Microfluidic Protein Detection through Genetically Engineered Bacterial Cells Sang-Hyun Oh,† Sang-Ho Lee,† Sophia A. Kenrick,†,‡ Patrick S. Daugherty,*,†,‡ and Hyongsok T. Soh*,†,§ California Nanosystems Institute and Biomolecular Science and Engineering Program; Department of Chemical Engineering; Department of Mechanical Engineering, University of California-Santa Barbara, California 93106 Received April 25, 2006
Abstract: Protein microarray technology, in which a large number of capture ligands are spatially arrayed at a high density, presents an attractive method for high-throughput proteomic analysis. Toward this end, we demonstrate the first cell-based protein detection in a microsystem, wherein Escherichia coli cells are genetically engineered to express the desired capture proteins on the membrane surface and are spatially arrayed as sensing elements in a microfluidic device. An E. coli clone expressing peptide ligands with high affinity and high specificity for target molecules was isolated a priori. Then these cells were electrokinetically immobilized on gold electrodes using dielectrophoresis, thus allowing each sensor element to be electrically addressable. Flow cytometry and subsequent fluorescence analysis verified the highly specific capture and detection of target molecules by the bacteria. Finally, through the coexpression of peptide-based capture ligands on the cell surface and fluorescent protein in the cytoplasm, we demonstrate an effective means of directly linking the fluorescence intensity to the density of capture ligands. Keywords: biosensor • protein arrays • cell-based detection • microfluidics • dielectrophoresis
Introduction The capability to analyze complex protein mixtures in a quantitative and parallel manner remains an important goal in biotechnology.1 Such analyses are typically carried out using 2-D gel electrophoresis or liquid chromatography in conjunction with tandem mass spectrometry.1,2 Though powerful, 2-D gel electrophoresis only provides semiquantitative data, and mass spectrometry requires the use of special isobaric reagents3 to obtain quantitative concentration information. In addition, a proteomic database of the organism is usually needed to validate the match of the mass spectra of the peptide fragments.2 * Author to whom correspondence should be addressed. E-mail: tsoh@ engineering.ucsb.edu; Phone: +1 (805) 893-7985; Fax: +1 (805) 893-8651 (H.T.S.); E-mail:
[email protected]; Phone: +1 (805) 893-2610; Fax: +1 (805) 893-4731 (P.S.D.). † California Nanosystems Institute and Biomolecular Science and Engineering Program. ‡ Department of Chemical Engineering. § Department of Mechanical Engineering. 10.1021/pr060193a CCC: $33.50
2006 American Chemical Society
The concept of protein microarrays, in which a large number of capture ligands are arrayed at high spatial densities, presents an attractive approach toward high throughput proteomic analysis.4-7 In contrast to nucleic acid-based microarrays, however, some major technical challenges remain. First, the availability of monoclonal antibodies, which are the most widely used capture reagents, is limited, and their production is costly and time-consuming.1 Second, the immobilization of the capture ligand to a solid surface and its effect on the biochemical activity and shelf life remains a concern.7 The use of genetically engineered cells, wherein the cell’s own machinery is harnessed to express the desired capture proteins, poses an interesting solution because the capture ligands for target molecules may be presented on the membrane surface in their natural conformation. Furthermore, the immobilization of these whole-cell affinity reagents in an electrically addressable array format could enable a novel microsystem for versatile and programmable proteomic analysis. Because microfabrication allows precise engineering of fluidic and electrical fields and their gradients, microfluidics lends itself to efficient electrokinetic manipulation of cells. In this work, we present the first demonstration of cell-based microfluidic protein sensors that are arrayed through the use of dielectrophoresis (DEP). DEP forces originate from the difference in conductivity and permittivity between the cell and its suspension medium, which induces a net polarization on the cell-medium interface, and it has been widely applied since its description in 1978.8 The polarity of the force (i.e., the attraction of a cell toward or its repulsion away from regions of high electric field gradients) and the frequency response are determined by the Clausius-Mossotti factor (fCM), which can be expressed as fCM(ω) )
/p(ω) - /m(ω) /p(ω) + 2/m(ω)
(1)
where /p and /m are the complex permittivities of the particle and the medium, respectively. The time-averaged DEP force on a spherical particle of radius rp is given by8 B F DEP ) 2πmr3pRe(fCM(ω))∇|Erms|2
(2)
where m is the permittivity of the medium, ω is the angular frequency, and Erms is the root-mean-square value of the applied electric field. In recent literature, negative DEP (repulsion of particles away from high electric field gradients) has Journal of Proteome Research 2006, 5, 3433-3437
3433
Published on Web 10/26/2006
Microfluidic Protein Detection
technical notes
been used for the spatial confinement of cells,9 and both positive DEP (attraction of particles toward high electric field gradients) and negative DEP have been utilized for the capture and fractionation of mammalian cells.10,11 In addition, the use of positive DEP has enabled the electrokinetic concentration of target cells for pathogen detection.12,13 In our work, Escherichia coli cells are selected and engineered such that each clone displays on its outer membrane surface thousands of copies of a unique peptide sequence, which functions as the capture ligand for the target molecule.14 Then different strains of such “probe cells” are spatially organized and immobilized onto a microfabricated electrode array using positive DEP. In this way, fluorescently labeled target molecules in the sample mixture bound to the probe cell surface can be detected directly through optical measurements.1 This approach is particularly useful because the biosynthetic machinery of the cell displays the capture ligands on the outer membrane, obviating the need for protein purification and their surface functionalization. Our results indicate that immobilized whole-cell microfluidic arrays provide an efficient and scalable platform for versatile detection of multiple analytes.
Materials and Methods Probe Cell Design. All experiments were performed with E. coli MC1061.14 Simultaneous expression of both green fluorescent protein (GFP) and an outer membrane protein containing the streptavidin-binding peptide CPX-SA1 was achieved through a bicistronic cassette under the control of an arabinoseinducible promoter.14,15 The SA1 peptide was fused to the extracellular N-terminus of circularly permuted outer membrane protein X (CPX) following the native outer membrane protein X signaling sequence and an ATST linker sequence.16 The gene encoding CPX-SA1 was digested with KpnI/HindIII and ligated into a similarly digested pBAD33 vector. The gene encoding the AlajGFP variant17 was likewise inserted between SacI and KpnI sites. A similar vector was constructed as a negative control containing the genes encoding AlajGFP and the wild-type CPX display scaffold. Cells were grown overnight at 37 °C in Luria-Bertani (LB) medium with 34 µg/mL chloramphenicol (CM) (Sigma, St. Louis, MO). The cells were subcultured at a 1:50 dilution for 2 h at 37 °C. Subsequently, the expression of AlajGFP and SAbinding peptide was induced by adding L-arabinose (0.02% final concentration) to the culture media at room temperature for 2 h. Finally, ∼108 cells were pelleted by centrifugation for 5 min at 5000 rpm followed by resuspension in 1 mL 0.05× phosphate-buffered saline (PBS; pH 7.4) prior to their DEP immobilization on the chip. Flow Cytometric Analysis of the Probe Cells. Flow cytometry (FACSAria, BD Biosciences, San Jose, CA) was used to confirm the affinity of the capture ligand for the target protein, which was streptavidin R-phycoerythrin (SAPE; Molecular Probes, Carlsbad, CA). The probe cells and negative control cells displaying wild-type CPX sequence, which do not bind to the target, were incubated with 66 nM SAPE in 0.05× PBS solution at room temperature for 30 min and washed once before their resuspension in 0.05× PBS. The cells were interrogated using a 20 mW (λ ) 488 nm) laser, and the fluorescence data were collected, compensated, and analyzed using BD FACSDiva Software (BD Biosciences, San Jose, CA). Microfluidic Device Design and Fabrication. PDMS soft lithography18 was used to fabricate the microfluidic device on a glass substrate. An array of Ti-Au electrodes was fabricated 3434
Journal of Proteome Research • Vol. 5, No. 12, 2006
Figure 1. The integrated microfluidic system: The probe cells, target proteins, and wash buffer are introduced sequentially through inlets (1), (2), and (3), respectively. Waste exits the chamber through common outlets (4). Probe cells are immobilized on the sensor elements. (Inset) Probe cells are immobilized on the interdigitated Ti-Au electrodes with 5 µm width, 5 µm spacing. PDMS walls separate adjacent columns of sensor elements.
with standard optical lithography and lift-off techniques on a 500-µm thick, 4-in borosilicate glass wafer (University Wafers, South Boston, MA). The interdigitated electrodes were 0.5 µm in thickness, 5 µm in width, and designed at a pitch of 10 µm. A computer-controlled CNC mill (Flashcut CNC, Menlo Park, CA) was used to drill fluidic reservoirs that were 750 µm diameter into the glass wafer. The negative-tone master mold of the microfluidic channels was fabricated on a 4-in silicon wafer using a deep reactive-ion-etcher (SLR-770, Plasmatherm, St. Petersburg, FL), which produced 20-µm deep channels. The height of the mold was confirmed using a surface profilometer. Subsequently, the surface of the silicon master was treated with a fluoro-carbon to prevent an irreversible bonding to PDMS during the replication step. The PDMS replicas of the silicon master mold were fabricated by applying a precursor (Sylgard 184, Dow-Corning Inc.; 10 part base resin:1 part curing agent) to the silicon master, followed by curing at 70 °C for 3 h. The PDMS replica and the electrode chip were cleaned in acetone and oxidized in a UV-ozone chamber prior to their bonding in a flip-chip aligner (Research Devices Inc., M8A, Piscataway, NJ). Covalent bonding sealed the two surfaces after a few hours, and no visible leaks were observed under the experimental
technical notes conditions. Finally, fabrication was completed by attaching brass eyelets to the device using an epoxy to form fluidic inlets/ outlets (Figure 1). The device was inserted into a custom cardedge connector for electrical interconnections. Two adjacent microchannels, one for the probe cells and the other for negative control cells, were used in the experiment. The length of each electrode was 200 µm, with the gap and width of 5 µm along the flow direction (Figure 1, inset). The area of each active element was 200 × 200 µm2, and the enclosed volume of the space above active element was approximately 0.8 nL. To prevent sample cross-contamination, each chip was discarded after a single usage. Cell Capture and Protein Detection. The experiment was performed at room temperature under continuous-flow conditions. Syringe pumps (PHD 2000, Harvard Apparatus, Holliston, MA) were utilized to deliver the probe cells, target molecules, and wash buffer (0.5× PBS) through inlets 1, 2, and 3, respectively (Figure 1). The waste stream was collected using a common outlet. Two microchannels were loaded with two cell types simultaneously: the probe cells expressing CPX-SA1 and the negative control cells expressing wild-type CPX. Other negative control cells expressing an irrelevant sequence (TCLRGPQQTRWCISR) were also tried, but they did not show significant binding to the target molecules, as measured by the resulting SAPE fluorescence amplitude (data not shown). The probe and negative control cells were immobilized in 10 min at a flow rate of 50 µL/hour and a capture voltage of 3.3 Vpeak-to-peak at 5 MHz. Following the probe cell immobilization, a solution containing the target molecules (330-nM SAPE in 0.05X PBS) was introduced into each channel and incubated for 10 min at a flow rate of 5 µL/hour. Subsequently, the channels were washed with 0.05× PBS for 10 min at 50 µL/ hour. The fluorescence analysis was performed by an epifluorescence microscope (DM 4000, LEICA Microsystems AG, Wetzlar, Germany) using 10× objective lens at 0.4 NA. The excitation source was a mercury arc lamp, and fluorescence imaging was performed with the narrow band-pass filter sets that were specifically designed for either FITC (excitation 480/ 30; emission 535/40; used for GFP imaging) or R-phycoerythrin (excitation 546/12; emission 585/40; used for SAPE). Fluorescence images were obtained with a CCD camera (ORCA-AG model; Hamamatsu Corporation, Bridgewater, NJ), using a software-controlled electronic shutter (C-imaging, Compix Inc, Cranberry Township, PA) to minimize photobleaching. The auto-fluorescence from PDMS was far below the detected fluorescence signals from the assay (data not shown).
Results DEP Immobilization of Probe Cells. The probe cells were attracted to and subsequently immobilized on the areas of high electric field gradients (i.e., in the vicinity of the electrodes) through positive DEP. The generation of the positive DEP force requires Re(fCM (ω)) > 0, as described in eq 2, and experimentally, this condition was achieved by decreasing the conductivity of the suspension medium to be less than that of the plasma membrane.11 To model the DEP immobilization conditions, the electrokinetic properties of E. coli were approximated by a dielectric multi-shell model19 using empirical parameters provided by Holzel.20 The frequency dependence of the DEP force on the cells in PBS of varying ionic strength was simulated (Figure 2A). Experimentally, it was determined that 0.05× PBS (σ ) 0.1 S/m at 20 °C) provided a suitable tradeoff between DEP capture efficiency and osmotic balance across the probe
Oh et al.
Figure 2. (A) The calculated Clausius-Mossotti factor of E. coli in PBS media of varying ionic strengths. (B) Epifluorescence micrographs taken at various time intervals of probe cell immobilization occurring on the active elements. The probe cells were suspended in 0.05X PBS at a concentration of 108/mL and injected at 100 µL/hour while a 5 MHz 3.3 Vpeak-to-peak bias was applied to the microelectrodes.
cell membrane. Simulation results for operation in 0.05× PBS was in agreement with the observed experimental trend of increasing capture force as we further increased the frequency up to 10 MHz. A rapid, efficient, and uniform immobilization of the probe cells was achieved at the operating frequency of 5 MHz (Figure 2B). As expected, the cells were initially attracted to the leading edge of each electrode where the electric field gradient is the highest and gradually filled up the entire active region in less than 10 min. The minimum threshold voltage for probe cell immobilization was ∼3.0 Vpeak-to-peak at a flow rate of 100 µL/ hour, and a nominal bias voltage of 3.3 Vpeak-to-peak was used for the immobilization. The electrolysis of the suspension medium did not occur at the operating conditions. Furthermore, it is apparent that (1) nonspecific binding of the probe cells to the microchannel surfaces is negligible, (2) DEP immobilization of the probe cells occurs only when the electrodes are powered, and (3) the density of probe cells across the powered electrode area is high. At the immobilization conditions, the influence of DEP forces on the target molecule was not observed (data not shown). Because the size of the target molecule (MW of SAPE: 353 kDa) is much smaller than that of the E. coli, it is hypothesized that the DEP force, which is proportional to the volume, becomes comparatively negligible.8 To prevent irreversible membrane damage and electropermeation of analytes into the cells due to the electric fields,11 the total assay time was limited to 30 min. However, the assay time may be significantly extended Journal of Proteome Research • Vol. 5, No. 12, 2006 3435
technical notes
Microfluidic Protein Detection
Figure 3. Flow cytometric analysis of probe and negative control cells. (A) After incubation with 66 nM SAPE and wash in 0.05X PBS, both probe and negative control cells show similar GFP intensity, which allows for a direct calibration of the number of immobilized cells. (B) In contrast, the 19-fold difference in red fluorescence between the probe and negative control cells confirms the specificity of the target binding.
as we have observed that the biochemical activity of the capture ligand is not affected by the viability of the host cell, as long as the integrity of the membrane structure is preserved (data not shown). Target Protein Detection. After incubation with the target molecule, streptavidin-phycoerythrin (SAPE), the red (SAPE) and green (GFP) fluorescence intensities of probe cells and negative control cells were analyzed with flow cytometry (Figure 3). Both cell types expressed similar levels of GFP, thus indicating similar expression levels of the displayed ligand, as the two are coexpressed through the bicistronic cassette (Figure 3A). The probe cells, however, demonstrate an approximately 19-fold increase in red fluorescence intensity, indicating effective and specific capture of the target molecules compared to that of the negative control cells in solution (Figure 3B). For on-chip analysis, the fact that both probe and negative control cells express similar levels of GFP provides a means of quantitatively measuring the number of cells immobilized to one electrode relative to another. This information can then be used to normalize the fluorescence of the target molecules to quantify the affinity and specificity. For example, by comparing the green fluorescence intensities at the two electrode areas (Figure 4A), it is apparent that there are approximately 2.3-fold more negative control cells immobilized (Figure 4A, bottom right) compared to the probe cells (Figure 4A, bottom left). When the target molecules are introduced, their red fluorescence can be normalized by green fluorescence of the cells for each sensor area. In the case of SAPE, a 23-fold increase in the red fluorescence (Figure 4B, bottom left) was observed 3436
Journal of Proteome Research • Vol. 5, No. 12, 2006
Figure 4. Fluorescence micrographs of on-chip target protein detection. (A) Green fluorescence levels confirm that nonspecific binding of the probe cells do not occur on unpowered electrodes (top) or on microchannel walls. The 2.3× higher GFP intensity from the negative control cells (bottom right) indicates that more cells were immobilized on that electrode compared to the probe cells on the bottom left electrode. (B) After a 10 min incubation and wash of the target molecule, a slight increase in nonspecific binding is visible on the unpowered electrode (top) and microchannel walls as indicated by the background red fluorescence. However, a 10-fold higher red fluorescence signal is detected from the probe cells (bottom left) compared to the negative control cells, verifying the efficiency of the assay.
compared to that of the negative control cells (Figure 4B, bottom right) after the normalization by their respective green fluorescence. Figure 4B further confirms the minimal nonspecific binding of the target molecules to the device as well as to the negative control cells.
Discussion and Conclusions Peptide arrays constructed using peptide-displaying bacteriophage particles have demonstrated utility for diagnostic proteome analysis.21 Likewise, arrays based upon whole-cell bacterial reagents would allow rapid reagent discovery and affinity optimization,22 and subsequent large-scale reagent production would be orders of magnitude less expensive than monoclonal antibody production. Although some proteins in the plasma proteome are known to bind bacterial surfaces, bacteria that do not display peptides can be used as separation reagents to remove these interfering species prior to sample analysis (unpublished data).
technical notes The uniform dielectrophoretic patterning of probe cells in microfluidic devices provides a useful means of quantitative proteomic analysis with inexpensive whole-cell capture reagents for a variety of diagnostic applications. The use of wholecell reagents enables expression of multiple copies of the capture ligand in its native conformation; each E. coli probe cell displays roughly 5000 peptides acting as capture ligands on its outer surface [unpublished]. Combined with the fact that up to ∼20 000 E. coli cells (∼1 × 2 µm2 cross-section) can fit into a 200 × 200 µm2 electrode area in a single monolayer, this allows efficient immobilization of over 2 × 108 capture ligands to be displayed in a compact, electrically addressable area. The actual number of displayed ligands may be even higher, as the formation of multiple layer stacks of E. coli cells have been observed (data not shown). In the current implementation, we anticipate that the dimension of a single active element may be reduced to ∼10 × 10 µm2 without sacrificing the reliability of the assay. The design of probe cells that coexpress GFP and the capture ligand through a bicistronic expression vector allows a direct and proportional relationship between the level of green fluorescence intensity and the capture ligand density.23 This direct linkage is particularly useful because it allows the calibration of the fluorescence signal from the target protein since it can be normalized by the number of available capture ligands. Because each array element contains ∼108 capture ligands with high affinity for the target proteins (KD ∼ 1-10 nM14), the speed of the assay is mainly limited by the diffusion of the target molecules to immobilized probe cells. Microfluidic systems, in which the spatial dimensions are miniaturized, offer a significant reduction in diffusion distance and thus provides a shorter incubation time as (time) ≈ (distance)2/(2 × diffusivity).24 In conclusion, we demonstrate a continuous-flow protein detection system that utilizes peptide-displaying bacteria as capture reagents. This work represents an important step toward the application of inexpensive, self-renewing, wholecell reagents as sensor elements, and such approach could be extended toward proteins that are difficult to purify and immobilize, such as transmembrane proteins, which could be overexpressed in an appropriate host cell in their native environment and conformation. This method may be applicable to multiple affinity ligand display methods, including bacteriophage25 and yeast26 display technologies. Finally, the extension of this work into larger arrays will require suitable means to multiplex, deliver, and capture the appropriate bacterial clones to the electrodes. Previously, large microfluidic circuits for multiplexing and delivery have been demonstrated using elastomeric valves,27 and scalable architectures for electrically addressing DEP trap arrays have been recently reported.28 By combining such existing techniques, it would be feasible to extend our methodology toward operation of larger detection arrays.
Acknowledgment. We thank the financial support from ONR-Young Investigator Program (N000140410456), AROInstitute for Collaborative Biotechnologies (DAAD1903D004),
Oh et al.
US Army grant (W911NF-0510083), DARPA/DMEA-CNID grant (H94003-05-2-0503), and gift funds from Samsung Advanced Institute of Technology. We thank Paul Bessette, Eric Lagally, and Sandra Hu for helpful discussions. Microfabrication was carried out in the Nanofabrication Facility at UC Santa Barbara.
References (1) Kodadek, T. Chem. Biol. 2001, 8, 105-115. (2) Hanash, S. Nature 2003, 422, 226-232. (3) Ross, P. L.; Huang, Y. N.; Marchese, J. N.; Williamson, B.; Parker, K.; Hattan, S.; Khainovski, N.; Pillai, S.; Dey, S.; Daniels, S.; Purkayastha, S.; Juhasz, P.; Martin, S.; Bartlet-Jones, M.; He, F.; Jacobson, A.; Pappin, D. J. Mol. Cell. Proteomics 2004, 3, 115469. (4) MacBeath, G.; Schreiber, S. L. Science 2000, 289, 1760-1763. (5) Zhu, H.; Klemic, J. F.; Chang, S.; Bertone, P.; Casamayor, A.; Klemic, K. G.; Smith, D.; Gerstein, M.; Reed, M. A.; Snyder, M. Nat. Genet. 2000, 26, 283-289. (6) Ramachandran, N.; Hainsworth, E.; Bhullar, B.; Eisenstein, S.; Rosen, B.; Lau, A. Y.; Walter, J. C.; LaBaer, J. Science 2004, 305, 86-90. (7) LaBaer, J.; Ramachandran, N. Curr. Opin. Chem. Biol. 2005, 9, 14-9. (8) Pohl, H. A. Dielectrophoresis: The Behavior of Neutral Matter in Nonuniform Electric Fields; Cambridge University Press: Cambridge, UK, 1978. (9) Muller, T.; Pfennig, A.; Klein, P.; Gradl, G.; Jager, M.; Schnelle, T. IEEE Eng. Med. Biol. Mag. 2003, 22, 51-61. (10) Becker, F. F.; Wang, X. B.; Huang, Y.; Pethig, R.; Vykoukal, J.; Gascoyne, P. R. Proc. Nat. Acad. Sci. U.S.A. 1995, 92, 860-4. (11) Gascoyne, P. R. C.; Vykoukal, J. V. Proc. IEEE 2004, 92, 22-42. (12) Huang, Y.; Ewalt, K. L.; Tirado, M.; Haigis, T. R.; Forster, A.; Ackley, D.; Heller, M. J.; O’Connell, J. P.; Krihak, M. Anal. Chem. 2001, 73, 1549-1559. (13) Lagally, E. T.; Lee, S. H.; Soh, H. T. Lab on a Chip 2005, 5, 10531058. (14) Bessette, P. H.; Rice, J. J.; Daugherty, P. S. Protein Eng. Des. Sel. 2004, 17, 731-739. (15) Guzman, L. M.; Belin, D.; Carson, M. J.; Beckwith, J. J. Bacteriol. 1995, 177, 4121-30. (16) Rice, J. J.; Schohn, A.; Bessette, P. H.; Boulware, K. T.; Daugherty, P. S. Bacterial display using circularly permuted outer membrane protein OmpX yields high affinity peptide ligands. Protein Sci., 2006, 15 (4), 825-836. (17) Bessette, P. H.; Daugherty, P. S. Biotechnol. Prog. 2004, 20, 963967. (18) Xia, Y. N.; Whitesides, G. M. Angew. Chem. Int. Ed. 1998, 37, 551575. (19) Chan, K. L.; Gascoyne, P. R. C.; Becker, F. F.; Pethig, R. Biochim. Biophys. Acta 1997, 1349, 182-196. (20) Holzel, R. Biochim. Biophys. Acta Mol. Cell Res. 1999, 1450, 5360. (21) Wang, X.; Yu, J.; Sreekumar, A.; Varambally, S.; Shen, R.; Giacherio, D.; Mehra, R.; Montie, J. E.; Pienta, K. J.; Sanda, M. G.; Kantoff, P. W.; Rubin, M. A.; Wei, J. T.; Ghosh, D.; Chinnaiyan, A. M. N. Engl. J. Med. 2005, 353, 1224-35. (22) Rice, J. J.; Schohn, A.; Bessette, P. H.; Boulware, K. T.; Daugherty, P. S. Protein Sci. 2006, 15, 825-36. (23) Dane, K. Y.; Chan, L. A.; Rice, J. J.; Daugherty, P. S. J. Immunol. Methods 2006, 309, 120-129. (24) Sia, S. K.; Whitesides, G. M. Electrophoresis 2003, 24, 3563-3576. (25) Smith, G. P. Science 1985, 228, 1315-7. (26) Boder, E. T.; Wittrup, K. D. Nat. Biotechnol. 1997, 15, 553-557. (27) Thorsen, T.; Maerkl, S. J.; Quake, S. R. Science 2002, 298, 580584. (28) Taff, B. M.; Voldman, J. Anal. Chem. 2005, 77, 7976-7983.
PR060193A
Journal of Proteome Research • Vol. 5, No. 12, 2006 3437
