Pressure Injection on a Valved Microdevice for Electrophoretic

The data presented show the ability to alternate between injection types and pressure-inject a representative sample volume after a .... Xing Yue (Lar...
0 downloads 0 Views 187KB Size
Anal. Chem. 2005, 77, 3637-3643

Pressure Injection on a Valved Microdevice for Electrophoretic Analysis of Submicroliter Samples James M. Karlinsey,† Jennifer Monahan,† Daniel J. Marchiarullo,† Jerome P. Ferrance,† and James P. Landers*,†,‡

Department of Chemistry, University of Virginia, Charlottesville, Virginia 22904, and Department of Pathology, University of Virginia Health Science Center, Charlottesville, Virginia 22903

A recent report describes a reversible valve that can be used in series to achieve diaphragm pumping on chip (Grover, W. H.; Skelley, A. M.; Liu, C. N.; Lagally, E. T.; and Mathies, R. A. Sens. Actuators, B 2003, 89, 315323). Here, the functionality of an integrated diaphragm pump on a hybrid PDMS-glass microchip to perform pressure injections for electrophoretic separations is demonstrated. A chip design that can perform both pressure and electrokinetic (EK) injection is described, and a mixture of fluorescein and ROX dyes in borate buffer is utilized as a model sample system. Multiple electrophoretic separations of sample injected with pressure and voltage are compared. Over multiple EK injections, an electrophoretic bias is observed and the injected analytes are not representative of the sample, with the peak area ratio changing 20% after 20 runs. Over multiple pressure injections, however, the sample composition is maintained, with a 3.6% CV over 20 runs. The data presented show the ability to alternate between injection types and pressure-inject a representative sample volume after a bias has already been observed with multiple EK injections. Multiple pressure injections have been performed on sample volumes as low as 500 nL while maintaining sample composition, supporting its use in integrated systems for small-volume sampling. Much of the analytical technology that has been developed for capillary electrophoresis separations is being transferred to a microchip format in an effort to realize a micro total analysis system. While there has been much success in this endeavor,1-3 with improvements coming in speed and reduced sample volumes, there are still limitations. Most notable is the limitation on the type of sample injection that can be achieved on a microchip. Injection into a capillary is either performed by electrokinetic or hydrostatic means, where sample is hydrostatically introduced by suction, pressure, or gravity. On a microchip, however, injections * Tp whom correspondence should be addressed. Phone: (434)-243-8658. Fax: (434)-924-3048. E-mail: [email protected]. † University of Virginia. ‡ University of Virginia Health Science Center. (1) Reyes, D. R.; Iossifidis, D.; Auroux, P.-A.; Manz, A. Anal. Chem. 2002, 74, 2623-2636. (2) Auroux, P.-A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2002, 74, 2637-2652. (3) Vilkner, T.; Janasek, D.; Manz, A. Anal. Chem. 2004, 76, 3373-3386. 10.1021/ac048115z CCC: $30.25 Published on Web 04/16/2005

© 2005 American Chemical Society

are predominantly electrokinetic. This is in part because of the difficulty in isolating the flow streams in multiple channels, but also because electrokinetic injections have proved to be adequate and effective for many samples. Electrokinetic injection techniques have been developed for “T”,4 “double-T”,5 and “cross-T”6,7 injectors. With a floating injection,6 the sample is pumped across the separation channel, allowing variable volume as a result of diffusion. For a gated injection,7 sample is continuously pumped with EOF. The sample is brought alongside the separation channel and allowed to diffuse into the channel for a variable-volume injection. In a pinched injection,6 buffer is pumped from the separation channel along with the sample channel, confining the sample plug for a constantvolume injection. However, each of these injections has the potential to suffer from some type of electophoretic bias. For analytes traveling with EOF, an electrokinetic bias has been illustrated on capillary8 where high-mobility analytes are preferentially introduced into the separation channel. The amount of sample injected is dependent upon the concentration of the analyte, the cross-sectional area of the capillary or channel, and the length of the sample zone. Since the length of the zone is equal to the total mobility multiplied by the electric field, EOF pumping favors higher mobility compounds. For a gated injection scheme, a transradial electrokinetic selection bias has also been described that favors analytes having lower electrophoretic mobilities.9 Sample bias on chips may be overcome by using longer injection times. However this increases the overall analysis time and accelerates sample depletion, thus limiting the number of injections that can be performed. The sample matrix and volume in turn become limiting factors as the analyte spends more time exposed to the electric field. In some solutions, e.g., those of high ionic strength, reproducible electrokinetic (EK) injections are not even feasible.10 Beyond this, applying a high voltage to the sample (4) Harrison, D. J.; Manz, A.; Fan, Z. H.; Ludi, H.; Widmer, H. M. Anal. Chem. 1992, 64, 1926-1932. (5) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 26372642. (6) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (7) Jacobson, S. C.; Hergenroder, R.; Moore, A. W.; Ramsey, J. M. Anal. Chem. 1994, 66, 4127-4132. (8) Huang, X.; Gordon, M. J.; Zare, R. N. Anal. Chem. 1988, 60, 375-377. (9) Slentz, B. E.; Penner, N. A.; Regnier, F. Anal. Chem. 2002, 74, 4835-4840. (10) Satow, T.; Machida, A.; Funakushi, I. C.; Palmieri, R. J. High Resolut. Chromatogr. 1991, 14, 276.

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005 3637

reservoir can induce electrolysis of the solution.11-13 Such electrolysis can have a significant effect on the buffering capacity and pH of the solution. While the majority of microchip electrophoresis techniques are based on CE, the capillary format still offers the advantage of flexibility, capable of either EK or pressure (PR) injections. For a microdevice to be capable of pressure injections, it must have the ability to pump fluid in a controlled manner. Recent literature has presented two methods for valving-off fluid microchannels using poly(dimethylsiloxane) (PDMS). The first method, developed by the Quake group, consists of multiple PDMS layers used to build active systems of valves and pumps.14,15 While versatile, these valves are designed in the normally open state, requiring pressure to control fluid contents. The second valving technique, developed by the Mathies group, consists of a PDMS membrane that can be incorporated into a glass device. These valves operate in a normally closed state and can be used to valve and pump with fairly low actuation pressures.16,17 The microdevice presented here is based on the Mathies pumping design and is capable of both EK and pressure injections. In short, three normally closed valves are arranged in series; opening them allows fluid to flow into a chamber of known volume. The pattern in which the valves are opened and closed determines the direction of flow, and the volume of the center valve defines the volume displaced per cycle. This pumping mechanism has been shown in three-and four-layer hybrid microdevices where a thin PDMS membrane is placed between glass layers, one containing fluidic channels and another containing manifold channels for valve access and actuation.16,17 Recently, there have been reports of hydrodynamic sample injection in the absence of electric field. Solignac and Gijs18 used a PDMS membrane seal on the sample reservoir to generate a pressure pulse that forced sample into the injection channel. The setup was able to show sequential injections with high repeatability and no bias, as well as variable plug length. To compensate for any back flow of solution into the sample reservoir, an additional sample waste channel is included in the design for electrokinetic pullback. However, the device is limited with respect to integrated applications, in that it requires a separate reservoir for actuation (eliminating the possibility of an enclosed reaction chamber) and offers no directionality when the reservoir is exposed to multiple channels. The Lunte group demonstrated a hydrodynamic injection using a modified gated injection on a double-T chip.19 The system is easy to implement and does not require precise voltage control, compared to a gated cross-T injection. However, like a gated injection, the sample is continuously pumped by EOF from (11) Guttman, A.; Schwartz, H. E. Anal. Chem. 1995, 67, 2279-2283. (12) Timperman, A.; Tracht, S. E.; Sweedler, J. V. Anal. Chem. 1996, 68, 26932698. (13) de Jesus, D. P.; Brito-Neto, J. G. A.; Richter, E. M.; Angnes, L.; Gutz, I. G. R.; do Lago, C. L. Anal. Chem. 2005, 77, 607-614. (14) Unger, M. A.; Chou, H.-P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, 113-116. (15) Studer, V.; Hang, G.; Pandolfi, A.; Ortiz, M.; Anderson, W. F.; Quake, S. R. J. Appl. Phys. 2004, 95, 393-398. (16) Grover, W. H.; Skelley, A. M.; Liu, C. N.; Lagally, E. T.; Mathies, R. A. Sens. Actuators, B 2003, 89, 315-323. (17) Skelley, A. M.; Scherer, J. R.; Aubrey, A. D.; Grover, W. H.; Ivester, R. H. C.; Ehrenfreund, P.; Grunthaner, F. G.; Bada, J. L.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 1041-1046. (18) Solignac, D.; Gijs, M. A. M. Anal. Chem. 2003, 75, 1652-1657. (19) Backofen, U.; Matysik, F. M.; Lunte, C. E. Anal. Chem. 2002, 74, 40544059.

3638

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

sample to sample waste during the separation, creating limitations on the amount of sample required and the duration of the analysis. It is perhaps important to consider whether an electrokinetic bias is even a problem worth addressing. The capillary electrophoresis literature frequently alludes to this, yet the plethora of publications on chip separations rarely mentions sample bias. Indeed, for qualitative analysis, it is the presence of peak patterns that are important, not reproducible peak area, and so bias is a problem only when the S/N limit is approached. For quantitative analysis, however, it is imperative that the sample injected be representative of the sample volume. Looking to complex devices, where precise control of analyte flow is required, sample bias is likely to pose a problem as well. This paper presents a microdevice for electrophoretic separations that is able to perform both EK and PR injections. Direct comparison of the two injection protocols can be made on the same device by actuating the valves in a controlled manner. We demonstrate the ability to reproducibly inject a representative sample successfully with subsecond injection times over multiple runs with pressure where EK fails. The pressure injection is also demonstrated for repeated introduction from a submicroliter sample volume. EXPERIMENTAL SECTION Reagents. Ultrapure water (Barnstead, Dubuque, IA) was used to prepare all aqueous solutions. Borate buffer was prepared from boric acid (Sigma-Aldrich, St. Louis, MO) and titrated to pH 8.1 with sodium hydroxide (Fisher Scientific, Pittsburgh, PA); stock 100 mM buffer was filtered, degassed, and stored at 4 °C for up to two weeks. Fresh 20 mM solutions of run buffer were prepared daily by diluting the stock. A 100 µM sodium fluorescein (FL) (Sigma-Aldrich) and 6-carboxy-X-rhodamine (ROX) (Fluka, Milwaukee, WI) solution was dissolved in run buffer and stored at -20 °C when not in use. Sodium hydroxide solutions were prepared at 0.5 M for chip conditioning. Microchip Design and Fabrication. The hybrid microdevices used in this experiment mimic the three-layer devices demonstrated by Grover et al.16 The top and bottom layers are borofloat glass, with channel features etched using standard wet chemical etching.20 The middle layer is a commercially available PDMS membrane (HT-6240, Bisco Silicones, Rogers Corp., Carol Stream, CT) with a thickness of 254 µm. Isotropically etched fluidic channels with a typical cross-T design (Figure 1A) were patterned into the top layer, with an initial width of 50 µm and depth of 50 µm. The separation channel (81 mm) is continuous throughout the chip; the injection channel (18 mm) is discontinuous in the sample arm where the valves are located, with 0.75-mm-wide gaps. The intersection of the channels occurs 6 mm from the buffer reservoir (BR). Manifold channels were patterned into the bottom layer with initial widths of 50 µm and terminated in the valve seats. Special attention was paid to the placement of the channels to avoid any overlap on the two layers. The elliptical shape of the valve seat was chosen based on previously published data.16 The gate valves had a long diameter of 2 mm and a short diameter of 1 mm prior to etching, and the diaphragm valve 3 and 1.5 mm. The manifold layer was etched to a depth of ∼60 µm, giving (20) Manz, A.; Fettinger, J. C.; Verpoorte, E.; Ludi, H.; Widmer, H. M.; Harrison, D. J. Trends Anal. Chem. 1991, 10, 144-149.

Figure 1. Device schematics. (A) Channel design and dimensions. Distance from cross-T intersection to reservoirs: BR, 0.6 cm; BW, 7.9 cm; SR and SW, 0.9 cm. Fluid channels (blue) etched 50 µm deep (50 µm original mask width). Valve channels (red) etched 60 µm deep. Volume of valve seats: gate, ∼100 nL; diaphragm, ∼200 nL. (B) Valve actuation sequence used to generate forward fluid pumping. Pressure (PR) injections presented as the number of cycles with the appropriate gate/diaphragm actuation times. (C) Assembled chip and pressure actuation manifold.

maximum volumes of 100 nL for the gate valves and 200 nL for the diaphragm based on an isotropic etch model. Access holes for both layers were drilled with 1.1-mm “triple ripple” diamondtipped bits (Abrasive Technology, Lewis Center, OH). After cleaning the patterned side of the fluidic layer with 2-propanol, the chip was assembled in a class 10 000 cleanroom. The PDMS membrane was rolled onto the patterned fluid layer to ensure good contact around the channel, creating a reversible seal. The patterned side of the manifold layer was then cleaned and brought into contact with the exposed side of the sealed membrane, positioning the valve seats over the gaps in the channel. The resulting channels in the assembled device have three glass walls and one PDMS wall, with a volume of ∼250 pL at the channel intersection. The PDMS film, sandwiched between the two glass layers, acts as a flexible membrane valve. A diagram of the assembled microdevice is shown in Figure 1A. Nanoport reservoirs (Upchurch, Oak Harbor, WA) were placed around the fluidic access holes and attached with the accompanying epoxy rings. Before use, the chip was then tested for proper valve function. Methods. Valve Actuation. The valves are normally in a closed position and actuate under vacuum. In the closed state, the PDMS membrane completely seals the gaps in the sample arm, preventing continuous fluid flow. To open, or actuate the valve, atmospheric pressure in the valve seat below each gap is reduced. This reduced pressure causes the PDMS membrane to deform into the valve seat and pull away from the fluid layer at the channel gap. This deformation allows liquid to flow around the channel

gap generating a continuous channel in the sample arm (“open state”). To close the valve once again, pressure in the valve seat is increased causing the PDMS membrane to push fluid into the channel and reseal the gap. The three valves in Figure 1A are arranged linearly to perform diaphragm pumping across the sample channel. In accordance with the pump described by Grover et al.,16 the central diaphragm valve defines the volume pumped, and the gate valves on either side enable directionality. The diaphragm pump was programmed to operate under four different modes: (1) pump open, where all of the valves opened simultaneously; (2) pump closed, where all of the valves closed simultaneously; (3) pump forward, where the gate valve closest to the sample reservoir acted as the entry valve; and (4) pump backward, where the gate valve closest to the cross-T intersection acted as the entry valve. The forward and reverse pumping was performed in six sequential steps, with forward pumping shown in Figure 1B. The dwell time for each of the steps was independently controlled, and any number of pump cycles could be performed. For this work, the valve actuation times were constant for both open and closed states of a pump sequence. Vacuum applied to the membrane was 60 kPa and pressure was 15 kPa. Injection and Separation. Both the pressure and the electrokinetic injections were performed on the same device. Fresh PDMS was used at the start of every day, resealing against a clean fluid layer. The fluidic channels were conditioned by pumping 0.5 M NaOH from the sample reservoir into the remaining channels. It was observed that the BR and sample waste (SW) reservoirs began to fill with solution, while the level in the buffer waste (BW) reservoir did not change after the initial filling of the channel. This is probably due to the length of the separation channel and increased flow resistance. To ensure thorough conditioning, 100 µL of fresh 0.5 M NaOH was placed in all reservoirs and electrokinetically flushed across the injection channel and down the separation channel. This was done with low field strengths (∼50 V/cm) for 3-5 min. The NaOH was then removed from the reservoirs, and the conditioning was repeated with run buffer using increased field strengths (∼300 V/cm). This was done at least twice; current was monitored during the conditioning to indicate whether additional buffer conditioning was required. The injection and separation conditions for the sequential injection series are displayed in Table 1 for both EK and PR injections. For EK injections, valves were maintained in an open state, generating a standard cross-T. (Because of three-dimensional nature of the valve seat, out of the plane of the separation channel, field strengths on the sample arm are estimates.) Except for the small sample volume experiment, all of the reservoirs were filled with the same solution volume (50 or 100 µL). Between runs, the buffer in the B, BW, and SW reservoirs was replaced. Initial experiments involved changing the valve actuation times and the duration of the EK injection to get similar sample plug volumes. In each case, the plug was defined primarily by the injection cross, but diffusion over long EK injections and displacement with increased actuation times had the potential to lengthen the plug. For later experiments where exaggerated bias was desired, the EK injection time was increased from 10 to 30 s; the pressure injection was also increased then, from 10 and 25 ms for the gate and diaphragm valve actuation times to 50 and 100 ms. Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3639

Table 1. Pressure/Voltage Programs for Injection and Separation

electrokinetic injection

time (s) 10

pressure injection

channel length (cm) SR SW BR BW

applied voltage (V) SR SW BR BW

0.9

0

no. cycles

0.9

separations

90

7.5

diaphragm valve actuation (ms)

10

25

channel length (cm) SR SW BR BW 0.9

-540

gate valve actuation (ms)

1

time (s)

0.6

0.9

0.6

7.5

SR

applied voltage (V) SW BR BW -390

250

-2400

Although plug shaping is typically used to focus the analyte zone,6 the BR and BW reservoirs were left floating during both injection types. This was done to isolate the effect of voltage in the injection step so that it would be present during the EK injection and completely absent during the pressure injection. The separation was performed with the pump closed so that voltage was only applied to the sample reservoir during the injection. After injection, a pullback voltage was applied to the sample waste arm in both cases. It should be noted that simply actuating the gate valve closest to the channel can generate a small injection even without the full pump procedure depicted in Figure 1B. Small Sample Volumes. For the small sample volume experiment, solution was removed from the Nanoport reservoir and the drilled access hole after conditioning the channel and injecting a buffer blank. The sample was then loaded directly into the access hole (volume ∼750 nL), and the reservoir was capped to reduce evaporation. A small hole was made in the cap to eliminate the possibility of a back pressure. Since the low sample volume generates little fluid pressure, diaphragm actuation times of 100 ms did not have enough time to fill the valve chamber and injections were limited at best (data not shown). For this reason, the actuation time was increased to 300 ms with small samples. Experimental Setup. A valve controller was built in-house to control the on-chip pumping. Pressure and vacuum chambers were included to ensure nonpulsatile flow from the external rotary pump, and 1/16-in. plastic tubing was split from each of the chambers to three-way miniature solenoid valves (LHDA 0533115H, Lee Co., Westbrook, CT). The output from the solenoid valves was connected to Sure-Lok fittings on a manifold stage made out of Plexiglas. Air channels were machined into the Plexiglas, and O-rings were embedded onto the stage corresponding to the glass manifold layer access holes. A second piece of Plexiglas was machined to fit above the microdevice, securing it to the manifold stage and sealing the O-rings against the chip valve layer. Platinum electrodes were placed into the fluid layer reservoirs for voltage control. A schematic of the assembled chip on the manifold stage is shown in Figure 1C. A LabVIEW (National Instruments, Austin, TX) program written in-house controlled the switching of the solenoid valves, the voltage application of four 5-kV power supplies (Spellman, Hauppauge, NY), and the data acquisition. The fluorescent dyes 3640 Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

Figure 2. Electrokinetic separations of FL and ROX. Sample, 100 µM ROX and 100 µM FL in 50 mM borate buffer. Sample volume, 50 µL. Overlay of EK injection (blue) and PR injection (red) demonstrate similar plug lengths. Chip dimensions and injection and separation conditions as described in Table 1.

were excited with the 488-nm line from an argon ion laser (Laser Physics, West Jordan, UT) using a conventional confocal detection setup21 with a 16× objective and a 1-mm pinhole, and emission was collected with a PMT (5784-01, Hamamatsu, Bridgewater, CT). The two dyes used in this experiment were selected for their different mobilities. Detection was performed with a 600-nm filter, where an overlap in emission was observed. It should be noted that the fluorescein exhibits a high quantum yield at the chosen excitation wavelength, while the ROX emits maximally in the detection region (λROX ) 578/604 nm, λFL ) 490/513 nm). Data processing was performed using Cutter 5 software.22 Safety Concerns. The instrumental setup uses high voltages and laser radiation; thus appropriate precautions were employed. RESULTS AND DISCUSSION Before beginning a direct comparison of electrokinetic and pressure-based injections, it was necessary to optimize the injection protocols for these microfluidic systems. Figure 2 demonstrates the qualitatively similar plug lengths achieved under experimental conditions listed in Table 1. The field strength for the EK injection was 300 V/cm. However, due to the threedimensional nature and increased volume of the valve seats, field strengths on the sample arm are expected to differ from those in a standard glass cross-T chip. Consequently, the exact electric field in the sample arm is not known. For both types of injection, the sample arm is closed during the separation, and voltages are applied to the remaining arms. Five control injections under these conditions generated peak areas for ROX and FL that differed by ∼10% for independent EK injections, while the differences in the peak area for independent pressure injections was ∼5% (Data not shown). A series of 20 sequential EK injections and separations is shown in Figure 3A. Field strengths are as described in Table 1. During separation, the field applied down the separation channel was 300 V/cm. Similarly, a 300 V/cm pull back was used on SW to prevent sample leakage into the channel during the separation. An overlay of runs 1, 5, 10, 15, and 20 are shown in Figure 3B. The two analytes, ROX and FL, are separated with a resolution of ∼1.2 for runs 1-20. As the more negatively charged species at this pH, FL migrates more slowly than ROX (µeof ≈ 7 × 10-4 cm2/ (21) Ocvirk, G.; Tang, T.; Harrison, D. J. Analyst 1998, 123, 1429-1434. (22) Shackman, J. G.; Watson, C. J.; Kennedy, R. T. J. Chromatogr., A 2004, 1040, 273-282.

Figure 3. Sequential electrokinetic separations of FL and ROX. Sample the same as in Figure 2. BR, AW, and BW reservoir buffers refreshed between runs. Chip dimensions and injection and separation conditions as described in Table 1. Valves closed after injection. (A) 20 sequential EK injections. (B) Overlay of runs 1 (red), 5 (orange), 10 (green), 15 (blue), and 20 (violet) for EK injection (denoted by arrows in A). (C) 20 sequential PR injections. (D) Overlay of runs 1 (red), 5 (orange), 10 (green), 15 (blue), and 20 (violet) for PR injection (denoted by arrows in C).

V‚s from the current monitoring method,23 µROX ≈ -3 × 10-4 cm2/ V‚s, µFL ≈ -5 × 10-4 cm2/V‚s). Of note is the fact that the ratio of ROX/FL injected in run 1 is clearly different from the ratio injected in run 20. In fact, the peak area of FL is seen to steadily decrease over time. With longer injection times (up to 60 s, data not shown), this difference was even more exaggerated. This is a good example of electrokinetic bias, a phenomenon that can cause nonrepresentative sample injections in CE. Figure 3C shows the same experiment conducted with valve-controlled pressure injections. Over the course of 20 runs, the sample composition is far more uniform and the resolution remains steady at Rs ∼1. It is clear from the overlaid traces in Figure 3D that there is little or no differentiation between runs. The peak area for both ROX and FL remained constant during the pressure injections, with the ROX/FL ratio exhibiting a 3.6% CV over the course of 20 runs. In comparison, for EK injections, the area of the FL peak decreased substantially over time while the ROX decreased to a lesser degree. The ROX/FL ratio varied over the first 10 runs with a 6.2% CV before increasing from 0.15 to 0.18 after 20 runs, a 20% change overall. In a complex sample mixture containing positive, negative, and neutral species, this change in sample matrix would be magnified. The ability of the diaphragm pump to reproducibly inject a representative sample, even after a series of EK injections, is illustrated in Figure 4. The first pair of peaks shown in Figure 4 represents a pressure injection of the fluorescein/ROX sample. (The injections in Figure 4 were preceded by a series of 10 EK (23) Huang, X.; Gordon, M. J.; Zare, R. N. Anal. Chem. 1988, 60, 1837-1838.

Figure 4. Alternating pressure and EK injections. Sample the same as in Figure 2. EK injection conditions (30 s at 300 V/cm) were chosen to accentuate the bias, and PR conditions (1 cycle with 50/100 ms actuation times) were selected to yield a similar peak area. Chip dimensions and separation conditions as described in Table 1. The initial PR injection displayed was followed by a series of five EK injections (red), where a decrease in the fluorescence intensity of the sample is observed. The PR injection was then repeated, showing a return to higher signal. This sequence was repeated two more times (EK shown in green, violet) to illustrate the trend.

injections and separations that resulted in a depletion of FL in the injections, similar to those shown in Figure 3A.) Immediately following the pressure injection and separation, five consecutive separations were performed on sample injected electrokinetically, with a sixth then using sample injected with pressure. This sequence was repeated twice, confirming the representative Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3641

Figure 5. ROX/FL peak area ratio plotted for the peaks in Figure 4. The ratio for each run is normalized against the initial PR injection. EK injections are represented by the squares and PR by the diamonds. The positive slope of the EK fit (blue) suggests a bias toward ROX during the injection, while the slightly negative slope of the PR fit (red) suggests a bias toward FL in the sample reservoir.

sample for pressure injections after a series of EK injections. This experiment also demonstrates the ability to perform both types of injections, easily alternating back and forth between the two types. Since fluorescein absorption and emission are known to be pH sensitive,24 it is possible that the decreased FL signal shown in Figures 3 and 4 could be caused by electrolysis. Electrolysis induces a pH shift in the reservoir, which in turn can affect the quantum efficiency of the fluorophore. By monitoring the pH of all reservoirs during the course of 20 sequential injections, it was determined that the pH of the sample decreased by ∼1 pH unit (data not shown, conditions as in Figure 3). While the pH shift in the sample reservoir may contribute to the decreasing FL signal observed for EK injections, the sequence of runs conducted in Figure 4 demonstrates that FL depletion is not exclusively due to a decreased quantum efficiency. When a series of EK runs are interrupted by a PR injection, a single pump cycle is sufficient to observe an increased fluorescence signal. Figure 5 compares the peak area ratio of ROX/FL obtained under pressure injections with the ratios obtained under EK injection. The values are normalized against the pressure injection data in the first run displayed. The ROX/FL ratio steadily increases with each EK injection, but this trend is not observed for the pressure. It is expected that if the decreased FL signal was entirely due to changing reservoir pH, the ratio of ROX/FL for both the EK and pressure runs would follow the same trend. It is interesting to note that, while successive EK injections are biased toward ROX, the slope of the PR data suggests a bias toward FL. This is not unexpected when one considers that an EK-biased sample injection will result in a biased sample composition in the reservoir. Not only is the sample injected not representative, but the sample composition is changing over successive EK injections. While we cannot definitively determine the cause of the decrease in FL signal, the observed bias appears to be a combination of electrolysis effects and analyte mobility. When applying voltage to the sample reservoir for multiple electrokinetic injections, there are a number of factors that will influence the ability to maintain representative and reproducible injections. Among these are the ionic strength and buffering (24) Sjoback, R.; Nygren, J.; Kubista, M. Spectrochim. Acta A 1995, 51, L7-L21.

3642 Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

Figure 6. Twelve sequential PR injections on a 1-µL sample volume. Sample the same as in Figure 2. One cycle PR injection applied with 50/300 ms actuation times for the gate/diaphragm valves. Chip dimensions and separation conditions same as Table 1. The ROX/FL peak area ratio (9) demonstrates that a representative amount of sample is injected even before the pump has become fully primed.

capacity of the sample buffer and electrolysis products such as pH change and bubble formation. Their effects are magnified as the sample volume becomes considerably smaller (e.g.,