Anal. Chem. 2006, 78, 5513-5524
Microfluidic Liquid Chromatography System for Proteomic Applications and Biomarker Screening Iulia M. Lazar,*,†,‡ Phichet Trisiripisal,†,§ and Hetal A. Sarvaiya†,|
Virginia Bioinformatics Institute, Department of Biological Sciences, Department of Electrical and Computer Engineering, and Department of Biomedical Engineering, Virginia Polytechnic Institute and State University, Washington St. Bio II/283, Blacksburg, Virginia 24061
A microfluidic liquid chromatography (LC) system for proteomic investigations that integrates all the necessary components for stand-alone operation, i.e., pump, valve, separation column, and electrospray interface, is described in this paper. The overall size of the LC device is small enough to enable the integration of two fully functional separation systems on a 3 in. × 1 in. glass microchip. A multichannel architecture that uses electroosmotic pumping principles provides the necessary functionality for eluent propulsion and sample valving. The flow rates generated within these chips are fully consistent with the requirements of nano-LC platforms that are routinely used in proteomic applications. The microfluidic device was evaluated for the analysis of a protein digest obtained from the MCF7 breast cancer cell line. The cytosolic protein extract was processed according to a shotgun protocol, and after tryptic digestion and prefractionation using strong cation exchange chromatography (SCX), selected sample subfractions were analyzed with conventional and microfluidic LC platforms. Using similar experimental conditions, the performance of the microchip LC was comparable to that obtained with benchtop instrumentation, providing an overlap of 75% in proteins that were identified by more than two unique peptides. The microfluidic LC analysis of a protein-rich SCX fraction enabled the confident identification of 77 proteins by using conventional data filtering parameters, of 39 proteins with p < 0.001, and of 5 proteins that are known to be cancer-specific biomarkers, demonstrating thus the potential applicability of these chips for future high-throughput biomarker screening applications. Advanced mass spectrometry (MS) detection enables a comprehensive characterization of the proteomic molecular cell profile. Relevant answers can be provided for questions that relate to the identity and expression level of the existing proteins, the nature, site, and number of the posttranslational modifications, and the specific functions associated with these proteins. The generated information is essential for differentiating normal versus diseased * To whom correspondence should be addressed. E-mail:
[email protected]. † Virginia Bioinformatics Institute. ‡ Department of Biological Sciences. § Department of Electrical and Computer Engineering. | Department of Biomedical Engineering. 10.1021/ac060434y CCC: $33.50 Published on Web 07/04/2006
© 2006 American Chemical Society
cell states. The challenges associated with the analysis of proteomic samples are, however, numerous: complexity (thousands of proteins/sample), wide range of concentrations (dynamic range of 1:106), low-level expression for certain components (less than 1000 copies/cell), dynamic composition (different sets of proteins are expressed in various stages of cell development), and availability. The preparation of these samples for MS analysis involves a complex workflow that often concludes with the generation of tens or hundreds of minimal sample subfractions (sample volumes of 5-10 µL, and/or low pM-µM concentration ranges). The development of multiplexed analytical platforms that enable highthroughput explorations is essential to provide a timely solution to proteomic investigations. Microfluidic devices have emerged as powerful and reliable analysis platforms.1-3 The miniature format and the capability to manipulate small sample amounts (0.1-1 nL) result in short analysis times and significantly reduced analysis costs. Sample injection, separation, labeling, and detection can be performed routinely in a few minutes or even seconds.4 The ability to perform precise and accurate sample handling operations enables process control and automation and, consequently, the generation of reliable and high-quality data. Microfabrication enables large-scale integration, multiplexing, and high-throughput analysis. In addition, a variety of novel analytical principles and configurations become possible only if they are executed in a microfabricated format. The microdomain environment enables the emergence of unique physical events. Surface-driven phenomena are dominating in the microscale world, and electroosmotic flow (EOF) represents a relevant example. EOF is commonly utilized for fluidic manipulations but can be effectively generated only in capillaries with dimensions in the micrometer domain, as larger dimensions do not support this fluid transport mechanism. In this context, a microfluidic pumping and valving system was developed.5 The configuration of the pumping device consists of hundreds of parallel micro/nanochannels that generate EOF. While the mechanism of operation relies on electroosmotic (1) Jakeway, S. C.; De Mello, A. J.; Russell, E. L. Fresenius J. Anal. Chem. 2000, 366, 525-539. (2) Greenwood, P. A.; Greenway, G. M. Trends Anal. Chem. 2002, 21 (11), 726-740. (3) Boone, T. D.; Fan, Z. H.; Hooper. H. H.; Ricco; A. J.; Tan, H.; Williams, S. J. Anal. Chem. 2002, 74, 78A-86A. (4) Jacobson; S. C.; Culbertson; C. T.; Daler; J. E.; Ramsey, J. M. Anal. Chem. 1998, 70, 3476-3480. (5) Lazar, I. M.; Karger, B. L. Anal. Chem. 2002, 74, 6259-6268.
Analytical Chemistry, Vol. 78, No. 15, August 1, 2006 5513
pumping principles, the pump enables stable flow and pressure generation in electrical field-free regions on the chip. In addition, the same configuration can be used for effective EOF valving, as well. The utility of the present electroosmotic pumping system is especially important for the generation of small flow rates, in the low-nanoliter per minute region, where conventional benchtop instrumentation is not always operational. The EOF pump flow range is fully consistent with the requirements of conventional nano-LC systems or micro/nanoelectrospray sources (30-100 nL/min, 100-200 psi). The pump occupies an area of only a few square millimeters and can be easily interfaced with other functional elements of a micro total analysis system. In this article, we describe the implementation of the multichannel EOF pumping technique into a microchip integrated LC system. The implementation of pressure-driven separations with superior performance and loading capacity such as LC, which is most often used as the final separation step prior to MS analysis, will significantly boost the applicability of microfluidic platforms for proteomic applications.6 Previously reported LC systems comprise separation channels fabricated in silicon, polyimide, or glass; these microfluidic devices are, however, typically connected to external pumping and valving systems and autosamplers.7-9 An alternative approach for performing microchip-LC with a higher level of integration is based on electrochemical pumping principles; this device was demonstrated for the analysis of peptide mixtures.10,11 The main advantage of the microfluidic LC system that we have devised is that it incorporates in a small, flat, 3 in. × 1 in. glass slide the entire components necessary to perform pressure-driven separations, i.e., separation channel, micropump, valve, and ESI interface. Fluidic propulsion and valving is accomplished with a simple multichannel configuration that can be reproducibly fabricated by using standard photolithography and wet chemical etching; no additional microfabrication steps are necessary for the implementation of the pump and valve. The system functions as a completely stand-alone unit. The small footprint enables multiplexing and possibly the fabrication of costeffective, disposable devices. The performance of the microchip integrated LC system is similar to a commercial µ-LC instrument, if identical experimental conditions are used. We demonstrate the applicability of this microfluidic LC chip for the detection of cancer biomarkers in cellular extracts generated from the MCF7 breast cancer cell line. To the authors’ best knowledge, this is the first demonstration of the applicability of a fully integrated microfluidic LC system for the detection of multiple disease-specific biomarkers. The unique architecture and disposable format of these miniaturized platforms will be particularly useful in the clinical setting for high-throughput discovery and screening of biomarker components. (6) Lazar, I. M.; Sarvaiya, H.; Trisiripisal, P.; Jung Hae, Y. 53rd Conference on Mass Spectrometry and Allied Topics, San Antonio, TX, June 5-9, 2005. (7) McEnery, M.; Tan, A. M.; Alderman, J.; Patterson, J.; O’Mathuna, S. C.; Glennon, J. D. Analyst 1999, 125, 25-27. (8) Fortier, M. H.; Bonneil, E.; Goodley, P.; Thibault, P. Anal. Chem. 2005, 77, 1631-1640. (9) Yin, H.; Killeen, K.; Brennen, R.; Sobek, D.; Werlich, M.; van de Goor, T. Anal. Chem. 2005, 77, 527-533. (10) Xie, J.; Miao, Y.; Shih, J.; He, Q.; Liu, J.; Tai, Y.-C.; Lee, T. D. Anal. Chem. 2004, 76, 3756-3763. (11) Xie, J.; Miao, Y.; Shih, J.; Tai, Y.-C.; Lee, T. D. Anal. Chem. 2005, 77, 69476953.
5514 Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
Figure 1. Schematic representation of a microfluidic LC system. Key: (1) pumping channels; (2A) and (2B) eluent inlet reservoirs; (3) eluent outlet reservoir; (4) double-T injector that contains the sample plug; (5) separation channel; (6) sample reservoir; (7) sample waste reservoir; (8) sample inlet channels; (9) sample outlet channels; (10) ESI capillary emitter; (11) LC waste reservoir. (A) Sample loading; (B) sample analysis. Note: arrows indicate the main flow pattern through the system.
EXPERIMENTAL SECTION Microfluidic LC System. Microfluidic devices were fabricated from glass using established photolithography and wet chemical etching.12,13 Glass substrates (1.6 mm thick, Nanofilm, Shelton, CA) sputtered with chrome and photoresist were etched to the desired channel depth and, after cutting and drilling access holes (∼1 mm), were thermally bonded to a cover plate by gradually increasing the temperature to 550 °C. Sample manipulation and LC channels were etched in the substrate, while pumping/valving channels were etched in the cover plate. The microfluidic layout of the photomask was prepared with AutoCAD software. Sample and eluent flows were visualized with a Nikon epifluorescence microscope (Melville, NY). The microchip integrated LC system (Figure 1) comprised two EOF pumps, a valving component, a separation channel with an on-column preconcentrator, and an ESI interface. The separation channel (5) was 2 cm long with a depth of ∼50 µm. A slurry of reversed-phase packing material, Zorbax SB-C18/dp ) 5 µm (Agilent Technologies), was loaded manually in the channel from the LC waste reservoir (11) with the aid of a 250 µL syringe. The slurry, ∼5 mg/mL, was prepared in i-PrOH. The packing material was retained in the separation channel or the preconcentrator with the aid of short (∼100 µm in length), multichannel structures, with dimensions similar to that of the pump (∼1.5-1.8 µm deep). The two EOF pumps (1A and 1B) consisted each of 200 nanochannels (2 cm long, ∼1.5-1.8 µm deep) and had different inlet reservoirs (2A and 2B) and a common outlet reservoir (3). The voltage for EOF generation in the pumps was applied to reservoirs 2 and 3. The voltage applied to reservoir 3 represented also the voltage for electrospray generation. EOF leakage in the outlet reservoir 3 was prevented by a porous glass disk (5-mm diameter, 0.8-1-mm width, 40-50-Å pore size) purchased from Chand Associates (Worcester, MA).5 The disk was secured to the bottom of the reservoir and enabled only the exchange of ions but not of bulk flow. Sample loading was accomplished through a double-T injector (4) with the aid of a multichannel EOF valving structure consisting of 100 nanochannels on each arm (2 cm long, ∼1.5-1.8 µm deep). A fused-silica capillary (10 mm long, 20 µm i.d. × 90 µm o.d.) from (12) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (13) Madou, M. Fundamentals of Microfabrication; CRC Press: Boca Raton, FL, 1997; p 405.
Polymicro Technologies (Phoenix, AZ) was inserted into the LC channel for ESI generation (10). The overall flow rate through the LC channel was measured by sliding a 200-µm-i.d. capillary (10 cm long) over the 90-µm-o.d. spray capillary inserted in the chip and by monitoring the displacement of the liquid meniscus versus time. Mass Spectrometry. Mass spectra were acquired with an LTQ ion trap mass spectrometer (Thermo Electron Corp., San Jose, CA). Data-dependent MS acquisition conditions were described in a previous paper.14 Briefly, one MS scan (5 microscans averaged) was followed by one zoom scan and one MS2 on the top five most intense peaks; zoom scan width was (5 m/z, dynamic exclusion was enabled at repeat count 1, repeat duration was 30 s, exclusion duration was 60 s, and exclusion mass width was (1.5 m/z. Collision-induced dissociation parameters were set at the following values: isolation width at 3 m/z, normalized collision energy at 35%, activation Q at 0.25, and activation time at 30 ms. Protein database searching was performed with the BioWorks 3.2 software (Thermo Electron Corp., San Jose, CA) against a human database extracted from the NCBI nr.gz database downloaded on 08/26/05 (included fields were “human” and “sapiens”; excluded field was “virus”). The database contained 131 585 entries. Only fully tryptic fragments (with two missed cleavages) were accepted for database searching. The peptide tolerance was set at 2 amu and the fragment ion tolerance at 1 amu. Chemical and posttranslational modifications were not included in the search. Data filtering was accomplished with XCorr versus charge-state values set at XCorr ) 1.9 for z ) 1, Xcorr ) 2.2 for z ) 2, and Xcorr ) 3.8 for z g 3, respectively. Sample Preparation. MCF-7 breast cancer cells (ATCC, Manassas, VA) were cultured in Eagle’s minimum essential medium supplemented with 10% fetal bovine serum and 0.1% bovine insulin. At 70% confluence, the cells were harvested and lysed and the supernatant was digested with trypsin and prefractioneated with strong cation exchange chromatography (SCX). The protocol is described in detail in a previous work.14 SCX fraction 7 was subjected to microfluidic LC-MS analysis. Reagents. Samples were prepared in high-purity solvents. HPLC grade methanol was purchased from Fisher Scientific (Fair Lawn, NJ). Deionized water (18 MΩ‚cm) was generated in-house using a MilliQ ultrapure water system (Millipore, Bedford, MA). Sequencing grade modified trypsin was purchased from Promega Corp. (Madison, WI), ammonium bicarbonate was from Aldrich (Milwaukee, WI), and Rhodamine 610 chloride was obtained from Exciton (Dayton, OH). Cell culturing reagents were from ATCC (Manassas, VA), and RIPA lysis buffer was obtained from Upstate (Lake Placid, NY). All other reagents and standards were purchased from Sigma (St. Louis, MO). RESULTS AND DISCUSSION EOF Pumping and Valving System. The choice for an EOF pumping system to run the microfluidic LC was dictated by several reasons: (1) the EOF pumps are the only miniaturized pumps that can generate high pressures (hundreds/thousands of bars),15 (2) the manufacturing of these pumps is extremely simple and reliable, (3) the same structure can be effectively utilized for (14) Sarvaiya, H.; Yoon, J. H.; Lazar, I. M. Submitted. (15) Paul, P. H.; Arnold, D. W.; Rakestraw, D. J. Proceedings of the Micro Total Analysis Systems Workshop, Banff, Canada, October 13-16, 1998.
Figure 2. Schematic diagram that illustrates the pumping and valving capabilities of multiple open channel configurations. A large number (100-1000) of microchannels, with very small diameter (d1) and large hydraulic resistance, are connecting a series of reservoirs (R) to a large diameter (d2) channel (C) on the chip. The pressure in the reservoirs is 1 bar and in the main channel is 10 bar. As a result of the large hydraulic resistance of the pumping/valving channels, material transport can only occur through an electrically driven mechanism, but not through a pressure-driven mechanism.
sample loading and valving, and (4) the design enables standalone operation, multiplexing, and high-throughput analysis. Accurate calculations enable the fabrication of pumps that can sustain the flows and pressures necessary for moving fluids through an entire microfluidic network. The equations that describe the fluidic propulsion, the principle of operation, and the performance of this pumping structure were described in detail in a previous work.5 The integration of this structure within a complex work flow such as an LC system is explained in Figure 2. Consider a channel (C) with diameter d2 that is connected to a few reservoirs (R) through a series of shallow microchannels with diameter d1 (Figure 2A). If a potential differential (∆V) is applied between a reservoir (R) and the channel (C), EOF will be generated through the microchannels; if the hydraulic resistance of these microchannels is sufficiently high, eluent will be pumped from the reservoir into the channel even if the channel is pressurized, e.g., at 10 bar. The large hydraulic resistance of the pumping microchannels will impede flow leakage back into the reservoir. Typical configurations in our designs include microchannels that are ∼1-2 µm deep and 5-20 mm long and which are capable of delivering flow rates in the 10-400 nL/min range. A valving structure composed of narrow microchannels similar to the ones used for pumping can be used for injecting and processing the sample in a pressurized environment. As the multiple open channel configuration has a much larger hydraulic resistance than any of the other functional elements on the chip, it can basically act as a valve that is open to material transport through an electrically driven mechanism but is closed to material transport through a pressure-driven mechanism. The same multichannel structure can be used as an EOF pump for eluents and as an EOF valve for sample introduction into a pressurized microfluidic system. If the depth of the microchannels is small enough, the hydraulic resistance is so large that one set of microchannels can be used for pumping and several other sets for valving (Figure 2B). Sample will be introduced and removed from the main channel on the chip only when a potential differential is applied between adequate sample reservoirs, at appropriate moments Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
5515
Figure 3. SEM images of pumping/valving channels. (A) Top view; (B) cross section.
during the analysis. We have shown that if d1/d2 ∼ 0.01, and the number of valving channels is equal or less than the number of pumping channels, fluid flow can be effectively delivered through a large hydraulic resistance element, with minimal pressure driven fluid loss through the nonpumping channels.5 This would be the case of packed channels used for high-performance LC (HPLC) that have hundreds of times larger hydraulic resistance than an open channel. Up to date, for microfluidic chips filled with packing material, we were able to generate sufficient flow for stable electrospray generation with pumping designs that had only onethird of the pumping channels functional. Thus, it is evident that the implementation on a chip of an LC system with a short, packed separation channel, and a fully integrated pumping and valving system, is possible. Scanning electron microscope (SEM) images of cross sections through the pumping channels are shown in Figure 3. The pumping/valving channels were placed 25 µm apart and were etched to a depth of ∼1.5-1.8 µm. Wet chemical isotropic etching resulted in somewhat less than optimal microchannel dimensions, as the 2-µm-wide channels on the photomask translated into ∼7-10-µm-wide channels on the chip, somewhat lower hydraulic resistance than desired, and thus more flow leakage through the pump and less flow through the LC channel. These microchannels were etched relatively fast, within 2-2.5 min, with an interchannel variability in etch depth of ∼5-10%. The etch time generally needed slight adjustments, to compensate for variations in room temperature and the freshness of the BOE etching solution. Correspondingly, for a given eluent composition and pumping electric field, the variability between the electrical currents measured in the two pumps on a new chip was in the 5-15% range. RSD values of flow rates measured within a given experiment were 8, enabling electrospray ionization in positive ion mode. While a different set of peptides will produce intense peaks when electrospraying from high pH solutions in comparison to low pH ones, the net gain arises from the fact that background noise at high pH in positive ESI mode is smaller, as background ions may not have the capability to acquire a positive charge at high pH such as peptides do by a double-protonation process.18 In addition, at high pH, the interaction between negatively charged peptides and negatively charged active sites on the surface of the separation packing material is reduced and results in improved peak shapes. Using this microfluidic arrangement, 77 proteins were identified in SCX fraction 7 by searching a human protein database with 131 585 entries and using charge-dependent cross correlation scores of 1.9, 2.2, and 3.75, respectively, as minimum acceptance data filtering criteria. Of these, 68 proteins were identified with p-values p < 0.1 and 39 proteins with p < 0.001 (Table 1). These proteins were matched by a total of 101 peptides (50 with p < 0.001). All these protein were top choice matches, i.e., the first best protein match for a given set of peptides. The total number of proteins identified from the same fraction using micro-HPLC (17) Lazar, I. M.; Ramsey, R. S.; Ramsey J. M. Anal. Chem. 2001, 73, 17331739. (18) Mansoori, B. A.; Volmer, D. A.; Boyd, R. K. Rapid Commun. Mass Spectrom. 1997, 11, 1120-1130.
Table 3. List of Putative Cancer Biomarkers Identified with the Microfluidic LC-MS Platforma reference 1 2 3 4
5
gi|2914387| gi|12653819| gi|7594732| gi|62913980
gi|4503143|
E chain E, human PCNA K.FSASGELGNGNIK.L keratin 18 (Homo sapiens) R.AQIFANTVDNAR.I R.QAQEYEALLNIK.V keratin 19 (H. sapiens) K.AALEDTLAETEAR.F R.IVLQIDNAR.L KRT8 (H. sapiens) R.ASLEAAIADAEQR.G K.LSELEAALQR.A R.LEGLTDEINFLR.Q cathepsin D preprotein R.VGFAEAARL
MH+
z
1293.6
2
1319.6 1419.7
2 2
1389.6 1041.6
2 2
1344.7 1129.6 1419.7
2 2 2
933.5
2
P (pro) P (pep)
Sf Sf
score XCorr
coverage DeltaCn
MWSp
6.6 × 10-8 6.6 × 10-8 2.5 × 10-7 2.5 × 10-7 6.3 × 10-6 3.7 × 10-7 3.7 × 10-7 6.8 × 10-4 1.5 × 10-5 1.5 × 10-5 3.0 × 10-5 2.6 × 10-4 2.2 × 10-5 2.2 × 10-5
0.61 0.61 2.04 0.94 0.78 1.72 0.96 0.77 5.36 0.93 0.97 0.93 0.77 0.77
10.14 2.87 30.19 3.88 3.20 20.21 4.15 2.91 70.22 4.14 4.31 3.55 10.12 2.38
5.00 0.31 7.70 0.48 0.44 10.60 0.48 0.12 20.30 0.31 0.19 0.46 2.20 0.33
28732.3 484.7 48002.6 997.8 297.7 23330.9 1343.5 728.7 41082.7 1148.0 2277.8 885.8 44523.7 745.7
RSp 1 1 1 1 1 1 1 1 1
peptides ions 1 (10000) 15/36 2 (20000) 20/33 16/33 2 (20000) 22/36 16/24 3 (30000) 20/36 23/27 22/33 1 (10000) 15/24
a MH+, protonated molecular ion; z, peptide charge state; p, probability of a random match; Sf, final score; XCorr, cross-correlation score between virtual and experimental spectrum; coverage, protein sequence coverage; DeltaCn, degree by which the lower ranked peptide scores differ from the correlation score of the best match; MW, protein molecular weight; Sp, preliminary score; RSp, rank of preliminary score; peptides, unique peptide hits per protein; ions, number of matched ions in the spectrum.
(100 µm × 12 cm columns packed with 5-µm Zorbax SB-C18, eluent H2O/CH3CN acidified with 0.01% TFA, flow rate ∼170 nL/min) was 935 (754 with p < 0.1 and 573 with p < 0.001). What concerns the proteins identified with high confidence (p < 0.001), a larger than 10× drop is observed when switching from the benchtop HPLC to the microfluidic platform. However, repeating the analysis with the benchtop system, and using conditions similar to the chip (2-cm separation column, basic buffer, ∼1-µL sample injection volumes), the total number of identified proteins was similar to the number obtained from the chip (see Table 1), i.e., 91 protein matches (76 with p < 0.1 and 48 with p < 0.001). Moreover, there was ∼75% overlap in proteins identified by two unique peptides between the benchtop and microfluidic LC. A detailed study was conducted to identify the reasons for the drop in the proteins identified with the microfluidic platform. Another MCF7 SCX extract was analyzed using various conditions (Table 2). The major factor that affected the number of identified proteins in going from typical HPLC conditions to experimental conditions that mimicked the microfluidic environment was the sample amount (volume) subjected to analysis. Changing the column length or pH conditions had a much smaller effect than decreasing the sample injection volume. By decreasing the volume from 16 to 4 µL, and then to 1 µL, the number of identified proteins with p < 0.001 decreased from 444 to about 160-180, and to 16, respectively. Other experiments, which involved the use of the conventional HPLC platform, were performed to optimize the conditions for MCF7 analysis and have confirmed this outcome: the analysis of the entire batch of 16 SCX fractions yielded 1960 protein matches for the 8-µL injections and 3799 protein matches for the 40-µL injections.14 Data filtering parameters for all these proteins were the same as the ones used for the chip experiments. The overall dimensions of the microchip integrated LC system were 0.5 in. × 2.5 in., enabling the integration of two LC systems on a 1 in. × 3 in. chip, or of six LC systems on a 3 in. × 3 in. chip substrate (Figure 8). The fabrication of high aspect ratio pumping channels by using dry etching techniques will enable the fabrication of pumps with increased density channels and overall smaller dimensions, that in return will result in the integration of an even larger number of LC systems on one microfluidic platform. These
Figure 8. Multiplexed LC microfluidic device. A 3 in. × 3 in. glass substrate contains six fully integrated LC systems.
chips will be investigated in the future for high-throughput sequential LC-ESI-MS analysis or parallel MALDI-MS. Biomarker Detection on a Chip. There is a vast amount of published information regarding differential protein expression analysis in cancerous versus normal cells. These studies have generated hundreds of potential biomarker proteins that could be used after further validation for population screening. The identification of protein coexpression patterns, of a series of biomarkers instead of just one, is extremely important in the clinical setting where high-sensitivity/high-specificity tests must be developed to justify the implementation of novel, large-scale population screening platforms. While differential protein expression analysis was well beyond the scope of this study, benchtop LC-MS analysis of all 16 SCX fractions enabled the identification of ∼25 proteins14 that were previously described in the literature to be up- or downregulated in cancerous cell states and, thus, are believed to represent good biomarker candidates.19-21 Five putative biomarker proteins that were identified in SCX fraction 7 using conventional HPLC-MS were also identified with the microfluidic (19) Hondermarck, H.; Vercoutter-Edouart, A.-S.; Re´villion, F.; Lemoine, J.; ElYazidi-Belkoura, I.; Nurcombe V.; Peyrat, J.-P. Proteomics 2001, 1, 12161232. (20) Esteva, F. J.; Hortobagyi, G. N. Breast Cancer Res. 2004, 6 (3), 109-118. (21) Ross, J. S.; Linette, G. P.; Stec, J.; Clark, E.; Ayers, M.; Leschly, N.; Symmans, W. F.; Hortobagyi, G. N.; Pusztai, L. Expert Rev. Mol. Diagn. 2004, 4 (2), 169-188.
Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
5521
Figure 9. Tandem mass spectra of a “PCNA” peptide generated from (A) microfluidic LC-MS platform and (B) benchtop HPLC-MS platform. 5522
Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
Figure 10. Tandem mass spectra of a “cathepsin D” peptide generated from (A) microfluidic LC-MS platform and (B) benchtop HPLC-MS platform.
Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
5523
LC-MS system: proliferating cell nuclear antigen (PCNA), cathepsin D, and keratins 8, 18, and 19.19 The Sequest output parameters for these protein biomarkers are given in Table 3. All the corresponding peptides had p < 0.001, XCorr scores above the set threshold, DeltaCn > 0.1, and good preliminary scores (Sp), and all of them identified their corresponding biomarker protein as a top choice match. Keratins K8, K18, and K19 are often found to be upregulated in cancer cells. As they are abundant and highly antigenic, keratins may present certain advantages for use as biomarker proteins.19,22,23 All three keratins were identified with both, the benchtop HPLC and the chip, and generated good-quality MS2 data. Unlike tumor cells, normal cells produce the keratins K5, K6, K7, K14, and K17, which are often identified as contaminants by MS techniques. In our study, we identified K8, K18. and K19 only in experiments that involved the analysis of cancer cells. MS2 spectra for the other two proteins that were identified by only a single peptide are given in Figures 9 and 10. PCNA is a protein involved in cell proliferation, DNA replication. and cell cycle regulation.19,24 It is often associated with breast, lung. and pancreatic cancers (upregulated). It was identified from the chip by one peptide with p ) 6.6 × 10-8. The benchtop HPLC-MS experiment yielded 7 unique peptide matches for this protein. MS2 spectra for the common peptide are shown in Figure 9, as acquired from the chip platform and the benchtop HPLC that used a 12cm separation column. Major peaks are common for both these spectra. Complete y-ion series and characteristic b-ions are observed. Cathepsin D is an estrogen-regulated protease involved in protein metabolism and cancer cell migration.19,21,25 High cathepsin D levels are associated with poor prognosis in human primary breast cancers (worse relapse-free and survival rate scenarios). It was identified from the microchip by a peptide with p < 2.2 × 10-5. MS2 spectra are given in Figure 10. A complete y-ion series is observable in both spectra. All biomarker peptide sequences were searched manually in the database to ensure that no incidental match to other proteins could occur. CONCLUSIONS The development of adequate instrumentation and analysis protocols that prevent sample carryover and maintain sample (22) Trask, D. K.; Band, V.; Zajchowski, D. A.; Yaswen, P.; Suh, T.; Sager, R. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 2319-2323. (23) Barak, v.; Goike, H.; Panaretakis, K. W.; Einarsson, R. Clin. Biochem. 2004, 37, 529-540. (24) Chu, J. S.; Huang, C. S.; Chang, K. J. Cancer Lett. 1998, 131 (2), 145-152. (25) Henry, J. A.; McCarthy, A. I.; Angus, B. Cancer 1990, 65, 265-271.
5524
Analytical Chemistry, Vol. 78, No. 15, August 1, 2006
losses at a minimal level is critical to enable the detection of lowlevel components in complex proteomic samples. We have devised a microfluidic LC system that facilitates data-dependent MS analysis of protein digest from cellular extracts and enables the detection of protein coexpression patterns relevant for cancer biomarker screening. The microfluidic format is prone for multiplexing and high-throughput analysis. A number of analytical tools are designed to be disposable in order to prevent sample contamination, carryover, and false positive identifications. While at the present time the fabrication of microfluidic devices is a rather expensive process, and their use as disposable devices may be considered exotic, the experience provided by the microelectronic industry points otherwise. The apparatus that is necessary for the fabrication of these chips is similar to the one that is utilized for the fabrication of microelectronic equipment, and parallelization and large-scale integration of analytical processing steps will result in a similar decrease of investment in the manufacturing of these chips. Once that an efficient and reliable workflow is designed, implemented, tested, and demonstrated to perform adequately on the chip, an entire process can be replicated with the same ease and effort as required for the fabrication of a single, isolated, processing component. Thus, the idea of a “disposable lab-on-achip” device becomes acceptable. The advance of low-cost, disposable microfluidic platforms will enable the biomedical research community to perform contamination-free, high-throughput proteomic investigations. Cost-effective discovery and screening for prognostic/diagnostic biomarker components that can be used collectively with increased specificity and sensitivity will significantly enhance our capacity to intervene in disease detection, prevention, and therapy. With further developments in MS instrumentation, e.g., miniaturized and multiplexed mass spectrometry detection systems, and high-throughput matrix-assisted laser desorption/ionization techniques with improved sensitivity, the possible implementation of the microfluidic-MS technology for large-scale population screening is envisioned. ACKNOWLEDGMENT This work was supported by NSF grant Career BES-0448840. The authors thank Tom Wertalik from the Chemistry glass shop at Virginia Tech for support with the fabrications of the microchips. Received for review March 8, 2006. Accepted June 2, 2006. AC060434Y
