same mass spectrometer and annotation settings. Autolytic trypsin fragments are indicated by T.

tion/desalting by reversed-phase chromatography are packed to a volume of 10 nL on top of a geometrical restriction in each channel, using Source 15RPC (Amersham Biosciences) with a particle size of 15 µm. The columns were conditioned with 50% acetonitrile in water, and after an additional rinse with 0.1% TFA, the tryptic digest was added to the sample inlet (Figure 1). To load the column, the spinning of the disk was adjusted to optimize the liquid flow. The peptides were retained on the reversed-phase column, while the solution with salts and other polar components passed through the exit hole from the disk into a waste collector. The column was washed with 10% ethanol and 0.1% TFA, using the same liquid addition procedure (Figure 1). A common distribution channel was used to add wash solution (200 nL) simultaneously to each group of 16 CD columns. The volume was defined by a two-step spinning procedure (Figure 1) in which an overflow function in the common distribution channel is activated. To elute peptides from the column, a volume of 200 nL of 50% acetonitrile, with 1 mg/mL R-cyano-4-hydroxycinnamic acid (CHCA) and 0.1% TFA, was distributed via the common channel. The eluate was transported via spinning to a MALDI target area of 200 ×

400 µm for solvent evaporation and crystallization of the peptide/ matrix mixture. The total time required to process a CD with 96 samples is ∼40 min. CD Technology. In contrast to traditional microcentrifugal approaches,17 the CD surfaces were modified by oxygen gas plasma treatment18 making the inner walls of the microchannels hydrophilic to ensure capillary action-driven filling of the structures. Hydrophobic breaks were employed to stop the liquid from filling the entire structure.19,20 Increasing the centrifugal force increases the pressure and the liquid is pushed over the hydrophobic break. When peptides and matrix are eluted from the CD columns, the speed of rotation of the CD is adjusted to ensure that the eluate enters the MALDI target area at a rate compatible with the rate of evaporation, since the eluted volume (17) Scott, C. D.; Burtis, C. A. Anal. Chem. 1973, 45, 327A-340A. (18) Larsson, A.; De´rand, H. J. Colloid Interface Sci. 2002, 246, 214-221. (19) Ekstrand, G.; Holmquist, C.; Edman O ¨ rlefors, A.; Hellman, B.; Larsson, A.; Andersson, P. 2000 MicroTotal Analysis Systems, Proceedings, 2000; pp 311-314. (20) Tiensuu, A.-L.; O ¨ hman, O.; Lundbladh, L.; Larsson, O. 2000 MicroTotal Analysis Systems, Proceedings, 2000; pp 575-578.

Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

347

is larger than the volume of the desorption area. A steady-state situation is typically created at 1000 rpm. Once the solvent/ peptide/matrix mixture is in the open area, it becomes exposed to air, leading to solvent evaporation and crystallization, a procedure with a typical duration of 2 min. Type of solvent and concentration of the MALDI matrix are important factors for crystal growth and laser desorption/ionization properties. On microscopic inspection, the crystals were similar in size and shape to crystals prepared by conventional techniques. MALDI Mass Spectrometry and Database Searches. Tryptic digests were eluted from the microcolumns together with the CHCA matrix for on-CD analysis. For comparison, in-gel digests were either directly mixed with matrix and applied to standard steel targets using the dried-droplet technique21 or treated with C18 ZipTips (Millipore) in the MassPREP robotic workstation (Micromass). For mass spectrometry, Voyager DE-Pro (Applied Biosystems) and Biflex IV (Bruker) instruments were employed, operated in reflectron mode. Before on-CD MALDI analysis, the CD was cut into two (Biflex IV) or 6 (Voyager DE-Pro) pieces using a precision cutter to be able to accommodate the disk in the mass spectrometer. For database searches and protein identification, the web site of Mascot (http://www.matrixscience. com) was used. RESULTS Femtomole Processing of Protein Digests. The CD technology works routinely for concentration, desalting, crystallization and MALDI-MS of in-gel-digested protein samples, as demonstrated by BSA at the subpicomole level (Figure 2). After SDSPAGE and staining with Coomassie blue, the BSA band was excised and digested with trypsin, resulting in a final peptide extract at maximally 12 fmol/µL (based on the amount of protein applied to the gel before electrophoresis, staining, excision, digestion, and extraction). Of this solution, 1 µL was on-CD concentrated, desalted, crystallized, and analyzed by MALDI-MS (Figure 2A), and the results were compared to those from analysis of the same amount (1 µL) directly applied to a standard steel target and mixed with 1 µL of 3 mg/mL CHCA in 70% acetonitrile containing 0.1% TFA (Figure 2B). Using the CD approach (Figure 2A), the desalted sample is focused on a narrow desorption spot with dimensions 200 × 400 µm, while in direct application using the dried-droplet method (Figure 2B), the sample contains gel and buffer contaminants and is crystallized over a relatively wide area in the range 1-1.5 mm. A comparison of the resulting MALDI mass data reveals that 10 additional peptides are detected for the CD processed sample in relation to the number observed for standard application to steel-target analysis. A database search with the mass mapping data from analysis of the two samples, using the same annotation threshold and Mascot settings, identified both as being BSA at the highest individual score. However, the score was 159 based on 17 peptides with a sequence coverage of 29% in the CD case (Figure 2A) and only 79 on 7 peptides with a sequence coverage of 12% in the standard method (Figure 2B). In a survey of cellular proteins separated by 2-D gel electrophoresis, 48 spots were analyzed using both CD technology (2030 min processing time excluding the MALDI analysis) and the (21) Doktycz, S. J.; Savickas, P. J.; Krueger, D. A. Rapid Commun. Mass Spectrom. 1991, 5, 145-148.

348

Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

standard method with C18 ZipTips (∼10-fold longer processing time when carried out manually). The proteins varied in size, 16110 kDa, and in pI, 4.75-8.3. In both routes, a robotic protein handling system was employed to reduce, alkylate, and in-gel digest the proteins. The data reveal that 45 of the totally 48 spots investigated resulted in protein identification using the CD approach, while 24 spots generated a positive identification after standard C18 ZipTip treatment. Based on these numbers, the success rate was 94% for the CD technique and 50% for the standard ZipTip procedure. To estimate the CD interreproducibility, identifications were carried out in duplicate resulting in sequence coverages that on average varied by 5%. The proportion of protein spots yielding a high sequence coverage (in excess of 30%) was greater with the CD technology (Table 1). Sensitivity and Reproducibility. The high sensitivity achieved with the CD procedure was tested using in-solution tryptic digests of BSA diluted to low concentrations by addition of 0.1% TFA. Samples of 400, 200, and 50 amol were on-CD concentrated, desalted, crystallized, and analyzed by MALDI-MS. Using these amounts, 11, 9, and 3 BSA tryptic peptides were detected, respectively (Figure 3). When the peak list resulting from each sample was submitted to Mascot and searched against SwissProt, it was found that both the 400-amol and the 200-amol levels were sufficient for protein identification with scores of 107 and 90 and sequence coverages of 20 and 15%, respectively (Figure 3A and B). Even at the 50-amol level, BSA was still the top suggestion from the database, but the score was only 42 and the sequence coverage 6%, which is not considered sufficient for an independent identification in a real case (Figure 3C). For the 400- and 200amol samples, the database was searched for all entries and was restricted to one miscleavage and a mass accuracy of 120 ppm including carbamidomethylcysteine and oxidized methionine modifications. For the 50-amol sample, the database search was limited to zero miscleavage and restricted to mammals only. In the pretreatments, pipet tips, Eppendorf tubes, and microtiter plates were coated with BSA to prevent adsorption of peptides (procedure according to Nalge Nunc International). The number of peaks detected, the signal-to-noise ratios observed, and the mass accuracies obtained in the spectra examined after on-CD analysis (Figure 3) are compatible with identification of low-abundance proteins in gel separations. The design of the MALDI/CD contains 6 sections of 16 individual sample preparation microstructures linked by a common distribution channel (cf. Figure 1). Each section concentrates and desalts 16 individual samples, and the reproducible operation of these microstructures therefore becomes an important issue. To test the overall reproducibility, aliquots corresponding to 1 fmol of an in-solution BSA tryptic digest were analyzed to verify the inter- and intradisk reproducibility. The MALDI mass spectra after on-CD analysis then turned out to be close to identical, giving rise to the same number of peaks for database searches and protein identification (Figure 4). DISCUSSION Successful CD methodology was demonstrated using newly developed instrumentation for cleanup, concentration, and MALDIMS of protein digests. Results are superior to those achieved with conventional methodology and reach sensitivities below the 100amol level. The signal intensity in MALDI-MS after on-CD

Figure 3. MALDI mass spectra from analysis of an in-solution tryptic digest of BSA from which aliquots in the attomole range were applied to concentration, desalting, crystallization, and MALDI-MS on the CD. Panel A: 1 µL (400 pM) or 400 amol, resulted in 11 peptides identified (asterisks) when the peak list was searched against SwissProt using Mascot. Panel B: 1 µL (200 pM) or 200 amol, resulted in nine peptides identified (asterisks, SwissProt and Mascot). Panel C: 250 nL (200 pM) or 50 amol, resulted in three peptides identified (asterisks, SwissProt and Mascot). In the pre-CD treatments, pipet tips, Eppendorf tubes, and microtiter plates were coated with BSA to prevent adsorption losses of low-abundance proteolytic peptides (procedure according to Nalge Nunc International).

Table 1. Comparison of the CD Technology and Standard C18 ZipTip Processing Applied to Cellular Proteins Separated by 2-D Gel Electrophoresisa sequence coverage CD ZipTip

10-20%

20-30%

>30%

5 4

11 7

29 13

a The study was carried out with 48 spots that were processed using the CD technology and standard C18 ZipTips resulting in 45 identifications by the CD approach and 24 after ZipTip treatment.

processing therefore enables the recovery of extensive peak lists, improving the database searches and protein identification. After tuning the acceleration potentials in the mass spectrometer, the mass accuracy and resolution are equal to those obtained for standard MALDI targets. The CD can perform multiple steps in an integrated fashion. Concentration of samples results from two steps within the process: (i) a fairly large sample volume (one to several microliters) is adsorbed onto the CD column (10 nL), similar to the approach of ZipTips (Millipore), followed by elution; and (ii)

crystallization in a restricted area (typically 200 × 400 µm), leading to a sample concentration of ∼10-fold in comparison with a 1-mm standard target. To achieve the high sensitivity required for detection of lowabundance proteins, it is crucial to minimize adsorptive losses. The present integration of several sample pretreatments and analytical steps within a single microstructure minimizes surface contacts and thereby sample losses. A number of surface chemistries were tested during the development of the CD system, and oxygen gas plasma treatment was found to give the polymer surface a hydrophilic character with low tendency to peptide adsorption and therefore suitable for microfluidic function and analytical performance. Improved performance in the CD technology over that of ZipTip treatment and standard MALDI steel targets is illustrated by tests with cellular protein extracts separated by 2-D gel electrophoresis. After Coomassie staining, 48 spots excised and analyzed using both the CD technology and the standard method with ZipTips, in both cases with a robotic protein handling system, identified 45 proteins using the CD approach but only 24 proteins Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

349

Figure 4. MALDI mass spectra in the m/z region 900-2100 from analysis of an in-solution tryptic digest of BSA demonstrating high analytical reproducibility. Aliquots corresponding to 1 fmol from the same digest were applied to all individual microstructures of several CDs and then processed and analyzed by MALDI-MS. The 10 spectra shown are from desorption of samples from 10 different CD microstructures. BSA fragment positions are indicated by asterisks.

using the ZipTip treatment. Based on these numbers, the success rate with the CD technique was about twice that with the standard procedure and the total time for CD processing before MALDI analysis was just 20-30 min. The CD intervariability regarding sequence coverage for individual proteins was 5%. More protein spots were identified at higher sequence coverage as compared to standard treatment (Table 1). In conclusion, the CD system allows 96 parallel applications of samples for simultaneous processing, where concentration, cleanup, crystallization, and actual MALDI-MS are integrated within a single CD format. The CD approach enables analysis at the femtomole/attomole level, high-throughput processing, and good reproducibility, all important factors in proteome characterization.

350 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

ACKNOWLEDGMENT We are grateful to Martin Schu¨renberg (Bruker, Bremen) and Phil Savickas (Applied Biosystems, Framingham) for help and fruitful discussions. This work was supported by grants from the Swedish Research Council (projects 03X-3532, B5101-879/2001, and K5104-20005891), the Swedish Cancer Society (project 4159), and Karolinska Institutet.

Received for review May 13, 2003. Accepted October 20, 2003. AC030194B