Protein Sizing on a Microchip - American Chemical Society

Feb 8, 2001 - Caliper Technologies Corp., 605 Fairchild Drive, Mountain View, ... samples, with sizing accuracy better than 5% and high .... was recov...
0 downloads 0 Views 175KB Size
Download PDF
Anal. Chem. 2001, 73, 1207-1212

Protein Sizing on a Microchip Luc Bousse,* Stephane Mouradian, Abdel Minalla, Herman Yee, Kathi Williams, and Robert Dubrow

Caliper Technologies Corp., 605 Fairchild Drive, Mountain View, California 94043-2234

The current standard method for protein sizing, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), is a labor-intensive technique that has not substantially changed in the 30 years since its introduction.1-3 Much work has been done to develop more automated and instrumentally based methods such as capillary gel electrophoresis (SDS-CGE). The first reports of SDS-CGE for protein sizing used UV detection and both cross-linked4,5 and non-cross-linked6-13 polymeric sieving

matrixes. These CE-based methods did not displace SDS-PAGE as the most commonly used protein sizing technique because they were still relatively slow (10-40 min for each sample, or 2-8 h for the same number of samples as on a gel) and not highly sensitive. To improve the detection limits, several authors14-16 used fluorescent labeling of proteins to achieve detection limits considerably better than those obtained with UV detection. The labeling process itself, however, used protein solutions at high concentration, and therefore, this detection limit does not represent the true assay detection limit.17 In addition, this covalent labeling process introduces several other problems. First, the process is time-consuming and cumbersome and can require purification steps after labeling. Second, the number of labels attached to the protein is variable, which will broaden out the peaks in separations. Finally, most labeling chemistries depend on lysine residues, and therefore, the labeling efficiency depends on the number of lysines present, which can vary considerably in different proteins. Recently, Pinto et al. achieved high assay sensitivities using a sheath flow detector for protein separations in the submicellar regime,17 but a precolumn covalent labeling reaction is still required. Another approach is to avoid the complications of covalent labeling altogether, by using fluoresecent dyes that bind to SDSprotein complexes.18 In the work described by Harvey et al.,19 this resulted in high sensitivities, but also poor separations and a high fluorescent background, since these dyes also fluoresce when bound to SDS micelles without protein present. In this work, we present a new method for protein sizing based on a microfabricated and miniaturized analytical device on a glass chip that automates and integrates the required assay steps. Many microfabricated biochemical analysis systems have been reported recently, capable of performing rapid separations,20-23 with ap-

* Corresponding author: (e-mail) [email protected]; (phone) (650) 623-0712; (fax) (650) 623-0500. (1) Shapiro, A. L.; Vinuela, E.; Maizel, J. V., Jr. Biochem. Biophys. Res. Commun. 1967, 28, 815-20. (2) Laemmli, U. K. Nature 1970, 227, 680-5. (3) Neville, D. M., Jr. J. Biol. Chem. 1971, 246, 6328-34. (4) Cohen, A. S.; Karger, B. L. J. Chromatogr. 1987, 397, 409-17. (5) Hjerten, S.; Elenbring, K.; Kilar, F.; Liao, J. L.; Chen, A. J.; Siebert, C. J.; Zhu, M. D. J. Chromatogr. 1987, 403, 47-61. (6) Zhu, M.; Hanson, D. L.; Burd, S.; Garrison, F. J. Chromatogr. 1989, 480, 311-9. (7) Widhalm, A.; Schwer, C.; Blaas, D.; Kenndler, E. J. Chromatogr. 1991, 549, 446-51. (8) Wu, D.; Regnier, F. E. J. Chromatogr. 1992, 608, 349-56. (9) Ganzler, K.; Greve, K. S.; Cohen, A. S.; Karger, B. L.; Guttman, A.; Cooke, N. C. Anal. Chem. 1992, 64, 2665-71. (10) Werner, W. E.; Demorest, D. M.; Stevens, J.; Wiktorowicz, J. E. Anal. Biochem. 1993, 212, 253-8. (11) Nakatani, M.; Shibukawa, A.; Nakagawa, T. Biol. Pharm. Bull. 1993, 16, 1185-8.

(12) Karim, M. R.; Janson, J. C.; Takagi, T. Electrophoresis 1994, 15, 1531-4. (13) Guttman, A.; Nolan, J.; Cooke, N. J. Chromatogr. 1993, 632, 171-5. (14) Gump, E. L.; Monnig, C. A. J. Chromatogr., A 1995, 715, 167-77. (15) Wise, E. T.; Singh, N.; Hogan, B. L. J. Chromatogr., A 1996, 746, 109-21. (16) Hunt, G.; Nashabeh, W. Anal. Chem. 1999, 71, 2390-7. (17) Pinto, D. M.; Arriaga, E. A.; Craig, D.; Angelova, J.; Sharma, N.; Ahmadzadeh, H.; Dovichi, N. J.; Boulet, C. A. Anal. Chem. 1997, 69, 3015-21. (18) Steinberg, T. H.; Jones, L. J.; Haugland, R. P.; Singer, V. L. Anal. Biochem. 1996, 239, 223-37. (19) Harvey, M. D.; Bandilla, D.; Banks, P. R. Electrophoresis 1998, 19, 216974. (20) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895-7. (21) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114-8. (22) Woolley, A. T.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 1134852. (23) Bousse, L.; Dubrow, B.; Ulfelder, K. In Micro Total Analysis Systems ′98; Harrison, D. J., van den Berg, A., Eds.; Kluwer Academic Publishers: Dordrecht, 1998; pp 271-5.

We have developed a microfabricated analytical device on a glass chip that performs a protein sizing assay, by integrating the required separation, staining, virtual destaining, and detection steps. To obtain a universal noncovalent fluorescent labeling method, we have combined on-chip dye staining with a novel electrophoretic dilution step. Denatured protein-sodium dodecyl sulfate (SDS) complexes are loaded on a chip and bind a fluorescent dye as the separation begins. At the end of the separation channel, an intersection is used to dilute the SDS below its critical micelle concentration before the detection point. This strongly reduces the background due to dye molecules bound to SDS micelles and also increases the peak amplitude by 1 order of magnitude. Both the on-chip staining and SDS dilution steps occur in the 100-ms time scale and are ∼104 times faster than their conventional counterparts in SDS-PAGE. This represents a much greater speed increase due to microfabrication than has been obtained in other assay steps such as electrophoretic separations. We have designed and tested a microchip capable of sequentially analyzing 11 different samples, with sizing accuracy better than 5% and high sensitivity (30 nM for carbonic anhydrase).

10.1021/ac0012492 CCC: $20.00 Published on Web 02/08/2001

© 2001 American Chemical Society

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001 1207

Figure 1. Chip design. The location of the wells is shown in light gray. Well D4 is the SDS dilution well and is connected to both sides of the dilution intersection. Wells A4 and C4 are the separation buffer and waste wells, and B4 and D3 are used as load wells. All other wells contain samples.

plications mostly centered on DNA separations,24,25 integrated immunoassays on a chip,26 and integrated amplification and sizing of DNA.27 Recently, a protein sizing assay on a microchip using precolumn covalently labeled protein samples was described.28 Our assay combines the best features of SDS-PAGE gels (sensitivity, unpretreated multiple samples, detection with staining/destaining) with the speed and immediate quantitation of microchip separations. To achieve this, we use noncovalently bound fluorescent dyes that bind to the SDS-protein complexes on the chip. To overcome the problem of high fluorescent background, we make use of the possibility of integrating multiple analytical operations on a microchip by adding an additional channel intersection in the microfluidic network. This channel is used for a novel SDS dilution step which not only strongly reduces the background fluorescence from SDS-dye complexes but also increases the signal amplitude by 1 order of magnitude. EXPERIMENTAL SECTION Microchip Fabrication. The chips were made from soda lime glass, with photolithographically defined wet-etched channels, fabricated using procedures reported previously.23 The channel width on the mask was 10 µm, which after the etch resulted in channels 13 µm deep and 36 µm wide. A lid with 2-mm-diameter holes drilled in it was thermally bonded to the lower plate to complete the channels. After dicing, the size of each chip is ∼17.5 mm square. Figure 1 shows the chip design with which most of the data presented here was obtained. (24) Simpson, P. C.; Roach, D.; Woolley, A. T.; Thorsen, T.; Johnston, R.; Sensabaugh, G. F.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 2256-61. (25) Shi, Y.; Simpson, P. C.; Scherer, J. R.; Wexler, D.; Skibola, C.; Smith, M. T.; Mathies, R. A. Anal. Chem. 1999, 71, 5354-61. (26) Chiem, N. H.; Harrison, D. J. Clin. Chem. 1998, 44, 591-8. (27) Burns, M. A.; Johnson, B. N.; Brahmasandra, S. N.; Handique, K.; Webster, J. R.; Krishnan, M.; Sammarco, T. S.; Man, P. M.; Jones, D.; Heldsinger, D.; Mastrangelo, C. H.; Burke, D. T. Science 1998, 282, 484-7. (28) Yao, S.; Anex, D. S.; Caldwell, W. B.; Arnold, D. W.; Smith, K. B.; Schultz, P. G. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 5372-7.

1208

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

Sample Preparation. The BioRad protein ladder (BioRad No. 148-2015) was diluted 1:10 in 1× PBS buffer (Irvine Scientific No. 9240). The BRL protein ladder (GibcoBRL No. 16001-018) was reconstituted according to the manufacturer’s instruction. The BioRad and BRL ladder contain similar bands except for an additional band at 116 kDa for the BioRad ladder. For the sensitivity study (Figure 7), a carbonic anhydrase sample (Sigma No. C2273) was used. Gel electrophoresis was performed using precast gels (Invitrogen Novex No. EC6025, power supply No. EI8600). Protein samples were denatured by mixing in a 2:1 ratio with denaturing buffer (4% SDS, 290 µg/mL myosin internal marker, and 3% β-mercaptoethanol) and heated to 100 °C for 5 min. Following denaturation, the samples for the experiments shown in Figures 4 and 5 were diluted 1:10 in deionized water. Samples for Figures 6 and 7 were diluted 1:15 in a 10% solution of lower marker dye (Agilent Technologies No. 5065-4430, excitation/ emission 650/680 nm, chemical name not available) in deionized water. Cell lysate samples were obtained following conventional protocols29 by centrifugation at 12000g for 10 min of 1 mL of Escherichia coli cell culture. The supernatant was removed, and 0.5 mL of 50 mM Tris-HCl (pH 7.4) was added. The washed pellet was recovered by centrifugation at 12000g for 10 min. After removing the supernatant, 25 µL of the above-described denaturing buffer was added to the pellet and the resultant mixture heated for 5 min at 100 °C. Chip Preparation. For all separation experiments, every channel in the glass chip was filled by loading sieving matrix into well A4 (See Figure 1) and applying pressure for 1 min. The sieving matrix used was a polymer based on polydimethylacrylamide30 at 3.25% in a Tris-Tricine buffer at pH 7.6 (120 mM Tricine, 42 mM Tris), containing 0.25% SDS (8.7 mM final concentration) and 4 µM of the same dye used as a lower marker, henceforth to be called the Agilent dye. Wells A4, B4, C4, and D3 were filled with this solution. The SDS dilution well (D4) contains only the sieving matrix and the Tris-Tricine buffer. Protein samples were applied to all remaining wells on the chip: A1-A3, B1-B3, C1-C3, D1, and D2. For the microphotographs of proteins on the chip (Figures 2 and 3), the same sieving matrix was used, except for a different dye, namely, SYPRO Orange (Molecular Probes No. S6650, chemical name and stock concentration not available) at 200:1 dilution in deionized water. This dye is excited at 300 nm and emits at 470 nm, which facilitates visualization of the process compared to the Agilent dye. The protein sample for these figures was 2 mg/mL bovine serum albumin (BSA). Measurement Instrument. Separation and detection was done in an Agilent 2100 Bioanalyzer instrument (Agilent Technologies G2940AA), which uses epifluorescent detection with a 10-mW semiconductor laser that emits at 630 nm. The instrument also contains 16 individually programmable high-voltage power supplies. For the microphotographs of proteins on the chip, a microscopebased setup was used as described previously.23 Images were recorded on a Nikon Diaphot 200 inverted fluorescent microscope, using an intensified CCD camera (Hammamatsu C2400-60). (29) Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular Cloning, a Laboratory Manual, 2nd ed.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 1989. (30) Dubrow, R. S. U.S. patent 5,948,227, 1999.

Figure 2. Series of photographs showing the on-column staining process at the intersection of two 36-µm-wide channels. Image a shows the sample loading from right to left, immediately before a sample was injected into the separation channel. Images b-d were recorded at 0.1-s intervals after (a) and show the injection downward into the separation column. The binding of protein with the SYPRO dye present in the gel occurs as soon as the sample enters the separation channel.

Critical Micelle Concentration (cmc) Measurement Method. The cmc was determined by measuring the increased fluorescence as dye binds to SDS micelles, following Brito and Vaz.31 We used the same Agilent dye as was used for the separation experiments. These measurements were carried out on a Perkin-Elmer LS50B luminescence spectrometer. Two series of samples were prepared. The first series contained the 3.25% sieving matrix, the assay buffer, 4 µM Agilent dye, and an SDS concentration varying from 0 to 4.6 mM. The second series was similar, except that the sieving matrix was omitted. RESULTS AND DISCUSSION CMC Determinations. The cmc was found to be 1.2 mM without the sieving matrix and 1.7 mM with the sieving matrix present. These values agree with the literature,31-33 in which a range of 1-2 mM is found for the cmc of SDS in buffers with ionic strengths in the range of 100-200 mM. Chip Loading and On-Chip Dye Labeling. As described in the Experimental Section, labeling on the chip is achieved by adding staining dye directly to the matrix filling the channel. Both the SYPRO and the Agilent dye exhibit a fluorescent enhancement upon binding to SDS micelles or SDS-protein complexes. These dyes act as generic stains that appear to bind to the SDS coating the protein, rather than to any specific amino acids.18,34 The polydimethylacrylamide-based sieving matrix was observed to reduce electroosmotic flow (EOF) sufficiently that the (31) Brito, R. M.; Vaz, W. L. Anal. Biochem. 1986, 152, 250-5. (32) Chattopadhyay, A.; London, E. Anal. Biochem. 1984, 139, 408-12. (33) Neugebauer, J. M. Methods Enzymol. 1990, 182, 239-53. (34) Steinberg, T. H.; Haugland, R. P.; Singer, V. L. Anal. Biochem. 1996, 239, 238-45.

Figure 3. Photographs of the SDS dilution process at two dilution ratios. Conditions are the same as for Figure 2. Image a shows a dilution ratio of 4 and image b a ratio of 12. It can be seen that some fluorescence remains, even in (b) after the SDS is diluted to below its cmc.

chips did not need to be coated. We measured the EOF mobility in the assay buffer and sieving matrix using a neutral marker dye and found a value of 5 × 10-6 cm2/V‚s, which is ∼60 times lower than expected without the sieving matrix. A reduction in EOF caused by the sieving matrix was previously observed by Wu and Regnier,8 who ascribed it to a combination of viscosity increase at the interface and adsorption of the polymer to the capillary wall. As a result, the operation of the microchip assay does not involve significant flow of the solutions in the chip and is due to the electrophoretic movement of ions. This is the same situation that typically prevails in microchip DNA separations.22,23 In the following discussion, the word “current” will refer to the electrical current carried by ions in the channels, since there is no bulk flow of solution in the chip. The first operation in the microchip is loading the sample toward an intersection, where it is confined by electrical currents coming from both sides in the separation channel.35 Typically, we observe that the loading channels of the chip are quickly depleted of dye as they are replaced by sample buffer during loading to the intersection. The microscope photographs in Figure 2 show that initially the sample is dark because it is still unlabeled, but the edges are brighter at the intersection where the sample meets the confining electrical currents because SYPRO dye is present in the separation channel. When all protein components have reached this intersection, the fields are switched to inject a small plug of the sample into the separation channel. The sample proteins immediately become brightly fluorescent as they pick up dye present in the separation channel. Figure 2b shows that this staining occurs in less than 100 ms, and the brightness does not increase further as the plug travels down the channel. In comparison, staining of proteins in a gel generally takes a minimum of 1 h. Given the channel dimensions (36 µm wide and 13 µm deep), we estimate the volume of this sample plug to be 25 pL. During injection and separation, small electrical currents are applied in both sides of the loading (35) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-13.

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

1209

Figure 5. Electropherogram of the separation of a BioRad protein ladder in optimized conditions (separation field 274 V/cm, and dilution ratio of 9).

Figure 4. Data showing a series of separations of a BioRad protein ladder at various dilution ratios (DR) 1-14. The electric field ranged from 219 to 221 V/cm. The BioRad ladder preparation and the buffers are described in the Experimental Section. The total protein concentration is 0.4 mg/mL for eight fragments. Graph a shows the total fluorescence without baseline correction, and graph b shows the amplitude above baseline of the 29 kDa peak.

channel to pull the sample still present in that channel away from the separation channel, as described by Jacobson.35 This simultaneously creates a well-defined sample plug and avoids cross contamination by clearing the loading channel before the next sample is loaded. Separation and Postcolumn SDS Dilution. The next step is separation in the presence of SDS, which occurs at a field of 200-300 V/cm, in a channel 1.25 cm long. We have found that separations, especially of complex mixtures such as cell lysates, require SDS levels significantly above the cmc to achieve high efficiencies, presumably due to the need to maintain denaturation. Since the labeling dye fluoresces when bound to either SDS micelles or SDS-protein complexes, we expect a high background and low signal if we simply detect the fluorescence at the end of the separation channel, as has been noted by Harvey et al.19 The only way to eliminate this high background is to reduce the SDS concentration below the cmc at a second, postcolumn, intersection, as shown in Figure 3. Two channels on each side contain the polymeric separation matrix and the buffer, but not SDS or dye. The electrical current from those side channels confines the SDS in the separation matrix to a thin stream in the middle of the channel. The ratio of the electrical current after this intersection to the separation current is called the dilution ratio. In this case, a dilution is achieved without bulk liquid flow but through the electrical current carried by ions out of well D4. The material in 1210 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

the middle of the separation channel then diffuses out to the edge, as can be seen in Figure 3 for the fluorescent dye-protein-SDS complex, and thereby reduces the SDS concentration by the dilution ratio. The on-chip SDS dilution process can be analyzed in terms of the diffusion coefficient of free SDS molecules (6 × 10-6 cm2/s according to Orfi et al.36). After the dilution intersection, the electrical current, field, and velocity are multiplied by the dilution ratio. Since we use a symmetrical design at this intersection, the distance for SDS molecules to travel is at most half the channel width, i.e., 18 µm, which requires a diffusion time of 280 ms. In our design, the detector is placed 0.15 cm downstream of the dilution intersection. At a typical dilution ratio of 9 and a separation field of 260 V/cm, the time needed for a relatively small protein (14 kDa) to travel this distance is 332 ms. This means that there is sufficient time to reach a near-uniform SDS concentration for all protein components. The effect of the SDS dilution ratio on the protein separation data is shown in Figure 4. For this experiment and all remaining figures, the Agilent dye was used. At ratios from 1 (no dilution) to 7, peak amplitudes are below 20 units, compared to a background of 300-1200 units. As the dilution ratio increases, the background decreases gradually as the concentration of SDS micelles decreases. The peak amplitude, on the other hand, remains low until a dilution ratio of ∼8 is reached, where it rapidly increases by ∼1 order of magnitude. This transition occurs at an SDS concentration after dilution of ∼1 mM, which is a factor of ∼2 below the cmc. These results therefore suggest that as the SDS micelles are broken up before the detector is reached, the dye that was bound to them is now available to bind to the protein-SDS complexes. Dilution ratios higher than 10 provide only limited additional benefits, probably because they increase the velocity between the dilution intersection and the detection point and, thus, reduce the available time for the SDS to diffuse out from the middle of the channel. This SDS dilution step can be compared to the destaining in conventional SDS gels. In both cases, the result is to reduce the (36) Orfi, L.; Lin, M.; Larive, C. K. Anal. Chem. 1998, 70, 1339-45.

Figure 6. Separations on a multisample chip design. Gel-style display of 10 separations plus a ladder on this chip. The ladder lane (labeled L) contains a BRL protein ladder; total protein concentration is 1 mg/mL for the seven fragments. Lanes 1, 3, 5, 7, 9, and 10 contain replicates of the BioRad ladder; total protein concentration is 0.4 mg/mL for eight fragments. Lanes 2, 4, 6, and 8 contain replicates of an E. coli lysate. Each lysate lane contains ∼25 resolvable bands. In each lane, the top band at 200 kDa is the myosin upper marker spiked in the denaturing buffer while the first bottom band is the lower marker Agilent dye added after denaturation. The lower marker is detected as it is in excess over the amount of dye already contained in the matrix. The markers are lined up by software for sizing and quantitation. An extra band at ∼9 kDa is ascribed to the Agilent dye bound to SDS micelles. The presence of this extra band effectively reduces our sizing range to 9-200 kDa.

fluorescent background and allow the protein peaks to be revealed, and in that sense, the SDS dilution acts as a virtual destaining. One key difference is, however, that a gel destain removes unbound dye molecules by diffusion, whereas SDS dilution breaks up the surfactant micelles, thereby allowing more dye molecules to bind to the protein. This rearrangement leads to our unexpected finding that the peak amplitude increases rapidly when the SDS concentration is lowered to just below the cmc. If we compare the speed of a microchip SDS dilution (0.3 s) with the speed of a conventional SDS-PAGE gel destain (about 1 h), we see that the microscale operation is 104 times faster. This represents a much greater speed increase than is obtained with microscale separations, which are typically 10-100 times faster than conventional capillary electrophoresis. The speed of the SDS dilution on a microchip is critically dependent on the narrow channels, and systems using larger dimensions would be much slower and require long channels after the dilution intersection. It therefore appears that this type of postcolumn dilution in analytical systems is most practical in microscale systems. Separation of a Protein Ladder. An expanded electropherogram of a separation of the same protein ladder is shown in Figure 5, using optimized conditions. In addition to the main eight peaks,

Figure 7. Assay limit of detection of carbonic anhydrase as a function of salt concentration in the sample. All limits of detection are expressed as three standard deviations above background. The concentrations are those of the protein before it is denatured and then diluted further. The limits of detection in terms of the actual protein concentration in the wells on the chip are thus 15 times lower. Gel data were obtained with Novex Gels and Coomassie Brilliant Blue stain. The gels were imaged with a Molecular Dynamics Personal Densitometer SI and scans analyzed with ImageQuant software.

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

1211

many other small peaks representing impurities are also clearly resolved. The separation efficiency for the larger peaks ranges from 7 × 104 to 1.2 × 105 theoretical plates, which is somewhat higher than the range typically seen in conventional SDSCGE.4,14,15 This efficiency can also be expressed as 107 plates/m, or 3750 plates/s, and in these measures of efficiency, this assay is 1-2 orders of magnitude better than SDS-CGE, and also ∼1 order of magnitude better than previous reports of on-chip protein separations in the presence of SDS.28 The main factor enabling these rapid high-efficiency separations is the use of a very small sample plug, on the order of the channel width.37 Other contributing factors are the use of sufficiently high SDS levels to ensure complete denaturation and the absence of covalent labeling reactions, which typically broaden peaks by inhomogeneous labeling. Chip Design for Multiple Samples. To accommodate multiple samples, we have designed the microchip shown in Figure 1. It has 16 wells in a 4 by 4 array, 5 of which are needed for the assay (2 load wells, separation waste, separation buffer, and dilution), leaving 11 available for samples. Since our goal is to be comparable with SDS-PAGE in terms of the number of samples than can be analyzed, we chose a design that optimizes that number. As a consequence, this design requires the use of a single separation channel in which samples are analyzed serially. The schemes for parallel separation that have been proposed24,25 would allow no more than 5 samples with 16 wells, given the need for SDS dilution in each channel. Our design places the wells in a densely packed array and routes all channels around those wells. Figure 6 shows our typical data output in a “gel view” mode, which displays the peak data on a gray scale to mimic a gel appearance. Bands at 6, 9, and 200 kDa (myosin) are lane markers that are used as reference peaks for automated sizing and normalization of peak areas and correct for fluctuations in the separation field. In these data, one protein ladder is used as a reference by the software to size another set of standards that have an additional band at 116 kDa (β-galactosidase). Accuracy of the automated sizing is ∼5% between 14 and 200 kDa, with respect to the expected sizes. To assess the usefulness of this technique, we have tested our assay with protein samples from various origins. Cell lysates (Figure 6) containing detergents such as SDS, Chaps, and Triton were successfully analyzed as well as HPLC column fractions in a variety of buffers and salt concentrations. In extensive comparisons between our microchip data and SDS-PAGE, we have found that the band patterns for lysates obtained with both techniques were similar, with major bands detected at the same molecular weight in PAGE and on the chip (data not shown). Assay Sensitivity. The sensitivity we observe is dependent on the ionic strength of the protein sample, as is also the case with SDS-CGE. This is due to sample stacking: the concentration of the protein sample entering the chip is essentially multiplied by the ratio of the conductivity in the chip and the conductivity in the sample.38,39 In SDS-PAGE, it is standard to use a stacking gel, which greatly reduces this sensitivity to the sample ionic strength. We have measured detection limits by measuring signal levels and background noise in regions where no protein peaks were present. The same procedure was applied to the same (37) Bousse, L.; Cohen, C.; Nikiforov, T.; Chow, A.; Kopf-Sill, A. R.; Dubrow, R.; Parce, J. W. Annu. Rev. Biophys. Biomol. Struct. 2000, 29, 155-81. (38) Chien, R.-L.; Burgi, D. S. Anal. Chem. 1992, 64, 489A-96A. (39) Chien, R.-L.; Helmer, J. C. Anal. Chem. 1991, 63, 1354-62.

1212

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

samples run in SDS-PAGE gels, which were then scanned to produce a line plot. The resulting detection limits are shown in Figure 7. At low sample ionic strength, the detection limit of the microchip assay (at 3 standard deviations) for carbonic anhydrase is 30 nM. With 150 and 500 mM NaCl added to the sample, the detection limit goes up to 80 and 200 nM, respectively. The detection limit for carbonic anhydrase in SDS-PAGE was found to be 140 nM, and only weakly dependent on the salt concentration in the sample. We conclude that, at low salt levels, the microchip assay is 4-5 times more sensitive than a standard Coomassie Bluestained gel, and at physiological salt levels, the chip is about twice as sensitive as the gel. It should be noted that the assay detection limit reported here does not include the sample dilution by a factor of 15 in water before the sample is loaded in the chip. Thus, the detection limits in terms of the concentrations actually in the well are 15 times lower, and these values are the ones that should be compared to the various reported sensitivities in SDS-CGE. In this assay, it is to be expected that, at some high concentration of protein, insufficient dye will be present to bind in a quantitative fashion to the SDS-protein complexes. Accordingly, we have tested the linearity of the quantitation and found that the assay is linear from the lower detection limit up to 2000 µg/ mL BSA. In experiments where we loaded up to 4000 µg/mL BSA, we observed that the linearity is gradually lost past 2000 µg/mL. Nevertheless, a linear protein assay from about 10 to 2000 µg/ mL covers well the range of concentrations typically used in conventional methods such as gel electrophoresis or Bradford quantitation assays. CONCLUSION This assay is an illustration of the power of integrating multiple operations on a chip. It allows a novel procedure for staining and diluting protein samples on the fly, leading to a procedure as general as gel electrophoresis but orders of magnitude faster and more automated. Interestingly, it is in the staining and SDS dilution operations, which are 4 orders of magnitude faster than their conventional counterparts, that the greatest speed gains are realized. We can anticipate many possible improvements in this assay. For instance, combining it with nanoliter-level sample accession methods,40 on-chip dilution to add the markers,41 and some automation will allow a very large number of samples to be analyzed without user intervention. In addition, using parallel separations will allow sample throughput to be increased. ACKNOWLEDGMENT We thank J. W. Parce, Calvin Chow, Carlton Brooks, Mac McReynolds, George Derbalian, Richard Haugland, Vicky Singer, Monika Dittmann, and Klaus Witt for their assistance.

Received for review December 17, 2000.

October

23,

2000.

Accepted

AC0012492 (40) Sundberg, S. A. Curr. Opin. Biotechnol. 2000, 11, 47-53. (41) Chow, A.; Kopf-Sill, A.; Nikiforov, T.; Zhou, A.; Coffin, J.; Wada, G.; Alajoki, L.; Yurkovetsky, Y.; Sundberg, S.; Parce, J. W. In Micro Total Analysis Systems 2000; van den Berg, A., Olthuis, W., Bergveld, P., Eds.; Kluwer Academic Publishers: Dordrecht, 2000; pp 489-92.